xref: /openbmc/linux/include/net/ip_vs.h (revision a06c488d)
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 #define IP_VS_HDR_INVERSE	1
33 #define IP_VS_HDR_ICMP		2
34 
35 /* Generic access of ipvs struct */
36 static inline struct netns_ipvs *net_ipvs(struct net* net)
37 {
38 	return net->ipvs;
39 }
40 
41 /* This one needed for single_open_net since net is stored directly in
42  * private not as a struct i.e. seq_file_net can't be used.
43  */
44 static inline struct net *seq_file_single_net(struct seq_file *seq)
45 {
46 #ifdef CONFIG_NET_NS
47 	return (struct net *)seq->private;
48 #else
49 	return &init_net;
50 #endif
51 }
52 
53 /* Connections' size value needed by ip_vs_ctl.c */
54 extern int ip_vs_conn_tab_size;
55 
56 struct ip_vs_iphdr {
57 	int hdr_flags;	/* ipvs flags */
58 	__u32 off;	/* Where IP or IPv4 header starts */
59 	__u32 len;	/* IPv4 simply where L4 starts
60 			 * IPv6 where L4 Transport Header starts */
61 	__u16 fragoffs; /* IPv6 fragment offset, 0 if first frag (or not frag)*/
62 	__s16 protocol;
63 	__s32 flags;
64 	union nf_inet_addr saddr;
65 	union nf_inet_addr daddr;
66 };
67 
68 static inline void *frag_safe_skb_hp(const struct sk_buff *skb, int offset,
69 				      int len, void *buffer,
70 				      const struct ip_vs_iphdr *ipvsh)
71 {
72 	return skb_header_pointer(skb, offset, len, buffer);
73 }
74 
75 /* This function handles filling *ip_vs_iphdr, both for IPv4 and IPv6.
76  * IPv6 requires some extra work, as finding proper header position,
77  * depend on the IPv6 extension headers.
78  */
79 static inline int
80 ip_vs_fill_iph_skb_off(int af, const struct sk_buff *skb, int offset,
81 		       int hdr_flags, struct ip_vs_iphdr *iphdr)
82 {
83 	iphdr->hdr_flags = hdr_flags;
84 	iphdr->off = offset;
85 
86 #ifdef CONFIG_IP_VS_IPV6
87 	if (af == AF_INET6) {
88 		struct ipv6hdr _iph;
89 		const struct ipv6hdr *iph = skb_header_pointer(
90 			skb, offset, sizeof(_iph), &_iph);
91 		if (!iph)
92 			return 0;
93 
94 		iphdr->saddr.in6 = iph->saddr;
95 		iphdr->daddr.in6 = iph->daddr;
96 		/* ipv6_find_hdr() updates len, flags */
97 		iphdr->len	 = offset;
98 		iphdr->flags	 = 0;
99 		iphdr->protocol  = ipv6_find_hdr(skb, &iphdr->len, -1,
100 						 &iphdr->fragoffs,
101 						 &iphdr->flags);
102 		if (iphdr->protocol < 0)
103 			return 0;
104 	} else
105 #endif
106 	{
107 		struct iphdr _iph;
108 		const struct iphdr *iph = skb_header_pointer(
109 			skb, offset, sizeof(_iph), &_iph);
110 		if (!iph)
111 			return 0;
112 
113 		iphdr->len	= offset + iph->ihl * 4;
114 		iphdr->fragoffs	= 0;
115 		iphdr->protocol	= iph->protocol;
116 		iphdr->saddr.ip	= iph->saddr;
117 		iphdr->daddr.ip	= iph->daddr;
118 	}
119 
120 	return 1;
121 }
122 
123 static inline int
124 ip_vs_fill_iph_skb_icmp(int af, const struct sk_buff *skb, int offset,
125 			bool inverse, struct ip_vs_iphdr *iphdr)
126 {
127 	int hdr_flags = IP_VS_HDR_ICMP;
128 
129 	if (inverse)
130 		hdr_flags |= IP_VS_HDR_INVERSE;
131 
132 	return ip_vs_fill_iph_skb_off(af, skb, offset, hdr_flags, iphdr);
133 }
134 
135 static inline int
136 ip_vs_fill_iph_skb(int af, const struct sk_buff *skb, bool inverse,
137 		   struct ip_vs_iphdr *iphdr)
138 {
139 	int hdr_flags = 0;
140 
141 	if (inverse)
142 		hdr_flags |= IP_VS_HDR_INVERSE;
143 
144 	return ip_vs_fill_iph_skb_off(af, skb, skb_network_offset(skb),
145 				      hdr_flags, iphdr);
146 }
147 
148 static inline bool
149 ip_vs_iph_inverse(const struct ip_vs_iphdr *iph)
150 {
151 	return !!(iph->hdr_flags & IP_VS_HDR_INVERSE);
152 }
153 
154 static inline bool
155 ip_vs_iph_icmp(const struct ip_vs_iphdr *iph)
156 {
157 	return !!(iph->hdr_flags & IP_VS_HDR_ICMP);
158 }
159 
160 static inline void ip_vs_addr_copy(int af, union nf_inet_addr *dst,
161 				   const union nf_inet_addr *src)
162 {
163 #ifdef CONFIG_IP_VS_IPV6
164 	if (af == AF_INET6)
165 		dst->in6 = src->in6;
166 	else
167 #endif
168 	dst->ip = src->ip;
169 }
170 
171 static inline void ip_vs_addr_set(int af, union nf_inet_addr *dst,
172 				  const union nf_inet_addr *src)
173 {
174 #ifdef CONFIG_IP_VS_IPV6
175 	if (af == AF_INET6) {
176 		dst->in6 = src->in6;
177 		return;
178 	}
179 #endif
180 	dst->ip = src->ip;
181 	dst->all[1] = 0;
182 	dst->all[2] = 0;
183 	dst->all[3] = 0;
184 }
185 
186 static inline int ip_vs_addr_equal(int af, const union nf_inet_addr *a,
187 				   const union nf_inet_addr *b)
188 {
189 #ifdef CONFIG_IP_VS_IPV6
190 	if (af == AF_INET6)
191 		return ipv6_addr_equal(&a->in6, &b->in6);
192 #endif
193 	return a->ip == b->ip;
194 }
195 
196 #ifdef CONFIG_IP_VS_DEBUG
197 #include <linux/net.h>
198 
199 int ip_vs_get_debug_level(void);
