xref: /linux/include/net/ip_vs.h (revision c0e297dc61f8d4453e07afbea1fa8d0e67cd4a34)
1 /* IP Virtual Server
2  * data structure and functionality definitions
3  */
4 
5 #ifndef _NET_IP_VS_H
6 #define _NET_IP_VS_H
7 
8 #include <linux/ip_vs.h>                /* definitions shared with userland */
9 
10 #include <asm/types.h>                  /* for __uXX types */
11 
12 #include <linux/list.h>                 /* for struct list_head */
13 #include <linux/spinlock.h>             /* for struct rwlock_t */
14 #include <linux/atomic.h>               /* for struct atomic_t */
15 #include <linux/compiler.h>
16 #include <linux/timer.h>
17 #include <linux/bug.h>
18 
19 #include <net/checksum.h>
20 #include <linux/netfilter.h>		/* for union nf_inet_addr */
21 #include <linux/ip.h>
22 #include <linux/ipv6.h>			/* for struct ipv6hdr */
23 #include <net/ipv6.h>
24 #if IS_ENABLED(CONFIG_IP_VS_IPV6)
25 #include <linux/netfilter_ipv6/ip6_tables.h>
26 #endif
27 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
28 #include <net/netfilter/nf_conntrack.h>
29 #endif
30 #include <net/net_namespace.h>		/* Netw namespace */
31 
32 /* Generic access of ipvs struct */
33 static inline struct netns_ipvs *net_ipvs(struct net* net)
34 {
35 	return net->ipvs;
36 }
37 
38 /* Get net ptr from skb in traffic cases
39  * use skb_sknet when call is from userland (ioctl or netlink)
40  */
41 static inline struct net *skb_net(const struct sk_buff *skb)
42 {
43 #ifdef CONFIG_NET_NS
44 #ifdef CONFIG_IP_VS_DEBUG
45 	/*
46 	 * This is used for debug only.
47 	 * Start with the most likely hit
48 	 * End with BUG
49 	 */
50 	if (likely(skb->dev && dev_net(skb->dev)))
51 		return dev_net(skb->dev);
52 	if (skb_dst(skb) && skb_dst(skb)->dev)
53 		return dev_net(skb_dst(skb)->dev);
54 	WARN(skb->sk, "Maybe skb_sknet should be used in %s() at line:%d\n",
55 		      __func__, __LINE__);
56 	if (likely(skb->sk && sock_net(skb->sk)))
57 		return sock_net(skb->sk);
58 	pr_err("There is no net ptr to find in the skb in %s() line:%d\n",
59 		__func__, __LINE__);
60 	BUG();
61 #else
62 	return dev_net(skb->dev ? : skb_dst(skb)->dev);
63 #endif
64 #else
65 	return &init_net;
66 #endif
67 }
68 
69 static inline struct net *skb_sknet(const struct sk_buff *skb)
70 {
71 #ifdef CONFIG_NET_NS
72 #ifdef CONFIG_IP_VS_DEBUG
73 	/* Start with the most likely hit */
74 	if (likely(skb->sk && sock_net(skb->sk)))
75 		return sock_net(skb->sk);
76 	WARN(skb->dev, "Maybe skb_net should be used instead in %s() line:%d\n",
77 		       __func__, __LINE__);
78 	if (likely(skb->dev && dev_net(skb->dev)))
79 		return dev_net(skb->dev);
80 	pr_err("There is no net ptr to find in the skb in %s() line:%d\n",
81 		__func__, __LINE__);
82 	BUG();
83 #else
84 	return sock_net(skb->sk);
85 #endif
86 #else
87 	return &init_net;
88 #endif
89 }
90 
91 /* This one needed for single_open_net since net is stored directly in
92  * private not as a struct i.e. seq_file_net can't be used.
93  */
94 static inline struct net *seq_file_single_net(struct seq_file *seq)
95 {
96 #ifdef CONFIG_NET_NS
97 	return (struct net *)seq->private;
98 #else
99 	return &init_net;
100 #endif
101 }
102 
103 /* Connections' size value needed by ip_vs_ctl.c */
104 extern int ip_vs_conn_tab_size;
105 
106 struct ip_vs_iphdr {
107 	__u32 len;	/* IPv4 simply where L4 starts
108 			 * IPv6 where L4 Transport Header starts */
109 	__u16 fragoffs; /* IPv6 fragment offset, 0 if first frag (or not frag)*/
110 	__s16 protocol;
111 	__s32 flags;
112 	union nf_inet_addr saddr;
113 	union nf_inet_addr daddr;
114 };
115 
116 static inline void *frag_safe_skb_hp(const struct sk_buff *skb, int offset,
117 				      int len, void *buffer,
118 				      const struct ip_vs_iphdr *ipvsh)
119 {
120 	return skb_header_pointer(skb, offset, len, buffer);
121 }
122 
123 static inline void
124 ip_vs_fill_ip4hdr(const void *nh, struct ip_vs_iphdr *iphdr)
125 {
126 	const struct iphdr *iph = nh;
127 
128 	iphdr->len	= iph->ihl * 4;
129 	iphdr->fragoffs	= 0;
130 	iphdr->protocol	= iph->protocol;
131 	iphdr->saddr.ip	= iph->saddr;
132 	iphdr->daddr.ip	= iph->daddr;
133 }
134 
135 /* This function handles filling *ip_vs_iphdr, both for IPv4 and IPv6.
136  * IPv6 requires some extra work, as finding proper header position,
137  * depend on the IPv6 extension headers.
