1 /* SPDX-License-Identifier: GPL-2.0 */
3 * data structure and functionality definitions
9 #include <linux/ip_vs.h> /* definitions shared with userland */
11 #include <asm/types.h> /* for __uXX types */
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>
19 #include <linux/compiler.h>
20 #include <linux/timer.h>
21 #include <linux/bug.h>
23 #include <net/checksum.h>
24 #include <linux/netfilter.h> /* for union nf_inet_addr */
26 #include <linux/ipv6.h> /* for struct ipv6hdr */
28 #if IS_ENABLED(CONFIG_IP_VS_IPV6)
29 #include <linux/netfilter_ipv6/ip6_tables.h>
31 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
32 #include <net/netfilter/nf_conntrack.h>
34 #include <net/net_namespace.h> /* Netw namespace */
36 #define IP_VS_HDR_INVERSE 1
37 #define IP_VS_HDR_ICMP 2
39 /* Generic access of ipvs struct */
40 static inline struct netns_ipvs *net_ipvs(struct net* net)
45 /* Connections' size value needed by ip_vs_ctl.c */
46 extern int ip_vs_conn_tab_size;
49 int hdr_flags; /* ipvs flags */
50 __u32 off; /* Where IP or IPv4 header starts */
51 __u32 len; /* IPv4 simply where L4 starts
52 * IPv6 where L4 Transport Header starts */
53 __u16 fragoffs; /* IPv6 fragment offset, 0 if first frag (or not frag)*/
56 union nf_inet_addr saddr;
57 union nf_inet_addr daddr;
60 static inline void *frag_safe_skb_hp(const struct sk_buff *skb, int offset,
61 int len, void *buffer)
63 return skb_header_pointer(skb, offset, len, buffer);
66 /* This function handles filling *ip_vs_iphdr, both for IPv4 and IPv6.
67 * IPv6 requires some extra work, as finding proper header position,
68 * depend on the IPv6 extension headers.
71 ip_vs_fill_iph_skb_off(int af, const struct sk_buff *skb, int offset,
72 int hdr_flags, struct ip_vs_iphdr *iphdr)
74 iphdr->hdr_flags = hdr_flags;
77 #ifdef CONFIG_IP_VS_IPV6
80 const struct ipv6hdr *iph = skb_header_pointer(
81 skb, offset, sizeof(_iph), &_iph);
85 iphdr->saddr.in6 = iph->saddr;
86 iphdr->daddr.in6 = iph->daddr;
87 /* ipv6_find_hdr() updates len, flags */
90 iphdr->protocol = ipv6_find_hdr(skb, &iphdr->len, -1,
93 if (iphdr->protocol < 0)
99 const struct iphdr *iph = skb_header_pointer(
100 skb, offset, sizeof(_iph), &_iph);
104 iphdr->len = offset + iph->ihl * 4;
106 iphdr->protocol = iph->protocol;
107 iphdr->saddr.ip = iph->saddr;
108 iphdr->daddr.ip = iph->daddr;
115 ip_vs_fill_iph_skb_icmp(int af, const struct sk_buff *skb, int offset,
116 bool inverse, struct ip_vs_iphdr *iphdr)
118 int hdr_flags = IP_VS_HDR_ICMP;
121 hdr_flags |= IP_VS_HDR_INVERSE;
123 return ip_vs_fill_iph_skb_off(af, skb, offset, hdr_flags, iphdr);
127 ip_vs_fill_iph_skb(int af, const struct sk_buff *skb, bool inverse,
128 struct ip_vs_iphdr *iphdr)
133 hdr_flags |= IP_VS_HDR_INVERSE;
135 return ip_vs_fill_iph_skb_off(af, skb, skb_network_offset(skb),
140 ip_vs_iph_inverse(const struct ip_vs_iphdr *iph)
142 return !!(iph->hdr_flags & IP_VS_HDR_INVERSE);
146 ip_vs_iph_icmp(const struct ip_vs_iphdr *iph)
148 return !!(iph->hdr_flags & IP_VS_HDR_ICMP);
151 static inline void ip_vs_addr_copy(int af, union nf_inet_addr *dst,
152 const union nf_inet_addr *src)
154 #ifdef CONFIG_IP_VS_IPV6
162 static inline void ip_vs_addr_set(int af, union nf_inet_addr *dst,
163 const union nf_inet_addr *src)
165 #ifdef CONFIG_IP_VS_IPV6
166 if (af == AF_INET6) {
177 static inline int ip_vs_addr_equal(int af, const union nf_inet_addr *a,
178 const union nf_inet_addr *b)
180 #ifdef CONFIG_IP_VS_IPV6
182 return ipv6_addr_equal(&a->in6, &b->in6);
184 return a->ip == b->ip;
187 #ifdef CONFIG_IP_VS_DEBUG
188 #include <linux/net.h>
190 int ip_vs_get_debug_level(void);
192 static inline const char *ip_vs_dbg_addr(int af, char *buf, size_t buf_len,
193 const union nf_inet_addr *addr,
197 #ifdef CONFIG_IP_VS_IPV6
199 len = snprintf(&buf[*idx], buf_len - *idx, "[%pI6c]",
203 len = snprintf(&buf[*idx], buf_len - *idx, "%pI4",
207 BUG_ON(*idx > buf_len + 1);
208 return &buf[*idx - len];
211 #define IP_VS_DBG_BUF(level, msg, ...) \
213 char ip_vs_dbg_buf[160]; \
214 int ip_vs_dbg_idx = 0; \
215 if (level <= ip_vs_get_debug_level()) \
216 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
218 #define IP_VS_ERR_BUF(msg...) \
220 char ip_vs_dbg_buf[160]; \
221 int ip_vs_dbg_idx = 0; \
225 /* Only use from within IP_VS_DBG_BUF() or IP_VS_ERR_BUF macros */
226 #define IP_VS_DBG_ADDR(af, addr) \
227 ip_vs_dbg_addr(af, ip_vs_dbg_buf, \
228 sizeof(ip_vs_dbg_buf), addr, \
231 #define IP_VS_DBG(level, msg, ...) \
233 if (level <= ip_vs_get_debug_level()) \
234 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
236 #define IP_VS_DBG_RL(msg, ...) \
238 if (net_ratelimit()) \
