1 /* SPDX-License-Identifier: GPL-2.0 */ 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/refcount.h> /* for struct refcount_t */ 17 18 #include <linux/compiler.h> 19 #include <linux/timer.h> 20 #include <linux/bug.h> 21 22 #include <net/checksum.h> 23 #include <linux/netfilter.h> /* for union nf_inet_addr */ 24 #include <linux/ip.h> 25 #include <linux/ipv6.h> /* for struct ipv6hdr */ 26 #include <net/ipv6.h> 27 #if IS_ENABLED(CONFIG_IP_VS_IPV6) 28 #include <linux/netfilter_ipv6/ip6_tables.h> 29 #endif 30 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 31 #include <net/netfilter/nf_conntrack.h> 32 #endif 33 #include <net/net_namespace.h> /* Netw namespace */ 34 35 #define IP_VS_HDR_INVERSE 1 36 #define IP_VS_HDR_ICMP 2 37 38 /* Generic access of ipvs struct */ 39 static inline struct netns_ipvs *net_ipvs(struct net* net) 40 { 41 return net->ipvs; 42 } 43 44 /* Connections' size value needed by ip_vs_ctl.c */ 45 extern int ip_vs_conn_tab_size; 46 47 struct ip_vs_iphdr { 48 int hdr_flags; /* ipvs flags */ 49 __u32 off; /* Where IP or IPv4 header starts */ 50 __u32 len; /* IPv4 simply where L4 starts 51 * IPv6 where L4 Transport Header starts */ 52 __u16 fragoffs; /* IPv6 fragment offset, 0 if first frag (or not frag)*/ 53 __s16 protocol; 54 __s32 flags; 55 union nf_inet_addr saddr; 56 union nf_inet_addr daddr; 57 }; 58 59 static inline void *frag_safe_skb_hp(const struct sk_buff *skb, int offset, 60 int len, void *buffer) 61 { 62 return skb_header_pointer(skb, offset, len, buffer); 63 } 64 65 /* This function handles filling *ip_vs_iphdr, both for IPv4 and IPv6. 66 * IPv6 requires some extra work, as finding proper header position, 67 * depend on the IPv6 extension headers. 68 */ 69 static inline int 70 ip_vs_fill_iph_skb_off(int af, const struct sk_buff *skb, int offset, 71 int hdr_flags, struct ip_vs_iphdr *iphdr) 72 { 73 iphdr->hdr_flags = hdr_flags; 74 iphdr->off = offset; 75 76 #ifdef CONFIG_IP_VS_IPV6 77 if (af == AF_INET6) { 78 struct ipv6hdr _iph; 79 const struct ipv6hdr *iph = skb_header_pointer( 80 skb, offset, sizeof(_iph), &_iph); 81 if (!iph) 82 return 0; 83 84 iphdr->saddr.in6 = iph->saddr; 85 iphdr->daddr.in6 = iph->daddr; 86 /* ipv6_find_hdr() updates len, flags */ 87 iphdr->len = offset; 88 iphdr->flags = 0; 89 iphdr->protocol = ipv6_find_hdr(skb, &iphdr->len, -1, 90 &iphdr->fragoffs, 91 &iphdr->flags); 92 if (iphdr->protocol < 0) 93 return 0; 94 } else 95 #endif 96 { 97 struct iphdr _iph; 98 const struct iphdr *iph = skb_header_pointer( 99 skb, offset, sizeof(_iph), &_iph); 100 if (!iph) 101 return 0; 102 103 iphdr->len = offset + iph->ihl * 4; 104 iphdr->fragoffs = 0; 105 iphdr->protocol = iph->protocol; 106 iphdr->saddr.ip = iph->saddr; 107 iphdr->daddr.ip = iph->daddr; 108 } 109 110 return 1; 111 } 112 113 static inline int 114 ip_vs_fill_iph_skb_icmp(int af, const struct sk_buff *skb, int offset, 115 bool inverse, struct ip_vs_iphdr *iphdr) 116 { 117 int hdr_flags = IP_VS_HDR_ICMP; 118 119 if (inverse) 120 hdr_flags |= IP_VS_HDR_INVERSE; 121 122 return ip_vs_fill_iph_skb_off(af, skb, offset, hdr_flags, iphdr); 123 } 124 125 static inline int 126 ip_vs_fill_iph_skb(int af, const struct sk_buff *skb, bool inverse, 127 struct ip_vs_iphdr *iphdr) 128 { 129 int hdr_flags = 0; 130 131 if (inverse) 132 hdr_flags |= IP_VS_HDR_INVERSE; 133 134 return ip_vs_fill_iph_skb_off(af, skb, skb_network_offset(skb), 135 hdr_flags, iphdr); 136 } 137 138 static inline bool 139 ip_vs_iph_inverse(const struct ip_vs_iphdr *iph) 140 { 141 return !!(iph->hdr_flags & IP_VS_HDR_INVERSE); 142 } 143 144 static inline bool 145 ip_vs_iph_icmp(const struct ip_vs_iphdr *iph) 146 { 147 return !!(iph->hdr_flags & IP_VS_HDR_ICMP); 148 } 149 150 static inline void ip_vs_addr_copy(int af, union nf_inet_addr *dst, 151 const union nf_inet_addr *src) 152 { 153 #ifdef CONFIG_IP_VS_IPV6 154 if (af == AF_INET6) 155 dst->in6 = src->in6; 156 else 157 #endif 158 dst->ip = src->ip; 159 } 160 161 static inline void ip_vs_addr_set(int af, union nf_inet_addr *dst, 162 const union nf_inet_addr *src) 163 { 164 #ifdef CONFIG_IP_VS_IPV6 165 if (af == AF_INET6) { 166 dst->in6 = src->in6; 167 return; 168 } 169 #endif 170 dst->ip = src->ip; 171 dst->all[1] = 0; 172 dst->all[2] = 0; 173 dst->all[3] = 0; 174 } 175 176 static inline int ip_vs_addr_equal(int af, const union nf_inet_addr *a, 177 const union nf_inet_addr *b) 178 { 179 #ifdef CONFIG_IP_VS_IPV6 180 if (af == AF_INET6) 181 return ipv6_addr_equal(&a->in6, &b->in6); 182 #endif 183 return a->ip == b->ip; 184 } 185 186 #ifdef CONFIG_IP_VS_DEBUG 187 #include <linux/net.h> 188 189 int ip_vs_get_debug_level(void); 190 191 static inline const char *ip_vs_dbg_addr(int af, char *buf, size_t buf_len, 192 const union nf_inet_addr *addr, 193 int *idx) 194 { 195 int len; 196 #ifdef CONFIG_IP_VS_IPV6 197 if (af == AF_INET6) 198 len = snprintf(&buf[*idx], buf_len - *idx, "[%pI6c]", 199 &addr->in6) + 1; 200 else 201 #endif 202 len = snprintf(&buf[*idx], buf_len - *idx, "%pI4", 203 &addr->ip) + 1; 204 205 *idx += len; 206 BUG_ON(*idx > buf_len + 1); 207 return &buf[*idx - len]; 208 } 209 210 #define IP_VS_DBG_BUF(level, msg, ...) \ 211 do { \ 212 char ip_vs_dbg_buf[160]; \ 213 int ip_vs_dbg_idx = 0; \ 214 if (level <= ip_vs_get_debug_level()) \ 215 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \ 216 } while (0) 217 #define IP_VS_ERR_BUF(msg...) \ 218 do { \ 219 char ip_vs_dbg_buf[160]; \ 220 int ip_vs_dbg_idx = 0; \ 221 pr_err(msg); \ 222 } while (0) 223 224 /* Only use from within IP_VS_DBG_BUF() or IP_VS_ERR_BUF macros */ 225 #define IP_VS_DBG_ADDR(af, addr) \ 226 ip_vs_dbg_addr(af, ip_vs_dbg_buf, \ 227 sizeof(ip_vs_dbg_buf), addr, \ 228 &ip_vs_dbg_idx) 229 230 #define IP_VS_DBG(level, msg, ...) \ 231 do { \ 232 if (level <= ip_vs_get_debug_level()) \ 233 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \ 234 } while (0) 235 #define IP_VS_DBG_RL(msg, ...) \ 236 do { \ 237 if (net_ratelimit()) \ 238 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \ 239 } while (0) 240 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) \ 241 do { \ 242 if (level <= ip_vs_get_debug_level()) \ 243 pp->debug_packet(af, pp, skb, ofs, msg); \ 244 } while (0) 245 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) \ 246 do { \ 247 if (level <= ip_vs_get_debug_level() && \ 248 net_ratelimit()) \ 