Loading...
1/* IP Virtual Server
2 * data structure and functionality definitions
3 */
4
5#ifndef _NET_IP_VS_H
6#define _NET_IP_VS_H
7
8#include <linux/ip_vs.h> /* definitions shared with userland */
9
10#include <asm/types.h> /* for __uXX types */
11
12#include <linux/list.h> /* for struct list_head */
13#include <linux/spinlock.h> /* for struct rwlock_t */
14#include <linux/atomic.h> /* for struct atomic_t */
15#include <linux/compiler.h>
16#include <linux/timer.h>
17#include <linux/bug.h>
18
19#include <net/checksum.h>
20#include <linux/netfilter.h> /* for union nf_inet_addr */
21#include <linux/ip.h>
22#include <linux/ipv6.h> /* for struct ipv6hdr */
23#include <net/ipv6.h>
24#if IS_ENABLED(CONFIG_IP_VS_IPV6)
25#include <linux/netfilter_ipv6/ip6_tables.h>
26#endif
27#if IS_ENABLED(CONFIG_NF_CONNTRACK)
28#include <net/netfilter/nf_conntrack.h>
29#endif
30#include <net/net_namespace.h> /* Netw namespace */
31
32#define IP_VS_HDR_INVERSE 1
33#define IP_VS_HDR_ICMP 2
34
35/* Generic access of ipvs struct */
36static inline struct netns_ipvs *net_ipvs(struct net* net)
37{
38 return net->ipvs;
39}
40
41/* This one needed for single_open_net since net is stored directly in
42 * private not as a struct i.e. seq_file_net can't be used.
43 */
44static inline struct net *seq_file_single_net(struct seq_file *seq)
45{
46#ifdef CONFIG_NET_NS
47 return (struct net *)seq->private;
48#else
49 return &init_net;
50#endif
51}
52
53/* Connections' size value needed by ip_vs_ctl.c */
54extern int ip_vs_conn_tab_size;
55
56struct ip_vs_iphdr {
57 int hdr_flags; /* ipvs flags */
58 __u32 off; /* Where IP or IPv4 header starts */
59 __u32 len; /* IPv4 simply where L4 starts
60 * IPv6 where L4 Transport Header starts */
61 __u16 fragoffs; /* IPv6 fragment offset, 0 if first frag (or not frag)*/
62 __s16 protocol;
63 __s32 flags;
64 union nf_inet_addr saddr;
65 union nf_inet_addr daddr;
66};
67
68static inline void *frag_safe_skb_hp(const struct sk_buff *skb, int offset,
69 int len, void *buffer,
70 const struct ip_vs_iphdr *ipvsh)
71{
72 return skb_header_pointer(skb, offset, len, buffer);
73}
74
75/* This function handles filling *ip_vs_iphdr, both for IPv4 and IPv6.
76 * IPv6 requires some extra work, as finding proper header position,
77 * depend on the IPv6 extension headers.
78 */
79static inline int
80ip_vs_fill_iph_skb_off(int af, const struct sk_buff *skb, int offset,
81 int hdr_flags, struct ip_vs_iphdr *iphdr)
82{
83 iphdr->hdr_flags = hdr_flags;
84 iphdr->off = offset;
85
86#ifdef CONFIG_IP_VS_IPV6
87 if (af == AF_INET6) {
88 struct ipv6hdr _iph;
89 const struct ipv6hdr *iph = skb_header_pointer(
90 skb, offset, sizeof(_iph), &_iph);
91 if (!iph)
92 return 0;
93
94 iphdr->saddr.in6 = iph->saddr;
95 iphdr->daddr.in6 = iph->daddr;
96 /* ipv6_find_hdr() updates len, flags */
97 iphdr->len = offset;
98 iphdr->flags = 0;
99 iphdr->protocol = ipv6_find_hdr(skb, &iphdr->len, -1,
100 &iphdr->fragoffs,
101 &iphdr->flags);
102 if (iphdr->protocol < 0)
103 return 0;
104 } else
105#endif
106 {
107 struct iphdr _iph;
108 const struct iphdr *iph = skb_header_pointer(
109 skb, offset, sizeof(_iph), &_iph);
110 if (!iph)
111 return 0;
112
113 iphdr->len = offset + iph->ihl * 4;
114 iphdr->fragoffs = 0;
115 iphdr->protocol = iph->protocol;
116 iphdr->saddr.ip = iph->saddr;
117 iphdr->daddr.ip = iph->daddr;
118 }
119
120 return 1;
121}
122
123static inline int
124ip_vs_fill_iph_skb_icmp(int af, const struct sk_buff *skb, int offset,
125 bool inverse, struct ip_vs_iphdr *iphdr)
126{
127 int hdr_flags = IP_VS_HDR_ICMP;
128
129 if (inverse)
130 hdr_flags |= IP_VS_HDR_INVERSE;
131
132 return ip_vs_fill_iph_skb_off(af, skb, offset, hdr_flags, iphdr);
133}
134
135static inline int
136ip_vs_fill_iph_skb(int af, const struct sk_buff *skb, bool inverse,
137 struct ip_vs_iphdr *iphdr)
138{
139 int hdr_flags = 0;
140
141 if (inverse)
142 hdr_flags |= IP_VS_HDR_INVERSE;
143
144 return ip_vs_fill_iph_skb_off(af, skb, skb_network_offset(skb),
145 hdr_flags, iphdr);
146}
147
148static inline bool
149ip_vs_iph_inverse(const struct ip_vs_iphdr *iph)
150{
151 return !!(iph->hdr_flags & IP_VS_HDR_INVERSE);
152}
153
154static inline bool
155ip_vs_iph_icmp(const struct ip_vs_iphdr *iph)
156{
157 return !!(iph->hdr_flags & IP_VS_HDR_ICMP);
158}
159
160static inline void ip_vs_addr_copy(int af, union nf_inet_addr *dst,
161 const union nf_inet_addr *src)
162{
163#ifdef CONFIG_IP_VS_IPV6
164 if (af == AF_INET6)
165 dst->in6 = src->in6;
166 else
167#endif
168 dst->ip = src->ip;
169}
170
171static inline void ip_vs_addr_set(int af, union nf_inet_addr *dst,
172 const union nf_inet_addr *src)
173{
174#ifdef CONFIG_IP_VS_IPV6
175 if (af == AF_INET6) {
176 dst->in6 = src->in6;
177 return;
178 }
179#endif
180 dst->ip = src->ip;
181 dst->all[1] = 0;
182 dst->all[2] = 0;
183 dst->all[3] = 0;
184}
185
186static inline int ip_vs_addr_equal(int af, const union nf_inet_addr *a,
187 const union nf_inet_addr *b)
188{
189#ifdef CONFIG_IP_VS_IPV6
190 if (af == AF_INET6)
191 return ipv6_addr_equal(&a->in6, &b->in6);
192#endif
193 return a->ip == b->ip;
194}
195
196#ifdef CONFIG_IP_VS_DEBUG
197#include <linux/net.h>
198
199int ip_vs_get_debug_level(void);
200
201static inline const char *ip_vs_dbg_addr(int af, char *buf, size_t buf_len,
202 const union nf_inet_addr *addr,
203 int *idx)
204{
205 int len;
206#ifdef CONFIG_IP_VS_IPV6
207 if (af == AF_INET6)
208 len = snprintf(&buf[*idx], buf_len - *idx, "[%pI6c]",
209 &addr->in6) + 1;
210 else
211#endif
212 len = snprintf(&buf[*idx], buf_len - *idx, "%pI4",
213 &addr->ip) + 1;
214
215 *idx += len;
216 BUG_ON(*idx > buf_len + 1);
217 return &buf[*idx - len];
218}
219
220#define IP_VS_DBG_BUF(level, msg, ...) \
221 do { \
222 char ip_vs_dbg_buf[160]; \
223 int ip_vs_dbg_idx = 0; \
224 if (level <= ip_vs_get_debug_level()) \
225 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
226 } while (0)
227#define IP_VS_ERR_BUF(msg...) \
228 do { \
229 char ip_vs_dbg_buf[160]; \
230 int ip_vs_dbg_idx = 0; \
231 pr_err(msg); \
232 } while (0)
233
234/* Only use from within IP_VS_DBG_BUF() or IP_VS_ERR_BUF macros */
235#define IP_VS_DBG_ADDR(af, addr) \
236 ip_vs_dbg_addr(af, ip_vs_dbg_buf, \
237 sizeof(ip_vs_dbg_buf), addr, \
238 &ip_vs_dbg_idx)
239
240#define IP_VS_DBG(level, msg, ...) \
241 do { \
242 if (level <= ip_vs_get_debug_level()) \
243 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
244 } while (0)
245#define IP_VS_DBG_RL(msg, ...) \
246 do { \
247 if (net_ratelimit()) \
248 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
249 } while (0)
250#define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) \
251 do { \
252 if (level <= ip_vs_get_debug_level()) \
253 pp->debug_packet(af, pp, skb, ofs, msg); \
254 } while (0)
255#define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) \
256 do { \
257 if (level <= ip_vs_get_debug_level() && \
258 net_ratelimit()) \
259 pp->debug_packet(af, pp, skb, ofs, msg); \
260 } while (0)
261#else /* NO DEBUGGING at ALL */
262#define IP_VS_DBG_BUF(level, msg...) do {} while (0)
263#define IP_VS_ERR_BUF(msg...) do {} while (0)
264#define IP_VS_DBG(level, msg...) do {} while (0)
265#define IP_VS_DBG_RL(msg...) do {} while (0)
266#define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
267#define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
268#endif
269
270#define IP_VS_BUG() BUG()
271#define IP_VS_ERR_RL(msg, ...) \
272 do { \
273 if (net_ratelimit()) \
274 pr_err(msg, ##__VA_ARGS__); \
275 } while (0)
276
277#ifdef CONFIG_IP_VS_DEBUG
278#define EnterFunction(level) \
279 do { \
280 if (level <= ip_vs_get_debug_level()) \
281 printk(KERN_DEBUG \
282 pr_fmt("Enter: %s, %s line %i\n"), \
283 __func__, __FILE__, __LINE__); \
284 } while (0)
285#define LeaveFunction(level) \
286 do { \
287 if (level <= ip_vs_get_debug_level()) \
288 printk(KERN_DEBUG \
289 pr_fmt("Leave: %s, %s line %i\n"), \
290 __func__, __FILE__, __LINE__); \
291 } while (0)
292#else
293#define EnterFunction(level) do {} while (0)
294#define LeaveFunction(level) do {} while (0)
295#endif
296
297/* The port number of FTP service (in network order). */
298#define FTPPORT cpu_to_be16(21)
299#define FTPDATA cpu_to_be16(20)
300
301/* TCP State Values */
302enum {
303 IP_VS_TCP_S_NONE = 0,
304 IP_VS_TCP_S_ESTABLISHED,
305 IP_VS_TCP_S_SYN_SENT,
306 IP_VS_TCP_S_SYN_RECV,
307 IP_VS_TCP_S_FIN_WAIT,
308 IP_VS_TCP_S_TIME_WAIT,
309 IP_VS_TCP_S_CLOSE,
310 IP_VS_TCP_S_CLOSE_WAIT,
311 IP_VS_TCP_S_LAST_ACK,
312 IP_VS_TCP_S_LISTEN,
313 IP_VS_TCP_S_SYNACK,
314 IP_VS_TCP_S_LAST
315};
316
317/* UDP State Values */
318enum {
319 IP_VS_UDP_S_NORMAL,
320 IP_VS_UDP_S_LAST,
321};
322
323/* ICMP State Values */
324enum {
325 IP_VS_ICMP_S_NORMAL,
326 IP_VS_ICMP_S_LAST,
327};
328
329/* SCTP State Values */
330enum ip_vs_sctp_states {
331 IP_VS_SCTP_S_NONE,
332 IP_VS_SCTP_S_INIT1,
333 IP_VS_SCTP_S_INIT,
334 IP_VS_SCTP_S_COOKIE_SENT,
335 IP_VS_SCTP_S_COOKIE_REPLIED,
336 IP_VS_SCTP_S_COOKIE_WAIT,
337 IP_VS_SCTP_S_COOKIE,
338 IP_VS_SCTP_S_COOKIE_ECHOED,
339 IP_VS_SCTP_S_ESTABLISHED,
340 IP_VS_SCTP_S_SHUTDOWN_SENT,
341 IP_VS_SCTP_S_SHUTDOWN_RECEIVED,
342 IP_VS_SCTP_S_SHUTDOWN_ACK_SENT,
343 IP_VS_SCTP_S_REJECTED,
344 IP_VS_SCTP_S_CLOSED,
345 IP_VS_SCTP_S_LAST
346};
347
348/* Delta sequence info structure
349 * Each ip_vs_conn has 2 (output AND input seq. changes).
350 * Only used in the VS/NAT.
