1 /* IP Virtual Server 2 * data structure and functionality definitions 3 */ 4 5 #ifndef _NET_IP_VS_H 6 #define _NET_IP_VS_H 7 8 #include <linux/ip_vs.h> /* definitions shared with userland */ 9 10 #include <asm/types.h> /* for __uXX types */ 11 12 #include <linux/list.h> /* for struct list_head */ 13 #include <linux/spinlock.h> /* for struct rwlock_t */ 14 #include <linux/atomic.h> /* for struct atomic_t */ 15 #include <linux/compiler.h> 16 #include <linux/timer.h> 17 #include <linux/bug.h> 18 19 #include <net/checksum.h> 20 #include <linux/netfilter.h> /* for union nf_inet_addr */ 21 #include <linux/ip.h> 22 #include <linux/ipv6.h> /* for struct ipv6hdr */ 23 #include <net/ipv6.h> 24 #if IS_ENABLED(CONFIG_IP_VS_IPV6) 25 #include <linux/netfilter_ipv6/ip6_tables.h> 26 #endif 27 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 28 #include <net/netfilter/nf_conntrack.h> 29 #endif 30 #include <net/net_namespace.h> /* Netw namespace */ 31 32 /* Generic access of ipvs struct */ 33 static inline struct netns_ipvs *net_ipvs(struct net* net) 34 { 35 return net->ipvs; 36 } 37 38 /* Get net ptr from skb in traffic cases 39 * use skb_sknet when call is from userland (ioctl or netlink) 40 */ 41 static inline struct net *skb_net(const struct sk_buff *skb) 42 { 43 #ifdef CONFIG_NET_NS 44 #ifdef CONFIG_IP_VS_DEBUG 45 /* 46 * This is used for debug only. 47 * Start with the most likely hit 48 * End with BUG 49 */ 50 if (likely(skb->dev && skb->dev->nd_net)) 51 return dev_net(skb->dev); 52 if (skb_dst(skb) && skb_dst(skb)->dev) 53 return dev_net(skb_dst(skb)->dev); 54 WARN(skb->sk, "Maybe skb_sknet should be used in %s() at line:%d\n", 55 __func__, __LINE__); 56 if (likely(skb->sk && skb->sk->sk_net)) 57 return sock_net(skb->sk); 58 pr_err("There is no net ptr to find in the skb in %s() line:%d\n", 59 __func__, __LINE__); 60 BUG(); 61 #else 62 return dev_net(skb->dev ? : skb_dst(skb)->dev); 63 #endif 64 #else 65 return &init_net; 66 #endif 67 } 68 69 static inline struct net *skb_sknet(const struct sk_buff *skb) 70 { 71 #ifdef CONFIG_NET_NS 72 #ifdef CONFIG_IP_VS_DEBUG 73 /* Start with the most likely hit */ 74 if (likely(skb->sk && skb->sk->sk_net)) 75 return sock_net(skb->sk); 76 WARN(skb->dev, "Maybe skb_net should be used instead in %s() line:%d\n", 77 __func__, __LINE__); 78 if (likely(skb->dev && skb->dev->nd_net)) 79 return dev_net(skb->dev); 80 pr_err("There is no net ptr to find in the skb in %s() line:%d\n", 81 __func__, __LINE__); 82 BUG(); 83 #else 84 return sock_net(skb->sk); 85 #endif 86 #else 87 return &init_net; 88 #endif 89 } 90 91 /* This one needed for single_open_net since net is stored directly in 92 * private not as a struct i.e. seq_file_net can't be used. 93 */ 94 static inline struct net *seq_file_single_net(struct seq_file *seq) 95 { 96 #ifdef CONFIG_NET_NS 97 return (struct net *)seq->private; 98 #else 99 return &init_net; 100 #endif 101 } 102 103 /* Connections' size value needed by ip_vs_ctl.c */ 104 extern int ip_vs_conn_tab_size; 105 106 struct ip_vs_iphdr { 107 __u32 len; /* IPv4 simply where L4 starts 108 * IPv6 where L4 Transport Header starts */ 109 __u16 fragoffs; /* IPv6 fragment offset, 0 if first frag (or not frag)*/ 110 __s16 protocol; 111 __s32 flags; 112 union nf_inet_addr saddr; 113 union nf_inet_addr daddr; 114 }; 115 116 static inline void *frag_safe_skb_hp(const struct sk_buff *skb, int offset, 117 int len, void *buffer, 118 const struct ip_vs_iphdr *ipvsh) 119 { 120 return skb_header_pointer(skb, offset, len, buffer); 121 } 122 123 static inline void 124 ip_vs_fill_ip4hdr(const void *nh, struct ip_vs_iphdr *iphdr) 125 { 126 const struct iphdr *iph = nh; 127 128 iphdr->len = iph->ihl * 4; 129 iphdr->fragoffs = 0; 130 iphdr->protocol = iph->protocol; 131 iphdr->saddr.ip = iph->saddr; 132 iphdr->daddr.ip = iph->daddr; 133 } 134 135 /* This function handles filling *ip_vs_iphdr, both for IPv4 and IPv6. 136 * IPv6 requires some extra work, as finding proper header position, 137 * depend on the IPv6 extension headers. 138 */ 139 static inline void 140 ip_vs_fill_iph_skb(int af, const struct sk_buff *skb, struct ip_vs_iphdr *iphdr) 141 { 142 #ifdef CONFIG_IP_VS_IPV6 143 if (af == AF_INET6) { 144 const struct ipv6hdr *iph = 145 (struct ipv6hdr *)skb_network_header(skb); 146 iphdr->saddr.in6 = iph->saddr; 147 iphdr->daddr.in6 = iph->daddr; 148 /* ipv6_find_hdr() updates len, flags */ 149 iphdr->len = 0; 150 iphdr->flags = 0; 151 iphdr->protocol = ipv6_find_hdr(skb, &iphdr->len, -1, 152 &iphdr->fragoffs, 153 &iphdr->flags); 154 } else 155 #endif 156 { 157 const struct iphdr *iph = 158 (struct iphdr *)skb_network_header(skb); 159 iphdr->len = iph->ihl * 4; 160 iphdr->fragoffs = 0; 161 iphdr->protocol = iph->protocol; 162 iphdr->saddr.ip = iph->saddr; 163 iphdr->daddr.ip = iph->daddr; 164 } 165 } 166 167 static inline void ip_vs_addr_copy(int af, union nf_inet_addr *dst, 168 const union nf_inet_addr *src) 169 { 170 #ifdef CONFIG_IP_VS_IPV6 171 if (af == AF_INET6) 172 dst->in6 = src->in6; 173 else 174 #endif 175 dst->ip = src->ip; 176 } 177 178 static inline void ip_vs_addr_set(int af, union nf_inet_addr *dst, 179 const union nf_inet_addr *src) 180 { 181 #ifdef CONFIG_IP_VS_IPV6 182 if (af == AF_INET6) { 183 dst->in6 = src->in6; 184 return; 185 } 186 #endif 187 dst->ip = src->ip; 188 dst->all[1] = 0; 189 dst->all[2] = 0; 190 dst->all[3] = 0; 191 } 192 193 static inline int ip_vs_addr_equal(int af, const union nf_inet_addr *a, 194 const union nf_inet_addr *b) 195 { 196 #ifdef CONFIG_IP_VS_IPV6 197 if (af == AF_INET6) 198 return ipv6_addr_equal(&a->in6, &b->in6); 199 #endif 200 return a->ip == b->ip; 201 } 202 203 #ifdef CONFIG_IP_VS_DEBUG 204 #include <linux/net.h> 205 206 int ip_vs_get_debug_level(void); 207 208 static inline const char *ip_vs_dbg_addr(int af, char *buf, size_t buf_len, 209 const union nf_inet_addr *addr, 210 int *idx) 211 { 212 int len; 213 #ifdef CONFIG_IP_VS_IPV6 214 if (af == AF_INET6) 215 len = snprintf(&buf[*idx], buf_len - *idx, "[%pI6c]", 216 &addr->in6) + 1; 217 else 218 #endif 219 len = snprintf(&buf[*idx], buf_len - *idx, "%pI4", 220 &addr->ip) + 1; 221 222 *idx += len; 223 BUG_ON(*idx > buf_len + 1); 224 