1 /* 2 * IP Virtual Server 3 * data structure and functionality definitions 4 */ 5 6 #ifndef _NET_IP_VS_H 7 #define _NET_IP_VS_H 8 9 #include <linux/ip_vs.h> /* definitions shared with userland */ 10 11 #include <asm/types.h> /* for __uXX types */ 12 13 #include <linux/list.h> /* for struct list_head */ 14 #include <linux/spinlock.h> /* for struct rwlock_t */ 15 #include <linux/atomic.h> /* for struct atomic_t */ 16 #include <linux/compiler.h> 17 #include <linux/timer.h> 18 #include <linux/bug.h> 19 20 #include <net/checksum.h> 21 #include <linux/netfilter.h> /* for union nf_inet_addr */ 22 #include <linux/ip.h> 23 #include <linux/ipv6.h> /* for struct ipv6hdr */ 24 #include <net/ipv6.h> 25 #if IS_ENABLED(CONFIG_IP_VS_IPV6) 26 #include <linux/netfilter_ipv6/ip6_tables.h> 27 #endif 28 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 29 #include <net/netfilter/nf_conntrack.h> 30 #endif 31 #include <net/net_namespace.h> /* Netw namespace */ 32 33 /* 34 * Generic access of ipvs struct 35 */ 36 static inline struct netns_ipvs *net_ipvs(struct net* net) 37 { 38 return net->ipvs; 39 } 40 /* 41 * Get net ptr from skb in traffic cases 42 * use skb_sknet when call is from userland (ioctl or netlink) 43 */ 44 static inline struct net *skb_net(const struct sk_buff *skb) 45 { 46 #ifdef CONFIG_NET_NS 47 #ifdef CONFIG_IP_VS_DEBUG 48 /* 49 * This is used for debug only. 50 * Start with the most likely hit 51 * End with BUG 52 */ 53 if (likely(skb->dev && skb->dev->nd_net)) 54 return dev_net(skb->dev); 55 if (skb_dst(skb) && skb_dst(skb)->dev) 56 return dev_net(skb_dst(skb)->dev); 57 WARN(skb->sk, "Maybe skb_sknet should be used in %s() at line:%d\n", 58 __func__, __LINE__); 59 if (likely(skb->sk && skb->sk->sk_net)) 60 return sock_net(skb->sk); 61 pr_err("There is no net ptr to find in the skb in %s() line:%d\n", 62 __func__, __LINE__); 63 BUG(); 64 #else 65 return dev_net(skb->dev ? : skb_dst(skb)->dev); 66 #endif 67 #else 68 return &init_net; 69 #endif 70 } 71 72 static inline struct net *skb_sknet(const struct sk_buff *skb) 73 { 74 #ifdef CONFIG_NET_NS 75 #ifdef CONFIG_IP_VS_DEBUG 76 /* Start with the most likely hit */ 77 if (likely(skb->sk && skb->sk->sk_net)) 78 return sock_net(skb->sk); 79 WARN(skb->dev, "Maybe skb_net should be used instead in %s() line:%d\n", 80 __func__, __LINE__); 81 if (likely(skb->dev && skb->dev->nd_net)) 82 return dev_net(skb->dev); 83 pr_err("There is no net ptr to find in the skb in %s() line:%d\n", 84 __func__, __LINE__); 85 BUG(); 86 #else 87 return sock_net(skb->sk); 88 #endif 89 #else 90 return &init_net; 91 #endif 92 } 93 /* 94 * This one needed for single_open_net since net is stored directly in 95 * private not as a struct i.e. seq_file_net can't be used. 96 */ 97 static inline struct net *seq_file_single_net(struct seq_file *seq) 98 { 99 #ifdef CONFIG_NET_NS 100 return (struct net *)seq->private; 101 #else 102 return &init_net; 103 #endif 104 } 105 106 /* Connections' size value needed by ip_vs_ctl.c */ 107 extern int ip_vs_conn_tab_size; 108 109 struct ip_vs_iphdr { 110 __u32 len; /* IPv4 simply where L4 starts 111 IPv6 where L4 Transport Header starts */ 112 __u16 fragoffs; /* IPv6 fragment offset, 0 if first frag (or not frag)*/ 113 __s16 protocol; 114 __s32 flags; 115 union nf_inet_addr saddr; 116 union nf_inet_addr daddr; 117 }; 118 119 static inline void *frag_safe_skb_hp(const struct sk_buff *skb, int offset, 120 int len, void *buffer, 121 const struct ip_vs_iphdr *ipvsh) 122 { 123 return skb_header_pointer(skb, offset, len, buffer); 124 } 125 126 static inline void 127 ip_vs_fill_ip4hdr(const void *nh, struct ip_vs_iphdr *iphdr) 128 { 129 const struct iphdr *iph = nh; 130 131 iphdr->len = iph->ihl * 4; 132 iphdr->fragoffs = 0; 133 iphdr->protocol = iph->protocol; 134 iphdr->saddr.ip = iph->saddr; 135 iphdr->daddr.ip = iph->daddr; 136 } 137 138 /* This function handles filling *ip_vs_iphdr, both for IPv4 and IPv6. 139 * IPv6 requires some extra work, as finding proper header position, 140 * depend on the IPv6 extension headers. 141 */ 142 static inline void 143 ip_vs_fill_iph_skb(int af, const struct sk_buff *skb, struct ip_vs_iphdr *iphdr) 144 { 145 #ifdef CONFIG_IP_VS_IPV6 146 if (af == AF_INET6) { 147 const struct ipv6hdr *iph = 148 (struct ipv6hdr *)skb_network_header(skb); 149 iphdr->saddr.in6 = iph->saddr; 150 iphdr->daddr.in6 = iph->daddr; 151 /* ipv6_find_hdr() updates len, flags */ 152 iphdr->len = 0; 153 iphdr->flags = 0; 154 iphdr->protocol = ipv6_find_hdr(skb, &iphdr->len, -1, 155 &iphdr->fragoffs, 156 &iphdr->flags); 157 } else 158 #endif 159 { 160 const struct iphdr *iph = 161 (struct iphdr *)skb_network_header(skb); 162 iphdr->len = iph->ihl * 4; 163 iphdr->fragoffs = 0; 164 iphdr->protocol = iph->protocol; 165 iphdr->saddr.ip = iph->saddr; 166 iphdr->daddr.ip = iph->daddr; 167 } 168 } 169 170 static inline void ip_vs_addr_copy(int af, union nf_inet_addr *dst, 171 const union nf_inet_addr *src) 172 { 173 #ifdef CONFIG_IP_VS_IPV6 174 if (af == AF_INET6) 175 dst->in6 = src->in6; 176 else 177 #endif 178 dst->ip = src->ip; 179 } 180 181 static inline void ip_vs_addr_set(int af, union nf_inet_addr *dst, 182 const union nf_inet_addr *src) 183 { 184 #ifdef CONFIG_IP_VS_IPV6 185 if (af == AF_INET6) { 186 dst->in6 = src->in6; 187 return; 188 } 189 #endif 190 dst->ip = src->ip; 191 dst->all[1] = 0; 192 dst->all[2] = 0; 193 dst->all[3] = 0; 194 } 195 196 static inline int ip_vs_addr_equal(int af, const union nf_inet_addr *a, 197 const union nf_inet_addr *b) 198 { 199 #ifdef CONFIG_IP_VS_IPV6 200 if (af == AF_INET6) 201 return ipv6_addr_equal(&a->in6, &b->in6); 202 #endif 203 return a->ip == b->ip; 204 } 205 206 #ifdef CONFIG_IP_VS_DEBUG 207 #include <linux/net.h> 208 209 int ip_vs_get_debug_level(void); 210 211 static inline const char *ip_vs_dbg_addr(int af, char *buf, size_t buf_len, 212 const union nf_inet_addr *addr, 213 int *idx) 214 { 215 int len; 216 #ifdef CONFIG_IP_VS_IPV6 217 if (af == AF_INET6) 218 len = snprintf(&buf[*idx], buf_len - *idx, "[%pI6c]", 219 &addr->in6) + 1; 220 else 221 #endif 222 len = snprintf(&buf[*idx], buf_len - *idx, "%pI4", 223 &addr->ip) + 1; 224 225 *idx += len; 226 BUG_ON(*idx > buf_len + 1); 227 return &buf[*idx - len]; 228 } 229 230 #define IP_VS_DBG_BUF(level, msg, ...) \ 231 do { \ 232 char ip_vs_dbg_buf[160]; \ 233 int ip_vs_dbg_idx = 0; \ 234 if (level <= ip_vs_get_debug_level()) \ 235 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \ 236 } while (0) 237 #define IP_VS_ERR_BUF(msg...) \ 238 do { \ 239 char ip_vs_dbg_buf[160]; \ 240 int ip_vs_dbg_idx = 0; \ 241 pr_err(msg); \ 242 } while (0) 243 244 /* Only use from within IP_VS_DBG_BUF() or IP_VS_ERR_BUF macros */ 245 #define IP_VS_DBG_ADDR(af, addr) \ 246 ip_vs_dbg_addr(af, ip_vs_dbg_buf, \ 247 sizeof(ip_vs_dbg_buf), addr, \ 248 &ip_vs_dbg_idx) 249 250 #define IP_VS_DBG(level, msg, ...) \ 251 do { \ 252 if (level <= ip_vs_get_debug_level()) \ 253 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \ 254 } while (0) 255 #define IP_VS_DBG_RL(msg, ...) \ 256 do { \ 257 if (net_ratelimit()) \ 258 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \ 259 } while (0) 260 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) \ 261 do { \ 262 if (level <= ip_vs_get_debug_level()) \ 263 pp->debug_packet(af, pp, skb, ofs, msg); \ 264 } while (0) 265 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) \ 266 do { \ 267 if (level <= ip_vs_get_debug_level() && \ 268 net_ratelimit()) \ 269 pp->debug_packet(af, pp, skb, ofs, msg); \ 270 } while (0) 271 #else /* NO DEBUGGING at ALL */ 272 #define IP_VS_DBG_BUF(level, msg...) do {} while (0) 273 #define IP_VS_ERR_BUF(msg...) do {} while (0) 274 #define IP_VS_DBG(level, msg...) do {} while (0) 275 #define IP_VS_DBG_RL(msg...) do {} while (0) 276 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) do {} while (0) 277 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) do {} while (0) 278 #endif 279 280 #define IP_VS_BUG() BUG() 281 #define IP_VS_ERR_RL(msg, ...) \ 282 do { \ 283 if (net_ratelimit()) \ 284 pr_err(msg, ##__VA_ARGS__); \ 285 } while (0) 286 287 #ifdef CONFIG_IP_VS_DEBUG 288 #define EnterFunction(level) \ 289 do { \ 290 if (level <= ip_vs_get_debug_level()) \ 291 printk(KERN_DEBUG \ 292 pr_fmt("Enter: %s, %s line %i\n"), \ 293 __func__, __FILE__, __LINE__); \ 294 } while (0) 295 #define LeaveFunction(level) \ 296 do { \ 297 if (level <= ip_vs_get_debug_level()) \ 298 printk(KERN_DEBUG \ 299 pr_fmt("Leave: %s, %s line %i\n"), \ 300 __func__, __FILE__, __LINE__); \ 301 } while (0) 302 #else 303 #define EnterFunction(level) do {} while (0) 304 #define LeaveFunction(level) do {} while (0) 305 #endif 306 307 308 /* 309 * The port number of FTP service (in network order). 310 */ 311 #define FTPPORT cpu_to_be16(21) 312 #define FTPDATA cpu_to_be16(20) 313 314 /* 315 * TCP State Values 316 */ 317 enum { 318 IP_VS_TCP_S_NONE = 0, 319 IP_VS_TCP_S_ESTABLISHED, 320 IP_VS_TCP_S_SYN_SENT, 321 IP_VS_TCP_S_SYN_RECV, 322 IP_VS_TCP_S_FIN_WAIT, 323 IP_VS_TCP_S_TIME_WAIT, 324 IP_VS_TCP_S_CLOSE, 325 IP_VS_TCP_S_CLOSE_WAIT, 326 IP_VS_TCP_S_LAST_ACK, 327 IP_VS_TCP_S_LISTEN, 328 IP_VS_TCP_S_SYNACK, 329 IP_VS_TCP_S_LAST 330 }; 331 332 /* 333 * UDP State Values 334 */ 335 enum { 336 IP_VS_UDP_S_NORMAL, 337 IP_VS_UDP_S_LAST, 338 }; 339 340 /* 341 * ICMP State Values 342 */ 343 enum { 344 IP_VS_ICMP_S_NORMAL, 345 IP_VS_ICMP_S_LAST, 346 }; 347 348 /* 349 * SCTP State Values 350 */ 351 enum ip_vs_sctp_states { 352 IP_VS_SCTP_S_NONE, 353 IP_VS_SCTP_S_INIT1, 354 IP_VS_SCTP_S_INIT, 355 IP_VS_SCTP_S_COOKIE_SENT, 356 IP_VS_SCTP_S_COOKIE_REPLIED, 357 IP_VS_SCTP_S_COOKIE_WAIT, 358 IP_VS_SCTP_S_COOKIE, 359 IP_VS_SCTP_S_COOKIE_ECHOED, 360 IP_VS_SCTP_S_ESTABLISHED, 361 IP_VS_SCTP_S_SHUTDOWN_SENT, 362 IP_VS_SCTP_S_SHUTDOWN_RECEIVED, 363 IP_VS_SCTP_S_SHUTDOWN_ACK_SENT, 364 IP_VS_SCTP_S_REJECTED, 365 IP_VS_SCTP_S_CLOSED, 366 IP_VS_SCTP_S_LAST 367 }; 368 369 /* 370 * Delta sequence info structure 371 * Each ip_vs_conn has 2 (output AND input seq. changes). 372 * Only used in the VS/NAT. 373 */ 374 struct ip_vs_seq { 375 __u32 init_seq; /* Add delta from this seq */ 376 __u32 delta; /* Delta in sequence numbers */ 377 __u32 previous_delta; /* Delta in sequence numbers 378 before last resized pkt */ 379 }; 380 381 /* 382 * counters per cpu 383 */ 384 struct ip_vs_counters { 385 __u32 conns; /* connections scheduled */ 386 __u32 inpkts; /* incoming packets */ 387 __u32 outpkts; /* outgoing packets */ 388 __u64 inbytes; /* incoming bytes */ 389 __u64 outbytes; /* outgoing bytes */ 390 }; 391 /* 392 * Stats per cpu 393 */ 394 struct ip_vs_cpu_stats { 395 struct ip_vs_counters ustats; 396 struct u64_stats_sync syncp; 397 }; 398 399 /* 400 * IPVS statistics objects 401 */ 402 struct ip_vs_estimator { 403 struct list_head list; 404 405 u64 last_inbytes; 406 u64 last_outbytes; 407 u32 last_conns; 408 u32 last_inpkts; 409 u32 last_outpkts; 410 411 u32 cps; 412 u32 inpps; 413 u32 outpps; 414 u32 inbps; 415 u32 outbps; 416 }; 417 418 struct ip_vs_stats { 419 struct ip_vs_stats_user ustats; /* statistics */ 420 struct ip_vs_estimator est; /* estimator */ 421 struct ip_vs_cpu_stats __percpu *cpustats; /* per cpu counters */ 422 spinlock_t lock; /* spin lock */ 423 struct ip_vs_stats_user ustats0; /* reset values */ 424 }; 425 426 struct dst_entry; 427 struct iphdr; 428 struct ip_vs_conn; 429 struct ip_vs_app; 430 struct sk_buff; 431 struct ip_vs_proto_data; 432 433 struct ip_vs_protocol { 434 struct ip_vs_protocol *next; 435 char *name; 436 u16 protocol; 437 u16 num_states; 438 int dont_defrag; 439 440 void (*init)(struct ip_vs_protocol *pp); 441 442 void (*exit)(struct ip_vs_protocol *pp); 443 444 int (*init_netns)(struct net *net, struct ip_vs_proto_data *pd); 445 446 void (*exit_netns)(struct net *net, struct ip_vs_proto_data *pd); 447 448 int (*conn_schedule)(int af, struct sk_buff *skb, 449 struct ip_vs_proto_data *pd, 450 int *verdict, struct ip_vs_conn **cpp, 451 struct ip_vs_iphdr *iph); 452 453 struct ip_vs_conn * 454 (*conn_in_get)(int af, 455 const struct sk_buff *skb, 456 const struct ip_vs_iphdr *iph, 457 int inverse); 458 459 struct ip_vs_conn * 460 (*conn_out_get)(int af, 461 const struct sk_buff *skb, 462 const struct ip_vs_iphdr *iph, 463 int inverse); 464 465 int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp, 466 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph); 