351 */
352struct ip_vs_seq {
353 __u32 init_seq; /* Add delta from this seq */
354 __u32 delta; /* Delta in sequence numbers */
355 __u32 previous_delta; /* Delta in sequence numbers
356 * before last resized pkt */
357};
358
359/* counters per cpu */
360struct ip_vs_counters {
361 __u64 conns; /* connections scheduled */
362 __u64 inpkts; /* incoming packets */
363 __u64 outpkts; /* outgoing packets */
364 __u64 inbytes; /* incoming bytes */
365 __u64 outbytes; /* outgoing bytes */
366};
367/* Stats per cpu */
368struct ip_vs_cpu_stats {
369 struct ip_vs_counters cnt;
370 struct u64_stats_sync syncp;
371};
372
373/* IPVS statistics objects */
374struct ip_vs_estimator {
375 struct list_head list;
376
377 u64 last_inbytes;
378 u64 last_outbytes;
379 u64 last_conns;
380 u64 last_inpkts;
381 u64 last_outpkts;
382
383 u64 cps;
384 u64 inpps;
385 u64 outpps;
386 u64 inbps;
387 u64 outbps;
388};
389
390/*
391 * IPVS statistics object, 64-bit kernel version of struct ip_vs_stats_user
392 */
393struct ip_vs_kstats {
394 u64 conns; /* connections scheduled */
395 u64 inpkts; /* incoming packets */
396 u64 outpkts; /* outgoing packets */
397 u64 inbytes; /* incoming bytes */
398 u64 outbytes; /* outgoing bytes */
399
400 u64 cps; /* current connection rate */
401 u64 inpps; /* current in packet rate */
402 u64 outpps; /* current out packet rate */
403 u64 inbps; /* current in byte rate */
404 u64 outbps; /* current out byte rate */
405};
406
407struct ip_vs_stats {
408 struct ip_vs_kstats kstats; /* kernel statistics */
409 struct ip_vs_estimator est; /* estimator */
410 struct ip_vs_cpu_stats __percpu *cpustats; /* per cpu counters */
411 spinlock_t lock; /* spin lock */
412 struct ip_vs_kstats kstats0; /* reset values */
413};
414
415struct dst_entry;
416struct iphdr;
417struct ip_vs_conn;
418struct ip_vs_app;
419struct sk_buff;
420struct ip_vs_proto_data;
421
422struct ip_vs_protocol {
423 struct ip_vs_protocol *next;
424 char *name;
425 u16 protocol;
426 u16 num_states;
427 int dont_defrag;
428
429 void (*init)(struct ip_vs_protocol *pp);
430
431 void (*exit)(struct ip_vs_protocol *pp);
432
433 int (*init_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
434
435 void (*exit_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
436
437 int (*conn_schedule)(struct netns_ipvs *ipvs,
438 int af, struct sk_buff *skb,
439 struct ip_vs_proto_data *pd,
440 int *verdict, struct ip_vs_conn **cpp,
441 struct ip_vs_iphdr *iph);
442
443 struct ip_vs_conn *
444 (*conn_in_get)(struct netns_ipvs *ipvs,
445 int af,
446 const struct sk_buff *skb,
447 const struct ip_vs_iphdr *iph);
448
449 struct ip_vs_conn *
450 (*conn_out_get)(struct netns_ipvs *ipvs,
451 int af,
452 const struct sk_buff *skb,
453 const struct ip_vs_iphdr *iph);
454
455 int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
456 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
457
458 int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
459 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
460
461 int (*csum_check)(int af, struct sk_buff *skb,
462 struct ip_vs_protocol *pp);
463
464 const char *(*state_name)(int state);
465
466 void (*state_transition)(struct ip_vs_conn *cp, int direction,
467 const struct sk_buff *skb,
468 struct ip_vs_proto_data *pd);
469
470 int (*register_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
471
472 void (*unregister_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
473
474 int (*app_conn_bind)(struct ip_vs_conn *cp);
475
476 void (*debug_packet)(int af, struct ip_vs_protocol *pp,
477 const struct sk_buff *skb,
478 int offset,
479 const char *msg);
480
481 void (*timeout_change)(struct ip_vs_proto_data *pd, int flags);
482};
483
484/* protocol data per netns */
485struct ip_vs_proto_data {
486 struct ip_vs_proto_data *next;
487 struct ip_vs_protocol *pp;
488 int *timeout_table; /* protocol timeout table */
489 atomic_t appcnt; /* counter of proto app incs. */
490 struct tcp_states_t *tcp_state_table;
491};
492
493struct ip_vs_protocol *ip_vs_proto_get(unsigned short proto);
494struct ip_vs_proto_data *ip_vs_proto_data_get(struct netns_ipvs *ipvs,
495 unsigned short proto);
496
497struct ip_vs_conn_param {
498 struct netns_ipvs *ipvs;
499 const union nf_inet_addr *caddr;
500 const union nf_inet_addr *vaddr;
501 __be16 cport;
502 __be16 vport;
503 __u16 protocol;
504 u16 af;
505
506 const struct ip_vs_pe *pe;
507 char *pe_data;
508 __u8 pe_data_len;
509};
510
511/* IP_VS structure allocated for each dynamically scheduled connection */
512struct ip_vs_conn {
513 struct hlist_node c_list; /* hashed list heads */
514 /* Protocol, addresses and port numbers */
515 __be16 cport;
516 __be16 dport;
517 __be16 vport;
518 u16 af; /* address family */
519 union nf_inet_addr caddr; /* client address */
520 union nf_inet_addr vaddr; /* virtual address */
521 union nf_inet_addr daddr; /* destination address */
522 volatile __u32 flags; /* status flags */
523 __u16 protocol; /* Which protocol (TCP/UDP) */
524 __u16 daf; /* Address family of the dest */
525 struct netns_ipvs *ipvs;
526
527 /* counter and timer */
528 atomic_t refcnt; /* reference count */
529 struct timer_list timer; /* Expiration timer */
530 volatile unsigned long timeout; /* timeout */
531
532 /* Flags and state transition */
533 spinlock_t lock; /* lock for state transition */
534 volatile __u16 state; /* state info */
535 volatile __u16 old_state; /* old state, to be used for
536 * state transition triggerd
537 * synchronization
538 */
539 __u32 fwmark; /* Fire wall mark from skb */
540 unsigned long sync_endtime; /* jiffies + sent_retries */
541
542 /* Control members */
543 struct ip_vs_conn *control; /* Master control connection */
544 atomic_t n_control; /* Number of controlled ones */
545 struct ip_vs_dest *dest; /* real server */
546 atomic_t in_pkts; /* incoming packet counter */
547
548 /* Packet transmitter for different forwarding methods. If it
549 * mangles the packet, it must return NF_DROP or better NF_STOLEN,
550 * otherwise this must be changed to a sk_buff **.
551 * NF_ACCEPT can be returned when destination is local.
552 */
553 int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp,
554 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
555
556 /* Note: we can group the following members into a structure,
557 * in order to save more space, and the following members are
558 * only used in VS/NAT anyway
559 */
560 struct ip_vs_app *app; /* bound ip_vs_app object */
561 void *app_data; /* Application private data */
562 struct ip_vs_seq in_seq; /* incoming seq. struct */
563 struct ip_vs_seq out_seq; /* outgoing seq. struct */
564
565 const struct ip_vs_pe *pe;
566 char *pe_data;
567 __u8 pe_data_len;
568
569 struct rcu_head rcu_head;
570};
571
572/* Extended internal versions of struct ip_vs_service_user and ip_vs_dest_user
573 * for IPv6 support.
574 *
575 * We need these to conveniently pass around service and destination
576 * options, but unfortunately, we also need to keep the old definitions to
577 * maintain userspace backwards compatibility for the setsockopt interface.
578 */
579struct ip_vs_service_user_kern {
580 /* virtual service addresses */
581 u16 af;
582 u16 protocol;
583 union nf_inet_addr addr; /* virtual ip address */
584 __be16 port;
585 u32 fwmark; /* firwall mark of service */
586
587 /* virtual service options */
588 char *sched_name;
589 char *pe_name;
590 unsigned int flags; /* virtual service flags */
591 unsigned int timeout; /* persistent timeout in sec */
592 __be32 netmask; /* persistent netmask or plen */
593};
594
595
596struct ip_vs_dest_user_kern {
597 /* destination server address */
598 union nf_inet_addr addr;
599 __be16 port;
600
601 /* real server options */
602 unsigned int conn_flags; /* connection flags */
603 int weight; /* destination weight */
604
605 /* thresholds for active connections */
606 u32 u_threshold; /* upper threshold */
607 u32 l_threshold; /* lower threshold */
608
609 /* Address family of addr */
610 u16 af;
611};
612
613
614/*
615 * The information about the virtual service offered to the net and the
616 * forwarding entries.
617 */
618struct ip_vs_service {
619 struct hlist_node s_list; /* for normal service table */
620 struct hlist_node f_list; /* for fwmark-based service table */
621 atomic_t refcnt; /* reference counter */
622
623 u16 af; /* address family */
624 __u16 protocol; /* which protocol (TCP/UDP) */
625 union nf_inet_addr addr; /* IP address for virtual service */
626 __be16 port; /* port number for the service */
627 __u32 fwmark; /* firewall mark of the service */
628 unsigned int flags; /* service status flags */
629 unsigned int timeout; /* persistent timeout in ticks */
630 __be32 netmask; /* grouping granularity, mask/plen */
631 struct netns_ipvs *ipvs;
632
633 struct list_head destinations; /* real server d-linked list */
634 __u32 num_dests; /* number of servers */
635 struct ip_vs_stats stats; /* statistics for the service */
636
637 /* for scheduling */
638 struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */
639 spinlock_t sched_lock; /* lock sched_data */
640 void *sched_data; /* scheduler application data */
641
642 /* alternate persistence engine */
643 struct ip_vs_pe __rcu *pe;
644
645 struct rcu_head rcu_head;
646};
647
648/* Information for cached dst */
649struct ip_vs_dest_dst {
650 struct dst_entry *dst_cache; /* destination cache entry */
651 u32 dst_cookie;
652 union nf_inet_addr dst_saddr;
653 struct rcu_head rcu_head;
654};
655
656/* The real server destination forwarding entry with ip address, port number,
657 * and so on.
658 */
659struct ip_vs_dest {
660 struct list_head n_list; /* for the dests in the service */
661 struct hlist_node d_list; /* for table with all the dests */
662
663 u16 af; /* address family */
664 __be16 port; /* port number of the server */
665 union nf_inet_addr addr; /* IP address of the server */
666 volatile unsigned int flags; /* dest status flags */
667 atomic_t conn_flags; /* flags to copy to conn */
668 atomic_t weight; /* server weight */
669
670 atomic_t refcnt; /* reference counter */
671 struct ip_vs_stats stats; /* statistics */
672 unsigned long idle_start; /* start time, jiffies */
673
674 /* connection counters and thresholds */
675 atomic_t activeconns; /* active connections */
676 atomic_t inactconns; /* inactive connections */
677 atomic_t persistconns; /* persistent connections */
678 __u32 u_threshold; /* upper threshold */
679 __u32 l_threshold; /* lower threshold */
680
681 /* for destination cache */
682 spinlock_t dst_lock; /* lock of dst_cache */
683 struct ip_vs_dest_dst __rcu *dest_dst; /* cached dst info */
684
685 /* for virtual service */
686 struct ip_vs_service __rcu *svc; /* service it belongs to */
687 __u16 protocol; /* which protocol (TCP/UDP) */
688 __be16 vport; /* virtual port number */
689 union nf_inet_addr vaddr; /* virtual IP address */
690 __u32 vfwmark; /* firewall mark of service */
691
692 struct list_head t_list; /* in dest_trash */
693 unsigned int in_rs_table:1; /* we are in rs_table */
694};
695
696/* The scheduler object */
697struct ip_vs_scheduler {
698 struct list_head n_list; /* d-linked list head */
699 char *name; /* scheduler name */
700 atomic_t refcnt; /* reference counter */
701 struct module *module; /* THIS_MODULE/NULL */
702
703 /* scheduler initializing service */
704 int (*init_service)(struct ip_vs_service *svc);
705 /* scheduling service finish */
706 void (*done_service)(struct ip_vs_service *svc);
707 /* dest is linked */
708 int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
709 /* dest is unlinked */
710 int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
711 /* dest is updated */
712 int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
713
714 /* selecting a server from the given service */
715 struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc,
716 const struct sk_buff *skb,
717 struct ip_vs_iphdr *iph);
718};
719
720/* The persistence engine object */
721struct ip_vs_pe {
722 struct list_head n_list; /* d-linked list head */
723 char *name; /* scheduler name */
724 atomic_t refcnt; /* reference counter */
725 struct module *module; /* THIS_MODULE/NULL */
726
727 /* get the connection template, if any */
728 int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb);
729 bool (*ct_match)(const struct ip_vs_conn_param *p,
730 struct ip_vs_conn *ct);
731 u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval,
732 bool inverse);
733 int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf);
734};
735
736/* The application module object (a.k.a. app incarnation) */
737struct ip_vs_app {
738 struct list_head a_list; /* member in app list */
739 int type; /* IP_VS_APP_TYPE_xxx */
740 char *name; /* application module name */
741 __u16 protocol;
742 struct module *module; /* THIS_MODULE/NULL */
743 struct list_head incs_list; /* list of incarnations */
744
745 /* members for application incarnations */
746 struct list_head p_list; /* member in proto app list */
747 struct ip_vs_app *app; /* its real application */
748 __be16 port; /* port number in net order */
749 atomic_t usecnt; /* usage counter */
750 struct rcu_head rcu_head;
751
752 /* output hook: Process packet in inout direction, diff set for TCP.