return &buf[*idx - len]; 225 } 226 227 #define IP_VS_DBG_BUF(level, msg, ...) \ 228 do { \ 229 char ip_vs_dbg_buf[160]; \ 230 int ip_vs_dbg_idx = 0; \ 231 if (level <= ip_vs_get_debug_level()) \ 232 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \ 233 } while (0) 234 #define IP_VS_ERR_BUF(msg...) \ 235 do { \ 236 char ip_vs_dbg_buf[160]; \ 237 int ip_vs_dbg_idx = 0; \ 238 pr_err(msg); \ 239 } while (0) 240 241 /* Only use from within IP_VS_DBG_BUF() or IP_VS_ERR_BUF macros */ 242 #define IP_VS_DBG_ADDR(af, addr) \ 243 ip_vs_dbg_addr(af, ip_vs_dbg_buf, \ 244 sizeof(ip_vs_dbg_buf), addr, \ 245 &ip_vs_dbg_idx) 246 247 #define IP_VS_DBG(level, msg, ...) \ 248 do { \ 249 if (level <= ip_vs_get_debug_level()) \ 250 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \ 251 } while (0) 252 #define IP_VS_DBG_RL(msg, ...) \ 253 do { \ 254 if (net_ratelimit()) \ 255 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \ 256 } while (0) 257 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) \ 258 do { \ 259 if (level <= ip_vs_get_debug_level()) \ 260 pp->debug_packet(af, pp, skb, ofs, msg); \ 261 } while (0) 262 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) \ 263 do { \ 264 if (level <= ip_vs_get_debug_level() && \ 265 net_ratelimit()) \ 266 pp->debug_packet(af, pp, skb, ofs, msg); \ 267 } while (0) 268 #else /* NO DEBUGGING at ALL */ 269 #define IP_VS_DBG_BUF(level, msg...) do {} while (0) 270 #define IP_VS_ERR_BUF(msg...) do {} while (0) 271 #define IP_VS_DBG(level, msg...) do {} while (0) 272 #define IP_VS_DBG_RL(msg...) do {} while (0) 273 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) do {} while (0) 274 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) do {} while (0) 275 #endif 276 277 #define IP_VS_BUG() BUG() 278 #define IP_VS_ERR_RL(msg, ...) \ 279 do { \ 280 if (net_ratelimit()) \ 281 pr_err(msg, ##__VA_ARGS__); \ 282 } while (0) 283 284 #ifdef CONFIG_IP_VS_DEBUG 285 #define EnterFunction(level) \ 286 do { \ 287 if (level <= ip_vs_get_debug_level()) \ 288 printk(KERN_DEBUG \ 289 pr_fmt("Enter: %s, %s line %i\n"), \ 290 __func__, __FILE__, __LINE__); \ 291 } while (0) 292 #define LeaveFunction(level) \ 293 do { \ 294 if (level <= ip_vs_get_debug_level()) \ 295 printk(KERN_DEBUG \ 296 pr_fmt("Leave: %s, %s line %i\n"), \ 297 __func__, __FILE__, __LINE__); \ 298 } while (0) 299 #else 300 #define EnterFunction(level) do {} while (0) 301 #define LeaveFunction(level) do {} while (0) 302 #endif 303 304 /* The port number of FTP service (in network order). */ 305 #define FTPPORT cpu_to_be16(21) 306 #define FTPDATA cpu_to_be16(20) 307 308 /* TCP State Values */ 309 enum { 310 IP_VS_TCP_S_NONE = 0, 311 IP_VS_TCP_S_ESTABLISHED, 312 IP_VS_TCP_S_SYN_SENT, 313 IP_VS_TCP_S_SYN_RECV, 314 IP_VS_TCP_S_FIN_WAIT, 315 IP_VS_TCP_S_TIME_WAIT, 316 IP_VS_TCP_S_CLOSE, 317 IP_VS_TCP_S_CLOSE_WAIT, 318 IP_VS_TCP_S_LAST_ACK, 319 IP_VS_TCP_S_LISTEN, 320 IP_VS_TCP_S_SYNACK, 321 IP_VS_TCP_S_LAST 322 }; 323 324 /* UDP State Values */ 325 enum { 326 IP_VS_UDP_S_NORMAL, 327 IP_VS_UDP_S_LAST, 328 }; 329 330 /* ICMP State Values */ 331 enum { 332 IP_VS_ICMP_S_NORMAL, 333 IP_VS_ICMP_S_LAST, 334 }; 335 336 /* SCTP State Values */ 337 enum ip_vs_sctp_states { 338 IP_VS_SCTP_S_NONE, 339 IP_VS_SCTP_S_INIT1, 340 IP_VS_SCTP_S_INIT, 341 IP_VS_SCTP_S_COOKIE_SENT, 342 IP_VS_SCTP_S_COOKIE_REPLIED, 343 IP_VS_SCTP_S_COOKIE_WAIT, 344 IP_VS_SCTP_S_COOKIE, 345 IP_VS_SCTP_S_COOKIE_ECHOED, 346 IP_VS_SCTP_S_ESTABLISHED, 347 IP_VS_SCTP_S_SHUTDOWN_SENT, 348 IP_VS_SCTP_S_SHUTDOWN_RECEIVED, 349 IP_VS_SCTP_S_SHUTDOWN_ACK_SENT, 350 IP_VS_SCTP_S_REJECTED, 351 IP_VS_SCTP_S_CLOSED, 352 IP_VS_SCTP_S_LAST 353 }; 354 355 /* Delta sequence info structure 356 * Each ip_vs_conn has 2 (output AND input seq. changes). 357 * Only used in the VS/NAT. 358 */ 359 struct ip_vs_seq { 360 __u32 init_seq; /* Add delta from this seq */ 361 __u32 delta; /* Delta in sequence numbers */ 362 __u32 previous_delta; /* Delta in sequence numbers 363 * before last resized pkt */ 364 }; 365 366 /* counters per cpu */ 367 struct ip_vs_counters { 368 __u32 conns; /* connections scheduled */ 369 __u32 inpkts; /* incoming packets */ 370 __u32 outpkts; /* outgoing packets */ 371 __u64 inbytes; /* incoming bytes */ 372 __u64 outbytes; /* outgoing bytes */ 373 }; 374 /* Stats per cpu */ 375 struct ip_vs_cpu_stats { 376 struct ip_vs_counters ustats; 377 struct u64_stats_sync syncp; 378 }; 379 380 /* IPVS statistics objects */ 381 struct ip_vs_estimator { 382 struct list_head list; 383 384 u64 last_inbytes; 385 u64 last_outbytes; 386 u32 last_conns; 387 u32 last_inpkts; 388 u32 last_outpkts; 389 390 u32 cps; 391 u32 inpps; 392 u32 outpps; 393 u32 inbps; 394 u32 outbps; 395 }; 396 397 struct ip_vs_stats { 398 struct ip_vs_stats_user ustats; /* statistics */ 399 struct ip_vs_estimator est; /* estimator */ 400 struct ip_vs_cpu_stats __percpu *cpustats; /* per cpu counters */ 401 spinlock_t lock; /* spin lock */ 402 struct ip_vs_stats_user ustats0; /* reset values */ 403 }; 404 405 struct dst_entry; 406 struct iphdr; 407 struct ip_vs_conn; 408 struct ip_vs_app; 409 struct sk_buff; 410 struct ip_vs_proto_data; 411 412 struct ip_vs_protocol { 413 struct ip_vs_protocol *next; 414 char *name; 415 u16 protocol; 416 u16 num_states; 417 int dont_defrag; 418 419 void (*init)(struct ip_vs_protocol *pp); 420 421 void (*exit)(struct ip_vs_protocol *pp); 422 423 int (*init_netns)(struct net *net, struct ip_vs_proto_data *pd); 424 425 void (*exit_netns)(struct net *net, struct ip_vs_proto_data *pd); 426 427 int (*conn_schedule)(int af, struct sk_buff *skb, 428 struct ip_vs_proto_data *pd, 429 int *verdict, struct ip_vs_conn **cpp, 430 struct ip_vs_iphdr *iph); 431 432 struct ip_vs_conn * 433 (*conn_in_get)(int af, 434 const struct sk_buff *skb, 435 const struct ip_vs_iphdr *iph, 436 int inverse); 437 438 struct ip_vs_conn * 439 (*conn_out_get)(int af, 440 const struct sk_buff *skb, 441 const struct ip_vs_iphdr *iph, 442 int inverse); 443 444 int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp, 445 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph); 446 447 int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp, 448 