200 
201 static inline const char *ip_vs_dbg_addr(int af, char *buf, size_t buf_len,
202 					 const union nf_inet_addr *addr,
203 					 int *idx)
204 {
205 	int len;
206 #ifdef CONFIG_IP_VS_IPV6
207 	if (af == AF_INET6)
208 		len = snprintf(&buf[*idx], buf_len - *idx, "[%pI6c]",
209 			       &addr->in6) + 1;
210 	else
211 #endif
212 		len = snprintf(&buf[*idx], buf_len - *idx, "%pI4",
213 			       &addr->ip) + 1;
214 
215 	*idx += len;
216 	BUG_ON(*idx > buf_len + 1);
217 	return &buf[*idx - len];
218 }
219 
220 #define IP_VS_DBG_BUF(level, msg, ...)					\
221 	do {								\
222 		char ip_vs_dbg_buf[160];				\
223 		int ip_vs_dbg_idx = 0;					\
224 		if (level <= ip_vs_get_debug_level())			\
225 			printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__);	\
226 	} while (0)
227 #define IP_VS_ERR_BUF(msg...)						\
228 	do {								\
229 		char ip_vs_dbg_buf[160];				\
230 		int ip_vs_dbg_idx = 0;					\
231 		pr_err(msg);						\
232 	} while (0)
233 
234 /* Only use from within IP_VS_DBG_BUF() or IP_VS_ERR_BUF macros */
235 #define IP_VS_DBG_ADDR(af, addr)					\
236 	ip_vs_dbg_addr(af, ip_vs_dbg_buf,				\
237 		       sizeof(ip_vs_dbg_buf), addr,			\
238 		       &ip_vs_dbg_idx)
239 
240 #define IP_VS_DBG(level, msg, ...)					\
241 	do {								\
242 		if (level <= ip_vs_get_debug_level())			\
243 			printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__);	\
244 	} while (0)
245 #define IP_VS_DBG_RL(msg, ...)						\
246 	do {								\
247 		if (net_ratelimit())					\
248 			printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__);	\
249 	} while (0)
250 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg)			\
251 	do {								\
252 		if (level <= ip_vs_get_debug_level())			\
253 			pp->debug_packet(af, pp, skb, ofs, msg);	\
254 	} while (0)
255 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg)			\
256 	do {								\
257 		if (level <= ip_vs_get_debug_level() &&			\
258 		    net_ratelimit())					\
259 			pp->debug_packet(af, pp, skb, ofs, msg);	\
260 	} while (0)
261 #else	/* NO DEBUGGING at ALL */
262 #define IP_VS_DBG_BUF(level, msg...)  do {} while (0)
263 #define IP_VS_ERR_BUF(msg...)  do {} while (0)
264 #define IP_VS_DBG(level, msg...)  do {} while (0)
265 #define IP_VS_DBG_RL(msg...)  do {} while (0)
266 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg)	do {} while (0)
267 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg)	do {} while (0)
268 #endif
269 
270 #define IP_VS_BUG() BUG()
271 #define IP_VS_ERR_RL(msg, ...)						\
272 	do {								\
273 		if (net_ratelimit())					\
274 			pr_err(msg, ##__VA_ARGS__);			\
275 	} while (0)
276 
277 #ifdef CONFIG_IP_VS_DEBUG
278 #define EnterFunction(level)						\
279 	do {								\
280 		if (level <= ip_vs_get_debug_level())			\
281 			printk(KERN_DEBUG				\
282 			       pr_fmt("Enter: %s, %s line %i\n"),	\
283 			       __func__, __FILE__, __LINE__);		\
284 	} while (0)
285 #define LeaveFunction(level)						\
286 	do {								\
287 		if (level <= ip_vs_get_debug_level())			\
288 			printk(KERN_DEBUG				\
289 			       pr_fmt("Leave: %s, %s line %i\n"),	\
290 			       __func__, __FILE__, __LINE__);		\
291 	} while (0)
292 #else
293 #define EnterFunction(level)   do {} while (0)
294 #define LeaveFunction(level)   do {} while (0)
295 #endif
296 
297 /* The port number of FTP service (in network order). */
298 #define FTPPORT  cpu_to_be16(21)
299 #define FTPDATA  cpu_to_be16(20)
300 
301 /* TCP State Values */
302 enum {
303 	IP_VS_TCP_S_NONE = 0,
304 	IP_VS_TCP_S_ESTABLISHED,
305 	IP_VS_TCP_S_SYN_SENT,
306 	IP_VS_TCP_S_SYN_RECV,
307 	IP_VS_TCP_S_FIN_WAIT,
308 	IP_VS_TCP_S_TIME_WAIT,
309 	IP_VS_TCP_S_CLOSE,
310 	IP_VS_TCP_S_CLOSE_WAIT,
311 	IP_VS_TCP_S_LAST_ACK,
312 	IP_VS_TCP_S_LISTEN,
313 	IP_VS_TCP_S_SYNACK,
314 	IP_VS_TCP_S_LAST
315 };
316 
317 /* UDP State Values */
318 enum {
319 	IP_VS_UDP_S_NORMAL,
320 	IP_VS_UDP_S_LAST,
321 };
322 
323 /* ICMP State Values */
324 enum {
325 	IP_VS_ICMP_S_NORMAL,
326 	IP_VS_ICMP_S_LAST,
327 };
328 
329 /* SCTP State Values */
330 enum ip_vs_sctp_states {
331 	IP_VS_SCTP_S_NONE,
332 	IP_VS_SCTP_S_INIT1,
333 	IP_VS_SCTP_S_INIT,
334 	IP_VS_SCTP_S_COOKIE_SENT,
335 	IP_VS_SCTP_S_COOKIE_REPLIED,
336 	IP_VS_SCTP_S_COOKIE_WAIT,
337 	IP_VS_SCTP_S_COOKIE,
338 	IP_VS_SCTP_S_COOKIE_ECHOED,
339 	IP_VS_SCTP_S_ESTABLISHED,
340 	IP_VS_SCTP_S_SHUTDOWN_SENT,
341 	IP_VS_SCTP_S_SHUTDOWN_RECEIVED,
342 	IP_VS_SCTP_S_SHUTDOWN_ACK_SENT,
343 	IP_VS_SCTP_S_REJECTED,
344 	IP_VS_SCTP_S_CLOSED,
345 	IP_VS_SCTP_S_LAST
346 };
347 
348 /* Delta sequence info structure
349  * Each ip_vs_conn has 2 (output AND input seq. changes).
350  * Only used in the VS/NAT.