753 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
754 * 2=Mangled but checksum was not updated
755 */
756 int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *,
757 struct sk_buff *, int *diff);
758
759 /* input hook: Process packet in outin direction, diff set for TCP.
760 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
761 * 2=Mangled but checksum was not updated
762 */
763 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *,
764 struct sk_buff *, int *diff);
765
766 /* ip_vs_app initializer */
767 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *);
768
769 /* ip_vs_app finish */
770 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *);
771
772
773 /* not used now */
774 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *,
775 struct ip_vs_protocol *);
776
777 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *);
778
779 int * timeout_table;
780 int * timeouts;
781 int timeouts_size;
782
783 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app,
784 int *verdict, struct ip_vs_conn **cpp);
785
786 struct ip_vs_conn *
787 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app,
788 const struct iphdr *iph, int inverse);
789
790 struct ip_vs_conn *
791 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app,
792 const struct iphdr *iph, int inverse);
793
794 int (*state_transition)(struct ip_vs_conn *cp, int direction,
795 const struct sk_buff *skb,
796 struct ip_vs_app *app);
797
798 void (*timeout_change)(struct ip_vs_app *app, int flags);
799};
800
801struct ipvs_master_sync_state {
802 struct list_head sync_queue;
803 struct ip_vs_sync_buff *sync_buff;
804 unsigned long sync_queue_len;
805 unsigned int sync_queue_delay;
806 struct task_struct *master_thread;
807 struct delayed_work master_wakeup_work;
808 struct netns_ipvs *ipvs;
809};
810
811/* How much time to keep dests in trash */
812#define IP_VS_DEST_TRASH_PERIOD (120 * HZ)
813
814struct ipvs_sync_daemon_cfg {
815 union nf_inet_addr mcast_group;
816 int syncid;
817 u16 sync_maxlen;
818 u16 mcast_port;
819 u8 mcast_af;
820 u8 mcast_ttl;
821 /* multicast interface name */
822 char mcast_ifn[IP_VS_IFNAME_MAXLEN];
823};
824
825/* IPVS in network namespace */
826struct netns_ipvs {
827 int gen; /* Generation */
828 int enable; /* enable like nf_hooks do */
829 /* Hash table: for real service lookups */
830 #define IP_VS_RTAB_BITS 4
831 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
832 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
833
834 struct hlist_head rs_table[IP_VS_RTAB_SIZE];
835 /* ip_vs_app */
836 struct list_head app_list;
837 /* ip_vs_proto */
838 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */
839 struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE];
840 /* ip_vs_proto_tcp */
841#ifdef CONFIG_IP_VS_PROTO_TCP
842 #define TCP_APP_TAB_BITS 4
843 #define TCP_APP_TAB_SIZE (1 << TCP_APP_TAB_BITS)
844 #define TCP_APP_TAB_MASK (TCP_APP_TAB_SIZE - 1)
845 struct list_head tcp_apps[TCP_APP_TAB_SIZE];
846#endif
847 /* ip_vs_proto_udp */
848#ifdef CONFIG_IP_VS_PROTO_UDP
849 #define UDP_APP_TAB_BITS 4
850 #define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS)
851 #define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1)
852 struct list_head udp_apps[UDP_APP_TAB_SIZE];
853#endif
854 /* ip_vs_proto_sctp */
855#ifdef CONFIG_IP_VS_PROTO_SCTP
856 #define SCTP_APP_TAB_BITS 4
857 #define SCTP_APP_TAB_SIZE (1 << SCTP_APP_TAB_BITS)
858 #define SCTP_APP_TAB_MASK (SCTP_APP_TAB_SIZE - 1)
859 /* Hash table for SCTP application incarnations */
860 struct list_head sctp_apps[SCTP_APP_TAB_SIZE];
861#endif
862 /* ip_vs_conn */
863 atomic_t conn_count; /* connection counter */
864
865 /* ip_vs_ctl */
866 struct ip_vs_stats tot_stats; /* Statistics & est. */
867
868 int num_services; /* no of virtual services */
869
870 /* Trash for destinations */
871 struct list_head dest_trash;
872 spinlock_t dest_trash_lock;
873 struct timer_list dest_trash_timer; /* expiration timer */
874 /* Service counters */
875 atomic_t ftpsvc_counter;
876 atomic_t nullsvc_counter;
877
878#ifdef CONFIG_SYSCTL
879 /* 1/rate drop and drop-entry variables */
880 struct delayed_work defense_work; /* Work handler */
881 int drop_rate;
882 int drop_counter;
883 atomic_t dropentry;
884 /* locks in ctl.c */
885 spinlock_t dropentry_lock; /* drop entry handling */
886 spinlock_t droppacket_lock; /* drop packet handling */
887 spinlock_t securetcp_lock; /* state and timeout tables */
888
889 /* sys-ctl struct */
890 struct ctl_table_header *sysctl_hdr;
891 struct ctl_table *sysctl_tbl;
892#endif
893
894 /* sysctl variables */
895 int sysctl_amemthresh;
896 int sysctl_am_droprate;
897 int sysctl_drop_entry;
898 int sysctl_drop_packet;
899 int sysctl_secure_tcp;
900#ifdef CONFIG_IP_VS_NFCT
901 int sysctl_conntrack;
902#endif
903 int sysctl_snat_reroute;
904 int sysctl_sync_ver;
905 int sysctl_sync_ports;
906 int sysctl_sync_persist_mode;
907 unsigned long sysctl_sync_qlen_max;
908 int sysctl_sync_sock_size;
909 int sysctl_cache_bypass;
910 int sysctl_expire_nodest_conn;
911 int sysctl_sloppy_tcp;
912 int sysctl_sloppy_sctp;
913 int sysctl_expire_quiescent_template;
914 int sysctl_sync_threshold[2];
915 unsigned int sysctl_sync_refresh_period;
916 int sysctl_sync_retries;
917 int sysctl_nat_icmp_send;
918 int sysctl_pmtu_disc;
919 int sysctl_backup_only;
920 int sysctl_conn_reuse_mode;
921 int sysctl_schedule_icmp;
922 int sysctl_ignore_tunneled;
923
924 /* ip_vs_lblc */
925 int sysctl_lblc_expiration;
926 struct ctl_table_header *lblc_ctl_header;
927 struct ctl_table *lblc_ctl_table;
928 /* ip_vs_lblcr */
929 int sysctl_lblcr_expiration;
930 struct ctl_table_header *lblcr_ctl_header;
931 struct ctl_table *lblcr_ctl_table;
932 /* ip_vs_est */
933 struct list_head est_list; /* estimator list */
934 spinlock_t est_lock;
935 struct timer_list est_timer; /* Estimation timer */
936 /* ip_vs_sync */
937 spinlock_t sync_lock;
938 struct ipvs_master_sync_state *ms;
939 spinlock_t sync_buff_lock;
940 struct task_struct **backup_threads;
941 int threads_mask;
942 volatile int sync_state;
943 struct mutex sync_mutex;
944 struct ipvs_sync_daemon_cfg mcfg; /* Master Configuration */
945 struct ipvs_sync_daemon_cfg bcfg; /* Backup Configuration */
946 /* net name space ptr */
947 struct net *net; /* Needed by timer routines */
948 /* Number of heterogeneous destinations, needed becaus heterogeneous
949 * are not supported when synchronization is enabled.
950 */
951 unsigned int mixed_address_family_dests;
952};
953
954#define DEFAULT_SYNC_THRESHOLD 3
955#define DEFAULT_SYNC_PERIOD 50
956#define DEFAULT_SYNC_VER 1
957#define DEFAULT_SLOPPY_TCP 0
958#define DEFAULT_SLOPPY_SCTP 0
959#define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ)
960#define DEFAULT_SYNC_RETRIES 0
961#define IPVS_SYNC_WAKEUP_RATE 8
962#define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4)
963#define IPVS_SYNC_SEND_DELAY (HZ / 50)
964#define IPVS_SYNC_CHECK_PERIOD HZ
965#define IPVS_SYNC_FLUSH_TIME (HZ * 2)
966#define IPVS_SYNC_PORTS_MAX (1 << 6)
967
968#ifdef CONFIG_SYSCTL
969
970static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
971{
972 return ipvs->sysctl_sync_threshold[0];
973}
974
975static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
976{
977 return ACCESS_ONCE(ipvs->sysctl_sync_threshold[1]);
978}
979
980static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
981{
982 return ACCESS_ONCE(ipvs->sysctl_sync_refresh_period);
983}
984
985static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
986{
987 return ipvs->sysctl_sync_retries;
988}
989
990static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
991{
992 return ipvs->sysctl_sync_ver;
993}
994
995static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
996{
997 return ipvs->sysctl_sloppy_tcp;
998}
999
1000static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1001{
1002 return ipvs->sysctl_sloppy_sctp;
1003}
1004
1005static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1006{
1007 return ACCESS_ONCE(ipvs->sysctl_sync_ports);
1008}
1009
1010static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1011{
1012 return ipvs->sysctl_sync_persist_mode;
1013}
1014
1015static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1016{
1017 return ipvs->sysctl_sync_qlen_max;
1018}
1019
1020static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1021{
1022 return ipvs->sysctl_sync_sock_size;
1023}
1024
1025static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1026{
1027 return ipvs->sysctl_pmtu_disc;
1028}
1029
1030static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1031{
1032 return ipvs->sync_state & IP_VS_STATE_BACKUP &&
1033 ipvs->sysctl_backup_only;
1034}
1035
1036static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1037{
1038 return ipvs->sysctl_conn_reuse_mode;
1039}
1040
1041static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1042{
1043 return ipvs->sysctl_schedule_icmp;
1044}
1045
1046static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1047{
1048 return ipvs->sysctl_ignore_tunneled;
1049}
1050
1051static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1052{
1053 return ipvs->sysctl_cache_bypass;
1054}
1055
1056#else
1057
1058static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1059{
1060 return DEFAULT_SYNC_THRESHOLD;
1061}
1062
1063static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1064{
1065 return DEFAULT_SYNC_PERIOD;
1066}
1067
1068static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1069{
1070 return DEFAULT_SYNC_REFRESH_PERIOD;
1071}
1072
1073static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1074{
1075 return DEFAULT_SYNC_RETRIES & 3;
1076}
1077
1078static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1079{
1080 return DEFAULT_SYNC_VER;
1081}
1082
1083static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1084{
1085 return DEFAULT_SLOPPY_TCP;
1086}
1087
1088static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1089{
1090 return DEFAULT_SLOPPY_SCTP;
1091}
1092
1093static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1094{
1095 return 1;
1096}
1097
1098static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1099{
1100 return 0;
1101}
1102
1103static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1104{
1105 return IPVS_SYNC_QLEN_MAX;
1106}
1107
1108static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1109{
1110 return 0;
1111}
1112
1113static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1114{
1115 return 1;
1116}
1117
1118static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1119{
1120 return 0;
1121}
1122
1123static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1124{
1125 return 1;
1126}
1127
1128static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1129{
1130 return 0;
1131}
1132
1133static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1134{
1135 return 0;
1136}
1137
1138static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1139{
1140 return 0;
1141}
1142
1143#endif
1144
1145/* IPVS core functions
1146 * (from ip_vs_core.c)
1147 */
1148const char *ip_vs_proto_name(unsigned int proto);
1149void ip_vs_init_hash_table(struct list_head *table, int rows);
1150#define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t)))
1151
1152#define IP_VS_APP_TYPE_FTP 1
1153
1154/* ip_vs_conn handling functions
1155 * (from ip_vs_conn.c)
1156 */
1157enum {
1158 IP_VS_DIR_INPUT = 0,
1159 IP_VS_DIR_OUTPUT,
1160 IP_VS_DIR_INPUT_ONLY,
1161 IP_VS_DIR_LAST,
1162};
1163
1164static inline void ip_vs_conn_fill_param(struct netns_ipvs *ipvs, int af, int protocol,
1165 const union nf_inet_addr *caddr,
1166 __be16 cport,
1167 const union nf_inet_addr *vaddr,
1168 __be16 vport,
1169 struct ip_vs_conn_param *p)
1170{
1171 p->ipvs = ipvs;
1172 p->af = af;
1173 p->protocol = protocol;
1174 p->caddr = caddr;
1175 p->cport = cport;
1176 p->vaddr = vaddr;
1177 p->vport = vport;
1178 p->pe = NULL;
1179 p->pe_data = NULL;
1180}
1181
1182struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p);
1183struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p);
1184
1185struct ip_vs_conn * ip_vs_conn_in_get_proto(struct netns_ipvs *ipvs, int af,
1186 const struct sk_buff *skb,
1187 const struct ip_vs_iphdr *iph);
1188
1189struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p);
1190
1191struct ip_vs_conn * ip_vs_conn_out_get_proto(struct netns_ipvs *ipvs, int af,
1192 const struct sk_buff *skb,
1193 const struct ip_vs_iphdr *iph);
1194
1195/* Get reference to gain full access to conn.
