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 */
32 #include <linux/sched/isolation.h>
34 #define IP_VS_HDR_INVERSE 1
35 #define IP_VS_HDR_ICMP 2
37 /* Generic access of ipvs struct */
38 static inline struct netns_ipvs *net_ipvs(struct net* net)
43 /* Connections' size value needed by ip_vs_ctl.c */
44 extern int ip_vs_conn_tab_size;
46 extern struct mutex __ip_vs_mutex;
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_stats_t conns; /* connections scheduled */
358 u64_stats_t inpkts; /* incoming packets */
359 u64_stats_t outpkts; /* outgoing packets */
360 u64_stats_t inbytes; /* incoming bytes */
361 u64_stats_t outbytes; /* outgoing bytes */
364 struct ip_vs_cpu_stats {
365 struct ip_vs_counters cnt;
366 struct u64_stats_sync syncp;
369 /* Default nice for estimator kthreads */
370 #define IPVS_EST_NICE 0
372 /* IPVS statistics objects */
373 struct ip_vs_estimator {
374 struct hlist_node list;
388 s32 ktid:16, /* kthread ID, -1=temp list */
389 ktrow:8, /* row/tick ID for kthread */
390 ktcid:8; /* chain ID for kthread tick */
394 * IPVS statistics object, 64-bit kernel version of struct ip_vs_stats_user
396 struct ip_vs_kstats {
397 u64 conns; /* connections scheduled */
398 u64 inpkts; /* incoming packets */
399 u64 outpkts; /* outgoing packets */
400 u64 inbytes; /* incoming bytes */
401 u64 outbytes; /* outgoing bytes */
403 u64 cps; /* current connection rate */
404 u64 inpps; /* current in packet rate */
405 u64 outpps; /* current out packet rate */
406 u64 inbps; /* current in byte rate */
407 u64 outbps; /* current out byte rate */
411 struct ip_vs_kstats kstats; /* kernel statistics */
412 struct ip_vs_estimator est; /* estimator */
413 struct ip_vs_cpu_stats __percpu *cpustats; /* per cpu counters */
414 spinlock_t lock; /* spin lock */
415 struct ip_vs_kstats kstats0; /* reset values */
418 struct ip_vs_stats_rcu {
419 struct ip_vs_stats s;
420 struct rcu_head rcu_head;
423 int ip_vs_stats_init_alloc(struct ip_vs_stats *s);
424 struct ip_vs_stats *ip_vs_stats_alloc(void);
425 void ip_vs_stats_release(struct ip_vs_stats *stats);
426 void ip_vs_stats_free(struct ip_vs_stats *stats);
428 /* Process estimators in multiple timer ticks (20/50/100, see ktrow) */
429 #define IPVS_EST_NTICKS 50
430 /* Estimation uses a 2-second period containing ticks (in jiffies) */
431 #define IPVS_EST_TICK ((2 * HZ) / IPVS_EST_NTICKS)
433 /* Limit of CPU load per kthread (8 for 12.5%), ratio of CPU capacity (1/C).
434 * Value of 4 and above ensures kthreads will take work without exceeding
435 * the CPU capacity under different circumstances.
437 #define IPVS_EST_LOAD_DIVISOR 8
439 /* Kthreads should not have work that exceeds the CPU load above 50% */
440 #define IPVS_EST_CPU_KTHREADS (IPVS_EST_LOAD_DIVISOR / 2)
442 /* Desired number of chains per timer tick (chain load factor in 100us units),
443 * 48=4.8ms of 40ms tick (12% CPU usage):
444 * 2 sec * 1000 ms in sec * 10 (100us in ms) / 8 (12.5%) / 50
446 #define IPVS_EST_CHAIN_FACTOR \
447 ALIGN_DOWN(2 * 1000 * 10 / IPVS_EST_LOAD_DIVISOR / IPVS_EST_NTICKS, 8)
449 /* Compiled number of chains per tick
450 * The defines should match cond_resched_rcu
452 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
453 #define IPVS_EST_TICK_CHAINS IPVS_EST_CHAIN_FACTOR
455 #define IPVS_EST_TICK_CHAINS 1
458 #if IPVS_EST_NTICKS > 127
459 #error Too many timer ticks for ktrow
462 /* Multiple chains processed in same tick */
463 struct ip_vs_est_tick_data {
464 struct hlist_head chains[IPVS_EST_TICK_CHAINS];
465 DECLARE_BITMAP(present, IPVS_EST_TICK_CHAINS);
466 DECLARE_BITMAP(full, IPVS_EST_TICK_CHAINS);
467 int chain_len[IPVS_EST_TICK_CHAINS];
470 /* Context for estimation kthread */
471 struct ip_vs_est_kt_data {
472 struct netns_ipvs *ipvs;
473 struct task_struct *task; /* task if running */
474 struct ip_vs_est_tick_data __rcu *ticks[IPVS_EST_NTICKS];
475 DECLARE_BITMAP(avail, IPVS_EST_NTICKS); /* tick has space for ests */
476 unsigned long est_timer; /* estimation timer (jiffies) */
477 struct ip_vs_stats *calc_stats; /* Used for calculation */
478 int tick_len[IPVS_EST_NTICKS]; /* est count */
479 int id; /* ktid per netns */
480 int chain_max; /* max ests per tick chain */
481 int tick_max; /* max ests per tick */
482 int est_count; /* attached ests to kthread */
483 int est_max_count; /* max ests per kthread */
484 int add_row; /* row for new ests */
485 int est_row; /* estimated row */
493 struct ip_vs_proto_data;
495 struct ip_vs_protocol {
496 struct ip_vs_protocol *next;
502 void (*init)(struct ip_vs_protocol *pp);
504 void (*exit)(struct ip_vs_protocol *pp);
506 int (*init_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
508 void (*exit_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
510 int (*conn_schedule)(struct netns_ipvs *ipvs,
511 int af, struct sk_buff *skb,
512 struct ip_vs_proto_data *pd,
513 int *verdict, struct ip_vs_conn **cpp,
514 struct ip_vs_iphdr *iph);
517 (*conn_in_get)(struct netns_ipvs *ipvs,