138  */
139 static inline void
140 ip_vs_fill_iph_skb(int af, const struct sk_buff *skb, struct ip_vs_iphdr *iphdr)
141 {
142 #ifdef CONFIG_IP_VS_IPV6
143 	if (af == AF_INET6) {
144 		const struct ipv6hdr *iph =
145 			(struct ipv6hdr *)skb_network_header(skb);
146 		iphdr->saddr.in6 = iph->saddr;
147 		iphdr->daddr.in6 = iph->daddr;
148 		/* ipv6_find_hdr() updates len, flags */
149 		iphdr->len	 = 0;
150 		iphdr->flags	 = 0;
151 		iphdr->protocol  = ipv6_find_hdr(skb, &iphdr->len, -1,
152 						 &iphdr->fragoffs,
153 						 &iphdr->flags);
154 	} else
155 #endif
156 	{
157 		const struct iphdr *iph =
158 			(struct iphdr *)skb_network_header(skb);
159 		iphdr->len	= iph->ihl * 4;
160 		iphdr->fragoffs	= 0;
161 		iphdr->protocol	= iph->protocol;
162 		iphdr->saddr.ip	= iph->saddr;
163 		iphdr->daddr.ip	= iph->daddr;
164 	}
165 }
166 
167 static inline void ip_vs_addr_copy(int af, union nf_inet_addr *dst,
168 				   const union nf_inet_addr *src)
169 {
170 #ifdef CONFIG_IP_VS_IPV6
171 	if (af == AF_INET6)
172 		dst->in6 = src->in6;
173 	else
174 #endif
175 	dst->ip = src->ip;
176 }
177 
178 static inline void ip_vs_addr_set(int af, union nf_inet_addr *dst,
179 				  const union nf_inet_addr *src)
180 {
181 #ifdef CONFIG_IP_VS_IPV6
182 	if (af == AF_INET6) {
183 		dst->in6 = src->in6;
184 		return;
185 	}
186 #endif
187 	dst->ip = src->ip;
188 	dst->all[1] = 0;
189 	dst->all[2] = 0;
190 	dst->all[3] = 0;
191 }
192 
193 static inline int ip_vs_addr_equal(int af, const union nf_inet_addr *a,
194 				   const union nf_inet_addr *b)
195 {
196 #ifdef CONFIG_IP_VS_IPV6
197 	if (af == AF_INET6)
198 		return ipv6_addr_equal(&a->in6, &b->in6);
199 #endif
200 	return a->ip == b->ip;
201 }
202 
203 #ifdef CONFIG_IP_VS_DEBUG
204 #include <linux/net.h>
205 
206 int ip_vs_get_debug_level(void);
207 
208 static inline const char *ip_vs_dbg_addr(int af, char *buf, size_t buf_len,
209 					 const union nf_inet_addr *addr,
210 					 int *idx)
211 {
212 	int len;
213 #ifdef CONFIG_IP_VS_IPV6
214 	if (af == AF_INET6)
215 		len = snprintf(&buf[*idx], buf_len - *idx, "[%pI6c]",
216 			       &addr->in6) + 1;
217 	else
218 #endif
219 		len = snprintf(&buf[*idx], buf_len - *idx, "%pI4",
220 			       &addr->ip) + 1;
221 
222 	*idx += len;
223 	BUG_ON(*idx > buf_len + 1);
224 	return &buf[*idx - len];
225 }
226 
227 #define IP_VS_DBG_BUF(level, msg, ...)					\
228 	do {								\
229 		char ip_vs_dbg_buf[160];				\
230 		int ip_vs_dbg_idx = 0;					\
231 		if (level <= ip_vs_get_debug_level())			\
232 			printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__);	\
233 	} while (0)
234 #define IP_VS_ERR_BUF(msg...)						\
235 	do {								\
236 		char ip_vs_dbg_buf[160];				\
237 		int ip_vs_dbg_idx = 0;					\
238 		pr_err(msg);						\
239 	} while (0)
240 
241 /* Only use from within IP_VS_DBG_BUF() or IP_VS_ERR_BUF macros */
242 #define IP_VS_DBG_ADDR(af, addr)					\
243 	ip_vs_dbg_addr(af, ip_vs_dbg_buf,				\
244 		       sizeof(ip_vs_dbg_buf), addr,			\
245 		       &ip_vs_dbg_idx)
246 
247 #define IP_VS_DBG(level, msg, ...)					\
248 	do {								\
249 		if (level <= ip_vs_get_debug_level())			\
250 			printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__);	\
251 	} while (0)
252 #define IP_VS_DBG_RL(msg, ...)						\
253 	do {								\
254 		if (net_ratelimit())					\
255 			printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__);	\
256 	} while (0)
257 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg)			\
258 	do {								\
259 		if (level <= ip_vs_get_debug_level())			\
260 			pp->debug_packet(af, pp, skb, ofs, msg);	\
261 	} while (0)
262 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg)			\
263 	do {								\
264 		if (level <= ip_vs_get_debug_level() &&			\
265 		    net_ratelimit())					\
266 			pp->debug_packet(af, pp, skb, ofs, msg);	\
267 	} while (0)
268 #else	/* NO DEBUGGING at ALL */
269 #define IP_VS_DBG_BUF(level, msg...)  do {} while (0)
270 #define IP_VS_ERR_BUF(msg...)  do {} while (0)
271 #define IP_VS_DBG(level, msg...)  do {} while (0)
272 #define IP_VS_DBG_RL(msg...)  do {} while (0)
273 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg)	do {} while (0)
274 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg)	do {} while (0)
275 #endif
276 
277 #define IP_VS_BUG() BUG()
278 #define IP_VS_ERR_RL(msg, ...)						\
279 	do {								\
280 		if (net_ratelimit())					\
281 			pr_err(msg, ##__VA_ARGS__);			\
282 	} while (0)
283 
284 #ifdef CONFIG_IP_VS_DEBUG
285 #define EnterFunction(level)						\
286 	do {								\
287 		if (level <= ip_vs_get_debug_level())			\
288 			printk(KERN_DEBUG				\
289 			       pr_fmt("Enter: %s, %s line %i\n"),	\
290 			       __func__, __FILE__, __LINE__);		\
291 	} while (0)
292 #define LeaveFunction(level)						\
293 	do {								\
294 		if (level <= ip_vs_get_debug_level())			\
295 			printk(KERN_DEBUG				\
296 			       pr_fmt("Leave: %s, %s line %i\n"),	\
297 			       __func__, __FILE__, __LINE__);		\
298 	} while (0)
299 #else
300 #define EnterFunction(level)   do {} while (0)
301 #define LeaveFunction(level)   do {} while (0)
302 #endif
303 
304 /* The port number of FTP service (in network order). */
305 #define FTPPORT  cpu_to_be16(21)
306 #define FTPDATA  cpu_to_be16(20)
307 
308 /* TCP State Values */
309 enum {
310 	IP_VS_TCP_S_NONE = 0,
311 	IP_VS_TCP_S_ESTABLISHED,
312 	IP_VS_TCP_S_SYN_SENT,
313 	IP_VS_TCP_S_SYN_RECV,
314 	IP_VS_TCP_S_FIN_WAIT,
315 	IP_VS_TCP_S_TIME_WAIT,
316 	IP_VS_TCP_S_CLOSE,
317 	IP_VS_TCP_S_CLOSE_WAIT,
318 	IP_VS_TCP_S_LAST_ACK,
319 	IP_VS_TCP_S_LISTEN,
320 	IP_VS_TCP_S_SYNACK,
321 	IP_VS_TCP_S_LAST
322 };
323 
324 /* UDP State Values */
325 enum {
326 	IP_VS_UDP_S_NORMAL,
327 	IP_VS_UDP_S_LAST,
328 };
329 
330 /* ICMP State Values */
331 enum {
332 	IP_VS_ICMP_S_NORMAL,
333 	IP_VS_ICMP_S_LAST,
334 };
335 
336 /* SCTP State Values */
337 enum ip_vs_sctp_states {
338 	IP_VS_SCTP_S_NONE,
339 	IP_VS_SCTP_S_INIT1,
340 	IP_VS_SCTP_S_INIT,
341 	IP_VS_SCTP_S_COOKIE_SENT,
342 	IP_VS_SCTP_S_COOKIE_REPLIED,
343 	IP_VS_SCTP_S_COOKIE_WAIT,
344 	IP_VS_SCTP_S_COOKIE,
345 	IP_VS_SCTP_S_COOKIE_ECHOED,
346 	IP_VS_SCTP_S_ESTABLISHED,
347 	IP_VS_SCTP_S_SHUTDOWN_SENT,
348 	IP_VS_SCTP_S_SHUTDOWN_RECEIVED,
349 	IP_VS_SCTP_S_SHUTDOWN_ACK_SENT,
350 	IP_VS_SCTP_S_REJECTED,
351 	IP_VS_SCTP_S_CLOSED,
352 	IP_VS_SCTP_S_LAST
353 };
354 
355 /* Delta sequence info structure
356  * Each ip_vs_conn has 2 (output AND input seq. changes).
357  * Only used in the VS/NAT.
