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_NF_CONNTRACK)
29 #include <net/netfilter/nf_conntrack.h>
31 #include <net/net_namespace.h> /* Netw namespace */
33 #define IP_VS_HDR_INVERSE 1
34 #define IP_VS_HDR_ICMP 2
36 /* Generic access of ipvs struct */
37 static inline struct netns_ipvs *net_ipvs(struct net* net)
42 /* Connections' size value needed by ip_vs_ctl.c */
43 extern int ip_vs_conn_tab_size;
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)*/
53 union nf_inet_addr saddr;
54 union nf_inet_addr daddr;
57 static inline void *frag_safe_skb_hp(const struct sk_buff *skb, int offset,
58 int len, void *buffer)
60 return skb_header_pointer(skb, offset, len, buffer);
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.
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)
71 iphdr->hdr_flags = hdr_flags;
74 #ifdef CONFIG_IP_VS_IPV6
77 const struct ipv6hdr *iph = skb_header_pointer(
78 skb, offset, sizeof(_iph), &_iph);
82 iphdr->saddr.in6 = iph->saddr;
83 iphdr->daddr.in6 = iph->daddr;
84 /* ipv6_find_hdr() updates len, flags */
87 iphdr->protocol = ipv6_find_hdr(skb, &iphdr->len, -1,
90 if (iphdr->protocol < 0)
96 const struct iphdr *iph = skb_header_pointer(
97 skb, offset, sizeof(_iph), &_iph);
101 iphdr->len = offset + iph->ihl * 4;
103 iphdr->protocol = iph->protocol;
104 iphdr->saddr.ip = iph->saddr;
105 iphdr->daddr.ip = iph->daddr;
112 ip_vs_fill_iph_skb_icmp(int af, const struct sk_buff *skb, int offset,
113 bool inverse, struct ip_vs_iphdr *iphdr)
115 int hdr_flags = IP_VS_HDR_ICMP;
118 hdr_flags |= IP_VS_HDR_INVERSE;
120 return ip_vs_fill_iph_skb_off(af, skb, offset, hdr_flags, iphdr);
124 ip_vs_fill_iph_skb(int af, const struct sk_buff *skb, bool inverse,
125 struct ip_vs_iphdr *iphdr)
130 hdr_flags |= IP_VS_HDR_INVERSE;
132 return ip_vs_fill_iph_skb_off(af, skb, skb_network_offset(skb),
137 ip_vs_iph_inverse(const struct ip_vs_iphdr *iph)
139 return !!(iph->hdr_flags & IP_VS_HDR_INVERSE);
143 ip_vs_iph_icmp(const struct ip_vs_iphdr *iph)
145 return !!(iph->hdr_flags & IP_VS_HDR_ICMP);
148 static inline void ip_vs_addr_copy(int af, union nf_inet_addr *dst,
149 const union nf_inet_addr *src)
151 #ifdef CONFIG_IP_VS_IPV6
159 static inline void ip_vs_addr_set(int af, union nf_inet_addr *dst,
160 const union nf_inet_addr *src)
162 #ifdef CONFIG_IP_VS_IPV6
163 if (af == AF_INET6) {
174 static inline int ip_vs_addr_equal(int af, const union nf_inet_addr *a,
175 const union nf_inet_addr *b)
177 #ifdef CONFIG_IP_VS_IPV6
179 return ipv6_addr_equal(&a->in6, &b->in6);
181 return a->ip == b->ip;
184 #ifdef CONFIG_IP_VS_DEBUG
185 #include <linux/net.h>
187 int ip_vs_get_debug_level(void);
189 static inline const char *ip_vs_dbg_addr(int af, char *buf, size_t buf_len,
190 const union nf_inet_addr *addr,
194 #ifdef CONFIG_IP_VS_IPV6
196 len = snprintf(&buf[*idx], buf_len - *idx, "[%pI6c]",
200 len = snprintf(&buf[*idx], buf_len - *idx, "%pI4",
204 BUG_ON(*idx > buf_len + 1);
205 return &buf[*idx - len];
208 #define IP_VS_DBG_BUF(level, msg, ...) \
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__); \
215 #define IP_VS_ERR_BUF(msg...) \
217 char ip_vs_dbg_buf[160]; \
218 int ip_vs_dbg_idx = 0; \
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, \
228 #define IP_VS_DBG(level, msg, ...) \
230 if (level <= ip_vs_get_debug_level()) \
231 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
233 #define IP_VS_DBG_RL(msg, ...) \
235 if (net_ratelimit()) \
236 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
238 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) \
240 if (level <= ip_vs_get_debug_level()) \
241 pp->debug_packet(af, pp, skb, ofs, msg); \
243 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) \
245 if (level <= ip_vs_get_debug_level() && \
247 pp->debug_packet(af, pp, skb, ofs, msg); \
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)
258 #define IP_VS_BUG() BUG()
259 #define IP_VS_ERR_RL(msg, ...) \
261 if (net_ratelimit()) \
