xref: /linux/include/net/ip_vs.h (revision c75c5ab575af7db707689cdbb5a5c458e9a034bb)
1 /*
2  *      IP Virtual Server
3  *      data structure and functionality definitions
4  */
5 
6 #ifndef _NET_IP_VS_H
7 #define _NET_IP_VS_H
8 
9 #include <linux/ip_vs.h>                /* definitions shared with userland */
10 
11 #include <asm/types.h>                  /* for __uXX types */
12 
13 #include <linux/list.h>                 /* for struct list_head */
14 #include <linux/spinlock.h>             /* for struct rwlock_t */
15 #include <linux/atomic.h>                 /* for struct atomic_t */
16 #include <linux/compiler.h>
17 #include <linux/timer.h>
18 #include <linux/bug.h>
19 
20 #include <net/checksum.h>
21 #include <linux/netfilter.h>		/* for union nf_inet_addr */
22 #include <linux/ip.h>
23 #include <linux/ipv6.h>			/* for struct ipv6hdr */
24 #include <net/ipv6.h>
25 #if IS_ENABLED(CONFIG_IP_VS_IPV6)
26 #include <linux/netfilter_ipv6/ip6_tables.h>
27 #endif
28 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
29 #include <net/netfilter/nf_conntrack.h>
30 #endif
31 #include <net/net_namespace.h>		/* Netw namespace */
32 
33 /*
34  * Generic access of ipvs struct
35  */
36 static inline struct netns_ipvs *net_ipvs(struct net* net)
37 {
38 	return net->ipvs;
39 }
40 /*
41  * Get net ptr from skb in traffic cases
42  * use skb_sknet when call is from userland (ioctl or netlink)
43  */
44 static inline struct net *skb_net(const struct sk_buff *skb)
45 {
46 #ifdef CONFIG_NET_NS
47 #ifdef CONFIG_IP_VS_DEBUG
48 	/*
49 	 * This is used for debug only.
50 	 * Start with the most likely hit
51 	 * End with BUG
52 	 */
53 	if (likely(skb->dev && skb->dev->nd_net))
54 		return dev_net(skb->dev);
55 	if (skb_dst(skb) && skb_dst(skb)->dev)
56 		return dev_net(skb_dst(skb)->dev);
57 	WARN(skb->sk, "Maybe skb_sknet should be used in %s() at line:%d\n",
58 		      __func__, __LINE__);
59 	if (likely(skb->sk && skb->sk->sk_net))
60 		return sock_net(skb->sk);
61 	pr_err("There is no net ptr to find in the skb in %s() line:%d\n",
62 		__func__, __LINE__);
63 	BUG();
64 #else
65 	return dev_net(skb->dev ? : skb_dst(skb)->dev);
66 #endif
67 #else
68 	return &init_net;
69 #endif
70 }
71 
72 static inline struct net *skb_sknet(const struct sk_buff *skb)
73 {
74 #ifdef CONFIG_NET_NS
75 #ifdef CONFIG_IP_VS_DEBUG
76 	/* Start with the most likely hit */
77 	if (likely(skb->sk && skb->sk->sk_net))
78 		return sock_net(skb->sk);
79 	WARN(skb->dev, "Maybe skb_net should be used instead in %s() line:%d\n",
80 		       __func__, __LINE__);
81 	if (likely(skb->dev && skb->dev->nd_net))
82 		return dev_net(skb->dev);
83 	pr_err("There is no net ptr to find in the skb in %s() line:%d\n",
84 		__func__, __LINE__);
85 	BUG();
86 #else
87 	return sock_net(skb->sk);
88 #endif
89 #else
90 	return &init_net;
91 #endif
92 }
93 /*
94  * This one needed for single_open_net since net is stored directly in
95  * private not as a struct i.e. seq_file_net can't be used.
96  */
97 static inline struct net *seq_file_single_net(struct seq_file *seq)
98 {
99 #ifdef CONFIG_NET_NS
100 	return (struct net *)seq->private;
101 #else
102 	return &init_net;
103 #endif
104 }
105 
106 /* Connections' size value needed by ip_vs_ctl.c */
107 extern int ip_vs_conn_tab_size;
108 
109 struct ip_vs_iphdr {
110 	__u32 len;	/* IPv4 simply where L4 starts
111 			   IPv6 where L4 Transport Header starts */
112 	__u32 thoff_reasm; /* Transport Header Offset in nfct_reasm skb */
113 	__u16 fragoffs; /* IPv6 fragment offset, 0 if first frag (or not frag)*/
114 	__s16 protocol;
115 	__s32 flags;
116 	union nf_inet_addr saddr;
117 	union nf_inet_addr daddr;
118 };
119 
120 /* Dependency to module: nf_defrag_ipv6 */
121 #if defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE)
122 static inline struct sk_buff *skb_nfct_reasm(const struct sk_buff *skb)
123 {
124 	return skb->nfct_reasm;
125 }
126 static inline void *frag_safe_skb_hp(const struct sk_buff *skb, int offset,
127 				      int len, void *buffer,
128 				      const struct ip_vs_iphdr *ipvsh)
129 {
130 	if (unlikely(ipvsh->fragoffs && skb_nfct_reasm(skb)))
131 		return skb_header_pointer(skb_nfct_reasm(skb),
132 					  ipvsh->thoff_reasm, len, buffer);
133 
134 	return skb_header_pointer(skb, offset, len, buffer);
135 }
136 #else
137 static inline struct sk_buff *skb_nfct_reasm(const struct sk_buff *skb)
138 {
139 	return NULL;
140 }
141 static inline void *frag_safe_skb_hp(const struct sk_buff *skb, int offset,
142 				      int len, void *buffer,
143 				      const struct ip_vs_iphdr *ipvsh)
144 {
145 	return skb_header_pointer(skb, offset, len, buffer);
146 }
147 #endif
148 
149 static inline void
150 ip_vs_fill_ip4hdr(const void *nh, struct ip_vs_iphdr *iphdr)
151 {
152 	const struct iphdr *iph = nh;
153 
154 	iphdr->len	= iph->ihl * 4;
155 	iphdr->fragoffs	= 0;
156 	iphdr->protocol	= iph->protocol;
157 	iphdr->saddr.ip	= iph->saddr;
158 	iphdr->daddr.ip	= iph->daddr;
159 }
160 
161 /* This function handles filling *ip_vs_iphdr, both for IPv4 and IPv6.
162  * IPv6 requires some extra work, as finding proper header position,
163  * depend on the IPv6 extension headers.
164  */
165 static inline void
166 ip_vs_fill_iph_skb(int af, const struct sk_buff *skb, struct ip_vs_iphdr *iphdr)
167 {
168 #ifdef CONFIG_IP_VS_IPV6
169 	if (af == AF_INET6) {
170 		const struct ipv6hdr *iph =
171 			(struct ipv6hdr *)skb_network_header(skb);
172 		iphdr->saddr.in6 = iph->saddr;
173 		iphdr->daddr.in6 = iph->daddr;
174 		/* ipv6_find_hdr() updates len, flags, thoff_reasm */
175 		iphdr->thoff_reasm = 0;
176 		iphdr->len	 = 0;
177 		iphdr->flags	 = 0;
178 		iphdr->protocol  = ipv6_find_hdr(skb, &iphdr->len, -1,
179 						 &iphdr->fragoffs,
180 						 &iphdr->flags);
181 		/* get proto from re-assembled packet and it's offset */
182 		if (skb_nfct_reasm(skb))
183 			iphdr->protocol = ipv6_find_hdr(skb_nfct_reasm(skb),
184 							&iphdr->thoff_reasm,
185 							-1, NULL, NULL);
186 
187 	} else
188 #endif
189 	{
190 		const struct iphdr *iph =
191 			(struct iphdr *)skb_network_header(skb);
192 		iphdr->len	= iph->ihl * 4;
193 		iphdr->fragoffs	= 0;
194 		iphdr->protocol	= iph->protocol;
195 		iphdr->saddr.ip	= iph->saddr;
196 		iphdr->daddr.ip	= iph->daddr;
197 	}
198 }
199 
200 /* This function is a faster version of ip_vs_fill_iph_skb().
