blob: a0d2e0bb9a94b90da24b22a327253dc6e9c0e8ff [file] [log] [blame]
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001/* SPDX-License-Identifier: GPL-2.0 */
2/* IP Virtual Server
3 * data structure and functionality definitions
4 */
5
6#ifndef _NET_IP_VS_H
7#define _NET_IP_VS_H
8
9#include <linux/ip_vs.h> /* definitions shared with userland */
10
11#include <asm/types.h> /* for __uXX types */
12
13#include <linux/list.h> /* for struct list_head */
14#include <linux/spinlock.h> /* for struct rwlock_t */
15#include <linux/atomic.h> /* for struct atomic_t */
16#include <linux/refcount.h> /* for struct refcount_t */
17
18#include <linux/compiler.h>
19#include <linux/timer.h>
20#include <linux/bug.h>
21
22#include <net/checksum.h>
23#include <linux/netfilter.h> /* for union nf_inet_addr */
24#include <linux/ip.h>
25#include <linux/ipv6.h> /* for struct ipv6hdr */
26#include <net/ipv6.h>
27#if IS_ENABLED(CONFIG_IP_VS_IPV6)
28#include <linux/netfilter_ipv6/ip6_tables.h>
29#endif
30#if IS_ENABLED(CONFIG_NF_CONNTRACK)
31#include <net/netfilter/nf_conntrack.h>
32#endif
33#include <net/net_namespace.h> /* Netw namespace */
34
35#define IP_VS_HDR_INVERSE 1
36#define IP_VS_HDR_ICMP 2
37
38/* Generic access of ipvs struct */
39static inline struct netns_ipvs *net_ipvs(struct net* net)
40{
41 return net->ipvs;
42}
43
44/* Connections' size value needed by ip_vs_ctl.c */
45extern int ip_vs_conn_tab_size;
46
47struct ip_vs_iphdr {
48 int hdr_flags; /* ipvs flags */
49 __u32 off; /* Where IP or IPv4 header starts */
50 __u32 len; /* IPv4 simply where L4 starts
51 * IPv6 where L4 Transport Header starts */
52 __u16 fragoffs; /* IPv6 fragment offset, 0 if first frag (or not frag)*/
53 __s16 protocol;
54 __s32 flags;
55 union nf_inet_addr saddr;
56 union nf_inet_addr daddr;
57};
58
59static inline void *frag_safe_skb_hp(const struct sk_buff *skb, int offset,
60 int len, void *buffer)
61{
62 return skb_header_pointer(skb, offset, len, buffer);
63}
64
65/* This function handles filling *ip_vs_iphdr, both for IPv4 and IPv6.
66 * IPv6 requires some extra work, as finding proper header position,
67 * depend on the IPv6 extension headers.
68 */
69static inline int
70ip_vs_fill_iph_skb_off(int af, const struct sk_buff *skb, int offset,
71 int hdr_flags, struct ip_vs_iphdr *iphdr)
72{
73 iphdr->hdr_flags = hdr_flags;
74 iphdr->off = offset;
75
76#ifdef CONFIG_IP_VS_IPV6
77 if (af == AF_INET6) {
78 struct ipv6hdr _iph;
79 const struct ipv6hdr *iph = skb_header_pointer(
80 skb, offset, sizeof(_iph), &_iph);
81 if (!iph)
82 return 0;
83
84 iphdr->saddr.in6 = iph->saddr;
85 iphdr->daddr.in6 = iph->daddr;
86 /* ipv6_find_hdr() updates len, flags */
87 iphdr->len = offset;
88 iphdr->flags = 0;
89 iphdr->protocol = ipv6_find_hdr(skb, &iphdr->len, -1,
90 &iphdr->fragoffs,
91 &iphdr->flags);
92 if (iphdr->protocol < 0)
93 return 0;
94 } else
95#endif
96 {
97 struct iphdr _iph;
98 const struct iphdr *iph = skb_header_pointer(
99 skb, offset, sizeof(_iph), &_iph);
100 if (!iph)
101 return 0;
102
103 iphdr->len = offset + iph->ihl * 4;
104 iphdr->fragoffs = 0;
105 iphdr->protocol = iph->protocol;
106 iphdr->saddr.ip = iph->saddr;
107 iphdr->daddr.ip = iph->daddr;
108 }
109
110 return 1;
111}
112
113static inline int
114ip_vs_fill_iph_skb_icmp(int af, const struct sk_buff *skb, int offset,
115 bool inverse, struct ip_vs_iphdr *iphdr)
116{
117 int hdr_flags = IP_VS_HDR_ICMP;
118
119 if (inverse)
120 hdr_flags |= IP_VS_HDR_INVERSE;
121
122 return ip_vs_fill_iph_skb_off(af, skb, offset, hdr_flags, iphdr);
123}
124
125static inline int
126ip_vs_fill_iph_skb(int af, const struct sk_buff *skb, bool inverse,
127 struct ip_vs_iphdr *iphdr)
128{
129 int hdr_flags = 0;
130
131 if (inverse)
132 hdr_flags |= IP_VS_HDR_INVERSE;
133
134 return ip_vs_fill_iph_skb_off(af, skb, skb_network_offset(skb),
135 hdr_flags, iphdr);
136}
137
138static inline bool
139ip_vs_iph_inverse(const struct ip_vs_iphdr *iph)
140{
141 return !!(iph->hdr_flags & IP_VS_HDR_INVERSE);
142}
143
144static inline bool
145ip_vs_iph_icmp(const struct ip_vs_iphdr *iph)
146{
147 return !!(iph->hdr_flags & IP_VS_HDR_ICMP);
148}
149
150static inline void ip_vs_addr_copy(int af, union nf_inet_addr *dst,
151 const union nf_inet_addr *src)
152{
153#ifdef CONFIG_IP_VS_IPV6
154 if (af == AF_INET6)
155 dst->in6 = src->in6;
156 else
157#endif
158 dst->ip = src->ip;
159}
160
161static inline void ip_vs_addr_set(int af, union nf_inet_addr *dst,
162 const union nf_inet_addr *src)
163{
164#ifdef CONFIG_IP_VS_IPV6
165 if (af == AF_INET6) {
166 dst->in6 = src->in6;
167 return;
168 }
169#endif
170 dst->ip = src->ip;
171 dst->all[1] = 0;
172 dst->all[2] = 0;
173 dst->all[3] = 0;
174}
175
