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 __u32 thoff_reasm; /* Transport Header Offset in nfct_reasm skb */ 113 __u16 fragoffs; /* IPv6 fragment offset, 0 if first frag (or not frag)*/ 114 __s16 protocol; 115 __s32 flags; 116 union nf_inet_addr saddr; 117 union nf_inet_addr daddr; 118 }; 119 120 /* Dependency to module: nf_defrag_ipv6 */ 121 #if defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE) 122 static inline struct sk_buff *skb_nfct_reasm(const struct sk_buff *skb) 123 { 124 return skb->nfct_reasm; 125 } 126 static inline void *frag_safe_skb_hp(const struct sk_buff *skb, int offset, 127 int len, void *buffer, 128 const struct ip_vs_iphdr *ipvsh) 129 { 130 if (unlikely(ipvsh->fragoffs && skb_nfct_reasm(skb))) 131 return skb_header_pointer(skb_nfct_reasm(skb), 132 ipvsh->thoff_reasm, len, buffer); 133 134 return skb_header_pointer(skb, offset, len, buffer); 135 } 136 #else 137 static inline struct sk_buff *skb_nfct_reasm(const struct sk_buff *skb) 138 { 139 return NULL; 140 } 141 static inline void *frag_safe_skb_hp(const struct sk_buff *skb, int offset, 142 int len, void *buffer, 143 const struct ip_vs_iphdr *ipvsh) 144 { 145 return skb_header_pointer(skb, offset, len, buffer); 146 } 147 #endif 148 149 static inline void 150 ip_vs_fill_ip4hdr(const void *nh, struct ip_vs_iphdr *iphdr) 151 { 152 const struct iphdr *iph = nh; 153 154 iphdr->len = iph->ihl * 4; 155 iphdr->fragoffs = 0; 156 iphdr->protocol = iph->protocol; 157 iphdr->saddr.ip = iph->saddr; 158 iphdr->daddr.ip = iph->daddr; 159 } 160 161 /* This function handles filling *ip_vs_iphdr, both for IPv4 and IPv6. 162 * IPv6 requires some extra work, as finding proper header position, 163 * depend on the IPv6 extension headers. 164 */ 165 static inline void 166 ip_vs_fill_iph_skb(int af, const struct sk_buff *skb, struct ip_vs_iphdr *iphdr) 167 { 168 #ifdef CONFIG_IP_VS_IPV6 169 if (af == AF_INET6) { 170 const struct ipv6hdr *iph = 171 (struct ipv6hdr *)skb_network_header(skb); 172 iphdr->saddr.in6 = iph->saddr; 173 iphdr->daddr.in6 = iph->daddr; 174 /* ipv6_find_hdr() updates len, flags, thoff_reasm */ 175 iphdr->thoff_reasm = 0; 176 iphdr->len = 0; 177 iphdr->flags = 0; 178 iphdr->protocol = ipv6_find_hdr(skb, &iphdr->len, -1, 179 &iphdr->fragoffs, 180 &iphdr->flags); 181 /* get proto from re-assembled packet and it's offset */ 182 if (skb_nfct_reasm(skb)) 183 iphdr->protocol = ipv6_find_hdr(skb_nfct_reasm(skb), 184 &iphdr->thoff_reasm, 185 -1, NULL, NULL); 186 187 } else 188 #endif 189 { 190 const struct iphdr *iph = 191 (struct iphdr *)skb_network_header(skb); 192 iphdr->len = iph->ihl * 4; 193 iphdr->fragoffs = 0; 194 iphdr->protocol = iph->protocol; 195 iphdr->saddr.ip = iph->saddr; 196 iphdr->daddr.ip = iph->daddr; 197 } 198 } 199 200 static inline void ip_vs_addr_copy(int af, union nf_inet_addr *dst, 201 const union nf_inet_addr *src) 202 { 203 #ifdef CONFIG_IP_VS_IPV6 204 if (af == AF_INET6) 205 dst->in6 = src->in6; 206 else 207 #endif 208 dst->ip = src->ip; 209 } 210 211 static inline void ip_vs_addr_set(int af, union nf_inet_addr *dst, 212 const union nf_inet_addr *src) 213 { 214 #ifdef CONFIG_IP_VS_IPV6 215 if (af == AF_INET6) { 216 dst->in6 = src->in6; 217 return; 218 } 219 #endif 220 dst->ip = src->ip; 221 dst->all[1] = 0; 222 dst->all[2] = 0; 223 dst->all[3] = 0; 224 } 225 226 static inline int ip_vs_addr_equal(int af, const union nf_inet_addr *a, 227 const union nf_inet_addr *b) 228 { 229 #ifdef CONFIG_IP_VS_IPV6 230 if (af == AF_INET6) 231 return ipv6_addr_equal(&a->in6, &b->in6); 232 #endif 233 return a->ip == b->ip; 234 } 235 236 #ifdef CONFIG_IP_VS_DEBUG 237 #include <linux/net.h> 238 239 extern int ip_vs_get_debug_level(void); 240 241 static inline const char *ip_vs_dbg_addr(int af, char *buf, size_t buf_len, 242 const union nf_inet_addr *addr, 243 int *idx) 244 { 245 int len; 246 #ifdef CONFIG_IP_VS_IPV6 247 if (af == AF_INET6) 248 len = snprintf(&buf[*idx], buf_len - *idx, "[%pI6c]", 249 &addr->in6) + 1; 250 else 251 #endif 252 len = snprintf(&buf[*idx], buf_len - *idx, "%pI4", 253 &addr->ip) + 1; 254 255 *idx += len; 256 BUG_ON(*idx > buf_len + 1); 257 return &buf[*idx - len]; 258 } 259 260 #define IP_VS_DBG_BUF(level, msg, ...) \ 261 do { \ 262 char ip_vs_dbg_buf[160]; \ 263 int ip_vs_dbg_idx = 0; \ 264 if (level <= ip_vs_get_debug_level()) \ 265 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \ 266 } while (0) 267 #define IP_VS_ERR_BUF(msg...) \ 268 do { \ 269 char ip_vs_dbg_buf[160]; \ 270 int ip_vs_dbg_idx = 0; \ 271 pr_err(msg); \ 272 } while (0) 273 274 /* Only use from within IP_VS_DBG_BUF() or IP_VS_ERR_BUF macros */ 275 #define IP_VS_DBG_ADDR(af, addr) \ 276 ip_vs_dbg_addr(af, ip_vs_dbg_buf, \ 277 sizeof(ip_vs_dbg_buf), addr, \ 278 &ip_vs_dbg_idx) 279 280 #define IP_VS_DBG(level, msg, ...) \ 281 do { \ 282 if (level <= ip_vs_get_debug_level()) \ 283 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \ 284 } while (0) 285 #define IP_VS_DBG_RL(msg, ...) \ 286 do { \ 287 if (net_ratelimit()) \ 288 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \ 289 } while (0) 290 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) \ 291 do { \ 292 if (level <= ip_vs_get_debug_level()) \ 293 pp->debug_packet(af, pp, skb, ofs, msg); \ 294 } while (0) 295 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) \ 296 do { \ 297 if (level <= ip_vs_get_debug_level() && \ 298 net_ratelimit()) \ 299 pp->debug_packet(af, pp, skb, ofs, msg); \ 300 } while (0) 301 #else /* NO DEBUGGING at ALL */ 302 #define IP_VS_DBG_BUF(level, msg...) do {} while (0) 303 #define IP_VS_ERR_BUF(msg...) do {} while (0) 304 #define IP_VS_DBG(level, msg...) do {} while (0) 305 #define IP_VS_DBG_RL(msg...) do {} while (0) 306 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) do {} while (0) 307 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) do {} while (0) 308 #endif 309 310 #define IP_VS_BUG() BUG() 311 #define IP_VS_ERR_RL(msg, ...) \ 312 do { \ 313 if (net_ratelimit()) \ 314 pr_err(msg, ##__VA_ARGS__); \ 315 } while (0) 316 317 #ifdef CONFIG_IP_VS_DEBUG 318 #define EnterFunction(level) \ 319 do { \ 320 if (level <= ip_vs_get_debug_level()) \ 321 printk(KERN_DEBUG \ 322 pr_fmt("Enter: %s, %s line %i\n"), \ 323 __func__, __FILE__, __LINE__); \ 324 } while (0) 325 #define LeaveFunction(level) \ 326 do { \ 327 if (level <= ip_vs_get_debug_level()) \ 328 printk(KERN_DEBUG \ 329 pr_fmt("Leave: %s, %s line %i\n"), \ 330 __func__, __FILE__, __LINE__); \ 331 } while (0) 332 #else 333 #define EnterFunction(level) do {} while (0) 334 #define LeaveFunction(level) do {} while (0) 335 #endif 336 337 338 /* 339 * The port number of FTP service (in network order). 