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 /* The port number of FTP service (in network order). */
269 #define FTPPORT cpu_to_be16(21)
270 #define FTPDATA cpu_to_be16(20)
272 /* TCP State Values */
274 IP_VS_TCP_S_NONE = 0,
275 IP_VS_TCP_S_ESTABLISHED,
276 IP_VS_TCP_S_SYN_SENT,
277 IP_VS_TCP_S_SYN_RECV,
278 IP_VS_TCP_S_FIN_WAIT,
279 IP_VS_TCP_S_TIME_WAIT,
281 IP_VS_TCP_S_CLOSE_WAIT,
282 IP_VS_TCP_S_LAST_ACK,
288 /* UDP State Values */
294 /* ICMP State Values */
300 /* SCTP State Values */
301 enum ip_vs_sctp_states {
305 IP_VS_SCTP_S_COOKIE_SENT,
306 IP_VS_SCTP_S_COOKIE_REPLIED,
307 IP_VS_SCTP_S_COOKIE_WAIT,
309 IP_VS_SCTP_S_COOKIE_ECHOED,
310 IP_VS_SCTP_S_ESTABLISHED,
311 IP_VS_SCTP_S_SHUTDOWN_SENT,
312 IP_VS_SCTP_S_SHUTDOWN_RECEIVED,
313 IP_VS_SCTP_S_SHUTDOWN_ACK_SENT,
314 IP_VS_SCTP_S_REJECTED,
319 /* Connection templates use bits from state */
320 #define IP_VS_CTPL_S_NONE 0x0000
321 #define IP_VS_CTPL_S_ASSURED 0x0001
322 #define IP_VS_CTPL_S_LAST 0x0002
324 /* Delta sequence info structure
325 * Each ip_vs_conn has 2 (output AND input seq. changes).
326 * Only used in the VS/NAT.
329 __u32 init_seq; /* Add delta from this seq */
330 __u32 delta; /* Delta in sequence numbers */
331 __u32 previous_delta; /* Delta in sequence numbers
332 * before last resized pkt */
335 /* counters per cpu */
336 struct ip_vs_counters {
337 u64_stats_t conns; /* connections scheduled */
338 u64_stats_t inpkts; /* incoming packets */
339 u64_stats_t outpkts; /* outgoing packets */
340 u64_stats_t inbytes; /* incoming bytes */
341 u64_stats_t outbytes; /* outgoing bytes */
344 struct ip_vs_cpu_stats {
345 struct ip_vs_counters cnt;
346 struct u64_stats_sync syncp;
349 /* Default nice for estimator kthreads */
350 #define IPVS_EST_NICE 0
352 /* IPVS statistics objects */
353 struct ip_vs_estimator {
354 struct hlist_node list;
368 s32 ktid:16, /* kthread ID, -1=temp list */
369 ktrow:8, /* row/tick ID for kthread */
370 ktcid:8; /* chain ID for kthread tick */
374 * IPVS statistics object, 64-bit kernel version of struct ip_vs_stats_user
376 struct ip_vs_kstats {
377 u64 conns; /* connections scheduled */
378 u64 inpkts; /* incoming packets */
379 u64 outpkts; /* outgoing packets */
380 u64 inbytes; /* incoming bytes */
381 u64 outbytes; /* outgoing bytes */
383 u64 cps; /* current connection rate */
384 u64 inpps; /* current in packet rate */
385 u64 outpps; /* current out packet rate */
386 u64 inbps; /* current in byte rate */
387 u64 outbps; /* current out byte rate */
391 struct ip_vs_kstats kstats; /* kernel statistics */
392 struct ip_vs_estimator est; /* estimator */
393 struct ip_vs_cpu_stats __percpu *cpustats; /* per cpu counters */
394 spinlock_t lock; /* spin lock */
395 struct ip_vs_kstats kstats0; /* reset values */
398 struct ip_vs_stats_rcu {
399 struct ip_vs_stats s;
400 struct rcu_head rcu_head;
403 int ip_vs_stats_init_alloc(struct ip_vs_stats *s);
404 struct ip_vs_stats *ip_vs_stats_alloc(void);
405 void ip_vs_stats_release(struct ip_vs_stats *stats);
406 void ip_vs_stats_free(struct ip_vs_stats *stats);
408 /* Process estimators in multiple timer ticks (20/50/100, see ktrow) */
409 #define IPVS_EST_NTICKS 50
410 /* Estimation uses a 2-second period containing ticks (in jiffies) */
411 #define IPVS_EST_TICK ((2 * HZ) / IPVS_EST_NTICKS)
413 /* Limit of CPU load per kthread (8 for 12.5%), ratio of CPU capacity (1/C).
414 * Value of 4 and above ensures kthreads will take work without exceeding
415 * the CPU capacity under different circumstances.