176static inline int ip_vs_addr_equal(int af, const union nf_inet_addr *a,
177 const union nf_inet_addr *b)
178{
179#ifdef CONFIG_IP_VS_IPV6
180 if (af == AF_INET6)
181 return ipv6_addr_equal(&a->in6, &b->in6);
182#endif
183 return a->ip == b->ip;
184}
185
186#ifdef CONFIG_IP_VS_DEBUG
187#include <linux/net.h>
188
189int ip_vs_get_debug_level(void);
190
191static inline const char *ip_vs_dbg_addr(int af, char *buf, size_t buf_len,
192 const union nf_inet_addr *addr,
193 int *idx)
194{
195 int len;
196#ifdef CONFIG_IP_VS_IPV6
197 if (af == AF_INET6)
198 len = snprintf(&buf[*idx], buf_len - *idx, "[%pI6c]",
199 &addr->in6) + 1;
200 else
201#endif
202 len = snprintf(&buf[*idx], buf_len - *idx, "%pI4",
203 &addr->ip) + 1;
204
205 *idx += len;
206 BUG_ON(*idx > buf_len + 1);
207 return &buf[*idx - len];
208}
209
210#define IP_VS_DBG_BUF(level, msg, ...) \
211 do { \
212 char ip_vs_dbg_buf[160]; \
213 int ip_vs_dbg_idx = 0; \
214 if (level <= ip_vs_get_debug_level()) \
215 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
216 } while (0)
217#define IP_VS_ERR_BUF(msg...) \
218 do { \
219 char ip_vs_dbg_buf[160]; \
220 int ip_vs_dbg_idx = 0; \
221 pr_err(msg); \
222 } while (0)
223
224/* Only use from within IP_VS_DBG_BUF() or IP_VS_ERR_BUF macros */
225#define IP_VS_DBG_ADDR(af, addr) \
226 ip_vs_dbg_addr(af, ip_vs_dbg_buf, \
227 sizeof(ip_vs_dbg_buf), addr, \
228 &ip_vs_dbg_idx)
229
230#define IP_VS_DBG(level, msg, ...) \
231 do { \
232 if (level <= ip_vs_get_debug_level()) \
233 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
234 } while (0)
235#define IP_VS_DBG_RL(msg, ...) \
236 do { \
237 if (net_ratelimit()) \
238 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
239 } while (0)
240#define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) \
241 do { \
242 if (level <= ip_vs_get_debug_level()) \
243 pp->debug_packet(af, pp, skb, ofs, msg); \
244 } while (0)
245#define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) \
246 do { \
247 if (level <= ip_vs_get_debug_level() && \
248 net_ratelimit()) \
249 pp->debug_packet(af, pp, skb, ofs, msg); \
250 } while (0)
251#else /* NO DEBUGGING at ALL */
252#define IP_VS_DBG_BUF(level, msg...) do {} while (0)
253#define IP_VS_ERR_BUF(msg...) do {} while (0)
254#define IP_VS_DBG(level, msg...) do {} while (0)
255#define IP_VS_DBG_RL(msg...) do {} while (0)
256#define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
257#define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
258#endif
259
260#define IP_VS_BUG() BUG()
261#define IP_VS_ERR_RL(msg, ...) \
262 do { \
263 if (net_ratelimit()) \
264 pr_err(msg, ##__VA_ARGS__); \
265 } while (0)
266
267#ifdef CONFIG_IP_VS_DEBUG
268#define EnterFunction(level) \
269 do { \
270 if (level <= ip_vs_get_debug_level()) \
271 printk(KERN_DEBUG \
272 pr_fmt("Enter: %s, %s line %i\n"), \
273 __func__, __FILE__, __LINE__); \
274 } while (0)
275#define LeaveFunction(level) \
276 do { \
277 if (level <= ip_vs_get_debug_level()) \
278 printk(KERN_DEBUG \
279 pr_fmt("Leave: %s, %s line %i\n"), \
280 __func__, __FILE__, __LINE__); \
281 } while (0)
282#else
283#define EnterFunction(level) do {} while (0)
284#define LeaveFunction(level) do {} while (0)
285#endif
286
287/* The port number of FTP service (in network order). */
288#define FTPPORT cpu_to_be16(21)
289#define FTPDATA cpu_to_be16(20)
290
291/* TCP State Values */
292enum {
293 IP_VS_TCP_S_NONE = 0,
294 IP_VS_TCP_S_ESTABLISHED,
295 IP_VS_TCP_S_SYN_SENT,
296 IP_VS_TCP_S_SYN_RECV,
297 IP_VS_TCP_S_FIN_WAIT,
298 IP_VS_TCP_S_TIME_WAIT,
299 IP_VS_TCP_S_CLOSE,
300 IP_VS_TCP_S_CLOSE_WAIT,
301 IP_VS_TCP_S_LAST_ACK,
302 IP_VS_TCP_S_LISTEN,
303 IP_VS_TCP_S_SYNACK,
304 IP_VS_TCP_S_LAST
305};
306
307/* UDP State Values */
308enum {
309 IP_VS_UDP_S_NORMAL,
310 IP_VS_UDP_S_LAST,
311};
312
313/* ICMP State Values */
314enum {
315 IP_VS_ICMP_S_NORMAL,
316 IP_VS_ICMP_S_LAST,
317};
318
319/* SCTP State Values */
320enum ip_vs_sctp_states {
321 IP_VS_SCTP_S_NONE,
322 IP_VS_SCTP_S_INIT1,
323 IP_VS_SCTP_S_INIT,
324 IP_VS_SCTP_S_COOKIE_SENT,
325 IP_VS_SCTP_S_COOKIE_REPLIED,
326 IP_VS_SCTP_S_COOKIE_WAIT,
327 IP_VS_SCTP_S_COOKIE,
328 IP_VS_SCTP_S_COOKIE_ECHOED,
329 IP_VS_SCTP_S_ESTABLISHED,
330 IP_VS_SCTP_S_SHUTDOWN_SENT,
331 IP_VS_SCTP_S_SHUTDOWN_RECEIVED,
332 IP_VS_SCTP_S_SHUTDOWN_ACK_SENT,
333 IP_VS_SCTP_S_REJECTED,
334 IP_VS_SCTP_S_CLOSED,
335 IP_VS_SCTP_S_LAST
336};
337
338/* Connection templates use bits from state */
339#define IP_VS_CTPL_S_NONE 0x0000
340#define IP_VS_CTPL_S_ASSURED 0x0001
341#define IP_VS_CTPL_S_LAST 0x0002
342
343/* Delta sequence info structure
344 * Each ip_vs_conn has 2 (output AND input seq. changes).
345 * Only used in the VS/NAT.
