kernel/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         pr_info("%s: enter\n", __func__);
1380
1381         /* Count only IPv4 services for old get/setsockopt interface */
1382         if (svc->af == AF_INET)
1383                 ipvs->num_services--;
1384
1385         ip_vs_stop_estimator(svc->ipvs, &svc->stats);
1386
1387         /* Unbind scheduler */
1388         old_sched = rcu_dereference_protected(svc->scheduler, 1);
1389         ip_vs_unbind_scheduler(svc, old_sched);
1390         ip_vs_scheduler_put(old_sched);
1391
1392         /* Unbind persistence engine, keep svc->pe */
1393         old_pe = rcu_dereference_protected(svc->pe, 1);
1394         ip_vs_pe_put(old_pe);
1395
1396         /*
1397          *    Unlink the whole destination list
1398          */
1399         list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1400                 __ip_vs_unlink_dest(svc, dest, 0);
1401                 __ip_vs_del_dest(svc->ipvs, dest, cleanup);
1402         }
1403
1404         /*
1405          *    Update the virtual service counters
1406          */
1407         if (svc->port == FTPPORT)
1408                 atomic_dec(&ipvs->ftpsvc_counter);
1409         else if (svc->port == 0)
1410                 atomic_dec(&ipvs->nullsvc_counter);
1411
1412         /*
1413          *    Free the service if nobody refers to it
1414          */
1415         __ip_vs_svc_put(svc, true);
1416
1417         /* decrease the module use count */
1418         ip_vs_use_count_dec();
1419 }
1420
1421 /*
1422  * Unlink a service from list and try to delete it if its refcnt reached 0
1423  */
1424 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1425 {
1426         /* Hold svc to avoid double release from dest_trash */
1427         atomic_inc(&svc->refcnt);
1428         /*
1429          * Unhash it from the service table
1430          */
1431         ip_vs_svc_unhash(svc);
1432
1433         __ip_vs_del_service(svc, cleanup);
1434 }
1435
1436 /*
1437  *      Delete a service from the service list
1438  */
1439 static int ip_vs_del_service(struct ip_vs_service *svc)
1440 {
1441         if (svc == NULL)
1442                 return -EEXIST;
1443         ip_vs_unlink_service(svc, false);
1444
1445         return 0;
1446 }
1447
1448
1449 /*
1450  *      Flush all the virtual services
1451  */
1452 static int ip_vs_flush(struct netns_ipvs *ipvs, bool cleanup)
1453 {
1454         int idx;
1455         struct ip_vs_service *svc;
1456         struct hlist_node *n;
1457
1458         /*
1459          * Flush the service table hashed by <netns,protocol,addr,port>
1460          */
1461         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1462                 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1463                                           s_list) {
1464                         if (svc->ipvs == ipvs)
1465                                 ip_vs_unlink_service(svc, cleanup);
1466                 }
1467         }
1468
1469         /*
1470          * Flush the service table hashed by fwmark
1471          */
1472         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1473                 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1474                                           f_list) {
1475                         if (svc->ipvs == ipvs)
1476                                 ip_vs_unlink_service(svc, cleanup);
1477                 }
1478         }
1479
1480         return 0;
1481 }
1482
1483 /*
1484  *      Delete service by {netns} in the service table.
1485  *      Called by __ip_vs_cleanup()
1486  */
1487 void ip_vs_service_net_cleanup(struct netns_ipvs *ipvs)
1488 {
1489         EnterFunction(2);
1490         /* Check for "full" addressed entries */
1491         mutex_lock(&__ip_vs_mutex);
1492         ip_vs_flush(ipvs, true);
1493         mutex_unlock(&__ip_vs_mutex);
1494         LeaveFunction(2);
1495 }
1496
1497 /* Put all references for device (dst_cache) */
1498 static inline void
1499 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1500 {
1501         struct ip_vs_dest_dst *dest_dst;
1502
1503         spin_lock_bh(&dest->dst_lock);
1504         dest_dst = rcu_dereference_protected(dest->dest_dst, 1);
1505         if (dest_dst && dest_dst->dst_cache->dev == dev) {
1506                 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1507                               dev->name,
1508                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1509                               ntohs(dest->port),
1510                               atomic_read(&dest->refcnt));
1511                 __ip_vs_dst_cache_reset(dest);
1512         }
1513         spin_unlock_bh(&dest->dst_lock);
1514
1515 }
1516 /* Netdev event receiver
1517  * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1518  */
1519 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1520                            void *ptr)
1521 {
1522         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1523         struct net *net = dev_net(dev);
1524         struct netns_ipvs *ipvs = net_ipvs(net);
1525         struct ip_vs_service *svc;
1526         struct ip_vs_dest *dest;
1527         unsigned int idx;
1528
1529         if (event != NETDEV_DOWN || !ipvs)
1530                 return NOTIFY_DONE;
1531         IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1532         EnterFunction(2);
1533         mutex_lock(&__ip_vs_mutex);
1534         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1535                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1536                         if (svc->ipvs == ipvs) {
1537                                 list_for_each_entry(dest, &svc->destinations,
1538                                                     n_list) {
1539                                         ip_vs_forget_dev(dest, dev);
1540                                 }
1541                         }
1542                 }
1543
1544                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1545                         if (svc->ipvs == ipvs) {
1546                                 list_for_each_entry(dest, &svc->destinations,
1547                                                     n_list) {
1548                                         ip_vs_forget_dev(dest, dev);
1549                                 }
1550                         }
1551
1552                 }
1553         }
1554
1555         spin_lock_bh(&ipvs->dest_trash_lock);
1556         list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1557                 ip_vs_forget_dev(dest, dev);
1558         }
1559         spin_unlock_bh(&ipvs->dest_trash_lock);
1560         mutex_unlock(&__ip_vs_mutex);
1561         LeaveFunction(2);
1562         return NOTIFY_DONE;
1563 }
1564
1565 /*
1566  *      Zero counters in a service or all services
1567  */
1568 static int ip_vs_zero_service(struct ip_vs_service *svc)
1569 {
1570         struct ip_vs_dest *dest;
1571
1572         list_for_each_entry(dest, &svc->destinations, n_list) {
1573                 ip_vs_zero_stats(&dest->stats);
1574         }
1575         ip_vs_zero_stats(&svc->stats);
1576         return 0;
1577 }
1578
1579 static int ip_vs_zero_all(struct netns_ipvs *ipvs)
1580 {
1581         int idx;
1582         struct ip_vs_service *svc;
1583
1584         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1585                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1586                         if (svc->ipvs == ipvs)
1587                                 ip_vs_zero_service(svc);
1588                 }
1589         }
1590
1591         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1592                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1593                         if (svc->ipvs == ipvs)
1594                                 ip_vs_zero_service(svc);
1595                 }
1596         }
1597
1598         ip_vs_zero_stats(&ipvs->tot_stats);
1599         return 0;
1600 }
1601
1602 #ifdef CONFIG_SYSCTL
1603
1604 static int zero;
1605 static int three = 3;
1606
1607 static int
1608 proc_do_defense_mode(struct ctl_table *table, int write,
1609                      void __user *buffer, size_t *lenp, loff_t *ppos)
1610 {
1611         struct netns_ipvs *ipvs = table->extra2;
1612         int *valp = table->data;
1613         int val = *valp;
1614         int rc;
1615
1616         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1617         if (write && (*valp != val)) {
1618                 if ((*valp < 0) || (*valp > 3)) {
1619                         /* Restore the correct value */
1620                         *valp = val;
1621                 } else {
1622                         update_defense_level(ipvs);
1623                 }
1624         }
1625         return rc;
1626 }
1627
1628 static int
1629 proc_do_sync_threshold(struct ctl_table *table, int write,
1630                        void __user *buffer, size_t *lenp, loff_t *ppos)
1631 {
1632         int *valp = table->data;
1633         int val[2];
1634         int rc;
1635
1636         /* backup the value first */
1637         memcpy(val, valp, sizeof(val));
1638
1639         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1640         if (write && (valp[0] < 0 || valp[1] < 0 ||
1641             (valp[0] >= valp[1] && valp[1]))) {
1642                 /* Restore the correct value */
1643                 memcpy(valp, val, sizeof(val));
1644         }
1645         return rc;
1646 }
1647
1648 static int
1649 proc_do_sync_mode(struct ctl_table *table, int write,
1650                      void __user *buffer, size_t *lenp, loff_t *ppos)
1651 {
1652         int *valp = table->data;
1653         int val = *valp;
1654         int rc;
1655
1656         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1657         if (write && (*valp != val)) {
1658                 if ((*valp < 0) || (*valp > 1)) {
1659                         /* Restore the correct value */
1660                         *valp = val;
1661                 }
1662         }
1663         return rc;
1664 }
1665
1666 static int
1667 proc_do_sync_ports(struct ctl_table *table, int write,
1668                    void __user *buffer, size_t *lenp, loff_t *ppos)
1669 {
1670         int *valp = table->data;
1671         int val = *valp;
1672         int rc;
1673
1674         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1675         if (write && (*valp != val)) {
1676                 if (*valp < 1 || !is_power_of_2(*valp)) {
1677                         /* Restore the correct value */
1678                         *valp = val;
1679                 }
1680         }
1681         return rc;
1682 }
1683
1684 /*
1685  *      IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1686  *      Do not change order or insert new entries without
1687  *      align with netns init in ip_vs_control_net_init()
1688  */
1689
1690 static struct ctl_table vs_vars[] = {
1691         {
1692                 .procname       = "amemthresh",
1693                 .maxlen         = sizeof(int),
1694                 .mode           = 0644,
1695                 .proc_handler   = proc_dointvec,
1696         },
1697         {
1698                 .procname       = "am_droprate",
1699                 .maxlen         = sizeof(int),
1700                 .mode           = 0644,
1701                 .proc_handler   = proc_dointvec,
1702         },
1703         {
1704                 .procname       = "drop_entry",
1705                 .maxlen         = sizeof(int),
1706                 .mode           = 0644,
1707                 .proc_handler   = proc_do_defense_mode,
1708         },
1709         {
1710                 .procname       = "drop_packet",
1711                 .maxlen         = sizeof(int),
1712                 .mode           = 0644,
1713                 .proc_handler   = proc_do_defense_mode,
1714         },
1715 #ifdef CONFIG_IP_VS_NFCT
1716         {
1717                 .procname       = "conntrack",
1718                 .maxlen         = sizeof(int),
1719                 .mode           = 0644,
1720                 .proc_handler   = &proc_dointvec,
1721         },
1722 #endif
1723         {
1724                 .procname       = "secure_tcp",
1725                 .maxlen         = sizeof(int),
1726                 .mode           = 0644,
1727                 .proc_handler   = proc_do_defense_mode,
1728         },
1729         {
1730                 .procname       = "snat_reroute",
1731                 .maxlen         = sizeof(int),
1732                 .mode           = 0644,
1733                 .proc_handler   = &proc_dointvec,
1734         },
1735         {
1736                 .procname       = "sync_version",
1737                 .maxlen         = sizeof(int),
1738                 .mode           = 0644,
1739                 .proc_handler   = &proc_do_sync_mode,
1740         },
1741         {
1742                 .procname       = "sync_ports",
1743                 .maxlen         = sizeof(int),
1744                 .mode           = 0644,
1745                 .proc_handler   = &proc_do_sync_ports,
1746         },
1747         {
1748                 .procname       = "sync_persist_mode",
1749                 .maxlen         = sizeof(int),
1750                 .mode           = 0644,
1751                 .proc_handler   = proc_dointvec,
1752         },
1753         {
1754                 .procname       = "sync_qlen_max",
1755                 .maxlen         = sizeof(unsigned long),
1756                 .mode           = 0644,
1757                 .proc_handler   = proc_doulongvec_minmax,
1758         },
1759         {
1760                 .procname       = "sync_sock_size",
1761                 .maxlen         = sizeof(int),
1762                 .mode           = 0644,
1763                 .proc_handler   = proc_dointvec,
1764         },
1765         {
1766                 .procname       = "cache_bypass",
1767                 .maxlen         = sizeof(int),
1768                 .mode           = 0644,
1769                 .proc_handler   = proc_dointvec,
1770         },
1771         {
1772                 .procname       = "expire_nodest_conn",
1773                 .maxlen         = sizeof(int),
1774                 .mode           = 0644,
1775                 .proc_handler   = proc_dointvec,
1776         },
1777         {
1778                 .procname       = "sloppy_tcp",
1779                 .maxlen         = sizeof(int),
1780                 .mode           = 0644,
1781                 .proc_handler   = proc_dointvec,
1782         },
1783         {
1784                 .procname       = "sloppy_sctp",
1785                 .maxlen         = sizeof(int),
1786                 .mode           = 0644,
1787                 .proc_handler   = proc_dointvec,
1788         },
1789         {
1790                 .procname       = "expire_quiescent_template",
1791                 .maxlen         = sizeof(int),
1792                 .mode           = 0644,
1793                 .proc_handler   = proc_dointvec,
1794         },
1795         {
1796                 .procname       = "sync_threshold",
1797                 .maxlen         =
1798                         sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1799                 .mode           = 0644,
1800                 .proc_handler   = proc_do_sync_threshold,
1801         },
1802         {
1803                 .procname       = "sync_refresh_period",
1804                 .maxlen         = sizeof(int),
1805                 .mode           = 0644,
1806                 .proc_handler   = proc_dointvec_jiffies,
1807         },
1808         {
1809                 .procname       = "sync_retries",
1810                 .maxlen         = sizeof(int),
1811                 .mode           = 0644,
1812                 .proc_handler   = proc_dointvec_minmax,
1813                 .extra1         = &zero,
1814                 .extra2         = &three,
