Loading...
1/* 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 const char *(*state_name)(int state);
457
458 void (*state_transition)(struct ip_vs_conn *cp, int direction,
459 const struct sk_buff *skb,
460 struct ip_vs_proto_data *pd);
461
462 int (*register_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
463
464 void (*unregister_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
465
466 int (*app_conn_bind)(struct ip_vs_conn *cp);
467
468 void (*debug_packet)(int af, struct ip_vs_protocol *pp,
469 const struct sk_buff *skb,
470 int offset,
471 const char *msg);
472
473 void (*timeout_change)(struct ip_vs_proto_data *pd, int flags);
474};
475
476/* protocol data per netns */
477struct ip_vs_proto_data {
478 struct ip_vs_proto_data *next;
479 struct ip_vs_protocol *pp;
480 int *timeout_table; /* protocol timeout table */
481 atomic_t appcnt; /* counter of proto app incs. */
482 struct tcp_states_t *tcp_state_table;
483};
484
485struct ip_vs_protocol *ip_vs_proto_get(unsigned short proto);
486struct ip_vs_proto_data *ip_vs_proto_data_get(struct netns_ipvs *ipvs,
487 unsigned short proto);
488
489struct ip_vs_conn_param {
490 struct netns_ipvs *ipvs;
491 const union nf_inet_addr *caddr;
492 const union nf_inet_addr *vaddr;
493 __be16 cport;
494 __be16 vport;
495 __u16 protocol;
496 u16 af;
497
498 const struct ip_vs_pe *pe;
499 char *pe_data;
500 __u8 pe_data_len;
501};
502
503/* IP_VS structure allocated for each dynamically scheduled connection */
504struct ip_vs_conn {
505 struct hlist_node c_list; /* hashed list heads */
506 /* Protocol, addresses and port numbers */
507 __be16 cport;
508 __be16 dport;
509 __be16 vport;
510 u16 af; /* address family */
511 union nf_inet_addr caddr; /* client address */
512 union nf_inet_addr vaddr; /* virtual address */
513 union nf_inet_addr daddr; /* destination address */
514 volatile __u32 flags; /* status flags */
515 __u16 protocol; /* Which protocol (TCP/UDP) */
516 __u16 daf; /* Address family of the dest */
517 struct netns_ipvs *ipvs;
518
519 /* counter and timer */
520 refcount_t refcnt; /* reference count */
521 struct timer_list timer; /* Expiration timer */
522 volatile unsigned long timeout; /* timeout */
523
524 /* Flags and state transition */
525 spinlock_t lock; /* lock for state transition */
526 volatile __u16 state; /* state info */
527 volatile __u16 old_state; /* old state, to be used for
528 * state transition triggerd
529 * synchronization
530 */
531 __u32 fwmark; /* Fire wall mark from skb */
532 unsigned long sync_endtime; /* jiffies + sent_retries */
533
534 /* Control members */
535 struct ip_vs_conn *control; /* Master control connection */
536 atomic_t n_control; /* Number of controlled ones */
537 struct ip_vs_dest *dest; /* real server */
538 atomic_t in_pkts; /* incoming packet counter */
539
540 /* Packet transmitter for different forwarding methods. If it
541 * mangles the packet, it must return NF_DROP or better NF_STOLEN,
542 * otherwise this must be changed to a sk_buff **.
543 * NF_ACCEPT can be returned when destination is local.
544 */
545 int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp,
546 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
547
548 /* Note: we can group the following members into a structure,
549 * in order to save more space, and the following members are
550 * only used in VS/NAT anyway
551 */
552 struct ip_vs_app *app; /* bound ip_vs_app object */
553 void *app_data; /* Application private data */
554 struct ip_vs_seq in_seq; /* incoming seq. struct */
555 struct ip_vs_seq out_seq; /* outgoing seq. struct */
556
557 const struct ip_vs_pe *pe;
558 char *pe_data;
559 __u8 pe_data_len;
560
561 struct rcu_head rcu_head;
562};
563
564/* Extended internal versions of struct ip_vs_service_user and ip_vs_dest_user
565 * for IPv6 support.
566 *
567 * We need these to conveniently pass around service and destination
568 * options, but unfortunately, we also need to keep the old definitions to
569 * maintain userspace backwards compatibility for the setsockopt interface.
570 */
571struct ip_vs_service_user_kern {
572 /* virtual service addresses */
573 u16 af;
574 u16 protocol;
575 union nf_inet_addr addr; /* virtual ip address */
576 __be16 port;
577 u32 fwmark; /* firwall mark of service */
578
579 /* virtual service options */
580 char *sched_name;
581 char *pe_name;
582 unsigned int flags; /* virtual service flags */
583 unsigned int timeout; /* persistent timeout in sec */
584 __be32 netmask; /* persistent netmask or plen */
585};
586
587
588struct ip_vs_dest_user_kern {
589 /* destination server address */
590 union nf_inet_addr addr;
591 __be16 port;
592
593 /* real server options */
594 unsigned int conn_flags; /* connection flags */
595 int weight; /* destination weight */
596
597 /* thresholds for active connections */
598 u32 u_threshold; /* upper threshold */
599 u32 l_threshold; /* lower threshold */
600
601 /* Address family of addr */
602 u16 af;
603
604 u16 tun_type; /* tunnel type */
605 __be16 tun_port; /* tunnel port */
606 u16 tun_flags; /* tunnel flags */
607};
608
609
610/*
611 * The information about the virtual service offered to the net and the
612 * forwarding entries.
613 */
614struct ip_vs_service {
615 struct hlist_node s_list; /* for normal service table */
616 struct hlist_node f_list; /* for fwmark-based service table */
617 atomic_t refcnt; /* reference counter */
618
619 u16 af; /* address family */
620 __u16 protocol; /* which protocol (TCP/UDP) */
621 union nf_inet_addr addr; /* IP address for virtual service */
622 __be16 port; /* port number for the service */
623 __u32 fwmark; /* firewall mark of the service */
624 unsigned int flags; /* service status flags */
625 unsigned int timeout; /* persistent timeout in ticks */
626 __be32 netmask; /* grouping granularity, mask/plen */
627 struct netns_ipvs *ipvs;
628
629 struct list_head destinations; /* real server d-linked list */
630 __u32 num_dests; /* number of servers */
631 struct ip_vs_stats stats; /* statistics for the service */
632
633 /* for scheduling */
634 struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */
635 spinlock_t sched_lock; /* lock sched_data */
636 void *sched_data; /* scheduler application data */
637
638 /* alternate persistence engine */
639 struct ip_vs_pe __rcu *pe;
640 int conntrack_afmask;
641
642 struct rcu_head rcu_head;
643};
644
645/* Information for cached dst */
646struct ip_vs_dest_dst {
647 struct dst_entry *dst_cache; /* destination cache entry */
648 u32 dst_cookie;
649 union nf_inet_addr dst_saddr;
650 struct rcu_head rcu_head;
651};
652
653/* The real server destination forwarding entry with ip address, port number,
654 * and so on.
655 */
656struct ip_vs_dest {
657 struct list_head n_list; /* for the dests in the service */
658 struct hlist_node d_list; /* for table with all the dests */
659
660 u16 af; /* address family */
661 __be16 port; /* port number of the server */
662 union nf_inet_addr addr; /* IP address of the server */
663 volatile unsigned int flags; /* dest status flags */
664 atomic_t conn_flags; /* flags to copy to conn */
665 atomic_t weight; /* server weight */
666 atomic_t last_weight; /* server latest weight */
667 __u16 tun_type; /* tunnel type */
668 __be16 tun_port; /* tunnel port */
669 __u16 tun_flags; /* tunnel flags */
670
671 refcount_t refcnt; /* reference counter */
672 struct ip_vs_stats stats; /* statistics */
673 unsigned long idle_start; /* start time, jiffies */
674
675 /* connection counters and thresholds */
676 atomic_t activeconns; /* active connections */
677 atomic_t inactconns; /* inactive connections */
678 atomic_t persistconns; /* persistent connections */
679 __u32 u_threshold; /* upper threshold */
680 __u32 l_threshold; /* lower threshold */
681
682 /* for destination cache */
683 spinlock_t dst_lock; /* lock of dst_cache */
684 struct ip_vs_dest_dst __rcu *dest_dst; /* cached dst info */
685
686 /* for virtual service */
687 struct ip_vs_service __rcu *svc; /* service it belongs to */
688 __u16 protocol; /* which protocol (TCP/UDP) */
689 __be16 vport; /* virtual port number */
690 union nf_inet_addr vaddr; /* virtual IP address */
691 __u32 vfwmark; /* firewall mark of service */
692
693 struct list_head t_list; /* in dest_trash */
694 unsigned int in_rs_table:1; /* we are in rs_table */
695};
696
697/* The scheduler object */
698struct ip_vs_scheduler {
699 struct list_head n_list; /* d-linked list head */
700 char *name; /* scheduler name */
701 atomic_t refcnt; /* reference counter */
702 struct module *module; /* THIS_MODULE/NULL */
703
704 /* scheduler initializing service */
705 int (*init_service)(struct ip_vs_service *svc);
706 /* scheduling service finish */
707 void (*done_service)(struct ip_vs_service *svc);
708 /* dest is linked */
709 int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
710 /* dest is unlinked */
711 int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
712 /* dest is updated */
713 int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
714
715 /* selecting a server from the given service */
716 struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc,
717 const struct sk_buff *skb,
718 struct ip_vs_iphdr *iph);
719};
720
721/* The persistence engine object */
722struct ip_vs_pe {
723 struct list_head n_list; /* d-linked list head */
724 char *name; /* scheduler name */
725 atomic_t refcnt; /* reference counter */
726 struct module *module; /* THIS_MODULE/NULL */
727
728 /* get the connection template, if any */
729 int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb);
730 bool (*ct_match)(const struct ip_vs_conn_param *p,
731 struct ip_vs_conn *ct);
732 u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval,
733 bool inverse);
734 int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf);
735 /* create connections for real-server outgoing packets */
736 struct ip_vs_conn* (*conn_out)(struct ip_vs_service *svc,
737 struct ip_vs_dest *dest,
738 struct sk_buff *skb,
739 const struct ip_vs_iphdr *iph,
740 __be16 dport, __be16 cport);
741};
742
743/* The application module object (a.k.a. app incarnation) */
744struct ip_vs_app {
745 struct list_head a_list; /* member in app list */
746 int type; /* IP_VS_APP_TYPE_xxx */
747 char *name; /* application module name */
748 __u16 protocol;
749 struct module *module; /* THIS_MODULE/NULL */
750 struct list_head incs_list; /* list of incarnations */
751
752 /* members for application incarnations */
753 struct list_head p_list; /* member in proto app list */
754 struct ip_vs_app *app; /* its real application */
755 __be16 port; /* port number in net order */
756 atomic_t usecnt; /* usage counter */
757 struct rcu_head rcu_head;
758
759 /* output hook: Process packet in inout 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_out)(struct ip_vs_app *, struct ip_vs_conn *,
764 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh);
765
766 /* input hook: Process packet in outin direction, diff set for TCP.