519 const struct sk_buff *skb,
520 const struct ip_vs_iphdr *iph);
523 (*conn_out_get)(struct netns_ipvs *ipvs,
525 const struct sk_buff *skb,
526 const struct ip_vs_iphdr *iph);
528 int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
529 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
531 int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
532 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
534 const char *(*state_name)(int state);
536 void (*state_transition)(struct ip_vs_conn *cp, int direction,
537 const struct sk_buff *skb,
538 struct ip_vs_proto_data *pd);
540 int (*register_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
542 void (*unregister_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
544 int (*app_conn_bind)(struct ip_vs_conn *cp);
546 void (*debug_packet)(int af, struct ip_vs_protocol *pp,
547 const struct sk_buff *skb,
551 void (*timeout_change)(struct ip_vs_proto_data *pd, int flags);
554 /* protocol data per netns */
555 struct ip_vs_proto_data {
556 struct ip_vs_proto_data *next;
557 struct ip_vs_protocol *pp;
558 int *timeout_table; /* protocol timeout table */
559 atomic_t appcnt; /* counter of proto app incs. */
560 struct tcp_states_t *tcp_state_table;
563 struct ip_vs_protocol *ip_vs_proto_get(unsigned short proto);
564 struct ip_vs_proto_data *ip_vs_proto_data_get(struct netns_ipvs *ipvs,
565 unsigned short proto);
567 struct ip_vs_conn_param {
568 struct netns_ipvs *ipvs;
569 const union nf_inet_addr *caddr;
570 const union nf_inet_addr *vaddr;
576 const struct ip_vs_pe *pe;
581 /* IP_VS structure allocated for each dynamically scheduled connection */
583 struct hlist_node c_list; /* hashed list heads */
584 /* Protocol, addresses and port numbers */
588 u16 af; /* address family */
589 union nf_inet_addr caddr; /* client address */
590 union nf_inet_addr vaddr; /* virtual address */
591 union nf_inet_addr daddr; /* destination address */
592 volatile __u32 flags; /* status flags */
593 __u16 protocol; /* Which protocol (TCP/UDP) */
594 __u16 daf; /* Address family of the dest */
595 struct netns_ipvs *ipvs;
597 /* counter and timer */
598 refcount_t refcnt; /* reference count */
599 struct timer_list timer; /* Expiration timer */
600 volatile unsigned long timeout; /* timeout */
602 /* Flags and state transition */
603 spinlock_t lock; /* lock for state transition */
604 volatile __u16 state; /* state info */
605 volatile __u16 old_state; /* old state, to be used for
606 * state transition triggerd
609 __u32 fwmark; /* Fire wall mark from skb */
610 unsigned long sync_endtime; /* jiffies + sent_retries */
612 /* Control members */
613 struct ip_vs_conn *control; /* Master control connection */
614 atomic_t n_control; /* Number of controlled ones */
615 struct ip_vs_dest *dest; /* real server */
616 atomic_t in_pkts; /* incoming packet counter */
618 /* Packet transmitter for different forwarding methods. If it
619 * mangles the packet, it must return NF_DROP or better NF_STOLEN,
620 * otherwise this must be changed to a sk_buff **.
621 * NF_ACCEPT can be returned when destination is local.
623 int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp,
624 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
626 /* Note: we can group the following members into a structure,
627 * in order to save more space, and the following members are
628 * only used in VS/NAT anyway
630 struct ip_vs_app *app; /* bound ip_vs_app object */
631 void *app_data; /* Application private data */
632 struct ip_vs_seq in_seq; /* incoming seq. struct */
633 struct ip_vs_seq out_seq; /* outgoing seq. struct */
635 const struct ip_vs_pe *pe;
639 struct rcu_head rcu_head;
642 /* Extended internal versions of struct ip_vs_service_user and ip_vs_dest_user
645 * We need these to conveniently pass around service and destination
646 * options, but unfortunately, we also need to keep the old definitions to
647 * maintain userspace backwards compatibility for the setsockopt interface.
649 struct ip_vs_service_user_kern {
650 /* virtual service addresses */
653 union nf_inet_addr addr; /* virtual ip address */
655 u32 fwmark; /* firwall mark of service */
657 /* virtual service options */
660 unsigned int flags; /* virtual service flags */
661 unsigned int timeout; /* persistent timeout in sec */
662 __be32 netmask; /* persistent netmask or plen */
666 struct ip_vs_dest_user_kern {
667 /* destination server address */
668 union nf_inet_addr addr;
671 /* real server options */
672 unsigned int conn_flags; /* connection flags */
673 int weight; /* destination weight */
675 /* thresholds for active connections */
676 u32 u_threshold; /* upper threshold */
677 u32 l_threshold; /* lower threshold */
679 /* Address family of addr */
682 u16 tun_type; /* tunnel type */
683 __be16 tun_port; /* tunnel port */
684 u16 tun_flags; /* tunnel flags */
689 * The information about the virtual service offered to the net and the
690 * forwarding entries.
