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 #include <linux/workqueue.h> 18 19 #include <linux/compiler.h> 20 #include <linux/timer.h> 21 #include <linux/bug.h> 22 23 #include <net/checksum.h> 24 #include <linux/netfilter.h> /* for union nf_inet_addr */ 25 #include <linux/ip.h> 26 #include <linux/ipv6.h> /* for struct ipv6hdr */ 27 #include <net/ipv6.h> 28 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 29 #include <net/netfilter/nf_conntrack.h> 30 #endif 31 #include <net/net_namespace.h> /* Netw namespace */ 32 33 #define IP_VS_HDR_INVERSE 1 34 #define IP_VS_HDR_ICMP 2 35 36 /* Generic access of ipvs struct */ 37 static inline struct netns_ipvs *net_ipvs(struct net* net) 38 { 39 return net->ipvs; 40 } 41 42 /* Connections' size value needed by ip_vs_ctl.c */ 43 extern int ip_vs_conn_tab_size; 44 45 struct ip_vs_iphdr { 46 int hdr_flags; /* ipvs flags */ 47 __u32 off; /* Where IP or IPv4 header starts */ 48 __u32 len; /* IPv4 simply where L4 starts 49 * IPv6 where L4 Transport Header starts */ 50 __u16 fragoffs; /* IPv6 fragment offset, 0 if first frag (or not frag)*/ 51 __s16 protocol; 52 __s32 flags; 53 union nf_inet_addr saddr; 54 union nf_inet_addr daddr; 55 }; 56 57 static inline void *frag_safe_skb_hp(const struct sk_buff *skb, int offset, 58 int len, void *buffer) 59 { 60 return skb_header_pointer(skb, offset, len, buffer); 61 } 62 63 /* This function handles filling *ip_vs_iphdr, both for IPv4 and IPv6. 64 * IPv6 requires some extra work, as finding proper header position, 65 * depend on the IPv6 extension headers. 66 */ 67 static inline int 68 ip_vs_fill_iph_skb_off(int af, const struct sk_buff *skb, int offset, 69 int hdr_flags, struct ip_vs_iphdr *iphdr) 70 { 71 iphdr->hdr_flags = hdr_flags; 72 iphdr->off = offset; 73 74 #ifdef CONFIG_IP_VS_IPV6 75 if (af == AF_INET6) { 76 struct ipv6hdr _iph; 77 const struct ipv6hdr *iph = skb_header_pointer( 78 skb, offset, sizeof(_iph), &_iph); 79 if (!iph) 80 return 0; 81 82 iphdr->saddr.in6 = iph->saddr; 83 iphdr->daddr.in6 = iph->daddr; 84 /* ipv6_find_hdr() updates len, flags */ 85 iphdr->len = offset; 86 iphdr->flags = 0; 87 iphdr->protocol = ipv6_find_hdr(skb, &iphdr->len, -1, 88 &iphdr->fragoffs, 89 &iphdr->flags); 90 if (iphdr->protocol < 0) 91 return 0; 92 } else 93 #endif 94 { 95 struct iphdr _iph; 96 const struct iphdr *iph = skb_header_pointer( 97 skb, offset, sizeof(_iph), &_iph); 98 if (!iph) 99 return 0; 100 101 iphdr->len = offset + iph->ihl * 4; 102 iphdr->fragoffs = 0; 103 iphdr->protocol = iph->protocol; 104 iphdr->saddr.ip = iph->saddr; 105 iphdr->daddr.ip = iph->daddr; 106 } 107 108 return 1; 109 } 110 111 static inline int 112 ip_vs_fill_iph_skb_icmp(int af, const struct sk_buff *skb, int offset, 113 bool inverse, struct ip_vs_iphdr *iphdr) 114 { 115 int hdr_flags = IP_VS_HDR_ICMP; 116 117 if (inverse) 118 hdr_flags |= IP_VS_HDR_INVERSE; 119 120 return ip_vs_fill_iph_skb_off(af, skb, offset, hdr_flags, iphdr); 121 } 122 123 static inline int 124 ip_vs_fill_iph_skb(int af, const struct sk_buff *skb, bool inverse, 125 struct ip_vs_iphdr *iphdr) 126 { 127 int hdr_flags = 0; 128 129 if (inverse) 130 hdr_flags |= IP_VS_HDR_INVERSE; 131 132 return ip_vs_fill_iph_skb_off(af, skb, skb_network_offset(skb), 133 hdr_flags, iphdr); 134 } 135 136 static inline bool 137 ip_vs_iph_inverse(const struct ip_vs_iphdr *iph) 138 { 139 return !!(iph->hdr_flags & IP_VS_HDR_INVERSE); 140 } 141 142 static inline bool 143 ip_vs_iph_icmp(const struct ip_vs_iphdr *iph) 144 { 145 return !!(iph->hdr_flags & IP_VS_HDR_ICMP); 146 } 147 148 static inline void ip_vs_addr_copy(int af, union nf_inet_addr *dst, 149 const union nf_inet_addr *src) 150 { 151 #ifdef CONFIG_IP_VS_IPV6 152 if (af == AF_INET6) 153 dst->in6 = src->in6; 154 else 155 #endif 156 dst->ip = src->ip; 157 } 158 159 static inline void ip_vs_addr_set(int af, union nf_inet_addr *dst, 160 const union nf_inet_addr *src) 161 { 162 #ifdef CONFIG_IP_VS_IPV6 163 if (af == AF_INET6) { 164 dst->in6 = src->in6; 165 return; 166 } 167 #endif 168 dst->ip = src->ip; 169 dst->all[1] = 0; 170 dst->all[2] = 0; 171 dst->all[3] = 0; 172 } 173 174 static inline int ip_vs_addr_equal(int af, const union nf_inet_addr *a, 175 const union nf_inet_addr *b) 176 { 177 #ifdef CONFIG_IP_VS_IPV6 178 if (af == AF_INET6) 179 return ipv6_addr_equal(&a->in6, &b->in6); 180 #endif 181 return a->ip == b->ip; 182 } 183 184 #ifdef CONFIG_IP_VS_DEBUG 185 #include <linux/net.h> 186 187 int ip_vs_get_debug_level(void); 188 189 static inline const char *ip_vs_dbg_addr(int af, char *buf, size_t buf_len, 190 const union nf_inet_addr *addr, 191 int *idx) 192 { 193 int len; 194 #ifdef CONFIG_IP_VS_IPV6 195 if (af == AF_INET6) 196 len = snprintf(&buf[*idx], buf_len - *idx, "[%pI6c]", 197 &addr->in6) + 1; 198 else 199 #endif 200 len = snprintf(&buf[*idx], buf_len - *idx, "%pI4", 201 &addr->ip) + 1; 202 203 *idx += len; 204 BUG_ON(*idx > buf_len + 1); 205 return &buf[*idx - len]; 206 } 207 208 #define IP_VS_DBG_BUF(level, msg, ...) \ 209 do { \ 210 char ip_vs_dbg_buf[160]; \ 211 int ip_vs_dbg_idx = 0; \ 212 if (level <= ip_vs_get_debug_level()) \ 213 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \ 214 } while (0) 215 #define IP_VS_ERR_BUF(msg...) \ 216 do { \ 217 char ip_vs_dbg_buf[160]; \ 218 int ip_vs_dbg_idx = 0; \ 219 pr_err(msg); \ 220 } while (0) 221 222 /* Only use from within IP_VS_DBG_BUF() or IP_VS_ERR_BUF macros */ 223 #define IP_VS_DBG_ADDR(af, addr) \ 224 ip_vs_dbg_addr(af, ip_vs_dbg_buf, \ 225 sizeof(ip_vs_dbg_buf), addr, \ 226 &ip_vs_dbg_idx) 227 228 #define IP_VS_DBG(level, msg, ...) \ 229 do { \ 230 if (level <= ip_vs_get_debug_level()) \ 231 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \ 232 } while (0) 233 #define IP_VS_DBG_RL(msg, ...) \ 234 do { \ 235 if (net_ratelimit()) \ 236 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \ 237 } while (0) 238 