767 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
768 * 2=Mangled but checksum was not updated
769 */
770 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *,
771 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh);
772
773 /* ip_vs_app initializer */
774 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *);
775
776 /* ip_vs_app finish */
777 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *);
778
779
780 /* not used now */
781 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *,
782 struct ip_vs_protocol *);
783
784 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *);
785
786 int * timeout_table;
787 int * timeouts;
788 int timeouts_size;
789
790 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app,
791 int *verdict, struct ip_vs_conn **cpp);
792
793 struct ip_vs_conn *
794 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app,
795 const struct iphdr *iph, int inverse);
796
797 struct ip_vs_conn *
798 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app,
799 const struct iphdr *iph, int inverse);
800
801 int (*state_transition)(struct ip_vs_conn *cp, int direction,
802 const struct sk_buff *skb,
803 struct ip_vs_app *app);
804
805 void (*timeout_change)(struct ip_vs_app *app, int flags);
806};
807
808struct ipvs_master_sync_state {
809 struct list_head sync_queue;
810 struct ip_vs_sync_buff *sync_buff;
811 unsigned long sync_queue_len;
812 unsigned int sync_queue_delay;
813 struct delayed_work master_wakeup_work;
814 struct netns_ipvs *ipvs;
815};
816
817struct ip_vs_sync_thread_data;
818
819/* How much time to keep dests in trash */
820#define IP_VS_DEST_TRASH_PERIOD (120 * HZ)
821
822struct ipvs_sync_daemon_cfg {
823 union nf_inet_addr mcast_group;
824 int syncid;
825 u16 sync_maxlen;
826 u16 mcast_port;
827 u8 mcast_af;
828 u8 mcast_ttl;
829 /* multicast interface name */
830 char mcast_ifn[IP_VS_IFNAME_MAXLEN];
831};
832
833/* IPVS in network namespace */
834struct netns_ipvs {
835 int gen; /* Generation */
836 int enable; /* enable like nf_hooks do */
837 /* Hash table: for real service lookups */
838 #define IP_VS_RTAB_BITS 4
839 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
840 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
841
842 struct hlist_head rs_table[IP_VS_RTAB_SIZE];
843 /* ip_vs_app */
844 struct list_head app_list;
845 /* ip_vs_proto */
846 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */
847 struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE];
848 /* ip_vs_proto_tcp */
849#ifdef CONFIG_IP_VS_PROTO_TCP
850 #define TCP_APP_TAB_BITS 4
851 #define TCP_APP_TAB_SIZE (1 << TCP_APP_TAB_BITS)
852 #define TCP_APP_TAB_MASK (TCP_APP_TAB_SIZE - 1)
853 struct list_head tcp_apps[TCP_APP_TAB_SIZE];
854#endif
855 /* ip_vs_proto_udp */
856#ifdef CONFIG_IP_VS_PROTO_UDP
857 #define UDP_APP_TAB_BITS 4
858 #define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS)
859 #define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1)
860 struct list_head udp_apps[UDP_APP_TAB_SIZE];
861#endif
862 /* ip_vs_proto_sctp */
863#ifdef CONFIG_IP_VS_PROTO_SCTP
864 #define SCTP_APP_TAB_BITS 4
865 #define SCTP_APP_TAB_SIZE (1 << SCTP_APP_TAB_BITS)
866 #define SCTP_APP_TAB_MASK (SCTP_APP_TAB_SIZE - 1)
867 /* Hash table for SCTP application incarnations */
868 struct list_head sctp_apps[SCTP_APP_TAB_SIZE];
869#endif
870 /* ip_vs_conn */
871 atomic_t conn_count; /* connection counter */
872
873 /* ip_vs_ctl */
874 struct ip_vs_stats tot_stats; /* Statistics & est. */
875
876 int num_services; /* no of virtual services */
877
878 /* Trash for destinations */
879 struct list_head dest_trash;
880 spinlock_t dest_trash_lock;
881 struct timer_list dest_trash_timer; /* expiration timer */
882 /* Service counters */
883 atomic_t ftpsvc_counter;
884 atomic_t nullsvc_counter;
885 atomic_t conn_out_counter;
886
887#ifdef CONFIG_SYSCTL
888 /* 1/rate drop and drop-entry variables */
889 struct delayed_work defense_work; /* Work handler */
890 int drop_rate;
891 int drop_counter;
892 int old_secure_tcp;
893 atomic_t dropentry;
894 /* locks in ctl.c */
895 spinlock_t dropentry_lock; /* drop entry handling */
896 spinlock_t droppacket_lock; /* drop packet handling */
897 spinlock_t securetcp_lock; /* state and timeout tables */
898
899 /* sys-ctl struct */
900 struct ctl_table_header *sysctl_hdr;
901 struct ctl_table *sysctl_tbl;
902#endif
903
904 /* sysctl variables */
905 int sysctl_amemthresh;
906 int sysctl_am_droprate;
907 int sysctl_drop_entry;
908 int sysctl_drop_packet;
909 int sysctl_secure_tcp;
910#ifdef CONFIG_IP_VS_NFCT
911 int sysctl_conntrack;
912#endif
913 int sysctl_snat_reroute;
914 int sysctl_sync_ver;
915 int sysctl_sync_ports;
916 int sysctl_sync_persist_mode;
917 unsigned long sysctl_sync_qlen_max;
918 int sysctl_sync_sock_size;
919 int sysctl_cache_bypass;
920 int sysctl_expire_nodest_conn;
921 int sysctl_sloppy_tcp;
922 int sysctl_sloppy_sctp;
923 int sysctl_expire_quiescent_template;
924 int sysctl_sync_threshold[2];
925 unsigned int sysctl_sync_refresh_period;
926 int sysctl_sync_retries;
927 int sysctl_nat_icmp_send;
928 int sysctl_pmtu_disc;
929 int sysctl_backup_only;
930 int sysctl_conn_reuse_mode;
931 int sysctl_schedule_icmp;
932 int sysctl_ignore_tunneled;
933
934 /* ip_vs_lblc */
935 int sysctl_lblc_expiration;
936 struct ctl_table_header *lblc_ctl_header;
937 struct ctl_table *lblc_ctl_table;
938 /* ip_vs_lblcr */
939 int sysctl_lblcr_expiration;
940 struct ctl_table_header *lblcr_ctl_header;
941 struct ctl_table *lblcr_ctl_table;
942 /* ip_vs_est */
943 struct list_head est_list; /* estimator list */
944 spinlock_t est_lock;
945 struct timer_list est_timer; /* Estimation timer */
946 /* ip_vs_sync */
947 spinlock_t sync_lock;
948 struct ipvs_master_sync_state *ms;
949 spinlock_t sync_buff_lock;
950 struct ip_vs_sync_thread_data *master_tinfo;
951 struct ip_vs_sync_thread_data *backup_tinfo;
952 int threads_mask;
953 volatile int sync_state;
954 struct mutex sync_mutex;
955 struct ipvs_sync_daemon_cfg mcfg; /* Master Configuration */
956 struct ipvs_sync_daemon_cfg bcfg; /* Backup Configuration */
957 /* net name space ptr */
958 struct net *net; /* Needed by timer routines */
959 /* Number of heterogeneous destinations, needed becaus heterogeneous
960 * are not supported when synchronization is enabled.
961 */
962 unsigned int mixed_address_family_dests;
963};
964
965#define DEFAULT_SYNC_THRESHOLD 3
966#define DEFAULT_SYNC_PERIOD 50
967#define DEFAULT_SYNC_VER 1
968#define DEFAULT_SLOPPY_TCP 0
969#define DEFAULT_SLOPPY_SCTP 0
970#define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ)
971#define DEFAULT_SYNC_RETRIES 0
972#define IPVS_SYNC_WAKEUP_RATE 8
973#define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4)
974#define IPVS_SYNC_SEND_DELAY (HZ / 50)
975#define IPVS_SYNC_CHECK_PERIOD HZ
976#define IPVS_SYNC_FLUSH_TIME (HZ * 2)
977#define IPVS_SYNC_PORTS_MAX (1 << 6)
978
979#ifdef CONFIG_SYSCTL
980
981static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
982{
983 return ipvs->sysctl_sync_threshold[0];
984}
985
986static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
987{
988 return READ_ONCE(ipvs->sysctl_sync_threshold[1]);
989}
990
991static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
992{
993 return READ_ONCE(ipvs->sysctl_sync_refresh_period);
994}
995
996static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
997{
998 return ipvs->sysctl_sync_retries;
999}
1000
1001static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1002{
1003 return ipvs->sysctl_sync_ver;
1004}
1005
1006static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1007{
1008 return ipvs->sysctl_sloppy_tcp;
1009}
1010
1011static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1012{
1013 return ipvs->sysctl_sloppy_sctp;
1014}
1015
1016static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1017{
1018 return READ_ONCE(ipvs->sysctl_sync_ports);
1019}
1020
1021static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1022{
1023 return ipvs->sysctl_sync_persist_mode;
1024}
1025
1026static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1027{
1028 return ipvs->sysctl_sync_qlen_max;
1029}
1030
1031static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1032{
1033 return ipvs->sysctl_sync_sock_size;
1034}
1035
1036static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1037{
1038 return ipvs->sysctl_pmtu_disc;
1039}
1040
1041static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1042{
1043 return ipvs->sync_state & IP_VS_STATE_BACKUP &&
1044 ipvs->sysctl_backup_only;
1045}
1046
1047static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1048{
1049 return ipvs->sysctl_conn_reuse_mode;
1050}
1051
1052static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1053{
1054 return ipvs->sysctl_schedule_icmp;
1055}
1056
1057static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1058{
1059 return ipvs->sysctl_ignore_tunneled;
1060}
1061
1062static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1063{
1064 return ipvs->sysctl_cache_bypass;
1065}
1066
1067#else
1068
1069static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1070{
1071 return DEFAULT_SYNC_THRESHOLD;
1072}
1073
1074static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1075{
1076 return DEFAULT_SYNC_PERIOD;
1077}
1078
1079static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1080{
1081 return DEFAULT_SYNC_REFRESH_PERIOD;
1082}
1083
1084static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1085{
1086 return DEFAULT_SYNC_RETRIES & 3;
1087}
1088
1089static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1090{
1091 return DEFAULT_SYNC_VER;
1092}
1093
1094static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1095{
1096 return DEFAULT_SLOPPY_TCP;
1097}
1098
1099static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1100{
1101 return DEFAULT_SLOPPY_SCTP;
1102}
1103
1104static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1105{
1106 return 1;
1107}
1108
1109static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1110{
1111 return 0;
1112}
1113
1114static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1115{
1116 return IPVS_SYNC_QLEN_MAX;
1117}
1118
1119static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1120{
1121 return 0;
1122}
1123
1124static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1125{
1126 return 1;
1127}
1128
1129static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1130{
1131 return 0;
1132}
1133
1134static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1135{
1136 return 1;
1137}
1138
1139static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1140{
1141 return 0;
1142}
1143
1144static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1145{
1146 return 0;
1147}
1148
1149static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1150{
1151 return 0;
1152}
1153
1154#endif
1155
1156/* IPVS core functions
1157 * (from ip_vs_core.c)
1158 */
1159const char *ip_vs_proto_name(unsigned int proto);
1160void ip_vs_init_hash_table(struct list_head *table, int rows);
1161struct ip_vs_conn *ip_vs_new_conn_out(struct ip_vs_service *svc,
1162 struct ip_vs_dest *dest,
1163 struct sk_buff *skb,
1164 const struct ip_vs_iphdr *iph,
1165 __be16 dport,
1166 __be16 cport);
1167#define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t)))
1168
1169#define IP_VS_APP_TYPE_FTP 1
1170
1171/* ip_vs_conn handling functions
1172 * (from ip_vs_conn.c)
1173 */
1174enum {
1175 IP_VS_DIR_INPUT = 0,
1176 IP_VS_DIR_OUTPUT,
1177 IP_VS_DIR_INPUT_ONLY,
1178 IP_VS_DIR_LAST,
1179};
1180
1181static inline void ip_vs_conn_fill_param(struct netns_ipvs *ipvs, int af, int protocol,
1182 const union nf_inet_addr *caddr,
1183 __be16 cport,
1184 const union nf_inet_addr *vaddr,
1185 __be16 vport,
1186 struct ip_vs_conn_param *p)
1187{
1188 p->ipvs = ipvs;
1189 p->af = af;
1190 p->protocol = protocol;
1191 p->caddr = caddr;
1192 p->cport = cport;
1193 p->vaddr = vaddr;
1194 p->vport = vport;
1195 p->pe = NULL;
1196 p->pe_data = NULL;
1197}
1198
1199struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p);
1200struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p);
1201
1202struct ip_vs_conn * ip_vs_conn_in_get_proto(struct netns_ipvs *ipvs, int af,
1203 const struct sk_buff *skb,
1204 const struct ip_vs_iphdr *iph);
1205
1206struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p);
1207
1208struct ip_vs_conn * ip_vs_conn_out_get_proto(struct netns_ipvs *ipvs, int af,
1209 const struct sk_buff *skb,
1210 const struct ip_vs_iphdr *iph);
1211
1212/* Get reference to gain full access to conn.
1213 * By default, RCU read-side critical sections have access only to
1214 * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference.
