1 /* IP Virtual Server 2 * data structure and functionality definitions 3 */ 4 5 #ifndef _NET_IP_VS_H 6 #define _NET_IP_VS_H 7 8 #include <linux/ip_vs.h> /* definitions shared with userland */ 9 10 #include <asm/types.h> /* for __uXX types */ 11 12 #include <linux/list.h> /* for struct list_head */ 13 #include <linux/spinlock.h> /* for struct rwlock_t */ 14 #include <linux/atomic.h> /* for struct atomic_t */ 15 #include <linux/compiler.h> 16 #include <linux/timer.h> 17 #include <linux/bug.h> 18 19 #include <net/checksum.h> 20 #include <linux/netfilter.h> /* for union nf_inet_addr */ 21 #include <linux/ip.h> 22 #include <linux/ipv6.h> /* for struct ipv6hdr */ 23 #include <net/ipv6.h> 24 #if IS_ENABLED(CONFIG_IP_VS_IPV6) 25 #include <linux/netfilter_ipv6/ip6_tables.h> 26 #endif 27 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 28 #include <net/netfilter/nf_conntrack.h> 29 #endif 30 #include <net/net_namespace.h> /* Netw namespace */ 31 32 /* Generic access of ipvs struct */ 33 static inline struct netns_ipvs *net_ipvs(struct net* net) 34 { 35 return net->ipvs; 36 } 37 38 /* Get net ptr from skb in traffic cases 39 * use skb_sknet when call is from userland (ioctl or netlink) 40 */ 41 static inline struct net *skb_net(const struct sk_buff *skb) 42 { 43 #ifdef CONFIG_NET_NS 44 #ifdef CONFIG_IP_VS_DEBUG 45 /* 46 * This is used for debug only. 47 * Start with the most likely hit 48 * End with BUG 49 */ 50 if (likely(skb->dev && dev_net(skb->dev))) 51 return dev_net(skb->dev); 52 if (skb_dst(skb) && skb_dst(skb)->dev) 53 return dev_net(skb_dst(skb)->dev); 54 WARN(skb->sk, "Maybe skb_sknet should be used in %s() at line:%d\n", 55 __func__, __LINE__); 56 if (likely(skb->sk && sock_net(skb->sk))) 57 return sock_net(skb->sk); 58 pr_err("There is no net ptr to find in the skb in %s() line:%d\n", 59 __func__, __LINE__); 60 BUG(); 61 #else 62 return dev_net(skb->dev ? : skb_dst(skb)->dev); 63 #endif 64 #else 65 return &init_net; 66 #endif 67 } 68 69 static inline struct net *skb_sknet(const struct sk_buff *skb) 70 { 71 #ifdef CONFIG_NET_NS 72 #ifdef CONFIG_IP_VS_DEBUG 73 /* Start with the most likely hit */ 74 if (likely(skb->sk && sock_net(skb->sk))) 75 return sock_net(skb->sk); 76 WARN(skb->dev, "Maybe skb_net should be used instead in %s() line:%d\n", 77 __func__, __LINE__); 78 if (likely(skb->dev && dev_net(skb->dev))) 79 return dev_net(skb->dev); 80 pr_err("There is no net ptr to find in the skb in %s() line:%d\n", 81 __func__, __LINE__); 82 BUG(); 83 #else 84 return sock_net(skb->sk); 85 #endif 86 #else 87 return &init_net; 88 #endif 89 } 90 91 /* This one needed for single_open_net since net is stored directly in 92 * private not as a struct i.e. seq_file_net can't be used. 93 */ 94 static inline struct net *seq_file_single_net(struct seq_file *seq) 95 { 96 #ifdef CONFIG_NET_NS 97 return (struct net *)seq->private; 98 #else 99 return &init_net; 100 #endif 101 } 102 103 /* Connections' size value needed by ip_vs_ctl.c */ 104 extern int ip_vs_conn_tab_size; 105 106 struct ip_vs_iphdr { 107 __u32 len; /* IPv4 simply where L4 starts 108 * IPv6 where L4 Transport Header starts */ 109 __u16 fragoffs; /* IPv6 fragment offset, 0 if first frag (or not frag)*/ 110 __s16 protocol; 111 __s32 flags; 112 union nf_inet_addr saddr; 113 union nf_inet_addr daddr; 114 }; 115 116 static inline void *frag_safe_skb_hp(const struct sk_buff *skb, int offset, 117 int len, void *buffer, 118 const struct ip_vs_iphdr *ipvsh) 119 { 120 return skb_header_pointer(skb, offset, len, buffer); 121 } 122 123 static inline void 124 ip_vs_fill_ip4hdr(const void *nh, struct ip_vs_iphdr *iphdr) 125 { 126 const struct iphdr *iph = nh; 127 128 iphdr->len = iph->ihl * 4; 129 iphdr->fragoffs = 0; 130 iphdr->protocol = iph->protocol; 131 iphdr->saddr.ip = iph->saddr; 132 iphdr->daddr.ip = iph->daddr; 133 } 134 135 /* This function handles filling *ip_vs_iphdr, both for IPv4 and IPv6. 136 * IPv6 requires some extra work, as finding proper header position, 137 * depend on the IPv6 extension headers. 138 */ 139 static inline void 140 ip_vs_fill_iph_skb(int af, const struct sk_buff *skb, struct ip_vs_iphdr *iphdr) 141 { 142 #ifdef CONFIG_IP_VS_IPV6 143 if (af == AF_INET6) { 144 const struct ipv6hdr *iph = 145 (struct ipv6hdr *)skb_network_header(skb); 146 iphdr->saddr.in6 = iph->saddr; 147 iphdr->daddr.in6 = iph->daddr; 148 /* ipv6_find_hdr() updates len, flags */ 149 iphdr->len = 0; 150 iphdr->flags = 0; 151 iphdr->protocol = ipv6_find_hdr(skb, &iphdr->len, -1, 152 &iphdr->fragoffs, 153 &iphdr->flags); 154 } else 155 #endif 156 { 157 const struct iphdr *iph = 158 (struct iphdr *)skb_network_header(skb); 159 iphdr->len = iph->ihl * 4; 160 iphdr->fragoffs = 0; 161 iphdr->protocol = iph->protocol; 162 iphdr->saddr.ip = iph->saddr; 163 iphdr->daddr.ip = iph->daddr; 164 } 165 } 166 167 static inline void ip_vs_addr_copy(int af, union nf_inet_addr *dst, 168 const union nf_inet_addr *src) 169 { 170 #ifdef CONFIG_IP_VS_IPV6 171 if (af == AF_INET6) 172 dst->in6 = src->in6; 173 else 174 #endif 175 dst->ip = src->ip; 176 } 177 178 static inline void ip_vs_addr_set(int af, union nf_inet_addr *dst, 179 const union nf_inet_addr *src) 180 { 181 #ifdef CONFIG_IP_VS_IPV6 182 if (af == AF_INET6) { 183 dst->in6 = src->in6; 184 return; 185 } 186 #endif 187 dst->ip = src->ip; 188 dst->all[1] = 0; 189 dst->all[2] = 0; 190 dst->all[3] = 0; 191 } 192 193 static inline int ip_vs_addr_equal(int af, const union nf_inet_addr *a, 194 const union nf_inet_addr *b) 195 { 196 #ifdef CONFIG_IP_VS_IPV6 197 if (af == AF_INET6) 198 return ipv6_addr_equal(&a->in6, &b->in6); 199 #endif 200 return a->ip == b->ip; 201 } 202 203 #ifdef CONFIG_IP_VS_DEBUG 204 #include <linux/net.h> 205 206 int ip_vs_get_debug_level(void); 207 208 static inline const char *ip_vs_dbg_addr(int af, char *buf, size_t buf_len, 209 const union nf_inet_addr *addr, 210 int *idx) 211 { 212 int len; 213 #ifdef CONFIG_IP_VS_IPV6 214 if (af == AF_INET6) 215 len = snprintf(&buf[*idx], buf_len - *idx, "[%pI6c]", 216 &addr->in6) + 1; 217 else 218 #endif 219 len = snprintf(&buf[*idx], buf_len - *idx, "%pI4", 220 &addr->ip) + 1; 221 222 *idx += len; 223 BUG_ON(*idx > buf_len + 1); 224 