1815         },
1816         {
1817                 .procname       = "nat_icmp_send",
1818                 .maxlen         = sizeof(int),
1819                 .mode           = 0644,
1820                 .proc_handler   = proc_dointvec,
1821         },
1822         {
1823                 .procname       = "pmtu_disc",
1824                 .maxlen         = sizeof(int),
1825                 .mode           = 0644,
1826                 .proc_handler   = proc_dointvec,
1827         },
1828         {
1829                 .procname       = "backup_only",
1830                 .maxlen         = sizeof(int),
1831                 .mode           = 0644,
1832                 .proc_handler   = proc_dointvec,
1833         },
1834         {
1835                 .procname       = "conn_reuse_mode",
1836                 .maxlen         = sizeof(int),
1837                 .mode           = 0644,
1838                 .proc_handler   = proc_dointvec,
1839         },
1840         {
1841                 .procname       = "schedule_icmp",
1842                 .maxlen         = sizeof(int),
1843                 .mode           = 0644,
1844                 .proc_handler   = proc_dointvec,
1845         },
1846         {
1847                 .procname       = "ignore_tunneled",
1848                 .maxlen         = sizeof(int),
1849                 .mode           = 0644,
1850                 .proc_handler   = proc_dointvec,
1851         },
1852 #ifdef CONFIG_IP_VS_DEBUG
1853         {
1854                 .procname       = "debug_level",
1855                 .data           = &sysctl_ip_vs_debug_level,
1856                 .maxlen         = sizeof(int),
1857                 .mode           = 0644,
1858                 .proc_handler   = proc_dointvec,
1859         },
1860 #endif
1861         { }
1862 };
1863
1864 #endif
1865
1866 #ifdef CONFIG_PROC_FS
1867
1868 struct ip_vs_iter {
1869         struct seq_net_private p;  /* Do not move this, netns depends upon it*/
1870         struct hlist_head *table;
1871         int bucket;
1872 };
1873
1874 /*
1875  *      Write the contents of the VS rule table to a PROCfs file.
1876  *      (It is kept just for backward compatibility)
1877  */
1878 static inline const char *ip_vs_fwd_name(unsigned int flags)
1879 {
1880         switch (flags & IP_VS_CONN_F_FWD_MASK) {
1881         case IP_VS_CONN_F_LOCALNODE:
1882                 return "Local";
1883         case IP_VS_CONN_F_TUNNEL:
1884                 return "Tunnel";
1885         case IP_VS_CONN_F_DROUTE:
1886                 return "Route";
1887         default:
1888                 return "Masq";
1889         }
1890 }
1891
1892
1893 /* Get the Nth entry in the two lists */
1894 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1895 {
1896         struct net *net = seq_file_net(seq);
1897         struct netns_ipvs *ipvs = net_ipvs(net);
1898         struct ip_vs_iter *iter = seq->private;
1899         int idx;
1900         struct ip_vs_service *svc;
1901
1902         /* look in hash by protocol */
1903         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1904                 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
1905                         if ((svc->ipvs == ipvs) && pos-- == 0) {
1906                                 iter->table = ip_vs_svc_table;
1907                                 iter->bucket = idx;
1908                                 return svc;
1909                         }
1910                 }
1911         }
1912
1913         /* keep looking in fwmark */
1914         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1915                 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
1916                                          f_list) {
1917                         if ((svc->ipvs == ipvs) && pos-- == 0) {
1918                                 iter->table = ip_vs_svc_fwm_table;
1919                                 iter->bucket = idx;
1920                                 return svc;
1921                         }
1922                 }
1923         }
1924
1925         return NULL;
1926 }
1927
1928 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1929         __acquires(RCU)
1930 {
1931         rcu_read_lock();
1932         return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1933 }
1934
1935
1936 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1937 {
1938         struct hlist_node *e;
1939         struct ip_vs_iter *iter;
1940         struct ip_vs_service *svc;
1941
1942         ++*pos;
1943         if (v == SEQ_START_TOKEN)
1944                 return ip_vs_info_array(seq,0);
1945
1946         svc = v;
1947         iter = seq->private;
1948
1949         if (iter->table == ip_vs_svc_table) {
1950                 /* next service in table hashed by protocol */
1951                 e = rcu_dereference(hlist_next_rcu(&svc->s_list));
1952                 if (e)
1953                         return hlist_entry(e, struct ip_vs_service, s_list);
1954
1955                 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1956                         hlist_for_each_entry_rcu(svc,
1957                                                  &ip_vs_svc_table[iter->bucket],
1958                                                  s_list) {
1959                                 return svc;
1960                         }
1961                 }
1962
1963                 iter->table = ip_vs_svc_fwm_table;
1964                 iter->bucket = -1;
1965                 goto scan_fwmark;
1966         }
1967
1968         /* next service in hashed by fwmark */
1969         e = rcu_dereference(hlist_next_rcu(&svc->f_list));
1970         if (e)
1971                 return hlist_entry(e, struct ip_vs_service, f_list);
1972
1973  scan_fwmark:
1974         while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1975                 hlist_for_each_entry_rcu(svc,
1976                                          &ip_vs_svc_fwm_table[iter->bucket],
1977                                          f_list)
1978                         return svc;
1979         }
1980
1981         return NULL;
1982 }
1983
1984 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
1985         __releases(RCU)
1986 {
1987         rcu_read_unlock();
1988 }
1989
1990
1991 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
1992 {
1993         if (v == SEQ_START_TOKEN) {
1994                 seq_printf(seq,
1995                         "IP Virtual Server version %d.%d.%d (size=%d)\n",
1996                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
1997                 seq_puts(seq,
1998                          "Prot LocalAddress:Port Scheduler Flags\n");
1999                 seq_puts(seq,
2000                          "  -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2001         } else {
2002                 const struct ip_vs_service *svc = v;
2003                 const struct ip_vs_iter *iter = seq->private;
2004                 const struct ip_vs_dest *dest;
2005                 struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler);
2006                 char *sched_name = sched ? sched->name : "none";
2007
2008                 if (iter->table == ip_vs_svc_table) {
2009 #ifdef CONFIG_IP_VS_IPV6
2010                         if (svc->af == AF_INET6)
2011                                 seq_printf(seq, "%s  [%pI6]:%04X %s ",
2012                                            ip_vs_proto_name(svc->protocol),
2013                                            &svc->addr.in6,
2014                                            ntohs(svc->port),
2015                                            sched_name);
2016                         else
2017 #endif
2018                                 seq_printf(seq, "%s  %08X:%04X %s %s ",
2019                                            ip_vs_proto_name(svc->protocol),
2020                                            ntohl(svc->addr.ip),
2021                                            ntohs(svc->port),
2022                                            sched_name,
2023                                            (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2024                 } else {
2025                         seq_printf(seq, "FWM  %08X %s %s",
2026                                    svc->fwmark, sched_name,
2027                                    (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2028                 }
2029
2030                 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2031                         seq_printf(seq, "persistent %d %08X\n",
2032                                 svc->timeout,
2033                                 ntohl(svc->netmask));
2034                 else
2035                         seq_putc(seq, '\n');
2036
2037                 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
2038 #ifdef CONFIG_IP_VS_IPV6
2039                         if (dest->af == AF_INET6)
2040                                 seq_printf(seq,
2041                                            "  -> [%pI6]:%04X"
2042                                            "      %-7s %-6d %-10d %-10d\n",
2043                                            &dest->addr.in6,
2044                                            ntohs(dest->port),
2045                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2046                                            atomic_read(&dest->weight),
2047                                            atomic_read(&dest->activeconns),
2048                                            atomic_read(&dest->inactconns));
2049                         else
2050 #endif
2051                                 seq_printf(seq,
2052                                            "  -> %08X:%04X      "
2053                                            "%-7s %-6d %-10d %-10d\n",
2054                                            ntohl(dest->addr.ip),
2055                                            ntohs(dest->port),
2056                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2057                                            atomic_read(&dest->weight),
2058                                            atomic_read(&dest->activeconns),
2059                                            atomic_read(&dest->inactconns));
2060
2061                 }
2062         }
2063         return 0;
2064 }
2065
2066 static const struct seq_operations ip_vs_info_seq_ops = {
2067         .start = ip_vs_info_seq_start,
2068         .next  = ip_vs_info_seq_next,
2069         .stop  = ip_vs_info_seq_stop,
2070         .show  = ip_vs_info_seq_show,
2071 };
2072
2073 static int ip_vs_info_open(struct inode *inode, struct file *file)
2074 {
2075         return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2076                         sizeof(struct ip_vs_iter));
2077 }
2078
2079 static const struct file_operations ip_vs_info_fops = {
2080         .owner   = THIS_MODULE,
2081         .open    = ip_vs_info_open,
2082         .read    = seq_read,
2083         .llseek  = seq_lseek,
2084         .release = seq_release_net,
2085 };
2086
2087 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2088 {
2089         struct net *net = seq_file_single_net(seq);
2090         struct ip_vs_kstats show;
2091
2092 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
2093         seq_puts(seq,
2094                  "   Total Incoming Outgoing         Incoming         Outgoing\n");
2095         seq_printf(seq,
2096                    "   Conns  Packets  Packets            Bytes            Bytes\n");
2097
2098         ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2099         seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n\n",
2100                    (unsigned long long)show.conns,
2101                    (unsigned long long)show.inpkts,
2102                    (unsigned long long)show.outpkts,
2103                    (unsigned long long)show.inbytes,
2104                    (unsigned long long)show.outbytes);
2105
2106 /*                01234567 01234567 01234567 0123456701234567 0123456701234567*/
2107         seq_puts(seq,
2108                  " Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
2109         seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n",
2110                    (unsigned long long)show.cps,
2111                    (unsigned long long)show.inpps,
2112                    (unsigned long long)show.outpps,
2113                    (unsigned long long)show.inbps,
2114                    (unsigned long long)show.outbps);
2115
2116         return 0;
2117 }
2118
2119 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2120 {
2121         return single_open_net(inode, file, ip_vs_stats_show);
2122 }
2123
2124 static const struct file_operations ip_vs_stats_fops = {
2125         .owner = THIS_MODULE,
2126         .open = ip_vs_stats_seq_open,
2127         .read = seq_read,
2128         .llseek = seq_lseek,
2129         .release = single_release_net,
2130 };
2131
2132 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2133 {
2134         struct net *net = seq_file_single_net(seq);
2135         struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2136         struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats;
2137         struct ip_vs_kstats kstats;
2138         int i;
2139
2140 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
2141         seq_puts(seq,
2142                  "       Total Incoming Outgoing         Incoming         Outgoing\n");
2143         seq_printf(seq,
2144                    "CPU    Conns  Packets  Packets            Bytes            Bytes\n");
2145
2146         for_each_possible_cpu(i) {
2147                 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2148                 unsigned int start;
2149                 u64 conns, inpkts, outpkts, inbytes, outbytes;
2150
2151                 do {
2152                         start = u64_stats_fetch_begin_irq(&u->syncp);
2153                         conns = u->cnt.conns;
2154                         inpkts = u->cnt.inpkts;
2155                         outpkts = u->cnt.outpkts;
2156                         inbytes = u->cnt.inbytes;
2157                         outbytes = u->cnt.outbytes;
2158                 } while (u64_stats_fetch_retry_irq(&u->syncp, start));
2159
2160                 seq_printf(seq, "%3X %8LX %8LX %8LX %16LX %16LX\n",
2161                            i, (u64)conns, (u64)inpkts,
2162                            (u64)outpkts, (u64)inbytes,
2163                            (u64)outbytes);
2164         }
2165
2166         ip_vs_copy_stats(&kstats, tot_stats);
2167
2168         seq_printf(seq, "  ~ %8LX %8LX %8LX %16LX %16LX\n\n",
2169                    (unsigned long long)kstats.conns,
2170                    (unsigned long long)kstats.inpkts,
2171                    (unsigned long long)kstats.outpkts,
2172                    (unsigned long long)kstats.inbytes,
2173                    (unsigned long long)kstats.outbytes);
2174
2175 /*                ... 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2176         seq_puts(seq,
2177                  "     Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
2178         seq_printf(seq, "    %8LX %8LX %8LX %16LX %16LX\n",
2179                    kstats.cps,
2180                    kstats.inpps,
2181                    kstats.outpps,
2182                    kstats.inbps,
2183                    kstats.outbps);
2184
2185         return 0;
2186 }
2187
2188 static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2189 {
2190         return single_open_net(inode, file, ip_vs_stats_percpu_show);
2191 }
2192
2193 static const struct file_operations ip_vs_stats_percpu_fops = {
2194         .owner = THIS_MODULE,
2195         .open = ip_vs_stats_percpu_seq_open,
2196         .read = seq_read,
2197         .llseek = seq_lseek,
2198         .release = single_release_net,
2199 };
2200 #endif
2201
2202 /*
2203  *      Set timeout values for tcp tcpfin udp in the timeout_table.
