netfilter/ipvs/ip_vs_ctl.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  * IPVS         An implementation of the IP virtual server support for the
   4  *              LINUX operating system.  IPVS is now implemented as a module
   5  *              over the NetFilter framework. IPVS can be used to build a
   6  *              high-performance and highly available server based on a
   7  *              cluster of servers.
   8  *
   9  * Authors:     Wensong Zhang <wensong@linuxvirtualserver.org>
  10  *              Peter Kese <peter.kese@ijs.si>
  11  *              Julian Anastasov <ja@ssi.bg>
  12  *
  13  * Changes:
  14  */
  15
  16 #define KMSG_COMPONENT "IPVS"
  17 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
  18
  19 #include <linux/module.h>
  20 #include <linux/init.h>
  21 #include <linux/types.h>
  22 #include <linux/capability.h>
  23 #include <linux/fs.h>
  24 #include <linux/sysctl.h>
  25 #include <linux/proc_fs.h>
  26 #include <linux/workqueue.h>
  27 #include <linux/seq_file.h>
  28 #include <linux/slab.h>
  29
  30 #include <linux/netfilter.h>
  31 #include <linux/netfilter_ipv4.h>
  32 #include <linux/mutex.h>
  33
  34 #include <net/net_namespace.h>
  35 #include <linux/nsproxy.h>
  36 #include <net/ip.h>
  37 #ifdef CONFIG_IP_VS_IPV6
  38 #include <net/ipv6.h>
  39 #include <net/ip6_route.h>
  40 #include <net/netfilter/ipv6/nf_defrag_ipv6.h>
  41 #endif
  42 #include <net/route.h>
  43 #include <net/sock.h>
  44 #include <net/genetlink.h>
  45
  46 #include <linux/uaccess.h>
  47
  48 #include <net/ip_vs.h>
  49
  50 MODULE_ALIAS_GENL_FAMILY(IPVS_GENL_NAME);
  51
  52 /* semaphore for IPVS sockopts. And, [gs]etsockopt may sleep. */
  53 static DEFINE_MUTEX(__ip_vs_mutex);
  54
  55 /* sysctl variables */
  56
  57 #ifdef CONFIG_IP_VS_DEBUG
  58 static int sysctl_ip_vs_debug_level = 0;
  59
  60 int ip_vs_get_debug_level(void)
  61 {
  62         return sysctl_ip_vs_debug_level;
  63 }
  64 #endif
  65
  66
  67 /*  Protos */
  68 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup);
  69
  70
  71 #ifdef CONFIG_IP_VS_IPV6
  72 /* Taken from rt6_fill_node() in net/ipv6/route.c, is there a better way? */
  73 static bool __ip_vs_addr_is_local_v6(struct net *net,
  74                                      const struct in6_addr *addr)
  75 {
  76         struct flowi6 fl6 = {
  77                 .daddr = *addr,
  78         };
  79         struct dst_entry *dst = ip6_route_output(net, NULL, &fl6);
  80         bool is_local;
  81
  82         is_local = !dst->error && dst->dev && (dst->dev->flags & IFF_LOOPBACK);
  83
  84         dst_release(dst);
  85         return is_local;
  86 }
  87 #endif
  88
  89 #ifdef CONFIG_SYSCTL
  90 /*
  91  *      update_defense_level is called from keventd and from sysctl,
  92  *      so it needs to protect itself from softirqs
  93  */
  94 static void update_defense_level(struct netns_ipvs *ipvs)
  95 {
  96         struct sysinfo i;
  97         int availmem;
  98         int nomem;
  99         int to_change = -1;
 100
 101         /* we only count free and buffered memory (in pages) */
 102         si_meminfo(&i);
 103         availmem = i.freeram + i.bufferram;
 104         /* however in linux 2.5 the i.bufferram is total page cache size,
 105            we need adjust it */
 106         /* si_swapinfo(&i); */
 107         /* availmem = availmem - (i.totalswap - i.freeswap); */
 108
 109         nomem = (availmem < ipvs->sysctl_amemthresh);
 110
 111         local_bh_disable();
 112
 113         /* drop_entry */
 114         spin_lock(&ipvs->dropentry_lock);
 115         switch (ipvs->sysctl_drop_entry) {
 116         case 0:
 117                 atomic_set(&ipvs->dropentry, 0);
 118                 break;
 119         case 1:
 120                 if (nomem) {
 121                         atomic_set(&ipvs->dropentry, 1);
 122                         ipvs->sysctl_drop_entry = 2;
 123                 } else {
 124                         atomic_set(&ipvs->dropentry, 0);
 125                 }
 126                 break;
 127         case 2:
 128                 if (nomem) {
 129                         atomic_set(&ipvs->dropentry, 1);
 130                 } else {
 131                         atomic_set(&ipvs->dropentry, 0);
 132                         ipvs->sysctl_drop_entry = 1;
 133                 }
 134                 break;
 135         case 3:
 136                 atomic_set(&ipvs->dropentry, 1);
 137                 break;
 138         }
 139         spin_unlock(&ipvs->dropentry_lock);
 140
 141         /* drop_packet */
 142         spin_lock(&ipvs->droppacket_lock);
 143         switch (ipvs->sysctl_drop_packet) {
 144         case 0:
 145                 ipvs->drop_rate = 0;
 146                 break;
 147         case 1:
 148                 if (nomem) {
 149                         ipvs->drop_rate = ipvs->drop_counter
 150                                 = ipvs->sysctl_amemthresh /
 151                                 (ipvs->sysctl_amemthresh-availmem);
 152                         ipvs->sysctl_drop_packet = 2;
 153                 } else {
 154                         ipvs->drop_rate = 0;
 155                 }
 156                 break;
 157         case 2:
 158                 if (nomem) {
 159                         ipvs->drop_rate = ipvs->drop_counter
 160                                 = ipvs->sysctl_amemthresh /
 161                                 (ipvs->sysctl_amemthresh-availmem);
 162                 } else {
 163                         ipvs->drop_rate = 0;
 164                         ipvs->sysctl_drop_packet = 1;
 165                 }
 166                 break;
 167         case 3:
 168                 ipvs->drop_rate = ipvs->sysctl_am_droprate;
 169                 break;
 170         }
 171         spin_unlock(&ipvs->droppacket_lock);
 172
 173         /* secure_tcp */
 174         spin_lock(&ipvs->securetcp_lock);
 175         switch (ipvs->sysctl_secure_tcp) {
 176         case 0:
 177                 if (ipvs->old_secure_tcp >= 2)
 178                         to_change = 0;
 179                 break;
 180         case 1:
 181                 if (nomem) {
 182                         if (ipvs->old_secure_tcp < 2)
 183                                 to_change = 1;
 184                         ipvs->sysctl_secure_tcp = 2;
 185                 } else {
 186                         if (ipvs->old_secure_tcp >= 2)
 187                                 to_change = 0;
 188                 }
 189                 break;
 190         case 2:
 191                 if (nomem) {
 192                         if (ipvs->old_secure_tcp < 2)
 193                                 to_change = 1;
 194                 } else {
 195                         if (ipvs->old_secure_tcp >= 2)
 196                                 to_change = 0;
 197                         ipvs->sysctl_secure_tcp = 1;
 198                 }
 199                 break;
 200         case 3:
 201                 if (ipvs->old_secure_tcp < 2)
 202                         to_change = 1;
 203                 break;
 204         }
 205         ipvs->old_secure_tcp = ipvs->sysctl_secure_tcp;
 206         if (to_change >= 0)
 207                 ip_vs_protocol_timeout_change(ipvs,
 208                                               ipvs->sysctl_secure_tcp > 1);
 209         spin_unlock(&ipvs->securetcp_lock);
 210
 211         local_bh_enable();
 212 }
 213
 214 /* Handler for delayed work for expiring no
 215  * destination connections
 216  */
 217 static void expire_nodest_conn_handler(struct work_struct *work)
 218 {
 219         struct netns_ipvs *ipvs;
 220
 221         ipvs = container_of(work, struct netns_ipvs,
 222                             expire_nodest_conn_work.work);
 223         ip_vs_expire_nodest_conn_flush(ipvs);
 224 }
 225
 226 /*
 227  *      Timer for checking the defense
 228  */
 229 #define DEFENSE_TIMER_PERIOD    1*HZ
 230
 231 static void defense_work_handler(struct work_struct *work)
 232 {
 233         struct netns_ipvs *ipvs =
 234                 container_of(work, struct netns_ipvs, defense_work.work);
 235
 236         update_defense_level(ipvs);
 237         if (atomic_read(&ipvs->dropentry))
 238                 ip_vs_random_dropentry(ipvs);
 239         queue_delayed_work(system_long_wq, &ipvs->defense_work,
 240                            DEFENSE_TIMER_PERIOD);
 241 }
 242 #endif
 243
 244 int
 245 ip_vs_use_count_inc(void)
 246 {
 247         return try_module_get(THIS_MODULE);
 248 }
 249
 250 void
 251 ip_vs_use_count_dec(void)
 252 {
 253         module_put(THIS_MODULE);
 254 }
 255
 256
 257 /*
 258  *      Hash table: for virtual service lookups
 259  */
 260 #define IP_VS_SVC_TAB_BITS 8
 261 #define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
 262 #define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)
 263
 264 /* the service table hashed by <protocol, addr, port> */
 265 static struct hlist_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE];
 266 /* the service table hashed by fwmark */
 267 static struct hlist_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE];
 268
 269
 270 /*
 271  *      Returns hash value for virtual service
 272  */
 273 static inline unsigned int
 274 ip_vs_svc_hashkey(struct netns_ipvs *ipvs, int af, unsigned int proto,
 275                   const union nf_inet_addr *addr, __be16 port)
 276 {
 277         unsigned int porth = ntohs(port);
 278         __be32 addr_fold = addr->ip;
 279         __u32 ahash;
 280
 281 #ifdef CONFIG_IP_VS_IPV6
 282         if (af == AF_INET6)
 283                 addr_fold = addr->ip6[0]^addr->ip6[1]^
 284                             addr->ip6[2]^addr->ip6[3];
 285 #endif
 286         ahash = ntohl(addr_fold);
 287         ahash ^= ((size_t) ipvs >> 8);
 288
 289         return (proto ^ ahash ^ (porth >> IP_VS_SVC_TAB_BITS) ^ porth) &
 290                IP_VS_SVC_TAB_MASK;
 291 }
 292
 293 /*
 294  *      Returns hash value of fwmark for virtual service lookup
 295  */
 296 static inline unsigned int ip_vs_svc_fwm_hashkey(struct netns_ipvs *ipvs, __u32 fwmark)
 297 {
 298         return (((size_t)ipvs>>8) ^ fwmark) & IP_VS_SVC_TAB_MASK;
 299 }
 300
 301 /*
 302  *      Hashes a service in the ip_vs_svc_table by <netns,proto,addr,port>
 303  *      or in the ip_vs_svc_fwm_table by fwmark.
 304  *      Should be called with locked tables.
 305  */
 306 static int ip_vs_svc_hash(struct ip_vs_service *svc)
 307 {
 308         unsigned int hash;
 309
 310         if (svc->flags & IP_VS_SVC_F_HASHED) {
 311                 pr_err("%s(): request for already hashed, called from %pS\n",
 312                        __func__, __builtin_return_address(0));
 313                 return 0;
 314         }
 315
 316         if (svc->fwmark == 0) {
 317                 /*
 318                  *  Hash it by <netns,protocol,addr,port> in ip_vs_svc_table
 319                  */
 320                 hash = ip_vs_svc_hashkey(svc->ipvs, svc->af, svc->protocol,
 321                                          &svc->addr, svc->port);
 322                 hlist_add_head_rcu(&svc->s_list, &ip_vs_svc_table[hash]);
 323         } else {
 324                 /*
 325                  *  Hash it by fwmark in svc_fwm_table
 326                  */
 327                 hash = ip_vs_svc_fwm_hashkey(svc->ipvs, svc->fwmark);
 328                 hlist_add_head_rcu(&svc->f_list, &ip_vs_svc_fwm_table[hash]);
 329         }
 330
 331         svc->flags |= IP_VS_SVC_F_HASHED;
 332         /* increase its refcnt because it is referenced by the svc table */
 333         atomic_inc(&svc->refcnt);
 334         return 1;
 335 }
 336
 337
 338 /*
 339  *      Unhashes a service from svc_table / svc_fwm_table.
 340  *      Should be called with locked tables.
 341  */
 342 static int ip_vs_svc_unhash(struct ip_vs_service *svc)
 343 {
 344         if (!(svc->flags & IP_VS_SVC_F_HASHED)) {
 345                 pr_err("%s(): request for unhash flagged, called from %pS\n",
 346                        __func__, __builtin_return_address(0));
 347                 return 0;
 348         }
 349
 350         if (svc->fwmark == 0) {
 351                 /* Remove it from the svc_table table */
 352                 hlist_del_rcu(&svc->s_list);
 353         } else {
 354                 /* Remove it from the svc_fwm_table table */
 355                 hlist_del_rcu(&svc->f_list);
 356         }
 357
 358         svc->flags &= ~IP_VS_SVC_F_HASHED;
 359         atomic_dec(&svc->refcnt);
 360         return 1;
 361 }
 362
 363
 364 /*
 365  *      Get service by {netns, proto,addr,port} in the service table.
 366  */
 367 static inline struct ip_vs_service *
 368 __ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u16 protocol,
 369                      const union nf_inet_addr *vaddr, __be16 vport)
 370 {
 371         unsigned int hash;
 372         struct ip_vs_service *svc;
 373
 374         /* Check for "full" addressed entries */
 375         hash = ip_vs_svc_hashkey(ipvs, af, protocol, vaddr, vport);
 376
 377         hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[hash], s_list) {
 378                 if ((svc->af == af)
 379                     && ip_vs_addr_equal(af, &svc->addr, vaddr)
 380                     && (svc->port == vport)
 381                     && (svc->protocol == protocol)
 382                     && (svc->ipvs == ipvs)) {
 383                         /* HIT */
 384                         return svc;
 385                 }
 386         }
 387
 388         return NULL;
 389 }
 390
 391
 392 /*
 393  *      Get service by {fwmark} in the service table.
 394  */
 395 static inline struct ip_vs_service *
 396 __ip_vs_svc_fwm_find(struct netns_ipvs *ipvs, int af, __u32 fwmark)
 397 {
 398         unsigned int hash;
 399         struct ip_vs_service *svc;
 400
 401         /* Check for fwmark addressed entries */
 402         hash = ip_vs_svc_fwm_hashkey(ipvs, fwmark);
 403
 404         hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[hash], f_list) {
 405                 if (svc->fwmark == fwmark && svc->af == af
 406                     && (svc->ipvs == ipvs)) {
 407                         /* HIT */
 408                         return svc;
 409                 }
 410         }
 411
 412         return NULL;
 413 }
 414
 415 /* Find service, called under RCU lock */
 416 struct ip_vs_service *
 417 ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
 418                    const union nf_inet_addr *vaddr, __be16 vport)
 419 {
 420         struct ip_vs_service *svc;
 421
 422         /*
 423          *      Check the table hashed by fwmark first
 424          */
 425         if (fwmark) {
 426                 svc = __ip_vs_svc_fwm_find(ipvs, af, fwmark);
 427                 if (svc)
 428                         goto out;
 429         }
 430
 431         /*
 432          *      Check the table hashed by <protocol,addr,port>
 433          *      for "full" addressed entries
 434          */
 435         svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, vport);
 436
 437         if (!svc && protocol == IPPROTO_TCP &&
 438             atomic_read(&ipvs->ftpsvc_counter) &&
 439             (vport == FTPDATA || !inet_port_requires_bind_service(ipvs->net, ntohs(vport)))) {
 440                 /*
 441                  * Check if ftp service entry exists, the packet
 442                  * might belong to FTP data connections.
 443                  */
 444                 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, FTPPORT);
 445         }
 446
 447         if (svc == NULL
 448             && atomic_read(&ipvs->nullsvc_counter)) {
 449                 /*
 450                  * Check if the catch-all port (port zero) exists
 451                  */
 452                 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, 0);
 453         }
 454
 455   out:
 456         IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
 457                       fwmark, ip_vs_proto_name(protocol),
 458                       IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
 459                       svc ? "hit" : "not hit");
 460
 461         return svc;
 462 }
 463
 464
 465 static inline void
 466 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
 467 {
 468         atomic_inc(&svc->refcnt);
 469         rcu_assign_pointer(dest->svc, svc);
 470 }
 471
 472 static void ip_vs_service_free(struct ip_vs_service *svc)
 473 {
 474         free_percpu(svc->stats.cpustats);
 475         kfree(svc);
 476 }
 477
 478 static void ip_vs_service_rcu_free(struct rcu_head *head)
 479 {
 480         struct ip_vs_service *svc;
 481
 482         svc = container_of(head, struct ip_vs_service, rcu_head);
 483         ip_vs_service_free(svc);
 484 }
 485
 486 static void __ip_vs_svc_put(struct ip_vs_service *svc, bool do_delay)
 487 {
 488         if (atomic_dec_and_test(&svc->refcnt)) {
 489                 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n",
 490                               svc->fwmark,
 491                               IP_VS_DBG_ADDR(svc->af, &svc->addr),
 492                               ntohs(svc->port));
 493                 if (do_delay)
 494                         call_rcu(&svc->rcu_head, ip_vs_service_rcu_free);
 495                 else
 496                         ip_vs_service_free(svc);
 497         }
 498 }
 499
 500
 501 /*
 502  *      Returns hash value for real service
 503  */
 504 static inline unsigned int ip_vs_rs_hashkey(int af,
 505                                             const union nf_inet_addr *addr,
 506                                             __be16 port)
 507 {
 508         unsigned int porth = ntohs(port);
 509         __be32 addr_fold = addr->ip;
 510
 511 #ifdef CONFIG_IP_VS_IPV6
 512         if (af == AF_INET6)
 513                 addr_fold = addr->ip6[0]^addr->ip6[1]^
 514                             addr->ip6[2]^addr->ip6[3];
 515 #endif
 516
 517         return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
 518                 & IP_VS_RTAB_MASK;
 519 }
 520
 521 /* Hash ip_vs_dest in rs_table by <proto,addr,port>. */
 522 static void ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
 523 {
 524         unsigned int hash;
 525         __be16 port;
 526
 527         if (dest->in_rs_table)
 528                 return;
 529
 530         switch (IP_VS_DFWD_METHOD(dest)) {
 531         case IP_VS_CONN_F_MASQ:
 532                 port = dest->port;
 533                 break;
 534         case IP_VS_CONN_F_TUNNEL:
 535                 switch (dest->tun_type) {
 536                 case IP_VS_CONN_F_TUNNEL_TYPE_GUE:
 537                         port = dest->tun_port;
 538                         break;
 539                 case IP_VS_CONN_F_TUNNEL_TYPE_IPIP:
 540                 case IP_VS_CONN_F_TUNNEL_TYPE_GRE:
 541                         port = 0;
 542                         break;
 543                 default:
 544                         return;
 545                 }
 546                 break;
 547         default:
 548                 return;
 549         }
 550
 551         /*
 552          *      Hash by proto,addr,port,
 553          *      which are the parameters of the real service.