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) \ 239 do { \ 240 if (level <= ip_vs_get_debug_level()) \ 241 pp->debug_packet(af, pp, skb, ofs, msg); \ 242 } while (0) 243 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) \ 244 do { \ 245 if (level <= ip_vs_get_debug_level() && \ 246 net_ratelimit()) \ 247 pp->debug_packet(af, pp, skb, ofs, msg); \ 248 } while (0) 249 #else /* NO DEBUGGING at ALL */ 250 #define IP_VS_DBG_BUF(level, msg...) do {} while (0) 251 #define IP_VS_ERR_BUF(msg...) do {} while (0) 252 #define IP_VS_DBG(level, msg...) do {} while (0) 253 #define IP_VS_DBG_RL(msg...) do {} while (0) 254 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) do {} while (0) 255 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) do {} while (0) 256 #endif 257 258 #define IP_VS_BUG() BUG() 259 #define IP_VS_ERR_RL(msg, ...) \ 260 do { \ 261 if (net_ratelimit()) \ 262 pr_err(msg, ##__VA_ARGS__); \ 263 } while (0) 264 265 #ifdef CONFIG_IP_VS_DEBUG 266 #define EnterFunction(level) \ 267 do { \ 268 if (level <= ip_vs_get_debug_level()) \ 269 printk(KERN_DEBUG \ 270 pr_fmt("Enter: %s, %s line %i\n"), \ 271 __func__, __FILE__, __LINE__); \ 272 } while (0) 273 #define LeaveFunction(level) \ 274 do { \ 275 if (level <= ip_vs_get_debug_level()) \ 276 printk(KERN_DEBUG \ 277 pr_fmt("Leave: %s, %s line %i\n"), \ 278 __func__, __FILE__, __LINE__); \ 279 } while (0) 280 #else 281 #define EnterFunction(level) do {} while (0) 282 #define LeaveFunction(level) do {} while (0) 283 #endif 284 285 /* The port number of FTP service (in network order). */ 286 #define FTPPORT cpu_to_be16(21) 287 #define FTPDATA cpu_to_be16(20) 288 289 /* TCP State Values */ 290 enum { 291 IP_VS_TCP_S_NONE = 0, 292 IP_VS_TCP_S_ESTABLISHED, 293 IP_VS_TCP_S_SYN_SENT, 294 IP_VS_TCP_S_SYN_RECV, 295 IP_VS_TCP_S_FIN_WAIT, 296 IP_VS_TCP_S_TIME_WAIT, 297 IP_VS_TCP_S_CLOSE, 298 IP_VS_TCP_S_CLOSE_WAIT, 299 IP_VS_TCP_S_LAST_ACK, 300 IP_VS_TCP_S_LISTEN, 301 IP_VS_TCP_S_SYNACK, 302 IP_VS_TCP_S_LAST 303 }; 304 305 /* UDP State Values */ 306 enum { 307 IP_VS_UDP_S_NORMAL, 308 IP_VS_UDP_S_LAST, 309 }; 310 311 /* ICMP State Values */ 312 enum { 313 IP_VS_ICMP_S_NORMAL, 314 IP_VS_ICMP_S_LAST, 315 }; 316 317 /* SCTP State Values */ 318 enum ip_vs_sctp_states { 319 IP_VS_SCTP_S_NONE, 320 IP_VS_SCTP_S_INIT1, 321 IP_VS_SCTP_S_INIT, 322 IP_VS_SCTP_S_COOKIE_SENT, 323 IP_VS_SCTP_S_COOKIE_REPLIED, 324 IP_VS_SCTP_S_COOKIE_WAIT, 325 IP_VS_SCTP_S_COOKIE, 326 IP_VS_SCTP_S_COOKIE_ECHOED, 327 IP_VS_SCTP_S_ESTABLISHED, 328 IP_VS_SCTP_S_SHUTDOWN_SENT, 329 IP_VS_SCTP_S_SHUTDOWN_RECEIVED, 330 IP_VS_SCTP_S_SHUTDOWN_ACK_SENT, 331 IP_VS_SCTP_S_REJECTED, 332 IP_VS_SCTP_S_CLOSED, 333 IP_VS_SCTP_S_LAST 334 }; 335 336 /* Connection templates use bits from state */ 337 #define IP_VS_CTPL_S_NONE 0x0000 338 #define IP_VS_CTPL_S_ASSURED 0x0001 339 #define IP_VS_CTPL_S_LAST 0x0002 340 341 /* Delta sequence info structure 342 * Each ip_vs_conn has 2 (output AND input seq. changes). 343 * Only used in the VS/NAT. 344 */ 345 struct ip_vs_seq { 346 __u32 init_seq; /* Add delta from this seq */ 347 __u32 delta; /* Delta in sequence numbers */ 348 __u32 previous_delta; /* Delta in sequence numbers 349 * before last resized pkt */ 350 }; 351 352 /* counters per cpu */ 353 struct ip_vs_counters { 354 __u64 conns; /* connections scheduled */ 355 __u64 inpkts; /* incoming packets */ 356 __u64 outpkts; /* outgoing packets */ 357 __u64 inbytes; /* incoming bytes */ 358 __u64 outbytes; /* outgoing bytes */ 359 }; 360 /* Stats per cpu */ 361 struct ip_vs_cpu_stats { 362 struct ip_vs_counters cnt; 363 struct u64_stats_sync syncp; 364 }; 365 366 /* IPVS statistics objects */ 367 struct ip_vs_estimator { 368 struct list_head list; 369 370 u64 last_inbytes; 371 u64 last_outbytes; 372 u64 last_conns; 373 u64 last_inpkts; 374 u64 last_outpkts; 375 376 u64 cps; 377 u64 inpps; 378 u64 outpps; 379 u64 inbps; 380 u64 outbps; 381 }; 382 383 /* 384 * IPVS statistics object, 64-bit kernel version of struct ip_vs_stats_user 385 */ 386 struct ip_vs_kstats { 387 u64 conns; /* connections scheduled */ 388 u64 inpkts; /* incoming packets */ 389 u64 outpkts; /* outgoing packets */ 390 u64 inbytes; /* incoming bytes */ 391 u64 outbytes; /* outgoing bytes */ 392 393 u64 cps; /* current connection rate */ 394 u64 inpps; /* current in packet rate */ 395 u64 outpps; /* current out packet rate */ 396 u64 inbps; /* current in byte rate */ 397 u64 outbps; /* current out byte rate */ 398 }; 399 400 struct ip_vs_stats { 401 struct ip_vs_kstats kstats; /* kernel statistics */ 402 struct ip_vs_estimator est; /* estimator */ 403 struct ip_vs_cpu_stats __percpu *cpustats; /* per cpu counters */ 404 spinlock_t lock; /* spin lock */ 405 struct ip_vs_kstats kstats0; /* reset values */ 406 }; 407 408 struct dst_entry; 409 struct iphdr; 410 struct ip_vs_conn; 411 struct ip_vs_app; 412 struct sk_buff; 413 struct ip_vs_proto_data; 414 415 struct ip_vs_protocol { 416 struct ip_vs_protocol *next; 417 char *name; 418 u16 protocol; 419 u16 num_states; 420 int dont_defrag; 421 422 void (*init)(struct ip_vs_protocol *pp); 423 424 void (*exit)(struct ip_vs_protocol *pp); 425 426 int (*init_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd); 427 428 void (*exit_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd); 429 430 int (*conn_schedule)(struct netns_ipvs *ipvs, 431 int af, struct sk_buff *skb, 432 struct ip_vs_proto_data *pd, 433 int *verdict, struct ip_vs_conn **cpp, 434 struct ip_vs_iphdr *iph); 435 436 struct ip_vs_conn * 437 (*conn_in_get)(struct netns_ipvs *ipvs, 438 int af, 439 const struct sk_buff *skb, 440 const struct ip_vs_iphdr *iph); 441 442 struct ip_vs_conn * 443 (*conn_out_get)(struct netns_ipvs *ipvs, 444 int af, 445 const struct sk_buff *skb, 446 const struct ip_vs_iphdr *iph); 447 448 int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp, 449 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph); 450 451 int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp, 452 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph); 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 u16 tun_type; /* tunnel type */ 603 __be16 tun_port; /* tunnel port */ 604 u16 tun_flags; /* tunnel flags */ 605 }; 606 607 608 /* 609 * The information about the virtual service offered to the net and the 610 * forwarding entries. 