1215 */
1216static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp)
1217{
1218 return refcount_inc_not_zero(&cp->refcnt);
1219}
1220
1221/* put back the conn without restarting its timer */
1222static inline void __ip_vs_conn_put(struct ip_vs_conn *cp)
1223{
1224 smp_mb__before_atomic();
1225 refcount_dec(&cp->refcnt);
1226}
1227void ip_vs_conn_put(struct ip_vs_conn *cp);
1228void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport);
1229
1230struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, int dest_af,
1231 const union nf_inet_addr *daddr,
1232 __be16 dport, unsigned int flags,
1233 struct ip_vs_dest *dest, __u32 fwmark);
1234void ip_vs_conn_expire_now(struct ip_vs_conn *cp);
1235
1236const char *ip_vs_state_name(const struct ip_vs_conn *cp);
1237
1238void ip_vs_tcp_conn_listen(struct ip_vs_conn *cp);
1239int ip_vs_check_template(struct ip_vs_conn *ct, struct ip_vs_dest *cdest);
1240void ip_vs_random_dropentry(struct netns_ipvs *ipvs);
1241int ip_vs_conn_init(void);
1242void ip_vs_conn_cleanup(void);
1243
1244static inline void ip_vs_control_del(struct ip_vs_conn *cp)
1245{
1246 struct ip_vs_conn *ctl_cp = cp->control;
1247 if (!ctl_cp) {
1248 IP_VS_ERR_BUF("request control DEL for uncontrolled: "
1249 "%s:%d to %s:%d\n",
1250 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1251 ntohs(cp->cport),
1252 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1253 ntohs(cp->vport));
1254
1255 return;
1256 }
1257
1258 IP_VS_DBG_BUF(7, "DELeting control for: "
1259 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1260 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1261 ntohs(cp->cport),
1262 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1263 ntohs(ctl_cp->cport));
1264
1265 cp->control = NULL;
1266 if (atomic_read(&ctl_cp->n_control) == 0) {
1267 IP_VS_ERR_BUF("BUG control DEL with n=0 : "
1268 "%s:%d to %s:%d\n",
1269 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1270 ntohs(cp->cport),
1271 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1272 ntohs(cp->vport));
1273
1274 return;
1275 }
1276 atomic_dec(&ctl_cp->n_control);
1277}
1278
1279static inline void
1280ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp)
1281{
1282 if (cp->control) {
1283 IP_VS_ERR_BUF("request control ADD for already controlled: "
1284 "%s:%d to %s:%d\n",
1285 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1286 ntohs(cp->cport),
1287 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1288 ntohs(cp->vport));
1289
1290 ip_vs_control_del(cp);
1291 }
1292
1293 IP_VS_DBG_BUF(7, "ADDing control for: "
1294 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1295 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1296 ntohs(cp->cport),
1297 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1298 ntohs(ctl_cp->cport));
1299
1300 cp->control = ctl_cp;
1301 atomic_inc(&ctl_cp->n_control);
1302}
1303
1304/* Mark our template as assured */
1305static inline void
1306ip_vs_control_assure_ct(struct ip_vs_conn *cp)
1307{
1308 struct ip_vs_conn *ct = cp->control;
1309
1310 if (ct && !(ct->state & IP_VS_CTPL_S_ASSURED) &&
1311 (ct->flags & IP_VS_CONN_F_TEMPLATE))
1312 ct->state |= IP_VS_CTPL_S_ASSURED;
1313}
1314
1315/* IPVS netns init & cleanup functions */
1316int ip_vs_estimator_net_init(struct netns_ipvs *ipvs);
1317int ip_vs_control_net_init(struct netns_ipvs *ipvs);
1318int ip_vs_protocol_net_init(struct netns_ipvs *ipvs);
1319int ip_vs_app_net_init(struct netns_ipvs *ipvs);
1320int ip_vs_conn_net_init(struct netns_ipvs *ipvs);
1321int ip_vs_sync_net_init(struct netns_ipvs *ipvs);
1322void ip_vs_conn_net_cleanup(struct netns_ipvs *ipvs);
1323void ip_vs_app_net_cleanup(struct netns_ipvs *ipvs);
1324void ip_vs_protocol_net_cleanup(struct netns_ipvs *ipvs);
1325void ip_vs_control_net_cleanup(struct netns_ipvs *ipvs);
1326void ip_vs_estimator_net_cleanup(struct netns_ipvs *ipvs);
1327void ip_vs_sync_net_cleanup(struct netns_ipvs *ipvs);
1328void ip_vs_service_net_cleanup(struct netns_ipvs *ipvs);
1329
1330/* IPVS application functions
1331 * (from ip_vs_app.c)
1332 */
1333#define IP_VS_APP_MAX_PORTS 8
1334struct ip_vs_app *register_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1335void unregister_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1336int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1337void ip_vs_unbind_app(struct ip_vs_conn *cp);
1338int register_ip_vs_app_inc(struct netns_ipvs *ipvs, struct ip_vs_app *app, __u16 proto,
1339 __u16 port);
1340int ip_vs_app_inc_get(struct ip_vs_app *inc);
1341void ip_vs_app_inc_put(struct ip_vs_app *inc);
1342
1343int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb,
1344 struct ip_vs_iphdr *ipvsh);
1345int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb,
1346 struct ip_vs_iphdr *ipvsh);
1347
1348int register_ip_vs_pe(struct ip_vs_pe *pe);
1349int unregister_ip_vs_pe(struct ip_vs_pe *pe);
1350struct ip_vs_pe *ip_vs_pe_getbyname(const char *name);
1351struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name);
1352
1353/* Use a #define to avoid all of module.h just for these trivial ops */
1354#define ip_vs_pe_get(pe) \
1355 if (pe && pe->module) \
1356 __module_get(pe->module);
1357
1358#define ip_vs_pe_put(pe) \
1359 if (pe && pe->module) \
1360 module_put(pe->module);
1361
1362/* IPVS protocol functions (from ip_vs_proto.c) */
1363int ip_vs_protocol_init(void);
1364void ip_vs_protocol_cleanup(void);
1365void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags);
1366int *ip_vs_create_timeout_table(int *table, int size);
1367void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
1368 const struct sk_buff *skb, int offset,
1369 const char *msg);
1370
1371extern struct ip_vs_protocol ip_vs_protocol_tcp;
1372extern struct ip_vs_protocol ip_vs_protocol_udp;
1373extern struct ip_vs_protocol ip_vs_protocol_icmp;
1374extern struct ip_vs_protocol ip_vs_protocol_esp;
1375extern struct ip_vs_protocol ip_vs_protocol_ah;
1376extern struct ip_vs_protocol ip_vs_protocol_sctp;
1377
1378/* Registering/unregistering scheduler functions
1379 * (from ip_vs_sched.c)
1380 */
1381int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1382int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1383int ip_vs_bind_scheduler(struct ip_vs_service *svc,
1384 struct ip_vs_scheduler *scheduler);
1385void ip_vs_unbind_scheduler(struct ip_vs_service *svc,
1386 struct ip_vs_scheduler *sched);
1387struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name);
1388void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler);
1389struct ip_vs_conn *
1390ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb,
1391 struct ip_vs_proto_data *pd, int *ignored,
1392 struct ip_vs_iphdr *iph);
1393int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb,
1394 struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph);
1395
1396void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg);
1397
1398/* IPVS control data and functions (from ip_vs_ctl.c) */
1399extern struct ip_vs_stats ip_vs_stats;
1400extern int sysctl_ip_vs_sync_ver;
1401
1402struct ip_vs_service *
1403ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
1404 const union nf_inet_addr *vaddr, __be16 vport);
1405
1406bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1407 const union nf_inet_addr *daddr, __be16 dport);
1408
1409struct ip_vs_dest *
1410ip_vs_find_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1411 const union nf_inet_addr *daddr, __be16 dport);
1412struct ip_vs_dest *ip_vs_find_tunnel(struct netns_ipvs *ipvs, int af,
1413 const union nf_inet_addr *daddr,
1414 __be16 tun_port);
1415
1416int ip_vs_use_count_inc(void);
1417void ip_vs_use_count_dec(void);
1418int ip_vs_register_nl_ioctl(void);
1419void ip_vs_unregister_nl_ioctl(void);
1420int ip_vs_control_init(void);
1421void ip_vs_control_cleanup(void);
1422struct ip_vs_dest *
1423ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
1424 const union nf_inet_addr *daddr, __be16 dport,
1425 const union nf_inet_addr *vaddr, __be16 vport,
1426 __u16 protocol, __u32 fwmark, __u32 flags);
1427void ip_vs_try_bind_dest(struct ip_vs_conn *cp);
1428
1429static inline void ip_vs_dest_hold(struct ip_vs_dest *dest)
1430{
1431 refcount_inc(&dest->refcnt);
1432}
1433
1434static inline void ip_vs_dest_put(struct ip_vs_dest *dest)
1435{
1436 smp_mb__before_atomic();
1437 refcount_dec(&dest->refcnt);
1438}
1439
1440static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest)
1441{
1442 if (refcount_dec_and_test(&dest->refcnt))
1443 kfree(dest);
1444}
1445
1446/* IPVS sync daemon data and function prototypes
1447 * (from ip_vs_sync.c)
1448 */
1449int start_sync_thread(struct netns_ipvs *ipvs, struct ipvs_sync_daemon_cfg *cfg,
1450 int state);
1451int stop_sync_thread(struct netns_ipvs *ipvs, int state);
1452void ip_vs_sync_conn(struct netns_ipvs *ipvs, struct ip_vs_conn *cp, int pkts);
1453
1454/* IPVS rate estimator prototypes (from ip_vs_est.c) */
1455void ip_vs_start_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1456void ip_vs_stop_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1457void ip_vs_zero_estimator(struct ip_vs_stats *stats);
1458void ip_vs_read_estimator(struct ip_vs_kstats *dst, struct ip_vs_stats *stats);
1459
1460/* Various IPVS packet transmitters (from ip_vs_xmit.c) */
1461int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1462 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1463int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1464 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1465int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1466 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1467int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1468 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1469int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1470 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1471int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1472 struct ip_vs_protocol *pp, int offset,
1473 unsigned int hooknum, struct ip_vs_iphdr *iph);
1474void ip_vs_dest_dst_rcu_free(struct rcu_head *head);
1475
1476#ifdef CONFIG_IP_VS_IPV6
1477int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1478 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1479int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1480 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1481int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1482 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1483int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1484 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1485int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1486 struct ip_vs_protocol *pp, int offset,
1487 unsigned int hooknum, struct ip_vs_iphdr *iph);
1488#endif
1489
1490#ifdef CONFIG_SYSCTL
1491/* This is a simple mechanism to ignore packets when
1492 * we are loaded. Just set ip_vs_drop_rate to 'n' and
1493 * we start to drop 1/rate of the packets
1494 */
1495static inline int ip_vs_todrop(struct netns_ipvs *ipvs)
1496{
1497 if (!ipvs->drop_rate)
1498 return 0;
1499 if (--ipvs->drop_counter > 0)
1500 return 0;
1501 ipvs->drop_counter = ipvs->drop_rate;
1502 return 1;
1503}
1504#else
1505static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; }
1506#endif
1507
1508#define IP_VS_DFWD_METHOD(dest) (atomic_read(&(dest)->conn_flags) & \
1509 IP_VS_CONN_F_FWD_MASK)
1510
1511/* ip_vs_fwd_tag returns the forwarding tag of the connection */
1512#define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK)
1513
1514static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp)
1515{
1516 char fwd;
1517
1518 switch (IP_VS_FWD_METHOD(cp)) {
1519 case IP_VS_CONN_F_MASQ:
1520 fwd = 'M'; break;
1521 case IP_VS_CONN_F_LOCALNODE:
1522 fwd = 'L'; break;
1523 case IP_VS_CONN_F_TUNNEL:
1524 fwd = 'T'; break;
1525 case IP_VS_CONN_F_DROUTE:
1526 fwd = 'R'; break;
1527 case IP_VS_CONN_F_BYPASS:
1528 fwd = 'B'; break;
1529 default:
1530 fwd = '?'; break;
1531 }
1532 return fwd;
1533}
1534
1535void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp,
1536 struct ip_vs_conn *cp, int dir);
1537
1538#ifdef CONFIG_IP_VS_IPV6
1539void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp,
1540 struct ip_vs_conn *cp, int dir);
1541#endif
1542
1543__sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset);
1544
1545static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum)
1546{
1547 __be32 diff[2] = { ~old, new };
1548
1549 return csum_partial(diff, sizeof(diff), oldsum);
1550}
1551
1552#ifdef CONFIG_IP_VS_IPV6
1553static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new,
1554 __wsum oldsum)
1555{
1556 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0],
1557 new[3], new[2], new[1], new[0] };
1558
1559 return csum_partial(diff, sizeof(diff), oldsum);
1560}
1561#endif
1562
1563static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum)
1564{
1565 __be16 diff[2] = { ~old, new };
1566
1567 return csum_partial(diff, sizeof(diff), oldsum);
1568}
1569
1570/* Forget current conntrack (unconfirmed) and attach notrack entry */
1571static inline void ip_vs_notrack(struct sk_buff *skb)
1572{
1573#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1574 enum ip_conntrack_info ctinfo;
1575 struct nf_conn *ct = nf_ct_get(skb, &ctinfo);
1576
1577 if (ct) {
1578 nf_conntrack_put(&ct->ct_general);
1579 nf_ct_set(skb, NULL, IP_CT_UNTRACKED);
1580 }
1581#endif
1582}
1583
1584#ifdef CONFIG_IP_VS_NFCT
1585/* Netfilter connection tracking
1586 * (from ip_vs_nfct.c)
1587 */
1588static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1589{
1590#ifdef CONFIG_SYSCTL
1591 return ipvs->sysctl_conntrack;
1592#else
1593 return 0;
1594#endif
1595}
1596
1597void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp,
1598 int outin);
1599int ip_vs_confirm_conntrack(struct sk_buff *skb);
1600void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct,
1601 struct ip_vs_conn *cp, u_int8_t proto,
1602 const __be16 port, int from_rs);
1603void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp);
1604
1605#else
1606
1607static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1608{
1609 return 0;
1610}
1611
1612static inline void ip_vs_update_conntrack(struct sk_buff *skb,
1613 struct ip_vs_conn *cp, int outin)
1614{
1615}
1616
1617static inline int ip_vs_confirm_conntrack(struct sk_buff *skb)
1618{
1619 return NF_ACCEPT;
1620}
1621
1622static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp)
1623{
1624}
1625#endif /* CONFIG_IP_VS_NFCT */
1626
1627/* Really using conntrack? */
1628static inline bool ip_vs_conn_uses_conntrack(struct ip_vs_conn *cp,
1629 struct sk_buff *skb)
1630{
1631#ifdef CONFIG_IP_VS_NFCT
1632 enum ip_conntrack_info ctinfo;
1633 struct nf_conn *ct;
1634
1635 if (!(cp->flags & IP_VS_CONN_F_NFCT))
1636 return false;
1637 ct = nf_ct_get(skb, &ctinfo);
1638 if (ct)
1639 return true;
1640#endif
1641 return false;
1642}
1643
1644static inline int ip_vs_register_conntrack(struct ip_vs_service *svc)
1645{
1646#if IS_ENABLED(CONFIG_NF_CONNTRACK)
1647 int afmask = (svc->af == AF_INET6) ? 2 : 1;
1648 int ret = 0;
1649
1650 if (!(svc->conntrack_afmask & afmask)) {
1651 ret = nf_ct_netns_get(svc->ipvs->net, svc->af);
1652 if (ret >= 0)
1653 svc->conntrack_afmask |= afmask;
1654 }
1655 return ret;
1656#else
1657 return 0;
1658#endif
1659}
1660
1661static inline void ip_vs_unregister_conntrack(struct ip_vs_service *svc)
1662{
1663#if IS_ENABLED(CONFIG_NF_CONNTRACK)
1664 int afmask = (svc->af == AF_INET6) ? 2 : 1;
1665
1666 if (svc->conntrack_afmask & afmask) {
1667 nf_ct_netns_put(svc->ipvs->net, svc->af);
1668 svc->conntrack_afmask &= ~afmask;
1669 }
1670#endif
1671}
1672
1673static inline int
1674ip_vs_dest_conn_overhead(struct ip_vs_dest *dest)
1675{
1676 /* We think the overhead of processing active connections is 256
1677 * times higher than that of inactive connections in average. (This
1678 * 256 times might not be accurate, we will change it later) We
1679 * use the following formula to estimate the overhead now:
1680 * dest->activeconns*256 + dest->inactconns
1681 */
1682 return (atomic_read(&dest->activeconns) << 8) +
1683 atomic_read(&dest->inactconns);
1684}
1685
1686#endif /* _NET_IP_VS_H */
1/* 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#include <linux/workqueue.h>
18
19#include <linux/compiler.h>
20#include <linux/timer.h>
21#include <linux/bug.h>
22
23#include <net/checksum.h>
24#include <linux/netfilter.h> /* for union nf_inet_addr */
25#include <linux/ip.h>
26#include <linux/ipv6.h> /* for struct ipv6hdr */
27#include <net/ipv6.h>
28#if IS_ENABLED(CONFIG_NF_CONNTRACK)
29#include <net/netfilter/nf_conntrack.h>
30#endif
31#include <net/net_namespace.h> /* Netw namespace */
32#include <linux/sched/isolation.h>
33
34#define IP_VS_HDR_INVERSE 1
35#define IP_VS_HDR_ICMP 2
36
37/* Generic access of ipvs struct */
38static inline struct netns_ipvs *net_ipvs(struct net* net)
39{
40 return net->ipvs;
41}
42
43/* Connections' size value needed by ip_vs_ctl.c */
44extern int ip_vs_conn_tab_size;
45
46extern struct mutex __ip_vs_mutex;
47
48struct ip_vs_iphdr {
49 int hdr_flags; /* ipvs flags */
50 __u32 off; /* Where IP or IPv4 header starts */
51 __u32 len; /* IPv4 simply where L4 starts
52 * IPv6 where L4 Transport Header starts */
53 __u16 fragoffs; /* IPv6 fragment offset, 0 if first frag (or not frag)*/
54 __s16 protocol;
55 __s32 flags;
56 union nf_inet_addr saddr;
57 union nf_inet_addr daddr;
58};
59
60static inline void *frag_safe_skb_hp(const struct sk_buff *skb, int offset,
61 int len, void *buffer)
62{
63 return skb_header_pointer(skb, offset, len, buffer);
64}
65
66/* This function handles filling *ip_vs_iphdr, both for IPv4 and IPv6.