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph); 449 450 int (*csum_check)(int af, struct sk_buff *skb, 451 struct ip_vs_protocol *pp); 452 453 const char *(*state_name)(int state); 454 455 void (*state_transition)(struct ip_vs_conn *cp, int direction, 456 const struct sk_buff *skb, 457 struct ip_vs_proto_data *pd); 458 459 int (*register_app)(struct net *net, struct ip_vs_app *inc); 460 461 void (*unregister_app)(struct net *net, struct ip_vs_app *inc); 462 463 int (*app_conn_bind)(struct ip_vs_conn *cp); 464 465 void (*debug_packet)(int af, struct ip_vs_protocol *pp, 466 const struct sk_buff *skb, 467 int offset, 468 const char *msg); 469 470 void (*timeout_change)(struct ip_vs_proto_data *pd, int flags); 471 }; 472 473 /* protocol data per netns */ 474 struct ip_vs_proto_data { 475 struct ip_vs_proto_data *next; 476 struct ip_vs_protocol *pp; 477 int *timeout_table; /* protocol timeout table */ 478 atomic_t appcnt; /* counter of proto app incs. */ 479 struct tcp_states_t *tcp_state_table; 480 }; 481 482 struct ip_vs_protocol *ip_vs_proto_get(unsigned short proto); 483 struct ip_vs_proto_data *ip_vs_proto_data_get(struct net *net, 484 unsigned short proto); 485 486 struct ip_vs_conn_param { 487 struct net *net; 488 const union nf_inet_addr *caddr; 489 const union nf_inet_addr *vaddr; 490 __be16 cport; 491 __be16 vport; 492 __u16 protocol; 493 u16 af; 494 495 const struct ip_vs_pe *pe; 496 char *pe_data; 497 __u8 pe_data_len; 498 }; 499 500 /* IP_VS structure allocated for each dynamically scheduled connection */ 501 struct ip_vs_conn { 502 struct hlist_node c_list; /* hashed list heads */ 503 /* Protocol, addresses and port numbers */ 504 __be16 cport; 505 __be16 dport; 506 __be16 vport; 507 u16 af; /* address family */ 508 union nf_inet_addr caddr; /* client address */ 509 union nf_inet_addr vaddr; /* virtual address */ 510 union nf_inet_addr daddr; /* destination address */ 511 volatile __u32 flags; /* status flags */ 512 __u16 protocol; /* Which protocol (TCP/UDP) */ 513 __u16 daf; /* Address family of the dest */ 514 #ifdef CONFIG_NET_NS 515 struct net *net; /* Name space */ 516 #endif 517 518 /* counter and timer */ 519 atomic_t refcnt; /* reference count */ 520 struct timer_list timer; /* Expiration timer */ 521 volatile unsigned long timeout; /* timeout */ 522 523 /* Flags and state transition */ 524 spinlock_t lock; /* lock for state transition */ 525 volatile __u16 state; /* state info */ 526 volatile __u16 old_state; /* old state, to be used for 527 * state transition triggerd 528 * synchronization 529 */ 530 __u32 fwmark; /* Fire wall mark from skb */ 531 unsigned long sync_endtime; /* jiffies + sent_retries */ 532 533 /* Control members */ 534 struct ip_vs_conn *control; /* Master control connection */ 535 atomic_t n_control; /* Number of controlled ones */ 536 struct ip_vs_dest *dest; /* real server */ 537 atomic_t in_pkts; /* incoming packet counter */ 538 539 /* Packet transmitter for different forwarding methods. If it 540 * mangles the packet, it must return NF_DROP or better NF_STOLEN, 541 * otherwise this must be changed to a sk_buff **. 542 * NF_ACCEPT can be returned when destination is local. 543 */ 544 int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp, 545 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 546 547 /* Note: we can group the following members into a structure, 548 * in order to save more space, and the following members are 549 * only used in VS/NAT anyway 550 */ 551 struct ip_vs_app *app; /* bound ip_vs_app object */ 552 void *app_data; /* Application private data */ 553 struct ip_vs_seq in_seq; /* incoming seq. struct */ 554 struct ip_vs_seq out_seq; /* outgoing seq. struct */ 555 556 const struct ip_vs_pe *pe; 557 char *pe_data; 558 __u8 pe_data_len; 559 560 struct rcu_head rcu_head; 561 }; 562 563 /* To save some memory in conn table when name space is disabled. */ 564 static inline struct net *ip_vs_conn_net(const struct ip_vs_conn *cp) 565 { 566 #ifdef CONFIG_NET_NS 567 return cp->net; 568 #else 569 return &init_net; 570 #endif 571 } 572 573 static inline void ip_vs_conn_net_set(struct ip_vs_conn *cp, struct net *net) 574 { 575 #ifdef CONFIG_NET_NS 576 cp->net = net; 577 #endif 578 } 579 580 static inline int ip_vs_conn_net_eq(const struct ip_vs_conn *cp, 581 struct net *net) 582 { 583 #ifdef CONFIG_NET_NS 584 return cp->net == net; 585 #else 586 return 1; 587 #endif 588 } 589 590 /* Extended internal versions of struct ip_vs_service_user and ip_vs_dest_user 591 * for IPv6 support. 592 * 593 * We need these to conveniently pass around service and destination 594 * options, but unfortunately, we also need to keep the old definitions to 595 * maintain userspace backwards compatibility for the setsockopt interface. 596 */ 597 struct ip_vs_service_user_kern { 598 /* virtual service addresses */ 599 u16 af; 600 u16 protocol; 601 union nf_inet_addr addr; /* virtual ip address */ 602 __be16 port; 603 u32 fwmark; /* firwall mark of service */ 604 605 /* virtual service options */ 606 char *sched_name; 607 char *pe_name; 608 unsigned int flags; /* virtual service flags */ 609 unsigned int timeout; /* persistent timeout in sec */ 610 __be32 netmask; /* persistent netmask or plen */ 611 }; 612 613 614 struct ip_vs_dest_user_kern { 615 /* destination server address */ 616 union nf_inet_addr addr; 617 __be16 port; 618 619 /* real server options */ 620 unsigned int conn_flags; /* connection flags */ 621 int weight; /* destination weight */ 622 623 /* thresholds for active connections */ 624 u32 u_threshold; /* upper threshold */ 625 u32 l_threshold; /* lower threshold */ 626 627 /* Address family of addr */ 628 u16 af; 629 }; 630 631 632 /* 633 * The information about the virtual service offered to the net and the 634 * forwarding entries. 