2204  */
2205 static int ip_vs_set_timeout(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2206 {
2207 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2208         struct ip_vs_proto_data *pd;
2209 #endif
2210
2211         IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2212                   u->tcp_timeout,
2213                   u->tcp_fin_timeout,
2214                   u->udp_timeout);
2215
2216 #ifdef CONFIG_IP_VS_PROTO_TCP
2217         if (u->tcp_timeout) {
2218                 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2219                 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2220                         = u->tcp_timeout * HZ;
2221         }
2222
2223         if (u->tcp_fin_timeout) {
2224                 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2225                 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2226                         = u->tcp_fin_timeout * HZ;
2227         }
2228 #endif
2229
2230 #ifdef CONFIG_IP_VS_PROTO_UDP
2231         if (u->udp_timeout) {
2232                 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2233                 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2234                         = u->udp_timeout * HZ;
2235         }
2236 #endif
2237         return 0;
2238 }
2239
2240 #define CMDID(cmd)              (cmd - IP_VS_BASE_CTL)
2241
2242 struct ip_vs_svcdest_user {
2243         struct ip_vs_service_user       s;
2244         struct ip_vs_dest_user          d;
2245 };
2246
2247 static const unsigned char set_arglen[CMDID(IP_VS_SO_SET_MAX) + 1] = {
2248         [CMDID(IP_VS_SO_SET_ADD)]         = sizeof(struct ip_vs_service_user),
2249         [CMDID(IP_VS_SO_SET_EDIT)]        = sizeof(struct ip_vs_service_user),
2250         [CMDID(IP_VS_SO_SET_DEL)]         = sizeof(struct ip_vs_service_user),
2251         [CMDID(IP_VS_SO_SET_ADDDEST)]     = sizeof(struct ip_vs_svcdest_user),
2252         [CMDID(IP_VS_SO_SET_DELDEST)]     = sizeof(struct ip_vs_svcdest_user),
2253         [CMDID(IP_VS_SO_SET_EDITDEST)]    = sizeof(struct ip_vs_svcdest_user),
2254         [CMDID(IP_VS_SO_SET_TIMEOUT)]     = sizeof(struct ip_vs_timeout_user),
2255         [CMDID(IP_VS_SO_SET_STARTDAEMON)] = sizeof(struct ip_vs_daemon_user),
2256         [CMDID(IP_VS_SO_SET_STOPDAEMON)]  = sizeof(struct ip_vs_daemon_user),
2257         [CMDID(IP_VS_SO_SET_ZERO)]        = sizeof(struct ip_vs_service_user),
2258 };
2259
2260 union ip_vs_set_arglen {
2261         struct ip_vs_service_user       field_IP_VS_SO_SET_ADD;
2262         struct ip_vs_service_user       field_IP_VS_SO_SET_EDIT;
2263         struct ip_vs_service_user       field_IP_VS_SO_SET_DEL;
2264         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_ADDDEST;
2265         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_DELDEST;
2266         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_EDITDEST;
2267         struct ip_vs_timeout_user       field_IP_VS_SO_SET_TIMEOUT;
2268         struct ip_vs_daemon_user        field_IP_VS_SO_SET_STARTDAEMON;
2269         struct ip_vs_daemon_user        field_IP_VS_SO_SET_STOPDAEMON;
2270         struct ip_vs_service_user       field_IP_VS_SO_SET_ZERO;
2271 };
2272
2273 #define MAX_SET_ARGLEN  sizeof(union ip_vs_set_arglen)
2274
2275 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2276                                   struct ip_vs_service_user *usvc_compat)
2277 {
2278         memset(usvc, 0, sizeof(*usvc));
2279
2280         usvc->af                = AF_INET;
2281         usvc->protocol          = usvc_compat->protocol;
2282         usvc->addr.ip           = usvc_compat->addr;
2283         usvc->port              = usvc_compat->port;
2284         usvc->fwmark            = usvc_compat->fwmark;
2285
2286         /* Deep copy of sched_name is not needed here */
2287         usvc->sched_name        = usvc_compat->sched_name;
2288
2289         usvc->flags             = usvc_compat->flags;
2290         usvc->timeout           = usvc_compat->timeout;
2291         usvc->netmask           = usvc_compat->netmask;
2292 }
2293
2294 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2295                                    struct ip_vs_dest_user *udest_compat)
2296 {
2297         memset(udest, 0, sizeof(*udest));
2298
2299         udest->addr.ip          = udest_compat->addr;
2300         udest->port             = udest_compat->port;
2301         udest->conn_flags       = udest_compat->conn_flags;
2302         udest->weight           = udest_compat->weight;
2303         udest->u_threshold      = udest_compat->u_threshold;
2304         udest->l_threshold      = udest_compat->l_threshold;
2305         udest->af               = AF_INET;
2306 }
2307
2308 static int
2309 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2310 {
2311         struct net *net = sock_net(sk);
2312         int ret;
2313         unsigned char arg[MAX_SET_ARGLEN];
2314         struct ip_vs_service_user *usvc_compat;
2315         struct ip_vs_service_user_kern usvc;
2316         struct ip_vs_service *svc;
2317         struct ip_vs_dest_user *udest_compat;
2318         struct ip_vs_dest_user_kern udest;
2319         struct netns_ipvs *ipvs = net_ipvs(net);
2320
2321         BUILD_BUG_ON(sizeof(arg) > 255);
2322         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2323                 return -EPERM;
2324
2325         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2326                 return -EINVAL;
2327         if (len != set_arglen[CMDID(cmd)]) {
2328                 IP_VS_DBG(1, "set_ctl: len %u != %u\n",
2329                           len, set_arglen[CMDID(cmd)]);
2330                 return -EINVAL;
2331         }
2332
2333         if (copy_from_user(arg, user, len) != 0)
2334                 return -EFAULT;
2335
2336         /* increase the module use count */
2337         ip_vs_use_count_inc();
2338
2339         /* Handle daemons since they have another lock */
2340         if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2341             cmd == IP_VS_SO_SET_STOPDAEMON) {
2342                 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2343
2344                 if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2345                         struct ipvs_sync_daemon_cfg cfg;
2346
2347                         memset(&cfg, 0, sizeof(cfg));
2348                         strlcpy(cfg.mcast_ifn, dm->mcast_ifn,
2349                                 sizeof(cfg.mcast_ifn));
2350                         cfg.syncid = dm->syncid;
2351                         rtnl_lock();
2352                         mutex_lock(&ipvs->sync_mutex);
2353                         ret = start_sync_thread(ipvs, &cfg, dm->state);
2354                         mutex_unlock(&ipvs->sync_mutex);
2355                         rtnl_unlock();
2356                 } else {
2357                         mutex_lock(&ipvs->sync_mutex);
2358                         ret = stop_sync_thread(ipvs, dm->state);
2359                         mutex_unlock(&ipvs->sync_mutex);
2360                 }
2361                 goto out_dec;
2362         }
2363
2364         mutex_lock(&__ip_vs_mutex);
2365         if (cmd == IP_VS_SO_SET_FLUSH) {
2366                 /* Flush the virtual service */
2367                 ret = ip_vs_flush(ipvs, false);
2368                 goto out_unlock;
2369         } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2370                 /* Set timeout values for (tcp tcpfin udp) */
2371                 ret = ip_vs_set_timeout(ipvs, (struct ip_vs_timeout_user *)arg);
2372                 goto out_unlock;
2373         }
2374
2375         usvc_compat = (struct ip_vs_service_user *)arg;
2376         udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2377
2378         /* We only use the new structs internally, so copy userspace compat
2379          * structs to extended internal versions */
2380         ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2381         ip_vs_copy_udest_compat(&udest, udest_compat);
2382
2383         if (cmd == IP_VS_SO_SET_ZERO) {
2384                 /* if no service address is set, zero counters in all */
2385                 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2386                         ret = ip_vs_zero_all(ipvs);
2387                         goto out_unlock;
2388                 }
2389         }
2390
2391         /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2392         if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2393             usvc.protocol != IPPROTO_SCTP) {
2394                 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2395                        usvc.protocol, &usvc.addr.ip,
2396                        ntohs(usvc.port), usvc.sched_name);
2397                 ret = -EFAULT;
2398                 goto out_unlock;
2399         }
2400
2401         /* Lookup the exact service by <protocol, addr, port> or fwmark */
2402         rcu_read_lock();
2403         if (usvc.fwmark == 0)
2404                 svc = __ip_vs_service_find(ipvs, usvc.af, usvc.protocol,
2405                                            &usvc.addr, usvc.port);
2406         else
2407                 svc = __ip_vs_svc_fwm_find(ipvs, usvc.af, usvc.fwmark);
2408         rcu_read_unlock();
2409
2410         if (cmd != IP_VS_SO_SET_ADD
2411             && (svc == NULL || svc->protocol != usvc.protocol)) {
2412                 ret = -ESRCH;
2413                 goto out_unlock;
2414         }
2415
2416         switch (cmd) {
2417         case IP_VS_SO_SET_ADD:
2418                 if (svc != NULL)
2419                         ret = -EEXIST;
2420                 else
2421                         ret = ip_vs_add_service(ipvs, &usvc, &svc);
2422                 break;
2423         case IP_VS_SO_SET_EDIT:
2424                 ret = ip_vs_edit_service(svc, &usvc);
2425                 break;
2426         case IP_VS_SO_SET_DEL:
2427                 ret = ip_vs_del_service(svc);
2428                 if (!ret)
2429                         goto out_unlock;
2430                 break;
2431         case IP_VS_SO_SET_ZERO:
2432                 ret = ip_vs_zero_service(svc);
2433                 break;
2434         case IP_VS_SO_SET_ADDDEST:
2435                 ret = ip_vs_add_dest(svc, &udest);
2436                 break;
2437         case IP_VS_SO_SET_EDITDEST:
2438                 ret = ip_vs_edit_dest(svc, &udest);
2439                 break;
2440         case IP_VS_SO_SET_DELDEST:
2441                 ret = ip_vs_del_dest(svc, &udest);
2442                 break;
2443         default:
2444                 ret = -EINVAL;
2445         }
2446
2447   out_unlock:
2448         mutex_unlock(&__ip_vs_mutex);
2449   out_dec:
2450         /* decrease the module use count */
2451         ip_vs_use_count_dec();
2452
2453         return ret;
2454 }
2455
2456
2457 static void
2458 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2459 {
2460         struct ip_vs_scheduler *sched;
2461         struct ip_vs_kstats kstats;
2462         char *sched_name;
2463
2464         sched = rcu_dereference_protected(src->scheduler, 1);
2465         sched_name = sched ? sched->name : "none";
2466         dst->protocol = src->protocol;
2467         dst->addr = src->addr.ip;
2468         dst->port = src->port;
2469         dst->fwmark = src->fwmark;
2470         strlcpy(dst->sched_name, sched_name, sizeof(dst->sched_name));
2471         dst->flags = src->flags;
2472         dst->timeout = src->timeout / HZ;
2473         dst->netmask = src->netmask;
2474         dst->num_dests = src->num_dests;
2475         ip_vs_copy_stats(&kstats, &src->stats);
2476         ip_vs_export_stats_user(&dst->stats, &kstats);
2477 }
2478
2479 static inline int
2480 __ip_vs_get_service_entries(struct netns_ipvs *ipvs,
2481                             const struct ip_vs_get_services *get,
2482                             struct ip_vs_get_services __user *uptr)
2483 {
2484         int idx, count=0;
2485         struct ip_vs_service *svc;
2486         struct ip_vs_service_entry entry;
2487         int ret = 0;
2488
2489         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2490                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2491                         /* Only expose IPv4 entries to old interface */
2492                         if (svc->af != AF_INET || (svc->ipvs != ipvs))
2493                                 continue;
2494
2495                         if (count >= get->num_services)
2496                                 goto out;
2497                         memset(&entry, 0, sizeof(entry));
2498                         ip_vs_copy_service(&entry, svc);
2499                         if (copy_to_user(&uptr->entrytable[count],