67 * IPv6 requires some extra work, as finding proper header position,
68 * depend on the IPv6 extension headers.
69 */
70static inline int
71ip_vs_fill_iph_skb_off(int af, const struct sk_buff *skb, int offset,
72 int hdr_flags, struct ip_vs_iphdr *iphdr)
73{
74 iphdr->hdr_flags = hdr_flags;
75 iphdr->off = offset;
76
77#ifdef CONFIG_IP_VS_IPV6
78 if (af == AF_INET6) {
79 struct ipv6hdr _iph;
80 const struct ipv6hdr *iph = skb_header_pointer(
81 skb, offset, sizeof(_iph), &_iph);
82 if (!iph)
83 return 0;
84
85 iphdr->saddr.in6 = iph->saddr;
86 iphdr->daddr.in6 = iph->daddr;
87 /* ipv6_find_hdr() updates len, flags */
88 iphdr->len = offset;
89 iphdr->flags = 0;
90 iphdr->protocol = ipv6_find_hdr(skb, &iphdr->len, -1,
91 &iphdr->fragoffs,
92 &iphdr->flags);
93 if (iphdr->protocol < 0)
94 return 0;
95 } else
96#endif
97 {
98 struct iphdr _iph;
99 const struct iphdr *iph = skb_header_pointer(
100 skb, offset, sizeof(_iph), &_iph);
101 if (!iph)
102 return 0;
103
104 iphdr->len = offset + iph->ihl * 4;
105 iphdr->fragoffs = 0;
106 iphdr->protocol = iph->protocol;
107 iphdr->saddr.ip = iph->saddr;
108 iphdr->daddr.ip = iph->daddr;
109 }
110
111 return 1;
112}
113
114static inline int
115ip_vs_fill_iph_skb_icmp(int af, const struct sk_buff *skb, int offset,
116 bool inverse, struct ip_vs_iphdr *iphdr)
117{
118 int hdr_flags = IP_VS_HDR_ICMP;
119
120 if (inverse)
121 hdr_flags |= IP_VS_HDR_INVERSE;
122
123 return ip_vs_fill_iph_skb_off(af, skb, offset, hdr_flags, iphdr);
124}
125
126static inline int
127ip_vs_fill_iph_skb(int af, const struct sk_buff *skb, bool inverse,
128 struct ip_vs_iphdr *iphdr)
129{
130 int hdr_flags = 0;
131
132 if (inverse)
133 hdr_flags |= IP_VS_HDR_INVERSE;
134
135 return ip_vs_fill_iph_skb_off(af, skb, skb_network_offset(skb),
136 hdr_flags, iphdr);
137}
138
139static inline bool
140ip_vs_iph_inverse(const struct ip_vs_iphdr *iph)
141{
142 return !!(iph->hdr_flags & IP_VS_HDR_INVERSE);
143}
144
145static inline bool
146ip_vs_iph_icmp(const struct ip_vs_iphdr *iph)
147{
148 return !!(iph->hdr_flags & IP_VS_HDR_ICMP);
149}
150
151static inline void ip_vs_addr_copy(int af, union nf_inet_addr *dst,
152 const union nf_inet_addr *src)
153{
154#ifdef CONFIG_IP_VS_IPV6
155 if (af == AF_INET6)
156 dst->in6 = src->in6;
157 else
158#endif
159 dst->ip = src->ip;
160}
161
162static inline void ip_vs_addr_set(int af, union nf_inet_addr *dst,
163 const union nf_inet_addr *src)
164{
165#ifdef CONFIG_IP_VS_IPV6
166 if (af == AF_INET6) {
167 dst->in6 = src->in6;
168 return;
169 }
170#endif
171 dst->ip = src->ip;
172 dst->all[1] = 0;
173 dst->all[2] = 0;
174 dst->all[3] = 0;
175}
176
177static inline int ip_vs_addr_equal(int af, const union nf_inet_addr *a,
178 const union nf_inet_addr *b)
179{
180#ifdef CONFIG_IP_VS_IPV6
181 if (af == AF_INET6)
182 return ipv6_addr_equal(&a->in6, &b->in6);
183#endif
184 return a->ip == b->ip;
185}
186
187#ifdef CONFIG_IP_VS_DEBUG
188#include <linux/net.h>
189
190int ip_vs_get_debug_level(void);
191
192static inline const char *ip_vs_dbg_addr(int af, char *buf, size_t buf_len,
193 const union nf_inet_addr *addr,
194 int *idx)
195{
196 int len;
197#ifdef CONFIG_IP_VS_IPV6
198 if (af == AF_INET6)
199 len = snprintf(&buf[*idx], buf_len - *idx, "[%pI6c]",
200 &addr->in6) + 1;
201 else
202#endif
203 len = snprintf(&buf[*idx], buf_len - *idx, "%pI4",
204 &addr->ip) + 1;
205
206 *idx += len;
207 BUG_ON(*idx > buf_len + 1);
208 return &buf[*idx - len];
209}
210
211#define IP_VS_DBG_BUF(level, msg, ...) \
212 do { \
213 char ip_vs_dbg_buf[160]; \
214 int ip_vs_dbg_idx = 0; \
215 if (level <= ip_vs_get_debug_level()) \
216 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
217 } while (0)
218#define IP_VS_ERR_BUF(msg...) \
219 do { \
220 char ip_vs_dbg_buf[160]; \
221 int ip_vs_dbg_idx = 0; \
222 pr_err(msg); \
223 } while (0)
224
225/* Only use from within IP_VS_DBG_BUF() or IP_VS_ERR_BUF macros */
226#define IP_VS_DBG_ADDR(af, addr) \
227 ip_vs_dbg_addr(af, ip_vs_dbg_buf, \
228 sizeof(ip_vs_dbg_buf), addr, \
229 &ip_vs_dbg_idx)
230
231#define IP_VS_DBG(level, msg, ...) \
232 do { \
233 if (level <= ip_vs_get_debug_level()) \
234 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
235 } while (0)
236#define IP_VS_DBG_RL(msg, ...) \
237 do { \
238 if (net_ratelimit()) \
239 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
240 } while (0)
241#define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) \
242 do { \
243 if (level <= ip_vs_get_debug_level()) \
244 pp->debug_packet(af, pp, skb, ofs, msg); \
245 } while (0)
246#define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) \
247 do { \
248 if (level <= ip_vs_get_debug_level() && \
249 net_ratelimit()) \
250 pp->debug_packet(af, pp, skb, ofs, msg); \
251 } while (0)
252#else /* NO DEBUGGING at ALL */
253#define IP_VS_DBG_BUF(level, msg...) do {} while (0)
254#define IP_VS_ERR_BUF(msg...) do {} while (0)
255#define IP_VS_DBG(level, msg...) do {} while (0)
256#define IP_VS_DBG_RL(msg...) do {} while (0)
257#define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
258#define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
259#endif
260
261#define IP_VS_BUG() BUG()
262#define IP_VS_ERR_RL(msg, ...) \
263 do { \
264 if (net_ratelimit()) \
265 pr_err(msg, ##__VA_ARGS__); \
266 } while (0)
267
268#ifdef CONFIG_IP_VS_DEBUG
269#define EnterFunction(level) \
270 do { \
271 if (level <= ip_vs_get_debug_level()) \
272 printk(KERN_DEBUG \
273 pr_fmt("Enter: %s, %s line %i\n"), \
274 __func__, __FILE__, __LINE__); \
275 } while (0)
276#define LeaveFunction(level) \
277 do { \
278 if (level <= ip_vs_get_debug_level()) \
279 printk(KERN_DEBUG \
280 pr_fmt("Leave: %s, %s line %i\n"), \
281 __func__, __FILE__, __LINE__); \
282 } while (0)
283#else
284#define EnterFunction(level) do {} while (0)
285#define LeaveFunction(level) do {} while (0)
286#endif
287
288/* The port number of FTP service (in network order). */
289#define FTPPORT cpu_to_be16(21)
290#define FTPDATA cpu_to_be16(20)
291
292/* TCP State Values */
293enum {
294 IP_VS_TCP_S_NONE = 0,
295 IP_VS_TCP_S_ESTABLISHED,
296 IP_VS_TCP_S_SYN_SENT,
297 IP_VS_TCP_S_SYN_RECV,
298 IP_VS_TCP_S_FIN_WAIT,
299 IP_VS_TCP_S_TIME_WAIT,
300 IP_VS_TCP_S_CLOSE,
301 IP_VS_TCP_S_CLOSE_WAIT,
302 IP_VS_TCP_S_LAST_ACK,
303 IP_VS_TCP_S_LISTEN,
304 IP_VS_TCP_S_SYNACK,
305 IP_VS_TCP_S_LAST
306};
307
308/* UDP State Values */
309enum {
310 IP_VS_UDP_S_NORMAL,
311 IP_VS_UDP_S_LAST,
312};
313
314/* ICMP State Values */
315enum {
316 IP_VS_ICMP_S_NORMAL,
317 IP_VS_ICMP_S_LAST,
318};
319
320/* SCTP State Values */
321enum ip_vs_sctp_states {
322 IP_VS_SCTP_S_NONE,
323 IP_VS_SCTP_S_INIT1,
324 IP_VS_SCTP_S_INIT,
325 IP_VS_SCTP_S_COOKIE_SENT,
326 IP_VS_SCTP_S_COOKIE_REPLIED,
327 IP_VS_SCTP_S_COOKIE_WAIT,
328 IP_VS_SCTP_S_COOKIE,
329 IP_VS_SCTP_S_COOKIE_ECHOED,
330 IP_VS_SCTP_S_ESTABLISHED,
331 IP_VS_SCTP_S_SHUTDOWN_SENT,
332 IP_VS_SCTP_S_SHUTDOWN_RECEIVED,
333 IP_VS_SCTP_S_SHUTDOWN_ACK_SENT,
334 IP_VS_SCTP_S_REJECTED,
335 IP_VS_SCTP_S_CLOSED,
336 IP_VS_SCTP_S_LAST
337};
338
339/* Connection templates use bits from state */
340#define IP_VS_CTPL_S_NONE 0x0000
341#define IP_VS_CTPL_S_ASSURED 0x0001
342#define IP_VS_CTPL_S_LAST 0x0002
343
344/* Delta sequence info structure
345 * Each ip_vs_conn has 2 (output AND input seq. changes).
346 * Only used in the VS/NAT.