346 */
347struct ip_vs_seq {
348 __u32 init_seq; /* Add delta from this seq */
349 __u32 delta; /* Delta in sequence numbers */
350 __u32 previous_delta; /* Delta in sequence numbers
351 * before last resized pkt */
352};
353
354/* counters per cpu */
355struct ip_vs_counters {
356 __u64 conns; /* connections scheduled */
357 __u64 inpkts; /* incoming packets */
358 __u64 outpkts; /* outgoing packets */
359 __u64 inbytes; /* incoming bytes */
360 __u64 outbytes; /* outgoing bytes */
361};
362/* Stats per cpu */
363struct ip_vs_cpu_stats {
364 struct ip_vs_counters cnt;
365 struct u64_stats_sync syncp;
366};
367
368/* IPVS statistics objects */
369struct ip_vs_estimator {
370 struct list_head list;
371
372 u64 last_inbytes;
373 u64 last_outbytes;
374 u64 last_conns;
375 u64 last_inpkts;
376 u64 last_outpkts;
377
378 u64 cps;
379 u64 inpps;
380 u64 outpps;
381 u64 inbps;
382 u64 outbps;
383};
384
385/*
386 * IPVS statistics object, 64-bit kernel version of struct ip_vs_stats_user
387 */
388struct ip_vs_kstats {
389 u64 conns; /* connections scheduled */
390 u64 inpkts; /* incoming packets */
391 u64 outpkts; /* outgoing packets */
392 u64 inbytes; /* incoming bytes */
393 u64 outbytes; /* outgoing bytes */
394
395 u64 cps; /* current connection rate */
396 u64 inpps; /* current in packet rate */
397 u64 outpps; /* current out packet rate */
398 u64 inbps; /* current in byte rate */
399 u64 outbps; /* current out byte rate */
400};
401
402struct ip_vs_stats {
403 struct ip_vs_kstats kstats; /* kernel statistics */
404 struct ip_vs_estimator est; /* estimator */
405 struct ip_vs_cpu_stats __percpu *cpustats; /* per cpu counters */
406 spinlock_t lock; /* spin lock */
407 struct ip_vs_kstats kstats0; /* reset values */
408};
409
410struct dst_entry;
411struct iphdr;
412struct ip_vs_conn;
413struct ip_vs_app;
414struct sk_buff;
415struct ip_vs_proto_data;
416
417struct ip_vs_protocol {
418 struct ip_vs_protocol *next;
419 char *name;
420 u16 protocol;
421 u16 num_states;
422 int dont_defrag;
423
424 void (*init)(struct ip_vs_protocol *pp);
425
426 void (*exit)(struct ip_vs_protocol *pp);
427
428 int (*init_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
429
430 void (*exit_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
431
432 int (*conn_schedule)(struct netns_ipvs *ipvs,
433 int af, struct sk_buff *skb,
434 struct ip_vs_proto_data *pd,
435 int *verdict, struct ip_vs_conn **cpp,
436 struct ip_vs_iphdr *iph);
437
438 struct ip_vs_conn *
439 (*conn_in_get)(struct netns_ipvs *ipvs,
440 int af,
441 const struct sk_buff *skb,
442 const struct ip_vs_iphdr *iph);
443
444 struct ip_vs_conn *
445 (*conn_out_get)(struct netns_ipvs *ipvs,
446 int af,
447 const struct sk_buff *skb,
448 const struct ip_vs_iphdr *iph);
449
450 int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
451 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
452
453 int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
454 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
455
456 int (*csum_check)(int af, struct sk_buff *skb,
457 struct ip_vs_protocol *pp);
458
459 const char *(*state_name)(int state);
460
461 void (*state_transition)(struct ip_vs_conn *cp, int direction,
462 const struct sk_buff *skb,
463 struct ip_vs_proto_data *pd);
464
465 int (*register_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
466
467 void (*unregister_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
468
469 int (*app_conn_bind)(struct ip_vs_conn *cp);
470
471 void (*debug_packet)(int af, struct ip_vs_protocol *pp,
472 const struct sk_buff *skb,
473 int offset,
474 const char *msg);
475
476 void (*timeout_change)(struct ip_vs_proto_data *pd, int flags);
477};
478
479/* protocol data per netns */
480struct ip_vs_proto_data {
481 struct ip_vs_proto_data *next;
482 struct ip_vs_protocol *pp;
483 int *timeout_table; /* protocol timeout table */
484 atomic_t appcnt; /* counter of proto app incs. */
485 struct tcp_states_t *tcp_state_table;
486};
487
488struct ip_vs_protocol *ip_vs_proto_get(unsigned short proto);
489struct ip_vs_proto_data *ip_vs_proto_data_get(struct netns_ipvs *ipvs,
490 unsigned short proto);
491
492struct ip_vs_conn_param {
493 struct netns_ipvs *ipvs;
494 const union nf_inet_addr *caddr;
495 const union nf_inet_addr *vaddr;
496 __be16 cport;
497 __be16 vport;
498 __u16 protocol;
499 u16 af;
500
501 const struct ip_vs_pe *pe;
502 char *pe_data;
503 __u8 pe_data_len;
504};
505
506/* IP_VS structure allocated for each dynamically scheduled connection */
507struct ip_vs_conn {
508 struct hlist_node c_list; /* hashed list heads */
509 /* Protocol, addresses and port numbers */
510 __be16 cport;
511 __be16 dport;
512 __be16 vport;
513 u16 af; /* address family */
514 union nf_inet_addr caddr; /* client address */
515 union nf_inet_addr vaddr; /* virtual address */
516 union nf_inet_addr daddr; /* destination address */
517 volatile __u32 flags; /* status flags */
518 __u16 protocol; /* Which protocol (TCP/UDP) */
519 __u16 daf; /* Address family of the dest */
520 struct netns_ipvs *ipvs;
521
522 /* counter and timer */
523 refcount_t refcnt; /* reference count */
524 struct timer_list timer; /* Expiration timer */
525 volatile unsigned long timeout; /* timeout */
526
527 /* Flags and state transition */
528 spinlock_t lock; /* lock for state transition */
529 volatile __u16 state; /* state info */
530 volatile __u16 old_state; /* old state, to be used for
531 * state transition triggerd
532 * synchronization
533 */
534 __u32 fwmark; /* Fire wall mark from skb */
535 unsigned long sync_endtime; /* jiffies + sent_retries */
536
537 /* Control members */
538 struct ip_vs_conn *control; /* Master control connection */
539 atomic_t n_control; /* Number of controlled ones */
540 struct ip_vs_dest *dest; /* real server */
541 atomic_t in_pkts; /* incoming packet counter */
542
543 /* Packet transmitter for different forwarding methods. If it
544 * mangles the packet, it must return NF_DROP or better NF_STOLEN,
545 * otherwise this must be changed to a sk_buff **.
546 * NF_ACCEPT can be returned when destination is local.
547 */
548 int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp,
549 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
550
551 /* Note: we can group the following members into a structure,
552 * in order to save more space, and the following members are
553 * only used in VS/NAT anyway
554 */
555 struct ip_vs_app *app; /* bound ip_vs_app object */
556 void *app_data; /* Application private data */
557 struct ip_vs_seq in_seq; /* incoming seq. struct */
558 struct ip_vs_seq out_seq; /* outgoing seq. struct */
559
560 const struct ip_vs_pe *pe;
561 char *pe_data;
562 __u8 pe_data_len;
563
564 struct rcu_head rcu_head;
565};
566
567/* Extended internal versions of struct ip_vs_service_user and ip_vs_dest_user
568 * for IPv6 support.
569 *
570 * We need these to conveniently pass around service and destination
571 * options, but unfortunately, we also need to keep the old definitions to
572 * maintain userspace backwards compatibility for the setsockopt interface.
573 */
574struct ip_vs_service_user_kern {
575 /* virtual service addresses */
576 u16 af;
577 u16 protocol;
578 union nf_inet_addr addr; /* virtual ip address */
579 __be16 port;
580 u32 fwmark; /* firwall mark of service */
581
582 /* virtual service options */
583 char *sched_name;
584 char *pe_name;
585 unsigned int flags; /* virtual service flags */
586 unsigned int timeout; /* persistent timeout in sec */
587 __be32 netmask; /* persistent netmask or plen */
588};
589
590
591struct ip_vs_dest_user_kern {
592 /* destination server address */
593 union nf_inet_addr addr;
594 __be16 port;
595
596 /* real server options */
597 unsigned int conn_flags; /* connection flags */
598 int weight; /* destination weight */
599
600 /* thresholds for active connections */
601 u32 u_threshold; /* upper threshold */
602 u32 l_threshold; /* lower threshold */
603
604 /* Address family of addr */
605 u16 af;
606};
607
608
609/*
610 * The information about the virtual service offered to the net and the
611 * forwarding entries.