358  */
359 struct ip_vs_seq {
360 	__u32			init_seq;	/* Add delta from this seq */
361 	__u32			delta;		/* Delta in sequence numbers */
362 	__u32			previous_delta;	/* Delta in sequence numbers
363 						 * before last resized pkt */
364 };
365 
366 /* counters per cpu */
367 struct ip_vs_counters {
368 	__u64		conns;		/* connections scheduled */
369 	__u64		inpkts;		/* incoming packets */
370 	__u64		outpkts;	/* outgoing packets */
371 	__u64		inbytes;	/* incoming bytes */
372 	__u64		outbytes;	/* outgoing bytes */
373 };
374 /* Stats per cpu */
375 struct ip_vs_cpu_stats {
376 	struct ip_vs_counters   cnt;
377 	struct u64_stats_sync   syncp;
378 };
379 
380 /* IPVS statistics objects */
381 struct ip_vs_estimator {
382 	struct list_head	list;
383 
384 	u64			last_inbytes;
385 	u64			last_outbytes;
386 	u64			last_conns;
387 	u64			last_inpkts;
388 	u64			last_outpkts;
389 
390 	u64			cps;
391 	u64			inpps;
392 	u64			outpps;
393 	u64			inbps;
394 	u64			outbps;
395 };
396 
397 /*
398  * IPVS statistics object, 64-bit kernel version of struct ip_vs_stats_user
399  */
400 struct ip_vs_kstats {
401 	u64			conns;		/* connections scheduled */
402 	u64			inpkts;		/* incoming packets */
403 	u64			outpkts;	/* outgoing packets */
404 	u64			inbytes;	/* incoming bytes */
405 	u64			outbytes;	/* outgoing bytes */
406 
407 	u64			cps;		/* current connection rate */
408 	u64			inpps;		/* current in packet rate */
409 	u64			outpps;		/* current out packet rate */
410 	u64			inbps;		/* current in byte rate */
411 	u64			outbps;		/* current out byte rate */
412 };
413 
414 struct ip_vs_stats {
415 	struct ip_vs_kstats	kstats;		/* kernel statistics */
416 	struct ip_vs_estimator	est;		/* estimator */
417 	struct ip_vs_cpu_stats __percpu	*cpustats;	/* per cpu counters */
418 	spinlock_t		lock;		/* spin lock */
419 	struct ip_vs_kstats	kstats0;	/* reset values */
420 };
421 
422 struct dst_entry;
423 struct iphdr;
424 struct ip_vs_conn;
425 struct ip_vs_app;
426 struct sk_buff;
427 struct ip_vs_proto_data;
428 
429 struct ip_vs_protocol {
430 	struct ip_vs_protocol	*next;
431 	char			*name;
432 	u16			protocol;
433 	u16			num_states;
434 	int			dont_defrag;
435 
436 	void (*init)(struct ip_vs_protocol *pp);
437 
438 	void (*exit)(struct ip_vs_protocol *pp);
439 
440 	int (*init_netns)(struct net *net, struct ip_vs_proto_data *pd);
441 
442 	void (*exit_netns)(struct net *net, struct ip_vs_proto_data *pd);
443 
444 	int (*conn_schedule)(int af, struct sk_buff *skb,
445 			     struct ip_vs_proto_data *pd,
446 			     int *verdict, struct ip_vs_conn **cpp,
447 			     struct ip_vs_iphdr *iph);
448 
449 	struct ip_vs_conn *
450 	(*conn_in_get)(int af,
451 		       const struct sk_buff *skb,
452 		       const struct ip_vs_iphdr *iph,
453 		       int inverse);
454 
455 	struct ip_vs_conn *
456 	(*conn_out_get)(int af,
457 			const struct sk_buff *skb,
458 			const struct ip_vs_iphdr *iph,
459 			int inverse);
460 
461 	int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
462 			    struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
463 
464 	int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
465 			    struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
466 
467 	int (*csum_check)(int af, struct sk_buff *skb,
468 			  struct ip_vs_protocol *pp);
469 
470 	const char *(*state_name)(int state);
471 
472 	void (*state_transition)(struct ip_vs_conn *cp, int direction,
473 				 const struct sk_buff *skb,
474 				 struct ip_vs_proto_data *pd);
475 
476 	int (*register_app)(struct net *net, struct ip_vs_app *inc);
477 
478 	void (*unregister_app)(struct net *net, struct ip_vs_app *inc);
479 
480 	int (*app_conn_bind)(struct ip_vs_conn *cp);
481 
482 	void (*debug_packet)(int af, struct ip_vs_protocol *pp,
483 			     const struct sk_buff *skb,
484 			     int offset,
485 			     const char *msg);
486 
487 	void (*timeout_change)(struct ip_vs_proto_data *pd, int flags);
488 };
489 
490 /* protocol data per netns */
491 struct ip_vs_proto_data {
492 	struct ip_vs_proto_data	*next;
493 	struct ip_vs_protocol	*pp;
494 	int			*timeout_table;	/* protocol timeout table */
495 	atomic_t		appcnt;		/* counter of proto app incs. */
496 	struct tcp_states_t	*tcp_state_table;
497 };
498 
499 struct ip_vs_protocol   *ip_vs_proto_get(unsigned short proto);
500 struct ip_vs_proto_data *ip_vs_proto_data_get(struct net *net,
501 					      unsigned short proto);
502 
503 struct ip_vs_conn_param {
504 	struct net			*net;
505 	const union nf_inet_addr	*caddr;
506 	const union nf_inet_addr	*vaddr;
507 	__be16				cport;
508 	__be16				vport;
509 	__u16				protocol;
510 	u16				af;
511 
512 	const struct ip_vs_pe		*pe;
513 	char				*pe_data;
514 	__u8				pe_data_len;
515 };
516 
517 /* IP_VS structure allocated for each dynamically scheduled connection */
518 struct ip_vs_conn {
519 	struct hlist_node	c_list;         /* hashed list heads */
520 	/* Protocol, addresses and port numbers */
521 	__be16                  cport;
522 	__be16                  dport;
523 	__be16                  vport;
524 	u16			af;		/* address family */
525 	union nf_inet_addr      caddr;          /* client address */
526 	union nf_inet_addr      vaddr;          /* virtual address */
527 	union nf_inet_addr      daddr;          /* destination address */
528 	volatile __u32          flags;          /* status flags */
529 	__u16                   protocol;       /* Which protocol (TCP/UDP) */
530 	__u16			daf;		/* Address family of the dest */
531 #ifdef CONFIG_NET_NS
532 	struct net              *net;           /* Name space */
533 #endif
534 
535 	/* counter and timer */
536 	atomic_t		refcnt;		/* reference count */
537 	struct timer_list	timer;		/* Expiration timer */
538 	volatile unsigned long	timeout;	/* timeout */
539 
540 	/* Flags and state transition */
541 	spinlock_t              lock;           /* lock for state transition */
542 	volatile __u16          state;          /* state info */
543 	volatile __u16          old_state;      /* old state, to be used for
544 						 * state transition triggerd
545 						 * synchronization
546 						 */
547 	__u32			fwmark;		/* Fire wall mark from skb */
548 	unsigned long		sync_endtime;	/* jiffies + sent_retries */
549 
550 	/* Control members */
551 	struct ip_vs_conn       *control;       /* Master control connection */
552 	atomic_t                n_control;      /* Number of controlled ones */
553 	struct ip_vs_dest       *dest;          /* real server */
554 	atomic_t                in_pkts;        /* incoming packet counter */
555 
556 	/* Packet transmitter for different forwarding methods.  If it
557 	 * mangles the packet, it must return NF_DROP or better NF_STOLEN,
558 	 * otherwise this must be changed to a sk_buff **.
559 	 * NF_ACCEPT can be returned when destination is local.