347 */
348struct ip_vs_seq {
349 __u32 init_seq; /* Add delta from this seq */
350 __u32 delta; /* Delta in sequence numbers */
351 __u32 previous_delta; /* Delta in sequence numbers
352 * before last resized pkt */
353};
354
355/* counters per cpu */
356struct ip_vs_counters {
357 u64_stats_t conns; /* connections scheduled */
358 u64_stats_t inpkts; /* incoming packets */
359 u64_stats_t outpkts; /* outgoing packets */
360 u64_stats_t inbytes; /* incoming bytes */
361 u64_stats_t outbytes; /* outgoing bytes */
362};
363/* Stats per cpu */
364struct ip_vs_cpu_stats {
365 struct ip_vs_counters cnt;
366 struct u64_stats_sync syncp;
367};
368
369/* Default nice for estimator kthreads */
370#define IPVS_EST_NICE 0
371
372/* IPVS statistics objects */
373struct ip_vs_estimator {
374 struct hlist_node list;
375
376 u64 last_inbytes;
377 u64 last_outbytes;
378 u64 last_conns;
379 u64 last_inpkts;
380 u64 last_outpkts;
381
382 u64 cps;
383 u64 inpps;
384 u64 outpps;
385 u64 inbps;
386 u64 outbps;
387
388 s32 ktid:16, /* kthread ID, -1=temp list */
389 ktrow:8, /* row/tick ID for kthread */
390 ktcid:8; /* chain ID for kthread tick */
391};
392
393/*
394 * IPVS statistics object, 64-bit kernel version of struct ip_vs_stats_user
395 */
396struct ip_vs_kstats {
397 u64 conns; /* connections scheduled */
398 u64 inpkts; /* incoming packets */
399 u64 outpkts; /* outgoing packets */
400 u64 inbytes; /* incoming bytes */
401 u64 outbytes; /* outgoing bytes */
402
403 u64 cps; /* current connection rate */
404 u64 inpps; /* current in packet rate */
405 u64 outpps; /* current out packet rate */
406 u64 inbps; /* current in byte rate */
407 u64 outbps; /* current out byte rate */
408};
409
410struct ip_vs_stats {
411 struct ip_vs_kstats kstats; /* kernel statistics */
412 struct ip_vs_estimator est; /* estimator */
413 struct ip_vs_cpu_stats __percpu *cpustats; /* per cpu counters */
414 spinlock_t lock; /* spin lock */
415 struct ip_vs_kstats kstats0; /* reset values */
416};
417
418struct ip_vs_stats_rcu {
419 struct ip_vs_stats s;
420 struct rcu_head rcu_head;
421};
422
423int ip_vs_stats_init_alloc(struct ip_vs_stats *s);
424struct ip_vs_stats *ip_vs_stats_alloc(void);
425void ip_vs_stats_release(struct ip_vs_stats *stats);
426void ip_vs_stats_free(struct ip_vs_stats *stats);
427
428/* Process estimators in multiple timer ticks (20/50/100, see ktrow) */
429#define IPVS_EST_NTICKS 50
430/* Estimation uses a 2-second period containing ticks (in jiffies) */
431#define IPVS_EST_TICK ((2 * HZ) / IPVS_EST_NTICKS)
432
433/* Limit of CPU load per kthread (8 for 12.5%), ratio of CPU capacity (1/C).
434 * Value of 4 and above ensures kthreads will take work without exceeding
435 * the CPU capacity under different circumstances.
436 */
437#define IPVS_EST_LOAD_DIVISOR 8
438
439/* Kthreads should not have work that exceeds the CPU load above 50% */
440#define IPVS_EST_CPU_KTHREADS (IPVS_EST_LOAD_DIVISOR / 2)
441
442/* Desired number of chains per timer tick (chain load factor in 100us units),
443 * 48=4.8ms of 40ms tick (12% CPU usage):
444 * 2 sec * 1000 ms in sec * 10 (100us in ms) / 8 (12.5%) / 50
445 */
446#define IPVS_EST_CHAIN_FACTOR \
447 ALIGN_DOWN(2 * 1000 * 10 / IPVS_EST_LOAD_DIVISOR / IPVS_EST_NTICKS, 8)
448
449/* Compiled number of chains per tick
450 * The defines should match cond_resched_rcu
451 */
452#if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
453#define IPVS_EST_TICK_CHAINS IPVS_EST_CHAIN_FACTOR
454#else
455#define IPVS_EST_TICK_CHAINS 1
456#endif
457
458#if IPVS_EST_NTICKS > 127
459#error Too many timer ticks for ktrow
460#endif
461
462/* Multiple chains processed in same tick */
463struct ip_vs_est_tick_data {
464 struct hlist_head chains[IPVS_EST_TICK_CHAINS];
465 DECLARE_BITMAP(present, IPVS_EST_TICK_CHAINS);
466 DECLARE_BITMAP(full, IPVS_EST_TICK_CHAINS);
467 int chain_len[IPVS_EST_TICK_CHAINS];
468};
469
470/* Context for estimation kthread */
471struct ip_vs_est_kt_data {
472 struct netns_ipvs *ipvs;
473 struct task_struct *task; /* task if running */
474 struct ip_vs_est_tick_data __rcu *ticks[IPVS_EST_NTICKS];
475 DECLARE_BITMAP(avail, IPVS_EST_NTICKS); /* tick has space for ests */
476 unsigned long est_timer; /* estimation timer (jiffies) */
477 struct ip_vs_stats *calc_stats; /* Used for calculation */
478 int tick_len[IPVS_EST_NTICKS]; /* est count */
479 int id; /* ktid per netns */
480 int chain_max; /* max ests per tick chain */
481 int tick_max; /* max ests per tick */
482 int est_count; /* attached ests to kthread */
483 int est_max_count; /* max ests per kthread */
484 int add_row; /* row for new ests */
485 int est_row; /* estimated row */
486};
487
488struct dst_entry;
489struct iphdr;
490struct ip_vs_conn;
491struct ip_vs_app;
492struct sk_buff;
493struct ip_vs_proto_data;
494
495struct ip_vs_protocol {
496 struct ip_vs_protocol *next;
497 char *name;
498 u16 protocol;
499 u16 num_states;
500 int dont_defrag;
501
502 void (*init)(struct ip_vs_protocol *pp);
503
504 void (*exit)(struct ip_vs_protocol *pp);
505
506 int (*init_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
507
508 void (*exit_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
509
510 int (*conn_schedule)(struct netns_ipvs *ipvs,
511 int af, struct sk_buff *skb,
512 struct ip_vs_proto_data *pd,
513 int *verdict, struct ip_vs_conn **cpp,
514 struct ip_vs_iphdr *iph);
515
516 struct ip_vs_conn *
517 (*conn_in_get)(struct netns_ipvs *ipvs,
518 int af,
519 const struct sk_buff *skb,
520 const struct ip_vs_iphdr *iph);
521
522 struct ip_vs_conn *
523 (*conn_out_get)(struct netns_ipvs *ipvs,
524 int af,
525 const struct sk_buff *skb,
526 const struct ip_vs_iphdr *iph);
527
528 int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
529 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
530
531 int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
532 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
533
534 const char *(*state_name)(int state);
535
536 void (*state_transition)(struct ip_vs_conn *cp, int direction,
537 const struct sk_buff *skb,
538 struct ip_vs_proto_data *pd);
539
540 int (*register_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
541
542 void (*unregister_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
543
544 int (*app_conn_bind)(struct ip_vs_conn *cp);
545
546 void (*debug_packet)(int af, struct ip_vs_protocol *pp,
547 const struct sk_buff *skb,
548 int offset,
549 const char *msg);
550
551 void (*timeout_change)(struct ip_vs_proto_data *pd, int flags);
552};
553
554/* protocol data per netns */
555struct ip_vs_proto_data {
556 struct ip_vs_proto_data *next;
557 struct ip_vs_protocol *pp;
558 int *timeout_table; /* protocol timeout table */
559 atomic_t appcnt; /* counter of proto app incs. */
560 struct tcp_states_t *tcp_state_table;
561};
562
563struct ip_vs_protocol *ip_vs_proto_get(unsigned short proto);
564struct ip_vs_proto_data *ip_vs_proto_data_get(struct netns_ipvs *ipvs,
565 unsigned short proto);
566
567struct ip_vs_conn_param {
568 struct netns_ipvs *ipvs;
569 const union nf_inet_addr *caddr;
570 const union nf_inet_addr *vaddr;
571 __be16 cport;
572 __be16 vport;
573 __u16 protocol;
574 u16 af;
575
576 const struct ip_vs_pe *pe;
577 char *pe_data;
578 __u8 pe_data_len;
579};
580
581/* IP_VS structure allocated for each dynamically scheduled connection */
582struct ip_vs_conn {
583 struct hlist_node c_list; /* hashed list heads */
584 /* Protocol, addresses and port numbers */
585 __be16 cport;
586 __be16 dport;
587 __be16 vport;
588 u16 af; /* address family */
589 union nf_inet_addr caddr; /* client address */
590 union nf_inet_addr vaddr; /* virtual address */
591 union nf_inet_addr daddr; /* destination address */
592 volatile __u32 flags; /* status flags */
593 __u16 protocol; /* Which protocol (TCP/UDP) */
594 __u16 daf; /* Address family of the dest */
595 struct netns_ipvs *ipvs;
596
597 /* counter and timer */
598 refcount_t refcnt; /* reference count */
599 struct timer_list timer; /* Expiration timer */
600 volatile unsigned long timeout; /* timeout */
601
602 /* Flags and state transition */
603 spinlock_t lock; /* lock for state transition */
604 volatile __u16 state; /* state info */
605 volatile __u16 old_state; /* old state, to be used for
606 * state transition triggerd
607 * synchronization
608 */
609 __u32 fwmark; /* Fire wall mark from skb */
610 unsigned long sync_endtime; /* jiffies + sent_retries */
611
612 /* Control members */
613 struct ip_vs_conn *control; /* Master control connection */
614 atomic_t n_control; /* Number of controlled ones */
615 struct ip_vs_dest *dest; /* real server */
616 atomic_t in_pkts; /* incoming packet counter */
617
618 /* Packet transmitter for different forwarding methods. If it
619 * mangles the packet, it must return NF_DROP or better NF_STOLEN,
620 * otherwise this must be changed to a sk_buff **.
621 * NF_ACCEPT can be returned when destination is local.
622 */
623 int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp,
624 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
625
626 /* Note: we can group the following members into a structure,
627 * in order to save more space, and the following members are
628 * only used in VS/NAT anyway
629 */
630 struct ip_vs_app *app; /* bound ip_vs_app object */
631 void *app_data; /* Application private data */
632 struct ip_vs_seq in_seq; /* incoming seq. struct */
633 struct ip_vs_seq out_seq; /* outgoing seq. struct */
634
635 const struct ip_vs_pe *pe;
636 char *pe_data;
637 __u8 pe_data_len;
638
639 struct rcu_head rcu_head;
640};
641
642/* Extended internal versions of struct ip_vs_service_user and ip_vs_dest_user
643 * for IPv6 support.
644 *
645 * We need these to conveniently pass around service and destination
646 * options, but unfortunately, we also need to keep the old definitions to
647 * maintain userspace backwards compatibility for the setsockopt interface.
648 */
649struct ip_vs_service_user_kern {
650 /* virtual service addresses */
651 u16 af;
652 u16 protocol;
653 union nf_inet_addr addr; /* virtual ip address */
654 __be16 port;
655 u32 fwmark; /* firwall mark of service */
656
657 /* virtual service options */
658 char *sched_name;
659 char *pe_name;
660 unsigned int flags; /* virtual service flags */
661 unsigned int timeout; /* persistent timeout in sec */
662 __be32 netmask; /* persistent netmask or plen */
663};
664
665
666struct ip_vs_dest_user_kern {
667 /* destination server address */
668 union nf_inet_addr addr;
669 __be16 port;
670
671 /* real server options */
672 unsigned int conn_flags; /* connection flags */
673 int weight; /* destination weight */
674
675 /* thresholds for active connections */
676 u32 u_threshold; /* upper threshold */
677 u32 l_threshold; /* lower threshold */
678
679 /* Address family of addr */
680 u16 af;
681
682 u16 tun_type; /* tunnel type */
683 __be16 tun_port; /* tunnel port */
684 u16 tun_flags; /* tunnel flags */
685};
686
687
688/*
689 * The information about the virtual service offered to the net and the
690 * forwarding entries.
691 */
692struct ip_vs_service {
693 struct hlist_node s_list; /* for normal service table */
694 struct hlist_node f_list; /* for fwmark-based service table */
695 atomic_t refcnt; /* reference counter */
696
697 u16 af; /* address family */
698 __u16 protocol; /* which protocol (TCP/UDP) */
699 union nf_inet_addr addr; /* IP address for virtual service */
700 __be16 port; /* port number for the service */
701 __u32 fwmark; /* firewall mark of the service */
702 unsigned int flags; /* service status flags */
703 unsigned int timeout; /* persistent timeout in ticks */
704 __be32 netmask; /* grouping granularity, mask/plen */
705 struct netns_ipvs *ipvs;
706
707 struct list_head destinations; /* real server d-linked list */
708 __u32 num_dests; /* number of servers */
709 struct ip_vs_stats stats; /* statistics for the service */
710
711 /* for scheduling */
712 struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */
713 spinlock_t sched_lock; /* lock sched_data */
714 void *sched_data; /* scheduler application data */
715
716 /* alternate persistence engine */
717 struct ip_vs_pe __rcu *pe;
718 int conntrack_afmask;
719
720 struct rcu_head rcu_head;
721};
722
723/* Information for cached dst */
724struct ip_vs_dest_dst {
725 struct dst_entry *dst_cache; /* destination cache entry */
726 u32 dst_cookie;
727 union nf_inet_addr dst_saddr;
728 struct rcu_head rcu_head;
729};
730
731/* The real server destination forwarding entry with ip address, port number,
732 * and so on.