return &buf[*idx - len]; 225 } 226 227 #define IP_VS_DBG_BUF(level, msg, ...) \ 228 do { \ 229 char ip_vs_dbg_buf[160]; \ 230 int ip_vs_dbg_idx = 0; \ 231 if (level <= ip_vs_get_debug_level()) \ 232 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \ 233 } while (0) 234 #define IP_VS_ERR_BUF(msg...) \ 235 do { \ 236 char ip_vs_dbg_buf[160]; \ 237 int ip_vs_dbg_idx = 0; \ 238 pr_err(msg); \ 239 } while (0) 240 241 /* Only use from within IP_VS_DBG_BUF() or IP_VS_ERR_BUF macros */ 242 #define IP_VS_DBG_ADDR(af, addr) \ 243 ip_vs_dbg_addr(af, ip_vs_dbg_buf, \ 244 sizeof(ip_vs_dbg_buf), addr, \ 245 &ip_vs_dbg_idx) 246 247 #define IP_VS_DBG(level, msg, ...) \ 248 do { \ 249 if (level <= ip_vs_get_debug_level()) \ 250 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \ 251 } while (0) 252 #define IP_VS_DBG_RL(msg, ...) \ 253 do { \ 254 if (net_ratelimit()) \ 255 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \ 256 } while (0) 257 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) \ 258 do { \ 259 if (level <= ip_vs_get_debug_level()) \ 260 pp->debug_packet(af, pp, skb, ofs, msg); \ 261 } while (0) 262 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) \ 263 do { \ 264 if (level <= ip_vs_get_debug_level() && \ 265 net_ratelimit()) \ 266 pp->debug_packet(af, pp, skb, ofs, msg); \ 267 } while (0) 268 #else /* NO DEBUGGING at ALL */ 269 #define IP_VS_DBG_BUF(level, msg...) do {} while (0) 270 #define IP_VS_ERR_BUF(msg...) do {} while (0) 271 #define IP_VS_DBG(level, msg...) do {} while (0) 272 #define IP_VS_DBG_RL(msg...) do {} while (0) 273 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) do {} while (0) 274 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) do {} while (0) 275 #endif 276 277 #define IP_VS_BUG() BUG() 278 #define IP_VS_ERR_RL(msg, ...) \ 279 do { \ 280 if (net_ratelimit()) \ 281 pr_err(msg, ##__VA_ARGS__); \ 282 } while (0) 283 284 #ifdef CONFIG_IP_VS_DEBUG 285 #define EnterFunction(level) \ 286 do { \ 287 if (level <= ip_vs_get_debug_level()) \ 288 printk(KERN_DEBUG \ 289 pr_fmt("Enter: %s, %s line %i\n"), \ 290 __func__, __FILE__, __LINE__); \ 291 } while (0) 292 #define LeaveFunction(level) \ 293 do { \ 294 if (level <= ip_vs_get_debug_level()) \ 295 printk(KERN_DEBUG \ 296 pr_fmt("Leave: %s, %s line %i\n"), \ 297 __func__, __FILE__, __LINE__); \ 298 } while (0) 299 #else 300 #define EnterFunction(level) do {} while (0) 301 #define LeaveFunction(level) do {} while (0) 302 #endif 303 304 /* The port number of FTP service (in network order). */ 305 #define FTPPORT cpu_to_be16(21) 306 #define FTPDATA cpu_to_be16(20) 307 308 /* TCP State Values */ 309 enum { 310 IP_VS_TCP_S_NONE = 0, 311 IP_VS_TCP_S_ESTABLISHED, 312 IP_VS_TCP_S_SYN_SENT, 313 IP_VS_TCP_S_SYN_RECV, 314 IP_VS_TCP_S_FIN_WAIT, 315 IP_VS_TCP_S_TIME_WAIT, 316 IP_VS_TCP_S_CLOSE, 317 IP_VS_TCP_S_CLOSE_WAIT, 318 IP_VS_TCP_S_LAST_ACK, 319 IP_VS_TCP_S_LISTEN, 320 IP_VS_TCP_S_SYNACK, 321 IP_VS_TCP_S_LAST 322 }; 323 324 /* UDP State Values */ 325 enum { 326 IP_VS_UDP_S_NORMAL, 327 IP_VS_UDP_S_LAST, 328 }; 329 330 /* ICMP State Values */ 331 enum { 332 IP_VS_ICMP_S_NORMAL, 333 IP_VS_ICMP_S_LAST, 334 }; 335 336 /* SCTP State Values */ 337 enum ip_vs_sctp_states { 338 IP_VS_SCTP_S_NONE, 339 IP_VS_SCTP_S_INIT1, 340 IP_VS_SCTP_S_INIT, 341 IP_VS_SCTP_S_COOKIE_SENT, 342 IP_VS_SCTP_S_COOKIE_REPLIED, 343 IP_VS_SCTP_S_COOKIE_WAIT, 344 IP_VS_SCTP_S_COOKIE, 345 IP_VS_SCTP_S_COOKIE_ECHOED, 346 IP_VS_SCTP_S_ESTABLISHED, 347 IP_VS_SCTP_S_SHUTDOWN_SENT, 348 IP_VS_SCTP_S_SHUTDOWN_RECEIVED, 349 IP_VS_SCTP_S_SHUTDOWN_ACK_SENT, 350 IP_VS_SCTP_S_REJECTED, 351 IP_VS_SCTP_S_CLOSED, 352 IP_VS_SCTP_S_LAST 353 }; 354 355 /* Delta sequence info structure 356 * Each ip_vs_conn has 2 (output AND input seq. changes). 357 * Only used in the VS/NAT. 358 */ 359 struct ip_vs_seq { 360 __u32 init_seq; /* Add delta from this seq */ 361 __u32 delta; /* Delta in sequence numbers */ 362 __u32 previous_delta; /* Delta in sequence numbers 363 * before last resized pkt */ 364 }; 365 366 /* counters per cpu */ 367 struct ip_vs_counters { 368 __u64 conns; /* connections scheduled */ 369 __u64 inpkts; /* incoming packets */ 370 __u64 outpkts; /* outgoing packets */ 371 __u64 inbytes; /* incoming bytes */ 372 __u64 outbytes; /* outgoing bytes */ 373 }; 374 /* Stats per cpu */ 375 struct ip_vs_cpu_stats { 376 struct ip_vs_counters cnt; 377 struct u64_stats_sync syncp; 378 }; 379 380 /* IPVS statistics objects */ 381 struct ip_vs_estimator { 382 struct list_head list; 383 384 u64 last_inbytes; 385 u64 last_outbytes; 386 u64 last_conns; 387 u64 last_inpkts; 388 u64 last_outpkts; 389 390 u64 cps; 391 u64 inpps; 392 u64 outpps; 393 u64 inbps; 394 u64 outbps; 395 }; 396 397 /* 398 * IPVS statistics object, 64-bit kernel version of struct ip_vs_stats_user 399 */ 400 struct ip_vs_kstats { 401 u64 conns; /* connections scheduled */ 402 u64 inpkts; /* incoming packets */ 403 u64 outpkts; /* outgoing packets */ 404 u64 inbytes; /* incoming bytes */ 405 u64 outbytes; /* outgoing bytes */ 406 407 u64 cps; /* current connection rate */ 408 u64 inpps; /* current in packet rate */ 409 u64 outpps; /* current out packet rate */ 410 u64 inbps; /* current in byte rate */ 411 u64 outbps; /* current out byte rate */ 412 }; 413 414 struct ip_vs_stats { 415 struct ip_vs_kstats kstats; /* kernel statistics */ 416 struct ip_vs_estimator est; /* estimator */ 417 struct ip_vs_cpu_stats __percpu *cpustats; /* per cpu counters */ 418 spinlock_t lock; /* spin lock */ 419 struct ip_vs_kstats kstats0; /* reset values */ 420 }; 421 422 struct dst_entry; 423 struct iphdr; 424 struct ip_vs_conn; 425 struct ip_vs_app; 426 struct sk_buff; 427 struct ip_vs_proto_data; 428 429 struct ip_vs_protocol { 430 struct ip_vs_protocol *next; 431 char *name; 432 u16 protocol; 433 u16 num_states; 434 int dont_defrag; 435 436 void (*init)(struct ip_vs_protocol *pp); 437 438 void (*exit)(struct ip_vs_protocol *pp); 439 440 int (*init_netns)(struct net *net, struct ip_vs_proto_data *pd); 441 442 void (*exit_netns)(struct net *net, struct ip_vs_proto_data *pd); 443 444 int (*conn_schedule)(int af, struct sk_buff *skb, 445 struct ip_vs_proto_data *pd, 446 int *verdict, struct