340 */ 341 #define FTPPORT cpu_to_be16(21) 342 #define FTPDATA cpu_to_be16(20) 343 344 /* 345 * TCP State Values 346 */ 347 enum { 348 IP_VS_TCP_S_NONE = 0, 349 IP_VS_TCP_S_ESTABLISHED, 350 IP_VS_TCP_S_SYN_SENT, 351 IP_VS_TCP_S_SYN_RECV, 352 IP_VS_TCP_S_FIN_WAIT, 353 IP_VS_TCP_S_TIME_WAIT, 354 IP_VS_TCP_S_CLOSE, 355 IP_VS_TCP_S_CLOSE_WAIT, 356 IP_VS_TCP_S_LAST_ACK, 357 IP_VS_TCP_S_LISTEN, 358 IP_VS_TCP_S_SYNACK, 359 IP_VS_TCP_S_LAST 360 }; 361 362 /* 363 * UDP State Values 364 */ 365 enum { 366 IP_VS_UDP_S_NORMAL, 367 IP_VS_UDP_S_LAST, 368 }; 369 370 /* 371 * ICMP State Values 372 */ 373 enum { 374 IP_VS_ICMP_S_NORMAL, 375 IP_VS_ICMP_S_LAST, 376 }; 377 378 /* 379 * SCTP State Values 380 */ 381 enum ip_vs_sctp_states { 382 IP_VS_SCTP_S_NONE, 383 IP_VS_SCTP_S_INIT1, 384 IP_VS_SCTP_S_INIT, 385 IP_VS_SCTP_S_COOKIE_SENT, 386 IP_VS_SCTP_S_COOKIE_REPLIED, 387 IP_VS_SCTP_S_COOKIE_WAIT, 388 IP_VS_SCTP_S_COOKIE, 389 IP_VS_SCTP_S_COOKIE_ECHOED, 390 IP_VS_SCTP_S_ESTABLISHED, 391 IP_VS_SCTP_S_SHUTDOWN_SENT, 392 IP_VS_SCTP_S_SHUTDOWN_RECEIVED, 393 IP_VS_SCTP_S_SHUTDOWN_ACK_SENT, 394 IP_VS_SCTP_S_REJECTED, 395 IP_VS_SCTP_S_CLOSED, 396 IP_VS_SCTP_S_LAST 397 }; 398 399 /* 400 * Delta sequence info structure 401 * Each ip_vs_conn has 2 (output AND input seq. changes). 402 * Only used in the VS/NAT. 403 */ 404 struct ip_vs_seq { 405 __u32 init_seq; /* Add delta from this seq */ 406 __u32 delta; /* Delta in sequence numbers */ 407 __u32 previous_delta; /* Delta in sequence numbers 408 before last resized pkt */ 409 }; 410 411 /* 412 * counters per cpu 413 */ 414 struct ip_vs_counters { 415 __u32 conns; /* connections scheduled */ 416 __u32 inpkts; /* incoming packets */ 417 __u32 outpkts; /* outgoing packets */ 418 __u64 inbytes; /* incoming bytes */ 419 __u64 outbytes; /* outgoing bytes */ 420 }; 421 /* 422 * Stats per cpu 423 */ 424 struct ip_vs_cpu_stats { 425 struct ip_vs_counters ustats; 426 struct u64_stats_sync syncp; 427 }; 428 429 /* 430 * IPVS statistics objects 431 */ 432 struct ip_vs_estimator { 433 struct list_head list; 434 435 u64 last_inbytes; 436 u64 last_outbytes; 437 u32 last_conns; 438 u32 last_inpkts; 439 u32 last_outpkts; 440 441 u32 cps; 442 u32 inpps; 443 u32 outpps; 444 u32 inbps; 445 u32 outbps; 446 }; 447 448 struct ip_vs_stats { 449 struct ip_vs_stats_user ustats; /* statistics */ 450 struct ip_vs_estimator est; /* estimator */ 451 struct ip_vs_cpu_stats __percpu *cpustats; /* per cpu counters */ 452 spinlock_t lock; /* spin lock */ 453 struct ip_vs_stats_user ustats0; /* reset values */ 454 }; 455 456 struct dst_entry; 457 struct iphdr; 458 struct ip_vs_conn; 459 struct ip_vs_app; 460 struct sk_buff; 461 struct ip_vs_proto_data; 462 463 struct ip_vs_protocol { 464 struct ip_vs_protocol *next; 465 char *name; 466 u16 protocol; 467 u16 num_states; 468 int dont_defrag; 469 470 void (*init)(struct ip_vs_protocol *pp); 471 472 void (*exit)(struct ip_vs_protocol *pp); 473 474 int (*init_netns)(struct net *net, struct ip_vs_proto_data *pd); 475 476 void (*exit_netns)(struct net *net, struct ip_vs_proto_data *pd); 477 478 int (*conn_schedule)(int af, struct sk_buff *skb, 479 struct ip_vs_proto_data *pd, 480 int *verdict, struct ip_vs_conn **cpp, 481 struct ip_vs_iphdr *iph); 482 483 struct ip_vs_conn * 484 (*conn_in_get)(int af, 485 const struct sk_buff *skb, 486 const struct ip_vs_iphdr *iph, 487 int inverse); 488 489 struct ip_vs_conn * 490 (*conn_out_get)(int af, 491 const struct sk_buff *skb, 492 const struct ip_vs_iphdr *iph, 493 int inverse); 494 495 int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp, 496 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph); 497 498 int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp, 499 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph); 500 501 int (*csum_check)(int af, struct sk_buff *skb, 502 struct ip_vs_protocol *pp); 503 504 const char *(*state_name)(int state); 505 506 void (*state_transition)(struct ip_vs_conn *cp, int direction, 507 const struct sk_buff *skb, 508 struct ip_vs_proto_data *pd); 509 510 int (*register_app)(struct net *net, struct ip_vs_app *inc); 511 512 void (*unregister_app)(struct net *net, struct ip_vs_app *inc); 513 514 int (*app_conn_bind)(struct ip_vs_conn *cp); 515 516 void (*debug_packet)(int af, struct ip_vs_protocol *pp, 517 const struct sk_buff *skb, 518 int offset, 519 const char *msg); 520 521 void (*timeout_change)(struct ip_vs_proto_data *pd, int flags); 522 }; 523 524 /* 525 * protocol data per netns 526 */ 527 struct ip_vs_proto_data { 528 struct ip_vs_proto_data *next; 529 struct ip_vs_protocol *pp; 530 int *timeout_table; /* protocol timeout table */ 531 atomic_t appcnt; /* counter of proto app incs. */ 532 struct tcp_states_t *tcp_state_table; 533 }; 534 535 extern struct ip_vs_protocol *ip_vs_proto_get(unsigned short proto); 536 extern struct ip_vs_proto_data *ip_vs_proto_data_get(struct net *net, 537 unsigned short proto); 538 539 struct ip_vs_conn_param { 540 struct net *net; 541 const union nf_inet_addr *caddr; 542 const union nf_inet_addr *vaddr; 543 __be16 cport; 544 __be16 vport; 545 __u16 protocol; 546 u16 af; 547 548 const struct ip_vs_pe *pe; 549 char *pe_data; 550 __u8 pe_data_len; 551 }; 552 553 /* 554 * IP_VS structure allocated for each dynamically scheduled connection 555 */ 556 struct ip_vs_conn { 557 struct hlist_node