635 */ 636 struct ip_vs_service { 637 struct hlist_node s_list; /* for normal service table */ 638 struct hlist_node f_list; /* for fwmark-based service table */ 639 atomic_t refcnt; /* reference counter */ 640 641 u16 af; /* address family */ 642 __u16 protocol; /* which protocol (TCP/UDP) */ 643 union nf_inet_addr addr; /* IP address for virtual service */ 644 __be16 port; /* port number for the service */ 645 __u32 fwmark; /* firewall mark of the service */ 646 unsigned int flags; /* service status flags */ 647 unsigned int timeout; /* persistent timeout in ticks */ 648 __be32 netmask; /* grouping granularity, mask/plen */ 649 struct net *net; 650 651 struct list_head destinations; /* real server d-linked list */ 652 __u32 num_dests; /* number of servers */ 653 struct ip_vs_stats stats; /* statistics for the service */ 654 655 /* for scheduling */ 656 struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */ 657 spinlock_t sched_lock; /* lock sched_data */ 658 void *sched_data; /* scheduler application data */ 659 660 /* alternate persistence engine */ 661 struct ip_vs_pe __rcu *pe; 662 663 struct rcu_head rcu_head; 664 }; 665 666 /* Information for cached dst */ 667 struct ip_vs_dest_dst { 668 struct dst_entry *dst_cache; /* destination cache entry */ 669 u32 dst_cookie; 670 union nf_inet_addr dst_saddr; 671 struct rcu_head rcu_head; 672 }; 673 674 /* The real server destination forwarding entry with ip address, port number, 675 * and so on. 676 */ 677 struct ip_vs_dest { 678 struct list_head n_list; /* for the dests in the service */ 679 struct hlist_node d_list; /* for table with all the dests */ 680 681 u16 af; /* address family */ 682 __be16 port; /* port number of the server */ 683 union nf_inet_addr addr; /* IP address of the server */ 684 volatile unsigned int flags; /* dest status flags */ 685 atomic_t conn_flags; /* flags to copy to conn */ 686 atomic_t weight; /* server weight */ 687 688 atomic_t refcnt; /* reference counter */ 689 struct ip_vs_stats stats; /* statistics */ 690 unsigned long idle_start; /* start time, jiffies */ 691 692 /* connection counters and thresholds */ 693 atomic_t activeconns; /* active connections */ 694 atomic_t inactconns; /* inactive connections */ 695 atomic_t persistconns; /* persistent connections */ 696 __u32 u_threshold; /* upper threshold */ 697 __u32 l_threshold; /* lower threshold */ 698 699 /* for destination cache */ 700 spinlock_t dst_lock; /* lock of dst_cache */ 701 struct ip_vs_dest_dst __rcu *dest_dst; /* cached dst info */ 702 703 /* for virtual service */ 704 struct ip_vs_service __rcu *svc; /* service it belongs to */ 705 __u16 protocol; /* which protocol (TCP/UDP) */ 706 __be16 vport; /* virtual port number */ 707 union nf_inet_addr vaddr; /* virtual IP address */ 708 __u32 vfwmark; /* firewall mark of service */ 709 710 struct list_head t_list; /* in dest_trash */ 711 unsigned int in_rs_table:1; /* we are in rs_table */ 712 }; 713 714 /* The scheduler object */ 715 struct ip_vs_scheduler { 716 struct list_head n_list; /* d-linked list head */ 717 char *name; /* scheduler name */ 718 atomic_t refcnt; /* reference counter */ 719 struct module *module; /* THIS_MODULE/NULL */ 720 721 /* scheduler initializing service */ 722 int (*init_service)(struct ip_vs_service *svc); 723 /* scheduling service finish */ 724 void (*done_service)(struct ip_vs_service *svc); 725 /* dest is linked */ 726 int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest); 727 /* dest is unlinked */ 728 int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest); 729 /* dest is updated */ 730 int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest); 731 732 /* selecting a server from the given service */ 733 struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc, 734 const struct sk_buff *skb, 735 struct ip_vs_iphdr *iph); 736 }; 737 738 /* The persistence engine object */ 739 struct ip_vs_pe { 740 struct list_head n_list; /* d-linked list head */ 741 char *name; /* scheduler name */ 742 atomic_t refcnt; /* reference counter */ 743 struct module *module; /* THIS_MODULE/NULL */ 744 745 /* get the connection template, if any */ 746 int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb); 747 bool (*ct_match)(const struct ip_vs_conn_param *p, 748 struct ip_vs_conn *ct); 749 u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval, 750 bool inverse); 751 int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf); 752 }; 753 754 /* The application module object (a.k.a. app incarnation) */ 755 struct ip_vs_app { 756 struct list_head a_list; /* member in app list */ 757 int type; /* IP_VS_APP_TYPE_xxx */ 758 char *name; /* application module name */ 759 __u16 protocol; 760 struct module *module; /* THIS_MODULE/NULL */ 761 struct list_head incs_list; /* list of incarnations */ 762 763 /* members for application incarnations */ 764 struct list_head p_list; /* member in proto app list */ 765 struct ip_vs_app *app; /* its real application */ 766 __be16 port; /* port number in net order */ 767 atomic_t usecnt; /* usage counter */ 768 struct rcu_head rcu_head; 769 770 /* output hook: Process packet in inout direction, diff set for TCP. 771 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok, 772 * 2=Mangled but checksum was not updated 773 */ 774 int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *, 775 struct sk_buff *, int *diff); 776 777 /* input hook: Process packet in outin direction, diff set for TCP. 778 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok, 779 * 2=Mangled but checksum was not updated 780 */ 781 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *, 782 struct sk_buff *, int *diff); 783 784 /* ip_vs_app initializer */ 785 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *); 786 787 /* ip_vs_app finish */ 788 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *); 789 790 791 /* not used now */ 792 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *, 793 struct ip_vs_protocol *); 794 795 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *); 796 797 int * timeout_table; 798 int * timeouts; 799 int timeouts_size; 800 801 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app, 802 int *verdict, struct ip_vs_conn **cpp); 803 804 struct ip_vs_conn * 805 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app, 806 const struct iphdr *iph, int inverse); 807 808 struct ip_vs_conn * 809 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app, 810 const struct iphdr *iph, int inverse); 811 812 int (*state_transition)(struct ip_vs_conn *cp, int direction, 