612 */
613struct ip_vs_service {
614 struct hlist_node s_list; /* for normal service table */
615 struct hlist_node f_list; /* for fwmark-based service table */
616 atomic_t refcnt; /* reference counter */
617
618 u16 af; /* address family */
619 __u16 protocol; /* which protocol (TCP/UDP) */
620 union nf_inet_addr addr; /* IP address for virtual service */
621 __be16 port; /* port number for the service */
622 __u32 fwmark; /* firewall mark of the service */
623 unsigned int flags; /* service status flags */
624 unsigned int timeout; /* persistent timeout in ticks */
625 __be32 netmask; /* grouping granularity, mask/plen */
626 struct netns_ipvs *ipvs;
627
628 struct list_head destinations; /* real server d-linked list */
629 __u32 num_dests; /* number of servers */
630 struct ip_vs_stats stats; /* statistics for the service */
631
632 /* for scheduling */
633 struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */
634 spinlock_t sched_lock; /* lock sched_data */
635 void *sched_data; /* scheduler application data */
636
637 /* alternate persistence engine */
638 struct ip_vs_pe __rcu *pe;
639 int conntrack_afmask;
640
641 struct rcu_head rcu_head;
642};
643
644/* Information for cached dst */
645struct ip_vs_dest_dst {
646 struct dst_entry *dst_cache; /* destination cache entry */
647 u32 dst_cookie;
648 union nf_inet_addr dst_saddr;
649 struct rcu_head rcu_head;
650};
651
652/* The real server destination forwarding entry with ip address, port number,
653 * and so on.
654 */
655struct ip_vs_dest {
656 struct list_head n_list; /* for the dests in the service */
657 struct hlist_node d_list; /* for table with all the dests */
658
659 u16 af; /* address family */
660 __be16 port; /* port number of the server */
661 union nf_inet_addr addr; /* IP address of the server */
662 volatile unsigned int flags; /* dest status flags */
663 atomic_t conn_flags; /* flags to copy to conn */
664 atomic_t weight; /* server weight */
665 atomic_t last_weight; /* server latest weight */
666
667 refcount_t refcnt; /* reference counter */
668 struct ip_vs_stats stats; /* statistics */
669 unsigned long idle_start; /* start time, jiffies */
670
671 /* connection counters and thresholds */
672 atomic_t activeconns; /* active connections */
673 atomic_t inactconns; /* inactive connections */
674 atomic_t persistconns; /* persistent connections */
675 __u32 u_threshold; /* upper threshold */
676 __u32 l_threshold; /* lower threshold */
677
678 /* for destination cache */
679 spinlock_t dst_lock; /* lock of dst_cache */
680 struct ip_vs_dest_dst __rcu *dest_dst; /* cached dst info */
681
682 /* for virtual service */
683 struct ip_vs_service __rcu *svc; /* service it belongs to */
684 __u16 protocol; /* which protocol (TCP/UDP) */
685 __be16 vport; /* virtual port number */
686 union nf_inet_addr vaddr; /* virtual IP address */
687 __u32 vfwmark; /* firewall mark of service */
688
689 struct list_head t_list; /* in dest_trash */
690 unsigned int in_rs_table:1; /* we are in rs_table */
691};
692
693/* The scheduler object */
694struct ip_vs_scheduler {
695 struct list_head n_list; /* d-linked list head */
696 char *name; /* scheduler name */
697 atomic_t refcnt; /* reference counter */
698 struct module *module; /* THIS_MODULE/NULL */
699
700 /* scheduler initializing service */
701 int (*init_service)(struct ip_vs_service *svc);
702 /* scheduling service finish */
703 void (*done_service)(struct ip_vs_service *svc);
704 /* dest is linked */
705 int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
706 /* dest is unlinked */
707 int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
708 /* dest is updated */
709 int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
710
711 /* selecting a server from the given service */
712 struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc,
713 const struct sk_buff *skb,
714 struct ip_vs_iphdr *iph);
715};
716
717/* The persistence engine object */
718struct ip_vs_pe {
719 struct list_head n_list; /* d-linked list head */
720 char *name; /* scheduler name */
721 atomic_t refcnt; /* reference counter */
722 struct module *module; /* THIS_MODULE/NULL */
723
724 /* get the connection template, if any */
725 int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb);
726 bool (*ct_match)(const struct ip_vs_conn_param *p,
727 struct ip_vs_conn *ct);
728 u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval,
729 bool inverse);
730 int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf);
731 /* create connections for real-server outgoing packets */
732 struct ip_vs_conn* (*conn_out)(struct ip_vs_service *svc,
733 struct ip_vs_dest *dest,
734 struct sk_buff *skb,
735 const struct ip_vs_iphdr *iph,
736 __be16 dport, __be16 cport);
737};
738
739/* The application module object (a.k.a. app incarnation) */
740struct ip_vs_app {
741 struct list_head a_list; /* member in app list */
742 int type; /* IP_VS_APP_TYPE_xxx */
743 char *name; /* application module name */
744 __u16 protocol;
745 struct module *module; /* THIS_MODULE/NULL */
746 struct list_head incs_list; /* list of incarnations */
747
748 /* members for application incarnations */
749 struct list_head p_list; /* member in proto app list */
750 struct ip_vs_app *app; /* its real application */
751 __be16 port; /* port number in net order */
752 atomic_t usecnt; /* usage counter */
753 struct rcu_head rcu_head;
754
755 /* output hook: Process packet in inout direction, diff set for TCP.
756 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
757 * 2=Mangled but checksum was not updated
758 */
759 int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *,
760 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh);
761
762 /* input hook: Process packet in outin direction, diff set for TCP.