733 */
734struct ip_vs_dest {
735 struct list_head n_list; /* for the dests in the service */
736 struct hlist_node d_list; /* for table with all the dests */
737
738 u16 af; /* address family */
739 __be16 port; /* port number of the server */
740 union nf_inet_addr addr; /* IP address of the server */
741 volatile unsigned int flags; /* dest status flags */
742 atomic_t conn_flags; /* flags to copy to conn */
743 atomic_t weight; /* server weight */
744 atomic_t last_weight; /* server latest weight */
745 __u16 tun_type; /* tunnel type */
746 __be16 tun_port; /* tunnel port */
747 __u16 tun_flags; /* tunnel flags */
748
749 refcount_t refcnt; /* reference counter */
750 struct ip_vs_stats stats; /* statistics */
751 unsigned long idle_start; /* start time, jiffies */
752
753 /* connection counters and thresholds */
754 atomic_t activeconns; /* active connections */
755 atomic_t inactconns; /* inactive connections */
756 atomic_t persistconns; /* persistent connections */
757 __u32 u_threshold; /* upper threshold */
758 __u32 l_threshold; /* lower threshold */
759
760 /* for destination cache */
761 spinlock_t dst_lock; /* lock of dst_cache */
762 struct ip_vs_dest_dst __rcu *dest_dst; /* cached dst info */
763
764 /* for virtual service */
765 struct ip_vs_service __rcu *svc; /* service it belongs to */
766 __u16 protocol; /* which protocol (TCP/UDP) */
767 __be16 vport; /* virtual port number */
768 union nf_inet_addr vaddr; /* virtual IP address */
769 __u32 vfwmark; /* firewall mark of service */
770
771 struct rcu_head rcu_head;
772 struct list_head t_list; /* in dest_trash */
773 unsigned int in_rs_table:1; /* we are in rs_table */
774};
775
776/* The scheduler object */
777struct ip_vs_scheduler {
778 struct list_head n_list; /* d-linked list head */
779 char *name; /* scheduler name */
780 atomic_t refcnt; /* reference counter */
781 struct module *module; /* THIS_MODULE/NULL */
782
783 /* scheduler initializing service */
784 int (*init_service)(struct ip_vs_service *svc);
785 /* scheduling service finish */
786 void (*done_service)(struct ip_vs_service *svc);
787 /* dest is linked */
788 int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
789 /* dest is unlinked */
790 int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
791 /* dest is updated */
792 int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
793
794 /* selecting a server from the given service */
795 struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc,
796 const struct sk_buff *skb,
797 struct ip_vs_iphdr *iph);
798};
799
800/* The persistence engine object */
801struct ip_vs_pe {
802 struct list_head n_list; /* d-linked list head */
803 char *name; /* scheduler name */
804 atomic_t refcnt; /* reference counter */
805 struct module *module; /* THIS_MODULE/NULL */
806
807 /* get the connection template, if any */
808 int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb);
809 bool (*ct_match)(const struct ip_vs_conn_param *p,
810 struct ip_vs_conn *ct);
811 u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval,
812 bool inverse);
813 int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf);
814 /* create connections for real-server outgoing packets */
815 struct ip_vs_conn* (*conn_out)(struct ip_vs_service *svc,
816 struct ip_vs_dest *dest,
817 struct sk_buff *skb,
818 const struct ip_vs_iphdr *iph,
819 __be16 dport, __be16 cport);
820};
821
822/* The application module object (a.k.a. app incarnation) */
823struct ip_vs_app {
824 struct list_head a_list; /* member in app list */
825 int type; /* IP_VS_APP_TYPE_xxx */
826 char *name; /* application module name */
827 __u16 protocol;
828 struct module *module; /* THIS_MODULE/NULL */
829 struct list_head incs_list; /* list of incarnations */
830
831 /* members for application incarnations */
832 struct list_head p_list; /* member in proto app list */
833 struct ip_vs_app *app; /* its real application */
834 __be16 port; /* port number in net order */
835 atomic_t usecnt; /* usage counter */
836 struct rcu_head rcu_head;
837
838 /* output hook: Process packet in inout direction, diff set for TCP.
839 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
840 * 2=Mangled but checksum was not updated
841 */
842 int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *,
843 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh);
844
845 /* input hook: Process packet in outin direction, diff set for TCP.
846 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
847 * 2=Mangled but checksum was not updated
848 */
849 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *,
850 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh);
851
852 /* ip_vs_app initializer */
853 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *);
854
855 /* ip_vs_app finish */
856 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *);
857
858
859 /* not used now */
860 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *,
861 struct ip_vs_protocol *);
862
863 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *);
864
865 int * timeout_table;
866 int * timeouts;
867 int timeouts_size;
868
869 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app,
870 int *verdict, struct ip_vs_conn **cpp);
871
872 struct ip_vs_conn *
873 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app,
874 const struct iphdr *iph, int inverse);
875
876 struct ip_vs_conn *
877 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app,
878 const struct iphdr *iph, int inverse);
879
880 int (*state_transition)(struct ip_vs_conn *cp, int direction,
881 const struct sk_buff *skb,
882 struct ip_vs_app *app);
883
884 void (*timeout_change)(struct ip_vs_app *app, int flags);
885};
886
887struct ipvs_master_sync_state {
888 struct list_head sync_queue;
889 struct ip_vs_sync_buff *sync_buff;
890 unsigned long sync_queue_len;
891 unsigned int sync_queue_delay;
892 struct delayed_work master_wakeup_work;
893 struct netns_ipvs *ipvs;
894};
895
896struct ip_vs_sync_thread_data;
897
898/* How much time to keep dests in trash */
899#define IP_VS_DEST_TRASH_PERIOD (120 * HZ)
900
901struct ipvs_sync_daemon_cfg {
902 union nf_inet_addr mcast_group;
903 int syncid;
904 u16 sync_maxlen;
905 u16 mcast_port;
906 u8 mcast_af;
907 u8 mcast_ttl;
908 /* multicast interface name */
909 char mcast_ifn[IP_VS_IFNAME_MAXLEN];
910};
911
912/* IPVS in network namespace */
913struct netns_ipvs {
914 int gen; /* Generation */
915 int enable; /* enable like nf_hooks do */
916 /* Hash table: for real service lookups */
917 #define IP_VS_RTAB_BITS 4
918 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
919 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
920
921 struct hlist_head rs_table[IP_VS_RTAB_SIZE];
922 /* ip_vs_app */
923 struct list_head app_list;
924 /* ip_vs_proto */
925 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */
926 struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE];
927 /* ip_vs_proto_tcp */
928#ifdef CONFIG_IP_VS_PROTO_TCP
929 #define TCP_APP_TAB_BITS 4
930 #define TCP_APP_TAB_SIZE (1 << TCP_APP_TAB_BITS)
931 #define TCP_APP_TAB_MASK (TCP_APP_TAB_SIZE - 1)
932 struct list_head tcp_apps[TCP_APP_TAB_SIZE];
933#endif
934 /* ip_vs_proto_udp */
935#ifdef CONFIG_IP_VS_PROTO_UDP
936 #define UDP_APP_TAB_BITS 4
937 #define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS)
938 #define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1)
939 struct list_head udp_apps[UDP_APP_TAB_SIZE];
940#endif
941 /* ip_vs_proto_sctp */
942#ifdef CONFIG_IP_VS_PROTO_SCTP
943 #define SCTP_APP_TAB_BITS 4
944 #define SCTP_APP_TAB_SIZE (1 << SCTP_APP_TAB_BITS)
945 #define SCTP_APP_TAB_MASK (SCTP_APP_TAB_SIZE - 1)
946 /* Hash table for SCTP application incarnations */
947 struct list_head sctp_apps[SCTP_APP_TAB_SIZE];
948#endif
949 /* ip_vs_conn */
950 atomic_t conn_count; /* connection counter */
951
952 /* ip_vs_ctl */
953 struct ip_vs_stats_rcu *tot_stats; /* Statistics & est. */
954
955 int num_services; /* no of virtual services */
956 int num_services6; /* IPv6 virtual services */
957
958 /* Trash for destinations */
959 struct list_head dest_trash;
960 spinlock_t dest_trash_lock;
961 struct timer_list dest_trash_timer; /* expiration timer */
962 /* Service counters */
963 atomic_t ftpsvc_counter;
964 atomic_t nullsvc_counter;
965 atomic_t conn_out_counter;
966
967#ifdef CONFIG_SYSCTL
968 /* delayed work for expiring no dest connections */
969 struct delayed_work expire_nodest_conn_work;
970 /* 1/rate drop and drop-entry variables */
971 struct delayed_work defense_work; /* Work handler */
972 int drop_rate;
973 int drop_counter;
974 int old_secure_tcp;
975 atomic_t dropentry;
976 /* locks in ctl.c */
977 spinlock_t dropentry_lock; /* drop entry handling */
978 spinlock_t droppacket_lock; /* drop packet handling */
979 spinlock_t securetcp_lock; /* state and timeout tables */
980
981 /* sys-ctl struct */
982 struct ctl_table_header *sysctl_hdr;
983 struct ctl_table *sysctl_tbl;
984#endif
985
986 /* sysctl variables */
987 int sysctl_amemthresh;
988 int sysctl_am_droprate;
989 int sysctl_drop_entry;
990 int sysctl_drop_packet;
991 int sysctl_secure_tcp;
992#ifdef CONFIG_IP_VS_NFCT
993 int sysctl_conntrack;
994#endif
995 int sysctl_snat_reroute;
996 int sysctl_sync_ver;
997 int sysctl_sync_ports;
998 int sysctl_sync_persist_mode;
999 unsigned long sysctl_sync_qlen_max;
1000 int sysctl_sync_sock_size;
1001 int sysctl_cache_bypass;
1002 int sysctl_expire_nodest_conn;
1003 int sysctl_sloppy_tcp;
1004 int sysctl_sloppy_sctp;
1005 int sysctl_expire_quiescent_template;
1006 int sysctl_sync_threshold[2];
1007 unsigned int sysctl_sync_refresh_period;
1008 int sysctl_sync_retries;
1009 int sysctl_nat_icmp_send;
1010 int sysctl_pmtu_disc;
1011 int sysctl_backup_only;
1012 int sysctl_conn_reuse_mode;
1013 int sysctl_schedule_icmp;
1014 int sysctl_ignore_tunneled;
1015 int sysctl_run_estimation;
1016#ifdef CONFIG_SYSCTL
1017 cpumask_var_t sysctl_est_cpulist; /* kthread cpumask */
1018 int est_cpulist_valid; /* cpulist set */
1019 int sysctl_est_nice; /* kthread nice */
1020 int est_stopped; /* stop tasks */
1021#endif
1022
1023 /* ip_vs_lblc */
1024 int sysctl_lblc_expiration;
1025 struct ctl_table_header *lblc_ctl_header;
1026 struct ctl_table *lblc_ctl_table;
1027 /* ip_vs_lblcr */
1028 int sysctl_lblcr_expiration;
1029 struct ctl_table_header *lblcr_ctl_header;
1030 struct ctl_table *lblcr_ctl_table;
1031 /* ip_vs_est */
1032 struct delayed_work est_reload_work;/* Reload kthread tasks */
1033 struct mutex est_mutex; /* protect kthread tasks */
1034 struct hlist_head est_temp_list; /* Ests during calc phase */
1035 struct ip_vs_est_kt_data **est_kt_arr; /* Array of kthread data ptrs */
1036 unsigned long est_max_threads;/* Hard limit of kthreads */
1037 int est_calc_phase; /* Calculation phase */
1038 int est_chain_max; /* Calculated chain_max */
1039 int est_kt_count; /* Allocated ptrs */
1040 int est_add_ktid; /* ktid where to add ests */
1041 atomic_t est_genid; /* kthreads reload genid */
1042 atomic_t est_genid_done; /* applied genid */
1043 /* ip_vs_sync */
1044 spinlock_t sync_lock;
1045 struct ipvs_master_sync_state *ms;
1046 spinlock_t sync_buff_lock;
1047 struct ip_vs_sync_thread_data *master_tinfo;
1048 struct ip_vs_sync_thread_data *backup_tinfo;
1049 int threads_mask;
1050 volatile int sync_state;
1051 struct mutex sync_mutex;
1052 struct ipvs_sync_daemon_cfg mcfg; /* Master Configuration */
1053 struct ipvs_sync_daemon_cfg bcfg; /* Backup Configuration */
1054 /* net name space ptr */
1055 struct net *net; /* Needed by timer routines */
1056 /* Number of heterogeneous destinations, needed becaus heterogeneous
1057 * are not supported when synchronization is enabled.