2500                                          &entry, sizeof(entry))) {
2501                                 ret = -EFAULT;
2502                                 goto out;
2503                         }
2504                         count++;
2505                 }
2506         }
2507
2508         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2509                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2510                         /* Only expose IPv4 entries to old interface */
2511                         if (svc->af != AF_INET || (svc->ipvs != ipvs))
2512                                 continue;
2513
2514                         if (count >= get->num_services)
2515                                 goto out;
2516                         memset(&entry, 0, sizeof(entry));
2517                         ip_vs_copy_service(&entry, svc);
2518                         if (copy_to_user(&uptr->entrytable[count],
2519                                          &entry, sizeof(entry))) {
2520                                 ret = -EFAULT;
2521                                 goto out;
2522                         }
2523                         count++;
2524                 }
2525         }
2526 out:
2527         return ret;
2528 }
2529
2530 static inline int
2531 __ip_vs_get_dest_entries(struct netns_ipvs *ipvs, const struct ip_vs_get_dests *get,
2532                          struct ip_vs_get_dests __user *uptr)
2533 {
2534         struct ip_vs_service *svc;
2535         union nf_inet_addr addr = { .ip = get->addr };
2536         int ret = 0;
2537
2538         rcu_read_lock();
2539         if (get->fwmark)
2540                 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, get->fwmark);
2541         else
2542                 svc = __ip_vs_service_find(ipvs, AF_INET, get->protocol, &addr,
2543                                            get->port);
2544         rcu_read_unlock();
2545
2546         if (svc) {
2547                 int count = 0;
2548                 struct ip_vs_dest *dest;
2549                 struct ip_vs_dest_entry entry;
2550                 struct ip_vs_kstats kstats;
2551
2552                 memset(&entry, 0, sizeof(entry));
2553                 list_for_each_entry(dest, &svc->destinations, n_list) {
2554                         if (count >= get->num_dests)
2555                                 break;
2556
2557                         /* Cannot expose heterogeneous members via sockopt
2558                          * interface
2559                          */
2560                         if (dest->af != svc->af)
2561                                 continue;
2562
2563                         entry.addr = dest->addr.ip;
2564                         entry.port = dest->port;
2565                         entry.conn_flags = atomic_read(&dest->conn_flags);
2566                         entry.weight = atomic_read(&dest->weight);
2567                         entry.u_threshold = dest->u_threshold;
2568                         entry.l_threshold = dest->l_threshold;
2569                         entry.activeconns = atomic_read(&dest->activeconns);
2570                         entry.inactconns = atomic_read(&dest->inactconns);
2571                         entry.persistconns = atomic_read(&dest->persistconns);
2572                         ip_vs_copy_stats(&kstats, &dest->stats);
2573                         ip_vs_export_stats_user(&entry.stats, &kstats);
2574                         if (copy_to_user(&uptr->entrytable[count],
2575                                          &entry, sizeof(entry))) {
2576                                 ret = -EFAULT;
2577                                 break;
2578                         }
2579                         count++;
2580                 }
2581         } else
2582                 ret = -ESRCH;
2583         return ret;
2584 }
2585
2586 static inline void
2587 __ip_vs_get_timeouts(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2588 {
2589 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2590         struct ip_vs_proto_data *pd;
2591 #endif
2592
2593         memset(u, 0, sizeof (*u));
2594
2595 #ifdef CONFIG_IP_VS_PROTO_TCP
2596         pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2597         u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2598         u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2599 #endif
2600 #ifdef CONFIG_IP_VS_PROTO_UDP
2601         pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2602         u->udp_timeout =
2603                         pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2604 #endif
2605 }
2606
2607 static const unsigned char get_arglen[CMDID(IP_VS_SO_GET_MAX) + 1] = {
2608         [CMDID(IP_VS_SO_GET_VERSION)]  = 64,
2609         [CMDID(IP_VS_SO_GET_INFO)]     = sizeof(struct ip_vs_getinfo),
2610         [CMDID(IP_VS_SO_GET_SERVICES)] = sizeof(struct ip_vs_get_services),
2611         [CMDID(IP_VS_SO_GET_SERVICE)]  = sizeof(struct ip_vs_service_entry),
2612         [CMDID(IP_VS_SO_GET_DESTS)]    = sizeof(struct ip_vs_get_dests),
2613         [CMDID(IP_VS_SO_GET_TIMEOUT)]  = sizeof(struct ip_vs_timeout_user),
2614         [CMDID(IP_VS_SO_GET_DAEMON)]   = 2 * sizeof(struct ip_vs_daemon_user),
2615 };
2616
2617 union ip_vs_get_arglen {
2618         char                            field_IP_VS_SO_GET_VERSION[64];
2619         struct ip_vs_getinfo            field_IP_VS_SO_GET_INFO;
2620         struct ip_vs_get_services       field_IP_VS_SO_GET_SERVICES;
2621         struct ip_vs_service_entry      field_IP_VS_SO_GET_SERVICE;
2622         struct ip_vs_get_dests          field_IP_VS_SO_GET_DESTS;
2623         struct ip_vs_timeout_user       field_IP_VS_SO_GET_TIMEOUT;
2624         struct ip_vs_daemon_user        field_IP_VS_SO_GET_DAEMON[2];
2625 };
2626
2627 #define MAX_GET_ARGLEN  sizeof(union ip_vs_get_arglen)
2628
2629 static int
2630 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2631 {
2632         unsigned char arg[MAX_GET_ARGLEN];
2633         int ret = 0;
2634         unsigned int copylen;
2635         struct net *net = sock_net(sk);
2636         struct netns_ipvs *ipvs = net_ipvs(net);
2637
2638         BUG_ON(!net);
2639         BUILD_BUG_ON(sizeof(arg) > 255);
2640         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2641                 return -EPERM;
2642
2643         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2644                 return -EINVAL;
2645
2646         copylen = get_arglen[CMDID(cmd)];
2647         if (*len < (int) copylen) {
2648                 IP_VS_DBG(1, "get_ctl: len %d < %u\n", *len, copylen);
2649                 return -EINVAL;
2650         }
2651
2652         if (copy_from_user(arg, user, copylen) != 0)
2653                 return -EFAULT;
2654         /*
2655          * Handle daemons first since it has its own locking
2656          */
2657         if (cmd == IP_VS_SO_GET_DAEMON) {
2658                 struct ip_vs_daemon_user d[2];
2659
2660                 memset(&d, 0, sizeof(d));
2661                 mutex_lock(&ipvs->sync_mutex);
2662                 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2663                         d[0].state = IP_VS_STATE_MASTER;
2664                         strlcpy(d[0].mcast_ifn, ipvs->mcfg.mcast_ifn,
2665                                 sizeof(d[0].mcast_ifn));
2666                         d[0].syncid = ipvs->mcfg.syncid;
2667                 }
2668                 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2669                         d[1].state = IP_VS_STATE_BACKUP;
2670                         strlcpy(d[1].mcast_ifn, ipvs->bcfg.mcast_ifn,
2671                                 sizeof(d[1].mcast_ifn));
2672                         d[1].syncid = ipvs->bcfg.syncid;
2673                 }
2674                 if (copy_to_user(user, &d, sizeof(d)) != 0)
2675                         ret = -EFAULT;
2676                 mutex_unlock(&ipvs->sync_mutex);
2677                 return ret;
2678         }
2679
2680         mutex_lock(&__ip_vs_mutex);
2681         switch (cmd) {
2682         case IP_VS_SO_GET_VERSION:
2683         {
2684                 char buf[64];
2685
2686                 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2687                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2688                 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2689                         ret = -EFAULT;
2690                         goto out;
2691                 }
2692                 *len = strlen(buf)+1;
2693         }
2694         break;
2695
2696         case IP_VS_SO_GET_INFO:
2697         {
2698                 struct ip_vs_getinfo info;
2699                 info.version = IP_VS_VERSION_CODE;
2700                 info.size = ip_vs_conn_tab_size;
2701                 info.num_services = ipvs->num_services;
2702                 if (copy_to_user(user, &info, sizeof(info)) != 0)
2703                         ret = -EFAULT;
2704         }
2705         break;
2706
2707         case IP_VS_SO_GET_SERVICES:
2708         {
2709                 struct ip_vs_get_services *get;
2710                 int size;
2711
2712                 get = (struct ip_vs_get_services *)arg;
2713                 size = sizeof(*get) +
2714                         sizeof(struct ip_vs_service_entry) * get->num_services;
2715                 if (*len != size) {
2716                         pr_err("length: %u != %u\n", *len, size);
2717                         ret = -EINVAL;
2718                         goto out;
2719                 }
2720                 ret = __ip_vs_get_service_entries(ipvs, get, user);
2721         }
2722         break;
2723
2724         case IP_VS_SO_GET_SERVICE:
2725         {
2726                 struct ip_vs_service_entry *entry;
2727                 struct ip_vs_service *svc;
2728                 union nf_inet_addr addr;
2729
2730                 entry = (struct ip_vs_service_entry *)arg;
2731                 addr.ip = entry->addr;
2732                 rcu_read_lock();
2733                 if (entry->fwmark)
2734                         svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, entry->fwmark);
2735                 else
2736                         svc = __ip_vs_service_find(ipvs, AF_INET,
2737                                                    entry->protocol, &addr,
2738                                                    entry->port);
2739                 rcu_read_unlock();
2740                 if (svc) {
2741                         ip_vs_copy_service(entry, svc);
2742                         if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2743                                 ret = -EFAULT;
2744                 } else
2745                         ret = -ESRCH;
2746         }
2747         break;
2748
2749         case IP_VS_SO_GET_DESTS:
2750         {
2751                 struct ip_vs_get_dests *get;
2752                 int size;
2753
2754                 get = (struct ip_vs_get_dests *)arg;
2755                 size = sizeof(*get) +
2756                         sizeof(struct ip_vs_dest_entry) * get->num_dests;
2757                 if (*len != size) {
2758                         pr_err("length: %u != %u\n", *len, size);
2759                         ret = -EINVAL;
2760                         goto out;
2761                 }
2762                 ret = __ip_vs_get_dest_entries(ipvs, get, user);
2763         }
2764         break;
2765
2766         case IP_VS_SO_GET_TIMEOUT:
2767         {
2768                 struct ip_vs_timeout_user t;
2769
2770                 __ip_vs_get_timeouts(ipvs, &t);
2771                 if (copy_to_user(user, &t, sizeof(t)) != 0)
2772                         ret = -EFAULT;
2773         }
2774         break;
2775
2776         default:
2777                 ret = -EINVAL;
2778         }
2779
2780 out:
2781         mutex_unlock(&__ip_vs_mutex);
2782         return ret;
2783 }
2784
2785
2786 static struct nf_sockopt_ops ip_vs_sockopts = {
2787         .pf             = PF_INET,
2788         .set_optmin     = IP_VS_BASE_CTL,
2789         .set_optmax     = IP_VS_SO_SET_MAX+1,
2790         .set            = do_ip_vs_set_ctl,
2791         .get_optmin     = IP_VS_BASE_CTL,
2792         .get_optmax     = IP_VS_SO_GET_MAX+1,
2793         .get            = do_ip_vs_get_ctl,
2794         .owner          = THIS_MODULE,
2795 };
2796
2797 /*
2798  * Generic Netlink interface
2799  */
2800
2801 /* IPVS genetlink family */
2802 static struct genl_family ip_vs_genl_family = {
2803         .id             = GENL_ID_GENERATE,
2804         .hdrsize        = 0,
2805         .name           = IPVS_GENL_NAME,
2806         .version        = IPVS_GENL_VERSION,
2807         .maxattr        = IPVS_CMD_MAX,
