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 <linux/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 %pS\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 %pS\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 && protocol == IPPROTO_TCP &&
 430             atomic_read(&ipvs->ftpsvc_counter) &&
 431             (vport == FTPDATA || ntohs(vport) >= inet_prot_sock(ipvs->net))) {
 432                 /*
 433                  * Check if ftp service entry exists, the packet
 434                  * might belong to FTP data connections.
 435                  */
 436                 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, FTPPORT);
 437         }
 438
 439         if (svc == NULL
 440             && atomic_read(&ipvs->nullsvc_counter)) {
 441                 /*
 442                  * Check if the catch-all port (port zero) exists
 443                  */
 444                 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, 0);
 445         }
 446
 447   out:
 448         IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
 449                       fwmark, ip_vs_proto_name(protocol),
 450                       IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
 451                       svc ? "hit" : "not hit");
 452
 453         return svc;
 454 }
 455
 456
 457 static inline void
 458 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
 459 {
 460         atomic_inc(&svc->refcnt);
 461         rcu_assign_pointer(dest->svc, svc);
 462 }
 463
 464 static void ip_vs_service_free(struct ip_vs_service *svc)
 465 {
 466         free_percpu(svc->stats.cpustats);
 467         kfree(svc);
 468 }
 469
 470 static void ip_vs_service_rcu_free(struct rcu_head *head)
 471 {
 472         struct ip_vs_service *svc;
 473
 474         svc = container_of(head, struct ip_vs_service, rcu_head);
 475         ip_vs_service_free(svc);
 476 }
 477
 478 static void __ip_vs_svc_put(struct ip_vs_service *svc, bool do_delay)
 479 {
 480         if (atomic_dec_and_test(&svc->refcnt)) {
 481                 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n",
 482                               svc->fwmark,
 483                               IP_VS_DBG_ADDR(svc->af, &svc->addr),
 484                               ntohs(svc->port));
 485                 if (do_delay)
 486                         call_rcu(&svc->rcu_head, ip_vs_service_rcu_free);
 487                 else
 488                         ip_vs_service_free(svc);
 489         }
 490 }
 491
 492
 493 /*
 494  *      Returns hash value for real service
 495  */
 496 static inline unsigned int ip_vs_rs_hashkey(int af,
 497                                             const union nf_inet_addr *addr,
 498                                             __be16 port)
 499 {
 500         register unsigned int porth = ntohs(port);
 501         __be32 addr_fold = addr->ip;
 502
 503 #ifdef CONFIG_IP_VS_IPV6
 504         if (af == AF_INET6)
 505                 addr_fold = addr->ip6[0]^addr->ip6[1]^
 506                             addr->ip6[2]^addr->ip6[3];
 507 #endif
 508
 509         return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
 510                 & IP_VS_RTAB_MASK;
 511 }
 512
 513 /* Hash ip_vs_dest in rs_table by <proto,addr,port>. */
 514 static void ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
 515 {
 516         unsigned int hash;
 517
 518         if (dest->in_rs_table)
 519                 return;
 520
 521         /*
 522          *      Hash by proto,addr,port,
 523          *      which are the parameters of the real service.
 524          */
 525         hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
 526
 527         hlist_add_head_rcu(&dest->d_list, &ipvs->rs_table[hash]);
 528         dest->in_rs_table = 1;
 529 }
 530
 531 /* Unhash ip_vs_dest from rs_table. */
 532 static void ip_vs_rs_unhash(struct ip_vs_dest *dest)
 533 {
 534         /*
 535          * Remove it from the rs_table table.
 536          */
 537         if (dest->in_rs_table) {
 538                 hlist_del_rcu(&dest->d_list);
 539                 dest->in_rs_table = 0;
 540         }
 541 }
 542
 543 /* Check if real service by <proto,addr,port> is present */
 544 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
 545                             const union nf_inet_addr *daddr, __be16 dport)
 546 {
 547         unsigned int hash;
 548         struct ip_vs_dest *dest;
 549
 550         /* Check for "full" addressed entries */
 551         hash = ip_vs_rs_hashkey(af, daddr, dport);
 552
 553         hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
 554                 if (dest->port == dport &&
 555                     dest->af == af &&
 556                     ip_vs_addr_equal(af, &dest->addr, daddr) &&
 557                     (dest->protocol == protocol || dest->vfwmark)) {
 558                         /* HIT */
 559                         return true;
 560                 }
 561         }
 562
 563         return false;
 564 }
 565
 566 /* Find real service record by <proto,addr,port>.
 567  * In case of multiple records with the same <proto,addr,port>, only
 568  * the first found record is returned.
 569  *
 570  * To be called under RCU lock.
 571  */
 572 struct ip_vs_dest *ip_vs_find_real_service(struct netns_ipvs *ipvs, int af,
 573                                            __u16 protocol,
 574                                            const union nf_inet_addr *daddr,
 575                                            __be16 dport)
 576 {
 577         unsigned int hash;
 578         struct ip_vs_dest *dest;
 579
 580         /* Check for "full" addressed entries */
 581         hash = ip_vs_rs_hashkey(af, daddr, dport);
 582
 583         hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
 584                 if (dest->port == dport &&
 585                     dest->af == af &&
 586                     ip_vs_addr_equal(af, &dest->addr, daddr) &&
 587                         (dest->protocol == protocol || dest->vfwmark)) {
 588                         /* HIT */
 589                         return dest;
 590                 }
 591         }
 592
 593         return NULL;
 594 }
 595
 596 /* Lookup destination by {addr,port} in the given service
 597  * Called under RCU lock.
 598  */
 599 static struct ip_vs_dest *
 600 ip_vs_lookup_dest(struct ip_vs_service *svc, int dest_af,
 601                   const union nf_inet_addr *daddr, __be16 dport)
 602 {
 603         struct ip_vs_dest *dest;
 604
 605         /*
 606          * Find the destination for the given service
 607          */
 608         list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
 609                 if ((dest->af == dest_af) &&
 610                     ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
 611                     (dest->port == dport)) {
 612                         /* HIT */
 613                         return dest;
 614                 }
 615         }
 616
 617         return NULL;
 618 }
 619
 620 /*
 621  * Find destination by {daddr,dport,vaddr,protocol}
 622  * Created to be used in ip_vs_process_message() in
 623  * the backup synchronization daemon. It finds the
 624  * destination to be bound to the received connection
 625  * on the backup.
 626  * Called under RCU lock, no refcnt is returned.
 627  */
 628 struct ip_vs_dest *ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
 629                                    const union nf_inet_addr *daddr,
 630                                    __be16 dport,
 631                                    const union nf_inet_addr *vaddr,
 632                                    __be16 vport, __u16 protocol, __u32 fwmark,
 633                                    __u32 flags)
 634 {
 635         struct ip_vs_dest *dest;
 636         struct ip_vs_service *svc;
 637         __be16 port = dport;
 638
 639         svc = ip_vs_service_find(ipvs, svc_af, fwmark, protocol, vaddr, vport);
 640         if (!svc)
 641                 return NULL;
 642         if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
 643                 port = 0;
 644         dest = ip_vs_lookup_dest(svc, dest_af, daddr, port);
 645         if (!dest)
 646                 dest = ip_vs_lookup_dest(svc, dest_af, daddr, port ^ dport);
 647         return dest;
 648 }
 649
 650 void ip_vs_dest_dst_rcu_free(struct rcu_head *head)
 651 {
 652         struct ip_vs_dest_dst *dest_dst = container_of(head,
 653                                                        struct ip_vs_dest_dst,
 654                                                        rcu_head);
 655
 656         dst_release(dest_dst->dst_cache);
 657         kfree(dest_dst);
 658 }
 659
 660 /* Release dest_dst and dst_cache for dest in user context */
 661 static void __ip_vs_dst_cache_reset(struct ip_vs_dest *dest)
 662 {
 663         struct ip_vs_dest_dst *old;
 664
 665         old = rcu_dereference_protected(dest->dest_dst, 1);
 666         if (old) {
 667                 RCU_INIT_POINTER(dest->dest_dst, NULL);
 668                 call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
 669         }
 670 }
 671
 672 /*
 673  *  Lookup dest by {svc,addr,port} in the destination trash.
 674  *  The destination trash is used to hold the destinations that are removed
 675  *  from the service table but are still referenced by some conn entries.
 676  *  The reason to add the destination trash is when the dest is temporary
 677  *  down (either by administrator or by monitor program), the dest can be
 678  *  picked back from the trash, the remaining connections to the dest can
 679  *  continue, and the counting information of the dest is also useful for
 680  *  scheduling.
 681  */
 682 static struct ip_vs_dest *
 683 ip_vs_trash_get_dest(struct ip_vs_service *svc, int dest_af,
 684                      const union nf_inet_addr *daddr, __be16 dport)
 685 {
 686         struct ip_vs_dest *dest;
 687         struct netns_ipvs *ipvs = svc->ipvs;
 688
 689         /*
 690          * Find the destination in trash
 691          */
 692         spin_lock_bh(&ipvs->dest_trash_lock);
 693         list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
 694                 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
 695                               "dest->refcnt=%d\n",
 696                               dest->vfwmark,
 697                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
 698                               ntohs(dest->port),
 699                               refcount_read(&dest->refcnt));
 700                 if (dest->af == dest_af &&
 701                     ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
 702                     dest->port == dport &&
 703                     dest->vfwmark == svc->fwmark &&
 704                     dest->protocol == svc->protocol &&
 705                     (svc->fwmark ||
 706                      (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
 707                       dest->vport == svc->port))) {
 708                         /* HIT */
 709                         list_del(&dest->t_list);
 710                         goto out;
 711                 }
 712         }
 713
 714         dest = NULL;
 715
 716 out:
 717         spin_unlock_bh(&ipvs->dest_trash_lock);
 718
 719         return dest;
 720 }
 721
 722 static void ip_vs_dest_free(struct ip_vs_dest *dest)
 723 {
 724         struct ip_vs_service *svc = rcu_dereference_protected(dest->svc, 1);
 725
 726         __ip_vs_dst_cache_reset(dest);
 727         __ip_vs_svc_put(svc, false);
 728         free_percpu(dest->stats.cpustats);
 729         ip_vs_dest_put_and_free(dest);
 730 }
 731
 732 /*
 733  *  Clean up all the destinations in the trash
 734  *  Called by the ip_vs_control_cleanup()
 735  *
 736  *  When the ip_vs_control_clearup is activated by ipvs module exit,
 737  *  the service tables must have been flushed and all the connections
 738  *  are expired, and the refcnt of each destination in the trash must
 739  *  be 1, so we simply release them here.
 740  */
 741 static void ip_vs_trash_cleanup(struct netns_ipvs *ipvs)
 742 {
 743         struct ip_vs_dest *dest, *nxt;
 744
 745         del_timer_sync(&ipvs->dest_trash_timer);
 746         /* No need to use dest_trash_lock */
 747         list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, t_list) {
 748                 list_del(&dest->t_list);
 749                 ip_vs_dest_free(dest);
 750         }
 751 }
 752
 753 static void
 754 ip_vs_copy_stats(struct ip_vs_kstats *dst, struct ip_vs_stats *src)
 755 {
 756 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->kstats.c - src->kstats0.c
 757
 758         spin_lock_bh(&src->lock);
 759
 760         IP_VS_SHOW_STATS_COUNTER(conns);
 761         IP_VS_SHOW_STATS_COUNTER(inpkts);
 762         IP_VS_SHOW_STATS_COUNTER(outpkts);
 763         IP_VS_SHOW_STATS_COUNTER(inbytes);
 764         IP_VS_SHOW_STATS_COUNTER(outbytes);
 765
 766         ip_vs_read_estimator(dst, src);
 767
 768         spin_unlock_bh(&src->lock);
 769 }
 770
 771 static void
 772 ip_vs_export_stats_user(struct ip_vs_stats_user *dst, struct ip_vs_kstats *src)
 773 {
 774         dst->conns = (u32)src->conns;
 775         dst->inpkts = (u32)src->inpkts;
 776         dst->outpkts = (u32)src->outpkts;
 777         dst->inbytes = src->inbytes;
 778         dst->outbytes = src->outbytes;
 779         dst->cps = (u32)src->cps;
 780         dst->inpps = (u32)src->inpps;
 781         dst->outpps = (u32)src->outpps;
 782         dst->inbps = (u32)src->inbps;
 783         dst->outbps = (u32)src->outbps;
 784 }
 785
 786 static void
 787 ip_vs_zero_stats(struct ip_vs_stats *stats)
 788 {
 789         spin_lock_bh(&stats->lock);
 790
 791         /* get current counters as zero point, rates are zeroed */
 792
 793 #define IP_VS_ZERO_STATS_COUNTER(c) stats->kstats0.c = stats->kstats.c
 794
 795         IP_VS_ZERO_STATS_COUNTER(conns);
 796         IP_VS_ZERO_STATS_COUNTER(inpkts);
 797         IP_VS_ZERO_STATS_COUNTER(outpkts);
 798         IP_VS_ZERO_STATS_COUNTER(inbytes);
 799         IP_VS_ZERO_STATS_COUNTER(outbytes);
 800
 801         ip_vs_zero_estimator(stats);
 802
 803         spin_unlock_bh(&stats->lock);
 804 }
 805
 806 /*
 807  *      Update a destination in the given service
 808  */
 809 static void
 810 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
 811                     struct ip_vs_dest_user_kern *udest, int add)
 812 {
 813         struct netns_ipvs *ipvs = svc->ipvs;
 814         struct ip_vs_service *old_svc;
 815         struct ip_vs_scheduler *sched;
 816         int conn_flags;
 817
 818         /* We cannot modify an address and change the address family */
 819         BUG_ON(!add && udest->af != dest->af);
 820
 821         if (add && udest->af != svc->af)
 822                 ipvs->mixed_address_family_dests++;
 823
 824         /* set the weight and the flags */
 825         atomic_set(&dest->weight, udest->weight);
 826         conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
 827         conn_flags |= IP_VS_CONN_F_INACTIVE;
 828
 829         /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
 830         if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
 831                 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
 832         } else {
 833                 /*
 834                  *    Put the real service in rs_table if not present.
 835                  *    For now only for NAT!