201  * Where we only populate {s,d}addr (and avoid calling ipv6_find_hdr()).
202  * This is used by the some of the ip_vs_*_schedule() functions.
203  * (Mostly done to avoid ABI breakage of external schedulers)
204  */
205 static inline void
206 ip_vs_fill_iph_addr_only(int af, const struct sk_buff *skb,
207 			 struct ip_vs_iphdr *iphdr)
208 {
209 #ifdef CONFIG_IP_VS_IPV6
210 	if (af == AF_INET6) {
211 		const struct ipv6hdr *iph =
212 			(struct ipv6hdr *)skb_network_header(skb);
213 		iphdr->saddr.in6 = iph->saddr;
214 		iphdr->daddr.in6 = iph->daddr;
215 	} else
216 #endif
217 	{
218 		const struct iphdr *iph =
219 			(struct iphdr *)skb_network_header(skb);
220 		iphdr->saddr.ip = iph->saddr;
221 		iphdr->daddr.ip = iph->daddr;
222 	}
223 }
224 
225 static inline void ip_vs_addr_copy(int af, union nf_inet_addr *dst,
226 				   const union nf_inet_addr *src)
227 {
228 #ifdef CONFIG_IP_VS_IPV6
229 	if (af == AF_INET6)
230 		dst->in6 = src->in6;
231 	else
232 #endif
233 	dst->ip = src->ip;
234 }
235 
236 static inline int ip_vs_addr_equal(int af, const union nf_inet_addr *a,
237 				   const union nf_inet_addr *b)
238 {
239 #ifdef CONFIG_IP_VS_IPV6
240 	if (af == AF_INET6)
241 		return ipv6_addr_equal(&a->in6, &b->in6);
242 #endif
243 	return a->ip == b->ip;
244 }
245 
246 #ifdef CONFIG_IP_VS_DEBUG
247 #include <linux/net.h>
248 
249 extern int ip_vs_get_debug_level(void);
250 
251 static inline const char *ip_vs_dbg_addr(int af, char *buf, size_t buf_len,
252 					 const union nf_inet_addr *addr,
253 					 int *idx)
254 {
255 	int len;
256 #ifdef CONFIG_IP_VS_IPV6
257 	if (af == AF_INET6)
258 		len = snprintf(&buf[*idx], buf_len - *idx, "[%pI6c]",
259 			       &addr->in6) + 1;
260 	else
261 #endif
262 		len = snprintf(&buf[*idx], buf_len - *idx, "%pI4",
263 			       &addr->ip) + 1;
264 
265 	*idx += len;
266 	BUG_ON(*idx > buf_len + 1);
267 	return &buf[*idx - len];
268 }
269 
270 #define IP_VS_DBG_BUF(level, msg, ...)					\
271 	do {								\
272 		char ip_vs_dbg_buf[160];				\
273 		int ip_vs_dbg_idx = 0;					\
274 		if (level <= ip_vs_get_debug_level())			\
275 			printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__);	\
276 	} while (0)
277 #define IP_VS_ERR_BUF(msg...)						\
278 	do {								\
279 		char ip_vs_dbg_buf[160];				\
280 		int ip_vs_dbg_idx = 0;					\
281 		pr_err(msg);						\
282 	} while (0)
283 
284 /* Only use from within IP_VS_DBG_BUF() or IP_VS_ERR_BUF macros */
285 #define IP_VS_DBG_ADDR(af, addr)					\
286 	ip_vs_dbg_addr(af, ip_vs_dbg_buf,				\
287 		       sizeof(ip_vs_dbg_buf), addr,			\
288 		       &ip_vs_dbg_idx)
289 
290 #define IP_VS_DBG(level, msg, ...)					\
291 	do {								\
292 		if (level <= ip_vs_get_debug_level())			\
293 			printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__);	\
294 	} while (0)
295 #define IP_VS_DBG_RL(msg, ...)						\
296 	do {								\
297 		if (net_ratelimit())					\
298 			printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__);	\
299 	} while (0)
300 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg)			\
301 	do {								\
302 		if (level <= ip_vs_get_debug_level())			\
303 			pp->debug_packet(af, pp, skb, ofs, msg);	\
304 	} while (0)
305 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg)			\
306 	do {								\
307 		if (level <= ip_vs_get_debug_level() &&			\
308 		    net_ratelimit())					\
309 			pp->debug_packet(af, pp, skb, ofs, msg);	\
310 	} while (0)
311 #else	/* NO DEBUGGING at ALL */
312 #define IP_VS_DBG_BUF(level, msg...)  do {} while (0)
313 #define IP_VS_ERR_BUF(msg...)  do {} while (0)
314 #define IP_VS_DBG(level, msg...)  do {} while (0)
315 #define IP_VS_DBG_RL(msg...)  do {} while (0)
316 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg)	do {} while (0)
317 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg)	do {} while (0)
318 #endif
319 
320 #define IP_VS_BUG() BUG()
321 #define IP_VS_ERR_RL(msg, ...)						\
322 	do {								\
323 		if (net_ratelimit())					\
324 			pr_err(msg, ##__VA_ARGS__);			\
325 	} while (0)
326 
327 #ifdef CONFIG_IP_VS_DEBUG
328 #define EnterFunction(level)						\
329 	do {								\
330 		if (level <= ip_vs_get_debug_level())			\
331 			printk(KERN_DEBUG				\
332 			       pr_fmt("Enter: %s, %s line %i\n"),	\
333 			       __func__, __FILE__, __LINE__);		\
334 	} while (0)
335 #define LeaveFunction(level)						\
336 	do {								\
337 		if (level <= ip_vs_get_debug_level())			\
338 			printk(KERN_DEBUG				\
339 			       pr_fmt("Leave: %s, %s line %i\n"),	\
340 			       __func__, __FILE__, __LINE__);		\
341 	} while (0)
342 #else
343 #define EnterFunction(level)   do {} while (0)
344 #define LeaveFunction(level)   do {} while (0)
345 #endif
346 
347 #define	IP_VS_WAIT_WHILE(expr)	while (expr) { cpu_relax(); }
348 
349 
350 /*
351  *      The port number of FTP service (in network order).