417 #define IPVS_EST_LOAD_DIVISOR 8
419 /* Kthreads should not have work that exceeds the CPU load above 50% */
420 #define IPVS_EST_CPU_KTHREADS (IPVS_EST_LOAD_DIVISOR / 2)
422 /* Desired number of chains per timer tick (chain load factor in 100us units),
423 * 48=4.8ms of 40ms tick (12% CPU usage):
424 * 2 sec * 1000 ms in sec * 10 (100us in ms) / 8 (12.5%) / 50
426 #define IPVS_EST_CHAIN_FACTOR \
427 ALIGN_DOWN(2 * 1000 * 10 / IPVS_EST_LOAD_DIVISOR / IPVS_EST_NTICKS, 8)
429 /* Compiled number of chains per tick
430 * The defines should match cond_resched_rcu
432 #if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
433 #define IPVS_EST_TICK_CHAINS IPVS_EST_CHAIN_FACTOR
435 #define IPVS_EST_TICK_CHAINS 1
438 #if IPVS_EST_NTICKS > 127
439 #error Too many timer ticks for ktrow
442 /* Multiple chains processed in same tick */
443 struct ip_vs_est_tick_data {
444 struct rcu_head rcu_head;
445 struct hlist_head chains[IPVS_EST_TICK_CHAINS];
446 DECLARE_BITMAP(present, IPVS_EST_TICK_CHAINS);
447 DECLARE_BITMAP(full, IPVS_EST_TICK_CHAINS);
448 int chain_len[IPVS_EST_TICK_CHAINS];
451 /* Context for estimation kthread */
452 struct ip_vs_est_kt_data {
453 struct netns_ipvs *ipvs;
454 struct task_struct *task; /* task if running */
455 struct ip_vs_est_tick_data __rcu *ticks[IPVS_EST_NTICKS];
456 DECLARE_BITMAP(avail, IPVS_EST_NTICKS); /* tick has space for ests */
457 unsigned long est_timer; /* estimation timer (jiffies) */
458 struct ip_vs_stats *calc_stats; /* Used for calculation */
459 int tick_len[IPVS_EST_NTICKS]; /* est count */
460 int id; /* ktid per netns */
461 int chain_max; /* max ests per tick chain */
462 int tick_max; /* max ests per tick */
463 int est_count; /* attached ests to kthread */
464 int est_max_count; /* max ests per kthread */
465 int add_row; /* row for new ests */
466 int est_row; /* estimated row */
474 struct ip_vs_proto_data;
476 struct ip_vs_protocol {
477 struct ip_vs_protocol *next;
483 void (*init)(struct ip_vs_protocol *pp);
485 void (*exit)(struct ip_vs_protocol *pp);
487 int (*init_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
489 void (*exit_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
491 int (*conn_schedule)(struct netns_ipvs *ipvs,
492 int af, struct sk_buff *skb,
493 struct ip_vs_proto_data *pd,
494 int *verdict, struct ip_vs_conn **cpp,
495 struct ip_vs_iphdr *iph);
498 (*conn_in_get)(struct netns_ipvs *ipvs,
500 const struct sk_buff *skb,
501 const struct ip_vs_iphdr *iph);
504 (*conn_out_get)(struct netns_ipvs *ipvs,
506 const struct sk_buff *skb,
507 const struct ip_vs_iphdr *iph);
509 int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
510 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
512 int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
513 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
515 const char *(*state_name)(int state);
517 void (*state_transition)(struct ip_vs_conn *cp, int direction,
518 const struct sk_buff *skb,
519 struct ip_vs_proto_data *pd);
521 int (*register_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
523 void (*unregister_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
525 int (*app_conn_bind)(struct ip_vs_conn *cp);
527 void (*debug_packet)(int af, struct ip_vs_protocol *pp,
528 const struct sk_buff *skb,
532 void (*timeout_change)(struct ip_vs_proto_data *pd, int flags);
535 /* protocol data per netns */
536 struct ip_vs_proto_data {
537 struct ip_vs_proto_data *next;
538 struct ip_vs_protocol *pp;
539 int *timeout_table; /* protocol timeout table */
540 atomic_t appcnt; /* counter of proto app incs. */
541 struct tcp_states_t *tcp_state_table;
544 struct ip_vs_protocol *ip_vs_proto_get(unsigned short proto);
545 struct ip_vs_proto_data *ip_vs_proto_data_get(struct netns_ipvs *ipvs,
546 unsigned short proto);
548 struct ip_vs_conn_param {
549 struct netns_ipvs *ipvs;
550 const union nf_inet_addr *caddr;
551 const union nf_inet_addr *vaddr;
557 const struct ip_vs_pe *pe;
562 /* IP_VS structure allocated for each dynamically scheduled connection */
564 struct hlist_node c_list; /* hashed list heads */
565 /* Protocol, addresses and port numbers */
569 u16 af; /* address family */
570 union nf_inet_addr caddr; /* client address */
571 union nf_inet_addr vaddr; /* virtual address */
572 union nf_inet_addr daddr; /* destination address */
573 volatile __u32 flags; /* status flags */
574 __u16 protocol; /* Which protocol (TCP/UDP) */
575 __u16 daf; /* Address family of the dest */
576 struct netns_ipvs *ipvs;
578 /* counter and timer */
579 refcount_t refcnt; /* reference count */
580 struct timer_list timer; /* Expiration timer */
581 volatile unsigned long timeout; /* timeout */
583 /* Flags and state transition */
584 spinlock_t lock; /* lock for state transition */
585 volatile __u16 state; /* state info */
586 volatile __u16 old_state; /* old state, to be used for
587 * state transition triggered
590 __u32 fwmark; /* Fire wall mark from skb */
591 unsigned long sync_endtime; /* jiffies + sent_retries */
593 /* Control members */
594 struct ip_vs_conn *control; /* Master control connection */
595 atomic_t n_control; /* Number of controlled ones */
596 struct ip_vs_dest *dest; /* real server */
597 atomic_t in_pkts; /* incoming packet counter */
599 /* Packet transmitter for different forwarding methods. If it
600 * mangles the packet, it must return NF_DROP or better NF_STOLEN,
601 * otherwise this must be changed to a sk_buff **.
602 * NF_ACCEPT can be returned when destination is local.
604 int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp,
605 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
607 /* Note: we can group the following members into a structure,
608 * in order to save more space, and the following members are
609 * only used in VS/NAT anyway
611 struct ip_vs_app *app; /* bound ip_vs_app object */
612 void *app_data; /* Application private data */
613 struct_group(sync_conn_opt,
614 struct ip_vs_seq in_seq; /* incoming seq. struct */
615 struct ip_vs_seq out_seq; /* outgoing seq. struct */
618 const struct ip_vs_pe *pe;
622 struct rcu_head rcu_head;
625 /* Extended internal versions of struct ip_vs_service_user and ip_vs_dest_user
628 * We need these to conveniently pass around service and destination
629 * options, but unfortunately, we also need to keep the old definitions to
630 * maintain userspace backwards compatibility for the setsockopt interface.