611 */ 612 struct ip_vs_service { 613 struct hlist_node s_list; /* for normal service table */ 614 struct hlist_node f_list; /* for fwmark-based service table */ 615 atomic_t refcnt; /* reference counter */ 616 617 u16 af; /* address family */ 618 __u16 protocol; /* which protocol (TCP/UDP) */ 619 union nf_inet_addr addr; /* IP address for virtual service */ 620 __be16 port; /* port number for the service */ 621 __u32 fwmark; /* firewall mark of the service */ 622 unsigned int flags; /* service status flags */ 623 unsigned int timeout; /* persistent timeout in ticks */ 624 __be32 netmask; /* grouping granularity, mask/plen */ 625 struct netns_ipvs *ipvs; 626 627 struct list_head destinations; /* real server d-linked list */ 628 __u32 num_dests; /* number of servers */ 629 struct ip_vs_stats stats; /* statistics for the service */ 630 631 /* for scheduling */ 632 struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */ 633 spinlock_t sched_lock; /* lock sched_data */ 634 void *sched_data; /* scheduler application data */ 635 636 /* alternate persistence engine */ 637 struct ip_vs_pe __rcu *pe; 638 int conntrack_afmask; 639 640 struct rcu_head rcu_head; 641 }; 642 643 /* Information for cached dst */ 644 struct ip_vs_dest_dst { 645 struct dst_entry *dst_cache; /* destination cache entry */ 646 u32 dst_cookie; 647 union nf_inet_addr dst_saddr; 648 struct rcu_head rcu_head; 649 }; 650 651 /* The real server destination forwarding entry with ip address, port number, 652 * and so on. 653 */ 654 struct ip_vs_dest { 655 struct list_head n_list; /* for the dests in the service */ 656 struct hlist_node d_list; /* for table with all the dests */ 657 658 u16 af; /* address family */ 659 __be16 port; /* port number of the server */ 660 union nf_inet_addr addr; /* IP address of the server */ 661 volatile unsigned int flags; /* dest status flags */ 662 atomic_t conn_flags; /* flags to copy to conn */ 663 atomic_t weight; /* server weight */ 664 atomic_t last_weight; /* server latest weight */ 665 __u16 tun_type; /* tunnel type */ 666 __be16 tun_port; /* tunnel port */ 667 __u16 tun_flags; /* tunnel flags */ 668 669 refcount_t refcnt; /* reference counter */ 670 struct ip_vs_stats stats; /* statistics */ 671 unsigned long idle_start; /* start time, jiffies */ 672 673 /* connection counters and thresholds */ 674 atomic_t activeconns; /* active connections */ 675 atomic_t inactconns; /* inactive connections */ 676 atomic_t persistconns; /* persistent connections */ 677 __u32 u_threshold; /* upper threshold */ 678 __u32 l_threshold; /* lower threshold */ 679 680 /* for destination cache */ 681 spinlock_t dst_lock; /* lock of dst_cache */ 682 struct ip_vs_dest_dst __rcu *dest_dst; /* cached dst info */ 683 684 /* for virtual service */ 685 struct ip_vs_service __rcu *svc; /* service it belongs to */ 686 __u16 protocol; /* which protocol (TCP/UDP) */ 687 __be16 vport; /* virtual port number */ 688 union nf_inet_addr vaddr; /* virtual IP address */ 689 __u32 vfwmark; /* firewall mark of service */ 690 691 struct list_head t_list; /* in dest_trash */ 692 unsigned int in_rs_table:1; /* we are in rs_table */ 693 }; 694 695 /* The scheduler object */ 696 struct ip_vs_scheduler { 697 struct list_head n_list; /* d-linked list head */ 698 char *name; /* scheduler name */ 699 atomic_t refcnt; /* reference counter */ 700 struct module *module; /* THIS_MODULE/NULL */ 701 702 /* scheduler initializing service */ 703 int (*init_service)(struct ip_vs_service *svc); 704 /* scheduling service finish */ 705 void (*done_service)(struct ip_vs_service *svc); 706 /* dest is linked */ 707 int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest); 708 /* dest is unlinked */ 709 int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest); 710 /* dest is updated */ 711 int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest); 712 713 /* selecting a server from the given service */ 714 struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc, 715 const struct sk_buff *skb, 716 struct ip_vs_iphdr *iph); 717 }; 718 719 /* The persistence engine object */ 720 struct ip_vs_pe { 721 struct list_head n_list; /* d-linked list head */ 722 char *name; /* scheduler name */ 723 atomic_t refcnt; /* reference counter */ 724 struct module *module; /* THIS_MODULE/NULL */ 725 726 /* get the connection template, if any */ 727 int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb); 728 bool (*ct_match)(const struct ip_vs_conn_param *p, 729 struct ip_vs_conn *ct); 730 u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval, 731 bool inverse); 732 int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf); 733 /* create connections for real-server outgoing packets */ 734 struct ip_vs_conn* (*conn_out)(struct ip_vs_service *svc, 735 struct ip_vs_dest *dest, 736 struct sk_buff *skb, 737 const struct ip_vs_iphdr *iph, 738 __be16 dport, __be16 cport); 739 }; 740 741 /* The application module object (a.k.a. app incarnation) */ 742 struct ip_vs_app { 743 struct list_head a_list; /* member in app list */ 744 int type; /* IP_VS_APP_TYPE_xxx */ 745 char *name; /* application module name */ 746 __u16 protocol; 747 struct module *module; /* THIS_MODULE/NULL */ 748 struct list_head incs_list; /* list of incarnations */ 749 750 /* members for application incarnations */ 751 struct list_head p_list; /* member in proto app list */ 752 struct ip_vs_app *app; /* its real application */ 753 __be16 port; /* port