c_list; /* hashed list heads */ 558 /* Protocol, addresses and port numbers */ 559 __be16 cport; 560 __be16 dport; 561 __be16 vport; 562 u16 af; /* address family */ 563 union nf_inet_addr caddr; /* client address */ 564 union nf_inet_addr vaddr; /* virtual address */ 565 union nf_inet_addr daddr; /* destination address */ 566 volatile __u32 flags; /* status flags */ 567 __u16 protocol; /* Which protocol (TCP/UDP) */ 568 #ifdef CONFIG_NET_NS 569 struct net *net; /* Name space */ 570 #endif 571 572 /* counter and timer */ 573 atomic_t refcnt; /* reference count */ 574 struct timer_list timer; /* Expiration timer */ 575 volatile unsigned long timeout; /* timeout */ 576 577 /* Flags and state transition */ 578 spinlock_t lock; /* lock for state transition */ 579 volatile __u16 state; /* state info */ 580 volatile __u16 old_state; /* old state, to be used for 581 * state transition triggerd 582 * synchronization 583 */ 584 __u32 fwmark; /* Fire wall mark from skb */ 585 unsigned long sync_endtime; /* jiffies + sent_retries */ 586 587 /* Control members */ 588 struct ip_vs_conn *control; /* Master control connection */ 589 atomic_t n_control; /* Number of controlled ones */ 590 struct ip_vs_dest *dest; /* real server */ 591 atomic_t in_pkts; /* incoming packet counter */ 592 593 /* packet transmitter for different forwarding methods. If it 594 mangles the packet, it must return NF_DROP or better NF_STOLEN, 595 otherwise this must be changed to a sk_buff **. 596 NF_ACCEPT can be returned when destination is local. 597 */ 598 int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp, 599 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 600 601 /* Note: we can group the following members into a structure, 602 in order to save more space, and the following members are 603 only used in VS/NAT anyway */ 604 struct ip_vs_app *app; /* bound ip_vs_app object */ 605 void *app_data; /* Application private data */ 606 struct ip_vs_seq in_seq; /* incoming seq. struct */ 607 struct ip_vs_seq out_seq; /* outgoing seq. struct */ 608 609 const struct ip_vs_pe *pe; 610 char *pe_data; 611 __u8 pe_data_len; 612 613 struct rcu_head rcu_head; 614 }; 615 616 /* 617 * To save some memory in conn table when name space is disabled. 618 */ 619 static inline struct net *ip_vs_conn_net(const struct ip_vs_conn *cp) 620 { 621 #ifdef CONFIG_NET_NS 622 return cp->net; 623 #else 624 return &init_net; 625 #endif 626 } 627 static inline void ip_vs_conn_net_set(struct ip_vs_conn *cp, struct net *net) 628 { 629 #ifdef CONFIG_NET_NS 630 cp->net = net; 631 #endif 632 } 633 634 static inline int ip_vs_conn_net_eq(const struct ip_vs_conn *cp, 635 struct net *net) 636 { 637 #ifdef CONFIG_NET_NS 638 return cp->net == net; 639 #else 640 return 1; 641 #endif 642 } 643 644 /* 645 * Extended internal versions of struct ip_vs_service_user and 646 * ip_vs_dest_user for IPv6 support. 647 * 648 * We need these to conveniently pass around service and destination 649 * options, but unfortunately, we also need to keep the old definitions to 650 * maintain userspace backwards compatibility for the setsockopt interface. 651 */ 652 struct ip_vs_service_user_kern { 653 /* virtual service addresses */ 654 u16 af; 655 u16 protocol; 656 union nf_inet_addr addr; /* virtual ip address */ 657 __be16 port; 658 u32 fwmark; /* firwall mark of service */ 659 660 /* virtual service options */ 661 char *sched_name; 662 char *pe_name; 663 unsigned int flags; /* virtual service flags */ 664 unsigned int timeout; /* persistent timeout in sec */ 665 __be32 netmask; /* persistent netmask or plen */ 666 }; 667 668 669 struct ip_vs_dest_user_kern { 670 /* destination server address */ 671 union nf_inet_addr addr; 672 __be16 port; 673 674 /* real server options */ 675 unsigned int conn_flags; /* connection flags */ 676 int weight; /* destination weight */ 677 678 /* thresholds for active connections */ 679 u32 u_threshold; /* upper threshold */ 680 u32 l_threshold; /* lower threshold */ 681 }; 682 683 684 /* 685 * The information about the virtual service offered to the net 686 * and the forwarding entries 687 */ 688 struct ip_vs_service { 689 struct hlist_node s_list; /* for normal service table */ 690 struct hlist_node f_list; /* for fwmark-based service table */ 691 atomic_t refcnt; /* reference counter */ 692 693 u16 af; /* address family */ 694 __u16 protocol; /* which protocol (TCP/UDP) */ 695 union nf_inet_addr addr; /* IP address for virtual service */ 696 __be16 port; /* port number for the service */ 697 __u32 fwmark; /* firewall mark of the service */ 698 unsigned int flags; /* service status flags */ 699 unsigned int timeout; /* persistent timeout in ticks */ 700 __be32 netmask; /* grouping granularity, mask/plen */ 701 struct net *net; 702 703 struct list_head destinations; /* real server d-linked list */ 704 __u32 num_dests; /* number of servers */ 705 struct ip_vs_stats stats; /* statistics for the service */ 706 707 /* for scheduling */ 708 struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */ 709 spinlock_t sched_lock; /* lock sched_data */ 710 void *sched_data; /* scheduler application data */ 711 712 /* alternate persistence engine */ 713 struct ip_vs_pe __rcu *pe; 714 715 struct rcu_head rcu_head; 716 }; 717 718 /* Information for cached dst */ 719 struct ip_vs_dest_dst { 720 struct dst_entry *dst_cache; /* destination cache entry */ 721 u32 dst_cookie; 722 union nf_inet_addr dst_saddr; 723 struct rcu_head rcu_head; 724 }; 725 726 /* In grace period after removing */ 727 #define IP_VS_DEST_STATE_REMOVING 0x01 728 /* 729 * The real server destination forwarding entry 730 * with ip address, port number, and so on. 731 */ 732 struct ip_vs_dest { 733 struct list_head n_list; /* for the dests in the service */ 734 struct hlist_node d_list; /* for table with all the dests */ 735 736 u16 af; /* address family */ 737 __be16 port; /* port number of the server */ 738 union nf_inet_addr addr; /* IP address of the server */ 739 volatile unsigned int