 554          */
 555         hash = ip_vs_rs_hashkey(dest->af, &dest->addr, port);
 556
 557         hlist_add_head_rcu(&dest->d_list, &ipvs->rs_table[hash]);
 558         dest->in_rs_table = 1;
 559 }
 560
 561 /* Unhash ip_vs_dest from rs_table. */
 562 static void ip_vs_rs_unhash(struct ip_vs_dest *dest)
 563 {
 564         /*
 565          * Remove it from the rs_table table.
 566          */
 567         if (dest->in_rs_table) {
 568                 hlist_del_rcu(&dest->d_list);
 569                 dest->in_rs_table = 0;
 570         }
 571 }
 572
 573 /* Check if real service by <proto,addr,port> is present */
 574 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
 575                             const union nf_inet_addr *daddr, __be16 dport)
 576 {
 577         unsigned int hash;
 578         struct ip_vs_dest *dest;
 579
 580         /* Check for "full" addressed entries */
 581         hash = ip_vs_rs_hashkey(af, daddr, dport);
 582
 583         hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
 584                 if (dest->port == dport &&
 585                     dest->af == af &&
 586                     ip_vs_addr_equal(af, &dest->addr, daddr) &&
 587                     (dest->protocol == protocol || dest->vfwmark) &&
 588                     IP_VS_DFWD_METHOD(dest) == IP_VS_CONN_F_MASQ) {
 589                         /* HIT */
 590                         return true;
 591                 }
 592         }
 593
 594         return false;
 595 }
 596
 597 /* Find real service record by <proto,addr,port>.
 598  * In case of multiple records with the same <proto,addr,port>, only
 599  * the first found record is returned.
 600  *
 601  * To be called under RCU lock.
 602  */
 603 struct ip_vs_dest *ip_vs_find_real_service(struct netns_ipvs *ipvs, int af,
 604                                            __u16 protocol,
 605                                            const union nf_inet_addr *daddr,
 606                                            __be16 dport)
 607 {
 608         unsigned int hash;
 609         struct ip_vs_dest *dest;
 610
 611         /* Check for "full" addressed entries */
 612         hash = ip_vs_rs_hashkey(af, daddr, dport);
 613
 614         hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
 615                 if (dest->port == dport &&
 616                     dest->af == af &&
 617                     ip_vs_addr_equal(af, &dest->addr, daddr) &&
 618                     (dest->protocol == protocol || dest->vfwmark) &&
 619                     IP_VS_DFWD_METHOD(dest) == IP_VS_CONN_F_MASQ) {
 620                         /* HIT */
 621                         return dest;
 622                 }
 623         }
 624
 625         return NULL;
 626 }
 627
 628 /* Find real service record by <af,addr,tun_port>.
 629  * In case of multiple records with the same <af,addr,tun_port>, only
 630  * the first found record is returned.
 631  *
 632  * To be called under RCU lock.
 633  */
 634 struct ip_vs_dest *ip_vs_find_tunnel(struct netns_ipvs *ipvs, int af,
 635                                      const union nf_inet_addr *daddr,
 636                                      __be16 tun_port)
 637 {
 638         struct ip_vs_dest *dest;
 639         unsigned int hash;
 640
 641         /* Check for "full" addressed entries */
 642         hash = ip_vs_rs_hashkey(af, daddr, tun_port);
 643
 644         hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
 645                 if (dest->tun_port == tun_port &&
 646                     dest->af == af &&
 647                     ip_vs_addr_equal(af, &dest->addr, daddr) &&
 648                     IP_VS_DFWD_METHOD(dest) == IP_VS_CONN_F_TUNNEL) {
 649                         /* HIT */
 650                         return dest;
 651                 }
 652         }
 653
 654         return NULL;
 655 }
 656
 657 /* Lookup destination by {addr,port} in the given service
 658  * Called under RCU lock.
 659  */
 660 static struct ip_vs_dest *
 661 ip_vs_lookup_dest(struct ip_vs_service *svc, int dest_af,
 662                   const union nf_inet_addr *daddr, __be16 dport)
 663 {
 664         struct ip_vs_dest *dest;
 665
 666         /*
 667          * Find the destination for the given service
 668          */
 669         list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
 670                 if ((dest->af == dest_af) &&
 671                     ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
 672                     (dest->port == dport)) {
 673                         /* HIT */
 674                         return dest;
 675                 }
 676         }
 677
 678         return NULL;
 679 }
 680
 681 /*
 682  * Find destination by {daddr,dport,vaddr,protocol}
 683  * Created to be used in ip_vs_process_message() in
 684  * the backup synchronization daemon. It finds the
 685  * destination to be bound to the received connection
 686  * on the backup.
 687  * Called under RCU lock, no refcnt is returned.
 688  */
 689 struct ip_vs_dest *ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
 690                                    const union nf_inet_addr *daddr,
 691                                    __be16 dport,
 692                                    const union nf_inet_addr *vaddr,
 693                                    __be16 vport, __u16 protocol, __u32 fwmark,
 694                                    __u32 flags)
 695 {
 696         struct ip_vs_dest *dest;
 697         struct ip_vs_service *svc;
 698         __be16 port = dport;
 699
 700         svc = ip_vs_service_find(ipvs, svc_af, fwmark, protocol, vaddr, vport);
 701         if (!svc)
 702                 return NULL;
 703         if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
 704                 port = 0;
 705         dest = ip_vs_lookup_dest(svc, dest_af, daddr, port);
 706         if (!dest)
 707                 dest = ip_vs_lookup_dest(svc, dest_af, daddr, port ^ dport);
 708         return dest;
 709 }
 710
 711 void ip_vs_dest_dst_rcu_free(struct rcu_head *head)
 712 {
 713         struct ip_vs_dest_dst *dest_dst = container_of(head,
 714                                                        struct ip_vs_dest_dst,
 715                                                        rcu_head);
 716
 717         dst_release(dest_dst->dst_cache);
 718         kfree(dest_dst);
 719 }
 720
 721 /* Release dest_dst and dst_cache for dest in user context */
 722 static void __ip_vs_dst_cache_reset(struct ip_vs_dest *dest)
 723 {
 724         struct ip_vs_dest_dst *old;
 725
 726         old = rcu_dereference_protected(dest->dest_dst, 1);
 727         if (old) {
 728                 RCU_INIT_POINTER(dest->dest_dst, NULL);
 729                 call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
 730         }
 731 }
 732
 733 /*
 734  *  Lookup dest by {svc,addr,port} in the destination trash.
 735  *  The destination trash is used to hold the destinations that are removed
 736  *  from the service table but are still referenced by some conn entries.
 737  *  The reason to add the destination trash is when the dest is temporary
 738  *  down (either by administrator or by monitor program), the dest can be
 739  *  picked back from the trash, the remaining connections to the dest can
 740  *  continue, and the counting information of the dest is also useful for
 741  *  scheduling.
 742  */
 743 static struct ip_vs_dest *
 744 ip_vs_trash_get_dest(struct ip_vs_service *svc, int dest_af,
 745                      const union nf_inet_addr *daddr, __be16 dport)
 746 {
 747         struct ip_vs_dest *dest;
 748         struct netns_ipvs *ipvs = svc->ipvs;
 749
 750         /*
 751          * Find the destination in trash
 752          */
 753         spin_lock_bh(&ipvs->dest_trash_lock);
 754         list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
 755                 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
 756                               "dest->refcnt=%d\n",
 757                               dest->vfwmark,
 758                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
 759                               ntohs(dest->port),
 760                               refcount_read(&dest->refcnt));
 761                 if (dest->af == dest_af &&
 762                     ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
 763                     dest->port == dport &&
 764                     dest->vfwmark == svc->fwmark &&
 765                     dest->protocol == svc->protocol &&
 766                     (svc->fwmark ||
 767                      (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
 768                       dest->vport == svc->port))) {
 769                         /* HIT */
 770                         list_del(&dest->t_list);
 771                         goto out;
 772                 }
 773         }
 774
 775         dest = NULL;
 776
 777 out:
 778         spin_unlock_bh(&ipvs->dest_trash_lock);
 779
 780         return dest;
 781 }
 782
 783 static void ip_vs_dest_free(struct ip_vs_dest *dest)
 784 {
 785         struct ip_vs_service *svc = rcu_dereference_protected(dest->svc, 1);
 786
 787         __ip_vs_dst_cache_reset(dest);
 788         __ip_vs_svc_put(svc, false);
 789         free_percpu(dest->stats.cpustats);
 790         ip_vs_dest_put_and_free(dest);
 791 }
 792
 793 /*
 794  *  Clean up all the destinations in the trash
 795  *  Called by the ip_vs_control_cleanup()
 796  *
 797  *  When the ip_vs_control_clearup is activated by ipvs module exit,
 798  *  the service tables must have been flushed and all the connections
 799  *  are expired, and the refcnt of each destination in the trash must
 800  *  be 1, so we simply release them here.
 801  */
 802 static void ip_vs_trash_cleanup(struct netns_ipvs *ipvs)
 803 {
 804         struct ip_vs_dest *dest, *nxt;
 805
 806         del_timer_sync(&ipvs->dest_trash_timer);
 807         /* No need to use dest_trash_lock */
 808         list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, t_list) {
 809                 list_del(&dest->t_list);
 810                 ip_vs_dest_free(dest);
 811         }
 812 }
 813
 814 static void
 815 ip_vs_copy_stats(struct ip_vs_kstats *dst, struct ip_vs_stats *src)
 816 {
 817 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->kstats.c - src->kstats0.c
 818
 819         spin_lock_bh(&src->lock);
 820
 821         IP_VS_SHOW_STATS_COUNTER(conns);
 822         IP_VS_SHOW_STATS_COUNTER(inpkts);
 823         IP_VS_SHOW_STATS_COUNTER(outpkts);
 824         IP_VS_SHOW_STATS_COUNTER(inbytes);
 825         IP_VS_SHOW_STATS_COUNTER(outbytes);
 826
 827         ip_vs_read_estimator(dst, src);
 828
 829         spin_unlock_bh(&src->lock);
 830 }
 831
 832 static void
 833 ip_vs_export_stats_user(struct ip_vs_stats_user *dst, struct ip_vs_kstats *src)
 834 {
 835         dst->conns = (u32)src->conns;
 836         dst->inpkts = (u32)src->inpkts;
 837         dst->outpkts = (u32)src->outpkts;
 838         dst->inbytes = src->inbytes;
 839         dst->outbytes = src->outbytes;
 840         dst->cps = (u32)src->cps;
 841         dst->inpps = (u32)src->inpps;
 842         dst->outpps = (u32)src->outpps;
 843         dst->inbps = (u32)src->inbps;
 844         dst->outbps = (u32)src->outbps;
 845 }
 846
 847 static void
 848 ip_vs_zero_stats(struct ip_vs_stats *stats)
 849 {
 850         spin_lock_bh(&stats->lock);
 851
 852         /* get current counters as zero point, rates are zeroed */
 853
 854 #define IP_VS_ZERO_STATS_COUNTER(c) stats->kstats0.c = stats->kstats.c
 855
 856         IP_VS_ZERO_STATS_COUNTER(conns);
 857         IP_VS_ZERO_STATS_COUNTER(inpkts);
 858         IP_VS_ZERO_STATS_COUNTER(outpkts);
 859         IP_VS_ZERO_STATS_COUNTER(inbytes);
 860         IP_VS_ZERO_STATS_COUNTER(outbytes);
 861
 862         ip_vs_zero_estimator(stats);
 863
 864         spin_unlock_bh(&stats->lock);
 865 }
 866
 867 /*
 868  *      Update a destination in the given service
 869  */
 870 static void
 871 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
 872                     struct ip_vs_dest_user_kern *udest, int add)
 873 {
 874         struct netns_ipvs *ipvs = svc->ipvs;
 875         struct ip_vs_service *old_svc;
 876         struct ip_vs_scheduler *sched;
 877         int conn_flags;
 878
 879         /* We cannot modify an address and change the address family */
 880         BUG_ON(!add && udest->af != dest->af);
 881
 882         if (add && udest->af != svc->af)
 883                 ipvs->mixed_address_family_dests++;
 884
 885         /* keep the last_weight with latest non-0 weight */
 886         if (add || udest->weight != 0)
 887                 atomic_set(&dest->last_weight, udest->weight);
 888
 889         /* set the weight and the flags */
 890         atomic_set(&dest->weight, udest->weight);
 891         conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
 892         conn_flags |= IP_VS_CONN_F_INACTIVE;
 893
 894         /* Need to rehash? */
 895         if ((udest->conn_flags & IP_VS_CONN_F_FWD_MASK) !=
 896             IP_VS_DFWD_METHOD(dest) ||
 897             udest->tun_type != dest->tun_type ||
 898             udest->tun_port != dest->tun_port)
 899                 ip_vs_rs_unhash(dest);
 900
 901         /* set the tunnel info */
 902         dest->tun_type = udest->tun_type;
 903         dest->tun_port = udest->tun_port;
 904         dest->tun_flags = udest->tun_flags;
 905
 906         /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
 907         if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
 908                 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
 909         } else {
 910                 /* FTP-NAT requires conntrack for mangling */
 911                 if (svc->port == FTPPORT)
 912                         ip_vs_register_conntrack(svc);
 913         }
 914         atomic_set(&dest->conn_flags, conn_flags);
 915         /* Put the real service in rs_table if not present. */
 916         ip_vs_rs_hash(ipvs, dest);
 917
 918         /* bind the service */
 919         old_svc = rcu_dereference_protected(dest->svc, 1);
 920         if (!old_svc) {
 921                 __ip_vs_bind_svc(dest, svc);
 922         } else {
 923                 if (old_svc != svc) {
 924                         ip_vs_zero_stats(&dest->stats);
 925                         __ip_vs_bind_svc(dest, svc);
 926                         __ip_vs_svc_put(old_svc, true);
 927                 }
 928         }
 929
 930         /* set the dest status flags */
 931         dest->flags |= IP_VS_DEST_F_AVAILABLE;
 932
 933         if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
 934                 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
 935         dest->u_threshold = udest->u_threshold;
 936         dest->l_threshold = udest->l_threshold;
 937
 938         dest->af = udest->af;
 939
 940         spin_lock_bh(&dest->dst_lock);
 941         __ip_vs_dst_cache_reset(dest);
 942         spin_unlock_bh(&dest->dst_lock);
 943
 944         if (add) {
 945                 ip_vs_start_estimator(svc->ipvs, &dest->stats);
 946                 list_add_rcu(&dest->n_list, &svc->destinations);
 947                 svc->num_dests++;
 948                 sched = rcu_dereference_protected(svc->scheduler, 1);
 949                 if (sched && sched->add_dest)
 950                         sched->add_dest(svc, dest);
 951         } else {
 952                 sched = rcu_dereference_protected(svc->scheduler, 1);
 953                 if (sched && sched->upd_dest)
 954                         sched->upd_dest(svc, dest);
 955         }
 956 }
 957
 958
 959 /*
 960  *      Create a destination for the given service
 961  */
 962 static int
 963 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
 964 {
 965         struct ip_vs_dest *dest;
 966         unsigned int atype, i;
 967
 968         EnterFunction(2);
 969
 970 #ifdef CONFIG_IP_VS_IPV6
 971         if (udest->af == AF_INET6) {
 972                 int ret;
 973
 974                 atype = ipv6_addr_type(&udest->addr.in6);
 975                 if ((!(atype & IPV6_ADDR_UNICAST) ||
 976                         atype & IPV6_ADDR_LINKLOCAL) &&
 977                         !__ip_vs_addr_is_local_v6(svc->ipvs->net, &udest->addr.in6))
 978                         return -EINVAL;
 979
 980                 ret = nf_defrag_ipv6_enable(svc->ipvs->net);
 981                 if (ret)
 982                         return ret;
 983         } else
 984 #endif
 985         {
 986                 atype = inet_addr_type(svc->ipvs->net, udest->addr.ip);
 987                 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
 988                         return -EINVAL;
 989         }
 990
 991         dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
 992         if (dest == NULL)
 993                 return -ENOMEM;
 994
 995         dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
 996         if (!dest->stats.cpustats)
 997                 goto err_alloc;
 998
 999         for_each_possible_cpu(i) {
1000                 struct ip_vs_cpu_stats *ip_vs_dest_stats;
1001                 ip_vs_dest_stats = per_cpu_ptr(dest->stats.cpustats, i);
1002                 u64_stats_init(&ip_vs_dest_stats->syncp);
1003         }
1004
1005         dest->af = udest->af;
1006         dest->protocol = svc->protocol;
1007         dest->vaddr = svc->addr;
1008         dest->vport = svc->port;
1009         dest->vfwmark = svc->fwmark;
1010         ip_vs_addr_copy(udest->af, &dest->addr, &udest->addr);
1011         dest->port = udest->port;
1012
1013         atomic_set(&dest->activeconns, 0);
1014         atomic_set(&dest->inactconns, 0);
1015         atomic_set(&dest->persistconns, 0);
1016         refcount_set(&dest->refcnt, 1);
1017
1018         INIT_HLIST_NODE(&dest->d_list);
1019         spin_lock_init(&dest->dst_lock);
1020         spin_lock_init(&dest->stats.lock);
1021         __ip_vs_update_dest(svc, dest, udest, 1);
1022
1023         LeaveFunction(2);
1024         return 0;
1025
1026 err_alloc:
1027         kfree(dest);
1028         return -ENOMEM;
1029 }
1030
1031
1032 /*
1033  *      Add a destination into an existing service
1034  */
1035 static int
1036 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1037 {
1038         struct ip_vs_dest *dest;
1039         union nf_inet_addr daddr;
1040         __be16 dport = udest->port;
1041         int ret;
1042
1043         EnterFunction(2);
1044
1045         if (udest->weight < 0) {
1046                 pr_err("%s(): server weight less than zero\n", __func__);
1047                 return -ERANGE;
1048         }
1049
1050         if (udest->l_threshold > udest->u_threshold) {
1051                 pr_err("%s(): lower threshold is higher than upper threshold\n",
1052                         __func__);
1053                 return -ERANGE;
1054         }
1055
1056         if (udest->tun_type == IP_VS_CONN_F_TUNNEL_TYPE_GUE) {
1057                 if (udest->tun_port == 0) {
1058                         pr_err("%s(): tunnel port is zero\n", __func__);
1059                         return -EINVAL;
1060                 }
1061         }
1062
1063         ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
1064
1065         /* We use function that requires RCU lock */
1066         rcu_read_lock();
1067         dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
1068         rcu_read_unlock();
1069
1070         if (dest != NULL) {
1071                 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
1072                 return -EEXIST;
1073         }
1074
1075         /*
1076          * Check if the dest already exists in the trash and
1077          * is from the same service
1078          */
1079         dest = ip_vs_trash_get_dest(svc, udest->af, &daddr, dport);
1080
1081         if (dest != NULL) {
1082                 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
1083                               "dest->refcnt=%d, service %u/%s:%u\n",
1084                               IP_VS_DBG_ADDR(udest->af, &daddr), ntohs(dport),
1085                               refcount_read(&dest->refcnt),
1086                               dest->vfwmark,
1087                               IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
1088                               ntohs(dest->vport));
1089
1090                 __ip_vs_update_dest(svc, dest, udest, 1);
1091                 ret = 0;
1092         } else {
1093                 /*
1094                  * Allocate and initialize the dest structure
1095                  */
1096                 ret = ip_vs_new_dest(svc, udest);
1097         }
1098         LeaveFunction(2);
1099
1100         return ret;
1101 }
1102
1103
1104 /*
1105  *      Edit a destination in the given service
1106  */
1107 static int
1108 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1109 {
1110         struct ip_vs_dest *dest;
1111         union nf_inet_addr daddr;
1112         __be16 dport = udest->port;
1113
1114         EnterFunction(2);
1115
1116         if (udest->weight < 0) {
1117                 pr_err("%s(): server weight less than zero\n", __func__);
1118                 return -ERANGE;
1119         }
1120
1121         if (udest->l_threshold > udest->u_threshold) {
1122                 pr_err("%s(): lower threshold is higher than upper threshold\n",
1123                         __func__);
1124                 return -ERANGE;
1125         }
1126
1127         if (udest->tun_type == IP_VS_CONN_F_TUNNEL_TYPE_GUE) {
1128                 if (udest->tun_port == 0) {
1129                         pr_err("%s(): tunnel port is zero\n", __func__);
1130                         return -EINVAL;
1131                 }
1132         }
1133
1134         ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
1135
1136         /* We use function that requires RCU lock */
1137         rcu_read_lock();
1138         dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
1139         rcu_read_unlock();
1140
1141         if (dest == NULL) {
1142                 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1143                 return -ENOENT;
1144         }
1145
1146         __ip_vs_update_dest(svc, dest, udest, 0);
1147         LeaveFunction(2);
1148
1149         return 0;
1150 }
1151
1152 /*
1153  *      Delete a destination (must be already unlinked from the service)
1154  */
1155 static void __ip_vs_del_dest(struct netns_ipvs *ipvs, struct ip_vs_dest *dest,
1156                              bool cleanup)
1157 {
1158         ip_vs_stop_estimator(ipvs, &dest->stats);
1159
1160         /*
1161          *  Remove it from the d-linked list with the real services.