number in net order */ 754 atomic_t usecnt; /* usage counter */ 755 struct rcu_head rcu_head; 756 757 /* output hook: Process packet in inout direction, diff set for TCP. 758 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok, 759 * 2=Mangled but checksum was not updated 760 */ 761 int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *, 762 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh); 763 764 /* input hook: Process packet in outin direction, diff set for TCP. 765 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok, 766 * 2=Mangled but checksum was not updated 767 */ 768 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *, 769 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh); 770 771 /* ip_vs_app initializer */ 772 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *); 773 774 /* ip_vs_app finish */ 775 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *); 776 777 778 /* not used now */ 779 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *, 780 struct ip_vs_protocol *); 781 782 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *); 783 784 int * timeout_table; 785 int * timeouts; 786 int timeouts_size; 787 788 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app, 789 int *verdict, struct ip_vs_conn **cpp); 790 791 struct ip_vs_conn * 792 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app, 793 const struct iphdr *iph, int inverse); 794 795 struct ip_vs_conn * 796 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app, 797 const struct iphdr *iph, int inverse); 798 799 int (*state_transition)(struct ip_vs_conn *cp, int direction, 800 const struct sk_buff *skb, 801 struct ip_vs_app *app); 802 803 void (*timeout_change)(struct ip_vs_app *app, int flags); 804 }; 805 806 struct ipvs_master_sync_state { 807 struct list_head sync_queue; 808 struct ip_vs_sync_buff *sync_buff; 809 unsigned long sync_queue_len; 810 unsigned int sync_queue_delay; 811 struct delayed_work master_wakeup_work; 812 struct netns_ipvs *ipvs; 813 }; 814 815 struct ip_vs_sync_thread_data; 816 817 /* How much time to keep dests in trash */ 818 #define IP_VS_DEST_TRASH_PERIOD (120 * HZ) 819 820 struct ipvs_sync_daemon_cfg { 821 union nf_inet_addr mcast_group; 822 int syncid; 823 u16 sync_maxlen; 824 u16 mcast_port; 825 u8 mcast_af; 826 u8 mcast_ttl; 827 /* multicast interface name */ 828 char mcast_ifn[IP_VS_IFNAME_MAXLEN]; 829 }; 830 831 /* IPVS in network namespace */ 832 struct netns_ipvs { 833 int gen; /* Generation */ 834 int enable; /* enable like nf_hooks do */ 835 /* Hash table: for real service lookups */ 836 #define IP_VS_RTAB_BITS 4 837 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS) 838 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1) 839 840 struct hlist_head rs_table[IP_VS_RTAB_SIZE]; 841 /* ip_vs_app */ 842 struct list_head app_list; 843 /* ip_vs_proto */ 844 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */ 845 struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE]; 846 /* ip_vs_proto_tcp */ 847 #ifdef CONFIG_IP_VS_PROTO_TCP 848 #define TCP_APP_TAB_BITS 4 849 #define TCP_APP_TAB_SIZE (1 << TCP_APP_TAB_BITS) 850 #define TCP_APP_TAB_MASK (TCP_APP_TAB_SIZE - 1) 851 struct list_head tcp_apps[TCP_APP_TAB_SIZE]; 852 #endif 853 /* ip_vs_proto_udp */ 854 #ifdef CONFIG_IP_VS_PROTO_UDP 855 #define UDP_APP_TAB_BITS 4 856 #define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS) 857 #define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1) 858 struct list_head udp_apps[UDP_APP_TAB_SIZE]; 859 #endif 860 /* ip_vs_proto_sctp */ 861 #ifdef CONFIG_IP_VS_PROTO_SCTP 862 #define SCTP_APP_TAB_BITS 4 863 #define SCTP_APP_TAB_SIZE (1 << SCTP_APP_TAB_BITS) 864 #define SCTP_APP_TAB_MASK (SCTP_APP_TAB_SIZE - 1) 865 /* Hash table for SCTP application incarnations */ 866 struct list_head sctp_apps[SCTP_APP_TAB_SIZE]; 867 #endif 868 /* ip_vs_conn */ 869 atomic_t conn_count; /* connection counter */ 870 871 /* ip_vs_ctl */ 872 struct ip_vs_stats tot_stats; /* Statistics & est. */ 873 874 int num_services; /* no of virtual services */ 875 int num_services6; /* IPv6 virtual services */ 876 877 /* Trash for destinations */ 878 struct list_head dest_trash; 879 spinlock_t dest_trash_lock; 880 struct timer_list dest_trash_timer; /* expiration timer */ 881 /* Service counters */ 882 atomic_t ftpsvc_counter; 883 atomic_t nullsvc_counter; 884 atomic_t conn_out_counter; 885 886 #ifdef CONFIG_SYSCTL 887 /* delayed work for expiring no dest connections */ 888 struct delayed_work expire_nodest_conn_work; 889 /* 1/rate drop and drop-entry variables */ 890 struct delayed_work defense_work; /* Work handler */ 891 int drop_rate; 892 int drop_counter; 893 int old_secure_tcp; 894 atomic_t dropentry; 895 /* locks in ctl.c */ 896 spinlock_t dropentry_lock; /* drop entry handling */ 897 spinlock_t droppacket_lock; /* drop packet handling */ 898 spinlock_t securetcp_lock; /* state and timeout tables */ 899 900 /* sys-ctl struct */ 901 struct ctl_table_header *sysctl_hdr; 902 struct ctl_table *sysctl_tbl; 903 #endif 904 905 /* sysctl variables */ 906 int sysctl_amemthresh; 907 int sysctl_am_droprate; 908 int sysctl_drop_entry; 909 int sysctl_drop_packet; 910 int sysctl_secure_tcp; 911 #ifdef CONFIG_IP_VS_NFCT 912 int sysctl_conntrack; 913 #endif 914 int sysctl_snat_reroute; 915 int sysctl_sync_ver; 916 int sysctl_sync_ports; 917 int sysctl_sync_persist_mode; 918 unsigned long sysctl_sync_qlen_max; 919 int sysctl_sync_sock_size; 920 int sysctl_cache_bypass; 921 int sysctl_expire_nodest_conn; 922 int sysctl_sloppy_tcp; 923 int sysctl_sloppy_sctp; 924 int sysctl_expire_quiescent_template; 925 int sysctl_sync_threshold[2]; 926 unsigned int sysctl_sync_refresh_period; 927 int sysctl_sync_retries; 928 int sysctl_nat_icmp_send; 929 int sysctl_pmtu_disc; 930 int sysctl_backup_only; 931 int sysctl_conn_reuse_mode; 932 int sysctl_schedule_icmp; 933 int