1058 */
1059 unsigned int mixed_address_family_dests;
1060 unsigned int hooks_afmask; /* &1=AF_INET, &2=AF_INET6 */
1061};
1062
1063#define DEFAULT_SYNC_THRESHOLD 3
1064#define DEFAULT_SYNC_PERIOD 50
1065#define DEFAULT_SYNC_VER 1
1066#define DEFAULT_SLOPPY_TCP 0
1067#define DEFAULT_SLOPPY_SCTP 0
1068#define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ)
1069#define DEFAULT_SYNC_RETRIES 0
1070#define IPVS_SYNC_WAKEUP_RATE 8
1071#define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4)
1072#define IPVS_SYNC_SEND_DELAY (HZ / 50)
1073#define IPVS_SYNC_CHECK_PERIOD HZ
1074#define IPVS_SYNC_FLUSH_TIME (HZ * 2)
1075#define IPVS_SYNC_PORTS_MAX (1 << 6)
1076
1077#ifdef CONFIG_SYSCTL
1078
1079static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1080{
1081 return ipvs->sysctl_sync_threshold[0];
1082}
1083
1084static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1085{
1086 return READ_ONCE(ipvs->sysctl_sync_threshold[1]);
1087}
1088
1089static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1090{
1091 return READ_ONCE(ipvs->sysctl_sync_refresh_period);
1092}
1093
1094static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1095{
1096 return ipvs->sysctl_sync_retries;
1097}
1098
1099static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1100{
1101 return ipvs->sysctl_sync_ver;
1102}
1103
1104static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1105{
1106 return ipvs->sysctl_sloppy_tcp;
1107}
1108
1109static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1110{
1111 return ipvs->sysctl_sloppy_sctp;
1112}
1113
1114static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1115{
1116 return READ_ONCE(ipvs->sysctl_sync_ports);
1117}
1118
1119static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1120{
1121 return ipvs->sysctl_sync_persist_mode;
1122}
1123
1124static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1125{
1126 return ipvs->sysctl_sync_qlen_max;
1127}
1128
1129static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1130{
1131 return ipvs->sysctl_sync_sock_size;
1132}
1133
1134static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1135{
1136 return ipvs->sysctl_pmtu_disc;
1137}
1138
1139static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1140{
1141 return ipvs->sync_state & IP_VS_STATE_BACKUP &&
1142 ipvs->sysctl_backup_only;
1143}
1144
1145static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1146{
1147 return ipvs->sysctl_conn_reuse_mode;
1148}
1149
1150static inline int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs)
1151{
1152 return ipvs->sysctl_expire_nodest_conn;
1153}
1154
1155static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1156{
1157 return ipvs->sysctl_schedule_icmp;
1158}
1159
1160static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1161{
1162 return ipvs->sysctl_ignore_tunneled;
1163}
1164
1165static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1166{
1167 return ipvs->sysctl_cache_bypass;
1168}
1169
1170static inline int sysctl_run_estimation(struct netns_ipvs *ipvs)
1171{
1172 return ipvs->sysctl_run_estimation;
1173}
1174
1175static inline const struct cpumask *sysctl_est_cpulist(struct netns_ipvs *ipvs)
1176{
1177 if (ipvs->est_cpulist_valid)
1178 return ipvs->sysctl_est_cpulist;
1179 else
1180 return housekeeping_cpumask(HK_TYPE_KTHREAD);
1181}
1182
1183static inline int sysctl_est_nice(struct netns_ipvs *ipvs)
1184{
1185 return ipvs->sysctl_est_nice;
1186}
1187
1188#else
1189
1190static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1191{
1192 return DEFAULT_SYNC_THRESHOLD;
1193}
1194
1195static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1196{
1197 return DEFAULT_SYNC_PERIOD;
1198}
1199
1200static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1201{
1202 return DEFAULT_SYNC_REFRESH_PERIOD;
1203}
1204
1205static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1206{
1207 return DEFAULT_SYNC_RETRIES & 3;
1208}
1209
1210static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1211{
1212 return DEFAULT_SYNC_VER;
1213}
1214
1215static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1216{
1217 return DEFAULT_SLOPPY_TCP;
1218}
1219
1220static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1221{
1222 return DEFAULT_SLOPPY_SCTP;
1223}
1224
1225static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1226{
1227 return 1;
1228}
1229
1230static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1231{
1232 return 0;
1233}
1234
1235static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1236{
1237 return IPVS_SYNC_QLEN_MAX;
1238}
1239
1240static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1241{
1242 return 0;
1243}
1244
1245static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1246{
1247 return 1;
1248}
1249
1250static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1251{
1252 return 0;
1253}
1254
1255static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1256{
1257 return 1;
1258}
1259
1260static inline int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs)
1261{
1262 return 0;
1263}
1264
1265static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1266{
1267 return 0;
1268}
1269
1270static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1271{
1272 return 0;
1273}
1274
1275static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1276{
1277 return 0;
1278}
1279
1280static inline int sysctl_run_estimation(struct netns_ipvs *ipvs)
1281{
1282 return 1;
1283}
1284
1285static inline const struct cpumask *sysctl_est_cpulist(struct netns_ipvs *ipvs)
1286{
1287 return housekeeping_cpumask(HK_TYPE_KTHREAD);
1288}
1289
1290static inline int sysctl_est_nice(struct netns_ipvs *ipvs)
1291{
1292 return IPVS_EST_NICE;
1293}
1294
1295#endif
1296
1297/* IPVS core functions
1298 * (from ip_vs_core.c)
1299 */
1300const char *ip_vs_proto_name(unsigned int proto);
1301void ip_vs_init_hash_table(struct list_head *table, int rows);
1302struct ip_vs_conn *ip_vs_new_conn_out(struct ip_vs_service *svc,
1303 struct ip_vs_dest *dest,
1304 struct sk_buff *skb,
1305 const struct ip_vs_iphdr *iph,
1306 __be16 dport,
1307 __be16 cport);
1308#define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t)))
1309
1310#define IP_VS_APP_TYPE_FTP 1
1311
1312/* ip_vs_conn handling functions
1313 * (from ip_vs_conn.c)
1314 */
1315enum {
1316 IP_VS_DIR_INPUT = 0,
1317 IP_VS_DIR_OUTPUT,
1318 IP_VS_DIR_INPUT_ONLY,
1319 IP_VS_DIR_LAST,
1320};
1321
1322static inline void ip_vs_conn_fill_param(struct netns_ipvs *ipvs, int af, int protocol,
1323 const union nf_inet_addr *caddr,
1324 __be16 cport,
1325 const union nf_inet_addr *vaddr,
1326 __be16 vport,
1327 struct ip_vs_conn_param *p)
1328{
1329 p->ipvs = ipvs;
1330 p->af = af;
1331 p->protocol = protocol;
1332 p->caddr = caddr;
1333 p->cport = cport;
1334 p->vaddr = vaddr;
1335 p->vport = vport;
1336 p->pe = NULL;
1337 p->pe_data = NULL;
1338}
1339
1340struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p);
1341struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p);
1342
1343struct ip_vs_conn * ip_vs_conn_in_get_proto(struct netns_ipvs *ipvs, int af,
1344 const struct sk_buff *skb,
1345 const struct ip_vs_iphdr *iph);
1346
1347struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p);
1348
1349struct ip_vs_conn * ip_vs_conn_out_get_proto(struct netns_ipvs *ipvs, int af,
1350 const struct sk_buff *skb,
1351 const struct ip_vs_iphdr *iph);
1352
1353/* Get reference to gain full access to conn.
1354 * By default, RCU read-side critical sections have access only to
1355 * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference.
1356 */
1357static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp)
1358{
1359 return refcount_inc_not_zero(&cp->refcnt);
1360}
1361
1362/* put back the conn without restarting its timer */
1363static inline void __ip_vs_conn_put(struct ip_vs_conn *cp)
1364{
1365 smp_mb__before_atomic();
1366 refcount_dec(&cp->refcnt);
1367}
1368void ip_vs_conn_put(struct ip_vs_conn *cp);
1369void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport);
1370
1371struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, int dest_af,
1372 const union nf_inet_addr *daddr,
1373 __be16 dport, unsigned int flags,
1374 struct ip_vs_dest *dest, __u32 fwmark);
1375void ip_vs_conn_expire_now(struct ip_vs_conn *cp);
1376
1377const char *ip_vs_state_name(const struct ip_vs_conn *cp);
1378
1379void ip_vs_tcp_conn_listen(struct ip_vs_conn *cp);
1380int ip_vs_check_template(struct ip_vs_conn *ct, struct ip_vs_dest *cdest);
1381void ip_vs_random_dropentry(struct netns_ipvs *ipvs);
1382int ip_vs_conn_init(void);
1383void ip_vs_conn_cleanup(void);
1384
1385static inline void ip_vs_control_del(struct ip_vs_conn *cp)
1386{
1387 struct ip_vs_conn *ctl_cp = cp->control;
1388 if (!ctl_cp) {
1389 IP_VS_ERR_BUF("request control DEL for uncontrolled: "
1390 "%s:%d to %s:%d\n",
1391 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1392 ntohs(cp->cport),
1393 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1394 ntohs(cp->vport));
1395
1396 return;
1397 }
1398
1399 IP_VS_DBG_BUF(7, "DELeting control for: "
1400 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1401 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1402 ntohs(cp->cport),
1403 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1404 ntohs(ctl_cp->cport));
1405
1406 cp->control = NULL;
1407 if (atomic_read(&ctl_cp->n_control) == 0) {
1408 IP_VS_ERR_BUF("BUG control DEL with n=0 : "
1409 "%s:%d to %s:%d\n",
1410 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1411 ntohs(cp->cport),
1412 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1413 ntohs(cp->vport));
1414
1415 return;
1416 }
1417 atomic_dec(&ctl_cp->n_control);
1418}
1419
1420static inline void
1421ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp)
1422{
1423 if (cp->control) {
1424 IP_VS_ERR_BUF("request control ADD for already controlled: "
1425 "%s:%d to %s:%d\n",
1426 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1427 ntohs(cp->cport),
1428 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1429 ntohs(cp->vport));
1430
1431 ip_vs_control_del(cp);
1432 }
1433
1434 IP_VS_DBG_BUF(7, "ADDing control for: "
1435 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1436 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1437 ntohs(cp->cport),
1438 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1439 ntohs(ctl_cp->cport));
1440
1441 cp->control = ctl_cp;
1442 atomic_inc(&ctl_cp->n_control);
1443}
1444
1445/* Mark our template as assured */
1446static inline void
1447ip_vs_control_assure_ct(struct ip_vs_conn *cp)
1448{
1449 struct ip_vs_conn *ct = cp->control;
1450
1451 if (ct && !(ct->state & IP_VS_CTPL_S_ASSURED) &&
1452 (ct->flags & IP_VS_CONN_F_TEMPLATE))
1453 ct->state |= IP_VS_CTPL_S_ASSURED;
1454}
1455
1456/* IPVS netns init & cleanup functions */
1457int ip_vs_estimator_net_init(struct netns_ipvs *ipvs);
1458int ip_vs_control_net_init(struct netns_ipvs *ipvs);
1459int ip_vs_protocol_net_init(struct netns_ipvs *ipvs);
1460int ip_vs_app_net_init(struct netns_ipvs *ipvs);
1461int ip_vs_conn_net_init(struct netns_ipvs *ipvs);
1462int ip_vs_sync_net_init(struct netns_ipvs *ipvs);
1463void ip_vs_conn_net_cleanup(struct netns_ipvs *ipvs);
1464void ip_vs_app_net_cleanup(struct netns_ipvs *ipvs);
1465void ip_vs_protocol_net_cleanup(struct netns_ipvs *ipvs);
1466void ip_vs_control_net_cleanup(struct netns_ipvs *ipvs);
1467void ip_vs_estimator_net_cleanup(struct netns_ipvs *ipvs);
1468void ip_vs_sync_net_cleanup(struct netns_ipvs *ipvs);
1469void ip_vs_service_nets_cleanup(struct list_head *net_list);
1470
1471/* IPVS application functions
1472 * (from ip_vs_app.c)
1473 */
1474#define IP_VS_APP_MAX_PORTS 8
1475struct ip_vs_app *register_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1476void unregister_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1477int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1478void ip_vs_unbind_app(struct ip_vs_conn *cp);
1479int register_ip_vs_app_inc(struct netns_ipvs *ipvs, struct ip_vs_app *app, __u16 proto,
1480 __u16 port);
1481int ip_vs_app_inc_get(struct ip_vs_app *inc);
1482void ip_vs_app_inc_put(struct ip_vs_app *inc);
1483
1484int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb,
1485 struct ip_vs_iphdr *ipvsh);
1486int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb,
1487 struct ip_vs_iphdr *ipvsh);
1488
1489int register_ip_vs_pe(struct ip_vs_pe *pe);
1490int unregister_ip_vs_pe(struct ip_vs_pe *pe);
1491struct ip_vs_pe *ip_vs_pe_getbyname(const char *name);
1492struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name);
1493
1494/* Use a #define to avoid all of module.h just for these trivial ops */
1495#define ip_vs_pe_get(pe) \
1496 if (pe && pe->module) \
1497 __module_get(pe->module);
1498
1499#define ip_vs_pe_put(pe) \
1500 if (pe && pe->module) \
1501 module_put(pe->module);
1502
1503/* IPVS protocol functions (from ip_vs_proto.c) */