813 const struct sk_buff *skb, 814 struct ip_vs_app *app); 815 816 void (*timeout_change)(struct ip_vs_app *app, int flags); 817 }; 818 819 struct ipvs_master_sync_state { 820 struct list_head sync_queue; 821 struct ip_vs_sync_buff *sync_buff; 822 unsigned long sync_queue_len; 823 unsigned int sync_queue_delay; 824 struct task_struct *master_thread; 825 struct delayed_work master_wakeup_work; 826 struct netns_ipvs *ipvs; 827 }; 828 829 /* How much time to keep dests in trash */ 830 #define IP_VS_DEST_TRASH_PERIOD (120 * HZ) 831 832 /* IPVS in network namespace */ 833 struct netns_ipvs { 834 int gen; /* Generation */ 835 int enable; /* enable like nf_hooks do */ 836 /* Hash table: for real service lookups */ 837 #define IP_VS_RTAB_BITS 4 838 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS) 839 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1) 840 841 struct hlist_head rs_table[IP_VS_RTAB_SIZE]; 842 /* ip_vs_app */ 843 struct list_head app_list; 844 /* ip_vs_proto */ 845 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */ 846 struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE]; 847 /* ip_vs_proto_tcp */ 848 #ifdef CONFIG_IP_VS_PROTO_TCP 849 #define TCP_APP_TAB_BITS 4 850 #define TCP_APP_TAB_SIZE (1 << TCP_APP_TAB_BITS) 851 #define TCP_APP_TAB_MASK (TCP_APP_TAB_SIZE - 1) 852 struct list_head tcp_apps[TCP_APP_TAB_SIZE]; 853 #endif 854 /* ip_vs_proto_udp */ 855 #ifdef CONFIG_IP_VS_PROTO_UDP 856 #define UDP_APP_TAB_BITS 4 857 #define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS) 858 #define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1) 859 struct list_head udp_apps[UDP_APP_TAB_SIZE]; 860 #endif 861 /* ip_vs_proto_sctp */ 862 #ifdef CONFIG_IP_VS_PROTO_SCTP 863 #define SCTP_APP_TAB_BITS 4 864 #define SCTP_APP_TAB_SIZE (1 << SCTP_APP_TAB_BITS) 865 #define SCTP_APP_TAB_MASK (SCTP_APP_TAB_SIZE - 1) 866 /* Hash table for SCTP application incarnations */ 867 struct list_head sctp_apps[SCTP_APP_TAB_SIZE]; 868 #endif 869 /* ip_vs_conn */ 870 atomic_t conn_count; /* connection counter */ 871 872 /* ip_vs_ctl */ 873 struct ip_vs_stats tot_stats; /* Statistics & est. */ 874 875 int num_services; /* no of virtual services */ 876 877 /* Trash for destinations */ 878 struct list_head dest_trash; 879 spinlock_t dest_trash_lock; 880 struct timer_list dest_trash_timer; /* expiration timer */ 881 /* Service counters */ 882 atomic_t ftpsvc_counter; 883 atomic_t nullsvc_counter; 884 885 #ifdef CONFIG_SYSCTL 886 /* 1/rate drop and drop-entry variables */ 887 struct delayed_work defense_work; /* Work handler */ 888 int drop_rate; 889 int drop_counter; 890 atomic_t dropentry; 891 /* locks in ctl.c */ 892 spinlock_t dropentry_lock; /* drop entry handling */ 893 spinlock_t droppacket_lock; /* drop packet handling */ 894 spinlock_t securetcp_lock; /* state and timeout tables */ 895 896 /* sys-ctl struct */ 897 struct ctl_table_header *sysctl_hdr; 898 struct ctl_table *sysctl_tbl; 899 #endif 900 901 /* sysctl variables */ 902 int sysctl_amemthresh; 903 int sysctl_am_droprate; 904 int sysctl_drop_entry; 905 int sysctl_drop_packet; 906 int sysctl_secure_tcp; 907 #ifdef CONFIG_IP_VS_NFCT 908 int sysctl_conntrack; 909 #endif 910 int sysctl_snat_reroute; 911 int sysctl_sync_ver; 912 int sysctl_sync_ports; 913 int sysctl_sync_persist_mode; 914 unsigned long sysctl_sync_qlen_max; 915 int sysctl_sync_sock_size; 916 int sysctl_cache_bypass; 917 int sysctl_expire_nodest_conn; 918 int sysctl_sloppy_tcp; 919 int sysctl_sloppy_sctp; 920 int sysctl_expire_quiescent_template; 921 int sysctl_sync_threshold[2]; 922 unsigned int sysctl_sync_refresh_period; 923 int sysctl_sync_retries; 924 int sysctl_nat_icmp_send; 925 int sysctl_pmtu_disc; 926 int sysctl_backup_only; 927 928 /* ip_vs_lblc */ 929 int sysctl_lblc_expiration; 930 struct ctl_table_header *lblc_ctl_header; 931 struct ctl_table *lblc_ctl_table; 932 /* ip_vs_lblcr */ 933 int sysctl_lblcr_expiration; 934 struct ctl_table_header *lblcr_ctl_header; 935 struct ctl_table *lblcr_ctl_table; 936 /* ip_vs_est */ 937 struct list_head est_list; /* estimator list */ 938 spinlock_t est_lock; 939 struct timer_list est_timer; /* Estimation timer */ 940 /* ip_vs_sync */ 941 spinlock_t sync_lock; 942 struct ipvs_master_sync_state *ms; 943 spinlock_t sync_buff_lock; 944 struct task_struct **backup_threads; 945 int threads_mask; 946 int send_mesg_maxlen; 947 int recv_mesg_maxlen; 948 volatile int sync_state; 949 volatile int master_syncid; 950 volatile int backup_syncid; 951 struct mutex sync_mutex; 952 /* multicast interface name */ 953 char master_mcast_ifn[IP_VS_IFNAME_MAXLEN]; 954 char backup_mcast_ifn[IP_VS_IFNAME_MAXLEN]; 955 /* net name space ptr */ 956 struct net *net; /* Needed by timer routines */ 957 /* Number of heterogeneous destinations, needed becaus heterogeneous 958 * are not supported when synchronization is enabled. 959 */ 960 unsigned int mixed_address_family_dests; 961 }; 962 963 #define DEFAULT_SYNC_THRESHOLD 3 964 #define DEFAULT_SYNC_PERIOD 50 965 #define DEFAULT_SYNC_VER 1 966 #define DEFAULT_SLOPPY_TCP 0 967 #define DEFAULT_SLOPPY_SCTP 0 968 #define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ) 969 #define DEFAULT_SYNC_RETRIES 0 970 #define IPVS_SYNC_WAKEUP_RATE 8 971 #define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4) 972 #define IPVS_SYNC_SEND_DELAY (HZ / 50) 973 #define IPVS_SYNC_CHECK_PERIOD HZ 974 #define IPVS_SYNC_FLUSH_TIME (HZ * 2) 975 #define IPVS_SYNC_PORTS_MAX (1 << 6) 976 977 #ifdef CONFIG_SYSCTL 978 979 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs) 980 { 981 return ipvs->sysctl_sync_threshold[0]; 982 } 983 984 static inline int sysctl_sync_period(struct netns_ipvs *ipvs) 985 { 986 return ACCESS_ONCE(ipvs->sysctl_sync_threshold[1]); 987 } 988 989 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs) 990 { 991 return ACCESS_ONCE(ipvs->sysctl_sync_refresh_period); 992 } 993 994 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs) 995 { 996 return ipvs->sysctl_sync_retries; 997 } 998 999 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs) 1000 { 1001 return ipvs->sysctl_sync_ver; 1002 } 1003 1004 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs) 