632 struct ip_vs_service_user_kern {
633 /* virtual service addresses */
636 union nf_inet_addr addr; /* virtual ip address */
638 u32 fwmark; /* firewall mark of service */
640 /* virtual service options */
643 unsigned int flags; /* virtual service flags */
644 unsigned int timeout; /* persistent timeout in sec */
645 __be32 netmask; /* persistent netmask or plen */
649 struct ip_vs_dest_user_kern {
650 /* destination server address */
651 union nf_inet_addr addr;
654 /* real server options */
655 unsigned int conn_flags; /* connection flags */
656 int weight; /* destination weight */
658 /* thresholds for active connections */
659 u32 u_threshold; /* upper threshold */
660 u32 l_threshold; /* lower threshold */
662 /* Address family of addr */
665 u16 tun_type; /* tunnel type */
666 __be16 tun_port; /* tunnel port */
667 u16 tun_flags; /* tunnel flags */
672 * The information about the virtual service offered to the net and the
673 * forwarding entries.
675 struct ip_vs_service {
676 struct hlist_node s_list; /* for normal service table */
677 struct hlist_node f_list; /* for fwmark-based service table */
678 atomic_t refcnt; /* reference counter */
680 u16 af; /* address family */
681 __u16 protocol; /* which protocol (TCP/UDP) */
682 union nf_inet_addr addr; /* IP address for virtual service */
683 __be16 port; /* port number for the service */
684 __u32 fwmark; /* firewall mark of the service */
685 unsigned int flags; /* service status flags */
686 unsigned int timeout; /* persistent timeout in ticks */
687 __be32 netmask; /* grouping granularity, mask/plen */
688 struct netns_ipvs *ipvs;
690 struct list_head destinations; /* real server d-linked list */
691 __u32 num_dests; /* number of servers */
692 struct ip_vs_stats stats; /* statistics for the service */
695 struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */
696 spinlock_t sched_lock; /* lock sched_data */
697 void *sched_data; /* scheduler application data */
699 /* alternate persistence engine */
700 struct ip_vs_pe __rcu *pe;
701 int conntrack_afmask;
703 struct rcu_head rcu_head;
706 /* Information for cached dst */
707 struct ip_vs_dest_dst {
708 struct dst_entry *dst_cache; /* destination cache entry */
710 union nf_inet_addr dst_saddr;
711 struct rcu_head rcu_head;
714 /* The real server destination forwarding entry with ip address, port number,
718 struct list_head n_list; /* for the dests in the service */
719 struct hlist_node d_list; /* for table with all the dests */
721 u16 af; /* address family */
722 __be16 port; /* port number of the server */
723 union nf_inet_addr addr; /* IP address of the server */
724 volatile unsigned int flags; /* dest status flags */
725 atomic_t conn_flags; /* flags to copy to conn */
726 atomic_t weight; /* server weight */
727 atomic_t last_weight; /* server latest weight */
728 __u16 tun_type; /* tunnel type */
729 __be16 tun_port; /* tunnel port */
730 __u16 tun_flags; /* tunnel flags */
732 refcount_t refcnt; /* reference counter */
733 struct ip_vs_stats stats; /* statistics */
734 unsigned long idle_start; /* start time, jiffies */
736 /* connection counters and thresholds */
737 atomic_t activeconns; /* active connections */
738 atomic_t inactconns; /* inactive connections */
739 atomic_t persistconns; /* persistent connections */
740 __u32 u_threshold; /* upper threshold */
741 __u32 l_threshold; /* lower threshold */
743 /* for destination cache */
744 spinlock_t dst_lock; /* lock of dst_cache */
745 struct ip_vs_dest_dst __rcu *dest_dst; /* cached dst info */
747 /* for virtual service */
748 struct ip_vs_service __rcu *svc; /* service it belongs to */
749 __u16 protocol; /* which protocol (TCP/UDP) */
750 __be16 vport; /* virtual port number */
751 union nf_inet_addr vaddr; /* virtual IP address */
752 __u32 vfwmark; /* firewall mark of service */
754 struct rcu_head rcu_head;
755 struct list_head t_list; /* in dest_trash */
756 unsigned int in_rs_table:1; /* we are in rs_table */
759 /* The scheduler object */
760 struct ip_vs_scheduler {
761 struct list_head n_list; /* d-linked list head */
762 char *name; /* scheduler name */
763 atomic_t refcnt; /* reference counter */
764 struct module *module; /* THIS_MODULE/NULL */
766 /* scheduler initializing service */
767 int (*init_service)(struct ip_vs_service *svc);
768 /* scheduling service finish */
769 void (*done_service)(struct ip_vs_service *svc);
771 int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
772 /* dest is unlinked */
773 int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
774 /* dest is updated */
775 int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
777 /* selecting a server from the given service */
778 struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc,
779 const struct sk_buff *skb,
780 struct ip_vs_iphdr *iph);
783 /* The persistence engine object */
785 struct list_head n_list; /* d-linked list head */
786 char *name; /* scheduler name */
787 atomic_t refcnt; /* reference counter */
788 struct module *module; /* THIS_MODULE/NULL */
790 /* get the connection template, if any */
791 int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb);
792 bool (*ct_match)(const struct ip_vs_conn_param *p,
793 struct ip_vs_conn *ct);
794 u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval,
796 int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf);
797 /* create connections for real-server outgoing packets */
798 struct ip_vs_conn* (*conn_out)(struct ip_vs_service *svc,
799 struct ip_vs_dest *dest,
801 const struct ip_vs_iphdr *iph,
802 __be16 dport, __be16 cport);
805 /* The application module object (a.k.a. app incarnation) */
807 struct list_head a_list; /* member in app list */
808 int type; /* IP_VS_APP_TYPE_xxx */
809 char *name; /* application module name */
811 struct module *module; /* THIS_MODULE/NULL */
812 struct list_head incs_list; /* list of incarnations */
814 /* members for application incarnations */
815 struct list_head p_list; /* member in proto app list */
816 struct ip_vs_app *app; /* its real application */
817 __be16 port; /* port number in net order */
818 atomic_t usecnt; /* usage counter */
819 struct rcu_head rcu_head;
821 /* output hook: Process packet in inout direction, diff set for TCP.