692 struct ip_vs_service {
693 struct hlist_node s_list; /* for normal service table */
694 struct hlist_node f_list; /* for fwmark-based service table */
695 atomic_t refcnt; /* reference counter */
697 u16 af; /* address family */
698 __u16 protocol; /* which protocol (TCP/UDP) */
699 union nf_inet_addr addr; /* IP address for virtual service */
700 __be16 port; /* port number for the service */
701 __u32 fwmark; /* firewall mark of the service */
702 unsigned int flags; /* service status flags */
703 unsigned int timeout; /* persistent timeout in ticks */
704 __be32 netmask; /* grouping granularity, mask/plen */
705 struct netns_ipvs *ipvs;
707 struct list_head destinations; /* real server d-linked list */
708 __u32 num_dests; /* number of servers */
709 struct ip_vs_stats stats; /* statistics for the service */
712 struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */
713 spinlock_t sched_lock; /* lock sched_data */
714 void *sched_data; /* scheduler application data */
716 /* alternate persistence engine */
717 struct ip_vs_pe __rcu *pe;
718 int conntrack_afmask;
720 struct rcu_head rcu_head;
723 /* Information for cached dst */
724 struct ip_vs_dest_dst {
725 struct dst_entry *dst_cache; /* destination cache entry */
727 union nf_inet_addr dst_saddr;
728 struct rcu_head rcu_head;
731 /* The real server destination forwarding entry with ip address, port number,
735 struct list_head n_list; /* for the dests in the service */
736 struct hlist_node d_list; /* for table with all the dests */
738 u16 af; /* address family */
739 __be16 port; /* port number of the server */
740 union nf_inet_addr addr; /* IP address of the server */
741 volatile unsigned int flags; /* dest status flags */
742 atomic_t conn_flags; /* flags to copy to conn */
743 atomic_t weight; /* server weight */
744 atomic_t last_weight; /* server latest weight */
745 __u16 tun_type; /* tunnel type */
746 __be16 tun_port; /* tunnel port */
747 __u16 tun_flags; /* tunnel flags */
749 refcount_t refcnt; /* reference counter */
750 struct ip_vs_stats stats; /* statistics */
751 unsigned long idle_start; /* start time, jiffies */
753 /* connection counters and thresholds */
754 atomic_t activeconns; /* active connections */
755 atomic_t inactconns; /* inactive connections */
756 atomic_t persistconns; /* persistent connections */
757 __u32 u_threshold; /* upper threshold */
758 __u32 l_threshold; /* lower threshold */
760 /* for destination cache */
761 spinlock_t dst_lock; /* lock of dst_cache */
762 struct ip_vs_dest_dst __rcu *dest_dst; /* cached dst info */
764 /* for virtual service */
765 struct ip_vs_service __rcu *svc; /* service it belongs to */
766 __u16 protocol; /* which protocol (TCP/UDP) */
767 __be16 vport; /* virtual port number */
768 union nf_inet_addr vaddr; /* virtual IP address */
769 __u32 vfwmark; /* firewall mark of service */
771 struct rcu_head rcu_head;
772 struct list_head t_list; /* in dest_trash */
773 unsigned int in_rs_table:1; /* we are in rs_table */
776 /* The scheduler object */
777 struct ip_vs_scheduler {
778 struct list_head n_list; /* d-linked list head */
779 char *name; /* scheduler name */
780 atomic_t refcnt; /* reference counter */
781 struct module *module; /* THIS_MODULE/NULL */
783 /* scheduler initializing service */
784 int (*init_service)(struct ip_vs_service *svc);
785 /* scheduling service finish */
786 void (*done_service)(struct ip_vs_service *svc);
788 int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
789 /* dest is unlinked */
790 int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
791 /* dest is updated */
792 int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
794 /* selecting a server from the given service */
795 struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc,
796 const struct sk_buff *skb,
797 struct ip_vs_iphdr *iph);
800 /* The persistence engine object */
802 struct list_head n_list; /* d-linked list head */
803 char *name; /* scheduler name */
804 atomic_t refcnt; /* reference counter */
805 struct module *module; /* THIS_MODULE/NULL */
807 /* get the connection template, if any */
808 int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb);
809 bool (*ct_match)(const struct ip_vs_conn_param *p,
810 struct ip_vs_conn *ct);
811 u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval,
813 int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf);
814 /* create connections for real-server outgoing packets */
815 struct ip_vs_conn* (*conn_out)(struct ip_vs_service *svc,
816 struct ip_vs_dest *dest,
818 const struct ip_vs_iphdr *iph,
819 __be16 dport, __be16 cport);
822 /* The application module object (a.k.a. app incarnation) */
824 struct list_head a_list; /* member in app list */
825 int type; /* IP_VS_APP_TYPE_xxx */
826 char *name; /* application module name */
828 struct module *module; /* THIS_MODULE/NULL */
829 struct list_head incs_list; /* list of incarnations */
831 /* members for application incarnations */
832 struct list_head p_list; /* member in proto app list */
833 struct ip_vs_app *app; /* its real application */
834 __be16 port; /* port number in net order */
835 atomic_t usecnt; /* usage counter */
836 struct rcu_head rcu_head;
838 /* output hook: Process packet in inout direction, diff set for TCP.
839 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
840 * 2=Mangled but checksum was not updated
842 int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *,
843 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh);
845 /* input hook: Process packet in outin direction, diff set for TCP.