1162          */
1163         ip_vs_rs_unhash(dest);
1164
1165         spin_lock_bh(&ipvs->dest_trash_lock);
1166         IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1167                       IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port),
1168                       refcount_read(&dest->refcnt));
1169         if (list_empty(&ipvs->dest_trash) && !cleanup)
1170                 mod_timer(&ipvs->dest_trash_timer,
1171                           jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1172         /* dest lives in trash with reference */
1173         list_add(&dest->t_list, &ipvs->dest_trash);
1174         dest->idle_start = 0;
1175         spin_unlock_bh(&ipvs->dest_trash_lock);
1176
1177         /* Queue up delayed work to expire all no destination connections.
1178          * No-op when CONFIG_SYSCTL is disabled.
1179          */
1180         if (!cleanup)
1181                 ip_vs_enqueue_expire_nodest_conns(ipvs);
1182 }
1183
1184
1185 /*
1186  *      Unlink a destination from the given service
1187  */
1188 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1189                                 struct ip_vs_dest *dest,
1190                                 int svcupd)
1191 {
1192         dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1193
1194         /*
1195          *  Remove it from the d-linked destination list.
1196          */
1197         list_del_rcu(&dest->n_list);
1198         svc->num_dests--;
1199
1200         if (dest->af != svc->af)
1201                 svc->ipvs->mixed_address_family_dests--;
1202
1203         if (svcupd) {
1204                 struct ip_vs_scheduler *sched;
1205
1206                 sched = rcu_dereference_protected(svc->scheduler, 1);
1207                 if (sched && sched->del_dest)
1208                         sched->del_dest(svc, dest);
1209         }
1210 }
1211
1212
1213 /*
1214  *      Delete a destination server in the given service
1215  */
1216 static int
1217 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1218 {
1219         struct ip_vs_dest *dest;
1220         __be16 dport = udest->port;
1221
1222         EnterFunction(2);
1223
1224         /* We use function that requires RCU lock */
1225         rcu_read_lock();
1226         dest = ip_vs_lookup_dest(svc, udest->af, &udest->addr, dport);
1227         rcu_read_unlock();
1228
1229         if (dest == NULL) {
1230                 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1231                 return -ENOENT;
1232         }
1233
1234         /*
1235          *      Unlink dest from the service
1236          */
1237         __ip_vs_unlink_dest(svc, dest, 1);
1238
1239         /*
1240          *      Delete the destination
1241          */
1242         __ip_vs_del_dest(svc->ipvs, dest, false);
1243
1244         LeaveFunction(2);
1245
1246         return 0;
1247 }
1248
1249 static void ip_vs_dest_trash_expire(struct timer_list *t)
1250 {
1251         struct netns_ipvs *ipvs = from_timer(ipvs, t, dest_trash_timer);
1252         struct ip_vs_dest *dest, *next;
1253         unsigned long now = jiffies;
1254
1255         spin_lock(&ipvs->dest_trash_lock);
1256         list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) {
1257                 if (refcount_read(&dest->refcnt) > 1)
1258                         continue;
1259                 if (dest->idle_start) {
1260                         if (time_before(now, dest->idle_start +
1261                                              IP_VS_DEST_TRASH_PERIOD))
1262                                 continue;
1263                 } else {
1264                         dest->idle_start = max(1UL, now);
1265                         continue;
1266                 }
1267                 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1268                               dest->vfwmark,
1269                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1270                               ntohs(dest->port));
1271                 list_del(&dest->t_list);
1272                 ip_vs_dest_free(dest);
1273         }
1274         if (!list_empty(&ipvs->dest_trash))
1275                 mod_timer(&ipvs->dest_trash_timer,
1276                           jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1277         spin_unlock(&ipvs->dest_trash_lock);
1278 }
1279
1280 /*
1281  *      Add a service into the service hash table
1282  */
1283 static int
1284 ip_vs_add_service(struct netns_ipvs *ipvs, struct ip_vs_service_user_kern *u,
1285                   struct ip_vs_service **svc_p)
1286 {
1287         int ret = 0, i;
1288         struct ip_vs_scheduler *sched = NULL;
1289         struct ip_vs_pe *pe = NULL;
1290         struct ip_vs_service *svc = NULL;
1291         int ret_hooks = -1;
1292
1293         /* increase the module use count */
1294         if (!ip_vs_use_count_inc())
1295                 return -ENOPROTOOPT;
1296
1297         /* Lookup the scheduler by 'u->sched_name' */
1298         if (strcmp(u->sched_name, "none")) {
1299                 sched = ip_vs_scheduler_get(u->sched_name);
1300                 if (!sched) {
1301                         pr_info("Scheduler module ip_vs_%s not found\n",
1302                                 u->sched_name);
1303                         ret = -ENOENT;
1304                         goto out_err;
1305                 }
1306         }
1307
1308         if (u->pe_name && *u->pe_name) {
1309                 pe = ip_vs_pe_getbyname(u->pe_name);
1310                 if (pe == NULL) {
1311                         pr_info("persistence engine module ip_vs_pe_%s "
1312                                 "not found\n", u->pe_name);
1313                         ret = -ENOENT;
1314                         goto out_err;
1315                 }
1316         }
1317
1318 #ifdef CONFIG_IP_VS_IPV6
1319         if (u->af == AF_INET6) {
1320                 __u32 plen = (__force __u32) u->netmask;
1321
1322                 if (plen < 1 || plen > 128) {
1323                         ret = -EINVAL;
1324                         goto out_err;
1325                 }
1326
1327                 ret = nf_defrag_ipv6_enable(ipvs->net);
1328                 if (ret)
1329                         goto out_err;
1330         }
1331 #endif
1332
1333         if ((u->af == AF_INET && !ipvs->num_services) ||
1334             (u->af == AF_INET6 && !ipvs->num_services6)) {
1335                 ret = ip_vs_register_hooks(ipvs, u->af);
1336                 if (ret < 0)
1337                         goto out_err;
1338                 ret_hooks = ret;
1339         }
1340
1341         svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1342         if (svc == NULL) {
1343                 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1344                 ret = -ENOMEM;
1345                 goto out_err;
1346         }
1347         svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1348         if (!svc->stats.cpustats) {
1349                 ret = -ENOMEM;
1350                 goto out_err;
1351         }
1352
1353         for_each_possible_cpu(i) {
1354                 struct ip_vs_cpu_stats *ip_vs_stats;
1355                 ip_vs_stats = per_cpu_ptr(svc->stats.cpustats, i);
1356                 u64_stats_init(&ip_vs_stats->syncp);
1357         }
1358
1359
1360         /* I'm the first user of the service */
1361         atomic_set(&svc->refcnt, 0);
1362
1363         svc->af = u->af;
1364         svc->protocol = u->protocol;
1365         ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1366         svc->port = u->port;
1367         svc->fwmark = u->fwmark;
1368         svc->flags = u->flags & ~IP_VS_SVC_F_HASHED;
1369         svc->timeout = u->timeout * HZ;
1370         svc->netmask = u->netmask;
1371         svc->ipvs = ipvs;
1372
1373         INIT_LIST_HEAD(&svc->destinations);
1374         spin_lock_init(&svc->sched_lock);
1375         spin_lock_init(&svc->stats.lock);
1376
1377         /* Bind the scheduler */
1378         if (sched) {
1379                 ret = ip_vs_bind_scheduler(svc, sched);
1380                 if (ret)
1381                         goto out_err;
1382                 sched = NULL;
1383         }
1384
1385         /* Bind the ct retriever */
1386         RCU_INIT_POINTER(svc->pe, pe);
1387         pe = NULL;
1388
1389         /* Update the virtual service counters */
1390         if (svc->port == FTPPORT)
1391                 atomic_inc(&ipvs->ftpsvc_counter);
1392         else if (svc->port == 0)
1393                 atomic_inc(&ipvs->nullsvc_counter);
1394         if (svc->pe && svc->pe->conn_out)
1395                 atomic_inc(&ipvs->conn_out_counter);
1396
1397         ip_vs_start_estimator(ipvs, &svc->stats);
1398
1399         /* Count only IPv4 services for old get/setsockopt interface */
1400         if (svc->af == AF_INET)
1401                 ipvs->num_services++;
1402         else if (svc->af == AF_INET6)
1403                 ipvs->num_services6++;
1404
1405         /* Hash the service into the service table */
1406         ip_vs_svc_hash(svc);
1407
1408         *svc_p = svc;
1409         /* Now there is a service - full throttle */
1410         ipvs->enable = 1;
1411         return 0;
1412
1413
1414  out_err:
1415         if (ret_hooks >= 0)
1416                 ip_vs_unregister_hooks(ipvs, u->af);
1417         if (svc != NULL) {
1418                 ip_vs_unbind_scheduler(svc, sched);
1419                 ip_vs_service_free(svc);
1420         }
1421         ip_vs_scheduler_put(sched);
1422         ip_vs_pe_put(pe);
1423
1424         /* decrease the module use count */
1425         ip_vs_use_count_dec();
1426
1427         return ret;
1428 }
1429
1430
1431 /*
1432  *      Edit a service and bind it with a new scheduler
1433  */
1434 static int
1435 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1436 {
1437         struct ip_vs_scheduler *sched = NULL, *old_sched;
1438         struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1439         int ret = 0;
1440         bool new_pe_conn_out, old_pe_conn_out;
1441
1442         /*
1443          * Lookup the scheduler, by 'u->sched_name'
1444          */
1445         if (strcmp(u->sched_name, "none")) {
1446                 sched = ip_vs_scheduler_get(u->sched_name);
1447                 if (!sched) {
1448                         pr_info("Scheduler module ip_vs_%s not found\n",
1449                                 u->sched_name);
1450                         return -ENOENT;
1451                 }
1452         }
1453         old_sched = sched;
1454
1455         if (u->pe_name && *u->pe_name) {
1456                 pe = ip_vs_pe_getbyname(u->pe_name);
1457                 if (pe == NULL) {
1458                         pr_info("persistence engine module ip_vs_pe_%s "
1459                                 "not found\n", u->pe_name);
1460                         ret = -ENOENT;
1461                         goto out;
1462                 }
1463                 old_pe = pe;
1464         }
1465
1466 #ifdef CONFIG_IP_VS_IPV6
1467         if (u->af == AF_INET6) {
1468                 __u32 plen = (__force __u32) u->netmask;
1469
1470                 if (plen < 1 || plen > 128) {
1471                         ret = -EINVAL;
1472                         goto out;
1473                 }
1474         }
1475 #endif
1476
1477         old_sched = rcu_dereference_protected(svc->scheduler, 1);
1478         if (sched != old_sched) {
1479                 if (old_sched) {
1480                         ip_vs_unbind_scheduler(svc, old_sched);
1481                         RCU_INIT_POINTER(svc->scheduler, NULL);
1482                         /* Wait all svc->sched_data users */
1483                         synchronize_rcu();
1484                 }
1485                 /* Bind the new scheduler */
1486                 if (sched) {
1487                         ret = ip_vs_bind_scheduler(svc, sched);
1488                         if (ret) {
1489                                 ip_vs_scheduler_put(sched);
1490                                 goto out;
1491                         }
1492                 }
1493         }
1494
1495         /*
1496          * Set the flags and timeout value
1497          */
1498         svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1499         svc->timeout = u->timeout * HZ;
1500         svc->netmask = u->netmask;
1501
1502         old_pe = rcu_dereference_protected(svc->pe, 1);
1503         if (pe != old_pe) {
1504                 rcu_assign_pointer(svc->pe, pe);
1505                 /* check for optional methods in new pe */
1506                 new_pe_conn_out = (pe && pe->conn_out) ? true : false;
1507                 old_pe_conn_out = (old_pe && old_pe->conn_out) ? true : false;
1508                 if (new_pe_conn_out && !old_pe_conn_out)
1509                         atomic_inc(&svc->ipvs->conn_out_counter);
1510                 if (old_pe_conn_out && !new_pe_conn_out)
1511                         atomic_dec(&svc->ipvs->conn_out_counter);
1512         }
1513
1514 out:
1515         ip_vs_scheduler_put(old_sched);
1516         ip_vs_pe_put(old_pe);
1517         return ret;
1518 }
1519
1520 /*
1521  *      Delete a service from the service list
1522  *      - The service must be unlinked, unlocked and not referenced!
1523  *      - We are called under _bh lock
1524  */
1525 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1526 {
1527         struct ip_vs_dest *dest, *nxt;
1528         struct ip_vs_scheduler *old_sched;
1529         struct ip_vs_pe *old_pe;
1530         struct netns_ipvs *ipvs = svc->ipvs;
1531
1532         if (svc->af == AF_INET) {
1533                 ipvs->num_services--;
1534                 if (!ipvs->num_services)
1535                         ip_vs_unregister_hooks(ipvs, svc->af);
1536         } else if (svc->af == AF_INET6) {
1537                 ipvs->num_services6--;
1538                 if (!ipvs->num_services6)
1539                         ip_vs_unregister_hooks(ipvs, svc->af);
1540         }
1541
1542         ip_vs_stop_estimator(svc->ipvs, &svc->stats);
1543
1544         /* Unbind scheduler */
1545         old_sched = rcu_dereference_protected(svc->scheduler, 1);
1546         ip_vs_unbind_scheduler(svc, old_sched);
1547         ip_vs_scheduler_put(old_sched);
1548
1549         /* Unbind persistence engine, keep svc->pe */
1550         old_pe = rcu_dereference_protected(svc->pe, 1);
1551         if (old_pe && old_pe->conn_out)
1552                 atomic_dec(&ipvs->conn_out_counter);
1553         ip_vs_pe_put(old_pe);
1554
1555         /*
1556          *    Unlink the whole destination list
1557          */
1558         list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1559                 __ip_vs_unlink_dest(svc, dest, 0);
1560                 __ip_vs_del_dest(svc->ipvs, dest, cleanup);
1561         }
1562
1563         /*
1564          *    Update the virtual service counters
1565          */
1566         if (svc->port == FTPPORT)
1567                 atomic_dec(&ipvs->ftpsvc_counter);
1568         else if (svc->port == 0)
1569                 atomic_dec(&ipvs->nullsvc_counter);
1570
1571         /*
1572          *    Free the service if nobody refers to it
1573          */
1574         __ip_vs_svc_put(svc, true);
1575
1576         /* decrease the module use count */
1577         ip_vs_use_count_dec();
1578 }
1579
1580 /*
1581  * Unlink a service from list and try to delete it if its refcnt reached 0
1582  */
1583 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1584 {
1585         ip_vs_unregister_conntrack(svc);
1586         /* Hold svc to avoid double release from dest_trash */
1587         atomic_inc(&svc->refcnt);
1588         /*
1589          * Unhash it from the service table
1590          */
1591         ip_vs_svc_unhash(svc);
1592
1593         __ip_vs_del_service(svc, cleanup);
1594 }
1595
1596 /*
1597  *      Delete a service from the service list
1598  */
1599 static int ip_vs_del_service(struct ip_vs_service *svc)
1600 {
1601         if (svc == NULL)
1602                 return -EEXIST;
1603         ip_vs_unlink_service(svc, false);
1604
1605         return 0;
1606 }
1607
1608
1609 /*
1610  *      Flush all the virtual services
1611  */
1612 static int ip_vs_flush(struct netns_ipvs *ipvs, bool cleanup)
1613 {
1614         int idx;
1615         struct ip_vs_service *svc;
1616         struct hlist_node *n;
1617
1618         /*
1619          * Flush the service table hashed by <netns,protocol,addr,port>
1620          */
1621         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1622                 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1623                                           s_list) {
1624                         if (svc->ipvs == ipvs)
1625                                 ip_vs_unlink_service(svc, cleanup);
1626                 }
1627         }
1628
1629         /*
1630          * Flush the service table hashed by fwmark
1631          */
1632         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1633                 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1634                                           f_list) {
1635                         if (svc->ipvs == ipvs)
1636                                 ip_vs_unlink_service(svc, cleanup);
1637                 }
1638         }
1639
1640         return 0;
1641 }
1642
1643 /*
1644  *      Delete service by {netns} in the service table.