sysctl_ignore_tunneled; 934 935 /* ip_vs_lblc */ 936 int sysctl_lblc_expiration; 937 struct ctl_table_header *lblc_ctl_header; 938 struct ctl_table *lblc_ctl_table; 939 /* ip_vs_lblcr */ 940 int sysctl_lblcr_expiration; 941 struct ctl_table_header *lblcr_ctl_header; 942 struct ctl_table *lblcr_ctl_table; 943 /* ip_vs_est */ 944 struct list_head est_list; /* estimator list */ 945 spinlock_t est_lock; 946 struct timer_list est_timer; /* Estimation timer */ 947 /* ip_vs_sync */ 948 spinlock_t sync_lock; 949 struct ipvs_master_sync_state *ms; 950 spinlock_t sync_buff_lock; 951 struct ip_vs_sync_thread_data *master_tinfo; 952 struct ip_vs_sync_thread_data *backup_tinfo; 953 int threads_mask; 954 volatile int sync_state; 955 struct mutex sync_mutex; 956 struct ipvs_sync_daemon_cfg mcfg; /* Master Configuration */ 957 struct ipvs_sync_daemon_cfg bcfg; /* Backup Configuration */ 958 /* net name space ptr */ 959 struct net *net; /* Needed by timer routines */ 960 /* Number of heterogeneous destinations, needed becaus heterogeneous 961 * are not supported when synchronization is enabled. 962 */ 963 unsigned int mixed_address_family_dests; 964 unsigned int hooks_afmask; /* &1=AF_INET, &2=AF_INET6 */ 965 }; 966 967 #define DEFAULT_SYNC_THRESHOLD 3 968 #define DEFAULT_SYNC_PERIOD 50 969 #define DEFAULT_SYNC_VER 1 970 #define DEFAULT_SLOPPY_TCP 0 971 #define DEFAULT_SLOPPY_SCTP 0 972 #define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ) 973 #define DEFAULT_SYNC_RETRIES 0 974 #define IPVS_SYNC_WAKEUP_RATE 8 975 #define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4) 976 #define IPVS_SYNC_SEND_DELAY (HZ / 50) 977 #define IPVS_SYNC_CHECK_PERIOD HZ 978 #define IPVS_SYNC_FLUSH_TIME (HZ * 2) 979 #define IPVS_SYNC_PORTS_MAX (1 << 6) 980 981 #ifdef CONFIG_SYSCTL 982 983 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs) 984 { 985 return ipvs->sysctl_sync_threshold[0]; 986 } 987 988 static inline int sysctl_sync_period(struct netns_ipvs *ipvs) 989 { 990 return READ_ONCE(ipvs->sysctl_sync_threshold[1]); 991 } 992 993 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs) 994 { 995 return READ_ONCE(ipvs->sysctl_sync_refresh_period); 996 } 997 998 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs) 999 { 1000 return ipvs->sysctl_sync_retries; 1001 } 1002 1003 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs) 1004 { 1005 return ipvs->sysctl_sync_ver; 1006 } 1007 1008 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs) 1009 { 1010 return ipvs->sysctl_sloppy_tcp; 1011 } 1012 1013 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs) 1014 { 1015 return ipvs->sysctl_sloppy_sctp; 1016 } 1017 1018 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs) 1019 { 1020 return READ_ONCE(ipvs->sysctl_sync_ports); 1021 } 1022 1023 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs) 1024 { 1025 return ipvs->sysctl_sync_persist_mode; 1026 } 1027 1028 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs) 1029 { 1030 return ipvs->sysctl_sync_qlen_max; 1031 } 1032 1033 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs) 1034 { 1035 return ipvs->sysctl_sync_sock_size; 1036 } 1037 1038 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs) 1039 { 1040 return ipvs->sysctl_pmtu_disc; 1041 } 1042 1043 static inline int sysctl_backup_only(struct netns_ipvs *ipvs) 1044 { 1045 return ipvs->sync_state & IP_VS_STATE_BACKUP && 1046 ipvs->sysctl_backup_only; 1047 } 1048 1049 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs) 1050 { 1051 return ipvs->sysctl_conn_reuse_mode; 1052 } 1053 1054 static inline int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs) 1055 { 1056 return ipvs->sysctl_expire_nodest_conn; 1057 } 1058 1059 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs) 1060 { 1061 return ipvs->sysctl_schedule_icmp; 1062 } 1063 1064 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs) 1065 { 1066 return ipvs->sysctl_ignore_tunneled; 1067 } 1068 1069 static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs) 1070 { 1071 return ipvs->sysctl_cache_bypass; 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 static inline int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs) 1147 { 1148 return 0; 1149 } 1150 1151 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs) 1152 { 1153 return 0; 1154 } 1155 1156 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs) 1157 { 1158 return 0; 1159 } 1160 1161 static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs) 1162 { 1163 return 0; 1164 } 1165 1166 #endif 1167 1168 /* IPVS core functions 1169 * (from ip_vs_core.c) 1170 */ 1171 const char *ip_vs_proto_name(unsigned int proto); 1172 void ip_vs_init_hash_table(struct list_head *table, int rows); 1173 struct ip_vs_conn *ip_vs_new_conn_out(struct ip_vs_service *svc, 1174 struct ip_vs_dest *dest, 1175 struct sk_buff *skb, 1176 const struct ip_vs_iphdr *iph, 1177 __be16 dport, 1178 __be16 cport); 1179 #define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t))) 1180 1181 #define IP_VS_APP_TYPE_FTP 1 1182 1183 /* ip_vs_conn handling functions 1184 * (from ip_vs_conn.c) 1185 */ 1186 enum { 1187 IP_VS_DIR_INPUT = 0, 1188 IP_VS_DIR_OUTPUT, 1189 IP_VS_DIR_INPUT_ONLY, 1190 IP_VS_DIR_LAST, 1191 }; 1192 1193 static inline void ip_vs_conn_fill_param(struct netns_ipvs *ipvs, int af, int protocol, 1194 const union nf_inet_addr *caddr, 1195 __be16 cport, 1196 const union nf_inet_addr *vaddr, 1197 __be16 vport, 1198 struct ip_vs_conn_param *p) 1199 { 1200 p->ipvs = ipvs; 1201 p->af = af; 1202 p->protocol = protocol; 1203 p->caddr = caddr; 1204 p->cport = cport; 1205 p->vaddr = vaddr; 1206 p->vport = vport; 1207 p->pe = NULL; 1208 p->pe_data = NULL; 1209 } 1210 1211 struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p); 1212 struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p); 1213 1214 struct ip_vs_conn * ip_vs_conn_in_get_proto(struct netns_ipvs *ipvs, int af, 1215 const struct sk_buff *skb, 1216 const struct ip_vs_iphdr *iph); 1217 1218 struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p); 1219 1220 struct ip_vs_conn * ip_vs_conn_out_get_proto(struct netns_ipvs *ipvs, int af, 1221 const struct sk_buff *skb, 1222 const struct ip_vs_iphdr *iph); 1223 1224 /* Get reference to gain full access to conn. 1225 * By default, RCU read-side critical sections have access only to 1226 * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference. 