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 #include <net/netfilter/ipv6/nf_defrag_ipv6.h>
  47 #endif
  48 #include <net/route.h>
  49 #include <net/sock.h>
  50 #include <net/genetlink.h>
  51
  52 #include <linux/uaccess.h>
  53
  54 #include <net/ip_vs.h>
  55
  56 /* semaphore for IPVS sockopts. And, [gs]etsockopt may sleep. */
  57 static DEFINE_MUTEX(__ip_vs_mutex);
  58
  59 /* sysctl variables */
  60
  61 #ifdef CONFIG_IP_VS_DEBUG
  62 static int sysctl_ip_vs_debug_level = 0;
  63
  64 int ip_vs_get_debug_level(void)
  65 {
  66         return sysctl_ip_vs_debug_level;
  67 }
  68 #endif
  69
  70
  71 /*  Protos */
  72 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup);
  73
  74
  75 #ifdef CONFIG_IP_VS_IPV6
  76 /* Taken from rt6_fill_node() in net/ipv6/route.c, is there a better way? */
  77 static bool __ip_vs_addr_is_local_v6(struct net *net,
  78                                      const struct in6_addr *addr)
  79 {
  80         struct flowi6 fl6 = {
  81                 .daddr = *addr,
  82         };
  83         struct dst_entry *dst = ip6_route_output(net, NULL, &fl6);
  84         bool is_local;
  85
  86         is_local = !dst->error && dst->dev && (dst->dev->flags & IFF_LOOPBACK);
  87
  88         dst_release(dst);
  89         return is_local;
  90 }
  91 #endif
  92
  93 #ifdef CONFIG_SYSCTL
  94 /*
  95  *      update_defense_level is called from keventd and from sysctl,
  96  *      so it needs to protect itself from softirqs
  97  */
  98 static void update_defense_level(struct netns_ipvs *ipvs)
  99 {
 100         struct sysinfo i;
 101         static int old_secure_tcp = 0;
 102         int availmem;
 103         int nomem;
 104         int to_change = -1;
 105
 106         /* we only count free and buffered memory (in pages) */
 107         si_meminfo(&i);
 108         availmem = i.freeram + i.bufferram;
 109         /* however in linux 2.5 the i.bufferram is total page cache size,
 110            we need adjust it */
 111         /* si_swapinfo(&i); */
 112         /* availmem = availmem - (i.totalswap - i.freeswap); */
 113
 114         nomem = (availmem < ipvs->sysctl_amemthresh);
 115
 116         local_bh_disable();
 117
 118         /* drop_entry */
 119         spin_lock(&ipvs->dropentry_lock);
 120         switch (ipvs->sysctl_drop_entry) {
 121         case 0:
 122                 atomic_set(&ipvs->dropentry, 0);
 123                 break;
 124         case 1:
 125                 if (nomem) {
 126                         atomic_set(&ipvs->dropentry, 1);
 127                         ipvs->sysctl_drop_entry = 2;
 128                 } else {
 129                         atomic_set(&ipvs->dropentry, 0);
 130                 }
 131                 break;
 132         case 2:
 133                 if (nomem) {
 134                         atomic_set(&ipvs->dropentry, 1);
 135                 } else {
 136                         atomic_set(&ipvs->dropentry, 0);
 137                         ipvs->sysctl_drop_entry = 1;
 138                 }
 139                 break;
 140         case 3:
 141                 atomic_set(&ipvs->dropentry, 1);
 142                 break;
 143         }
 144         spin_unlock(&ipvs->dropentry_lock);
 145
 146         /* drop_packet */
 147         spin_lock(&ipvs->droppacket_lock);
 148         switch (ipvs->sysctl_drop_packet) {
 149         case 0:
 150                 ipvs->drop_rate = 0;
 151                 break;
 152         case 1:
 153                 if (nomem) {
 154                         ipvs->drop_rate = ipvs->drop_counter
 155                                 = ipvs->sysctl_amemthresh /
 156                                 (ipvs->sysctl_amemthresh-availmem);
 157                         ipvs->sysctl_drop_packet = 2;
 158                 } else {
 159                         ipvs->drop_rate = 0;
 160                 }
 161                 break;
 162         case 2:
 163                 if (nomem) {
 164                         ipvs->drop_rate = ipvs->drop_counter
 165                                 = ipvs->sysctl_amemthresh /
 166                                 (ipvs->sysctl_amemthresh-availmem);
 167                 } else {
 168                         ipvs->drop_rate = 0;
 169                         ipvs->sysctl_drop_packet = 1;
 170                 }
 171                 break;
 172         case 3:
 173                 ipvs->drop_rate = ipvs->sysctl_am_droprate;
 174                 break;
 175         }
 176         spin_unlock(&ipvs->droppacket_lock);
 177
 178         /* secure_tcp */
 179         spin_lock(&ipvs->securetcp_lock);
 180         switch (ipvs->sysctl_secure_tcp) {
 181         case 0:
 182                 if (old_secure_tcp >= 2)
 183                         to_change = 0;
 184                 break;
 185         case 1:
 186                 if (nomem) {
 187                         if (old_secure_tcp < 2)
 188                                 to_change = 1;
 189                         ipvs->sysctl_secure_tcp = 2;
 190                 } else {
 191                         if (old_secure_tcp >= 2)
 192                                 to_change = 0;
 193                 }
 194                 break;
 195         case 2:
 196                 if (nomem) {
 197                         if (old_secure_tcp < 2)
 198                                 to_change = 1;
 199                 } else {
 200                         if (old_secure_tcp >= 2)
 201                                 to_change = 0;
 202                         ipvs->sysctl_secure_tcp = 1;
 203                 }
 204                 break;
 205         case 3:
 206                 if (old_secure_tcp < 2)
 207                         to_change = 1;
 208                 break;
 209         }
 210         old_secure_tcp = ipvs->sysctl_secure_tcp;
 211         if (to_change >= 0)
 212                 ip_vs_protocol_timeout_change(ipvs,
 213                                               ipvs->sysctl_secure_tcp > 1);
 214         spin_unlock(&ipvs->securetcp_lock);
 215
 216         local_bh_enable();
 217 }
 218
 219
 220 /*
 221  *      Timer for checking the defense
 222  */
 223 #define DEFENSE_TIMER_PERIOD    1*HZ
 224
 225 static void defense_work_handler(struct work_struct *work)
 226 {
 227         struct netns_ipvs *ipvs =
 228                 container_of(work, struct netns_ipvs, defense_work.work);
 229
 230         update_defense_level(ipvs);
 231         if (atomic_read(&ipvs->dropentry))
 232                 ip_vs_random_dropentry(ipvs);
 233         schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
 234 }
 235 #endif
 236
 237 int
 238 ip_vs_use_count_inc(void)
 239 {
 240         return try_module_get(THIS_MODULE);
 241 }
 242
 243 void
 244 ip_vs_use_count_dec(void)
 245 {
 246         module_put(THIS_MODULE);
 247 }
 248
 249
 250 /*
 251  *      Hash table: for virtual service lookups
 252  */
 253 #define IP_VS_SVC_TAB_BITS 8
 254 #define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
 255 #define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)
 256
 257 /* the service table hashed by <protocol, addr, port> */
 258 static struct hlist_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE];
 259 /* the service table hashed by fwmark */
 260 static struct hlist_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE];
 261
 262
 263 /*
 264  *      Returns hash value for virtual service
 265  */
 266 static inline unsigned int
 267 ip_vs_svc_hashkey(struct netns_ipvs *ipvs, int af, unsigned int proto,
 268                   const union nf_inet_addr *addr, __be16 port)
 269 {
 270         register unsigned int porth = ntohs(port);
 271         __be32 addr_fold = addr->ip;
 272         __u32 ahash;
 273
 274 #ifdef CONFIG_IP_VS_IPV6
 275         if (af == AF_INET6)
 276                 addr_fold = addr->ip6[0]^addr->ip6[1]^
 277                             addr->ip6[2]^addr->ip6[3];
 278 #endif
 279         ahash = ntohl(addr_fold);
 280         ahash ^= ((size_t) ipvs >> 8);
 281
 282         return (proto ^ ahash ^ (porth >> IP_VS_SVC_TAB_BITS) ^ porth) &
 283                IP_VS_SVC_TAB_MASK;
 284 }
 285
 286 /*
 287  *      Returns hash value of fwmark for virtual service lookup
 288  */
 289 static inline unsigned int ip_vs_svc_fwm_hashkey(struct netns_ipvs *ipvs, __u32 fwmark)
 290 {
 291         return (((size_t)ipvs>>8) ^ fwmark) & IP_VS_SVC_TAB_MASK;
 292 }
 293
 294 /*
 295  *      Hashes a service in the ip_vs_svc_table by <netns,proto,addr,port>
 296  *      or in the ip_vs_svc_fwm_table by fwmark.
 297  *      Should be called with locked tables.
 298  */
 299 static int ip_vs_svc_hash(struct ip_vs_service *svc)
 300 {
 301         unsigned int hash;
 302
 303         if (svc->flags & IP_VS_SVC_F_HASHED) {
 304                 pr_err("%s(): request for already hashed, called from %pS\n",
 305                        __func__, __builtin_return_address(0));
 306                 return 0;
 307         }
 308
 309         if (svc->fwmark == 0) {
 310                 /*
 311                  *  Hash it by <netns,protocol,addr,port> in ip_vs_svc_table
 312                  */
 313                 hash = ip_vs_svc_hashkey(svc->ipvs, svc->af, svc->protocol,
 314                                          &svc->addr, svc->port);
 315                 hlist_add_head_rcu(&svc->s_list, &ip_vs_svc_table[hash]);
 316         } else {
 317                 /*
 318                  *  Hash it by fwmark in svc_fwm_table
 319                  */
 320                 hash = ip_vs_svc_fwm_hashkey(svc->ipvs, svc->fwmark);
 321                 hlist_add_head_rcu(&svc->f_list, &ip_vs_svc_fwm_table[hash]);
 322         }
 323
 324         svc->flags |= IP_VS_SVC_F_HASHED;
 325         /* increase its refcnt because it is referenced by the svc table */
 326         atomic_inc(&svc->refcnt);
 327         return 1;
 328 }
 329
 330
 331 /*
 332  *      Unhashes a service from svc_table / svc_fwm_table.
 333  *      Should be called with locked tables.
 334  */
 335 static int ip_vs_svc_unhash(struct ip_vs_service *svc)
 336 {
 337         if (!(svc->flags & IP_VS_SVC_F_HASHED)) {
 338                 pr_err("%s(): request for unhash flagged, called from %pS\n",
 339                        __func__, __builtin_return_address(0));
 340                 return 0;
 341         }
 342
 343         if (svc->fwmark == 0) {
 344                 /* Remove it from the svc_table table */
 345                 hlist_del_rcu(&svc->s_list);
 346         } else {
 347                 /* Remove it from the svc_fwm_table table */
 348                 hlist_del_rcu(&svc->f_list);
 349         }
 350
 351         svc->flags &= ~IP_VS_SVC_F_HASHED;
 352         atomic_dec(&svc->refcnt);
 353         return 1;
 354 }
 355
 356
 357 /*
 358  *      Get service by {netns, proto,addr,port} in the service table.
 359  */
 360 static inline struct ip_vs_service *
 361 __ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u16 protocol,
 362                      const union nf_inet_addr *vaddr, __be16 vport)
 363 {
 364         unsigned int hash;
 365         struct ip_vs_service *svc;
 366
 367         /* Check for "full" addressed entries */
 368         hash = ip_vs_svc_hashkey(ipvs, af, protocol, vaddr, vport);
 369
 370         hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[hash], s_list) {
 371                 if ((svc->af == af)
 372                     && ip_vs_addr_equal(af, &svc->addr, vaddr)
 373                     && (svc->port == vport)
 374                     && (svc->protocol == protocol)
 375                     && (svc->ipvs == ipvs)) {
 376                         /* HIT */
 377                         return svc;
 378                 }
 379         }
 380
 381         return NULL;
 382 }
 383
 384
 385 /*
 386  *      Get service by {fwmark} in the service table.
 387  */
 388 static inline struct ip_vs_service *
 389 __ip_vs_svc_fwm_find(struct netns_ipvs *ipvs, int af, __u32 fwmark)
 390 {
 391         unsigned int hash;
 392         struct ip_vs_service *svc;
 393
 394         /* Check for fwmark addressed entries */
 395         hash = ip_vs_svc_fwm_hashkey(ipvs, fwmark);
 396
 397         hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[hash], f_list) {
 398                 if (svc->fwmark == fwmark && svc->af == af
 399                     && (svc->ipvs == ipvs)) {
 400                         /* HIT */
 401                         return svc;
 402                 }
 403         }
 404
 405         return NULL;
 406 }
 407
 408 /* Find service, called under RCU lock */
 409 struct ip_vs_service *
 410 ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
 411                    const union nf_inet_addr *vaddr, __be16 vport)
 412 {
 413         struct ip_vs_service *svc;
 414
 415         /*
 416          *      Check the table hashed by fwmark first
 417          */
 418         if (fwmark) {
 419                 svc = __ip_vs_svc_fwm_find(ipvs, af, fwmark);
 420                 if (svc)
 421                         goto out;
 422         }
 423
 424         /*
 425          *      Check the table hashed by <protocol,addr,port>
 426          *      for "full" addressed entries
 427          */
 428         svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, vport);
 429
 430         if (!svc && protocol == IPPROTO_TCP &&
 431             atomic_read(&ipvs->ftpsvc_counter) &&
 432             (vport == FTPDATA || ntohs(vport) >= inet_prot_sock(ipvs->net))) {
 433                 /*
 434                  * Check if ftp service entry exists, the packet
 435                  * might belong to FTP data connections.
 436                  */
 437                 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, FTPPORT);
 438         }
 439
 440         if (svc == NULL
 441             && atomic_read(&ipvs->nullsvc_counter)) {
 442                 /*
 443                  * Check if the catch-all port (port zero) exists
 444                  */
 445                 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, 0);
 446         }
 447
 448   out:
 449         IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
 450                       fwmark, ip_vs_proto_name(protocol),
 451                       IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
 452                       svc ? "hit" : "not hit");
 453
 454         return svc;
 455 }
 456
 457
 458 static inline void
 459 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
 460 {
 461         atomic_inc(&svc->refcnt);
 462         rcu_assign_pointer(dest->svc, svc);
 463 }
 464
 465 static void ip_vs_service_free(struct ip_vs_service *svc)
 466 {
 467         free_percpu(svc->stats.cpustats);
 468         kfree(svc);
 469 }
 470
 471 static void ip_vs_service_rcu_free(struct rcu_head *head)
 472 {
 473         struct ip_vs_service *svc;
 474
 475         svc = container_of(head, struct ip_vs_service, rcu_head);
 476         ip_vs_service_free(svc);
 477 }
 478
 479 static void __ip_vs_svc_put(struct ip_vs_service *svc, bool do_delay)
 480 {
 481         if (atomic_dec_and_test(&svc->refcnt)) {
 482                 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n",
 483                               svc->fwmark,
 484                               IP_VS_DBG_ADDR(svc->af, &svc->addr),
 485                               ntohs(svc->port));
 486                 if (do_delay)
 487                         call_rcu(&svc->rcu_head, ip_vs_service_rcu_free);
 488                 else
 489                         ip_vs_service_free(svc);
 490         }
 491 }
 492
 493
 494 /*
 495  *      Returns hash value for real service
 496  */
 497 static inline unsigned int ip_vs_rs_hashkey(int af,
 498                                             const union nf_inet_addr *addr,
 499                                             __be16 port)
 500 {
 501         register unsigned int porth = ntohs(port);
 502         __be32 addr_fold = addr->ip;
 503
 504 #ifdef CONFIG_IP_VS_IPV6
 505         if (af == AF_INET6)
 506                 addr_fold = addr->ip6[0]^addr->ip6[1]^
 507                             addr->ip6[2]^addr->ip6[3];
 508 #endif
 509
 510         return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
 511                 & IP_VS_RTAB_MASK;
 512 }
 513
 514 /* Hash ip_vs_dest in rs_table by <proto,addr,port>. */
 515 static void ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
 516 {
 517         unsigned int hash;
 518
 519         if (dest->in_rs_table)
 520                 return;
 521
 522         /*
 523          *      Hash by proto,addr,port,
 524          *      which are the parameters of the real service.
 525          */
 526         hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
 527
 528         hlist_add_head_rcu(&dest->d_list, &ipvs->rs_table[hash]);
 529         dest->in_rs_table = 1;
 530 }
 531
 532 /* Unhash ip_vs_dest from rs_table. */
 533 static void ip_vs_rs_unhash(struct ip_vs_dest *dest)
 534 {
 535         /*
 536          * Remove it from the rs_table table.
 537          */
 538         if (dest->in_rs_table) {
 539                 hlist_del_rcu(&dest->d_list);
 540                 dest->in_rs_table = 0;
 541         }
 542 }
 543
 544 /* Check if real service by <proto,addr,port> is present */
 545 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
 546                             const union nf_inet_addr *daddr, __be16 dport)
 547 {
 548         unsigned int hash;
 549         struct ip_vs_dest *dest;
 550
 551         /* Check for "full" addressed entries */
 552         hash = ip_vs_rs_hashkey(af, daddr, dport);
 553
 554         hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
 555                 if (dest->port == dport &&
 556                     dest->af == af &&
 557                     ip_vs_addr_equal(af, &dest->addr, daddr) &&
 558                     (dest->protocol == protocol || dest->vfwmark)) {
 559                         /* HIT */
 560                         return true;
 561                 }
 562         }
 563
 564         return false;
 565 }
 566
 567 /* Find real service record by <proto,addr,port>.
 568  * In case of multiple records with the same <proto,addr,port>, only
 569  * the first found record is returned.
 570  *
 571  * To be called under RCU lock.