2808         .netnsok        = true,         /* Make ipvsadm to work on netns */
2809 };
2810
2811 /* Policy used for first-level command attributes */
2812 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2813         [IPVS_CMD_ATTR_SERVICE]         = { .type = NLA_NESTED },
2814         [IPVS_CMD_ATTR_DEST]            = { .type = NLA_NESTED },
2815         [IPVS_CMD_ATTR_DAEMON]          = { .type = NLA_NESTED },
2816         [IPVS_CMD_ATTR_TIMEOUT_TCP]     = { .type = NLA_U32 },
2817         [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2818         [IPVS_CMD_ATTR_TIMEOUT_UDP]     = { .type = NLA_U32 },
2819 };
2820
2821 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2822 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2823         [IPVS_DAEMON_ATTR_STATE]        = { .type = NLA_U32 },
2824         [IPVS_DAEMON_ATTR_MCAST_IFN]    = { .type = NLA_NUL_STRING,
2825                                             .len = IP_VS_IFNAME_MAXLEN },
2826         [IPVS_DAEMON_ATTR_SYNC_ID]      = { .type = NLA_U32 },
2827         [IPVS_DAEMON_ATTR_SYNC_MAXLEN]  = { .type = NLA_U16 },
2828         [IPVS_DAEMON_ATTR_MCAST_GROUP]  = { .type = NLA_U32 },
2829         [IPVS_DAEMON_ATTR_MCAST_GROUP6] = { .len = sizeof(struct in6_addr) },
2830         [IPVS_DAEMON_ATTR_MCAST_PORT]   = { .type = NLA_U16 },
2831         [IPVS_DAEMON_ATTR_MCAST_TTL]    = { .type = NLA_U8 },
2832 };
2833
2834 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2835 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2836         [IPVS_SVC_ATTR_AF]              = { .type = NLA_U16 },
2837         [IPVS_SVC_ATTR_PROTOCOL]        = { .type = NLA_U16 },
2838         [IPVS_SVC_ATTR_ADDR]            = { .type = NLA_BINARY,
2839                                             .len = sizeof(union nf_inet_addr) },
2840         [IPVS_SVC_ATTR_PORT]            = { .type = NLA_U16 },
2841         [IPVS_SVC_ATTR_FWMARK]          = { .type = NLA_U32 },
2842         [IPVS_SVC_ATTR_SCHED_NAME]      = { .type = NLA_NUL_STRING,
2843                                             .len = IP_VS_SCHEDNAME_MAXLEN },
2844         [IPVS_SVC_ATTR_PE_NAME]         = { .type = NLA_NUL_STRING,
2845                                             .len = IP_VS_PENAME_MAXLEN },
2846         [IPVS_SVC_ATTR_FLAGS]           = { .type = NLA_BINARY,
2847                                             .len = sizeof(struct ip_vs_flags) },
2848         [IPVS_SVC_ATTR_TIMEOUT]         = { .type = NLA_U32 },
2849         [IPVS_SVC_ATTR_NETMASK]         = { .type = NLA_U32 },
2850         [IPVS_SVC_ATTR_STATS]           = { .type = NLA_NESTED },
2851 };
2852
2853 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2854 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2855         [IPVS_DEST_ATTR_ADDR]           = { .type = NLA_BINARY,
2856                                             .len = sizeof(union nf_inet_addr) },
2857         [IPVS_DEST_ATTR_PORT]           = { .type = NLA_U16 },
2858         [IPVS_DEST_ATTR_FWD_METHOD]     = { .type = NLA_U32 },
2859         [IPVS_DEST_ATTR_WEIGHT]         = { .type = NLA_U32 },
2860         [IPVS_DEST_ATTR_U_THRESH]       = { .type = NLA_U32 },
2861         [IPVS_DEST_ATTR_L_THRESH]       = { .type = NLA_U32 },
2862         [IPVS_DEST_ATTR_ACTIVE_CONNS]   = { .type = NLA_U32 },
2863         [IPVS_DEST_ATTR_INACT_CONNS]    = { .type = NLA_U32 },
2864         [IPVS_DEST_ATTR_PERSIST_CONNS]  = { .type = NLA_U32 },
2865         [IPVS_DEST_ATTR_STATS]          = { .type = NLA_NESTED },
2866         [IPVS_DEST_ATTR_ADDR_FAMILY]    = { .type = NLA_U16 },
2867 };
2868
2869 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2870                                  struct ip_vs_kstats *kstats)
2871 {
2872         struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2873
2874         if (!nl_stats)
2875                 return -EMSGSIZE;
2876
2877         if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, (u32)kstats->conns) ||
2878             nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, (u32)kstats->inpkts) ||
2879             nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, (u32)kstats->outpkts) ||
2880             nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes) ||
2881             nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes) ||
2882             nla_put_u32(skb, IPVS_STATS_ATTR_CPS, (u32)kstats->cps) ||
2883             nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, (u32)kstats->inpps) ||
2884             nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, (u32)kstats->outpps) ||
2885             nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, (u32)kstats->inbps) ||
2886             nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, (u32)kstats->outbps))
2887                 goto nla_put_failure;
2888         nla_nest_end(skb, nl_stats);
2889
2890         return 0;
2891
2892 nla_put_failure:
2893         nla_nest_cancel(skb, nl_stats);
2894         return -EMSGSIZE;
2895 }
2896
2897 static int ip_vs_genl_fill_stats64(struct sk_buff *skb, int container_type,
2898                                    struct ip_vs_kstats *kstats)
2899 {
2900         struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2901
2902         if (!nl_stats)
2903                 return -EMSGSIZE;
2904
2905         if (nla_put_u64(skb, IPVS_STATS_ATTR_CONNS, kstats->conns) ||
2906             nla_put_u64(skb, IPVS_STATS_ATTR_INPKTS, kstats->inpkts) ||
2907             nla_put_u64(skb, IPVS_STATS_ATTR_OUTPKTS, kstats->outpkts) ||
2908             nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes) ||
2909             nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes) ||
2910             nla_put_u64(skb, IPVS_STATS_ATTR_CPS, kstats->cps) ||
2911             nla_put_u64(skb, IPVS_STATS_ATTR_INPPS, kstats->inpps) ||
2912             nla_put_u64(skb, IPVS_STATS_ATTR_OUTPPS, kstats->outpps) ||
2913             nla_put_u64(skb, IPVS_STATS_ATTR_INBPS, kstats->inbps) ||
2914             nla_put_u64(skb, IPVS_STATS_ATTR_OUTBPS, kstats->outbps))
2915                 goto nla_put_failure;
2916         nla_nest_end(skb, nl_stats);
2917
2918         return 0;
2919
2920 nla_put_failure:
2921         nla_nest_cancel(skb, nl_stats);
2922         return -EMSGSIZE;
2923 }
2924
2925 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2926                                    struct ip_vs_service *svc)
2927 {
2928         struct ip_vs_scheduler *sched;
2929         struct ip_vs_pe *pe;
2930         struct nlattr *nl_service;
2931         struct ip_vs_flags flags = { .flags = svc->flags,
2932                                      .mask = ~0 };
2933         struct ip_vs_kstats kstats;
2934         char *sched_name;
2935
2936         nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2937         if (!nl_service)
2938                 return -EMSGSIZE;
2939
2940         if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2941                 goto nla_put_failure;
2942         if (svc->fwmark) {
2943                 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2944                         goto nla_put_failure;
2945         } else {
2946                 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2947                     nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2948                     nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2949                         goto nla_put_failure;
2950         }
2951
2952         sched = rcu_dereference_protected(svc->scheduler, 1);
2953         sched_name = sched ? sched->name : "none";
2954         pe = rcu_dereference_protected(svc->pe, 1);
2955         if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched_name) ||
2956             (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
2957             nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2958             nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2959             nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
2960                 goto nla_put_failure;
2961         ip_vs_copy_stats(&kstats, &svc->stats);
2962         if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &kstats))
2963                 goto nla_put_failure;
2964         if (ip_vs_genl_fill_stats64(skb, IPVS_SVC_ATTR_STATS64, &kstats))
2965                 goto nla_put_failure;
2966
2967         nla_nest_end(skb, nl_service);
2968
2969         return 0;
2970
2971 nla_put_failure:
2972         nla_nest_cancel(skb, nl_service);
2973         return -EMSGSIZE;
2974 }
2975
2976 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2977                                    struct ip_vs_service *svc,
2978                                    struct netlink_callback *cb)
2979 {
2980         void *hdr;
2981
2982         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2983                           &ip_vs_genl_family, NLM_F_MULTI,
2984                           IPVS_CMD_NEW_SERVICE);
2985         if (!hdr)
2986                 return -EMSGSIZE;
2987
2988         if (ip_vs_genl_fill_service(skb, svc) < 0)
2989                 goto nla_put_failure;
2990
2991         genlmsg_end(skb, hdr);
2992         return 0;
2993
2994 nla_put_failure:
2995         genlmsg_cancel(skb, hdr);
2996         return -EMSGSIZE;
2997 }
2998
2999 static int ip_vs_genl_dump_services(struct sk_buff *skb,
3000                                     struct netlink_callback *cb)
3001 {
3002         int idx = 0, i;
3003         int start = cb->args[0];
3004         struct ip_vs_service *svc;
3005         struct net *net = sock_net(skb->sk);
3006         struct netns_ipvs *ipvs = net_ipvs(net);
3007
3008         mutex_lock(&__ip_vs_mutex);
3009         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3010                 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
3011                         if (++idx <= start || (svc->ipvs != ipvs))
3012                                 continue;
3013                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3014                                 idx--;
3015                                 goto nla_put_failure;
3016                         }
3017                 }
3018         }
3019
3020         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3021                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
3022                         if (++idx <= start || (svc->ipvs != ipvs))
3023                                 continue;
3024                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3025                                 idx--;
3026                                 goto nla_put_failure;
3027                         }
3028                 }
3029         }
3030
3031 nla_put_failure:
3032         mutex_unlock(&__ip_vs_mutex);
3033         cb->args[0] = idx;
3034
3035         return skb->len;
3036 }
3037
3038 static int ip_vs_genl_parse_service(struct netns_ipvs *ipvs,
3039                                     struct ip_vs_service_user_kern *usvc,
3040                                     struct nlattr *nla, int full_entry,
3041                                     struct ip_vs_service **ret_svc)
3042 {
3043         struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
3044         struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
3045         struct ip_vs_service *svc;
3046
3047         /* Parse mandatory identifying service fields first */
3048         if (nla == NULL ||
3049             nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
3050                 return -EINVAL;
3051
3052         nla_af          = attrs[IPVS_SVC_ATTR_AF];
3053         nla_protocol    = attrs[IPVS_SVC_ATTR_PROTOCOL];
3054         nla_addr        = attrs[IPVS_SVC_ATTR_ADDR];
3055         nla_port        = attrs[IPVS_SVC_ATTR_PORT];
3056         nla_fwmark      = attrs[IPVS_SVC_ATTR_FWMARK];
3057
3058         if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3059                 return -EINVAL;
3060
3061         memset(usvc, 0, sizeof(*usvc));
3062
3063         usvc->af = nla_get_u16(nla_af);
3064 #ifdef CONFIG_IP_VS_IPV6
3065         if (usvc->af != AF_INET && usvc->af != AF_INET6)
3066 #else
3067         if (usvc->af != AF_INET)
3068 #endif
3069                 return -EAFNOSUPPORT;
3070
3071         if (nla_fwmark) {
3072                 usvc->protocol = IPPROTO_TCP;
3073                 usvc->fwmark = nla_get_u32(nla_fwmark);
3074         } else {
3075                 usvc->protocol = nla_get_u16(nla_protocol);
3076                 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3077                 usvc->port = nla_get_be16(nla_port);
3078                 usvc->fwmark = 0;
3079         }