352  */
353 #define FTPPORT  cpu_to_be16(21)
354 #define FTPDATA  cpu_to_be16(20)
355 
356 /*
357  *      TCP State Values
358  */
359 enum {
360 	IP_VS_TCP_S_NONE = 0,
361 	IP_VS_TCP_S_ESTABLISHED,
362 	IP_VS_TCP_S_SYN_SENT,
363 	IP_VS_TCP_S_SYN_RECV,
364 	IP_VS_TCP_S_FIN_WAIT,
365 	IP_VS_TCP_S_TIME_WAIT,
366 	IP_VS_TCP_S_CLOSE,
367 	IP_VS_TCP_S_CLOSE_WAIT,
368 	IP_VS_TCP_S_LAST_ACK,
369 	IP_VS_TCP_S_LISTEN,
370 	IP_VS_TCP_S_SYNACK,
371 	IP_VS_TCP_S_LAST
372 };
373 
374 /*
375  *	UDP State Values
376  */
377 enum {
378 	IP_VS_UDP_S_NORMAL,
379 	IP_VS_UDP_S_LAST,
380 };
381 
382 /*
383  *	ICMP State Values
384  */
385 enum {
386 	IP_VS_ICMP_S_NORMAL,
387 	IP_VS_ICMP_S_LAST,
388 };
389 
390 /*
391  *	SCTP State Values
392  */
393 enum ip_vs_sctp_states {
394 	IP_VS_SCTP_S_NONE,
395 	IP_VS_SCTP_S_INIT_CLI,
396 	IP_VS_SCTP_S_INIT_SER,
397 	IP_VS_SCTP_S_INIT_ACK_CLI,
398 	IP_VS_SCTP_S_INIT_ACK_SER,
399 	IP_VS_SCTP_S_ECHO_CLI,
400 	IP_VS_SCTP_S_ECHO_SER,
401 	IP_VS_SCTP_S_ESTABLISHED,
402 	IP_VS_SCTP_S_SHUT_CLI,
403 	IP_VS_SCTP_S_SHUT_SER,
404 	IP_VS_SCTP_S_SHUT_ACK_CLI,
405 	IP_VS_SCTP_S_SHUT_ACK_SER,
406 	IP_VS_SCTP_S_CLOSED,
407 	IP_VS_SCTP_S_LAST
408 };
409 
410 /*
411  *	Delta sequence info structure
412  *	Each ip_vs_conn has 2 (output AND input seq. changes).
413  *      Only used in the VS/NAT.
414  */
415 struct ip_vs_seq {
416 	__u32			init_seq;	/* Add delta from this seq */
417 	__u32			delta;		/* Delta in sequence numbers */
418 	__u32			previous_delta;	/* Delta in sequence numbers
419 						   before last resized pkt */
420 };
421 
422 /*
423  * counters per cpu
424  */
425 struct ip_vs_counters {
426 	__u32		conns;		/* connections scheduled */
427 	__u32		inpkts;		/* incoming packets */
428 	__u32		outpkts;	/* outgoing packets */
429 	__u64		inbytes;	/* incoming bytes */
430 	__u64		outbytes;	/* outgoing bytes */
431 };
432 /*
433  * Stats per cpu
434  */
435 struct ip_vs_cpu_stats {
436 	struct ip_vs_counters   ustats;
437 	struct u64_stats_sync   syncp;
438 };
439 
440 /*
441  *	IPVS statistics objects
442  */
443 struct ip_vs_estimator {
444 	struct list_head	list;
445 
446 	u64			last_inbytes;
447 	u64			last_outbytes;
448 	u32			last_conns;
449 	u32			last_inpkts;
450 	u32			last_outpkts;
451 
452 	u32			cps;
453 	u32			inpps;
454 	u32			outpps;
455 	u32			inbps;
456 	u32			outbps;
457 };
458 
459 struct ip_vs_stats {
460 	struct ip_vs_stats_user	ustats;		/* statistics */
461 	struct ip_vs_estimator	est;		/* estimator */
462 	struct ip_vs_cpu_stats	*cpustats;	/* per cpu counters */
463 	spinlock_t		lock;		/* spin lock */
464 	struct ip_vs_stats_user	ustats0;	/* reset values */
465 };
466 
467 struct dst_entry;
468 struct iphdr;
469 struct ip_vs_conn;
470 struct ip_vs_app;
471 struct sk_buff;
472 struct ip_vs_proto_data;
473 
474 struct ip_vs_protocol {
475 	struct ip_vs_protocol	*next;
476 	char			*name;
477 	u16			protocol;
478 	u16			num_states;
479 	int			dont_defrag;
480 
481 	void (*init)(struct ip_vs_protocol *pp);
482 
483 	void (*exit)(struct ip_vs_protocol *pp);
484 
485 	int (*init_netns)(struct net *net, struct ip_vs_proto_data *pd);
486 
487 	void (*exit_netns)(struct net *net, struct ip_vs_proto_data *pd);
488 
489 	int (*conn_schedule)(int af, struct sk_buff *skb,
490 			     struct ip_vs_proto_data *pd,
491 			     int *verdict, struct ip_vs_conn **cpp,
492 			     struct ip_vs_iphdr *iph);
493 
494 	struct ip_vs_conn *
495 	(*conn_in_get)(int af,
496 		       const struct sk_buff *skb,
497 		       const struct ip_vs_iphdr *iph,
498 		       int inverse);
499 
500 	struct ip_vs_conn *
501 	(*conn_out_get)(int af,
502 			const struct sk_buff *skb,
503 			const struct ip_vs_iphdr *iph,
504 			int inverse);
505 
506 	int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
507 			    struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
508 
509 	int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
510 			    struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
511 
512 	int (*csum_check)(int af, struct sk_buff *skb,
513 			  struct ip_vs_protocol *pp);
514 
515 	const char *(*state_name)(int state);
516 
517 	void (*state_transition)(struct ip_vs_conn *cp, int direction,
518 				 const struct sk_buff *skb,
519 				 struct ip_vs_proto_data *pd);
520 
521 	int (*register_app)(struct net *net, struct ip_vs_app *inc);
522 
523 	void (*unregister_app)(struct net *net, struct ip_vs_app *inc);
524 
525 	int (*app_conn_bind)(struct ip_vs_conn *cp);
526 
527 	void (*debug_packet)(int af, struct ip_vs_protocol *pp,
528 			     const struct sk_buff *skb,
529 			     int offset,
530 			     const char *msg);
531 
532 	void (*timeout_change)(struct ip_vs_proto_data *pd, int flags);
533 };
534 
535 /*
536  * protocol data per netns
537  */
538 struct ip_vs_proto_data {
539 	struct ip_vs_proto_data	*next;
540 	struct ip_vs_protocol	*pp;
541 	int			*timeout_table;	/* protocol timeout table */
542 	atomic_t		appcnt;		/* counter of proto app incs. */
543 	struct tcp_states_t	*tcp_state_table;
544 };
545 
546 extern struct ip_vs_protocol   *ip_vs_proto_get(unsigned short proto);
547 extern struct ip_vs_proto_data *ip_vs_proto_data_get(struct net *net,
548 						     unsigned short proto);
549 
550 struct ip_vs_conn_param {
551 	struct net			*net;
552 	const union nf_inet_addr	*caddr;
553 	const union nf_inet_addr	*vaddr;
554 	__be16				cport;
555 	__be16				vport;
556 	__u16				protocol;
557 	u16				af;
558 
559 	const struct ip_vs_pe		*pe;
560 	char				*pe_data;
561 	__u8				pe_data_len;
562 };
563 
564 /*
565  *	IP_VS structure allocated for each dynamically scheduled connection
566  */
567 struct ip_vs_conn {
568 	struct hlist_node	c_list;         /* hashed list heads */
569 #ifdef CONFIG_NET_NS
570 	struct net              *net;           /* Name space */
571 #endif
572 	/* Protocol, addresses and port numbers */
573 	u16                     af;             /* address family */
574 	__be16                  cport;
575 	__be16                  vport;
576 	__be16                  dport;
577 	__u32                   fwmark;         /* Fire wall mark from skb */
578 	union nf_inet_addr      caddr;          /* client address */
579 	union nf_inet_addr      vaddr;          /* virtual address */
580 	union nf_inet_addr      daddr;          /* destination address */
581 	volatile __u32          flags;          /* status flags */
582 	__u16                   protocol;       /* Which protocol (TCP/UDP) */
583 
584 	/* counter and timer */
585 	atomic_t		refcnt;		/* reference count */
586 	struct timer_list	timer;		/* Expiration timer */
587 	volatile unsigned long	timeout;	/* timeout */
588 
589 	/* Flags and state transition */
590 	spinlock_t              lock;           /* lock for state transition */
591 	volatile __u16          state;          /* state info */
592 	volatile __u16          old_state;      /* old state, to be used for
593 						 * state transition triggerd
594 						 * synchronization
595 						 */
596 	unsigned long		sync_endtime;	/* jiffies + sent_retries */
597 
598 	/* Control members */
599 	struct ip_vs_conn       *control;       /* Master control connection */
600 	atomic_t                n_control;      /* Number of controlled ones */
601 	struct ip_vs_dest       *dest;          /* real server */
602 	atomic_t                in_pkts;        /* incoming packet counter */
603 
604 	/* packet transmitter for different forwarding methods.  If it
605 	   mangles the packet, it must return NF_DROP or better NF_STOLEN,
606 	   otherwise this must be changed to a sk_buff **.