 572  */
 573 struct ip_vs_dest *ip_vs_find_real_service(struct netns_ipvs *ipvs, int af,
 574                                            __u16 protocol,
 575                                            const union nf_inet_addr *daddr,
 576                                            __be16 dport)
 577 {
 578         unsigned int hash;
 579         struct ip_vs_dest *dest;
 580
 581         /* Check for "full" addressed entries */
 582         hash = ip_vs_rs_hashkey(af, daddr, dport);
 583
 584         hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
 585                 if (dest->port == dport &&
 586                     dest->af == af &&
 587                     ip_vs_addr_equal(af, &dest->addr, daddr) &&
 588                         (dest->protocol == protocol || dest->vfwmark)) {
 589                         /* HIT */
 590                         return dest;
 591                 }
 592         }
 593
 594         return NULL;
 595 }
 596
 597 /* Lookup destination by {addr,port} in the given service
 598  * Called under RCU lock.
 599  */
 600 static struct ip_vs_dest *
 601 ip_vs_lookup_dest(struct ip_vs_service *svc, int dest_af,
 602                   const union nf_inet_addr *daddr, __be16 dport)
 603 {
 604         struct ip_vs_dest *dest;
 605
 606         /*
 607          * Find the destination for the given service
 608          */
 609         list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
 610                 if ((dest->af == dest_af) &&
 611                     ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
 612                     (dest->port == dport)) {
 613                         /* HIT */
 614                         return dest;
 615                 }
 616         }
 617
 618         return NULL;
 619 }
 620
 621 /*
 622  * Find destination by {daddr,dport,vaddr,protocol}
 623  * Created to be used in ip_vs_process_message() in
 624  * the backup synchronization daemon. It finds the
 625  * destination to be bound to the received connection
 626  * on the backup.
 627  * Called under RCU lock, no refcnt is returned.
 628  */
 629 struct ip_vs_dest *ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
 630                                    const union nf_inet_addr *daddr,
 631                                    __be16 dport,
 632                                    const union nf_inet_addr *vaddr,
 633                                    __be16 vport, __u16 protocol, __u32 fwmark,
 634                                    __u32 flags)
 635 {
 636         struct ip_vs_dest *dest;
 637         struct ip_vs_service *svc;
 638         __be16 port = dport;
 639
 640         svc = ip_vs_service_find(ipvs, svc_af, fwmark, protocol, vaddr, vport);
 641         if (!svc)
 642                 return NULL;
 643         if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
 644                 port = 0;
 645         dest = ip_vs_lookup_dest(svc, dest_af, daddr, port);
 646         if (!dest)
 647                 dest = ip_vs_lookup_dest(svc, dest_af, daddr, port ^ dport);
 648         return dest;
 649 }
 650
 651 void ip_vs_dest_dst_rcu_free(struct rcu_head *head)
 652 {
 653         struct ip_vs_dest_dst *dest_dst = container_of(head,
 654                                                        struct ip_vs_dest_dst,
 655                                                        rcu_head);
 656
 657         dst_release(dest_dst->dst_cache);
 658         kfree(dest_dst);
 659 }
 660
 661 /* Release dest_dst and dst_cache for dest in user context */
 662 static void __ip_vs_dst_cache_reset(struct ip_vs_dest *dest)
 663 {
 664         struct ip_vs_dest_dst *old;
 665
 666         old = rcu_dereference_protected(dest->dest_dst, 1);
 667         if (old) {
 668                 RCU_INIT_POINTER(dest->dest_dst, NULL);
 669                 call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
 670         }
 671 }
 672
 673 /*
 674  *  Lookup dest by {svc,addr,port} in the destination trash.
 675  *  The destination trash is used to hold the destinations that are removed
 676  *  from the service table but are still referenced by some conn entries.
 677  *  The reason to add the destination trash is when the dest is temporary
 678  *  down (either by administrator or by monitor program), the dest can be
 679  *  picked back from the trash, the remaining connections to the dest can
 680  *  continue, and the counting information of the dest is also useful for
 681  *  scheduling.
 682  */
 683 static struct ip_vs_dest *
 684 ip_vs_trash_get_dest(struct ip_vs_service *svc, int dest_af,
 685                      const union nf_inet_addr *daddr, __be16 dport)
 686 {
 687         struct ip_vs_dest *dest;
 688         struct netns_ipvs *ipvs = svc->ipvs;
 689
 690         /*
 691          * Find the destination in trash
 692          */
 693         spin_lock_bh(&ipvs->dest_trash_lock);
 694         list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
 695                 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
 696                               "dest->refcnt=%d\n",
 697                               dest->vfwmark,
 698                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
 699                               ntohs(dest->port),
 700                               refcount_read(&dest->refcnt));
 701                 if (dest->af == dest_af &&
 702                     ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
 703                     dest->port == dport &&
 704                     dest->vfwmark == svc->fwmark &&
 705                     dest->protocol == svc->protocol &&
 706                     (svc->fwmark ||
 707                      (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
 708                       dest->vport == svc->port))) {
 709                         /* HIT */
 710                         list_del(&dest->t_list);
 711                         goto out;
 712                 }
 713         }
 714
 715         dest = NULL;
 716
 717 out:
 718         spin_unlock_bh(&ipvs->dest_trash_lock);
 719
 720         return dest;
 721 }
 722
 723 static void ip_vs_dest_free(struct ip_vs_dest *dest)
 724 {
 725         struct ip_vs_service *svc = rcu_dereference_protected(dest->svc, 1);
 726
 727         __ip_vs_dst_cache_reset(dest);
 728         __ip_vs_svc_put(svc, false);
 729         free_percpu(dest->stats.cpustats);
 730         ip_vs_dest_put_and_free(dest);
 731 }
 732
 733 /*
 734  *  Clean up all the destinations in the trash
 735  *  Called by the ip_vs_control_cleanup()
 736  *
 737  *  When the ip_vs_control_clearup is activated by ipvs module exit,
 738  *  the service tables must have been flushed and all the connections
 739  *  are expired, and the refcnt of each destination in the trash must
 740  *  be 1, so we simply release them here.
 741  */
 742 static void ip_vs_trash_cleanup(struct netns_ipvs *ipvs)
 743 {
 744         struct ip_vs_dest *dest, *nxt;
 745
 746         del_timer_sync(&ipvs->dest_trash_timer);
 747         /* No need to use dest_trash_lock */
 748         list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, t_list) {
 749                 list_del(&dest->t_list);
 750                 ip_vs_dest_free(dest);
 751         }
 752 }
 753
 754 static void
 755 ip_vs_copy_stats(struct ip_vs_kstats *dst, struct ip_vs_stats *src)
 756 {
 757 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->kstats.c - src->kstats0.c
 758
 759         spin_lock_bh(&src->lock);
 760
 761         IP_VS_SHOW_STATS_COUNTER(conns);
 762         IP_VS_SHOW_STATS_COUNTER(inpkts);
 763         IP_VS_SHOW_STATS_COUNTER(outpkts);
 764         IP_VS_SHOW_STATS_COUNTER(inbytes);
 765         IP_VS_SHOW_STATS_COUNTER(outbytes);
 766
 767         ip_vs_read_estimator(dst, src);
 768
 769         spin_unlock_bh(&src->lock);
 770 }
 771
 772 static void
 773 ip_vs_export_stats_user(struct ip_vs_stats_user *dst, struct ip_vs_kstats *src)
 774 {
 775         dst->conns = (u32)src->conns;
 776         dst->inpkts = (u32)src->inpkts;
 777         dst->outpkts = (u32)src->outpkts;
 778         dst->inbytes = src->inbytes;
 779         dst->outbytes = src->outbytes;
 780         dst->cps = (u32)src->cps;
 781         dst->inpps = (u32)src->inpps;
 782         dst->outpps = (u32)src->outpps;
 783         dst->inbps = (u32)src->inbps;
 784         dst->outbps = (u32)src->outbps;
 785 }
 786
 787 static void
 788 ip_vs_zero_stats(struct ip_vs_stats *stats)
 789 {
 790         spin_lock_bh(&stats->lock);
 791
 792         /* get current counters as zero point, rates are zeroed */
 793
 794 #define IP_VS_ZERO_STATS_COUNTER(c) stats->kstats0.c = stats->kstats.c
 795
 796         IP_VS_ZERO_STATS_COUNTER(conns);
 797         IP_VS_ZERO_STATS_COUNTER(inpkts);
 798         IP_VS_ZERO_STATS_COUNTER(outpkts);
 799         IP_VS_ZERO_STATS_COUNTER(inbytes);
 800         IP_VS_ZERO_STATS_COUNTER(outbytes);
 801
 802         ip_vs_zero_estimator(stats);
 803
 804         spin_unlock_bh(&stats->lock);
 805 }
 806
 807 /*
 808  *      Update a destination in the given service
 809  */
 810 static void
 811 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
 812                     struct ip_vs_dest_user_kern *udest, int add)
 813 {
 814         struct netns_ipvs *ipvs = svc->ipvs;
 815         struct ip_vs_service *old_svc;
 816         struct ip_vs_scheduler *sched;
 817         int conn_flags;
 818
 819         /* We cannot modify an address and change the address family */
 820         BUG_ON(!add && udest->af != dest->af);
 821
 822         if (add && udest->af != svc->af)
 823                 ipvs->mixed_address_family_dests++;
 824
 825         /* keep the last_weight with latest non-0 weight */
 826         if (add || udest->weight != 0)
 827                 atomic_set(&dest->last_weight, udest->weight);
 828
 829         /* set the weight and the flags */
 830         atomic_set(&dest->weight, udest->weight);
 831         conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
 832         conn_flags |= IP_VS_CONN_F_INACTIVE;
 833
 834         /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
 835         if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
 836                 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
 837         } else {
 838                 /*
 839                  *    Put the real service in rs_table if not present.
 840                  *    For now only for NAT!
 841                  */
 842                 ip_vs_rs_hash(ipvs, dest);
 843                 /* FTP-NAT requires conntrack for mangling */
 844                 if (svc->port == FTPPORT)
 845                         ip_vs_register_conntrack(svc);
 846         }
 847         atomic_set(&dest->conn_flags, conn_flags);
 848
 849         /* bind the service */
 850         old_svc = rcu_dereference_protected(dest->svc, 1);
 851         if (!old_svc) {
 852                 __ip_vs_bind_svc(dest, svc);
 853         } else {
 854                 if (old_svc != svc) {
 855                         ip_vs_zero_stats(&dest->stats);
 856                         __ip_vs_bind_svc(dest, svc);
 857                         __ip_vs_svc_put(old_svc, true);
 858                 }
 859         }
 860
 861         /* set the dest status flags */
 862         dest->flags |= IP_VS_DEST_F_AVAILABLE;
 863
 864         if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
 865                 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
 866         dest->u_threshold = udest->u_threshold;
 867         dest->l_threshold = udest->l_threshold;
 868
 869         dest->af = udest->af;
 870
 871         spin_lock_bh(&dest->dst_lock);
 872         __ip_vs_dst_cache_reset(dest);
 873         spin_unlock_bh(&dest->dst_lock);
 874
 875         if (add) {
 876                 ip_vs_start_estimator(svc->ipvs, &dest->stats);
 877                 list_add_rcu(&dest->n_list, &svc->destinations);
 878                 svc->num_dests++;
 879                 sched = rcu_dereference_protected(svc->scheduler, 1);
 880                 if (sched && sched->add_dest)
 881                         sched->add_dest(svc, dest);
 882         } else {
 883                 sched = rcu_dereference_protected(svc->scheduler, 1);
 884                 if (sched && sched->upd_dest)
 885                         sched->upd_dest(svc, dest);
 886         }
 887 }
 888
 889
 890 /*
 891  *      Create a destination for the given service
 892  */
 893 static int
 894 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
 895                struct ip_vs_dest **dest_p)
 896 {
 897         struct ip_vs_dest *dest;
 898         unsigned int atype, i;
 899
 900         EnterFunction(2);
 901
 902 #ifdef CONFIG_IP_VS_IPV6
 903         if (udest->af == AF_INET6) {
 904                 int ret;
 905
 906                 atype = ipv6_addr_type(&udest->addr.in6);
 907                 if ((!(atype & IPV6_ADDR_UNICAST) ||
 908                         atype & IPV6_ADDR_LINKLOCAL) &&
 909                         !__ip_vs_addr_is_local_v6(svc->ipvs->net, &udest->addr.in6))
 910                         return -EINVAL;
 911
 912                 ret = nf_defrag_ipv6_enable(svc->ipvs->net);
 913                 if (ret)
 914                         return ret;
 915         } else
 916 #endif
 917         {
 918                 atype = inet_addr_type(svc->ipvs->net, udest->addr.ip);
 919                 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
 920                         return -EINVAL;
 921         }
 922
 923         dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
 924         if (dest == NULL)
 925                 return -ENOMEM;
 926
 927         dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
 928         if (!dest->stats.cpustats)
 929                 goto err_alloc;
 930
 931         for_each_possible_cpu(i) {
 932                 struct ip_vs_cpu_stats *ip_vs_dest_stats;
 933                 ip_vs_dest_stats = per_cpu_ptr(dest->stats.cpustats, i);
 934                 u64_stats_init(&ip_vs_dest_stats->syncp);
 935         }
 936
 937         dest->af = udest->af;
 938         dest->protocol = svc->protocol;
 939         dest->vaddr = svc->addr;
 940         dest->vport = svc->port;
 941         dest->vfwmark = svc->fwmark;
 942         ip_vs_addr_copy(udest->af, &dest->addr, &udest->addr);
 943         dest->port = udest->port;
 944
 945         atomic_set(&dest->activeconns, 0);
 946         atomic_set(&dest->inactconns, 0);
 947         atomic_set(&dest->persistconns, 0);
 948         refcount_set(&dest->refcnt, 1);
 949
 950         INIT_HLIST_NODE(&dest->d_list);
 951         spin_lock_init(&dest->dst_lock);
 952         spin_lock_init(&dest->stats.lock);
 953         __ip_vs_update_dest(svc, dest, udest, 1);
 954
 955         *dest_p = dest;
 956
 957         LeaveFunction(2);
 958         return 0;
 959
 960 err_alloc:
 961         kfree(dest);
 962         return -ENOMEM;
 963 }
 964
 965
 966 /*
 967  *      Add a destination into an existing service
 968  */
 969 static int
 970 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
 971 {
 972         struct ip_vs_dest *dest;
 973         union nf_inet_addr daddr;
 974         __be16 dport = udest->port;
 975         int ret;
 976
 977         EnterFunction(2);
 978
 979         if (udest->weight < 0) {
 980                 pr_err("%s(): server weight less than zero\n", __func__);
 981                 return -ERANGE;
 982         }
 983
 984         if (udest->l_threshold > udest->u_threshold) {
 985                 pr_err("%s(): lower threshold is higher than upper threshold\n",
 986                         __func__);
 987                 return -ERANGE;
 988         }
 989
 990         ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
 991
 992         /* We use function that requires RCU lock */
 993         rcu_read_lock();
 994         dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
 995         rcu_read_unlock();
 996
 997         if (dest != NULL) {
 998                 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
 999                 return -EEXIST;
1000         }
1001
1002         /*
1003          * Check if the dest already exists in the trash and
1004          * is from the same service
1005          */
1006         dest = ip_vs_trash_get_dest(svc, udest->af, &daddr, dport);
1007
1008         if (dest != NULL) {
1009                 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
1010                               "dest->refcnt=%d, service %u/%s:%u\n",
1011                               IP_VS_DBG_ADDR(udest->af, &daddr), ntohs(dport),
1012                               refcount_read(&dest->refcnt),
1013                               dest->vfwmark,
1014                               IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
1015                               ntohs(dest->vport));
1016
1017                 __ip_vs_update_dest(svc, dest, udest, 1);
1018                 ret = 0;
1019         } else {
1020                 /*
1021                  * Allocate and initialize the dest structure
1022                  */
1023                 ret = ip_vs_new_dest(svc, udest, &dest);
1024         }
1025         LeaveFunction(2);
1026
1027         return ret;
1028 }
1029
1030
1031 /*
1032  *      Edit a destination in the given service
1033  */
1034 static int
1035 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1036 {
1037         struct ip_vs_dest *dest;
1038         union nf_inet_addr daddr;
1039         __be16 dport = udest->port;
1040
1041         EnterFunction(2);
1042
1043         if (udest->weight < 0) {
1044                 pr_err("%s(): server weight less than zero\n", __func__);
1045                 return -ERANGE;
1046         }
1047
1048         if (udest->l_threshold > udest->u_threshold) {
1049                 pr_err("%s(): lower threshold is higher than upper threshold\n",
1050                         __func__);
1051                 return -ERANGE;
1052         }
1053
1054         ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
1055
1056         /* We use function that requires RCU lock */
1057         rcu_read_lock();
1058         dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
1059         rcu_read_unlock();
1060
1061         if (dest == NULL) {
1062                 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1063                 return -ENOENT;
1064         }
1065
1066         __ip_vs_update_dest(svc, dest, udest, 0);
1067         LeaveFunction(2);
1068
1069         return 0;
1070 }
1071
1072 /*
1073  *      Delete a destination (must be already unlinked from the service)
1074  */
1075 static void __ip_vs_del_dest(struct netns_ipvs *ipvs, struct ip_vs_dest *dest,
1076                              bool cleanup)
1077 {
1078         ip_vs_stop_estimator(ipvs, &dest->stats);
1079
1080         /*
1081          *  Remove it from the d-linked list with the real services.
1082          */
1083         ip_vs_rs_unhash(dest);
1084
1085         spin_lock_bh(&ipvs->dest_trash_lock);
1086         IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1087                       IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port),
1088                       refcount_read(&dest->refcnt));
1089         if (list_empty(&ipvs->dest_trash) && !cleanup)
1090                 mod_timer(&ipvs->dest_trash_timer,
1091                           jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1092         /* dest lives in trash with reference */
1093         list_add(&dest->t_list, &ipvs->dest_trash);
1094         dest->idle_start = 0;
1095         spin_unlock_bh(&ipvs->dest_trash_lock);
1096 }
1097
1098
1099 /*
1100  *      Unlink a destination from the given service
1101  */
1102 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1103                                 struct ip_vs_dest *dest,
1104                                 int svcupd)
1105 {
1106         dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1107
1108         /*
1109          *  Remove it from the d-linked destination list.