822 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
823 * 2=Mangled but checksum was not updated
825 int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *,
826 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh);
828 /* input hook: Process packet in outin direction, diff set for TCP.
829 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
830 * 2=Mangled but checksum was not updated
832 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *,
833 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh);
835 /* ip_vs_app initializer */
836 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *);
838 /* ip_vs_app finish */
839 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *);
843 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *,
844 struct ip_vs_protocol *);
846 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *);
852 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app,
853 int *verdict, struct ip_vs_conn **cpp);
856 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app,
857 const struct iphdr *iph, int inverse);
860 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app,
861 const struct iphdr *iph, int inverse);
863 int (*state_transition)(struct ip_vs_conn *cp, int direction,
864 const struct sk_buff *skb,
865 struct ip_vs_app *app);
867 void (*timeout_change)(struct ip_vs_app *app, int flags);
870 struct ipvs_master_sync_state {
871 struct list_head sync_queue;
872 struct ip_vs_sync_buff *sync_buff;
873 unsigned long sync_queue_len;
874 unsigned int sync_queue_delay;
875 struct delayed_work master_wakeup_work;
876 struct netns_ipvs *ipvs;
879 struct ip_vs_sync_thread_data;
881 /* How much time to keep dests in trash */
882 #define IP_VS_DEST_TRASH_PERIOD (120 * HZ)
884 struct ipvs_sync_daemon_cfg {
885 union nf_inet_addr mcast_group;
891 /* multicast interface name */
892 char mcast_ifn[IP_VS_IFNAME_MAXLEN];
895 /* IPVS in network namespace */
897 int gen; /* Generation */
898 int enable; /* enable like nf_hooks do */
899 /* Hash table: for real service lookups */
900 #define IP_VS_RTAB_BITS 4
901 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
902 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
904 struct hlist_head rs_table[IP_VS_RTAB_SIZE];
906 struct list_head app_list;
908 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */
909 struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE];
910 /* ip_vs_proto_tcp */
911 #ifdef CONFIG_IP_VS_PROTO_TCP
912 #define TCP_APP_TAB_BITS 4
913 #define TCP_APP_TAB_SIZE (1 << TCP_APP_TAB_BITS)
914 #define TCP_APP_TAB_MASK (TCP_APP_TAB_SIZE - 1)
915 struct list_head tcp_apps[TCP_APP_TAB_SIZE];
917 /* ip_vs_proto_udp */
918 #ifdef CONFIG_IP_VS_PROTO_UDP
919 #define UDP_APP_TAB_BITS 4
920 #define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS)
921 #define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1)
922 struct list_head udp_apps[UDP_APP_TAB_SIZE];
924 /* ip_vs_proto_sctp */
925 #ifdef CONFIG_IP_VS_PROTO_SCTP
926 #define SCTP_APP_TAB_BITS 4
927 #define SCTP_APP_TAB_SIZE (1 << SCTP_APP_TAB_BITS)
928 #define SCTP_APP_TAB_MASK (SCTP_APP_TAB_SIZE - 1)
929 /* Hash table for SCTP application incarnations */
930 struct list_head sctp_apps[SCTP_APP_TAB_SIZE];
933 atomic_t conn_count; /* connection counter */
936 struct ip_vs_stats_rcu *tot_stats; /* Statistics & est. */
938 int num_services; /* no of virtual services */
939 int num_services6; /* IPv6 virtual services */
941 /* Trash for destinations */
942 struct list_head dest_trash;
943 spinlock_t dest_trash_lock;
944 struct timer_list dest_trash_timer; /* expiration timer */
945 /* Service counters */
946 atomic_t ftpsvc_counter;
947 atomic_t nullsvc_counter;
948 atomic_t conn_out_counter;
951 /* delayed work for expiring no dest connections */
952 struct delayed_work expire_nodest_conn_work;
953 /* 1/rate drop and drop-entry variables */
954 struct delayed_work defense_work; /* Work handler */
960 spinlock_t dropentry_lock; /* drop entry handling */
961 spinlock_t droppacket_lock; /* drop packet handling */
962 spinlock_t securetcp_lock; /* state and timeout tables */
965 struct ctl_table_header *sysctl_hdr;
966 struct ctl_table *sysctl_tbl;
969 /* sysctl variables */
970 int sysctl_amemthresh;
971 int sysctl_am_droprate;
972 int sysctl_drop_entry;
973 int sysctl_drop_packet;
974 int sysctl_secure_tcp;
975 #ifdef CONFIG_IP_VS_NFCT
976 int sysctl_conntrack;
978 int sysctl_snat_reroute;
980 int sysctl_sync_ports;
981 int sysctl_sync_persist_mode;
982 unsigned long sysctl_sync_qlen_max;
983 int sysctl_sync_sock_size;
984 int sysctl_cache_bypass;
985 int sysctl_expire_nodest_conn;
986 int sysctl_sloppy_tcp;
987 int sysctl_sloppy_sctp;
988 int sysctl_expire_quiescent_template;
989 int sysctl_sync_threshold[2];
990 unsigned int sysctl_sync_refresh_period;
991 int sysctl_sync_retries;
992 int sysctl_nat_icmp_send;
993 int sysctl_pmtu_disc;
994 int sysctl_backup_only;
995 int sysctl_conn_reuse_mode;
996 int sysctl_schedule_icmp;
997 int sysctl_ignore_tunneled;
998 int sysctl_run_estimation;
1000 cpumask_var_t sysctl_est_cpulist; /* kthread cpumask */
1001 int est_cpulist_valid; /* cpulist set */
1002 int sysctl_est_nice; /* kthread nice */
1003 int est_stopped; /* stop tasks */
1007 int sysctl_lblc_expiration;
1008 struct ctl_table_header *lblc_ctl_header;
1009 struct ctl_table *lblc_ctl_table;
1011 int sysctl_lblcr_expiration;
1012 struct ctl_table_header *lblcr_ctl_header;
1013 struct ctl_table *lblcr_ctl_table;
1015 struct delayed_work est_reload_work;/* Reload kthread tasks */
1016 struct mutex est_mutex; /* protect kthread tasks */
1017 struct hlist_head est_temp_list; /* Ests during calc phase */
1018 struct ip_vs_est_kt_data **est_kt_arr; /* Array of kthread data ptrs */
1019 unsigned long est_max_threads;/* Hard limit of kthreads */
1020 int est_calc_phase; /* Calculation phase */
1021 int est_chain_max; /* Calculated chain_max */
1022 int est_kt_count; /* Allocated ptrs */
1023 int est_add_ktid; /* ktid where to add ests */
1024 atomic_t est_genid; /* kthreads reload genid */
1025 atomic_t est_genid_done; /* applied genid */
1027 spinlock_t sync_lock;
1028 struct ipvs_master_sync_state *ms;
1029 spinlock_t sync_buff_lock;
1030 struct ip_vs_sync_thread_data *master_tinfo;
1031 struct ip_vs_sync_thread_data *backup_tinfo;
1033 volatile int sync_state;
1034 struct mutex sync_mutex;
1035 struct ipvs_sync_daemon_cfg mcfg; /* Master Configuration */
1036 struct ipvs_sync_daemon_cfg bcfg; /* Backup Configuration */
1037 /* net name space ptr */
1038 struct net *net; /* Needed by timer routines */
1039 /* Number of heterogeneous destinations, needed because heterogeneous
1040 * are not supported when synchronization is enabled.