560 	 */
561 	int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp,
562 			   struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
563 
564 	/* Note: we can group the following members into a structure,
565 	 * in order to save more space, and the following members are
566 	 * only used in VS/NAT anyway
567 	 */
568 	struct ip_vs_app        *app;           /* bound ip_vs_app object */
569 	void                    *app_data;      /* Application private data */
570 	struct ip_vs_seq        in_seq;         /* incoming seq. struct */
571 	struct ip_vs_seq        out_seq;        /* outgoing seq. struct */
572 
573 	const struct ip_vs_pe	*pe;
574 	char			*pe_data;
575 	__u8			pe_data_len;
576 
577 	struct rcu_head		rcu_head;
578 };
579 
580 /* To save some memory in conn table when name space is disabled. */
581 static inline struct net *ip_vs_conn_net(const struct ip_vs_conn *cp)
582 {
583 #ifdef CONFIG_NET_NS
584 	return cp->net;
585 #else
586 	return &init_net;
587 #endif
588 }
589 
590 static inline void ip_vs_conn_net_set(struct ip_vs_conn *cp, struct net *net)
591 {
592 #ifdef CONFIG_NET_NS
593 	cp->net = net;
594 #endif
595 }
596 
597 static inline int ip_vs_conn_net_eq(const struct ip_vs_conn *cp,
598 				    struct net *net)
599 {
600 #ifdef CONFIG_NET_NS
601 	return cp->net == net;
602 #else
603 	return 1;
604 #endif
605 }
606 
607 /* Extended internal versions of struct ip_vs_service_user and ip_vs_dest_user
608  * for IPv6 support.
609  *
610  * We need these to conveniently pass around service and destination
611  * options, but unfortunately, we also need to keep the old definitions to
612  * maintain userspace backwards compatibility for the setsockopt interface.
613  */
614 struct ip_vs_service_user_kern {
615 	/* virtual service addresses */
616 	u16			af;
617 	u16			protocol;
618 	union nf_inet_addr	addr;		/* virtual ip address */
619 	__be16			port;
620 	u32			fwmark;		/* firwall mark of service */
621 
622 	/* virtual service options */
623 	char			*sched_name;
624 	char			*pe_name;
625 	unsigned int		flags;		/* virtual service flags */
626 	unsigned int		timeout;	/* persistent timeout in sec */
627 	__be32			netmask;	/* persistent netmask or plen */
628 };
629 
630 
631 struct ip_vs_dest_user_kern {
632 	/* destination server address */
633 	union nf_inet_addr	addr;
634 	__be16			port;
635 
636 	/* real server options */
637 	unsigned int		conn_flags;	/* connection flags */
638 	int			weight;		/* destination weight */
639 
640 	/* thresholds for active connections */
641 	u32			u_threshold;	/* upper threshold */
642 	u32			l_threshold;	/* lower threshold */
643 
644 	/* Address family of addr */
645 	u16			af;
646 };
647 
648 
649 /*
650  * The information about the virtual service offered to the net and the
651  * forwarding entries.
652  */
653 struct ip_vs_service {
654 	struct hlist_node	s_list;   /* for normal service table */
655 	struct hlist_node	f_list;   /* for fwmark-based service table */
656 	atomic_t		refcnt;   /* reference counter */
657 
658 	u16			af;       /* address family */
659 	__u16			protocol; /* which protocol (TCP/UDP) */
660 	union nf_inet_addr	addr;	  /* IP address for virtual service */
661 	__be16			port;	  /* port number for the service */
662 	__u32                   fwmark;   /* firewall mark of the service */
663 	unsigned int		flags;	  /* service status flags */
664 	unsigned int		timeout;  /* persistent timeout in ticks */
665 	__be32			netmask;  /* grouping granularity, mask/plen */
666 	struct net		*net;
667 
668 	struct list_head	destinations;  /* real server d-linked list */
669 	__u32			num_dests;     /* number of servers */
670 	struct ip_vs_stats      stats;         /* statistics for the service */
671 
672 	/* for scheduling */
673 	struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */
674 	spinlock_t		sched_lock;    /* lock sched_data */
675 	void			*sched_data;   /* scheduler application data */
676 
677 	/* alternate persistence engine */
678 	struct ip_vs_pe __rcu	*pe;
679 
680 	struct rcu_head		rcu_head;
681 };
682 
683 /* Information for cached dst */
684 struct ip_vs_dest_dst {
685 	struct dst_entry	*dst_cache;	/* destination cache entry */
686 	u32			dst_cookie;
687 	union nf_inet_addr	dst_saddr;
688 	struct rcu_head		rcu_head;
689 };
690 
691 /* The real server destination forwarding entry with ip address, port number,
692  * and so on.
693  */
694 struct ip_vs_dest {
695 	struct list_head	n_list;   /* for the dests in the service */
696 	struct hlist_node	d_list;   /* for table with all the dests */
697 
698 	u16			af;		/* address family */
699 	__be16			port;		/* port number of the server */
700 	union nf_inet_addr	addr;		/* IP address of the server */
701 	volatile unsigned int	flags;		/* dest status flags */
702 	atomic_t		conn_flags;	/* flags to copy to conn */
703 	atomic_t		weight;		/* server weight */
704 
705 	atomic_t		refcnt;		/* reference counter */
706 	struct ip_vs_stats      stats;          /* statistics */
707 	unsigned long		idle_start;	/* start time, jiffies */
708 
709 	/* connection counters and thresholds */
710 	atomic_t		activeconns;	/* active connections */
711 	atomic_t		inactconns;	/* inactive connections */
712 	atomic_t		persistconns;	/* persistent connections */
713 	__u32			u_threshold;	/* upper threshold */
714 	__u32			l_threshold;	/* lower threshold */
715 
716 	/* for destination cache */
717 	spinlock_t		dst_lock;	/* lock of dst_cache */
718 	struct ip_vs_dest_dst __rcu *dest_dst;	/* cached dst info */
719 
720 	/* for virtual service */
721 	struct ip_vs_service __rcu *svc;	/* service it belongs to */
722 	__u16			protocol;	/* which protocol (TCP/UDP) */
723 	__be16			vport;		/* virtual port number */
724 	union nf_inet_addr	vaddr;		/* virtual IP address */
725 	__u32			vfwmark;	/* firewall mark of service */
726 
727 	struct list_head	t_list;		/* in dest_trash */
728 	unsigned int		in_rs_table:1;	/* we are in rs_table */
729 };
730 
731 /* The scheduler object */
732 struct ip_vs_scheduler {
733 	struct list_head	n_list;		/* d-linked list head */
734 	char			*name;		/* scheduler name */
735 	atomic_t		refcnt;		/* reference counter */
736 	struct module		*module;	/* THIS_MODULE/NULL */
737 
738 	/* scheduler initializing service */
739 	int (*init_service)(struct ip_vs_service *svc);
740 	/* scheduling service finish */
741 	void (*done_service)(struct ip_vs_service *svc);
742 	/* dest is linked */
743 	int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
744 	/* dest is unlinked */
745 	int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
746 	/* dest is updated */
747 	int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
748 
749 	/* selecting a server from the given service */
750 	struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc,
751 				       const struct sk_buff *skb,
752 				       struct ip_vs_iphdr *iph);
753 };
754 
755 /* The persistence engine object */
756 struct ip_vs_pe {
757 	struct list_head	n_list;		/* d-linked list head */
758 	char			*name;		/* scheduler name */
759 	atomic_t		refcnt;		/* reference counter */
760 	struct module		*module;	/* THIS_MODULE/NULL */
761 
762 	/* get the connection template, if any */
763 	int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb);
764 	bool (*ct_match)(const struct ip_vs_conn_param *p,
765 			 struct ip_vs_conn *ct);
766 	u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval,
767 			   bool inverse);
768 	int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf);
769 };
770 
771 /* The application module object (a.k.a. app incarnation) */
772 struct ip_vs_app {
773 	struct list_head	a_list;		/* member in app list */
774 	int			type;		/* IP_VS_APP_TYPE_xxx */
775 	char			*name;		/* application module name */
776 	__u16			protocol;
777 	struct module		*module;	/* THIS_MODULE/NULL */
778 	struct list_head	incs_list;	/* list of incarnations */
779 
780 	/* members for application incarnations */
781 	struct list_head	p_list;		/* member in proto app list */
782 	struct ip_vs_app	*app;		/* its real application */
783 	__be16			port;		/* port number in net order */
784 	atomic_t		usecnt;		/* usage counter */
785 	struct rcu_head		rcu_head;
786 
787 	/* output hook: Process packet in inout direction, diff set for TCP.