flags; /* dest status flags */ 740 atomic_t conn_flags; /* flags to copy to conn */ 741 atomic_t weight; /* server weight */ 742 743 atomic_t refcnt; /* reference counter */ 744 struct ip_vs_stats stats; /* statistics */ 745 unsigned long state; /* state flags */ 746 747 /* connection counters and thresholds */ 748 atomic_t activeconns; /* active connections */ 749 atomic_t inactconns; /* inactive connections */ 750 atomic_t persistconns; /* persistent connections */ 751 __u32 u_threshold; /* upper threshold */ 752 __u32 l_threshold; /* lower threshold */ 753 754 /* for destination cache */ 755 spinlock_t dst_lock; /* lock of dst_cache */ 756 struct ip_vs_dest_dst __rcu *dest_dst; /* cached dst info */ 757 758 /* for virtual service */ 759 struct ip_vs_service *svc; /* service it belongs to */ 760 __u16 protocol; /* which protocol (TCP/UDP) */ 761 __be16 vport; /* virtual port number */ 762 union nf_inet_addr vaddr; /* virtual IP address */ 763 __u32 vfwmark; /* firewall mark of service */ 764 765 struct list_head t_list; /* in dest_trash */ 766 struct rcu_head rcu_head; 767 unsigned int in_rs_table:1; /* we are in rs_table */ 768 }; 769 770 771 /* 772 * The scheduler object 773 */ 774 struct ip_vs_scheduler { 775 struct list_head n_list; /* d-linked list head */ 776 char *name; /* scheduler name */ 777 atomic_t refcnt; /* reference counter */ 778 struct module *module; /* THIS_MODULE/NULL */ 779 780 /* scheduler initializing service */ 781 int (*init_service)(struct ip_vs_service *svc); 782 /* scheduling service finish */ 783 void (*done_service)(struct ip_vs_service *svc); 784 /* dest is linked */ 785 int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest); 786 /* dest is unlinked */ 787 int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest); 788 /* dest is updated */ 789 int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest); 790 791 /* selecting a server from the given service */ 792 struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc, 793 const struct sk_buff *skb, 794 struct ip_vs_iphdr *iph); 795 }; 796 797 /* The persistence engine object */ 798 struct ip_vs_pe { 799 struct list_head n_list; /* d-linked list head */ 800 char *name; /* scheduler name */ 801 atomic_t refcnt; /* reference counter */ 802 struct module *module; /* THIS_MODULE/NULL */ 803 804 /* get the connection template, if any */ 805 int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb); 806 bool (*ct_match)(const struct ip_vs_conn_param *p, 807 struct ip_vs_conn *ct); 808 u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval, 809 bool inverse); 810 int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf); 811 }; 812 813 /* 814 * The application module object (a.k.a. app incarnation) 815 */ 816 struct ip_vs_app { 817 struct list_head a_list; /* member in app list */ 818 int type; /* IP_VS_APP_TYPE_xxx */ 819 char *name; /* application module name */ 820 __u16 protocol; 821 struct module *module; /* THIS_MODULE/NULL */ 822 struct list_head incs_list; /* list of incarnations */ 823 824 /* members for application incarnations */ 825 struct list_head p_list; /* member in proto app list */ 826 struct ip_vs_app *app; /* its real application */ 827 __be16 port; /* port number in net order */ 828 atomic_t usecnt; /* usage counter */ 829 struct rcu_head rcu_head; 830 831 /* 832 * output hook: Process packet in inout direction, diff set for TCP. 833 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok, 834 * 2=Mangled but checksum was not updated 835 */ 836 int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *, 837 struct sk_buff *, int *diff); 838 839 /* 840 * input hook: Process packet in outin direction, diff set for TCP. 841 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok, 842 * 2=Mangled but checksum was not updated 843 */ 844 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *, 845 struct sk_buff *, int *diff); 846 847 /* ip_vs_app initializer */ 848 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *); 849 850 /* ip_vs_app finish */ 851 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *); 852 853 854 /* not used now */ 855 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *, 856 struct ip_vs_protocol *); 857 858 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *); 859 860 int * timeout_table; 861 int * timeouts; 862 int timeouts_size; 863 864 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app, 865 int *verdict, struct ip_vs_conn **cpp); 866 867 struct ip_vs_conn * 868 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app, 869 const struct iphdr *iph, int inverse); 870 871 struct ip_vs_conn * 872 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app, 873 const struct iphdr *iph, int inverse); 874 875 int (*state_transition)(struct ip_vs_conn *cp, int direction, 876 const struct sk_buff *skb, 877 struct ip_vs_app *app); 878 879 void (*timeout_change)(struct ip_vs_app *app, int flags); 880 }; 881 882 struct ipvs_master_sync_state { 883 struct list_head sync_queue; 884 struct ip_vs_sync_buff *sync_buff; 885 unsigned long sync_queue_len; 886 unsigned int sync_queue_delay; 887 struct task_struct *master_thread; 888 struct delayed_work master_wakeup_work; 889 struct netns_ipvs *ipvs; 890 }; 891 892 /* How much time to keep dests in trash */ 893 #define IP_VS_DEST_TRASH_PERIOD (120 * HZ) 894 895 /* IPVS in network namespace */ 896 struct netns_ipvs { 897 int gen; /* Generation */ 898 int enable; /* enable like nf_hooks do */ 899 /* 900 * Hash table: for real service lookups 901 */ 902 #define IP_VS_RTAB_BITS 4 903 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS) 904 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1) 905 906 struct hlist_head rs_table[IP_VS_RTAB_SIZE]; 907 /* ip_vs_app */ 908 struct list_head app_list; 909 /* ip_vs_proto */ 910 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */ 911 struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE]; 912 /* ip_vs_proto_tcp */ 913 #ifdef CONFIG_IP_VS_PROTO_TCP 914 #define TCP_APP_TAB_BITS 