 836                  */
 837                 ip_vs_rs_hash(ipvs, dest);
 838         }
 839         atomic_set(&dest->conn_flags, conn_flags);
 840
 841         /* bind the service */
 842         old_svc = rcu_dereference_protected(dest->svc, 1);
 843         if (!old_svc) {
 844                 __ip_vs_bind_svc(dest, svc);
 845         } else {
 846                 if (old_svc != svc) {
 847                         ip_vs_zero_stats(&dest->stats);
 848                         __ip_vs_bind_svc(dest, svc);
 849                         __ip_vs_svc_put(old_svc, true);
 850                 }
 851         }
 852
 853         /* set the dest status flags */
 854         dest->flags |= IP_VS_DEST_F_AVAILABLE;
 855
 856         if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
 857                 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
 858         dest->u_threshold = udest->u_threshold;
 859         dest->l_threshold = udest->l_threshold;
 860
 861         dest->af = udest->af;
 862
 863         spin_lock_bh(&dest->dst_lock);
 864         __ip_vs_dst_cache_reset(dest);
 865         spin_unlock_bh(&dest->dst_lock);
 866
 867         if (add) {
 868                 ip_vs_start_estimator(svc->ipvs, &dest->stats);
 869                 list_add_rcu(&dest->n_list, &svc->destinations);
 870                 svc->num_dests++;
 871                 sched = rcu_dereference_protected(svc->scheduler, 1);
 872                 if (sched && sched->add_dest)
 873                         sched->add_dest(svc, dest);
 874         } else {
 875                 sched = rcu_dereference_protected(svc->scheduler, 1);
 876                 if (sched && sched->upd_dest)
 877                         sched->upd_dest(svc, dest);
 878         }
 879 }
 880
 881
 882 /*
 883  *      Create a destination for the given service
 884  */
 885 static int
 886 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
 887                struct ip_vs_dest **dest_p)
 888 {
 889         struct ip_vs_dest *dest;
 890         unsigned int atype, i;
 891
 892         EnterFunction(2);
 893
 894 #ifdef CONFIG_IP_VS_IPV6
 895         if (udest->af == AF_INET6) {
 896                 atype = ipv6_addr_type(&udest->addr.in6);
 897                 if ((!(atype & IPV6_ADDR_UNICAST) ||
 898                         atype & IPV6_ADDR_LINKLOCAL) &&
 899                         !__ip_vs_addr_is_local_v6(svc->ipvs->net, &udest->addr.in6))
 900                         return -EINVAL;
 901         } else
 902 #endif
 903         {
 904                 atype = inet_addr_type(svc->ipvs->net, udest->addr.ip);
 905                 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
 906                         return -EINVAL;
 907         }
 908
 909         dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
 910         if (dest == NULL)
 911                 return -ENOMEM;
 912
 913         dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
 914         if (!dest->stats.cpustats)
 915                 goto err_alloc;
 916
 917         for_each_possible_cpu(i) {
 918                 struct ip_vs_cpu_stats *ip_vs_dest_stats;
 919                 ip_vs_dest_stats = per_cpu_ptr(dest->stats.cpustats, i);
 920                 u64_stats_init(&ip_vs_dest_stats->syncp);
 921         }
 922
 923         dest->af = udest->af;
 924         dest->protocol = svc->protocol;
 925         dest->vaddr = svc->addr;
 926         dest->vport = svc->port;
 927         dest->vfwmark = svc->fwmark;
 928         ip_vs_addr_copy(udest->af, &dest->addr, &udest->addr);
 929         dest->port = udest->port;
 930
 931         atomic_set(&dest->activeconns, 0);
 932         atomic_set(&dest->inactconns, 0);
 933         atomic_set(&dest->persistconns, 0);
 934         refcount_set(&dest->refcnt, 1);
 935
 936         INIT_HLIST_NODE(&dest->d_list);
 937         spin_lock_init(&dest->dst_lock);
 938         spin_lock_init(&dest->stats.lock);
 939         __ip_vs_update_dest(svc, dest, udest, 1);
 940
 941         *dest_p = dest;
 942
 943         LeaveFunction(2);
 944         return 0;
 945
 946 err_alloc:
 947         kfree(dest);
 948         return -ENOMEM;
 949 }
 950
 951
 952 /*
 953  *      Add a destination into an existing service
 954  */
 955 static int
 956 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
 957 {
 958         struct ip_vs_dest *dest;
 959         union nf_inet_addr daddr;
 960         __be16 dport = udest->port;
 961         int ret;
 962
 963         EnterFunction(2);
 964
 965         if (udest->weight < 0) {
 966                 pr_err("%s(): server weight less than zero\n", __func__);
 967                 return -ERANGE;
 968         }
 969
 970         if (udest->l_threshold > udest->u_threshold) {
 971                 pr_err("%s(): lower threshold is higher than upper threshold\n",
 972                         __func__);
 973                 return -ERANGE;
 974         }
 975
 976         ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
 977
 978         /* We use function that requires RCU lock */
 979         rcu_read_lock();
 980         dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
 981         rcu_read_unlock();
 982
 983         if (dest != NULL) {
 984                 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
 985                 return -EEXIST;
 986         }
 987
 988         /*
 989          * Check if the dest already exists in the trash and
 990          * is from the same service
 991          */
 992         dest = ip_vs_trash_get_dest(svc, udest->af, &daddr, dport);
 993
 994         if (dest != NULL) {
 995                 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
 996                               "dest->refcnt=%d, service %u/%s:%u\n",
 997                               IP_VS_DBG_ADDR(udest->af, &daddr), ntohs(dport),
 998                               refcount_read(&dest->refcnt),
 999                               dest->vfwmark,
1000                               IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
1001                               ntohs(dest->vport));
1002
1003                 __ip_vs_update_dest(svc, dest, udest, 1);
1004                 ret = 0;
1005         } else {
1006                 /*
1007                  * Allocate and initialize the dest structure
1008                  */
1009                 ret = ip_vs_new_dest(svc, udest, &dest);
1010         }
1011         LeaveFunction(2);
1012
1013         return ret;
1014 }
1015
1016
1017 /*
1018  *      Edit a destination in the given service
1019  */
1020 static int
1021 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1022 {
1023         struct ip_vs_dest *dest;
1024         union nf_inet_addr daddr;
1025         __be16 dport = udest->port;
1026
1027         EnterFunction(2);
1028
1029         if (udest->weight < 0) {
1030                 pr_err("%s(): server weight less than zero\n", __func__);
1031                 return -ERANGE;
1032         }
1033
1034         if (udest->l_threshold > udest->u_threshold) {
1035                 pr_err("%s(): lower threshold is higher than upper threshold\n",
1036                         __func__);
1037                 return -ERANGE;
1038         }
1039
1040         ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
1041
1042         /* We use function that requires RCU lock */
1043         rcu_read_lock();
1044         dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
1045         rcu_read_unlock();
1046
1047         if (dest == NULL) {
1048                 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1049                 return -ENOENT;
1050         }
1051
1052         __ip_vs_update_dest(svc, dest, udest, 0);
1053         LeaveFunction(2);
1054
1055         return 0;
1056 }
1057
1058 /*
1059  *      Delete a destination (must be already unlinked from the service)
1060  */
1061 static void __ip_vs_del_dest(struct netns_ipvs *ipvs, struct ip_vs_dest *dest,
1062                              bool cleanup)
1063 {
1064         ip_vs_stop_estimator(ipvs, &dest->stats);
1065
1066         /*
1067          *  Remove it from the d-linked list with the real services.
1068          */
1069         ip_vs_rs_unhash(dest);
1070
1071         spin_lock_bh(&ipvs->dest_trash_lock);
1072         IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1073                       IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port),
1074                       refcount_read(&dest->refcnt));
1075         if (list_empty(&ipvs->dest_trash) && !cleanup)
1076                 mod_timer(&ipvs->dest_trash_timer,
1077                           jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1078         /* dest lives in trash with reference */
1079         list_add(&dest->t_list, &ipvs->dest_trash);
1080         dest->idle_start = 0;
1081         spin_unlock_bh(&ipvs->dest_trash_lock);
1082 }
1083
1084
1085 /*
1086  *      Unlink a destination from the given service
1087  */
1088 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1089                                 struct ip_vs_dest *dest,
1090                                 int svcupd)
1091 {
1092         dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1093
1094         /*
1095          *  Remove it from the d-linked destination list.
1096          */
1097         list_del_rcu(&dest->n_list);
1098         svc->num_dests--;
1099
1100         if (dest->af != svc->af)
1101                 svc->ipvs->mixed_address_family_dests--;
1102
1103         if (svcupd) {
1104                 struct ip_vs_scheduler *sched;
1105
1106                 sched = rcu_dereference_protected(svc->scheduler, 1);
1107                 if (sched && sched->del_dest)
1108                         sched->del_dest(svc, dest);
1109         }
1110 }
1111
1112
1113 /*
1114  *      Delete a destination server in the given service
1115  */
1116 static int
1117 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1118 {
1119         struct ip_vs_dest *dest;
1120         __be16 dport = udest->port;
1121
1122         EnterFunction(2);
1123
1124         /* We use function that requires RCU lock */
1125         rcu_read_lock();
1126         dest = ip_vs_lookup_dest(svc, udest->af, &udest->addr, dport);
1127         rcu_read_unlock();
1128
1129         if (dest == NULL) {
1130                 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1131                 return -ENOENT;
1132         }
1133
1134         /*
1135          *      Unlink dest from the service
1136          */
1137         __ip_vs_unlink_dest(svc, dest, 1);
1138
1139         /*
1140          *      Delete the destination
1141          */
1142         __ip_vs_del_dest(svc->ipvs, dest, false);
1143
1144         LeaveFunction(2);
1145
1146         return 0;
1147 }
1148
1149 static void ip_vs_dest_trash_expire(struct timer_list *t)
1150 {
1151         struct netns_ipvs *ipvs = from_timer(ipvs, t, dest_trash_timer);
1152         struct ip_vs_dest *dest, *next;
1153         unsigned long now = jiffies;
1154
1155         spin_lock(&ipvs->dest_trash_lock);
1156         list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) {
1157                 if (refcount_read(&dest->refcnt) > 1)
1158                         continue;
1159                 if (dest->idle_start) {
1160                         if (time_before(now, dest->idle_start +
1161                                              IP_VS_DEST_TRASH_PERIOD))
1162                                 continue;
1163                 } else {
1164                         dest->idle_start = max(1UL, now);
1165                         continue;
1166                 }
1167                 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1168                               dest->vfwmark,
1169                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1170                               ntohs(dest->port));
1171                 list_del(&dest->t_list);
1172                 ip_vs_dest_free(dest);
1173         }
1174         if (!list_empty(&ipvs->dest_trash))
1175                 mod_timer(&ipvs->dest_trash_timer,
1176                           jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1177         spin_unlock(&ipvs->dest_trash_lock);
1178 }
1179
1180 /*
1181  *      Add a service into the service hash table
1182  */
1183 static int
1184 ip_vs_add_service(struct netns_ipvs *ipvs, struct ip_vs_service_user_kern *u,
1185                   struct ip_vs_service **svc_p)
1186 {
1187         int ret = 0, i;
1188         struct ip_vs_scheduler *sched = NULL;
1189         struct ip_vs_pe *pe = NULL;
1190         struct ip_vs_service *svc = NULL;
1191
1192         /* increase the module use count */
1193         ip_vs_use_count_inc();
1194
1195         /* Lookup the scheduler by 'u->sched_name' */
1196         if (strcmp(u->sched_name, "none")) {
1197                 sched = ip_vs_scheduler_get(u->sched_name);
1198                 if (!sched) {
1199                         pr_info("Scheduler module ip_vs_%s not found\n",
1200                                 u->sched_name);
1201                         ret = -ENOENT;
1202                         goto out_err;
1203                 }
1204         }
1205
1206         if (u->pe_name && *u->pe_name) {
1207                 pe = ip_vs_pe_getbyname(u->pe_name);
1208                 if (pe == NULL) {
1209                         pr_info("persistence engine module ip_vs_pe_%s "
1210                                 "not found\n", u->pe_name);
1211                         ret = -ENOENT;
1212                         goto out_err;
1213                 }
1214         }
1215
1216 #ifdef CONFIG_IP_VS_IPV6
1217         if (u->af == AF_INET6) {
1218                 __u32 plen = (__force __u32) u->netmask;
1219
1220                 if (plen < 1 || plen > 128) {
1221                         ret = -EINVAL;
1222                         goto out_err;
1223                 }
1224         }
1225 #endif
1226
1227         svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1228         if (svc == NULL) {
1229                 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1230                 ret = -ENOMEM;
1231                 goto out_err;
1232         }
1233         svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1234         if (!svc->stats.cpustats) {
1235                 ret = -ENOMEM;
1236                 goto out_err;
1237         }
1238
1239         for_each_possible_cpu(i) {
1240                 struct ip_vs_cpu_stats *ip_vs_stats;
1241                 ip_vs_stats = per_cpu_ptr(svc->stats.cpustats, i);
1242                 u64_stats_init(&ip_vs_stats->syncp);
1243         }
1244
1245
1246         /* I'm the first user of the service */
1247         atomic_set(&svc->refcnt, 0);
1248
1249         svc->af = u->af;
1250         svc->protocol = u->protocol;
1251         ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1252         svc->port = u->port;
1253         svc->fwmark = u->fwmark;
1254         svc->flags = u->flags;
1255         svc->timeout = u->timeout * HZ;
1256         svc->netmask = u->netmask;
1257         svc->ipvs = ipvs;
1258
1259         INIT_LIST_HEAD(&svc->destinations);
1260         spin_lock_init(&svc->sched_lock);
1261         spin_lock_init(&svc->stats.lock);
1262
1263         /* Bind the scheduler */
1264         if (sched) {
1265                 ret = ip_vs_bind_scheduler(svc, sched);
1266                 if (ret)
1267                         goto out_err;
1268                 sched = NULL;
1269         }
1270
1271         /* Bind the ct retriever */
1272         RCU_INIT_POINTER(svc->pe, pe);
1273         pe = NULL;
1274
1275         /* Update the virtual service counters */
1276         if (svc->port == FTPPORT)
1277                 atomic_inc(&ipvs->ftpsvc_counter);
1278         else if (svc->port == 0)
1279                 atomic_inc(&ipvs->nullsvc_counter);
1280         if (svc->pe && svc->pe->conn_out)
1281                 atomic_inc(&ipvs->conn_out_counter);
1282
1283         ip_vs_start_estimator(ipvs, &svc->stats);
1284
1285         /* Count only IPv4 services for old get/setsockopt interface */
1286         if (svc->af == AF_INET)
1287                 ipvs->num_services++;
1288
1289         /* Hash the service into the service table */
1290         ip_vs_svc_hash(svc);
1291
1292         *svc_p = svc;
1293         /* Now there is a service - full throttle */
1294         ipvs->enable = 1;
1295         return 0;
1296
1297
1298  out_err:
1299         if (svc != NULL) {
1300                 ip_vs_unbind_scheduler(svc, sched);
1301                 ip_vs_service_free(svc);
1302         }
1303         ip_vs_scheduler_put(sched);
1304         ip_vs_pe_put(pe);
1305
1306         /* decrease the module use count */
1307         ip_vs_use_count_dec();
1308
1309         return ret;
1310 }
1311
1312
1313 /*
1314  *      Edit a service and bind it with a new scheduler
1315  */
1316 static int
1317 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1318 {
1319         struct ip_vs_scheduler *sched = NULL, *old_sched;
1320         struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1321         int ret = 0;
1322         bool new_pe_conn_out, old_pe_conn_out;
1323
1324         /*
1325          * Lookup the scheduler, by 'u->sched_name'
1326          */
1327         if (strcmp(u->sched_name, "none")) {
1328                 sched = ip_vs_scheduler_get(u->sched_name);
1329                 if (!sched) {
1330                         pr_info("Scheduler module ip_vs_%s not found\n",
1331                                 u->sched_name);
1332                         return -ENOENT;
1333                 }
1334         }
1335         old_sched = sched;
1336
1337         if (u->pe_name && *u->pe_name) {
1338                 pe = ip_vs_pe_getbyname(u->pe_name);
1339                 if (pe == NULL) {
1340                         pr_info("persistence engine module ip_vs_pe_%s "
1341                                 "not found\n", u->pe_name);
1342                         ret = -ENOENT;
1343                         goto out;
1344                 }
1345                 old_pe = pe;
1346         }
1347
1348 #ifdef CONFIG_IP_VS_IPV6
1349         if (u->af == AF_INET6) {
1350                 __u32 plen = (__force __u32) u->netmask;
1351
1352                 if (plen < 1 || plen > 128) {
1353                         ret = -EINVAL;
1354                         goto out;
1355                 }
1356         }
1357 #endif
1358
1359         old_sched = rcu_dereference_protected(svc->scheduler, 1);
1360         if (sched != old_sched) {
1361                 if (old_sched) {
1362                         ip_vs_unbind_scheduler(svc, old_sched);
1363                         RCU_INIT_POINTER(svc->scheduler, NULL);
1364                         /* Wait all svc->sched_data users */
1365                         synchronize_rcu();
1366                 }
1367                 /* Bind the new scheduler */
1368                 if (sched) {
1369                         ret = ip_vs_bind_scheduler(svc, sched);
1370                         if (ret) {
1371                                 ip_vs_scheduler_put(sched);
1372                                 goto out;
1373                         }
1374                 }
1375         }
1376
1377         /*
1378          * Set the flags and timeout value
1379          */
1380         svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1381         svc->timeout = u->timeout * HZ;
1382         svc->netmask = u->netmask;
1383
1384         old_pe = rcu_dereference_protected(svc->pe, 1);
1385         if (pe != old_pe) {
1386                 rcu_assign_pointer(svc->pe, pe);
1387                 /* check for optional methods in new pe */
1388                 new_pe_conn_out = (pe && pe->conn_out) ? true : false;
1389                 old_pe_conn_out = (old_pe && old_pe->conn_out) ? true : false;
1390                 if (new_pe_conn_out && !old_pe_conn_out)
1391                         atomic_inc(&svc->ipvs->conn_out_counter);
1392                 if (old_pe_conn_out && !new_pe_conn_out)
1393                         atomic_dec(&svc->ipvs->conn_out_counter);
1394         }
1395
1396 out:
1397         ip_vs_scheduler_put(old_sched);
1398         ip_vs_pe_put(old_pe);
1399         return ret;
1400 }
1401
1402 /*
1403  *      Delete a service from the service list
1404  *      - The service must be unlinked, unlocked and not referenced!