ip_vs_conn **cpp, 447 struct ip_vs_iphdr *iph); 448 449 struct ip_vs_conn * 450 (*conn_in_get)(int af, 451 const struct sk_buff *skb, 452 const struct ip_vs_iphdr *iph, 453 int inverse); 454 455 struct ip_vs_conn * 456 (*conn_out_get)(int af, 457 const struct sk_buff *skb, 458 const struct ip_vs_iphdr *iph, 459 int inverse); 460 461 int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp, 462 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph); 463 464 int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp, 465 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph); 466 467 int (*csum_check)(int af, struct sk_buff *skb, 468 struct ip_vs_protocol *pp); 469 470 const char *(*state_name)(int state); 471 472 void (*state_transition)(struct ip_vs_conn *cp, int direction, 473 const struct sk_buff *skb, 474 struct ip_vs_proto_data *pd); 475 476 int (*register_app)(struct net *net, struct ip_vs_app *inc); 477 478 void (*unregister_app)(struct net *net, struct ip_vs_app *inc); 479 480 int (*app_conn_bind)(struct ip_vs_conn *cp); 481 482 void (*debug_packet)(int af, struct ip_vs_protocol *pp, 483 const struct sk_buff *skb, 484 int offset, 485 const char *msg); 486 487 void (*timeout_change)(struct ip_vs_proto_data *pd, int flags); 488 }; 489 490 /* protocol data per netns */ 491 struct ip_vs_proto_data { 492 struct ip_vs_proto_data *next; 493 struct ip_vs_protocol *pp; 494 int *timeout_table; /* protocol timeout table */ 495 atomic_t appcnt; /* counter of proto app incs. */ 496 struct tcp_states_t *tcp_state_table; 497 }; 498 499 struct ip_vs_protocol *ip_vs_proto_get(unsigned short proto); 500 struct ip_vs_proto_data *ip_vs_proto_data_get(struct net *net, 501 unsigned short proto); 502 503 struct ip_vs_conn_param { 504 struct net *net; 505 const union nf_inet_addr *caddr; 506 const union nf_inet_addr *vaddr; 507 __be16 cport; 508 __be16 vport; 509 __u16 protocol; 510 u16 af; 511 512 const struct ip_vs_pe *pe; 513 char *pe_data; 514 __u8 pe_data_len; 515 }; 516 517 /* IP_VS structure allocated for each dynamically scheduled connection */ 518 struct ip_vs_conn { 519 struct hlist_node c_list; /* hashed list heads */ 520 /* Protocol, addresses and port numbers */ 521 __be16 cport; 522 __be16 dport; 523 __be16 vport; 524 u16 af; /* address family */ 525 union nf_inet_addr caddr; /* client address */ 526 union nf_inet_addr vaddr; /* virtual address */ 527 union nf_inet_addr daddr; /* destination address */ 528 volatile __u32 flags; /* status flags */ 529 __u16 protocol; /* Which protocol (TCP/UDP) */ 530 __u16 daf; /* Address family of the dest */ 531 #ifdef CONFIG_NET_NS 532 struct net *net; /* Name space */ 533 #endif 534 535 /* counter and timer */ 536 atomic_t refcnt; /* reference count */ 537 struct timer_list timer; /* Expiration timer */ 538 volatile unsigned long timeout; /* timeout */ 539 540 /* Flags and state transition */ 541 spinlock_t lock; /* lock for state transition */ 542 volatile __u16 state; /* state info */ 543 volatile __u16 old_state; /* old state, to be used for 544 * state transition triggerd 545 * synchronization 546 */ 547 __u32 fwmark; /* Fire wall mark from skb */ 548 unsigned long sync_endtime; /* jiffies + sent_retries */ 549 550 /* Control members */ 551 struct ip_vs_conn *control; /* Master control connection */ 552 atomic_t n_control; /* Number of controlled ones */ 553 struct ip_vs_dest *dest; /* real server */ 554 atomic_t in_pkts; /* incoming packet counter */ 555 556 /* Packet transmitter for different forwarding methods. If it 557 * mangles the packet, it must return NF_DROP or better NF_STOLEN, 558 * otherwise this must be changed to a sk_buff **. 559 * NF_ACCEPT can be returned when destination is local. 560 */ 561 int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp, 562 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 563 564 /* Note: we can group the following members into a structure, 565 * in order to save more space, and the following members are 566 * only used in VS/NAT anyway 567 */ 568 struct ip_vs_app *app; /* bound ip_vs_app object */ 569 void *app_data; /* Application private data */ 570 struct ip_vs_seq in_seq; /* incoming seq. struct */ 571 struct ip_vs_seq out_seq; /* outgoing seq. struct */ 572 573 const struct ip_vs_pe *pe; 574 char *pe_data; 575 __u8 pe_data_len; 576 577 struct rcu_head rcu_head; 578 }; 579 580 /* To save some memory in conn table when name space is disabled. */ 581 static inline struct net *ip_vs_conn_net(const struct ip_vs_conn *cp) 582 { 583 #ifdef CONFIG_NET_NS 584 return cp->net; 585 #else 586 return &init_net; 587 #endif 588 } 589 590 static inline void ip_vs_conn_net_set(struct ip_vs_conn *cp, struct net *net) 591 { 592 #ifdef CONFIG_NET_NS 593 cp->net = net; 594 #endif 595 } 596 597 static inline int ip_vs_conn_net_eq(const struct ip_vs_conn *cp, 598 struct net *net) 599 { 600 #ifdef CONFIG_NET_NS 601 return cp->net == net; 602 #else 603 return 1; 604 #endif 605 } 606 607 /* Extended internal versions of struct ip_vs_service_user and ip_vs_dest_user 608 * for IPv6 support. 609 * 610 * We need these to conveniently pass around service and destination 611 * options, but unfortunately, we also need to keep the old definitions to 612 * maintain userspace backwards compatibility for the setsockopt interface. 613 */ 614 struct ip_vs_service_user_kern { 615 /* virtual service addresses */ 616 u16 af; 617 u16 protocol; 618 union nf_inet_addr addr; /* virtual ip address */ 619 __be16 port; 620 u32 fwmark; /* firwall mark of service */ 621 622 /* virtual service options */ 623 char *sched_name; 624 char *pe_name; 625 unsigned int flags; /* virtual service flags */ 626 unsigned int timeout; /* persistent timeout in sec */ 627 __be32 netmask; /* persistent netmask or plen */ 628 }; 629 630 631 struct ip_vs_dest_user_kern { 632 /* destination server address */ 633 union nf_inet_addr addr; 634 __be16 port; 635 636 /* real server options */ 637 unsigned int conn_flags; /* connection flags */ 638 int weight; /* destination weight */ 639 640 /* thresholds for active connections */ 641 u32 u_threshold; /* upper threshold */ 642 u32 l_threshold; /* lower threshold */ 643 644 /* Address family of addr */ 645 u16 af; 646 }; 647 648 649 /* 650 * The information about the virtual service offered to the net and the 651 * forwarding entries. 