239 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
241 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) \
243 if (level <= ip_vs_get_debug_level()) \
244 pp->debug_packet(af, pp, skb, ofs, msg); \
246 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) \
248 if (level <= ip_vs_get_debug_level() && \
250 pp->debug_packet(af, pp, skb, ofs, msg); \
252 #else /* NO DEBUGGING at ALL */
253 #define IP_VS_DBG_BUF(level, msg...) do {} while (0)
254 #define IP_VS_ERR_BUF(msg...) do {} while (0)
255 #define IP_VS_DBG(level, msg...) do {} while (0)
256 #define IP_VS_DBG_RL(msg...) do {} while (0)
257 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
258 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
261 #define IP_VS_BUG() BUG()
262 #define IP_VS_ERR_RL(msg, ...) \
264 if (net_ratelimit()) \
265 pr_err(msg, ##__VA_ARGS__); \
268 #ifdef CONFIG_IP_VS_DEBUG
269 #define EnterFunction(level) \
271 if (level <= ip_vs_get_debug_level()) \
273 pr_fmt("Enter: %s, %s line %i\n"), \
274 __func__, __FILE__, __LINE__); \
276 #define LeaveFunction(level) \
278 if (level <= ip_vs_get_debug_level()) \
280 pr_fmt("Leave: %s, %s line %i\n"), \
281 __func__, __FILE__, __LINE__); \
284 #define EnterFunction(level) do {} while (0)
285 #define LeaveFunction(level) do {} while (0)
288 /* The port number of FTP service (in network order). */
289 #define FTPPORT cpu_to_be16(21)
290 #define FTPDATA cpu_to_be16(20)
292 /* TCP State Values */
294 IP_VS_TCP_S_NONE = 0,
295 IP_VS_TCP_S_ESTABLISHED,
296 IP_VS_TCP_S_SYN_SENT,
297 IP_VS_TCP_S_SYN_RECV,
298 IP_VS_TCP_S_FIN_WAIT,
299 IP_VS_TCP_S_TIME_WAIT,
301 IP_VS_TCP_S_CLOSE_WAIT,
302 IP_VS_TCP_S_LAST_ACK,
308 /* UDP State Values */
314 /* ICMP State Values */
320 /* SCTP State Values */
321 enum ip_vs_sctp_states {
325 IP_VS_SCTP_S_COOKIE_SENT,
326 IP_VS_SCTP_S_COOKIE_REPLIED,
327 IP_VS_SCTP_S_COOKIE_WAIT,
329 IP_VS_SCTP_S_COOKIE_ECHOED,
330 IP_VS_SCTP_S_ESTABLISHED,
331 IP_VS_SCTP_S_SHUTDOWN_SENT,
332 IP_VS_SCTP_S_SHUTDOWN_RECEIVED,
333 IP_VS_SCTP_S_SHUTDOWN_ACK_SENT,
334 IP_VS_SCTP_S_REJECTED,
339 /* Connection templates use bits from state */
340 #define IP_VS_CTPL_S_NONE 0x0000
341 #define IP_VS_CTPL_S_ASSURED 0x0001
342 #define IP_VS_CTPL_S_LAST 0x0002
344 /* Delta sequence info structure
345 * Each ip_vs_conn has 2 (output AND input seq. changes).
346 * Only used in the VS/NAT.
349 __u32 init_seq; /* Add delta from this seq */
350 __u32 delta; /* Delta in sequence numbers */
351 __u32 previous_delta; /* Delta in sequence numbers
352 * before last resized pkt */
355 /* counters per cpu */
356 struct ip_vs_counters {
357 __u64 conns; /* connections scheduled */
358 __u64 inpkts; /* incoming packets */
359 __u64 outpkts; /* outgoing packets */
360 __u64 inbytes; /* incoming bytes */
361 __u64 outbytes; /* outgoing bytes */
364 struct ip_vs_cpu_stats {
365 struct ip_vs_counters cnt;
366 struct u64_stats_sync syncp;
369 /* IPVS statistics objects */
370 struct ip_vs_estimator {
371 struct list_head list;
387 * IPVS statistics object, 64-bit kernel version of struct ip_vs_stats_user
389 struct ip_vs_kstats {
390 u64 conns; /* connections scheduled */
391 u64 inpkts; /* incoming packets */
392 u64 outpkts; /* outgoing packets */
393 u64 inbytes; /* incoming bytes */
394 u64 outbytes; /* outgoing bytes */
396 u64 cps; /* current connection rate */
397 u64 inpps; /* current in packet rate */
398 u64 outpps; /* current out packet rate */
399 u64 inbps; /* current in byte rate */
400 u64 outbps; /* current out byte rate */
404 struct ip_vs_kstats kstats; /* kernel statistics */
405 struct ip_vs_estimator est; /* estimator */
406 struct ip_vs_cpu_stats __percpu *cpustats; /* per cpu counters */
407 spinlock_t lock; /* spin lock */
408 struct ip_vs_kstats kstats0; /* reset values */
416 struct ip_vs_proto_data;
418 struct ip_vs_protocol {
419 struct ip_vs_protocol *next;
425 void (*init)(struct ip_vs_protocol *pp);
427 void (*exit)(struct ip_vs_protocol *pp);
429 int (*init_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
431 void (*exit_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
433 int (*conn_schedule)(struct netns_ipvs *ipvs,
434 int af, struct sk_buff *skb,
435 struct ip_vs_proto_data *pd,
436 int *verdict, struct ip_vs_conn **cpp,
437 struct ip_vs_iphdr *iph);
440 (*conn_in_get)(struct netns_ipvs *ipvs,
442 const struct sk_buff *skb,
443 const struct ip_vs_iphdr *iph);
446 (*conn_out_get)(struct netns_ipvs *ipvs,
448 const struct sk_buff *skb,
449 const struct ip_vs_iphdr *iph);
451 int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
452 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