249 pp->debug_packet(af, pp, skb, ofs, msg); \ 250 } while (0) 251 #else /* NO DEBUGGING at ALL */ 252 #define IP_VS_DBG_BUF(level, msg...) do {} while (0) 253 #define IP_VS_ERR_BUF(msg...) do {} while (0) 254 #define IP_VS_DBG(level, msg...) do {} while (0) 255 #define IP_VS_DBG_RL(msg...) do {} while (0) 256 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) do {} while (0) 257 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) do {} while (0) 258 #endif 259 260 #define IP_VS_BUG() BUG() 261 #define IP_VS_ERR_RL(msg, ...) \ 262 do { \ 263 if (net_ratelimit()) \ 264 pr_err(msg, ##__VA_ARGS__); \ 265 } while (0) 266 267 #ifdef CONFIG_IP_VS_DEBUG 268 #define EnterFunction(level) \ 269 do { \ 270 if (level <= ip_vs_get_debug_level()) \ 271 printk(KERN_DEBUG \ 272 pr_fmt("Enter: %s, %s line %i\n"), \ 273 __func__, __FILE__, __LINE__); \ 274 } while (0) 275 #define LeaveFunction(level) \ 276 do { \ 277 if (level <= ip_vs_get_debug_level()) \ 278 printk(KERN_DEBUG \ 279 pr_fmt("Leave: %s, %s line %i\n"), \ 280 __func__, __FILE__, __LINE__); \ 281 } while (0) 282 #else 283 #define EnterFunction(level) do {} while (0) 284 #define LeaveFunction(level) do {} while (0) 285 #endif 286 287 /* The port number of FTP service (in network order). */ 288 #define FTPPORT cpu_to_be16(21) 289 #define FTPDATA cpu_to_be16(20) 290 291 /* TCP State Values */ 292 enum { 293 IP_VS_TCP_S_NONE = 0, 294 IP_VS_TCP_S_ESTABLISHED, 295 IP_VS_TCP_S_SYN_SENT, 296 IP_VS_TCP_S_SYN_RECV, 297 IP_VS_TCP_S_FIN_WAIT, 298 IP_VS_TCP_S_TIME_WAIT, 299 IP_VS_TCP_S_CLOSE, 300 IP_VS_TCP_S_CLOSE_WAIT, 301 IP_VS_TCP_S_LAST_ACK, 302 IP_VS_TCP_S_LISTEN, 303 IP_VS_TCP_S_SYNACK, 304 IP_VS_TCP_S_LAST 305 }; 306 307 /* UDP State Values */ 308 enum { 309 IP_VS_UDP_S_NORMAL, 310 IP_VS_UDP_S_LAST, 311 }; 312 313 /* ICMP State Values */ 314 enum { 315 IP_VS_ICMP_S_NORMAL, 316 IP_VS_ICMP_S_LAST, 317 }; 318 319 /* SCTP State Values */ 320 enum ip_vs_sctp_states { 321 IP_VS_SCTP_S_NONE, 322 IP_VS_SCTP_S_INIT1, 323 IP_VS_SCTP_S_INIT, 324 IP_VS_SCTP_S_COOKIE_SENT, 325 IP_VS_SCTP_S_COOKIE_REPLIED, 326 IP_VS_SCTP_S_COOKIE_WAIT, 327 IP_VS_SCTP_S_COOKIE, 328 IP_VS_SCTP_S_COOKIE_ECHOED, 329 IP_VS_SCTP_S_ESTABLISHED, 330 IP_VS_SCTP_S_SHUTDOWN_SENT, 331 IP_VS_SCTP_S_SHUTDOWN_RECEIVED, 332 IP_VS_SCTP_S_SHUTDOWN_ACK_SENT, 333 IP_VS_SCTP_S_REJECTED, 334 IP_VS_SCTP_S_CLOSED, 335 IP_VS_SCTP_S_LAST 336 }; 337 338 /* Delta sequence info structure 339 * Each ip_vs_conn has 2 (output AND input seq. changes). 340 * Only used in the VS/NAT. 341 */ 342 struct ip_vs_seq { 343 __u32 init_seq; /* Add delta from this seq */ 344 __u32 delta; /* Delta in sequence numbers */ 345 __u32 previous_delta; /* Delta in sequence numbers 346 * before last resized pkt */ 347 }; 348 349 /* counters per cpu */ 350 struct ip_vs_counters { 351 __u64 conns; /* connections scheduled */ 352 __u64 inpkts; /* incoming packets */ 353 __u64 outpkts; /* outgoing packets */ 354 __u64 inbytes; /* incoming bytes */ 355 __u64 outbytes; /* outgoing bytes */ 356 }; 357 /* Stats per cpu */ 358 struct ip_vs_cpu_stats { 359 struct ip_vs_counters cnt; 360 struct u64_stats_sync syncp; 361 }; 362 363 /* IPVS statistics objects */ 364 struct ip_vs_estimator { 365 struct list_head list; 366 367 u64 last_inbytes; 368 u64 last_outbytes; 369 u64 last_conns; 370 u64 last_inpkts; 371 u64 last_outpkts; 372 373 u64 cps; 374 u64 inpps; 375 u64 outpps; 376 u64 inbps; 377 u64 outbps; 378 }; 379 380 /* 381 * IPVS statistics object, 64-bit kernel version of struct ip_vs_stats_user 382 */ 383 struct ip_vs_kstats { 384 u64 conns; /* connections scheduled */ 385 u64 inpkts; /* incoming packets */ 386 u64 outpkts; /* outgoing packets */ 387 u64 inbytes; /* incoming bytes */ 388 u64 outbytes; /* outgoing bytes */ 389 390 u64 cps; /* current connection rate */ 391 u64 inpps; /* current in packet rate */ 392 u64 outpps; /* current out packet rate */ 393 u64 inbps; /* current in byte rate */ 394 u64 outbps; /* current out byte rate */ 395 }; 396 397 struct ip_vs_stats { 398 struct ip_vs_kstats kstats; /* kernel statistics */ 399 struct ip_vs_estimator est; /* estimator */ 400 struct ip_vs_cpu_stats __percpu *cpustats; /* per cpu counters */ 401 spinlock_t lock; /* spin lock */ 402 struct ip_vs_kstats kstats0; /* reset values */ 403 }; 404 405 struct dst_entry; 406 struct iphdr; 407 struct ip_vs_conn; 408 struct ip_vs_app; 409 struct sk_buff; 410 struct ip_vs_proto_data; 411 412 struct ip_vs_protocol { 413 struct ip_vs_protocol *next; 414 char *name; 415 u16 protocol; 416 u16 num_states; 417 int dont_defrag; 418 419 void (*init)(struct ip_vs_protocol *pp); 420 421 void (*exit)(struct ip_vs_protocol *pp); 422 423 int (*init_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd); 424 425 void (*exit_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd); 426 427 int (*conn_schedule)(struct netns_ipvs *ipvs, 428 int af, struct sk_buff *skb, 429 struct ip_vs_proto_data *pd, 430 int *verdict, struct ip_vs_conn **cpp, 431 struct ip_vs_iphdr *iph); 432 433 struct ip_vs_conn * 434 (*conn_in_get)(struct netns_ipvs *ipvs, 435 int af, 436 const struct sk_buff *skb, 437 const struct ip_vs_iphdr *iph); 438 439 struct ip_vs_conn * 440 (*conn_out_get)(struct netns_ipvs *ipvs, 441 int af, 442 const struct sk_buff *skb, 443 const struct ip_vs_iphdr *iph); 444 445 int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp, 446 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph); 447 448 int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp, 449 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph); 450 451 int (*csum_check)(int af, struct sk_buff *skb, 452 struct ip_vs_protocol *pp); 453 454 const char *(*state_name)(int state); 455 456 void (*state_transition)(struct ip_vs_conn *cp, int direction, 457 const struct sk_buff *skb, 458 struct ip_vs_proto_data *pd); 459 460 int (*register_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc); 461 462 void (*unregister_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc); 463 464 int (*app_conn_bind)(struct ip_vs_conn *cp); 465 466 void (*debug_packet)(int af, struct ip_vs_protocol *pp, 467 const struct sk_buff *skb, 468 int offset, 469 const char *msg); 470 471 void (*timeout_change)(struct ip_vs_proto_data *pd, int flags); 472 }; 473 474 /* protocol data per netns */ 475 struct