1405  *      - We are called under _bh lock
1406  */
1407 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1408 {
1409         struct ip_vs_dest *dest, *nxt;
1410         struct ip_vs_scheduler *old_sched;
1411         struct ip_vs_pe *old_pe;
1412         struct netns_ipvs *ipvs = svc->ipvs;
1413
1414         /* Count only IPv4 services for old get/setsockopt interface */
1415         if (svc->af == AF_INET)
1416                 ipvs->num_services--;
1417
1418         ip_vs_stop_estimator(svc->ipvs, &svc->stats);
1419
1420         /* Unbind scheduler */
1421         old_sched = rcu_dereference_protected(svc->scheduler, 1);
1422         ip_vs_unbind_scheduler(svc, old_sched);
1423         ip_vs_scheduler_put(old_sched);
1424
1425         /* Unbind persistence engine, keep svc->pe */
1426         old_pe = rcu_dereference_protected(svc->pe, 1);
1427         if (old_pe && old_pe->conn_out)
1428                 atomic_dec(&ipvs->conn_out_counter);
1429         ip_vs_pe_put(old_pe);
1430
1431         /*
1432          *    Unlink the whole destination list
1433          */
1434         list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1435                 __ip_vs_unlink_dest(svc, dest, 0);
1436                 __ip_vs_del_dest(svc->ipvs, dest, cleanup);
1437         }
1438
1439         /*
1440          *    Update the virtual service counters
1441          */
1442         if (svc->port == FTPPORT)
1443                 atomic_dec(&ipvs->ftpsvc_counter);
1444         else if (svc->port == 0)
1445                 atomic_dec(&ipvs->nullsvc_counter);
1446
1447         /*
1448          *    Free the service if nobody refers to it
1449          */
1450         __ip_vs_svc_put(svc, true);
1451
1452         /* decrease the module use count */
1453         ip_vs_use_count_dec();
1454 }
1455
1456 /*
1457  * Unlink a service from list and try to delete it if its refcnt reached 0
1458  */
1459 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1460 {
1461         /* Hold svc to avoid double release from dest_trash */
1462         atomic_inc(&svc->refcnt);
1463         /*
1464          * Unhash it from the service table
1465          */
1466         ip_vs_svc_unhash(svc);
1467
1468         __ip_vs_del_service(svc, cleanup);
1469 }
1470
1471 /*
1472  *      Delete a service from the service list
1473  */
1474 static int ip_vs_del_service(struct ip_vs_service *svc)
1475 {
1476         if (svc == NULL)
1477                 return -EEXIST;
1478         ip_vs_unlink_service(svc, false);
1479
1480         return 0;
1481 }
1482
1483
1484 /*
1485  *      Flush all the virtual services
1486  */
1487 static int ip_vs_flush(struct netns_ipvs *ipvs, bool cleanup)
1488 {
1489         int idx;
1490         struct ip_vs_service *svc;
1491         struct hlist_node *n;
1492
1493         /*
1494          * Flush the service table hashed by <netns,protocol,addr,port>
1495          */
1496         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1497                 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1498                                           s_list) {
1499                         if (svc->ipvs == ipvs)
1500                                 ip_vs_unlink_service(svc, cleanup);
1501                 }
1502         }
1503
1504         /*
1505          * Flush the service table hashed by fwmark
1506          */
1507         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1508                 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1509                                           f_list) {
1510                         if (svc->ipvs == ipvs)
1511                                 ip_vs_unlink_service(svc, cleanup);
1512                 }
1513         }
1514
1515         return 0;
1516 }
1517
1518 /*
1519  *      Delete service by {netns} in the service table.
1520  *      Called by __ip_vs_cleanup()
1521  */
1522 void ip_vs_service_net_cleanup(struct netns_ipvs *ipvs)
1523 {
1524         EnterFunction(2);
1525         /* Check for "full" addressed entries */
1526         mutex_lock(&__ip_vs_mutex);
1527         ip_vs_flush(ipvs, true);
1528         mutex_unlock(&__ip_vs_mutex);
1529         LeaveFunction(2);
1530 }
1531
1532 /* Put all references for device (dst_cache) */
1533 static inline void
1534 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1535 {
1536         struct ip_vs_dest_dst *dest_dst;
1537
1538         spin_lock_bh(&dest->dst_lock);
1539         dest_dst = rcu_dereference_protected(dest->dest_dst, 1);
1540         if (dest_dst && dest_dst->dst_cache->dev == dev) {
1541                 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1542                               dev->name,
1543                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1544                               ntohs(dest->port),
1545                               refcount_read(&dest->refcnt));
1546                 __ip_vs_dst_cache_reset(dest);
1547         }
1548         spin_unlock_bh(&dest->dst_lock);
1549
1550 }
1551 /* Netdev event receiver
1552  * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1553  */
1554 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1555                            void *ptr)
1556 {
1557         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1558         struct net *net = dev_net(dev);
1559         struct netns_ipvs *ipvs = net_ipvs(net);
1560         struct ip_vs_service *svc;
1561         struct ip_vs_dest *dest;
1562         unsigned int idx;
1563
1564         if (event != NETDEV_DOWN || !ipvs)
1565                 return NOTIFY_DONE;
1566         IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1567         EnterFunction(2);
1568         mutex_lock(&__ip_vs_mutex);
1569         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1570                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1571                         if (svc->ipvs == ipvs) {
1572                                 list_for_each_entry(dest, &svc->destinations,
1573                                                     n_list) {
1574                                         ip_vs_forget_dev(dest, dev);
1575                                 }
1576                         }
1577                 }
1578
1579                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1580                         if (svc->ipvs == ipvs) {
1581                                 list_for_each_entry(dest, &svc->destinations,
1582                                                     n_list) {
1583                                         ip_vs_forget_dev(dest, dev);
1584                                 }
1585                         }
1586
1587                 }
1588         }
1589
1590         spin_lock_bh(&ipvs->dest_trash_lock);
1591         list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1592                 ip_vs_forget_dev(dest, dev);
1593         }
1594         spin_unlock_bh(&ipvs->dest_trash_lock);
1595         mutex_unlock(&__ip_vs_mutex);
1596         LeaveFunction(2);
1597         return NOTIFY_DONE;
1598 }
1599
1600 /*
1601  *      Zero counters in a service or all services
1602  */
1603 static int ip_vs_zero_service(struct ip_vs_service *svc)
1604 {
1605         struct ip_vs_dest *dest;
1606
1607         list_for_each_entry(dest, &svc->destinations, n_list) {
1608                 ip_vs_zero_stats(&dest->stats);
1609         }
1610         ip_vs_zero_stats(&svc->stats);
1611         return 0;
1612 }
1613
1614 static int ip_vs_zero_all(struct netns_ipvs *ipvs)
1615 {
1616         int idx;
1617         struct ip_vs_service *svc;
1618
1619         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1620                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1621                         if (svc->ipvs == ipvs)
1622                                 ip_vs_zero_service(svc);
1623                 }
1624         }
1625
1626         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1627                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1628                         if (svc->ipvs == ipvs)
1629                                 ip_vs_zero_service(svc);
1630                 }
1631         }
1632
1633         ip_vs_zero_stats(&ipvs->tot_stats);
1634         return 0;
1635 }
1636
1637 #ifdef CONFIG_SYSCTL
1638
1639 static int zero;
1640 static int three = 3;
1641
1642 static int
1643 proc_do_defense_mode(struct ctl_table *table, int write,
1644                      void __user *buffer, size_t *lenp, loff_t *ppos)
1645 {
1646         struct netns_ipvs *ipvs = table->extra2;
1647         int *valp = table->data;
1648         int val = *valp;
1649         int rc;
1650
1651         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1652         if (write && (*valp != val)) {
1653                 if ((*valp < 0) || (*valp > 3)) {
1654                         /* Restore the correct value */
1655                         *valp = val;
1656                 } else {
1657                         update_defense_level(ipvs);
1658                 }
1659         }
1660         return rc;
1661 }
1662
1663 static int
1664 proc_do_sync_threshold(struct ctl_table *table, int write,
1665                        void __user *buffer, size_t *lenp, loff_t *ppos)
1666 {
1667         int *valp = table->data;
1668         int val[2];
1669         int rc;
1670
1671         /* backup the value first */
1672         memcpy(val, valp, sizeof(val));
1673
1674         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1675         if (write && (valp[0] < 0 || valp[1] < 0 ||
1676             (valp[0] >= valp[1] && valp[1]))) {
1677                 /* Restore the correct value */
1678                 memcpy(valp, val, sizeof(val));
1679         }
1680         return rc;
1681 }
1682
1683 static int
1684 proc_do_sync_mode(struct ctl_table *table, int write,
1685                      void __user *buffer, size_t *lenp, loff_t *ppos)
1686 {
1687         int *valp = table->data;
1688         int val = *valp;
1689         int rc;
1690
1691         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1692         if (write && (*valp != val)) {
1693                 if ((*valp < 0) || (*valp > 1)) {
1694                         /* Restore the correct value */
1695                         *valp = val;
1696                 }
1697         }
1698         return rc;
1699 }
1700
1701 static int
1702 proc_do_sync_ports(struct ctl_table *table, int write,
1703                    void __user *buffer, size_t *lenp, loff_t *ppos)
1704 {
1705         int *valp = table->data;
1706         int val = *valp;
1707         int rc;
1708
1709         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1710         if (write && (*valp != val)) {
1711                 if (*valp < 1 || !is_power_of_2(*valp)) {
1712                         /* Restore the correct value */
1713                         *valp = val;
1714                 }
1715         }
1716         return rc;
1717 }
1718
1719 /*
1720  *      IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1721  *      Do not change order or insert new entries without
1722  *      align with netns init in ip_vs_control_net_init()
1723  */
1724
1725 static struct ctl_table vs_vars[] = {
1726         {
1727                 .procname       = "amemthresh",
1728                 .maxlen         = sizeof(int),
1729                 .mode           = 0644,
1730                 .proc_handler   = proc_dointvec,
1731         },
1732         {
1733                 .procname       = "am_droprate",
1734                 .maxlen         = sizeof(int),
1735                 .mode           = 0644,
1736                 .proc_handler   = proc_dointvec,
1737         },
1738         {
1739                 .procname       = "drop_entry",
1740                 .maxlen         = sizeof(int),
1741                 .mode           = 0644,
1742                 .proc_handler   = proc_do_defense_mode,
1743         },
1744         {
1745                 .procname       = "drop_packet",
1746                 .maxlen         = sizeof(int),
1747                 .mode           = 0644,
1748                 .proc_handler   = proc_do_defense_mode,
1749         },
1750 #ifdef CONFIG_IP_VS_NFCT
1751         {
1752                 .procname       = "conntrack",
1753                 .maxlen         = sizeof(int),
1754                 .mode           = 0644,
1755                 .proc_handler   = &proc_dointvec,
1756         },
1757 #endif
1758         {
1759                 .procname       = "secure_tcp",
1760                 .maxlen         = sizeof(int),
1761                 .mode           = 0644,
1762                 .proc_handler   = proc_do_defense_mode,
1763         },
1764         {
1765                 .procname       = "snat_reroute",
1766                 .maxlen         = sizeof(int),
1767                 .mode           = 0644,
1768                 .proc_handler   = &proc_dointvec,
1769         },
1770         {
1771                 .procname       = "sync_version",
1772                 .maxlen         = sizeof(int),
1773                 .mode           = 0644,
1774                 .proc_handler   = proc_do_sync_mode,
1775         },
1776         {
1777                 .procname       = "sync_ports",
1778                 .maxlen         = sizeof(int),
1779                 .mode           = 0644,
1780                 .proc_handler   = proc_do_sync_ports,
1781         },
1782         {
1783                 .procname       = "sync_persist_mode",
1784                 .maxlen         = sizeof(int),
1785                 .mode           = 0644,
1786                 .proc_handler   = proc_dointvec,
1787         },
1788         {
1789                 .procname       = "sync_qlen_max",
1790                 .maxlen         = sizeof(unsigned long),
1791                 .mode           = 0644,
1792                 .proc_handler   = proc_doulongvec_minmax,
1793         },
1794         {
1795                 .procname       = "sync_sock_size",
1796                 .maxlen         = sizeof(int),
1797                 .mode           = 0644,
1798                 .proc_handler   = proc_dointvec,
1799         },
1800         {
1801                 .procname       = "cache_bypass",
1802                 .maxlen         = sizeof(int),
1803                 .mode           = 0644,
1804                 .proc_handler   = proc_dointvec,
1805         },
1806         {
1807                 .procname       = "expire_nodest_conn",
1808                 .maxlen         = sizeof(int),
1809                 .mode           = 0644,
1810                 .proc_handler   = proc_dointvec,
1811         },
1812         {
1813                 .procname       = "sloppy_tcp",
1814                 .maxlen         = sizeof(int),
1815                 .mode           = 0644,
1816                 .proc_handler   = proc_dointvec,
1817         },
1818         {
1819                 .procname       = "sloppy_sctp",
1820                 .maxlen         = sizeof(int),
1821                 .mode           = 0644,
1822                 .proc_handler   = proc_dointvec,
1823         },
1824         {
1825                 .procname       = "expire_quiescent_template",
1826                 .maxlen         = sizeof(int),
1827                 .mode           = 0644,
1828                 .proc_handler   = proc_dointvec,
1829         },
1830         {
1831                 .procname       = "sync_threshold",
1832                 .maxlen         =
1833                         sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1834                 .mode           = 0644,
1835                 .proc_handler   = proc_do_sync_threshold,
1836         },
1837         {
1838                 .procname       = "sync_refresh_period",
1839                 .maxlen         = sizeof(int),
1840                 .mode           = 0644,
1841                 .proc_handler   = proc_dointvec_jiffies,
1842         },
1843         {
1844                 .procname       = "sync_retries",
1845                 .maxlen         = sizeof(int),
1846                 .mode           = 0644,
1847                 .proc_handler   = proc_dointvec_minmax,
1848                 .extra1         = &zero,
1849                 .extra2         = &three,
1850         },
1851         {
1852                 .procname       = "nat_icmp_send",
1853                 .maxlen         = sizeof(int),
1854                 .mode           = 0644,
1855                 .proc_handler   = proc_dointvec,
1856         },
1857         {
1858                 .procname       = "pmtu_disc",
1859                 .maxlen         = sizeof(int),
1860                 .mode           = 0644,
1861                 .proc_handler   = proc_dointvec,
1862         },
1863         {
1864                 .procname       = "backup_only",
1865                 .maxlen         = sizeof(int),
1866                 .mode           = 0644,
1867                 .proc_handler   = proc_dointvec,
1868         },
1869         {
1870                 .procname       = "conn_reuse_mode",
1871                 .maxlen         = sizeof(int),
1872                 .mode           = 0644,
1873                 .proc_handler   = proc_dointvec,
1874         },
1875         {
1876                 .procname       = "schedule_icmp",
1877                 .maxlen         = sizeof(int),
1878                 .mode           = 0644,
1879                 .proc_handler   = proc_dointvec,
1880         },
1881         {
1882                 .procname       = "ignore_tunneled",
1883                 .maxlen         = sizeof(int),
1884                 .mode           = 0644,
1885                 .proc_handler   = proc_dointvec,
1886         },
1887 #ifdef CONFIG_IP_VS_DEBUG
1888         {
1889                 .procname       = "debug_level",
1890                 .data           = &sysctl_ip_vs_debug_level,
1891                 .maxlen         = sizeof(int),
1892                 .mode           = 0644,
1893                 .proc_handler   = proc_dointvec,
1894         },
1895 #endif
1896         { }
1897 };
1898
1899 #endif
1900
1901 #ifdef CONFIG_PROC_FS
1902
1903 struct ip_vs_iter {
1904         struct seq_net_private p;  /* Do not move this, netns depends upon it*/
1905         struct hlist_head *table;
1906         int bucket;
1907 };
1908
1909 /*
1910  *      Write the contents of the VS rule table to a PROCfs file.