788 	 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
789 	 *	   2=Mangled but checksum was not updated
790 	 */
791 	int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *,
792 		       struct sk_buff *, int *diff);
793 
794 	/* input hook: Process packet in outin direction, diff set for TCP.
795 	 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
796 	 *	   2=Mangled but checksum was not updated
797 	 */
798 	int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *,
799 		      struct sk_buff *, int *diff);
800 
801 	/* ip_vs_app initializer */
802 	int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *);
803 
804 	/* ip_vs_app finish */
805 	int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *);
806 
807 
808 	/* not used now */
809 	int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *,
810 			 struct ip_vs_protocol *);
811 
812 	void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *);
813 
814 	int *			timeout_table;
815 	int *			timeouts;
816 	int			timeouts_size;
817 
818 	int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app,
819 			     int *verdict, struct ip_vs_conn **cpp);
820 
821 	struct ip_vs_conn *
822 	(*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app,
823 		       const struct iphdr *iph, int inverse);
824 
825 	struct ip_vs_conn *
826 	(*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app,
827 			const struct iphdr *iph, int inverse);
828 
829 	int (*state_transition)(struct ip_vs_conn *cp, int direction,
830 				const struct sk_buff *skb,
831 				struct ip_vs_app *app);
832 
833 	void (*timeout_change)(struct ip_vs_app *app, int flags);
834 };
835 
836 struct ipvs_master_sync_state {
837 	struct list_head	sync_queue;
838 	struct ip_vs_sync_buff	*sync_buff;
839 	unsigned long		sync_queue_len;
840 	unsigned int		sync_queue_delay;
841 	struct task_struct	*master_thread;
842 	struct delayed_work	master_wakeup_work;
843 	struct netns_ipvs	*ipvs;
844 };
845 
846 /* How much time to keep dests in trash */
847 #define IP_VS_DEST_TRASH_PERIOD		(120 * HZ)
848 
849 /* IPVS in network namespace */
850 struct netns_ipvs {
851 	int			gen;		/* Generation */
852 	int			enable;		/* enable like nf_hooks do */
853 	/* Hash table: for real service lookups */
854 	#define IP_VS_RTAB_BITS 4
855 	#define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
856 	#define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
857 
858 	struct hlist_head	rs_table[IP_VS_RTAB_SIZE];
859 	/* ip_vs_app */
860 	struct list_head	app_list;
861 	/* ip_vs_proto */
862 	#define IP_VS_PROTO_TAB_SIZE	32	/* must be power of 2 */
863 	struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE];
864 	/* ip_vs_proto_tcp */
865 #ifdef CONFIG_IP_VS_PROTO_TCP
866 	#define	TCP_APP_TAB_BITS	4
867 	#define	TCP_APP_TAB_SIZE	(1 << TCP_APP_TAB_BITS)
868 	#define	TCP_APP_TAB_MASK	(TCP_APP_TAB_SIZE - 1)
869 	struct list_head	tcp_apps[TCP_APP_TAB_SIZE];
870 #endif
871 	/* ip_vs_proto_udp */
872 #ifdef CONFIG_IP_VS_PROTO_UDP
873 	#define	UDP_APP_TAB_BITS	4
874 	#define	UDP_APP_TAB_SIZE	(1 << UDP_APP_TAB_BITS)
875 	#define	UDP_APP_TAB_MASK	(UDP_APP_TAB_SIZE - 1)
876 	struct list_head	udp_apps[UDP_APP_TAB_SIZE];
877 #endif
878 	/* ip_vs_proto_sctp */
879 #ifdef CONFIG_IP_VS_PROTO_SCTP
880 	#define SCTP_APP_TAB_BITS	4
881 	#define SCTP_APP_TAB_SIZE	(1 << SCTP_APP_TAB_BITS)
882 	#define SCTP_APP_TAB_MASK	(SCTP_APP_TAB_SIZE - 1)
883 	/* Hash table for SCTP application incarnations	 */
884 	struct list_head	sctp_apps[SCTP_APP_TAB_SIZE];
885 #endif
886 	/* ip_vs_conn */
887 	atomic_t		conn_count;      /* connection counter */
888 
889 	/* ip_vs_ctl */
890 	struct ip_vs_stats		tot_stats;  /* Statistics & est. */
891 
892 	int			num_services;    /* no of virtual services */
893 
894 	/* Trash for destinations */
895 	struct list_head	dest_trash;
896 	spinlock_t		dest_trash_lock;
897 	struct timer_list	dest_trash_timer; /* expiration timer */
898 	/* Service counters */
899 	atomic_t		ftpsvc_counter;
900 	atomic_t		nullsvc_counter;
901 
902 #ifdef CONFIG_SYSCTL
903 	/* 1/rate drop and drop-entry variables */
904 	struct delayed_work	defense_work;   /* Work handler */
905 	int			drop_rate;
906 	int			drop_counter;
907 	atomic_t		dropentry;
908 	/* locks in ctl.c */
909 	spinlock_t		dropentry_lock;  /* drop entry handling */
910 	spinlock_t		droppacket_lock; /* drop packet handling */
911 	spinlock_t		securetcp_lock;  /* state and timeout tables */
912 
913 	/* sys-ctl struct */
914 	struct ctl_table_header	*sysctl_hdr;
915 	struct ctl_table	*sysctl_tbl;
916 #endif
917 
918 	/* sysctl variables */
919 	int			sysctl_amemthresh;
920 	int			sysctl_am_droprate;
921 	int			sysctl_drop_entry;
922 	int			sysctl_drop_packet;
923 	int			sysctl_secure_tcp;
924 #ifdef CONFIG_IP_VS_NFCT
925 	int			sysctl_conntrack;
926 #endif
927 	int			sysctl_snat_reroute;
928 	int			sysctl_sync_ver;
929 	int			sysctl_sync_ports;
930 	int			sysctl_sync_persist_mode;
931 	unsigned long		sysctl_sync_qlen_max;
932 	int			sysctl_sync_sock_size;
933 	int			sysctl_cache_bypass;
934 	int			sysctl_expire_nodest_conn;
935 	int			sysctl_sloppy_tcp;
936 	int			sysctl_sloppy_sctp;
937 	int			sysctl_expire_quiescent_template;
938 	int			sysctl_sync_threshold[2];
939 	unsigned int		sysctl_sync_refresh_period;
940 	int			sysctl_sync_retries;
941 	int			sysctl_nat_icmp_send;
942 	int			sysctl_pmtu_disc;
943 	int			sysctl_backup_only;
944 	int			sysctl_conn_reuse_mode;
945 
946 	/* ip_vs_lblc */
947 	int			sysctl_lblc_expiration;
948 	struct ctl_table_header	*lblc_ctl_header;
949 	struct ctl_table	*lblc_ctl_table;
950 	/* ip_vs_lblcr */
951 	int			sysctl_lblcr_expiration;
952 	struct ctl_table_header	*lblcr_ctl_header;
953 	struct ctl_table	*lblcr_ctl_table;
954 	/* ip_vs_est */
955 	struct list_head	est_list;	/* estimator list */
956 	spinlock_t		est_lock;
957 	struct timer_list	est_timer;	/* Estimation timer */
958 	/* ip_vs_sync */
959 	spinlock_t		sync_lock;
960 	struct ipvs_master_sync_state *ms;
961 	spinlock_t		sync_buff_lock;
962 	struct task_struct	**backup_threads;
963 	int			threads_mask;
964 	int			send_mesg_maxlen;
965 	int			recv_mesg_maxlen;
966 	volatile int		sync_state;
967 	volatile int		master_syncid;
968 	volatile int		backup_syncid;
969 	struct mutex		sync_mutex;
970 	/* multicast interface name */
971 	char			master_mcast_ifn[IP_VS_IFNAME_MAXLEN];
972 	char			backup_mcast_ifn[IP_VS_IFNAME_MAXLEN];
973 	/* net name space ptr */
974 	struct net		*net;            /* Needed by timer routines */
975 	/* Number of heterogeneous destinations, needed becaus heterogeneous
976 	 * are not supported when synchronization is enabled.