3080
3081         rcu_read_lock();
3082         if (usvc->fwmark)
3083                 svc = __ip_vs_svc_fwm_find(ipvs, usvc->af, usvc->fwmark);
3084         else
3085                 svc = __ip_vs_service_find(ipvs, usvc->af, usvc->protocol,
3086                                            &usvc->addr, usvc->port);
3087         rcu_read_unlock();
3088         *ret_svc = svc;
3089
3090         /* If a full entry was requested, check for the additional fields */
3091         if (full_entry) {
3092                 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3093                               *nla_netmask;
3094                 struct ip_vs_flags flags;
3095
3096                 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3097                 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3098                 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3099                 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3100                 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3101
3102                 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3103                         return -EINVAL;
3104
3105                 nla_memcpy(&flags, nla_flags, sizeof(flags));
3106
3107                 /* prefill flags from service if it already exists */
3108                 if (svc)
3109                         usvc->flags = svc->flags;
3110
3111                 /* set new flags from userland */
3112                 usvc->flags = (usvc->flags & ~flags.mask) |
3113                               (flags.flags & flags.mask);
3114                 usvc->sched_name = nla_data(nla_sched);
3115                 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3116                 usvc->timeout = nla_get_u32(nla_timeout);
3117                 usvc->netmask = nla_get_be32(nla_netmask);
3118         }
3119
3120         return 0;
3121 }
3122
3123 static struct ip_vs_service *ip_vs_genl_find_service(struct netns_ipvs *ipvs,
3124                                                      struct nlattr *nla)
3125 {
3126         struct ip_vs_service_user_kern usvc;
3127         struct ip_vs_service *svc;
3128         int ret;
3129
3130         ret = ip_vs_genl_parse_service(ipvs, &usvc, nla, 0, &svc);
3131         return ret ? ERR_PTR(ret) : svc;
3132 }
3133
3134 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3135 {
3136         struct nlattr *nl_dest;
3137         struct ip_vs_kstats kstats;
3138
3139         nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3140         if (!nl_dest)
3141                 return -EMSGSIZE;
3142
3143         if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3144             nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3145             nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3146                         (atomic_read(&dest->conn_flags) &
3147                          IP_VS_CONN_F_FWD_MASK)) ||
3148             nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3149                         atomic_read(&dest->weight)) ||
3150             nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3151             nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3152             nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3153                         atomic_read(&dest->activeconns)) ||
3154             nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3155                         atomic_read(&dest->inactconns)) ||
3156             nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3157                         atomic_read(&dest->persistconns)) ||
3158             nla_put_u16(skb, IPVS_DEST_ATTR_ADDR_FAMILY, dest->af))
3159                 goto nla_put_failure;
3160         ip_vs_copy_stats(&kstats, &dest->stats);
3161         if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &kstats))
3162                 goto nla_put_failure;
3163         if (ip_vs_genl_fill_stats64(skb, IPVS_DEST_ATTR_STATS64, &kstats))
3164                 goto nla_put_failure;
3165
3166         nla_nest_end(skb, nl_dest);
3167
3168         return 0;
3169
3170 nla_put_failure:
3171         nla_nest_cancel(skb, nl_dest);
3172         return -EMSGSIZE;
3173 }
3174
3175 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3176                                 struct netlink_callback *cb)
3177 {
3178         void *hdr;
3179
3180         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3181                           &ip_vs_genl_family, NLM_F_MULTI,
3182                           IPVS_CMD_NEW_DEST);
3183         if (!hdr)
3184                 return -EMSGSIZE;
3185
3186         if (ip_vs_genl_fill_dest(skb, dest) < 0)
3187                 goto nla_put_failure;
3188
3189         genlmsg_end(skb, hdr);
3190         return 0;
3191
3192 nla_put_failure:
3193         genlmsg_cancel(skb, hdr);
3194         return -EMSGSIZE;
3195 }
3196
3197 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3198                                  struct netlink_callback *cb)
3199 {
3200         int idx = 0;
3201         int start = cb->args[0];
3202         struct ip_vs_service *svc;
3203         struct ip_vs_dest *dest;
3204         struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3205         struct net *net = sock_net(skb->sk);
3206         struct netns_ipvs *ipvs = net_ipvs(net);
3207
3208         mutex_lock(&__ip_vs_mutex);
3209
3210         /* Try to find the service for which to dump destinations */
3211         if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3212                         IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3213                 goto out_err;
3214
3215
3216         svc = ip_vs_genl_find_service(ipvs, attrs[IPVS_CMD_ATTR_SERVICE]);
3217         if (IS_ERR(svc) || svc == NULL)
3218                 goto out_err;
3219
3220         /* Dump the destinations */
3221         list_for_each_entry(dest, &svc->destinations, n_list) {
3222                 if (++idx <= start)
3223                         continue;
3224                 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3225                         idx--;
3226                         goto nla_put_failure;
3227                 }
3228         }
3229
3230 nla_put_failure:
3231         cb->args[0] = idx;
3232
3233 out_err:
3234         mutex_unlock(&__ip_vs_mutex);
3235
3236         return skb->len;
3237 }
3238
3239 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3240                                  struct nlattr *nla, int full_entry)
3241 {
3242         struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3243         struct nlattr *nla_addr, *nla_port;
3244         struct nlattr *nla_addr_family;
3245
3246         /* Parse mandatory identifying destination fields first */
3247         if (nla == NULL ||
3248             nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3249                 return -EINVAL;
3250
3251         nla_addr        = attrs[IPVS_DEST_ATTR_ADDR];
3252         nla_port        = attrs[IPVS_DEST_ATTR_PORT];
3253         nla_addr_family = attrs[IPVS_DEST_ATTR_ADDR_FAMILY];
3254
3255         if (!(nla_addr && nla_port))
3256                 return -EINVAL;
3257
3258         memset(udest, 0, sizeof(*udest));
3259
3260         nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3261         udest->port = nla_get_be16(nla_port);
3262
3263         if (nla_addr_family)
3264                 udest->af = nla_get_u16(nla_addr_family);
3265         else
3266                 udest->af = 0;
3267
3268         /* If a full entry was requested, check for the additional fields */
3269         if (full_entry) {
3270                 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3271                               *nla_l_thresh;
3272
3273                 nla_fwd         = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3274                 nla_weight      = attrs[IPVS_DEST_ATTR_WEIGHT];
3275                 nla_u_thresh    = attrs[IPVS_DEST_ATTR_U_THRESH];
3276                 nla_l_thresh    = attrs[IPVS_DEST_ATTR_L_THRESH];
3277
3278                 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3279                         return -EINVAL;
3280
3281                 udest->conn_flags = nla_get_u32(nla_fwd)
3282                                     & IP_VS_CONN_F_FWD_MASK;
3283                 udest->weight = nla_get_u32(nla_weight);
3284                 udest->u_threshold = nla_get_u32(nla_u_thresh);
3285                 udest->l_threshold = nla_get_u32(nla_l_thresh);
3286         }
3287
3288         return 0;
3289 }
3290
3291 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3292                                   struct ipvs_sync_daemon_cfg *c)
3293 {
3294         struct nlattr *nl_daemon;
3295
3296         nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3297         if (!nl_daemon)
3298                 return -EMSGSIZE;
3299
3300         if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3301             nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, c->mcast_ifn) ||
3302             nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, c->syncid) ||
3303             nla_put_u16(skb, IPVS_DAEMON_ATTR_SYNC_MAXLEN, c->sync_maxlen) ||
3304             nla_put_u16(skb, IPVS_DAEMON_ATTR_MCAST_PORT, c->mcast_port) ||
3305             nla_put_u8(skb, IPVS_DAEMON_ATTR_MCAST_TTL, c->mcast_ttl))
3306                 goto nla_put_failure;
3307 #ifdef CONFIG_IP_VS_IPV6
3308         if (c->mcast_af == AF_INET6) {
3309                 if (nla_put_in6_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP6,
3310                                      &c->mcast_group.in6))
3311                         goto nla_put_failure;
3312         } else
3313 #endif
3314                 if (c->mcast_af == AF_INET &&
3315                     nla_put_in_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP,
3316                                     c->mcast_group.ip))
3317                         goto nla_put_failure;
3318         nla_nest_end(skb, nl_daemon);
3319
3320         return 0;
3321
3322 nla_put_failure:
3323         nla_nest_cancel(skb, nl_daemon);
3324         return -EMSGSIZE;
3325 }
3326
3327 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3328                                   struct ipvs_sync_daemon_cfg *c,
3329                                   struct netlink_callback *cb)
3330 {
3331         void *hdr;
3332         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3333                           &ip_vs_genl_family, NLM_F_MULTI,
3334                           IPVS_CMD_NEW_DAEMON);
3335         if (!hdr)
3336                 return -EMSGSIZE;
3337
3338         if (ip_vs_genl_fill_daemon(skb, state, c))
3339                 goto nla_put_failure;
3340
3341         genlmsg_end(skb, hdr);
3342         return 0;
3343
3344 nla_put_failure:
3345         genlmsg_cancel(skb, hdr);
3346         return -EMSGSIZE;
3347 }
3348
3349 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3350                                    struct netlink_callback *cb)
3351 {
3352         struct net *net = sock_net(skb->sk);
3353         struct netns_ipvs *ipvs = net_ipvs(net);
3354
3355         mutex_lock(&ipvs->sync_mutex);
3356         if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3357                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3358                                            &ipvs->mcfg, cb) < 0)
3359                         goto nla_put_failure;
3360
3361                 cb->args[0] = 1;
3362         }
3363
3364         if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3365                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3366                                            &ipvs->bcfg, cb) < 0)
3367                         goto nla_put_failure;
3368
3369                 cb->args[1] = 1;
3370         }
3371
3372 nla_put_failure:
3373         mutex_unlock(&ipvs->sync_mutex);
3374
3375         return skb->len;
3376 }
3377
3378 static int ip_vs_genl_new_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3379 {
3380         struct ipvs_sync_daemon_cfg c;
3381         struct nlattr *a;
3382         int ret;
3383
3384         memset(&c, 0, sizeof(c));
3385         if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3386               attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3387               attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3388                 return -EINVAL;
3389         strlcpy(c.mcast_ifn, nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3390                 sizeof(c.mcast_ifn));
3391         c.syncid = nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]);
3392
3393         a = attrs[IPVS_DAEMON_ATTR_SYNC_MAXLEN];