652 */ 653 struct ip_vs_service { 654 struct hlist_node s_list; /* for normal service table */ 655 struct hlist_node f_list; /* for fwmark-based service table */ 656 atomic_t refcnt; /* reference counter */ 657 658 u16 af; /* address family */ 659 __u16 protocol; /* which protocol (TCP/UDP) */ 660 union nf_inet_addr addr; /* IP address for virtual service */ 661 __be16 port; /* port number for the service */ 662 __u32 fwmark; /* firewall mark of the service */ 663 unsigned int flags; /* service status flags */ 664 unsigned int timeout; /* persistent timeout in ticks */ 665 __be32 netmask; /* grouping granularity, mask/plen */ 666 struct net *net; 667 668 struct list_head destinations; /* real server d-linked list */ 669 __u32 num_dests; /* number of servers */ 670 struct ip_vs_stats stats; /* statistics for the service */ 671 672 /* for scheduling */ 673 struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */ 674 spinlock_t sched_lock; /* lock sched_data */ 675 void *sched_data; /* scheduler application data */ 676 677 /* alternate persistence engine */ 678 struct ip_vs_pe __rcu *pe; 679 680 struct rcu_head rcu_head; 681 }; 682 683 /* Information for cached dst */ 684 struct ip_vs_dest_dst { 685 struct dst_entry *dst_cache; /* destination cache entry */ 686 u32 dst_cookie; 687 union nf_inet_addr dst_saddr; 688 struct rcu_head rcu_head; 689 }; 690 691 /* The real server destination forwarding entry with ip address, port number, 692 * and so on. 693 */ 694 struct ip_vs_dest { 695 struct list_head n_list; /* for the dests in the service */ 696 struct hlist_node d_list; /* for table with all the dests */ 697 698 u16 af; /* address family */ 699 __be16 port; /* port number of the server */ 700 union nf_inet_addr addr; /* IP address of the server */ 701 volatile unsigned int flags; /* dest status flags */ 702 atomic_t conn_flags; /* flags to copy to conn */ 703 atomic_t weight; /* server weight */ 704 705 atomic_t refcnt; /* reference counter */ 706 struct ip_vs_stats stats; /* statistics */ 707 unsigned long idle_start; /* start time, jiffies */ 708 709 /* connection counters and thresholds */ 710 atomic_t activeconns; /* active connections */ 711 atomic_t inactconns; /* inactive connections */ 712 atomic_t persistconns; /* persistent connections */ 713 __u32 u_threshold; /* upper threshold */ 714 __u32 l_threshold; /* lower threshold */ 715 716 /* for destination cache */ 717 spinlock_t dst_lock; /* lock of dst_cache */ 718 struct ip_vs_dest_dst __rcu *dest_dst; /* cached dst info */ 719 720 /* for virtual service */ 721 struct ip_vs_service __rcu *svc; /* service it belongs to */ 722 __u16 protocol; /* which protocol (TCP/UDP) */ 723 __be16 vport; /* virtual port number */ 724 union nf_inet_addr vaddr; /* virtual IP address */ 725 __u32 vfwmark; /* firewall mark of service */ 726 727 struct list_head t_list; /* in dest_trash */ 728 unsigned int in_rs_table:1; /* we are in rs_table */ 729 }; 730 731 /* The scheduler object */ 732 struct ip_vs_scheduler { 733 struct list_head n_list; /* d-linked list head */ 734 char *name; /* scheduler name */ 735 atomic_t refcnt; /* reference counter */ 736 struct module *module; /* THIS_MODULE/NULL */ 737 738 /* scheduler initializing service */ 739 int (*init_service)(struct ip_vs_service *svc); 740 /* scheduling service finish */ 741 void (*done_service)(struct ip_vs_service *svc); 742 /* dest is linked */ 743 int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest); 744 /* dest is unlinked */ 745 int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest); 746 /* dest is updated */ 747 int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest); 748 749 /* selecting a server from the given service */ 750 struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc, 751 const struct sk_buff *skb, 752 struct ip_vs_iphdr *iph); 753 }; 754 755 /* The persistence engine object */ 756 struct ip_vs_pe { 757 struct list_head n_list; /* d-linked list head */ 758 char *name; /* scheduler name */ 759 atomic_t refcnt; /* reference counter */ 760 struct module *module; /* THIS_MODULE/NULL */ 761 762 /* get the connection template, if any */ 763 int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb); 764 bool (*ct_match)(const struct ip_vs_conn_param *p, 765 struct ip_vs_conn *ct); 766 u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval, 767 bool inverse); 768 int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf); 769 }; 770 771 /* The application module object (a.k.a. app incarnation) */ 772 struct ip_vs_app { 773 struct list_head a_list; /* member in app list */ 774 int type; /* IP_VS_APP_TYPE_xxx */ 775 char *name; /* application module name */ 776 __u16 protocol; 777 struct module *module; /* THIS_MODULE/NULL */ 778 struct list_head incs_list; /* list of incarnations */ 779 780 /* members for application incarnations */ 781 struct list_head p_list; /* member in proto app list */ 782 struct ip_vs_app *app; /* its real application */ 783 __be16 port; /* port number in net order */ 784 atomic_t usecnt; /* usage counter */ 785 struct rcu_head rcu_head; 786 787 /* output hook: Process packet in inout direction, diff set for TCP. 788 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok, 789 * 2=Mangled but checksum was not updated 790 */ 791 int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *, 792 struct sk_buff *, int *diff); 793 794 /* input hook: Process packet in outin direction, diff set for TCP. 795 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok, 796 * 2=Mangled but checksum was not updated 797 */ 798 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *, 799 struct sk_buff *, int *diff); 800 801 /* ip_vs_app initializer */ 802 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *); 803 804 /* ip_vs_app finish */ 805 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *); 806 807 808 /* not used now */ 809 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *, 810 struct ip_vs_protocol *); 811 812 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *); 