1227 */ 1228 static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp) 1229 { 1230 return refcount_inc_not_zero(&cp->refcnt); 1231 } 1232 1233 /* put back the conn without restarting its timer */ 1234 static inline void __ip_vs_conn_put(struct ip_vs_conn *cp) 1235 { 1236 smp_mb__before_atomic(); 1237 refcount_dec(&cp->refcnt); 1238 } 1239 void ip_vs_conn_put(struct ip_vs_conn *cp); 1240 void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport); 1241 1242 struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, int dest_af, 1243 const union nf_inet_addr *daddr, 1244 __be16 dport, unsigned int flags, 1245 struct ip_vs_dest *dest, __u32 fwmark); 1246 void ip_vs_conn_expire_now(struct ip_vs_conn *cp); 1247 1248 const char *ip_vs_state_name(const struct ip_vs_conn *cp); 1249 1250 void ip_vs_tcp_conn_listen(struct ip_vs_conn *cp); 1251 int ip_vs_check_template(struct ip_vs_conn *ct, struct ip_vs_dest *cdest); 1252 void ip_vs_random_dropentry(struct netns_ipvs *ipvs); 1253 int ip_vs_conn_init(void); 1254 void ip_vs_conn_cleanup(void); 1255 1256 static inline void ip_vs_control_del(struct ip_vs_conn *cp) 1257 { 1258 struct ip_vs_conn *ctl_cp = cp->control; 1259 if (!ctl_cp) { 1260 IP_VS_ERR_BUF("request control DEL for uncontrolled: " 1261 "%s:%d to %s:%d\n", 1262 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1263 ntohs(cp->cport), 1264 IP_VS_DBG_ADDR(cp->af, &cp->vaddr), 1265 ntohs(cp->vport)); 1266 1267 return; 1268 } 1269 1270 IP_VS_DBG_BUF(7, "DELeting control for: " 1271 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n", 1272 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1273 ntohs(cp->cport), 1274 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr), 1275 ntohs(ctl_cp->cport)); 1276 1277 cp->control = NULL; 1278 if (atomic_read(&ctl_cp->n_control) == 0) { 1279 IP_VS_ERR_BUF("BUG control DEL with n=0 : " 1280 "%s:%d to %s:%d\n", 1281 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1282 ntohs(cp->cport), 1283 IP_VS_DBG_ADDR(cp->af, &cp->vaddr), 1284 ntohs(cp->vport)); 1285 1286 return; 1287 } 1288 atomic_dec(&ctl_cp->n_control); 1289 } 1290 1291 static inline void 1292 ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp) 1293 { 1294 if (cp->control) { 1295 IP_VS_ERR_BUF("request control ADD for already controlled: " 1296 "%s:%d to %s:%d\n", 1297 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1298 ntohs(cp->cport), 1299 IP_VS_DBG_ADDR(cp->af, &cp->vaddr), 1300 ntohs(cp->vport)); 1301 1302 ip_vs_control_del(cp); 1303 } 1304 1305 IP_VS_DBG_BUF(7, "ADDing control for: " 1306 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n", 1307 IP_VS_DBG_ADDR(cp->af, &cp->caddr), 1308 ntohs(cp->cport), 1309 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr), 1310 ntohs(ctl_cp->cport)); 1311 1312 cp->control = ctl_cp; 1313 atomic_inc(&ctl_cp->n_control); 1314 } 1315 1316 /* Mark our template as assured */ 1317 static inline void 1318 ip_vs_control_assure_ct(struct ip_vs_conn *cp) 1319 { 1320 struct ip_vs_conn *ct = cp->control; 1321 1322 if (ct && !(ct->state & IP_VS_CTPL_S_ASSURED) && 1323 (ct->flags & IP_VS_CONN_F_TEMPLATE)) 1324 ct->state |= IP_VS_CTPL_S_ASSURED; 1325 } 1326 1327 /* IPVS netns init & cleanup functions */ 1328 int ip_vs_estimator_net_init(struct netns_ipvs *ipvs); 1329 int ip_vs_control_net_init(struct netns_ipvs *ipvs); 1330 int ip_vs_protocol_net_init(struct netns_ipvs *ipvs); 1331 int ip_vs_app_net_init(struct netns_ipvs *ipvs); 1332 int ip_vs_conn_net_init(struct netns_ipvs *ipvs); 1333 int ip_vs_sync_net_init(struct netns_ipvs *ipvs); 1334 void ip_vs_conn_net_cleanup(struct netns_ipvs *ipvs); 1335 void ip_vs_app_net_cleanup(struct netns_ipvs *ipvs); 1336 void ip_vs_protocol_net_cleanup(struct netns_ipvs *ipvs); 1337 void ip_vs_control_net_cleanup(struct netns_ipvs *ipvs); 1338 void ip_vs_estimator_net_cleanup(struct netns_ipvs *ipvs); 1339 void ip_vs_sync_net_cleanup(struct netns_ipvs *ipvs); 1340 void ip_vs_service_nets_cleanup(struct list_head *net_list); 1341 1342 /* IPVS application functions 1343 * (from ip_vs_app.c) 1344 */ 1345 #define IP_VS_APP_MAX_PORTS 8 1346 struct ip_vs_app *register_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app); 1347 void unregister_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app); 1348 int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp); 1349 void ip_vs_unbind_app(struct ip_vs_conn *cp); 1350 int register_ip_vs_app_inc(struct netns_ipvs *ipvs, struct ip_vs_app *app, __u16 proto, 1351 __u16 port); 1352 int ip_vs_app_inc_get(struct ip_vs_app *inc); 1353 void ip_vs_app_inc_put(struct ip_vs_app *inc); 1354 1355 int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb, 1356 struct ip_vs_iphdr *ipvsh); 1357 int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb, 1358 struct ip_vs_iphdr *ipvsh); 1359 1360 int register_ip_vs_pe(struct ip_vs_pe *pe); 1361 int unregister_ip_vs_pe(struct ip_vs_pe *pe); 1362 struct ip_vs_pe *ip_vs_pe_getbyname(const char *name); 1363 struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name); 1364 1365 /* Use a #define to avoid all of module.h just for these trivial ops */ 1366 #define ip_vs_pe_get(pe) \ 1367 if (pe && pe->module) \ 1368 __module_get(pe->module); 1369 1370 #define ip_vs_pe_put(pe) \ 1371 if (pe && pe->module) \ 1372 module_put(pe->module); 1373 1374 /* IPVS protocol functions (from ip_vs_proto.c) */ 1375 int ip_vs_protocol_init(void); 