1005 { 1006 return ipvs->sysctl_sloppy_tcp; 1007 } 1008 1009 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs) 1010 { 1011 return ipvs->sysctl_sloppy_sctp; 1012 } 1013 1014 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs) 1015 { 1016 return ACCESS_ONCE(ipvs->sysctl_sync_ports); 1017 } 1018 1019 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs) 1020 { 1021 return ipvs->sysctl_sync_persist_mode; 1022 } 1023 1024 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs) 1025 { 1026 return ipvs->sysctl_sync_qlen_max; 1027 } 1028 1029 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs) 1030 { 1031 return ipvs->sysctl_sync_sock_size; 1032 } 1033 1034 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs) 1035 { 1036 return ipvs->sysctl_pmtu_disc; 1037 } 1038 1039 static inline int sysctl_backup_only(struct netns_ipvs *ipvs) 1040 { 1041 return ipvs->sync_state & IP_VS_STATE_BACKUP && 1042 ipvs->sysctl_backup_only; 1043 } 1044 1045 #else 1046 1047 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs) 1048 { 1049 return DEFAULT_SYNC_THRESHOLD; 1050 } 1051 1052 static inline int sysctl_sync_period(struct netns_ipvs *ipvs) 1053 { 1054 return DEFAULT_SYNC_PERIOD; 1055 } 1056 1057 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs) 1058 { 1059 return DEFAULT_SYNC_REFRESH_PERIOD; 1060 } 1061 1062 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs) 1063 { 1064 return DEFAULT_SYNC_RETRIES & 3; 1065 } 1066 1067 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs) 1068 { 1069 return DEFAULT_SYNC_VER; 1070 } 1071 1072 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs) 1073 { 1074 return DEFAULT_SLOPPY_TCP; 1075 } 1076 1077 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs) 1078 { 1079 return DEFAULT_SLOPPY_SCTP; 1080 } 1081 1082 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs) 1083 { 1084 return 1; 1085 } 1086 1087 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs) 1088 { 1089 return 0; 1090 } 1091 1092 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs) 1093 { 1094 return IPVS_SYNC_QLEN_MAX; 1095 } 1096 1097 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs) 1098 { 1099 return 0; 1100 } 1101 1102 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs) 1103 { 1104 return 1; 1105 } 1106 1107 static inline int sysctl_backup_only(struct netns_ipvs *ipvs) 1108 { 1109 return 0; 1110 } 1111 1112 #endif 1113 1114 /* IPVS core functions 1115 * (from ip_vs_core.c) 1116 */ 1117 const char *ip_vs_proto_name(unsigned int proto); 1118 void ip_vs_init_hash_table(struct list_head *table, int rows); 1119 #define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t))) 1120 1121 #define IP_VS_APP_TYPE_FTP 1 1122 1123 /* ip_vs_conn handling functions 1124 * (from ip_vs_conn.c) 1125 */ 1126 enum { 1127 IP_VS_DIR_INPUT = 0, 1128 IP_VS_DIR_OUTPUT, 1129 IP_VS_DIR_INPUT_ONLY, 1130 IP_VS_DIR_LAST, 1131 }; 1132 1133 static inline void ip_vs_conn_fill_param(struct net *net, int af, int protocol, 1134 const union nf_inet_addr *caddr, 1135 __be16 cport, 1136 const union nf_inet_addr *vaddr, 1137 __be16 vport, 1138 struct ip_vs_conn_param *p) 1139 { 1140 p->net = net; 1141 p->af = af; 1142 p->protocol = protocol; 1143 p->caddr = caddr; 1144 p->cport = cport; 1145 p->vaddr = vaddr; 1146 p->vport = vport; 1147 p->pe = NULL; 1148 p->pe_data = NULL; 1149 } 1150 1151 struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p); 1152 struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p); 1153 1154 struct ip_vs_conn * ip_vs_conn_in_get_proto(int af, const struct sk_buff *skb, 1155 const struct ip_vs_iphdr *iph, 1156 int inverse); 1157 1158 struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p); 1159 1160 struct ip_vs_conn * ip_vs_conn_out_get_proto(int af, const struct sk_buff *skb, 1161 const struct ip_vs_iphdr *iph, 1162 int inverse); 1163 1164 /* Get reference to gain full access to conn. 1165 * By default, RCU read-side critical sections have access only to 1166 * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference. 1167 */ 1168 static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp) 1169 { 1170 return atomic_inc_not_zero(&cp->refcnt); 1171 } 1172 1173 /* put back the conn without restarting its timer */ 1174 static inline void __ip_vs_conn_put(struct ip_vs_conn *cp) 1175 { 1176 smp_mb__before_atomic(); 1177 atomic_dec(&cp->refcnt); 1178 } 1179 void ip_vs_conn_put(struct ip_vs_conn *cp); 1180 void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport); 1181 1182 struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, int dest_af, 1183 const union nf_inet_addr *daddr, 1184 __be16 dport, unsigned int flags, 1185 struct ip_vs_dest *dest, __u32 fwmark); 1186 void ip_vs_conn_expire_now(struct ip_vs_conn *cp); 1187 1188 const char *ip_vs_state_name(__u16 proto, int state); 1189 1190 void ip_vs_tcp_conn_listen(struct net *net, struct ip_vs_conn *cp); 1191 int ip_vs_check_template(struct ip_vs_conn *ct); 1192 void ip_vs_random_dropentry(struct net *net); 1193 int ip_vs_conn_init(void); 1194 void ip_vs_conn_cleanup(void); 1195 1196 static inline void ip_vs_control_del(struct ip_vs_conn *cp) 1197 { 1198 struct ip_vs_conn *ctl_cp = cp->control; 1199 if (!ctl_cp) { 1200 IP_VS_ERR_BUF("request control DEL for uncontrolled: " 1201 "%s:%d to %s:%d\n", 1202 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1203 ntohs(cp->cport), 1204 IP_VS_DBG_ADDR(cp->af, &cp->vaddr), 1205 ntohs(cp->vport)); 1206 1207 return; 1208 } 1209 1210 IP_VS_DBG_BUF(7, "DELeting control for: " 1211 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n", 1212 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1213 ntohs(cp->cport), 1214 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr), 1215 ntohs(ctl_cp->cport)); 1216 1217 cp->control = NULL; 1218 if (atomic_read(&ctl_cp->n_control) == 0) { 1219 IP_VS_ERR_BUF("BUG control DEL with n=0 : " 1220 "%s:%d to %s:%d\n", 1221 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1222 ntohs(cp->cport), 1223 IP_VS_DBG_ADDR(cp->af, &cp->vaddr), 1224 ntohs(cp->vport)); 1225 1226 return; 1227 } 1228 atomic_dec(&ctl_cp->n_control); 1229 } 1230 1231 static inline void 1232 ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp) 