1110          */
1111         list_del_rcu(&dest->n_list);
1112         svc->num_dests--;
1113
1114         if (dest->af != svc->af)
1115                 svc->ipvs->mixed_address_family_dests--;
1116
1117         if (svcupd) {
1118                 struct ip_vs_scheduler *sched;
1119
1120                 sched = rcu_dereference_protected(svc->scheduler, 1);
1121                 if (sched && sched->del_dest)
1122                         sched->del_dest(svc, dest);
1123         }
1124 }
1125
1126
1127 /*
1128  *      Delete a destination server in the given service
1129  */
1130 static int
1131 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1132 {
1133         struct ip_vs_dest *dest;
1134         __be16 dport = udest->port;
1135
1136         EnterFunction(2);
1137
1138         /* We use function that requires RCU lock */
1139         rcu_read_lock();
1140         dest = ip_vs_lookup_dest(svc, udest->af, &udest->addr, dport);
1141         rcu_read_unlock();
1142
1143         if (dest == NULL) {
1144                 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1145                 return -ENOENT;
1146         }
1147
1148         /*
1149          *      Unlink dest from the service
1150          */
1151         __ip_vs_unlink_dest(svc, dest, 1);
1152
1153         /*
1154          *      Delete the destination
1155          */
1156         __ip_vs_del_dest(svc->ipvs, dest, false);
1157
1158         LeaveFunction(2);
1159
1160         return 0;
1161 }
1162
1163 static void ip_vs_dest_trash_expire(struct timer_list *t)
1164 {
1165         struct netns_ipvs *ipvs = from_timer(ipvs, t, dest_trash_timer);
1166         struct ip_vs_dest *dest, *next;
1167         unsigned long now = jiffies;
1168
1169         spin_lock(&ipvs->dest_trash_lock);
1170         list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) {
1171                 if (refcount_read(&dest->refcnt) > 1)
1172                         continue;
1173                 if (dest->idle_start) {
1174                         if (time_before(now, dest->idle_start +
1175                                              IP_VS_DEST_TRASH_PERIOD))
1176                                 continue;
1177                 } else {
1178                         dest->idle_start = max(1UL, now);
1179                         continue;
1180                 }
1181                 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1182                               dest->vfwmark,
1183                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1184                               ntohs(dest->port));
1185                 list_del(&dest->t_list);
1186                 ip_vs_dest_free(dest);
1187         }
1188         if (!list_empty(&ipvs->dest_trash))
1189                 mod_timer(&ipvs->dest_trash_timer,
1190                           jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1191         spin_unlock(&ipvs->dest_trash_lock);
1192 }
1193
1194 /*
1195  *      Add a service into the service hash table
1196  */
1197 static int
1198 ip_vs_add_service(struct netns_ipvs *ipvs, struct ip_vs_service_user_kern *u,
1199                   struct ip_vs_service **svc_p)
1200 {
1201         int ret = 0, i;
1202         struct ip_vs_scheduler *sched = NULL;
1203         struct ip_vs_pe *pe = NULL;
1204         struct ip_vs_service *svc = NULL;
1205
1206         /* increase the module use count */
1207         ip_vs_use_count_inc();
1208
1209         /* Lookup the scheduler by 'u->sched_name' */
1210         if (strcmp(u->sched_name, "none")) {
1211                 sched = ip_vs_scheduler_get(u->sched_name);
1212                 if (!sched) {
1213                         pr_info("Scheduler module ip_vs_%s not found\n",
1214                                 u->sched_name);
1215                         ret = -ENOENT;
1216                         goto out_err;
1217                 }
1218         }
1219
1220         if (u->pe_name && *u->pe_name) {
1221                 pe = ip_vs_pe_getbyname(u->pe_name);
1222                 if (pe == NULL) {
1223                         pr_info("persistence engine module ip_vs_pe_%s "
1224                                 "not found\n", u->pe_name);
1225                         ret = -ENOENT;
1226                         goto out_err;
1227                 }
1228         }
1229
1230 #ifdef CONFIG_IP_VS_IPV6
1231         if (u->af == AF_INET6) {
1232                 __u32 plen = (__force __u32) u->netmask;
1233
1234                 if (plen < 1 || plen > 128) {
1235                         ret = -EINVAL;
1236                         goto out_err;
1237                 }
1238
1239                 ret = nf_defrag_ipv6_enable(ipvs->net);
1240                 if (ret)
1241                         goto out_err;
1242         }
1243 #endif
1244
1245         svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1246         if (svc == NULL) {
1247                 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1248                 ret = -ENOMEM;
1249                 goto out_err;
1250         }
1251         svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1252         if (!svc->stats.cpustats) {
1253                 ret = -ENOMEM;
1254                 goto out_err;
1255         }
1256
1257         for_each_possible_cpu(i) {
1258                 struct ip_vs_cpu_stats *ip_vs_stats;
1259                 ip_vs_stats = per_cpu_ptr(svc->stats.cpustats, i);
1260                 u64_stats_init(&ip_vs_stats->syncp);
1261         }
1262
1263
1264         /* I'm the first user of the service */
1265         atomic_set(&svc->refcnt, 0);
1266
1267         svc->af = u->af;
1268         svc->protocol = u->protocol;
1269         ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1270         svc->port = u->port;
1271         svc->fwmark = u->fwmark;
1272         svc->flags = u->flags;
1273         svc->timeout = u->timeout * HZ;
1274         svc->netmask = u->netmask;
1275         svc->ipvs = ipvs;
1276
1277         INIT_LIST_HEAD(&svc->destinations);
1278         spin_lock_init(&svc->sched_lock);
1279         spin_lock_init(&svc->stats.lock);
1280
1281         /* Bind the scheduler */
1282         if (sched) {
1283                 ret = ip_vs_bind_scheduler(svc, sched);
1284                 if (ret)
1285                         goto out_err;
1286                 sched = NULL;
1287         }
1288
1289         /* Bind the ct retriever */
1290         RCU_INIT_POINTER(svc->pe, pe);
1291         pe = NULL;
1292
1293         /* Update the virtual service counters */
1294         if (svc->port == FTPPORT)
1295                 atomic_inc(&ipvs->ftpsvc_counter);
1296         else if (svc->port == 0)
1297                 atomic_inc(&ipvs->nullsvc_counter);
1298         if (svc->pe && svc->pe->conn_out)
1299                 atomic_inc(&ipvs->conn_out_counter);
1300
1301         ip_vs_start_estimator(ipvs, &svc->stats);
1302
1303         /* Count only IPv4 services for old get/setsockopt interface */
1304         if (svc->af == AF_INET)
1305                 ipvs->num_services++;
1306
1307         /* Hash the service into the service table */
1308         ip_vs_svc_hash(svc);
1309
1310         *svc_p = svc;
1311         /* Now there is a service - full throttle */
1312         ipvs->enable = 1;
1313         return 0;
1314
1315
1316  out_err:
1317         if (svc != NULL) {
1318                 ip_vs_unbind_scheduler(svc, sched);
1319                 ip_vs_service_free(svc);
1320         }
1321         ip_vs_scheduler_put(sched);
1322         ip_vs_pe_put(pe);
1323
1324         /* decrease the module use count */
1325         ip_vs_use_count_dec();
1326
1327         return ret;
1328 }
1329
1330
1331 /*
1332  *      Edit a service and bind it with a new scheduler
1333  */
1334 static int
1335 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1336 {
1337         struct ip_vs_scheduler *sched = NULL, *old_sched;
1338         struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1339         int ret = 0;
1340         bool new_pe_conn_out, old_pe_conn_out;
1341
1342         /*
1343          * Lookup the scheduler, by 'u->sched_name'
1344          */
1345         if (strcmp(u->sched_name, "none")) {
1346                 sched = ip_vs_scheduler_get(u->sched_name);
1347                 if (!sched) {
1348                         pr_info("Scheduler module ip_vs_%s not found\n",
1349                                 u->sched_name);
1350                         return -ENOENT;
1351                 }
1352         }
1353         old_sched = sched;
1354
1355         if (u->pe_name && *u->pe_name) {
1356                 pe = ip_vs_pe_getbyname(u->pe_name);
1357                 if (pe == NULL) {
1358                         pr_info("persistence engine module ip_vs_pe_%s "
1359                                 "not found\n", u->pe_name);
1360                         ret = -ENOENT;
1361                         goto out;
1362                 }
1363                 old_pe = pe;
1364         }
1365
1366 #ifdef CONFIG_IP_VS_IPV6
1367         if (u->af == AF_INET6) {
1368                 __u32 plen = (__force __u32) u->netmask;
1369
1370                 if (plen < 1 || plen > 128) {
1371                         ret = -EINVAL;
1372                         goto out;
1373                 }
1374         }
1375 #endif
1376
1377         old_sched = rcu_dereference_protected(svc->scheduler, 1);
1378         if (sched != old_sched) {
1379                 if (old_sched) {
1380                         ip_vs_unbind_scheduler(svc, old_sched);
1381                         RCU_INIT_POINTER(svc->scheduler, NULL);
1382                         /* Wait all svc->sched_data users */
1383                         synchronize_rcu();
1384                 }
1385                 /* Bind the new scheduler */
1386                 if (sched) {
1387                         ret = ip_vs_bind_scheduler(svc, sched);
1388                         if (ret) {
1389                                 ip_vs_scheduler_put(sched);
1390                                 goto out;
1391                         }
1392                 }
1393         }
1394
1395         /*
1396          * Set the flags and timeout value
1397          */
1398         svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1399         svc->timeout = u->timeout * HZ;
1400         svc->netmask = u->netmask;
1401
1402         old_pe = rcu_dereference_protected(svc->pe, 1);
1403         if (pe != old_pe) {
1404                 rcu_assign_pointer(svc->pe, pe);
1405                 /* check for optional methods in new pe */
1406                 new_pe_conn_out = (pe && pe->conn_out) ? true : false;
1407                 old_pe_conn_out = (old_pe && old_pe->conn_out) ? true : false;
1408                 if (new_pe_conn_out && !old_pe_conn_out)
1409                         atomic_inc(&svc->ipvs->conn_out_counter);
1410                 if (old_pe_conn_out && !new_pe_conn_out)
1411                         atomic_dec(&svc->ipvs->conn_out_counter);
1412         }
1413
1414 out:
1415         ip_vs_scheduler_put(old_sched);
1416         ip_vs_pe_put(old_pe);
1417         return ret;
1418 }
1419
1420 /*
1421  *      Delete a service from the service list
1422  *      - The service must be unlinked, unlocked and not referenced!
1423  *      - We are called under _bh lock
1424  */
1425 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1426 {
1427         struct ip_vs_dest *dest, *nxt;
1428         struct ip_vs_scheduler *old_sched;
1429         struct ip_vs_pe *old_pe;
1430         struct netns_ipvs *ipvs = svc->ipvs;
1431
1432         /* Count only IPv4 services for old get/setsockopt interface */
1433         if (svc->af == AF_INET)
1434                 ipvs->num_services--;
1435
1436         ip_vs_stop_estimator(svc->ipvs, &svc->stats);
1437
1438         /* Unbind scheduler */
1439         old_sched = rcu_dereference_protected(svc->scheduler, 1);
1440         ip_vs_unbind_scheduler(svc, old_sched);
1441         ip_vs_scheduler_put(old_sched);
1442
1443         /* Unbind persistence engine, keep svc->pe */
1444         old_pe = rcu_dereference_protected(svc->pe, 1);
1445         if (old_pe && old_pe->conn_out)
1446                 atomic_dec(&ipvs->conn_out_counter);
1447         ip_vs_pe_put(old_pe);
1448
1449         /*
1450          *    Unlink the whole destination list
1451          */
1452         list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1453                 __ip_vs_unlink_dest(svc, dest, 0);
1454                 __ip_vs_del_dest(svc->ipvs, dest, cleanup);
1455         }
1456
1457         /*
1458          *    Update the virtual service counters
1459          */
1460         if (svc->port == FTPPORT)
1461                 atomic_dec(&ipvs->ftpsvc_counter);
1462         else if (svc->port == 0)
1463                 atomic_dec(&ipvs->nullsvc_counter);
1464
1465         /*
1466          *    Free the service if nobody refers to it
1467          */
1468         __ip_vs_svc_put(svc, true);
1469
1470         /* decrease the module use count */
1471         ip_vs_use_count_dec();
1472 }
1473
1474 /*
1475  * Unlink a service from list and try to delete it if its refcnt reached 0
1476  */
1477 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1478 {
1479         ip_vs_unregister_conntrack(svc);
1480         /* Hold svc to avoid double release from dest_trash */
1481         atomic_inc(&svc->refcnt);
1482         /*
1483          * Unhash it from the service table
1484          */
1485         ip_vs_svc_unhash(svc);
1486
1487         __ip_vs_del_service(svc, cleanup);
1488 }
1489
1490 /*
1491  *      Delete a service from the service list
1492  */
1493 static int ip_vs_del_service(struct ip_vs_service *svc)
1494 {
1495         if (svc == NULL)
1496                 return -EEXIST;
1497         ip_vs_unlink_service(svc, false);
1498
1499         return 0;
1500 }
1501
1502
1503 /*
1504  *      Flush all the virtual services
1505  */
1506 static int ip_vs_flush(struct netns_ipvs *ipvs, bool cleanup)
1507 {
1508         int idx;
1509         struct ip_vs_service *svc;
1510         struct hlist_node *n;
1511
1512         /*
1513          * Flush the service table hashed by <netns,protocol,addr,port>
1514          */
1515         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1516                 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1517                                           s_list) {
1518                         if (svc->ipvs == ipvs)
1519                                 ip_vs_unlink_service(svc, cleanup);
1520                 }
1521         }
1522
1523         /*
1524          * Flush the service table hashed by fwmark
1525          */
1526         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1527                 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1528                                           f_list) {
1529                         if (svc->ipvs == ipvs)
1530                                 ip_vs_unlink_service(svc, cleanup);
1531                 }
1532         }
1533
1534         return 0;
1535 }
1536
1537 /*
1538  *      Delete service by {netns} in the service table.