813 814 int * timeout_table; 815 int * timeouts; 816 int timeouts_size; 817 818 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app, 819 int *verdict, struct ip_vs_conn **cpp); 820 821 struct ip_vs_conn * 822 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app, 823 const struct iphdr *iph, int inverse); 824 825 struct ip_vs_conn * 826 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app, 827 const struct iphdr *iph, int inverse); 828 829 int (*state_transition)(struct ip_vs_conn *cp, int direction, 830 const struct sk_buff *skb, 831 struct ip_vs_app *app); 832 833 void (*timeout_change)(struct ip_vs_app *app, int flags); 834 }; 835 836 struct ipvs_master_sync_state { 837 struct list_head sync_queue; 838 struct ip_vs_sync_buff *sync_buff; 839 unsigned long sync_queue_len; 840 unsigned int sync_queue_delay; 841 struct task_struct *master_thread; 842 struct delayed_work master_wakeup_work; 843 struct netns_ipvs *ipvs; 844 }; 845 846 /* How much time to keep dests in trash */ 847 #define IP_VS_DEST_TRASH_PERIOD (120 * HZ) 848 849 struct ipvs_sync_daemon_cfg { 850 union nf_inet_addr mcast_group; 851 int syncid; 852 u16 sync_maxlen; 853 u16 mcast_port; 854 u8 mcast_af; 855 u8 mcast_ttl; 856 /* multicast interface name */ 857 char mcast_ifn[IP_VS_IFNAME_MAXLEN]; 858 }; 859 860 /* IPVS in network namespace */ 861 struct netns_ipvs { 862 int gen; /* Generation */ 863 int enable; /* enable like nf_hooks do */ 864 /* Hash table: for real service lookups */ 865 #define IP_VS_RTAB_BITS 4 866 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS) 867 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1) 868 869 struct hlist_head rs_table[IP_VS_RTAB_SIZE]; 870 /* ip_vs_app */ 871 struct list_head app_list; 872 /* ip_vs_proto */ 873 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */ 874 struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE]; 875 /* ip_vs_proto_tcp */ 876 #ifdef CONFIG_IP_VS_PROTO_TCP 877 #define TCP_APP_TAB_BITS 4 878 #define TCP_APP_TAB_SIZE (1 << TCP_APP_TAB_BITS) 879 #define TCP_APP_TAB_MASK (TCP_APP_TAB_SIZE - 1) 880 struct list_head tcp_apps[TCP_APP_TAB_SIZE]; 881 #endif 882 /* ip_vs_proto_udp */ 883 #ifdef CONFIG_IP_VS_PROTO_UDP 884 #define UDP_APP_TAB_BITS 4 885 #define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS) 886 #define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1) 887 struct list_head udp_apps[UDP_APP_TAB_SIZE]; 888 #endif 889 /* ip_vs_proto_sctp */ 890 #ifdef CONFIG_IP_VS_PROTO_SCTP 891 #define SCTP_APP_TAB_BITS 4 892 #define SCTP_APP_TAB_SIZE (1 << SCTP_APP_TAB_BITS) 893 #define SCTP_APP_TAB_MASK (SCTP_APP_TAB_SIZE - 1) 894 /* Hash table for SCTP application incarnations */ 895 struct list_head sctp_apps[SCTP_APP_TAB_SIZE]; 896 #endif 897 /* ip_vs_conn */ 898 atomic_t conn_count; /* connection counter */ 899 900 /* ip_vs_ctl */ 901 struct ip_vs_stats tot_stats; /* Statistics & est. */ 902 903 int num_services; /* no of virtual services */ 904 905 /* Trash for destinations */ 906 struct list_head dest_trash; 907 spinlock_t dest_trash_lock; 908 struct timer_list dest_trash_timer; /* expiration timer */ 909 /* Service counters */ 910 atomic_t ftpsvc_counter; 911 atomic_t nullsvc_counter; 912 913 #ifdef CONFIG_SYSCTL 914 /* 1/rate drop and drop-entry variables */ 915 struct delayed_work defense_work; /* Work handler */ 916 int drop_rate; 917 int drop_counter; 918 atomic_t dropentry; 919 /* locks in ctl.c */ 920 spinlock_t dropentry_lock; /* drop entry handling */ 921 spinlock_t droppacket_lock; /* drop packet handling */ 922 spinlock_t securetcp_lock; /* state and timeout tables */ 923 924 /* sys-ctl struct */ 925 struct ctl_table_header *sysctl_hdr; 926 struct ctl_table *sysctl_tbl; 927 #endif 928 929 /* sysctl variables */ 930 int sysctl_amemthresh; 931 int sysctl_am_droprate; 932 int sysctl_drop_entry; 933 int sysctl_drop_packet; 934 int sysctl_secure_tcp; 935 #ifdef CONFIG_IP_VS_NFCT 936 int sysctl_conntrack; 937 #endif 938 int sysctl_snat_reroute; 939 int sysctl_sync_ver; 940 int sysctl_sync_ports; 941 int sysctl_sync_persist_mode; 942 unsigned long sysctl_sync_qlen_max; 943 int sysctl_sync_sock_size; 944 int sysctl_cache_bypass; 945 int sysctl_expire_nodest_conn; 946 int sysctl_sloppy_tcp; 947 int sysctl_sloppy_sctp; 948 int sysctl_expire_quiescent_template; 949 int sysctl_sync_threshold[2]; 950 unsigned int sysctl_sync_refresh_period; 951 int sysctl_sync_retries; 952 int sysctl_nat_icmp_send; 953 int sysctl_pmtu_disc; 954 int sysctl_backup_only; 955 int sysctl_conn_reuse_mode; 956 957 /* ip_vs_lblc */ 958 int sysctl_lblc_expiration; 959 struct ctl_table_header *lblc_ctl_header; 960 struct ctl_table *lblc_ctl_table; 961 /* ip_vs_lblcr */ 962 int sysctl_lblcr_expiration; 963 struct ctl_table_header *lblcr_ctl_header; 964 struct ctl_table *lblcr_ctl_table; 965 /* ip_vs_est */ 966 struct list_head est_list; /* estimator list */ 967 spinlock_t est_lock; 968 struct timer_list est_timer; /* Estimation timer */ 969 /* ip_vs_sync */ 970 spinlock_t sync_lock; 971 struct ipvs_master_sync_state *ms; 972 spinlock_t sync_buff_lock; 973 struct task_struct **backup_threads; 974 int threads_mask; 975 volatile int sync_state; 976 struct mutex sync_mutex; 977 struct ipvs_sync_daemon_cfg mcfg; /* Master Configuration */ 978 struct ipvs_sync_daemon_cfg bcfg; /* Backup Configuration */ 979 /* net name space ptr */ 980 struct net *net; /* Needed by timer routines */ 981 /* Number of heterogeneous destinations, needed becaus heterogeneous 982 * are not supported when synchronization is enabled. 