351  */
352 struct ip_vs_seq {
353 	__u32			init_seq;	/* Add delta from this seq */
354 	__u32			delta;		/* Delta in sequence numbers */
355 	__u32			previous_delta;	/* Delta in sequence numbers
356 						 * before last resized pkt */
357 };
358 
359 /* counters per cpu */
360 struct ip_vs_counters {
361 	__u64		conns;		/* connections scheduled */
362 	__u64		inpkts;		/* incoming packets */
363 	__u64		outpkts;	/* outgoing packets */
364 	__u64		inbytes;	/* incoming bytes */
365 	__u64		outbytes;	/* outgoing bytes */
366 };
367 /* Stats per cpu */
368 struct ip_vs_cpu_stats {
369 	struct ip_vs_counters   cnt;
370 	struct u64_stats_sync   syncp;
371 };
372 
373 /* IPVS statistics objects */
374 struct ip_vs_estimator {
375 	struct list_head	list;
376 
377 	u64			last_inbytes;
378 	u64			last_outbytes;
379 	u64			last_conns;
380 	u64			last_inpkts;
381 	u64			last_outpkts;
382 
383 	u64			cps;
384 	u64			inpps;
385 	u64			outpps;
386 	u64			inbps;
387 	u64			outbps;
388 };
389 
390 /*
391  * IPVS statistics object, 64-bit kernel version of struct ip_vs_stats_user
392  */
393 struct ip_vs_kstats {
394 	u64			conns;		/* connections scheduled */
395 	u64			inpkts;		/* incoming packets */
396 	u64			outpkts;	/* outgoing packets */
397 	u64			inbytes;	/* incoming bytes */
398 	u64			outbytes;	/* outgoing bytes */
399 
400 	u64			cps;		/* current connection rate */
401 	u64			inpps;		/* current in packet rate */
402 	u64			outpps;		/* current out packet rate */
403 	u64			inbps;		/* current in byte rate */
404 	u64			outbps;		/* current out byte rate */
405 };
406 
407 struct ip_vs_stats {
408 	struct ip_vs_kstats	kstats;		/* kernel statistics */
409 	struct ip_vs_estimator	est;		/* estimator */
410 	struct ip_vs_cpu_stats __percpu	*cpustats;	/* per cpu counters */
411 	spinlock_t		lock;		/* spin lock */
412 	struct ip_vs_kstats	kstats0;	/* reset values */
413 };
414 
415 struct dst_entry;
416 struct iphdr;
417 struct ip_vs_conn;
418 struct ip_vs_app;
419 struct sk_buff;
420 struct ip_vs_proto_data;
421 
422 struct ip_vs_protocol {
423 	struct ip_vs_protocol	*next;
424 	char			*name;
425 	u16			protocol;
426 	u16			num_states;
427 	int			dont_defrag;
428 
429 	void (*init)(struct ip_vs_protocol *pp);
430 
431 	void (*exit)(struct ip_vs_protocol *pp);
432 
433 	int (*init_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
434 
435 	void (*exit_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
436 
437 	int (*conn_schedule)(struct netns_ipvs *ipvs,
438 			     int af, struct sk_buff *skb,
439 			     struct ip_vs_proto_data *pd,
440 			     int *verdict, struct ip_vs_conn **cpp,
441 			     struct ip_vs_iphdr *iph);
442 
443 	struct ip_vs_conn *
444 	(*conn_in_get)(struct netns_ipvs *ipvs,
445 		       int af,
446 		       const struct sk_buff *skb,
447 		       const struct ip_vs_iphdr *iph);
448 
449 	struct ip_vs_conn *
450 	(*conn_out_get)(struct netns_ipvs *ipvs,
451 			int af,
452 			const struct sk_buff *skb,
453 			const struct ip_vs_iphdr *iph);
454 
455 	int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
456 			    struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
457 
458 	int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
459 			    struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
460 
461 	int (*csum_check)(int af, struct sk_buff *skb,
462 			  struct ip_vs_protocol *pp);
463 
464 	const char *(*state_name)(int state);
465 
466 	void (*state_transition)(struct ip_vs_conn *cp, int direction,
467 				 const struct sk_buff *skb,
468 				 struct ip_vs_proto_data *pd);
469 
470 	int (*register_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
471 
472 	void (*unregister_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
473 
474 	int (*app_conn_bind)(struct ip_vs_conn *cp);
475 
476 	void (*debug_packet)(int af, struct ip_vs_protocol *pp,
477 			     const struct sk_buff *skb,
478 			     int offset,
479 			     const char *msg);
480 
481 	void (*timeout_change)(struct ip_vs_proto_data *pd, int flags);
482 };
483 
484 /* protocol data per netns */
485 struct ip_vs_proto_data {
486 	struct ip_vs_proto_data	*next;
487 	struct ip_vs_protocol	*pp;
488 	int			*timeout_table;	/* protocol timeout table */
489 	atomic_t		appcnt;		/* counter of proto app incs. */
490 	struct tcp_states_t	*tcp_state_table;
491 };
492 
493 struct ip_vs_protocol   *ip_vs_proto_get(unsigned short proto);
494 struct ip_vs_proto_data *ip_vs_proto_data_get(struct netns_ipvs *ipvs,
495 					      unsigned short proto);
496 
497 struct ip_vs_conn_param {
498 	struct netns_ipvs		*ipvs;
499 	const union nf_inet_addr	*caddr;
500 	const union nf_inet_addr	*vaddr;
501 	__be16				cport;
502 	__be16				vport;
503 	__u16				protocol;
504 	u16				af;
505 
506 	const struct ip_vs_pe		*pe;
507 	char				*pe_data;
508 	__u8				pe_data_len;
509 };
510 
511 /* IP_VS structure allocated for each dynamically scheduled connection */
512 struct ip_vs_conn {
513 	struct hlist_node	c_list;         /* hashed list heads */
514 	/* Protocol, addresses and port numbers */
515 	__be16                  cport;
516 	__be16                  dport;
517 	__be16                  vport;
518 	u16			af;		/* address family */
519 	union nf_inet_addr      caddr;          /* client address */
520 	union nf_inet_addr      vaddr;          /* virtual address */
521 	union nf_inet_addr      daddr;          /* destination address */
522 	volatile __u32          flags;          /* status flags */
523 	__u16                   protocol;       /* Which protocol (TCP/UDP) */
524 	__u16			daf;		/* Address family of the dest */
525 	struct netns_ipvs	*ipvs;
526 
527 	/* counter and timer */
528 	atomic_t		refcnt;		/* reference count */
529 	struct timer_list	timer;		/* Expiration timer */
530 	volatile unsigned long	timeout;	/* timeout */
531 
532 	/* Flags and state transition */
533 	spinlock_t              lock;           /* lock for state transition */
534 	volatile __u16          state;          /* state info */
535 	volatile __u16          old_state;      /* old state, to be used for
536 						 * state transition triggerd
537 						 * synchronization
538 						 */
539 	__u32			fwmark;		/* Fire wall mark from skb */
540 	unsigned long		sync_endtime;	/* jiffies + sent_retries */
541 
542 	/* Control members */
543 	struct ip_vs_conn       *control;       /* Master control connection */
544 	atomic_t                n_control;      /* Number of controlled ones */
545 	struct ip_vs_dest       *dest;          /* real server */
546 	atomic_t                in_pkts;        /* incoming packet counter */
547 
548 	/* Packet transmitter for different forwarding methods.  If it
549 	 * mangles the packet, it must return NF_DROP or better NF_STOLEN,
550 	 * otherwise this must be changed to a sk_buff **.
551 	 * NF_ACCEPT can be returned when destination is local.
552 	 */
553 	int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp,
554 			   struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
555 
556 	/* Note: we can group the following members into a structure,
557 	 * in order to save more space, and the following members are
558 	 * only used in VS/NAT anyway
559 	 */
560 	struct ip_vs_app        *app;           /* bound ip_vs_app object */
561 	void                    *app_data;      /* Application private data */
562 	struct ip_vs_seq        in_seq;         /* incoming seq. struct */
563 	struct ip_vs_seq        out_seq;        /* outgoing seq. struct */
564 
565 	const struct ip_vs_pe	*pe;
566 	char			*pe_data;
567 	__u8			pe_data_len;
568 
569 	struct rcu_head		rcu_head;
570 };
571 
572 /* Extended internal versions of struct ip_vs_service_user and ip_vs_dest_user
573  * for IPv6 support.
574  *
575  * We need these to conveniently pass around service and destination
576  * options, but unfortunately, we also need to keep the old definitions to
577  * maintain userspace backwards compatibility for the setsockopt interface.
578  */
579 struct ip_vs_service_user_kern {
580 	/* virtual service addresses */
581 	u16			af;
582 	u16			protocol;
583 	union nf_inet_addr	addr;		/* virtual ip address */
584 	__be16			port;
585 	u32			fwmark;		/* firwall mark of service */
586 
587 	/* virtual service options */
588 	char			*sched_name;
589 	char			*pe_name;
590 	unsigned int		flags;		/* virtual service flags */
591 	unsigned int		timeout;	/* persistent timeout in sec */
592 	__be32			netmask;	/* persistent netmask or plen */
593 };
594 
595 
596 struct ip_vs_dest_user_kern {
597 	/* destination server address */
598 	union nf_inet_addr	addr;
599 	__be16			port;
600 
601 	/* real server options */
602 	unsigned int		conn_flags;	/* connection flags */
603 	int			weight;		/* destination weight */
604 
605 	/* thresholds for active connections */
606 	u32			u_threshold;	/* upper threshold */
607 	u32			l_threshold;	/* lower threshold */
608 
609 	/* Address family of addr */
610 	u16			af;
611 };
612 
613 
614 /*
615  * The information about the virtual service offered to the net and the
616  * forwarding entries.