763 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
764 * 2=Mangled but checksum was not updated
765 */
766 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *,
767 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh);
768
769 /* ip_vs_app initializer */
770 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *);
771
772 /* ip_vs_app finish */
773 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *);
774
775
776 /* not used now */
777 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *,
778 struct ip_vs_protocol *);
779
780 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *);
781
782 int * timeout_table;
783 int * timeouts;
784 int timeouts_size;
785
786 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app,
787 int *verdict, struct ip_vs_conn **cpp);
788
789 struct ip_vs_conn *
790 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app,
791 const struct iphdr *iph, int inverse);
792
793 struct ip_vs_conn *
794 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app,
795 const struct iphdr *iph, int inverse);
796
797 int (*state_transition)(struct ip_vs_conn *cp, int direction,
798 const struct sk_buff *skb,
799 struct ip_vs_app *app);
800
801 void (*timeout_change)(struct ip_vs_app *app, int flags);
802};
803
804struct ipvs_master_sync_state {
805 struct list_head sync_queue;
806 struct ip_vs_sync_buff *sync_buff;
807 unsigned long sync_queue_len;
808 unsigned int sync_queue_delay;
809 struct task_struct *master_thread;
810 struct delayed_work master_wakeup_work;
811 struct netns_ipvs *ipvs;
812};
813
814/* How much time to keep dests in trash */
815#define IP_VS_DEST_TRASH_PERIOD (120 * HZ)
816
817struct ipvs_sync_daemon_cfg {
818 union nf_inet_addr mcast_group;
819 int syncid;
820 u16 sync_maxlen;
821 u16 mcast_port;
822 u8 mcast_af;
823 u8 mcast_ttl;
824 /* multicast interface name */
825 char mcast_ifn[IP_VS_IFNAME_MAXLEN];
826};
827
828/* IPVS in network namespace */
829struct netns_ipvs {
830 int gen; /* Generation */
831 int enable; /* enable like nf_hooks do */
832 /* Hash table: for real service lookups */
833 #define IP_VS_RTAB_BITS 4
834 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
835 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
836
837 struct hlist_head rs_table[IP_VS_RTAB_SIZE];
838 /* ip_vs_app */
839 struct list_head app_list;
840 /* ip_vs_proto */
841 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */
842 struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE];
843 /* ip_vs_proto_tcp */
844#ifdef CONFIG_IP_VS_PROTO_TCP
845 #define TCP_APP_TAB_BITS 4
846 #define TCP_APP_TAB_SIZE (1 << TCP_APP_TAB_BITS)
847 #define TCP_APP_TAB_MASK (TCP_APP_TAB_SIZE - 1)
848 struct list_head tcp_apps[TCP_APP_TAB_SIZE];
849#endif
850 /* ip_vs_proto_udp */
851#ifdef CONFIG_IP_VS_PROTO_UDP
852 #define UDP_APP_TAB_BITS 4
853 #define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS)
854 #define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1)
855 struct list_head udp_apps[UDP_APP_TAB_SIZE];
856#endif
857 /* ip_vs_proto_sctp */
858#ifdef CONFIG_IP_VS_PROTO_SCTP
859 #define SCTP_APP_TAB_BITS 4
860 #define SCTP_APP_TAB_SIZE (1 << SCTP_APP_TAB_BITS)
861 #define SCTP_APP_TAB_MASK (SCTP_APP_TAB_SIZE - 1)
862 /* Hash table for SCTP application incarnations */
863 struct list_head sctp_apps[SCTP_APP_TAB_SIZE];
864#endif
865 /* ip_vs_conn */
866 atomic_t conn_count; /* connection counter */
867
868 /* ip_vs_ctl */
869 struct ip_vs_stats tot_stats; /* Statistics & est. */
870
871 int num_services; /* no of virtual services */
872
873 /* Trash for destinations */
874 struct list_head dest_trash;
875 spinlock_t dest_trash_lock;
876 struct timer_list dest_trash_timer; /* expiration timer */
877 /* Service counters */
878 atomic_t ftpsvc_counter;
879 atomic_t nullsvc_counter;
880 atomic_t conn_out_counter;
881
882#ifdef CONFIG_SYSCTL
883 /* 1/rate drop and drop-entry variables */
884 struct delayed_work defense_work; /* Work handler */
885 int drop_rate;
886 int drop_counter;
887 atomic_t dropentry;
888 /* locks in ctl.c */
889 spinlock_t dropentry_lock; /* drop entry handling */
890 spinlock_t droppacket_lock; /* drop packet handling */
891 spinlock_t securetcp_lock; /* state and timeout tables */
892
893 /* sys-ctl struct */
894 struct ctl_table_header *sysctl_hdr;
895 struct ctl_table *sysctl_tbl;
896#endif
897
898 /* sysctl variables */
899 int sysctl_amemthresh;
900 int sysctl_am_droprate;
901 int sysctl_drop_entry;
902 int sysctl_drop_packet;
903 int sysctl_secure_tcp;
904#ifdef CONFIG_IP_VS_NFCT
905 int sysctl_conntrack;
906#endif
907 int sysctl_snat_reroute;
908 int sysctl_sync_ver;
909 int sysctl_sync_ports;
910 int sysctl_sync_persist_mode;
911 unsigned long sysctl_sync_qlen_max;
912 int sysctl_sync_sock_size;
913 int sysctl_cache_bypass;
914 int sysctl_expire_nodest_conn;
915 int sysctl_sloppy_tcp;
916 int sysctl_sloppy_sctp;
917 int sysctl_expire_quiescent_template;
918 int sysctl_sync_threshold[2];
919 unsigned int sysctl_sync_refresh_period;
920 int sysctl_sync_retries;
921 int sysctl_nat_icmp_send;
922 int sysctl_pmtu_disc;
923 int sysctl_backup_only;
924 int sysctl_conn_reuse_mode;
925 int sysctl_schedule_icmp;
926 int sysctl_ignore_tunneled;
927
928 /* ip_vs_lblc */
929 int sysctl_lblc_expiration;
930 struct ctl_table_header *lblc_ctl_header;
931 struct ctl_table *lblc_ctl_table;
932 /* ip_vs_lblcr */
933 int sysctl_lblcr_expiration;
934 struct ctl_table_header *lblcr_ctl_header;
935 struct ctl_table *lblcr_ctl_table;
936 /* ip_vs_est */
937 struct list_head est_list; /* estimator list */
938 spinlock_t est_lock;
939 struct timer_list est_timer; /* Estimation timer */
940 /* ip_vs_sync */
941 spinlock_t sync_lock;
942 struct ipvs_master_sync_state *ms;
943 spinlock_t sync_buff_lock;
944 struct task_struct **backup_threads;
945 int threads_mask;
946 volatile int sync_state;
947 struct mutex sync_mutex;
948 struct ipvs_sync_daemon_cfg mcfg; /* Master Configuration */
949 struct ipvs_sync_daemon_cfg bcfg; /* Backup Configuration */
950 /* net name space ptr */
951 struct net *net; /* Needed by timer routines */
952 /* Number of heterogeneous destinations, needed becaus heterogeneous
953 * are not supported when synchronization is enabled.