1042 unsigned int mixed_address_family_dests;
1043 unsigned int hooks_afmask; /* &1=AF_INET, &2=AF_INET6 */
1046 #define DEFAULT_SYNC_THRESHOLD 3
1047 #define DEFAULT_SYNC_PERIOD 50
1048 #define DEFAULT_SYNC_VER 1
1049 #define DEFAULT_SLOPPY_TCP 0
1050 #define DEFAULT_SLOPPY_SCTP 0
1051 #define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ)
1052 #define DEFAULT_SYNC_RETRIES 0
1053 #define IPVS_SYNC_WAKEUP_RATE 8
1054 #define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4)
1055 #define IPVS_SYNC_SEND_DELAY (HZ / 50)
1056 #define IPVS_SYNC_CHECK_PERIOD HZ
1057 #define IPVS_SYNC_FLUSH_TIME (HZ * 2)
1058 #define IPVS_SYNC_PORTS_MAX (1 << 6)
1060 #ifdef CONFIG_SYSCTL
1062 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1064 return ipvs->sysctl_sync_threshold[0];
1067 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1069 return READ_ONCE(ipvs->sysctl_sync_threshold[1]);
1072 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1074 return READ_ONCE(ipvs->sysctl_sync_refresh_period);
1077 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1079 return ipvs->sysctl_sync_retries;
1082 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1084 return ipvs->sysctl_sync_ver;
1087 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1089 return ipvs->sysctl_sloppy_tcp;
1092 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1094 return ipvs->sysctl_sloppy_sctp;
1097 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1099 return READ_ONCE(ipvs->sysctl_sync_ports);
1102 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1104 return ipvs->sysctl_sync_persist_mode;
1107 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1109 return ipvs->sysctl_sync_qlen_max;
1112 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1114 return ipvs->sysctl_sync_sock_size;
1117 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1119 return ipvs->sysctl_pmtu_disc;
1122 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1124 return ipvs->sync_state & IP_VS_STATE_BACKUP &&
1125 ipvs->sysctl_backup_only;
1128 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1130 return ipvs->sysctl_conn_reuse_mode;
1133 static inline int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs)
1135 return ipvs->sysctl_expire_nodest_conn;
1138 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1140 return ipvs->sysctl_schedule_icmp;
1143 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1145 return ipvs->sysctl_ignore_tunneled;
1148 static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1150 return ipvs->sysctl_cache_bypass;
1153 static inline int sysctl_run_estimation(struct netns_ipvs *ipvs)
1155 return ipvs->sysctl_run_estimation;
1158 static inline const struct cpumask *sysctl_est_cpulist(struct netns_ipvs *ipvs)
1160 if (ipvs->est_cpulist_valid)
1161 return ipvs->sysctl_est_cpulist;
1163 return housekeeping_cpumask(HK_TYPE_KTHREAD);
1166 static inline int sysctl_est_nice(struct netns_ipvs *ipvs)
1168 return ipvs->sysctl_est_nice;
1173 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1175 return DEFAULT_SYNC_THRESHOLD;
1178 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1180 return DEFAULT_SYNC_PERIOD;
1183 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1185 return DEFAULT_SYNC_REFRESH_PERIOD;
1188 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1190 return DEFAULT_SYNC_RETRIES & 3;
1193 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1195 return DEFAULT_SYNC_VER;
1198 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1200 return DEFAULT_SLOPPY_TCP;
1203 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1205 return DEFAULT_SLOPPY_SCTP;
1208 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1213 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1218 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1220 return IPVS_SYNC_QLEN_MAX;
1223 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1228 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1233 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1238 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1243 static inline int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs)
1248 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1253 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1258 static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1263 static inline int sysctl_run_estimation(struct netns_ipvs *ipvs)
1268 static inline const struct cpumask *sysctl_est_cpulist(struct netns_ipvs *ipvs)
1270 return housekeeping_cpumask(HK_TYPE_KTHREAD);
1273 static inline int sysctl_est_nice(struct netns_ipvs *ipvs)
1275 return IPVS_EST_NICE;
1280 /* IPVS core functions
1281 * (from ip_vs_core.c)
1283 const char *ip_vs_proto_name(unsigned int proto);
1284 void ip_vs_init_hash_table(struct list_head *table, int rows);
1285 struct ip_vs_conn *ip_vs_new_conn_out(struct ip_vs_service *svc,
1286 struct ip_vs_dest *dest,
1287 struct sk_buff *skb,
1288 const struct ip_vs_iphdr *iph,
1291 #define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t)))
1293 #define IP_VS_APP_TYPE_FTP 1
1295 /* ip_vs_conn handling functions
1296 * (from ip_vs_conn.c)
1299 IP_VS_DIR_INPUT = 0,
1301 IP_VS_DIR_INPUT_ONLY,
1305 static inline void ip_vs_conn_fill_param(struct netns_ipvs *ipvs, int af, int protocol,
1306 const union nf_inet_addr *caddr,
1308 const union nf_inet_addr *vaddr,
1310 struct ip_vs_conn_param *p)
1314 p->protocol = protocol;
1323 struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p);
1324 struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p);
1326 struct ip_vs_conn * ip_vs_conn_in_get_proto(struct netns_ipvs *ipvs, int af,
1327 const struct sk_buff *skb,
1328 const struct ip_vs_iphdr *iph);
1330 struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p);
1332 struct ip_vs_conn * ip_vs_conn_out_get_proto(struct netns_ipvs *ipvs, int af,
1333 const struct sk_buff *skb,
1334 const struct ip_vs_iphdr *iph);
1336 /* Get reference to gain full access to conn.