262 pr_err(msg, ##__VA_ARGS__); \
265 #ifdef CONFIG_IP_VS_DEBUG
266 #define EnterFunction(level) \
268 if (level <= ip_vs_get_debug_level()) \
270 pr_fmt("Enter: %s, %s line %i\n"), \
271 __func__, __FILE__, __LINE__); \
273 #define LeaveFunction(level) \
275 if (level <= ip_vs_get_debug_level()) \
277 pr_fmt("Leave: %s, %s line %i\n"), \
278 __func__, __FILE__, __LINE__); \
281 #define EnterFunction(level) do {} while (0)
282 #define LeaveFunction(level) do {} while (0)
285 /* The port number of FTP service (in network order). */
286 #define FTPPORT cpu_to_be16(21)
287 #define FTPDATA cpu_to_be16(20)
289 /* TCP State Values */
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,
298 IP_VS_TCP_S_CLOSE_WAIT,
299 IP_VS_TCP_S_LAST_ACK,
305 /* UDP State Values */
311 /* ICMP State Values */
317 /* SCTP State Values */
318 enum ip_vs_sctp_states {
322 IP_VS_SCTP_S_COOKIE_SENT,
323 IP_VS_SCTP_S_COOKIE_REPLIED,
324 IP_VS_SCTP_S_COOKIE_WAIT,
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,
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
341 /* Delta sequence info structure
342 * Each ip_vs_conn has 2 (output AND input seq. changes).
343 * Only used in the VS/NAT.
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 */
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 */
361 struct ip_vs_cpu_stats {
362 struct ip_vs_counters cnt;
363 struct u64_stats_sync syncp;
366 /* IPVS statistics objects */
367 struct ip_vs_estimator {
368 struct list_head list;
384 * IPVS statistics object, 64-bit kernel version of struct ip_vs_stats_user
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 */
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 */
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 */
413 struct ip_vs_proto_data;
415 struct ip_vs_protocol {
416 struct ip_vs_protocol *next;
422 void (*init)(struct ip_vs_protocol *pp);
424 void (*exit)(struct ip_vs_protocol *pp);
426 int (*init_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
428 void (*exit_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
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);
437 (*conn_in_get)(struct netns_ipvs *ipvs,
439 const struct sk_buff *skb,
440 const struct ip_vs_iphdr *iph);
443 (*conn_out_get)(struct netns_ipvs *ipvs,
445 const struct sk_buff *skb,
446 const struct ip_vs_iphdr *iph);
448 int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
449 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
451 int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
452 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
454 const char *(*state_name)(int state);
456 void (*state_transition)(struct ip_vs_conn *cp, int direction,
457 const struct sk_buff *skb,
458 struct ip_vs_proto_data *pd);
460 int (*register_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
462 void (*unregister_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
464 int (*app_conn_bind)(struct ip_vs_conn *cp);
466 void (*debug_packet)(int af, struct ip_vs_protocol *pp,
467 const struct sk_buff *skb,
471 void (*timeout_change)(struct ip_vs_proto_data *pd, int flags);
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;
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);
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;
496 const struct ip_vs_pe *pe;
501 /* IP_VS structure allocated for each dynamically scheduled connection */
503 struct hlist_node c_list; /* hashed list heads */
504 /* Protocol, addresses and port numbers */
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;
517 /* counter and timer */
518 refcount_t refcnt; /* reference count */
519 struct timer_list timer; /* Expiration timer */
520 volatile unsigned long timeout; /* timeout */
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
529 __u32 fwmark; /* Fire wall mark from skb */
530 unsigned long sync_endtime; /* jiffies + sent_retries */
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 */
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.