ip_vs_proto_data { 476 struct ip_vs_proto_data *next; 477 struct ip_vs_protocol *pp; 478 int *timeout_table; /* protocol timeout table */ 479 atomic_t appcnt; /* counter of proto app incs. */ 480 struct tcp_states_t *tcp_state_table; 481 }; 482 483 struct ip_vs_protocol *ip_vs_proto_get(unsigned short proto); 484 struct ip_vs_proto_data *ip_vs_proto_data_get(struct netns_ipvs *ipvs, 485 unsigned short proto); 486 487 struct ip_vs_conn_param { 488 struct netns_ipvs *ipvs; 489 const union nf_inet_addr *caddr; 490 const union nf_inet_addr *vaddr; 491 __be16 cport; 492 __be16 vport; 493 __u16 protocol; 494 u16 af; 495 496 const struct ip_vs_pe *pe; 497 char *pe_data; 498 __u8 pe_data_len; 499 }; 500 501 /* IP_VS structure allocated for each dynamically scheduled connection */ 502 struct ip_vs_conn { 503 struct hlist_node c_list; /* hashed list heads */ 504 /* Protocol, addresses and port numbers */ 505 __be16 cport; 506 __be16 dport; 507 __be16 vport; 508 u16 af; /* address family */ 509 union nf_inet_addr caddr; /* client address */ 510 union nf_inet_addr vaddr; /* virtual address */ 511 union nf_inet_addr daddr; /* destination address */ 512 volatile __u32 flags; /* status flags */ 513 __u16 protocol; /* Which protocol (TCP/UDP) */ 514 __u16 daf; /* Address family of the dest */ 515 struct netns_ipvs *ipvs; 516 517 /* counter and timer */ 518 refcount_t refcnt; /* reference count */ 519 struct timer_list timer; /* Expiration timer */ 520 volatile unsigned long timeout; /* timeout */ 521 522 /* Flags and state transition */ 523 spinlock_t lock; /* lock for state transition */ 524 volatile __u16 state; /* state info */ 525 volatile __u16 old_state; /* old state, to be used for 526 * state transition triggerd 527 * synchronization 528 */ 529 __u32 fwmark; /* Fire wall mark from skb */ 530 unsigned long sync_endtime; /* jiffies + sent_retries */ 531 532 /* Control members */ 533 struct ip_vs_conn *control; /* Master control connection */ 534 atomic_t n_control; /* Number of controlled ones */ 535 struct ip_vs_dest *dest; /* real server */ 536 atomic_t in_pkts; /* incoming packet counter */ 537 538 /* Packet transmitter for different forwarding methods. If it 539 * mangles the packet, it must return NF_DROP or better NF_STOLEN, 540 * otherwise this must be changed to a sk_buff **. 541 * NF_ACCEPT can be returned when destination is local. 542 */ 543 int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp, 544 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 545 546 /* Note: we can group the following members into a structure, 547 * in order to save more space, and the following members are 548 * only used in VS/NAT anyway 549 */ 550 struct ip_vs_app *app; /* bound ip_vs_app object */ 551 void *app_data; /* Application private data */ 552 struct ip_vs_seq in_seq; /* incoming seq. struct */ 553 struct ip_vs_seq out_seq; /* outgoing seq. struct */ 554 555 const struct ip_vs_pe *pe; 556 char *pe_data; 557 __u8 pe_data_len; 558 559 struct rcu_head rcu_head; 560 }; 561 562 /* Extended internal versions of struct ip_vs_service_user and ip_vs_dest_user 563 * for IPv6 support. 564 * 565 * We need these to conveniently pass around service and destination 566 * options, but unfortunately, we also need to keep the old definitions to 567 * maintain userspace backwards compatibility for the setsockopt interface. 568 */ 569 struct ip_vs_service_user_kern { 570 /* virtual service addresses */ 571 u16 af; 572 u16 protocol; 573 union nf_inet_addr addr; /* virtual ip address */ 574 __be16 port; 575 u32 fwmark; /* firwall mark of service */ 576 577 /* virtual service options */ 578 char *sched_name; 579 char *pe_name; 580 unsigned int flags; /* virtual service flags */ 581 unsigned int timeout; /* persistent timeout in sec */ 582 __be32 netmask; /* persistent netmask or plen */ 583 }; 584 585 586 struct ip_vs_dest_user_kern { 587 /* destination server address */ 588 union nf_inet_addr addr; 589 __be16 port; 590 591 /* real server options */ 592 unsigned int conn_flags; /* connection flags */ 593 int weight; /* destination weight */ 594 595 /* thresholds for active connections */ 596 u32 u_threshold; /* upper threshold */ 597 u32 l_threshold; /* lower threshold */ 598 599 /* Address family of addr */ 600 u16 af; 601 }; 602 603 604 /* 605 * The information about the virtual service offered to the net and the 606 * forwarding entries. 607 */ 608 struct ip_vs_service { 609 struct hlist_node s_list; /* for normal service table */ 610 struct hlist_node f_list; /* for fwmark-based service table */ 611 atomic_t refcnt; /* reference counter */ 612 613 u16 af; /* address family */ 614 __u16 protocol; /* which protocol (TCP/UDP) */ 615 union nf_inet_addr addr; /* IP address for virtual service */ 616 __be16 port; /* port number for the service */ 617 __u32 fwmark; /* firewall mark of the service */ 618 unsigned int flags; /* service status flags */ 619 unsigned int timeout; /* persistent timeout in ticks */ 620 __be32 netmask; /* grouping granularity, mask/plen */ 621 struct netns_ipvs *ipvs; 622 623 struct list_head destinations; /* real server d-linked list */ 624 __u32 num_dests; /* number of servers */ 625 struct ip_vs_stats stats; /* statistics for the service */ 626 627 /* for scheduling */ 628 struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */ 629 spinlock_t sched_lock; /* lock sched_data */ 630 void *sched_data; /* scheduler application data */ 631 632 /* alternate persistence engine */ 633 struct ip_vs_pe __rcu *pe; 634 635 struct rcu_head rcu_head; 636 }; 637 638 /* Information for cached dst */ 639 struct ip_vs_dest_dst { 640 struct dst_entry *dst_cache; /* destination cache entry */ 641 u32 dst_cookie; 642 union nf_inet_addr dst_saddr; 643 struct rcu_head rcu_head; 644 }; 645 646 /* The real server destination forwarding entry with ip address, port number, 647 * and so on. 648 */ 649 struct ip_vs_dest { 650 struct list_head n_list; /* for the dests in the service */ 651 struct hlist_node d_list; /* for table with all the dests */ 652 653 u16 af; /* address family */ 654 __be16 port; /* port number of the server */ 655 union nf_inet_addr addr; /* IP address of the server */ 656 volatile unsigned int flags; /* dest status flags */ 657 atomic_t conn_flags; /* flags to copy to conn */ 658 atomic_t weight; /* server weight */ 659 atomic_t last_weight; /* server latest weight */ 660 661 refcount_t refcnt; /* reference counter */ 662 struct ip_vs_stats stats; /* statistics */ 663 unsigned long idle_start; /* start time, jiffies */ 664 665 /* connection counters and thresholds */ 666 atomic_t activeconns; /* active connections */ 667 atomic_t inactconns; /* inactive connections */ 668 atomic_t persistconns; /* persistent connections */ 669 __u32 u_threshold; /* upper threshold */ 670 __u32 l_threshold; /* lower threshold */ 671 672 /* for destination cache */ 673 spinlock_t dst_lock; /* lock of dst_cache */ 674 struct ip_vs_dest_dst __rcu *dest_dst; /* cached dst info */ 675 676 /* for virtual service */ 677 struct ip_vs_service __rcu *svc; /* service it belongs to */ 678 __u16 protocol; /* which protocol (TCP/UDP) */ 679 __be16 vport; /* virtual port number */ 680 union nf_inet_addr vaddr; /* virtual IP address */ 681 __u32 vfwmark; /* firewall mark of service */ 682 683 struct list_head t_list; /* in dest_trash */ 684 unsigned int in_rs_table:1; /* we are in rs_table */ 685 }; 686 687 /* The scheduler object */ 688 struct ip_vs_scheduler { 689 struct list_head n_list; /* d-linked list head */ 690 char *name; /* scheduler name */ 691 atomic_t refcnt; /* reference counter */ 692 struct module *module; /* THIS_MODULE/NULL */ 693 694 /* scheduler initializing service */ 695 int (*init_service)(struct ip_vs_service *svc); 696 /* scheduling service finish */ 697 void (*done_service)(struct ip_vs_service *svc); 698 /* dest is linked */ 699 int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest); 700 /* dest is unlinked */ 701 int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest); 702 /* dest is updated */ 703 int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest); 704 705 /* selecting a server from the given service */ 706 struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc, 707 const struct sk_buff *skb, 708 struct ip_vs_iphdr *iph); 709 }; 710 711 /* The persistence engine object */ 712 struct ip_vs_pe { 713 struct list_head n_list; /* d-linked list head */ 714 char *name; /* scheduler name */ 715 atomic_t refcnt; /* reference counter */ 716 struct module *module; /* THIS_MODULE/NULL */ 717 718 /* get the connection template, if any */ 719 int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb); 720 bool (*ct_match)(const struct ip_vs_conn_param *p, 721 struct ip_vs_conn *ct); 722 u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval, 723 bool inverse); 724 int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf); 725 /* create connections for real-server outgoing packets */ 726 struct ip_vs_conn* (*conn_out)(struct ip_vs_service *svc, 727 struct ip_vs_dest *dest, 728 struct sk_buff *skb, 729 const struct ip_vs_iphdr *iph, 730 __be16 dport, __be16 cport); 731 }; 732 733 /* The application module object (a.k.a. app incarnation) */ 734 struct ip_vs_app { 735 struct list_head a_list; /* member in app list */ 736 int type; /* IP_VS_APP_TYPE_xxx */ 737 char *name; /* application module name */ 738 __u16 protocol; 739 struct module *module; /* THIS_MODULE/NULL */ 740 struct list_head incs_list; /* list of incarnations */ 741 742 /* members for application incarnations */ 743 struct list_head p_list; /* member in proto app list */ 744 struct ip_vs_app *app; /* its real application */ 745 __be16 port; /* port number in net order */ 746 atomic_t usecnt; /* usage counter */ 747 struct rcu_head rcu_head; 748 749 /* output hook: Process packet in inout direction, diff set for TCP. 750 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok, 751 * 2=Mangled but checksum was not updated 752 */ 753 int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *, 754 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh); 755 756 /* input hook: Process packet in outin direction, diff set for TCP. 757 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok, 758 * 2=Mangled but checksum was not updated 759 */ 760 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *, 761 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh); 762 763 /* ip_vs_app initializer */ 764 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *); 765 766 /* ip_vs_app finish */ 767 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *); 768 769 770 /* not used now */ 771 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *, 772 struct ip_vs_protocol *); 773 774 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *); 775 776 int * timeout_table; 777 int * timeouts; 778 int timeouts_size; 779 780 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app, 781 int *verdict, struct ip_vs_conn **cpp); 782 783 struct ip_vs_conn * 784 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app, 785 const struct iphdr *iph, int inverse); 786 787 struct ip_vs_conn * 788 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app, 789 const struct iphdr *iph, int inverse); 790 791 int (*state_transition)(struct ip_vs_conn *cp, int direction, 792 const struct sk_buff *skb, 793 struct ip_vs_app *app); 794 795 void (*timeout_change)(struct ip_vs_app *app, int flags); 796 }; 797 798 struct ipvs_master_sync_state { 799 struct list_head sync_queue; 800 struct ip_vs_sync_buff *sync_buff; 801 unsigned long sync_queue_len; 802 unsigned int sync_queue_delay; 803 struct task_struct *master_thread; 804 struct delayed_work master_wakeup_work; 805 struct netns_ipvs *ipvs; 806 }; 807 808 /* How much time to keep dests in trash */ 809 #define IP_VS_DEST_TRASH_PERIOD (120 * HZ) 810 811 struct ipvs_sync_daemon_cfg { 812 union nf_inet_addr mcast_group; 813 int syncid; 814 u16 sync_maxlen; 815 u16 mcast_port; 816 u8 mcast_af; 817 u8 mcast_ttl; 818 /* multicast interface name */ 819 char mcast_ifn[IP_VS_IFNAME_MAXLEN]; 820 }; 821 822 /* IPVS in network namespace */ 823 struct netns_ipvs { 824 int gen; /* Generation */ 825 int enable; /* enable like nf_hooks do */ 826 /* Hash table: for real service lookups */ 827 #define IP_VS_RTAB_BITS 4 828 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS) 829 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1) 830 831 struct hlist_head rs_table[IP_VS_RTAB_SIZE]; 832 /* ip_vs_app */ 833 struct list_head app_list; 834 /* ip_vs_proto */ 835 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */ 836 struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE]; 837 /* ip_vs_proto_tcp */ 838 #ifdef CONFIG_IP_VS_PROTO_TCP 839 #define TCP_APP_TAB_BITS 4 840 #define TCP_APP_TAB_SIZE (1 << TCP_APP_TAB_BITS) 841 #define TCP_APP_TAB_MASK (TCP_APP_TAB_SIZE - 1) 842 struct list_head tcp_apps[TCP_APP_TAB_SIZE]; 843 #endif 844 /* ip_vs_proto_udp */ 845 #ifdef CONFIG_IP_VS_PROTO_UDP 846 #define UDP_APP_TAB_BITS 4 847 #define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS) 848 #define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1) 849 struct list_head udp_apps[UDP_APP_TAB_SIZE]; 850 #endif 851 /* ip_vs_proto_sctp */ 852 #ifdef CONFIG_IP_VS_PROTO_SCTP 853 #define SCTP_APP_TAB_BITS 4 854 #define SCTP_APP_TAB_SIZE (1 << SCTP_APP_TAB_BITS) 855 #define SCTP_APP_TAB_MASK (SCTP_APP_TAB_SIZE - 1) 856 /* Hash table for SCTP application incarnations */ 857 struct list_head sctp_apps[SCTP_APP_TAB_SIZE]; 858 #endif 859 /* ip_vs_conn */ 860 atomic_t conn_count; /* connection counter */ 861 862 /* ip_vs_ctl */ 863 struct ip_vs_stats tot_stats; /* Statistics & est. */ 864 865 int num_services; /* no of virtual services */ 866 867 /* Trash for destinations */ 868 struct list_head dest_trash; 869 spinlock_t dest_trash_lock; 870 struct timer_list dest_trash_timer; /* expiration timer */ 871 /* Service counters */ 872 atomic_t ftpsvc_counter; 873 atomic_t nullsvc_counter; 874 atomic_t conn_out_counter; 875 876 #ifdef CONFIG_SYSCTL 877 /* 1/rate drop and drop-entry variables */ 878 struct delayed_work defense_work; /* Work handler */ 879 int drop_rate; 880 int drop_counter; 881 atomic_t dropentry; 882 /* locks in ctl.c */ 883 spinlock_t dropentry_lock; /* drop entry handling */ 884 spinlock_t droppacket_lock; /* drop packet handling */ 885 spinlock_t securetcp_lock; /* state and timeout tables */ 886 887 /* sys-ctl struct */ 888 struct ctl_table_header *sysctl_hdr; 889 struct ctl_table *sysctl_tbl; 890 #endif 891 892 /* sysctl variables */ 893 int sysctl_amemthresh; 894 int sysctl_am_droprate; 895 int sysctl_drop_entry; 896 int sysctl_drop_packet; 897 int sysctl_secure_tcp; 898 #ifdef CONFIG_IP_VS_NFCT 899 int sysctl_conntrack; 900 #endif 901 int sysctl_snat_reroute; 902 int sysctl_sync_ver; 903 int sysctl_sync_ports; 904 int sysctl_sync_persist_mode; 905 unsigned long sysctl_sync_qlen_max; 906 int sysctl_sync_sock_size; 907 int sysctl_cache_bypass; 908 int sysctl_expire_nodest_conn; 909 int sysctl_sloppy_tcp; 910 int sysctl_sloppy_sctp; 911 int sysctl_expire_quiescent_template; 912 int sysctl_sync_threshold[2]; 913 unsigned int sysctl_sync_refresh_period; 914 int sysctl_sync_retries; 915 int sysctl_nat_icmp_send; 916 int sysctl_pmtu_disc; 917 int sysctl_backup_only; 918 int sysctl_conn_reuse_mode; 919 int sysctl_schedule_icmp; 920 int sysctl_ignore_tunneled; 921 922 /* ip_vs_lblc */ 923 int sysctl_lblc_expiration; 924 struct ctl_table_header *lblc_ctl_header; 925 struct ctl_table *lblc_ctl_table; 926 /* ip_vs_lblcr */ 927 int sysctl_lblcr_expiration; 928 struct ctl_table_header *lblcr_ctl_header; 929 struct ctl_table *lblcr_ctl_table; 930 /* ip_vs_est */ 931 struct list_head est_list; /* estimator list */ 932 spinlock_t est_lock; 933 struct timer_list est_timer; /* Estimation timer */ 934 /* ip_vs_sync */ 935 spinlock_t sync_lock; 936 struct ipvs_master_sync_state *ms; 937 spinlock_t sync_buff_lock; 938 struct task_struct **backup_threads; 939 int threads_mask; 940 volatile int sync_state; 941 struct mutex sync_mutex; 942 struct ipvs_sync_daemon_cfg mcfg; /* Master Configuration */ 943 struct ipvs_sync_daemon_cfg bcfg; /* Backup Configuration */ 944 /* net name space ptr */ 945 struct net *net; /* Needed by timer routines */ 946 /* Number of heterogeneous destinations, needed becaus heterogeneous 947 * are not supported when synchronization is enabled. 948 */ 949 unsigned int mixed_address_family_dests; 950 }; 951 952 #define DEFAULT_SYNC_THRESHOLD 3 953 #define DEFAULT_SYNC_PERIOD 50 954 #define DEFAULT_SYNC_VER 1 955 #define DEFAULT_SLOPPY_TCP 0 956 #define DEFAULT_SLOPPY_SCTP 0 957 #define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ) 958 #define DEFAULT_SYNC_RETRIES 0 959 #define IPVS_SYNC_WAKEUP_RATE 8 960 #define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4) 961 #define IPVS_SYNC_SEND_DELAY (HZ / 50) 962 #define IPVS_SYNC_CHECK_PERIOD HZ 963 #define IPVS_SYNC_FLUSH_TIME (HZ * 2) 964 #define IPVS_SYNC_PORTS_MAX (1 << 6) 965 966 #ifdef CONFIG_SYSCTL 967 968 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs) 969 { 970 return ipvs->sysctl_sync_threshold[0]; 971 } 972 973 static inline int sysctl_sync_period(struct netns_ipvs *ipvs) 974 { 975 return READ_ONCE(ipvs->sysctl_sync_threshold[1]); 976 } 977 978 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs) 979 { 980 return READ_ONCE(ipvs->sysctl_sync_refresh_period); 981 } 982 983 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs) 984 { 985 return ipvs->sysctl_sync_retries; 986 } 987 988 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs) 989 { 990 return ipvs->sysctl_sync_ver; 991 } 992 993 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs) 994 { 995 return ipvs->sysctl_sloppy_tcp; 996 } 997 998 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs) 999 { 1000 return ipvs->sysctl_sloppy_sctp; 1001 } 1002 1003 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs) 1004 { 1005 return READ_ONCE(ipvs->sysctl_sync_ports); 1006 } 1007 1008 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs) 1009 { 1010 return ipvs->sysctl_sync_persist_mode; 1011 } 1012 1013 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs) 1014 { 1015 return ipvs->sysctl_sync_qlen_max; 1016 } 1017 1018 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs) 1019 { 1020 return ipvs->sysctl_sync_sock_size; 1021 } 1022 1023 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs) 1024 { 1025 return ipvs->sysctl_pmtu_disc; 1026 } 1027 1028 static inline int sysctl_backup_only(struct netns_ipvs *ipvs) 1029 { 1030 return ipvs->sync_state & IP_VS_STATE_BACKUP && 1031 ipvs->sysctl_backup_only; 1032 } 1033 1034 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs) 1035 { 1036 return ipvs->sysctl_conn_reuse_mode; 1037 } 1038 1039 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs) 1040 { 1041 return ipvs->sysctl_schedule_icmp; 1042 } 1043 1044 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs) 