846 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
847 * 2=Mangled but checksum was not updated
849 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *,
850 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh);
852 /* ip_vs_app initializer */
853 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *);
855 /* ip_vs_app finish */
856 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *);
860 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *,
861 struct ip_vs_protocol *);
863 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *);
869 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app,
870 int *verdict, struct ip_vs_conn **cpp);
873 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app,
874 const struct iphdr *iph, int inverse);
877 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app,
878 const struct iphdr *iph, int inverse);
880 int (*state_transition)(struct ip_vs_conn *cp, int direction,
881 const struct sk_buff *skb,
882 struct ip_vs_app *app);
884 void (*timeout_change)(struct ip_vs_app *app, int flags);
887 struct ipvs_master_sync_state {
888 struct list_head sync_queue;
889 struct ip_vs_sync_buff *sync_buff;
890 unsigned long sync_queue_len;
891 unsigned int sync_queue_delay;
892 struct delayed_work master_wakeup_work;
893 struct netns_ipvs *ipvs;
896 struct ip_vs_sync_thread_data;
898 /* How much time to keep dests in trash */
899 #define IP_VS_DEST_TRASH_PERIOD (120 * HZ)
901 struct ipvs_sync_daemon_cfg {
902 union nf_inet_addr mcast_group;
908 /* multicast interface name */
909 char mcast_ifn[IP_VS_IFNAME_MAXLEN];
912 /* IPVS in network namespace */
914 int gen; /* Generation */
915 int enable; /* enable like nf_hooks do */
916 /* Hash table: for real service lookups */
917 #define IP_VS_RTAB_BITS 4
918 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
919 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
921 struct hlist_head rs_table[IP_VS_RTAB_SIZE];
923 struct list_head app_list;
925 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */
926 struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE];
927 /* ip_vs_proto_tcp */
928 #ifdef CONFIG_IP_VS_PROTO_TCP
929 #define TCP_APP_TAB_BITS 4
930 #define TCP_APP_TAB_SIZE (1 << TCP_APP_TAB_BITS)
931 #define TCP_APP_TAB_MASK (TCP_APP_TAB_SIZE - 1)
932 struct list_head tcp_apps[TCP_APP_TAB_SIZE];
934 /* ip_vs_proto_udp */
935 #ifdef CONFIG_IP_VS_PROTO_UDP
936 #define UDP_APP_TAB_BITS 4
937 #define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS)
938 #define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1)
939 struct list_head udp_apps[UDP_APP_TAB_SIZE];
941 /* ip_vs_proto_sctp */
942 #ifdef CONFIG_IP_VS_PROTO_SCTP
943 #define SCTP_APP_TAB_BITS 4
944 #define SCTP_APP_TAB_SIZE (1 << SCTP_APP_TAB_BITS)
945 #define SCTP_APP_TAB_MASK (SCTP_APP_TAB_SIZE - 1)
946 /* Hash table for SCTP application incarnations */
947 struct list_head sctp_apps[SCTP_APP_TAB_SIZE];
950 atomic_t conn_count; /* connection counter */
953 struct ip_vs_stats_rcu *tot_stats; /* Statistics & est. */
955 int num_services; /* no of virtual services */
956 int num_services6; /* IPv6 virtual services */
958 /* Trash for destinations */
959 struct list_head dest_trash;
960 spinlock_t dest_trash_lock;
961 struct timer_list dest_trash_timer; /* expiration timer */
962 /* Service counters */
963 atomic_t ftpsvc_counter;
964 atomic_t nullsvc_counter;
965 atomic_t conn_out_counter;
968 /* delayed work for expiring no dest connections */
969 struct delayed_work expire_nodest_conn_work;
970 /* 1/rate drop and drop-entry variables */
971 struct delayed_work defense_work; /* Work handler */
977 spinlock_t dropentry_lock; /* drop entry handling */
978 spinlock_t droppacket_lock; /* drop packet handling */
979 spinlock_t securetcp_lock; /* state and timeout tables */
982 struct ctl_table_header *sysctl_hdr;
983 struct ctl_table *sysctl_tbl;
986 /* sysctl variables */
987 int sysctl_amemthresh;
988 int sysctl_am_droprate;
989 int sysctl_drop_entry;
990 int sysctl_drop_packet;
991 int sysctl_secure_tcp;
992 #ifdef CONFIG_IP_VS_NFCT
993 int sysctl_conntrack;
995 int sysctl_snat_reroute;
997 int sysctl_sync_ports;
998 int sysctl_sync_persist_mode;
999 unsigned long sysctl_sync_qlen_max;
1000 int sysctl_sync_sock_size;
1001 int sysctl_cache_bypass;
1002 int sysctl_expire_nodest_conn;
1003 int sysctl_sloppy_tcp;
1004 int sysctl_sloppy_sctp;
1005 int sysctl_expire_quiescent_template;
1006 int sysctl_sync_threshold[2];
1007 unsigned int sysctl_sync_refresh_period;
1008 int sysctl_sync_retries;
1009 int sysctl_nat_icmp_send;
1010 int sysctl_pmtu_disc;
1011 int sysctl_backup_only;
1012 int sysctl_conn_reuse_mode;
1013 int sysctl_schedule_icmp;
1014 int sysctl_ignore_tunneled;
1015 int sysctl_run_estimation;
1016 #ifdef CONFIG_SYSCTL
1017 cpumask_var_t sysctl_est_cpulist; /* kthread cpumask */
1018 int est_cpulist_valid; /* cpulist set */
1019 int sysctl_est_nice; /* kthread nice */
1020 int est_stopped; /* stop tasks */
1024 int sysctl_lblc_expiration;
1025 struct ctl_table_header *lblc_ctl_header;
1026 struct ctl_table *lblc_ctl_table;
1028 int sysctl_lblcr_expiration;
1029 struct ctl_table_header *lblcr_ctl_header;
1030 struct ctl_table *lblcr_ctl_table;
1032 struct delayed_work est_reload_work;/* Reload kthread tasks */
1033 struct mutex est_mutex; /* protect kthread tasks */
1034 struct hlist_head est_temp_list; /* Ests during calc phase */
1035 struct ip_vs_est_kt_data **est_kt_arr; /* Array of kthread data ptrs */
1036 unsigned long est_max_threads;/* Hard limit of kthreads */
1037 int est_calc_phase; /* Calculation phase */
1038 int est_chain_max; /* Calculated chain_max */
1039 int est_kt_count; /* Allocated ptrs */
1040 int est_add_ktid; /* ktid where to add ests */
1041 atomic_t est_genid; /* kthreads reload genid */
1042 atomic_t est_genid_done; /* applied genid */
1044 spinlock_t sync_lock;
1045 struct ipvs_master_sync_state *ms;
1046 spinlock_t sync_buff_lock;
1047 struct ip_vs_sync_thread_data *master_tinfo;
1048 struct ip_vs_sync_thread_data *backup_tinfo;
1050 volatile int sync_state;
1051 struct mutex sync_mutex;
1052 struct ipvs_sync_daemon_cfg mcfg; /* Master Configuration */
1053 struct ipvs_sync_daemon_cfg bcfg; /* Backup Configuration */
1054 /* net name space ptr */
1055 struct net *net; /* Needed by timer routines */
1056 /* Number of heterogeneous destinations, needed becaus heterogeneous
1057 * are not supported when synchronization is enabled.