1645  *      Called by __ip_vs_batch_cleanup()
1646  */
1647 void ip_vs_service_nets_cleanup(struct list_head *net_list)
1648 {
1649         struct netns_ipvs *ipvs;
1650         struct net *net;
1651
1652         EnterFunction(2);
1653         /* Check for "full" addressed entries */
1654         mutex_lock(&__ip_vs_mutex);
1655         list_for_each_entry(net, net_list, exit_list) {
1656                 ipvs = net_ipvs(net);
1657                 ip_vs_flush(ipvs, true);
1658         }
1659         mutex_unlock(&__ip_vs_mutex);
1660         LeaveFunction(2);
1661 }
1662
1663 /* Put all references for device (dst_cache) */
1664 static inline void
1665 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1666 {
1667         struct ip_vs_dest_dst *dest_dst;
1668
1669         spin_lock_bh(&dest->dst_lock);
1670         dest_dst = rcu_dereference_protected(dest->dest_dst, 1);
1671         if (dest_dst && dest_dst->dst_cache->dev == dev) {
1672                 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1673                               dev->name,
1674                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1675                               ntohs(dest->port),
1676                               refcount_read(&dest->refcnt));
1677                 __ip_vs_dst_cache_reset(dest);
1678         }
1679         spin_unlock_bh(&dest->dst_lock);
1680
1681 }
1682 /* Netdev event receiver
1683  * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1684  */
1685 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1686                            void *ptr)
1687 {
1688         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1689         struct net *net = dev_net(dev);
1690         struct netns_ipvs *ipvs = net_ipvs(net);
1691         struct ip_vs_service *svc;
1692         struct ip_vs_dest *dest;
1693         unsigned int idx;
1694
1695         if (event != NETDEV_DOWN || !ipvs)
1696                 return NOTIFY_DONE;
1697         IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1698         EnterFunction(2);
1699         mutex_lock(&__ip_vs_mutex);
1700         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1701                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1702                         if (svc->ipvs == ipvs) {
1703                                 list_for_each_entry(dest, &svc->destinations,
1704                                                     n_list) {
1705                                         ip_vs_forget_dev(dest, dev);
1706                                 }
1707                         }
1708                 }
1709
1710                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1711                         if (svc->ipvs == ipvs) {
1712                                 list_for_each_entry(dest, &svc->destinations,
1713                                                     n_list) {
1714                                         ip_vs_forget_dev(dest, dev);
1715                                 }
1716                         }
1717
1718                 }
1719         }
1720
1721         spin_lock_bh(&ipvs->dest_trash_lock);
1722         list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1723                 ip_vs_forget_dev(dest, dev);
1724         }
1725         spin_unlock_bh(&ipvs->dest_trash_lock);
1726         mutex_unlock(&__ip_vs_mutex);
1727         LeaveFunction(2);
1728         return NOTIFY_DONE;
1729 }
1730
1731 /*
1732  *      Zero counters in a service or all services
1733  */
1734 static int ip_vs_zero_service(struct ip_vs_service *svc)
1735 {
1736         struct ip_vs_dest *dest;
1737
1738         list_for_each_entry(dest, &svc->destinations, n_list) {
1739                 ip_vs_zero_stats(&dest->stats);
1740         }
1741         ip_vs_zero_stats(&svc->stats);
1742         return 0;
1743 }
1744
1745 static int ip_vs_zero_all(struct netns_ipvs *ipvs)
1746 {
1747         int idx;
1748         struct ip_vs_service *svc;
1749
1750         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1751                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1752                         if (svc->ipvs == ipvs)
1753                                 ip_vs_zero_service(svc);
1754                 }
1755         }
1756
1757         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1758                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1759                         if (svc->ipvs == ipvs)
1760                                 ip_vs_zero_service(svc);
1761                 }
1762         }
1763
1764         ip_vs_zero_stats(&ipvs->tot_stats);
1765         return 0;
1766 }
1767
1768 #ifdef CONFIG_SYSCTL
1769
1770 static int
1771 proc_do_defense_mode(struct ctl_table *table, int write,
1772                      void *buffer, size_t *lenp, loff_t *ppos)
1773 {
1774         struct netns_ipvs *ipvs = table->extra2;
1775         int *valp = table->data;
1776         int val = *valp;
1777         int rc;
1778
1779         struct ctl_table tmp = {
1780                 .data = &val,
1781                 .maxlen = sizeof(int),
1782                 .mode = table->mode,
1783         };
1784
1785         rc = proc_dointvec(&tmp, write, buffer, lenp, ppos);
1786         if (write && (*valp != val)) {
1787                 if (val < 0 || val > 3) {
1788                         rc = -EINVAL;
1789                 } else {
1790                         *valp = val;
1791                         update_defense_level(ipvs);
1792                 }
1793         }
1794         return rc;
1795 }
1796
1797 static int
1798 proc_do_sync_threshold(struct ctl_table *table, int write,
1799                        void *buffer, size_t *lenp, loff_t *ppos)
1800 {
1801         struct netns_ipvs *ipvs = table->extra2;
1802         int *valp = table->data;
1803         int val[2];
1804         int rc;
1805         struct ctl_table tmp = {
1806                 .data = &val,
1807                 .maxlen = table->maxlen,
1808                 .mode = table->mode,
1809         };
1810
1811         mutex_lock(&ipvs->sync_mutex);
1812         memcpy(val, valp, sizeof(val));
1813         rc = proc_dointvec(&tmp, write, buffer, lenp, ppos);
1814         if (write) {
1815                 if (val[0] < 0 || val[1] < 0 ||
1816                     (val[0] >= val[1] && val[1]))
1817                         rc = -EINVAL;
1818                 else
1819                         memcpy(valp, val, sizeof(val));
1820         }
1821         mutex_unlock(&ipvs->sync_mutex);
1822         return rc;
1823 }
1824
1825 static int
1826 proc_do_sync_ports(struct ctl_table *table, int write,
1827                    void *buffer, size_t *lenp, loff_t *ppos)
1828 {
1829         int *valp = table->data;
1830         int val = *valp;
1831         int rc;
1832
1833         struct ctl_table tmp = {
1834                 .data = &val,
1835                 .maxlen = sizeof(int),
1836                 .mode = table->mode,
1837         };
1838
1839         rc = proc_dointvec(&tmp, write, buffer, lenp, ppos);
1840         if (write && (*valp != val)) {
1841                 if (val < 1 || !is_power_of_2(val))
1842                         rc = -EINVAL;
1843                 else
1844                         *valp = val;
1845         }
1846         return rc;
1847 }
1848
1849 /*
1850  *      IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1851  *      Do not change order or insert new entries without
1852  *      align with netns init in ip_vs_control_net_init()
1853  */
1854
1855 static struct ctl_table vs_vars[] = {
1856         {
1857                 .procname       = "amemthresh",
1858                 .maxlen         = sizeof(int),
1859                 .mode           = 0644,
1860                 .proc_handler   = proc_dointvec,
1861         },
1862         {
1863                 .procname       = "am_droprate",
1864                 .maxlen         = sizeof(int),
1865                 .mode           = 0644,
1866                 .proc_handler   = proc_dointvec,
1867         },
1868         {
1869                 .procname       = "drop_entry",
1870                 .maxlen         = sizeof(int),
1871                 .mode           = 0644,
1872                 .proc_handler   = proc_do_defense_mode,
1873         },
1874         {
1875                 .procname       = "drop_packet",
1876                 .maxlen         = sizeof(int),
1877                 .mode           = 0644,
1878                 .proc_handler   = proc_do_defense_mode,
1879         },
1880 #ifdef CONFIG_IP_VS_NFCT
1881         {
1882                 .procname       = "conntrack",
1883                 .maxlen         = sizeof(int),
1884                 .mode           = 0644,
1885                 .proc_handler   = &proc_dointvec,
1886         },
1887 #endif
1888         {
1889                 .procname       = "secure_tcp",
1890                 .maxlen         = sizeof(int),
1891                 .mode           = 0644,
1892                 .proc_handler   = proc_do_defense_mode,
1893         },
1894         {
1895                 .procname       = "snat_reroute",
1896                 .maxlen         = sizeof(int),
1897                 .mode           = 0644,
1898                 .proc_handler   = &proc_dointvec,
1899         },
1900         {
1901                 .procname       = "sync_version",
1902                 .maxlen         = sizeof(int),
1903                 .mode           = 0644,
1904                 .proc_handler   = proc_dointvec_minmax,
1905                 .extra1         = SYSCTL_ZERO,
1906                 .extra2         = SYSCTL_ONE,
1907         },
1908         {
1909                 .procname       = "sync_ports",
1910                 .maxlen         = sizeof(int),
1911                 .mode           = 0644,
1912                 .proc_handler   = proc_do_sync_ports,
1913         },
1914         {
1915                 .procname       = "sync_persist_mode",
1916                 .maxlen         = sizeof(int),
1917                 .mode           = 0644,
1918                 .proc_handler   = proc_dointvec,
1919         },
1920         {
1921                 .procname       = "sync_qlen_max",
1922                 .maxlen         = sizeof(unsigned long),
1923                 .mode           = 0644,
1924                 .proc_handler   = proc_doulongvec_minmax,
1925         },
1926         {
1927                 .procname       = "sync_sock_size",
1928                 .maxlen         = sizeof(int),
1929                 .mode           = 0644,
1930                 .proc_handler   = proc_dointvec,
1931         },
1932         {
1933                 .procname       = "cache_bypass",
1934                 .maxlen         = sizeof(int),
1935                 .mode           = 0644,
1936                 .proc_handler   = proc_dointvec,
1937         },
1938         {
1939                 .procname       = "expire_nodest_conn",
1940                 .maxlen         = sizeof(int),
1941                 .mode           = 0644,
1942                 .proc_handler   = proc_dointvec,
1943         },
1944         {
1945                 .procname       = "sloppy_tcp",
1946                 .maxlen         = sizeof(int),
1947                 .mode           = 0644,
1948                 .proc_handler   = proc_dointvec,
1949         },
1950         {
1951                 .procname       = "sloppy_sctp",
1952                 .maxlen         = sizeof(int),
1953                 .mode           = 0644,
1954                 .proc_handler   = proc_dointvec,
1955         },
1956         {
1957                 .procname       = "expire_quiescent_template",
1958                 .maxlen         = sizeof(int),
1959                 .mode           = 0644,
1960                 .proc_handler   = proc_dointvec,
1961         },
1962         {
1963                 .procname       = "sync_threshold",
1964                 .maxlen         =
1965                         sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1966                 .mode           = 0644,
1967                 .proc_handler   = proc_do_sync_threshold,
1968         },
1969         {
1970                 .procname       = "sync_refresh_period",
1971                 .maxlen         = sizeof(int),
1972                 .mode           = 0644,
1973                 .proc_handler   = proc_dointvec_jiffies,
1974         },
1975         {
1976                 .procname       = "sync_retries",
1977                 .maxlen         = sizeof(int),
1978                 .mode           = 0644,
1979                 .proc_handler   = proc_dointvec_minmax,
1980                 .extra1         = SYSCTL_ZERO,
1981                 .extra2         = SYSCTL_THREE,
1982         },
1983         {
1984                 .procname       = "nat_icmp_send",
1985                 .maxlen         = sizeof(int),
1986                 .mode           = 0644,
1987                 .proc_handler   = proc_dointvec,
1988         },
1989         {
1990                 .procname       = "pmtu_disc",
1991                 .maxlen         = sizeof(int),
1992                 .mode           = 0644,
1993                 .proc_handler   = proc_dointvec,
1994         },
1995         {
1996                 .procname       = "backup_only",
1997                 .maxlen         = sizeof(int),
1998                 .mode           = 0644,
1999                 .proc_handler   = proc_dointvec,
2000         },
2001         {
2002                 .procname       = "conn_reuse_mode",
2003                 .maxlen         = sizeof(int),
2004                 .mode           = 0644,
2005                 .proc_handler   = proc_dointvec,
2006         },
2007         {
2008                 .procname       = "schedule_icmp",
2009                 .maxlen         = sizeof(int),
2010                 .mode           = 0644,
2011                 .proc_handler   = proc_dointvec,
2012         },
2013         {
2014                 .procname       = "ignore_tunneled",
2015                 .maxlen         = sizeof(int),
2016                 .mode           = 0644,
2017                 .proc_handler   = proc_dointvec,
2018         },
2019         {
2020                 .procname       = "run_estimation",
2021                 .maxlen         = sizeof(int),
2022                 .mode           = 0644,
2023                 .proc_handler   = proc_dointvec,
2024         },
2025 #ifdef CONFIG_IP_VS_DEBUG
2026         {
2027                 .procname       = "debug_level",
2028                 .data           = &sysctl_ip_vs_debug_level,
2029                 .maxlen         = sizeof(int),
2030                 .mode           = 0644,
2031                 .proc_handler   = proc_dointvec,
2032         },
2033 #endif
2034         { }
2035 };
2036
2037 #endif
2038
2039 #ifdef CONFIG_PROC_FS
2040
2041 struct ip_vs_iter {
2042         struct seq_net_private p;  /* Do not move this, netns depends upon it*/
2043         struct hlist_head *table;
2044         int bucket;
2045 };
2046
2047 /*
2048  *      Write the contents of the VS rule table to a PROCfs file.
2049  *      (It is kept just for backward compatibility)
2050  */
2051 static inline const char *ip_vs_fwd_name(unsigned int flags)
2052 {
2053         switch (flags & IP_VS_CONN_F_FWD_MASK) {
2054         case IP_VS_CONN_F_LOCALNODE:
2055                 return "Local";
2056         case IP_VS_CONN_F_TUNNEL:
2057                 return "Tunnel";
2058         case IP_VS_CONN_F_DROUTE:
2059                 return "Route";
2060         default:
2061                 return "Masq";
2062         }
2063 }
2064
2065
2066 /* Get the Nth entry in the two lists */
2067 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
2068 {
2069         struct net *net = seq_file_net(seq);
2070         struct netns_ipvs *ipvs = net_ipvs(net);
2071         struct ip_vs_iter *iter = seq->private;
2072         int idx;
2073         struct ip_vs_service *svc;
2074
2075         /* look in hash by protocol */
2076         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2077                 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
2078                         if ((svc->ipvs == ipvs) && pos-- == 0) {
2079                                 iter->table = ip_vs_svc_table;
2080                                 iter->bucket = idx;
2081                                 return svc;
2082                         }
2083                 }
2084         }
2085
2086         /* keep looking in fwmark */
2087         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2088                 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
2089                                          f_list) {
2090                         if ((svc->ipvs == ipvs) && pos-- == 0) {
2091                                 iter->table = ip_vs_svc_fwm_table;
2092                                 iter->bucket = idx;
2093                                 return svc;
2094                         }
2095                 }
2096         }
2097
2098         return NULL;
2099 }
2100
2101 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
2102         __acquires(RCU)
2103 {
2104         rcu_read_lock();
2105         return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
2106 }
2107
2108
2109 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2110 {
2111         struct hlist_node *e;
2112         struct ip_vs_iter *iter;
2113         struct ip_vs_service *svc;
2114
2115         ++*pos;
2116         if (v == SEQ_START_TOKEN)
2117                 return ip_vs_info_array(seq,0);
2118
2119         svc = v;
2120         iter = seq->private;
2121
2122         if (iter->table == ip_vs_svc_table) {
2123                 /* next service in table hashed by protocol */
2124                 e = rcu_dereference(hlist_next_rcu(&svc->s_list));
2125                 if (e)
2126                         return hlist_entry(e, struct ip_vs_service, s_list);
2127
2128                 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2129                         hlist_for_each_entry_rcu(svc,
2130                                                  &ip_vs_svc_table[iter->bucket],
2131                                                  s_list) {
2132                                 return svc;
2133                         }
2134                 }
2135
2136                 iter->table = ip_vs_svc_fwm_table;
2137                 iter->bucket = -1;
2138                 goto scan_fwmark;
2139         }
2140
2141         /* next service in hashed by fwmark */
2142         e = rcu_dereference(hlist_next_rcu(&svc->f_list));
2143         if (e)
2144                 return hlist_entry(e, struct ip_vs_service, f_list);
2145
2146  scan_fwmark:
2147         while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2148                 hlist_for_each_entry_rcu(svc,
2149                                          &ip_vs_svc_fwm_table[iter->bucket],
2150                                          f_list)
2151                         return svc;
2152         }
2153
2154         return NULL;
2155 }
2156
2157 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
2158         __releases(RCU)
2159 {
2160         rcu_read_unlock();
2161 }
2162
2163
2164 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
2165 {
2166         if (v == SEQ_START_TOKEN) {
2167                 seq_printf(seq,
2168                         "IP Virtual Server version %d.%d.%d (size=%d)\n",
2169                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2170                 seq_puts(seq,
2171                          "Prot LocalAddress:Port Scheduler Flags\n");
2172                 seq_puts(seq,
2173                          "  -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2174         } else {
2175                 struct net *net = seq_file_net(seq);
2176                 struct netns_ipvs *ipvs = net_ipvs(net);
2177                 const struct ip_vs_service *svc = v;
2178                 const struct ip_vs_iter *iter = seq->private;
2179                 const struct ip_vs_dest *dest;
2180                 struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler);
2181                 char *sched_name = sched ? sched->name : "none";
2182
2183                 if (svc->ipvs != ipvs)
2184                         return 0;
2185                 if (iter->table == ip_vs_svc_table) {
2186 #ifdef CONFIG_IP_VS_IPV6
2187                         if (svc->af == AF_INET6)
2188                                 seq_printf(seq, "%s  [%pI6]:%04X %s ",
2189                                            ip_vs_proto_name(svc->protocol),
2190                                            &svc->addr.in6,
2191                                            ntohs(svc->port),
2192                                            sched_name);
2193                         else
2194 #endif
2195                                 seq_printf(seq, "%s  %08X:%04X %s %s ",
2196                                            ip_vs_proto_name(svc->protocol),
2197                                            ntohl(svc->addr.ip),
2198                                            ntohs(svc->port),
2199                                            sched_name,
2200                                            (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2201                 } else {
2202                         seq_printf(seq, "FWM  %08X %s %s",
2203                                    svc->fwmark, sched_name,
2204                                    (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2205                 }
2206
2207                 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2208                         seq_printf(seq, "persistent %d %08X\n",
2209                                 svc->timeout,
2210                                 ntohl(svc->netmask));
2211                 else
2212                         seq_putc(seq, '\n');
2213
2214                 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
2215 #ifdef CONFIG_IP_VS_IPV6
2216                         if (dest->af == AF_INET6)
2217                                 seq_printf(seq,
2218                                            "  -> [%pI6]:%04X"
2219                                            "      %-7s %-6d %-10d %-10d\n",
2220                                            &dest->addr.in6,
2221                                            ntohs(dest->port),
2222                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2223                                            atomic_read(&dest->weight),
2224                                            atomic_read(&dest->activeconns),
2225                                            atomic_read(&dest->inactconns));
2226                         else
2227 #endif
2228                                 seq_printf(seq,
2229                                            "  -> %08X:%04X      "
2230                                            "%-7s %-6d %-10d %-10d\n",
2231                                            ntohl(dest->addr.ip),
2232                                            ntohs(dest->port),
2233                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2234                                            atomic_read(&dest->weight),
2235                                            atomic_read(&dest->activeconns),
2236                                            atomic_read(&dest->inactconns));
2237
2238                 }
2239         }
2240         return 0;
2241 }
2242
2243 static const struct seq_operations ip_vs_info_seq_ops = {
2244         .start = ip_vs_info_seq_start,
2245         .next  = ip_vs_info_seq_next,
2246         .stop  = ip_vs_info_seq_stop,
2247         .show  = ip_vs_info_seq_show,
2248 };
2249
2250 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2251 {
2252         struct net *net = seq_file_single_net(seq);
2253         struct ip_vs_kstats show;
2254
2255 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
2256         seq_puts(seq,
2257                  "   Total Incoming Outgoing         Incoming         Outgoing\n");
2258         seq_puts(seq,
2259                  "   Conns  Packets  Packets            Bytes            Bytes\n");
2260
2261         ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2262         seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n\n",
2263                    (unsigned long long)show.conns,
2264                    (unsigned long long)show.inpkts,
2265                    (unsigned long long)show.outpkts,
2266                    (unsigned long long)show.inbytes,
2267                    (unsigned long long)show.outbytes);
2268
2269 /*                01234567 01234567 01234567 0123456701234567 0123456701234567*/
2270         seq_puts(seq,
2271                  " Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
2272         seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n",
2273                    (unsigned long long)show.cps,
2274                    (unsigned long long)show.inpps,
2275                    (unsigned long long)show.outpps,
2276                    (unsigned long long)show.inbps,
2277                    (unsigned long long)show.outbps);
2278
2279         return 0;
2280 }
2281
2282 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2283 {
2284         struct net *net = seq_file_single_net(seq);
2285         struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2286         struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats;
2287         struct ip_vs_kstats kstats;
2288         int i;
2289
2290 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
2291         seq_puts(seq,
2292                  "       Total Incoming Outgoing         Incoming         Outgoing\n");
2293         seq_puts(seq,
2294                  "CPU    Conns  Packets  Packets            Bytes            Bytes\n");
2295
2296         for_each_possible_cpu(i) {
2297                 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2298                 unsigned int start;
2299                 u64 conns, inpkts, outpkts, inbytes, outbytes;
2300
2301                 do {
2302                         start = u64_stats_fetch_begin_irq(&u->syncp);
2303                         conns = u64_stats_read(&u->cnt.conns);
2304                         inpkts = u64_stats_read(&u->cnt.inpkts);
2305                         outpkts = u64_stats_read(&u->cnt.outpkts);
2306                         inbytes = u64_stats_read(&u->cnt.inbytes);
2307                         outbytes = u64_stats_read(&u->cnt.outbytes);
2308                 } while (u64_stats_fetch_retry_irq(&u->syncp, start));
2309
2310                 seq_printf(seq, "%3X %8LX %8LX %8LX %16LX %16LX\n",
2311                            i, (u64)conns, (u64)inpkts,
2312                            (u64)outpkts, (u64)inbytes,
2313                            (u64)outbytes);
2314         }
2315
2316         ip_vs_copy_stats(&kstats, tot_stats);
2317
2318         seq_printf(seq, "  ~ %8LX %8LX %8LX %16LX %16LX\n\n",
2319                    (unsigned long long)kstats.conns,
2320                    (unsigned long long)kstats.inpkts,
2321                    (unsigned long long)kstats.outpkts,
2322                    (unsigned long long)kstats.inbytes,
2323                    (unsigned long long)kstats.outbytes);
2324
2325 /*                ... 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2326         seq_puts(seq,
2327                  "     Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
2328         seq_printf(seq, "    %8LX %8LX %8LX %16LX %16LX\n",
2329                    kstats.cps,
2330                    kstats.inpps,
2331                    kstats.outpps,
2332                    kstats.inbps,
2333                    kstats.outbps);
2334
2335         return 0;
2336 }
2337 #endif
2338
2339 /*
2340  *      Set timeout values for tcp tcpfin udp in the timeout_table.