467 468 int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp, 469 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph); 470 471 int (*csum_check)(int af, struct sk_buff *skb, 472 struct ip_vs_protocol *pp); 473 474 const char *(*state_name)(int state); 475 476 void (*state_transition)(struct ip_vs_conn *cp, int direction, 477 const struct sk_buff *skb, 478 struct ip_vs_proto_data *pd); 479 480 int (*register_app)(struct net *net, struct ip_vs_app *inc); 481 482 void (*unregister_app)(struct net *net, struct ip_vs_app *inc); 483 484 int (*app_conn_bind)(struct ip_vs_conn *cp); 485 486 void (*debug_packet)(int af, struct ip_vs_protocol *pp, 487 const struct sk_buff *skb, 488 int offset, 489 const char *msg); 490 491 void (*timeout_change)(struct ip_vs_proto_data *pd, int flags); 492 }; 493 494 /* 495 * protocol data per netns 496 */ 497 struct ip_vs_proto_data { 498 struct ip_vs_proto_data *next; 499 struct ip_vs_protocol *pp; 500 int *timeout_table; /* protocol timeout table */ 501 atomic_t appcnt; /* counter of proto app incs. */ 502 struct tcp_states_t *tcp_state_table; 503 }; 504 505 struct ip_vs_protocol *ip_vs_proto_get(unsigned short proto); 506 struct ip_vs_proto_data *ip_vs_proto_data_get(struct net *net, 507 unsigned short proto); 508 509 struct ip_vs_conn_param { 510 struct net *net; 511 const union nf_inet_addr *caddr; 512 const union nf_inet_addr *vaddr; 513 __be16 cport; 514 __be16 vport; 515 __u16 protocol; 516 u16 af; 517 518 const struct ip_vs_pe *pe; 519 char *pe_data; 520 __u8 pe_data_len; 521 }; 522 523 /* 524 * IP_VS structure allocated for each dynamically scheduled connection 525 */ 526 struct ip_vs_conn { 527 struct hlist_node c_list; /* hashed list heads */ 528 /* Protocol, addresses and port numbers */ 529 __be16 cport; 530 __be16 dport; 531 __be16 vport; 532 u16 af; /* address family */ 533 union nf_inet_addr caddr; /* client address */ 534 union nf_inet_addr vaddr; /* virtual address */ 535 union nf_inet_addr daddr; /* destination address */ 536 volatile __u32 flags; /* status flags */ 537 __u16 protocol; /* Which protocol (TCP/UDP) */ 538 #ifdef CONFIG_NET_NS 539 struct net *net; /* Name space */ 540 #endif 541 542 /* counter and timer */ 543 atomic_t refcnt; /* reference count */ 544 struct timer_list timer; /* Expiration timer */ 545 volatile unsigned long timeout; /* timeout */ 546 547 /* Flags and state transition */ 548 spinlock_t lock; /* lock for state transition */ 549 volatile __u16 state; /* state info */ 550 volatile __u16 old_state; /* old state, to be used for 551 * state transition triggerd 552 * synchronization 553 */ 554 __u32 fwmark; /* Fire wall mark from skb */ 555 unsigned long sync_endtime; /* jiffies + sent_retries */ 556 557 /* Control members */ 558 struct ip_vs_conn *control; /* Master control connection */ 559 atomic_t n_control; /* Number of controlled ones */ 560 struct ip_vs_dest *dest; /* real server */ 561 atomic_t in_pkts; /* incoming packet counter */ 562 563 /* packet transmitter for different forwarding methods. If it 564 mangles the packet, it must return NF_DROP or better NF_STOLEN, 565 otherwise this must be changed to a sk_buff **. 566 NF_ACCEPT can be returned when destination is local. 567 */ 568 int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp, 569 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 570 571 /* Note: we can group the following members into a structure, 572 in order to save more space, and the following members are 573 only used in VS/NAT anyway */ 574 struct ip_vs_app *app; /* bound ip_vs_app object */ 575 void *app_data; /* Application private data */ 576 struct ip_vs_seq in_seq; /* incoming seq. struct */ 577 struct ip_vs_seq out_seq; /* outgoing seq. struct */ 578 579 const struct ip_vs_pe *pe; 580 char *pe_data; 581 __u8 pe_data_len; 582 583 struct rcu_head rcu_head; 584 }; 585 586 /* 587 * To save some memory in conn table when name space is disabled. 588 */ 589 static inline struct net *ip_vs_conn_net(const struct ip_vs_conn *cp) 590 { 591 #ifdef CONFIG_NET_NS 592 return cp->net; 593 #else 594 return &init_net; 595 #endif 596 } 597 static inline void ip_vs_conn_net_set(struct ip_vs_conn *cp, struct net *net) 598 { 599 #ifdef CONFIG_NET_NS 600 cp->net = net; 601 #endif 602 } 603 604 static inline int ip_vs_conn_net_eq(const struct ip_vs_conn *cp, 605 struct net *net) 606 { 607 #ifdef CONFIG_NET_NS 608 return cp->net == net; 609 #else 610 return 1; 611 #endif 612 } 613 614 /* 615 * Extended internal versions of struct ip_vs_service_user and 616 * ip_vs_dest_user for IPv6 support. 617 * 618 * We need these to conveniently pass around service and destination 619 * options, but unfortunately, we also need to keep the old definitions to 620 * maintain userspace backwards compatibility for the setsockopt interface. 621 */ 622 struct ip_vs_service_user_kern { 623 /* virtual service addresses */ 624 u16 af; 625 u16 protocol; 626 union nf_inet_addr addr; /* virtual ip address */ 627 __be16 port; 628 u32 fwmark; /* firwall mark of service */ 629 630 /* virtual service options */ 631 char *sched_name; 632 char *pe_name; 633 unsigned int flags; /* virtual service flags */ 634 unsigned int timeout; /* persistent timeout in sec */ 635 __be32 netmask; /* persistent netmask or plen */ 636 }; 637 638 639 struct ip_vs_dest_user_kern { 640 /* destination server address */ 641 union nf_inet_addr addr; 642 __be16 port; 643 644 /* real server options */ 645 unsigned int conn_flags; /* connection flags */ 646 int weight; /* destination weight */ 647 648 /* thresholds for active connections */ 649 u32 u_threshold; /* upper threshold */ 650 u32 l_threshold; /* lower threshold */ 651 }; 652 653 654 /* 655 * The information about the virtual service offered to the net 656 * and the forwarding entries 657 */ 658 struct ip_vs_service { 659 struct hlist_node s_list; /* for normal service table */ 660 struct hlist_node f_list; /* for fwmark-based service table */ 661 atomic_t refcnt; /* reference counter */ 662 663 u16 af; /* address family */ 664 __u16 protocol; /* which protocol (TCP/UDP) */ 665 union nf_inet_addr addr; /* IP address for virtual service */ 666 __be16 port; /* port