1059 unsigned int mixed_address_family_dests;
1060 unsigned int hooks_afmask; /* &1=AF_INET, &2=AF_INET6 */
1063 #define DEFAULT_SYNC_THRESHOLD 3
1064 #define DEFAULT_SYNC_PERIOD 50
1065 #define DEFAULT_SYNC_VER 1
1066 #define DEFAULT_SLOPPY_TCP 0
1067 #define DEFAULT_SLOPPY_SCTP 0
1068 #define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ)
1069 #define DEFAULT_SYNC_RETRIES 0
1070 #define IPVS_SYNC_WAKEUP_RATE 8
1071 #define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4)
1072 #define IPVS_SYNC_SEND_DELAY (HZ / 50)
1073 #define IPVS_SYNC_CHECK_PERIOD HZ
1074 #define IPVS_SYNC_FLUSH_TIME (HZ * 2)
1075 #define IPVS_SYNC_PORTS_MAX (1 << 6)
1077 #ifdef CONFIG_SYSCTL
1079 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1081 return ipvs->sysctl_sync_threshold[0];
1084 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1086 return READ_ONCE(ipvs->sysctl_sync_threshold[1]);
1089 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1091 return READ_ONCE(ipvs->sysctl_sync_refresh_period);
1094 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1096 return ipvs->sysctl_sync_retries;
1099 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1101 return ipvs->sysctl_sync_ver;
1104 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1106 return ipvs->sysctl_sloppy_tcp;
1109 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1111 return ipvs->sysctl_sloppy_sctp;
1114 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1116 return READ_ONCE(ipvs->sysctl_sync_ports);
1119 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1121 return ipvs->sysctl_sync_persist_mode;
1124 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1126 return ipvs->sysctl_sync_qlen_max;
1129 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1131 return ipvs->sysctl_sync_sock_size;
1134 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1136 return ipvs->sysctl_pmtu_disc;
1139 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1141 return ipvs->sync_state & IP_VS_STATE_BACKUP &&
1142 ipvs->sysctl_backup_only;
1145 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1147 return ipvs->sysctl_conn_reuse_mode;
1150 static inline int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs)
1152 return ipvs->sysctl_expire_nodest_conn;
1155 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1157 return ipvs->sysctl_schedule_icmp;
1160 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1162 return ipvs->sysctl_ignore_tunneled;
1165 static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1167 return ipvs->sysctl_cache_bypass;
1170 static inline int sysctl_run_estimation(struct netns_ipvs *ipvs)
1172 return ipvs->sysctl_run_estimation;
1175 static inline const struct cpumask *sysctl_est_cpulist(struct netns_ipvs *ipvs)
1177 if (ipvs->est_cpulist_valid)
1178 return ipvs->sysctl_est_cpulist;
1180 return housekeeping_cpumask(HK_TYPE_KTHREAD);
1183 static inline int sysctl_est_nice(struct netns_ipvs *ipvs)
1185 return ipvs->sysctl_est_nice;
1190 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1192 return DEFAULT_SYNC_THRESHOLD;
1195 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1197 return DEFAULT_SYNC_PERIOD;
1200 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1202 return DEFAULT_SYNC_REFRESH_PERIOD;
1205 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1207 return DEFAULT_SYNC_RETRIES & 3;
1210 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1212 return DEFAULT_SYNC_VER;
1215 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1217 return DEFAULT_SLOPPY_TCP;
1220 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1222 return DEFAULT_SLOPPY_SCTP;
1225 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1230 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1235 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1237 return IPVS_SYNC_QLEN_MAX;
1240 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1245 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1250 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1255 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1260 static inline int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs)
1265 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1270 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1275 static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1280 static inline int sysctl_run_estimation(struct netns_ipvs *ipvs)
1285 static inline const struct cpumask *sysctl_est_cpulist(struct netns_ipvs *ipvs)
1287 return housekeeping_cpumask(HK_TYPE_KTHREAD);
1290 static inline int sysctl_est_nice(struct netns_ipvs *ipvs)
1292 return IPVS_EST_NICE;
1297 /* IPVS core functions
1298 * (from ip_vs_core.c)
1300 const char *ip_vs_proto_name(unsigned int proto);
1301 void ip_vs_init_hash_table(struct list_head *table, int rows);
1302 struct ip_vs_conn *ip_vs_new_conn_out(struct ip_vs_service *svc,
1303 struct ip_vs_dest *dest,
1304 struct sk_buff *skb,
1305 const struct ip_vs_iphdr *iph,
1308 #define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t)))
1310 #define IP_VS_APP_TYPE_FTP 1
1312 /* ip_vs_conn handling functions
1313 * (from ip_vs_conn.c)
1316 IP_VS_DIR_INPUT = 0,
1318 IP_VS_DIR_INPUT_ONLY,
1322 static inline void ip_vs_conn_fill_param(struct netns_ipvs *ipvs, int af, int protocol,
1323 const union nf_inet_addr *caddr,
1325 const union nf_inet_addr *vaddr,
1327 struct ip_vs_conn_param *p)
1331 p->protocol = protocol;
1340 struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p);
1341 struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p);
1343 struct ip_vs_conn * ip_vs_conn_in_get_proto(struct netns_ipvs *ipvs, int af,
1344 const struct sk_buff *skb,
1345 const struct ip_vs_iphdr *iph);
1347 struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p);
1349 struct ip_vs_conn * ip_vs_conn_out_get_proto(struct netns_ipvs *ipvs, int af,
1350 const struct sk_buff *skb,
1351 const struct ip_vs_iphdr *iph);
1353 /* Get reference to gain full access to conn.