543 int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp,
544 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
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
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 */
555 const struct ip_vs_pe *pe;
559 struct rcu_head rcu_head;
562 /* Extended internal versions of struct ip_vs_service_user and ip_vs_dest_user
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.
569 struct ip_vs_service_user_kern {
570 /* virtual service addresses */
573 union nf_inet_addr addr; /* virtual ip address */
575 u32 fwmark; /* firwall mark of service */
577 /* virtual service options */
580 unsigned int flags; /* virtual service flags */
581 unsigned int timeout; /* persistent timeout in sec */
582 __be32 netmask; /* persistent netmask or plen */
586 struct ip_vs_dest_user_kern {
587 /* destination server address */
588 union nf_inet_addr addr;
591 /* real server options */
592 unsigned int conn_flags; /* connection flags */
593 int weight; /* destination weight */
595 /* thresholds for active connections */
596 u32 u_threshold; /* upper threshold */
597 u32 l_threshold; /* lower threshold */
599 /* Address family of addr */
602 u16 tun_type; /* tunnel type */
603 __be16 tun_port; /* tunnel port */
604 u16 tun_flags; /* tunnel flags */
609 * The information about the virtual service offered to the net and the
610 * forwarding entries.
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 */
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;
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 */
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 */
636 /* alternate persistence engine */
637 struct ip_vs_pe __rcu *pe;
638 int conntrack_afmask;
640 struct rcu_head rcu_head;
643 /* Information for cached dst */
644 struct ip_vs_dest_dst {
645 struct dst_entry *dst_cache; /* destination cache entry */
647 union nf_inet_addr dst_saddr;
648 struct rcu_head rcu_head;
651 /* The real server destination forwarding entry with ip address, port number,
655 struct list_head n_list; /* for the dests in the service */
656 struct hlist_node d_list; /* for table with all the dests */
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 */
669 refcount_t refcnt; /* reference counter */
670 struct ip_vs_stats stats; /* statistics */
671 unsigned long idle_start; /* start time, jiffies */
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 */
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 */
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 */
691 struct list_head t_list; /* in dest_trash */
692 unsigned int in_rs_table:1; /* we are in rs_table */
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 */
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);
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);
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);
719 /* The persistence engine object */
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 */
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,
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,
737 const struct ip_vs_iphdr *iph,
738 __be16 dport, __be16 cport);
741 /* The application module object (a.k.a. app incarnation) */
743 struct list_head a_list; /* member in app list */
744 int type; /* IP_VS_APP_TYPE_xxx */
745 char *name; /* application module name */
747 struct module *module; /* THIS_MODULE/NULL */
748 struct list_head incs_list; /* list of incarnations */
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;
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
761 int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *,
762 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh);
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
768 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *,
769 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh);
771 /* ip_vs_app initializer */
772 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *);
774 /* ip_vs_app finish */
775 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *);
779 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *,
780 struct ip_vs_protocol *);
782 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *);
788 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app,
789 int *verdict, struct ip_vs_conn **cpp);
792 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app,
793 const struct iphdr *iph, int inverse);
796 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app,
797 const struct iphdr *iph, int inverse);
799 int (*state_transition)(struct ip_vs_conn *cp, int direction,
800 const struct sk_buff *skb,
801 struct ip_vs_app *app);
803 void (*timeout_change)(struct ip_vs_app *app, int flags);
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;
815 struct ip_vs_sync_thread_data;
817 /* How much time to keep dests in trash */
818 #define IP_VS_DEST_TRASH_PERIOD (120 * HZ)
820 struct ipvs_sync_daemon_cfg {
821 union nf_inet_addr mcast_group;
827 /* multicast interface name */
828 char mcast_ifn[IP_VS_IFNAME_MAXLEN];
831 /* IPVS in network namespace */
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)
840 struct hlist_head rs_table[IP_VS_RTAB_SIZE];
842 struct list_head app_list;
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];
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];
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];
869 atomic_t conn_count; /* connection counter */
872 struct ip_vs_stats tot_stats; /* Statistics & est. */
874 int num_services; /* no of virtual services */
875 int num_services6; /* IPv6 virtual services */
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;
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 */
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 */
901 struct ctl_table_header *sysctl_hdr;
902 struct ctl_table *sysctl_tbl;
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;
914 int sysctl_snat_reroute;
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 int sysctl_run_estimation;
937 int sysctl_lblc_expiration;
938 struct ctl_table_header *lblc_ctl_header;
939 struct ctl_table *lblc_ctl_table;
941 int sysctl_lblcr_expiration;
942 struct ctl_table_header *lblcr_ctl_header;
943 struct ctl_table *lblcr_ctl_table;
945 struct list_head est_list; /* estimator list */
947 struct timer_list est_timer; /* Estimation timer */
949 spinlock_t sync_lock;
950 struct ipvs_master_sync_state *ms;
951 spinlock_t sync_buff_lock;
952 struct ip_vs_sync_thread_data *master_tinfo;
953 struct ip_vs_sync_thread_data *backup_tinfo;
955 volatile int sync_state;
956 struct mutex sync_mutex;
957 struct ipvs_sync_daemon_cfg mcfg; /* Master Configuration */
958 struct ipvs_sync_daemon_cfg bcfg; /* Backup Configuration */
959 /* net name space ptr */
960 struct net *net; /* Needed by timer routines */
961 /* Number of heterogeneous destinations, needed becaus heterogeneous
962 * are not supported when synchronization is enabled.