number for the service */ 667 __u32 fwmark; /* firewall mark of the service */ 668 unsigned int flags; /* service status flags */ 669 unsigned int timeout; /* persistent timeout in ticks */ 670 __be32 netmask; /* grouping granularity, mask/plen */ 671 struct net *net; 672 673 struct list_head destinations; /* real server d-linked list */ 674 __u32 num_dests; /* number of servers */ 675 struct ip_vs_stats stats; /* statistics for the service */ 676 677 /* for scheduling */ 678 struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */ 679 spinlock_t sched_lock; /* lock sched_data */ 680 void *sched_data; /* scheduler application data */ 681 682 /* alternate persistence engine */ 683 struct ip_vs_pe __rcu *pe; 684 685 struct rcu_head rcu_head; 686 }; 687 688 /* Information for cached dst */ 689 struct ip_vs_dest_dst { 690 struct dst_entry *dst_cache; /* destination cache entry */ 691 u32 dst_cookie; 692 union nf_inet_addr dst_saddr; 693 struct rcu_head rcu_head; 694 }; 695 696 /* 697 * The real server destination forwarding entry 698 * with ip address, port number, and so on. 699 */ 700 struct ip_vs_dest { 701 struct list_head n_list; /* for the dests in the service */ 702 struct hlist_node d_list; /* for table with all the dests */ 703 704 u16 af; /* address family */ 705 __be16 port; /* port number of the server */ 706 union nf_inet_addr addr; /* IP address of the server */ 707 volatile unsigned int flags; /* dest status flags */ 708 atomic_t conn_flags; /* flags to copy to conn */ 709 atomic_t weight; /* server weight */ 710 711 atomic_t refcnt; /* reference counter */ 712 struct ip_vs_stats stats; /* statistics */ 713 unsigned long idle_start; /* start time, jiffies */ 714 715 /* connection counters and thresholds */ 716 atomic_t activeconns; /* active connections */ 717 atomic_t inactconns; /* inactive connections */ 718 atomic_t persistconns; /* persistent connections */ 719 __u32 u_threshold; /* upper threshold */ 720 __u32 l_threshold; /* lower threshold */ 721 722 /* for destination cache */ 723 spinlock_t dst_lock; /* lock of dst_cache */ 724 struct ip_vs_dest_dst __rcu *dest_dst; /* cached dst info */ 725 726 /* for virtual service */ 727 struct ip_vs_service __rcu *svc; /* service it belongs to */ 728 __u16 protocol; /* which protocol (TCP/UDP) */ 729 __be16 vport; /* virtual port number */ 730 union nf_inet_addr vaddr; /* virtual IP address */ 731 __u32 vfwmark; /* firewall mark of service */ 732 733 struct list_head t_list; /* in dest_trash */ 734 unsigned int in_rs_table:1; /* we are in rs_table */ 735 }; 736 737 738 /* 739 * The scheduler object 740 */ 741 struct ip_vs_scheduler { 742 struct list_head n_list; /* d-linked list head */ 743 char *name; /* scheduler name */ 744 atomic_t refcnt; /* reference counter */ 745 struct module *module; /* THIS_MODULE/NULL */ 746 747 /* scheduler initializing service */ 748 int (*init_service)(struct ip_vs_service *svc); 749 /* scheduling service finish */ 750 void (*done_service)(struct ip_vs_service *svc); 751 /* dest is linked */ 752 int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest); 753 /* dest is unlinked */ 754 int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest); 755 /* dest is updated */ 756 int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest); 757 758 /* selecting a server from the given service */ 759 struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc, 760 const struct sk_buff *skb, 761 struct ip_vs_iphdr *iph); 762 }; 763 764 /* The persistence engine object */ 765 struct ip_vs_pe { 766 struct list_head n_list; /* d-linked list head */ 767 char *name; /* scheduler name */ 768 atomic_t refcnt; /* reference counter */ 769 struct module *module; /* THIS_MODULE/NULL */ 770 771 /* get the connection template, if any */ 772 int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb); 773 bool (*ct_match)(const struct ip_vs_conn_param *p, 774 struct ip_vs_conn *ct); 775 u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval, 776 bool inverse); 777 int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf); 778 }; 779 780 /* 781 * The application module object (a.k.a. app incarnation) 782 */ 783 struct ip_vs_app { 784 struct list_head a_list; /* member in app list */ 785 int type; /* IP_VS_APP_TYPE_xxx */ 786 char *name; /* application module name */ 787 __u16 protocol; 788 struct module *module; /* THIS_MODULE/NULL */ 789 struct list_head incs_list; /* list of incarnations */ 790 791 /* members for application incarnations */ 792 struct list_head p_list; /* member in proto app list */ 793 struct ip_vs_app *app; /* its real application */ 794 __be16 port; /* port number in net order */ 795 atomic_t usecnt; /* usage counter */ 796 struct rcu_head rcu_head; 797 798 /* 799 * output hook: Process packet in inout direction, diff set for TCP. 800 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok, 801 * 2=Mangled but checksum was not updated 802 */ 803 int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *, 804 struct sk_buff *, int *diff); 805 806 /* 807 * input hook: Process packet in outin direction, diff set for TCP. 808 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok, 809 * 2=Mangled but checksum was not updated 810 */ 811 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *, 812 struct sk_buff *, int *diff); 813 814 /* ip_vs_app initializer */ 815 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *); 816 817 /* ip_vs_app finish */ 818 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *); 819 820 821 /* not used now */ 822 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *, 823 struct ip_vs_protocol *); 824 825 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *); 826 827 int * timeout_table; 828 int * timeouts; 829 int timeouts_size; 830 831 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app, 832 int *verdict, struct ip_vs_conn **cpp); 833 834 struct ip_vs_conn * 835 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app, 836 const struct iphdr *iph, int inverse); 837 838 struct ip_vs_conn * 839 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app, 840 const struct iphdr *iph, int inverse); 841 842 int (*state_transition)(struct