954 */
955 unsigned int mixed_address_family_dests;
956};
957
958#define DEFAULT_SYNC_THRESHOLD 3
959#define DEFAULT_SYNC_PERIOD 50
960#define DEFAULT_SYNC_VER 1
961#define DEFAULT_SLOPPY_TCP 0
962#define DEFAULT_SLOPPY_SCTP 0
963#define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ)
964#define DEFAULT_SYNC_RETRIES 0
965#define IPVS_SYNC_WAKEUP_RATE 8
966#define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4)
967#define IPVS_SYNC_SEND_DELAY (HZ / 50)
968#define IPVS_SYNC_CHECK_PERIOD HZ
969#define IPVS_SYNC_FLUSH_TIME (HZ * 2)
970#define IPVS_SYNC_PORTS_MAX (1 << 6)
971
972#ifdef CONFIG_SYSCTL
973
974static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
975{
976 return ipvs->sysctl_sync_threshold[0];
977}
978
979static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
980{
981 return READ_ONCE(ipvs->sysctl_sync_threshold[1]);
982}
983
984static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
985{
986 return READ_ONCE(ipvs->sysctl_sync_refresh_period);
987}
988
989static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
990{
991 return ipvs->sysctl_sync_retries;
992}
993
994static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
995{
996 return ipvs->sysctl_sync_ver;
997}
998
999static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1000{
1001 return ipvs->sysctl_sloppy_tcp;
1002}
1003
1004static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1005{
1006 return ipvs->sysctl_sloppy_sctp;
1007}
1008
1009static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1010{
1011 return READ_ONCE(ipvs->sysctl_sync_ports);
1012}
1013
1014static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1015{
1016 return ipvs->sysctl_sync_persist_mode;
1017}
1018
1019static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1020{
1021 return ipvs->sysctl_sync_qlen_max;
1022}
1023
1024static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1025{
1026 return ipvs->sysctl_sync_sock_size;
1027}
1028
1029static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1030{
1031 return ipvs->sysctl_pmtu_disc;
1032}
1033
1034static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1035{
1036 return ipvs->sync_state & IP_VS_STATE_BACKUP &&
1037 ipvs->sysctl_backup_only;
1038}
1039
1040static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1041{
1042 return ipvs->sysctl_conn_reuse_mode;
1043}
1044
1045static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1046{
1047 return ipvs->sysctl_schedule_icmp;
1048}
1049
1050static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1051{
1052 return ipvs->sysctl_ignore_tunneled;
1053}
1054
1055static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1056{
1057 return ipvs->sysctl_cache_bypass;
1058}
1059
1060#else
1061
1062static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1063{
1064 return DEFAULT_SYNC_THRESHOLD;
1065}
1066
1067static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1068{
1069 return DEFAULT_SYNC_PERIOD;
1070}
1071
1072static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1073{
1074 return DEFAULT_SYNC_REFRESH_PERIOD;
1075}
1076
1077static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1078{
1079 return DEFAULT_SYNC_RETRIES & 3;
1080}
1081
1082static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1083{
1084 return DEFAULT_SYNC_VER;
1085}
1086
1087static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1088{
1089 return DEFAULT_SLOPPY_TCP;
1090}
1091
1092static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1093{
1094 return DEFAULT_SLOPPY_SCTP;
1095}
1096
1097static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1098{
1099 return 1;
1100}
1101
1102static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1103{
1104 return 0;
1105}
1106
1107static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1108{
1109 return IPVS_SYNC_QLEN_MAX;
1110}
1111
1112static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1113{
1114 return 0;
1115}
1116
1117static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1118{
1119 return 1;
1120}
1121
1122static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1123{
1124 return 0;
1125}
1126
1127static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1128{
1129 return 1;
1130}
1131
1132static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1133{
1134 return 0;
1135}
1136
1137static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1138{
1139 return 0;
1140}
1141
1142static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1143{
1144 return 0;
1145}
1146
1147#endif
1148
1149/* IPVS core functions
1150 * (from ip_vs_core.c)
1151 */
1152const char *ip_vs_proto_name(unsigned int proto);
1153void ip_vs_init_hash_table(struct list_head *table, int rows);
1154struct ip_vs_conn *ip_vs_new_conn_out(struct ip_vs_service *svc,
1155 struct ip_vs_dest *dest,
1156 struct sk_buff *skb,
1157 const struct ip_vs_iphdr *iph,
1158 __be16 dport,
1159 __be16 cport);
1160#define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t)))
1161
1162#define IP_VS_APP_TYPE_FTP 1
1163
1164/* ip_vs_conn handling functions
1165 * (from ip_vs_conn.c)
1166 */
1167enum {
1168 IP_VS_DIR_INPUT = 0,
1169 IP_VS_DIR_OUTPUT,
1170 IP_VS_DIR_INPUT_ONLY,
1171 IP_VS_DIR_LAST,
1172};
1173
1174static inline void ip_vs_conn_fill_param(struct netns_ipvs *ipvs, int af, int protocol,
1175 const union nf_inet_addr *caddr,
1176 __be16 cport,
1177 const union nf_inet_addr *vaddr,
1178 __be16 vport,
1179 struct ip_vs_conn_param *p)
1180{
1181 p->ipvs = ipvs;
1182 p->af = af;
1183 p->protocol = protocol;
1184 p->caddr = caddr;
1185 p->cport = cport;
1186 p->vaddr = vaddr;
1187 p->vport = vport;
1188 p->pe = NULL;
1189 p->pe_data = NULL;
1190}
1191
1192struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p);
1193struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p);
1194
1195struct ip_vs_conn * ip_vs_conn_in_get_proto(struct netns_ipvs *ipvs, int af,
1196 const struct sk_buff *skb,
1197 const struct ip_vs_iphdr *iph);
1198
1199struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p);
1200
1201struct ip_vs_conn * ip_vs_conn_out_get_proto(struct netns_ipvs *ipvs, int af,
1202 const struct sk_buff *skb,
1203 const struct ip_vs_iphdr *iph);
1204
1205/* Get reference to gain full access to conn.
1206 * By default, RCU read-side critical sections have access only to
1207 * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference.
1208 */
1209static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp)
1210{
1211 return refcount_inc_not_zero(&cp->refcnt);
1212}
1213
1214/* put back the conn without restarting its timer */
1215static inline void __ip_vs_conn_put(struct ip_vs_conn *cp)
1216{
1217 smp_mb__before_atomic();
1218 refcount_dec(&cp->refcnt);
1219}
1220void ip_vs_conn_put(struct ip_vs_conn *cp);
1221void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport);
1222
1223struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, int dest_af,
1224 const union nf_inet_addr *daddr,
1225 __be16 dport, unsigned int flags,
1226 struct ip_vs_dest *dest, __u32 fwmark);
1227void ip_vs_conn_expire_now(struct ip_vs_conn *cp);
1228
1229const char *ip_vs_state_name(const struct ip_vs_conn *cp);
1230
1231void ip_vs_tcp_conn_listen(struct ip_vs_conn *cp);
1232int ip_vs_check_template(struct ip_vs_conn *ct, struct ip_vs_dest *cdest);
1233void ip_vs_random_dropentry(struct netns_ipvs *ipvs);
1234int ip_vs_conn_init(void);
1235void ip_vs_conn_cleanup(void);
1236