2341  */
2342 static int ip_vs_set_timeout(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2343 {
2344 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2345         struct ip_vs_proto_data *pd;
2346 #endif
2347
2348         IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2349                   u->tcp_timeout,
2350                   u->tcp_fin_timeout,
2351                   u->udp_timeout);
2352
2353 #ifdef CONFIG_IP_VS_PROTO_TCP
2354         if (u->tcp_timeout < 0 || u->tcp_timeout > (INT_MAX / HZ) ||
2355             u->tcp_fin_timeout < 0 || u->tcp_fin_timeout > (INT_MAX / HZ)) {
2356                 return -EINVAL;
2357         }
2358 #endif
2359
2360 #ifdef CONFIG_IP_VS_PROTO_UDP
2361         if (u->udp_timeout < 0 || u->udp_timeout > (INT_MAX / HZ))
2362                 return -EINVAL;
2363 #endif
2364
2365 #ifdef CONFIG_IP_VS_PROTO_TCP
2366         if (u->tcp_timeout) {
2367                 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2368                 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2369                         = u->tcp_timeout * HZ;
2370         }
2371
2372         if (u->tcp_fin_timeout) {
2373                 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2374                 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2375                         = u->tcp_fin_timeout * HZ;
2376         }
2377 #endif
2378
2379 #ifdef CONFIG_IP_VS_PROTO_UDP
2380         if (u->udp_timeout) {
2381                 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2382                 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2383                         = u->udp_timeout * HZ;
2384         }
2385 #endif
2386         return 0;
2387 }
2388
2389 #define CMDID(cmd)              (cmd - IP_VS_BASE_CTL)
2390
2391 struct ip_vs_svcdest_user {
2392         struct ip_vs_service_user       s;
2393         struct ip_vs_dest_user          d;
2394 };
2395
2396 static const unsigned char set_arglen[CMDID(IP_VS_SO_SET_MAX) + 1] = {
2397         [CMDID(IP_VS_SO_SET_ADD)]         = sizeof(struct ip_vs_service_user),
2398         [CMDID(IP_VS_SO_SET_EDIT)]        = sizeof(struct ip_vs_service_user),
2399         [CMDID(IP_VS_SO_SET_DEL)]         = sizeof(struct ip_vs_service_user),
2400         [CMDID(IP_VS_SO_SET_ADDDEST)]     = sizeof(struct ip_vs_svcdest_user),
2401         [CMDID(IP_VS_SO_SET_DELDEST)]     = sizeof(struct ip_vs_svcdest_user),
2402         [CMDID(IP_VS_SO_SET_EDITDEST)]    = sizeof(struct ip_vs_svcdest_user),
2403         [CMDID(IP_VS_SO_SET_TIMEOUT)]     = sizeof(struct ip_vs_timeout_user),
2404         [CMDID(IP_VS_SO_SET_STARTDAEMON)] = sizeof(struct ip_vs_daemon_user),
2405         [CMDID(IP_VS_SO_SET_STOPDAEMON)]  = sizeof(struct ip_vs_daemon_user),
2406         [CMDID(IP_VS_SO_SET_ZERO)]        = sizeof(struct ip_vs_service_user),
2407 };
2408
2409 union ip_vs_set_arglen {
2410         struct ip_vs_service_user       field_IP_VS_SO_SET_ADD;
2411         struct ip_vs_service_user       field_IP_VS_SO_SET_EDIT;
2412         struct ip_vs_service_user       field_IP_VS_SO_SET_DEL;
2413         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_ADDDEST;
2414         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_DELDEST;
2415         struct ip_vs_svcdest_user       field_IP_VS_SO_SET_EDITDEST;
2416         struct ip_vs_timeout_user       field_IP_VS_SO_SET_TIMEOUT;
2417         struct ip_vs_daemon_user        field_IP_VS_SO_SET_STARTDAEMON;
2418         struct ip_vs_daemon_user        field_IP_VS_SO_SET_STOPDAEMON;
2419         struct ip_vs_service_user       field_IP_VS_SO_SET_ZERO;
2420 };
2421
2422 #define MAX_SET_ARGLEN  sizeof(union ip_vs_set_arglen)
2423
2424 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2425                                   struct ip_vs_service_user *usvc_compat)
2426 {
2427         memset(usvc, 0, sizeof(*usvc));
2428
2429         usvc->af                = AF_INET;
2430         usvc->protocol          = usvc_compat->protocol;
2431         usvc->addr.ip           = usvc_compat->addr;
2432         usvc->port              = usvc_compat->port;
2433         usvc->fwmark            = usvc_compat->fwmark;
2434
2435         /* Deep copy of sched_name is not needed here */
2436         usvc->sched_name        = usvc_compat->sched_name;
2437
2438         usvc->flags             = usvc_compat->flags;
2439         usvc->timeout           = usvc_compat->timeout;
2440         usvc->netmask           = usvc_compat->netmask;
2441 }
2442
2443 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2444                                    struct ip_vs_dest_user *udest_compat)
2445 {
2446         memset(udest, 0, sizeof(*udest));
2447
2448         udest->addr.ip          = udest_compat->addr;
2449         udest->port             = udest_compat->port;
2450         udest->conn_flags       = udest_compat->conn_flags;
2451         udest->weight           = udest_compat->weight;
2452         udest->u_threshold      = udest_compat->u_threshold;
2453         udest->l_threshold      = udest_compat->l_threshold;
2454         udest->af               = AF_INET;
2455         udest->tun_type         = IP_VS_CONN_F_TUNNEL_TYPE_IPIP;
2456 }
2457
2458 static int
2459 do_ip_vs_set_ctl(struct sock *sk, int cmd, sockptr_t ptr, unsigned int len)
2460 {
2461         struct net *net = sock_net(sk);
2462         int ret;
2463         unsigned char arg[MAX_SET_ARGLEN];
2464         struct ip_vs_service_user *usvc_compat;
2465         struct ip_vs_service_user_kern usvc;
2466         struct ip_vs_service *svc;
2467         struct ip_vs_dest_user *udest_compat;
2468         struct ip_vs_dest_user_kern udest;
2469         struct netns_ipvs *ipvs = net_ipvs(net);
2470
2471         BUILD_BUG_ON(sizeof(arg) > 255);
2472         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2473                 return -EPERM;
2474
2475         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2476                 return -EINVAL;
2477         if (len != set_arglen[CMDID(cmd)]) {
2478                 IP_VS_DBG(1, "set_ctl: len %u != %u\n",
2479                           len, set_arglen[CMDID(cmd)]);
2480                 return -EINVAL;
2481         }
2482
2483         if (copy_from_sockptr(arg, ptr, len) != 0)
2484                 return -EFAULT;
2485
2486         /* Handle daemons since they have another lock */
2487         if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2488             cmd == IP_VS_SO_SET_STOPDAEMON) {
2489                 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2490
2491                 if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2492                         struct ipvs_sync_daemon_cfg cfg;
2493
2494                         memset(&cfg, 0, sizeof(cfg));
2495                         ret = -EINVAL;
2496                         if (strscpy(cfg.mcast_ifn, dm->mcast_ifn,
2497                                     sizeof(cfg.mcast_ifn)) <= 0)
2498                                 return ret;
2499                         cfg.syncid = dm->syncid;
2500                         ret = start_sync_thread(ipvs, &cfg, dm->state);
2501                 } else {
2502                         ret = stop_sync_thread(ipvs, dm->state);
2503                 }
2504                 return ret;
2505         }
2506
2507         mutex_lock(&__ip_vs_mutex);
2508         if (cmd == IP_VS_SO_SET_FLUSH) {
2509                 /* Flush the virtual service */
2510                 ret = ip_vs_flush(ipvs, false);
2511                 goto out_unlock;
2512         } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2513                 /* Set timeout values for (tcp tcpfin udp) */
2514                 ret = ip_vs_set_timeout(ipvs, (struct ip_vs_timeout_user *)arg);
2515                 goto out_unlock;
2516         } else if (!len) {
2517                 /* No more commands with len == 0 below */
2518                 ret = -EINVAL;
2519                 goto out_unlock;
2520         }
2521
2522         usvc_compat = (struct ip_vs_service_user *)arg;
2523         udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2524
2525         /* We only use the new structs internally, so copy userspace compat
2526          * structs to extended internal versions */
2527         ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2528         ip_vs_copy_udest_compat(&udest, udest_compat);
2529
2530         if (cmd == IP_VS_SO_SET_ZERO) {
2531                 /* if no service address is set, zero counters in all */
2532                 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2533                         ret = ip_vs_zero_all(ipvs);
2534                         goto out_unlock;
2535                 }
2536         }
2537
2538         if ((cmd == IP_VS_SO_SET_ADD || cmd == IP_VS_SO_SET_EDIT) &&
2539             strnlen(usvc.sched_name, IP_VS_SCHEDNAME_MAXLEN) ==
2540             IP_VS_SCHEDNAME_MAXLEN) {
2541                 ret = -EINVAL;
2542                 goto out_unlock;
2543         }
2544
2545         /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2546         if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2547             usvc.protocol != IPPROTO_SCTP) {
2548                 pr_err("set_ctl: invalid protocol: %d %pI4:%d\n",
2549                        usvc.protocol, &usvc.addr.ip,
2550                        ntohs(usvc.port));
2551                 ret = -EFAULT;
2552                 goto out_unlock;
2553         }
2554
2555         /* Lookup the exact service by <protocol, addr, port> or fwmark */
2556         rcu_read_lock();
2557         if (usvc.fwmark == 0)
2558                 svc = __ip_vs_service_find(ipvs, usvc.af, usvc.protocol,
2559                                            &usvc.addr, usvc.port);
2560         else
2561                 svc = __ip_vs_svc_fwm_find(ipvs, usvc.af, usvc.fwmark);
2562         rcu_read_unlock();
2563
2564         if (cmd != IP_VS_SO_SET_ADD
2565             && (svc == NULL || svc->protocol != usvc.protocol)) {
2566                 ret = -ESRCH;
2567                 goto out_unlock;
2568         }
2569
2570         switch (cmd) {
2571         case IP_VS_SO_SET_ADD:
2572                 if (svc != NULL)
2573                         ret = -EEXIST;
2574                 else
2575                         ret = ip_vs_add_service(ipvs, &usvc, &svc);
2576                 break;
2577         case IP_VS_SO_SET_EDIT:
2578                 ret = ip_vs_edit_service(svc, &usvc);
2579                 break;
2580         case IP_VS_SO_SET_DEL:
2581                 ret = ip_vs_del_service(svc);
2582                 if (!ret)
2583                         goto out_unlock;
2584                 break;
2585         case IP_VS_SO_SET_ZERO:
2586                 ret = ip_vs_zero_service(svc);
2587                 break;
2588         case IP_VS_SO_SET_ADDDEST:
2589                 ret = ip_vs_add_dest(svc, &udest);
2590                 break;
2591         case IP_VS_SO_SET_EDITDEST:
2592                 ret = ip_vs_edit_dest(svc, &udest);
2593                 break;
2594         case IP_VS_SO_SET_DELDEST:
2595                 ret = ip_vs_del_dest(svc, &udest);
2596         }
2597
2598   out_unlock:
2599         mutex_unlock(&__ip_vs_mutex);
2600         return ret;
2601 }
2602
2603
2604 static void
2605 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2606 {
2607         struct ip_vs_scheduler *sched;
2608         struct ip_vs_kstats kstats;
2609         char *sched_name;
2610
2611         sched = rcu_dereference_protected(src->scheduler, 1);
2612         sched_name = sched ? sched->name : "none";
2613         dst->protocol = src->protocol;
2614         dst->addr = src->addr.ip;
2615         dst->port = src->port;
2616         dst->fwmark = src->fwmark;
2617         strscpy(dst->sched_name, sched_name, sizeof(dst->sched_name));
2618         dst->flags = src->flags;
2619         dst->timeout = src->timeout / HZ;
2620         dst->netmask = src->netmask;
2621         dst->num_dests = src->num_dests;
2622         ip_vs_copy_stats(&kstats, &src->stats);
2623         ip_vs_export_stats_user(&dst->stats, &kstats);
2624 }
2625
2626 static inline int
2627 __ip_vs_get_service_entries(struct netns_ipvs *ipvs,
2628                             const struct ip_vs_get_services *get,
2629                             struct ip_vs_get_services __user *uptr)
2630 {
2631         int idx, count=0;
2632         struct ip_vs_service *svc;
2633         struct ip_vs_service_entry entry;
2634         int ret = 0;
2635
2636         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2637                 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2638                         /* Only expose IPv4 entries to old interface */
2639                         if (svc->af != AF_INET || (svc->ipvs != ipvs))
2640                                 continue;
2641
2642                         if (count >= get->num_services)
2643                                 goto out;
2644                         memset(&entry, 0, sizeof(entry));
2645                         ip_vs_copy_service(&entry, svc);
2646                         if (copy_to_user(&uptr->entrytable[count],
2647                                          &entry, sizeof(entry))) {
2648                                 ret = -EFAULT;
2649                                 goto out;
2650                         }
2651                         count++;
2652                 }
2653         }
2654
2655         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2656                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2657                         /* Only expose IPv4 entries to old interface */
2658                         if (svc->af != AF_INET || (svc->ipvs != ipvs))
2659                                 continue;
2660
2661                         if (count >= get->num_services)
2662                                 goto out;
2663                         memset(&entry, 0, sizeof(entry));
2664                         ip_vs_copy_service(&entry, svc);
2665                         if (copy_to_user(&uptr->entrytable[count],
2666                                          &entry, sizeof(entry))) {
2667                                 ret = -EFAULT;
2668                                 goto out;
2669                         }
2670                         count++;
2671                 }
2672         }
2673 out:
2674         return ret;
2675 }
2676
2677 static inline int
2678 __ip_vs_get_dest_entries(struct netns_ipvs *ipvs, const struct ip_vs_get_dests *get,
2679                          struct ip_vs_get_dests __user *uptr)
2680 {
2681         struct ip_vs_service *svc;
2682         union nf_inet_addr addr = { .ip = get->addr };
2683         int ret = 0;
2684
2685         rcu_read_lock();
2686         if (get->fwmark)
2687                 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, get->fwmark);
2688         else
2689                 svc = __ip_vs_service_find(ipvs, AF_INET, get->protocol, &addr,
2690                                            get->port);
2691         rcu_read_unlock();
2692
2693         if (svc) {
2694                 int count = 0;
2695                 struct ip_vs_dest *dest;
2696                 struct ip_vs_dest_entry entry;
2697                 struct ip_vs_kstats kstats;
2698
2699                 memset(&entry, 0, sizeof(entry));
2700                 list_for_each_entry(dest, &svc->destinations, n_list) {
2701                         if (count >= get->num_dests)
2702                                 break;
2703
2704                         /* Cannot expose heterogeneous members via sockopt
2705                          * interface
2706                          */
2707                         if (dest->af != svc->af)
2708                                 continue;
2709
2710                         entry.addr = dest->addr.ip;
2711                         entry.port = dest->port;
2712                         entry.conn_flags = atomic_read(&dest->conn_flags);
2713                         entry.weight = atomic_read(&dest->weight);
2714                         entry.u_threshold = dest->u_threshold;
2715                         entry.l_threshold = dest->l_threshold;
2716                         entry.activeconns = atomic_read(&dest->activeconns);
2717                         entry.inactconns = atomic_read(&dest->inactconns);
2718                         entry.persistconns = atomic_read(&dest->persistconns);
2719                         ip_vs_copy_stats(&kstats, &dest->stats);
2720                         ip_vs_export_stats_user(&entry.stats, &kstats);
2721                         if (copy_to_user(&uptr->entrytable[count],
2722                                          &entry, sizeof(entry))) {
2723                                 ret = -EFAULT;
2724                                 break;
2725                         }
2726                         count++;
2727                 }
2728         } else
2729                 ret = -ESRCH;
2730         return ret;
2731 }
2732
2733 static inline void
2734 __ip_vs_get_timeouts(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2735 {
2736 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2737         struct ip_vs_proto_data *pd;
2738 #endif
2739
2740         memset(u, 0, sizeof (*u));
2741
2742 #ifdef CONFIG_IP_VS_PROTO_TCP
2743         pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2744         u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2745         u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2746 #endif
2747 #ifdef CONFIG_IP_VS_PROTO_UDP
2748         pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2749         u->udp_timeout =
2750                         pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2751 #endif
2752 }
2753
2754 static const unsigned char get_arglen[CMDID(IP_VS_SO_GET_MAX) + 1] = {
2755         [CMDID(IP_VS_SO_GET_VERSION)]  = 64,
2756         [CMDID(IP_VS_SO_GET_INFO)]     = sizeof(struct ip_vs_getinfo),
2757         [CMDID(IP_VS_SO_GET_SERVICES)] = sizeof(struct ip_vs_get_services),
2758         [CMDID(IP_VS_SO_GET_SERVICE)]  = sizeof(struct ip_vs_service_entry),
2759         [CMDID(IP_VS_SO_GET_DESTS)]    = sizeof(struct ip_vs_get_dests),
2760         [CMDID(IP_VS_SO_GET_TIMEOUT)]  = sizeof(struct ip_vs_timeout_user),
2761         [CMDID(IP_VS_SO_GET_DAEMON)]   = 2 * sizeof(struct ip_vs_daemon_user),
2762 };
2763
2764 union ip_vs_get_arglen {
2765         char                            field_IP_VS_SO_GET_VERSION[64];
2766         struct ip_vs_getinfo            field_IP_VS_SO_GET_INFO;
2767         struct ip_vs_get_services       field_IP_VS_SO_GET_SERVICES;
2768         struct ip_vs_service_entry      field_IP_VS_SO_GET_SERVICE;
2769         struct ip_vs_get_dests          field_IP_VS_SO_GET_DESTS;
2770         struct ip_vs_timeout_user       field_IP_VS_SO_GET_TIMEOUT;
2771         struct ip_vs_daemon_user        field_IP_VS_SO_GET_DAEMON[2];
2772 };
2773
2774 #define MAX_GET_ARGLEN  sizeof(union ip_vs_get_arglen)
2775
2776 static int
2777 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2778 {
2779         unsigned char arg[MAX_GET_ARGLEN];
2780         int ret = 0;
2781         unsigned int copylen;
2782         struct net *net = sock_net(sk);
2783         struct netns_ipvs *ipvs = net_ipvs(net);
2784
2785         BUG_ON(!net);
2786         BUILD_BUG_ON(sizeof(arg) > 255);
2787         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2788                 return -EPERM;
2789
2790         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2791                 return -EINVAL;
2792
2793         copylen = get_arglen[CMDID(cmd)];
2794         if (*len < (int) copylen) {
2795                 IP_VS_DBG(1, "get_ctl: len %d < %u\n", *len, copylen);
2796                 return -EINVAL;
2797         }
2798
2799         if (copy_from_user(arg, user, copylen) != 0)
2800                 return -EFAULT;
2801         /*