617  */
618 struct ip_vs_service {
619 	struct hlist_node	s_list;   /* for normal service table */
620 	struct hlist_node	f_list;   /* for fwmark-based service table */
621 	atomic_t		refcnt;   /* reference counter */
622 
623 	u16			af;       /* address family */
624 	__u16			protocol; /* which protocol (TCP/UDP) */
625 	union nf_inet_addr	addr;	  /* IP address for virtual service */
626 	__be16			port;	  /* port number for the service */
627 	__u32                   fwmark;   /* firewall mark of the service */
628 	unsigned int		flags;	  /* service status flags */
629 	unsigned int		timeout;  /* persistent timeout in ticks */
630 	__be32			netmask;  /* grouping granularity, mask/plen */
631 	struct netns_ipvs	*ipvs;
632 
633 	struct list_head	destinations;  /* real server d-linked list */
634 	__u32			num_dests;     /* number of servers */
635 	struct ip_vs_stats      stats;         /* statistics for the service */
636 
637 	/* for scheduling */
638 	struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */
639 	spinlock_t		sched_lock;    /* lock sched_data */
640 	void			*sched_data;   /* scheduler application data */
641 
642 	/* alternate persistence engine */
643 	struct ip_vs_pe __rcu	*pe;
644 
645 	struct rcu_head		rcu_head;
646 };
647 
648 /* Information for cached dst */
649 struct ip_vs_dest_dst {
650 	struct dst_entry	*dst_cache;	/* destination cache entry */
651 	u32			dst_cookie;
652 	union nf_inet_addr	dst_saddr;
653 	struct rcu_head		rcu_head;
654 };
655 
656 /* The real server destination forwarding entry with ip address, port number,
657  * and so on.
658  */
659 struct ip_vs_dest {
660 	struct list_head	n_list;   /* for the dests in the service */
661 	struct hlist_node	d_list;   /* for table with all the dests */
662 
663 	u16			af;		/* address family */
664 	__be16			port;		/* port number of the server */
665 	union nf_inet_addr	addr;		/* IP address of the server */
666 	volatile unsigned int	flags;		/* dest status flags */
667 	atomic_t		conn_flags;	/* flags to copy to conn */
668 	atomic_t		weight;		/* server weight */
669 
670 	atomic_t		refcnt;		/* reference counter */
671 	struct ip_vs_stats      stats;          /* statistics */
672 	unsigned long		idle_start;	/* start time, jiffies */
673 
674 	/* connection counters and thresholds */
675 	atomic_t		activeconns;	/* active connections */
676 	atomic_t		inactconns;	/* inactive connections */
677 	atomic_t		persistconns;	/* persistent connections */
678 	__u32			u_threshold;	/* upper threshold */
679 	__u32			l_threshold;	/* lower threshold */
680 
681 	/* for destination cache */
682 	spinlock_t		dst_lock;	/* lock of dst_cache */
683 	struct ip_vs_dest_dst __rcu *dest_dst;	/* cached dst info */
684 
685 	/* for virtual service */
686 	struct ip_vs_service __rcu *svc;	/* service it belongs to */
687 	__u16			protocol;	/* which protocol (TCP/UDP) */
688 	__be16			vport;		/* virtual port number */
689 	union nf_inet_addr	vaddr;		/* virtual IP address */
690 	__u32			vfwmark;	/* firewall mark of service */
691 
692 	struct list_head	t_list;		/* in dest_trash */
693 	unsigned int		in_rs_table:1;	/* we are in rs_table */
694 };
695 
696 /* The scheduler object */
697 struct ip_vs_scheduler {
698 	struct list_head	n_list;		/* d-linked list head */
699 	char			*name;		/* scheduler name */
700 	atomic_t		refcnt;		/* reference counter */
701 	struct module		*module;	/* THIS_MODULE/NULL */
702 
703 	/* scheduler initializing service */
704 	int (*init_service)(struct ip_vs_service *svc);
705 	/* scheduling service finish */
706 	void (*done_service)(struct ip_vs_service *svc);
707 	/* dest is linked */
708 	int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
709 	/* dest is unlinked */
710 	int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
711 	/* dest is updated */
712 	int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
713 
714 	/* selecting a server from the given service */
715 	struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc,
716 				       const struct sk_buff *skb,
717 				       struct ip_vs_iphdr *iph);
718 };
719 
720 /* The persistence engine object */
721 struct ip_vs_pe {
722 	struct list_head	n_list;		/* d-linked list head */
723 	char			*name;		/* scheduler name */
724 	atomic_t		refcnt;		/* reference counter */
725 	struct module		*module;	/* THIS_MODULE/NULL */
726 
727 	/* get the connection template, if any */
728 	int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb);
729 	bool (*ct_match)(const struct ip_vs_conn_param *p,
730 			 struct ip_vs_conn *ct);
731 	u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval,
732 			   bool inverse);
733 	int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf);
734 };
735 
736 /* The application module object (a.k.a. app incarnation) */
737 struct ip_vs_app {
738 	struct list_head	a_list;		/* member in app list */
739 	int			type;		/* IP_VS_APP_TYPE_xxx */
740 	char			*name;		/* application module name */
741 	__u16			protocol;
742 	struct module		*module;	/* THIS_MODULE/NULL */
743 	struct list_head	incs_list;	/* list of incarnations */
744 
745 	/* members for application incarnations */
746 	struct list_head	p_list;		/* member in proto app list */
747 	struct ip_vs_app	*app;		/* its real application */
748 	__be16			port;		/* port number in net order */
749 	atomic_t		usecnt;		/* usage counter */
750 	struct rcu_head		rcu_head;
751 
752 	/* output hook: Process packet in inout direction, diff set for TCP.
753 	 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
754 	 *	   2=Mangled but checksum was not updated
755 	 */
756 	int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *,
757 		       struct sk_buff *, int *diff);
758 
759 	/* input hook: Process packet in outin direction, diff set for TCP.