1911  *      (It is kept just for backward compatibility)
1912  */
1913 static inline const char *ip_vs_fwd_name(unsigned int flags)
1914 {
1915         switch (flags & IP_VS_CONN_F_FWD_MASK) {
1916         case IP_VS_CONN_F_LOCALNODE:
1917                 return "Local";
1918         case IP_VS_CONN_F_TUNNEL:
1919                 return "Tunnel";
1920         case IP_VS_CONN_F_DROUTE:
1921                 return "Route";
1922         default:
1923                 return "Masq";
1924         }
1925 }
1926
1927
1928 /* Get the Nth entry in the two lists */
1929 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1930 {
1931         struct net *net = seq_file_net(seq);
1932         struct netns_ipvs *ipvs = net_ipvs(net);
1933         struct ip_vs_iter *iter = seq->private;
1934         int idx;
1935         struct ip_vs_service *svc;
1936
1937         /* look in hash by protocol */
1938         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1939                 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
1940                         if ((svc->ipvs == ipvs) && pos-- == 0) {
1941                                 iter->table = ip_vs_svc_table;
1942                                 iter->bucket = idx;
1943                                 return svc;
1944                         }
1945                 }
1946         }
1947
1948         /* keep looking in fwmark */
1949         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1950                 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
1951                                          f_list) {
1952                         if ((svc->ipvs == ipvs) && pos-- == 0) {
1953                                 iter->table = ip_vs_svc_fwm_table;
1954                                 iter->bucket = idx;
1955                                 return svc;
1956                         }
1957                 }
1958         }
1959
1960         return NULL;
1961 }
1962
1963 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1964         __acquires(RCU)
1965 {
1966         rcu_read_lock();
1967         return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1968 }
1969
1970
1971 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1972 {
1973         struct hlist_node *e;
1974         struct ip_vs_iter *iter;
1975         struct ip_vs_service *svc;
1976
1977         ++*pos;
1978         if (v == SEQ_START_TOKEN)
1979                 return ip_vs_info_array(seq,0);
1980
1981         svc = v;
1982         iter = seq->private;
1983
1984         if (iter->table == ip_vs_svc_table) {
1985                 /* next service in table hashed by protocol */
1986                 e = rcu_dereference(hlist_next_rcu(&svc->s_list));
1987                 if (e)
1988                         return hlist_entry(e, struct ip_vs_service, s_list);
1989
1990                 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1991                         hlist_for_each_entry_rcu(svc,
1992                                                  &ip_vs_svc_table[iter->bucket],
1993                                                  s_list) {
1994                                 return svc;
1995                         }
1996                 }
1997
1998                 iter->table = ip_vs_svc_fwm_table;
1999                 iter->bucket = -1;
2000                 goto scan_fwmark;
2001         }
2002
2003         /* next service in hashed by fwmark */
2004         e = rcu_dereference(hlist_next_rcu(&svc->f_list));
2005         if (e)
2006                 return hlist_entry(e, struct ip_vs_service, f_list);
2007
2008  scan_fwmark:
2009         while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2010                 hlist_for_each_entry_rcu(svc,
2011                                          &ip_vs_svc_fwm_table[iter->bucket],
2012                                          f_list)
2013                         return svc;
2014         }
2015
2016         return NULL;
2017 }
2018
2019 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
2020         __releases(RCU)
2021 {
2022         rcu_read_unlock();
2023 }
2024
2025
2026 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
2027 {
2028         if (v == SEQ_START_TOKEN) {
2029                 seq_printf(seq,
2030                         "IP Virtual Server version %d.%d.%d (size=%d)\n",
2031                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2032                 seq_puts(seq,
2033                          "Prot LocalAddress:Port Scheduler Flags\n");
2034                 seq_puts(seq,
2035                          "  -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2036         } else {
2037                 struct net *net = seq_file_net(seq);
2038                 struct netns_ipvs *ipvs = net_ipvs(net);
2039                 const struct ip_vs_service *svc = v;
2040                 const struct ip_vs_iter *iter = seq->private;
2041                 const struct ip_vs_dest *dest;
2042                 struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler);
2043                 char *sched_name = sched ? sched->name : "none";
2044
2045                 if (svc->ipvs != ipvs)
2046                         return 0;
2047                 if (iter->table == ip_vs_svc_table) {
2048 #ifdef CONFIG_IP_VS_IPV6
2049                         if (svc->af == AF_INET6)
2050                                 seq_printf(seq, "%s  [%pI6]:%04X %s ",
2051                                            ip_vs_proto_name(svc->protocol),
2052                                            &svc->addr.in6,
2053                                            ntohs(svc->port),
2054                                            sched_name);
2055                         else
2056 #endif
2057                                 seq_printf(seq, "%s  %08X:%04X %s %s ",
2058                                            ip_vs_proto_name(svc->protocol),
2059                                            ntohl(svc->addr.ip),
2060                                            ntohs(svc->port),
2061                                            sched_name,
2062                                            (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2063                 } else {
2064                         seq_printf(seq, "FWM  %08X %s %s",
2065                                    svc->fwmark, sched_name,
2066                                    (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2067                 }
2068
2069                 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2070                         seq_printf(seq, "persistent %d %08X\n",
2071                                 svc->timeout,
2072                                 ntohl(svc->netmask));
2073                 else
2074                         seq_putc(seq, '\n');
2075
2076                 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
2077 #ifdef CONFIG_IP_VS_IPV6
2078                         if (dest->af == AF_INET6)
2079                                 seq_printf(seq,
2080                                            "  -> [%pI6]:%04X"
2081                                            "      %-7s %-6d %-10d %-10d\n",
2082                                            &dest->addr.in6,
2083                                            ntohs(dest->port),
2084                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2085                                            atomic_read(&dest->weight),
2086                                            atomic_read(&dest->activeconns),
2087                                            atomic_read(&dest->inactconns));
2088                         else
2089 #endif
2090                                 seq_printf(seq,
2091                                            "  -> %08X:%04X      "
2092                                            "%-7s %-6d %-10d %-10d\n",
2093                                            ntohl(dest->addr.ip),
2094                                            ntohs(dest->port),
2095                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2096                                            atomic_read(&dest->weight),
2097                                            atomic_read(&dest->activeconns),
2098                                            atomic_read(&dest->inactconns));
2099
2100                 }
2101         }
2102         return 0;
2103 }
2104
2105 static const struct seq_operations ip_vs_info_seq_ops = {
2106         .start = ip_vs_info_seq_start,
2107         .next  = ip_vs_info_seq_next,
2108         .stop  = ip_vs_info_seq_stop,
2109         .show  = ip_vs_info_seq_show,
2110 };
2111
2112 static int ip_vs_info_open(struct inode *inode, struct file *file)
2113 {
2114         return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2115                         sizeof(struct ip_vs_iter));
2116 }
2117
2118 static const struct file_operations ip_vs_info_fops = {
2119         .open    = ip_vs_info_open,
2120         .read    = seq_read,
2121         .llseek  = seq_lseek,
2122         .release = seq_release_net,
2123 };
2124
2125 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2126 {
2127         struct net *net = seq_file_single_net(seq);
2128         struct ip_vs_kstats show;
2129
2130 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
2131         seq_puts(seq,
2132                  "   Total Incoming Outgoing         Incoming         Outgoing\n");
2133         seq_puts(seq,
2134                  "   Conns  Packets  Packets            Bytes            Bytes\n");
2135
2136         ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2137         seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n\n",
2138                    (unsigned long long)show.conns,
2139                    (unsigned long long)show.inpkts,
2140                    (unsigned long long)show.outpkts,
2141                    (unsigned long long)show.inbytes,
2142                    (unsigned long long)show.outbytes);
2143
2144 /*                01234567 01234567 01234567 0123456701234567 0123456701234567*/
2145         seq_puts(seq,
2146                  " Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
2147         seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n",
2148                    (unsigned long long)show.cps,
2149                    (unsigned long long)show.inpps,
2150                    (unsigned long long)show.outpps,
2151                    (unsigned long long)show.inbps,
2152                    (unsigned long long)show.outbps);
2153
2154         return 0;
2155 }
2156
2157 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2158 {
2159         return single_open_net(inode, file, ip_vs_stats_show);
2160 }
2161
2162 static const struct file_operations ip_vs_stats_fops = {
2163         .open = ip_vs_stats_seq_open,
2164         .read = seq_read,
2165         .llseek = seq_lseek,
2166         .release = single_release_net,
2167 };
2168
2169 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2170 {
2171         struct net *net = seq_file_single_net(seq);
2172         struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2173         struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats;
2174         struct ip_vs_kstats kstats;
2175         int i;
2176
2177 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
2178         seq_puts(seq,
2179                  "       Total Incoming Outgoing         Incoming         Outgoing\n");
2180         seq_puts(seq,
2181                  "CPU    Conns  Packets  Packets            Bytes            Bytes\n");
2182
2183         for_each_possible_cpu(i) {
2184                 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2185                 unsigned int start;
2186                 u64 conns, inpkts, outpkts, inbytes, outbytes;
2187
2188                 do {
2189                         start = u64_stats_fetch_begin_irq(&u->syncp);
2190                         conns = u->cnt.conns;
2191                         inpkts = u->cnt.inpkts;
2192                         outpkts = u->cnt.outpkts;
2193                         inbytes = u->cnt.inbytes;
2194                         outbytes = u->cnt.outbytes;
2195                 } while (u64_stats_fetch_retry_irq(&u->syncp, start));
2196
2197                 seq_printf(seq, "%3X %8LX %8LX %8LX %16LX %16LX\n",
2198                            i, (u64)conns, (u64)inpkts,
2199                            (u64)outpkts, (u64)inbytes,
2200                            (u64)outbytes);
2201         }
2202
2203         ip_vs_copy_stats(&kstats, tot_stats);
2204
2205         seq_printf(seq, "  ~ %8LX %8LX %8LX %16LX %16LX\n\n",
2206                    (unsigned long long)kstats.conns,
2207                    (unsigned long long)kstats.inpkts,
2208                    (unsigned long long)kstats.outpkts,
2209                    (unsigned long long)kstats.inbytes,
2210                    (unsigned long long)kstats.outbytes);
2211
2212 /*                ... 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2213         seq_puts(seq,
2214                  "     Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
2215         seq_printf(seq, "    %8LX %8LX %8LX %16LX %16LX\n",
2216                    kstats.cps,
2217                    kstats.inpps,
2218                    kstats.outpps,
2219                    kstats.inbps,
2220                    kstats.outbps);
2221
2222         return 0;
2223 }
2224
2225 static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2226 {
2227         return single_open_net(inode, file, ip_vs_stats_percpu_show);
2228 }
2229
2230 static const struct file_operations ip_vs_stats_percpu_fops = {
2231         .open = ip_vs_stats_percpu_seq_open,
2232         .read = seq_read,
2233         .llseek = seq_lseek,
2234         .release = single_release_net,
2235 };
2236 #endif
2237
2238 /*
2239  *      Set timeout values for tcp tcpfin udp in the timeout_table.
2240  */
2241 static int ip_vs_set_timeout(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2242 {
2243 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2244         struct ip_vs_proto_data *pd;
2245 #endif
2246
2247         IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2248                   u->tcp_timeout,
2249                   u->tcp_fin_timeout,
2250                   u->udp_timeout);
2251
2252 #ifdef CONFIG_IP_VS_PROTO_TCP
2253         if (u->tcp_timeout) {
2254                 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2255                 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2256                         = u->tcp_timeout * HZ;
2257         }
2258
2259         if (u->tcp_fin_timeout) {
2260                 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2261                 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2262                         = u->tcp_fin_timeout * HZ;
2263         }
2264 #endif
2265
2266 #ifdef CONFIG_IP_VS_PROTO_UDP
2267         if (u->udp_timeout) {
2268                 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2269                 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2270                         = u->udp_timeout * HZ;
2271         }
2272 #endif
2273         return 0;
2274 }
2275
2276 #define CMDID(cmd)              (cmd - IP_VS_BASE_CTL)
2277
2278 struct ip_vs_svcdest_user {
2279         struct ip_vs_service_user       s;
2280         struct ip_vs_dest_user          d;
2281 };
2282
2283 static const unsigned char set_arglen[CMDID(IP_VS_SO_SET_MAX) + 1] = {
2284         [CMDID(IP_VS_SO_SET_ADD)]         = sizeof(struct ip_vs_service_user),
2285         [CMDID(IP_VS_SO_SET_EDIT)]        = sizeof(struct ip_vs_service_user),
2286         [CMDID(IP_VS_SO_SET_DEL)]         = sizeof(struct ip_vs_service_user),
2287         [CMDID(IP_VS_SO_SET_ADDDEST)]     = sizeof(struct ip_vs_svcdest_user),
2288         [CMDID(IP_VS_SO_SET_DELDEST)]     = sizeof(struct ip_vs_svcdest_user),
2289         [CMDID(IP_VS_SO_SET_EDITDEST)]    = sizeof(struct ip_vs_svcdest_user),
2290         [CMDID(IP_VS_SO_SET_TIMEOUT)]     = sizeof(struct ip_vs_timeout_user),
2291         [CMDID(IP_VS_SO_SET_STARTDAEMON)] = sizeof(struct ip_vs_daemon_user),
2292         [CMDID(IP_VS_SO_SET_STOPDAEMON)]  = sizeof(struct ip_vs_daemon_user),
2293         [CMDID(IP_VS_SO_SET_ZERO)]        = sizeof(struct ip_vs_service_user),
2294 };
2295
2296 union ip_vs_set_arglen {
2297         struct ip_vs_service_user       field_IP_VS_SO_SET_ADD;
2298         struct ip_vs_service_user       field_IP_VS_SO_SET_EDIT;
2299         struct ip_vs_service_user       field_IP_VS_SO_SET_DEL;
2300         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_ADDDEST;
2301         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_DELDEST;
2302         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_EDITDEST;
2303         struct ip_vs_timeout_user       field_IP_VS_SO_SET_TIMEOUT;
2304         struct ip_vs_daemon_user        field_IP_VS_SO_SET_STARTDAEMON;
2305         struct ip_vs_daemon_user        field_IP_VS_SO_SET_STOPDAEMON;
2306         struct ip_vs_service_user       field_IP_VS_SO_SET_ZERO;
2307 };
2308
2309 #define MAX_SET_ARGLEN  sizeof(union ip_vs_set_arglen)
2310
2311 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2312                                   struct ip_vs_service_user *usvc_compat)
2313 {
2314         memset(usvc, 0, sizeof(*usvc));
2315
2316         usvc->af                = AF_INET;
2317         usvc->protocol          = usvc_compat->protocol;
2318         usvc->addr.ip           = usvc_compat->addr;
2319         usvc->port              = usvc_compat->port;
2320         usvc->fwmark            = usvc_compat->fwmark;
2321
2322         /* Deep copy of sched_name is not needed here */
2323         usvc->sched_name        = usvc_compat->sched_name;
2324
2325         usvc->flags             = usvc_compat->flags;
2326         usvc->timeout           = usvc_compat->timeout;
2327         usvc->netmask           = usvc_compat->netmask;
2328 }
2329