3394         if (a)
3395                 c.sync_maxlen = nla_get_u16(a);
3396
3397         a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP];
3398         if (a) {
3399                 c.mcast_af = AF_INET;
3400                 c.mcast_group.ip = nla_get_in_addr(a);
3401                 if (!ipv4_is_multicast(c.mcast_group.ip))
3402                         return -EINVAL;
3403         } else {
3404                 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP6];
3405                 if (a) {
3406 #ifdef CONFIG_IP_VS_IPV6
3407                         int addr_type;
3408
3409                         c.mcast_af = AF_INET6;
3410                         c.mcast_group.in6 = nla_get_in6_addr(a);
3411                         addr_type = ipv6_addr_type(&c.mcast_group.in6);
3412                         if (!(addr_type & IPV6_ADDR_MULTICAST))
3413                                 return -EINVAL;
3414 #else
3415                         return -EAFNOSUPPORT;
3416 #endif
3417                 }
3418         }
3419
3420         a = attrs[IPVS_DAEMON_ATTR_MCAST_PORT];
3421         if (a)
3422                 c.mcast_port = nla_get_u16(a);
3423
3424         a = attrs[IPVS_DAEMON_ATTR_MCAST_TTL];
3425         if (a)
3426                 c.mcast_ttl = nla_get_u8(a);
3427
3428         /* The synchronization protocol is incompatible with mixed family
3429          * services
3430          */
3431         if (ipvs->mixed_address_family_dests > 0)
3432                 return -EINVAL;
3433
3434         rtnl_lock();
3435         mutex_lock(&ipvs->sync_mutex);
3436         ret = start_sync_thread(ipvs, &c,
3437                                 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3438         mutex_unlock(&ipvs->sync_mutex);
3439         rtnl_unlock();
3440         return ret;
3441 }
3442
3443 static int ip_vs_genl_del_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3444 {
3445         int ret;
3446
3447         if (!attrs[IPVS_DAEMON_ATTR_STATE])
3448                 return -EINVAL;
3449
3450         mutex_lock(&ipvs->sync_mutex);
3451         ret = stop_sync_thread(ipvs,
3452                                nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3453         mutex_unlock(&ipvs->sync_mutex);
3454         return ret;
3455 }
3456
3457 static int ip_vs_genl_set_config(struct netns_ipvs *ipvs, struct nlattr **attrs)
3458 {
3459         struct ip_vs_timeout_user t;
3460
3461         __ip_vs_get_timeouts(ipvs, &t);
3462
3463         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3464                 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3465
3466         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3467                 t.tcp_fin_timeout =
3468                         nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3469
3470         if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3471                 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3472
3473         return ip_vs_set_timeout(ipvs, &t);
3474 }
3475
3476 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3477 {
3478         int ret = -EINVAL, cmd;
3479         struct net *net = sock_net(skb->sk);
3480         struct netns_ipvs *ipvs = net_ipvs(net);
3481
3482         cmd = info->genlhdr->cmd;
3483
3484         if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3485                 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3486
3487                 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3488                     nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3489                                      info->attrs[IPVS_CMD_ATTR_DAEMON],
3490                                      ip_vs_daemon_policy))
3491                         goto out;
3492
3493                 if (cmd == IPVS_CMD_NEW_DAEMON)
3494                         ret = ip_vs_genl_new_daemon(ipvs, daemon_attrs);
3495                 else
3496                         ret = ip_vs_genl_del_daemon(ipvs, daemon_attrs);
3497         }
3498
3499 out:
3500         return ret;
3501 }
3502
3503 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3504 {
3505         struct ip_vs_service *svc = NULL;
3506         struct ip_vs_service_user_kern usvc;
3507         struct ip_vs_dest_user_kern udest;
3508         int ret = 0, cmd;
3509         int need_full_svc = 0, need_full_dest = 0;
3510         struct net *net = sock_net(skb->sk);
3511         struct netns_ipvs *ipvs = net_ipvs(net);
3512
3513         cmd = info->genlhdr->cmd;
3514
3515         mutex_lock(&__ip_vs_mutex);
3516
3517         if (cmd == IPVS_CMD_FLUSH) {
3518                 ret = ip_vs_flush(ipvs, false);
3519                 goto out;
3520         } else if (cmd == IPVS_CMD_SET_CONFIG) {
3521                 ret = ip_vs_genl_set_config(ipvs, info->attrs);
3522                 goto out;
3523         } else if (cmd == IPVS_CMD_ZERO &&
3524                    !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3525                 ret = ip_vs_zero_all(ipvs);
3526                 goto out;
3527         }
3528
3529         /* All following commands require a service argument, so check if we
3530          * received a valid one. We need a full service specification when
3531          * adding / editing a service. Only identifying members otherwise. */
3532         if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3533                 need_full_svc = 1;
3534
3535         ret = ip_vs_genl_parse_service(ipvs, &usvc,
3536                                        info->attrs[IPVS_CMD_ATTR_SERVICE],
3537                                        need_full_svc, &svc);
3538         if (ret)
3539                 goto out;
3540
3541         /* Unless we're adding a new service, the service must already exist */
3542         if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3543                 ret = -ESRCH;
3544                 goto out;
3545         }
3546
3547         /* Destination commands require a valid destination argument. For
3548          * adding / editing a destination, we need a full destination
3549          * specification. */
3550         if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3551             cmd == IPVS_CMD_DEL_DEST) {
3552                 if (cmd != IPVS_CMD_DEL_DEST)
3553                         need_full_dest = 1;
3554
3555                 ret = ip_vs_genl_parse_dest(&udest,
3556                                             info->attrs[IPVS_CMD_ATTR_DEST],
3557                                             need_full_dest);
3558                 if (ret)
3559                         goto out;
3560
3561                 /* Old protocols did not allow the user to specify address
3562                  * family, so we set it to zero instead.  We also didn't
3563                  * allow heterogeneous pools in the old code, so it's safe
3564                  * to assume that this will have the same address family as
3565                  * the service.
3566                  */
3567                 if (udest.af == 0)
3568                         udest.af = svc->af;
3569
3570                 if (udest.af != svc->af && cmd != IPVS_CMD_DEL_DEST) {
3571                         /* The synchronization protocol is incompatible
3572                          * with mixed family services
3573                          */
3574                         if (ipvs->sync_state) {
3575                                 ret = -EINVAL;
3576                                 goto out;
3577                         }
3578
3579                         /* Which connection types do we support? */
3580                         switch (udest.conn_flags) {
3581                         case IP_VS_CONN_F_TUNNEL:
3582                                 /* We are able to forward this */
3583                                 break;
3584                         default:
3585                                 ret = -EINVAL;
3586                                 goto out;
3587                         }
3588                 }
3589         }
3590
3591         switch (cmd) {
3592         case IPVS_CMD_NEW_SERVICE:
3593                 if (svc == NULL)
3594                         ret = ip_vs_add_service(ipvs, &usvc, &svc);
3595                 else
3596                         ret = -EEXIST;
3597                 break;
3598         case IPVS_CMD_SET_SERVICE:
3599                 ret = ip_vs_edit_service(svc, &usvc);
3600                 break;
3601         case IPVS_CMD_DEL_SERVICE:
3602                 ret = ip_vs_del_service(svc);
3603                 /* do not use svc, it can be freed */
3604                 break;
3605         case IPVS_CMD_NEW_DEST:
3606                 ret = ip_vs_add_dest(svc, &udest);
3607                 break;
3608         case IPVS_CMD_SET_DEST:
3609                 ret = ip_vs_edit_dest(svc, &udest);
3610                 break;
3611         case IPVS_CMD_DEL_DEST:
3612                 ret = ip_vs_del_dest(svc, &udest);
3613                 break;
3614         case IPVS_CMD_ZERO:
3615                 ret = ip_vs_zero_service(svc);
3616                 break;
3617         default:
3618                 ret = -EINVAL;
3619         }
3620
3621 out:
3622         mutex_unlock(&__ip_vs_mutex);
3623
3624         return ret;
3625 }
3626
3627 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3628 {
3629         struct sk_buff *msg;
3630         void *reply;
3631         int ret, cmd, reply_cmd;
3632         struct net *net = sock_net(skb->sk);
3633         struct netns_ipvs *ipvs = net_ipvs(net);
3634
3635         cmd = info->genlhdr->cmd;
3636
3637         if (cmd == IPVS_CMD_GET_SERVICE)
3638                 reply_cmd = IPVS_CMD_NEW_SERVICE;
3639         else if (cmd == IPVS_CMD_GET_INFO)
3640                 reply_cmd = IPVS_CMD_SET_INFO;
3641         else if (cmd == IPVS_CMD_GET_CONFIG)
3642                 reply_cmd = IPVS_CMD_SET_CONFIG;
3643         else {
3644                 pr_err("unknown Generic Netlink command\n");
3645                 return -EINVAL;
3646         }
3647
3648         msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3649         if (!msg)
3650                 return -ENOMEM;
3651
3652         mutex_lock(&__ip_vs_mutex);
3653
3654         reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3655         if (reply == NULL)
3656                 goto nla_put_failure;
3657
3658         switch (cmd) {
3659         case IPVS_CMD_GET_SERVICE:
3660         {
3661                 struct ip_vs_service *svc;
3662
3663                 svc = ip_vs_genl_find_service(ipvs,
3664                                               info->attrs[IPVS_CMD_ATTR_SERVICE]);
3665                 if (IS_ERR(svc)) {
3666                         ret = PTR_ERR(svc);
3667                         goto out_err;
3668                 } else if (svc) {
3669                         ret = ip_vs_genl_fill_service(msg, svc);
3670                         if (ret)
3671                                 goto nla_put_failure;
3672                 } else {
3673                         ret = -ESRCH;
3674                         goto out_err;
3675                 }
3676
3677                 break;
3678         }
3679
3680         case IPVS_CMD_GET_CONFIG:
3681         {
3682                 struct ip_vs_timeout_user t;
3683
3684                 __ip_vs_get_timeouts(ipvs, &t);
3685 #ifdef CONFIG_IP_VS_PROTO_TCP
3686                 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3687                                 t.tcp_timeout) ||
3688                     nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3689                                 t.tcp_fin_timeout))
3690                         goto nla_put_failure;
3691 #endif
3692 #ifdef CONFIG_IP_VS_PROTO_UDP
3693                 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3694                         goto nla_put_failure;
3695 #endif
3696
3697                 break;
3698         }
3699
3700         case IPVS_CMD_GET_INFO:
3701                 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3702                                 IP_VS_VERSION_CODE) ||
3703                     nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3704                                 ip_vs_conn_tab_size))
3705                         goto nla_put_failure;
3706                 break;
3707         }
3708
3709         genlmsg_end(msg, reply);
3710         ret = genlmsg_reply(msg, info);
3711         goto out;
3712
3713 nla_put_failure:
3714         pr_err("not enough space in Netlink message\n");
3715         ret = -EMSGSIZE;
3716
3717 out_err:
3718         nlmsg_free(msg);
3719 out:
3720         mutex_unlock(&__ip_vs_mutex);
3721
3722         return ret;
3723 }
3724