1233 { 1234 if (cp->control) { 1235 IP_VS_ERR_BUF("request control ADD for already controlled: " 1236 "%s:%d to %s:%d\n", 1237 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1238 ntohs(cp->cport), 1239 IP_VS_DBG_ADDR(cp->af, &cp->vaddr), 1240 ntohs(cp->vport)); 1241 1242 ip_vs_control_del(cp); 1243 } 1244 1245 IP_VS_DBG_BUF(7, "ADDing control for: " 1246 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n", 1247 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1248 ntohs(cp->cport), 1249 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr), 1250 ntohs(ctl_cp->cport)); 1251 1252 cp->control = ctl_cp; 1253 atomic_inc(&ctl_cp->n_control); 1254 } 1255 1256 /* IPVS netns init & cleanup functions */ 1257 int ip_vs_estimator_net_init(struct net *net); 1258 int ip_vs_control_net_init(struct net *net); 1259 int ip_vs_protocol_net_init(struct net *net); 1260 int ip_vs_app_net_init(struct net *net); 1261 int ip_vs_conn_net_init(struct net *net); 1262 int ip_vs_sync_net_init(struct net *net); 1263 void ip_vs_conn_net_cleanup(struct net *net); 1264 void ip_vs_app_net_cleanup(struct net *net); 1265 void ip_vs_protocol_net_cleanup(struct net *net); 1266 void ip_vs_control_net_cleanup(struct net *net); 1267 void ip_vs_estimator_net_cleanup(struct net *net); 1268 void ip_vs_sync_net_cleanup(struct net *net); 1269 void ip_vs_service_net_cleanup(struct net *net); 1270 1271 /* IPVS application functions 1272 * (from ip_vs_app.c) 1273 */ 1274 #define IP_VS_APP_MAX_PORTS 8 1275 struct ip_vs_app *register_ip_vs_app(struct net *net, struct ip_vs_app *app); 1276 void unregister_ip_vs_app(struct net *net, struct ip_vs_app *app); 1277 int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp); 1278 void ip_vs_unbind_app(struct ip_vs_conn *cp); 1279 int register_ip_vs_app_inc(struct net *net, struct ip_vs_app *app, __u16 proto, 1280 __u16 port); 1281 int ip_vs_app_inc_get(struct ip_vs_app *inc); 1282 void ip_vs_app_inc_put(struct ip_vs_app *inc); 1283 1284 int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb); 1285 int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb); 1286 1287 int register_ip_vs_pe(struct ip_vs_pe *pe); 1288 int unregister_ip_vs_pe(struct ip_vs_pe *pe); 1289 struct ip_vs_pe *ip_vs_pe_getbyname(const char *name); 1290 struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name); 1291 1292 /* Use a #define to avoid all of module.h just for these trivial ops */ 1293 #define ip_vs_pe_get(pe) \ 1294 if (pe && pe->module) \ 1295 __module_get(pe->module); 1296 1297 #define ip_vs_pe_put(pe) \ 1298 if (pe && pe->module) \ 1299 module_put(pe->module); 1300 1301 /* IPVS protocol functions (from ip_vs_proto.c) */ 1302 int ip_vs_protocol_init(void); 1303 void ip_vs_protocol_cleanup(void); 1304 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags); 1305 int *ip_vs_create_timeout_table(int *table, int size); 1306 int ip_vs_set_state_timeout(int *table, int num, const char *const *names, 1307 const char *name, int to); 1308 void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp, 1309 const struct sk_buff *skb, int offset, 1310 const char *msg); 1311 1312 extern struct ip_vs_protocol ip_vs_protocol_tcp; 1313 extern struct ip_vs_protocol ip_vs_protocol_udp; 1314 extern struct ip_vs_protocol ip_vs_protocol_icmp; 1315 extern struct ip_vs_protocol ip_vs_protocol_esp; 1316 extern struct ip_vs_protocol ip_vs_protocol_ah; 1317 extern struct ip_vs_protocol ip_vs_protocol_sctp; 1318 1319 /* Registering/unregistering scheduler functions 1320 * (from ip_vs_sched.c) 1321 */ 1322 int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler); 1323 int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler); 1324 int ip_vs_bind_scheduler(struct ip_vs_service *svc, 1325 struct ip_vs_scheduler *scheduler); 1326 void ip_vs_unbind_scheduler(struct ip_vs_service *svc, 1327 struct ip_vs_scheduler *sched); 1328 struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name); 1329 void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler); 1330 struct ip_vs_conn * 1331 ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb, 1332 struct ip_vs_proto_data *pd, int *ignored, 1333 struct ip_vs_iphdr *iph); 1334 int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb, 1335 struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph); 1336 1337 void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg); 1338 1339 /* IPVS control data and functions (from ip_vs_ctl.c) */ 1340 extern struct ip_vs_stats ip_vs_stats; 1341 extern int sysctl_ip_vs_sync_ver; 1342 1343 struct ip_vs_service * 1344 ip_vs_service_find(struct net *net, int af, __u32 fwmark, __u16 protocol, 1345 const union nf_inet_addr *vaddr, __be16 vport); 1346 1347 bool ip_vs_has_real_service(struct net *net, int af, __u16 protocol, 1348 const union nf_inet_addr *daddr, __be16 dport); 1349 1350 int ip_vs_use_count_inc(void); 1351 void ip_vs_use_count_dec(void); 1352 int ip_vs_register_nl_ioctl(void); 1353 void ip_vs_unregister_nl_ioctl(void); 1354 int ip_vs_control_init(void); 1355 void ip_vs_control_cleanup(void); 1356 struct ip_vs_dest * 1357 ip_vs_find_dest(struct net *net, int svc_af, int dest_af, 1358 const union nf_inet_addr *daddr, __be16 dport, 1359 const union nf_inet_addr *vaddr, __be16 vport, 1360 __u16 protocol, __u32 fwmark, __u32 flags); 1361 void ip_vs_try_bind_dest(struct ip_vs_conn *cp); 1362 1363 static inline void ip_vs_dest_hold(struct ip_vs_dest *dest) 1364 { 1365 atomic_inc(&dest->refcnt); 1366 } 1367 1368 static inline void ip_vs_dest_put(struct ip_vs_dest *dest) 1369 { 1370 smp_mb__before_atomic(); 1371 atomic_dec(&dest->refcnt); 1372 } 1373 1374 static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest) 1375 { 1376 if (atomic_dec_return(&dest->refcnt) < 0) 1377 kfree(dest); 1378 } 1379 1380 /* IPVS sync daemon data and function prototypes 1381 * (from ip_vs_sync.c) 1382 */ 1383 int start_sync_thread(struct net *net, int state, char *mcast_ifn, __u8 syncid); 1384 int stop_sync_thread(struct net *net, int state); 1385 void ip_vs_sync_conn(struct net *net, struct ip_vs_conn *cp, int pkts); 1386 1387 /* IPVS rate estimator prototypes (from ip_vs_est.c) */ 1388 void