760 	 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
761 	 *	   2=Mangled but checksum was not updated
762 	 */
763 	int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *,
764 		      struct sk_buff *, int *diff);
765 
766 	/* ip_vs_app initializer */
767 	int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *);
768 
769 	/* ip_vs_app finish */
770 	int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *);
771 
772 
773 	/* not used now */
774 	int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *,
775 			 struct ip_vs_protocol *);
776 
777 	void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *);
778 
779 	int *			timeout_table;
780 	int *			timeouts;
781 	int			timeouts_size;
782 
783 	int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app,
784 			     int *verdict, struct ip_vs_conn **cpp);
785 
786 	struct ip_vs_conn *
787 	(*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app,
788 		       const struct iphdr *iph, int inverse);
789 
790 	struct ip_vs_conn *
791 	(*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app,
792 			const struct iphdr *iph, int inverse);
793 
794 	int (*state_transition)(struct ip_vs_conn *cp, int direction,
795 				const struct sk_buff *skb,
796 				struct ip_vs_app *app);
797 
798 	void (*timeout_change)(struct ip_vs_app *app, int flags);
799 };
800 
801 struct ipvs_master_sync_state {
802 	struct list_head	sync_queue;
803 	struct ip_vs_sync_buff	*sync_buff;
804 	unsigned long		sync_queue_len;
805 	unsigned int		sync_queue_delay;
806 	struct task_struct	*master_thread;
807 	struct delayed_work	master_wakeup_work;
808 	struct netns_ipvs	*ipvs;
809 };
810 
811 /* How much time to keep dests in trash */
812 #define IP_VS_DEST_TRASH_PERIOD		(120 * HZ)
813 
814 struct ipvs_sync_daemon_cfg {
815 	union nf_inet_addr	mcast_group;
816 	int			syncid;
817 	u16			sync_maxlen;
818 	u16			mcast_port;
819 	u8			mcast_af;
820 	u8			mcast_ttl;
821 	/* multicast interface name */
822 	char			mcast_ifn[IP_VS_IFNAME_MAXLEN];
823 };
824 
825 /* IPVS in network namespace */
826 struct netns_ipvs {
827 	int			gen;		/* Generation */
828 	int			enable;		/* enable like nf_hooks do */
829 	/* Hash table: for real service lookups */
830 	#define IP_VS_RTAB_BITS 4
831 	#define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
832 	#define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
833 
834 	struct hlist_head	rs_table[IP_VS_RTAB_SIZE];
835 	/* ip_vs_app */
836 	struct list_head	app_list;
837 	/* ip_vs_proto */
838 	#define IP_VS_PROTO_TAB_SIZE	32	/* must be power of 2 */
839 	struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE];
840 	/* ip_vs_proto_tcp */
841 #ifdef CONFIG_IP_VS_PROTO_TCP
842 	#define	TCP_APP_TAB_BITS	4
843 	#define	TCP_APP_TAB_SIZE	(1 << TCP_APP_TAB_BITS)
844 	#define	TCP_APP_TAB_MASK	(TCP_APP_TAB_SIZE - 1)
845 	struct list_head	tcp_apps[TCP_APP_TAB_SIZE];
846 #endif
847 	/* ip_vs_proto_udp */
848 #ifdef CONFIG_IP_VS_PROTO_UDP
849 	#define	UDP_APP_TAB_BITS	4
850 	#define	UDP_APP_TAB_SIZE	(1 << UDP_APP_TAB_BITS)
851 	#define	UDP_APP_TAB_MASK	(UDP_APP_TAB_SIZE - 1)
852 	struct list_head	udp_apps[UDP_APP_TAB_SIZE];
853 #endif
854 	/* ip_vs_proto_sctp */
855 #ifdef CONFIG_IP_VS_PROTO_SCTP
856 	#define SCTP_APP_TAB_BITS	4
857 	#define SCTP_APP_TAB_SIZE	(1 << SCTP_APP_TAB_BITS)
858 	#define SCTP_APP_TAB_MASK	(SCTP_APP_TAB_SIZE - 1)
859 	/* Hash table for SCTP application incarnations	 */
860 	struct list_head	sctp_apps[SCTP_APP_TAB_SIZE];
861 #endif
862 	/* ip_vs_conn */
863 	atomic_t		conn_count;      /* connection counter */
864 
865 	/* ip_vs_ctl */
866 	struct ip_vs_stats		tot_stats;  /* Statistics & est. */
867 
868 	int			num_services;    /* no of virtual services */
869 
870 	/* Trash for destinations */
871 	struct list_head	dest_trash;
872 	spinlock_t		dest_trash_lock;
873 	struct timer_list	dest_trash_timer; /* expiration timer */
874 	/* Service counters */
875 	atomic_t		ftpsvc_counter;
876 	atomic_t		nullsvc_counter;
877 
878 #ifdef CONFIG_SYSCTL
879 	/* 1/rate drop and drop-entry variables */
880 	struct delayed_work	defense_work;   /* Work handler */
881 	int			drop_rate;
882 	int			drop_counter;
883 	atomic_t		dropentry;
884 	/* locks in ctl.c */
885 	spinlock_t		dropentry_lock;  /* drop entry handling */
886 	spinlock_t		droppacket_lock; /* drop packet handling */
887 	spinlock_t		securetcp_lock;  /* state and timeout tables */
888 
889 	/* sys-ctl struct */
890 	struct ctl_table_header	*sysctl_hdr;
891 	struct ctl_table	*sysctl_tbl;
892 #endif
893 
894 	/* sysctl variables */
895 	int			sysctl_amemthresh;
896 	int			sysctl_am_droprate;
897 	int			sysctl_drop_entry;
898 	int			sysctl_drop_packet;
899 	int			sysctl_secure_tcp;
900 #ifdef CONFIG_IP_VS_NFCT
901 	int			sysctl_conntrack;
902 #endif
903 	int			sysctl_snat_reroute;
904 	int			sysctl_sync_ver;
905 	int			sysctl_sync_ports;
906 	int			sysctl_sync_persist_mode;
907 	unsigned long		sysctl_sync_qlen_max;
908 	int			sysctl_sync_sock_size;
909 	int			sysctl_cache_bypass;
910 	int			sysctl_expire_nodest_conn;
911 	int			sysctl_sloppy_tcp;
912 	int			sysctl_sloppy_sctp;
913 	int			sysctl_expire_quiescent_template;
914 	int			sysctl_sync_threshold[2];
915 	unsigned int		sysctl_sync_refresh_period;
916 	int			sysctl_sync_retries;
917 	int			sysctl_nat_icmp_send;
918 	int			sysctl_pmtu_disc;
919 	int			sysctl_backup_only;
920 	int			sysctl_conn_reuse_mode;
921 	int			sysctl_schedule_icmp;
922 	int			sysctl_ignore_tunneled;
923 
924 	/* ip_vs_lblc */
925 	int			sysctl_lblc_expiration;
926 	struct ctl_table_header	*lblc_ctl_header;
927 	struct ctl_table	*lblc_ctl_table;
928 	/* ip_vs_lblcr */
929 	int			sysctl_lblcr_expiration;
930 	struct ctl_table_header	*lblcr_ctl_header;
931 	struct ctl_table	*lblcr_ctl_table;
932 	/* ip_vs_est */
933 	struct list_head	est_list;	/* estimator list */
934 	spinlock_t		est_lock;
935 	struct timer_list	est_timer;	/* Estimation timer */
936 	/* ip_vs_sync */
937 	spinlock_t		sync_lock;
938 	struct ipvs_master_sync_state *ms;
939 	spinlock_t		sync_buff_lock;
940 	struct task_struct	**backup_threads;
941 	int			threads_mask;
942 	volatile int		sync_state;
943 	struct mutex		sync_mutex;
944 	struct ipvs_sync_daemon_cfg	mcfg;	/* Master Configuration */
945 	struct ipvs_sync_daemon_cfg	bcfg;	/* Backup Configuration */
946 	/* net name space ptr */
947 	struct net		*net;            /* Needed by timer routines */
948 	/* Number of heterogeneous destinations, needed becaus heterogeneous
949 	 * are not supported when synchronization is enabled.