1354 * By default, RCU read-side critical sections have access only to
1355 * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference.
1357 static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp)
1359 return refcount_inc_not_zero(&cp->refcnt);
1362 /* put back the conn without restarting its timer */
1363 static inline void __ip_vs_conn_put(struct ip_vs_conn *cp)
1365 smp_mb__before_atomic();
1366 refcount_dec(&cp->refcnt);
1368 void ip_vs_conn_put(struct ip_vs_conn *cp);
1369 void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport);
1371 struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, int dest_af,
1372 const union nf_inet_addr *daddr,
1373 __be16 dport, unsigned int flags,
1374 struct ip_vs_dest *dest, __u32 fwmark);
1375 void ip_vs_conn_expire_now(struct ip_vs_conn *cp);
1377 const char *ip_vs_state_name(const struct ip_vs_conn *cp);
1379 void ip_vs_tcp_conn_listen(struct ip_vs_conn *cp);
1380 int ip_vs_check_template(struct ip_vs_conn *ct, struct ip_vs_dest *cdest);
1381 void ip_vs_random_dropentry(struct netns_ipvs *ipvs);
1382 int ip_vs_conn_init(void);
1383 void ip_vs_conn_cleanup(void);
1385 static inline void ip_vs_control_del(struct ip_vs_conn *cp)
1387 struct ip_vs_conn *ctl_cp = cp->control;
1389 IP_VS_ERR_BUF("request control DEL for uncontrolled: "
1391 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1393 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1399 IP_VS_DBG_BUF(7, "DELeting control for: "
1400 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1401 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1403 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1404 ntohs(ctl_cp->cport));
1407 if (atomic_read(&ctl_cp->n_control) == 0) {
1408 IP_VS_ERR_BUF("BUG control DEL with n=0 : "
1410 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1412 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1417 atomic_dec(&ctl_cp->n_control);
1421 ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp)
1424 IP_VS_ERR_BUF("request control ADD for already controlled: "
1426 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1428 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1431 ip_vs_control_del(cp);
1434 IP_VS_DBG_BUF(7, "ADDing control for: "
1435 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1436 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1438 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1439 ntohs(ctl_cp->cport));
1441 cp->control = ctl_cp;
1442 atomic_inc(&ctl_cp->n_control);
1445 /* Mark our template as assured */
1447 ip_vs_control_assure_ct(struct ip_vs_conn *cp)
1449 struct ip_vs_conn *ct = cp->control;
1451 if (ct && !(ct->state & IP_VS_CTPL_S_ASSURED) &&
1452 (ct->flags & IP_VS_CONN_F_TEMPLATE))
1453 ct->state |= IP_VS_CTPL_S_ASSURED;
1456 /* IPVS netns init & cleanup functions */
1457 int ip_vs_estimator_net_init(struct netns_ipvs *ipvs);
1458 int ip_vs_control_net_init(struct netns_ipvs *ipvs);
1459 int ip_vs_protocol_net_init(struct netns_ipvs *ipvs);
1460 int ip_vs_app_net_init(struct netns_ipvs *ipvs);
1461 int ip_vs_conn_net_init(struct netns_ipvs *ipvs);
1462 int ip_vs_sync_net_init(struct netns_ipvs *ipvs);
1463 void ip_vs_conn_net_cleanup(struct netns_ipvs *ipvs);
1464 void ip_vs_app_net_cleanup(struct netns_ipvs *ipvs);
1465 void ip_vs_protocol_net_cleanup(struct netns_ipvs *ipvs);
1466 void ip_vs_control_net_cleanup(struct netns_ipvs *ipvs);
1467 void ip_vs_estimator_net_cleanup(struct netns_ipvs *ipvs);
1468 void ip_vs_sync_net_cleanup(struct netns_ipvs *ipvs);
1469 void ip_vs_service_nets_cleanup(struct list_head *net_list);
1471 /* IPVS application functions
1472 * (from ip_vs_app.c)
1474 #define IP_VS_APP_MAX_PORTS 8
1475 struct ip_vs_app *register_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1476 void unregister_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1477 int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1478 void ip_vs_unbind_app(struct ip_vs_conn *cp);
1479 int register_ip_vs_app_inc(struct netns_ipvs *ipvs, struct ip_vs_app *app, __u16 proto,
1481 int ip_vs_app_inc_get(struct ip_vs_app *inc);
1482 void ip_vs_app_inc_put(struct ip_vs_app *inc);
1484 int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb,
1485 struct ip_vs_iphdr *ipvsh);
1486 int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb,
1487 struct ip_vs_iphdr *ipvsh);
1489 int register_ip_vs_pe(struct ip_vs_pe *pe);
1490 int unregister_ip_vs_pe(struct ip_vs_pe *pe);
1491 struct ip_vs_pe *ip_vs_pe_getbyname(const char *name);
1492 struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name);
1494 /* Use a #define to avoid all of module.h just for these trivial ops */
1495 #define ip_vs_pe_get(pe) \
1496 if (pe && pe->module) \
1497 __module_get(pe->module);
1499 #define ip_vs_pe_put(pe) \
1500 if (pe && pe->module) \
1501 module_put(pe->module);
1503 /* IPVS protocol functions (from ip_vs_proto.c) */
1504 int ip_vs_protocol_init(void);
1505 void ip_vs_protocol_cleanup(void);
1506 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags);
1507 int *ip_vs_create_timeout_table(int *table, int size);
1508 void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
1509 const struct sk_buff *skb, int offset,
1512 extern struct ip_vs_protocol ip_vs_protocol_tcp;
1513 extern struct ip_vs_protocol ip_vs_protocol_udp;
1514 extern struct ip_vs_protocol ip_vs_protocol_icmp;
1515 extern struct ip_vs_protocol ip_vs_protocol_esp;
1516 extern struct ip_vs_protocol ip_vs_protocol_ah;