ip_vs_start_estimator(struct net *net, struct ip_vs_stats *stats); 1389 void ip_vs_stop_estimator(struct net *net, struct ip_vs_stats *stats); 1390 void ip_vs_zero_estimator(struct ip_vs_stats *stats); 1391 void ip_vs_read_estimator(struct ip_vs_stats_user *dst, 1392 struct ip_vs_stats *stats); 1393 1394 /* Various IPVS packet transmitters (from ip_vs_xmit.c) */ 1395 int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1396 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1397 int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1398 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1399 int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1400 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1401 int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1402 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1403 int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1404 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1405 int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1406 struct ip_vs_protocol *pp, int offset, 1407 unsigned int hooknum, struct ip_vs_iphdr *iph); 1408 void ip_vs_dest_dst_rcu_free(struct rcu_head *head); 1409 1410 #ifdef CONFIG_IP_VS_IPV6 1411 int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1412 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1413 int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1414 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1415 int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1416 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1417 int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1418 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1419 int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1420 struct ip_vs_protocol *pp, int offset, 1421 unsigned int hooknum, struct ip_vs_iphdr *iph); 1422 #endif 1423 1424 #ifdef CONFIG_SYSCTL 1425 /* This is a simple mechanism to ignore packets when 1426 * we are loaded. Just set ip_vs_drop_rate to 'n' and 1427 * we start to drop 1/rate of the packets 1428 */ 1429 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) 1430 { 1431 if (!ipvs->drop_rate) 1432 return 0; 1433 if (--ipvs->drop_counter > 0) 1434 return 0; 1435 ipvs->drop_counter = ipvs->drop_rate; 1436 return 1; 1437 } 1438 #else 1439 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; } 1440 #endif 1441 1442 /* ip_vs_fwd_tag returns the forwarding tag of the connection */ 1443 #define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK) 1444 1445 static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp) 1446 { 1447 char fwd; 1448 1449 switch (IP_VS_FWD_METHOD(cp)) { 1450 case IP_VS_CONN_F_MASQ: 1451 fwd = 'M'; break; 1452 case IP_VS_CONN_F_LOCALNODE: 1453 fwd = 'L'; break; 1454 case IP_VS_CONN_F_TUNNEL: 1455 fwd = 'T'; break; 1456 case IP_VS_CONN_F_DROUTE: 1457 fwd = 'R'; break; 1458 case IP_VS_CONN_F_BYPASS: 1459 fwd = 'B'; break; 1460 default: 1461 fwd = '?'; break; 1462 } 1463 return fwd; 1464 } 1465 1466 void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp, 1467 struct ip_vs_conn *cp, int dir); 1468 1469 #ifdef CONFIG_IP_VS_IPV6 1470 void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp, 1471 struct ip_vs_conn *cp, int dir); 1472 #endif 1473 1474 __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset); 1475 1476 static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum) 1477 { 1478 __be32 diff[2] = { ~old, new }; 1479 1480 return csum_partial(diff, sizeof(diff), oldsum); 1481 } 1482 1483 #ifdef CONFIG_IP_VS_IPV6 1484 static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new, 1485 __wsum oldsum) 1486 { 1487 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0], 1488 new[3], new[2], new[1], new[0] }; 1489 1490 return csum_partial(diff, sizeof(diff), oldsum); 1491 } 1492 #endif 1493 1494 static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum) 1495 { 1496 __be16 diff[2] = { ~old, new }; 1497 1498 return csum_partial(diff, sizeof(diff), oldsum); 1499 } 1500 1501 /* Forget current conntrack (unconfirmed) and attach notrack entry */ 1502 static inline void ip_vs_notrack(struct sk_buff *skb) 1503 { 1504 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 1505 enum ip_conntrack_info ctinfo; 1506 struct nf_conn *ct = nf_ct_get(skb, &ctinfo); 1507 1508 if (!ct || !nf_ct_is_untracked(ct)) { 1509 nf_conntrack_put(skb->nfct); 1510 skb->nfct = &nf_ct_untracked_get()->ct_general; 1511 skb->nfctinfo = IP_CT_NEW; 1512 nf_conntrack_get(skb->nfct); 1513 } 1514 #endif 1515 } 1516 1517 #ifdef CONFIG_IP_VS_NFCT 1518 /* Netfilter connection tracking 1519 * (from ip_vs_nfct.c) 1520 */ 1521 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs) 1522 { 1523 #ifdef CONFIG_SYSCTL 1524 return ipvs->sysctl_conntrack; 1525 #else 1526 return 0; 1527 #endif 1528 } 1529 1530 void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp, 1531 int outin); 1532 int ip_vs_confirm_conntrack(struct sk_buff *skb); 1533 void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct, 1534 struct ip_vs_conn *cp, u_int8_t proto, 1535 const __be16 port, int from_rs); 1536 void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp); 1537 1538 #else 1539 1540 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs) 1541 { 1542 return 0; 1543 } 1544 1545 static inline void ip_vs_update_conntrack(struct sk_buff *skb, 1546 struct ip_vs_conn *cp, int outin) 1547 { 1548 } 1549 1550 static inline int ip_vs_confirm_conntrack(struct sk_buff *skb) 1551 { 1552 return NF_ACCEPT; 1553 } 1554 1555 static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp) 1556 { 1557 } 1558 #endif /* CONFIG_IP_VS_NFCT */ 1559 1560 static inline int 1561 ip_vs_dest_conn_overhead(struct ip_vs_dest *dest) 1562 { 1563 /* We think the overhead of processing active connections is 256 1564 * times higher than that of inactive connections in average. (This 1565 * 256 times might not be accurate, we will change it later) We 1566 * use the following formula to estimate the overhead now: 1567 * dest->activeconns*256 + dest->inactconns 1568 */ 1569 return (atomic_read(&dest->activeconns) << 8) + 1570 atomic_read(&dest->inactconns); 1571 } 1572 1573 #endif /* _NET_IP_VS_H */ 1574