ip_vs_conn *cp, int direction, 843 const struct sk_buff *skb, 844 struct ip_vs_app *app); 845 846 void (*timeout_change)(struct ip_vs_app *app, int flags); 847 }; 848 849 struct ipvs_master_sync_state { 850 struct list_head sync_queue; 851 struct ip_vs_sync_buff *sync_buff; 852 unsigned long sync_queue_len; 853 unsigned int sync_queue_delay; 854 struct task_struct *master_thread; 855 struct delayed_work master_wakeup_work; 856 struct netns_ipvs *ipvs; 857 }; 858 859 /* How much time to keep dests in trash */ 860 #define IP_VS_DEST_TRASH_PERIOD (120 * HZ) 861 862 /* IPVS in network namespace */ 863 struct netns_ipvs { 864 int gen; /* Generation */ 865 int enable; /* enable like nf_hooks do */ 866 /* 867 * Hash table: for real service lookups 868 */ 869 #define IP_VS_RTAB_BITS 4 870 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS) 871 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1) 872 873 struct hlist_head rs_table[IP_VS_RTAB_SIZE]; 874 /* ip_vs_app */ 875 struct list_head app_list; 876 /* ip_vs_proto */ 877 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */ 878 struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE]; 879 /* ip_vs_proto_tcp */ 880 #ifdef CONFIG_IP_VS_PROTO_TCP 881 #define TCP_APP_TAB_BITS 4 882 #define TCP_APP_TAB_SIZE (1 << TCP_APP_TAB_BITS) 883 #define TCP_APP_TAB_MASK (TCP_APP_TAB_SIZE - 1) 884 struct list_head tcp_apps[TCP_APP_TAB_SIZE]; 885 #endif 886 /* ip_vs_proto_udp */ 887 #ifdef CONFIG_IP_VS_PROTO_UDP 888 #define UDP_APP_TAB_BITS 4 889 #define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS) 890 #define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1) 891 struct list_head udp_apps[UDP_APP_TAB_SIZE]; 892 #endif 893 /* ip_vs_proto_sctp */ 894 #ifdef CONFIG_IP_VS_PROTO_SCTP 895 #define SCTP_APP_TAB_BITS 4 896 #define SCTP_APP_TAB_SIZE (1 << SCTP_APP_TAB_BITS) 897 #define SCTP_APP_TAB_MASK (SCTP_APP_TAB_SIZE - 1) 898 /* Hash table for SCTP application incarnations */ 899 struct list_head sctp_apps[SCTP_APP_TAB_SIZE]; 900 #endif 901 /* ip_vs_conn */ 902 atomic_t conn_count; /* connection counter */ 903 904 /* ip_vs_ctl */ 905 struct ip_vs_stats tot_stats; /* Statistics & est. */ 906 907 int num_services; /* no of virtual services */ 908 909 /* Trash for destinations */ 910 struct list_head dest_trash; 911 spinlock_t dest_trash_lock; 912 struct timer_list dest_trash_timer; /* expiration timer */ 913 /* Service counters */ 914 atomic_t ftpsvc_counter; 915 atomic_t nullsvc_counter; 916 917 #ifdef CONFIG_SYSCTL 918 /* 1/rate drop and drop-entry variables */ 919 struct delayed_work defense_work; /* Work handler */ 920 int drop_rate; 921 int drop_counter; 922 atomic_t dropentry; 923 /* locks in ctl.c */ 924 spinlock_t dropentry_lock; /* drop entry handling */ 925 spinlock_t droppacket_lock; /* drop packet handling */ 926 spinlock_t securetcp_lock; /* state and timeout tables */ 927 928 /* sys-ctl struct */ 929 struct ctl_table_header *sysctl_hdr; 930 struct ctl_table *sysctl_tbl; 931 #endif 932 933 /* sysctl variables */ 934 int sysctl_amemthresh; 935 int sysctl_am_droprate; 936 int sysctl_drop_entry; 937 int sysctl_drop_packet; 938 int sysctl_secure_tcp; 939 #ifdef CONFIG_IP_VS_NFCT 940 int sysctl_conntrack; 941 #endif 942 int sysctl_snat_reroute; 943 int sysctl_sync_ver; 944 int sysctl_sync_ports; 945 int sysctl_sync_persist_mode; 946 unsigned long sysctl_sync_qlen_max; 947 int sysctl_sync_sock_size; 948 int sysctl_cache_bypass; 949 int sysctl_expire_nodest_conn; 950 int sysctl_sloppy_tcp; 951 int sysctl_sloppy_sctp; 952 int sysctl_expire_quiescent_template; 953 int sysctl_sync_threshold[2]; 954 unsigned int sysctl_sync_refresh_period; 955 int sysctl_sync_retries; 956 int sysctl_nat_icmp_send; 957 int sysctl_pmtu_disc; 958 int sysctl_backup_only; 959 960 /* ip_vs_lblc */ 961 int sysctl_lblc_expiration; 962 struct ctl_table_header *lblc_ctl_header; 963 struct ctl_table *lblc_ctl_table; 964 /* ip_vs_lblcr */ 965 int sysctl_lblcr_expiration; 966 struct ctl_table_header *lblcr_ctl_header; 967 struct ctl_table *lblcr_ctl_table; 968 /* ip_vs_est */ 969 struct list_head est_list; /* estimator list */ 970 spinlock_t est_lock; 971 struct timer_list est_timer; /* Estimation timer */ 972 /* ip_vs_sync */ 973 spinlock_t sync_lock; 974 struct ipvs_master_sync_state *ms; 975 spinlock_t sync_buff_lock; 976 struct task_struct **backup_threads; 977 int threads_mask; 978 int send_mesg_maxlen; 979 int recv_mesg_maxlen; 980 volatile int sync_state; 981 volatile int master_syncid; 982 volatile int backup_syncid; 983 struct mutex sync_mutex; 984 /* multicast interface name */ 985 char master_mcast_ifn[IP_VS_IFNAME_MAXLEN]; 986 char backup_mcast_ifn[IP_VS_IFNAME_MAXLEN]; 987 /* net name space ptr */ 988 struct net *net; /* Needed by timer routines */ 989 }; 990 991 #define DEFAULT_SYNC_THRESHOLD 3 992 #define DEFAULT_SYNC_PERIOD 50 993 #define DEFAULT_SYNC_VER 1 994 #define DEFAULT_SLOPPY_TCP 0 995 #define DEFAULT_SLOPPY_SCTP 0 996 #define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ) 997 #define DEFAULT_SYNC_RETRIES 0 998 #define IPVS_SYNC_WAKEUP_RATE 8 999 #define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4) 1000 #define IPVS_SYNC_SEND_DELAY (HZ / 50) 1001 #define IPVS_SYNC_CHECK_PERIOD HZ 1002 #define IPVS_SYNC_FLUSH_TIME (HZ * 2) 1003 #define IPVS_SYNC_PORTS_MAX (1 << 6) 1004 1005 #ifdef CONFIG_SYSCTL 1006 1007 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs) 1008 { 1009 return ipvs->sysctl_sync_threshold[0]; 1010 } 1011 1012 static inline int sysctl_sync_period(struct netns_ipvs *ipvs) 1013 { 1014 return ACCESS_ONCE(ipvs->sysctl_sync_threshold[1]); 1015 } 1016 1017 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs) 1018 { 1019 return ACCESS_ONCE(ipvs->sysctl_sync_refresh_period); 1020 } 1021 1022 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs) 1023 { 1024 return ipvs->sysctl_sync_retries; 1025 } 1026 1027 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs) 1028 { 1029 return ipvs->sysctl_sync_ver; 1030 } 1031 1032 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs) 1033 { 1034 return ipvs->sysctl_sloppy_tcp; 1035 } 1036 1037 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs) 