1539  *      Called by __ip_vs_cleanup()
1540  */
1541 void ip_vs_service_net_cleanup(struct netns_ipvs *ipvs)
1542 {
1543         EnterFunction(2);
1544         /* Check for "full" addressed entries */
1545         mutex_lock(&__ip_vs_mutex);
1546         ip_vs_flush(ipvs, true);
1547         mutex_unlock(&__ip_vs_mutex);
1548         LeaveFunction(2);
1549 }
1550
1551 /* Put all references for device (dst_cache) */
1552 static inline void
1553 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1554 {
1555         struct ip_vs_dest_dst *dest_dst;
1556
1557         spin_lock_bh(&dest->dst_lock);
1558         dest_dst = rcu_dereference_protected(dest->dest_dst, 1);
1559         if (dest_dst && dest_dst->dst_cache->dev == dev) {
1560                 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1561                               dev->name,
1562                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1563                               ntohs(dest->port),
1564                               refcount_read(&dest->refcnt));
1565                 __ip_vs_dst_cache_reset(dest);
1566         }
1567         spin_unlock_bh(&dest->dst_lock);
1568
1569 }
1570 /* Netdev event receiver
1571  * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1572  */
1573 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1574                            void *ptr)
1575 {
1576         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1577         struct net *net = dev_net(dev);
1578         struct netns_ipvs *ipvs = net_ipvs(net);
1579         struct ip_vs_service *svc;
1580         struct ip_vs_dest *dest;
1581         unsigned int idx;
1582
1583         if (event != NETDEV_DOWN || !ipvs)
1584                 return NOTIFY_DONE;
1585         IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1586         EnterFunction(2);
1587         mutex_lock(&__ip_vs_mutex);
1588         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1589                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1590                         if (svc->ipvs == ipvs) {
1591                                 list_for_each_entry(dest, &svc->destinations,
1592                                                     n_list) {
1593                                         ip_vs_forget_dev(dest, dev);
1594                                 }
1595                         }
1596                 }
1597
1598                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1599                         if (svc->ipvs == ipvs) {
1600                                 list_for_each_entry(dest, &svc->destinations,
1601                                                     n_list) {
1602                                         ip_vs_forget_dev(dest, dev);
1603                                 }
1604                         }
1605
1606                 }
1607         }
1608
1609         spin_lock_bh(&ipvs->dest_trash_lock);
1610         list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1611                 ip_vs_forget_dev(dest, dev);
1612         }
1613         spin_unlock_bh(&ipvs->dest_trash_lock);
1614         mutex_unlock(&__ip_vs_mutex);
1615         LeaveFunction(2);
1616         return NOTIFY_DONE;
1617 }
1618
1619 /*
1620  *      Zero counters in a service or all services
1621  */
1622 static int ip_vs_zero_service(struct ip_vs_service *svc)
1623 {
1624         struct ip_vs_dest *dest;
1625
1626         list_for_each_entry(dest, &svc->destinations, n_list) {
1627                 ip_vs_zero_stats(&dest->stats);
1628         }
1629         ip_vs_zero_stats(&svc->stats);
1630         return 0;
1631 }
1632
1633 static int ip_vs_zero_all(struct netns_ipvs *ipvs)
1634 {
1635         int idx;
1636         struct ip_vs_service *svc;
1637
1638         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1639                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1640                         if (svc->ipvs == ipvs)
1641                                 ip_vs_zero_service(svc);
1642                 }
1643         }
1644
1645         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1646                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1647                         if (svc->ipvs == ipvs)
1648                                 ip_vs_zero_service(svc);
1649                 }
1650         }
1651
1652         ip_vs_zero_stats(&ipvs->tot_stats);
1653         return 0;
1654 }
1655
1656 #ifdef CONFIG_SYSCTL
1657
1658 static int zero;
1659 static int three = 3;
1660
1661 static int
1662 proc_do_defense_mode(struct ctl_table *table, int write,
1663                      void __user *buffer, size_t *lenp, loff_t *ppos)
1664 {
1665         struct netns_ipvs *ipvs = table->extra2;
1666         int *valp = table->data;
1667         int val = *valp;
1668         int rc;
1669
1670         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1671         if (write && (*valp != val)) {
1672                 if ((*valp < 0) || (*valp > 3)) {
1673                         /* Restore the correct value */
1674                         *valp = val;
1675                 } else {
1676                         update_defense_level(ipvs);
1677                 }
1678         }
1679         return rc;
1680 }
1681
1682 static int
1683 proc_do_sync_threshold(struct ctl_table *table, int write,
1684                        void __user *buffer, size_t *lenp, loff_t *ppos)
1685 {
1686         int *valp = table->data;
1687         int val[2];
1688         int rc;
1689
1690         /* backup the value first */
1691         memcpy(val, valp, sizeof(val));
1692
1693         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1694         if (write && (valp[0] < 0 || valp[1] < 0 ||
1695             (valp[0] >= valp[1] && valp[1]))) {
1696                 /* Restore the correct value */
1697                 memcpy(valp, val, sizeof(val));
1698         }
1699         return rc;
1700 }
1701
1702 static int
1703 proc_do_sync_mode(struct ctl_table *table, int write,
1704                      void __user *buffer, size_t *lenp, loff_t *ppos)
1705 {
1706         int *valp = table->data;
1707         int val = *valp;
1708         int rc;
1709
1710         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1711         if (write && (*valp != val)) {
1712                 if ((*valp < 0) || (*valp > 1)) {
1713                         /* Restore the correct value */
1714                         *valp = val;
1715                 }
1716         }
1717         return rc;
1718 }
1719
1720 static int
1721 proc_do_sync_ports(struct ctl_table *table, int write,
1722                    void __user *buffer, size_t *lenp, loff_t *ppos)
1723 {
1724         int *valp = table->data;
1725         int val = *valp;
1726         int rc;
1727
1728         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1729         if (write && (*valp != val)) {
1730                 if (*valp < 1 || !is_power_of_2(*valp)) {
1731                         /* Restore the correct value */
1732                         *valp = val;
1733                 }
1734         }
1735         return rc;
1736 }
1737
1738 /*
1739  *      IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1740  *      Do not change order or insert new entries without
1741  *      align with netns init in ip_vs_control_net_init()
1742  */
1743
1744 static struct ctl_table vs_vars[] = {
1745         {
1746                 .procname       = "amemthresh",
1747                 .maxlen         = sizeof(int),
1748                 .mode           = 0644,
1749                 .proc_handler   = proc_dointvec,
1750         },
1751         {
1752                 .procname       = "am_droprate",
1753                 .maxlen         = sizeof(int),
1754                 .mode           = 0644,
1755                 .proc_handler   = proc_dointvec,
1756         },
1757         {
1758                 .procname       = "drop_entry",
1759                 .maxlen         = sizeof(int),
1760                 .mode           = 0644,
1761                 .proc_handler   = proc_do_defense_mode,
1762         },
1763         {
1764                 .procname       = "drop_packet",
1765                 .maxlen         = sizeof(int),
1766                 .mode           = 0644,
1767                 .proc_handler   = proc_do_defense_mode,
1768         },
1769 #ifdef CONFIG_IP_VS_NFCT
1770         {
1771                 .procname       = "conntrack",
1772                 .maxlen         = sizeof(int),
1773                 .mode           = 0644,
1774                 .proc_handler   = &proc_dointvec,
1775         },
1776 #endif
1777         {
1778                 .procname       = "secure_tcp",
1779                 .maxlen         = sizeof(int),
1780                 .mode           = 0644,
1781                 .proc_handler   = proc_do_defense_mode,
1782         },
1783         {
1784                 .procname       = "snat_reroute",
1785                 .maxlen         = sizeof(int),
1786                 .mode           = 0644,
1787                 .proc_handler   = &proc_dointvec,
1788         },
1789         {
1790                 .procname       = "sync_version",
1791                 .maxlen         = sizeof(int),
1792                 .mode           = 0644,
1793                 .proc_handler   = proc_do_sync_mode,
1794         },
1795         {
1796                 .procname       = "sync_ports",
1797                 .maxlen         = sizeof(int),
1798                 .mode           = 0644,
1799                 .proc_handler   = proc_do_sync_ports,
1800         },
1801         {
1802                 .procname       = "sync_persist_mode",
1803                 .maxlen         = sizeof(int),
1804                 .mode           = 0644,
1805                 .proc_handler   = proc_dointvec,
1806         },
1807         {
1808                 .procname       = "sync_qlen_max",
1809                 .maxlen         = sizeof(unsigned long),
1810                 .mode           = 0644,
1811                 .proc_handler   = proc_doulongvec_minmax,
1812         },
1813         {
1814                 .procname       = "sync_sock_size",
1815                 .maxlen         = sizeof(int),
1816                 .mode           = 0644,
1817                 .proc_handler   = proc_dointvec,
1818         },
1819         {
1820                 .procname       = "cache_bypass",
1821                 .maxlen         = sizeof(int),
1822                 .mode           = 0644,
1823                 .proc_handler   = proc_dointvec,
1824         },
1825         {
1826                 .procname       = "expire_nodest_conn",
1827                 .maxlen         = sizeof(int),
1828                 .mode           = 0644,
1829                 .proc_handler   = proc_dointvec,
1830         },
1831         {
1832                 .procname       = "sloppy_tcp",
1833                 .maxlen         = sizeof(int),
1834                 .mode           = 0644,
1835                 .proc_handler   = proc_dointvec,
1836         },
1837         {
1838                 .procname       = "sloppy_sctp",
1839                 .maxlen         = sizeof(int),
1840                 .mode           = 0644,
1841                 .proc_handler   = proc_dointvec,
1842         },
1843         {
1844                 .procname       = "expire_quiescent_template",
1845                 .maxlen         = sizeof(int),
1846                 .mode           = 0644,
1847                 .proc_handler   = proc_dointvec,
1848         },
1849         {
1850                 .procname       = "sync_threshold",
1851                 .maxlen         =
1852                         sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1853                 .mode           = 0644,
1854                 .proc_handler   = proc_do_sync_threshold,
1855         },
1856         {
1857                 .procname       = "sync_refresh_period",
1858                 .maxlen         = sizeof(int),
1859                 .mode           = 0644,
1860                 .proc_handler   = proc_dointvec_jiffies,
1861         },
1862         {
1863                 .procname       = "sync_retries",
1864                 .maxlen         = sizeof(int),
1865                 .mode           = 0644,
1866                 .proc_handler   = proc_dointvec_minmax,
1867                 .extra1         = &zero,
1868                 .extra2         = &three,
1869         },
1870         {
1871                 .procname       = "nat_icmp_send",
1872                 .maxlen         = sizeof(int),
1873                 .mode           = 0644,
1874                 .proc_handler   = proc_dointvec,
1875         },
1876         {
1877                 .procname       = "pmtu_disc",
1878                 .maxlen         = sizeof(int),
1879                 .mode           = 0644,
1880                 .proc_handler   = proc_dointvec,
1881         },
1882         {
1883                 .procname       = "backup_only",
1884                 .maxlen         = sizeof(int),
1885                 .mode           = 0644,
1886                 .proc_handler   = proc_dointvec,
1887         },
1888         {
1889                 .procname       = "conn_reuse_mode",
1890                 .maxlen         = sizeof(int),
1891                 .mode           = 0644,
1892                 .proc_handler   = proc_dointvec,
1893         },
1894         {
1895                 .procname       = "schedule_icmp",
1896                 .maxlen         = sizeof(int),
1897                 .mode           = 0644,
1898                 .proc_handler   = proc_dointvec,
1899         },
1900         {
1901                 .procname       = "ignore_tunneled",
1902                 .maxlen         = sizeof(int),
1903                 .mode           = 0644,
1904                 .proc_handler   = proc_dointvec,
1905         },
1906 #ifdef CONFIG_IP_VS_DEBUG
1907         {
1908                 .procname       = "debug_level",
1909                 .data           = &sysctl_ip_vs_debug_level,
1910                 .maxlen         = sizeof(int),
1911                 .mode           = 0644,
1912                 .proc_handler   = proc_dointvec,
1913         },
1914 #endif
1915         { }
1916 };
1917
1918 #endif
1919
1920 #ifdef CONFIG_PROC_FS
1921
1922 struct ip_vs_iter {
1923         struct seq_net_private p;  /* Do not move this, netns depends upon it*/
1924         struct hlist_head *table;
1925         int bucket;
1926 };
1927
1928 /*
1929  *      Write the contents of the VS rule table to a PROCfs file.
1930  *      (It is kept just for backward compatibility)
1931  */
1932 static inline const char *ip_vs_fwd_name(unsigned int flags)
1933 {
1934         switch (flags & IP_VS_CONN_F_FWD_MASK) {
1935         case IP_VS_CONN_F_LOCALNODE:
1936                 return "Local";
1937         case IP_VS_CONN_F_TUNNEL:
1938                 return "Tunnel";
1939         case IP_VS_CONN_F_DROUTE:
1940                 return "Route";
1941         default:
1942                 return "Masq";
1943         }
1944 }
1945
1946
1947 /* Get the Nth entry in the two lists */
1948 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1949 {
1950         struct net *net = seq_file_net(seq);
1951         struct netns_ipvs *ipvs = net_ipvs(net);
1952         struct ip_vs_iter *iter = seq->private;
1953         int idx;
1954         struct ip_vs_service *svc;
1955
1956         /* look in hash by protocol */
1957         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1958                 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
1959                         if ((svc->ipvs == ipvs) && pos-- == 0) {
1960                                 iter->table = ip_vs_svc_table;
1961                                 iter->bucket = idx;
1962                                 return svc;
1963                         }
1964                 }
1965         }
1966
1967         /* keep looking in fwmark */
1968         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1969                 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
1970                                          f_list) {
1971                         if ((svc->ipvs == ipvs) && pos-- == 0) {
1972                                 iter->table = ip_vs_svc_fwm_table;
1973                                 iter->bucket = idx;
1974                                 return svc;
1975                         }
1976                 }
1977         }
1978
1979         return NULL;
1980 }
1981
1982 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1983         __acquires(RCU)
1984 {
1985         rcu_read_lock();
1986         return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1987 }
1988
1989
1990 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1991 {
1992         struct hlist_node *e;
1993         struct ip_vs_iter *iter;
1994         struct ip_vs_service *svc;
1995
1996         ++*pos;
1997         if (v == SEQ_START_TOKEN)
1998                 return ip_vs_info_array(seq,0);
1999
2000         svc = v;
2001         iter = seq->private;
2002
2003         if (iter->table == ip_vs_svc_table) {
2004                 /* next service in table hashed by protocol */
2005                 e = rcu_dereference(hlist_next_rcu(&svc->s_list));
2006                 if (e)
2007                         return hlist_entry(e, struct ip_vs_service, s_list);
2008
2009                 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2010                         hlist_for_each_entry_rcu(svc,
2011                                                  &ip_vs_svc_table[iter->bucket],
2012                                                  s_list) {
2013                                 return svc;
2014                         }
2015                 }
2016
2017                 iter->table = ip_vs_svc_fwm_table;
2018                 iter->bucket = -1;
2019                 goto scan_fwmark;
2020         }
2021
2022         /* next service in hashed by fwmark */
2023         e = rcu_dereference(hlist_next_rcu(&svc->f_list));
2024         if (e)
2025                 return hlist_entry(e, struct ip_vs_service, f_list);
2026
2027  scan_fwmark:
2028         while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2029                 hlist_for_each_entry_rcu(svc,
2030                                          &ip_vs_svc_fwm_table[iter->bucket],
2031                                          f_list)
2032                         return svc;
2033         }
2034
2035         return NULL;
2036 }
2037
2038 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
2039         __releases(RCU)
2040 {
2041         rcu_read_unlock();
2042 }
2043
2044
2045 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
2046 {
2047         if (v == SEQ_START_TOKEN) {
2048                 seq_printf(seq,
2049                         "IP Virtual Server version %d.%d.%d (size=%d)\n",
2050                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2051                 seq_puts(seq,
2052                          "Prot LocalAddress:Port Scheduler Flags\n");
2053                 seq_puts(seq,
2054                          "  -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2055         } else {
2056                 struct net *net = seq_file_net(seq);
2057                 struct netns_ipvs *ipvs = net_ipvs(net);
2058                 const struct ip_vs_service *svc = v;
2059                 const struct ip_vs_iter *iter = seq->private;
2060                 const struct ip_vs_dest *dest;
2061                 struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler);
2062                 char *sched_name = sched ? sched->name : "none";
2063
2064                 if (svc->ipvs != ipvs)
2065                         return 0;
2066                 if (iter->table == ip_vs_svc_table) {
2067 #ifdef CONFIG_IP_VS_IPV6
2068                         if (svc->af == AF_INET6)
2069                                 seq_printf(seq, "%s  [%pI6]:%04X %s ",
2070                                            ip_vs_proto_name(svc->protocol),
2071                                            &svc->addr.in6,
2072                                            ntohs(svc->port),
2073                                            sched_name);
2074                         else
2075 #endif
2076                                 seq_printf(seq, "%s  %08X:%04X %s %s ",
2077                                            ip_vs_proto_name(svc->protocol),
2078                                            ntohl(svc->addr.ip),
2079                                            ntohs(svc->port),
2080                                            sched_name,
2081                                            (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2082                 } else {
2083                         seq_printf(seq, "FWM  %08X %s %s",
2084                                    svc->fwmark, sched_name,
2085                                    (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2086                 }
2087
2088                 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2089                         seq_printf(seq, "persistent %d %08X\n",
2090                                 svc->timeout,
2091                                 ntohl(svc->netmask));
2092                 else
2093                         seq_putc(seq, '\n');
2094
2095                 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
2096 #ifdef CONFIG_IP_VS_IPV6
2097                         if (dest->af == AF_INET6)
2098                                 seq_printf(seq,
2099                                            "  -> [%pI6]:%04X"
2100                                            "      %-7s %-6d %-10d %-10d\n",
2101                                            &dest->addr.in6,
2102                                            ntohs(dest->port),
2103                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2104                                            atomic_read(&dest->weight),
2105                                            atomic_read(&dest->activeconns),
2106                                            atomic_read(&dest->inactconns));
2107                         else
2108 #endif
2109                                 seq_printf(seq,
2110                                            "  -> %08X:%04X      "
2111                                            "%-7s %-6d %-10d %-10d\n",
2112                                            ntohl(dest->addr.ip),
2113                                            ntohs(dest->port),
2114                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2115                                            atomic_read(&dest->weight),
2116                                            atomic_read(&dest->activeconns),
2117                                            atomic_read(&dest->inactconns));
2118
2119                 }
2120         }
2121         return 0;
2122 }
2123
2124 static const struct seq_operations ip_vs_info_seq_ops = {
2125         .start = ip_vs_info_seq_start,
2126         .next  = ip_vs_info_seq_next,
2127         .stop  = ip_vs_info_seq_stop,
2128         .show  = ip_vs_info_seq_show,
2129 };
2130
2131 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2132 {
2133         struct net *net = seq_file_single_net(seq);
2134         struct ip_vs_kstats show;
2135
2136 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
2137         seq_puts(seq,
2138                  "   Total Incoming Outgoing         Incoming         Outgoing\n");
2139         seq_puts(seq,
2140                  "   Conns  Packets  Packets            Bytes            Bytes\n");
2141
2142         ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2143         seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n\n",
2144                    (unsigned long long)show.conns,
2145                    (unsigned long long)show.inpkts,
2146                    (unsigned long long)show.outpkts,
2147                    (unsigned long long)show.inbytes,
2148                    (unsigned long long)show.outbytes);
2149
2150 /*                01234567 01234567 01234567 0123456701234567 0123456701234567*/
2151         seq_puts(seq,
2152                  " Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
2153         seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n",
2154                    (unsigned long long)show.cps,
2155                    (unsigned long long)show.inpps,
2156                    (unsigned long long)show.outpps,
2157                    (unsigned long long)show.inbps,
2158                    (unsigned long long)show.outbps);
2159
2160         return 0;
2161 }
2162
2163 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2164 {
2165         struct net *net = seq_file_single_net(seq);
2166         struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2167         struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats;
2168         struct ip_vs_kstats kstats;
2169         int i;
2170
2171 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
2172         seq_puts(seq,
2173                  "       Total Incoming Outgoing         Incoming         Outgoing\n");
2174         seq_puts(seq,
2175                  "CPU    Conns  Packets  Packets            Bytes            Bytes\n");
2176
2177         for_each_possible_cpu(i) {
2178                 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2179                 unsigned int start;
2180                 u64 conns, inpkts, outpkts, inbytes, outbytes;
2181
2182                 do {
2183                         start = u64_stats_fetch_begin_irq(&u->syncp);
2184                         conns = u->cnt.conns;
2185                         inpkts = u->cnt.inpkts;
2186                         outpkts = u->cnt.outpkts;
2187                         inbytes = u->cnt.inbytes;
2188                         outbytes = u->cnt.outbytes;
2189                 } while (u64_stats_fetch_retry_irq(&u->syncp, start));
2190
2191                 seq_printf(seq, "%3X %8LX %8LX %8LX %16LX %16LX\n",
2192                            i, (u64)conns, (u64)inpkts,
2193                            (u64)outpkts, (u64)inbytes,
2194                            (u64)outbytes);
2195         }
2196
2197         ip_vs_copy_stats(&kstats, tot_stats);
2198
2199         seq_printf(seq, "  ~ %8LX %8LX %8LX %16LX %16LX\n\n",
2200                    (unsigned long long)kstats.conns,
2201                    (unsigned long long)kstats.inpkts,
2202                    (unsigned long long)kstats.outpkts,
2203                    (unsigned long long)kstats.inbytes,
2204                    (unsigned long long)kstats.outbytes);
2205
2206 /*                ... 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2207         seq_puts(seq,
2208                  "     Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
2209         seq_printf(seq, "    %8LX %8LX %8LX %16LX %16LX\n",
2210                    kstats.cps,
2211                    kstats.inpps,
2212                    kstats.outpps,
2213                    kstats.inbps,
2214                    kstats.outbps);
2215
2216         return 0;
2217 }
2218 #endif
2219
2220 /*
2221  *      Set timeout values for tcp tcpfin udp in the timeout_table.