454 int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
455 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
457 const char *(*state_name)(int state);
459 void (*state_transition)(struct ip_vs_conn *cp, int direction,
460 const struct sk_buff *skb,
461 struct ip_vs_proto_data *pd);
463 int (*register_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
465 void (*unregister_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
467 int (*app_conn_bind)(struct ip_vs_conn *cp);
469 void (*debug_packet)(int af, struct ip_vs_protocol *pp,
470 const struct sk_buff *skb,
474 void (*timeout_change)(struct ip_vs_proto_data *pd, int flags);
477 /* protocol data per netns */
478 struct ip_vs_proto_data {
479 struct ip_vs_proto_data *next;
480 struct ip_vs_protocol *pp;
481 int *timeout_table; /* protocol timeout table */
482 atomic_t appcnt; /* counter of proto app incs. */
483 struct tcp_states_t *tcp_state_table;
486 struct ip_vs_protocol *ip_vs_proto_get(unsigned short proto);
487 struct ip_vs_proto_data *ip_vs_proto_data_get(struct netns_ipvs *ipvs,
488 unsigned short proto);
490 struct ip_vs_conn_param {
491 struct netns_ipvs *ipvs;
492 const union nf_inet_addr *caddr;
493 const union nf_inet_addr *vaddr;
499 const struct ip_vs_pe *pe;
504 /* IP_VS structure allocated for each dynamically scheduled connection */
506 struct hlist_node c_list; /* hashed list heads */
507 /* Protocol, addresses and port numbers */
511 u16 af; /* address family */
512 union nf_inet_addr caddr; /* client address */
513 union nf_inet_addr vaddr; /* virtual address */
514 union nf_inet_addr daddr; /* destination address */
515 volatile __u32 flags; /* status flags */
516 __u16 protocol; /* Which protocol (TCP/UDP) */
517 __u16 daf; /* Address family of the dest */
518 struct netns_ipvs *ipvs;
520 /* counter and timer */
521 refcount_t refcnt; /* reference count */
522 struct timer_list timer; /* Expiration timer */
523 volatile unsigned long timeout; /* timeout */
525 /* Flags and state transition */
526 spinlock_t lock; /* lock for state transition */
527 volatile __u16 state; /* state info */
528 volatile __u16 old_state; /* old state, to be used for
529 * state transition triggerd
532 __u32 fwmark; /* Fire wall mark from skb */
533 unsigned long sync_endtime; /* jiffies + sent_retries */
535 /* Control members */
536 struct ip_vs_conn *control; /* Master control connection */
537 atomic_t n_control; /* Number of controlled ones */
538 struct ip_vs_dest *dest; /* real server */
539 atomic_t in_pkts; /* incoming packet counter */
541 /* Packet transmitter for different forwarding methods. If it
542 * mangles the packet, it must return NF_DROP or better NF_STOLEN,
543 * otherwise this must be changed to a sk_buff **.
544 * NF_ACCEPT can be returned when destination is local.
546 int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp,
547 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
549 /* Note: we can group the following members into a structure,
550 * in order to save more space, and the following members are
551 * only used in VS/NAT anyway
553 struct ip_vs_app *app; /* bound ip_vs_app object */
554 void *app_data; /* Application private data */
555 struct ip_vs_seq in_seq; /* incoming seq. struct */
556 struct ip_vs_seq out_seq; /* outgoing seq. struct */
558 const struct ip_vs_pe *pe;
562 struct rcu_head rcu_head;
565 /* Extended internal versions of struct ip_vs_service_user and ip_vs_dest_user
568 * We need these to conveniently pass around service and destination
569 * options, but unfortunately, we also need to keep the old definitions to
570 * maintain userspace backwards compatibility for the setsockopt interface.
572 struct ip_vs_service_user_kern {
573 /* virtual service addresses */
576 union nf_inet_addr addr; /* virtual ip address */
578 u32 fwmark; /* firwall mark of service */
580 /* virtual service options */
583 unsigned int flags; /* virtual service flags */
584 unsigned int timeout; /* persistent timeout in sec */
585 __be32 netmask; /* persistent netmask or plen */
589 struct ip_vs_dest_user_kern {
590 /* destination server address */
591 union nf_inet_addr addr;
594 /* real server options */
595 unsigned int conn_flags; /* connection flags */
596 int weight; /* destination weight */
598 /* thresholds for active connections */
599 u32 u_threshold; /* upper threshold */
600 u32 l_threshold; /* lower threshold */
602 /* Address family of addr */
605 u16 tun_type; /* tunnel type */
606 __be16 tun_port; /* tunnel port */
607 u16 tun_flags; /* tunnel flags */
612 * The information about the virtual service offered to the net and the
613 * forwarding entries.