607 	   NF_ACCEPT can be returned when destination is local.
608 	 */
609 	int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp,
610 			   struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
611 
612 	/* Note: we can group the following members into a structure,
613 	   in order to save more space, and the following members are
614 	   only used in VS/NAT anyway */
615 	struct ip_vs_app        *app;           /* bound ip_vs_app object */
616 	void                    *app_data;      /* Application private data */
617 	struct ip_vs_seq        in_seq;         /* incoming seq. struct */
618 	struct ip_vs_seq        out_seq;        /* outgoing seq. struct */
619 
620 	const struct ip_vs_pe	*pe;
621 	char			*pe_data;
622 	__u8			pe_data_len;
623 };
624 
625 /*
626  *  To save some memory in conn table when name space is disabled.
627  */
628 static inline struct net *ip_vs_conn_net(const struct ip_vs_conn *cp)
629 {
630 #ifdef CONFIG_NET_NS
631 	return cp->net;
632 #else
633 	return &init_net;
634 #endif
635 }
636 static inline void ip_vs_conn_net_set(struct ip_vs_conn *cp, struct net *net)
637 {
638 #ifdef CONFIG_NET_NS
639 	cp->net = net;
640 #endif
641 }
642 
643 static inline int ip_vs_conn_net_eq(const struct ip_vs_conn *cp,
644 				    struct net *net)
645 {
646 #ifdef CONFIG_NET_NS
647 	return cp->net == net;
648 #else
649 	return 1;
650 #endif
651 }
652 
653 /*
654  *	Extended internal versions of struct ip_vs_service_user and
655  *	ip_vs_dest_user for IPv6 support.
656  *
657  *	We need these to conveniently pass around service and destination
658  *	options, but unfortunately, we also need to keep the old definitions to
659  *	maintain userspace backwards compatibility for the setsockopt interface.
660  */
661 struct ip_vs_service_user_kern {
662 	/* virtual service addresses */
663 	u16			af;
664 	u16			protocol;
665 	union nf_inet_addr	addr;		/* virtual ip address */
666 	u16			port;
667 	u32			fwmark;		/* firwall mark of service */
668 
669 	/* virtual service options */
670 	char			*sched_name;
671 	char			*pe_name;
672 	unsigned int		flags;		/* virtual service flags */
673 	unsigned int		timeout;	/* persistent timeout in sec */
674 	u32			netmask;	/* persistent netmask */
675 };
676 
677 
678 struct ip_vs_dest_user_kern {
679 	/* destination server address */
680 	union nf_inet_addr	addr;
681 	u16			port;
682 
683 	/* real server options */
684 	unsigned int		conn_flags;	/* connection flags */
685 	int			weight;		/* destination weight */
686 
687 	/* thresholds for active connections */
688 	u32			u_threshold;	/* upper threshold */
689 	u32			l_threshold;	/* lower threshold */
690 };
691 
692 
693 /*
694  *	The information about the virtual service offered to the net
695  *	and the forwarding entries
696  */
697 struct ip_vs_service {
698 	struct list_head	s_list;   /* for normal service table */
699 	struct list_head	f_list;   /* for fwmark-based service table */
700 	atomic_t		refcnt;   /* reference counter */
701 	atomic_t		usecnt;   /* use counter */
702 
703 	u16			af;       /* address family */
704 	__u16			protocol; /* which protocol (TCP/UDP) */
705 	union nf_inet_addr	addr;	  /* IP address for virtual service */
706 	__be16			port;	  /* port number for the service */
707 	__u32                   fwmark;   /* firewall mark of the service */
708 	unsigned int		flags;	  /* service status flags */
709 	unsigned int		timeout;  /* persistent timeout in ticks */
710 	__be32			netmask;  /* grouping granularity */
711 	struct net		*net;
712 
713 	struct list_head	destinations;  /* real server d-linked list */
714 	__u32			num_dests;     /* number of servers */
715 	struct ip_vs_stats      stats;         /* statistics for the service */
716 	struct ip_vs_app	*inc;	  /* bind conns to this app inc */
717 
718 	/* for scheduling */
719 	struct ip_vs_scheduler	*scheduler;    /* bound scheduler object */
720 	rwlock_t		sched_lock;    /* lock sched_data */
721 	void			*sched_data;   /* scheduler application data */
722 
723 	/* alternate persistence engine */
724 	struct ip_vs_pe		*pe;
725 };
726 
727 
728 /*
729  *	The real server destination forwarding entry
730  *	with ip address, port number, and so on.
731  */
732 struct ip_vs_dest {
733 	struct list_head	n_list;   /* for the dests in the service */
734 	struct list_head	d_list;   /* for table with all the dests */
735 
736 	u16			af;		/* address family */
737 	__be16			port;		/* port number of the server */
738 	union nf_inet_addr	addr;		/* IP address of the server */
739 	volatile unsigned int	flags;		/* dest status flags */
740 	atomic_t		conn_flags;	/* flags to copy to conn */
741 	atomic_t		weight;		/* server weight */
742 
743 	atomic_t		refcnt;		/* reference counter */
744 	struct ip_vs_stats      stats;          /* statistics */
745 
746 	/* connection counters and thresholds */
747 	atomic_t		activeconns;	/* active connections */
748 	atomic_t		inactconns;	/* inactive connections */
749 	atomic_t		persistconns;	/* persistent connections */
750 	__u32			u_threshold;	/* upper threshold */
751 	__u32			l_threshold;	/* lower threshold */
752 
753 	/* for destination cache */
754 	spinlock_t		dst_lock;	/* lock of dst_cache */
755 	struct dst_entry	*dst_cache;	/* destination cache entry */
756 	u32			dst_rtos;	/* RT_TOS(tos) for dst */
757 	u32			dst_cookie;
758 	union nf_inet_addr	dst_saddr;
759 
760 	/* for virtual service */
761 	struct ip_vs_service	*svc;		/* service it belongs to */
762 	__u16			protocol;	/* which protocol (TCP/UDP) */
763 	__be16			vport;		/* virtual port number */
764 	union nf_inet_addr	vaddr;		/* virtual IP address */
765 	__u32			vfwmark;	/* firewall mark of service */
766 };
767 
768 
769 /*
770  *	The scheduler object
771  */
772 struct ip_vs_scheduler {
773 	struct list_head	n_list;		/* d-linked list head */
774 	char			*name;		/* scheduler name */
775 	atomic_t		refcnt;		/* reference counter */
776 	struct module		*module;	/* THIS_MODULE/NULL */
777 
778 	/* scheduler initializing service */
779 	int (*init_service)(struct ip_vs_service *svc);
780 	/* scheduling service finish */
781 	int (*done_service)(struct ip_vs_service *svc);
782 	/* scheduler updating service */
783 	int (*update_service)(struct ip_vs_service *svc);
784 
785 	/* selecting a server from the given service */
786 	struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc,
787 				       const struct sk_buff *skb);
788 };
789 
790 /* The persistence engine object */
791 struct ip_vs_pe {
792 	struct list_head	n_list;		/* d-linked list head */
793 	char			*name;		/* scheduler name */
794 	atomic_t		refcnt;		/* reference counter */
795 	struct module		*module;	/* THIS_MODULE/NULL */
796 
797 	/* get the connection template, if any */
798 	int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb);
799 	bool (*ct_match)(const struct ip_vs_conn_param *p,
800 			 struct ip_vs_conn *ct);
801 	u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval,
802 			   bool inverse);
803 	int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf);
804 };
805 
806 /*
807  *	The application module object (a.k.a. app incarnation)
808  */
809 struct ip_vs_app {
810 	struct list_head	a_list;		/* member in app list */
811 	int			type;		/* IP_VS_APP_TYPE_xxx */
812 	char			*name;		/* application module name */
813 	__u16			protocol;
814 	struct module		*module;	/* THIS_MODULE/NULL */
815 	struct list_head	incs_list;	/* list of incarnations */
816 
817 	/* members for application incarnations */
818 	struct list_head	p_list;		/* member in proto app list */
819 	struct ip_vs_app	*app;		/* its real application */
820 	__be16			port;		/* port number in net order */
821 	atomic_t		usecnt;		/* usage counter */
822 
823 	/*
824 	 * output hook: Process packet in inout direction, diff set for TCP.
