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