1337 * By default, RCU read-side critical sections have access only to
1338 * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference.
1340 static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp)
1342 return refcount_inc_not_zero(&cp->refcnt);
1345 /* put back the conn without restarting its timer */
1346 static inline void __ip_vs_conn_put(struct ip_vs_conn *cp)
1348 smp_mb__before_atomic();
1349 refcount_dec(&cp->refcnt);
1351 void ip_vs_conn_put(struct ip_vs_conn *cp);
1352 void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport);
1354 struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, int dest_af,
1355 const union nf_inet_addr *daddr,
1356 __be16 dport, unsigned int flags,
1357 struct ip_vs_dest *dest, __u32 fwmark);
1358 void ip_vs_conn_expire_now(struct ip_vs_conn *cp);
1360 const char *ip_vs_state_name(const struct ip_vs_conn *cp);
1362 void ip_vs_tcp_conn_listen(struct ip_vs_conn *cp);
1363 int ip_vs_check_template(struct ip_vs_conn *ct, struct ip_vs_dest *cdest);
1364 void ip_vs_random_dropentry(struct netns_ipvs *ipvs);
1365 int ip_vs_conn_init(void);
1366 void ip_vs_conn_cleanup(void);
1368 static inline void ip_vs_control_del(struct ip_vs_conn *cp)
1370 struct ip_vs_conn *ctl_cp = cp->control;
1372 IP_VS_ERR_BUF("request control DEL for uncontrolled: "
1374 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1376 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1382 IP_VS_DBG_BUF(7, "DELeting control for: "
1383 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1384 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1386 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1387 ntohs(ctl_cp->cport));
1390 if (atomic_read(&ctl_cp->n_control) == 0) {
1391 IP_VS_ERR_BUF("BUG control DEL with n=0 : "
1393 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1395 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1400 atomic_dec(&ctl_cp->n_control);
1404 ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp)
1407 IP_VS_ERR_BUF("request control ADD for already controlled: "
1409 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1411 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1414 ip_vs_control_del(cp);
1417 IP_VS_DBG_BUF(7, "ADDing control for: "
1418 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1419 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1421 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1422 ntohs(ctl_cp->cport));
1424 cp->control = ctl_cp;
1425 atomic_inc(&ctl_cp->n_control);
1428 /* Mark our template as assured */
1430 ip_vs_control_assure_ct(struct ip_vs_conn *cp)
1432 struct ip_vs_conn *ct = cp->control;
1434 if (ct && !(ct->state & IP_VS_CTPL_S_ASSURED) &&
1435 (ct->flags & IP_VS_CONN_F_TEMPLATE))
1436 ct->state |= IP_VS_CTPL_S_ASSURED;
1439 /* IPVS netns init & cleanup functions */
1440 int ip_vs_estimator_net_init(struct netns_ipvs *ipvs);
1441 int ip_vs_control_net_init(struct netns_ipvs *ipvs);
1442 int ip_vs_protocol_net_init(struct netns_ipvs *ipvs);
1443 int ip_vs_app_net_init(struct netns_ipvs *ipvs);
1444 int ip_vs_conn_net_init(struct netns_ipvs *ipvs);
1445 int ip_vs_sync_net_init(struct netns_ipvs *ipvs);
1446 void ip_vs_conn_net_cleanup(struct netns_ipvs *ipvs);
1447 void ip_vs_app_net_cleanup(struct netns_ipvs *ipvs);
1448 void ip_vs_protocol_net_cleanup(struct netns_ipvs *ipvs);
1449 void ip_vs_control_net_cleanup(struct netns_ipvs *ipvs);
1450 void ip_vs_estimator_net_cleanup(struct netns_ipvs *ipvs);
1451 void ip_vs_sync_net_cleanup(struct netns_ipvs *ipvs);
1452 void ip_vs_service_nets_cleanup(struct list_head *net_list);
1454 /* IPVS application functions
1455 * (from ip_vs_app.c)
1457 #define IP_VS_APP_MAX_PORTS 8
1458 struct ip_vs_app *register_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1459 void unregister_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1460 int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1461 void ip_vs_unbind_app(struct ip_vs_conn *cp);
1462 int register_ip_vs_app_inc(struct netns_ipvs *ipvs, struct ip_vs_app *app, __u16 proto,
1464 int ip_vs_app_inc_get(struct ip_vs_app *inc);
1465 void ip_vs_app_inc_put(struct ip_vs_app *inc);
1467 int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb,
1468 struct ip_vs_iphdr *ipvsh);
1469 int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb,
1470 struct ip_vs_iphdr *ipvsh);
1472 int register_ip_vs_pe(struct ip_vs_pe *pe);
1473 int unregister_ip_vs_pe(struct ip_vs_pe *pe);
1474 struct ip_vs_pe *ip_vs_pe_getbyname(const char *name);
1475 struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name);
1477 /* Use a #define to avoid all of module.h just for these trivial ops */
1478 #define ip_vs_pe_get(pe) \
1479 if (pe && pe->module) \
1480 __module_get(pe->module);
1482 #define ip_vs_pe_put(pe) \
1483 if (pe && pe->module) \
1484 module_put(pe->module);
1486 /* IPVS protocol functions (from ip_vs_proto.c) */
1487 int ip_vs_protocol_init(void);
1488 void ip_vs_protocol_cleanup(void);
1489 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags);
1490 int *ip_vs_create_timeout_table(int *table, int size);
1491 void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
1492 const struct sk_buff *skb, int offset,
1495 extern struct ip_vs_protocol ip_vs_protocol_tcp;
1496 extern struct ip_vs_protocol ip_vs_protocol_udp;
1497 extern struct ip_vs_protocol ip_vs_protocol_icmp;
1498 extern struct ip_vs_protocol ip_vs_protocol_esp;
1499 extern struct ip_vs_protocol ip_vs_protocol_ah;