825 	 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
826 	 *	   2=Mangled but checksum was not updated
827 	 */
828 	int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *,
829 		       struct sk_buff *, int *diff);
830 
831 	/*
832 	 * input hook: Process packet in outin direction, diff set for TCP.
833 	 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
834 	 *	   2=Mangled but checksum was not updated
835 	 */
836 	int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *,
837 		      struct sk_buff *, int *diff);
838 
839 	/* ip_vs_app initializer */
840 	int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *);
841 
842 	/* ip_vs_app finish */
843 	int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *);
844 
845 
846 	/* not used now */
847 	int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *,
848 			 struct ip_vs_protocol *);
849 
850 	void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *);
851 
852 	int *			timeout_table;
853 	int *			timeouts;
854 	int			timeouts_size;
855 
856 	int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app,
857 			     int *verdict, struct ip_vs_conn **cpp);
858 
859 	struct ip_vs_conn *
860 	(*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app,
861 		       const struct iphdr *iph, int inverse);
862 
863 	struct ip_vs_conn *
864 	(*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app,
865 			const struct iphdr *iph, int inverse);
866 
867 	int (*state_transition)(struct ip_vs_conn *cp, int direction,
868 				const struct sk_buff *skb,
869 				struct ip_vs_app *app);
870 
871 	void (*timeout_change)(struct ip_vs_app *app, int flags);
872 };
873 
874 struct ipvs_master_sync_state {
875 	struct list_head	sync_queue;
876 	struct ip_vs_sync_buff	*sync_buff;
877 	int			sync_queue_len;
878 	unsigned int		sync_queue_delay;
879 	struct task_struct	*master_thread;
880 	struct delayed_work	master_wakeup_work;
881 	struct netns_ipvs	*ipvs;
882 };
883 
884 /* IPVS in network namespace */
885 struct netns_ipvs {
886 	int			gen;		/* Generation */
887 	int			enable;		/* enable like nf_hooks do */
888 	/*
889 	 *	Hash table: for real service lookups
890 	 */
891 	#define IP_VS_RTAB_BITS 4
892 	#define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
893 	#define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
894 
895 	struct list_head	rs_table[IP_VS_RTAB_SIZE];
896 	/* ip_vs_app */
897 	struct list_head	app_list;
898 	/* ip_vs_proto */
899 	#define IP_VS_PROTO_TAB_SIZE	32	/* must be power of 2 */
900 	struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE];
901 	/* ip_vs_proto_tcp */
902 #ifdef CONFIG_IP_VS_PROTO_TCP
903 	#define	TCP_APP_TAB_BITS	4
904 	#define	TCP_APP_TAB_SIZE	(1 << TCP_APP_TAB_BITS)
905 	#define	TCP_APP_TAB_MASK	(TCP_APP_TAB_SIZE - 1)
906 	struct list_head	tcp_apps[TCP_APP_TAB_SIZE];
907 	spinlock_t		tcp_app_lock;
908 #endif
909 	/* ip_vs_proto_udp */
910 #ifdef CONFIG_IP_VS_PROTO_UDP
911 	#define	UDP_APP_TAB_BITS	4
912 	#define	UDP_APP_TAB_SIZE	(1 << UDP_APP_TAB_BITS)
913 	#define	UDP_APP_TAB_MASK	(UDP_APP_TAB_SIZE - 1)
914 	struct list_head	udp_apps[UDP_APP_TAB_SIZE];
915 	spinlock_t		udp_app_lock;
916 #endif
917 	/* ip_vs_proto_sctp */
918 #ifdef CONFIG_IP_VS_PROTO_SCTP
919 	#define SCTP_APP_TAB_BITS	4
920 	#define SCTP_APP_TAB_SIZE	(1 << SCTP_APP_TAB_BITS)
921 	#define SCTP_APP_TAB_MASK	(SCTP_APP_TAB_SIZE - 1)
922 	/* Hash table for SCTP application incarnations	 */
923 	struct list_head	sctp_apps[SCTP_APP_TAB_SIZE];
924 	spinlock_t		sctp_app_lock;
925 #endif
926 	/* ip_vs_conn */
927 	atomic_t		conn_count;      /*  connection counter */
928 
929 	/* ip_vs_ctl */
930 	struct ip_vs_stats		tot_stats;  /* Statistics & est. */
931 
932 	int			num_services;    /* no of virtual services */
933 
934 	rwlock_t		rs_lock;         /* real services table */
935 	/* Trash for destinations */
936 	struct list_head	dest_trash;
937 	/* Service counters */
938 	atomic_t		ftpsvc_counter;
939 	atomic_t		nullsvc_counter;
940 
941 #ifdef CONFIG_SYSCTL
942 	/* 1/rate drop and drop-entry variables */
943 	struct delayed_work	defense_work;   /* Work handler */
944 	int			drop_rate;
945 	int			drop_counter;
946 	atomic_t		dropentry;
947 	/* locks in ctl.c */
948 	spinlock_t		dropentry_lock;  /* drop entry handling */
949 	spinlock_t		droppacket_lock; /* drop packet handling */
950 	spinlock_t		securetcp_lock;  /* state and timeout tables */
951 
952 	/* sys-ctl struct */
953 	struct ctl_table_header	*sysctl_hdr;
954 	struct ctl_table	*sysctl_tbl;
955 #endif
956 
957 	/* sysctl variables */
958 	int			sysctl_amemthresh;
959 	int			sysctl_am_droprate;
960 	int			sysctl_drop_entry;
961 	int			sysctl_drop_packet;
962 	int			sysctl_secure_tcp;
963 #ifdef CONFIG_IP_VS_NFCT
964 	int			sysctl_conntrack;
965 #endif
966 	int			sysctl_snat_reroute;
967 	int			sysctl_sync_ver;
968 	int			sysctl_sync_ports;
969 	int			sysctl_sync_qlen_max;
970 	int			sysctl_sync_sock_size;
971 	int			sysctl_cache_bypass;
972 	int			sysctl_expire_nodest_conn;
973 	int			sysctl_expire_quiescent_template;
974 	int			sysctl_sync_threshold[2];
975 	unsigned int		sysctl_sync_refresh_period;
976 	int			sysctl_sync_retries;
977 	int			sysctl_nat_icmp_send;
978 	int			sysctl_pmtu_disc;
979 	int			sysctl_backup_only;
980 
981 	/* ip_vs_lblc */
982 	int			sysctl_lblc_expiration;
983 	struct ctl_table_header	*lblc_ctl_header;
984 	struct ctl_table	*lblc_ctl_table;
985 	/* ip_vs_lblcr */
986 	int			sysctl_lblcr_expiration;
987 	struct ctl_table_header	*lblcr_ctl_header;
988 	struct ctl_table	*lblcr_ctl_table;
989 	/* ip_vs_est */
990 	struct list_head	est_list;	/* estimator list */
991 	spinlock_t		est_lock;
992 	struct timer_list	est_timer;	/* Estimation timer */
993 	/* ip_vs_sync */
994 	spinlock_t		sync_lock;
995 	struct ipvs_master_sync_state *ms;
996 	spinlock_t		sync_buff_lock;
997 	struct task_struct	**backup_threads;
998 	int			threads_mask;
999 	int			send_mesg_maxlen;
1000 	int			recv_mesg_maxlen;
1001 	volatile int		sync_state;
1002 	volatile int		master_syncid;
1003 	volatile int		backup_syncid;
1004 	struct mutex		sync_mutex;
1005 	/* multicast interface name */
1006 	char			master_mcast_ifn[IP_VS_IFNAME_MAXLEN];
1007 	char			backup_mcast_ifn[IP_VS_IFNAME_MAXLEN];
1008 	/* net name space ptr */
1009 	struct net		*net;            /* Needed by timer routines */
1010 };
1011 
1012 #define DEFAULT_SYNC_THRESHOLD	3
1013 #define DEFAULT_SYNC_PERIOD	50
1014 #define DEFAULT_SYNC_VER	1
1015 #define DEFAULT_SYNC_REFRESH_PERIOD	(0U * HZ)
1016 #define DEFAULT_SYNC_RETRIES		0
1017 #define IPVS_SYNC_WAKEUP_RATE	8
1018 #define IPVS_SYNC_QLEN_MAX	(IPVS_SYNC_WAKEUP_RATE * 4)