1376 void ip_vs_protocol_cleanup(void); 1377 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags); 1378 int *ip_vs_create_timeout_table(int *table, int size); 1379 void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp, 1380 const struct sk_buff *skb, int offset, 1381 const char *msg); 1382 1383 extern struct ip_vs_protocol ip_vs_protocol_tcp; 1384 extern struct ip_vs_protocol ip_vs_protocol_udp; 1385 extern struct ip_vs_protocol ip_vs_protocol_icmp; 1386 extern struct ip_vs_protocol ip_vs_protocol_esp; 1387 extern struct ip_vs_protocol ip_vs_protocol_ah; 1388 extern struct ip_vs_protocol ip_vs_protocol_sctp; 1389 1390 /* Registering/unregistering scheduler functions 1391 * (from ip_vs_sched.c) 1392 */ 1393 int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler); 1394 int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler); 1395 int ip_vs_bind_scheduler(struct ip_vs_service *svc, 1396 struct ip_vs_scheduler *scheduler); 1397 void ip_vs_unbind_scheduler(struct ip_vs_service *svc, 1398 struct ip_vs_scheduler *sched); 1399 struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name); 1400 void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler); 1401 struct ip_vs_conn * 1402 ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb, 1403 struct ip_vs_proto_data *pd, int *ignored, 1404 struct ip_vs_iphdr *iph); 1405 int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb, 1406 struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph); 1407 1408 void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg); 1409 1410 /* IPVS control data and functions (from ip_vs_ctl.c) */ 1411 extern struct ip_vs_stats ip_vs_stats; 1412 extern int sysctl_ip_vs_sync_ver; 1413 1414 struct ip_vs_service * 1415 ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol, 1416 const union nf_inet_addr *vaddr, __be16 vport); 1417 1418 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol, 1419 const union nf_inet_addr *daddr, __be16 dport); 1420 1421 struct ip_vs_dest * 1422 ip_vs_find_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol, 1423 const union nf_inet_addr *daddr, __be16 dport); 1424 struct ip_vs_dest *ip_vs_find_tunnel(struct netns_ipvs *ipvs, int af, 1425 const union nf_inet_addr *daddr, 1426 __be16 tun_port); 1427 1428 int ip_vs_use_count_inc(void); 1429 void ip_vs_use_count_dec(void); 1430 int ip_vs_register_nl_ioctl(void); 1431 void ip_vs_unregister_nl_ioctl(void); 1432 int ip_vs_control_init(void); 1433 void ip_vs_control_cleanup(void); 1434 struct ip_vs_dest * 1435 ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af, 1436 const union nf_inet_addr *daddr, __be16 dport, 1437 const union nf_inet_addr *vaddr, __be16 vport, 1438 __u16 protocol, __u32 fwmark, __u32 flags); 1439 void ip_vs_try_bind_dest(struct ip_vs_conn *cp); 1440 1441 static inline void ip_vs_dest_hold(struct ip_vs_dest *dest) 1442 { 1443 refcount_inc(&dest->refcnt); 1444 } 1445 1446 static inline void ip_vs_dest_put(struct ip_vs_dest *dest) 1447 { 1448 smp_mb__before_atomic(); 1449 refcount_dec(&dest->refcnt); 1450 } 1451 1452 static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest) 1453 { 1454 if (refcount_dec_and_test(&dest->refcnt)) 1455 kfree(dest); 1456 } 1457 1458 /* IPVS sync daemon data and function prototypes 1459 * (from ip_vs_sync.c) 1460 */ 1461 int start_sync_thread(struct netns_ipvs *ipvs, struct ipvs_sync_daemon_cfg *cfg, 1462 int state); 1463 int stop_sync_thread(struct netns_ipvs *ipvs, int state); 1464 void ip_vs_sync_conn(struct netns_ipvs *ipvs, struct ip_vs_conn *cp, int pkts); 1465 1466 /* IPVS rate estimator prototypes (from ip_vs_est.c) */ 1467 void ip_vs_start_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats); 1468 void ip_vs_stop_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats); 1469 void ip_vs_zero_estimator(struct ip_vs_stats *stats); 1470 void ip_vs_read_estimator(struct ip_vs_kstats *dst, struct ip_vs_stats *stats); 1471 1472 /* Various IPVS packet transmitters (from ip_vs_xmit.c) */ 1473 int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1474 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1475 int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1476 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1477 int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1478 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1479 int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1480 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1481 int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1482 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1483 int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp, 1484 struct ip_vs_protocol *pp, int offset, 1485 unsigned int hooknum, struct ip_vs_iphdr *iph); 1486 void ip_vs_dest_dst_rcu_free(struct rcu_head *head); 1487 1488 #ifdef CONFIG_IP_VS_IPV6 1489 int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1490 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1491 int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1492 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1493 int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1494 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1495 int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1496 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph); 1497 int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp, 1498 struct ip_vs_protocol *pp, int offset, 1499 unsigned int hooknum, struct ip_vs_iphdr *iph); 1500 #endif 1501 1502 #ifdef CONFIG_SYSCTL 1503 /* This is a simple mechanism to ignore packets when 1504 * we are loaded. Just set ip_vs_drop_rate to 'n' and 1505 * we start to drop 1/rate of the packets 1506 */ 1507 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) 1508 { 1509 if (!ipvs->drop_rate) 1510 return 0; 1511 if (--ipvs->drop_counter > 0) 1512 return 0; 1513 