964 unsigned int mixed_address_family_dests;
965 unsigned int hooks_afmask; /* &1=AF_INET, &2=AF_INET6 */
968 #define DEFAULT_SYNC_THRESHOLD 3
969 #define DEFAULT_SYNC_PERIOD 50
970 #define DEFAULT_SYNC_VER 1
971 #define DEFAULT_SLOPPY_TCP 0
972 #define DEFAULT_SLOPPY_SCTP 0
973 #define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ)
974 #define DEFAULT_SYNC_RETRIES 0
975 #define IPVS_SYNC_WAKEUP_RATE 8
976 #define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4)
977 #define IPVS_SYNC_SEND_DELAY (HZ / 50)
978 #define IPVS_SYNC_CHECK_PERIOD HZ
979 #define IPVS_SYNC_FLUSH_TIME (HZ * 2)
980 #define IPVS_SYNC_PORTS_MAX (1 << 6)
984 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
986 return ipvs->sysctl_sync_threshold[0];
989 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
991 return READ_ONCE(ipvs->sysctl_sync_threshold[1]);
994 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
996 return READ_ONCE(ipvs->sysctl_sync_refresh_period);
999 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1001 return ipvs->sysctl_sync_retries;
1004 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1006 return ipvs->sysctl_sync_ver;
1009 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1011 return ipvs->sysctl_sloppy_tcp;
1014 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1016 return ipvs->sysctl_sloppy_sctp;
1019 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1021 return READ_ONCE(ipvs->sysctl_sync_ports);
1024 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1026 return ipvs->sysctl_sync_persist_mode;
1029 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1031 return ipvs->sysctl_sync_qlen_max;
1034 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1036 return ipvs->sysctl_sync_sock_size;
1039 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1041 return ipvs->sysctl_pmtu_disc;
1044 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1046 return ipvs->sync_state & IP_VS_STATE_BACKUP &&
1047 ipvs->sysctl_backup_only;
1050 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1052 return ipvs->sysctl_conn_reuse_mode;
1055 static inline int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs)
1057 return ipvs->sysctl_expire_nodest_conn;
1060 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1062 return ipvs->sysctl_schedule_icmp;
1065 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1067 return ipvs->sysctl_ignore_tunneled;
1070 static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1072 return ipvs->sysctl_cache_bypass;
1075 static inline int sysctl_run_estimation(struct netns_ipvs *ipvs)
1077 return ipvs->sysctl_run_estimation;
1082 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1084 return DEFAULT_SYNC_THRESHOLD;
1087 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1089 return DEFAULT_SYNC_PERIOD;
1092 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1094 return DEFAULT_SYNC_REFRESH_PERIOD;
1097 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1099 return DEFAULT_SYNC_RETRIES & 3;
1102 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1104 return DEFAULT_SYNC_VER;
1107 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1109 return DEFAULT_SLOPPY_TCP;
1112 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1114 return DEFAULT_SLOPPY_SCTP;
1117 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1122 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1127 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1129 return IPVS_SYNC_QLEN_MAX;
1132 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1137 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1142 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1147 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1152 static inline int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs)
1157 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1162 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1167 static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1172 static inline int sysctl_run_estimation(struct netns_ipvs *ipvs)
1179 /* IPVS core functions
1180 * (from ip_vs_core.c)
1182 const char *ip_vs_proto_name(unsigned int proto);
1183 void ip_vs_init_hash_table(struct list_head *table, int rows);