1500 extern struct ip_vs_protocol ip_vs_protocol_sctp;
1502 /* Registering/unregistering scheduler functions
1503 * (from ip_vs_sched.c)
1505 int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1506 int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1507 int ip_vs_bind_scheduler(struct ip_vs_service *svc,
1508 struct ip_vs_scheduler *scheduler);
1509 void ip_vs_unbind_scheduler(struct ip_vs_service *svc,
1510 struct ip_vs_scheduler *sched);
1511 struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name);
1512 void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler);
1514 ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb,
1515 struct ip_vs_proto_data *pd, int *ignored,
1516 struct ip_vs_iphdr *iph);
1517 int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb,
1518 struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph);
1520 void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg);
1522 /* IPVS control data and functions (from ip_vs_ctl.c) */
1523 extern struct ip_vs_stats ip_vs_stats;
1524 extern int sysctl_ip_vs_sync_ver;
1526 struct ip_vs_service *
1527 ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
1528 const union nf_inet_addr *vaddr, __be16 vport);
1530 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1531 const union nf_inet_addr *daddr, __be16 dport);
1534 ip_vs_find_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1535 const union nf_inet_addr *daddr, __be16 dport);
1536 struct ip_vs_dest *ip_vs_find_tunnel(struct netns_ipvs *ipvs, int af,
1537 const union nf_inet_addr *daddr,
1540 int ip_vs_use_count_inc(void);
1541 void ip_vs_use_count_dec(void);
1542 int ip_vs_register_nl_ioctl(void);
1543 void ip_vs_unregister_nl_ioctl(void);
1544 int ip_vs_control_init(void);
1545 void ip_vs_control_cleanup(void);
1547 ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
1548 const union nf_inet_addr *daddr, __be16 dport,
1549 const union nf_inet_addr *vaddr, __be16 vport,
1550 __u16 protocol, __u32 fwmark, __u32 flags);
1551 void ip_vs_try_bind_dest(struct ip_vs_conn *cp);
1553 static inline void ip_vs_dest_hold(struct ip_vs_dest *dest)
1555 refcount_inc(&dest->refcnt);
1558 static inline void ip_vs_dest_put(struct ip_vs_dest *dest)
1560 smp_mb__before_atomic();
1561 refcount_dec(&dest->refcnt);
1564 static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest)
1566 if (refcount_dec_and_test(&dest->refcnt))
1570 /* IPVS sync daemon data and function prototypes
1571 * (from ip_vs_sync.c)
1573 int start_sync_thread(struct netns_ipvs *ipvs, struct ipvs_sync_daemon_cfg *cfg,
1575 int stop_sync_thread(struct netns_ipvs *ipvs, int state);
1576 void ip_vs_sync_conn(struct netns_ipvs *ipvs, struct ip_vs_conn *cp, int pkts);
1578 /* IPVS rate estimator prototypes (from ip_vs_est.c) */
1579 int ip_vs_start_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1580 void ip_vs_stop_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1581 void ip_vs_zero_estimator(struct ip_vs_stats *stats);
1582 void ip_vs_read_estimator(struct ip_vs_kstats *dst, struct ip_vs_stats *stats);
1583 void ip_vs_est_reload_start(struct netns_ipvs *ipvs);
1584 int ip_vs_est_kthread_start(struct netns_ipvs *ipvs,
1585 struct ip_vs_est_kt_data *kd);
1586 void ip_vs_est_kthread_stop(struct ip_vs_est_kt_data *kd);
1588 static inline void ip_vs_est_stopped_recalc(struct netns_ipvs *ipvs)
1590 #ifdef CONFIG_SYSCTL
1591 /* Stop tasks while cpulist is empty or if disabled with flag */
1592 ipvs->est_stopped = !sysctl_run_estimation(ipvs) ||
1593 (ipvs->est_cpulist_valid &&
1594 cpumask_empty(sysctl_est_cpulist(ipvs)));
1598 static inline bool ip_vs_est_stopped(struct netns_ipvs *ipvs)
1600 #ifdef CONFIG_SYSCTL
1601 return ipvs->est_stopped;
1607 static inline int ip_vs_est_max_threads(struct netns_ipvs *ipvs)
1609 unsigned int limit = IPVS_EST_CPU_KTHREADS *
1610 cpumask_weight(sysctl_est_cpulist(ipvs));
1612 return max(1U, limit);
1615 /* Various IPVS packet transmitters (from ip_vs_xmit.c) */
1616 int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1617 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1618 int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1619 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1620 int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1621 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1622 int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1623 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1624 int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1625 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1626 int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1627 struct ip_vs_protocol *pp, int offset,
1628 unsigned int hooknum, struct ip_vs_iphdr *iph);
1629 void ip_vs_dest_dst_rcu_free(struct rcu_head *head);
1631 #ifdef CONFIG_IP_VS_IPV6
1632 int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1633 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1634 int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1635 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1636 int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1637 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1638 int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1639 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1640 int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1641 struct ip_vs_protocol *pp, int offset,
1642 unsigned int hooknum, struct ip_vs_iphdr *iph);
1645 #ifdef CONFIG_SYSCTL