1196 * By default, RCU read-side critical sections have access only to
1197 * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference.
1198 */
1199static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp)
1200{
1201 return atomic_inc_not_zero(&cp->refcnt);
1202}
1203
1204/* put back the conn without restarting its timer */
1205static inline void __ip_vs_conn_put(struct ip_vs_conn *cp)
1206{
1207 smp_mb__before_atomic();
1208 atomic_dec(&cp->refcnt);
1209}
1210void ip_vs_conn_put(struct ip_vs_conn *cp);
1211void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport);
1212
1213struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, int dest_af,
1214 const union nf_inet_addr *daddr,
1215 __be16 dport, unsigned int flags,
1216 struct ip_vs_dest *dest, __u32 fwmark);
1217void ip_vs_conn_expire_now(struct ip_vs_conn *cp);
1218
1219const char *ip_vs_state_name(__u16 proto, int state);
1220
1221void ip_vs_tcp_conn_listen(struct ip_vs_conn *cp);
1222int ip_vs_check_template(struct ip_vs_conn *ct);
1223void ip_vs_random_dropentry(struct netns_ipvs *ipvs);
1224int ip_vs_conn_init(void);
1225void ip_vs_conn_cleanup(void);
1226
1227static inline void ip_vs_control_del(struct ip_vs_conn *cp)
1228{
1229 struct ip_vs_conn *ctl_cp = cp->control;
1230 if (!ctl_cp) {
1231 IP_VS_ERR_BUF("request control DEL for uncontrolled: "
1232 "%s:%d to %s:%d\n",
1233 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1234 ntohs(cp->cport),
1235 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1236 ntohs(cp->vport));
1237
1238 return;
1239 }
1240
1241 IP_VS_DBG_BUF(7, "DELeting control for: "
1242 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1243 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1244 ntohs(cp->cport),
1245 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1246 ntohs(ctl_cp->cport));
1247
1248 cp->control = NULL;
1249 if (atomic_read(&ctl_cp->n_control) == 0) {
1250 IP_VS_ERR_BUF("BUG control DEL with n=0 : "
1251 "%s:%d to %s:%d\n",
1252 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1253 ntohs(cp->cport),
1254 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1255 ntohs(cp->vport));
1256
1257 return;
1258 }
1259 atomic_dec(&ctl_cp->n_control);
1260}
1261
1262static inline void
1263ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp)
1264{
1265 if (cp->control) {
1266 IP_VS_ERR_BUF("request control ADD for already controlled: "
1267 "%s:%d to %s:%d\n",
1268 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1269 ntohs(cp->cport),
1270 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1271 ntohs(cp->vport));
1272
1273 ip_vs_control_del(cp);
1274 }
1275
1276 IP_VS_DBG_BUF(7, "ADDing control for: "
1277 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1278 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1279 ntohs(cp->cport),
1280 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1281 ntohs(ctl_cp->cport));
1282
1283 cp->control = ctl_cp;
1284 atomic_inc(&ctl_cp->n_control);
1285}
1286
1287/* IPVS netns init & cleanup functions */
1288int ip_vs_estimator_net_init(struct netns_ipvs *ipvs);
1289int ip_vs_control_net_init(struct netns_ipvs *ipvs);
1290int ip_vs_protocol_net_init(struct netns_ipvs *ipvs);
1291int ip_vs_app_net_init(struct netns_ipvs *ipvs);
1292int ip_vs_conn_net_init(struct netns_ipvs *ipvs);
1293int ip_vs_sync_net_init(struct netns_ipvs *ipvs);
1294void ip_vs_conn_net_cleanup(struct netns_ipvs *ipvs);
1295void ip_vs_app_net_cleanup(struct netns_ipvs *ipvs);
1296void ip_vs_protocol_net_cleanup(struct netns_ipvs *ipvs);
1297void ip_vs_control_net_cleanup(struct netns_ipvs *ipvs);
1298void ip_vs_estimator_net_cleanup(struct netns_ipvs *ipvs);
1299void ip_vs_sync_net_cleanup(struct netns_ipvs *ipvs);
1300void ip_vs_service_net_cleanup(struct netns_ipvs *ipvs);
1301
1302/* IPVS application functions
1303 * (from ip_vs_app.c)
1304 */
1305#define IP_VS_APP_MAX_PORTS 8
1306struct ip_vs_app *register_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1307void unregister_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1308int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1309void ip_vs_unbind_app(struct ip_vs_conn *cp);
1310int register_ip_vs_app_inc(struct netns_ipvs *ipvs, struct ip_vs_app *app, __u16 proto,
1311 __u16 port);
1312int ip_vs_app_inc_get(struct ip_vs_app *inc);
1313void ip_vs_app_inc_put(struct ip_vs_app *inc);
1314
1315int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb);
1316int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb);
1317
1318int register_ip_vs_pe(struct ip_vs_pe *pe);
1319int unregister_ip_vs_pe(struct ip_vs_pe *pe);
1320struct ip_vs_pe *ip_vs_pe_getbyname(const char *name);
1321struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name);
1322
1323/* Use a #define to avoid all of module.h just for these trivial ops */
1324#define ip_vs_pe_get(pe) \
1325 if (pe && pe->module) \
1326 __module_get(pe->module);
1327
1328#define ip_vs_pe_put(pe) \
1329 if (pe && pe->module) \
1330 module_put(pe->module);
1331
1332/* IPVS protocol functions (from ip_vs_proto.c) */
1333int ip_vs_protocol_init(void);
1334void ip_vs_protocol_cleanup(void);
1335void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags);
1336int *ip_vs_create_timeout_table(int *table, int size);
1337int ip_vs_set_state_timeout(int *table, int num, const char *const *names,
1338 const char *name, int to);
1339void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
1340 const struct sk_buff *skb, int offset,
1341 const char *msg);
1342
1343extern struct ip_vs_protocol ip_vs_protocol_tcp;
1344extern struct ip_vs_protocol ip_vs_protocol_udp;
1345extern struct ip_vs_protocol ip_vs_protocol_icmp;
1346extern struct ip_vs_protocol ip_vs_protocol_esp;
1347extern struct ip_vs_protocol ip_vs_protocol_ah;
1348extern struct ip_vs_protocol ip_vs_protocol_sctp;
1349
1350/* Registering/unregistering scheduler functions
1351 * (from ip_vs_sched.c)
1352 */
1353int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1354int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1355int ip_vs_bind_scheduler(struct ip_vs_service *svc,
1356 struct ip_vs_scheduler *scheduler);
1357void ip_vs_unbind_scheduler(struct ip_vs_service *svc,
1358 struct ip_vs_scheduler *sched);
1359struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name);
1360void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler);
1361struct ip_vs_conn *
1362ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb,
1363 struct ip_vs_proto_data *pd, int *ignored,
1364 struct ip_vs_iphdr *iph);
1365int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb,
1366 struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph);
1367
1368void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg);
1369
1370/* IPVS control data and functions (from ip_vs_ctl.c) */
1371extern struct ip_vs_stats ip_vs_stats;
1372extern int sysctl_ip_vs_sync_ver;
1373
1374struct ip_vs_service *
1375ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
1376 const union nf_inet_addr *vaddr, __be16 vport);
1377
1378bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1379 const union nf_inet_addr *daddr, __be16 dport);
1380
1381int ip_vs_use_count_inc(void);
1382void ip_vs_use_count_dec(void);
1383int ip_vs_register_nl_ioctl(void);
1384void ip_vs_unregister_nl_ioctl(void);
1385int ip_vs_control_init(void);
1386void ip_vs_control_cleanup(void);
1387struct ip_vs_dest *
1388ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
1389 const union nf_inet_addr *daddr, __be16 dport,
1390 const union nf_inet_addr *vaddr, __be16 vport,
1391 __u16 protocol, __u32 fwmark, __u32 flags);
1392void ip_vs_try_bind_dest(struct ip_vs_conn *cp);
1393
1394static inline void ip_vs_dest_hold(struct ip_vs_dest *dest)
1395{
1396 atomic_inc(&dest->refcnt);
1397}
1398
1399static inline void ip_vs_dest_put(struct ip_vs_dest *dest)
1400{
1401 smp_mb__before_atomic();
1402 atomic_dec(&dest->refcnt);
1403}
1404
1405static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest)
1406{
1407 if (atomic_dec_return(&dest->refcnt) < 0)
1408 kfree(dest);
1409}
1410
1411/* IPVS sync daemon data and function prototypes
1412 * (from ip_vs_sync.c)
1413 */
1414int start_sync_thread(struct netns_ipvs *ipvs, struct ipvs_sync_daemon_cfg *cfg,
1415 int state);
1416int stop_sync_thread(struct netns_ipvs *ipvs, int state);
1417void ip_vs_sync_conn(struct netns_ipvs *ipvs, struct ip_vs_conn *cp, int pkts);
1418
1419/* IPVS rate estimator prototypes (from ip_vs_est.c) */
1420void ip_vs_start_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1421void ip_vs_stop_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1422void ip_vs_zero_estimator(struct ip_vs_stats *stats);
1423void ip_vs_read_estimator(struct ip_vs_kstats *dst, struct ip_vs_stats *stats);
1424
1425/* Various IPVS packet transmitters (from ip_vs_xmit.c) */
1426int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1427 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1428int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1429 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1430int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1431 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1432int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1433 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1434int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1435 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1436int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1437 struct ip_vs_protocol *pp, int offset,
1438 unsigned int hooknum, struct ip_vs_iphdr *iph);
1439void ip_vs_dest_dst_rcu_free(struct rcu_head *head);
1440
1441#ifdef CONFIG_IP_VS_IPV6
1442int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1443 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1444int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1445 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1446int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1447 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1448int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1449 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1450int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1451 struct ip_vs_protocol *pp, int offset,
1452 unsigned int hooknum, struct ip_vs_iphdr *iph);
1453#endif
1454
1455#ifdef CONFIG_SYSCTL
1456/* This is a simple mechanism to ignore packets when
1457 * we are loaded. Just set ip_vs_drop_rate to 'n' and
1458 * we start to drop 1/rate of the packets
1459 */
1460static inline int ip_vs_todrop(struct netns_ipvs *ipvs)
1461{
1462 if (!ipvs->drop_rate)
1463 return 0;
1464 if (--ipvs->drop_counter > 0)
1465 return 0;
1466 ipvs->drop_counter = ipvs->drop_rate;
1467 return 1;
1468}
1469#else
1470static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; }
1471#endif
1472
1473/* ip_vs_fwd_tag returns the forwarding tag of the connection */
1474#define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK)
1475
1476static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp)
1477{
1478 char fwd;
1479
1480 switch (IP_VS_FWD_METHOD(cp)) {
1481 case IP_VS_CONN_F_MASQ:
1482 fwd = 'M'; break;
1483 case IP_VS_CONN_F_LOCALNODE:
1484 fwd = 'L'; break;
1485 case IP_VS_CONN_F_TUNNEL:
1486 fwd = 'T'; break;
1487 case IP_VS_CONN_F_DROUTE:
1488 fwd = 'R'; break;
1489 case IP_VS_CONN_F_BYPASS:
1490 fwd = 'B'; break;
1491 default:
1492 fwd = '?'; break;
1493 }
1494 return fwd;
1495}
1496
1497void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp,
1498 struct ip_vs_conn *cp, int dir);
1499
1500#ifdef CONFIG_IP_VS_IPV6
1501void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp,
1502 struct ip_vs_conn *cp, int dir);
1503#endif
1504
1505__sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset);
1506
1507static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum)
1508{
1509 __be32 diff[2] = { ~old, new };
1510
1511 return csum_partial(diff, sizeof(diff), oldsum);
1512}
1513
1514#ifdef CONFIG_IP_VS_IPV6
1515static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new,
1516 __wsum oldsum)
1517{
1518 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0],
1519 new[3], new[2], new[1], new[0] };
1520
1521 return csum_partial(diff, sizeof(diff), oldsum);
1522}
1523#endif
1524
1525static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum)
1526{
1527 __be16 diff[2] = { ~old, new };
1528
1529 return csum_partial(diff, sizeof(diff), oldsum);
1530}
1531
1532/* Forget current conntrack (unconfirmed) and attach notrack entry */
1533static inline void ip_vs_notrack(struct sk_buff *skb)
1534{
1535#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1536 enum ip_conntrack_info ctinfo;
1537 struct nf_conn *ct = nf_ct_get(skb, &ctinfo);