4 915 #define TCP_APP_TAB_SIZE (1 << TCP_APP_TAB_BITS) 916 #define TCP_APP_TAB_MASK (TCP_APP_TAB_SIZE - 1) 917 struct list_head tcp_apps[TCP_APP_TAB_SIZE]; 918 #endif 919 /* ip_vs_proto_udp */ 920 #ifdef CONFIG_IP_VS_PROTO_UDP 921 #define UDP_APP_TAB_BITS 4 922 #define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS) 923 #define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1) 924 struct list_head udp_apps[UDP_APP_TAB_SIZE]; 925 #endif 926 /* ip_vs_proto_sctp */ 927 #ifdef CONFIG_IP_VS_PROTO_SCTP 928 #define SCTP_APP_TAB_BITS 4 929 #define SCTP_APP_TAB_SIZE (1 << SCTP_APP_TAB_BITS) 930 #define SCTP_APP_TAB_MASK (SCTP_APP_TAB_SIZE - 1) 931 /* Hash table for SCTP application incarnations */ 932 struct list_head sctp_apps[SCTP_APP_TAB_SIZE]; 933 #endif 934 /* ip_vs_conn */ 935 atomic_t conn_count; /* connection counter */ 936 937 /* ip_vs_ctl */ 938 struct ip_vs_stats tot_stats; /* Statistics & est. */ 939 940 int num_services; /* no of virtual services */ 941 942 /* Trash for destinations */ 943 struct list_head dest_trash; 944 spinlock_t dest_trash_lock; 945 struct timer_list dest_trash_timer; /* expiration timer */ 946 /* Service counters */ 947 atomic_t ftpsvc_counter; 948 atomic_t nullsvc_counter; 949 950 #ifdef CONFIG_SYSCTL 951 /* 1/rate drop and drop-entry variables */ 952 struct delayed_work defense_work; /* Work handler */ 953 int drop_rate; 954 int drop_counter; 955 atomic_t dropentry; 956 /* locks in ctl.c */ 957 spinlock_t dropentry_lock; /* drop entry handling */ 958 spinlock_t droppacket_lock; /* drop packet handling */ 959 spinlock_t securetcp_lock; /* state and timeout tables */ 960 961 /* sys-ctl struct */ 962 struct ctl_table_header *sysctl_hdr; 963 struct ctl_table *sysctl_tbl; 964 #endif 965 966 /* sysctl variables */ 967 int sysctl_amemthresh; 968 int sysctl_am_droprate; 969 int sysctl_drop_entry; 970 int sysctl_drop_packet; 971 int sysctl_secure_tcp; 972 #ifdef CONFIG_IP_VS_NFCT 973 int sysctl_conntrack; 974 #endif 975 int sysctl_snat_reroute; 976 int sysctl_sync_ver; 977 int sysctl_sync_ports; 978 int sysctl_sync_persist_mode; 979 unsigned long sysctl_sync_qlen_max; 980 int sysctl_sync_sock_size; 981 int sysctl_cache_bypass; 982 int sysctl_expire_nodest_conn; 983 int sysctl_sloppy_tcp; 984 int sysctl_sloppy_sctp; 985 int sysctl_expire_quiescent_template; 986 int sysctl_sync_threshold[2]; 987 unsigned int sysctl_sync_refresh_period; 988 int sysctl_sync_retries; 989 int sysctl_nat_icmp_send; 990 int sysctl_pmtu_disc; 991 int sysctl_backup_only; 992 993 /* ip_vs_lblc */ 994 int sysctl_lblc_expiration; 995 struct ctl_table_header *lblc_ctl_header; 996 struct ctl_table *lblc_ctl_table; 997 /* ip_vs_lblcr */ 998 int sysctl_lblcr_expiration; 999 struct ctl_table_header *lblcr_ctl_header; 1000 struct ctl_table *lblcr_ctl_table; 1001 /* ip_vs_est */ 1002 struct list_head est_list; /* estimator list */ 1003 spinlock_t est_lock; 1004 struct timer_list est_timer; /* Estimation timer */ 1005 /* ip_vs_sync */ 1006 spinlock_t sync_lock; 1007 struct ipvs_master_sync_state *ms; 1008 spinlock_t sync_buff_lock; 1009 struct task_struct **backup_threads; 1010 int threads_mask; 1011 int send_mesg_maxlen; 1012 int recv_mesg_maxlen; 1013 volatile int sync_state; 1014 volatile int master_syncid; 1015 volatile int backup_syncid; 1016 struct mutex sync_mutex; 1017 /* multicast interface name */ 1018 char master_mcast_ifn[IP_VS_IFNAME_MAXLEN]; 1019 char backup_mcast_ifn[IP_VS_IFNAME_MAXLEN]; 1020 /* net name space ptr */ 1021 struct net *net; /* Needed by timer routines */ 1022 }; 1023 1024 #define DEFAULT_SYNC_THRESHOLD 3 1025 #define DEFAULT_SYNC_PERIOD 50 1026 #define DEFAULT_SYNC_VER 1 1027 #define DEFAULT_SLOPPY_TCP 0 1028 #define DEFAULT_SLOPPY_SCTP 0 1029 #define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ) 1030 #define DEFAULT_SYNC_RETRIES 0 1031 #define IPVS_SYNC_WAKEUP_RATE 8 1032 #define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4) 1033 #define IPVS_SYNC_SEND_DELAY (HZ / 50) 1034 #define IPVS_SYNC_CHECK_PERIOD HZ 1035 #define IPVS_SYNC_FLUSH_TIME (HZ * 2) 1036 #define IPVS_SYNC_PORTS_MAX (1 << 6) 1037 1038 #ifdef CONFIG_SYSCTL 1039 1040 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs) 1041 { 1042 return ipvs->sysctl_sync_threshold[0]; 1043 } 1044 1045 static inline int sysctl_sync_period(struct netns_ipvs *ipvs) 1046 { 1047 return ACCESS_ONCE(ipvs->sysctl_sync_threshold[1]); 1048 } 1049 1050 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs) 1051 { 1052 return ACCESS_ONCE(ipvs->sysctl_sync_refresh_period); 1053 } 1054 1055 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs) 1056 { 1057 return ipvs->sysctl_sync_retries; 1058 } 1059 1060 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs) 1061 { 1062 return ipvs->sysctl_sync_ver; 1063 } 1064 1065 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs) 1066 { 1067 return ipvs->sysctl_sloppy_tcp; 1068 } 1069 1070 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs) 1071 { 1072 return ipvs->sysctl_sloppy_sctp; 1073 } 1074 1075 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs) 1076 { 1077 return ACCESS_ONCE(ipvs->sysctl_sync_ports); 1078 } 1079 1080 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs) 1081 { 1082 return ipvs->sysctl_sync_persist_mode; 1083 } 1084 1085 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs) 1086 { 1087 return ipvs->sysctl_sync_qlen_max; 1088 } 1089 1090 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs) 1091 { 1092 return ipvs->sysctl_sync_sock_size; 1093 } 1094 1095 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs) 1096 { 1097 return ipvs->sysctl_pmtu_disc; 1098 } 1099 1100 static inline int sysctl_backup_only(struct netns_ipvs *ipvs) 1101 { 1102 return ipvs->sync_state & IP_VS_STATE_BACKUP && 1103 ipvs->sysctl_backup_only; 1104 } 1105 1106 #else 1107 1108 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs) 1109 { 1110 return DEFAULT_SYNC_THRESHOLD; 1111 } 1112 1113 static inline int sysctl_sync_period(struct netns_ipvs *ipvs) 