1646 /* This is a simple mechanism to ignore packets when
1647 * we are loaded. Just set ip_vs_drop_rate to 'n' and
1648 * we start to drop 1/rate of the packets
1650 static inline int ip_vs_todrop(struct netns_ipvs *ipvs)
1652 if (!ipvs->drop_rate)
1654 if (--ipvs->drop_counter > 0)
1656 ipvs->drop_counter = ipvs->drop_rate;
1660 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; }
1663 #ifdef CONFIG_SYSCTL
1664 /* Enqueue delayed work for expiring no dest connections
1665 * Only run when sysctl_expire_nodest=1
1667 static inline void ip_vs_enqueue_expire_nodest_conns(struct netns_ipvs *ipvs)
1669 if (sysctl_expire_nodest_conn(ipvs))
1670 queue_delayed_work(system_long_wq,
1671 &ipvs->expire_nodest_conn_work, 1);
1674 void ip_vs_expire_nodest_conn_flush(struct netns_ipvs *ipvs);
1676 static inline void ip_vs_enqueue_expire_nodest_conns(struct netns_ipvs *ipvs) {}
1679 #define IP_VS_DFWD_METHOD(dest) (atomic_read(&(dest)->conn_flags) & \
1680 IP_VS_CONN_F_FWD_MASK)
1682 /* ip_vs_fwd_tag returns the forwarding tag of the connection */
1683 #define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK)
1685 static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp)
1689 switch (IP_VS_FWD_METHOD(cp)) {
1690 case IP_VS_CONN_F_MASQ:
1692 case IP_VS_CONN_F_LOCALNODE:
1694 case IP_VS_CONN_F_TUNNEL:
1696 case IP_VS_CONN_F_DROUTE:
1698 case IP_VS_CONN_F_BYPASS:
1706 void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp,
1707 struct ip_vs_conn *cp, int dir);
1709 #ifdef CONFIG_IP_VS_IPV6
1710 void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp,
1711 struct ip_vs_conn *cp, int dir);
1714 __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset);
1716 static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum)
1718 __be32 diff[2] = { ~old, new };
1720 return csum_partial(diff, sizeof(diff), oldsum);
1723 #ifdef CONFIG_IP_VS_IPV6
1724 static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new,
1727 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0],
1728 new[3], new[2], new[1], new[0] };
1730 return csum_partial(diff, sizeof(diff), oldsum);
1734 static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum)
1736 __be16 diff[2] = { ~old, new };
1738 return csum_partial(diff, sizeof(diff), oldsum);
1741 /* Forget current conntrack (unconfirmed) and attach notrack entry */
1742 static inline void ip_vs_notrack(struct sk_buff *skb)
1744 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1745 enum ip_conntrack_info ctinfo;
1746 struct nf_conn *ct = nf_ct_get(skb, &ctinfo);
1749 nf_conntrack_put(&ct->ct_general);
1750 nf_ct_set(skb, NULL, IP_CT_UNTRACKED);
1755 #ifdef CONFIG_IP_VS_NFCT
1756 /* Netfilter connection tracking
1757 * (from ip_vs_nfct.c)
1759 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1761 #ifdef CONFIG_SYSCTL
1762 return ipvs->sysctl_conntrack;
1768 void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp,
1770 int ip_vs_confirm_conntrack(struct sk_buff *skb);
1771 void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct,
1772 struct ip_vs_conn *cp, u_int8_t proto,
1773 const __be16 port, int from_rs);
1774 void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp);
1778 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1783 static inline void ip_vs_update_conntrack(struct sk_buff *skb,
1784 struct ip_vs_conn *cp, int outin)
1788 static inline int ip_vs_confirm_conntrack(struct sk_buff *skb)
1793 static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp)
1796 #endif /* CONFIG_IP_VS_NFCT */
1798 /* Using old conntrack that can not be redirected to another real server? */
1799 static inline bool ip_vs_conn_uses_old_conntrack(struct ip_vs_conn *cp,
1800 struct sk_buff *skb)
1802 #ifdef CONFIG_IP_VS_NFCT
1803 enum ip_conntrack_info ctinfo;
1806 ct = nf_ct_get(skb, &ctinfo);
1807 if (ct && nf_ct_is_confirmed(ct))
1813 static inline int ip_vs_register_conntrack(struct ip_vs_service *svc)
1815 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
1816 int afmask = (svc->af == AF_INET6) ? 2 : 1;
1819 if (!(svc->conntrack_afmask & afmask)) {
1820 ret = nf_ct_netns_get(svc->ipvs->net, svc->af);
1822 svc->conntrack_afmask |= afmask;
1830 static inline void ip_vs_unregister_conntrack(struct ip_vs_service *svc)
1832 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
1833 int afmask = (svc->af == AF_INET6) ? 2 : 1;
1835 if (svc->conntrack_afmask & afmask) {
1836 nf_ct_netns_put(svc->ipvs->net, svc->af);
1837 svc->conntrack_afmask &= ~afmask;
1842 int ip_vs_register_hooks(struct netns_ipvs *ipvs, unsigned int af);
1843 void ip_vs_unregister_hooks(struct netns_ipvs *ipvs, unsigned int af);
1846 ip_vs_dest_conn_overhead(struct ip_vs_dest *dest)
1848 /* We think the overhead of processing active connections is 256
1849 * times higher than that of inactive connections in average. (This
1850 * 256 times might not be accurate, we will change it later) We
1851 * use the following formula to estimate the overhead now:
1852 * dest->activeconns*256 + dest->inactconns
1854 return (atomic_read(&dest->activeconns) << 8) +
1855 atomic_read(&dest->inactconns);
1858 #ifdef CONFIG_IP_VS_PROTO_TCP
1859 INDIRECT_CALLABLE_DECLARE(int
1860 tcp_snat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
1861 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph));
1864 #ifdef CONFIG_IP_VS_PROTO_UDP
1865 INDIRECT_CALLABLE_DECLARE(int
1866 udp_snat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
1867 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph));
1869 #endif /* _NET_IP_VS_H */