1517 extern struct ip_vs_protocol ip_vs_protocol_sctp;
1519 /* Registering/unregistering scheduler functions
1520 * (from ip_vs_sched.c)
1522 int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1523 int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1524 int ip_vs_bind_scheduler(struct ip_vs_service *svc,
1525 struct ip_vs_scheduler *scheduler);
1526 void ip_vs_unbind_scheduler(struct ip_vs_service *svc,
1527 struct ip_vs_scheduler *sched);
1528 struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name);
1529 void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler);
1531 ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb,
1532 struct ip_vs_proto_data *pd, int *ignored,
1533 struct ip_vs_iphdr *iph);
1534 int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb,
1535 struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph);
1537 void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg);
1539 /* IPVS control data and functions (from ip_vs_ctl.c) */
1540 extern struct ip_vs_stats ip_vs_stats;
1541 extern int sysctl_ip_vs_sync_ver;
1543 struct ip_vs_service *
1544 ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
1545 const union nf_inet_addr *vaddr, __be16 vport);
1547 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1548 const union nf_inet_addr *daddr, __be16 dport);
1551 ip_vs_find_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1552 const union nf_inet_addr *daddr, __be16 dport);
1553 struct ip_vs_dest *ip_vs_find_tunnel(struct netns_ipvs *ipvs, int af,
1554 const union nf_inet_addr *daddr,
1557 int ip_vs_use_count_inc(void);
1558 void ip_vs_use_count_dec(void);
1559 int ip_vs_register_nl_ioctl(void);
1560 void ip_vs_unregister_nl_ioctl(void);
1561 int ip_vs_control_init(void);
1562 void ip_vs_control_cleanup(void);
1564 ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
1565 const union nf_inet_addr *daddr, __be16 dport,
1566 const union nf_inet_addr *vaddr, __be16 vport,
1567 __u16 protocol, __u32 fwmark, __u32 flags);
1568 void ip_vs_try_bind_dest(struct ip_vs_conn *cp);
1570 static inline void ip_vs_dest_hold(struct ip_vs_dest *dest)
1572 refcount_inc(&dest->refcnt);
1575 static inline void ip_vs_dest_put(struct ip_vs_dest *dest)
1577 smp_mb__before_atomic();
1578 refcount_dec(&dest->refcnt);
1581 static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest)
1583 if (refcount_dec_and_test(&dest->refcnt))
1587 /* IPVS sync daemon data and function prototypes
1588 * (from ip_vs_sync.c)
1590 int start_sync_thread(struct netns_ipvs *ipvs, struct ipvs_sync_daemon_cfg *cfg,
1592 int stop_sync_thread(struct netns_ipvs *ipvs, int state);
1593 void ip_vs_sync_conn(struct netns_ipvs *ipvs, struct ip_vs_conn *cp, int pkts);
1595 /* IPVS rate estimator prototypes (from ip_vs_est.c) */
1596 int ip_vs_start_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1597 void ip_vs_stop_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1598 void ip_vs_zero_estimator(struct ip_vs_stats *stats);
1599 void ip_vs_read_estimator(struct ip_vs_kstats *dst, struct ip_vs_stats *stats);
1600 void ip_vs_est_reload_start(struct netns_ipvs *ipvs);
1601 int ip_vs_est_kthread_start(struct netns_ipvs *ipvs,
1602 struct ip_vs_est_kt_data *kd);
1603 void ip_vs_est_kthread_stop(struct ip_vs_est_kt_data *kd);
1605 static inline void ip_vs_est_stopped_recalc(struct netns_ipvs *ipvs)
1607 #ifdef CONFIG_SYSCTL
1608 /* Stop tasks while cpulist is empty or if disabled with flag */
1609 ipvs->est_stopped = !sysctl_run_estimation(ipvs) ||
1610 (ipvs->est_cpulist_valid &&
1611 cpumask_empty(sysctl_est_cpulist(ipvs)));
1615 static inline bool ip_vs_est_stopped(struct netns_ipvs *ipvs)
1617 #ifdef CONFIG_SYSCTL
1618 return ipvs->est_stopped;
1624 static inline int ip_vs_est_max_threads(struct netns_ipvs *ipvs)
1626 unsigned int limit = IPVS_EST_CPU_KTHREADS *
1627 cpumask_weight(sysctl_est_cpulist(ipvs));
1629 return max(1U, limit);
1632 /* Various IPVS packet transmitters (from ip_vs_xmit.c) */
1633 int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1634 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1635 int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1636 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1637 int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1638 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1639 int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1640 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1641 int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1642 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1643 int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1644 struct ip_vs_protocol *pp, int offset,
1645 unsigned int hooknum, struct ip_vs_iphdr *iph);
1646 void ip_vs_dest_dst_rcu_free(struct rcu_head *head);
1648 #ifdef CONFIG_IP_VS_IPV6
1649 int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1650 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1651 int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1652 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1653 int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1654 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1655 int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1656 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1657 int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1658 struct ip_vs_protocol *pp, int offset,
1659 unsigned int hooknum, struct ip_vs_iphdr *iph);
1662 #ifdef CONFIG_SYSCTL