2222  */
2223 static int ip_vs_set_timeout(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2224 {
2225 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2226         struct ip_vs_proto_data *pd;
2227 #endif
2228
2229         IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2230                   u->tcp_timeout,
2231                   u->tcp_fin_timeout,
2232                   u->udp_timeout);
2233
2234 #ifdef CONFIG_IP_VS_PROTO_TCP
2235         if (u->tcp_timeout < 0 || u->tcp_timeout > (INT_MAX / HZ) ||
2236             u->tcp_fin_timeout < 0 || u->tcp_fin_timeout > (INT_MAX / HZ)) {
2237                 return -EINVAL;
2238         }
2239 #endif
2240
2241 #ifdef CONFIG_IP_VS_PROTO_UDP
2242         if (u->udp_timeout < 0 || u->udp_timeout > (INT_MAX / HZ))
2243                 return -EINVAL;
2244 #endif
2245
2246 #ifdef CONFIG_IP_VS_PROTO_TCP
2247         if (u->tcp_timeout) {
2248                 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2249                 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2250                         = u->tcp_timeout * HZ;
2251         }
2252
2253         if (u->tcp_fin_timeout) {
2254                 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2255                 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2256                         = u->tcp_fin_timeout * HZ;
2257         }
2258 #endif
2259
2260 #ifdef CONFIG_IP_VS_PROTO_UDP
2261         if (u->udp_timeout) {
2262                 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2263                 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2264                         = u->udp_timeout * HZ;
2265         }
2266 #endif
2267         return 0;
2268 }
2269
2270 #define CMDID(cmd)              (cmd - IP_VS_BASE_CTL)
2271
2272 struct ip_vs_svcdest_user {
2273         struct ip_vs_service_user       s;
2274         struct ip_vs_dest_user          d;
2275 };
2276
2277 static const unsigned char set_arglen[CMDID(IP_VS_SO_SET_MAX) + 1] = {
2278         [CMDID(IP_VS_SO_SET_ADD)]         = sizeof(struct ip_vs_service_user),
2279         [CMDID(IP_VS_SO_SET_EDIT)]        = sizeof(struct ip_vs_service_user),
2280         [CMDID(IP_VS_SO_SET_DEL)]         = sizeof(struct ip_vs_service_user),
2281         [CMDID(IP_VS_SO_SET_ADDDEST)]     = sizeof(struct ip_vs_svcdest_user),
2282         [CMDID(IP_VS_SO_SET_DELDEST)]     = sizeof(struct ip_vs_svcdest_user),
2283         [CMDID(IP_VS_SO_SET_EDITDEST)]    = sizeof(struct ip_vs_svcdest_user),
2284         [CMDID(IP_VS_SO_SET_TIMEOUT)]     = sizeof(struct ip_vs_timeout_user),
2285         [CMDID(IP_VS_SO_SET_STARTDAEMON)] = sizeof(struct ip_vs_daemon_user),
2286         [CMDID(IP_VS_SO_SET_STOPDAEMON)]  = sizeof(struct ip_vs_daemon_user),
2287         [CMDID(IP_VS_SO_SET_ZERO)]        = sizeof(struct ip_vs_service_user),
2288 };
2289
2290 union ip_vs_set_arglen {
2291         struct ip_vs_service_user       field_IP_VS_SO_SET_ADD;
2292         struct ip_vs_service_user       field_IP_VS_SO_SET_EDIT;
2293         struct ip_vs_service_user       field_IP_VS_SO_SET_DEL;
2294         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_ADDDEST;
2295         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_DELDEST;
2296         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_EDITDEST;
2297         struct ip_vs_timeout_user       field_IP_VS_SO_SET_TIMEOUT;
2298         struct ip_vs_daemon_user        field_IP_VS_SO_SET_STARTDAEMON;
2299         struct ip_vs_daemon_user        field_IP_VS_SO_SET_STOPDAEMON;
2300         struct ip_vs_service_user       field_IP_VS_SO_SET_ZERO;
2301 };
2302
2303 #define MAX_SET_ARGLEN  sizeof(union ip_vs_set_arglen)
2304
2305 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2306                                   struct ip_vs_service_user *usvc_compat)
2307 {
2308         memset(usvc, 0, sizeof(*usvc));
2309
2310         usvc->af                = AF_INET;
2311         usvc->protocol          = usvc_compat->protocol;
2312         usvc->addr.ip           = usvc_compat->addr;
2313         usvc->port              = usvc_compat->port;
2314         usvc->fwmark            = usvc_compat->fwmark;
2315
2316         /* Deep copy of sched_name is not needed here */
2317         usvc->sched_name        = usvc_compat->sched_name;
2318
2319         usvc->flags             = usvc_compat->flags;
2320         usvc->timeout           = usvc_compat->timeout;
2321         usvc->netmask           = usvc_compat->netmask;
2322 }
2323
2324 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2325                                    struct ip_vs_dest_user *udest_compat)
2326 {
2327         memset(udest, 0, sizeof(*udest));
2328
2329         udest->addr.ip          = udest_compat->addr;
2330         udest->port             = udest_compat->port;
2331         udest->conn_flags       = udest_compat->conn_flags;
2332         udest->weight           = udest_compat->weight;
2333         udest->u_threshold      = udest_compat->u_threshold;
2334         udest->l_threshold      = udest_compat->l_threshold;
2335         udest->af               = AF_INET;
2336 }
2337
2338 static int
2339 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2340 {
2341         struct net *net = sock_net(sk);
2342         int ret;
2343         unsigned char arg[MAX_SET_ARGLEN];
2344         struct ip_vs_service_user *usvc_compat;
2345         struct ip_vs_service_user_kern usvc;
2346         struct ip_vs_service *svc;
2347         struct ip_vs_dest_user *udest_compat;
2348         struct ip_vs_dest_user_kern udest;
2349         struct netns_ipvs *ipvs = net_ipvs(net);
2350
2351         BUILD_BUG_ON(sizeof(arg) > 255);
2352         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2353                 return -EPERM;
2354
2355         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2356                 return -EINVAL;
2357         if (len != set_arglen[CMDID(cmd)]) {
2358                 IP_VS_DBG(1, "set_ctl: len %u != %u\n",
2359                           len, set_arglen[CMDID(cmd)]);
2360                 return -EINVAL;
2361         }
2362
2363         if (copy_from_user(arg, user, len) != 0)
2364                 return -EFAULT;
2365
2366         /* increase the module use count */
2367         ip_vs_use_count_inc();
2368
2369         /* Handle daemons since they have another lock */
2370         if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2371             cmd == IP_VS_SO_SET_STOPDAEMON) {
2372                 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2373
2374                 if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2375                         struct ipvs_sync_daemon_cfg cfg;
2376
2377                         memset(&cfg, 0, sizeof(cfg));
2378                         ret = -EINVAL;
2379                         if (strscpy(cfg.mcast_ifn, dm->mcast_ifn,
2380                                     sizeof(cfg.mcast_ifn)) <= 0)
2381                                 goto out_dec;
2382                         cfg.syncid = dm->syncid;
2383                         ret = start_sync_thread(ipvs, &cfg, dm->state);
2384                 } else {
2385                         mutex_lock(&ipvs->sync_mutex);
2386                         ret = stop_sync_thread(ipvs, dm->state);
2387                         mutex_unlock(&ipvs->sync_mutex);
2388                 }
2389                 goto out_dec;
2390         }
2391
2392         mutex_lock(&__ip_vs_mutex);
2393         if (cmd == IP_VS_SO_SET_FLUSH) {
2394                 /* Flush the virtual service */
2395                 ret = ip_vs_flush(ipvs, false);
2396                 goto out_unlock;
2397         } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2398                 /* Set timeout values for (tcp tcpfin udp) */
2399                 ret = ip_vs_set_timeout(ipvs, (struct ip_vs_timeout_user *)arg);
2400                 goto out_unlock;
2401         }
2402
2403         usvc_compat = (struct ip_vs_service_user *)arg;
2404         udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2405
2406         /* We only use the new structs internally, so copy userspace compat
2407          * structs to extended internal versions */
2408         ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2409         ip_vs_copy_udest_compat(&udest, udest_compat);
2410
2411         if (cmd == IP_VS_SO_SET_ZERO) {
2412                 /* if no service address is set, zero counters in all */
2413                 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2414                         ret = ip_vs_zero_all(ipvs);
2415                         goto out_unlock;
2416                 }
2417         }
2418
2419         if ((cmd == IP_VS_SO_SET_ADD || cmd == IP_VS_SO_SET_EDIT) &&
2420             strnlen(usvc.sched_name, IP_VS_SCHEDNAME_MAXLEN) ==
2421             IP_VS_SCHEDNAME_MAXLEN) {
2422                 ret = -EINVAL;
2423                 goto out_unlock;
2424         }
2425
2426         /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2427         if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2428             usvc.protocol != IPPROTO_SCTP) {
2429                 pr_err("set_ctl: invalid protocol: %d %pI4:%d\n",
2430                        usvc.protocol, &usvc.addr.ip,
2431                        ntohs(usvc.port));
2432                 ret = -EFAULT;
2433                 goto out_unlock;
2434         }
2435
2436         /* Lookup the exact service by <protocol, addr, port> or fwmark */
2437         rcu_read_lock();
2438         if (usvc.fwmark == 0)
2439                 svc = __ip_vs_service_find(ipvs, usvc.af, usvc.protocol,
2440                                            &usvc.addr, usvc.port);
2441         else
2442                 svc = __ip_vs_svc_fwm_find(ipvs, usvc.af, usvc.fwmark);
2443         rcu_read_unlock();
2444
2445         if (cmd != IP_VS_SO_SET_ADD
2446             && (svc == NULL || svc->protocol != usvc.protocol)) {
2447                 ret = -ESRCH;
2448                 goto out_unlock;
2449         }
2450
2451         switch (cmd) {
2452         case IP_VS_SO_SET_ADD:
2453                 if (svc != NULL)
2454                         ret = -EEXIST;
2455                 else
2456                         ret = ip_vs_add_service(ipvs, &usvc, &svc);
2457                 break;
2458         case IP_VS_SO_SET_EDIT:
2459                 ret = ip_vs_edit_service(svc, &usvc);
2460                 break;
2461         case IP_VS_SO_SET_DEL:
2462                 ret = ip_vs_del_service(svc);
2463                 if (!ret)
2464                         goto out_unlock;
2465                 break;
2466         case IP_VS_SO_SET_ZERO:
2467                 ret = ip_vs_zero_service(svc);
2468                 break;
2469         case IP_VS_SO_SET_ADDDEST:
2470                 ret = ip_vs_add_dest(svc, &udest);
2471                 break;
2472         case IP_VS_SO_SET_EDITDEST:
2473                 ret = ip_vs_edit_dest(svc, &udest);
2474                 break;
2475         case IP_VS_SO_SET_DELDEST:
2476                 ret = ip_vs_del_dest(svc, &udest);
2477                 break;
2478         default:
2479                 ret = -EINVAL;
2480         }
2481
2482   out_unlock:
2483         mutex_unlock(&__ip_vs_mutex);
2484   out_dec:
2485         /* decrease the module use count */
2486         ip_vs_use_count_dec();
2487
2488         return ret;
2489 }
2490
2491
2492 static void
2493 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2494 {
2495         struct ip_vs_scheduler *sched;
2496         struct ip_vs_kstats kstats;
2497         char *sched_name;
2498
2499         sched = rcu_dereference_protected(src->scheduler, 1);
2500         sched_name = sched ? sched->name : "none";
2501         dst->protocol = src->protocol;
2502         dst->addr = src->addr.ip;
2503         dst->port = src->port;
2504         dst->fwmark = src->fwmark;
2505         strlcpy(dst->sched_name, sched_name, sizeof(dst->sched_name));
2506         dst->flags = src->flags;
2507         dst->timeout = src->timeout / HZ;
2508         dst->netmask = src->netmask;
2509         dst->num_dests = src->num_dests;
2510         ip_vs_copy_stats(&kstats, &src->stats);
2511         ip_vs_export_stats_user(&dst->stats, &kstats);
2512 }
2513
2514 static inline int
2515 __ip_vs_get_service_entries(struct netns_ipvs *ipvs,
2516                             const struct ip_vs_get_services *get,
2517                             struct ip_vs_get_services __user *uptr)
2518 {
2519         int idx, count=0;
2520         struct ip_vs_service *svc;
2521         struct ip_vs_service_entry entry;
2522         int ret = 0;
2523
2524         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2525                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2526                         /* Only expose IPv4 entries to old interface */
2527                         if (svc->af != AF_INET || (svc->ipvs != ipvs))
2528                                 continue;
2529
2530                         if (count >= get->num_services)
2531                                 goto out;
2532                         memset(&entry, 0, sizeof(entry));
2533                         ip_vs_copy_service(&entry, svc);
2534                         if (copy_to_user(&uptr->entrytable[count],
2535                                          &entry, sizeof(entry))) {
2536                                 ret = -EFAULT;
2537                                 goto out;
2538                         }
2539                         count++;
2540                 }
2541         }
2542
2543         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2544                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2545                         /* Only expose IPv4 entries to old interface */
2546                         if (svc->af != AF_INET || (svc->ipvs != ipvs))
2547                                 continue;
2548
2549                         if (count >= get->num_services)
2550                                 goto out;
2551                         memset(&entry, 0, sizeof(entry));
2552                         ip_vs_copy_service(&entry, svc);
2553                         if (copy_to_user(&uptr->entrytable[count],
2554                                          &entry, sizeof(entry))) {
2555                                 ret = -EFAULT;
2556                                 goto out;
2557                         }
2558                         count++;
2559                 }
2560         }
2561 out:
2562         return ret;
2563 }
2564
2565 static inline int
2566 __ip_vs_get_dest_entries(struct netns_ipvs *ipvs, const struct ip_vs_get_dests *get,
2567                          struct ip_vs_get_dests __user *uptr)
2568 {
2569         struct ip_vs_service *svc;
2570         union nf_inet_addr addr = { .ip = get->addr };
2571         int ret = 0;
2572
2573         rcu_read_lock();
2574         if (get->fwmark)
2575                 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, get->fwmark);
2576         else
2577                 svc = __ip_vs_service_find(ipvs, AF_INET, get->protocol, &addr,
2578                                            get->port);
2579         rcu_read_unlock();
2580
2581         if (svc) {
2582                 int count = 0;
2583                 struct ip_vs_dest *dest;
2584                 struct ip_vs_dest_entry entry;
2585                 struct ip_vs_kstats kstats;
2586
2587                 memset(&entry, 0, sizeof(entry));
2588                 list_for_each_entry(dest, &svc->destinations, n_list) {
2589                         if (count >= get->num_dests)
2590                                 break;
2591
2592                         /* Cannot expose heterogeneous members via sockopt
2593                          * interface
2594                          */
2595                         if (dest->af != svc->af)
2596                                 continue;
2597
2598                         entry.addr = dest->addr.ip;
2599                         entry.port = dest->port;
2600                         entry.conn_flags = atomic_read(&dest->conn_flags);
2601                         entry.weight = atomic_read(&dest->weight);
2602                         entry.u_threshold = dest->u_threshold;
2603                         entry.l_threshold = dest->l_threshold;
2604                         entry.activeconns = atomic_read(&dest->activeconns);
2605                         entry.inactconns = atomic_read(&dest->inactconns);
2606                         entry.persistconns = atomic_read(&dest->persistconns);
2607                         ip_vs_copy_stats(&kstats, &dest->stats);
2608                         ip_vs_export_stats_user(&entry.stats, &kstats);
2609                         if (copy_to_user(&uptr->entrytable[count],
2610                                          &entry, sizeof(entry))) {
2611                                 ret = -EFAULT;
2612                                 break;
2613                         }
2614                         count++;
2615                 }
2616         } else
2617                 ret = -ESRCH;
2618         return ret;
2619 }
2620
2621 static inline void
2622 __ip_vs_get_timeouts(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2623 {
2624 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2625         struct ip_vs_proto_data *pd;
2626 #endif
2627
2628         memset(u, 0, sizeof (*u));
2629
2630 #ifdef CONFIG_IP_VS_PROTO_TCP
2631         pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2632         u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2633         u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2634 #endif
2635 #ifdef CONFIG_IP_VS_PROTO_UDP
2636         pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2637         u->udp_timeout =
2638                         pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2639 #endif
2640 }
2641
2642 static const unsigned char get_arglen[CMDID(IP_VS_SO_GET_MAX) + 1] = {
2643         [CMDID(IP_VS_SO_GET_VERSION)]  = 64,
2644         [CMDID(IP_VS_SO_GET_INFO)]     = sizeof(struct ip_vs_getinfo),
2645         [CMDID(IP_VS_SO_GET_SERVICES)] = sizeof(struct ip_vs_get_services),
2646         [CMDID(IP_VS_SO_GET_SERVICE)]  = sizeof(struct ip_vs_service_entry),
2647         [CMDID(IP_VS_SO_GET_DESTS)]    = sizeof(struct ip_vs_get_dests),
2648         [CMDID(IP_VS_SO_GET_TIMEOUT)]  = sizeof(struct ip_vs_timeout_user),
2649         [CMDID(IP_VS_SO_GET_DAEMON)]   = 2 * sizeof(struct ip_vs_daemon_user),
2650 };
2651
2652 union ip_vs_get_arglen {
2653         char                            field_IP_VS_SO_GET_VERSION[64];
2654         struct ip_vs_getinfo            field_IP_VS_SO_GET_INFO;
2655         struct ip_vs_get_services       field_IP_VS_SO_GET_SERVICES;
2656         struct ip_vs_service_entry      field_IP_VS_SO_GET_SERVICE;
2657         struct ip_vs_get_dests          field_IP_VS_SO_GET_DESTS;
2658         struct ip_vs_timeout_user       field_IP_VS_SO_GET_TIMEOUT;
2659         struct ip_vs_daemon_user        field_IP_VS_SO_GET_DAEMON[2];
2660 };
2661
2662 #define MAX_GET_ARGLEN  sizeof(union ip_vs_get_arglen)
2663
2664 static int
2665 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2666 {
2667         unsigned char arg[MAX_GET_ARGLEN];
2668         int ret = 0;
2669         unsigned int copylen;
2670         struct net *net = sock_net(sk);
2671         struct netns_ipvs *ipvs = net_ipvs(net);
2672
2673         BUG_ON(!net);
2674         BUILD_BUG_ON(sizeof(arg) > 255);
2675         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2676                 return -EPERM;
2677
2678         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2679                 return -EINVAL;