1504int ip_vs_protocol_init(void);
1505void ip_vs_protocol_cleanup(void);
1506void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags);
1507int *ip_vs_create_timeout_table(int *table, int size);
1508void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
1509 const struct sk_buff *skb, int offset,
1510 const char *msg);
1511
1512extern struct ip_vs_protocol ip_vs_protocol_tcp;
1513extern struct ip_vs_protocol ip_vs_protocol_udp;
1514extern struct ip_vs_protocol ip_vs_protocol_icmp;
1515extern struct ip_vs_protocol ip_vs_protocol_esp;
1516extern struct ip_vs_protocol ip_vs_protocol_ah;
1517extern struct ip_vs_protocol ip_vs_protocol_sctp;
1518
1519/* Registering/unregistering scheduler functions
1520 * (from ip_vs_sched.c)
1521 */
1522int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1523int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1524int ip_vs_bind_scheduler(struct ip_vs_service *svc,
1525 struct ip_vs_scheduler *scheduler);
1526void ip_vs_unbind_scheduler(struct ip_vs_service *svc,
1527 struct ip_vs_scheduler *sched);
1528struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name);
1529void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler);
1530struct ip_vs_conn *
1531ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb,
1532 struct ip_vs_proto_data *pd, int *ignored,
1533 struct ip_vs_iphdr *iph);
1534int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb,
1535 struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph);
1536
1537void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg);
1538
1539/* IPVS control data and functions (from ip_vs_ctl.c) */
1540extern struct ip_vs_stats ip_vs_stats;
1541extern int sysctl_ip_vs_sync_ver;
1542
1543struct ip_vs_service *
1544ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
1545 const union nf_inet_addr *vaddr, __be16 vport);
1546
1547bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1548 const union nf_inet_addr *daddr, __be16 dport);
1549
1550struct ip_vs_dest *
1551ip_vs_find_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1552 const union nf_inet_addr *daddr, __be16 dport);
1553struct ip_vs_dest *ip_vs_find_tunnel(struct netns_ipvs *ipvs, int af,
1554 const union nf_inet_addr *daddr,
1555 __be16 tun_port);
1556
1557int ip_vs_use_count_inc(void);
1558void ip_vs_use_count_dec(void);
1559int ip_vs_register_nl_ioctl(void);
1560void ip_vs_unregister_nl_ioctl(void);
1561int ip_vs_control_init(void);
1562void ip_vs_control_cleanup(void);
1563struct ip_vs_dest *
1564ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
1565 const union nf_inet_addr *daddr, __be16 dport,
1566 const union nf_inet_addr *vaddr, __be16 vport,
1567 __u16 protocol, __u32 fwmark, __u32 flags);
1568void ip_vs_try_bind_dest(struct ip_vs_conn *cp);
1569
1570static inline void ip_vs_dest_hold(struct ip_vs_dest *dest)
1571{
1572 refcount_inc(&dest->refcnt);
1573}
1574
1575static inline void ip_vs_dest_put(struct ip_vs_dest *dest)
1576{
1577 smp_mb__before_atomic();
1578 refcount_dec(&dest->refcnt);
1579}
1580
1581static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest)
1582{
1583 if (refcount_dec_and_test(&dest->refcnt))
1584 kfree(dest);
1585}
1586
1587/* IPVS sync daemon data and function prototypes
1588 * (from ip_vs_sync.c)
1589 */
1590int start_sync_thread(struct netns_ipvs *ipvs, struct ipvs_sync_daemon_cfg *cfg,
1591 int state);
1592int stop_sync_thread(struct netns_ipvs *ipvs, int state);
1593void ip_vs_sync_conn(struct netns_ipvs *ipvs, struct ip_vs_conn *cp, int pkts);
1594
1595/* IPVS rate estimator prototypes (from ip_vs_est.c) */
1596int ip_vs_start_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1597void ip_vs_stop_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1598void ip_vs_zero_estimator(struct ip_vs_stats *stats);
1599void ip_vs_read_estimator(struct ip_vs_kstats *dst, struct ip_vs_stats *stats);
1600void ip_vs_est_reload_start(struct netns_ipvs *ipvs);
1601int ip_vs_est_kthread_start(struct netns_ipvs *ipvs,
1602 struct ip_vs_est_kt_data *kd);
1603void ip_vs_est_kthread_stop(struct ip_vs_est_kt_data *kd);
1604
1605static inline void ip_vs_est_stopped_recalc(struct netns_ipvs *ipvs)
1606{
1607#ifdef CONFIG_SYSCTL
1608 /* Stop tasks while cpulist is empty or if disabled with flag */
1609 ipvs->est_stopped = !sysctl_run_estimation(ipvs) ||
1610 (ipvs->est_cpulist_valid &&
1611 cpumask_empty(sysctl_est_cpulist(ipvs)));
1612#endif
1613}
1614
1615static inline bool ip_vs_est_stopped(struct netns_ipvs *ipvs)
1616{
1617#ifdef CONFIG_SYSCTL
1618 return ipvs->est_stopped;
1619#else
1620 return false;
1621#endif
1622}
1623
1624static inline int ip_vs_est_max_threads(struct netns_ipvs *ipvs)
1625{
1626 unsigned int limit = IPVS_EST_CPU_KTHREADS *
1627 cpumask_weight(sysctl_est_cpulist(ipvs));
1628
1629 return max(1U, limit);
1630}
1631
1632/* Various IPVS packet transmitters (from ip_vs_xmit.c) */
1633int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1634 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1635int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1636 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1637int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1638 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1639int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1640 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1641int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1642 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1643int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1644 struct ip_vs_protocol *pp, int offset,
1645 unsigned int hooknum, struct ip_vs_iphdr *iph);
1646void ip_vs_dest_dst_rcu_free(struct rcu_head *head);
1647
1648#ifdef CONFIG_IP_VS_IPV6
1649int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1650 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1651int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1652 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1653int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1654 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1655int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1656 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1657int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1658 struct ip_vs_protocol *pp, int offset,
1659 unsigned int hooknum, struct ip_vs_iphdr *iph);
1660#endif
1661
1662#ifdef CONFIG_SYSCTL
1663/* This is a simple mechanism to ignore packets when
1664 * we are loaded. Just set ip_vs_drop_rate to 'n' and
1665 * we start to drop 1/rate of the packets
1666 */
1667static inline int ip_vs_todrop(struct netns_ipvs *ipvs)
1668{
1669 if (!ipvs->drop_rate)
1670 return 0;
1671 if (--ipvs->drop_counter > 0)
1672 return 0;
1673 ipvs->drop_counter = ipvs->drop_rate;
1674 return 1;
1675}
1676#else
1677static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; }
1678#endif
1679
1680#ifdef CONFIG_SYSCTL
1681/* Enqueue delayed work for expiring no dest connections
1682 * Only run when sysctl_expire_nodest=1
1683 */
1684static inline void ip_vs_enqueue_expire_nodest_conns(struct netns_ipvs *ipvs)
1685{
1686 if (sysctl_expire_nodest_conn(ipvs))
1687 queue_delayed_work(system_long_wq,
1688 &ipvs->expire_nodest_conn_work, 1);
1689}
1690
1691void ip_vs_expire_nodest_conn_flush(struct netns_ipvs *ipvs);
1692#else
1693static inline void ip_vs_enqueue_expire_nodest_conns(struct netns_ipvs *ipvs) {}
1694#endif
1695
1696#define IP_VS_DFWD_METHOD(dest) (atomic_read(&(dest)->conn_flags) & \
1697 IP_VS_CONN_F_FWD_MASK)
1698
1699/* ip_vs_fwd_tag returns the forwarding tag of the connection */
1700#define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK)
1701
1702static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp)
1703{
1704 char fwd;
1705
1706 switch (IP_VS_FWD_METHOD(cp)) {
1707 case IP_VS_CONN_F_MASQ:
1708 fwd = 'M'; break;
1709 case IP_VS_CONN_F_LOCALNODE:
1710 fwd = 'L'; break;
1711 case IP_VS_CONN_F_TUNNEL:
1712 fwd = 'T'; break;
1713 case IP_VS_CONN_F_DROUTE:
1714 fwd = 'R'; break;
1715 case IP_VS_CONN_F_BYPASS:
1716 fwd = 'B'; break;
1717 default:
1718 fwd = '?'; break;
1719 }
1720 return fwd;
1721}
1722
1723void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp,
1724 struct ip_vs_conn *cp, int dir);
1725
1726#ifdef CONFIG_IP_VS_IPV6
1727void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp,
1728 struct ip_vs_conn *cp, int dir);
1729#endif
1730
1731__sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset);
1732
1733static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum)
1734{
1735 __be32 diff[2] = { ~old, new };
1736
1737 return csum_partial(diff, sizeof(diff), oldsum);
1738}
1739
1740#ifdef CONFIG_IP_VS_IPV6
1741static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new,
1742 __wsum oldsum)
1743{
1744 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0],
1745 new[3], new[2], new[1], new[0] };
1746
1747 return csum_partial(diff, sizeof(diff), oldsum);
1748}
1749#endif
1750
1751static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum)
1752{
1753 __be16 diff[2] = { ~old, new };
1754
1755 return csum_partial(diff, sizeof(diff), oldsum);
1756}
1757
1758/* Forget current conntrack (unconfirmed) and attach notrack entry */
1759static inline void ip_vs_notrack(struct sk_buff *skb)
1760{
1761#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1762 enum ip_conntrack_info ctinfo;
1763 struct nf_conn *ct = nf_ct_get(skb, &ctinfo);
1764
1765 if (ct) {
1766 nf_conntrack_put(&ct->ct_general);
1767 nf_ct_set(skb, NULL, IP_CT_UNTRACKED);
1768 }
1769#endif
1770}
1771
1772#ifdef CONFIG_IP_VS_NFCT
1773/* Netfilter connection tracking
1774 * (from ip_vs_nfct.c)
1775 */
1776static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1777{
1778#ifdef CONFIG_SYSCTL
1779 return ipvs->sysctl_conntrack;
1780#else
1781 return 0;
1782#endif
1783}
1784
1785void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp,
1786 int outin);
1787int ip_vs_confirm_conntrack(struct sk_buff *skb);
1788void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct,
1789 struct ip_vs_conn *cp, u_int8_t proto,
1790 const __be16 port, int from_rs);
1791void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp);
1792
1793#else
1794
1795static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1796{
1797 return 0;
1798}
1799
1800static inline void ip_vs_update_conntrack(struct sk_buff *skb,
1801 struct ip_vs_conn *cp, int outin)
1802{
1803}
1804
1805static inline int ip_vs_confirm_conntrack(struct sk_buff *skb)
1806{
1807 return NF_ACCEPT;
1808}
1809
1810static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp)
1811{
1812}
1813#endif /* CONFIG_IP_VS_NFCT */
1814
1815/* Using old conntrack that can not be redirected to another real server? */
1816static inline bool ip_vs_conn_uses_old_conntrack(struct ip_vs_conn *cp,
1817 struct sk_buff *skb)
1818{
1819#ifdef CONFIG_IP_VS_NFCT
1820 enum ip_conntrack_info ctinfo;
1821 struct nf_conn *ct;
1822
1823 ct = nf_ct_get(skb, &ctinfo);
1824 if (ct && nf_ct_is_confirmed(ct))
1825 return true;
1826#endif
1827 return false;
1828}
1829
1830static inline int ip_vs_register_conntrack(struct ip_vs_service *svc)
1831{
1832#if IS_ENABLED(CONFIG_NF_CONNTRACK)
1833 int afmask = (svc->af == AF_INET6) ? 2 : 1;
1834 int ret = 0;
1835
1836 if (!(svc->conntrack_afmask & afmask)) {
1837 ret = nf_ct_netns_get(svc->ipvs->net, svc->af);
1838 if (ret >= 0)
1839 svc->conntrack_afmask |= afmask;
1840 }
1841 return ret;
1842#else
1843 return 0;
1844#endif
1845}
1846
1847static inline void ip_vs_unregister_conntrack(struct ip_vs_service *svc)
1848{
1849#if IS_ENABLED(CONFIG_NF_CONNTRACK)
1850 int afmask = (svc->af == AF_INET6) ? 2 : 1;
1851
1852 if (svc->conntrack_afmask & afmask) {
1853 nf_ct_netns_put(svc->ipvs->net, svc->af);
1854 svc->conntrack_afmask &= ~afmask;
1855 }
1856#endif
1857}
1858
1859int ip_vs_register_hooks(struct netns_ipvs *ipvs, unsigned int af);
1860void ip_vs_unregister_hooks(struct netns_ipvs *ipvs, unsigned int af);
1861
1862static inline int
1863ip_vs_dest_conn_overhead(struct ip_vs_dest *dest)
1864{
1865 /* We think the overhead of processing active connections is 256
1866 * times higher than that of inactive connections in average. (This
1867 * 256 times might not be accurate, we will change it later) We
1868 * use the following formula to estimate the overhead now:
1869 * dest->activeconns*256 + dest->inactconns
1870 */
1871 return (atomic_read(&dest->activeconns) << 8) +
1872 atomic_read(&dest->inactconns);
1873}
1874
1875#ifdef CONFIG_IP_VS_PROTO_TCP
1876INDIRECT_CALLABLE_DECLARE(int
1877 tcp_snat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
1878 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph));
1879#endif
1880
1881#ifdef CONFIG_IP_VS_PROTO_UDP
1882INDIRECT_CALLABLE_DECLARE(int
1883 udp_snat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
1884 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph));
1885#endif
1886#endif /* _NET_IP_VS_H */