977 	 */
978 	unsigned int		mixed_address_family_dests;
979 };
980 
981 #define DEFAULT_SYNC_THRESHOLD	3
982 #define DEFAULT_SYNC_PERIOD	50
983 #define DEFAULT_SYNC_VER	1
984 #define DEFAULT_SLOPPY_TCP	0
985 #define DEFAULT_SLOPPY_SCTP	0
986 #define DEFAULT_SYNC_REFRESH_PERIOD	(0U * HZ)
987 #define DEFAULT_SYNC_RETRIES		0
988 #define IPVS_SYNC_WAKEUP_RATE	8
989 #define IPVS_SYNC_QLEN_MAX	(IPVS_SYNC_WAKEUP_RATE * 4)
990 #define IPVS_SYNC_SEND_DELAY	(HZ / 50)
991 #define IPVS_SYNC_CHECK_PERIOD	HZ
992 #define IPVS_SYNC_FLUSH_TIME	(HZ * 2)
993 #define IPVS_SYNC_PORTS_MAX	(1 << 6)
994 
995 #ifdef CONFIG_SYSCTL
996 
997 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
998 {
999 	return ipvs->sysctl_sync_threshold[0];
1000 }
1001 
1002 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1003 {
1004 	return ACCESS_ONCE(ipvs->sysctl_sync_threshold[1]);
1005 }
1006 
1007 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1008 {
1009 	return ACCESS_ONCE(ipvs->sysctl_sync_refresh_period);
1010 }
1011 
1012 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1013 {
1014 	return ipvs->sysctl_sync_retries;
1015 }
1016 
1017 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1018 {
1019 	return ipvs->sysctl_sync_ver;
1020 }
1021 
1022 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1023 {
1024 	return ipvs->sysctl_sloppy_tcp;
1025 }
1026 
1027 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1028 {
1029 	return ipvs->sysctl_sloppy_sctp;
1030 }
1031 
1032 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1033 {
1034 	return ACCESS_ONCE(ipvs->sysctl_sync_ports);
1035 }
1036 
1037 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1038 {
1039 	return ipvs->sysctl_sync_persist_mode;
1040 }
1041 
1042 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1043 {
1044 	return ipvs->sysctl_sync_qlen_max;
1045 }
1046 
1047 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1048 {
1049 	return ipvs->sysctl_sync_sock_size;
1050 }
1051 
1052 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1053 {
1054 	return ipvs->sysctl_pmtu_disc;
1055 }
1056 
1057 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1058 {
1059 	return ipvs->sync_state & IP_VS_STATE_BACKUP &&
1060 	       ipvs->sysctl_backup_only;
1061 }
1062 
1063 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1064 {
1065 	return ipvs->sysctl_conn_reuse_mode;
1066 }
1067 
1068 #else
1069 
1070 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1071 {
1072 	return DEFAULT_SYNC_THRESHOLD;
1073 }
1074 
1075 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1076 {
1077 	return DEFAULT_SYNC_PERIOD;
1078 }
1079 
1080 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1081 {
1082 	return DEFAULT_SYNC_REFRESH_PERIOD;
1083 }
1084 
1085 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1086 {
1087 	return DEFAULT_SYNC_RETRIES & 3;
1088 }
1089 
1090 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1091 {
1092 	return DEFAULT_SYNC_VER;
1093 }
1094 
1095 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1096 {
1097 	return DEFAULT_SLOPPY_TCP;
1098 }
1099 
1100 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1101 {
1102 	return DEFAULT_SLOPPY_SCTP;
1103 }
1104 
1105 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1106 {
1107 	return 1;
1108 }
1109 
1110 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1111 {
1112 	return 0;
1113 }
1114 
1115 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1116 {
1117 	return IPVS_SYNC_QLEN_MAX;
1118 }
1119 
1120 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1121 {
1122 	return 0;
1123 }
1124 
1125 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1126 {
1127 	return 1;
1128 }
1129 
1130 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1131 {
1132 	return 0;
1133 }
1134 
1135 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1136 {
1137 	return 1;
1138 }
1139 
1140 #endif
1141 
1142 /* IPVS core functions
1143  * (from ip_vs_core.c)
1144  */
1145 const char *ip_vs_proto_name(unsigned int proto);
1146 void ip_vs_init_hash_table(struct list_head *table, int rows);
1147 #define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t)))
1148 
1149 #define IP_VS_APP_TYPE_FTP	1
1150 
1151 /* ip_vs_conn handling functions
1152  * (from ip_vs_conn.c)
1153  */
1154 enum {
1155 	IP_VS_DIR_INPUT = 0,
1156 	IP_VS_DIR_OUTPUT,
1157 	IP_VS_DIR_INPUT_ONLY,
1158 	IP_VS_DIR_LAST,
1159 };
1160 
1161 static inline void ip_vs_conn_fill_param(struct net *net, int af, int protocol,
1162 					 const union nf_inet_addr *caddr,
1163 					 __be16 cport,
1164 					 const union nf_inet_addr *vaddr,
1165 					 __be16 vport,
1166 					 struct ip_vs_conn_param *p)
1167 {
1168 	p->net = net;
1169 	p->af = af;
1170 	p->protocol = protocol;
1171 	p->caddr = caddr;
1172 	p->cport = cport;
1173 	p->vaddr = vaddr;
1174 	p->vport = vport;
1175 	p->pe = NULL;
1176 	p->pe_data = NULL;
1177 }
1178 
1179 struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p);
1180 struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p);
1181 
1182 struct ip_vs_conn * ip_vs_conn_in_get_proto(int af, const struct sk_buff *skb,
1183 					    const struct ip_vs_iphdr *iph,
1184 					    int inverse);
1185 
1186 struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p);
1187 
1188 struct ip_vs_conn * ip_vs_conn_out_get_proto(int af, const struct sk_buff *skb,
1189 					     const struct ip_vs_iphdr *iph,
1190 					     int inverse);
1191 
1192 /* Get reference to gain full access to conn.
1193  * By default, RCU read-side critical sections have access only to
1194  * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference.