615 struct ip_vs_service {
616 struct hlist_node s_list; /* for normal service table */
617 struct hlist_node f_list; /* for fwmark-based service table */
618 atomic_t refcnt; /* reference counter */
620 u16 af; /* address family */
621 __u16 protocol; /* which protocol (TCP/UDP) */
622 union nf_inet_addr addr; /* IP address for virtual service */
623 __be16 port; /* port number for the service */
624 __u32 fwmark; /* firewall mark of the service */
625 unsigned int flags; /* service status flags */
626 unsigned int timeout; /* persistent timeout in ticks */
627 __be32 netmask; /* grouping granularity, mask/plen */
628 struct netns_ipvs *ipvs;
630 struct list_head destinations; /* real server d-linked list */
631 __u32 num_dests; /* number of servers */
632 struct ip_vs_stats stats; /* statistics for the service */
635 struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */
636 spinlock_t sched_lock; /* lock sched_data */
637 void *sched_data; /* scheduler application data */
639 /* alternate persistence engine */
640 struct ip_vs_pe __rcu *pe;
641 int conntrack_afmask;
643 struct rcu_head rcu_head;
646 /* Information for cached dst */
647 struct ip_vs_dest_dst {
648 struct dst_entry *dst_cache; /* destination cache entry */
650 union nf_inet_addr dst_saddr;
651 struct rcu_head rcu_head;
654 /* The real server destination forwarding entry with ip address, port number,
658 struct list_head n_list; /* for the dests in the service */
659 struct hlist_node d_list; /* for table with all the dests */
661 u16 af; /* address family */
662 __be16 port; /* port number of the server */
663 union nf_inet_addr addr; /* IP address of the server */
664 volatile unsigned int flags; /* dest status flags */
665 atomic_t conn_flags; /* flags to copy to conn */
666 atomic_t weight; /* server weight */
667 atomic_t last_weight; /* server latest weight */
668 __u16 tun_type; /* tunnel type */
669 __be16 tun_port; /* tunnel port */
670 __u16 tun_flags; /* tunnel flags */
672 refcount_t refcnt; /* reference counter */
673 struct ip_vs_stats stats; /* statistics */
674 unsigned long idle_start; /* start time, jiffies */
676 /* connection counters and thresholds */
677 atomic_t activeconns; /* active connections */
678 atomic_t inactconns; /* inactive connections */
679 atomic_t persistconns; /* persistent connections */
680 __u32 u_threshold; /* upper threshold */
681 __u32 l_threshold; /* lower threshold */
683 /* for destination cache */
684 spinlock_t dst_lock; /* lock of dst_cache */
685 struct ip_vs_dest_dst __rcu *dest_dst; /* cached dst info */
687 /* for virtual service */
688 struct ip_vs_service __rcu *svc; /* service it belongs to */
689 __u16 protocol; /* which protocol (TCP/UDP) */
690 __be16 vport; /* virtual port number */
691 union nf_inet_addr vaddr; /* virtual IP address */
692 __u32 vfwmark; /* firewall mark of service */
694 struct list_head t_list; /* in dest_trash */
695 unsigned int in_rs_table:1; /* we are in rs_table */
698 /* The scheduler object */
699 struct ip_vs_scheduler {
700 struct list_head n_list; /* d-linked list head */
701 char *name; /* scheduler name */
702 atomic_t refcnt; /* reference counter */
703 struct module *module; /* THIS_MODULE/NULL */
705 /* scheduler initializing service */
706 int (*init_service)(struct ip_vs_service *svc);
707 /* scheduling service finish */
708 void (*done_service)(struct ip_vs_service *svc);
710 int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
711 /* dest is unlinked */
712 int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
713 /* dest is updated */
714 int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
716 /* selecting a server from the given service */
717 struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc,
718 const struct sk_buff *skb,
719 struct ip_vs_iphdr *iph);
722 /* The persistence engine object */
724 struct list_head n_list; /* d-linked list head */
725 char *name; /* scheduler name */
726 atomic_t refcnt; /* reference counter */
727 struct module *module; /* THIS_MODULE/NULL */
729 /* get the connection template, if any */
730 int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb);
731 bool (*ct_match)(const struct ip_vs_conn_param *p,
732 struct ip_vs_conn *ct);
733 u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval,
735 int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf);
736 /* create connections for real-server outgoing packets */
737 struct ip_vs_conn* (*conn_out)(struct ip_vs_service *svc,
738 struct ip_vs_dest *dest,
740 const struct ip_vs_iphdr *iph,
741 __be16 dport, __be16 cport);
744 /* The application module object (a.k.a. app incarnation) */
746 struct list_head a_list; /* member in app list */
747 int type; /* IP_VS_APP_TYPE_xxx */
748 char *name; /* application module name */
750 struct module *module; /* THIS_MODULE/NULL */
751 struct list_head incs_list; /* list of incarnations */
753 /* members for application incarnations */
754 struct list_head p_list; /* member in proto app list */
755 struct ip_vs_app *app; /* its real application */
756 __be16 port; /* port number in net order */
757 atomic_t usecnt; /* usage counter */
758 struct rcu_head rcu_head;
760 /* output hook: Process packet in inout direction, diff set for TCP.
761 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
762 * 2=Mangled but checksum was not updated
764 int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *,
765 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh);
767 /* input hook: Process packet in outin direction, diff set for TCP.
768 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