2802          * Handle daemons first since it has its own locking
2803          */
2804         if (cmd == IP_VS_SO_GET_DAEMON) {
2805                 struct ip_vs_daemon_user d[2];
2806
2807                 memset(&d, 0, sizeof(d));
2808                 mutex_lock(&ipvs->sync_mutex);
2809                 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2810                         d[0].state = IP_VS_STATE_MASTER;
2811                         strscpy(d[0].mcast_ifn, ipvs->mcfg.mcast_ifn,
2812                                 sizeof(d[0].mcast_ifn));
2813                         d[0].syncid = ipvs->mcfg.syncid;
2814                 }
2815                 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2816                         d[1].state = IP_VS_STATE_BACKUP;
2817                         strscpy(d[1].mcast_ifn, ipvs->bcfg.mcast_ifn,
2818                                 sizeof(d[1].mcast_ifn));
2819                         d[1].syncid = ipvs->bcfg.syncid;
2820                 }
2821                 if (copy_to_user(user, &d, sizeof(d)) != 0)
2822                         ret = -EFAULT;
2823                 mutex_unlock(&ipvs->sync_mutex);
2824                 return ret;
2825         }
2826
2827         mutex_lock(&__ip_vs_mutex);
2828         switch (cmd) {
2829         case IP_VS_SO_GET_VERSION:
2830         {
2831                 char buf[64];
2832
2833                 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2834                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2835                 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2836                         ret = -EFAULT;
2837                         goto out;
2838                 }
2839                 *len = strlen(buf)+1;
2840         }
2841         break;
2842
2843         case IP_VS_SO_GET_INFO:
2844         {
2845                 struct ip_vs_getinfo info;
2846                 info.version = IP_VS_VERSION_CODE;
2847                 info.size = ip_vs_conn_tab_size;
2848                 info.num_services = ipvs->num_services;
2849                 if (copy_to_user(user, &info, sizeof(info)) != 0)
2850                         ret = -EFAULT;
2851         }
2852         break;
2853
2854         case IP_VS_SO_GET_SERVICES:
2855         {
2856                 struct ip_vs_get_services *get;
2857                 int size;
2858
2859                 get = (struct ip_vs_get_services *)arg;
2860                 size = struct_size(get, entrytable, get->num_services);
2861                 if (*len != size) {
2862                         pr_err("length: %u != %u\n", *len, size);
2863                         ret = -EINVAL;
2864                         goto out;
2865                 }
2866                 ret = __ip_vs_get_service_entries(ipvs, get, user);
2867         }
2868         break;
2869
2870         case IP_VS_SO_GET_SERVICE:
2871         {
2872                 struct ip_vs_service_entry *entry;
2873                 struct ip_vs_service *svc;
2874                 union nf_inet_addr addr;
2875
2876                 entry = (struct ip_vs_service_entry *)arg;
2877                 addr.ip = entry->addr;
2878                 rcu_read_lock();
2879                 if (entry->fwmark)
2880                         svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, entry->fwmark);
2881                 else
2882                         svc = __ip_vs_service_find(ipvs, AF_INET,
2883                                                    entry->protocol, &addr,
2884                                                    entry->port);
2885                 rcu_read_unlock();
2886                 if (svc) {
2887                         ip_vs_copy_service(entry, svc);
2888                         if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2889                                 ret = -EFAULT;
2890                 } else
2891                         ret = -ESRCH;
2892         }
2893         break;
2894
2895         case IP_VS_SO_GET_DESTS:
2896         {
2897                 struct ip_vs_get_dests *get;
2898                 int size;
2899
2900                 get = (struct ip_vs_get_dests *)arg;
2901                 size = struct_size(get, entrytable, get->num_dests);
2902                 if (*len != size) {
2903                         pr_err("length: %u != %u\n", *len, size);
2904                         ret = -EINVAL;
2905                         goto out;
2906                 }
2907                 ret = __ip_vs_get_dest_entries(ipvs, get, user);
2908         }
2909         break;
2910
2911         case IP_VS_SO_GET_TIMEOUT:
2912         {
2913                 struct ip_vs_timeout_user t;
2914
2915                 __ip_vs_get_timeouts(ipvs, &t);
2916                 if (copy_to_user(user, &t, sizeof(t)) != 0)
2917                         ret = -EFAULT;
2918         }
2919         break;
2920
2921         default:
2922                 ret = -EINVAL;
2923         }
2924
2925 out:
2926         mutex_unlock(&__ip_vs_mutex);
2927         return ret;
2928 }
2929
2930
2931 static struct nf_sockopt_ops ip_vs_sockopts = {
2932         .pf             = PF_INET,
2933         .set_optmin     = IP_VS_BASE_CTL,
2934         .set_optmax     = IP_VS_SO_SET_MAX+1,
2935         .set            = do_ip_vs_set_ctl,
2936         .get_optmin     = IP_VS_BASE_CTL,
2937         .get_optmax     = IP_VS_SO_GET_MAX+1,
2938         .get            = do_ip_vs_get_ctl,
2939         .owner          = THIS_MODULE,
2940 };
2941
2942 /*
2943  * Generic Netlink interface
2944  */
2945
2946 /* IPVS genetlink family */
2947 static struct genl_family ip_vs_genl_family;
2948
2949 /* Policy used for first-level command attributes */
2950 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2951         [IPVS_CMD_ATTR_SERVICE]         = { .type = NLA_NESTED },
2952         [IPVS_CMD_ATTR_DEST]            = { .type = NLA_NESTED },
2953         [IPVS_CMD_ATTR_DAEMON]          = { .type = NLA_NESTED },
2954         [IPVS_CMD_ATTR_TIMEOUT_TCP]     = { .type = NLA_U32 },
2955         [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2956         [IPVS_CMD_ATTR_TIMEOUT_UDP]     = { .type = NLA_U32 },
2957 };
2958
2959 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2960 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2961         [IPVS_DAEMON_ATTR_STATE]        = { .type = NLA_U32 },
2962         [IPVS_DAEMON_ATTR_MCAST_IFN]    = { .type = NLA_NUL_STRING,
2963                                             .len = IP_VS_IFNAME_MAXLEN - 1 },
2964         [IPVS_DAEMON_ATTR_SYNC_ID]      = { .type = NLA_U32 },
2965         [IPVS_DAEMON_ATTR_SYNC_MAXLEN]  = { .type = NLA_U16 },
2966         [IPVS_DAEMON_ATTR_MCAST_GROUP]  = { .type = NLA_U32 },
2967         [IPVS_DAEMON_ATTR_MCAST_GROUP6] = { .len = sizeof(struct in6_addr) },
2968         [IPVS_DAEMON_ATTR_MCAST_PORT]   = { .type = NLA_U16 },
2969         [IPVS_DAEMON_ATTR_MCAST_TTL]    = { .type = NLA_U8 },
2970 };
2971
2972 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2973 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2974         [IPVS_SVC_ATTR_AF]              = { .type = NLA_U16 },
2975         [IPVS_SVC_ATTR_PROTOCOL]        = { .type = NLA_U16 },
2976         [IPVS_SVC_ATTR_ADDR]            = { .type = NLA_BINARY,
2977                                             .len = sizeof(union nf_inet_addr) },
2978         [IPVS_SVC_ATTR_PORT]            = { .type = NLA_U16 },
2979         [IPVS_SVC_ATTR_FWMARK]          = { .type = NLA_U32 },
2980         [IPVS_SVC_ATTR_SCHED_NAME]      = { .type = NLA_NUL_STRING,
2981                                             .len = IP_VS_SCHEDNAME_MAXLEN - 1 },
2982         [IPVS_SVC_ATTR_PE_NAME]         = { .type = NLA_NUL_STRING,
2983                                             .len = IP_VS_PENAME_MAXLEN },
2984         [IPVS_SVC_ATTR_FLAGS]           = { .type = NLA_BINARY,
2985                                             .len = sizeof(struct ip_vs_flags) },
2986         [IPVS_SVC_ATTR_TIMEOUT]         = { .type = NLA_U32 },
2987         [IPVS_SVC_ATTR_NETMASK]         = { .type = NLA_U32 },
2988         [IPVS_SVC_ATTR_STATS]           = { .type = NLA_NESTED },
2989 };
2990
2991 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2992 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2993         [IPVS_DEST_ATTR_ADDR]           = { .type = NLA_BINARY,
2994                                             .len = sizeof(union nf_inet_addr) },
2995         [IPVS_DEST_ATTR_PORT]           = { .type = NLA_U16 },
2996         [IPVS_DEST_ATTR_FWD_METHOD]     = { .type = NLA_U32 },
2997         [IPVS_DEST_ATTR_WEIGHT]         = { .type = NLA_U32 },
2998         [IPVS_DEST_ATTR_U_THRESH]       = { .type = NLA_U32 },
2999         [IPVS_DEST_ATTR_L_THRESH]       = { .type = NLA_U32 },
3000         [IPVS_DEST_ATTR_ACTIVE_CONNS]   = { .type = NLA_U32 },
3001         [IPVS_DEST_ATTR_INACT_CONNS]    = { .type = NLA_U32 },
3002         [IPVS_DEST_ATTR_PERSIST_CONNS]  = { .type = NLA_U32 },
3003         [IPVS_DEST_ATTR_STATS]          = { .type = NLA_NESTED },
3004         [IPVS_DEST_ATTR_ADDR_FAMILY]    = { .type = NLA_U16 },
3005         [IPVS_DEST_ATTR_TUN_TYPE]       = { .type = NLA_U8 },
3006         [IPVS_DEST_ATTR_TUN_PORT]       = { .type = NLA_U16 },
3007         [IPVS_DEST_ATTR_TUN_FLAGS]      = { .type = NLA_U16 },
3008 };
3009
3010 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
3011                                  struct ip_vs_kstats *kstats)
3012 {
3013         struct nlattr *nl_stats = nla_nest_start_noflag(skb, container_type);
3014
3015         if (!nl_stats)
3016                 return -EMSGSIZE;
3017
3018         if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, (u32)kstats->conns) ||
3019             nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, (u32)kstats->inpkts) ||
3020             nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, (u32)kstats->outpkts) ||
3021             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes,
3022                               IPVS_STATS_ATTR_PAD) ||
3023             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes,
3024                               IPVS_STATS_ATTR_PAD) ||
3025             nla_put_u32(skb, IPVS_STATS_ATTR_CPS, (u32)kstats->cps) ||
3026             nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, (u32)kstats->inpps) ||
3027             nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, (u32)kstats->outpps) ||
3028             nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, (u32)kstats->inbps) ||
3029             nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, (u32)kstats->outbps))
3030                 goto nla_put_failure;
3031         nla_nest_end(skb, nl_stats);
3032
3033         return 0;
3034
3035 nla_put_failure:
3036         nla_nest_cancel(skb, nl_stats);
3037         return -EMSGSIZE;
3038 }
3039
3040 static int ip_vs_genl_fill_stats64(struct sk_buff *skb, int container_type,
3041                                    struct ip_vs_kstats *kstats)
3042 {
3043         struct nlattr *nl_stats = nla_nest_start_noflag(skb, container_type);
3044
3045         if (!nl_stats)
3046                 return -EMSGSIZE;
3047
3048         if (nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CONNS, kstats->conns,
3049                               IPVS_STATS_ATTR_PAD) ||
3050             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPKTS, kstats->inpkts,
3051                               IPVS_STATS_ATTR_PAD) ||
3052             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPKTS, kstats->outpkts,
3053                               IPVS_STATS_ATTR_PAD) ||
3054             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes,
3055                               IPVS_STATS_ATTR_PAD) ||
3056             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes,
3057                               IPVS_STATS_ATTR_PAD) ||
3058             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CPS, kstats->cps,
3059                               IPVS_STATS_ATTR_PAD) ||
3060             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPPS, kstats->inpps,
3061                               IPVS_STATS_ATTR_PAD) ||
3062             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPPS, kstats->outpps,
3063                               IPVS_STATS_ATTR_PAD) ||
3064             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBPS, kstats->inbps,
3065                               IPVS_STATS_ATTR_PAD) ||
3066             nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBPS, kstats->outbps,
3067                               IPVS_STATS_ATTR_PAD))
3068                 goto nla_put_failure;
3069         nla_nest_end(skb, nl_stats);
3070
3071         return 0;
3072
3073 nla_put_failure:
3074         nla_nest_cancel(skb, nl_stats);
3075         return -EMSGSIZE;
3076 }
3077
3078 static int ip_vs_genl_fill_service(struct sk_buff *skb,
3079                                    struct ip_vs_service *svc)
3080 {
3081         struct ip_vs_scheduler *sched;
3082         struct ip_vs_pe *pe;
3083         struct nlattr *nl_service;
3084         struct ip_vs_flags flags = { .flags = svc->flags,
3085                                      .mask = ~0 };
3086         struct ip_vs_kstats kstats;
3087         char *sched_name;
3088
3089         nl_service = nla_nest_start_noflag(skb, IPVS_CMD_ATTR_SERVICE);
3090         if (!nl_service)
3091                 return -EMSGSIZE;
3092
3093         if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
3094                 goto nla_put_failure;
3095         if (svc->fwmark) {
3096                 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
3097                         goto nla_put_failure;
3098         } else {
3099                 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
3100                     nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
3101                     nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
3102                         goto nla_put_failure;
3103         }
3104
3105         sched = rcu_dereference_protected(svc->scheduler, 1);
3106         sched_name = sched ? sched->name : "none";
3107         pe = rcu_dereference_protected(svc->pe, 1);
3108         if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched_name) ||
3109             (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
3110             nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
3111             nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
3112             nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
3113                 goto nla_put_failure;
3114         ip_vs_copy_stats(&kstats, &svc->stats);
3115         if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &kstats))
3116                 goto nla_put_failure;
3117         if (ip_vs_genl_fill_stats64(skb, IPVS_SVC_ATTR_STATS64, &kstats))
3118                 goto nla_put_failure;
3119
3120         nla_nest_end(skb, nl_service);
3121
3122         return 0;
3123
3124 nla_put_failure:
3125         nla_nest_cancel(skb, nl_service);
3126         return -EMSGSIZE;
3127 }
3128
3129 static int ip_vs_genl_dump_service(struct sk_buff *skb,
3130                                    struct ip_vs_service *svc,
3131                                    struct netlink_callback *cb)
3132 {
3133         void *hdr;
3134
3135         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3136                           &ip_vs_genl_family, NLM_F_MULTI,
3137                           IPVS_CMD_NEW_SERVICE);
3138         if (!hdr)
3139                 return -EMSGSIZE;
3140
3141         if (ip_vs_genl_fill_service(skb, svc) < 0)
3142                 goto nla_put_failure;
3143
3144         genlmsg_end(skb, hdr);
3145         return 0;
3146
3147 nla_put_failure:
3148         genlmsg_cancel(skb, hdr);
3149         return -EMSGSIZE;
3150 }
3151
3152 static int ip_vs_genl_dump_services(struct sk_buff *skb,
3153                                     struct netlink_callback *cb)
3154 {
3155         int idx = 0, i;
3156         int start = cb->args[0];
3157         struct ip_vs_service *svc;
3158         struct net *net = sock_net(skb->sk);
3159         struct netns_ipvs *ipvs = net_ipvs(net);
3160
3161         mutex_lock(&__ip_vs_mutex);
3162         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3163                 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
3164                         if (++idx <= start || (svc->ipvs != ipvs))
3165                                 continue;
3166                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3167                                 idx--;
3168                                 goto nla_put_failure;
3169                         }
3170                 }
3171         }
3172
3173         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3174                 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
3175                         if (++idx <= start || (svc->ipvs != ipvs))
3176                                 continue;
3177                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3178                                 idx--;
3179                                 goto nla_put_failure;
3180                         }
3181                 }
3182         }
3183
3184 nla_put_failure:
3185         mutex_unlock(&__ip_vs_mutex);
3186         cb->args[0] = idx;
3187
3188         return skb->len;
3189 }
3190
3191 static bool ip_vs_is_af_valid(int af)
3192 {
3193         if (af == AF_INET)
3194                 return true;
3195 #ifdef CONFIG_IP_VS_IPV6
3196         if (af == AF_INET6 && ipv6_mod_enabled())
3197                 return true;
3198 #endif
3199         return false;
3200 }
3201
3202 static int ip_vs_genl_parse_service(struct netns_ipvs *ipvs,
3203                                     struct ip_vs_service_user_kern *usvc,
3204                                     struct nlattr *nla, bool full_entry,
3205                                     struct ip_vs_service **ret_svc)
3206 {
3207         struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
3208         struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
3209         struct ip_vs_service *svc;
3210
3211         /* Parse mandatory identifying service fields first */
3212         if (nla == NULL ||
3213             nla_parse_nested_deprecated(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy, NULL))
3214                 return -EINVAL;
3215
3216         nla_af          = attrs[IPVS_SVC_ATTR_AF];
3217         nla_protocol    = attrs[IPVS_SVC_ATTR_PROTOCOL];
3218         nla_addr        = attrs[IPVS_SVC_ATTR_ADDR];
3219         nla_port        = attrs[IPVS_SVC_ATTR_PORT];
3220         nla_fwmark      = attrs[IPVS_SVC_ATTR_FWMARK];
3221
3222         if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3223                 return -EINVAL;
3224
3225         memset(usvc, 0, sizeof(*usvc));
3226
3227         usvc->af = nla_get_u16(nla_af);
3228         if (!ip_vs_is_af_valid(usvc->af))
3229                 return -EAFNOSUPPORT;
3230
3231         if (nla_fwmark) {
3232                 usvc->protocol = IPPROTO_TCP;
3233                 usvc->fwmark = nla_get_u32(nla_fwmark);
3234         } else {
3235                 usvc->protocol = nla_get_u16(nla_protocol);
3236                 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3237                 usvc->port = nla_get_be16(nla_port);
3238                 usvc->fwmark = 0;
3239         }
3240
3241         rcu_read_lock();
3242         if (usvc->fwmark)
3243                 svc = __ip_vs_svc_fwm_find(ipvs, usvc->af, usvc->fwmark);
3244         else
3245                 svc = __ip_vs_service_find(ipvs, usvc->af, usvc->protocol,
3246                                            &usvc->addr, usvc->port);
3247         rcu_read_unlock();
3248         *ret_svc = svc;
3249
3250         /* If a full entry was requested, check for the additional fields */