2680
2681         copylen = get_arglen[CMDID(cmd)];
2682         if (*len < (int) copylen) {
2683                 IP_VS_DBG(1, "get_ctl: len %d < %u\n", *len, copylen);
2684                 return -EINVAL;
2685         }
2686
2687         if (copy_from_user(arg, user, copylen) != 0)
2688                 return -EFAULT;
2689         /*
2690          * Handle daemons first since it has its own locking
2691          */
2692         if (cmd == IP_VS_SO_GET_DAEMON) {
2693                 struct ip_vs_daemon_user d[2];
2694
2695                 memset(&d, 0, sizeof(d));
2696                 mutex_lock(&ipvs->sync_mutex);
2697                 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2698                         d[0].state = IP_VS_STATE_MASTER;
2699                         strlcpy(d[0].mcast_ifn, ipvs->mcfg.mcast_ifn,
2700                                 sizeof(d[0].mcast_ifn));
2701                         d[0].syncid = ipvs->mcfg.syncid;
2702                 }
2703                 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2704                         d[1].state = IP_VS_STATE_BACKUP;
2705                         strlcpy(d[1].mcast_ifn, ipvs->bcfg.mcast_ifn,
2706                                 sizeof(d[1].mcast_ifn));
2707                         d[1].syncid = ipvs->bcfg.syncid;
2708                 }
2709                 if (copy_to_user(user, &d, sizeof(d)) != 0)
2710                         ret = -EFAULT;
2711                 mutex_unlock(&ipvs->sync_mutex);
2712                 return ret;
2713         }
2714
2715         mutex_lock(&__ip_vs_mutex);
2716         switch (cmd) {
2717         case IP_VS_SO_GET_VERSION:
2718         {
2719                 char buf[64];
2720
2721                 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2722                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2723                 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2724                         ret = -EFAULT;
2725                         goto out;
2726                 }
2727                 *len = strlen(buf)+1;
2728         }
2729         break;
2730
2731         case IP_VS_SO_GET_INFO:
2732         {
2733                 struct ip_vs_getinfo info;
2734                 info.version = IP_VS_VERSION_CODE;
2735                 info.size = ip_vs_conn_tab_size;
2736                 info.num_services = ipvs->num_services;
2737                 if (copy_to_user(user, &info, sizeof(info)) != 0)
2738                         ret = -EFAULT;
2739         }
2740         break;
2741
2742         case IP_VS_SO_GET_SERVICES:
2743         {
2744                 struct ip_vs_get_services *get;
2745                 int size;
2746
2747                 get = (struct ip_vs_get_services *)arg;
2748                 size = sizeof(*get) +
2749                         sizeof(struct ip_vs_service_entry) * get->num_services;
2750                 if (*len != size) {
2751                         pr_err("length: %u != %u\n", *len, size);
2752                         ret = -EINVAL;
2753                         goto out;
2754                 }
2755                 ret = __ip_vs_get_service_entries(ipvs, get, user);
2756         }
2757         break;
2758
2759         case IP_VS_SO_GET_SERVICE:
2760         {
2761                 struct ip_vs_service_entry *entry;
2762                 struct ip_vs_service *svc;
2763                 union nf_inet_addr addr;
2764
2765                 entry = (struct ip_vs_service_entry *)arg;
2766                 addr.ip = entry->addr;
2767                 rcu_read_lock();
2768                 if (entry->fwmark)
2769                         svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, entry->fwmark);
2770                 else
2771                         svc = __ip_vs_service_find(ipvs, AF_INET,
2772                                                    entry->protocol, &addr,
2773                                                    entry->port);
2774                 rcu_read_unlock();
2775                 if (svc) {
2776                         ip_vs_copy_service(entry, svc);
2777                         if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2778                                 ret = -EFAULT;
2779                 } else
2780                         ret = -ESRCH;
2781         }
2782         break;
2783
2784         case IP_VS_SO_GET_DESTS:
2785         {
2786                 struct ip_vs_get_dests *get;
2787                 int size;
2788
2789                 get = (struct ip_vs_get_dests *)arg;
2790                 size = sizeof(*get) +
2791                         sizeof(struct ip_vs_dest_entry) * get->num_dests;
2792                 if (*len != size) {
2793                         pr_err("length: %u != %u\n", *len, size);
2794                         ret = -EINVAL;
2795                         goto out;
2796                 }
2797                 ret = __ip_vs_get_dest_entries(ipvs, get, user);
2798         }
2799         break;
2800
2801         case IP_VS_SO_GET_TIMEOUT:
2802         {
2803                 struct ip_vs_timeout_user t;
2804
2805                 __ip_vs_get_timeouts(ipvs, &t);
2806                 if (copy_to_user(user, &t, sizeof(t)) != 0)
2807                         ret = -EFAULT;
2808         }
2809         break;
2810
2811         default:
2812                 ret = -EINVAL;
2813         }
2814
2815 out:
2816         mutex_unlock(&__ip_vs_mutex);
2817         return ret;
2818 }
2819
2820
2821 static struct nf_sockopt_ops ip_vs_sockopts = {
2822         .pf             = PF_INET,
2823         .set_optmin     = IP_VS_BASE_CTL,
2824         .set_optmax     = IP_VS_SO_SET_MAX+1,
2825         .set            = do_ip_vs_set_ctl,
2826         .get_optmin     = IP_VS_BASE_CTL,
2827         .get_optmax     = IP_VS_SO_GET_MAX+1,
2828         .get            = do_ip_vs_get_ctl,
2829         .owner          = THIS_MODULE,
2830 };
2831
2832 /*
2833  * Generic Netlink interface
2834  */
2835
2836 /* IPVS genetlink family */
2837 static struct genl_family ip_vs_genl_family;
2838
2839 /* Policy used for first-level command attributes */
2840 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2841         [IPVS_CMD_ATTR_SERVICE]         = { .type = NLA_NESTED },
2842         [IPVS_CMD_ATTR_DEST]            = { .type = NLA_NESTED },
2843         [IPVS_CMD_ATTR_DAEMON]          = { .type = NLA_NESTED },
2844         [IPVS_CMD_ATTR_TIMEOUT_TCP]     = { .type = NLA_U32 },
2845         [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2846         [IPVS_CMD_ATTR_TIMEOUT_UDP]     = { .type = NLA_U32 },
2847 };
2848
2849 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2850 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2851         [IPVS_DAEMON_ATTR_STATE]        = { .type = NLA_U32 },
2852         [IPVS_DAEMON_ATTR_MCAST_IFN]    = { .type = NLA_NUL_STRING,
2853                                             .len = IP_VS_IFNAME_MAXLEN - 1 },
2854         [IPVS_DAEMON_ATTR_SYNC_ID]      = { .type = NLA_U32 },
2855         [IPVS_DAEMON_ATTR_SYNC_MAXLEN]  = { .type = NLA_U16 },
2856         [IPVS_DAEMON_ATTR_MCAST_GROUP]  = { .type = NLA_U32 },
2857         [IPVS_DAEMON_ATTR_MCAST_GROUP6] = { .len = sizeof(struct in6_addr) },
2858         [IPVS_DAEMON_ATTR_MCAST_PORT]   = { .type = NLA_U16 },
2859         [IPVS_DAEMON_ATTR_MCAST_TTL]    = { .type = NLA_U8 },
2860 };
2861
2862 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2863 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2864         [IPVS_SVC_ATTR_AF]              = { .type = NLA_U16 },
2865         [IPVS_SVC_ATTR_PROTOCOL]        = { .type = NLA_U16 },
2866         [IPVS_SVC_ATTR_ADDR]            = { .type = NLA_BINARY,
2867                                             .len = sizeof(union nf_inet_addr) },
2868         [IPVS_SVC_ATTR_PORT]            = { .type = NLA_U16 },
2869         [IPVS_SVC_ATTR_FWMARK]          = { .type = NLA_U32 },
2870         [IPVS_SVC_ATTR_SCHED_NAME]      = { .type = NLA_NUL_STRING,
2871                                             .len = IP_VS_SCHEDNAME_MAXLEN - 1 },
2872         [IPVS_SVC_ATTR_PE_NAME]         = { .type = NLA_NUL_STRING,
2873                                             .len = IP_VS_PENAME_MAXLEN },
2874         [IPVS_SVC_ATTR_FLAGS]           = { .type = NLA_BINARY,
2875                                             .len = sizeof(struct ip_vs_flags) },
2876         [IPVS_SVC_ATTR_TIMEOUT]         = { .type = NLA_U32 },
2877         [IPVS_SVC_ATTR_NETMASK]         = { .type = NLA_U32 },
2878         [IPVS_SVC_ATTR_STATS]           = { .type = NLA_NESTED },
2879 };
2880
2881 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2882 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2883         [IPVS_DEST_ATTR_ADDR]           = { .type = NLA_BINARY,
2884                                             .len = sizeof(union nf_inet_addr) },
2885         [IPVS_DEST_ATTR_PORT]           = { .type = NLA_U16 },
2886         [IPVS_DEST_ATTR_FWD_METHOD]     = { .type = NLA_U32 },
2887         [IPVS_DEST_ATTR_WEIGHT]         = { .type = NLA_U32 },
2888         [IPVS_DEST_ATTR_U_THRESH]       = { .type = NLA_U32 },
2889         [IPVS_DEST_ATTR_L_THRESH]       = { .type = NLA_U32 },
2890         [IPVS_DEST_ATTR_ACTIVE_CONNS]   = { .type = NLA_U32 },
2891         [IPVS_DEST_ATTR_INACT_CONNS]    = { .type = NLA_U32 },
2892         [IPVS_DEST_ATTR_PERSIST_CONNS]  = { .type = NLA_U32 },
2893         [IPVS_DEST_ATTR_STATS]          = { .type = NLA_NESTED },
2894         [IPVS_DEST_ATTR_ADDR_FAMILY]    = { .type = NLA_U16 },
2895 };
2896
2897 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2898                                  struct ip_vs_kstats *kstats)
2899 {
2900         struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2901
2902         if (!nl_stats)
2903                 return -EMSGSIZE;
2904
2905         if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, (u32)kstats->conns) ||
2906             nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, (u32)kstats->inpkts) ||
2907             nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, (u32)kstats->outpkts) ||
2908             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes,
2909                               IPVS_STATS_ATTR_PAD) ||
2910             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes,
2911                               IPVS_STATS_ATTR_PAD) ||
2912             nla_put_u32(skb, IPVS_STATS_ATTR_CPS, (u32)kstats->cps) ||
2913             nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, (u32)kstats->inpps) ||
2914             nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, (u32)kstats->outpps) ||
2915             nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, (u32)kstats->inbps) ||
2916             nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, (u32)kstats->outbps))
2917                 goto nla_put_failure;
2918         nla_nest_end(skb, nl_stats);
2919
2920         return 0;
2921
2922 nla_put_failure:
2923         nla_nest_cancel(skb, nl_stats);
2924         return -EMSGSIZE;
2925 }
2926
2927 static int ip_vs_genl_fill_stats64(struct sk_buff *skb, int container_type,
2928                                    struct ip_vs_kstats *kstats)
2929 {
2930         struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2931
2932         if (!nl_stats)
2933                 return -EMSGSIZE;
2934
2935         if (nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CONNS, kstats->conns,
2936                               IPVS_STATS_ATTR_PAD) ||
2937             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPKTS, kstats->inpkts,
2938                               IPVS_STATS_ATTR_PAD) ||
2939             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPKTS, kstats->outpkts,
2940                               IPVS_STATS_ATTR_PAD) ||
2941             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes,
2942                               IPVS_STATS_ATTR_PAD) ||
2943             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes,
2944                               IPVS_STATS_ATTR_PAD) ||
2945             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CPS, kstats->cps,
2946                               IPVS_STATS_ATTR_PAD) ||
2947             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPPS, kstats->inpps,
2948                               IPVS_STATS_ATTR_PAD) ||
2949             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPPS, kstats->outpps,
2950                               IPVS_STATS_ATTR_PAD) ||
2951             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBPS, kstats->inbps,
2952                               IPVS_STATS_ATTR_PAD) ||
2953             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBPS, kstats->outbps,
2954                               IPVS_STATS_ATTR_PAD))
2955                 goto nla_put_failure;
2956         nla_nest_end(skb, nl_stats);
2957
2958         return 0;
2959
2960 nla_put_failure:
2961         nla_nest_cancel(skb, nl_stats);
2962         return -EMSGSIZE;
2963 }
2964
2965 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2966                                    struct ip_vs_service *svc)
2967 {
2968         struct ip_vs_scheduler *sched;
2969         struct ip_vs_pe *pe;
2970         struct nlattr *nl_service;
2971         struct ip_vs_flags flags = { .flags = svc->flags,
2972                                      .mask = ~0 };
2973         struct ip_vs_kstats kstats;
2974         char *sched_name;
2975
2976         nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2977         if (!nl_service)
2978                 return -EMSGSIZE;
2979
2980         if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2981                 goto nla_put_failure;
2982         if (svc->fwmark) {
2983                 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2984                         goto nla_put_failure;
2985         } else {
2986                 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2987                     nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2988                     nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2989                         goto nla_put_failure;
2990         }
2991
2992         sched = rcu_dereference_protected(svc->scheduler, 1);
2993         sched_name = sched ? sched->name : "none";
2994         pe = rcu_dereference_protected(svc->pe, 1);
2995         if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched_name) ||
2996             (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
2997             nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2998             nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2999             nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
3000                 goto nla_put_failure;
3001         ip_vs_copy_stats(&kstats, &svc->stats);
3002         if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &kstats))
3003                 goto nla_put_failure;
3004         if (ip_vs_genl_fill_stats64(skb, IPVS_SVC_ATTR_STATS64, &kstats))
3005                 goto nla_put_failure;
3006
3007         nla_nest_end(skb, nl_service);
3008
3009         return 0;
3010
3011 nla_put_failure:
3012         nla_nest_cancel(skb, nl_service);
3013         return -EMSGSIZE;
3014 }
3015
3016 static int ip_vs_genl_dump_service(struct sk_buff *skb,
3017                                    struct ip_vs_service *svc,
3018                                    struct netlink_callback *cb)
3019 {
3020         void *hdr;
3021
3022         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3023                           &ip_vs_genl_family, NLM_F_MULTI,
3024                           IPVS_CMD_NEW_SERVICE);
3025         if (!hdr)
3026                 return -EMSGSIZE;
3027
3028         if (ip_vs_genl_fill_service(skb, svc) < 0)
3029                 goto nla_put_failure;
3030
3031         genlmsg_end(skb, hdr);
3032         return 0;
3033
3034 nla_put_failure:
3035         genlmsg_cancel(skb, hdr);
3036         return -EMSGSIZE;
3037 }
3038
3039 static int ip_vs_genl_dump_services(struct sk_buff *skb,
3040                                     struct netlink_callback *cb)
3041 {
3042         int idx = 0, i;
3043         int start = cb->args[0];
3044         struct ip_vs_service *svc;
3045         struct net *net = sock_net(skb->sk);
3046         struct netns_ipvs *ipvs = net_ipvs(net);
3047
3048         mutex_lock(&__ip_vs_mutex);
3049         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3050                 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
3051                         if (++idx <= start || (svc->ipvs != ipvs))
3052                                 continue;
3053                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3054                                 idx--;
3055                                 goto nla_put_failure;
3056                         }
3057                 }
3058         }
3059
3060         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3061                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
3062                         if (++idx <= start || (svc->ipvs != ipvs))
3063                                 continue;
3064                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3065                                 idx--;
3066                                 goto nla_put_failure;
3067                         }
3068                 }
3069         }
3070
3071 nla_put_failure:
3072         mutex_unlock(&__ip_vs_mutex);
3073         cb->args[0] = idx;
3074
3075         return skb->len;
3076 }
3077
3078 static bool ip_vs_is_af_valid(int af)
3079 {
3080         if (af == AF_INET)
3081                 return true;
3082 #ifdef CONFIG_IP_VS_IPV6
3083         if (af == AF_INET6 && ipv6_mod_enabled())
3084                 return true;
3085 #endif
3086         return false;
3087 }
3088
3089 static int ip_vs_genl_parse_service(struct netns_ipvs *ipvs,
3090                                     struct ip_vs_service_user_kern *usvc,
3091                                     struct nlattr *nla, int full_entry,
3092                                     struct ip_vs_service **ret_svc)
3093 {
3094         struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
3095         struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
3096         struct ip_vs_service *svc;
3097
3098         /* Parse mandatory identifying service fields first */
3099         if (nla == NULL ||
3100             nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla,
3101                              ip_vs_svc_policy, NULL))
3102                 return -EINVAL;
3103
3104         nla_af          = attrs[IPVS_SVC_ATTR_AF];
3105         nla_protocol    = attrs[IPVS_SVC_ATTR_PROTOCOL];
3106         nla_addr        = attrs[IPVS_SVC_ATTR_ADDR];
3107         nla_port        = attrs[IPVS_SVC_ATTR_PORT];
3108         nla_fwmark      = attrs[IPVS_SVC_ATTR_FWMARK];
3109
3110         if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3111                 return -EINVAL;
3112
3113         memset(usvc, 0, sizeof(*usvc));
3114
3115         usvc->af = nla_get_u16(nla_af);
3116         if (!ip_vs_is_af_valid(usvc->af))
3117                 return -EAFNOSUPPORT;
3118
3119         if (nla_fwmark) {
3120                 usvc->protocol = IPPROTO_TCP;
3121                 usvc->fwmark = nla_get_u32(nla_fwmark);
3122         } else {
3123                 usvc->protocol = nla_get_u16(nla_protocol);
3124                 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3125                 usvc->port = nla_get_be16(nla_port);