1114 { 1115 return DEFAULT_SYNC_PERIOD; 1116 } 1117 1118 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs) 1119 { 1120 return DEFAULT_SYNC_REFRESH_PERIOD; 1121 } 1122 1123 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs) 1124 { 1125 return DEFAULT_SYNC_RETRIES & 3; 1126 } 1127 1128 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs) 1129 { 1130 return DEFAULT_SYNC_VER; 1131 } 1132 1133 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs) 1134 { 1135 return DEFAULT_SLOPPY_TCP; 1136 } 1137 1138 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs) 1139 { 1140 return DEFAULT_SLOPPY_SCTP; 1141 } 1142 1143 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs) 1144 { 1145 return 1; 1146 } 1147 1148 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs) 1149 { 1150 return 0; 1151 } 1152 1153 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs) 1154 { 1155 return IPVS_SYNC_QLEN_MAX; 1156 } 1157 1158 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs) 1159 { 1160 return 0; 1161 } 1162 1163 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs) 1164 { 1165 return 1; 1166 } 1167 1168 static inline int sysctl_backup_only(struct netns_ipvs *ipvs) 1169 { 1170 return 0; 1171 } 1172 1173 #endif 1174 1175 /* 1176 * IPVS core functions 1177 * (from ip_vs_core.c) 1178 */ 1179 extern const char *ip_vs_proto_name(unsigned int proto); 1180 extern void ip_vs_init_hash_table(struct list_head *table, int rows); 1181 #define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t))) 1182 1183 #define IP_VS_APP_TYPE_FTP 1 1184 1185 /* 1186 * ip_vs_conn handling functions 1187 * (from ip_vs_conn.c) 1188 */ 1189 1190 enum { 1191 IP_VS_DIR_INPUT = 0, 1192 IP_VS_DIR_OUTPUT, 1193 IP_VS_DIR_INPUT_ONLY, 1194 IP_VS_DIR_LAST, 1195 }; 1196 1197 static inline void ip_vs_conn_fill_param(struct net *net, int af, int protocol, 1198 const union nf_inet_addr *caddr, 1199 __be16 cport, 1200 const union nf_inet_addr *vaddr, 1201 __be16 vport, 1202 struct ip_vs_conn_param *p) 1203 { 1204 p->net = net; 1205 p->af = af; 1206 p->protocol = protocol; 1207 p->caddr = caddr; 1208 p->cport = cport; 1209 p->vaddr = vaddr; 1210 p->vport = vport; 1211 p->pe = NULL; 1212 p->pe_data = NULL; 1213 } 1214 1215 struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p); 1216 struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p); 1217 1218 struct ip_vs_conn * ip_vs_conn_in_get_proto(int af, const struct sk_buff *skb, 1219 const struct ip_vs_iphdr *iph, 1220 int inverse); 1221 1222 struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p); 1223 1224 struct ip_vs_conn * ip_vs_conn_out_get_proto(int af, const struct sk_buff *skb, 1225 const struct ip_vs_iphdr *iph, 1226 int inverse); 1227 1228 /* Get reference to gain full access to conn. 1229 * By default, RCU read-side critical sections have access only to 1230 * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference. 1231 */ 1232 static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp) 1233 { 1234 return atomic_inc_not_zero(&cp->refcnt); 1235 } 1236 1237 /* put back the conn without restarting its timer */ 1238 static inline void __ip_vs_conn_put(struct ip_vs_conn *cp) 1239 { 1240 smp_mb__before_atomic_dec(); 1241 atomic_dec(&cp->refcnt); 1242 } 1243 extern void ip_vs_conn_put(struct ip_vs_conn *cp); 1244 extern void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport); 1245 1246 struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, 1247 const union nf_inet_addr *daddr, 1248 __be16 dport, unsigned int flags, 1249 struct ip_vs_dest *dest, __u32 fwmark); 1250 extern void ip_vs_conn_expire_now(struct ip_vs_conn *cp); 1251 1252 extern const char * ip_vs_state_name(__u16 proto, int state); 1253 1254 extern void ip_vs_tcp_conn_listen(struct net *net, struct ip_vs_conn *cp); 1255 extern int ip_vs_check_template(struct ip_vs_conn *ct); 1256 extern void ip_vs_random_dropentry(struct net *net); 1257 extern int ip_vs_conn_init(void); 1258 extern void ip_vs_conn_cleanup(void); 1259 1260 static inline void ip_vs_control_del(struct ip_vs_conn *cp) 1261 { 1262 struct ip_vs_conn *ctl_cp = cp->control; 1263 if (!ctl_cp) { 1264 IP_VS_ERR_BUF("request control DEL for uncontrolled: " 1265 "%s:%d to %s:%d\n", 1266 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1267 ntohs(cp->cport), 1268 IP_VS_DBG_ADDR(cp->af, &cp->vaddr), 1269 ntohs(cp->vport)); 1270 1271 return; 1272 } 1273 1274 IP_VS_DBG_BUF(7, "DELeting control for: " 1275 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n", 1276 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1277 ntohs(cp->cport), 1278 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr), 1279 ntohs(ctl_cp->cport)); 1280 1281 cp->control = NULL; 1282 if (atomic_read(&ctl_cp->n_control) == 0) { 1283 IP_VS_ERR_BUF("BUG control DEL with n=0 : " 1284 "%s:%d to %s:%d\n", 1285 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1286 ntohs(cp->cport), 1287 IP_VS_DBG_ADDR(cp->af, &cp->vaddr), 1288 ntohs(cp->vport)); 1289 1290 return; 1291 } 1292 atomic_dec(&ctl_cp->n_control); 1293 } 1294 1295 static inline void 1296 ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp) 1297 { 1298 if (cp->control) { 1299 IP_VS_ERR_BUF("request control ADD for already controlled: " 1300 "%s:%d to %s:%d\n", 1301 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1302 ntohs(cp->cport), 1303 IP_VS_DBG_ADDR(cp->af, &cp->vaddr), 1304 ntohs(cp->vport)); 1305 1306 ip_vs_control_del(cp); 1307 } 1308 1309 IP_VS_DBG_BUF(7, "ADDing control for: " 1310 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n", 1311 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1312 ntohs(cp->cport), 1313 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr), 1314 ntohs(ctl_cp->cport)); 1315 1316 cp->control = ctl_cp; 1317 atomic_inc(&ctl_cp->n_control); 1318 } 1319 1320 /* 1321 * IPVS netns init & cleanup functions 1322 */ 1323 extern int ip_vs_estimator_net_init(struct net *net); 1324 extern int ip_vs_control_net_init(struct net *net); 1325 extern int