3251         if (full_entry) {
3252                 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3253                               *nla_netmask;
3254                 struct ip_vs_flags flags;
3255
3256                 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3257                 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3258                 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3259                 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3260                 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3261
3262                 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3263                         return -EINVAL;
3264
3265                 nla_memcpy(&flags, nla_flags, sizeof(flags));
3266
3267                 /* prefill flags from service if it already exists */
3268                 if (svc)
3269                         usvc->flags = svc->flags;
3270
3271                 /* set new flags from userland */
3272                 usvc->flags = (usvc->flags & ~flags.mask) |
3273                               (flags.flags & flags.mask);
3274                 usvc->sched_name = nla_data(nla_sched);
3275                 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3276                 usvc->timeout = nla_get_u32(nla_timeout);
3277                 usvc->netmask = nla_get_be32(nla_netmask);
3278         }
3279
3280         return 0;
3281 }
3282
3283 static struct ip_vs_service *ip_vs_genl_find_service(struct netns_ipvs *ipvs,
3284                                                      struct nlattr *nla)
3285 {
3286         struct ip_vs_service_user_kern usvc;
3287         struct ip_vs_service *svc;
3288         int ret;
3289
3290         ret = ip_vs_genl_parse_service(ipvs, &usvc, nla, false, &svc);
3291         return ret ? ERR_PTR(ret) : svc;
3292 }
3293
3294 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3295 {
3296         struct nlattr *nl_dest;
3297         struct ip_vs_kstats kstats;
3298
3299         nl_dest = nla_nest_start_noflag(skb, IPVS_CMD_ATTR_DEST);
3300         if (!nl_dest)
3301                 return -EMSGSIZE;
3302
3303         if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3304             nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3305             nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3306                         (atomic_read(&dest->conn_flags) &
3307                          IP_VS_CONN_F_FWD_MASK)) ||
3308             nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3309                         atomic_read(&dest->weight)) ||
3310             nla_put_u8(skb, IPVS_DEST_ATTR_TUN_TYPE,
3311                        dest->tun_type) ||
3312             nla_put_be16(skb, IPVS_DEST_ATTR_TUN_PORT,
3313                          dest->tun_port) ||
3314             nla_put_u16(skb, IPVS_DEST_ATTR_TUN_FLAGS,
3315                         dest->tun_flags) ||
3316             nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3317             nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3318             nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3319                         atomic_read(&dest->activeconns)) ||
3320             nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3321                         atomic_read(&dest->inactconns)) ||
3322             nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3323                         atomic_read(&dest->persistconns)) ||
3324             nla_put_u16(skb, IPVS_DEST_ATTR_ADDR_FAMILY, dest->af))
3325                 goto nla_put_failure;
3326         ip_vs_copy_stats(&kstats, &dest->stats);
3327         if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &kstats))
3328                 goto nla_put_failure;
3329         if (ip_vs_genl_fill_stats64(skb, IPVS_DEST_ATTR_STATS64, &kstats))
3330                 goto nla_put_failure;
3331
3332         nla_nest_end(skb, nl_dest);
3333
3334         return 0;
3335
3336 nla_put_failure:
3337         nla_nest_cancel(skb, nl_dest);
3338         return -EMSGSIZE;
3339 }
3340
3341 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3342                                 struct netlink_callback *cb)
3343 {
3344         void *hdr;
3345
3346         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3347                           &ip_vs_genl_family, NLM_F_MULTI,
3348                           IPVS_CMD_NEW_DEST);
3349         if (!hdr)
3350                 return -EMSGSIZE;
3351
3352         if (ip_vs_genl_fill_dest(skb, dest) < 0)
3353                 goto nla_put_failure;
3354
3355         genlmsg_end(skb, hdr);
3356         return 0;
3357
3358 nla_put_failure:
3359         genlmsg_cancel(skb, hdr);
3360         return -EMSGSIZE;
3361 }
3362
3363 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3364                                  struct netlink_callback *cb)
3365 {
3366         int idx = 0;
3367         int start = cb->args[0];
3368         struct ip_vs_service *svc;
3369         struct ip_vs_dest *dest;
3370         struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3371         struct net *net = sock_net(skb->sk);
3372         struct netns_ipvs *ipvs = net_ipvs(net);
3373
3374         mutex_lock(&__ip_vs_mutex);
3375
3376         /* Try to find the service for which to dump destinations */
3377         if (nlmsg_parse_deprecated(cb->nlh, GENL_HDRLEN, attrs, IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy, cb->extack))
3378                 goto out_err;
3379
3380
3381         svc = ip_vs_genl_find_service(ipvs, attrs[IPVS_CMD_ATTR_SERVICE]);
3382         if (IS_ERR_OR_NULL(svc))
3383                 goto out_err;
3384
3385         /* Dump the destinations */
3386         list_for_each_entry(dest, &svc->destinations, n_list) {
3387                 if (++idx <= start)
3388                         continue;
3389                 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3390                         idx--;
3391                         goto nla_put_failure;
3392                 }
3393         }
3394
3395 nla_put_failure:
3396         cb->args[0] = idx;
3397
3398 out_err:
3399         mutex_unlock(&__ip_vs_mutex);
3400
3401         return skb->len;
3402 }
3403
3404 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3405                                  struct nlattr *nla, bool full_entry)
3406 {
3407         struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3408         struct nlattr *nla_addr, *nla_port;
3409         struct nlattr *nla_addr_family;
3410
3411         /* Parse mandatory identifying destination fields first */
3412         if (nla == NULL ||
3413             nla_parse_nested_deprecated(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy, NULL))
3414                 return -EINVAL;
3415
3416         nla_addr        = attrs[IPVS_DEST_ATTR_ADDR];
3417         nla_port        = attrs[IPVS_DEST_ATTR_PORT];
3418         nla_addr_family = attrs[IPVS_DEST_ATTR_ADDR_FAMILY];
3419
3420         if (!(nla_addr && nla_port))
3421                 return -EINVAL;
3422
3423         memset(udest, 0, sizeof(*udest));
3424
3425         nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3426         udest->port = nla_get_be16(nla_port);
3427
3428         if (nla_addr_family)
3429                 udest->af = nla_get_u16(nla_addr_family);
3430         else
3431                 udest->af = 0;
3432
3433         /* If a full entry was requested, check for the additional fields */
3434         if (full_entry) {
3435                 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3436                               *nla_l_thresh, *nla_tun_type, *nla_tun_port,
3437                               *nla_tun_flags;
3438
3439                 nla_fwd         = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3440                 nla_weight      = attrs[IPVS_DEST_ATTR_WEIGHT];
3441                 nla_u_thresh    = attrs[IPVS_DEST_ATTR_U_THRESH];
3442                 nla_l_thresh    = attrs[IPVS_DEST_ATTR_L_THRESH];
3443                 nla_tun_type    = attrs[IPVS_DEST_ATTR_TUN_TYPE];
3444                 nla_tun_port    = attrs[IPVS_DEST_ATTR_TUN_PORT];
3445                 nla_tun_flags   = attrs[IPVS_DEST_ATTR_TUN_FLAGS];
3446
3447                 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3448                         return -EINVAL;
3449
3450                 udest->conn_flags = nla_get_u32(nla_fwd)
3451                                     & IP_VS_CONN_F_FWD_MASK;
3452                 udest->weight = nla_get_u32(nla_weight);
3453                 udest->u_threshold = nla_get_u32(nla_u_thresh);
3454                 udest->l_threshold = nla_get_u32(nla_l_thresh);
3455
3456                 if (nla_tun_type)
3457                         udest->tun_type = nla_get_u8(nla_tun_type);
3458
3459                 if (nla_tun_port)
3460                         udest->tun_port = nla_get_be16(nla_tun_port);
3461
3462                 if (nla_tun_flags)
3463                         udest->tun_flags = nla_get_u16(nla_tun_flags);
3464         }
3465
3466         return 0;
3467 }
3468
3469 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3470                                   struct ipvs_sync_daemon_cfg *c)
3471 {
3472         struct nlattr *nl_daemon;
3473
3474         nl_daemon = nla_nest_start_noflag(skb, IPVS_CMD_ATTR_DAEMON);
3475         if (!nl_daemon)
3476                 return -EMSGSIZE;
3477
3478         if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3479             nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, c->mcast_ifn) ||
3480             nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, c->syncid) ||
3481             nla_put_u16(skb, IPVS_DAEMON_ATTR_SYNC_MAXLEN, c->sync_maxlen) ||
3482             nla_put_u16(skb, IPVS_DAEMON_ATTR_MCAST_PORT, c->mcast_port) ||
3483             nla_put_u8(skb, IPVS_DAEMON_ATTR_MCAST_TTL, c->mcast_ttl))
3484                 goto nla_put_failure;
3485 #ifdef CONFIG_IP_VS_IPV6
3486         if (c->mcast_af == AF_INET6) {
3487                 if (nla_put_in6_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP6,
3488                                      &c->mcast_group.in6))
3489                         goto nla_put_failure;
3490         } else
3491 #endif
3492                 if (c->mcast_af == AF_INET &&
3493                     nla_put_in_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP,
3494                                     c->mcast_group.ip))
3495                         goto nla_put_failure;
3496         nla_nest_end(skb, nl_daemon);
3497
3498         return 0;
3499
3500 nla_put_failure:
3501         nla_nest_cancel(skb, nl_daemon);
3502         return -EMSGSIZE;
3503 }
3504
3505 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3506                                   struct ipvs_sync_daemon_cfg *c,
3507                                   struct netlink_callback *cb)
3508 {
3509         void *hdr;
3510         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3511                           &ip_vs_genl_family, NLM_F_MULTI,
3512                           IPVS_CMD_NEW_DAEMON);
3513         if (!hdr)
3514                 return -EMSGSIZE;
3515
3516         if (ip_vs_genl_fill_daemon(skb, state, c))
3517                 goto nla_put_failure;
3518
3519         genlmsg_end(skb, hdr);
3520         return 0;
3521
3522 nla_put_failure:
3523         genlmsg_cancel(skb, hdr);
3524         return -EMSGSIZE;
3525 }
3526
3527 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3528                                    struct netlink_callback *cb)
3529 {
3530         struct net *net = sock_net(skb->sk);
3531         struct netns_ipvs *ipvs = net_ipvs(net);
3532
3533         mutex_lock(&ipvs->sync_mutex);
3534         if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3535                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3536                                            &ipvs->mcfg, cb) < 0)
3537                         goto nla_put_failure;
3538
3539                 cb->args[0] = 1;
3540         }
3541
3542         if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3543                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3544                                            &ipvs->bcfg, cb) < 0)
3545                         goto nla_put_failure;
3546
3547                 cb->args[1] = 1;
3548         }
3549
3550 nla_put_failure:
3551         mutex_unlock(&ipvs->sync_mutex);
3552
3553         return skb->len;
3554 }
3555
3556 static int ip_vs_genl_new_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3557 {
3558         struct ipvs_sync_daemon_cfg c;
3559         struct nlattr *a;
3560         int ret;
3561
3562         memset(&c, 0, sizeof(c));
3563         if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3564               attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3565               attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3566                 return -EINVAL;
3567         strscpy(c.mcast_ifn, nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3568                 sizeof(c.mcast_ifn));
3569         c.syncid = nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]);
3570
3571         a = attrs[IPVS_DAEMON_ATTR_SYNC_MAXLEN];
3572         if (a)
3573                 c.sync_maxlen = nla_get_u16(a);
3574
3575         a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP];
3576         if (a) {
3577                 c.mcast_af = AF_INET;
3578                 c.mcast_group.ip = nla_get_in_addr(a);
3579                 if (!ipv4_is_multicast(c.mcast_group.ip))
3580                         return -EINVAL;
3581         } else {
3582                 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP6];
3583                 if (a) {
3584 #ifdef CONFIG_IP_VS_IPV6
3585                         int addr_type;
3586
3587                         c.mcast_af = AF_INET6;
3588                         c.mcast_group.in6 = nla_get_in6_addr(a);
3589                         addr_type = ipv6_addr_type(&c.mcast_group.in6);
3590                         if (!(addr_type & IPV6_ADDR_MULTICAST))
3591                                 return -EINVAL;
3592 #else
3593                         return -EAFNOSUPPORT;
3594 #endif
3595                 }
3596         }
3597
3598         a = attrs[IPVS_DAEMON_ATTR_MCAST_PORT];
3599         if (a)
3600                 c.mcast_port = nla_get_u16(a);
3601
3602         a = attrs[IPVS_DAEMON_ATTR_MCAST_TTL];
3603         if (a)
3604                 c.mcast_ttl = nla_get_u8(a);
3605
3606         /* The synchronization protocol is incompatible with mixed family
3607          * services
3608          */
3609         if (ipvs->mixed_address_family_dests > 0)
3610                 return -EINVAL;
3611
3612         ret = start_sync_thread(ipvs, &c,
3613                                 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3614         return ret;
3615 }
3616
3617 static int ip_vs_genl_del_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3618 {
3619         int ret;
3620
3621         if (!attrs[IPVS_DAEMON_ATTR_STATE])
3622                 return -EINVAL;
3623
3624         ret = stop_sync_thread(ipvs,
3625                                nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3626         return ret;
3627 }
3628
3629 static int ip_vs_genl_set_config(struct netns_ipvs *ipvs, struct nlattr **attrs)
3630 {
3631         struct ip_vs_timeout_user t;
3632
3633         __ip_vs_get_timeouts(ipvs, &t);
3634
3635         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3636                 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3637
3638         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3639                 t.tcp_fin_timeout =
3640                         nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3641
3642         if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3643                 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3644
3645         return ip_vs_set_timeout(ipvs, &t);
3646 }
3647
3648 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3649 {
3650         int ret = -EINVAL, cmd;
3651         struct net *net = sock_net(skb->sk);
3652         struct netns_ipvs *ipvs = net_ipvs(net);
3653
3654         cmd = info->genlhdr->cmd;
3655
3656         if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3657                 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3658
3659                 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3660                     nla_parse_nested_deprecated(daemon_attrs, IPVS_DAEMON_ATTR_MAX, info->attrs[IPVS_CMD_ATTR_DAEMON], ip_vs_daemon_policy, info->extack))
3661                         goto out;
3662
3663                 if (cmd == IPVS_CMD_NEW_DAEMON)
3664                         ret = ip_vs_genl_new_daemon(ipvs, daemon_attrs);
3665                 else
3666                         ret = ip_vs_genl_del_daemon(ipvs, daemon_attrs);
3667         }
3668
3669 out:
3670         return ret;
3671 }
3672
3673 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3674 {
3675         bool need_full_svc = false, need_full_dest = false;
3676         struct ip_vs_service *svc = NULL;
3677         struct ip_vs_service_user_kern usvc;
3678         struct ip_vs_dest_user_kern udest;
3679         int ret = 0, cmd;
3680         struct net *net = sock_net(skb->sk);
3681         struct netns_ipvs *ipvs = net_ipvs(net);
3682
3683         cmd = info->genlhdr->cmd;
3684
3685         mutex_lock(&__ip_vs_mutex);
3686
3687         if (cmd == IPVS_CMD_FLUSH) {
3688                 ret = ip_vs_flush(ipvs, false);
3689                 goto out;
3690         } else if (cmd == IPVS_CMD_SET_CONFIG) {
3691                 ret = ip_vs_genl_set_config(ipvs, info->attrs);
3692                 goto out;
3693         } else if (cmd == IPVS_CMD_ZERO &&
3694                    !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3695                 ret = ip_vs_zero_all(ipvs);
3696                 goto out;
3697         }
3698
3699         /* All following commands require a service argument, so check if we
3700          * received a valid one. We need a full service specification when
3701          * adding / editing a service. Only identifying members otherwise. */
3702         if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3703                 need_full_svc = true;
3704
3705         ret = ip_vs_genl_parse_service(ipvs, &usvc,
3706                                        info->attrs[IPVS_CMD_ATTR_SERVICE],
3707                                        need_full_svc, &svc);
3708         if (ret)
3709                 goto out;
3710
3711         /* Unless we're adding a new service, the service must already exist */
3712         if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3713                 ret = -ESRCH;
3714                 goto out;
3715         }
3716
3717         /* Destination commands require a valid destination argument. For
3718          * adding / editing a destination, we need a full destination
3719          * specification. */
3720         if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3721             cmd == IPVS_CMD_DEL_DEST) {
3722                 if (cmd != IPVS_CMD_DEL_DEST)
3723                         need_full_dest = true;
3724
3725                 ret = ip_vs_genl_parse_dest(&udest,
3726                                             info->attrs[IPVS_CMD_ATTR_DEST],
3727                                             need_full_dest);
3728                 if (ret)
3729                         goto out;
3730
3731                 /* Old protocols did not allow the user to specify address
3732                  * family, so we set it to zero instead.  We also didn't
3733                  * allow heterogeneous pools in the old code, so it's safe
3734                  * to assume that this will have the same address family as
3735                  * the service.