1237static inline void ip_vs_control_del(struct ip_vs_conn *cp)
1238{
1239 struct ip_vs_conn *ctl_cp = cp->control;
1240 if (!ctl_cp) {
1241 IP_VS_ERR_BUF("request control DEL for uncontrolled: "
1242 "%s:%d to %s:%d\n",
1243 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1244 ntohs(cp->cport),
1245 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1246 ntohs(cp->vport));
1247
1248 return;
1249 }
1250
1251 IP_VS_DBG_BUF(7, "DELeting control for: "
1252 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1253 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1254 ntohs(cp->cport),
1255 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1256 ntohs(ctl_cp->cport));
1257
1258 cp->control = NULL;
1259 if (atomic_read(&ctl_cp->n_control) == 0) {
1260 IP_VS_ERR_BUF("BUG control DEL with n=0 : "
1261 "%s:%d to %s:%d\n",
1262 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1263 ntohs(cp->cport),
1264 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1265 ntohs(cp->vport));
1266
1267 return;
1268 }
1269 atomic_dec(&ctl_cp->n_control);
1270}
1271
1272static inline void
1273ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp)
1274{
1275 if (cp->control) {
1276 IP_VS_ERR_BUF("request control ADD for already controlled: "
1277 "%s:%d to %s:%d\n",
1278 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1279 ntohs(cp->cport),
1280 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1281 ntohs(cp->vport));
1282
1283 ip_vs_control_del(cp);
1284 }
1285
1286 IP_VS_DBG_BUF(7, "ADDing control for: "
1287 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1288 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1289 ntohs(cp->cport),
1290 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1291 ntohs(ctl_cp->cport));
1292
1293 cp->control = ctl_cp;
1294 atomic_inc(&ctl_cp->n_control);
1295}
1296
1297/* Mark our template as assured */
1298static inline void
1299ip_vs_control_assure_ct(struct ip_vs_conn *cp)
1300{
1301 struct ip_vs_conn *ct = cp->control;
1302
1303 if (ct && !(ct->state & IP_VS_CTPL_S_ASSURED) &&
1304 (ct->flags & IP_VS_CONN_F_TEMPLATE))
1305 ct->state |= IP_VS_CTPL_S_ASSURED;
1306}
1307
1308/* IPVS netns init & cleanup functions */
1309int ip_vs_estimator_net_init(struct netns_ipvs *ipvs);
1310int ip_vs_control_net_init(struct netns_ipvs *ipvs);
1311int ip_vs_protocol_net_init(struct netns_ipvs *ipvs);
1312int ip_vs_app_net_init(struct netns_ipvs *ipvs);
1313int ip_vs_conn_net_init(struct netns_ipvs *ipvs);
1314int ip_vs_sync_net_init(struct netns_ipvs *ipvs);
1315void ip_vs_conn_net_cleanup(struct netns_ipvs *ipvs);
1316void ip_vs_app_net_cleanup(struct netns_ipvs *ipvs);
1317void ip_vs_protocol_net_cleanup(struct netns_ipvs *ipvs);
1318void ip_vs_control_net_cleanup(struct netns_ipvs *ipvs);
1319void ip_vs_estimator_net_cleanup(struct netns_ipvs *ipvs);
1320void ip_vs_sync_net_cleanup(struct netns_ipvs *ipvs);
1321void ip_vs_service_net_cleanup(struct netns_ipvs *ipvs);
1322
1323/* IPVS application functions
1324 * (from ip_vs_app.c)
1325 */
1326#define IP_VS_APP_MAX_PORTS 8
1327struct ip_vs_app *register_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1328void unregister_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1329int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1330void ip_vs_unbind_app(struct ip_vs_conn *cp);
1331int register_ip_vs_app_inc(struct netns_ipvs *ipvs, struct ip_vs_app *app, __u16 proto,
1332 __u16 port);
1333int ip_vs_app_inc_get(struct ip_vs_app *inc);
1334void ip_vs_app_inc_put(struct ip_vs_app *inc);
1335
1336int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb,
1337 struct ip_vs_iphdr *ipvsh);
1338int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb,
1339 struct ip_vs_iphdr *ipvsh);
1340
1341int register_ip_vs_pe(struct ip_vs_pe *pe);
1342int unregister_ip_vs_pe(struct ip_vs_pe *pe);
1343struct ip_vs_pe *ip_vs_pe_getbyname(const char *name);
1344struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name);
1345
1346/* Use a #define to avoid all of module.h just for these trivial ops */
1347#define ip_vs_pe_get(pe) \
1348 if (pe && pe->module) \
1349 __module_get(pe->module);
1350
1351#define ip_vs_pe_put(pe) \
1352 if (pe && pe->module) \
1353 module_put(pe->module);
1354
1355/* IPVS protocol functions (from ip_vs_proto.c) */
1356int ip_vs_protocol_init(void);
1357void ip_vs_protocol_cleanup(void);
1358void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags);
1359int *ip_vs_create_timeout_table(int *table, int size);
1360void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
1361 const struct sk_buff *skb, int offset,
1362 const char *msg);
1363
1364extern struct ip_vs_protocol ip_vs_protocol_tcp;
1365extern struct ip_vs_protocol ip_vs_protocol_udp;
1366extern struct ip_vs_protocol ip_vs_protocol_icmp;
1367extern struct ip_vs_protocol ip_vs_protocol_esp;
1368extern struct ip_vs_protocol ip_vs_protocol_ah;
1369extern struct ip_vs_protocol ip_vs_protocol_sctp;
1370
1371/* Registering/unregistering scheduler functions
1372 * (from ip_vs_sched.c)
1373 */
1374int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1375int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1376int ip_vs_bind_scheduler(struct ip_vs_service *svc,
1377 struct ip_vs_scheduler *scheduler);
1378void ip_vs_unbind_scheduler(struct ip_vs_service *svc,
1379 struct ip_vs_scheduler *sched);
1380struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name);
1381void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler);
1382struct ip_vs_conn *
1383ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb,
1384 struct ip_vs_proto_data *pd, int *ignored,
1385 struct ip_vs_iphdr *iph);
1386int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb,
1387 struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph);
1388
1389void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg);
1390
1391/* IPVS control data and functions (from ip_vs_ctl.c) */
1392extern struct ip_vs_stats ip_vs_stats;
1393extern int sysctl_ip_vs_sync_ver;
1394
1395struct ip_vs_service *
1396ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
1397 const union nf_inet_addr *vaddr, __be16 vport);
1398
1399bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1400 const union nf_inet_addr *daddr, __be16 dport);
1401
1402struct ip_vs_dest *
1403ip_vs_find_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1404 const union nf_inet_addr *daddr, __be16 dport);
1405
1406int ip_vs_use_count_inc(void);
1407void ip_vs_use_count_dec(void);
1408int ip_vs_register_nl_ioctl(void);
1409void ip_vs_unregister_nl_ioctl(void);
1410int ip_vs_control_init(void);
1411void ip_vs_control_cleanup(void);
1412struct ip_vs_dest *
1413ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
1414 const union nf_inet_addr *daddr, __be16 dport,
1415 const union nf_inet_addr *vaddr, __be16 vport,
1416 __u16 protocol, __u32 fwmark, __u32 flags);
1417void ip_vs_try_bind_dest(struct ip_vs_conn *cp);
1418
1419static inline void ip_vs_dest_hold(struct ip_vs_dest *dest)
1420{
1421 refcount_inc(&dest->refcnt);
1422}
1423
1424static inline void ip_vs_dest_put(struct ip_vs_dest *dest)
1425{
1426 smp_mb__before_atomic();
1427 refcount_dec(&dest->refcnt);
1428}
1429
1430static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest)
1431{
1432 if (refcount_dec_and_test(&dest->refcnt))
1433 kfree(dest);
1434}
1435
1436/* IPVS sync daemon data and function prototypes
1437 * (from ip_vs_sync.c)
1438 */