950 	 */
951 	unsigned int		mixed_address_family_dests;
952 };
953 
954 #define DEFAULT_SYNC_THRESHOLD	3
955 #define DEFAULT_SYNC_PERIOD	50
956 #define DEFAULT_SYNC_VER	1
957 #define DEFAULT_SLOPPY_TCP	0
958 #define DEFAULT_SLOPPY_SCTP	0
959 #define DEFAULT_SYNC_REFRESH_PERIOD	(0U * HZ)
960 #define DEFAULT_SYNC_RETRIES		0
961 #define IPVS_SYNC_WAKEUP_RATE	8
962 #define IPVS_SYNC_QLEN_MAX	(IPVS_SYNC_WAKEUP_RATE * 4)
963 #define IPVS_SYNC_SEND_DELAY	(HZ / 50)
964 #define IPVS_SYNC_CHECK_PERIOD	HZ
965 #define IPVS_SYNC_FLUSH_TIME	(HZ * 2)
966 #define IPVS_SYNC_PORTS_MAX	(1 << 6)
967 
968 #ifdef CONFIG_SYSCTL
969 
970 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
971 {
972 	return ipvs->sysctl_sync_threshold[0];
973 }
974 
975 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
976 {
977 	return ACCESS_ONCE(ipvs->sysctl_sync_threshold[1]);
978 }
979 
980 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
981 {
982 	return ACCESS_ONCE(ipvs->sysctl_sync_refresh_period);
983 }
984 
985 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
986 {
987 	return ipvs->sysctl_sync_retries;
988 }
989 
990 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
991 {
992 	return ipvs->sysctl_sync_ver;
993 }
994 
995 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
996 {
997 	return ipvs->sysctl_sloppy_tcp;
998 }
999 
1000 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1001 {
1002 	return ipvs->sysctl_sloppy_sctp;
1003 }
1004 
1005 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1006 {
1007 	return ACCESS_ONCE(ipvs->sysctl_sync_ports);
1008 }
1009 
1010 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1011 {
1012 	return ipvs->sysctl_sync_persist_mode;
1013 }
1014 
1015 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1016 {
1017 	return ipvs->sysctl_sync_qlen_max;
1018 }
1019 
1020 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1021 {
1022 	return ipvs->sysctl_sync_sock_size;
1023 }
1024 
1025 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1026 {
1027 	return ipvs->sysctl_pmtu_disc;
1028 }
1029 
1030 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1031 {
1032 	return ipvs->sync_state & IP_VS_STATE_BACKUP &&
1033 	       ipvs->sysctl_backup_only;
1034 }
1035 
1036 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1037 {
1038 	return ipvs->sysctl_conn_reuse_mode;
1039 }
1040 
1041 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1042 {
1043 	return ipvs->sysctl_schedule_icmp;
1044 }
1045 
1046 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1047 {
1048 	return ipvs->sysctl_ignore_tunneled;
1049 }
1050 
1051 static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1052 {
1053 	return ipvs->sysctl_cache_bypass;
1054 }
1055 
1056 #else
1057 
1058 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1059 {
1060 	return DEFAULT_SYNC_THRESHOLD;
1061 }
1062 
1063 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1064 {
1065 	return DEFAULT_SYNC_PERIOD;
1066 }
1067 
1068 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1069 {
1070 	return DEFAULT_SYNC_REFRESH_PERIOD;
1071 }
1072 
1073 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1074 {
1075 	return DEFAULT_SYNC_RETRIES & 3;
1076 }
1077 
1078 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1079 {
1080 	return DEFAULT_SYNC_VER;
1081 }
1082 
1083 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1084 {
1085 	return DEFAULT_SLOPPY_TCP;
1086 }
1087 
1088 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1089 {
1090 	return DEFAULT_SLOPPY_SCTP;
1091 }
1092 
1093 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1094 {
1095 	return 1;
1096 }
1097 
1098 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1099 {
1100 	return 0;
1101 }
1102 
1103 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1104 {
1105 	return IPVS_SYNC_QLEN_MAX;
1106 }
1107 
1108 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1109 {
1110 	return 0;
1111 }
1112 
1113 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1114 {
1115 	return 1;
1116 }
1117 
1118 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1119 {
1120 	return 0;
1121 }
1122 
1123 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1124 {
1125 	return 1;
1126 }
1127 
1128 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1129 {
1130 	return 0;
1131 }
1132 
1133 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1134 {
1135 	return 0;
1136 }
1137 
1138 static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1139 {
1140 	return 0;
1141 }
1142 
1143 #endif
1144 
1145 /* IPVS core functions
1146  * (from ip_vs_core.c)
1147  */
1148 const char *ip_vs_proto_name(unsigned int proto);
1149 void ip_vs_init_hash_table(struct list_head *table, int rows);
1150 #define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t)))
1151 
1152 #define IP_VS_APP_TYPE_FTP	1
1153 
1154 /* ip_vs_conn handling functions
1155  * (from ip_vs_conn.c)
1156  */
1157 enum {
1158 	IP_VS_DIR_INPUT = 0,
1159 	IP_VS_DIR_OUTPUT,
1160 	IP_VS_DIR_INPUT_ONLY,
1161 	IP_VS_DIR_LAST,
1162 };
1163 
1164 static inline void ip_vs_conn_fill_param(struct netns_ipvs *ipvs, int af, int protocol,
1165 					 const union nf_inet_addr *caddr,
1166 					 __be16 cport,
1167 					 const union nf_inet_addr *vaddr,
1168 					 __be16 vport,
1169 					 struct ip_vs_conn_param *p)
1170 {
1171 	p->ipvs = ipvs;
1172 	p->af = af;
1173 	p->protocol = protocol;
1174 	p->caddr = caddr;
1175 	p->cport = cport;
1176 	p->vaddr = vaddr;
1177 	p->vport = vport;
1178 	p->pe = NULL;
1179 	p->pe_data = NULL;
1180 }
1181 
1182 struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p);
1183 struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p);
1184 
1185 struct ip_vs_conn * ip_vs_conn_in_get_proto(struct netns_ipvs *ipvs, int af,
1186 					    const struct sk_buff *skb,
1187 					    const struct ip_vs_iphdr *iph);
1188 
1189 struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p);
1190 
1191 struct ip_vs_conn * ip_vs_conn_out_get_proto(struct netns_ipvs *ipvs, int af,
1192 					     const struct sk_buff *skb,
1193 					     const struct ip_vs_iphdr *iph);
1194 
1195 /* Get reference to gain full access to conn.
1196  * By default, RCU read-side critical sections have access only to
1197  * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference.