1538
1539 if (!ct || !nf_ct_is_untracked(ct)) {
1540 nf_conntrack_put(skb->nfct);
1541 skb->nfct = &nf_ct_untracked_get()->ct_general;
1542 skb->nfctinfo = IP_CT_NEW;
1543 nf_conntrack_get(skb->nfct);
1544 }
1545#endif
1546}
1547
1548#ifdef CONFIG_IP_VS_NFCT
1549/* Netfilter connection tracking
1550 * (from ip_vs_nfct.c)
1551 */
1552static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1553{
1554#ifdef CONFIG_SYSCTL
1555 return ipvs->sysctl_conntrack;
1556#else
1557 return 0;
1558#endif
1559}
1560
1561void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp,
1562 int outin);
1563int ip_vs_confirm_conntrack(struct sk_buff *skb);
1564void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct,
1565 struct ip_vs_conn *cp, u_int8_t proto,
1566 const __be16 port, int from_rs);
1567void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp);
1568
1569#else
1570
1571static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1572{
1573 return 0;
1574}
1575
1576static inline void ip_vs_update_conntrack(struct sk_buff *skb,
1577 struct ip_vs_conn *cp, int outin)
1578{
1579}
1580
1581static inline int ip_vs_confirm_conntrack(struct sk_buff *skb)
1582{
1583 return NF_ACCEPT;
1584}
1585
1586static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp)
1587{
1588}
1589#endif /* CONFIG_IP_VS_NFCT */
1590
1591/* Really using conntrack? */
1592static inline bool ip_vs_conn_uses_conntrack(struct ip_vs_conn *cp,
1593 struct sk_buff *skb)
1594{
1595#ifdef CONFIG_IP_VS_NFCT
1596 enum ip_conntrack_info ctinfo;
1597 struct nf_conn *ct;
1598
1599 if (!(cp->flags & IP_VS_CONN_F_NFCT))
1600 return false;
1601 ct = nf_ct_get(skb, &ctinfo);
1602 if (ct && !nf_ct_is_untracked(ct))
1603 return true;
1604#endif
1605 return false;
1606}
1607
1608static inline int
1609ip_vs_dest_conn_overhead(struct ip_vs_dest *dest)
1610{
1611 /* We think the overhead of processing active connections is 256
1612 * times higher than that of inactive connections in average. (This
1613 * 256 times might not be accurate, we will change it later) We
1614 * use the following formula to estimate the overhead now:
1615 * dest->activeconns*256 + dest->inactconns
1616 */
1617 return (atomic_read(&dest->activeconns) << 8) +
1618 atomic_read(&dest->inactconns);
1619}
1620
1621#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/* The port number of FTP service (in network order). */
269#define FTPPORT cpu_to_be16(21)
270#define FTPDATA cpu_to_be16(20)
271
272/* TCP State Values */
273enum {
274 IP_VS_TCP_S_NONE = 0,
275 IP_VS_TCP_S_ESTABLISHED,
276 IP_VS_TCP_S_SYN_SENT,
277 IP_VS_TCP_S_SYN_RECV,
278 IP_VS_TCP_S_FIN_WAIT,
279 IP_VS_TCP_S_TIME_WAIT,
280 IP_VS_TCP_S_CLOSE,
281 IP_VS_TCP_S_CLOSE_WAIT,
282 IP_VS_TCP_S_LAST_ACK,
283 IP_VS_TCP_S_LISTEN,
284 IP_VS_TCP_S_SYNACK,
285 IP_VS_TCP_S_LAST
286};
287
288/* UDP State Values */
289enum {
290 IP_VS_UDP_S_NORMAL,
291 IP_VS_UDP_S_LAST,
292};
293
294/* ICMP State Values */
295enum {
296 IP_VS_ICMP_S_NORMAL,
297 IP_VS_ICMP_S_LAST,
298};
299
300/* SCTP State Values */
301enum ip_vs_sctp_states {
302 IP_VS_SCTP_S_NONE,
303 IP_VS_SCTP_S_INIT1,
304 IP_VS_SCTP_S_INIT,
305 IP_VS_SCTP_S_COOKIE_SENT,
306 IP_VS_SCTP_S_COOKIE_REPLIED,
307 IP_VS_SCTP_S_COOKIE_WAIT,
308 IP_VS_SCTP_S_COOKIE,
309 IP_VS_SCTP_S_COOKIE_ECHOED,
310 IP_VS_SCTP_S_ESTABLISHED,
311 IP_VS_SCTP_S_SHUTDOWN_SENT,
312 IP_VS_SCTP_S_SHUTDOWN_RECEIVED,
313 IP_VS_SCTP_S_SHUTDOWN_ACK_SENT,
314 IP_VS_SCTP_S_REJECTED,
315 IP_VS_SCTP_S_CLOSED,
316 IP_VS_SCTP_S_LAST
317};
318
319/* Connection templates use bits from state */
320#define IP_VS_CTPL_S_NONE 0x0000
321#define IP_VS_CTPL_S_ASSURED 0x0001
322#define IP_VS_CTPL_S_LAST 0x0002
323
324/* Delta sequence info structure
325 * Each ip_vs_conn has 2 (output AND input seq. changes).
326 * Only used in the VS/NAT.
327 */
328struct ip_vs_seq {
329 __u32 init_seq; /* Add delta from this seq */
330 __u32 delta; /* Delta in sequence numbers */
331 __u32 previous_delta; /* Delta in sequence numbers
332 * before last resized pkt */
333};
334
335/* counters per cpu */
336struct ip_vs_counters {
337 u64_stats_t conns; /* connections scheduled */
338 u64_stats_t inpkts; /* incoming packets */
339 u64_stats_t outpkts; /* outgoing packets */
340 u64_stats_t inbytes; /* incoming bytes */
341 u64_stats_t outbytes; /* outgoing bytes */
342};
343/* Stats per cpu */
344struct ip_vs_cpu_stats {
345 struct ip_vs_counters cnt;
346 struct u64_stats_sync syncp;
347};
348
349/* Default nice for estimator kthreads */
350#define IPVS_EST_NICE 0
351
352/* IPVS statistics objects */
353struct ip_vs_estimator {
354 struct hlist_node list;
355
356 u64 last_inbytes;
357 u64 last_outbytes;
358 u64 last_conns;
359 u64 last_inpkts;
360 u64 last_outpkts;
361
362 u64 cps;
363 u64 inpps;
364 u64 outpps;
365 u64 inbps;
366 u64 outbps;
367
368 s32 ktid:16, /* kthread ID, -1=temp list */
369 ktrow:8, /* row/tick ID for kthread */
370 ktcid:8; /* chain ID for kthread tick */
371};
372
373/*
374 * IPVS statistics object, 64-bit kernel version of struct ip_vs_stats_user
375 */
376struct ip_vs_kstats {
377 u64 conns; /* connections scheduled */
378 u64 inpkts; /* incoming packets */
379 u64 outpkts; /* outgoing packets */
380 u64 inbytes; /* incoming bytes */
381 u64 outbytes; /* outgoing bytes */
382
383 u64 cps; /* current connection rate */
384 u64 inpps; /* current in packet rate */
385 u64 outpps; /* current out packet rate */
386 u64 inbps; /* current in byte rate */
387 u64 outbps; /* current out byte rate */
388};
389
390struct ip_vs_stats {
391 struct ip_vs_kstats kstats; /* kernel statistics */
392 struct ip_vs_estimator est; /* estimator */
393 struct ip_vs_cpu_stats __percpu *cpustats; /* per cpu counters */
394 spinlock_t lock; /* spin lock */
395 struct ip_vs_kstats kstats0; /* reset values */
396};
397
398struct ip_vs_stats_rcu {
399 struct ip_vs_stats s;
400 struct rcu_head rcu_head;
401};
402
403int ip_vs_stats_init_alloc(struct ip_vs_stats *s);
404struct ip_vs_stats *ip_vs_stats_alloc(void);
405void ip_vs_stats_release(struct ip_vs_stats *stats);
406void ip_vs_stats_free(struct ip_vs_stats *stats);
407
408/* Process estimators in multiple timer ticks (20/50/100, see ktrow) */
409#define IPVS_EST_NTICKS 50
410/* Estimation uses a 2-second period containing ticks (in jiffies) */
411#define IPVS_EST_TICK ((2 * HZ) / IPVS_EST_NTICKS)
412
413/* Limit of CPU load per kthread (8 for 12.5%), ratio of CPU capacity (1/C).
414 * Value of 4 and above ensures kthreads will take work without exceeding
415 * the CPU capacity under different circumstances.
416 */
417#define IPVS_EST_LOAD_DIVISOR 8
418
419/* Kthreads should not have work that exceeds the CPU load above 50% */
420#define IPVS_EST_CPU_KTHREADS (IPVS_EST_LOAD_DIVISOR / 2)
421
422/* Desired number of chains per timer tick (chain load factor in 100us units),
423 * 48=4.8ms of 40ms tick (12% CPU usage):
424 * 2 sec * 1000 ms in sec * 10 (100us in ms) / 8 (12.5%) / 50
425 */
426#define IPVS_EST_CHAIN_FACTOR \
427 ALIGN_DOWN(2 * 1000 * 10 / IPVS_EST_LOAD_DIVISOR / IPVS_EST_NTICKS, 8)
428
429/* Compiled number of chains per tick
430 * The defines should match cond_resched_rcu
431 */
432#if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
433#define IPVS_EST_TICK_CHAINS IPVS_EST_CHAIN_FACTOR
434#else
435#define IPVS_EST_TICK_CHAINS 1
436#endif
437
438#if IPVS_EST_NTICKS > 127
439#error Too many timer ticks for ktrow
440#endif
441
442/* Multiple chains processed in same tick */
443struct ip_vs_est_tick_data {
444 struct rcu_head rcu_head;
445 struct hlist_head chains[IPVS_EST_TICK_CHAINS];
446 DECLARE_BITMAP(present, IPVS_EST_TICK_CHAINS);
447 DECLARE_BITMAP(full, IPVS_EST_TICK_CHAINS);
448 int chain_len[IPVS_EST_TICK_CHAINS];
449};
450
451/* Context for estimation kthread */
452struct ip_vs_est_kt_data {
453 struct netns_ipvs *ipvs;
454 struct task_struct *task; /* task if running */
455 struct ip_vs_est_tick_data __rcu *ticks[IPVS_EST_NTICKS];
456 DECLARE_BITMAP(avail, IPVS_EST_NTICKS); /* tick has space for ests */
457 unsigned long est_timer; /* estimation timer (jiffies) */
458 struct ip_vs_stats *calc_stats; /* Used for calculation */
459 int tick_len[IPVS_EST_NTICKS]; /* est count */
460 int id; /* ktid per netns */
461 int chain_max; /* max ests per tick chain */
462 int tick_max; /* max ests per tick */
463 int est_count; /* attached ests to kthread */
464 int est_max_count; /* max ests per kthread */
465 int add_row; /* row for new ests */
466 int est_row; /* estimated row */
467};
468
469struct dst_entry;
470struct iphdr;
471struct ip_vs_conn;
472struct ip_vs_app;
473struct sk_buff;
474struct ip_vs_proto_data;
475
476struct ip_vs_protocol {
477 struct ip_vs_protocol *next;
478 char *name;
479 u16 protocol;
480 u16 num_states;
481 int dont_defrag;
482
483 void (*init)(struct ip_vs_protocol *pp);
484
485 void (*exit)(struct ip_vs_protocol *pp);
486
487 int (*init_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
488
489 void (*exit_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
490
491 int (*conn_schedule)(struct netns_ipvs *ipvs,
492 int af, struct sk_buff *skb,
493 struct ip_vs_proto_data *pd,
494 int *verdict, struct ip_vs_conn **cpp,
495 struct ip_vs_iphdr *iph);
496
497 struct ip_vs_conn *
498 (*conn_in_get)(struct netns_ipvs *ipvs,
499 int af,
500 const struct sk_buff *skb,
501 const struct ip_vs_iphdr *iph);
502
503 struct ip_vs_conn *
504 (*conn_out_get)(struct netns_ipvs *ipvs,
505 int af,
506 const struct sk_buff *skb,
507 const struct ip_vs_iphdr *iph);
508
509 int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
510 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
511
512 int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
513 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
514
515 const char *(*state_name)(int state);
516
517 void (*state_transition)(struct ip_vs_conn *cp, int direction,
518 const struct sk_buff *skb,
519 struct ip_vs_proto_data *pd);
520
521 int (*register_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
522
523 void (*unregister_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
524
525 int (*app_conn_bind)(struct ip_vs_conn *cp);
526
527 void (*debug_packet)(int af, struct ip_vs_protocol *pp,
528 const struct sk_buff *skb,
529 int offset,
530 const char *msg);
531
532 void (*timeout_change)(struct ip_vs_proto_data *pd, int flags);
533};
534
535/* protocol data per netns */
536struct ip_vs_proto_data {
537 struct ip_vs_proto_data *next;
538 struct ip_vs_protocol *pp;
539 int *timeout_table; /* protocol timeout table */
540 atomic_t appcnt; /* counter of proto app incs. */
541 struct tcp_states_t *tcp_state_table;
542};
543
544struct ip_vs_protocol *ip_vs_proto_get(unsigned short proto);
545struct ip_vs_proto_data *ip_vs_proto_data_get(struct netns_ipvs *ipvs,
546 unsigned short proto);
547
548struct ip_vs_conn_param {
549 struct netns_ipvs *ipvs;
550 const union nf_inet_addr *caddr;
551 const union nf_inet_addr *vaddr;
552 __be16 cport;
553 __be16 vport;
554 __u16 protocol;
555 u16 af;
556
557 const struct ip_vs_pe *pe;
558 char *pe_data;
559 __u8 pe_data_len;
560};
561
562/* IP_VS structure allocated for each dynamically scheduled connection */
563struct ip_vs_conn {
564 struct hlist_node c_list; /* hashed list heads */
565 /* Protocol, addresses and port numbers */
566 __be16 cport;
567 __be16 dport;
568 __be16 vport;
569 u16 af; /* address family */
570 union nf_inet_addr caddr; /* client address */
571 union nf_inet_addr vaddr; /* virtual address */
572 union nf_inet_addr daddr; /* destination address */
573 volatile __u32 flags; /* status flags */
574 __u16 protocol; /* Which protocol (TCP/UDP) */
575 __u16 daf; /* Address family of the dest */
576 struct netns_ipvs *ipvs;
577
578 /* counter and timer */
579 refcount_t refcnt; /* reference count */
580 struct timer_list timer; /* Expiration timer */
581 volatile unsigned long timeout; /* timeout */
582
583 /* Flags and state transition */
584 spinlock_t lock; /* lock for state transition */
585 volatile __u16 state; /* state info */
586 volatile __u16 old_state; /* old state, to be used for
587 * state transition triggered
588 * synchronization
589 */
590 __u32 fwmark; /* Fire wall mark from skb */
591 unsigned long sync_endtime; /* jiffies + sent_retries */
592
593 /* Control members */
594 struct ip_vs_conn *control; /* Master control connection */
595 atomic_t n_control; /* Number of controlled ones */
596 struct ip_vs_dest *dest; /* real server */
597 atomic_t in_pkts; /* incoming packet counter */
598
599 /* Packet transmitter for different forwarding methods. If it
600 * mangles the packet, it must return NF_DROP or better NF_STOLEN,
601 * otherwise this must be changed to a sk_buff **.