3126                 usvc->fwmark = 0;
3127         }
3128
3129         rcu_read_lock();
3130         if (usvc->fwmark)
3131                 svc = __ip_vs_svc_fwm_find(ipvs, usvc->af, usvc->fwmark);
3132         else
3133                 svc = __ip_vs_service_find(ipvs, usvc->af, usvc->protocol,
3134                                            &usvc->addr, usvc->port);
3135         rcu_read_unlock();
3136         *ret_svc = svc;
3137
3138         /* If a full entry was requested, check for the additional fields */
3139         if (full_entry) {
3140                 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3141                               *nla_netmask;
3142                 struct ip_vs_flags flags;
3143
3144                 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3145                 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3146                 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3147                 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3148                 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3149
3150                 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3151                         return -EINVAL;
3152
3153                 nla_memcpy(&flags, nla_flags, sizeof(flags));
3154
3155                 /* prefill flags from service if it already exists */
3156                 if (svc)
3157                         usvc->flags = svc->flags;
3158
3159                 /* set new flags from userland */
3160                 usvc->flags = (usvc->flags & ~flags.mask) |
3161                               (flags.flags & flags.mask);
3162                 usvc->sched_name = nla_data(nla_sched);
3163                 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3164                 usvc->timeout = nla_get_u32(nla_timeout);
3165                 usvc->netmask = nla_get_be32(nla_netmask);
3166         }
3167
3168         return 0;
3169 }
3170
3171 static struct ip_vs_service *ip_vs_genl_find_service(struct netns_ipvs *ipvs,
3172                                                      struct nlattr *nla)
3173 {
3174         struct ip_vs_service_user_kern usvc;
3175         struct ip_vs_service *svc;
3176         int ret;
3177
3178         ret = ip_vs_genl_parse_service(ipvs, &usvc, nla, 0, &svc);
3179         return ret ? ERR_PTR(ret) : svc;
3180 }
3181
3182 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3183 {
3184         struct nlattr *nl_dest;
3185         struct ip_vs_kstats kstats;
3186
3187         nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3188         if (!nl_dest)
3189                 return -EMSGSIZE;
3190
3191         if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3192             nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3193             nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3194                         (atomic_read(&dest->conn_flags) &
3195                          IP_VS_CONN_F_FWD_MASK)) ||
3196             nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3197                         atomic_read(&dest->weight)) ||
3198             nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3199             nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3200             nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3201                         atomic_read(&dest->activeconns)) ||
3202             nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3203                         atomic_read(&dest->inactconns)) ||
3204             nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3205                         atomic_read(&dest->persistconns)) ||
3206             nla_put_u16(skb, IPVS_DEST_ATTR_ADDR_FAMILY, dest->af))
3207                 goto nla_put_failure;
3208         ip_vs_copy_stats(&kstats, &dest->stats);
3209         if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &kstats))
3210                 goto nla_put_failure;
3211         if (ip_vs_genl_fill_stats64(skb, IPVS_DEST_ATTR_STATS64, &kstats))
3212                 goto nla_put_failure;
3213
3214         nla_nest_end(skb, nl_dest);
3215
3216         return 0;
3217
3218 nla_put_failure:
3219         nla_nest_cancel(skb, nl_dest);
3220         return -EMSGSIZE;
3221 }
3222
3223 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3224                                 struct netlink_callback *cb)
3225 {
3226         void *hdr;
3227
3228         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3229                           &ip_vs_genl_family, NLM_F_MULTI,
3230                           IPVS_CMD_NEW_DEST);
3231         if (!hdr)
3232                 return -EMSGSIZE;
3233
3234         if (ip_vs_genl_fill_dest(skb, dest) < 0)
3235                 goto nla_put_failure;
3236
3237         genlmsg_end(skb, hdr);
3238         return 0;
3239
3240 nla_put_failure:
3241         genlmsg_cancel(skb, hdr);
3242         return -EMSGSIZE;
3243 }
3244
3245 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3246                                  struct netlink_callback *cb)
3247 {
3248         int idx = 0;
3249         int start = cb->args[0];
3250         struct ip_vs_service *svc;
3251         struct ip_vs_dest *dest;
3252         struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3253         struct net *net = sock_net(skb->sk);
3254         struct netns_ipvs *ipvs = net_ipvs(net);
3255
3256         mutex_lock(&__ip_vs_mutex);
3257
3258         /* Try to find the service for which to dump destinations */
3259         if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs, IPVS_CMD_ATTR_MAX,
3260                         ip_vs_cmd_policy, cb->extack))
3261                 goto out_err;
3262
3263
3264         svc = ip_vs_genl_find_service(ipvs, attrs[IPVS_CMD_ATTR_SERVICE]);
3265         if (IS_ERR_OR_NULL(svc))
3266                 goto out_err;
3267
3268         /* Dump the destinations */
3269         list_for_each_entry(dest, &svc->destinations, n_list) {
3270                 if (++idx <= start)
3271                         continue;
3272                 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3273                         idx--;
3274                         goto nla_put_failure;
3275                 }
3276         }
3277
3278 nla_put_failure:
3279         cb->args[0] = idx;
3280
3281 out_err:
3282         mutex_unlock(&__ip_vs_mutex);
3283
3284         return skb->len;
3285 }
3286
3287 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3288                                  struct nlattr *nla, int full_entry)
3289 {
3290         struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3291         struct nlattr *nla_addr, *nla_port;
3292         struct nlattr *nla_addr_family;
3293
3294         /* Parse mandatory identifying destination fields first */
3295         if (nla == NULL ||
3296             nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla,
3297                              ip_vs_dest_policy, NULL))
3298                 return -EINVAL;
3299
3300         nla_addr        = attrs[IPVS_DEST_ATTR_ADDR];
3301         nla_port        = attrs[IPVS_DEST_ATTR_PORT];
3302         nla_addr_family = attrs[IPVS_DEST_ATTR_ADDR_FAMILY];
3303
3304         if (!(nla_addr && nla_port))
3305                 return -EINVAL;
3306
3307         memset(udest, 0, sizeof(*udest));
3308
3309         nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3310         udest->port = nla_get_be16(nla_port);
3311
3312         if (nla_addr_family)
3313                 udest->af = nla_get_u16(nla_addr_family);
3314         else
3315                 udest->af = 0;
3316
3317         /* If a full entry was requested, check for the additional fields */
3318         if (full_entry) {
3319                 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3320                               *nla_l_thresh;
3321
3322                 nla_fwd         = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3323                 nla_weight      = attrs[IPVS_DEST_ATTR_WEIGHT];
3324                 nla_u_thresh    = attrs[IPVS_DEST_ATTR_U_THRESH];
3325                 nla_l_thresh    = attrs[IPVS_DEST_ATTR_L_THRESH];
3326
3327                 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3328                         return -EINVAL;
3329
3330                 udest->conn_flags = nla_get_u32(nla_fwd)
3331                                     & IP_VS_CONN_F_FWD_MASK;
3332                 udest->weight = nla_get_u32(nla_weight);
3333                 udest->u_threshold = nla_get_u32(nla_u_thresh);
3334                 udest->l_threshold = nla_get_u32(nla_l_thresh);
3335         }
3336
3337         return 0;
3338 }
3339
3340 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3341                                   struct ipvs_sync_daemon_cfg *c)
3342 {
3343         struct nlattr *nl_daemon;
3344
3345         nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3346         if (!nl_daemon)
3347                 return -EMSGSIZE;
3348
3349         if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3350             nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, c->mcast_ifn) ||
3351             nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, c->syncid) ||
3352             nla_put_u16(skb, IPVS_DAEMON_ATTR_SYNC_MAXLEN, c->sync_maxlen) ||
3353             nla_put_u16(skb, IPVS_DAEMON_ATTR_MCAST_PORT, c->mcast_port) ||
3354             nla_put_u8(skb, IPVS_DAEMON_ATTR_MCAST_TTL, c->mcast_ttl))
3355                 goto nla_put_failure;
3356 #ifdef CONFIG_IP_VS_IPV6
3357         if (c->mcast_af == AF_INET6) {
3358                 if (nla_put_in6_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP6,
3359                                      &c->mcast_group.in6))
3360                         goto nla_put_failure;
3361         } else
3362 #endif
3363                 if (c->mcast_af == AF_INET &&
3364                     nla_put_in_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP,
3365                                     c->mcast_group.ip))
3366                         goto nla_put_failure;
3367         nla_nest_end(skb, nl_daemon);
3368
3369         return 0;
3370
3371 nla_put_failure:
3372         nla_nest_cancel(skb, nl_daemon);
3373         return -EMSGSIZE;
3374 }
3375
3376 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3377                                   struct ipvs_sync_daemon_cfg *c,
3378                                   struct netlink_callback *cb)
3379 {
3380         void *hdr;
3381         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3382                           &ip_vs_genl_family, NLM_F_MULTI,
3383                           IPVS_CMD_NEW_DAEMON);
3384         if (!hdr)
3385                 return -EMSGSIZE;
3386
3387         if (ip_vs_genl_fill_daemon(skb, state, c))
3388                 goto nla_put_failure;
3389
3390         genlmsg_end(skb, hdr);
3391         return 0;
3392
3393 nla_put_failure:
3394         genlmsg_cancel(skb, hdr);
3395         return -EMSGSIZE;
3396 }
3397
3398 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3399                                    struct netlink_callback *cb)
3400 {
3401         struct net *net = sock_net(skb->sk);
3402         struct netns_ipvs *ipvs = net_ipvs(net);
3403
3404         mutex_lock(&ipvs->sync_mutex);
3405         if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3406                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3407                                            &ipvs->mcfg, cb) < 0)
3408                         goto nla_put_failure;
3409
3410                 cb->args[0] = 1;
3411         }
3412
3413         if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3414                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3415                                            &ipvs->bcfg, cb) < 0)
3416                         goto nla_put_failure;
3417
3418                 cb->args[1] = 1;
3419         }
3420
3421 nla_put_failure:
3422         mutex_unlock(&ipvs->sync_mutex);
3423
3424         return skb->len;
3425 }
3426
3427 static int ip_vs_genl_new_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3428 {
3429         struct ipvs_sync_daemon_cfg c;
3430         struct nlattr *a;
3431         int ret;
3432
3433         memset(&c, 0, sizeof(c));
3434         if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3435               attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3436               attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3437                 return -EINVAL;
3438         strlcpy(c.mcast_ifn, nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3439                 sizeof(c.mcast_ifn));
3440         c.syncid = nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]);
3441
3442         a = attrs[IPVS_DAEMON_ATTR_SYNC_MAXLEN];
3443         if (a)
3444                 c.sync_maxlen = nla_get_u16(a);
3445
3446         a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP];
3447         if (a) {
3448                 c.mcast_af = AF_INET;
3449                 c.mcast_group.ip = nla_get_in_addr(a);
3450                 if (!ipv4_is_multicast(c.mcast_group.ip))
3451                         return -EINVAL;
3452         } else {
3453                 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP6];
3454                 if (a) {
3455 #ifdef CONFIG_IP_VS_IPV6
3456                         int addr_type;
3457
3458                         c.mcast_af = AF_INET6;
3459                         c.mcast_group.in6 = nla_get_in6_addr(a);
3460                         addr_type = ipv6_addr_type(&c.mcast_group.in6);
3461                         if (!(addr_type & IPV6_ADDR_MULTICAST))
3462                                 return -EINVAL;
3463 #else
3464                         return -EAFNOSUPPORT;
3465 #endif
3466                 }
3467         }
3468
3469         a = attrs[IPVS_DAEMON_ATTR_MCAST_PORT];
3470         if (a)
3471                 c.mcast_port = nla_get_u16(a);
3472
3473         a = attrs[IPVS_DAEMON_ATTR_MCAST_TTL];
3474         if (a)
3475                 c.mcast_ttl = nla_get_u8(a);
3476
3477         /* The synchronization protocol is incompatible with mixed family
3478          * services
3479          */
3480         if (ipvs->mixed_address_family_dests > 0)
3481                 return -EINVAL;
3482
3483         ret = start_sync_thread(ipvs, &c,
3484                                 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3485         return ret;
3486 }
3487
3488 static int ip_vs_genl_del_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3489 {
3490         int ret;
3491
3492         if (!attrs[IPVS_DAEMON_ATTR_STATE])
3493                 return -EINVAL;
3494
3495         mutex_lock(&ipvs->sync_mutex);
3496         ret = stop_sync_thread(ipvs,
3497                                nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3498         mutex_unlock(&ipvs->sync_mutex);
3499         return ret;
3500 }
3501
3502 static int ip_vs_genl_set_config(struct netns_ipvs *ipvs, struct nlattr **attrs)
3503 {
3504         struct ip_vs_timeout_user t;
3505
3506         __ip_vs_get_timeouts(ipvs, &t);
3507
3508         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3509                 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3510
3511         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3512                 t.tcp_fin_timeout =
3513                         nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3514
3515         if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3516                 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3517
3518         return ip_vs_set_timeout(ipvs, &t);
3519 }
3520
3521 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3522 {
3523         int ret = -EINVAL, cmd;
3524         struct net *net = sock_net(skb->sk);
3525         struct netns_ipvs *ipvs = net_ipvs(net);
3526
3527         cmd = info->genlhdr->cmd;
3528
3529         if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3530                 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3531
3532                 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3533                     nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3534                                      info->attrs[IPVS_CMD_ATTR_DAEMON],
3535                                      ip_vs_daemon_policy, info->extack))
3536                         goto out;
3537
3538                 if (cmd == IPVS_CMD_NEW_DAEMON)
3539                         ret = ip_vs_genl_new_daemon(ipvs, daemon_attrs);
3540                 else
3541                         ret = ip_vs_genl_del_daemon(ipvs, daemon_attrs);
3542         }
3543
3544 out:
3545         return ret;
3546 }
3547
3548 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3549 {
3550         struct ip_vs_service *svc = NULL;
3551         struct ip_vs_service_user_kern usvc;
3552         struct ip_vs_dest_user_kern udest;
3553         int ret = 0, cmd;
3554         int need_full_svc = 0, need_full_dest = 0;
3555         struct net *net = sock_net(skb->sk);
3556         struct netns_ipvs *ipvs = net_ipvs(net);
3557
3558         cmd = info->genlhdr->cmd;
3559
3560         mutex_lock(&__ip_vs_mutex);
3561
3562         if (cmd == IPVS_CMD_FLUSH) {
3563                 ret = ip_vs_flush(ipvs, false);
3564                 goto out;
3565         } else if (cmd == IPVS_CMD_SET_CONFIG) {
3566                 ret = ip_vs_genl_set_config(ipvs, info->attrs);
3567                 goto out;
3568         } else if (cmd == IPVS_CMD_ZERO &&
3569                    !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3570                 ret = ip_vs_zero_all(ipvs);
3571                 goto out;
3572         }
3573
3574         /* All following commands require a service argument, so check if we
3575          * received a valid one. We need a full service specification when
3576          * adding / editing a service. Only identifying members otherwise. */
3577         if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3578                 need_full_svc = 1;
3579
3580         ret = ip_vs_genl_parse_service(ipvs, &usvc,
3581                                        info->attrs[IPVS_CMD_ATTR_SERVICE],
3582                                        need_full_svc, &svc);
3583         if (ret)
3584                 goto out;
3585
3586         /* Unless we're adding a new service, the service must already exist */
3587         if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3588                 ret = -ESRCH;
3589                 goto out;
3590         }
3591
3592         /* Destination commands require a valid destination argument. For
3593          * adding / editing a destination, we need a full destination
3594          * specification. */
3595         if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3596             cmd == IPVS_CMD_DEL_DEST) {
3597                 if (cmd != IPVS_CMD_DEL_DEST)
3598                         need_full_dest = 1;
3599
3600                 ret = ip_vs_genl_parse_dest(&udest,
3601                                             info->attrs[IPVS_CMD_ATTR_DEST],
3602                                             need_full_dest);
3603                 if (ret)
3604                         goto out;
3605
3606                 /* Old protocols did not allow the user to specify address
3607                  * family, so we set it to zero instead.  We also didn't
3608                  * allow heterogeneous pools in the old code, so it's safe
3609                  * to assume that this will have the same address family as
3610                  * the service.