1198  */
1199 static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp)
1200 {
1201 	return atomic_inc_not_zero(&cp->refcnt);
1202 }
1203 
1204 /* put back the conn without restarting its timer */
1205 static inline void __ip_vs_conn_put(struct ip_vs_conn *cp)
1206 {
1207 	smp_mb__before_atomic();
1208 	atomic_dec(&cp->refcnt);
1209 }
1210 void ip_vs_conn_put(struct ip_vs_conn *cp);
1211 void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport);
1212 
1213 struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, int dest_af,
1214 				  const union nf_inet_addr *daddr,
1215 				  __be16 dport, unsigned int flags,
1216 				  struct ip_vs_dest *dest, __u32 fwmark);
1217 void ip_vs_conn_expire_now(struct ip_vs_conn *cp);
1218 
1219 const char *ip_vs_state_name(__u16 proto, int state);
1220 
1221 void ip_vs_tcp_conn_listen(struct ip_vs_conn *cp);
1222 int ip_vs_check_template(struct ip_vs_conn *ct);
1223 void ip_vs_random_dropentry(struct netns_ipvs *ipvs);
1224 int ip_vs_conn_init(void);
1225 void ip_vs_conn_cleanup(void);
1226 
1227 static inline void ip_vs_control_del(struct ip_vs_conn *cp)
1228 {
1229 	struct ip_vs_conn *ctl_cp = cp->control;
1230 	if (!ctl_cp) {
1231 		IP_VS_ERR_BUF("request control DEL for uncontrolled: "
1232 			      "%s:%d to %s:%d\n",
1233 			      IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1234 			      ntohs(cp->cport),
1235 			      IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1236 			      ntohs(cp->vport));
1237 
1238 		return;
1239 	}
1240 
1241 	IP_VS_DBG_BUF(7, "DELeting control for: "
1242 		      "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1243 		      IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1244 		      ntohs(cp->cport),
1245 		      IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1246 		      ntohs(ctl_cp->cport));
1247 
1248 	cp->control = NULL;
1249 	if (atomic_read(&ctl_cp->n_control) == 0) {
1250 		IP_VS_ERR_BUF("BUG control DEL with n=0 : "
1251 			      "%s:%d to %s:%d\n",
1252 			      IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1253 			      ntohs(cp->cport),
1254 			      IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1255 			      ntohs(cp->vport));
1256 
1257 		return;
1258 	}
1259 	atomic_dec(&ctl_cp->n_control);
1260 }
1261 
1262 static inline void
1263 ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp)
1264 {
1265 	if (cp->control) {
1266 		IP_VS_ERR_BUF("request control ADD for already controlled: "
1267 			      "%s:%d to %s:%d\n",
1268 			      IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1269 			      ntohs(cp->cport),
1270 			      IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1271 			      ntohs(cp->vport));
1272 
1273 		ip_vs_control_del(cp);
1274 	}
1275 
1276 	IP_VS_DBG_BUF(7, "ADDing control for: "
1277 		      "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1278 		      IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1279 		      ntohs(cp->cport),
1280 		      IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1281 		      ntohs(ctl_cp->cport));
1282 
1283 	cp->control = ctl_cp;
1284 	atomic_inc(&ctl_cp->n_control);
1285 }
1286 
1287 /* IPVS netns init & cleanup functions */
1288 int ip_vs_estimator_net_init(struct netns_ipvs *ipvs);
1289 int ip_vs_control_net_init(struct netns_ipvs *ipvs);
1290 int ip_vs_protocol_net_init(struct netns_ipvs *ipvs);
1291 int ip_vs_app_net_init(struct netns_ipvs *ipvs);
1292 int ip_vs_conn_net_init(struct netns_ipvs *ipvs);
1293 int ip_vs_sync_net_init(struct netns_ipvs *ipvs);
1294 void ip_vs_conn_net_cleanup(struct netns_ipvs *ipvs);
1295 void ip_vs_app_net_cleanup(struct netns_ipvs *ipvs);
1296 void ip_vs_protocol_net_cleanup(struct netns_ipvs *ipvs);
1297 void ip_vs_control_net_cleanup(struct netns_ipvs *ipvs);
1298 void ip_vs_estimator_net_cleanup(struct netns_ipvs *ipvs);
1299 void ip_vs_sync_net_cleanup(struct netns_ipvs *ipvs);
1300 void ip_vs_service_net_cleanup(struct netns_ipvs *ipvs);
1301 
1302 /* IPVS application functions
1303  * (from ip_vs_app.c)
1304  */
1305 #define IP_VS_APP_MAX_PORTS  8
1306 struct ip_vs_app *register_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1307 void unregister_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1308 int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1309 void ip_vs_unbind_app(struct ip_vs_conn *cp);
1310 int register_ip_vs_app_inc(struct netns_ipvs *ipvs, struct ip_vs_app *app, __u16 proto,
1311 			   __u16 port);
1312 int ip_vs_app_inc_get(struct ip_vs_app *inc);
1313 void ip_vs_app_inc_put(struct ip_vs_app *inc);
1314 
1315 int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb);
1316 int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb);
1317 
1318 int register_ip_vs_pe(struct ip_vs_pe *pe);
1319 int unregister_ip_vs_pe(struct ip_vs_pe *pe);
1320 struct ip_vs_pe *ip_vs_pe_getbyname(const char *name);
1321 struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name);
1322 
1323 /* Use a #define to avoid all of module.h just for these trivial ops */
1324 #define ip_vs_pe_get(pe)			\
1325 	if (pe && pe->module)			\
1326 		__module_get(pe->module);
1327 
1328 #define ip_vs_pe_put(pe)			\
1329 	if (pe && pe->module)			\
1330 		module_put(pe->module);
1331 
1332 /* IPVS protocol functions (from ip_vs_proto.c) */
1333 int ip_vs_protocol_init(void);
1334 void ip_vs_protocol_cleanup(void);
1335 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags);
1336 int *ip_vs_create_timeout_table(int *table, int size);
1337 int ip_vs_set_state_timeout(int *table, int num, const char *const *names,
1338 			    const char *name, int to);
1339 void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
1340 			       const struct sk_buff *skb, int offset,
1341 			       const char *msg);
1342 
1343 extern struct ip_vs_protocol ip_vs_protocol_tcp;
1344 extern struct ip_vs_protocol ip_vs_protocol_udp;
1345 extern struct ip_vs_protocol ip_vs_protocol_icmp;
1346 extern struct ip_vs_protocol ip_vs_protocol_esp;
1347 extern struct ip_vs_protocol ip_vs_protocol_ah;
1348 extern struct ip_vs_protocol ip_vs_protocol_sctp;
1349 
1350 /* Registering/unregistering scheduler functions
1351  * (from ip_vs_sched.c)
1352  */
1353 int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1354 int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1355 int ip_vs_bind_scheduler(struct ip_vs_service *svc,
1356 			 struct ip_vs_scheduler *scheduler);
1357 void ip_vs_unbind_scheduler(struct ip_vs_service *svc,
1358 			    struct ip_vs_scheduler *sched);
1359 struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name);
1360 void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler);
1361 struct ip_vs_conn *
1362 ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb,
1363 	       struct ip_vs_proto_data *pd, int *ignored,
1364 	       struct ip_vs_iphdr *iph);
1365 int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb,
1366 		struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph);
1367 
1368 void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg);
1369 
1370 /* IPVS control data and functions (from ip_vs_ctl.c) */
1371 extern struct ip_vs_stats ip_vs_stats;
1372 extern int sysctl_ip_vs_sync_ver;
1373 
1374 struct ip_vs_service *
1375 ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
1376 		  const union nf_inet_addr *vaddr, __be16 vport);
1377 
1378 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1379 			    const union nf_inet_addr *daddr, __be16 dport);
1380 