3725
3726 static const struct genl_ops ip_vs_genl_ops[] = {
3727         {
3728                 .cmd    = IPVS_CMD_NEW_SERVICE,
3729                 .flags  = GENL_ADMIN_PERM,
3730                 .policy = ip_vs_cmd_policy,
3731                 .doit   = ip_vs_genl_set_cmd,
3732         },
3733         {
3734                 .cmd    = IPVS_CMD_SET_SERVICE,
3735                 .flags  = GENL_ADMIN_PERM,
3736                 .policy = ip_vs_cmd_policy,
3737                 .doit   = ip_vs_genl_set_cmd,
3738         },
3739         {
3740                 .cmd    = IPVS_CMD_DEL_SERVICE,
3741                 .flags  = GENL_ADMIN_PERM,
3742                 .policy = ip_vs_cmd_policy,
3743                 .doit   = ip_vs_genl_set_cmd,
3744         },
3745         {
3746                 .cmd    = IPVS_CMD_GET_SERVICE,
3747                 .flags  = GENL_ADMIN_PERM,
3748                 .doit   = ip_vs_genl_get_cmd,
3749                 .dumpit = ip_vs_genl_dump_services,
3750                 .policy = ip_vs_cmd_policy,
3751         },
3752         {
3753                 .cmd    = IPVS_CMD_NEW_DEST,
3754                 .flags  = GENL_ADMIN_PERM,
3755                 .policy = ip_vs_cmd_policy,
3756                 .doit   = ip_vs_genl_set_cmd,
3757         },
3758         {
3759                 .cmd    = IPVS_CMD_SET_DEST,
3760                 .flags  = GENL_ADMIN_PERM,
3761                 .policy = ip_vs_cmd_policy,
3762                 .doit   = ip_vs_genl_set_cmd,
3763         },
3764         {
3765                 .cmd    = IPVS_CMD_DEL_DEST,
3766                 .flags  = GENL_ADMIN_PERM,
3767                 .policy = ip_vs_cmd_policy,
3768                 .doit   = ip_vs_genl_set_cmd,
3769         },
3770         {
3771                 .cmd    = IPVS_CMD_GET_DEST,
3772                 .flags  = GENL_ADMIN_PERM,
3773                 .policy = ip_vs_cmd_policy,
3774                 .dumpit = ip_vs_genl_dump_dests,
3775         },
3776         {
3777                 .cmd    = IPVS_CMD_NEW_DAEMON,
3778                 .flags  = GENL_ADMIN_PERM,
3779                 .policy = ip_vs_cmd_policy,
3780                 .doit   = ip_vs_genl_set_daemon,
3781         },
3782         {
3783                 .cmd    = IPVS_CMD_DEL_DAEMON,
3784                 .flags  = GENL_ADMIN_PERM,
3785                 .policy = ip_vs_cmd_policy,
3786                 .doit   = ip_vs_genl_set_daemon,
3787         },
3788         {
3789                 .cmd    = IPVS_CMD_GET_DAEMON,
3790                 .flags  = GENL_ADMIN_PERM,
3791                 .dumpit = ip_vs_genl_dump_daemons,
3792         },
3793         {
3794                 .cmd    = IPVS_CMD_SET_CONFIG,
3795                 .flags  = GENL_ADMIN_PERM,
3796                 .policy = ip_vs_cmd_policy,
3797                 .doit   = ip_vs_genl_set_cmd,
3798         },
3799         {
3800                 .cmd    = IPVS_CMD_GET_CONFIG,
3801                 .flags  = GENL_ADMIN_PERM,
3802                 .doit   = ip_vs_genl_get_cmd,
3803         },
3804         {
3805                 .cmd    = IPVS_CMD_GET_INFO,
3806                 .flags  = GENL_ADMIN_PERM,
3807                 .doit   = ip_vs_genl_get_cmd,
3808         },
3809         {
3810                 .cmd    = IPVS_CMD_ZERO,
3811                 .flags  = GENL_ADMIN_PERM,
3812                 .policy = ip_vs_cmd_policy,
3813                 .doit   = ip_vs_genl_set_cmd,
3814         },
3815         {
3816                 .cmd    = IPVS_CMD_FLUSH,
3817                 .flags  = GENL_ADMIN_PERM,
3818                 .doit   = ip_vs_genl_set_cmd,
3819         },
3820 };
3821
3822 static int __init ip_vs_genl_register(void)
3823 {
3824         return genl_register_family_with_ops(&ip_vs_genl_family,
3825                                              ip_vs_genl_ops);
3826 }
3827
3828 static void ip_vs_genl_unregister(void)
3829 {
3830         genl_unregister_family(&ip_vs_genl_family);
3831 }
3832
3833 /* End of Generic Netlink interface definitions */
3834
3835 /*
3836  * per netns intit/exit func.
3837  */
3838 #ifdef CONFIG_SYSCTL
3839 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs)
3840 {
3841         struct net *net = ipvs->net;
3842         int idx;
3843         struct ctl_table *tbl;
3844
3845         atomic_set(&ipvs->dropentry, 0);
3846         spin_lock_init(&ipvs->dropentry_lock);
3847         spin_lock_init(&ipvs->droppacket_lock);
3848         spin_lock_init(&ipvs->securetcp_lock);
3849
3850         if (!net_eq(net, &init_net)) {
3851                 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3852                 if (tbl == NULL)
3853                         return -ENOMEM;
3854
3855                 /* Don't export sysctls to unprivileged users */
3856                 if (net->user_ns != &init_user_ns)
3857                         tbl[0].procname = NULL;
3858         } else
3859                 tbl = vs_vars;
3860         /* Initialize sysctl defaults */
3861         for (idx = 0; idx < ARRAY_SIZE(vs_vars); idx++) {
3862                 if (tbl[idx].proc_handler == proc_do_defense_mode)
3863                         tbl[idx].extra2 = ipvs;
3864         }
3865         idx = 0;
3866         ipvs->sysctl_amemthresh = 1024;
3867         tbl[idx++].data = &ipvs->sysctl_amemthresh;
3868         ipvs->sysctl_am_droprate = 10;
3869         tbl[idx++].data = &ipvs->sysctl_am_droprate;
3870         tbl[idx++].data = &ipvs->sysctl_drop_entry;
3871         tbl[idx++].data = &ipvs->sysctl_drop_packet;
3872 #ifdef CONFIG_IP_VS_NFCT
3873         tbl[idx++].data = &ipvs->sysctl_conntrack;
3874 #endif
3875         tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3876         ipvs->sysctl_snat_reroute = 1;
3877         tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3878         ipvs->sysctl_sync_ver = 1;
3879         tbl[idx++].data = &ipvs->sysctl_sync_ver;
3880         ipvs->sysctl_sync_ports = 1;
3881         tbl[idx++].data = &ipvs->sysctl_sync_ports;
3882         tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
3883         ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3884         tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3885         ipvs->sysctl_sync_sock_size = 0;
3886         tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3887         tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3888         tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3889         tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
3890         tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
3891         tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3892         ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3893         ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3894         tbl[idx].data = &ipvs->sysctl_sync_threshold;
3895         tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3896         ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3897         tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3898         ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3899         tbl[idx++].data = &ipvs->sysctl_sync_retries;
3900         tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3901         ipvs->sysctl_pmtu_disc = 1;
3902         tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
3903         tbl[idx++].data = &ipvs->sysctl_backup_only;
3904         ipvs->sysctl_conn_reuse_mode = 1;
3905         tbl[idx++].data = &ipvs->sysctl_conn_reuse_mode;
3906         tbl[idx++].data = &ipvs->sysctl_schedule_icmp;
3907         tbl[idx++].data = &ipvs->sysctl_ignore_tunneled;
3908
3909         ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3910         if (ipvs->sysctl_hdr == NULL) {
3911                 if (!net_eq(net, &init_net))
3912                         kfree(tbl);
3913                 return -ENOMEM;
3914         }
3915         ip_vs_start_estimator(ipvs, &ipvs->tot_stats);
3916         ipvs->sysctl_tbl = tbl;
3917         /* Schedule defense work */
3918         INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3919         schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3920
3921         return 0;
3922 }
3923
3924 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs)
3925 {
3926         struct net *net = ipvs->net;
3927
3928         cancel_delayed_work_sync(&ipvs->defense_work);
3929         cancel_work_sync(&ipvs->defense_work.work);
3930         unregister_net_sysctl_table(ipvs->sysctl_hdr);
3931         ip_vs_stop_estimator(ipvs, &ipvs->tot_stats);
3932
3933         if (!net_eq(net, &init_net))
3934                 kfree(ipvs->sysctl_tbl);
3935 }
3936
3937 #else
3938
3939 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs) { return 0; }
3940 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs) { }
3941
3942 #endif
3943
3944 static struct notifier_block ip_vs_dst_notifier = {
3945         .notifier_call = ip_vs_dst_event,
3946 };
3947
3948 int __net_init ip_vs_control_net_init(struct netns_ipvs *ipvs)
3949 {
3950         struct net *net = ipvs->net;
3951         int i, idx;
3952
3953         /* Initialize rs_table */
3954         for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
3955                 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
3956
3957         INIT_LIST_HEAD(&ipvs->dest_trash);
3958         spin_lock_init(&ipvs->dest_trash_lock);
3959         setup_timer(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire,
3960                     (unsigned long) ipvs);
3961         atomic_set(&ipvs->ftpsvc_counter, 0);
3962         atomic_set(&ipvs->nullsvc_counter, 0);
3963
3964         /* procfs stats */
3965         ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
3966         if (!ipvs->tot_stats.cpustats)
3967                 return -ENOMEM;
3968
3969         for_each_possible_cpu(i) {
3970                 struct ip_vs_cpu_stats *ipvs_tot_stats;
3971                 ipvs_tot_stats = per_cpu_ptr(ipvs->tot_stats.cpustats, i);
3972                 u64_stats_init(&ipvs_tot_stats->syncp);
3973         }
3974
3975         spin_lock_init(&ipvs->tot_stats.lock);
3976
3977         proc_create("ip_vs", 0, net->proc_net, &ip_vs_info_fops);
3978         proc_create("ip_vs_stats", 0, net->proc_net, &ip_vs_stats_fops);
3979         proc_create("ip_vs_stats_percpu", 0, net->proc_net,
3980                     &ip_vs_stats_percpu_fops);
3981
3982         if (ip_vs_control_net_init_sysctl(ipvs))
3983                 goto err;
3984
3985         return 0;
3986
3987 err:
3988         free_percpu(ipvs->tot_stats.cpustats);
3989         return -ENOMEM;
3990 }
3991
3992 void __net_exit ip_vs_control_net_cleanup(struct netns_ipvs *ipvs)
3993 {
3994         struct net *net = ipvs->net;
3995
3996         ip_vs_trash_cleanup(ipvs);
3997         ip_vs_control_net_cleanup_sysctl(ipvs);
3998         remove_proc_entry("ip_vs_stats_percpu", net->proc_net);
3999         remove_proc_entry("ip_vs_stats", net->proc_net);
4000         remove_proc_entry("ip_vs", net->proc_net);
4001         free_percpu(ipvs->tot_stats.cpustats);
4002 }
4003
4004 int __init ip_vs_register_nl_ioctl(void)
4005 {
4006         int ret;
4007
4008         ret = nf_register_sockopt(&ip_vs_sockopts);
4009         if (ret) {
4010                 pr_err("cannot register sockopt.\n");
4011                 goto err_sock;
4012         }
4013
4014         ret = ip_vs_genl_register();
4015         if (ret) {
4016                 pr_err("cannot register Generic Netlink interface.\n");
4017                 goto err_genl;
4018         }
4019         return 0;
4020
4021 err_genl:
4022         nf_unregister_sockopt(&ip_vs_sockopts);
4023 err_sock:
4024         return ret;
4025 }
4026
4027 void ip_vs_unregister_nl_ioctl(void)
4028 {
4029         ip_vs_genl_unregister();
4030         nf_unregister_sockopt(&ip_vs_sockopts);
4031 }
4032
4033 int __init ip_vs_control_init(void)
4034 {
4035         int idx;
4036         int ret;
4037
4038         EnterFunction(2);
4039
4040         /* Initialize svc_table, ip_vs_svc_fwm_table */
4041         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
4042                 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
4043                 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
4044         }
4045
4046         smp_wmb();      /* Do we really need it now ? */
4047
4048         ret = register_netdevice_notifier(&ip_vs_dst_notifier);
4049         if (ret < 0)
4050                 return ret;
4051
4052         LeaveFunction(2);
4053         return 0;
4054 }
4055
4056
4057 void ip_vs_control_cleanup(void)
4058 {
4059         EnterFunction(2);
4060         unregister_netdevice_notifier(&ip_vs_dst_notifier);
4061         LeaveFunction(2);
4062 }