1195  */
1196 static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp)
1197 {
1198 	return atomic_inc_not_zero(&cp->refcnt);
1199 }
1200 
1201 /* put back the conn without restarting its timer */
1202 static inline void __ip_vs_conn_put(struct ip_vs_conn *cp)
1203 {
1204 	smp_mb__before_atomic();
1205 	atomic_dec(&cp->refcnt);
1206 }
1207 void ip_vs_conn_put(struct ip_vs_conn *cp);
1208 void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport);
1209 
1210 struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, int dest_af,
1211 				  const union nf_inet_addr *daddr,
1212 				  __be16 dport, unsigned int flags,
1213 				  struct ip_vs_dest *dest, __u32 fwmark);
1214 void ip_vs_conn_expire_now(struct ip_vs_conn *cp);
1215 
1216 const char *ip_vs_state_name(__u16 proto, int state);
1217 
1218 void ip_vs_tcp_conn_listen(struct net *net, struct ip_vs_conn *cp);
1219 int ip_vs_check_template(struct ip_vs_conn *ct);
1220 void ip_vs_random_dropentry(struct net *net);
1221 int ip_vs_conn_init(void);
1222 void ip_vs_conn_cleanup(void);
1223 
1224 static inline void ip_vs_control_del(struct ip_vs_conn *cp)
1225 {
1226 	struct ip_vs_conn *ctl_cp = cp->control;
1227 	if (!ctl_cp) {
1228 		IP_VS_ERR_BUF("request control DEL for uncontrolled: "
1229 			      "%s:%d to %s:%d\n",
1230 			      IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1231 			      ntohs(cp->cport),
1232 			      IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1233 			      ntohs(cp->vport));
1234 
1235 		return;
1236 	}
1237 
1238 	IP_VS_DBG_BUF(7, "DELeting control for: "
1239 		      "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1240 		      IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1241 		      ntohs(cp->cport),
1242 		      IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1243 		      ntohs(ctl_cp->cport));
1244 
1245 	cp->control = NULL;
1246 	if (atomic_read(&ctl_cp->n_control) == 0) {
1247 		IP_VS_ERR_BUF("BUG control DEL with n=0 : "
1248 			      "%s:%d to %s:%d\n",
1249 			      IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1250 			      ntohs(cp->cport),
1251 			      IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1252 			      ntohs(cp->vport));
1253 
1254 		return;
1255 	}
1256 	atomic_dec(&ctl_cp->n_control);
1257 }
1258 
1259 static inline void
1260 ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp)
1261 {
1262 	if (cp->control) {
1263 		IP_VS_ERR_BUF("request control ADD for already controlled: "
1264 			      "%s:%d to %s:%d\n",
1265 			      IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1266 			      ntohs(cp->cport),
1267 			      IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1268 			      ntohs(cp->vport));
1269 
1270 		ip_vs_control_del(cp);
1271 	}
1272 
1273 	IP_VS_DBG_BUF(7, "ADDing control for: "
1274 		      "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1275 		      IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1276 		      ntohs(cp->cport),
1277 		      IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1278 		      ntohs(ctl_cp->cport));
1279 
1280 	cp->control = ctl_cp;
1281 	atomic_inc(&ctl_cp->n_control);
1282 }
1283 
1284 /* IPVS netns init & cleanup functions */
1285 int ip_vs_estimator_net_init(struct net *net);
1286 int ip_vs_control_net_init(struct net *net);
1287 int ip_vs_protocol_net_init(struct net *net);
1288 int ip_vs_app_net_init(struct net *net);
1289 int ip_vs_conn_net_init(struct net *net);
1290 int ip_vs_sync_net_init(struct net *net);
1291 void ip_vs_conn_net_cleanup(struct net *net);
1292 void ip_vs_app_net_cleanup(struct net *net);
1293 void ip_vs_protocol_net_cleanup(struct net *net);
1294 void ip_vs_control_net_cleanup(struct net *net);
1295 void ip_vs_estimator_net_cleanup(struct net *net);
1296 void ip_vs_sync_net_cleanup(struct net *net);
1297 void ip_vs_service_net_cleanup(struct net *net);
1298 
1299 /* IPVS application functions
1300  * (from ip_vs_app.c)
1301  */
1302 #define IP_VS_APP_MAX_PORTS  8
1303 struct ip_vs_app *register_ip_vs_app(struct net *net, struct ip_vs_app *app);
1304 void unregister_ip_vs_app(struct net *net, struct ip_vs_app *app);
1305 int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1306 void ip_vs_unbind_app(struct ip_vs_conn *cp);
1307 int register_ip_vs_app_inc(struct net *net, struct ip_vs_app *app, __u16 proto,
1308 			   __u16 port);
1309 int ip_vs_app_inc_get(struct ip_vs_app *inc);
1310 void ip_vs_app_inc_put(struct ip_vs_app *inc);
1311 
1312 int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb);
1313 int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb);
1314 
1315 int register_ip_vs_pe(struct ip_vs_pe *pe);
1316 int unregister_ip_vs_pe(struct ip_vs_pe *pe);
1317 struct ip_vs_pe *ip_vs_pe_getbyname(const char *name);
1318 struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name);
1319 
1320 /* Use a #define to avoid all of module.h just for these trivial ops */
1321 #define ip_vs_pe_get(pe)			\
1322 	if (pe && pe->module)			\
1323 		__module_get(pe->module);
1324 
1325 #define ip_vs_pe_put(pe)			\
1326 	if (pe && pe->module)			\
1327 		module_put(pe->module);
1328 
1329 /* IPVS protocol functions (from ip_vs_proto.c) */
1330 int ip_vs_protocol_init(void);
1331 void ip_vs_protocol_cleanup(void);
1332 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags);
1333 int *ip_vs_create_timeout_table(int *table, int size);
1334 int ip_vs_set_state_timeout(int *table, int num, const char *const *names,
1335 			    const char *name, int to);
1336 void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
1337 			       const struct sk_buff *skb, int offset,
1338 			       const char *msg);
1339 
1340 extern struct ip_vs_protocol ip_vs_protocol_tcp;
1341 extern struct ip_vs_protocol ip_vs_protocol_udp;
1342 extern struct ip_vs_protocol ip_vs_protocol_icmp;
1343 extern struct ip_vs_protocol ip_vs_protocol_esp;
1344 extern struct ip_vs_protocol ip_vs_protocol_ah;
1345 extern struct ip_vs_protocol ip_vs_protocol_sctp;
1346 
1347 /* Registering/unregistering scheduler functions
1348  * (from ip_vs_sched.c)
1349  */
1350 int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1351 int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1352 int ip_vs_bind_scheduler(struct ip_vs_service *svc,
1353 			 struct ip_vs_scheduler *scheduler);
1354 void ip_vs_unbind_scheduler(struct ip_vs_service *svc,
1355 			    struct ip_vs_scheduler *sched);
1356 struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name);
1357 void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler);
1358 struct ip_vs_conn *
1359 ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb,
1360 	       struct ip_vs_proto_data *pd, int *ignored,
1361 	       struct ip_vs_iphdr *iph);
1362 int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb,
1363 		struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph);
1364 
1365 void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg);
1366 
1367 /* IPVS control data and functions (from ip_vs_ctl.c) */
1368 extern struct ip_vs_stats ip_vs_stats;
1369 extern int sysctl_ip_vs_sync_ver;
1370 
1371 struct ip_vs_service *
1372 ip_vs_service_find(struct net *net, int af, __u32 fwmark, __u16 protocol,
1373 		  const union nf_inet_addr *vaddr, __be16 vport);
1374 
1375 bool ip_vs_has_real_service(struct net *net, int af, __u16 protocol,
1376 			    const union nf_inet_addr *daddr, __be16 dport);
1377 
1378 int ip_vs_use_count_inc(void);
1379 void ip_vs_use_count_dec(void);
1380 int ip_vs_register_nl_ioctl(void);