2330 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2331                                    struct ip_vs_dest_user *udest_compat)
2332 {
2333         memset(udest, 0, sizeof(*udest));
2334
2335         udest->addr.ip          = udest_compat->addr;
2336         udest->port             = udest_compat->port;
2337         udest->conn_flags       = udest_compat->conn_flags;
2338         udest->weight           = udest_compat->weight;
2339         udest->u_threshold      = udest_compat->u_threshold;
2340         udest->l_threshold      = udest_compat->l_threshold;
2341         udest->af               = AF_INET;
2342 }
2343
2344 static int
2345 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2346 {
2347         struct net *net = sock_net(sk);
2348         int ret;
2349         unsigned char arg[MAX_SET_ARGLEN];
2350         struct ip_vs_service_user *usvc_compat;
2351         struct ip_vs_service_user_kern usvc;
2352         struct ip_vs_service *svc;
2353         struct ip_vs_dest_user *udest_compat;
2354         struct ip_vs_dest_user_kern udest;
2355         struct netns_ipvs *ipvs = net_ipvs(net);
2356
2357         BUILD_BUG_ON(sizeof(arg) > 255);
2358         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2359                 return -EPERM;
2360
2361         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2362                 return -EINVAL;
2363         if (len != set_arglen[CMDID(cmd)]) {
2364                 IP_VS_DBG(1, "set_ctl: len %u != %u\n",
2365                           len, set_arglen[CMDID(cmd)]);
2366                 return -EINVAL;
2367         }
2368
2369         if (copy_from_user(arg, user, len) != 0)
2370                 return -EFAULT;
2371
2372         /* increase the module use count */
2373         ip_vs_use_count_inc();
2374
2375         /* Handle daemons since they have another lock */
2376         if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2377             cmd == IP_VS_SO_SET_STOPDAEMON) {
2378                 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2379
2380                 if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2381                         struct ipvs_sync_daemon_cfg cfg;
2382
2383                         memset(&cfg, 0, sizeof(cfg));
2384                         strlcpy(cfg.mcast_ifn, dm->mcast_ifn,
2385                                 sizeof(cfg.mcast_ifn));
2386                         cfg.syncid = dm->syncid;
2387                         ret = start_sync_thread(ipvs, &cfg, dm->state);
2388                 } else {
2389                         mutex_lock(&ipvs->sync_mutex);
2390                         ret = stop_sync_thread(ipvs, dm->state);
2391                         mutex_unlock(&ipvs->sync_mutex);
2392                 }
2393                 goto out_dec;
2394         }
2395
2396         mutex_lock(&__ip_vs_mutex);
2397         if (cmd == IP_VS_SO_SET_FLUSH) {
2398                 /* Flush the virtual service */
2399                 ret = ip_vs_flush(ipvs, false);
2400                 goto out_unlock;
2401         } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2402                 /* Set timeout values for (tcp tcpfin udp) */
2403                 ret = ip_vs_set_timeout(ipvs, (struct ip_vs_timeout_user *)arg);
2404                 goto out_unlock;
2405         }
2406
2407         usvc_compat = (struct ip_vs_service_user *)arg;
2408         udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2409
2410         /* We only use the new structs internally, so copy userspace compat
2411          * structs to extended internal versions */
2412         ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2413         ip_vs_copy_udest_compat(&udest, udest_compat);
2414
2415         if (cmd == IP_VS_SO_SET_ZERO) {
2416                 /* if no service address is set, zero counters in all */
2417                 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2418                         ret = ip_vs_zero_all(ipvs);
2419                         goto out_unlock;
2420                 }
2421         }
2422
2423         /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2424         if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2425             usvc.protocol != IPPROTO_SCTP) {
2426                 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2427                        usvc.protocol, &usvc.addr.ip,
2428                        ntohs(usvc.port), usvc.sched_name);
2429                 ret = -EFAULT;
2430                 goto out_unlock;
2431         }
2432
2433         /* Lookup the exact service by <protocol, addr, port> or fwmark */
2434         rcu_read_lock();
2435         if (usvc.fwmark == 0)
2436                 svc = __ip_vs_service_find(ipvs, usvc.af, usvc.protocol,
2437                                            &usvc.addr, usvc.port);
2438         else
2439                 svc = __ip_vs_svc_fwm_find(ipvs, usvc.af, usvc.fwmark);
2440         rcu_read_unlock();
2441
2442         if (cmd != IP_VS_SO_SET_ADD
2443             && (svc == NULL || svc->protocol != usvc.protocol)) {
2444                 ret = -ESRCH;
2445                 goto out_unlock;
2446         }
2447
2448         switch (cmd) {
2449         case IP_VS_SO_SET_ADD:
2450                 if (svc != NULL)
2451                         ret = -EEXIST;
2452                 else
2453                         ret = ip_vs_add_service(ipvs, &usvc, &svc);
2454                 break;
2455         case IP_VS_SO_SET_EDIT:
2456                 ret = ip_vs_edit_service(svc, &usvc);
2457                 break;
2458         case IP_VS_SO_SET_DEL:
2459                 ret = ip_vs_del_service(svc);
2460                 if (!ret)
2461                         goto out_unlock;
2462                 break;
2463         case IP_VS_SO_SET_ZERO:
2464                 ret = ip_vs_zero_service(svc);
2465                 break;
2466         case IP_VS_SO_SET_ADDDEST:
2467                 ret = ip_vs_add_dest(svc, &udest);
2468                 break;
2469         case IP_VS_SO_SET_EDITDEST:
2470                 ret = ip_vs_edit_dest(svc, &udest);
2471                 break;
2472         case IP_VS_SO_SET_DELDEST:
2473                 ret = ip_vs_del_dest(svc, &udest);
2474                 break;
2475         default:
2476                 ret = -EINVAL;
2477         }
2478
2479   out_unlock:
2480         mutex_unlock(&__ip_vs_mutex);
2481   out_dec:
2482         /* decrease the module use count */
2483         ip_vs_use_count_dec();
2484
2485         return ret;
2486 }
2487
2488
2489 static void
2490 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2491 {
2492         struct ip_vs_scheduler *sched;
2493         struct ip_vs_kstats kstats;
2494         char *sched_name;
2495
2496         sched = rcu_dereference_protected(src->scheduler, 1);
2497         sched_name = sched ? sched->name : "none";
2498         dst->protocol = src->protocol;
2499         dst->addr = src->addr.ip;
2500         dst->port = src->port;
2501         dst->fwmark = src->fwmark;
2502         strlcpy(dst->sched_name, sched_name, sizeof(dst->sched_name));
2503         dst->flags = src->flags;
2504         dst->timeout = src->timeout / HZ;
2505         dst->netmask = src->netmask;
2506         dst->num_dests = src->num_dests;
2507         ip_vs_copy_stats(&kstats, &src->stats);
2508         ip_vs_export_stats_user(&dst->stats, &kstats);
2509 }
2510
2511 static inline int
2512 __ip_vs_get_service_entries(struct netns_ipvs *ipvs,
2513                             const struct ip_vs_get_services *get,
2514                             struct ip_vs_get_services __user *uptr)
2515 {
2516         int idx, count=0;
2517         struct ip_vs_service *svc;
2518         struct ip_vs_service_entry entry;
2519         int ret = 0;
2520
2521         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2522                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2523                         /* Only expose IPv4 entries to old interface */
2524                         if (svc->af != AF_INET || (svc->ipvs != ipvs))
2525                                 continue;
2526
2527                         if (count >= get->num_services)
2528                                 goto out;
2529                         memset(&entry, 0, sizeof(entry));
2530                         ip_vs_copy_service(&entry, svc);
2531                         if (copy_to_user(&uptr->entrytable[count],
2532                                          &entry, sizeof(entry))) {
2533                                 ret = -EFAULT;
2534                                 goto out;
2535                         }
2536                         count++;
2537                 }
2538         }
2539
2540         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2541                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2542                         /* Only expose IPv4 entries to old interface */
2543                         if (svc->af != AF_INET || (svc->ipvs != ipvs))
2544                                 continue;
2545
2546                         if (count >= get->num_services)
2547                                 goto out;
2548                         memset(&entry, 0, sizeof(entry));
2549                         ip_vs_copy_service(&entry, svc);
2550                         if (copy_to_user(&uptr->entrytable[count],
2551                                          &entry, sizeof(entry))) {
2552                                 ret = -EFAULT;
2553                                 goto out;
2554                         }
2555                         count++;
2556                 }
2557         }
2558 out:
2559         return ret;
2560 }
2561
2562 static inline int
2563 __ip_vs_get_dest_entries(struct netns_ipvs *ipvs, const struct ip_vs_get_dests *get,
2564                          struct ip_vs_get_dests __user *uptr)
2565 {
2566         struct ip_vs_service *svc;
2567         union nf_inet_addr addr = { .ip = get->addr };
2568         int ret = 0;
2569
2570         rcu_read_lock();
2571         if (get->fwmark)
2572                 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, get->fwmark);
2573         else
2574                 svc = __ip_vs_service_find(ipvs, AF_INET, get->protocol, &addr,
2575                                            get->port);
2576         rcu_read_unlock();
2577
2578         if (svc) {
2579                 int count = 0;
2580                 struct ip_vs_dest *dest;
2581                 struct ip_vs_dest_entry entry;
2582                 struct ip_vs_kstats kstats;
2583
2584                 memset(&entry, 0, sizeof(entry));
2585                 list_for_each_entry(dest, &svc->destinations, n_list) {
2586                         if (count >= get->num_dests)
2587                                 break;
2588
2589                         /* Cannot expose heterogeneous members via sockopt
2590                          * interface
2591                          */
2592                         if (dest->af != svc->af)
2593                                 continue;
2594
2595                         entry.addr = dest->addr.ip;
2596                         entry.port = dest->port;
2597                         entry.conn_flags = atomic_read(&dest->conn_flags);
2598                         entry.weight = atomic_read(&dest->weight);
2599                         entry.u_threshold = dest->u_threshold;
2600                         entry.l_threshold = dest->l_threshold;
2601                         entry.activeconns = atomic_read(&dest->activeconns);
2602                         entry.inactconns = atomic_read(&dest->inactconns);
2603                         entry.persistconns = atomic_read(&dest->persistconns);
2604                         ip_vs_copy_stats(&kstats, &dest->stats);
2605                         ip_vs_export_stats_user(&entry.stats, &kstats);
2606                         if (copy_to_user(&uptr->entrytable[count],
2607                                          &entry, sizeof(entry))) {
2608                                 ret = -EFAULT;
2609                                 break;
2610                         }
2611                         count++;
2612                 }
2613         } else
2614                 ret = -ESRCH;
2615         return ret;
2616 }
2617
2618 static inline void
2619 __ip_vs_get_timeouts(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2620 {
2621 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2622         struct ip_vs_proto_data *pd;
2623 #endif
2624
2625         memset(u, 0, sizeof (*u));
2626
2627 #ifdef CONFIG_IP_VS_PROTO_TCP
2628         pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2629         u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2630         u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2631 #endif
2632 #ifdef CONFIG_IP_VS_PROTO_UDP
2633         pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2634         u->udp_timeout =
2635                         pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2636 #endif
2637 }
2638
2639 static const unsigned char get_arglen[CMDID(IP_VS_SO_GET_MAX) + 1] = {
2640         [CMDID(IP_VS_SO_GET_VERSION)]  = 64,
2641         [CMDID(IP_VS_SO_GET_INFO)]     = sizeof(struct ip_vs_getinfo),
2642         [CMDID(IP_VS_SO_GET_SERVICES)] = sizeof(struct ip_vs_get_services),
2643         [CMDID(IP_VS_SO_GET_SERVICE)]  = sizeof(struct ip_vs_service_entry),
2644         [CMDID(IP_VS_SO_GET_DESTS)]    = sizeof(struct ip_vs_get_dests),
2645         [CMDID(IP_VS_SO_GET_TIMEOUT)]  = sizeof(struct ip_vs_timeout_user),
2646         [CMDID(IP_VS_SO_GET_DAEMON)]   = 2 * sizeof(struct ip_vs_daemon_user),
2647 };
2648
2649 union ip_vs_get_arglen {
2650         char                            field_IP_VS_SO_GET_VERSION[64];
2651         struct ip_vs_getinfo            field_IP_VS_SO_GET_INFO;
2652         struct ip_vs_get_services       field_IP_VS_SO_GET_SERVICES;
2653         struct ip_vs_service_entry      field_IP_VS_SO_GET_SERVICE;
2654         struct ip_vs_get_dests          field_IP_VS_SO_GET_DESTS;
2655         struct ip_vs_timeout_user       field_IP_VS_SO_GET_TIMEOUT;
2656         struct ip_vs_daemon_user        field_IP_VS_SO_GET_DAEMON[2];
2657 };
2658
2659 #define MAX_GET_ARGLEN  sizeof(union ip_vs_get_arglen)
2660
2661 static int
2662 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2663 {
2664         unsigned char arg[MAX_GET_ARGLEN];
2665         int ret = 0;
2666         unsigned int copylen;
2667         struct net *net = sock_net(sk);
2668         struct netns_ipvs *ipvs = net_ipvs(net);
2669
2670         BUG_ON(!net);
2671         BUILD_BUG_ON(sizeof(arg) > 255);
2672         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2673                 return -EPERM;
2674
2675         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2676                 return -EINVAL;
2677
2678         copylen = get_arglen[CMDID(cmd)];
2679         if (*len < (int) copylen) {
2680                 IP_VS_DBG(1, "get_ctl: len %d < %u\n", *len, copylen);
2681                 return -EINVAL;
2682         }
2683
2684         if (copy_from_user(arg, user, copylen) != 0)
2685                 return -EFAULT;
2686         /*
2687          * Handle daemons first since it has its own locking
2688          */
2689         if (cmd == IP_VS_SO_GET_DAEMON) {
2690                 struct ip_vs_daemon_user d[2];
2691
2692                 memset(&d, 0, sizeof(d));
2693                 mutex_lock(&ipvs->sync_mutex);
2694                 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2695                         d[0].state = IP_VS_STATE_MASTER;
2696                         strlcpy(d[0].mcast_ifn, ipvs->mcfg.mcast_ifn,
2697                                 sizeof(d[0].mcast_ifn));
2698                         d[0].syncid = ipvs->mcfg.syncid;
2699                 }
2700                 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2701                         d[1].state = IP_VS_STATE_BACKUP;
2702                         strlcpy(d[1].mcast_ifn, ipvs->bcfg.mcast_ifn,
2703                                 sizeof(d[1].mcast_ifn));
2704                         d[1].syncid = ipvs->bcfg.syncid;
2705                 }
2706                 if (copy_to_user(user, &d, sizeof(d)) != 0)
2707                         ret = -EFAULT;
2708                 mutex_unlock(&ipvs->sync_mutex);
2709                 return ret;
2710         }
2711
2712         mutex_lock(&__ip_vs_mutex);
2713         switch (cmd) {
2714         case IP_VS_SO_GET_VERSION:
2715         {
2716                 char buf[64];
2717
2718                 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2719                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2720                 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2721                         ret = -EFAULT;
2722                         goto out;
2723                 }
2724                 *len = strlen(buf)+1;
2725         }
2726         break;
2727
2728         case IP_VS_SO_GET_INFO:
2729         {
2730                 struct ip_vs_getinfo info;
2731                 info.version = IP_VS_VERSION_CODE;
2732                 info.size = ip_vs_conn_tab_size;
2733                 info.num_services = ipvs->num_services;
2734                 if (copy_to_user(user, &info, sizeof(info)) != 0)
2735                         ret = -EFAULT;
2736         }
2737         break;
2738
2739         case IP_VS_SO_GET_SERVICES:
2740         {
2741                 struct ip_vs_get_services *get;
2742                 int size;
2743
2744                 get = (struct ip_vs_get_services *)arg;
2745                 size = sizeof(*get) +
2746                         sizeof(struct ip_vs_service_entry) * get->num_services;
2747                 if (*len != size) {
2748                         pr_err("length: %u != %u\n", *len, size);
2749                         ret = -EINVAL;
2750                         goto out;
2751                 }
2752                 ret = __ip_vs_get_service_entries(ipvs, get, user);
2753         }
2754         break;
2755
2756         case IP_VS_SO_GET_SERVICE:
2757         {
2758                 struct ip_vs_service_entry *entry;
2759                 struct ip_vs_service *svc;
2760                 union nf_inet_addr addr;
2761
2762                 entry = (struct ip_vs_service_entry *)arg;
2763                 addr.ip = entry->addr;
2764                 rcu_read_lock();
2765                 if (entry->fwmark)
2766                         svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, entry->fwmark);
2767                 else
2768                         svc = __ip_vs_service_find(ipvs, AF_INET,
2769                                                    entry->protocol, &addr,
2770                                                    entry->port);
2771                 rcu_read_unlock();
2772                 if (svc) {
2773                         ip_vs_copy_service(entry, svc);
2774                         if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2775                                 ret = -EFAULT;
2776                 } else
2777                         ret = -ESRCH;
2778         }
2779         break;
2780
2781         case IP_VS_SO_GET_DESTS:
2782         {
2783                 struct ip_vs_get_dests *get;
2784                 int size;
2785
2786                 get = (struct ip_vs_get_dests *)arg;
2787                 size = sizeof(*get) +
2788                         sizeof(struct ip_vs_dest_entry) * get->num_dests;
2789                 if (*len != size) {
2790                         pr_err("length: %u != %u\n", *len, size);
2791                         ret = -EINVAL;
2792                         goto out;
2793                 }
2794                 ret = __ip_vs_get_dest_entries(ipvs, get, user);
2795         }
2796         break;
2797
2798         case IP_VS_SO_GET_TIMEOUT:
2799         {
2800                 struct ip_vs_timeout_user t;
2801
2802                 __ip_vs_get_timeouts(ipvs, &t);
2803                 if (copy_to_user(user, &t, sizeof(t)) != 0)
2804                         ret = -EFAULT;
2805         }
2806         break;
2807
2808         default:
2809                 ret = -EINVAL;
2810         }
2811
2812 out:
2813         mutex_unlock(&__ip_vs_mutex);
2814         return ret;
2815 }
2816
2817
2818 static struct nf_sockopt_ops ip_vs_sockopts = {
2819         .pf             = PF_INET,
2820         .set_optmin     = IP_VS_BASE_CTL,
2821         .set_optmax     = IP_VS_SO_SET_MAX+1,
2822         .set            = do_ip_vs_set_ctl,
2823         .get_optmin     = IP_VS_BASE_CTL,
2824         .get_optmax     = IP_VS_SO_GET_MAX+1,
2825         .get            = do_ip_vs_get_ctl,
2826         .owner          = THIS_MODULE,
2827 };
2828
2829 /*
2830  * Generic Netlink interface
2831  */
2832
2833 /* IPVS genetlink family */
2834 static struct genl_family ip_vs_genl_family;
2835
2836 /* Policy used for first-level command attributes */
2837 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2838         [IPVS_CMD_ATTR_SERVICE]         = { .type = NLA_NESTED },
2839         [IPVS_CMD_ATTR_DEST]            = { .type = NLA_NESTED },
2840         [IPVS_CMD_ATTR_DAEMON]          = { .type = NLA_NESTED },
2841         [IPVS_CMD_ATTR_TIMEOUT_TCP]     = { .type = NLA_U32 },
2842         [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2843         [IPVS_CMD_ATTR_TIMEOUT_UDP]     = { .type = NLA_U32 },
2844 };
2845
2846 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2847 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2848         [IPVS_DAEMON_ATTR_STATE]        = { .type = NLA_U32 },
2849         [IPVS_DAEMON_ATTR_MCAST_IFN]    = { .type = NLA_NUL_STRING,
2850                                             .len = IP_VS_IFNAME_MAXLEN },
2851         [IPVS_DAEMON_ATTR_SYNC_ID]      = { .type = NLA_U32 },
2852         [IPVS_DAEMON_ATTR_SYNC_MAXLEN]  = { .type = NLA_U16 },
2853         [IPVS_DAEMON_ATTR_MCAST_GROUP]  = { .type = NLA_U32 },
2854         [IPVS_DAEMON_ATTR_MCAST_GROUP6] = { .len = sizeof(struct in6_addr) },
2855         [IPVS_DAEMON_ATTR_MCAST_PORT]   = { .type = NLA_U16 },
2856         [IPVS_DAEMON_ATTR_MCAST_TTL]    = { .type = NLA_U8 },
2857 };
2858
2859 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2860 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2861         [IPVS_SVC_ATTR_AF]              = { .type = NLA_U16 },
2862         [IPVS_SVC_ATTR_PROTOCOL]        = { .type = NLA_U16 },
2863         [IPVS_SVC_ATTR_ADDR]            = { .type = NLA_BINARY,
2864                                             .len = sizeof(union nf_inet_addr) },
2865         [IPVS_SVC_ATTR_PORT]            = { .type = NLA_U16 },
2866         [IPVS_SVC_ATTR_FWMARK]          = { .type = NLA_U32 },
2867         [IPVS_SVC_ATTR_SCHED_NAME]      = { .type = NLA_NUL_STRING,
2868                                             .len = IP_VS_SCHEDNAME_MAXLEN },
2869         [IPVS_SVC_ATTR_PE_NAME]         = { .type = NLA_NUL_STRING,
2870                                             .len = IP_VS_PENAME_MAXLEN },
2871         [IPVS_SVC_ATTR_FLAGS]           = { .type = NLA_BINARY,
2872                                             .len = sizeof(struct ip_vs_flags) },
2873         [IPVS_SVC_ATTR_TIMEOUT]         = { .type = NLA_U32 },
2874         [IPVS_SVC_ATTR_NETMASK]         = { .type = NLA_U32 },
2875         [IPVS_SVC_ATTR_STATS]           = { .type = NLA_NESTED },
2876 };
2877
2878 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2879 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2880         [IPVS_DEST_ATTR_ADDR]           = { .type = NLA_BINARY,
2881                                             .len = sizeof(union nf_inet_addr) },
2882         [IPVS_DEST_ATTR_PORT]           = { .type = NLA_U16 },
2883         [IPVS_DEST_ATTR_FWD_METHOD]     = { .type = NLA_U32 },
2884         [IPVS_DEST_ATTR_WEIGHT]         = { .type = NLA_U32 },
2885         [IPVS_DEST_ATTR_U_THRESH]       = { .type = NLA_U32 },
2886         [IPVS_DEST_ATTR_L_THRESH]       = { .type = NLA_U32 },
2887         [IPVS_DEST_ATTR_ACTIVE_CONNS]   = { .type = NLA_U32 },
2888         [IPVS_DEST_ATTR_INACT_CONNS]    = { .type = NLA_U32 },
2889         [IPVS_DEST_ATTR_PERSIST_CONNS]  = { .type = NLA_U32 },
2890         [IPVS_DEST_ATTR_STATS]          = { .type = NLA_NESTED },
2891         [IPVS_DEST_ATTR_ADDR_FAMILY]    = { .type = NLA_U16 },
2892 };
2893
2894 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2895                                  struct ip_vs_kstats *kstats)
2896 {
2897         struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2898
2899         if (!nl_stats)
2900                 return -EMSGSIZE;
2901
2902         if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, (u32)kstats->conns) ||
2903             nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, (u32)kstats->inpkts) ||
2904             nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, (u32)kstats->outpkts) ||
2905             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes,
2906                               IPVS_STATS_ATTR_PAD) ||
2907             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes,
2908                               IPVS_STATS_ATTR_PAD) ||
2909             nla_put_u32(skb, IPVS_STATS_ATTR_CPS, (u32)kstats->cps) ||
2910             nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, (u32)kstats->inpps) ||
2911             nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, (u32)kstats->outpps) ||
2912             nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, (u32)kstats->inbps) ||
2913             nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, (u32)kstats->outbps))
2914                 goto nla_put_failure;
2915         nla_nest_end(skb, nl_stats);
2916
2917         return 0;
2918
2919 nla_put_failure:
2920         nla_nest_cancel(skb, nl_stats);
2921         return -EMSGSIZE;
2922 }
2923
2924 static int ip_vs_genl_fill_stats64(struct sk_buff *skb, int container_type,
2925                                    struct ip_vs_kstats *kstats)
2926 {
2927         struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2928
2929         if (!nl_stats)
2930                 return -EMSGSIZE;
2931
2932         if (nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CONNS, kstats->conns,
2933                               IPVS_STATS_ATTR_PAD) ||
2934             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPKTS, kstats->inpkts,
2935                               IPVS_STATS_ATTR_PAD) ||
2936             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPKTS, kstats->outpkts,
2937                               IPVS_STATS_ATTR_PAD) ||
2938             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes,
2939                               IPVS_STATS_ATTR_PAD) ||
2940             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes,
2941                               IPVS_STATS_ATTR_PAD) ||
2942             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CPS, kstats->cps,
2943                               IPVS_STATS_ATTR_PAD) ||
2944             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPPS, kstats->inpps,
2945                               IPVS_STATS_ATTR_PAD) ||
2946             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPPS, kstats->outpps,
2947                               IPVS_STATS_ATTR_PAD) ||
2948             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBPS, kstats->inbps,
2949                               IPVS_STATS_ATTR_PAD) ||
2950             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBPS, kstats->outbps,
2951                               IPVS_STATS_ATTR_PAD))
2952                 goto nla_put_failure;
2953         nla_nest_end(skb, nl_stats);
2954
2955         return 0;
2956
2957 nla_put_failure:
2958         nla_nest_cancel(skb, nl_stats);
2959         return -EMSGSIZE;
2960 }
2961
2962 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2963                                    struct ip_vs_service *svc)
2964 {
2965         struct ip_vs_scheduler *sched;
2966         struct ip_vs_pe *pe;
2967         struct nlattr *nl_service;
2968         struct ip_vs_flags flags = { .flags = svc->flags,
2969                                      .mask = ~0 };
2970         struct ip_vs_kstats kstats;
2971         char *sched_name;
2972
2973         nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2974         if (!nl_service)
2975                 return -EMSGSIZE;
2976
2977         if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2978                 goto nla_put_failure;
2979         if (svc->fwmark) {
2980                 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2981                         goto nla_put_failure;
2982         } else {
2983                 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2984                     nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2985                     nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2986                         goto nla_put_failure;
2987         }
2988
2989         sched = rcu_dereference_protected(svc->scheduler, 1);
2990         sched_name = sched ? sched->name : "none";
2991         pe = rcu_dereference_protected(svc->pe, 1);
2992         if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched_name) ||
2993             (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
2994             nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2995             nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2996             nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
2997                 goto nla_put_failure;
2998         ip_vs_copy_stats(&kstats, &svc->stats);
2999         if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &kstats))
3000                 goto nla_put_failure;
3001         if (ip_vs_genl_fill_stats64(skb, IPVS_SVC_ATTR_STATS64, &kstats))
3002                 goto nla_put_failure;
3003
3004         nla_nest_end(skb, nl_service);
3005
3006         return 0;
3007
3008 nla_put_failure:
3009         nla_nest_cancel(skb, nl_service);
3010         return -EMSGSIZE;
3011 }
3012
3013 static int ip_vs_genl_dump_service(struct sk_buff *skb,
3014                                    struct ip_vs_service *svc,
3015                                    struct netlink_callback *cb)
3016 {
3017         void *hdr;
3018
3019         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3020                           &ip_vs_genl_family, NLM_F_MULTI,
3021                           IPVS_CMD_NEW_SERVICE);
3022         if (!hdr)
3023                 return -EMSGSIZE;
3024
3025         if (ip_vs_genl_fill_service(skb, svc) < 0)
3026                 goto nla_put_failure;
3027
3028         genlmsg_end(skb, hdr);
3029         return 0;
3030
3031 nla_put_failure:
3032         genlmsg_cancel(skb, hdr);
3033         return -EMSGSIZE;
3034 }
3035
3036 static int ip_vs_genl_dump_services(struct sk_buff *skb,
3037                                     struct netlink_callback *cb)
3038 {
3039         int idx = 0, i;
3040         int start = cb->args[0];
3041         struct ip_vs_service *svc;
3042         struct net *net = sock_net(skb->sk);
3043         struct netns_ipvs *ipvs = net_ipvs(net);
3044
3045         mutex_lock(&__ip_vs_mutex);
3046         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3047                 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
3048                         if (++idx <= start || (svc->ipvs != ipvs))
3049                                 continue;
3050                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3051                                 idx--;
3052                                 goto nla_put_failure;
3053                         }
3054                 }
3055         }
3056
3057         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3058                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
3059                         if (++idx <= start || (svc->ipvs != ipvs))
3060                                 continue;
3061                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3062                                 idx--;
3063                                 goto nla_put_failure;
3064                         }
3065                 }
3066         }
3067
3068 nla_put_failure:
3069         mutex_unlock(&__ip_vs_mutex);
3070         cb->args[0] = idx;
3071
3072         return skb->len;
3073 }
3074
3075 static bool ip_vs_is_af_valid(int af)
3076 {
3077         if (af == AF_INET)
3078                 return true;
3079 #ifdef CONFIG_IP_VS_IPV6
3080         if (af == AF_INET6 && ipv6_mod_enabled())
3081                 return true;
3082 #endif
3083         return false;
3084 }
3085
3086 static int ip_vs_genl_parse_service(struct netns_ipvs *ipvs,
3087                                     struct ip_vs_service_user_kern *usvc,
3088                                     struct nlattr *nla, int full_entry,
3089                                     struct ip_vs_service **ret_svc)
3090 {
3091         struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
3092         struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
3093         struct ip_vs_service *svc;
3094
3095         /* Parse mandatory identifying service fields first */
3096         if (nla == NULL ||
3097             nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla,
3098                              ip_vs_svc_policy, NULL))
3099                 return -EINVAL;
3100
3101         nla_af          = attrs[IPVS_SVC_ATTR_AF];
3102         nla_protocol    = attrs[IPVS_SVC_ATTR_PROTOCOL];
3103         nla_addr        = attrs[IPVS_SVC_ATTR_ADDR];
3104         nla_port        = attrs[IPVS_SVC_ATTR_PORT];
3105         nla_fwmark      = attrs[IPVS_SVC_ATTR_FWMARK];
3106
3107         if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3108                 return -EINVAL;
3109
3110         memset(usvc, 0, sizeof(*usvc));
3111
3112         usvc->af = nla_get_u16(nla_af);
3113         if (!ip_vs_is_af_valid(usvc->af))
3114                 return -EAFNOSUPPORT;
3115
3116         if (nla_fwmark) {
3117                 usvc->protocol = IPPROTO_TCP;
3118                 usvc->fwmark = nla_get_u32(nla_fwmark);
3119         } else {
3120                 usvc->protocol = nla_get_u16(nla_protocol);
3121                 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3122                 usvc->port = nla_get_be16(nla_port);
3123                 usvc->fwmark = 0;
3124         }
3125
3126         rcu_read_lock();
3127         if (usvc->fwmark)
3128                 svc = __ip_vs_svc_fwm_find(ipvs, usvc->af, usvc->fwmark);
3129         else
3130                 svc = __ip_vs_service_find(ipvs, usvc->af, usvc->protocol,
3131                                            &usvc->addr, usvc->port);
3132         rcu_read_unlock();
3133         *ret_svc = svc;
3134
3135         /* If a full entry was requested, check for the additional fields */
3136         if (full_entry) {
3137                 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3138                               *nla_netmask;
3139                 struct ip_vs_flags flags;
3140
3141                 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3142                 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3143                 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3144                 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3145                 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3146
3147                 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3148                         return -EINVAL;
3149
3150                 nla_memcpy(&flags, nla_flags, sizeof(flags));
3151
3152                 /* prefill flags from service if it already exists */
3153                 if (svc)
3154                         usvc->flags = svc->flags;
3155
3156                 /* set new flags from userland */
3157                 usvc->flags = (usvc->flags & ~flags.mask) |
3158                               (flags.flags & flags.mask);