3736                  */
3737                 if (udest.af == 0)
3738                         udest.af = svc->af;
3739
3740                 if (!ip_vs_is_af_valid(udest.af)) {
3741                         ret = -EAFNOSUPPORT;
3742                         goto out;
3743                 }
3744
3745                 if (udest.af != svc->af && cmd != IPVS_CMD_DEL_DEST) {
3746                         /* The synchronization protocol is incompatible
3747                          * with mixed family services
3748                          */
3749                         if (ipvs->sync_state) {
3750                                 ret = -EINVAL;
3751                                 goto out;
3752                         }
3753
3754                         /* Which connection types do we support? */
3755                         switch (udest.conn_flags) {
3756                         case IP_VS_CONN_F_TUNNEL:
3757                                 /* We are able to forward this */
3758                                 break;
3759                         default:
3760                                 ret = -EINVAL;
3761                                 goto out;
3762                         }
3763                 }
3764         }
3765
3766         switch (cmd) {
3767         case IPVS_CMD_NEW_SERVICE:
3768                 if (svc == NULL)
3769                         ret = ip_vs_add_service(ipvs, &usvc, &svc);
3770                 else
3771                         ret = -EEXIST;
3772                 break;
3773         case IPVS_CMD_SET_SERVICE:
3774                 ret = ip_vs_edit_service(svc, &usvc);
3775                 break;
3776         case IPVS_CMD_DEL_SERVICE:
3777                 ret = ip_vs_del_service(svc);
3778                 /* do not use svc, it can be freed */
3779                 break;
3780         case IPVS_CMD_NEW_DEST:
3781                 ret = ip_vs_add_dest(svc, &udest);
3782                 break;
3783         case IPVS_CMD_SET_DEST:
3784                 ret = ip_vs_edit_dest(svc, &udest);
3785                 break;
3786         case IPVS_CMD_DEL_DEST:
3787                 ret = ip_vs_del_dest(svc, &udest);
3788                 break;
3789         case IPVS_CMD_ZERO:
3790                 ret = ip_vs_zero_service(svc);
3791                 break;
3792         default:
3793                 ret = -EINVAL;
3794         }
3795
3796 out:
3797         mutex_unlock(&__ip_vs_mutex);
3798
3799         return ret;
3800 }
3801
3802 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3803 {
3804         struct sk_buff *msg;
3805         void *reply;
3806         int ret, cmd, reply_cmd;
3807         struct net *net = sock_net(skb->sk);
3808         struct netns_ipvs *ipvs = net_ipvs(net);
3809
3810         cmd = info->genlhdr->cmd;
3811
3812         if (cmd == IPVS_CMD_GET_SERVICE)
3813                 reply_cmd = IPVS_CMD_NEW_SERVICE;
3814         else if (cmd == IPVS_CMD_GET_INFO)
3815                 reply_cmd = IPVS_CMD_SET_INFO;
3816         else if (cmd == IPVS_CMD_GET_CONFIG)
3817                 reply_cmd = IPVS_CMD_SET_CONFIG;
3818         else {
3819                 pr_err("unknown Generic Netlink command\n");
3820                 return -EINVAL;
3821         }
3822
3823         msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3824         if (!msg)
3825                 return -ENOMEM;
3826
3827         mutex_lock(&__ip_vs_mutex);
3828
3829         reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3830         if (reply == NULL)
3831                 goto nla_put_failure;
3832
3833         switch (cmd) {
3834         case IPVS_CMD_GET_SERVICE:
3835         {
3836                 struct ip_vs_service *svc;
3837
3838                 svc = ip_vs_genl_find_service(ipvs,
3839                                               info->attrs[IPVS_CMD_ATTR_SERVICE]);
3840                 if (IS_ERR(svc)) {
3841                         ret = PTR_ERR(svc);
3842                         goto out_err;
3843                 } else if (svc) {
3844                         ret = ip_vs_genl_fill_service(msg, svc);
3845                         if (ret)
3846                                 goto nla_put_failure;
3847                 } else {
3848                         ret = -ESRCH;
3849                         goto out_err;
3850                 }
3851
3852                 break;
3853         }
3854
3855         case IPVS_CMD_GET_CONFIG:
3856         {
3857                 struct ip_vs_timeout_user t;
3858
3859                 __ip_vs_get_timeouts(ipvs, &t);
3860 #ifdef CONFIG_IP_VS_PROTO_TCP
3861                 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3862                                 t.tcp_timeout) ||
3863                     nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3864                                 t.tcp_fin_timeout))
3865                         goto nla_put_failure;
3866 #endif
3867 #ifdef CONFIG_IP_VS_PROTO_UDP
3868                 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3869                         goto nla_put_failure;
3870 #endif
3871
3872                 break;
3873         }
3874
3875         case IPVS_CMD_GET_INFO:
3876                 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3877                                 IP_VS_VERSION_CODE) ||
3878                     nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3879                                 ip_vs_conn_tab_size))
3880                         goto nla_put_failure;
3881                 break;
3882         }
3883
3884         genlmsg_end(msg, reply);
3885         ret = genlmsg_reply(msg, info);
3886         goto out;
3887
3888 nla_put_failure:
3889         pr_err("not enough space in Netlink message\n");
3890         ret = -EMSGSIZE;
3891
3892 out_err:
3893         nlmsg_free(msg);
3894 out:
3895         mutex_unlock(&__ip_vs_mutex);
3896
3897         return ret;
3898 }
3899
3900
3901 static const struct genl_small_ops ip_vs_genl_ops[] = {
3902         {
3903                 .cmd    = IPVS_CMD_NEW_SERVICE,
3904                 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3905                 .flags  = GENL_ADMIN_PERM,
3906                 .doit   = ip_vs_genl_set_cmd,
3907         },
3908         {
3909                 .cmd    = IPVS_CMD_SET_SERVICE,
3910                 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3911                 .flags  = GENL_ADMIN_PERM,
3912                 .doit   = ip_vs_genl_set_cmd,
3913         },
3914         {
3915                 .cmd    = IPVS_CMD_DEL_SERVICE,
3916                 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3917                 .flags  = GENL_ADMIN_PERM,
3918                 .doit   = ip_vs_genl_set_cmd,
3919         },
3920         {
3921                 .cmd    = IPVS_CMD_GET_SERVICE,
3922                 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3923                 .flags  = GENL_ADMIN_PERM,
3924                 .doit   = ip_vs_genl_get_cmd,
3925                 .dumpit = ip_vs_genl_dump_services,
3926         },
3927         {
3928                 .cmd    = IPVS_CMD_NEW_DEST,
3929                 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3930                 .flags  = GENL_ADMIN_PERM,
3931                 .doit   = ip_vs_genl_set_cmd,
3932         },
3933         {
3934                 .cmd    = IPVS_CMD_SET_DEST,
3935                 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3936                 .flags  = GENL_ADMIN_PERM,
3937                 .doit   = ip_vs_genl_set_cmd,
3938         },
3939         {
3940                 .cmd    = IPVS_CMD_DEL_DEST,
3941                 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3942                 .flags  = GENL_ADMIN_PERM,
3943                 .doit   = ip_vs_genl_set_cmd,
3944         },
3945         {
3946                 .cmd    = IPVS_CMD_GET_DEST,
3947                 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3948                 .flags  = GENL_ADMIN_PERM,
3949                 .dumpit = ip_vs_genl_dump_dests,
3950         },
3951         {
3952                 .cmd    = IPVS_CMD_NEW_DAEMON,
3953                 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3954                 .flags  = GENL_ADMIN_PERM,
3955                 .doit   = ip_vs_genl_set_daemon,
3956         },
3957         {
3958                 .cmd    = IPVS_CMD_DEL_DAEMON,
3959                 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3960                 .flags  = GENL_ADMIN_PERM,
3961                 .doit   = ip_vs_genl_set_daemon,
3962         },
3963         {
3964                 .cmd    = IPVS_CMD_GET_DAEMON,
3965                 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3966                 .flags  = GENL_ADMIN_PERM,
3967                 .dumpit = ip_vs_genl_dump_daemons,
3968         },
3969         {
3970                 .cmd    = IPVS_CMD_SET_CONFIG,
3971                 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3972                 .flags  = GENL_ADMIN_PERM,
3973                 .doit   = ip_vs_genl_set_cmd,
3974         },
3975         {
3976                 .cmd    = IPVS_CMD_GET_CONFIG,
3977                 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3978                 .flags  = GENL_ADMIN_PERM,
3979                 .doit   = ip_vs_genl_get_cmd,
3980         },
3981         {
3982                 .cmd    = IPVS_CMD_GET_INFO,
3983                 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3984                 .flags  = GENL_ADMIN_PERM,
3985                 .doit   = ip_vs_genl_get_cmd,
3986         },
3987         {
3988                 .cmd    = IPVS_CMD_ZERO,
3989                 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3990                 .flags  = GENL_ADMIN_PERM,
3991                 .doit   = ip_vs_genl_set_cmd,
3992         },
3993         {
3994                 .cmd    = IPVS_CMD_FLUSH,
3995                 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3996                 .flags  = GENL_ADMIN_PERM,
3997                 .doit   = ip_vs_genl_set_cmd,
3998         },
3999 };
4000
4001 static struct genl_family ip_vs_genl_family __ro_after_init = {
4002         .hdrsize        = 0,
4003         .name           = IPVS_GENL_NAME,
4004         .version        = IPVS_GENL_VERSION,
4005         .maxattr        = IPVS_CMD_ATTR_MAX,
4006         .policy = ip_vs_cmd_policy,
4007         .netnsok        = true,         /* Make ipvsadm to work on netns */
4008         .module         = THIS_MODULE,
4009         .small_ops      = ip_vs_genl_ops,
4010         .n_small_ops    = ARRAY_SIZE(ip_vs_genl_ops),
4011         .resv_start_op  = IPVS_CMD_FLUSH + 1,
4012 };
4013
4014 static int __init ip_vs_genl_register(void)
4015 {
4016         return genl_register_family(&ip_vs_genl_family);
4017 }
4018
4019 static void ip_vs_genl_unregister(void)
4020 {
4021         genl_unregister_family(&ip_vs_genl_family);
4022 }
4023
4024 /* End of Generic Netlink interface definitions */
4025
4026 /*
4027  * per netns intit/exit func.
4028  */
4029 #ifdef CONFIG_SYSCTL
4030 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs)
4031 {
4032         struct net *net = ipvs->net;
4033         int idx;
4034         struct ctl_table *tbl;
4035
4036         atomic_set(&ipvs->dropentry, 0);
4037         spin_lock_init(&ipvs->dropentry_lock);
4038         spin_lock_init(&ipvs->droppacket_lock);
4039         spin_lock_init(&ipvs->securetcp_lock);
4040
4041         if (!net_eq(net, &init_net)) {
4042                 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
4043                 if (tbl == NULL)
4044                         return -ENOMEM;
4045
4046                 /* Don't export sysctls to unprivileged users */
4047                 if (net->user_ns != &init_user_ns)
4048                         tbl[0].procname = NULL;
4049         } else
4050                 tbl = vs_vars;
4051         /* Initialize sysctl defaults */
4052         for (idx = 0; idx < ARRAY_SIZE(vs_vars); idx++) {
4053                 if (tbl[idx].proc_handler == proc_do_defense_mode)
4054                         tbl[idx].extra2 = ipvs;
4055         }
4056         idx = 0;
4057         ipvs->sysctl_amemthresh = 1024;
4058         tbl[idx++].data = &ipvs->sysctl_amemthresh;
4059         ipvs->sysctl_am_droprate = 10;
4060         tbl[idx++].data = &ipvs->sysctl_am_droprate;
4061         tbl[idx++].data = &ipvs->sysctl_drop_entry;
4062         tbl[idx++].data = &ipvs->sysctl_drop_packet;
4063 #ifdef CONFIG_IP_VS_NFCT
4064         tbl[idx++].data = &ipvs->sysctl_conntrack;
4065 #endif
4066         tbl[idx++].data = &ipvs->sysctl_secure_tcp;
4067         ipvs->sysctl_snat_reroute = 1;
4068         tbl[idx++].data = &ipvs->sysctl_snat_reroute;
4069         ipvs->sysctl_sync_ver = 1;
4070         tbl[idx++].data = &ipvs->sysctl_sync_ver;
4071         ipvs->sysctl_sync_ports = 1;
4072         tbl[idx++].data = &ipvs->sysctl_sync_ports;
4073         tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
4074         ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
4075         tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
4076         ipvs->sysctl_sync_sock_size = 0;
4077         tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
4078         tbl[idx++].data = &ipvs->sysctl_cache_bypass;
4079         tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
4080         tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
4081         tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
4082         tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
4083         ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
4084         ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
4085         tbl[idx].data = &ipvs->sysctl_sync_threshold;
4086         tbl[idx].extra2 = ipvs;
4087         tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
4088         ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
4089         tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
4090         ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
4091         tbl[idx++].data = &ipvs->sysctl_sync_retries;
4092         tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
4093         ipvs->sysctl_pmtu_disc = 1;
4094         tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
4095         tbl[idx++].data = &ipvs->sysctl_backup_only;
4096         ipvs->sysctl_conn_reuse_mode = 1;
4097         tbl[idx++].data = &ipvs->sysctl_conn_reuse_mode;
4098         tbl[idx++].data = &ipvs->sysctl_schedule_icmp;
4099         tbl[idx++].data = &ipvs->sysctl_ignore_tunneled;
4100         ipvs->sysctl_run_estimation = 1;
4101         tbl[idx++].data = &ipvs->sysctl_run_estimation;
4102 #ifdef CONFIG_IP_VS_DEBUG
4103         /* Global sysctls must be ro in non-init netns */
4104         if (!net_eq(net, &init_net))
4105                 tbl[idx++].mode = 0444;
4106 #endif
4107
4108         ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
4109         if (ipvs->sysctl_hdr == NULL) {
4110                 if (!net_eq(net, &init_net))
4111                         kfree(tbl);
4112                 return -ENOMEM;
4113         }
4114         ip_vs_start_estimator(ipvs, &ipvs->tot_stats);
4115         ipvs->sysctl_tbl = tbl;
4116         /* Schedule defense work */
4117         INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
4118         queue_delayed_work(system_long_wq, &ipvs->defense_work,
4119                            DEFENSE_TIMER_PERIOD);
4120
4121         /* Init delayed work for expiring no dest conn */
4122         INIT_DELAYED_WORK(&ipvs->expire_nodest_conn_work,
4123                           expire_nodest_conn_handler);
4124
4125         return 0;
4126 }
4127
4128 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs)
4129 {
4130         struct net *net = ipvs->net;
4131
4132         cancel_delayed_work_sync(&ipvs->expire_nodest_conn_work);
4133         cancel_delayed_work_sync(&ipvs->defense_work);
4134         cancel_work_sync(&ipvs->defense_work.work);
4135         unregister_net_sysctl_table(ipvs->sysctl_hdr);
4136         ip_vs_stop_estimator(ipvs, &ipvs->tot_stats);
4137
4138         if (!net_eq(net, &init_net))
4139                 kfree(ipvs->sysctl_tbl);
4140 }
4141
4142 #else
4143
4144 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs) { return 0; }
4145 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs) { }
4146
4147 #endif
4148
4149 static struct notifier_block ip_vs_dst_notifier = {
4150         .notifier_call = ip_vs_dst_event,
4151 #ifdef CONFIG_IP_VS_IPV6
4152         .priority = ADDRCONF_NOTIFY_PRIORITY + 5,
4153 #endif
4154 };
4155
4156 int __net_init ip_vs_control_net_init(struct netns_ipvs *ipvs)
4157 {
4158         int i, idx;
4159
4160         /* Initialize rs_table */
4161         for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
4162                 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
4163
4164         INIT_LIST_HEAD(&ipvs->dest_trash);
4165         spin_lock_init(&ipvs->dest_trash_lock);
4166         timer_setup(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire, 0);
4167         atomic_set(&ipvs->ftpsvc_counter, 0);
4168         atomic_set(&ipvs->nullsvc_counter, 0);
4169         atomic_set(&ipvs->conn_out_counter, 0);
4170
4171         /* procfs stats */
4172         ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
4173         if (!ipvs->tot_stats.cpustats)
4174                 return -ENOMEM;
4175
4176         for_each_possible_cpu(i) {
4177                 struct ip_vs_cpu_stats *ipvs_tot_stats;
4178                 ipvs_tot_stats = per_cpu_ptr(ipvs->tot_stats.cpustats, i);
4179                 u64_stats_init(&ipvs_tot_stats->syncp);
4180         }
4181
4182         spin_lock_init(&ipvs->tot_stats.lock);
4183
4184 #ifdef CONFIG_PROC_FS
4185         if (!proc_create_net("ip_vs", 0, ipvs->net->proc_net,
4186                              &ip_vs_info_seq_ops, sizeof(struct ip_vs_iter)))
4187                 goto err_vs;
4188         if (!proc_create_net_single("ip_vs_stats", 0, ipvs->net->proc_net,
4189                                     ip_vs_stats_show, NULL))
4190                 goto err_stats;
4191         if (!proc_create_net_single("ip_vs_stats_percpu", 0,
4192                                     ipvs->net->proc_net,
4193                                     ip_vs_stats_percpu_show, NULL))
4194                 goto err_percpu;
4195 #endif
4196
4197         if (ip_vs_control_net_init_sysctl(ipvs))
4198                 goto err;
4199
4200         return 0;
4201
4202 err:
4203 #ifdef CONFIG_PROC_FS
4204         remove_proc_entry("ip_vs_stats_percpu", ipvs->net->proc_net);
4205
4206 err_percpu:
4207         remove_proc_entry("ip_vs_stats", ipvs->net->proc_net);
4208
4209 err_stats:
4210         remove_proc_entry("ip_vs", ipvs->net->proc_net);
4211
4212 err_vs:
4213 #endif
4214         free_percpu(ipvs->tot_stats.cpustats);
4215         return -ENOMEM;
4216 }
4217
4218 void __net_exit ip_vs_control_net_cleanup(struct netns_ipvs *ipvs)
4219 {
4220         ip_vs_trash_cleanup(ipvs);
4221         ip_vs_control_net_cleanup_sysctl(ipvs);
4222 #ifdef CONFIG_PROC_FS
4223         remove_proc_entry("ip_vs_stats_percpu", ipvs->net->proc_net);
4224         remove_proc_entry("ip_vs_stats", ipvs->net->proc_net);
4225         remove_proc_entry("ip_vs", ipvs->net->proc_net);
4226 #endif
4227         free_percpu(ipvs->tot_stats.cpustats);
4228 }
4229
4230 int __init ip_vs_register_nl_ioctl(void)
4231 {
4232         int ret;
4233
4234         ret = nf_register_sockopt(&ip_vs_sockopts);
4235         if (ret) {
4236                 pr_err("cannot register sockopt.\n");
4237                 goto err_sock;
4238         }
4239
4240         ret = ip_vs_genl_register();
4241         if (ret) {
4242                 pr_err("cannot register Generic Netlink interface.\n");
4243                 goto err_genl;
4244         }
4245         return 0;
4246
4247 err_genl:
4248         nf_unregister_sockopt(&ip_vs_sockopts);
4249 err_sock:
4250         return ret;
4251 }
4252
4253 void ip_vs_unregister_nl_ioctl(void)
4254 {
4255         ip_vs_genl_unregister();
4256         nf_unregister_sockopt(&ip_vs_sockopts);
4257 }
4258
4259 int __init ip_vs_control_init(void)
4260 {
4261         int idx;
4262         int ret;
4263
4264         EnterFunction(2);
4265
4266         /* Initialize svc_table, ip_vs_svc_fwm_table */
4267         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
4268                 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
4269                 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
4270         }
4271
4272         smp_wmb();      /* Do we really need it now ? */
4273
4274         ret = register_netdevice_notifier(&ip_vs_dst_notifier);
4275         if (ret < 0)
4276                 return ret;
4277
4278         LeaveFunction(2);
4279         return 0;
4280 }
4281
4282
4283 void ip_vs_control_cleanup(void)
4284 {
4285         EnterFunction(2);
4286         unregister_netdevice_notifier(&ip_vs_dst_notifier);
4287         LeaveFunction(2);
4288 }