1045 { 1046 return ipvs->sysctl_ignore_tunneled; 1047 } 1048 1049 static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs) 1050 { 1051 return ipvs->sysctl_cache_bypass; 1052 } 1053 1054 #else 1055 1056 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs) 1057 { 1058 return DEFAULT_SYNC_THRESHOLD; 1059 } 1060 1061 static inline int sysctl_sync_period(struct netns_ipvs *ipvs) 1062 { 1063 return DEFAULT_SYNC_PERIOD; 1064 } 1065 1066 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs) 1067 { 1068 return DEFAULT_SYNC_REFRESH_PERIOD; 1069 } 1070 1071 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs) 1072 { 1073 return DEFAULT_SYNC_RETRIES & 3; 1074 } 1075 1076 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs) 1077 { 1078 return DEFAULT_SYNC_VER; 1079 } 1080 1081 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs) 1082 { 1083 return DEFAULT_SLOPPY_TCP; 1084 } 1085 1086 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs) 1087 { 1088 return DEFAULT_SLOPPY_SCTP; 1089 } 1090 1091 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs) 1092 { 1093 return 1; 1094 } 1095 1096 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs) 1097 { 1098 return 0; 1099 } 1100 1101 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs) 1102 { 1103 return IPVS_SYNC_QLEN_MAX; 1104 } 1105 1106 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs) 1107 { 1108 return 0; 1109 } 1110 1111 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs) 1112 { 1113 return 1; 1114 } 1115 1116 static inline int sysctl_backup_only(struct netns_ipvs *ipvs) 1117 { 1118 return 0; 1119 } 1120 1121 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs) 1122 { 1123 return 1; 1124 } 1125 1126 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs) 1127 { 1128 return 0; 1129 } 1130 1131 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs) 1132 { 1133 return 0; 1134 } 1135 1136 static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs) 1137 { 1138 return 0; 1139 } 1140 1141 #endif 1142 1143 /* IPVS core functions 1144 * (from ip_vs_core.c) 1145 */ 1146 const char *ip_vs_proto_name(unsigned int proto); 1147 void ip_vs_init_hash_table(struct list_head *table, int rows); 1148 struct ip_vs_conn *ip_vs_new_conn_out(struct ip_vs_service *svc, 1149 struct ip_vs_dest *dest, 1150 struct sk_buff *skb, 1151 const struct ip_vs_iphdr *iph, 1152 __be16 dport, 1153 __be16 cport); 1154 #define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t))) 1155 1156 #define IP_VS_APP_TYPE_FTP 1 1157 1158 /* ip_vs_conn handling functions 1159 * (from ip_vs_conn.c) 1160 */ 1161 enum { 1162 IP_VS_DIR_INPUT = 0, 1163 IP_VS_DIR_OUTPUT, 1164 IP_VS_DIR_INPUT_ONLY, 1165 IP_VS_DIR_LAST, 1166 }; 1167 1168 static inline void ip_vs_conn_fill_param(struct netns_ipvs *ipvs, int af, int protocol, 1169 const union nf_inet_addr *caddr, 1170 __be16 cport, 1171 const union nf_inet_addr *vaddr, 1172 __be16 vport, 1173 struct ip_vs_conn_param *p) 1174 { 1175 p->ipvs = ipvs; 1176 p->af = af; 1177 p->protocol = protocol; 1178 p->caddr = caddr; 1179 p->cport = cport; 1180 p->vaddr = vaddr; 1181 p->vport = vport; 1182 p->pe = NULL; 1183 p->pe_data = NULL; 1184 } 1185 1186 struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p); 1187 struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p); 1188 1189 struct ip_vs_conn * ip_vs_conn_in_get_proto(struct netns_ipvs *ipvs, int af, 1190 const struct sk_buff *skb, 1191 const struct ip_vs_iphdr *iph); 1192 1193 struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p); 1194 1195 struct ip_vs_conn * ip_vs_conn_out_get_proto(struct netns_ipvs *ipvs, int af, 1196 const struct sk_buff *skb, 1197 const struct ip_vs_iphdr *iph); 1198 1199 /* Get reference to gain full access to conn. 1200 * By default, RCU read-side critical sections have access only to 1201 * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference. 1202 */ 1203 static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp) 1204 { 1205 return refcount_inc_not_zero(&cp->refcnt); 1206 } 1207 1208 /* put back the conn without restarting its timer */ 1209 static inline void __ip_vs_conn_put(struct ip_vs_conn *cp) 1210 { 1211 smp_mb__before_atomic(); 1212 refcount_dec(&cp->refcnt); 1213 } 1214 void ip_vs_conn_put(struct ip_vs_conn *cp); 1215 void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport); 1216 1217 struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, int dest_af, 1218 const union nf_inet_addr *daddr, 1219 __be16 dport, unsigned int flags, 1220 struct ip_vs_dest *dest, __u32 fwmark); 1221 void ip_vs_conn_expire_now(struct ip_vs_conn *cp); 1222 1223 const char *ip_vs_state_name(__u16 proto, int state); 1224 1225 void ip_vs_tcp_conn_listen(struct ip_vs_conn *cp); 1226 int ip_vs_check_template(struct ip_vs_conn *ct, struct ip_vs_dest *cdest); 1227 void ip_vs_random_dropentry(struct netns_ipvs *ipvs); 1228 int ip_vs_conn_init(void); 1229 void ip_vs_conn_cleanup(void); 1230 1231 static inline void ip_vs_control_del(struct ip_vs_conn *cp) 1232 { 1233 struct ip_vs_conn *ctl_cp = cp->control; 1234 if (!ctl_cp) { 1235 IP_VS_ERR_BUF("request control DEL for uncontrolled: " 1236 "%s:%d to %s:%d\n", 1237 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1238 ntohs(cp->cport), 1239 IP_VS_DBG_ADDR(cp->af, &cp->vaddr), 1240 ntohs(cp->vport)); 1241 1242 return; 1243 } 1244 1245 IP_VS_DBG_BUF(7, "DELeting control for: " 1246 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n", 1247 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1248 ntohs(cp->cport), 1249 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr), 1250 ntohs(ctl_cp->cport)); 1251 1252 cp->control = NULL; 1253 if (atomic_read(&ctl_cp->n_control) == 0) { 1254 IP_VS_ERR_BUF("BUG control DEL with n=0 : " 1255 "%s:%d to %s:%d\n", 1256 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1257 ntohs(cp->cport), 1258 IP_VS_DBG_ADDR(cp->af, &cp->vaddr), 1259 ntohs(cp->vport)); 1260 1261 return; 1262 } 1263 atomic_dec(&ctl_cp->n_control); 1264 } 1265 1266 static inline void 1267 ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp) 1268 { 1269 if (cp->control) { 1270 IP_VS_ERR_BUF("request control ADD for already controlled: " 1271 "%s:%d to %s:%d\n", 1272 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1273 ntohs(cp->cport), 1274 IP_VS_DBG_ADDR(cp->af, &cp->vaddr), 