3159                 usvc->sched_name = nla_data(nla_sched);
3160                 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3161                 usvc->timeout = nla_get_u32(nla_timeout);
3162                 usvc->netmask = nla_get_be32(nla_netmask);
3163         }
3164
3165         return 0;
3166 }
3167
3168 static struct ip_vs_service *ip_vs_genl_find_service(struct netns_ipvs *ipvs,
3169                                                      struct nlattr *nla)
3170 {
3171         struct ip_vs_service_user_kern usvc;
3172         struct ip_vs_service *svc;
3173         int ret;
3174
3175         ret = ip_vs_genl_parse_service(ipvs, &usvc, nla, 0, &svc);
3176         return ret ? ERR_PTR(ret) : svc;
3177 }
3178
3179 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3180 {
3181         struct nlattr *nl_dest;
3182         struct ip_vs_kstats kstats;
3183
3184         nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3185         if (!nl_dest)
3186                 return -EMSGSIZE;
3187
3188         if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3189             nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3190             nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3191                         (atomic_read(&dest->conn_flags) &
3192                          IP_VS_CONN_F_FWD_MASK)) ||
3193             nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3194                         atomic_read(&dest->weight)) ||
3195             nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3196             nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3197             nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3198                         atomic_read(&dest->activeconns)) ||
3199             nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3200                         atomic_read(&dest->inactconns)) ||
3201             nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3202                         atomic_read(&dest->persistconns)) ||
3203             nla_put_u16(skb, IPVS_DEST_ATTR_ADDR_FAMILY, dest->af))
3204                 goto nla_put_failure;
3205         ip_vs_copy_stats(&kstats, &dest->stats);
3206         if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &kstats))
3207                 goto nla_put_failure;
3208         if (ip_vs_genl_fill_stats64(skb, IPVS_DEST_ATTR_STATS64, &kstats))
3209                 goto nla_put_failure;
3210
3211         nla_nest_end(skb, nl_dest);
3212
3213         return 0;
3214
3215 nla_put_failure:
3216         nla_nest_cancel(skb, nl_dest);
3217         return -EMSGSIZE;
3218 }
3219
3220 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3221                                 struct netlink_callback *cb)
3222 {
3223         void *hdr;
3224
3225         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3226                           &ip_vs_genl_family, NLM_F_MULTI,
3227                           IPVS_CMD_NEW_DEST);
3228         if (!hdr)
3229                 return -EMSGSIZE;
3230
3231         if (ip_vs_genl_fill_dest(skb, dest) < 0)
3232                 goto nla_put_failure;
3233
3234         genlmsg_end(skb, hdr);
3235         return 0;
3236
3237 nla_put_failure:
3238         genlmsg_cancel(skb, hdr);
3239         return -EMSGSIZE;
3240 }
3241
3242 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3243                                  struct netlink_callback *cb)
3244 {
3245         int idx = 0;
3246         int start = cb->args[0];
3247         struct ip_vs_service *svc;
3248         struct ip_vs_dest *dest;
3249         struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3250         struct net *net = sock_net(skb->sk);
3251         struct netns_ipvs *ipvs = net_ipvs(net);
3252
3253         mutex_lock(&__ip_vs_mutex);
3254
3255         /* Try to find the service for which to dump destinations */
3256         if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs, IPVS_CMD_ATTR_MAX,
3257                         ip_vs_cmd_policy, NULL))
3258                 goto out_err;
3259
3260
3261         svc = ip_vs_genl_find_service(ipvs, attrs[IPVS_CMD_ATTR_SERVICE]);
3262         if (IS_ERR_OR_NULL(svc))
3263                 goto out_err;
3264
3265         /* Dump the destinations */
3266         list_for_each_entry(dest, &svc->destinations, n_list) {
3267                 if (++idx <= start)
3268                         continue;
3269                 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3270                         idx--;
3271                         goto nla_put_failure;
3272                 }
3273         }
3274
3275 nla_put_failure:
3276         cb->args[0] = idx;
3277
3278 out_err:
3279         mutex_unlock(&__ip_vs_mutex);
3280
3281         return skb->len;
3282 }
3283
3284 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3285                                  struct nlattr *nla, int full_entry)
3286 {
3287         struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3288         struct nlattr *nla_addr, *nla_port;
3289         struct nlattr *nla_addr_family;
3290
3291         /* Parse mandatory identifying destination fields first */
3292         if (nla == NULL ||
3293             nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla,
3294                              ip_vs_dest_policy, NULL))
3295                 return -EINVAL;
3296
3297         nla_addr        = attrs[IPVS_DEST_ATTR_ADDR];
3298         nla_port        = attrs[IPVS_DEST_ATTR_PORT];
3299         nla_addr_family = attrs[IPVS_DEST_ATTR_ADDR_FAMILY];
3300
3301         if (!(nla_addr && nla_port))
3302                 return -EINVAL;
3303
3304         memset(udest, 0, sizeof(*udest));
3305
3306         nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3307         udest->port = nla_get_be16(nla_port);
3308
3309         if (nla_addr_family)
3310                 udest->af = nla_get_u16(nla_addr_family);
3311         else
3312                 udest->af = 0;
3313
3314         /* If a full entry was requested, check for the additional fields */
3315         if (full_entry) {
3316                 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3317                               *nla_l_thresh;
3318
3319                 nla_fwd         = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3320                 nla_weight      = attrs[IPVS_DEST_ATTR_WEIGHT];
3321                 nla_u_thresh    = attrs[IPVS_DEST_ATTR_U_THRESH];
3322                 nla_l_thresh    = attrs[IPVS_DEST_ATTR_L_THRESH];
3323
3324                 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3325                         return -EINVAL;
3326
3327                 udest->conn_flags = nla_get_u32(nla_fwd)
3328                                     & IP_VS_CONN_F_FWD_MASK;
3329                 udest->weight = nla_get_u32(nla_weight);
3330                 udest->u_threshold = nla_get_u32(nla_u_thresh);
3331                 udest->l_threshold = nla_get_u32(nla_l_thresh);
3332         }
3333
3334         return 0;
3335 }
3336
3337 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3338                                   struct ipvs_sync_daemon_cfg *c)
3339 {
3340         struct nlattr *nl_daemon;
3341
3342         nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3343         if (!nl_daemon)
3344                 return -EMSGSIZE;
3345
3346         if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3347             nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, c->mcast_ifn) ||
3348             nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, c->syncid) ||
3349             nla_put_u16(skb, IPVS_DAEMON_ATTR_SYNC_MAXLEN, c->sync_maxlen) ||
3350             nla_put_u16(skb, IPVS_DAEMON_ATTR_MCAST_PORT, c->mcast_port) ||
3351             nla_put_u8(skb, IPVS_DAEMON_ATTR_MCAST_TTL, c->mcast_ttl))
3352                 goto nla_put_failure;
3353 #ifdef CONFIG_IP_VS_IPV6
3354         if (c->mcast_af == AF_INET6) {
3355                 if (nla_put_in6_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP6,
3356                                      &c->mcast_group.in6))
3357                         goto nla_put_failure;
3358         } else
3359 #endif
3360                 if (c->mcast_af == AF_INET &&
3361                     nla_put_in_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP,
3362                                     c->mcast_group.ip))
3363                         goto nla_put_failure;
3364         nla_nest_end(skb, nl_daemon);
3365
3366         return 0;
3367
3368 nla_put_failure:
3369         nla_nest_cancel(skb, nl_daemon);
3370         return -EMSGSIZE;
3371 }
3372
3373 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3374                                   struct ipvs_sync_daemon_cfg *c,
3375                                   struct netlink_callback *cb)
3376 {
3377         void *hdr;
3378         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3379                           &ip_vs_genl_family, NLM_F_MULTI,
3380                           IPVS_CMD_NEW_DAEMON);
3381         if (!hdr)
3382                 return -EMSGSIZE;
3383
3384         if (ip_vs_genl_fill_daemon(skb, state, c))
3385                 goto nla_put_failure;
3386
3387         genlmsg_end(skb, hdr);
3388         return 0;
3389
3390 nla_put_failure:
3391         genlmsg_cancel(skb, hdr);
3392         return -EMSGSIZE;
3393 }
3394
3395 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3396                                    struct netlink_callback *cb)
3397 {
3398         struct net *net = sock_net(skb->sk);
3399         struct netns_ipvs *ipvs = net_ipvs(net);
3400
3401         mutex_lock(&ipvs->sync_mutex);
3402         if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3403                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3404                                            &ipvs->mcfg, cb) < 0)
3405                         goto nla_put_failure;
3406
3407                 cb->args[0] = 1;
3408         }
3409
3410         if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3411                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3412                                            &ipvs->bcfg, cb) < 0)
3413                         goto nla_put_failure;
3414
3415                 cb->args[1] = 1;
3416         }
3417
3418 nla_put_failure:
3419         mutex_unlock(&ipvs->sync_mutex);
3420
3421         return skb->len;
3422 }
3423
3424 static int ip_vs_genl_new_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3425 {
3426         struct ipvs_sync_daemon_cfg c;
3427         struct nlattr *a;
3428         int ret;
3429
3430         memset(&c, 0, sizeof(c));
3431         if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3432               attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3433               attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3434                 return -EINVAL;
3435         strlcpy(c.mcast_ifn, nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3436                 sizeof(c.mcast_ifn));
3437         c.syncid = nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]);
3438
3439         a = attrs[IPVS_DAEMON_ATTR_SYNC_MAXLEN];
3440         if (a)
3441                 c.sync_maxlen = nla_get_u16(a);
3442
3443         a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP];
3444         if (a) {
3445                 c.mcast_af = AF_INET;
3446                 c.mcast_group.ip = nla_get_in_addr(a);
3447                 if (!ipv4_is_multicast(c.mcast_group.ip))
3448                         return -EINVAL;
3449         } else {
3450                 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP6];
3451                 if (a) {
3452 #ifdef CONFIG_IP_VS_IPV6
3453                         int addr_type;
3454
3455                         c.mcast_af = AF_INET6;
3456                         c.mcast_group.in6 = nla_get_in6_addr(a);
3457                         addr_type = ipv6_addr_type(&c.mcast_group.in6);
3458                         if (!(addr_type & IPV6_ADDR_MULTICAST))
3459                                 return -EINVAL;
3460 #else
3461                         return -EAFNOSUPPORT;
3462 #endif
3463                 }
3464         }
3465
3466         a = attrs[IPVS_DAEMON_ATTR_MCAST_PORT];
3467         if (a)
3468                 c.mcast_port = nla_get_u16(a);
3469
3470         a = attrs[IPVS_DAEMON_ATTR_MCAST_TTL];
3471         if (a)
3472                 c.mcast_ttl = nla_get_u8(a);
3473
3474         /* The synchronization protocol is incompatible with mixed family
3475          * services
3476          */
3477         if (ipvs->mixed_address_family_dests > 0)
3478                 return -EINVAL;
3479
3480         ret = start_sync_thread(ipvs, &c,
3481                                 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3482         return ret;
3483 }
3484
3485 static int ip_vs_genl_del_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3486 {
3487         int ret;
3488
3489         if (!attrs[IPVS_DAEMON_ATTR_STATE])
3490                 return -EINVAL;
3491
3492         mutex_lock(&ipvs->sync_mutex);
3493         ret = stop_sync_thread(ipvs,
3494                                nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3495         mutex_unlock(&ipvs->sync_mutex);
3496         return ret;
3497 }
3498
3499 static int ip_vs_genl_set_config(struct netns_ipvs *ipvs, struct nlattr **attrs)
3500 {
3501         struct ip_vs_timeout_user t;
3502
3503         __ip_vs_get_timeouts(ipvs, &t);
3504
3505         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3506                 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3507
3508         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3509                 t.tcp_fin_timeout =
3510                         nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3511
3512         if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3513                 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3514
3515         return ip_vs_set_timeout(ipvs, &t);
3516 }
3517
3518 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3519 {
3520         int ret = -EINVAL, cmd;
3521         struct net *net = sock_net(skb->sk);
3522         struct netns_ipvs *ipvs = net_ipvs(net);
3523
3524         cmd = info->genlhdr->cmd;
3525
3526         if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3527                 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3528
3529                 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3530                     nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3531                                      info->attrs[IPVS_CMD_ATTR_DAEMON],
3532                                      ip_vs_daemon_policy, info->extack))
3533                         goto out;
3534
3535                 if (cmd == IPVS_CMD_NEW_DAEMON)
3536                         ret = ip_vs_genl_new_daemon(ipvs, daemon_attrs);
3537                 else
3538                         ret = ip_vs_genl_del_daemon(ipvs, daemon_attrs);
3539         }
3540
3541 out:
3542         return ret;
3543 }
3544
3545 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3546 {
3547         struct ip_vs_service *svc = NULL;
3548         struct ip_vs_service_user_kern usvc;
3549         struct ip_vs_dest_user_kern udest;
3550         int ret = 0, cmd;
3551         int need_full_svc = 0, need_full_dest = 0;
3552         struct net *net = sock_net(skb->sk);
3553         struct netns_ipvs *ipvs = net_ipvs(net);
3554
3555         cmd = info->genlhdr->cmd;
3556
3557         mutex_lock(&__ip_vs_mutex);
3558
3559         if (cmd == IPVS_CMD_FLUSH) {
3560                 ret = ip_vs_flush(ipvs, false);
3561                 goto out;
3562         } else if (cmd == IPVS_CMD_SET_CONFIG) {
3563                 ret = ip_vs_genl_set_config(ipvs, info->attrs);
3564                 goto out;
3565         } else if (cmd == IPVS_CMD_ZERO &&
3566                    !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3567                 ret = ip_vs_zero_all(ipvs);
3568                 goto out;
3569         }
3570
3571         /* All following commands require a service argument, so check if we
3572          * received a valid one. We need a full service specification when
3573          * adding / editing a service. Only identifying members otherwise. */
3574         if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3575                 need_full_svc = 1;
3576
3577         ret = ip_vs_genl_parse_service(ipvs, &usvc,
3578                                        info->attrs[IPVS_CMD_ATTR_SERVICE],
3579                                        need_full_svc, &svc);
3580         if (ret)
3581                 goto out;
3582
3583         /* Unless we're adding a new service, the service must already exist */
3584         if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3585                 ret = -ESRCH;
3586                 goto out;
3587         }
3588
3589         /* Destination commands require a valid destination argument. For
3590          * adding / editing a destination, we need a full destination
3591          * specification. */
3592         if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3593             cmd == IPVS_CMD_DEL_DEST) {
3594                 if (cmd != IPVS_CMD_DEL_DEST)
3595                         need_full_dest = 1;
3596
3597                 ret = ip_vs_genl_parse_dest(&udest,
3598                                             info->attrs[IPVS_CMD_ATTR_DEST],
3599                                             need_full_dest);
3600                 if (ret)
3601                         goto out;
3602
3603                 /* Old protocols did not allow the user to specify address
3604                  * family, so we set it to zero instead.  We also didn't
3605                  * allow heterogeneous pools in the old code, so it's safe
3606                  * to assume that this will have the same address family as
3607                  * the service.