602 * NF_ACCEPT can be returned when destination is local.
603 */
604 int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp,
605 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
606
607 /* Note: we can group the following members into a structure,
608 * in order to save more space, and the following members are
609 * only used in VS/NAT anyway
610 */
611 struct ip_vs_app *app; /* bound ip_vs_app object */
612 void *app_data; /* Application private data */
613 struct_group(sync_conn_opt,
614 struct ip_vs_seq in_seq; /* incoming seq. struct */
615 struct ip_vs_seq out_seq; /* outgoing seq. struct */
616 );
617
618 const struct ip_vs_pe *pe;
619 char *pe_data;
620 __u8 pe_data_len;
621
622 struct rcu_head rcu_head;
623};
624
625/* Extended internal versions of struct ip_vs_service_user and ip_vs_dest_user
626 * for IPv6 support.
627 *
628 * We need these to conveniently pass around service and destination
629 * options, but unfortunately, we also need to keep the old definitions to
630 * maintain userspace backwards compatibility for the setsockopt interface.
631 */
632struct ip_vs_service_user_kern {
633 /* virtual service addresses */
634 u16 af;
635 u16 protocol;
636 union nf_inet_addr addr; /* virtual ip address */
637 __be16 port;
638 u32 fwmark; /* firewall mark of service */
639
640 /* virtual service options */
641 char *sched_name;
642 char *pe_name;
643 unsigned int flags; /* virtual service flags */
644 unsigned int timeout; /* persistent timeout in sec */
645 __be32 netmask; /* persistent netmask or plen */
646};
647
648
649struct ip_vs_dest_user_kern {
650 /* destination server address */
651 union nf_inet_addr addr;
652 __be16 port;
653
654 /* real server options */
655 unsigned int conn_flags; /* connection flags */
656 int weight; /* destination weight */
657
658 /* thresholds for active connections */
659 u32 u_threshold; /* upper threshold */
660 u32 l_threshold; /* lower threshold */
661
662 /* Address family of addr */
663 u16 af;
664
665 u16 tun_type; /* tunnel type */
666 __be16 tun_port; /* tunnel port */
667 u16 tun_flags; /* tunnel flags */
668};
669
670
671/*
672 * The information about the virtual service offered to the net and the
673 * forwarding entries.
674 */
675struct ip_vs_service {
676 struct hlist_node s_list; /* for normal service table */
677 struct hlist_node f_list; /* for fwmark-based service table */
678 atomic_t refcnt; /* reference counter */
679
680 u16 af; /* address family */
681 __u16 protocol; /* which protocol (TCP/UDP) */
682 union nf_inet_addr addr; /* IP address for virtual service */
683 __be16 port; /* port number for the service */
684 __u32 fwmark; /* firewall mark of the service */
685 unsigned int flags; /* service status flags */
686 unsigned int timeout; /* persistent timeout in ticks */
687 __be32 netmask; /* grouping granularity, mask/plen */
688 struct netns_ipvs *ipvs;
689
690 struct list_head destinations; /* real server d-linked list */
691 __u32 num_dests; /* number of servers */
692 struct ip_vs_stats stats; /* statistics for the service */
693
694 /* for scheduling */
695 struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */
696 spinlock_t sched_lock; /* lock sched_data */
697 void *sched_data; /* scheduler application data */
698
699 /* alternate persistence engine */
700 struct ip_vs_pe __rcu *pe;
701 int conntrack_afmask;
702
703 struct rcu_head rcu_head;
704};
705
706/* Information for cached dst */
707struct ip_vs_dest_dst {
708 struct dst_entry *dst_cache; /* destination cache entry */
709 u32 dst_cookie;
710 union nf_inet_addr dst_saddr;
711 struct rcu_head rcu_head;
712};
713
714/* The real server destination forwarding entry with ip address, port number,
715 * and so on.
716 */
717struct ip_vs_dest {
718 struct list_head n_list; /* for the dests in the service */
719 struct hlist_node d_list; /* for table with all the dests */
720
721 u16 af; /* address family */
722 __be16 port; /* port number of the server */
723 union nf_inet_addr addr; /* IP address of the server */
724 volatile unsigned int flags; /* dest status flags */
725 atomic_t conn_flags; /* flags to copy to conn */
726 atomic_t weight; /* server weight */
727 atomic_t last_weight; /* server latest weight */
728 __u16 tun_type; /* tunnel type */
729 __be16 tun_port; /* tunnel port */
730 __u16 tun_flags; /* tunnel flags */
731
732 refcount_t refcnt; /* reference counter */
733 struct ip_vs_stats stats; /* statistics */
734 unsigned long idle_start; /* start time, jiffies */
735
736 /* connection counters and thresholds */
737 atomic_t activeconns; /* active connections */
738 atomic_t inactconns; /* inactive connections */
739 atomic_t persistconns; /* persistent connections */
740 __u32 u_threshold; /* upper threshold */
741 __u32 l_threshold; /* lower threshold */
742
743 /* for destination cache */
744 spinlock_t dst_lock; /* lock of dst_cache */
745 struct ip_vs_dest_dst __rcu *dest_dst; /* cached dst info */
746
747 /* for virtual service */
748 struct ip_vs_service __rcu *svc; /* service it belongs to */
749 __u16 protocol; /* which protocol (TCP/UDP) */
750 __be16 vport; /* virtual port number */
751 union nf_inet_addr vaddr; /* virtual IP address */
752 __u32 vfwmark; /* firewall mark of service */
753
754 struct rcu_head rcu_head;
755 struct list_head t_list; /* in dest_trash */
756 unsigned int in_rs_table:1; /* we are in rs_table */
757};
758
759/* The scheduler object */
760struct ip_vs_scheduler {
761 struct list_head n_list; /* d-linked list head */
762 char *name; /* scheduler name */
763 atomic_t refcnt; /* reference counter */
764 struct module *module; /* THIS_MODULE/NULL */
765
766 /* scheduler initializing service */
767 int (*init_service)(struct ip_vs_service *svc);
768 /* scheduling service finish */
769 void (*done_service)(struct ip_vs_service *svc);
770 /* dest is linked */
771 int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
772 /* dest is unlinked */
773 int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
774 /* dest is updated */
775 int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
776
777 /* selecting a server from the given service */
778 struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc,
779 const struct sk_buff *skb,
780 struct ip_vs_iphdr *iph);
781};
782
783/* The persistence engine object */
784struct ip_vs_pe {
785 struct list_head n_list; /* d-linked list head */
786 char *name; /* scheduler name */
787 atomic_t refcnt; /* reference counter */
788 struct module *module; /* THIS_MODULE/NULL */
789
790 /* get the connection template, if any */
791 int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb);
792 bool (*ct_match)(const struct ip_vs_conn_param *p,
793 struct ip_vs_conn *ct);
794 u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval,
795 bool inverse);
796 int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf);
797 /* create connections for real-server outgoing packets */
798 struct ip_vs_conn* (*conn_out)(struct ip_vs_service *svc,
799 struct ip_vs_dest *dest,
800 struct sk_buff *skb,
801 const struct ip_vs_iphdr *iph,
802 __be16 dport, __be16 cport);
803};
804
805/* The application module object (a.k.a. app incarnation) */
806struct ip_vs_app {
807 struct list_head a_list; /* member in app list */
808 int type; /* IP_VS_APP_TYPE_xxx */
809 char *name; /* application module name */
810 __u16 protocol;
811 struct module *module; /* THIS_MODULE/NULL */
812 struct list_head incs_list; /* list of incarnations */
813
814 /* members for application incarnations */
815 struct list_head p_list; /* member in proto app list */
816 struct ip_vs_app *app; /* its real application */
817 __be16 port; /* port number in net order */
818 atomic_t usecnt; /* usage counter */
819 struct rcu_head rcu_head;
820
821 /* output hook: Process packet in inout direction, diff set for TCP.
822 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
823 * 2=Mangled but checksum was not updated
824 */
825 int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *,
826 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh);
827
828 /* input hook: Process packet in outin direction, diff set for TCP.