769 * 2=Mangled but checksum was not updated
771 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *,
772 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh);
774 /* ip_vs_app initializer */
775 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *);
777 /* ip_vs_app finish */
778 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *);
782 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *,
783 struct ip_vs_protocol *);
785 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *);
791 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app,
792 int *verdict, struct ip_vs_conn **cpp);
795 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app,
796 const struct iphdr *iph, int inverse);
799 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app,
800 const struct iphdr *iph, int inverse);
802 int (*state_transition)(struct ip_vs_conn *cp, int direction,
803 const struct sk_buff *skb,
804 struct ip_vs_app *app);
806 void (*timeout_change)(struct ip_vs_app *app, int flags);
809 struct ipvs_master_sync_state {
810 struct list_head sync_queue;
811 struct ip_vs_sync_buff *sync_buff;
812 unsigned long sync_queue_len;
813 unsigned int sync_queue_delay;
814 struct delayed_work master_wakeup_work;
815 struct netns_ipvs *ipvs;
818 struct ip_vs_sync_thread_data;
820 /* How much time to keep dests in trash */
821 #define IP_VS_DEST_TRASH_PERIOD (120 * HZ)
823 struct ipvs_sync_daemon_cfg {
824 union nf_inet_addr mcast_group;
830 /* multicast interface name */
831 char mcast_ifn[IP_VS_IFNAME_MAXLEN];
834 /* IPVS in network namespace */
836 int gen; /* Generation */
837 int enable; /* enable like nf_hooks do */
838 /* Hash table: for real service lookups */
839 #define IP_VS_RTAB_BITS 4
840 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
841 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
843 struct hlist_head rs_table[IP_VS_RTAB_SIZE];
845 struct list_head app_list;
847 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */
848 struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE];
849 /* ip_vs_proto_tcp */
850 #ifdef CONFIG_IP_VS_PROTO_TCP
851 #define TCP_APP_TAB_BITS 4
852 #define TCP_APP_TAB_SIZE (1 << TCP_APP_TAB_BITS)
853 #define TCP_APP_TAB_MASK (TCP_APP_TAB_SIZE - 1)
854 struct list_head tcp_apps[TCP_APP_TAB_SIZE];
856 /* ip_vs_proto_udp */
857 #ifdef CONFIG_IP_VS_PROTO_UDP
858 #define UDP_APP_TAB_BITS 4
859 #define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS)
860 #define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1)
861 struct list_head udp_apps[UDP_APP_TAB_SIZE];
863 /* ip_vs_proto_sctp */
864 #ifdef CONFIG_IP_VS_PROTO_SCTP
865 #define SCTP_APP_TAB_BITS 4
866 #define SCTP_APP_TAB_SIZE (1 << SCTP_APP_TAB_BITS)
867 #define SCTP_APP_TAB_MASK (SCTP_APP_TAB_SIZE - 1)
868 /* Hash table for SCTP application incarnations */
869 struct list_head sctp_apps[SCTP_APP_TAB_SIZE];
872 atomic_t conn_count; /* connection counter */
875 struct ip_vs_stats tot_stats; /* Statistics & est. */
877 int num_services; /* no of virtual services */
878 int num_services6; /* IPv6 virtual services */
880 /* Trash for destinations */
881 struct list_head dest_trash;
882 spinlock_t dest_trash_lock;
883 struct timer_list dest_trash_timer; /* expiration timer */
884 /* Service counters */
885 atomic_t ftpsvc_counter;
886 atomic_t nullsvc_counter;
887 atomic_t conn_out_counter;
890 /* delayed work for expiring no dest connections */
891 struct delayed_work expire_nodest_conn_work;
892 /* 1/rate drop and drop-entry variables */
893 struct delayed_work defense_work; /* Work handler */
899 spinlock_t dropentry_lock; /* drop entry handling */
900 spinlock_t droppacket_lock; /* drop packet handling */
901 spinlock_t securetcp_lock; /* state and timeout tables */
904 struct ctl_table_header *sysctl_hdr;
905 struct ctl_table *sysctl_tbl;
908 /* sysctl variables */
909 int sysctl_amemthresh;
910 int sysctl_am_droprate;
911 int sysctl_drop_entry;
912 int sysctl_drop_packet;
913 int sysctl_secure_tcp;
914 #ifdef CONFIG_IP_VS_NFCT
915 int sysctl_conntrack;
917 int sysctl_snat_reroute;
919 int sysctl_sync_ports;
920 int sysctl_sync_persist_mode;
921 unsigned long sysctl_sync_qlen_max;
922 int sysctl_sync_sock_size;
923 int sysctl_cache_bypass;
924 int sysctl_expire_nodest_conn;
925 int sysctl_sloppy_tcp;
926 int sysctl_sloppy_sctp;
927 int sysctl_expire_quiescent_template;
928 int sysctl_sync_threshold[2];
929 unsigned int sysctl_sync_refresh_period;
930 int sysctl_sync_retries;
931 int sysctl_nat_icmp_send;
932 int sysctl_pmtu_disc;
933 int sysctl_backup_only;
934 int sysctl_conn_reuse_mode;
935 int sysctl_schedule_icmp;
936 int sysctl_ignore_tunneled;
939 int sysctl_lblc_expiration;
940 struct ctl_table_header *lblc_ctl_header;
941 struct ctl_table *lblc_ctl_table;
943 int sysctl_lblcr_expiration;
944 struct ctl_table_header *lblcr_ctl_header;
945 struct ctl_table *lblcr_ctl_table;
947 struct list_head est_list; /* estimator list */
949 struct timer_list est_timer; /* Estimation timer */
951 spinlock_t sync_lock;
952 struct ipvs_master_sync_state *ms;
953 spinlock_t sync_buff_lock;
954 struct ip_vs_sync_thread_data *master_tinfo;
955 struct ip_vs_sync_thread_data *backup_tinfo;
957 volatile int sync_state;
958 struct mutex sync_mutex;
959 struct ipvs_sync_daemon_cfg mcfg; /* Master Configuration */
960 struct ipvs_sync_daemon_cfg bcfg; /* Backup Configuration */
961 /* net name space ptr */
962 struct net *net; /* Needed by timer routines */
963 /* Number of heterogeneous destinations, needed becaus heterogeneous
964 * are not supported when synchronization is enabled.