1019 #define IPVS_SYNC_SEND_DELAY	(HZ / 50)
1020 #define IPVS_SYNC_CHECK_PERIOD	HZ
1021 #define IPVS_SYNC_FLUSH_TIME	(HZ * 2)
1022 #define IPVS_SYNC_PORTS_MAX	(1 << 6)
1023 
1024 #ifdef CONFIG_SYSCTL
1025 
1026 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1027 {
1028 	return ipvs->sysctl_sync_threshold[0];
1029 }
1030 
1031 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1032 {
1033 	return ACCESS_ONCE(ipvs->sysctl_sync_threshold[1]);
1034 }
1035 
1036 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1037 {
1038 	return ACCESS_ONCE(ipvs->sysctl_sync_refresh_period);
1039 }
1040 
1041 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1042 {
1043 	return ipvs->sysctl_sync_retries;
1044 }
1045 
1046 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1047 {
1048 	return ipvs->sysctl_sync_ver;
1049 }
1050 
1051 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1052 {
1053 	return ACCESS_ONCE(ipvs->sysctl_sync_ports);
1054 }
1055 
1056 static inline int sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1057 {
1058 	return ipvs->sysctl_sync_qlen_max;
1059 }
1060 
1061 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1062 {
1063 	return ipvs->sysctl_sync_sock_size;
1064 }
1065 
1066 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1067 {
1068 	return ipvs->sysctl_pmtu_disc;
1069 }
1070 
1071 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1072 {
1073 	return ipvs->sync_state & IP_VS_STATE_BACKUP &&
1074 	       ipvs->sysctl_backup_only;
1075 }
1076 
1077 #else
1078 
1079 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1080 {
1081 	return DEFAULT_SYNC_THRESHOLD;
1082 }
1083 
1084 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1085 {
1086 	return DEFAULT_SYNC_PERIOD;
1087 }
1088 
1089 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1090 {
1091 	return DEFAULT_SYNC_REFRESH_PERIOD;
1092 }
1093 
1094 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1095 {
1096 	return DEFAULT_SYNC_RETRIES & 3;
1097 }
1098 
1099 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1100 {
1101 	return DEFAULT_SYNC_VER;
1102 }
1103 
1104 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1105 {
1106 	return 1;
1107 }
1108 
1109 static inline int sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1110 {
1111 	return IPVS_SYNC_QLEN_MAX;
1112 }
1113 
1114 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1115 {
1116 	return 0;
1117 }
1118 
1119 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1120 {
1121 	return 1;
1122 }
1123 
1124 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1125 {
1126 	return 0;
1127 }
1128 
1129 #endif
1130 
1131 /*
1132  *      IPVS core functions
1133  *      (from ip_vs_core.c)
1134  */
1135 extern const char *ip_vs_proto_name(unsigned int proto);
1136 extern void ip_vs_init_hash_table(struct list_head *table, int rows);
1137 #define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t)))
1138 
1139 #define IP_VS_APP_TYPE_FTP	1
1140 
1141 /*
1142  *     ip_vs_conn handling functions
1143  *     (from ip_vs_conn.c)
1144  */
1145 
1146 enum {
1147 	IP_VS_DIR_INPUT = 0,
1148 	IP_VS_DIR_OUTPUT,
1149 	IP_VS_DIR_INPUT_ONLY,
1150 	IP_VS_DIR_LAST,
1151 };
1152 
1153 static inline void ip_vs_conn_fill_param(struct net *net, int af, int protocol,
1154 					 const union nf_inet_addr *caddr,
1155 					 __be16 cport,
1156 					 const union nf_inet_addr *vaddr,
1157 					 __be16 vport,
1158 					 struct ip_vs_conn_param *p)
1159 {
1160 	p->net = net;
1161 	p->af = af;
1162 	p->protocol = protocol;
1163 	p->caddr = caddr;
1164 	p->cport = cport;
1165 	p->vaddr = vaddr;
1166 	p->vport = vport;
1167 	p->pe = NULL;
1168 	p->pe_data = NULL;
1169 }
1170 
1171 struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p);
1172 struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p);
1173 
1174 struct ip_vs_conn * ip_vs_conn_in_get_proto(int af, const struct sk_buff *skb,
1175 					    const struct ip_vs_iphdr *iph,
1176 					    int inverse);
1177 
1178 struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p);
1179 
1180 struct ip_vs_conn * ip_vs_conn_out_get_proto(int af, const struct sk_buff *skb,
1181 					     const struct ip_vs_iphdr *iph,
1182 					     int inverse);
1183 
1184 /* put back the conn without restarting its timer */
1185 static inline void __ip_vs_conn_put(struct ip_vs_conn *cp)
1186 {
1187 	atomic_dec(&cp->refcnt);
1188 }
1189 extern void ip_vs_conn_put(struct ip_vs_conn *cp);
1190 extern void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport);
1191 
1192 struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p,
1193 				  const union nf_inet_addr *daddr,
1194 				  __be16 dport, unsigned int flags,
1195 				  struct ip_vs_dest *dest, __u32 fwmark);
1196 extern void ip_vs_conn_expire_now(struct ip_vs_conn *cp);
1197 
1198 extern const char * ip_vs_state_name(__u16 proto, int state);
1199 
1200 extern void ip_vs_tcp_conn_listen(struct net *net, struct ip_vs_conn *cp);
1201 extern int ip_vs_check_template(struct ip_vs_conn *ct);
1202 extern void ip_vs_random_dropentry(struct net *net);
1203 extern int ip_vs_conn_init(void);
1204 extern void ip_vs_conn_cleanup(void);
1205 
1206 static inline void ip_vs_control_del(struct ip_vs_conn *cp)
1207 {
1208 	struct ip_vs_conn *ctl_cp = cp->control;
1209 	if (!ctl_cp) {
1210 		IP_VS_ERR_BUF("request control DEL for uncontrolled: "
1211 			      "%s:%d to %s:%d\n",
1212 			      IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1213 			      ntohs(cp->cport),
1214 			      IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1215 			      ntohs(cp->vport));
1216 
1217 		return;
1218 	}
1219 
1220 	IP_VS_DBG_BUF(7, "DELeting control for: "
1221 		      "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1222 		      IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1223 		      ntohs(cp->cport),
1224 		      IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1225 		      ntohs(ctl_cp->cport));
1226 
1227 	cp->control = NULL;
1228 	if (atomic_read(&ctl_cp->n_control) == 0) {
1229 		IP_VS_ERR_BUF("BUG control DEL with n=0 : "
1230 			      "%s:%d to %s:%d\n",
1231 			      IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1232 			      ntohs(cp->cport),
1233 			      IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1234 			      ntohs(cp->vport));
1235 
1236 		return;
1237 	}
1238 	atomic_dec(&ctl_cp->n_control);
1239 }
1240 
1241 static inline void
1242 ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp)
1243 {
1244 	if (cp->control) {
1245 		IP_VS_ERR_BUF("request control ADD for already controlled: "
1246 			      "%s:%d to %s:%d\n",