1439int start_sync_thread(struct netns_ipvs *ipvs, struct ipvs_sync_daemon_cfg *cfg,
1440 int state);
1441int stop_sync_thread(struct netns_ipvs *ipvs, int state);
1442void ip_vs_sync_conn(struct netns_ipvs *ipvs, struct ip_vs_conn *cp, int pkts);
1443
1444/* IPVS rate estimator prototypes (from ip_vs_est.c) */
1445void ip_vs_start_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1446void ip_vs_stop_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1447void ip_vs_zero_estimator(struct ip_vs_stats *stats);
1448void ip_vs_read_estimator(struct ip_vs_kstats *dst, struct ip_vs_stats *stats);
1449
1450/* Various IPVS packet transmitters (from ip_vs_xmit.c) */
1451int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1452 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1453int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1454 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1455int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1456 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1457int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1458 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1459int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1460 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1461int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1462 struct ip_vs_protocol *pp, int offset,
1463 unsigned int hooknum, struct ip_vs_iphdr *iph);
1464void ip_vs_dest_dst_rcu_free(struct rcu_head *head);
1465
1466#ifdef CONFIG_IP_VS_IPV6
1467int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1468 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1469int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1470 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1471int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1472 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1473int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1474 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1475int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1476 struct ip_vs_protocol *pp, int offset,
1477 unsigned int hooknum, struct ip_vs_iphdr *iph);
1478#endif
1479
1480#ifdef CONFIG_SYSCTL
1481/* This is a simple mechanism to ignore packets when
1482 * we are loaded. Just set ip_vs_drop_rate to 'n' and
1483 * we start to drop 1/rate of the packets
1484 */
1485static inline int ip_vs_todrop(struct netns_ipvs *ipvs)
1486{
1487 if (!ipvs->drop_rate)
1488 return 0;
1489 if (--ipvs->drop_counter > 0)
1490 return 0;
1491 ipvs->drop_counter = ipvs->drop_rate;
1492 return 1;
1493}
1494#else
1495static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; }
1496#endif
1497
1498/* ip_vs_fwd_tag returns the forwarding tag of the connection */
1499#define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK)
1500
1501static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp)
1502{
1503 char fwd;
1504
1505 switch (IP_VS_FWD_METHOD(cp)) {
1506 case IP_VS_CONN_F_MASQ:
1507 fwd = 'M'; break;
1508 case IP_VS_CONN_F_LOCALNODE:
1509 fwd = 'L'; break;
1510 case IP_VS_CONN_F_TUNNEL:
1511 fwd = 'T'; break;
1512 case IP_VS_CONN_F_DROUTE:
1513 fwd = 'R'; break;
1514 case IP_VS_CONN_F_BYPASS:
1515 fwd = 'B'; break;
1516 default:
1517 fwd = '?'; break;
1518 }
1519 return fwd;
1520}
1521
1522void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp,
1523 struct ip_vs_conn *cp, int dir);
1524
1525#ifdef CONFIG_IP_VS_IPV6
1526void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp,
1527 struct ip_vs_conn *cp, int dir);
1528#endif
1529
1530__sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset);
1531
1532static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum)
1533{
1534 __be32 diff[2] = { ~old, new };
1535
1536 return csum_partial(diff, sizeof(diff), oldsum);
1537}
1538
1539#ifdef CONFIG_IP_VS_IPV6
1540static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new,
1541 __wsum oldsum)
1542{
1543 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0],
1544 new[3], new[2], new[1], new[0] };
1545
1546 return csum_partial(diff, sizeof(diff), oldsum);
1547}
1548#endif
1549
1550static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum)
1551{
1552 __be16 diff[2] = { ~old, new };
1553
1554 return csum_partial(diff, sizeof(diff), oldsum);
1555}
1556
1557/* Forget current conntrack (unconfirmed) and attach notrack entry */
1558static inline void ip_vs_notrack(struct sk_buff *skb)
1559{
1560#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1561 enum ip_conntrack_info ctinfo;
1562 struct nf_conn *ct = nf_ct_get(skb, &ctinfo);
1563
1564 if (ct) {
1565 nf_conntrack_put(&ct->ct_general);
1566 nf_ct_set(skb, NULL, IP_CT_UNTRACKED);
1567 }
1568#endif
1569}
1570
1571#ifdef CONFIG_IP_VS_NFCT
1572/* Netfilter connection tracking
1573 * (from ip_vs_nfct.c)
1574 */
1575static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1576{
1577#ifdef CONFIG_SYSCTL
1578 return ipvs->sysctl_conntrack;
1579#else
1580 return 0;
1581#endif
1582}
1583
1584void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp,
1585 int outin);
1586int ip_vs_confirm_conntrack(struct sk_buff *skb);
1587void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct,
1588 struct ip_vs_conn *cp, u_int8_t proto,
1589 const __be16 port, int from_rs);
1590void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp);
1591
1592#else
1593
1594static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1595{
1596 return 0;
1597}
1598
1599static inline void ip_vs_update_conntrack(struct sk_buff *skb,
1600 struct ip_vs_conn *cp, int outin)
1601{
1602}
1603
1604static inline int ip_vs_confirm_conntrack(struct sk_buff *skb)
1605{
1606 return NF_ACCEPT;
1607}
1608
1609static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp)
1610{
1611}
1612#endif /* CONFIG_IP_VS_NFCT */
1613
1614/* Really using conntrack? */
1615static inline bool ip_vs_conn_uses_conntrack(struct ip_vs_conn *cp,
1616 struct sk_buff *skb)
1617{
1618#ifdef CONFIG_IP_VS_NFCT
1619 enum ip_conntrack_info ctinfo;
1620 struct nf_conn *ct;
1621
1622 if (!(cp->flags & IP_VS_CONN_F_NFCT))
1623 return false;
1624 ct = nf_ct_get(skb, &ctinfo);
1625 if (ct)
1626 return true;
1627#endif
1628 return false;
1629}
1630
1631static inline int ip_vs_register_conntrack(struct ip_vs_service *svc)
1632{
1633#if IS_ENABLED(CONFIG_NF_CONNTRACK)
1634 int afmask = (svc->af == AF_INET6) ? 2 : 1;
1635 int ret = 0;
1636
1637 if (!(svc->conntrack_afmask & afmask)) {
1638 ret = nf_ct_netns_get(svc->ipvs->net, svc->af);
1639 if (ret >= 0)
1640 svc->conntrack_afmask |= afmask;
1641 }
1642 return ret;
1643#else
1644 return 0;
1645#endif
1646}
1647
1648static inline void ip_vs_unregister_conntrack(struct ip_vs_service *svc)
1649{
1650#if IS_ENABLED(CONFIG_NF_CONNTRACK)
1651 int afmask = (svc->af == AF_INET6) ? 2 : 1;
1652
1653 if (svc->conntrack_afmask & afmask) {
1654 nf_ct_netns_put(svc->ipvs->net, svc->af);
1655 svc->conntrack_afmask &= ~afmask;
1656 }
1657#endif
1658}
1659
1660static inline int
1661ip_vs_dest_conn_overhead(struct ip_vs_dest *dest)
1662{
1663 /* We think the overhead of processing active connections is 256
1664 * times higher than that of inactive connections in average. (This
1665 * 256 times might not be accurate, we will change it later) We
1666 * use the following formula to estimate the overhead now:
1667 * dest->activeconns*256 + dest->inactconns
1668 */
1669 return (atomic_read(&dest->activeconns) << 8) +
1670 atomic_read(&dest->inactconns);
1671}
1672
1673#endif /* _NET_IP_VS_H */