1184 struct ip_vs_conn *ip_vs_new_conn_out(struct ip_vs_service *svc,
1185 struct ip_vs_dest *dest,
1186 struct sk_buff *skb,
1187 const struct ip_vs_iphdr *iph,
1190 #define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t)))
1192 #define IP_VS_APP_TYPE_FTP 1
1194 /* ip_vs_conn handling functions
1195 * (from ip_vs_conn.c)
1198 IP_VS_DIR_INPUT = 0,
1200 IP_VS_DIR_INPUT_ONLY,
1204 static inline void ip_vs_conn_fill_param(struct netns_ipvs *ipvs, int af, int protocol,
1205 const union nf_inet_addr *caddr,
1207 const union nf_inet_addr *vaddr,
1209 struct ip_vs_conn_param *p)
1213 p->protocol = protocol;
1222 struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p);
1223 struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p);
1225 struct ip_vs_conn * ip_vs_conn_in_get_proto(struct netns_ipvs *ipvs, int af,
1226 const struct sk_buff *skb,
1227 const struct ip_vs_iphdr *iph);
1229 struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p);
1231 struct ip_vs_conn * ip_vs_conn_out_get_proto(struct netns_ipvs *ipvs, int af,
1232 const struct sk_buff *skb,
1233 const struct ip_vs_iphdr *iph);
1235 /* Get reference to gain full access to conn.
1236 * By default, RCU read-side critical sections have access only to
1237 * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference.
1239 static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp)
1241 return refcount_inc_not_zero(&cp->refcnt);
1244 /* put back the conn without restarting its timer */
1245 static inline void __ip_vs_conn_put(struct ip_vs_conn *cp)
1247 smp_mb__before_atomic();
1248 refcount_dec(&cp->refcnt);
1250 void ip_vs_conn_put(struct ip_vs_conn *cp);
1251 void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport);
1253 struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, int dest_af,
1254 const union nf_inet_addr *daddr,
1255 __be16 dport, unsigned int flags,
1256 struct ip_vs_dest *dest, __u32 fwmark);
1257 void ip_vs_conn_expire_now(struct ip_vs_conn *cp);
1259 const char *ip_vs_state_name(const struct ip_vs_conn *cp);
1261 void ip_vs_tcp_conn_listen(struct ip_vs_conn *cp);
1262 int ip_vs_check_template(struct ip_vs_conn *ct, struct ip_vs_dest *cdest);
1263 void ip_vs_random_dropentry(struct netns_ipvs *ipvs);
1264 int ip_vs_conn_init(void);
1265 void ip_vs_conn_cleanup(void);
1267 static inline void ip_vs_control_del(struct ip_vs_conn *cp)
1269 struct ip_vs_conn *ctl_cp = cp->control;
1271 IP_VS_ERR_BUF("request control DEL for uncontrolled: "
1273 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1275 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1281 IP_VS_DBG_BUF(7, "DELeting control for: "
1282 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1283 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1285 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1286 ntohs(ctl_cp->cport));
1289 if (atomic_read(&ctl_cp->n_control) == 0) {
1290 IP_VS_ERR_BUF("BUG control DEL with n=0 : "
1292 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1294 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1299 atomic_dec(&ctl_cp->n_control);
1303 ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp)
1306 IP_VS_ERR_BUF("request control ADD for already controlled: "
1308 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1310 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1313 ip_vs_control_del(cp);
1316 IP_VS_DBG_BUF(7, "ADDing control for: "
1317 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1318 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1320 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1321 ntohs(ctl_cp->cport));
1323 cp->control = ctl_cp;
1324 atomic_inc(&ctl_cp->n_control);
1327 /* Mark our template as assured */
1329 ip_vs_control_assure_ct(struct ip_vs_conn *cp)
1331 struct ip_vs_conn *ct = cp->control;
1333 if (ct && !(ct->state & IP_VS_CTPL_S_ASSURED) &&
1334 (ct->flags & IP_VS_CONN_F_TEMPLATE))
1335 ct->state |= IP_VS_CTPL_S_ASSURED;
1338 /* IPVS netns init & cleanup functions */
1339 int ip_vs_estimator_net_init(struct netns_ipvs *ipvs);
1340 int ip_vs_control_net_init(struct netns_ipvs *ipvs);
1341 int ip_vs_protocol_net_init(struct netns_ipvs *ipvs);