1663 /* This is a simple mechanism to ignore packets when
1664 * we are loaded. Just set ip_vs_drop_rate to 'n' and
1665 * we start to drop 1/rate of the packets
1667 static inline int ip_vs_todrop(struct netns_ipvs *ipvs)
1669 if (!ipvs->drop_rate)
1671 if (--ipvs->drop_counter > 0)
1673 ipvs->drop_counter = ipvs->drop_rate;
1677 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; }
1680 #ifdef CONFIG_SYSCTL
1681 /* Enqueue delayed work for expiring no dest connections
1682 * Only run when sysctl_expire_nodest=1
1684 static inline void ip_vs_enqueue_expire_nodest_conns(struct netns_ipvs *ipvs)
1686 if (sysctl_expire_nodest_conn(ipvs))
1687 queue_delayed_work(system_long_wq,
1688 &ipvs->expire_nodest_conn_work, 1);
1691 void ip_vs_expire_nodest_conn_flush(struct netns_ipvs *ipvs);
1693 static inline void ip_vs_enqueue_expire_nodest_conns(struct netns_ipvs *ipvs) {}
1696 #define IP_VS_DFWD_METHOD(dest) (atomic_read(&(dest)->conn_flags) & \
1697 IP_VS_CONN_F_FWD_MASK)
1699 /* ip_vs_fwd_tag returns the forwarding tag of the connection */
1700 #define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK)
1702 static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp)
1706 switch (IP_VS_FWD_METHOD(cp)) {
1707 case IP_VS_CONN_F_MASQ:
1709 case IP_VS_CONN_F_LOCALNODE:
1711 case IP_VS_CONN_F_TUNNEL:
1713 case IP_VS_CONN_F_DROUTE:
1715 case IP_VS_CONN_F_BYPASS:
1723 void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp,
1724 struct ip_vs_conn *cp, int dir);
1726 #ifdef CONFIG_IP_VS_IPV6
1727 void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp,
1728 struct ip_vs_conn *cp, int dir);
1731 __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset);
1733 static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum)
1735 __be32 diff[2] = { ~old, new };
1737 return csum_partial(diff, sizeof(diff), oldsum);
1740 #ifdef CONFIG_IP_VS_IPV6
1741 static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new,
1744 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0],
1745 new[3], new[2], new[1], new[0] };
1747 return csum_partial(diff, sizeof(diff), oldsum);
1751 static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum)
1753 __be16 diff[2] = { ~old, new };
1755 return csum_partial(diff, sizeof(diff), oldsum);
1758 /* Forget current conntrack (unconfirmed) and attach notrack entry */
1759 static inline void ip_vs_notrack(struct sk_buff *skb)
1761 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1762 enum ip_conntrack_info ctinfo;
1763 struct nf_conn *ct = nf_ct_get(skb, &ctinfo);
1766 nf_conntrack_put(&ct->ct_general);
1767 nf_ct_set(skb, NULL, IP_CT_UNTRACKED);
1772 #ifdef CONFIG_IP_VS_NFCT
1773 /* Netfilter connection tracking
1774 * (from ip_vs_nfct.c)
1776 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1778 #ifdef CONFIG_SYSCTL
1779 return ipvs->sysctl_conntrack;
1785 void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp,
1787 int ip_vs_confirm_conntrack(struct sk_buff *skb);
1788 void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct,
1789 struct ip_vs_conn *cp, u_int8_t proto,
1790 const __be16 port, int from_rs);
1791 void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp);
1795 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1800 static inline void ip_vs_update_conntrack(struct sk_buff *skb,
1801 struct ip_vs_conn *cp, int outin)
1805 static inline int ip_vs_confirm_conntrack(struct sk_buff *skb)
1810 static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp)
1813 #endif /* CONFIG_IP_VS_NFCT */
1815 /* Using old conntrack that can not be redirected to another real server? */
1816 static inline bool ip_vs_conn_uses_old_conntrack(struct ip_vs_conn *cp,
1817 struct sk_buff *skb)
1819 #ifdef CONFIG_IP_VS_NFCT
1820 enum ip_conntrack_info ctinfo;
1823 ct = nf_ct_get(skb, &ctinfo);
1824 if (ct && nf_ct_is_confirmed(ct))
1830 static inline int ip_vs_register_conntrack(struct ip_vs_service *svc)
1832 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
1833 int afmask = (svc->af == AF_INET6) ? 2 : 1;
1836 if (!(svc->conntrack_afmask & afmask)) {
1837 ret = nf_ct_netns_get(svc->ipvs->net, svc->af);
1839 svc->conntrack_afmask |= afmask;
1847 static inline void ip_vs_unregister_conntrack(struct ip_vs_service *svc)
1849 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
1850 int afmask = (svc->af == AF_INET6) ? 2 : 1;
1852 if (svc->conntrack_afmask & afmask) {
1853 nf_ct_netns_put(svc->ipvs->net, svc->af);
1854 svc->conntrack_afmask &= ~afmask;
1859 int ip_vs_register_hooks(struct netns_ipvs *ipvs, unsigned int af);
1860 void ip_vs_unregister_hooks(struct netns_ipvs *ipvs, unsigned int af);
1863 ip_vs_dest_conn_overhead(struct ip_vs_dest *dest)
1865 /* We think the overhead of processing active connections is 256
1866 * times higher than that of inactive connections in average. (This
1867 * 256 times might not be accurate, we will change it later) We
1868 * use the following formula to estimate the overhead now:
1869 * dest->activeconns*256 + dest->inactconns
1871 return (atomic_read(&dest->activeconns) << 8) +
1872 atomic_read(&dest->inactconns);
1875 #ifdef CONFIG_IP_VS_PROTO_TCP
1876 INDIRECT_CALLABLE_DECLARE(int
1877 tcp_snat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
1878 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph));
1881 #ifdef CONFIG_IP_VS_PROTO_UDP
1882 INDIRECT_CALLABLE_DECLARE(int
1883 udp_snat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
1884 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph));
1886 #endif /* _NET_IP_VS_H */