1381 void ip_vs_unregister_nl_ioctl(void);
1382 int ip_vs_control_init(void);
1383 void ip_vs_control_cleanup(void);
1384 struct ip_vs_dest *
1385 ip_vs_find_dest(struct net *net, int svc_af, int dest_af,
1386 		const union nf_inet_addr *daddr, __be16 dport,
1387 		const union nf_inet_addr *vaddr, __be16 vport,
1388 		__u16 protocol, __u32 fwmark, __u32 flags);
1389 void ip_vs_try_bind_dest(struct ip_vs_conn *cp);
1390 
1391 static inline void ip_vs_dest_hold(struct ip_vs_dest *dest)
1392 {
1393 	atomic_inc(&dest->refcnt);
1394 }
1395 
1396 static inline void ip_vs_dest_put(struct ip_vs_dest *dest)
1397 {
1398 	smp_mb__before_atomic();
1399 	atomic_dec(&dest->refcnt);
1400 }
1401 
1402 static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest)
1403 {
1404 	if (atomic_dec_return(&dest->refcnt) < 0)
1405 		kfree(dest);
1406 }
1407 
1408 /* IPVS sync daemon data and function prototypes
1409  * (from ip_vs_sync.c)
1410  */
1411 int start_sync_thread(struct net *net, int state, char *mcast_ifn, __u8 syncid);
1412 int stop_sync_thread(struct net *net, int state);
1413 void ip_vs_sync_conn(struct net *net, struct ip_vs_conn *cp, int pkts);
1414 
1415 /* IPVS rate estimator prototypes (from ip_vs_est.c) */
1416 void ip_vs_start_estimator(struct net *net, struct ip_vs_stats *stats);
1417 void ip_vs_stop_estimator(struct net *net, struct ip_vs_stats *stats);
1418 void ip_vs_zero_estimator(struct ip_vs_stats *stats);
1419 void ip_vs_read_estimator(struct ip_vs_kstats *dst, struct ip_vs_stats *stats);
1420 
1421 /* Various IPVS packet transmitters (from ip_vs_xmit.c) */
1422 int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1423 		    struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1424 int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1425 		      struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1426 int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1427 		   struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1428 int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1429 		      struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1430 int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1431 		  struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1432 int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1433 		    struct ip_vs_protocol *pp, int offset,
1434 		    unsigned int hooknum, struct ip_vs_iphdr *iph);
1435 void ip_vs_dest_dst_rcu_free(struct rcu_head *head);
1436 
1437 #ifdef CONFIG_IP_VS_IPV6
1438 int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1439 			 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1440 int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1441 		      struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1442 int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1443 			 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1444 int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1445 		     struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1446 int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1447 		       struct ip_vs_protocol *pp, int offset,
1448 		       unsigned int hooknum, struct ip_vs_iphdr *iph);
1449 #endif
1450 
1451 #ifdef CONFIG_SYSCTL
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 static inline int ip_vs_todrop(struct netns_ipvs *ipvs)
1457 {
1458 	if (!ipvs->drop_rate)
1459 		return 0;
1460 	if (--ipvs->drop_counter > 0)
1461 		return 0;
1462 	ipvs->drop_counter = ipvs->drop_rate;
1463 	return 1;
1464 }
1465 #else
1466 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; }
1467 #endif
1468 
1469 /* ip_vs_fwd_tag returns the forwarding tag of the connection */
1470 #define IP_VS_FWD_METHOD(cp)  (cp->flags & IP_VS_CONN_F_FWD_MASK)
1471 
1472 static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp)
1473 {
1474 	char fwd;
1475 
1476 	switch (IP_VS_FWD_METHOD(cp)) {
1477 	case IP_VS_CONN_F_MASQ:
1478 		fwd = 'M'; break;
1479 	case IP_VS_CONN_F_LOCALNODE:
1480 		fwd = 'L'; break;
1481 	case IP_VS_CONN_F_TUNNEL:
1482 		fwd = 'T'; break;
1483 	case IP_VS_CONN_F_DROUTE:
1484 		fwd = 'R'; break;
1485 	case IP_VS_CONN_F_BYPASS:
1486 		fwd = 'B'; break;
1487 	default:
1488 		fwd = '?'; break;
1489 	}
1490 	return fwd;
1491 }
1492 
1493 void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp,
1494 		    struct ip_vs_conn *cp, int dir);
1495 
1496 #ifdef CONFIG_IP_VS_IPV6
1497 void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp,
1498 		       struct ip_vs_conn *cp, int dir);
1499 #endif
1500 
1501 __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset);
1502 
1503 static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum)
1504 {
1505 	__be32 diff[2] = { ~old, new };
1506 
1507 	return csum_partial(diff, sizeof(diff), oldsum);
1508 }
1509 
1510 #ifdef CONFIG_IP_VS_IPV6
1511 static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new,
1512 					__wsum oldsum)
1513 {
1514 	__be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0],
1515 			    new[3],  new[2],  new[1],  new[0] };
1516 
1517 	return csum_partial(diff, sizeof(diff), oldsum);
1518 }
1519 #endif
1520 
1521 static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum)
1522 {
1523 	__be16 diff[2] = { ~old, new };
1524 
1525 	return csum_partial(diff, sizeof(diff), oldsum);
1526 }
1527 
1528 /* Forget current conntrack (unconfirmed) and attach notrack entry */
1529 static inline void ip_vs_notrack(struct sk_buff *skb)
1530 {
1531 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1532 	enum ip_conntrack_info ctinfo;
1533 	struct nf_conn *ct = nf_ct_get(skb, &ctinfo);
1534 
1535 	if (!ct || !nf_ct_is_untracked(ct)) {
1536 		nf_conntrack_put(skb->nfct);
1537 		skb->nfct = &nf_ct_untracked_get()->ct_general;
1538 		skb->nfctinfo = IP_CT_NEW;
1539 		nf_conntrack_get(skb->nfct);
1540 	}
1541 #endif
1542 }
1543 
1544 #ifdef CONFIG_IP_VS_NFCT
1545 /* Netfilter connection tracking
1546  * (from ip_vs_nfct.c)
1547  */
1548 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1549 {
1550 #ifdef CONFIG_SYSCTL
1551 	return ipvs->sysctl_conntrack;
1552 #else
1553 	return 0;
1554 #endif
1555 }
1556 
1557 void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp,
1558 			    int outin);
1559 int ip_vs_confirm_conntrack(struct sk_buff *skb);
1560 void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct,
1561 			       struct ip_vs_conn *cp, u_int8_t proto,
1562 			       const __be16 port, int from_rs);
1563 void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp);
1564 
1565 #else
1566 
1567 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1568 {
1569 	return 0;
1570 }
1571 
1572 static inline void ip_vs_update_conntrack(struct sk_buff *skb,
1573 					  struct ip_vs_conn *cp, int outin)
1574 {
1575 }
1576 
1577 static inline int ip_vs_confirm_conntrack(struct sk_buff *skb)
1578 {
1579 	return NF_ACCEPT;
1580 }
1581 
1582 static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp)
1583 {
1584 }
1585 #endif /* CONFIG_IP_VS_NFCT */
1586 
1587 static inline int
1588 ip_vs_dest_conn_overhead(struct ip_vs_dest *dest)
1589 {
1590 	/* We think the overhead of processing active connections is 256
1591 	 * times higher than that of inactive connections in average. (This
1592 	 * 256 times might not be accurate, we will change it later) We
1593 	 * use the following formula to estimate the overhead now:
1594 	 *		  dest->activeconns*256 + dest->inactconns
1595 	 */
1596 	return (atomic_read(&dest->activeconns) << 8) +
1597 		atomic_read(&dest->inactconns);
1598 }
1599 
1600 #endif	/* _NET_IP_VS_H */
1601