3608                  */
3609                 if (udest.af == 0)
3610                         udest.af = svc->af;
3611
3612                 if (!ip_vs_is_af_valid(udest.af)) {
3613                         ret = -EAFNOSUPPORT;
3614                         goto out;
3615                 }
3616
3617                 if (udest.af != svc->af && cmd != IPVS_CMD_DEL_DEST) {
3618                         /* The synchronization protocol is incompatible
3619                          * with mixed family services
3620                          */
3621                         if (ipvs->sync_state) {
3622                                 ret = -EINVAL;
3623                                 goto out;
3624                         }
3625
3626                         /* Which connection types do we support? */
3627                         switch (udest.conn_flags) {
3628                         case IP_VS_CONN_F_TUNNEL:
3629                                 /* We are able to forward this */
3630                                 break;
3631                         default:
3632                                 ret = -EINVAL;
3633                                 goto out;
3634                         }
3635                 }
3636         }
3637
3638         switch (cmd) {
3639         case IPVS_CMD_NEW_SERVICE:
3640                 if (svc == NULL)
3641                         ret = ip_vs_add_service(ipvs, &usvc, &svc);
3642                 else
3643                         ret = -EEXIST;
3644                 break;
3645         case IPVS_CMD_SET_SERVICE:
3646                 ret = ip_vs_edit_service(svc, &usvc);
3647                 break;
3648         case IPVS_CMD_DEL_SERVICE:
3649                 ret = ip_vs_del_service(svc);
3650                 /* do not use svc, it can be freed */
3651                 break;
3652         case IPVS_CMD_NEW_DEST:
3653                 ret = ip_vs_add_dest(svc, &udest);
3654                 break;
3655         case IPVS_CMD_SET_DEST:
3656                 ret = ip_vs_edit_dest(svc, &udest);
3657                 break;
3658         case IPVS_CMD_DEL_DEST:
3659                 ret = ip_vs_del_dest(svc, &udest);
3660                 break;
3661         case IPVS_CMD_ZERO:
3662                 ret = ip_vs_zero_service(svc);
3663                 break;
3664         default:
3665                 ret = -EINVAL;
3666         }
3667
3668 out:
3669         mutex_unlock(&__ip_vs_mutex);
3670
3671         return ret;
3672 }
3673
3674 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3675 {
3676         struct sk_buff *msg;
3677         void *reply;
3678         int ret, cmd, reply_cmd;
3679         struct net *net = sock_net(skb->sk);
3680         struct netns_ipvs *ipvs = net_ipvs(net);
3681
3682         cmd = info->genlhdr->cmd;
3683
3684         if (cmd == IPVS_CMD_GET_SERVICE)
3685                 reply_cmd = IPVS_CMD_NEW_SERVICE;
3686         else if (cmd == IPVS_CMD_GET_INFO)
3687                 reply_cmd = IPVS_CMD_SET_INFO;
3688         else if (cmd == IPVS_CMD_GET_CONFIG)
3689                 reply_cmd = IPVS_CMD_SET_CONFIG;
3690         else {
3691                 pr_err("unknown Generic Netlink command\n");
3692                 return -EINVAL;
3693         }
3694
3695         msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3696         if (!msg)
3697                 return -ENOMEM;
3698
3699         mutex_lock(&__ip_vs_mutex);
3700
3701         reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3702         if (reply == NULL)
3703                 goto nla_put_failure;
3704
3705         switch (cmd) {
3706         case IPVS_CMD_GET_SERVICE:
3707         {
3708                 struct ip_vs_service *svc;
3709
3710                 svc = ip_vs_genl_find_service(ipvs,
3711                                               info->attrs[IPVS_CMD_ATTR_SERVICE]);
3712                 if (IS_ERR(svc)) {
3713                         ret = PTR_ERR(svc);
3714                         goto out_err;
3715                 } else if (svc) {
3716                         ret = ip_vs_genl_fill_service(msg, svc);
3717                         if (ret)
3718                                 goto nla_put_failure;
3719                 } else {
3720                         ret = -ESRCH;
3721                         goto out_err;
3722                 }
3723
3724                 break;
3725         }
3726
3727         case IPVS_CMD_GET_CONFIG:
3728         {
3729                 struct ip_vs_timeout_user t;
3730
3731                 __ip_vs_get_timeouts(ipvs, &t);
3732 #ifdef CONFIG_IP_VS_PROTO_TCP
3733                 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3734                                 t.tcp_timeout) ||
3735                     nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3736                                 t.tcp_fin_timeout))
3737                         goto nla_put_failure;
3738 #endif
3739 #ifdef CONFIG_IP_VS_PROTO_UDP
3740                 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3741                         goto nla_put_failure;
3742 #endif
3743
3744                 break;
3745         }
3746
3747         case IPVS_CMD_GET_INFO:
3748                 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3749                                 IP_VS_VERSION_CODE) ||
3750                     nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3751                                 ip_vs_conn_tab_size))
3752                         goto nla_put_failure;
3753                 break;
3754         }
3755
3756         genlmsg_end(msg, reply);
3757         ret = genlmsg_reply(msg, info);
3758         goto out;
3759
3760 nla_put_failure:
3761         pr_err("not enough space in Netlink message\n");
3762         ret = -EMSGSIZE;
3763
3764 out_err:
3765         nlmsg_free(msg);
3766 out:
3767         mutex_unlock(&__ip_vs_mutex);
3768
3769         return ret;
3770 }
3771
3772
3773 static const struct genl_ops ip_vs_genl_ops[] = {
3774         {
3775                 .cmd    = IPVS_CMD_NEW_SERVICE,
3776                 .flags  = GENL_ADMIN_PERM,
3777                 .policy = ip_vs_cmd_policy,
3778                 .doit   = ip_vs_genl_set_cmd,
3779         },
3780         {
3781                 .cmd    = IPVS_CMD_SET_SERVICE,
3782                 .flags  = GENL_ADMIN_PERM,
3783                 .policy = ip_vs_cmd_policy,
3784                 .doit   = ip_vs_genl_set_cmd,
3785         },
3786         {
3787                 .cmd    = IPVS_CMD_DEL_SERVICE,
3788                 .flags  = GENL_ADMIN_PERM,
3789                 .policy = ip_vs_cmd_policy,
3790                 .doit   = ip_vs_genl_set_cmd,
3791         },
3792         {
3793                 .cmd    = IPVS_CMD_GET_SERVICE,
3794                 .flags  = GENL_ADMIN_PERM,
3795                 .doit   = ip_vs_genl_get_cmd,
3796                 .dumpit = ip_vs_genl_dump_services,
3797                 .policy = ip_vs_cmd_policy,
3798         },
3799         {
3800                 .cmd    = IPVS_CMD_NEW_DEST,
3801                 .flags  = GENL_ADMIN_PERM,
3802                 .policy = ip_vs_cmd_policy,
3803                 .doit   = ip_vs_genl_set_cmd,
3804         },
3805         {
3806                 .cmd    = IPVS_CMD_SET_DEST,
3807                 .flags  = GENL_ADMIN_PERM,
3808                 .policy = ip_vs_cmd_policy,
3809                 .doit   = ip_vs_genl_set_cmd,
3810         },
3811         {
3812                 .cmd    = IPVS_CMD_DEL_DEST,
3813                 .flags  = GENL_ADMIN_PERM,
3814                 .policy = ip_vs_cmd_policy,
3815                 .doit   = ip_vs_genl_set_cmd,
3816         },
3817         {
3818                 .cmd    = IPVS_CMD_GET_DEST,
3819                 .flags  = GENL_ADMIN_PERM,
3820                 .policy = ip_vs_cmd_policy,
3821                 .dumpit = ip_vs_genl_dump_dests,
3822         },
3823         {
3824                 .cmd    = IPVS_CMD_NEW_DAEMON,
3825                 .flags  = GENL_ADMIN_PERM,
3826                 .policy = ip_vs_cmd_policy,
3827                 .doit   = ip_vs_genl_set_daemon,
3828         },
3829         {
3830                 .cmd    = IPVS_CMD_DEL_DAEMON,
3831                 .flags  = GENL_ADMIN_PERM,
3832                 .policy = ip_vs_cmd_policy,
3833                 .doit   = ip_vs_genl_set_daemon,
3834         },
3835         {
3836                 .cmd    = IPVS_CMD_GET_DAEMON,
3837                 .flags  = GENL_ADMIN_PERM,
3838                 .dumpit = ip_vs_genl_dump_daemons,
3839         },
3840         {
3841                 .cmd    = IPVS_CMD_SET_CONFIG,
3842                 .flags  = GENL_ADMIN_PERM,
3843                 .policy = ip_vs_cmd_policy,
3844                 .doit   = ip_vs_genl_set_cmd,
3845         },
3846         {
3847                 .cmd    = IPVS_CMD_GET_CONFIG,
3848                 .flags  = GENL_ADMIN_PERM,
3849                 .doit   = ip_vs_genl_get_cmd,
3850         },
3851         {
3852                 .cmd    = IPVS_CMD_GET_INFO,
3853                 .flags  = GENL_ADMIN_PERM,
3854                 .doit   = ip_vs_genl_get_cmd,
3855         },
3856         {
3857                 .cmd    = IPVS_CMD_ZERO,
3858                 .flags  = GENL_ADMIN_PERM,
3859                 .policy = ip_vs_cmd_policy,
3860                 .doit   = ip_vs_genl_set_cmd,
3861         },
3862         {
3863                 .cmd    = IPVS_CMD_FLUSH,
3864                 .flags  = GENL_ADMIN_PERM,
3865                 .doit   = ip_vs_genl_set_cmd,
3866         },
3867 };
3868
3869 static struct genl_family ip_vs_genl_family __ro_after_init = {
3870         .hdrsize        = 0,
3871         .name           = IPVS_GENL_NAME,
3872         .version        = IPVS_GENL_VERSION,
3873         .maxattr        = IPVS_CMD_ATTR_MAX,
3874         .netnsok        = true,         /* Make ipvsadm to work on netns */
3875         .module         = THIS_MODULE,
3876         .ops            = ip_vs_genl_ops,
3877         .n_ops          = ARRAY_SIZE(ip_vs_genl_ops),
3878 };
3879
3880 static int __init ip_vs_genl_register(void)
3881 {
3882         return genl_register_family(&ip_vs_genl_family);
3883 }
3884
3885 static void ip_vs_genl_unregister(void)
3886 {
3887         genl_unregister_family(&ip_vs_genl_family);
3888 }
3889
3890 /* End of Generic Netlink interface definitions */
3891
3892 /*
3893  * per netns intit/exit func.
3894  */
3895 #ifdef CONFIG_SYSCTL
3896 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs)
3897 {
3898         struct net *net = ipvs->net;
3899         int idx;
3900         struct ctl_table *tbl;
3901
3902         atomic_set(&ipvs->dropentry, 0);
3903         spin_lock_init(&ipvs->dropentry_lock);
3904         spin_lock_init(&ipvs->droppacket_lock);
3905         spin_lock_init(&ipvs->securetcp_lock);
3906
3907         if (!net_eq(net, &init_net)) {
3908                 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3909                 if (tbl == NULL)
3910                         return -ENOMEM;
3911
3912                 /* Don't export sysctls to unprivileged users */
3913                 if (net->user_ns != &init_user_ns)
3914                         tbl[0].procname = NULL;
3915         } else
3916                 tbl = vs_vars;
3917         /* Initialize sysctl defaults */
3918         for (idx = 0; idx < ARRAY_SIZE(vs_vars); idx++) {
3919                 if (tbl[idx].proc_handler == proc_do_defense_mode)
3920                         tbl[idx].extra2 = ipvs;
3921         }
3922         idx = 0;
3923         ipvs->sysctl_amemthresh = 1024;
3924         tbl[idx++].data = &ipvs->sysctl_amemthresh;
3925         ipvs->sysctl_am_droprate = 10;
3926         tbl[idx++].data = &ipvs->sysctl_am_droprate;
3927         tbl[idx++].data = &ipvs->sysctl_drop_entry;
3928         tbl[idx++].data = &ipvs->sysctl_drop_packet;
3929 #ifdef CONFIG_IP_VS_NFCT
3930         tbl[idx++].data = &ipvs->sysctl_conntrack;
3931 #endif
3932         tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3933         ipvs->sysctl_snat_reroute = 1;
3934         tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3935         ipvs->sysctl_sync_ver = 1;
3936         tbl[idx++].data = &ipvs->sysctl_sync_ver;
3937         ipvs->sysctl_sync_ports = 1;
3938         tbl[idx++].data = &ipvs->sysctl_sync_ports;
3939         tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
3940         ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3941         tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3942         ipvs->sysctl_sync_sock_size = 0;
3943         tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3944         tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3945         tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3946         tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
3947         tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
3948         tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3949         ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3950         ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3951         tbl[idx].data = &ipvs->sysctl_sync_threshold;
3952         tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3953         ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3954         tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3955         ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3956         tbl[idx++].data = &ipvs->sysctl_sync_retries;
3957         tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3958         ipvs->sysctl_pmtu_disc = 1;
3959         tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
3960         tbl[idx++].data = &ipvs->sysctl_backup_only;
3961         ipvs->sysctl_conn_reuse_mode = 1;
3962         tbl[idx++].data = &ipvs->sysctl_conn_reuse_mode;
3963         tbl[idx++].data = &ipvs->sysctl_schedule_icmp;
3964         tbl[idx++].data = &ipvs->sysctl_ignore_tunneled;
3965
3966         ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3967         if (ipvs->sysctl_hdr == NULL) {
3968                 if (!net_eq(net, &init_net))
3969                         kfree(tbl);
3970                 return -ENOMEM;
3971         }
3972         ip_vs_start_estimator(ipvs, &ipvs->tot_stats);
3973         ipvs->sysctl_tbl = tbl;
3974         /* Schedule defense work */
3975         INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3976         schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3977
3978         return 0;
3979 }
3980
3981 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs)
3982 {
3983         struct net *net = ipvs->net;
3984
3985         cancel_delayed_work_sync(&ipvs->defense_work);
3986         cancel_work_sync(&ipvs->defense_work.work);
3987         unregister_net_sysctl_table(ipvs->sysctl_hdr);
3988         ip_vs_stop_estimator(ipvs, &ipvs->tot_stats);
3989
3990         if (!net_eq(net, &init_net))
3991                 kfree(ipvs->sysctl_tbl);
3992 }
3993
3994 #else
3995
3996 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs) { return 0; }
3997 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs) { }
3998
3999 #endif
4000
4001 static struct notifier_block ip_vs_dst_notifier = {
4002         .notifier_call = ip_vs_dst_event,
4003 };
4004
4005 int __net_init ip_vs_control_net_init(struct netns_ipvs *ipvs)
4006 {
4007         int i, idx;
4008
4009         /* Initialize rs_table */
4010         for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
4011                 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
4012
4013         INIT_LIST_HEAD(&ipvs->dest_trash);
4014         spin_lock_init(&ipvs->dest_trash_lock);
4015         timer_setup(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire, 0);
4016         atomic_set(&ipvs->ftpsvc_counter, 0);
4017         atomic_set(&ipvs->nullsvc_counter, 0);
4018         atomic_set(&ipvs->conn_out_counter, 0);
4019
4020         /* procfs stats */
4021         ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
4022         if (!ipvs->tot_stats.cpustats)
4023                 return -ENOMEM;
4024
4025         for_each_possible_cpu(i) {
4026                 struct ip_vs_cpu_stats *ipvs_tot_stats;
4027                 ipvs_tot_stats = per_cpu_ptr(ipvs->tot_stats.cpustats, i);
4028                 u64_stats_init(&ipvs_tot_stats->syncp);
4029         }
4030
4031         spin_lock_init(&ipvs->tot_stats.lock);
4032
4033         proc_create("ip_vs", 0, ipvs->net->proc_net, &ip_vs_info_fops);
4034         proc_create("ip_vs_stats", 0, ipvs->net->proc_net, &ip_vs_stats_fops);
4035         proc_create("ip_vs_stats_percpu", 0, ipvs->net->proc_net,
4036                     &ip_vs_stats_percpu_fops);
4037
4038         if (ip_vs_control_net_init_sysctl(ipvs))
4039                 goto err;
4040
4041         return 0;
4042
4043 err:
4044         free_percpu(ipvs->tot_stats.cpustats);
4045         return -ENOMEM;
4046 }
4047
4048 void __net_exit ip_vs_control_net_cleanup(struct netns_ipvs *ipvs)
4049 {
4050         ip_vs_trash_cleanup(ipvs);
4051         ip_vs_control_net_cleanup_sysctl(ipvs);
4052         remove_proc_entry("ip_vs_stats_percpu", ipvs->net->proc_net);
4053         remove_proc_entry("ip_vs_stats", ipvs->net->proc_net);
4054         remove_proc_entry("ip_vs", ipvs->net->proc_net);
4055         free_percpu(ipvs->tot_stats.cpustats);
4056 }
4057
4058 int __init ip_vs_register_nl_ioctl(void)
4059 {
4060         int ret;
4061
4062         ret = nf_register_sockopt(&ip_vs_sockopts);
4063         if (ret) {
4064                 pr_err("cannot register sockopt.\n");
4065                 goto err_sock;
4066         }
4067
4068         ret = ip_vs_genl_register();
4069         if (ret) {
4070                 pr_err("cannot register Generic Netlink interface.\n");
4071                 goto err_genl;
4072         }
4073         return 0;
4074
4075 err_genl:
4076         nf_unregister_sockopt(&ip_vs_sockopts);
4077 err_sock:
4078         return ret;
4079 }
4080
4081 void ip_vs_unregister_nl_ioctl(void)
4082 {
4083         ip_vs_genl_unregister();
4084         nf_unregister_sockopt(&ip_vs_sockopts);
4085 }
4086
4087 int __init ip_vs_control_init(void)
4088 {
4089         int idx;
4090         int ret;
4091
4092         EnterFunction(2);
4093
4094         /* Initialize svc_table, ip_vs_svc_fwm_table */
4095         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
4096                 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
4097                 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
4098         }
4099
4100         smp_wmb();      /* Do we really need it now ? */
4101
4102         ret = register_netdevice_notifier(&ip_vs_dst_notifier);
4103         if (ret < 0)
4104                 return ret;
4105
4106         LeaveFunction(2);
4107         return 0;
4108 }
4109
4110
4111 void ip_vs_control_cleanup(void)
4112 {
4113         EnterFunction(2);
4114         unregister_netdevice_notifier(&ip_vs_dst_notifier);
4115         LeaveFunction(2);
4116 }