ip_vs_protocol_net_init(struct net *net); 1326 extern int ip_vs_app_net_init(struct net *net); 1327 extern int ip_vs_conn_net_init(struct net *net); 1328 extern int ip_vs_sync_net_init(struct net *net); 1329 extern void ip_vs_conn_net_cleanup(struct net *net); 1330 extern void ip_vs_app_net_cleanup(struct net *net); 1331 extern void ip_vs_protocol_net_cleanup(struct net *net); 1332 extern void ip_vs_control_net_cleanup(struct net *net); 1333 extern void ip_vs_estimator_net_cleanup(struct net *net); 1334 extern void ip_vs_sync_net_cleanup(struct net *net); 1335 extern void ip_vs_service_net_cleanup(struct net *net); 1336 1337 /* 1338 * IPVS application functions 1339 * (from ip_vs_app.c) 1340 */ 1341 #define IP_VS_APP_MAX_PORTS 8 1342 extern struct ip_vs_app *register_ip_vs_app(struct net *net, 1343 struct ip_vs_app *app); 1344 extern void unregister_ip_vs_app(struct net *net, struct ip_vs_app *app); 1345 extern int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp); 1346 extern void ip_vs_unbind_app(struct ip_vs_conn *cp); 1347 extern int register_ip_vs_app_inc(struct net *net, struct ip_vs_app *app, 1348 __u16 proto, __u16 port); 1349 extern int ip_vs_app_inc_get(struct ip_vs_app *inc); 1350 extern void ip_vs_app_inc_put(struct ip_vs_app *inc); 1351 1352 extern int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb); 1353 extern int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb); 1354 1355 int register_ip_vs_pe(struct ip_vs_pe *pe); 1356 int unregister_ip_vs_pe(struct ip_vs_pe *pe); 1357 struct ip_vs_pe *ip_vs_pe_getbyname(const char *name); 1358 struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name); 1359 1360 /* 1361 * Use a #define to avoid all of module.h just for these trivial ops 1362 */ 1363 #define ip_vs_pe_get(pe) \ 1364 if (pe && pe->module) \ 1365 __module_get(pe->module); 1366 1367 #define ip_vs_pe_put(pe) \ 1368 if (pe && pe->module) \ 1369 module_put(pe->module); 1370 1371 /* 1372 * IPVS protocol functions (from ip_vs_proto.c) 1373 */ 1374 extern int ip_vs_protocol_init(void); 1375 extern void ip_vs_protocol_cleanup(void); 1376 extern void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags); 1377 extern int *ip_vs_create_timeout_table(int *table, int size); 1378 extern int 1379 ip_vs_set_state_timeout(int *table, int num, const char *const *names, 1380 const char *name, int to); 1381 extern void 1382 ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp, 1383 const struct sk_buff *skb, 1384 int offset, const char *msg); 1385 1386 extern struct ip_vs_protocol ip_vs_protocol_tcp; 1387 extern struct ip_vs_protocol ip_vs_protocol_udp; 1388 extern struct ip_vs_protocol ip_vs_protocol_icmp; 1389 extern struct ip_vs_protocol ip_vs_protocol_esp; 1390 extern struct ip_vs_protocol ip_vs_protocol_ah; 1391 extern struct ip_vs_protocol ip_vs_protocol_sctp; 1392 1393 /* 1394 * Registering/unregistering scheduler functions 1395 * (from ip_vs_sched.c) 1396 */ 1397 extern int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler); 1398 extern int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler); 1399 extern int ip_vs_bind_scheduler(struct ip_vs_service *svc, 1400 struct ip_vs_scheduler *scheduler); 1401 extern void ip_vs_unbind_scheduler(struct ip_vs_service *svc, 1402 struct ip_vs_scheduler *sched); 1403 extern struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name); 1404 extern void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler); 1405 extern struct ip_vs_conn * 1406 ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb, 1407 struct ip_vs_proto_data *pd, int *ignored, 1408 struct ip_vs_iphdr *iph); 1409 extern int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb, 1410 struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph); 1411 1412 extern void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg); 1413 1414 1415 /* 1416 * IPVS control data and functions (from ip_vs_ctl.c) 1417 */ 1418 extern struct ip_vs_stats ip_vs_stats; 1419 extern int sysctl_ip_vs_sync_ver; 1420 1421 extern struct ip_vs_service * 1422 ip_vs_service_find(struct net *net, int af, __u32 fwmark, __u16 protocol, 1423 const union nf_inet_addr *vaddr, __be16 vport); 1424 1425 extern bool 1426 ip_vs_has_real_service(struct net *net, int af, __u16 protocol, 1427 const union nf_inet_addr *daddr, __be16 dport); 1428 1429 extern int ip_vs_use_count_inc(void); 1430 extern void ip_vs_use_count_dec(void); 1431 extern int ip_vs_register_nl_ioctl(void); 1432 extern void ip_vs_unregister_nl_ioctl(void); 1433 extern int ip_vs_control_init(void); 1434 extern void ip_vs_control_cleanup(void); 1435 extern struct ip_vs_dest * 1436 ip_vs_find_dest(struct net *net, int af, const union nf_inet_addr *daddr, 1437 __be16 dport, const union nf_inet_addr *vaddr, __be16 vport, 1438 __u16 protocol, __u32 fwmark, __u32 flags); 1439 extern void ip_vs_try_bind_dest(struct ip_vs_conn *cp); 1440 1441 static inline void ip_vs_dest_hold(struct ip_vs_dest *dest) 1442 { 1443 atomic_inc(&dest->refcnt); 1444 } 1445 1446 static inline void ip_vs_dest_put(struct ip_vs_dest *dest) 1447 { 1448 smp_mb__before_atomic_dec(); 1449 atomic_dec(&dest->refcnt); 1450 } 1451 1452 /* 1453 * IPVS sync daemon data and function prototypes 1454 * (from ip_vs_sync.c) 1455 */ 1456 extern int start_sync_thread(struct net *net, int state, char *mcast_ifn, 1457 __u8 syncid); 1458 extern int stop_sync_thread(struct net *net, int state); 1459 extern void ip_vs_sync_conn(struct net *net, struct ip_vs_conn *cp, int pkts); 1460 1461 1462 /* 1463 * IPVS rate estimator prototypes (from ip_vs_est.c) 1464 */ 1465 extern void ip_vs_start_estimator(struct net *net, struct ip_vs_stats *stats); 1466 extern void ip_vs_stop_estimator(struct net *net, struct ip_vs_stats *stats); 1467 extern void ip_vs_zero_estimator(struct ip_vs_stats *stats); 1468 extern void ip_vs_read_estimator(struct ip_vs_stats_user *dst, 1469 struct ip_vs_stats *stats); 1470 1471 /* 1472 * Various IPVS packet transmitters (from ip_vs_xmit.c) 1473 */ 1474 extern