1247 			      IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1248 			      ntohs(cp->cport),
1249 			      IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1250 			      ntohs(cp->vport));
1251 
1252 		ip_vs_control_del(cp);
1253 	}
1254 
1255 	IP_VS_DBG_BUF(7, "ADDing control for: "
1256 		      "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1257 		      IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1258 		      ntohs(cp->cport),
1259 		      IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1260 		      ntohs(ctl_cp->cport));
1261 
1262 	cp->control = ctl_cp;
1263 	atomic_inc(&ctl_cp->n_control);
1264 }
1265 
1266 /*
1267  * IPVS netns init & cleanup functions
1268  */
1269 extern int ip_vs_estimator_net_init(struct net *net);
1270 extern int ip_vs_control_net_init(struct net *net);
1271 extern int ip_vs_protocol_net_init(struct net *net);
1272 extern int ip_vs_app_net_init(struct net *net);
1273 extern int ip_vs_conn_net_init(struct net *net);
1274 extern int ip_vs_sync_net_init(struct net *net);
1275 extern void ip_vs_conn_net_cleanup(struct net *net);
1276 extern void ip_vs_app_net_cleanup(struct net *net);
1277 extern void ip_vs_protocol_net_cleanup(struct net *net);
1278 extern void ip_vs_control_net_cleanup(struct net *net);
1279 extern void ip_vs_estimator_net_cleanup(struct net *net);
1280 extern void ip_vs_sync_net_cleanup(struct net *net);
1281 extern void ip_vs_service_net_cleanup(struct net *net);
1282 
1283 /*
1284  *      IPVS application functions
1285  *      (from ip_vs_app.c)
1286  */
1287 #define IP_VS_APP_MAX_PORTS  8
1288 extern struct ip_vs_app *register_ip_vs_app(struct net *net,
1289 					    struct ip_vs_app *app);
1290 extern void unregister_ip_vs_app(struct net *net, struct ip_vs_app *app);
1291 extern int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1292 extern void ip_vs_unbind_app(struct ip_vs_conn *cp);
1293 extern int register_ip_vs_app_inc(struct net *net, struct ip_vs_app *app,
1294 				  __u16 proto, __u16 port);
1295 extern int ip_vs_app_inc_get(struct ip_vs_app *inc);
1296 extern void ip_vs_app_inc_put(struct ip_vs_app *inc);
1297 
1298 extern int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb);
1299 extern int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb);
1300 
1301 void ip_vs_bind_pe(struct ip_vs_service *svc, struct ip_vs_pe *pe);
1302 void ip_vs_unbind_pe(struct ip_vs_service *svc);
1303 int register_ip_vs_pe(struct ip_vs_pe *pe);
1304 int unregister_ip_vs_pe(struct ip_vs_pe *pe);
1305 struct ip_vs_pe *ip_vs_pe_getbyname(const char *name);
1306 struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name);
1307 
1308 /*
1309  * Use a #define to avoid all of module.h just for these trivial ops
1310  */
1311 #define ip_vs_pe_get(pe)			\
1312 	if (pe && pe->module)			\
1313 		__module_get(pe->module);
1314 
1315 #define ip_vs_pe_put(pe)			\
1316 	if (pe && pe->module)			\
1317 		module_put(pe->module);
1318 
1319 /*
1320  *	IPVS protocol functions (from ip_vs_proto.c)
1321  */
1322 extern int ip_vs_protocol_init(void);
1323 extern void ip_vs_protocol_cleanup(void);
1324 extern void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags);
1325 extern int *ip_vs_create_timeout_table(int *table, int size);
1326 extern int
1327 ip_vs_set_state_timeout(int *table, int num, const char *const *names,
1328 			const char *name, int to);
1329 extern void
1330 ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
1331 			  const struct sk_buff *skb,
1332 			  int offset, const char *msg);
1333 
1334 extern struct ip_vs_protocol ip_vs_protocol_tcp;
1335 extern struct ip_vs_protocol ip_vs_protocol_udp;
1336 extern struct ip_vs_protocol ip_vs_protocol_icmp;
1337 extern struct ip_vs_protocol ip_vs_protocol_esp;
1338 extern struct ip_vs_protocol ip_vs_protocol_ah;
1339 extern struct ip_vs_protocol ip_vs_protocol_sctp;
1340 
1341 /*
1342  *      Registering/unregistering scheduler functions
1343  *      (from ip_vs_sched.c)
1344  */
1345 extern int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1346 extern int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1347 extern int ip_vs_bind_scheduler(struct ip_vs_service *svc,
1348 				struct ip_vs_scheduler *scheduler);
1349 extern int ip_vs_unbind_scheduler(struct ip_vs_service *svc);
1350 extern struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name);
1351 extern void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler);
1352 extern struct ip_vs_conn *
1353 ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb,
1354 	       struct ip_vs_proto_data *pd, int *ignored,
1355 	       struct ip_vs_iphdr *iph);
1356 extern int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb,
1357 			struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph);
1358 
1359 extern void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg);
1360 
1361 
1362 /*
1363  *      IPVS control data and functions (from ip_vs_ctl.c)
1364  */
1365 extern struct ip_vs_stats ip_vs_stats;
1366 extern int sysctl_ip_vs_sync_ver;
1367 
1368 extern struct ip_vs_service *
1369 ip_vs_service_get(struct net *net, int af, __u32 fwmark, __u16 protocol,
1370 		  const union nf_inet_addr *vaddr, __be16 vport);
1371 
1372 static inline void ip_vs_service_put(struct ip_vs_service *svc)
1373 {
1374 	atomic_dec(&svc->usecnt);
1375 }
1376 
1377 extern struct ip_vs_dest *
1378 ip_vs_lookup_real_service(struct net *net, int af, __u16 protocol,
1379 			  const union nf_inet_addr *daddr, __be16 dport);
1380 
1381 extern int ip_vs_use_count_inc(void);
1382 extern void ip_vs_use_count_dec(void);
1383 extern int ip_vs_register_nl_ioctl(void);
1384 extern void ip_vs_unregister_nl_ioctl(void);
1385 extern int ip_vs_control_init(void);
1386 extern void ip_vs_control_cleanup(void);
1387 extern struct ip_vs_dest *
1388 ip_vs_find_dest(struct net *net, int af, const union nf_inet_addr *daddr,
1389 		__be16 dport, const union nf_inet_addr *vaddr, __be16 vport,
1390 		__u16 protocol, __u32 fwmark, __u32 flags);
1391 extern struct ip_vs_dest *ip_vs_try_bind_dest(struct ip_vs_conn *cp);
1392 
1393 
1394 /*
1395  *      IPVS sync daemon data and function prototypes
1396  *      (from ip_vs_sync.c)
1397  */
1398 extern int start_sync_thread(struct net *net, int state, char *mcast_ifn,
1399 			     __u8 syncid);
1400 extern int stop_sync_thread(struct net *net, int state);
1401 extern void ip_vs_sync_conn(struct net *net, struct ip_vs_conn *cp, int pkts);
1402 
1403 
1404 /*
1405  *      IPVS rate estimator prototypes (from ip_vs_est.c)
1406  */
1407 extern void ip_vs_start_estimator(struct net *net, struct ip_vs_stats *stats);
1408 extern void ip_vs_stop_estimator(struct net *net, struct ip_vs_stats *stats);