1381 int ip_vs_use_count_inc(void);
1382 void ip_vs_use_count_dec(void);
1383 int ip_vs_register_nl_ioctl(void);
1384 void ip_vs_unregister_nl_ioctl(void);
1385 int ip_vs_control_init(void);
1386 void ip_vs_control_cleanup(void);
1387 struct ip_vs_dest *
1388 ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
1389 		const union nf_inet_addr *daddr, __be16 dport,
1390 		const union nf_inet_addr *vaddr, __be16 vport,
1391 		__u16 protocol, __u32 fwmark, __u32 flags);
1392 void ip_vs_try_bind_dest(struct ip_vs_conn *cp);
1393 
1394 static inline void ip_vs_dest_hold(struct ip_vs_dest *dest)
1395 {
1396 	atomic_inc(&dest->refcnt);
1397 }
1398 
1399 static inline void ip_vs_dest_put(struct ip_vs_dest *dest)
1400 {
1401 	smp_mb__before_atomic();
1402 	atomic_dec(&dest->refcnt);
1403 }
1404 
1405 static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest)
1406 {
1407 	if (atomic_dec_return(&dest->refcnt) < 0)
1408 		kfree(dest);
1409 }
1410 
1411 /* IPVS sync daemon data and function prototypes
1412  * (from ip_vs_sync.c)
1413  */
1414 int start_sync_thread(struct netns_ipvs *ipvs, struct ipvs_sync_daemon_cfg *cfg,
1415 		      int state);
1416 int stop_sync_thread(struct netns_ipvs *ipvs, int state);
1417 void ip_vs_sync_conn(struct netns_ipvs *ipvs, struct ip_vs_conn *cp, int pkts);
1418 
1419 /* IPVS rate estimator prototypes (from ip_vs_est.c) */
1420 void ip_vs_start_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1421 void ip_vs_stop_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1422 void ip_vs_zero_estimator(struct ip_vs_stats *stats);
1423 void ip_vs_read_estimator(struct ip_vs_kstats *dst, struct ip_vs_stats *stats);
1424 
1425 /* Various IPVS packet transmitters (from ip_vs_xmit.c) */
1426 int ip_vs_null_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_bypass_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_nat_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_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1433 		      struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1434 int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1435 		  struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1436 int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1437 		    struct ip_vs_protocol *pp, int offset,
1438 		    unsigned int hooknum, struct ip_vs_iphdr *iph);
1439 void ip_vs_dest_dst_rcu_free(struct rcu_head *head);
1440 
1441 #ifdef CONFIG_IP_VS_IPV6
1442 int ip_vs_bypass_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_nat_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_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1447 			 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1448 int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1449 		     struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1450 int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1451 		       struct ip_vs_protocol *pp, int offset,
1452 		       unsigned int hooknum, struct ip_vs_iphdr *iph);
1453 #endif
1454 
1455 #ifdef CONFIG_SYSCTL
1456 /* This is a simple mechanism to ignore packets when
1457  * we are loaded. Just set ip_vs_drop_rate to 'n' and
1458  * we start to drop 1/rate of the packets
1459  */
1460 static inline int ip_vs_todrop(struct netns_ipvs *ipvs)
1461 {
1462 	if (!ipvs->drop_rate)
1463 		return 0;
1464 	if (--ipvs->drop_counter > 0)
1465 		return 0;
1466 	ipvs->drop_counter = ipvs->drop_rate;
1467 	return 1;
1468 }
1469 #else
1470 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; }
1471 #endif
1472 
1473 /* ip_vs_fwd_tag returns the forwarding tag of the connection */
1474 #define IP_VS_FWD_METHOD(cp)  (cp->flags & IP_VS_CONN_F_FWD_MASK)
1475 
1476 static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp)
1477 {
1478 	char fwd;
1479 
1480 	switch (IP_VS_FWD_METHOD(cp)) {
1481 	case IP_VS_CONN_F_MASQ:
1482 		fwd = 'M'; break;
1483 	case IP_VS_CONN_F_LOCALNODE:
1484 		fwd = 'L'; break;
1485 	case IP_VS_CONN_F_TUNNEL:
1486 		fwd = 'T'; break;
1487 	case IP_VS_CONN_F_DROUTE:
1488 		fwd = 'R'; break;
1489 	case IP_VS_CONN_F_BYPASS:
1490 		fwd = 'B'; break;
1491 	default:
1492 		fwd = '?'; break;
1493 	}
1494 	return fwd;
1495 }
1496 
1497 void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp,
1498 		    struct ip_vs_conn *cp, int dir);
1499 
1500 #ifdef CONFIG_IP_VS_IPV6
1501 void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp,
1502 		       struct ip_vs_conn *cp, int dir);
1503 #endif
1504 
1505 __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset);
1506 
1507 static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum)
1508 {
1509 	__be32 diff[2] = { ~old, new };
1510 
1511 	return csum_partial(diff, sizeof(diff), oldsum);
1512 }
1513 
1514 #ifdef CONFIG_IP_VS_IPV6
1515 static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new,
1516 					__wsum oldsum)
1517 {
1518 	__be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0],
1519 			    new[3],  new[2],  new[1],  new[0] };
1520 
1521 	return csum_partial(diff, sizeof(diff), oldsum);
1522 }
1523 #endif
1524 
1525 static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum)
1526 {
1527 	__be16 diff[2] = { ~old, new };
1528 
1529 	return csum_partial(diff, sizeof(diff), oldsum);
1530 }
1531 
1532 /* Forget current conntrack (unconfirmed) and attach notrack entry */
1533 static inline void ip_vs_notrack(struct sk_buff *skb)
1534 {
1535 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1536 	enum ip_conntrack_info ctinfo;
1537 	struct nf_conn *ct = nf_ct_get(skb, &ctinfo);
1538 
1539 	if (!ct || !nf_ct_is_untracked(ct)) {
1540 		nf_conntrack_put(skb->nfct);
1541 		skb->nfct = &nf_ct_untracked_get()->ct_general;
1542 		skb->nfctinfo = IP_CT_NEW;
1543 		nf_conntrack_get(skb->nfct);
1544 	}
1545 #endif
1546 }
1547 
1548 #ifdef CONFIG_IP_VS_NFCT
1549 /* Netfilter connection tracking
1550  * (from ip_vs_nfct.c)
1551  */
1552 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1553 {
1554 #ifdef CONFIG_SYSCTL
1555 	return ipvs->sysctl_conntrack;
1556 #else
1557 	return 0;
1558 #endif
1559 }
1560 
1561 void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp,
1562 			    int outin);
1563 int ip_vs_confirm_conntrack(struct sk_buff *skb);
1564 void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct,
1565 			       struct ip_vs_conn *cp, u_int8_t proto,
1566 			       const __be16 port, int from_rs);
1567 void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp);
1568 
1569 #else
1570 
1571 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1572 {
1573 	return 0;
1574 }
1575 
1576 static inline void ip_vs_update_conntrack(struct sk_buff *skb,
1577 					  struct ip_vs_conn *cp, int outin)
1578 {
1579 }
1580 
1581 static inline int ip_vs_confirm_conntrack(struct sk_buff *skb)
1582 {
1583 	return NF_ACCEPT;
1584 }
1585 
1586 static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp)
1587 {
1588 }
1589 #endif /* CONFIG_IP_VS_NFCT */
1590 
1591 static inline int
1592 ip_vs_dest_conn_overhead(struct ip_vs_dest *dest)
1593 {
1594 	/* We think the overhead of processing active connections is 256
1595 	 * times higher than that of inactive connections in average. (This
1596 	 * 256 times might not be accurate, we will change it later) We
1597 	 * use the following formula to estimate the overhead now:
1598 	 *		  dest->activeconns*256 + dest->inactconns
1599 	 */
1600 	return (atomic_read(&dest->activeconns) << 8) +
1601 		atomic_read(&dest->inactconns);
1602 }
1603 
1604 #endif	/* _NET_IP_VS_H */
1605