int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1475 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1476 extern int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1477 struct ip_vs_protocol *pp, 1478 struct ip_vs_iphdr *iph); 1479 extern int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1480 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1481 extern int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1482 struct ip_vs_protocol *pp, 1483 struct ip_vs_iphdr *iph); 1484 extern int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1485 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1486 extern int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1487 struct ip_vs_protocol *pp, int offset, 1488 unsigned int hooknum, struct ip_vs_iphdr *iph); 1489 extern void ip_vs_dest_dst_rcu_free(struct rcu_head *head); 1490 1491 #ifdef CONFIG_IP_VS_IPV6 1492 extern int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1493 struct ip_vs_protocol *pp, 1494 struct ip_vs_iphdr *iph); 1495 extern int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1496 struct ip_vs_protocol *pp, 1497 struct ip_vs_iphdr *iph); 1498 extern int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1499 struct ip_vs_protocol *pp, 1500 struct ip_vs_iphdr *iph); 1501 extern int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1502 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1503 extern int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1504 struct ip_vs_protocol *pp, int offset, 1505 unsigned int hooknum, struct ip_vs_iphdr *iph); 1506 #endif 1507 1508 #ifdef CONFIG_SYSCTL 1509 /* 1510 * This is a simple mechanism to ignore packets when 1511 * we are loaded. Just set ip_vs_drop_rate to 'n' and 1512 * we start to drop 1/rate of the packets 1513 */ 1514 1515 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) 1516 { 1517 if (!ipvs->drop_rate) 1518 return 0; 1519 if (--ipvs->drop_counter > 0) 1520 return 0; 1521 ipvs->drop_counter = ipvs->drop_rate; 1522 return 1; 1523 } 1524 #else 1525 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; } 1526 #endif 1527 1528 /* 1529 * ip_vs_fwd_tag returns the forwarding tag of the connection 1530 */ 1531 #define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK) 1532 1533 static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp) 1534 { 1535 char fwd; 1536 1537 switch (IP_VS_FWD_METHOD(cp)) { 1538 case IP_VS_CONN_F_MASQ: 1539 fwd = 'M'; break; 1540 case IP_VS_CONN_F_LOCALNODE: 1541 fwd = 'L'; break; 1542 case IP_VS_CONN_F_TUNNEL: 1543 fwd = 'T'; break; 1544 case IP_VS_CONN_F_DROUTE: 1545 fwd = 'R'; break; 1546 case IP_VS_CONN_F_BYPASS: 1547 fwd = 'B'; break; 1548 default: 1549 fwd = '?'; break; 1550 } 1551 return fwd; 1552 } 1553 1554 extern void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp, 1555 struct ip_vs_conn *cp, int dir); 1556 1557 #ifdef CONFIG_IP_VS_IPV6 1558 extern void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp, 1559 struct ip_vs_conn *cp, int dir); 1560 #endif 1561 1562 extern __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset); 1563 1564 static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum) 1565 { 1566 __be32 diff[2] = { ~old, new }; 1567 1568 return csum_partial(diff, sizeof(diff), oldsum); 1569 } 1570 1571 #ifdef CONFIG_IP_VS_IPV6 1572 static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new, 1573 __wsum oldsum) 1574 { 1575 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0], 1576 new[3], new[2], new[1], new[0] }; 1577 1578 return csum_partial(diff, sizeof(diff), oldsum); 1579 } 1580 #endif 1581 1582 static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum) 1583 { 1584 __be16 diff[2] = { ~old, new }; 1585 1586 return csum_partial(diff, sizeof(diff), oldsum); 1587 } 1588 1589 /* 1590 * Forget current conntrack (unconfirmed) and attach notrack entry 1591 */ 1592 static inline void ip_vs_notrack(struct sk_buff *skb) 1593 { 1594 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 1595 enum ip_conntrack_info ctinfo; 1596 struct nf_conn *ct = nf_ct_get(skb, &ctinfo); 1597 1598 if (!ct || !nf_ct_is_untracked(ct)) { 1599 nf_conntrack_put(skb->nfct); 1600 skb->nfct = &nf_ct_untracked_get()->ct_general; 1601 skb->nfctinfo = IP_CT_NEW; 1602 nf_conntrack_get(skb->nfct); 1603 } 1604 #endif 1605 } 1606 1607 #ifdef CONFIG_IP_VS_NFCT 1608 /* 1609 * Netfilter connection tracking 1610 * (from ip_vs_nfct.c) 1611 */ 1612 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs) 1613 { 1614 #ifdef CONFIG_SYSCTL 1615 return ipvs->sysctl_conntrack; 1616 #else 1617 return 0; 1618 #endif 1619 } 1620 1621 extern void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp, 1622 int outin); 1623 extern int ip_vs_confirm_conntrack(struct sk_buff *skb); 1624 extern void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct, 1625 struct ip_vs_conn *cp, u_int8_t proto, 1626 const __be16 port, int from_rs); 1627 extern void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp); 1628 1629 #else 1630 1631 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs) 1632 { 1633 return 0; 1634 } 1635 1636 static inline void ip_vs_update_conntrack(struct sk_buff *skb, 1637 struct ip_vs_conn *cp, int outin) 1638 { 1639 } 1640 1641 static inline int ip_vs_confirm_conntrack(struct sk_buff *skb) 1642 { 1643 return NF_ACCEPT; 1644 } 1645 1646 static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp) 1647 { 1648 } 1649 /* CONFIG_IP_VS_NFCT */ 1650 #endif 1651 1652 static inline unsigned int 1653 ip_vs_dest_conn_overhead(struct ip_vs_dest *dest) 1654 { 1655 /* 1656 * We think the overhead of processing active connections is 256 1657 * times higher than that of inactive connections in average. (This 1658 * 256 times might not be accurate, we will change it later) We 1659 * use the following formula to estimate the overhead now: 1660 * dest->activeconns*256 + dest->inactconns 1661 */ 1662 return (atomic_read(&dest->activeconns) << 8) + 1663 atomic_read(&dest->inactconns); 1664 } 1665 1666 #endif /* _NET_IP_VS_H */ 1667