1038 { 1039 return ipvs->sysctl_sloppy_sctp; 1040 } 1041 1042 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs) 1043 { 1044 return ACCESS_ONCE(ipvs->sysctl_sync_ports); 1045 } 1046 1047 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs) 1048 { 1049 return ipvs->sysctl_sync_persist_mode; 1050 } 1051 1052 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs) 1053 { 1054 return ipvs->sysctl_sync_qlen_max; 1055 } 1056 1057 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs) 1058 { 1059 return ipvs->sysctl_sync_sock_size; 1060 } 1061 1062 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs) 1063 { 1064 return ipvs->sysctl_pmtu_disc; 1065 } 1066 1067 static inline int sysctl_backup_only(struct netns_ipvs *ipvs) 1068 { 1069 return ipvs->sync_state & IP_VS_STATE_BACKUP && 1070 ipvs->sysctl_backup_only; 1071 } 1072 1073 #else 1074 1075 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs) 1076 { 1077 return DEFAULT_SYNC_THRESHOLD; 1078 } 1079 1080 static inline int sysctl_sync_period(struct netns_ipvs *ipvs) 1081 { 1082 return DEFAULT_SYNC_PERIOD; 1083 } 1084 1085 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs) 1086 { 1087 return DEFAULT_SYNC_REFRESH_PERIOD; 1088 } 1089 1090 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs) 1091 { 1092 return DEFAULT_SYNC_RETRIES & 3; 1093 } 1094 1095 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs) 1096 { 1097 return DEFAULT_SYNC_VER; 1098 } 1099 1100 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs) 1101 { 1102 return DEFAULT_SLOPPY_TCP; 1103 } 1104 1105 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs) 1106 { 1107 return DEFAULT_SLOPPY_SCTP; 1108 } 1109 1110 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs) 1111 { 1112 return 1; 1113 } 1114 1115 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs) 1116 { 1117 return 0; 1118 } 1119 1120 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs) 1121 { 1122 return IPVS_SYNC_QLEN_MAX; 1123 } 1124 1125 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs) 1126 { 1127 return 0; 1128 } 1129 1130 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs) 1131 { 1132 return 1; 1133 } 1134 1135 static inline int sysctl_backup_only(struct netns_ipvs *ipvs) 1136 { 1137 return 0; 1138 } 1139 1140 #endif 1141 1142 /* 1143 * IPVS core functions 1144 * (from ip_vs_core.c) 1145 */ 1146 const char *ip_vs_proto_name(unsigned int proto); 1147 void ip_vs_init_hash_table(struct list_head *table, int rows); 1148 #define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t))) 1149 1150 #define IP_VS_APP_TYPE_FTP 1 1151 1152 /* 1153 * ip_vs_conn handling functions 1154 * (from ip_vs_conn.c) 1155 */ 1156 1157 enum { 1158 IP_VS_DIR_INPUT = 0, 1159 IP_VS_DIR_OUTPUT, 1160 IP_VS_DIR_INPUT_ONLY, 1161 IP_VS_DIR_LAST, 1162 }; 1163 1164 static inline void ip_vs_conn_fill_param(struct net *net, int af, int protocol, 1165 const union nf_inet_addr *caddr, 1166 __be16 cport, 1167 const union nf_inet_addr *vaddr, 1168 __be16 vport, 1169 struct ip_vs_conn_param *p) 1170 { 1171 p->net = net; 1172 p->af = af; 1173 p->protocol = protocol; 1174 p->caddr = caddr; 1175 p->cport = cport; 1176 p->vaddr = vaddr; 1177 p->vport = vport; 1178 p->pe = NULL; 1179 p->pe_data = NULL; 1180 } 1181 1182 struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p); 1183 struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p); 1184 1185 struct ip_vs_conn * ip_vs_conn_in_get_proto(int af, const struct sk_buff *skb, 1186 const struct ip_vs_iphdr *iph, 1187 int inverse); 1188 1189 struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p); 1190 1191 struct ip_vs_conn * ip_vs_conn_out_get_proto(int af, const struct sk_buff *skb, 1192 const struct ip_vs_iphdr *iph, 1193 int inverse); 1194 1195 /* Get reference to gain full access to conn. 1196 * By default, RCU read-side critical sections have access only to 1197 * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference. 1198 */ 1199 static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp) 1200 { 1201 return atomic_inc_not_zero(&cp->refcnt); 1202 } 1203 1204 /* put back the conn without restarting its timer */ 1205 static inline void __ip_vs_conn_put(struct ip_vs_conn *cp) 1206 { 1207 smp_mb__before_atomic_dec(); 1208 atomic_dec(&cp->refcnt); 1209 } 1210 void ip_vs_conn_put(struct ip_vs_conn *cp); 1211 void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport); 1212 1213 struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, 1214 const union nf_inet_addr *daddr, 1215 __be16 dport, unsigned int flags, 1216 struct ip_vs_dest *dest, __u32 fwmark); 1217 void ip_vs_conn_expire_now(struct ip_vs_conn *cp); 1218 1219 const char *ip_vs_state_name(__u16 proto, int state); 1220 1221 void ip_vs_tcp_conn_listen(struct net *net, struct ip_vs_conn *cp); 1222 int ip_vs_check_template(struct ip_vs_conn *ct); 1223 void ip_vs_random_dropentry(struct net *net); 1224 int ip_vs_conn_init(void); 1225 void ip_vs_conn_cleanup(void); 1226 1227 static inline void ip_vs_control_del(struct ip_vs_conn *cp) 1228 { 1229 struct ip_vs_conn *ctl_cp = cp->control; 1230 if (!ctl_cp) { 1231 IP_VS_ERR_BUF("request control DEL for uncontrolled: " 1232 "%s:%d to %s:%d\n", 1233 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1234 ntohs(cp->cport), 1235 IP_VS_DBG_ADDR(cp->af, &cp->vaddr), 1236 ntohs(cp->vport)); 1237 1238 return; 1239 } 1240 1241 IP_VS_DBG_BUF(7, "DELeting control for: " 1242 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n", 1243 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1244 ntohs(cp->cport), 1245 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr), 1246 ntohs(ctl_cp->cport)); 1247 1248 cp->control = NULL; 1249 if (atomic_read(&ctl_cp->n_control) == 0) { 1250 IP_VS_ERR_BUF("BUG control DEL with n=0 : " 1251 "%s:%d to %s:%d\n", 1252 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1253 ntohs(cp->cport), 1254 IP_VS_DBG_ADDR(cp->af, &cp->vaddr), 1255 ntohs(cp->vport)); 1256 1257 return; 1258 } 1259 atomic_dec(&ctl_cp->n_control); 1260 } 1261 1262 static inline void 1263 ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp) 1264 { 1265 if (cp->control) { 1266 IP_VS_ERR_BUF("request control ADD for already controlled: " 1267 "%s:%d to %s:%d\n", 1268 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1269 ntohs(cp->cport), 1270 IP_VS_DBG_ADDR(cp->af, &cp->vaddr), 1271 ntohs(cp->vport)); 1272 1273 ip_vs_control_del(cp); 1274 } 1275 1276 IP_VS_DBG_BUF(7, "ADDing control for: " 1277 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n", 1278 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1279 ntohs(cp->cport), 1280 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr), 1281 ntohs(ctl_cp->cport)); 1282 1283 cp->control = ctl_cp; 1284 atomic_inc(&ctl_cp->n_control); 1285 } 1286 1287 /* 1288 * IPVS netns init & cleanup functions 1289 */ 1290 int ip_vs_estimator_net_init(struct net *net); 1291 int ip_vs_control_net_init(struct net *net); 1292 int ip_vs_protocol_net_init(struct net *net); 1293 int ip_vs_app_net_init(struct net *net); 1294 int ip_vs_conn_net_init(struct net *net); 1295 int ip_vs_sync_net_init(struct net *net); 1296 void ip_vs_conn_net_cleanup(struct net *net); 1297 void ip_vs_app_net_cleanup(struct net *net); 1298 void ip_vs_protocol_net_cleanup(struct net *net); 1299 void ip_vs_control_net_cleanup(struct net *net); 1300 void ip_vs_estimator_net_cleanup(struct net *net); 1301 void ip_vs_sync_net_cleanup(struct net *net); 1302 void ip_vs_service_net_cleanup(struct net *net); 1303 1304 /* 1305 * IPVS application functions 1306 * (from ip_vs_app.c) 1307 */ 1308 #define IP_VS_APP_MAX_PORTS 8 1309 struct ip_vs_app *register_ip_vs_app(struct net *net, struct ip_vs_app *app); 1310 void unregister_ip_vs_app(struct net *net, struct ip_vs_app *app); 1311 int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp); 1312 void ip_vs_unbind_app(struct ip_vs_conn *cp); 1313 int register_ip_vs_app_inc(struct net *net, struct ip_vs_app *app, __u16 proto, 1314 __u16 port); 1315 int ip_vs_app_inc_get(struct ip_vs_app *inc); 1316 void ip_vs_app_inc_put(struct ip_vs_app *inc); 1317 1318 int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb); 1319 int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb); 1320 1321 int register_ip_vs_pe(struct ip_vs_pe *pe); 1322 int unregister_ip_vs_pe(struct ip_vs_pe *pe); 1323 struct ip_vs_pe *ip_vs_pe_getbyname(const char *name); 1324 struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name); 1325 1326 /* 1327 * Use a #define to avoid all of module.h just for these trivial ops 1328 */ 1329 #define ip_vs_pe_get(pe) \ 1330 if (pe && pe->module) \ 1331 __module_get(pe->module); 1332 1333 #define ip_vs_pe_put(pe) \ 1334 if (pe && pe->module) \ 1335 module_put(pe->module); 1336 1337 /* 1338 * IPVS protocol functions (from ip_vs_proto.c) 1339 */ 1340 int ip_vs_protocol_init(void); 1341 void ip_vs_protocol_cleanup(void); 1342 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags); 1343 int *ip_vs_create_timeout_table(int *table, int size); 1344 int ip_vs_set_state_timeout(int *table, int num, const char *const *names, 1345 const char *name, int to); 1346 void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp, 1347 const struct sk_buff *skb, int offset, 1348 const char *msg); 1349 1350 extern struct ip_vs_protocol ip_vs_protocol_tcp; 1351 extern struct ip_vs_protocol ip_vs_protocol_udp; 1352 extern struct ip_vs_protocol ip_vs_protocol_icmp; 1353 extern struct ip_vs_protocol ip_vs_protocol_esp; 1354 extern struct ip_vs_protocol ip_vs_protocol_ah; 1355 extern struct ip_vs_protocol ip_vs_protocol_sctp; 1356 1357 /* 1358 * Registering/unregistering scheduler functions 1359 * (from ip_vs_sched.c) 1360 */ 1361 int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler); 1362 int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler); 1363 int ip_vs_bind_scheduler(struct ip_vs_service *svc, 1364 struct ip_vs_scheduler *scheduler); 1365 void ip_vs_unbind_scheduler(struct ip_vs_service *svc, 1366 struct ip_vs_scheduler *sched); 1367 struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name); 1368 void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler); 1369 struct ip_vs_conn * 1370 ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb, 1371 struct ip_vs_proto_data *pd, int *ignored, 1372 struct ip_vs_iphdr *iph); 1373 int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb, 1374 struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph); 1375 1376 void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg); 1377 1378 1379 /* 1380 * IPVS control data and functions (from ip_vs_ctl.c) 1381 */ 1382 extern struct ip_vs_stats ip_vs_stats; 1383 extern int sysctl_ip_vs_sync_ver; 1384 1385 struct ip_vs_service * 1386 ip_vs_service_find(struct net *net, int af, __u32 fwmark, __u16 protocol, 1387 const union nf_inet_addr *vaddr, __be16 vport); 1388 1389 bool ip_vs_has_real_service(struct net *net, int af, __u16 protocol, 1390 const union nf_inet_addr *daddr, __be16 dport); 1391 1392 int ip_vs_use_count_inc(void); 1393 void ip_vs_use_count_dec(void); 1394 int ip_vs_register_nl_ioctl(void); 1395 void ip_vs_unregister_nl_ioctl(void); 1396 int ip_vs_control_init(void); 1397 void ip_vs_control_cleanup(void); 1398 struct ip_vs_dest * 1399 ip_vs_find_dest(struct net *net, int af, const union nf_inet_addr *daddr, 1400 __be16 dport, const union nf_inet_addr *vaddr, __be16 vport, 1401 __u16 protocol, __u32 fwmark, __u32 flags); 1402 void ip_vs_try_bind_dest(struct ip_vs_conn *cp); 1403 1404 static inline void ip_vs_dest_hold(struct ip_vs_dest *dest) 1405 { 1406 atomic_inc(&dest->refcnt); 1407 } 1408 1409 static inline void ip_vs_dest_put(struct ip_vs_dest *dest) 1410 { 1411 smp_mb__before_atomic_dec(); 1412 atomic_dec(&dest->refcnt); 1413 } 1414 1415 static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest) 1416 { 1417 if (atomic_dec_return(&dest->refcnt) < 0) 1418 kfree(dest); 1419 } 1420 1421 /* 1422 * IPVS sync daemon data and function prototypes 1423 * (from ip_vs_sync.c) 1424 */ 1425 int start_sync_thread(struct net *net, int state, char *mcast_ifn, __u8 syncid); 1426 int stop_sync_thread(struct net *net, int state); 1427 void ip_vs_sync_conn(struct net *net, struct ip_vs_conn *cp, int pkts); 1428 1429 /* 1430 * IPVS rate estimator prototypes (from ip_vs_est.c) 1431 */ 1432 void ip_vs_start_estimator(struct net *net, struct ip_vs_stats *stats); 1433 void ip_vs_stop_estimator(struct net *net, struct ip_vs_stats *stats); 1434 void ip_vs_zero_estimator(struct ip_vs_stats *stats); 1435 void ip_vs_read_estimator(struct ip_vs_stats_user *dst, 1436 struct ip_vs_stats *stats); 1437 1438 /* 1439 * Various IPVS packet transmitters (from ip_vs_xmit.c) 1440 */ 1441 int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1442 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1443 int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1444 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1445 int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1446 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1447 int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1448 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1449 int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1450 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1451 int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1452 struct ip_vs_protocol *pp, int offset, 1453 unsigned int hooknum, struct ip_vs_iphdr *iph); 1454 void ip_vs_dest_dst_rcu_free(struct rcu_head *head); 1455 1456 #ifdef CONFIG_IP_VS_IPV6 1457 int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1458 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1459 int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1460 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1461 int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1462 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1463 int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1464 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1465 int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1466 struct ip_vs_protocol *pp, int offset, 1467 unsigned int hooknum, struct ip_vs_iphdr *iph); 1468 #endif 1469 1470 #ifdef CONFIG_SYSCTL 1471 /* 1472 * This is a simple mechanism to ignore packets when 1473 * we are loaded. Just set ip_vs_drop_rate to 'n' and 1474 * we start to drop 1/rate of the packets 1475 */ 1476 1477 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) 1478 { 1479 if (!ipvs->drop_rate) 1480 return 0; 1481 if (--ipvs->drop_counter > 0) 1482 return 0; 1483 ipvs->drop_counter = ipvs->drop_rate; 1484 return 1; 1485 } 1486 #else 1487 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; } 1488 #endif 1489 1490 /* 1491 * ip_vs_fwd_tag returns the forwarding tag of the connection 1492 */ 1493 #define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK) 1494 1495 static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp) 1496 { 1497 char fwd; 1498 1499 switch (IP_VS_FWD_METHOD(cp)) { 1500 case IP_VS_CONN_F_MASQ: 1501 fwd = 'M'; break; 1502 case IP_VS_CONN_F_LOCALNODE: 1503 fwd = 'L'; break; 1504 case IP_VS_CONN_F_TUNNEL: 1505 fwd = 'T'; break; 1506 case IP_VS_CONN_F_DROUTE: 1507 fwd = 'R'; break; 1508 case IP_VS_CONN_F_BYPASS: 1509 fwd = 'B'; break; 1510 default: 1511 fwd = '?'; break; 1512 } 1513 return fwd; 1514 } 1515 1516 void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp, 1517 struct ip_vs_conn *cp, int dir); 1518 1519 #ifdef CONFIG_IP_VS_IPV6 1520 void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp, 1521 struct ip_vs_conn *cp, int dir); 1522 #endif 1523 1524 __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset); 1525 1526 static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum) 1527 { 1528 __be32 diff[2] = { ~old, new }; 1529 1530 return csum_partial(diff, sizeof(diff), oldsum); 1531 } 1532 1533 #ifdef CONFIG_IP_VS_IPV6 1534 static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new, 1535 __wsum oldsum) 1536 { 1537 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0], 1538 new[3], new[2], new[1], new[0] }; 1539 1540 return csum_partial(diff, sizeof(diff), oldsum); 1541 } 1542 #endif 1543 1544 static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum) 1545 { 1546 __be16 diff[2] = { ~old, new }; 1547 1548 return csum_partial(diff, sizeof(diff), oldsum); 1549 } 1550 1551 /* 1552 * Forget current conntrack (unconfirmed) and attach notrack entry 1553 */ 1554 static inline void ip_vs_notrack(struct sk_buff *skb) 1555 { 1556 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 1557 enum ip_conntrack_info ctinfo; 1558 struct nf_conn *ct = nf_ct_get(skb, &ctinfo); 1559 1560 if (!ct || !nf_ct_is_untracked(ct)) { 1561 nf_conntrack_put(skb->nfct); 1562 skb->nfct = &nf_ct_untracked_get()->ct_general; 1563 skb->nfctinfo = IP_CT_NEW; 1564 nf_conntrack_get(skb->nfct); 1565 } 1566 #endif 1567 } 1568 1569 #ifdef CONFIG_IP_VS_NFCT 1570 /* 1571 * Netfilter connection tracking 1572 * (from ip_vs_nfct.c) 1573 */ 1574 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs) 1575 { 1576 #ifdef CONFIG_SYSCTL 1577 return ipvs->sysctl_conntrack; 1578 #else 1579 return 0; 1580 #endif 1581 } 1582 1583 void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp, 1584 int outin); 1585 int ip_vs_confirm_conntrack(struct sk_buff *skb); 1586 void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct, 1587 struct ip_vs_conn *cp, u_int8_t proto, 1588 const __be16 port, int from_rs); 1589 void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp); 1590 1591 #else 1592 1593 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs) 1594 { 1595 return 0; 1596 } 1597 1598 static inline void ip_vs_update_conntrack(struct sk_buff *skb, 1599 struct ip_vs_conn *cp, int outin) 1600 { 1601 } 1602 1603 static inline int ip_vs_confirm_conntrack(struct sk_buff *skb) 1604 { 1605 return NF_ACCEPT; 1606 } 1607 1608 static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp) 1609 { 1610 } 1611 /* CONFIG_IP_VS_NFCT */ 1612 #endif 1613 1614 static inline int 1615 ip_vs_dest_conn_overhead(struct ip_vs_dest *dest) 1616 { 1617 /* 1618 * We think the overhead of processing active connections is 256 1619 * times higher than that of inactive connections in average. (This 1620 * 256 times might not be accurate, we will change it later) We 1621 * use the following formula to estimate the overhead now: 1622 * dest->activeconns*256 + dest->inactconns 1623 */ 1624 return (atomic_read(&dest->activeconns) << 8) + 1625 atomic_read(&dest->inactconns); 1626 } 1627 1628 #endif /* _NET_IP_VS_H */ 1629