ipvs->drop_counter = ipvs->drop_rate; 1514 return 1; 1515 } 1516 #else 1517 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; } 1518 #endif 1519 1520 #ifdef CONFIG_SYSCTL 1521 /* Enqueue delayed work for expiring no dest connections 1522 * Only run when sysctl_expire_nodest=1 1523 */ 1524 static inline void ip_vs_enqueue_expire_nodest_conns(struct netns_ipvs *ipvs) 1525 { 1526 if (sysctl_expire_nodest_conn(ipvs)) 1527 queue_delayed_work(system_long_wq, 1528 &ipvs->expire_nodest_conn_work, 1); 1529 } 1530 1531 void ip_vs_expire_nodest_conn_flush(struct netns_ipvs *ipvs); 1532 #else 1533 static inline void ip_vs_enqueue_expire_nodest_conns(struct netns_ipvs *ipvs) {} 1534 #endif 1535 1536 #define IP_VS_DFWD_METHOD(dest) (atomic_read(&(dest)->conn_flags) & \ 1537 IP_VS_CONN_F_FWD_MASK) 1538 1539 /* ip_vs_fwd_tag returns the forwarding tag of the connection */ 1540 #define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK) 1541 1542 static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp) 1543 { 1544 char fwd; 1545 1546 switch (IP_VS_FWD_METHOD(cp)) { 1547 case IP_VS_CONN_F_MASQ: 1548 fwd = 'M'; break; 1549 case IP_VS_CONN_F_LOCALNODE: 1550 fwd = 'L'; break; 1551 case IP_VS_CONN_F_TUNNEL: 1552 fwd = 'T'; break; 1553 case IP_VS_CONN_F_DROUTE: 1554 fwd = 'R'; break; 1555 case IP_VS_CONN_F_BYPASS: 1556 fwd = 'B'; break; 1557 default: 1558 fwd = '?'; break; 1559 } 1560 return fwd; 1561 } 1562 1563 void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp, 1564 struct ip_vs_conn *cp, int dir); 1565 1566 #ifdef CONFIG_IP_VS_IPV6 1567 void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp, 1568 struct ip_vs_conn *cp, int dir); 1569 #endif 1570 1571 __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset); 1572 1573 static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum) 1574 { 1575 __be32 diff[2] = { ~old, new }; 1576 1577 return csum_partial(diff, sizeof(diff), oldsum); 1578 } 1579 1580 #ifdef CONFIG_IP_VS_IPV6 1581 static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new, 1582 __wsum oldsum) 1583 { 1584 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0], 1585 new[3], new[2], new[1], new[0] }; 1586 1587 return csum_partial(diff, sizeof(diff), oldsum); 1588 } 1589 #endif 1590 1591 static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum) 1592 { 1593 __be16 diff[2] = { ~old, new }; 1594 1595 return csum_partial(diff, sizeof(diff), oldsum); 1596 } 1597 1598 /* Forget current conntrack (unconfirmed) and attach notrack entry */ 1599 static inline void ip_vs_notrack(struct sk_buff *skb) 1600 { 1601 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) 1602 enum ip_conntrack_info ctinfo; 1603 struct nf_conn *ct = nf_ct_get(skb, &ctinfo); 1604 1605 if (ct) { 1606 nf_conntrack_put(&ct->ct_general); 1607 nf_ct_set(skb, NULL, IP_CT_UNTRACKED); 1608 } 1609 #endif 1610 } 1611 1612 #ifdef CONFIG_IP_VS_NFCT 1613 /* Netfilter connection tracking 1614 * (from ip_vs_nfct.c) 1615 */ 1616 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs) 1617 { 1618 #ifdef CONFIG_SYSCTL 1619 return ipvs->sysctl_conntrack; 1620 #else 1621 return 0; 1622 #endif 1623 } 1624 1625 void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp, 1626 int outin); 1627 int ip_vs_confirm_conntrack(struct sk_buff *skb); 1628 void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct, 1629 struct ip_vs_conn *cp, u_int8_t proto, 1630 const __be16 port, int from_rs); 1631 void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp); 1632 1633 #else 1634 1635 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs) 1636 { 1637 return 0; 1638 } 1639 1640 static inline void ip_vs_update_conntrack(struct sk_buff *skb, 1641 struct ip_vs_conn *cp, int outin) 1642 { 1643 } 1644 1645 static inline int ip_vs_confirm_conntrack(struct sk_buff *skb) 1646 { 1647 return NF_ACCEPT; 1648 } 1649 1650 static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp) 1651 { 1652 } 1653 #endif /* CONFIG_IP_VS_NFCT */ 1654 1655 /* Using old conntrack that can not be redirected to another real server? */ 1656 static inline bool ip_vs_conn_uses_old_conntrack(struct ip_vs_conn *cp, 1657 struct sk_buff *skb) 1658 { 1659 #ifdef CONFIG_IP_VS_NFCT 1660 enum ip_conntrack_info ctinfo; 1661 struct nf_conn *ct; 1662 1663 ct = nf_ct_get(skb, &ctinfo); 1664 if (ct && nf_ct_is_confirmed(ct)) 1665 return true; 1666 #endif 1667 return false; 1668 } 1669 1670 static inline int ip_vs_register_conntrack(struct ip_vs_service *svc) 1671 { 1672 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 1673 int afmask = (svc->af == AF_INET6) ? 2 : 1; 1674 int ret = 0; 1675 1676 if (!(svc->conntrack_afmask & afmask)) { 1677 ret = nf_ct_netns_get(svc->ipvs->net, svc->af); 1678 if (ret >= 0) 1679 svc->conntrack_afmask |= afmask; 1680 } 1681 return ret; 1682 #else 1683 return 0; 1684 #endif 1685 } 1686 1687 static inline void ip_vs_unregister_conntrack(struct ip_vs_service *svc) 1688 { 1689 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 1690 int afmask = (svc->af == AF_INET6) ? 2 : 1; 1691 1692 if (svc->conntrack_afmask & afmask) { 1693 nf_ct_netns_put(svc->ipvs->net, svc->af); 1694 svc->conntrack_afmask &= ~afmask; 1695 } 1696 #endif 1697 } 1698 1699 int ip_vs_register_hooks(struct netns_ipvs *ipvs, unsigned int af); 1700 void ip_vs_unregister_hooks(struct netns_ipvs *ipvs, unsigned int af); 1701 1702 static inline int 1703 ip_vs_dest_conn_overhead(struct ip_vs_dest *dest) 1704 { 1705 /* We think the overhead of processing active connections is 256 1706 * times higher than that of inactive connections in average. (This 1707 * 256 times might not be accurate, we will change it later) We 1708 * use the following formula to estimate the overhead now: 1709 * dest->activeconns*256 + dest->inactconns 1710 */ 1711 return (atomic_read(&dest->activeconns) << 8) + 1712 atomic_read(&dest->inactconns); 1713 } 1714 1715 #endif /* _NET_IP_VS_H */ 1716