829 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
830 * 2=Mangled but checksum was not updated
831 */
832 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *,
833 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh);
834
835 /* ip_vs_app initializer */
836 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *);
837
838 /* ip_vs_app finish */
839 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *);
840
841
842 /* not used now */
843 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *,
844 struct ip_vs_protocol *);
845
846 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *);
847
848 int * timeout_table;
849 int * timeouts;
850 int timeouts_size;
851
852 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app,
853 int *verdict, struct ip_vs_conn **cpp);
854
855 struct ip_vs_conn *
856 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app,
857 const struct iphdr *iph, int inverse);
858
859 struct ip_vs_conn *
860 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app,
861 const struct iphdr *iph, int inverse);
862
863 int (*state_transition)(struct ip_vs_conn *cp, int direction,
864 const struct sk_buff *skb,
865 struct ip_vs_app *app);
866
867 void (*timeout_change)(struct ip_vs_app *app, int flags);
868};
869
870struct ipvs_master_sync_state {
871 struct list_head sync_queue;
872 struct ip_vs_sync_buff *sync_buff;
873 unsigned long sync_queue_len;
874 unsigned int sync_queue_delay;
875 struct delayed_work master_wakeup_work;
876 struct netns_ipvs *ipvs;
877};
878
879struct ip_vs_sync_thread_data;
880
881/* How much time to keep dests in trash */
882#define IP_VS_DEST_TRASH_PERIOD (120 * HZ)
883
884struct ipvs_sync_daemon_cfg {
885 union nf_inet_addr mcast_group;
886 int syncid;
887 u16 sync_maxlen;
888 u16 mcast_port;
889 u8 mcast_af;
890 u8 mcast_ttl;
891 /* multicast interface name */
892 char mcast_ifn[IP_VS_IFNAME_MAXLEN];
893};
894
895/* IPVS in network namespace */
896struct netns_ipvs {
897 int gen; /* Generation */
898 int enable; /* enable like nf_hooks do */
899 /* Hash table: for real service lookups */
900 #define IP_VS_RTAB_BITS 4
901 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
902 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
903
904 struct hlist_head rs_table[IP_VS_RTAB_SIZE];
905 /* ip_vs_app */
906 struct list_head app_list;
907 /* ip_vs_proto */
908 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */
909 struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE];
910 /* ip_vs_proto_tcp */
911#ifdef CONFIG_IP_VS_PROTO_TCP
912 #define TCP_APP_TAB_BITS 4
913 #define TCP_APP_TAB_SIZE (1 << TCP_APP_TAB_BITS)
914 #define TCP_APP_TAB_MASK (TCP_APP_TAB_SIZE - 1)
915 struct list_head tcp_apps[TCP_APP_TAB_SIZE];
916#endif
917 /* ip_vs_proto_udp */
918#ifdef CONFIG_IP_VS_PROTO_UDP
919 #define UDP_APP_TAB_BITS 4
920 #define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS)
921 #define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1)
922 struct list_head udp_apps[UDP_APP_TAB_SIZE];
923#endif
924 /* ip_vs_proto_sctp */
925#ifdef CONFIG_IP_VS_PROTO_SCTP
926 #define SCTP_APP_TAB_BITS 4
927 #define SCTP_APP_TAB_SIZE (1 << SCTP_APP_TAB_BITS)
928 #define SCTP_APP_TAB_MASK (SCTP_APP_TAB_SIZE - 1)
929 /* Hash table for SCTP application incarnations */
930 struct list_head sctp_apps[SCTP_APP_TAB_SIZE];
931#endif
932 /* ip_vs_conn */
933 atomic_t conn_count; /* connection counter */
934
935 /* ip_vs_ctl */
936 struct ip_vs_stats_rcu *tot_stats; /* Statistics & est. */
937
938 int num_services; /* no of virtual services */
939 int num_services6; /* IPv6 virtual services */
940
941 /* Trash for destinations */
942 struct list_head dest_trash;
943 spinlock_t dest_trash_lock;
944 struct timer_list dest_trash_timer; /* expiration timer */
945 /* Service counters */
946 atomic_t ftpsvc_counter;
947 atomic_t nullsvc_counter;
948 atomic_t conn_out_counter;
949
950#ifdef CONFIG_SYSCTL
951 /* delayed work for expiring no dest connections */
952 struct delayed_work expire_nodest_conn_work;
953 /* 1/rate drop and drop-entry variables */
954 struct delayed_work defense_work; /* Work handler */
955 int drop_rate;
956 int drop_counter;
957 int old_secure_tcp;
958 atomic_t dropentry;
959 /* locks in ctl.c */
960 spinlock_t dropentry_lock; /* drop entry handling */
961 spinlock_t droppacket_lock; /* drop packet handling */
962 spinlock_t securetcp_lock; /* state and timeout tables */
963
964 /* sys-ctl struct */
965 struct ctl_table_header *sysctl_hdr;
966 struct ctl_table *sysctl_tbl;
967#endif
968
969 /* sysctl variables */
970 int sysctl_amemthresh;
971 int sysctl_am_droprate;
972 int sysctl_drop_entry;
973 int sysctl_drop_packet;
974 int sysctl_secure_tcp;
975#ifdef CONFIG_IP_VS_NFCT
976 int sysctl_conntrack;
977#endif
978 int sysctl_snat_reroute;
979 int sysctl_sync_ver;
980 int sysctl_sync_ports;
981 int sysctl_sync_persist_mode;
982 unsigned long sysctl_sync_qlen_max;
983 int sysctl_sync_sock_size;
984 int sysctl_cache_bypass;
985 int sysctl_expire_nodest_conn;
986 int sysctl_sloppy_tcp;
987 int sysctl_sloppy_sctp;
988 int sysctl_expire_quiescent_template;
989 int sysctl_sync_threshold[2];
990 unsigned int sysctl_sync_refresh_period;
991 int sysctl_sync_retries;
992 int sysctl_nat_icmp_send;
993 int sysctl_pmtu_disc;
994 int sysctl_backup_only;
995 int sysctl_conn_reuse_mode;
996 int sysctl_schedule_icmp;
997 int sysctl_ignore_tunneled;
998 int sysctl_run_estimation;
999#ifdef CONFIG_SYSCTL
1000 cpumask_var_t sysctl_est_cpulist; /* kthread cpumask */
1001 int est_cpulist_valid; /* cpulist set */
1002 int sysctl_est_nice; /* kthread nice */
1003 int est_stopped; /* stop tasks */
1004#endif
1005
1006 /* ip_vs_lblc */
1007 int sysctl_lblc_expiration;
1008 struct ctl_table_header *lblc_ctl_header;
1009 struct ctl_table *lblc_ctl_table;
1010 /* ip_vs_lblcr */
1011 int sysctl_lblcr_expiration;
1012 struct ctl_table_header *lblcr_ctl_header;
1013 struct ctl_table *lblcr_ctl_table;
1014 /* ip_vs_est */
1015 struct delayed_work est_reload_work;/* Reload kthread tasks */
1016 struct mutex est_mutex; /* protect kthread tasks */
1017 struct hlist_head est_temp_list; /* Ests during calc phase */
1018 struct ip_vs_est_kt_data **est_kt_arr; /* Array of kthread data ptrs */
1019 unsigned long est_max_threads;/* Hard limit of kthreads */
1020 int est_calc_phase; /* Calculation phase */
1021 int est_chain_max; /* Calculated chain_max */
1022 int est_kt_count; /* Allocated ptrs */
1023 int est_add_ktid; /* ktid where to add ests */
1024 atomic_t est_genid; /* kthreads reload genid */
1025 atomic_t est_genid_done; /* applied genid */
1026 /* ip_vs_sync */
1027 spinlock_t sync_lock;
1028 struct ipvs_master_sync_state *ms;
1029 spinlock_t sync_buff_lock;
1030 struct ip_vs_sync_thread_data *master_tinfo;
1031 struct ip_vs_sync_thread_data *backup_tinfo;
1032 int threads_mask;
1033 volatile int sync_state;
1034 struct mutex sync_mutex;
1035 struct ipvs_sync_daemon_cfg mcfg; /* Master Configuration */
1036 struct ipvs_sync_daemon_cfg bcfg; /* Backup Configuration */
1037 /* net name space ptr */
1038 struct net *net; /* Needed by timer routines */
1039 /* Number of heterogeneous destinations, needed because heterogeneous
1040 * are not supported when synchronization is enabled.
1041 */
1042 unsigned int mixed_address_family_dests;
1043 unsigned int hooks_afmask; /* &1=AF_INET, &2=AF_INET6 */
1044};
1045
1046#define DEFAULT_SYNC_THRESHOLD 3
1047#define DEFAULT_SYNC_PERIOD 50
1048#define DEFAULT_SYNC_VER 1
1049#define DEFAULT_SLOPPY_TCP 0
1050#define DEFAULT_SLOPPY_SCTP 0
1051#define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ)
1052#define DEFAULT_SYNC_RETRIES 0
1053#define IPVS_SYNC_WAKEUP_RATE 8
1054#define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4)
1055#define IPVS_SYNC_SEND_DELAY (HZ / 50)
1056#define IPVS_SYNC_CHECK_PERIOD HZ
1057#define IPVS_SYNC_FLUSH_TIME (HZ * 2)
1058#define IPVS_SYNC_PORTS_MAX (1 << 6)
1059
1060#ifdef CONFIG_SYSCTL
1061
1062static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1063{
1064 return ipvs->sysctl_sync_threshold[0];
1065}
1066
1067static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1068{
1069 return READ_ONCE(ipvs->sysctl_sync_threshold[1]);
1070}
1071
1072static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1073{
1074 return READ_ONCE(ipvs->sysctl_sync_refresh_period);
1075}
1076
1077static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1078{
1079 return ipvs->sysctl_sync_retries;
1080}
1081
1082static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1083{
1084 return ipvs->sysctl_sync_ver;
1085}
1086
1087static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1088{
1089 return ipvs->sysctl_sloppy_tcp;
1090}
1091
1092static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1093{
1094 return ipvs->sysctl_sloppy_sctp;
1095}
1096
1097static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1098{
1099 return READ_ONCE(ipvs->sysctl_sync_ports);
1100}
1101
1102static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1103{
1104 return ipvs->sysctl_sync_persist_mode;
1105}
1106
1107static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1108{
1109 return ipvs->sysctl_sync_qlen_max;
1110}
1111
1112static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1113{
1114 return ipvs->sysctl_sync_sock_size;
1115}
1116
1117static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1118{
1119 return ipvs->sysctl_pmtu_disc;
1120}
1121
1122static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1123{
1124 return ipvs->sync_state & IP_VS_STATE_BACKUP &&
1125 ipvs->sysctl_backup_only;
1126}
1127
1128static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1129{
1130 return ipvs->sysctl_conn_reuse_mode;
1131}
1132
1133static inline int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs)
1134{
1135 return ipvs->sysctl_expire_nodest_conn;
1136}
1137
1138static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1139{
1140 return ipvs->sysctl_schedule_icmp;
1141}
1142
1143static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1144{
1145 return ipvs->sysctl_ignore_tunneled;
1146}
1147
1148static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1149{
1150 return ipvs->sysctl_cache_bypass;
1151}
1152
1153static inline int sysctl_run_estimation(struct netns_ipvs *ipvs)
1154{
1155 return ipvs->sysctl_run_estimation;
1156}
1157
1158static inline const struct cpumask *sysctl_est_cpulist(struct netns_ipvs *ipvs)
1159{
1160 if (ipvs->est_cpulist_valid)
1161 return ipvs->sysctl_est_cpulist;
1162 else
1163 return housekeeping_cpumask(HK_TYPE_KTHREAD);
1164}
1165
1166static inline int sysctl_est_nice(struct netns_ipvs *ipvs)
1167{
1168 return ipvs->sysctl_est_nice;
1169}
1170
1171#else
1172
1173static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1174{
1175 return DEFAULT_SYNC_THRESHOLD;
1176}
1177
1178static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1179{
1180 return DEFAULT_SYNC_PERIOD;
1181}
1182
1183static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1184{
1185 return DEFAULT_SYNC_REFRESH_PERIOD;
1186}
1187
1188static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1189{
1190 return DEFAULT_SYNC_RETRIES & 3;
1191}
1192
1193static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1194{
1195 return DEFAULT_SYNC_VER;
1196}
1197
1198static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1199{
1200 return DEFAULT_SLOPPY_TCP;
1201}
1202
1203static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1204{
1205 return DEFAULT_SLOPPY_SCTP;
1206}
1207
1208static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1209{
1210 return 1;
1211}
1212
1213static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1214{
1215 return 0;
1216}
1217
1218static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1219{
1220 return IPVS_SYNC_QLEN_MAX;
1221}
1222
1223static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1224{
1225 return 0;
1226}
1227
1228static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1229{
1230 return 1;
1231}
1232
1233static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1234{
1235 return 0;
1236}
1237
1238static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1239{
1240 return 1;
1241}
1242
1243static inline int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs)
1244{
1245 return 0;
1246}
1247
1248static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1249{
1250 return 0;
1251}
1252
1253static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1254{
1255 return 0;
1256}
1257
1258static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1259{
1260 return 0;
1261}
1262
1263static inline int sysctl_run_estimation(struct netns_ipvs *ipvs)
1264{
1265 return 1;
1266}
1267
1268static inline const struct cpumask *sysctl_est_cpulist(struct netns_ipvs *ipvs)
1269{
1270 return housekeeping_cpumask(HK_TYPE_KTHREAD);
1271}
1272
1273static inline int sysctl_est_nice(struct netns_ipvs *ipvs)
1274{
1275 return IPVS_EST_NICE;
1276}
1277
1278#endif
1279
1280/* IPVS core functions
1281 * (from ip_vs_core.c)
1282 */
1283const char *ip_vs_proto_name(unsigned int proto);
1284void ip_vs_init_hash_table(struct list_head *table, int rows);
1285struct ip_vs_conn *ip_vs_new_conn_out(struct ip_vs_service *svc,
1286 struct ip_vs_dest *dest,
1287 struct sk_buff *skb,
1288 const struct ip_vs_iphdr *iph,
1289 __be16 dport,
1290 __be16 cport);
1291#define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t)))
1292
1293#define IP_VS_APP_TYPE_FTP 1
1294
1295/* ip_vs_conn handling functions
1296 * (from ip_vs_conn.c)
1297 */
1298enum {
1299 IP_VS_DIR_INPUT = 0,
1300 IP_VS_DIR_OUTPUT,
1301 IP_VS_DIR_INPUT_ONLY,
1302 IP_VS_DIR_LAST,
1303};
1304
1305static inline void ip_vs_conn_fill_param(struct netns_ipvs *ipvs, int af, int protocol,
1306 const union nf_inet_addr *caddr,
1307 __be16 cport,
1308 const union nf_inet_addr *vaddr,
1309 __be16 vport,
1310 struct ip_vs_conn_param *p)
1311{
1312 p->ipvs = ipvs;
1313 p->af = af;
1314 p->protocol = protocol;
1315 p->caddr = caddr;
1316 p->cport = cport;
1317 p->vaddr = vaddr;
1318 p->vport = vport;
1319 p->pe = NULL;
1320 p->pe_data = NULL;
1321}
1322
1323struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p);
1324struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p);
1325
1326struct ip_vs_conn * ip_vs_conn_in_get_proto(struct netns_ipvs *ipvs, int af,
1327 const struct sk_buff *skb,
1328 const struct ip_vs_iphdr *iph);
1329
1330struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p);
1331
1332struct ip_vs_conn * ip_vs_conn_out_get_proto(struct netns_ipvs *ipvs, int af,
1333 const struct sk_buff *skb,
1334 const struct ip_vs_iphdr *iph);
1335
1336/* Get reference to gain full access to conn.
1337 * By default, RCU read-side critical sections have access only to
1338 * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference.