983 */ 984 unsigned int mixed_address_family_dests; 985 }; 986 987 #define DEFAULT_SYNC_THRESHOLD 3 988 #define DEFAULT_SYNC_PERIOD 50 989 #define DEFAULT_SYNC_VER 1 990 #define DEFAULT_SLOPPY_TCP 0 991 #define DEFAULT_SLOPPY_SCTP 0 992 #define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ) 993 #define DEFAULT_SYNC_RETRIES 0 994 #define IPVS_SYNC_WAKEUP_RATE 8 995 #define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4) 996 #define IPVS_SYNC_SEND_DELAY (HZ / 50) 997 #define IPVS_SYNC_CHECK_PERIOD HZ 998 #define IPVS_SYNC_FLUSH_TIME (HZ * 2) 999 #define IPVS_SYNC_PORTS_MAX (1 << 6) 1000 1001 #ifdef CONFIG_SYSCTL 1002 1003 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs) 1004 { 1005 return ipvs->sysctl_sync_threshold[0]; 1006 } 1007 1008 static inline int sysctl_sync_period(struct netns_ipvs *ipvs) 1009 { 1010 return ACCESS_ONCE(ipvs->sysctl_sync_threshold[1]); 1011 } 1012 1013 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs) 1014 { 1015 return ACCESS_ONCE(ipvs->sysctl_sync_refresh_period); 1016 } 1017 1018 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs) 1019 { 1020 return ipvs->sysctl_sync_retries; 1021 } 1022 1023 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs) 1024 { 1025 return ipvs->sysctl_sync_ver; 1026 } 1027 1028 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs) 1029 { 1030 return ipvs->sysctl_sloppy_tcp; 1031 } 1032 1033 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs) 1034 { 1035 return ipvs->sysctl_sloppy_sctp; 1036 } 1037 1038 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs) 1039 { 1040 return ACCESS_ONCE(ipvs->sysctl_sync_ports); 1041 } 1042 1043 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs) 1044 { 1045 return ipvs->sysctl_sync_persist_mode; 1046 } 1047 1048 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs) 1049 { 1050 return ipvs->sysctl_sync_qlen_max; 1051 } 1052 1053 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs) 1054 { 1055 return ipvs->sysctl_sync_sock_size; 1056 } 1057 1058 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs) 1059 { 1060 return ipvs->sysctl_pmtu_disc; 1061 } 1062 1063 static inline int sysctl_backup_only(struct netns_ipvs *ipvs) 1064 { 1065 return ipvs->sync_state & IP_VS_STATE_BACKUP && 1066 ipvs->sysctl_backup_only; 1067 } 1068 1069 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs) 1070 { 1071 return ipvs->sysctl_conn_reuse_mode; 1072 } 1073 1074 #else 1075 1076 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs) 1077 { 1078 return DEFAULT_SYNC_THRESHOLD; 1079 } 1080 1081 static inline int sysctl_sync_period(struct netns_ipvs *ipvs) 1082 { 1083 return DEFAULT_SYNC_PERIOD; 1084 } 1085 1086 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs) 1087 { 1088 return DEFAULT_SYNC_REFRESH_PERIOD; 1089 } 1090 1091 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs) 1092 { 1093 return DEFAULT_SYNC_RETRIES & 3; 1094 } 1095 1096 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs) 1097 { 1098 return DEFAULT_SYNC_VER; 1099 } 1100 1101 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs) 1102 { 1103 return DEFAULT_SLOPPY_TCP; 1104 } 1105 1106 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs) 1107 { 1108 return DEFAULT_SLOPPY_SCTP; 1109 } 1110 1111 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs) 1112 { 1113 return 1; 1114 } 1115 1116 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs) 1117 { 1118 return 0; 1119 } 1120 1121 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs) 1122 { 1123 return IPVS_SYNC_QLEN_MAX; 1124 } 1125 1126 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs) 1127 { 1128 return 0; 1129 } 1130 1131 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs) 1132 { 1133 return 1; 1134 } 1135 1136 static inline int sysctl_backup_only(struct netns_ipvs *ipvs) 1137 { 1138 return 0; 1139 } 1140 1141 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs) 1142 { 1143 return 1; 1144 } 1145 1146 #endif 1147 1148 /* IPVS core functions 1149 * (from ip_vs_core.c) 1150 */ 1151 const char *ip_vs_proto_name(unsigned int proto); 1152 void ip_vs_init_hash_table(struct list_head *table, int rows); 1153 #define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t))) 1154 1155 #define IP_VS_APP_TYPE_FTP 1 1156 1157 /* ip_vs_conn handling functions 1158 * (from ip_vs_conn.c) 1159 */ 1160 enum { 1161 IP_VS_DIR_INPUT = 0, 1162 IP_VS_DIR_OUTPUT, 1163 IP_VS_DIR_INPUT_ONLY, 1164 IP_VS_DIR_LAST, 1165 }; 1166 1167 static inline void ip_vs_conn_fill_param(struct net *net, int af, int protocol, 1168 const union nf_inet_addr *caddr, 1169 __be16 cport, 1170 const union nf_inet_addr *vaddr, 1171 __be16 vport, 1172 struct ip_vs_conn_param *p) 1173 { 1174 p->net = net; 1175 p->af = af; 1176 p->protocol = protocol; 1177 p->caddr = caddr; 1178 p->cport = cport; 1179 p->vaddr = vaddr; 1180 p->vport = vport; 1181 p->pe = NULL; 1182 p->pe_data = NULL; 1183 } 1184 1185 struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p); 1186 struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p); 1187 1188 struct ip_vs_conn * ip_vs_conn_in_get_proto(int af, const struct sk_buff *skb, 1189 const struct ip_vs_iphdr *iph, 1190 int inverse); 1191 1192 struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p); 1193 1194 struct ip_vs_conn * ip_vs_conn_out_get_proto(int af, const struct sk_buff *skb, 1195 const struct ip_vs_iphdr *iph, 1196 int inverse); 1197 1198 /* Get reference to gain full access to conn. 1199 * By default, RCU read-side critical sections have access only to 1200 * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference. 1201 */ 1202 static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp) 1203 { 1204 return atomic_inc_not_zero(&cp->refcnt); 1205 } 1206 1207 /* put back the conn without restarting its timer */ 1208 static inline void __ip_vs_conn_put(struct ip_vs_conn *cp) 1209 { 1210 smp_mb__before_atomic(); 1211 atomic_dec(&cp->refcnt); 1212 } 1213 void ip_vs_conn_put(struct ip_vs_conn *cp); 1214 void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport); 1215 1216 struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, int dest_af, 1217 const union nf_inet_addr *daddr, 1218 __be16 dport, unsigned int flags, 1219 struct ip_vs_dest *dest, __u32 fwmark); 1220 void ip_vs_conn_expire_now(struct ip_vs_conn *cp); 1221 1222 const char *ip_vs_state_name(__u16 proto, int state); 1223 1224 void ip_vs_tcp_conn_listen(struct net *net, struct ip_vs_conn *cp); 1225 int ip_vs_check_template(struct ip_vs_conn *ct); 1226 void ip_vs_random_dropentry(struct net *net); 1227 int ip_vs_conn_init(void); 1228 void ip_vs_conn_cleanup(void); 1229 1230 static inline void ip_vs_control_del(struct ip_vs_conn *cp) 1231 { 1232 struct ip_vs_conn *ctl_cp = cp->control; 1233 if (!ctl_cp) { 1234 IP_VS_ERR_BUF("request control DEL for uncontrolled: " 1235 "%s:%d to %s:%d\n", 1236 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1237 ntohs(cp->cport), 1238 IP_VS_DBG_ADDR(cp->af, &cp->vaddr), 1239 ntohs(cp->vport)); 1240 1241 return; 1242 } 1243 1244 IP_VS_DBG_BUF(7, "DELeting control for: " 1245 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n", 1246 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1247 ntohs(cp->cport), 1248 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr), 1249 ntohs(ctl_cp->cport)); 1250 1251 cp->control = NULL; 1252 if (atomic_read(&ctl_cp->n_control) == 0) { 1253 IP_VS_ERR_BUF("BUG control DEL with n=0 : " 1254 "%s:%d to %s:%d\n", 1255 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1256 ntohs(cp->cport), 1257 IP_VS_DBG_ADDR(cp->af, &cp->vaddr), 1258 ntohs(cp->vport)); 1259 1260 return; 1261 } 1262 atomic_dec(&ctl_cp->n_control); 1263 } 1264 1265 static inline void 1266 ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp) 1267 { 1268 if (cp->control) { 1269 IP_VS_ERR_BUF("request control ADD for already controlled: " 1270 "%s:%d to %s:%d\n", 1271 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1272 ntohs(cp->cport), 1273 IP_VS_DBG_ADDR(cp->af, &cp->vaddr), 1274 ntohs(cp->vport)); 1275 1276 ip_vs_control_del(cp); 1277 } 1278 1279 IP_VS_DBG_BUF(7, "ADDing control for: " 1280 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n", 1281 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1282 ntohs(cp->cport), 1283 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr), 1284 ntohs(ctl_cp->cport)); 1285 1286 cp->control = ctl_cp; 1287 atomic_inc(&ctl_cp->n_control); 1288 } 1289 1290 /* IPVS netns init & cleanup functions */ 1291 int ip_vs_estimator_net_init(struct net *net); 1292 int ip_vs_control_net_init(struct net *net); 1293 int ip_vs_protocol_net_init(struct net *net); 1294 int ip_vs_app_net_init(struct net *net); 1295 int ip_vs_conn_net_init(struct net *net); 1296 int ip_vs_sync_net_init(struct net *net); 1297 void ip_vs_conn_net_cleanup(struct net *net); 1298 void ip_vs_app_net_cleanup(struct net *net); 1299 void ip_vs_protocol_net_cleanup(struct net *net); 1300 void ip_vs_control_net_cleanup(struct net *net); 1301 void ip_vs_estimator_net_cleanup(struct net *net); 1302 void ip_vs_sync_net_cleanup(struct net *net); 1303 void ip_vs_service_net_cleanup(struct net *net); 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 /* Use a #define to avoid all of module.h just for these trivial ops */ 1327 #define ip_vs_pe_get(pe) \ 1328 if (pe && pe->module) \ 1329 __module_get(pe->module); 1330 1331 #define ip_vs_pe_put(pe) \ 1332 if (pe && pe->module) \ 1333 module_put(pe->module); 1334 1335 /* IPVS protocol functions (from ip_vs_proto.c) */ 1336 int ip_vs_protocol_init(void); 1337 void ip_vs_protocol_cleanup(void); 1338 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags); 1339 int *ip_vs_create_timeout_table(int *table, int size); 1340 int ip_vs_set_state_timeout(int *table, int num, const char *const *names, 1341 const char *name, int to); 1342 void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp, 1343 const struct sk_buff *skb, int offset, 1344 const char *msg); 1345 1346 extern struct ip_vs_protocol ip_vs_protocol_tcp; 1347 extern struct ip_vs_protocol ip_vs_protocol_udp; 1348 extern struct ip_vs_protocol ip_vs_protocol_icmp; 1349 extern struct ip_vs_protocol ip_vs_protocol_esp; 1350 extern struct ip_vs_protocol ip_vs_protocol_ah; 1351 extern struct ip_vs_protocol ip_vs_protocol_sctp; 1352 1353 /* Registering/unregistering scheduler functions 1354 * (from ip_vs_sched.c) 1355 */ 1356 int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler); 1357 int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler); 1358 int ip_vs_bind_scheduler(struct ip_vs_service *svc, 1359 struct ip_vs_scheduler *scheduler); 1360 void ip_vs_unbind_scheduler(struct ip_vs_service *svc, 1361 struct ip_vs_scheduler *sched); 1362 struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name); 1363 void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler); 1364 struct ip_vs_conn * 1365 ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb, 1366 struct ip_vs_proto_data *pd, int *ignored, 1367 struct ip_vs_iphdr *iph); 1368 int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb, 1369 struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph); 1370 1371 void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg); 1372 1373 /* IPVS control data and functions (from ip_vs_ctl.c) */ 1374 extern struct ip_vs_stats ip_vs_stats; 1375 extern int sysctl_ip_vs_sync_ver; 1376 1377 struct ip_vs_service * 1378 ip_vs_service_find(struct net *net, int af, __u32 fwmark, __u16 protocol, 1379 const union nf_inet_addr *vaddr, __be16 vport); 1380 1381 bool ip_vs_has_real_service(struct net *net, int af, __u16 protocol, 1382 const union nf_inet_addr *daddr, __be16 dport); 1383 1384 int ip_vs_use_count_inc(void); 1385 void ip_vs_use_count_dec(void); 1386 int ip_vs_register_nl_ioctl(void); 1387 void ip_vs_unregister_nl_ioctl(void); 1388 int ip_vs_control_init(void); 1389 void ip_vs_control_cleanup(void); 1390 struct ip_vs_dest * 1391 ip_vs_find_dest(struct net *net, int svc_af, int dest_af, 1392 const union nf_inet_addr *daddr, __be16 dport, 1393 const union nf_inet_addr *vaddr, __be16 vport, 1394 __u16 protocol, __u32 fwmark, __u32 flags); 1395 void ip_vs_try_bind_dest(struct ip_vs_conn *cp); 1396 1397 static inline void ip_vs_dest_hold(struct ip_vs_dest *dest) 1398 { 1399 atomic_inc(&dest->refcnt); 1400 } 1401 1402 static inline void ip_vs_dest_put(struct ip_vs_dest *dest) 1403 { 1404 smp_mb__before_atomic(); 1405 atomic_dec(&dest->refcnt); 1406 } 1407 1408 static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest) 1409 { 1410 if (atomic_dec_return(&dest->refcnt) < 0) 1411 kfree(dest); 1412 } 1413 1414 /* IPVS sync daemon data and function prototypes 1415 * (from ip_vs_sync.c) 1416 */ 1417 int start_sync_thread(struct net *net, struct ipvs_sync_daemon_cfg *cfg, 1418 int state); 1419 int stop_sync_thread(struct net *net, int state); 1420 void ip_vs_sync_conn(struct net *net, struct ip_vs_conn *cp, int pkts); 1421 1422 /* IPVS rate estimator prototypes (from ip_vs_est.c) */ 