1342 int ip_vs_app_net_init(struct netns_ipvs *ipvs);
1343 int ip_vs_conn_net_init(struct netns_ipvs *ipvs);
1344 int ip_vs_sync_net_init(struct netns_ipvs *ipvs);
1345 void ip_vs_conn_net_cleanup(struct netns_ipvs *ipvs);
1346 void ip_vs_app_net_cleanup(struct netns_ipvs *ipvs);
1347 void ip_vs_protocol_net_cleanup(struct netns_ipvs *ipvs);
1348 void ip_vs_control_net_cleanup(struct netns_ipvs *ipvs);
1349 void ip_vs_estimator_net_cleanup(struct netns_ipvs *ipvs);
1350 void ip_vs_sync_net_cleanup(struct netns_ipvs *ipvs);
1351 void ip_vs_service_nets_cleanup(struct list_head *net_list);
1353 /* IPVS application functions
1354 * (from ip_vs_app.c)
1356 #define IP_VS_APP_MAX_PORTS 8
1357 struct ip_vs_app *register_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1358 void unregister_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1359 int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1360 void ip_vs_unbind_app(struct ip_vs_conn *cp);
1361 int register_ip_vs_app_inc(struct netns_ipvs *ipvs, struct ip_vs_app *app, __u16 proto,
1363 int ip_vs_app_inc_get(struct ip_vs_app *inc);
1364 void ip_vs_app_inc_put(struct ip_vs_app *inc);
1366 int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb,
1367 struct ip_vs_iphdr *ipvsh);
1368 int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb,
1369 struct ip_vs_iphdr *ipvsh);
1371 int register_ip_vs_pe(struct ip_vs_pe *pe);
1372 int unregister_ip_vs_pe(struct ip_vs_pe *pe);
1373 struct ip_vs_pe *ip_vs_pe_getbyname(const char *name);
1374 struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name);
1376 /* Use a #define to avoid all of module.h just for these trivial ops */
1377 #define ip_vs_pe_get(pe) \
1378 if (pe && pe->module) \
1379 __module_get(pe->module);
1381 #define ip_vs_pe_put(pe) \
1382 if (pe && pe->module) \
1383 module_put(pe->module);
1385 /* IPVS protocol functions (from ip_vs_proto.c) */
1386 int ip_vs_protocol_init(void);
1387 void ip_vs_protocol_cleanup(void);
1388 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags);
1389 int *ip_vs_create_timeout_table(int *table, int size);
1390 void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
1391 const struct sk_buff *skb, int offset,
1394 extern struct ip_vs_protocol ip_vs_protocol_tcp;
1395 extern struct ip_vs_protocol ip_vs_protocol_udp;
1396 extern struct ip_vs_protocol ip_vs_protocol_icmp;
1397 extern struct ip_vs_protocol ip_vs_protocol_esp;
1398 extern struct ip_vs_protocol ip_vs_protocol_ah;
1399 extern struct ip_vs_protocol ip_vs_protocol_sctp;
1401 /* Registering/unregistering scheduler functions
1402 * (from ip_vs_sched.c)
1404 int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1405 int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1406 int ip_vs_bind_scheduler(struct ip_vs_service *svc,
1407 struct ip_vs_scheduler *scheduler);
1408 void ip_vs_unbind_scheduler(struct ip_vs_service *svc,
1409 struct ip_vs_scheduler *sched);
1410 struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name);
1411 void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler);
1413 ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb,
1414 struct ip_vs_proto_data *pd, int *ignored,
1415 struct ip_vs_iphdr *iph);
1416 int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb,
1417 struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph);
1419 void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg);
1421 /* IPVS control data and functions (from ip_vs_ctl.c) */
1422 extern struct ip_vs_stats ip_vs_stats;
1423 extern int sysctl_ip_vs_sync_ver;
1425 struct ip_vs_service *
1426 ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
1427 const union nf_inet_addr *vaddr, __be16 vport);
1429 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1430 const union nf_inet_addr *daddr, __be16 dport);
1433 ip_vs_find_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1434 const union nf_inet_addr *daddr, __be16 dport);
1435 struct ip_vs_dest *ip_vs_find_tunnel(struct netns_ipvs *ipvs, int af,
1436 const union nf_inet_addr *daddr,
1439 int ip_vs_use_count_inc(void);
1440 void ip_vs_use_count_dec(void);
1441 int ip_vs_register_nl_ioctl(void);
1442 void ip_vs_unregister_nl_ioctl(void);