3611                  */
3612                 if (udest.af == 0)
3613                         udest.af = svc->af;
3614
3615                 if (!ip_vs_is_af_valid(udest.af)) {
3616                         ret = -EAFNOSUPPORT;
3617                         goto out;
3618                 }
3619
3620                 if (udest.af != svc->af && cmd != IPVS_CMD_DEL_DEST) {
3621                         /* The synchronization protocol is incompatible
3622                          * with mixed family services
3623                          */
3624                         if (ipvs->sync_state) {
3625                                 ret = -EINVAL;
3626                                 goto out;
3627                         }
3628
3629                         /* Which connection types do we support? */
3630                         switch (udest.conn_flags) {
3631                         case IP_VS_CONN_F_TUNNEL:
3632                                 /* We are able to forward this */
3633                                 break;
3634                         default:
3635                                 ret = -EINVAL;
3636                                 goto out;
3637                         }
3638                 }
3639         }
3640
3641         switch (cmd) {
3642         case IPVS_CMD_NEW_SERVICE:
3643                 if (svc == NULL)
3644                         ret = ip_vs_add_service(ipvs, &usvc, &svc);
3645                 else
3646                         ret = -EEXIST;
3647                 break;
3648         case IPVS_CMD_SET_SERVICE:
3649                 ret = ip_vs_edit_service(svc, &usvc);
3650                 break;
3651         case IPVS_CMD_DEL_SERVICE:
3652                 ret = ip_vs_del_service(svc);
3653                 /* do not use svc, it can be freed */
3654                 break;
3655         case IPVS_CMD_NEW_DEST:
3656                 ret = ip_vs_add_dest(svc, &udest);
3657                 break;
3658         case IPVS_CMD_SET_DEST:
3659                 ret = ip_vs_edit_dest(svc, &udest);
3660                 break;
3661         case IPVS_CMD_DEL_DEST:
3662                 ret = ip_vs_del_dest(svc, &udest);
3663                 break;
3664         case IPVS_CMD_ZERO:
3665                 ret = ip_vs_zero_service(svc);
3666                 break;
3667         default:
3668                 ret = -EINVAL;
3669         }
3670
3671 out:
3672         mutex_unlock(&__ip_vs_mutex);
3673
3674         return ret;
3675 }
3676
3677 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3678 {
3679         struct sk_buff *msg;
3680         void *reply;
3681         int ret, cmd, reply_cmd;
3682         struct net *net = sock_net(skb->sk);
3683         struct netns_ipvs *ipvs = net_ipvs(net);
3684
3685         cmd = info->genlhdr->cmd;
3686
3687         if (cmd == IPVS_CMD_GET_SERVICE)
3688                 reply_cmd = IPVS_CMD_NEW_SERVICE;
3689         else if (cmd == IPVS_CMD_GET_INFO)
3690                 reply_cmd = IPVS_CMD_SET_INFO;
3691         else if (cmd == IPVS_CMD_GET_CONFIG)
3692                 reply_cmd = IPVS_CMD_SET_CONFIG;
3693         else {
3694                 pr_err("unknown Generic Netlink command\n");
3695                 return -EINVAL;
3696         }
3697
3698         msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3699         if (!msg)
3700                 return -ENOMEM;
3701
3702         mutex_lock(&__ip_vs_mutex);
3703
3704         reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3705         if (reply == NULL)
3706                 goto nla_put_failure;
3707
3708         switch (cmd) {
3709         case IPVS_CMD_GET_SERVICE:
3710         {
3711                 struct ip_vs_service *svc;
3712
3713                 svc = ip_vs_genl_find_service(ipvs,
3714                                               info->attrs[IPVS_CMD_ATTR_SERVICE]);
3715                 if (IS_ERR(svc)) {
3716                         ret = PTR_ERR(svc);
3717                         goto out_err;
3718                 } else if (svc) {
3719                         ret = ip_vs_genl_fill_service(msg, svc);
3720                         if (ret)
3721                                 goto nla_put_failure;
3722                 } else {
3723                         ret = -ESRCH;
3724                         goto out_err;
3725                 }
3726
3727                 break;
3728         }
3729
3730         case IPVS_CMD_GET_CONFIG:
3731         {
3732                 struct ip_vs_timeout_user t;
3733
3734                 __ip_vs_get_timeouts(ipvs, &t);
3735 #ifdef CONFIG_IP_VS_PROTO_TCP
3736                 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3737                                 t.tcp_timeout) ||
3738                     nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3739                                 t.tcp_fin_timeout))
3740                         goto nla_put_failure;
3741 #endif
3742 #ifdef CONFIG_IP_VS_PROTO_UDP
3743                 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3744                         goto nla_put_failure;
3745 #endif
3746
3747                 break;
3748         }
3749
3750         case IPVS_CMD_GET_INFO:
3751                 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3752                                 IP_VS_VERSION_CODE) ||
3753                     nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3754                                 ip_vs_conn_tab_size))
3755                         goto nla_put_failure;
3756                 break;
3757         }
3758
3759         genlmsg_end(msg, reply);
3760         ret = genlmsg_reply(msg, info);
3761         goto out;
3762
3763 nla_put_failure:
3764         pr_err("not enough space in Netlink message\n");
3765         ret = -EMSGSIZE;
3766
3767 out_err:
3768         nlmsg_free(msg);
3769 out:
3770         mutex_unlock(&__ip_vs_mutex);
3771
3772         return ret;
3773 }
3774
3775
3776 static const struct genl_ops ip_vs_genl_ops[] = {
3777         {
3778                 .cmd    = IPVS_CMD_NEW_SERVICE,
3779                 .flags  = GENL_ADMIN_PERM,
3780                 .policy = ip_vs_cmd_policy,
3781                 .doit   = ip_vs_genl_set_cmd,
3782         },
3783         {
3784                 .cmd    = IPVS_CMD_SET_SERVICE,
3785                 .flags  = GENL_ADMIN_PERM,
3786                 .policy = ip_vs_cmd_policy,
3787                 .doit   = ip_vs_genl_set_cmd,
3788         },
3789         {
3790                 .cmd    = IPVS_CMD_DEL_SERVICE,
3791                 .flags  = GENL_ADMIN_PERM,
3792                 .policy = ip_vs_cmd_policy,
3793                 .doit   = ip_vs_genl_set_cmd,
3794         },
3795         {
3796                 .cmd    = IPVS_CMD_GET_SERVICE,
3797                 .flags  = GENL_ADMIN_PERM,
3798                 .doit   = ip_vs_genl_get_cmd,
3799                 .dumpit = ip_vs_genl_dump_services,
3800                 .policy = ip_vs_cmd_policy,
3801         },
3802         {
3803                 .cmd    = IPVS_CMD_NEW_DEST,
3804                 .flags  = GENL_ADMIN_PERM,
3805                 .policy = ip_vs_cmd_policy,
3806                 .doit   = ip_vs_genl_set_cmd,
3807         },
3808         {
3809                 .cmd    = IPVS_CMD_SET_DEST,
3810                 .flags  = GENL_ADMIN_PERM,
3811                 .policy = ip_vs_cmd_policy,
3812                 .doit   = ip_vs_genl_set_cmd,
3813         },
3814         {
3815                 .cmd    = IPVS_CMD_DEL_DEST,
3816                 .flags  = GENL_ADMIN_PERM,
3817                 .policy = ip_vs_cmd_policy,
3818                 .doit   = ip_vs_genl_set_cmd,
3819         },
3820         {
3821                 .cmd    = IPVS_CMD_GET_DEST,
3822                 .flags  = GENL_ADMIN_PERM,
3823                 .policy = ip_vs_cmd_policy,
3824                 .dumpit = ip_vs_genl_dump_dests,
3825         },
3826         {
3827                 .cmd    = IPVS_CMD_NEW_DAEMON,
3828                 .flags  = GENL_ADMIN_PERM,
3829                 .policy = ip_vs_cmd_policy,
3830                 .doit   = ip_vs_genl_set_daemon,
3831         },
3832         {
3833                 .cmd    = IPVS_CMD_DEL_DAEMON,
3834                 .flags  = GENL_ADMIN_PERM,
3835                 .policy = ip_vs_cmd_policy,
3836                 .doit   = ip_vs_genl_set_daemon,
3837         },
3838         {
3839                 .cmd    = IPVS_CMD_GET_DAEMON,
3840                 .flags  = GENL_ADMIN_PERM,
3841                 .dumpit = ip_vs_genl_dump_daemons,
3842         },
3843         {
3844                 .cmd    = IPVS_CMD_SET_CONFIG,
3845                 .flags  = GENL_ADMIN_PERM,
3846                 .policy = ip_vs_cmd_policy,
3847                 .doit   = ip_vs_genl_set_cmd,
3848         },
3849         {
3850                 .cmd    = IPVS_CMD_GET_CONFIG,
3851                 .flags  = GENL_ADMIN_PERM,
3852                 .doit   = ip_vs_genl_get_cmd,
3853         },
3854         {
3855                 .cmd    = IPVS_CMD_GET_INFO,
3856                 .flags  = GENL_ADMIN_PERM,
3857                 .doit   = ip_vs_genl_get_cmd,
3858         },
3859         {
3860                 .cmd    = IPVS_CMD_ZERO,
3861                 .flags  = GENL_ADMIN_PERM,
3862                 .policy = ip_vs_cmd_policy,
3863                 .doit   = ip_vs_genl_set_cmd,
3864         },
3865         {
3866                 .cmd    = IPVS_CMD_FLUSH,
3867                 .flags  = GENL_ADMIN_PERM,
3868                 .doit   = ip_vs_genl_set_cmd,
3869         },
3870 };
3871
3872 static struct genl_family ip_vs_genl_family __ro_after_init = {
3873         .hdrsize        = 0,
3874         .name           = IPVS_GENL_NAME,
3875         .version        = IPVS_GENL_VERSION,
3876         .maxattr        = IPVS_CMD_ATTR_MAX,
3877         .netnsok        = true,         /* Make ipvsadm to work on netns */
3878         .module         = THIS_MODULE,
3879         .ops            = ip_vs_genl_ops,
3880         .n_ops          = ARRAY_SIZE(ip_vs_genl_ops),
3881 };
3882
3883 static int __init ip_vs_genl_register(void)
3884 {
3885         return genl_register_family(&ip_vs_genl_family);
3886 }
3887
3888 static void ip_vs_genl_unregister(void)
3889 {
3890         genl_unregister_family(&ip_vs_genl_family);
3891 }
3892
3893 /* End of Generic Netlink interface definitions */
3894
3895 /*
3896  * per netns intit/exit func.
3897  */
3898 #ifdef CONFIG_SYSCTL
3899 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs)
3900 {
3901         struct net *net = ipvs->net;
3902         int idx;
3903         struct ctl_table *tbl;
3904
3905         atomic_set(&ipvs->dropentry, 0);
3906         spin_lock_init(&ipvs->dropentry_lock);
3907         spin_lock_init(&ipvs->droppacket_lock);
3908         spin_lock_init(&ipvs->securetcp_lock);
3909
3910         if (!net_eq(net, &init_net)) {
3911                 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3912                 if (tbl == NULL)
3913                         return -ENOMEM;
3914
3915                 /* Don't export sysctls to unprivileged users */
3916                 if (net->user_ns != &init_user_ns)
3917                         tbl[0].procname = NULL;
3918         } else
3919                 tbl = vs_vars;
3920         /* Initialize sysctl defaults */
3921         for (idx = 0; idx < ARRAY_SIZE(vs_vars); idx++) {
3922                 if (tbl[idx].proc_handler == proc_do_defense_mode)
3923                         tbl[idx].extra2 = ipvs;
3924         }
3925         idx = 0;
3926         ipvs->sysctl_amemthresh = 1024;
3927         tbl[idx++].data = &ipvs->sysctl_amemthresh;
3928         ipvs->sysctl_am_droprate = 10;
3929         tbl[idx++].data = &ipvs->sysctl_am_droprate;
3930         tbl[idx++].data = &ipvs->sysctl_drop_entry;
3931         tbl[idx++].data = &ipvs->sysctl_drop_packet;
3932 #ifdef CONFIG_IP_VS_NFCT
3933         tbl[idx++].data = &ipvs->sysctl_conntrack;
3934 #endif
3935         tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3936         ipvs->sysctl_snat_reroute = 1;
3937         tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3938         ipvs->sysctl_sync_ver = 1;
3939         tbl[idx++].data = &ipvs->sysctl_sync_ver;
3940         ipvs->sysctl_sync_ports = 1;
3941         tbl[idx++].data = &ipvs->sysctl_sync_ports;
3942         tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
3943         ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3944         tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3945         ipvs->sysctl_sync_sock_size = 0;
3946         tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3947         tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3948         tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3949         tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
3950         tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
3951         tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3952         ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3953         ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3954         tbl[idx].data = &ipvs->sysctl_sync_threshold;
3955         tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3956         ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3957         tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3958         ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3959         tbl[idx++].data = &ipvs->sysctl_sync_retries;
3960         tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3961         ipvs->sysctl_pmtu_disc = 1;
3962         tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
3963         tbl[idx++].data = &ipvs->sysctl_backup_only;
3964         ipvs->sysctl_conn_reuse_mode = 1;
3965         tbl[idx++].data = &ipvs->sysctl_conn_reuse_mode;
3966         tbl[idx++].data = &ipvs->sysctl_schedule_icmp;
3967         tbl[idx++].data = &ipvs->sysctl_ignore_tunneled;
3968
3969         ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3970         if (ipvs->sysctl_hdr == NULL) {
3971                 if (!net_eq(net, &init_net))
3972                         kfree(tbl);
3973                 return -ENOMEM;
3974         }
3975         ip_vs_start_estimator(ipvs, &ipvs->tot_stats);
3976         ipvs->sysctl_tbl = tbl;
3977         /* Schedule defense work */
3978         INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3979         schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3980
3981         return 0;
3982 }
3983
3984 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs)
3985 {
3986         struct net *net = ipvs->net;
3987
3988         cancel_delayed_work_sync(&ipvs->defense_work);
3989         cancel_work_sync(&ipvs->defense_work.work);
3990         unregister_net_sysctl_table(ipvs->sysctl_hdr);
3991         ip_vs_stop_estimator(ipvs, &ipvs->tot_stats);
3992
3993         if (!net_eq(net, &init_net))
3994                 kfree(ipvs->sysctl_tbl);
3995 }
3996
3997 #else
3998
3999 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs) { return 0; }
4000 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs) { }
4001
4002 #endif
4003
4004 static struct notifier_block ip_vs_dst_notifier = {
4005         .notifier_call = ip_vs_dst_event,
4006 #ifdef CONFIG_IP_VS_IPV6
4007         .priority = ADDRCONF_NOTIFY_PRIORITY + 5,
4008 #endif
4009 };
4010
4011 int __net_init ip_vs_control_net_init(struct netns_ipvs *ipvs)
4012 {
4013         int i, idx;
4014
4015         /* Initialize rs_table */
4016         for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
4017                 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
4018
4019         INIT_LIST_HEAD(&ipvs->dest_trash);
4020         spin_lock_init(&ipvs->dest_trash_lock);
4021         timer_setup(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire, 0);
4022         atomic_set(&ipvs->ftpsvc_counter, 0);
4023         atomic_set(&ipvs->nullsvc_counter, 0);
4024         atomic_set(&ipvs->conn_out_counter, 0);
4025
4026         /* procfs stats */
4027         ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
4028         if (!ipvs->tot_stats.cpustats)
4029                 return -ENOMEM;
4030
4031         for_each_possible_cpu(i) {
4032                 struct ip_vs_cpu_stats *ipvs_tot_stats;
4033                 ipvs_tot_stats = per_cpu_ptr(ipvs->tot_stats.cpustats, i);
4034                 u64_stats_init(&ipvs_tot_stats->syncp);
4035         }
4036
4037         spin_lock_init(&ipvs->tot_stats.lock);
4038
4039         proc_create_net("ip_vs", 0, ipvs->net->proc_net, &ip_vs_info_seq_ops,
4040                         sizeof(struct ip_vs_iter));
4041         proc_create_net_single("ip_vs_stats", 0, ipvs->net->proc_net,
4042                         ip_vs_stats_show, NULL);
4043         proc_create_net_single("ip_vs_stats_percpu", 0, ipvs->net->proc_net,
4044                         ip_vs_stats_percpu_show, NULL);
4045
4046         if (ip_vs_control_net_init_sysctl(ipvs))
4047                 goto err;
4048
4049         return 0;
4050
4051 err:
4052         free_percpu(ipvs->tot_stats.cpustats);
4053         return -ENOMEM;
4054 }
4055
4056 void __net_exit ip_vs_control_net_cleanup(struct netns_ipvs *ipvs)
4057 {
4058         ip_vs_trash_cleanup(ipvs);
4059         ip_vs_control_net_cleanup_sysctl(ipvs);
4060         remove_proc_entry("ip_vs_stats_percpu", ipvs->net->proc_net);
4061         remove_proc_entry("ip_vs_stats", ipvs->net->proc_net);
4062         remove_proc_entry("ip_vs", ipvs->net->proc_net);
4063         free_percpu(ipvs->tot_stats.cpustats);
4064 }
4065
4066 int __init ip_vs_register_nl_ioctl(void)
4067 {
4068         int ret;
4069
4070         ret = nf_register_sockopt(&ip_vs_sockopts);
4071         if (ret) {
4072                 pr_err("cannot register sockopt.\n");
4073                 goto err_sock;
4074         }
4075
4076         ret = ip_vs_genl_register();
4077         if (ret) {
4078                 pr_err("cannot register Generic Netlink interface.\n");
4079                 goto err_genl;
4080         }
4081         return 0;
4082
4083 err_genl:
4084         nf_unregister_sockopt(&ip_vs_sockopts);
4085 err_sock:
4086         return ret;
4087 }
4088
4089 void ip_vs_unregister_nl_ioctl(void)
4090 {
4091         ip_vs_genl_unregister();
4092         nf_unregister_sockopt(&ip_vs_sockopts);
4093 }
4094
4095 int __init ip_vs_control_init(void)
4096 {
4097         int idx;
4098         int ret;
4099
4100         EnterFunction(2);
4101
4102         /* Initialize svc_table, ip_vs_svc_fwm_table */
4103         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
4104                 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
4105                 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
4106         }
4107
4108         smp_wmb();      /* Do we really need it now ? */
4109
4110         ret = register_netdevice_notifier(&ip_vs_dst_notifier);
4111         if (ret < 0)
4112                 return ret;
4113
4114         LeaveFunction(2);
4115         return 0;
4116 }
4117
4118
4119 void ip_vs_control_cleanup(void)
4120 {
4121         EnterFunction(2);
4122         unregister_netdevice_notifier(&ip_vs_dst_notifier);
4123         LeaveFunction(2);
4124 }