1423 void ip_vs_start_estimator(struct net *net, struct ip_vs_stats *stats); 1424 void ip_vs_stop_estimator(struct net *net, struct ip_vs_stats *stats); 1425 void ip_vs_zero_estimator(struct ip_vs_stats *stats); 1426 void ip_vs_read_estimator(struct ip_vs_kstats *dst, struct ip_vs_stats *stats); 1427 1428 /* Various IPVS packet transmitters (from ip_vs_xmit.c) */ 1429 int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1430 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1431 int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1432 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1433 int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1434 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1435 int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1436 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1437 int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1438 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1439 int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1440 struct ip_vs_protocol *pp, int offset, 1441 unsigned int hooknum, struct ip_vs_iphdr *iph); 1442 void ip_vs_dest_dst_rcu_free(struct rcu_head *head); 1443 1444 #ifdef CONFIG_IP_VS_IPV6 1445 int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1446 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1447 int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1448 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1449 int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1450 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1451 int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1452 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1453 int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1454 struct ip_vs_protocol *pp, int offset, 1455 unsigned int hooknum, struct ip_vs_iphdr *iph); 1456 #endif 1457 1458 #ifdef CONFIG_SYSCTL 1459 /* This is a simple mechanism to ignore packets when 1460 * we are loaded. Just set ip_vs_drop_rate to 'n' and 1461 * we start to drop 1/rate of the packets 1462 */ 1463 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) 1464 { 1465 if (!ipvs->drop_rate) 1466 return 0; 1467 if (--ipvs->drop_counter > 0) 1468 return 0; 1469 ipvs->drop_counter = ipvs->drop_rate; 1470 return 1; 1471 } 1472 #else 1473 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; } 1474 #endif 1475 1476 /* ip_vs_fwd_tag returns the forwarding tag of the connection */ 1477 #define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK) 1478 1479 static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp) 1480 { 1481 char fwd; 1482 1483 switch (IP_VS_FWD_METHOD(cp)) { 1484 case IP_VS_CONN_F_MASQ: 1485 fwd = 'M'; break; 1486 case IP_VS_CONN_F_LOCALNODE: 1487 fwd = 'L'; break; 1488 case IP_VS_CONN_F_TUNNEL: 1489 fwd = 'T'; break; 1490 case IP_VS_CONN_F_DROUTE: 1491 fwd = 'R'; break; 1492 case IP_VS_CONN_F_BYPASS: 1493 fwd = 'B'; break; 1494 default: 1495 fwd = '?'; break; 1496 } 1497 return fwd; 1498 } 1499 1500 void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp, 1501 struct ip_vs_conn *cp, int dir); 1502 1503 #ifdef CONFIG_IP_VS_IPV6 1504 void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp, 1505 struct ip_vs_conn *cp, int dir); 1506 #endif 1507 1508 __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset); 1509 1510 static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum) 1511 { 1512 __be32 diff[2] = { ~old, new }; 1513 1514 return csum_partial(diff, sizeof(diff), oldsum); 1515 } 1516 1517 #ifdef CONFIG_IP_VS_IPV6 1518 static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new, 1519 __wsum oldsum) 1520 { 1521 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0], 1522 new[3], new[2], new[1], new[0] }; 1523 1524 return csum_partial(diff, sizeof(diff), oldsum); 1525 } 1526 #endif 1527 1528 static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum) 1529 { 1530 __be16 diff[2] = { ~old, new }; 1531 1532 return csum_partial(diff, sizeof(diff), oldsum); 1533 } 1534 1535 /* Forget current conntrack (unconfirmed) and attach notrack entry */ 1536 static inline void ip_vs_notrack(struct sk_buff *skb) 1537 { 1538 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 1539 enum ip_conntrack_info ctinfo; 1540 struct nf_conn *ct = nf_ct_get(skb, &ctinfo); 1541 1542 if (!ct || !nf_ct_is_untracked(ct)) { 1543 nf_conntrack_put(skb->nfct); 1544 skb->nfct = &nf_ct_untracked_get()->ct_general; 1545 skb->nfctinfo = IP_CT_NEW; 1546 nf_conntrack_get(skb->nfct); 1547 } 1548 #endif 1549 } 1550 1551 #ifdef CONFIG_IP_VS_NFCT 1552 /* Netfilter connection tracking 1553 * (from ip_vs_nfct.c) 1554 */ 1555 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs) 1556 { 1557 #ifdef CONFIG_SYSCTL 1558 return ipvs->sysctl_conntrack; 1559 #else 1560 return 0; 1561 #endif 1562 } 1563 1564 void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp, 1565 int outin); 1566 int ip_vs_confirm_conntrack(struct sk_buff *skb); 1567 void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct, 1568 struct ip_vs_conn *cp, u_int8_t proto, 1569 const __be16 port, int from_rs); 1570 void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp); 1571 1572 #else 1573 1574 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs) 1575 { 1576 return 0; 1577 } 1578 1579 static inline void ip_vs_update_conntrack(struct sk_buff *skb, 1580 struct ip_vs_conn *cp, int outin) 1581 { 1582 } 1583 1584 static inline int ip_vs_confirm_conntrack(struct sk_buff *skb) 1585 { 1586 return NF_ACCEPT; 1587 } 1588 1589 static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp) 1590 { 1591 } 1592 #endif /* CONFIG_IP_VS_NFCT */ 1593 1594 static inline int 1595 ip_vs_dest_conn_overhead(struct ip_vs_dest *dest) 1596 { 1597 /* We think the overhead of processing active connections is 256 1598 * times higher than that of inactive connections in average. (This 1599 * 256 times might not be accurate, we will change it later) We 1600 * use the following formula to estimate the overhead now: 1601 * dest->activeconns*256 + dest->inactconns 1602 */ 1603 return (atomic_read(&dest->activeconns) << 8) + 1604 atomic_read(&dest->inactconns); 1605 } 1606 1607 #endif /* _NET_IP_VS_H */ 1608