1275 ntohs(cp->vport)); 1276 1277 ip_vs_control_del(cp); 1278 } 1279 1280 IP_VS_DBG_BUF(7, "ADDing control for: " 1281 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n", 1282 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1283 ntohs(cp->cport), 1284 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr), 1285 ntohs(ctl_cp->cport)); 1286 1287 cp->control = ctl_cp; 1288 atomic_inc(&ctl_cp->n_control); 1289 } 1290 1291 /* IPVS netns init & cleanup functions */ 1292 int ip_vs_estimator_net_init(struct netns_ipvs *ipvs); 1293 int ip_vs_control_net_init(struct netns_ipvs *ipvs); 1294 int ip_vs_protocol_net_init(struct netns_ipvs *ipvs); 1295 int ip_vs_app_net_init(struct netns_ipvs *ipvs); 1296 int ip_vs_conn_net_init(struct netns_ipvs *ipvs); 1297 int ip_vs_sync_net_init(struct netns_ipvs *ipvs); 1298 void ip_vs_conn_net_cleanup(struct netns_ipvs *ipvs); 1299 void ip_vs_app_net_cleanup(struct netns_ipvs *ipvs); 1300 void ip_vs_protocol_net_cleanup(struct netns_ipvs *ipvs); 1301 void ip_vs_control_net_cleanup(struct netns_ipvs *ipvs); 1302 void ip_vs_estimator_net_cleanup(struct netns_ipvs *ipvs); 1303 void ip_vs_sync_net_cleanup(struct netns_ipvs *ipvs); 1304 void ip_vs_service_net_cleanup(struct netns_ipvs *ipvs); 1305 1306 /* IPVS application functions 1307 * (from ip_vs_app.c) 1308 */ 1309 #define IP_VS_APP_MAX_PORTS 8 1310 struct ip_vs_app *register_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app); 1311 void unregister_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app); 1312 int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp); 1313 void ip_vs_unbind_app(struct ip_vs_conn *cp); 1314 int register_ip_vs_app_inc(struct netns_ipvs *ipvs, struct ip_vs_app *app, __u16 proto, 1315 __u16 port); 1316 int ip_vs_app_inc_get(struct ip_vs_app *inc); 1317 void ip_vs_app_inc_put(struct ip_vs_app *inc); 1318 1319 int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb, 1320 struct ip_vs_iphdr *ipvsh); 1321 int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb, 1322 struct ip_vs_iphdr *ipvsh); 1323 1324 int register_ip_vs_pe(struct ip_vs_pe *pe); 1325 int unregister_ip_vs_pe(struct ip_vs_pe *pe); 1326 struct ip_vs_pe *ip_vs_pe_getbyname(const char *name); 1327 struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name); 1328 1329 /* Use a #define to avoid all of module.h just for these trivial ops */ 1330 #define ip_vs_pe_get(pe) \ 1331 if (pe && pe->module) \ 1332 __module_get(pe->module); 1333 1334 #define ip_vs_pe_put(pe) \ 1335 if (pe && pe->module) \ 1336 module_put(pe->module); 1337 1338 /* IPVS protocol functions (from ip_vs_proto.c) */ 1339 int ip_vs_protocol_init(void); 1340 void ip_vs_protocol_cleanup(void); 1341 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags); 1342 int *ip_vs_create_timeout_table(int *table, int size); 1343 void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp, 1344 const struct sk_buff *skb, int offset, 1345 const char *msg); 1346 1347 extern struct ip_vs_protocol ip_vs_protocol_tcp; 1348 extern struct ip_vs_protocol ip_vs_protocol_udp; 1349 extern struct ip_vs_protocol ip_vs_protocol_icmp; 1350 extern struct ip_vs_protocol ip_vs_protocol_esp; 1351 extern struct ip_vs_protocol ip_vs_protocol_ah; 1352 extern struct ip_vs_protocol ip_vs_protocol_sctp; 1353 1354 /* Registering/unregistering scheduler functions 1355 * (from ip_vs_sched.c) 1356 */ 1357 int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler); 1358 int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler); 1359 int ip_vs_bind_scheduler(struct ip_vs_service *svc, 1360 struct ip_vs_scheduler *scheduler); 1361 void ip_vs_unbind_scheduler(struct ip_vs_service *svc, 1362 struct ip_vs_scheduler *sched); 1363 struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name); 1364 void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler); 1365 struct ip_vs_conn * 1366 ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb, 1367 struct ip_vs_proto_data *pd, int *ignored, 1368 struct ip_vs_iphdr *iph); 1369 int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb, 1370 struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph); 1371 1372 void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg); 1373 1374 /* IPVS control data and functions (from ip_vs_ctl.c) */ 1375 extern struct ip_vs_stats ip_vs_stats; 1376 extern int sysctl_ip_vs_sync_ver; 1377 1378 struct ip_vs_service * 1379 ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol, 1380 const union nf_inet_addr *vaddr, __be16 vport); 1381 1382 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol, 1383 const union nf_inet_addr *daddr, __be16 dport); 1384 1385 struct ip_vs_dest * 1386 ip_vs_find_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol, 1387 const union nf_inet_addr *daddr, __be16 dport); 1388 1389 int ip_vs_use_count_inc(void); 1390 void ip_vs_use_count_dec(void); 1391 int ip_vs_register_nl_ioctl(void); 1392 void ip_vs_unregister_nl_ioctl(void); 1393 int ip_vs_control_init(void); 1394 void ip_vs_control_cleanup(void); 1395 struct ip_vs_dest * 1396 ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af, 1397 const union nf_inet_addr *daddr, __be16 dport, 1398 const union nf_inet_addr *vaddr, __be16 vport, 1399 __u16 protocol, __u32 fwmark, __u32 flags); 1400 void ip_vs_try_bind_dest(struct ip_vs_conn *cp); 1401 1402 static inline void ip_vs_dest_hold(struct ip_vs_dest *dest) 1403 { 1404 refcount_inc(&dest->refcnt); 1405 } 1406 1407 static inline void ip_vs_dest_put(struct ip_vs_dest *dest) 1408 { 1409 smp_mb__before_atomic(); 1410 refcount_dec(&dest->refcnt); 1411 } 1412 1413 static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest) 1414 { 1415 if (refcount_dec_and_test(&dest->refcnt)) 1416 kfree(dest); 1417 } 1418 1419 /* IPVS sync daemon data and function prototypes 1420 * (from ip_vs_sync.c) 1421 */ 1422 int start_sync_thread(struct netns_ipvs *ipvs, struct ipvs_sync_daemon_cfg *cfg, 1423 int state); 1424 int stop_sync_thread(struct netns_ipvs *ipvs, int state); 1425 void ip_vs_sync_conn(struct netns_ipvs *ipvs, struct ip_vs_conn *cp, int pkts); 1426 1427 /* IPVS rate estimator prototypes (from ip_vs_est.c) */ 1428 void ip_vs_start_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats); 1429 void ip_vs_stop_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats); 1430 void