966 unsigned int mixed_address_family_dests;
967 unsigned int hooks_afmask; /* &1=AF_INET, &2=AF_INET6 */
970 #define DEFAULT_SYNC_THRESHOLD 3
971 #define DEFAULT_SYNC_PERIOD 50
972 #define DEFAULT_SYNC_VER 1
973 #define DEFAULT_SLOPPY_TCP 0
974 #define DEFAULT_SLOPPY_SCTP 0
975 #define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ)
976 #define DEFAULT_SYNC_RETRIES 0
977 #define IPVS_SYNC_WAKEUP_RATE 8
978 #define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4)
979 #define IPVS_SYNC_SEND_DELAY (HZ / 50)
980 #define IPVS_SYNC_CHECK_PERIOD HZ
981 #define IPVS_SYNC_FLUSH_TIME (HZ * 2)
982 #define IPVS_SYNC_PORTS_MAX (1 << 6)
986 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
988 return ipvs->sysctl_sync_threshold[0];
991 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
993 return READ_ONCE(ipvs->sysctl_sync_threshold[1]);
996 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
998 return READ_ONCE(ipvs->sysctl_sync_refresh_period);
1001 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1003 return ipvs->sysctl_sync_retries;
1006 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1008 return ipvs->sysctl_sync_ver;
1011 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1013 return ipvs->sysctl_sloppy_tcp;
1016 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1018 return ipvs->sysctl_sloppy_sctp;
1021 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1023 return READ_ONCE(ipvs->sysctl_sync_ports);
1026 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1028 return ipvs->sysctl_sync_persist_mode;
1031 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1033 return ipvs->sysctl_sync_qlen_max;
1036 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1038 return ipvs->sysctl_sync_sock_size;
1041 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1043 return ipvs->sysctl_pmtu_disc;
1046 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1048 return ipvs->sync_state & IP_VS_STATE_BACKUP &&
1049 ipvs->sysctl_backup_only;
1052 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1054 return ipvs->sysctl_conn_reuse_mode;
1057 static inline int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs)
1059 return ipvs->sysctl_expire_nodest_conn;
1062 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1064 return ipvs->sysctl_schedule_icmp;
1067 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1069 return ipvs->sysctl_ignore_tunneled;
1072 static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1074 return ipvs->sysctl_cache_bypass;
1079 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1081 return DEFAULT_SYNC_THRESHOLD;
1084 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1086 return DEFAULT_SYNC_PERIOD;
1089 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1091 return DEFAULT_SYNC_REFRESH_PERIOD;
1094 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1096 return DEFAULT_SYNC_RETRIES & 3;
1099 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1101 return DEFAULT_SYNC_VER;
1104 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1106 return DEFAULT_SLOPPY_TCP;
1109 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1111 return DEFAULT_SLOPPY_SCTP;
1114 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1119 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1124 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1126 return IPVS_SYNC_QLEN_MAX;
1129 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1134 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1139 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1144 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1149 static inline int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs)
1154 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1159 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1164 static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1171 /* IPVS core functions
1172 * (from ip_vs_core.c)
1174 const char *ip_vs_proto_name(unsigned int proto);
1175 void ip_vs_init_hash_table(struct list_head *table, int rows);
1176 struct ip_vs_conn *ip_vs_new_conn_out(struct ip_vs_service *svc,
1177 struct ip_vs_dest *dest,
1178 struct sk_buff *skb,
1179 const struct ip_vs_iphdr *iph,
1182 #define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t)))
1184 #define IP_VS_APP_TYPE_FTP 1
1186 /* ip_vs_conn handling functions
1187 * (from ip_vs_conn.c)
1190 IP_VS_DIR_INPUT = 0,
1192 IP_VS_DIR_INPUT_ONLY,
1196 static inline void ip_vs_conn_fill_param(struct netns_ipvs *ipvs, int af, int protocol,
1197 const union nf_inet_addr *caddr,
1199 const union nf_inet_addr *vaddr,
1201 struct ip_vs_conn_param *p)
1205 p->protocol = protocol;
1214 struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p);
1215 struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p);
1217 struct ip_vs_conn * ip_vs_conn_in_get_proto(struct netns_ipvs *ipvs, int af,
1218 const struct sk_buff *skb,
1219 const struct ip_vs_iphdr *iph);
1221 struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p);
1223 struct ip_vs_conn * ip_vs_conn_out_get_proto(struct netns_ipvs *ipvs, int af,
1224 const struct sk_buff *skb,
1225 const struct ip_vs_iphdr *iph);
1227 /* Get reference to gain full access to conn.
1228 * By default, RCU read-side critical sections have access only to
1229 * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference.
1231 static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp)
1233 return refcount_inc_not_zero(&cp->refcnt);
1236 /* put back the conn without restarting its timer */
1237 static inline void __ip_vs_conn_put(struct ip_vs_conn *cp)
1239 smp_mb__before_atomic();
1240 refcount_dec(&cp->refcnt);
1242 void ip_vs_conn_put(struct ip_vs_conn *cp);
1243 void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport);
1245 struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, int dest_af,
1246 const union nf_inet_addr *daddr,
1247 __be16 dport, unsigned int flags,
1248 struct ip_vs_dest *dest, __u32 fwmark);
1249 void ip_vs_conn_expire_now(struct ip_vs_conn *cp);
1251 const char *ip_vs_state_name(const struct ip_vs_conn *cp);
1253 void ip_vs_tcp_conn_listen(struct ip_vs_conn *cp);
1254 int ip_vs_check_template(struct ip_vs_conn *ct, struct ip_vs_dest *cdest);
1255 void ip_vs_random_dropentry(struct netns_ipvs *ipvs);
1256 int ip_vs_conn_init(void);
1257 void ip_vs_conn_cleanup(void);
1259 static inline void ip_vs_control_del(struct ip_vs_conn *cp)
1261 struct ip_vs_conn *ctl_cp = cp->control;
1263 IP_VS_ERR_BUF("request control DEL for uncontrolled: "
1265 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1267 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1273 IP_VS_DBG_BUF(7, "DELeting control for: "
1274 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1275 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1277 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1278 ntohs(ctl_cp->cport));