1409 extern void ip_vs_zero_estimator(struct ip_vs_stats *stats);
1410 extern void ip_vs_read_estimator(struct ip_vs_stats_user *dst,
1411 				 struct ip_vs_stats *stats);
1412 
1413 /*
1414  *	Various IPVS packet transmitters (from ip_vs_xmit.c)
1415  */
1416 extern int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1417 			   struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1418 extern int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1419 			     struct ip_vs_protocol *pp,
1420 			     struct ip_vs_iphdr *iph);
1421 extern int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1422 			  struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1423 extern int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1424 			     struct ip_vs_protocol *pp,
1425 			     struct ip_vs_iphdr *iph);
1426 extern int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1427 			 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1428 extern int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1429 			   struct ip_vs_protocol *pp, int offset,
1430 			   unsigned int hooknum, struct ip_vs_iphdr *iph);
1431 extern void ip_vs_dst_reset(struct ip_vs_dest *dest);
1432 
1433 #ifdef CONFIG_IP_VS_IPV6
1434 extern int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1435 				struct ip_vs_protocol *pp,
1436 				struct ip_vs_iphdr *iph);
1437 extern int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1438 			     struct ip_vs_protocol *pp,
1439 			     struct ip_vs_iphdr *iph);
1440 extern int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1441 				struct ip_vs_protocol *pp,
1442 				struct ip_vs_iphdr *iph);
1443 extern int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1444 			    struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1445 extern int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1446 			      struct ip_vs_protocol *pp, int offset,
1447 			      unsigned int hooknum, struct ip_vs_iphdr *iph);
1448 #endif
1449 
1450 #ifdef CONFIG_SYSCTL
1451 /*
1452  *	This is a simple mechanism to ignore packets when
1453  *	we are loaded. Just set ip_vs_drop_rate to 'n' and
1454  *	we start to drop 1/rate of the packets
1455  */
1456 
1457 static inline int ip_vs_todrop(struct netns_ipvs *ipvs)
1458 {
1459 	if (!ipvs->drop_rate)
1460 		return 0;
1461 	if (--ipvs->drop_counter > 0)
1462 		return 0;
1463 	ipvs->drop_counter = ipvs->drop_rate;
1464 	return 1;
1465 }
1466 #else
1467 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; }
1468 #endif
1469 
1470 /*
1471  *      ip_vs_fwd_tag returns the forwarding tag of the connection
1472  */
1473 #define IP_VS_FWD_METHOD(cp)  (cp->flags & IP_VS_CONN_F_FWD_MASK)
1474 
1475 static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp)
1476 {
1477 	char fwd;
1478 
1479 	switch (IP_VS_FWD_METHOD(cp)) {
1480 	case IP_VS_CONN_F_MASQ:
1481 		fwd = 'M'; break;
1482 	case IP_VS_CONN_F_LOCALNODE:
1483 		fwd = 'L'; break;
1484 	case IP_VS_CONN_F_TUNNEL:
1485 		fwd = 'T'; break;
1486 	case IP_VS_CONN_F_DROUTE:
1487 		fwd = 'R'; break;
1488 	case IP_VS_CONN_F_BYPASS:
1489 		fwd = 'B'; break;
1490 	default:
1491 		fwd = '?'; break;
1492 	}
1493 	return fwd;
1494 }
1495 
1496 extern void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp,
1497 			   struct ip_vs_conn *cp, int dir);
1498 
1499 #ifdef CONFIG_IP_VS_IPV6
1500 extern void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp,
1501 			      struct ip_vs_conn *cp, int dir);
1502 #endif
1503 
1504 extern __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset);
1505 
1506 static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum)
1507 {
1508 	__be32 diff[2] = { ~old, new };
1509 
1510 	return csum_partial(diff, sizeof(diff), oldsum);
1511 }
1512 
1513 #ifdef CONFIG_IP_VS_IPV6
1514 static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new,
1515 					__wsum oldsum)
1516 {
1517 	__be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0],
1518 			    new[3],  new[2],  new[1],  new[0] };
1519 
1520 	return csum_partial(diff, sizeof(diff), oldsum);
1521 }
1522 #endif
1523 
1524 static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum)
1525 {
1526 	__be16 diff[2] = { ~old, new };
1527 
1528 	return csum_partial(diff, sizeof(diff), oldsum);
1529 }
1530 
1531 /*
1532  * Forget current conntrack (unconfirmed) and attach notrack entry
1533  */
1534 static inline void ip_vs_notrack(struct sk_buff *skb)
1535 {
1536 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1537 	enum ip_conntrack_info ctinfo;
1538 	struct nf_conn *ct = nf_ct_get(skb, &ctinfo);
1539 
1540 	if (!ct || !nf_ct_is_untracked(ct)) {
1541 		nf_conntrack_put(skb->nfct);
1542 		skb->nfct = &nf_ct_untracked_get()->ct_general;
1543 		skb->nfctinfo = IP_CT_NEW;
1544 		nf_conntrack_get(skb->nfct);
1545 	}
1546 #endif
1547 }
1548 
1549 #ifdef CONFIG_IP_VS_NFCT
1550 /*
1551  *      Netfilter connection tracking
1552  *      (from ip_vs_nfct.c)
1553  */
1554 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1555 {
1556 #ifdef CONFIG_SYSCTL
1557 	return ipvs->sysctl_conntrack;
1558 #else
1559 	return 0;
1560 #endif
1561 }
1562 
1563 extern void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp,
1564 				   int outin);
1565 extern int ip_vs_confirm_conntrack(struct sk_buff *skb);
1566 extern void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct,
1567 				      struct ip_vs_conn *cp, u_int8_t proto,
1568 				      const __be16 port, int from_rs);
1569 extern void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp);
1570 
1571 #else
1572 
1573 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1574 {
1575 	return 0;
1576 }
1577 
1578 static inline void ip_vs_update_conntrack(struct sk_buff *skb,
1579 					  struct ip_vs_conn *cp, int outin)
1580 {
1581 }
1582 
1583 static inline int ip_vs_confirm_conntrack(struct sk_buff *skb)
1584 {
1585 	return NF_ACCEPT;
1586 }
1587 
1588 static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp)
1589 {
1590 }
1591 /* CONFIG_IP_VS_NFCT */
1592 #endif
1593 
1594 static inline unsigned int
1595 ip_vs_dest_conn_overhead(struct ip_vs_dest *dest)
1596 {
1597 	/*
1598 	 * We think the overhead of processing active connections is 256
1599 	 * times higher than that of inactive connections in average. (This
1600 	 * 256 times might not be accurate, we will change it later) We
1601 	 * use the following formula to estimate the overhead now:
1602 	 *		  dest->activeconns*256 + dest->inactconns
1603 	 */
1604 	return (atomic_read(&dest->activeconns) << 8) +
1605 		atomic_read(&dest->inactconns);
1606 }
1607 
1608 #endif	/* _NET_IP_VS_H */
1609