ip_vs_zero_estimator(struct ip_vs_stats *stats); 1431 void ip_vs_read_estimator(struct ip_vs_kstats *dst, struct ip_vs_stats *stats); 1432 1433 /* Various IPVS packet transmitters (from ip_vs_xmit.c) */ 1434 int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1435 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1436 int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1437 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1438 int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1439 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1440 int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1441 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1442 int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1443 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1444 int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1445 struct ip_vs_protocol *pp, int offset, 1446 unsigned int hooknum, struct ip_vs_iphdr *iph); 1447 void ip_vs_dest_dst_rcu_free(struct rcu_head *head); 1448 1449 #ifdef CONFIG_IP_VS_IPV6 1450 int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1451 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1452 int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1453 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1454 int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1455 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1456 int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1457 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1458 int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1459 struct ip_vs_protocol *pp, int offset, 1460 unsigned int hooknum, struct ip_vs_iphdr *iph); 1461 #endif 1462 1463 #ifdef CONFIG_SYSCTL 1464 /* This is a simple mechanism to ignore packets when 1465 * we are loaded. Just set ip_vs_drop_rate to 'n' and 1466 * we start to drop 1/rate of the packets 1467 */ 1468 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) 1469 { 1470 if (!ipvs->drop_rate) 1471 return 0; 1472 if (--ipvs->drop_counter > 0) 1473 return 0; 1474 ipvs->drop_counter = ipvs->drop_rate; 1475 return 1; 1476 } 1477 #else 1478 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; } 1479 #endif 1480 1481 /* ip_vs_fwd_tag returns the forwarding tag of the connection */ 1482 #define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK) 1483 1484 static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp) 1485 { 1486 char fwd; 1487 1488 switch (IP_VS_FWD_METHOD(cp)) { 1489 case IP_VS_CONN_F_MASQ: 1490 fwd = 'M'; break; 1491 case IP_VS_CONN_F_LOCALNODE: 1492 fwd = 'L'; break; 1493 case IP_VS_CONN_F_TUNNEL: 1494 fwd = 'T'; break; 1495 case IP_VS_CONN_F_DROUTE: 1496 fwd = 'R'; break; 1497 case IP_VS_CONN_F_BYPASS: 1498 fwd = 'B'; break; 1499 default: 1500 fwd = '?'; break; 1501 } 1502 return fwd; 1503 } 1504 1505 void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp, 1506 struct ip_vs_conn *cp, int dir); 1507 1508 #ifdef CONFIG_IP_VS_IPV6 1509 void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp, 1510 struct ip_vs_conn *cp, int dir); 1511 #endif 1512 1513 __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset); 1514 1515 static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum) 1516 { 1517 __be32 diff[2] = { ~old, new }; 1518 1519 return csum_partial(diff, sizeof(diff), oldsum); 1520 } 1521 1522 #ifdef CONFIG_IP_VS_IPV6 1523 static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new, 1524 __wsum oldsum) 1525 { 1526 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0], 1527 new[3], new[2], new[1], new[0] }; 1528 1529 return csum_partial(diff, sizeof(diff), oldsum); 1530 } 1531 #endif 1532 1533 static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum) 1534 { 1535 __be16 diff[2] = { ~old, new }; 1536 1537 return csum_partial(diff, sizeof(diff), oldsum); 1538 } 1539 1540 /* Forget current conntrack (unconfirmed) and attach notrack entry */ 1541 static inline void ip_vs_notrack(struct sk_buff *skb) 1542 { 1543 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 1544 enum ip_conntrack_info ctinfo; 1545 struct nf_conn *ct = nf_ct_get(skb, &ctinfo); 1546 1547 if (ct) { 1548 nf_conntrack_put(&ct->ct_general); 1549 nf_ct_set(skb, NULL, IP_CT_UNTRACKED); 1550 } 1551 #endif 1552 } 1553 1554 #ifdef CONFIG_IP_VS_NFCT 1555 /* Netfilter connection tracking 1556 * (from ip_vs_nfct.c) 1557 */ 1558 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs) 1559 { 1560 #ifdef CONFIG_SYSCTL 1561 return ipvs->sysctl_conntrack; 1562 #else 1563 return 0; 1564 #endif 1565 } 1566 1567 void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp, 1568 int outin); 1569 int ip_vs_confirm_conntrack(struct sk_buff *skb); 1570 void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct, 1571 struct ip_vs_conn *cp, u_int8_t proto, 1572 const __be16 port, int from_rs); 1573 void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp); 1574 1575 #else 1576 1577 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs) 1578 { 1579 return 0; 1580 } 1581 1582 static inline void ip_vs_update_conntrack(struct sk_buff *skb, 1583 struct ip_vs_conn *cp, int outin) 1584 { 1585 } 1586 1587 static inline int ip_vs_confirm_conntrack(struct sk_buff *skb) 1588 { 1589 return NF_ACCEPT; 1590 } 1591 1592 static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp) 1593 { 1594 } 1595 #endif /* CONFIG_IP_VS_NFCT */ 1596 1597 /* Really using conntrack? */ 1598 static inline bool ip_vs_conn_uses_conntrack(struct ip_vs_conn *cp, 1599 struct sk_buff *skb) 1600 { 1601 #ifdef CONFIG_IP_VS_NFCT 1602 enum ip_conntrack_info ctinfo; 1603 struct nf_conn *ct; 1604 1605 if (!(cp->flags & IP_VS_CONN_F_NFCT)) 1606 return false; 1607 ct = nf_ct_get(skb, &ctinfo); 1608 if (ct) 1609 return true; 1610 #endif 1611 return false; 1612 } 1613 1614 static inline int 1615 ip_vs_dest_conn_overhead(struct ip_vs_dest *dest) 1616 { 1617 /* We think the overhead of processing active connections is 256 1618 * times higher than that of inactive connections in average. (This 1619 * 256 times might not be accurate, we will change it later) We 1620 * use the following formula to estimate the overhead now: 1621 * dest->activeconns*256 + dest->inactconns 1622 */ 1623 return (atomic_read(&dest->activeconns) << 8) + 1624 atomic_read(&dest->inactconns); 1625 } 1626 1627 #endif /* _NET_IP_VS_H */ 1628