1281 if (atomic_read(&ctl_cp->n_control) == 0) {
1282 IP_VS_ERR_BUF("BUG control DEL with n=0 : "
1284 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1286 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1291 atomic_dec(&ctl_cp->n_control);
1295 ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp)
1298 IP_VS_ERR_BUF("request control ADD for already controlled: "
1300 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1302 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1305 ip_vs_control_del(cp);
1308 IP_VS_DBG_BUF(7, "ADDing control for: "
1309 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1310 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1312 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1313 ntohs(ctl_cp->cport));
1315 cp->control = ctl_cp;
1316 atomic_inc(&ctl_cp->n_control);
1319 /* Mark our template as assured */
1321 ip_vs_control_assure_ct(struct ip_vs_conn *cp)
1323 struct ip_vs_conn *ct = cp->control;
1325 if (ct && !(ct->state & IP_VS_CTPL_S_ASSURED) &&
1326 (ct->flags & IP_VS_CONN_F_TEMPLATE))
1327 ct->state |= IP_VS_CTPL_S_ASSURED;
1330 /* IPVS netns init & cleanup functions */
1331 int ip_vs_estimator_net_init(struct netns_ipvs *ipvs);
1332 int ip_vs_control_net_init(struct netns_ipvs *ipvs);
1333 int ip_vs_protocol_net_init(struct netns_ipvs *ipvs);
1334 int ip_vs_app_net_init(struct netns_ipvs *ipvs);
1335 int ip_vs_conn_net_init(struct netns_ipvs *ipvs);
1336 int ip_vs_sync_net_init(struct netns_ipvs *ipvs);
1337 void ip_vs_conn_net_cleanup(struct netns_ipvs *ipvs);
1338 void ip_vs_app_net_cleanup(struct netns_ipvs *ipvs);
1339 void ip_vs_protocol_net_cleanup(struct netns_ipvs *ipvs);
1340 void ip_vs_control_net_cleanup(struct netns_ipvs *ipvs);
1341 void ip_vs_estimator_net_cleanup(struct netns_ipvs *ipvs);
1342 void ip_vs_sync_net_cleanup(struct netns_ipvs *ipvs);
1343 void ip_vs_service_nets_cleanup(struct list_head *net_list);
1345 /* IPVS application functions
1346 * (from ip_vs_app.c)
1348 #define IP_VS_APP_MAX_PORTS 8
1349 struct ip_vs_app *register_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1350 void unregister_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1351 int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1352 void ip_vs_unbind_app(struct ip_vs_conn *cp);
1353 int register_ip_vs_app_inc(struct netns_ipvs *ipvs, struct ip_vs_app *app, __u16 proto,
1355 int ip_vs_app_inc_get(struct ip_vs_app *inc);
1356 void ip_vs_app_inc_put(struct ip_vs_app *inc);
1358 int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb,
1359 struct ip_vs_iphdr *ipvsh);
1360 int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb,
1361 struct ip_vs_iphdr *ipvsh);
1363 int register_ip_vs_pe(struct ip_vs_pe *pe);
1364 int unregister_ip_vs_pe(struct ip_vs_pe *pe);
1365 struct ip_vs_pe *ip_vs_pe_getbyname(const char *name);
1366 struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name);
1368 /* Use a #define to avoid all of module.h just for these trivial ops */
1369 #define ip_vs_pe_get(pe) \
1370 if (pe && pe->module) \
1371 __module_get(pe->module);
1373 #define ip_vs_pe_put(pe) \
1374 if (pe && pe->module) \
1375 module_put(pe->module);
1377 /* IPVS protocol functions (from ip_vs_proto.c) */
1378 int ip_vs_protocol_init(void);
1379 void ip_vs_protocol_cleanup(void);
1380 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags);
1381 int *ip_vs_create_timeout_table(int *table, int size);
1382 void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
1383 const struct sk_buff *skb, int offset,
1386 extern struct ip_vs_protocol ip_vs_protocol_tcp;
1387 extern struct ip_vs_protocol ip_vs_protocol_udp;
1388 extern struct ip_vs_protocol ip_vs_protocol_icmp;
1389 extern struct ip_vs_protocol ip_vs_protocol_esp;
1390 extern struct ip_vs_protocol ip_vs_protocol_ah;
1391 extern struct ip_vs_protocol ip_vs_protocol_sctp;
1393 /* Registering/unregistering scheduler functions
1394 * (from ip_vs_sched.c)
1396 int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1397 int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1398 int ip_vs_bind_scheduler(struct ip_vs_service *svc,
1399 struct ip_vs_scheduler *scheduler);
1400 void ip_vs_unbind_scheduler(struct ip_vs_service *svc,
1401 struct ip_vs_scheduler *sched);
1402 struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name);
1403 void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler);
1405 ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb,
1406 struct ip_vs_proto_data *pd, int *ignored,
1407 struct ip_vs_iphdr *iph);
1408 int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb,
1409 struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph);
1411 void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg);
1413 /* IPVS control data and functions (from ip_vs_ctl.c) */
1414 extern struct ip_vs_stats ip_vs_stats;
1415 extern int sysctl_ip_vs_sync_ver;
1417 struct ip_vs_service *
1418 ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
1419 const union nf_inet_addr *vaddr, __be16 vport);
1421 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1422 const union nf_inet_addr *daddr, __be16 dport);
1425 ip_vs_find_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1426 const union nf_inet_addr *daddr, __be16 dport);
1427 struct ip_vs_dest *ip_vs_find_tunnel(struct netns_ipvs *ipvs, int af,
1428 const union nf_inet_addr *daddr,
1431 int ip_vs_use_count_inc(void);
1432 void ip_vs_use_count_dec(void);
1433 int ip_vs_register_nl_ioctl(void);
1434 void ip_vs_unregister_nl_ioctl(void);
1435 int ip_vs_control_init(void);
1436 void ip_vs_control_cleanup(void);
1438 ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
1439 const union nf_inet_addr *daddr, __be16 dport,
1440 const union nf_inet_addr *vaddr, __be16 vport,
1441 __u16 protocol, __u32 fwmark, __u32 flags);
1442 void ip_vs_try_bind_dest(struct ip_vs_conn *cp);
1444 static inline void ip_vs_dest_hold(struct ip_vs_dest *dest)
1446 refcount_inc(&dest->refcnt);
1449 static inline void ip_vs_dest_put(struct ip_vs_dest *dest)
1451 smp_mb__before_atomic();
1452 refcount_dec(&dest->refcnt);
1455 static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest)
1457 if (refcount_dec_and_test(&dest->refcnt))
1461 /* IPVS sync daemon data and function prototypes
1462 * (from ip_vs_sync.c)
1464 int start_sync_thread(struct netns_ipvs *ipvs, struct ipvs_sync_daemon_cfg *cfg,
1466 int stop_sync_thread(struct netns_ipvs *ipvs, int state);
1467 void ip_vs_sync_conn(struct netns_ipvs *ipvs, struct ip_vs_conn *cp, int pkts);
1469 /* IPVS rate estimator prototypes (from ip_vs_est.c) */
1470 void ip_vs_start_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1471 void ip_vs_stop_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1472 void ip_vs_zero_estimator(struct ip_vs_stats *stats);