1443 int ip_vs_control_init(void);
1444 void ip_vs_control_cleanup(void);
1446 ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
1447 const union nf_inet_addr *daddr, __be16 dport,
1448 const union nf_inet_addr *vaddr, __be16 vport,
1449 __u16 protocol, __u32 fwmark, __u32 flags);
1450 void ip_vs_try_bind_dest(struct ip_vs_conn *cp);
1452 static inline void ip_vs_dest_hold(struct ip_vs_dest *dest)
1454 refcount_inc(&dest->refcnt);
1457 static inline void ip_vs_dest_put(struct ip_vs_dest *dest)
1459 smp_mb__before_atomic();
1460 refcount_dec(&dest->refcnt);
1463 static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest)
1465 if (refcount_dec_and_test(&dest->refcnt))
1469 /* IPVS sync daemon data and function prototypes
1470 * (from ip_vs_sync.c)
1472 int start_sync_thread(struct netns_ipvs *ipvs, struct ipvs_sync_daemon_cfg *cfg,
1474 int stop_sync_thread(struct netns_ipvs *ipvs, int state);
1475 void ip_vs_sync_conn(struct netns_ipvs *ipvs, struct ip_vs_conn *cp, int pkts);
1477 /* IPVS rate estimator prototypes (from ip_vs_est.c) */
1478 void ip_vs_start_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1479 void ip_vs_stop_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1480 void ip_vs_zero_estimator(struct ip_vs_stats *stats);
1481 void ip_vs_read_estimator(struct ip_vs_kstats *dst, struct ip_vs_stats *stats);
1483 /* Various IPVS packet transmitters (from ip_vs_xmit.c) */
1484 int ip_vs_null_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_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1487 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1488 int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1489 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1490 int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1491 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1492 int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1493 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1494 int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1495 struct ip_vs_protocol *pp, int offset,
1496 unsigned int hooknum, struct ip_vs_iphdr *iph);
1497 void ip_vs_dest_dst_rcu_free(struct rcu_head *head);
1499 #ifdef CONFIG_IP_VS_IPV6
1500 int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1501 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1502 int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1503 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1504 int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1505 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1506 int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1507 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1508 int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1509 struct ip_vs_protocol *pp, int offset,
1510 unsigned int hooknum, struct ip_vs_iphdr *iph);
1513 #ifdef CONFIG_SYSCTL
1514 /* This is a simple mechanism to ignore packets when
1515 * we are loaded. Just set ip_vs_drop_rate to 'n' and
1516 * we start to drop 1/rate of the packets
1518 static inline int ip_vs_todrop(struct netns_ipvs *ipvs)
1520 if (!ipvs->drop_rate)
1522 if (--ipvs->drop_counter > 0)
1524 ipvs->drop_counter = ipvs->drop_rate;
1528 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; }
1531 #ifdef CONFIG_SYSCTL
1532 /* Enqueue delayed work for expiring no dest connections
1533 * Only run when sysctl_expire_nodest=1
1535 static inline void ip_vs_enqueue_expire_nodest_conns(struct netns_ipvs *ipvs)
1537 if (sysctl_expire_nodest_conn(ipvs))
1538 queue_delayed_work(system_long_wq,
1539 &ipvs->expire_nodest_conn_work, 1);
1542 void ip_vs_expire_nodest_conn_flush(struct netns_ipvs *ipvs);
1544 static inline void ip_vs_enqueue_expire_nodest_conns(struct netns_ipvs *ipvs) {}
1547 #define IP_VS_DFWD_METHOD(dest) (atomic_read(&(dest)->conn_flags) & \
1548 IP_VS_CONN_F_FWD_MASK)
1550 /* ip_vs_fwd_tag returns the forwarding tag of the connection */
1551 #define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK)
1553 static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp)
1557 switch (IP_VS_FWD_METHOD(cp)) {
1558 case IP_VS_CONN_F_MASQ:
1560 case IP_VS_CONN_F_LOCALNODE:
1562 case IP_VS_CONN_F_TUNNEL:
1564 case IP_VS_CONN_F_DROUTE:
1566 case IP_VS_CONN_F_BYPASS:
1574 void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp,
1575 struct ip_vs_conn *cp, int dir);
1577 #ifdef CONFIG_IP_VS_IPV6
1578 void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp,
1579 struct ip_vs_conn *cp, int dir);
1582 __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset);
1584 static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum)
1586 __be32 diff[2] = { ~old, new };
1588 return csum_partial(diff, sizeof(diff), oldsum);
1591 #ifdef CONFIG_IP_VS_IPV6
1592 static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new,
1595 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0],
1596 new[3], new[2], new[1], new[0] };
1598 return csum_partial(diff, sizeof(diff), oldsum);
1602 static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum)
1604 __be16 diff[2] = { ~old, new };
1606 return csum_partial(diff, sizeof(diff), oldsum);
1609 /* Forget current conntrack (unconfirmed) and attach notrack entry */
1610 static inline void ip_vs_notrack(struct sk_buff *skb)
1612 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1613 enum ip_conntrack_info ctinfo;
1614 struct nf_conn *ct = nf_ct_get(skb, &ctinfo);
1617 nf_conntrack_put(&ct->ct_general);
1618 nf_ct_set(skb, NULL, IP_CT_UNTRACKED);
1623 #ifdef CONFIG_IP_VS_NFCT
1624 /* Netfilter connection tracking
1625 * (from ip_vs_nfct.c)
1627 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1629 #ifdef CONFIG_SYSCTL
1630 return ipvs->sysctl_conntrack;
1636 void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp,
1638 int ip_vs_confirm_conntrack(struct sk_buff *skb);
1639 void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct,
1640 struct ip_vs_conn *cp, u_int8_t proto,
1641 const __be16 port, int from_rs);
1642 void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp);
1646 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1651 static inline void ip_vs_update_conntrack(struct sk_buff *skb,
1652 struct ip_vs_conn *cp, int outin)
1656 static inline int ip_vs_confirm_conntrack(struct sk_buff *skb)
1661 static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp)
1664 #endif /* CONFIG_IP_VS_NFCT */
1666 /* Using old conntrack that can not be redirected to another real server? */
1667 static inline bool ip_vs_conn_uses_old_conntrack(struct ip_vs_conn *cp,
1668 struct sk_buff *skb)
1670 #ifdef CONFIG_IP_VS_NFCT
1671 enum ip_conntrack_info ctinfo;
1674 ct = nf_ct_get(skb, &ctinfo);
1675 if (ct && nf_ct_is_confirmed(ct))
1681 static inline int ip_vs_register_conntrack(struct ip_vs_service *svc)
1683 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
1684 int afmask = (svc->af == AF_INET6) ? 2 : 1;
1687 if (!(svc->conntrack_afmask & afmask)) {
1688 ret = nf_ct_netns_get(svc->ipvs->net, svc->af);
1690 svc->conntrack_afmask |= afmask;
1698 static inline void ip_vs_unregister_conntrack(struct ip_vs_service *svc)
1700 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
1701 int afmask = (svc->af == AF_INET6) ? 2 : 1;
1703 if (svc->conntrack_afmask & afmask) {
1704 nf_ct_netns_put(svc->ipvs->net, svc->af);
1705 svc->conntrack_afmask &= ~afmask;
1710 int ip_vs_register_hooks(struct netns_ipvs *ipvs, unsigned int af);
1711 void ip_vs_unregister_hooks(struct netns_ipvs *ipvs, unsigned int af);
1714 ip_vs_dest_conn_overhead(struct ip_vs_dest *dest)
1716 /* We think the overhead of processing active connections is 256
1717 * times higher than that of inactive connections in average. (This
1718 * 256 times might not be accurate, we will change it later) We
1719 * use the following formula to estimate the overhead now:
1720 * dest->activeconns*256 + dest->inactconns
1722 return (atomic_read(&dest->activeconns) << 8) +
1723 atomic_read(&dest->inactconns);
1726 #ifdef CONFIG_IP_VS_PROTO_TCP
1727 INDIRECT_CALLABLE_DECLARE(int
1728 tcp_snat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
1729 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph));
1732 #ifdef CONFIG_IP_VS_PROTO_UDP
1733 INDIRECT_CALLABLE_DECLARE(int
1734 udp_snat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
1735 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph));
1737 #endif /* _NET_IP_VS_H */