1473 void ip_vs_read_estimator(struct ip_vs_kstats *dst, struct ip_vs_stats *stats);
1475 /* Various IPVS packet transmitters (from ip_vs_xmit.c) */
1476 int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1477 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1478 int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1479 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1480 int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1481 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1482 int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1483 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1484 int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1485 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1486 int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1487 struct ip_vs_protocol *pp, int offset,
1488 unsigned int hooknum, struct ip_vs_iphdr *iph);
1489 void ip_vs_dest_dst_rcu_free(struct rcu_head *head);
1491 #ifdef CONFIG_IP_VS_IPV6
1492 int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1493 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1494 int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1495 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1496 int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1497 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1498 int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1499 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1500 int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1501 struct ip_vs_protocol *pp, int offset,
1502 unsigned int hooknum, struct ip_vs_iphdr *iph);
1505 #ifdef CONFIG_SYSCTL
1506 /* This is a simple mechanism to ignore packets when
1507 * we are loaded. Just set ip_vs_drop_rate to 'n' and
1508 * we start to drop 1/rate of the packets
1510 static inline int ip_vs_todrop(struct netns_ipvs *ipvs)
1512 if (!ipvs->drop_rate)
1514 if (--ipvs->drop_counter > 0)
1516 ipvs->drop_counter = ipvs->drop_rate;
1520 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; }
1523 #ifdef CONFIG_SYSCTL
1524 /* Enqueue delayed work for expiring no dest connections
1525 * Only run when sysctl_expire_nodest=1
1527 static inline void ip_vs_enqueue_expire_nodest_conns(struct netns_ipvs *ipvs)
1529 if (sysctl_expire_nodest_conn(ipvs))
1530 queue_delayed_work(system_long_wq,
1531 &ipvs->expire_nodest_conn_work, 1);
1534 void ip_vs_expire_nodest_conn_flush(struct netns_ipvs *ipvs);
1536 static inline void ip_vs_enqueue_expire_nodest_conns(struct netns_ipvs *ipvs) {}
1539 #define IP_VS_DFWD_METHOD(dest) (atomic_read(&(dest)->conn_flags) & \
1540 IP_VS_CONN_F_FWD_MASK)
1542 /* ip_vs_fwd_tag returns the forwarding tag of the connection */
1543 #define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK)
1545 static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp)
1549 switch (IP_VS_FWD_METHOD(cp)) {
1550 case IP_VS_CONN_F_MASQ:
1552 case IP_VS_CONN_F_LOCALNODE:
1554 case IP_VS_CONN_F_TUNNEL:
1556 case IP_VS_CONN_F_DROUTE:
1558 case IP_VS_CONN_F_BYPASS:
1566 void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp,
1567 struct ip_vs_conn *cp, int dir);
1569 #ifdef CONFIG_IP_VS_IPV6
1570 void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp,
1571 struct ip_vs_conn *cp, int dir);
1574 __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset);
1576 static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum)
1578 __be32 diff[2] = { ~old, new };
1580 return csum_partial(diff, sizeof(diff), oldsum);
1583 #ifdef CONFIG_IP_VS_IPV6
1584 static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new,
1587 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0],
1588 new[3], new[2], new[1], new[0] };
1590 return csum_partial(diff, sizeof(diff), oldsum);
1594 static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum)
1596 __be16 diff[2] = { ~old, new };
1598 return csum_partial(diff, sizeof(diff), oldsum);
1601 /* Forget current conntrack (unconfirmed) and attach notrack entry */
1602 static inline void ip_vs_notrack(struct sk_buff *skb)
1604 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1605 enum ip_conntrack_info ctinfo;
1606 struct nf_conn *ct = nf_ct_get(skb, &ctinfo);
1609 nf_conntrack_put(&ct->ct_general);
1610 nf_ct_set(skb, NULL, IP_CT_UNTRACKED);
1615 #ifdef CONFIG_IP_VS_NFCT
1616 /* Netfilter connection tracking
1617 * (from ip_vs_nfct.c)
1619 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1621 #ifdef CONFIG_SYSCTL
1622 return ipvs->sysctl_conntrack;
1628 void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp,
1630 int ip_vs_confirm_conntrack(struct sk_buff *skb);
1631 void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct,
1632 struct ip_vs_conn *cp, u_int8_t proto,
1633 const __be16 port, int from_rs);
1634 void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp);
1638 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1643 static inline void ip_vs_update_conntrack(struct sk_buff *skb,
1644 struct ip_vs_conn *cp, int outin)
1648 static inline int ip_vs_confirm_conntrack(struct sk_buff *skb)
1653 static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp)
1656 #endif /* CONFIG_IP_VS_NFCT */
1658 /* Using old conntrack that can not be redirected to another real server? */
1659 static inline bool ip_vs_conn_uses_old_conntrack(struct ip_vs_conn *cp,
1660 struct sk_buff *skb)
1662 #ifdef CONFIG_IP_VS_NFCT
1663 enum ip_conntrack_info ctinfo;
1666 ct = nf_ct_get(skb, &ctinfo);
1667 if (ct && nf_ct_is_confirmed(ct))
1673 static inline int ip_vs_register_conntrack(struct ip_vs_service *svc)
1675 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
1676 int afmask = (svc->af == AF_INET6) ? 2 : 1;
1679 if (!(svc->conntrack_afmask & afmask)) {
1680 ret = nf_ct_netns_get(svc->ipvs->net, svc->af);
1682 svc->conntrack_afmask |= afmask;
1690 static inline void ip_vs_unregister_conntrack(struct ip_vs_service *svc)
1692 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
1693 int afmask = (svc->af == AF_INET6) ? 2 : 1;
1695 if (svc->conntrack_afmask & afmask) {
1696 nf_ct_netns_put(svc->ipvs->net, svc->af);
1697 svc->conntrack_afmask &= ~afmask;
1702 int ip_vs_register_hooks(struct netns_ipvs *ipvs, unsigned int af);
1703 void ip_vs_unregister_hooks(struct netns_ipvs *ipvs, unsigned int af);
1706 ip_vs_dest_conn_overhead(struct ip_vs_dest *dest)
1708 /* We think the overhead of processing active connections is 256
1709 * times higher than that of inactive connections in average. (This
1710 * 256 times might not be accurate, we will change it later) We
1711 * use the following formula to estimate the overhead now:
1712 * dest->activeconns*256 + dest->inactconns
1714 return (atomic_read(&dest->activeconns) << 8) +
1715 atomic_read(&dest->inactconns);
1718 #endif /* _NET_IP_VS_H */