1339 */
1340static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp)
1341{
1342 return refcount_inc_not_zero(&cp->refcnt);
1343}
1344
1345/* put back the conn without restarting its timer */
1346static inline void __ip_vs_conn_put(struct ip_vs_conn *cp)
1347{
1348 smp_mb__before_atomic();
1349 refcount_dec(&cp->refcnt);
1350}
1351void ip_vs_conn_put(struct ip_vs_conn *cp);
1352void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport);
1353
1354struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, int dest_af,
1355 const union nf_inet_addr *daddr,
1356 __be16 dport, unsigned int flags,
1357 struct ip_vs_dest *dest, __u32 fwmark);
1358void ip_vs_conn_expire_now(struct ip_vs_conn *cp);
1359
1360const char *ip_vs_state_name(const struct ip_vs_conn *cp);
1361
1362void ip_vs_tcp_conn_listen(struct ip_vs_conn *cp);
1363int ip_vs_check_template(struct ip_vs_conn *ct, struct ip_vs_dest *cdest);
1364void ip_vs_random_dropentry(struct netns_ipvs *ipvs);
1365int ip_vs_conn_init(void);
1366void ip_vs_conn_cleanup(void);
1367
1368static inline void ip_vs_control_del(struct ip_vs_conn *cp)
1369{
1370 struct ip_vs_conn *ctl_cp = cp->control;
1371 if (!ctl_cp) {
1372 IP_VS_ERR_BUF("request control DEL for uncontrolled: "
1373 "%s:%d to %s:%d\n",
1374 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1375 ntohs(cp->cport),
1376 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1377 ntohs(cp->vport));
1378
1379 return;
1380 }
1381
1382 IP_VS_DBG_BUF(7, "DELeting control for: "
1383 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1384 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1385 ntohs(cp->cport),
1386 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1387 ntohs(ctl_cp->cport));
1388
1389 cp->control = NULL;
1390 if (atomic_read(&ctl_cp->n_control) == 0) {
1391 IP_VS_ERR_BUF("BUG control DEL with n=0 : "
1392 "%s:%d to %s:%d\n",
1393 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1394 ntohs(cp->cport),
1395 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1396 ntohs(cp->vport));
1397
1398 return;
1399 }
1400 atomic_dec(&ctl_cp->n_control);
1401}
1402
1403static inline void
1404ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp)
1405{
1406 if (cp->control) {
1407 IP_VS_ERR_BUF("request control ADD for already controlled: "
1408 "%s:%d to %s:%d\n",
1409 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1410 ntohs(cp->cport),
1411 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1412 ntohs(cp->vport));
1413
1414 ip_vs_control_del(cp);
1415 }
1416
1417 IP_VS_DBG_BUF(7, "ADDing control for: "
1418 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1419 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1420 ntohs(cp->cport),
1421 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1422 ntohs(ctl_cp->cport));
1423
1424 cp->control = ctl_cp;
1425 atomic_inc(&ctl_cp->n_control);
1426}
1427
1428/* Mark our template as assured */
1429static inline void
1430ip_vs_control_assure_ct(struct ip_vs_conn *cp)
1431{
1432 struct ip_vs_conn *ct = cp->control;
1433
1434 if (ct && !(ct->state & IP_VS_CTPL_S_ASSURED) &&
1435 (ct->flags & IP_VS_CONN_F_TEMPLATE))
1436 ct->state |= IP_VS_CTPL_S_ASSURED;
1437}
1438
1439/* IPVS netns init & cleanup functions */
1440int ip_vs_estimator_net_init(struct netns_ipvs *ipvs);
1441int ip_vs_control_net_init(struct netns_ipvs *ipvs);
1442int ip_vs_protocol_net_init(struct netns_ipvs *ipvs);
1443int ip_vs_app_net_init(struct netns_ipvs *ipvs);
1444int ip_vs_conn_net_init(struct netns_ipvs *ipvs);
1445int ip_vs_sync_net_init(struct netns_ipvs *ipvs);
1446void ip_vs_conn_net_cleanup(struct netns_ipvs *ipvs);
1447void ip_vs_app_net_cleanup(struct netns_ipvs *ipvs);
1448void ip_vs_protocol_net_cleanup(struct netns_ipvs *ipvs);
1449void ip_vs_control_net_cleanup(struct netns_ipvs *ipvs);
1450void ip_vs_estimator_net_cleanup(struct netns_ipvs *ipvs);
1451void ip_vs_sync_net_cleanup(struct netns_ipvs *ipvs);
1452void ip_vs_service_nets_cleanup(struct list_head *net_list);
1453
1454/* IPVS application functions
1455 * (from ip_vs_app.c)
1456 */
1457#define IP_VS_APP_MAX_PORTS 8
1458struct ip_vs_app *register_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1459void unregister_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1460int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1461void ip_vs_unbind_app(struct ip_vs_conn *cp);
1462int register_ip_vs_app_inc(struct netns_ipvs *ipvs, struct ip_vs_app *app, __u16 proto,
1463 __u16 port);
1464int ip_vs_app_inc_get(struct ip_vs_app *inc);
1465void ip_vs_app_inc_put(struct ip_vs_app *inc);
1466
1467int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb,
1468 struct ip_vs_iphdr *ipvsh);
1469int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb,
1470 struct ip_vs_iphdr *ipvsh);
1471
1472int register_ip_vs_pe(struct ip_vs_pe *pe);
1473int unregister_ip_vs_pe(struct ip_vs_pe *pe);
1474struct ip_vs_pe *ip_vs_pe_getbyname(const char *name);
1475struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name);
1476
1477/* Use a #define to avoid all of module.h just for these trivial ops */
1478#define ip_vs_pe_get(pe) \
1479 if (pe && pe->module) \
1480 __module_get(pe->module);
1481
1482#define ip_vs_pe_put(pe) \
1483 if (pe && pe->module) \
1484 module_put(pe->module);
1485
1486/* IPVS protocol functions (from ip_vs_proto.c) */
1487int ip_vs_protocol_init(void);
1488void ip_vs_protocol_cleanup(void);
1489void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags);
1490int *ip_vs_create_timeout_table(int *table, int size);
1491void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
1492 const struct sk_buff *skb, int offset,
1493 const char *msg);
1494
1495extern struct ip_vs_protocol ip_vs_protocol_tcp;
1496extern struct ip_vs_protocol ip_vs_protocol_udp;
1497extern struct ip_vs_protocol ip_vs_protocol_icmp;
1498extern struct ip_vs_protocol ip_vs_protocol_esp;
1499extern struct ip_vs_protocol ip_vs_protocol_ah;
1500extern struct ip_vs_protocol ip_vs_protocol_sctp;
1501
1502/* Registering/unregistering scheduler functions
1503 * (from ip_vs_sched.c)
1504 */
1505int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1506int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1507int ip_vs_bind_scheduler(struct ip_vs_service *svc,
1508 struct ip_vs_scheduler *scheduler);
1509void ip_vs_unbind_scheduler(struct ip_vs_service *svc,
1510 struct ip_vs_scheduler *sched);
1511struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name);
1512void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler);
1513struct ip_vs_conn *
1514ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb,
1515 struct ip_vs_proto_data *pd, int *ignored,
1516 struct ip_vs_iphdr *iph);
1517int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb,
1518 struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph);
1519
1520void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg);
1521
1522/* IPVS control data and functions (from ip_vs_ctl.c) */
1523extern struct ip_vs_stats ip_vs_stats;
1524extern int sysctl_ip_vs_sync_ver;
1525
1526struct ip_vs_service *
1527ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
1528 const union nf_inet_addr *vaddr, __be16 vport);
1529
1530bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1531 const union nf_inet_addr *daddr, __be16 dport);
1532
1533struct ip_vs_dest *
1534ip_vs_find_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1535 const union nf_inet_addr *daddr, __be16 dport);
1536struct ip_vs_dest *ip_vs_find_tunnel(struct netns_ipvs *ipvs, int af,
1537 const union nf_inet_addr *daddr,
1538 __be16 tun_port);
1539
1540int ip_vs_use_count_inc(void);
1541void ip_vs_use_count_dec(void);
1542int ip_vs_register_nl_ioctl(void);
1543void ip_vs_unregister_nl_ioctl(void);
1544int ip_vs_control_init(void);
1545void ip_vs_control_cleanup(void);
1546struct ip_vs_dest *
1547ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
1548 const union nf_inet_addr *daddr, __be16 dport,
1549 const union nf_inet_addr *vaddr, __be16 vport,
1550 __u16 protocol, __u32 fwmark, __u32 flags);
1551void ip_vs_try_bind_dest(struct ip_vs_conn *cp);
1552
1553static inline void ip_vs_dest_hold(struct ip_vs_dest *dest)
1554{
1555 refcount_inc(&dest->refcnt);
1556}
1557
1558static inline void ip_vs_dest_put(struct ip_vs_dest *dest)
1559{
1560 smp_mb__before_atomic();
1561 refcount_dec(&dest->refcnt);
1562}
1563
1564static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest)
1565{
1566 if (refcount_dec_and_test(&dest->refcnt))
1567 kfree(dest);
1568}
1569
1570/* IPVS sync daemon data and function prototypes
1571 * (from ip_vs_sync.c)
1572 */
1573int start_sync_thread(struct netns_ipvs *ipvs, struct ipvs_sync_daemon_cfg *cfg,
1574 int state);
1575int stop_sync_thread(struct netns_ipvs *ipvs, int state);
1576void ip_vs_sync_conn(struct netns_ipvs *ipvs, struct ip_vs_conn *cp, int pkts);
1577
1578/* IPVS rate estimator prototypes (from ip_vs_est.c) */
1579int ip_vs_start_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1580void ip_vs_stop_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1581void ip_vs_zero_estimator(struct ip_vs_stats *stats);
1582void ip_vs_read_estimator(struct ip_vs_kstats *dst, struct ip_vs_stats *stats);
1583void ip_vs_est_reload_start(struct netns_ipvs *ipvs);
1584int ip_vs_est_kthread_start(struct netns_ipvs *ipvs,
1585 struct ip_vs_est_kt_data *kd);
1586void ip_vs_est_kthread_stop(struct ip_vs_est_kt_data *kd);
1587
1588static inline void ip_vs_est_stopped_recalc(struct netns_ipvs *ipvs)
1589{
1590#ifdef CONFIG_SYSCTL
1591 /* Stop tasks while cpulist is empty or if disabled with flag */
1592 ipvs->est_stopped = !sysctl_run_estimation(ipvs) ||
1593 (ipvs->est_cpulist_valid &&
1594 cpumask_empty(sysctl_est_cpulist(ipvs)));
1595#endif
1596}
1597
1598static inline bool ip_vs_est_stopped(struct netns_ipvs *ipvs)
1599{
1600#ifdef CONFIG_SYSCTL
1601 return ipvs->est_stopped;
1602#else
1603 return false;
1604#endif
1605}
1606
1607static inline int ip_vs_est_max_threads(struct netns_ipvs *ipvs)
1608{
1609 unsigned int limit = IPVS_EST_CPU_KTHREADS *
1610 cpumask_weight(sysctl_est_cpulist(ipvs));
1611
1612 return max(1U, limit);
1613}
1614
1615/* Various IPVS packet transmitters (from ip_vs_xmit.c) */
1616int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1617 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1618int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1619 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1620int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1621 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1622int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1623 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1624int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1625 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1626int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1627 struct ip_vs_protocol *pp, int offset,
1628 unsigned int hooknum, struct ip_vs_iphdr *iph);
1629void ip_vs_dest_dst_rcu_free(struct rcu_head *head);
1630
1631#ifdef CONFIG_IP_VS_IPV6
1632int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1633 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1634int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1635 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1636int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1637 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1638int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1639 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1640int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1641 struct ip_vs_protocol *pp, int offset,
1642 unsigned int hooknum, struct ip_vs_iphdr *iph);
1643#endif
1644
1645#ifdef CONFIG_SYSCTL
1646/* This is a simple mechanism to ignore packets when
1647 * we are loaded. Just set ip_vs_drop_rate to 'n' and
1648 * we start to drop 1/rate of the packets
1649 */
1650static inline int ip_vs_todrop(struct netns_ipvs *ipvs)
1651{
1652 if (!ipvs->drop_rate)
1653 return 0;
1654 if (--ipvs->drop_counter > 0)
1655 return 0;
1656 ipvs->drop_counter = ipvs->drop_rate;
1657 return 1;
1658}
1659#else
1660static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; }
1661#endif
1662
1663#ifdef CONFIG_SYSCTL
1664/* Enqueue delayed work for expiring no dest connections
1665 * Only run when sysctl_expire_nodest=1
1666 */
1667static inline void ip_vs_enqueue_expire_nodest_conns(struct netns_ipvs *ipvs)
1668{
1669 if (sysctl_expire_nodest_conn(ipvs))
1670 queue_delayed_work(system_long_wq,
1671 &ipvs->expire_nodest_conn_work, 1);
1672}
1673
1674void ip_vs_expire_nodest_conn_flush(struct netns_ipvs *ipvs);
1675#else
1676static inline void ip_vs_enqueue_expire_nodest_conns(struct netns_ipvs *ipvs) {}
1677#endif
1678
1679#define IP_VS_DFWD_METHOD(dest) (atomic_read(&(dest)->conn_flags) & \
1680 IP_VS_CONN_F_FWD_MASK)
1681
1682/* ip_vs_fwd_tag returns the forwarding tag of the connection */
1683#define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK)
1684
1685static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp)
1686{
1687 char fwd;
1688
1689 switch (IP_VS_FWD_METHOD(cp)) {
1690 case IP_VS_CONN_F_MASQ:
1691 fwd = 'M'; break;
1692 case IP_VS_CONN_F_LOCALNODE:
1693 fwd = 'L'; break;
1694 case IP_VS_CONN_F_TUNNEL:
1695 fwd = 'T'; break;
1696 case IP_VS_CONN_F_DROUTE:
1697 fwd = 'R'; break;
1698 case IP_VS_CONN_F_BYPASS:
1699 fwd = 'B'; break;
1700 default:
1701 fwd = '?'; break;
1702 }
1703 return fwd;
1704}
1705
1706void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp,
1707 struct ip_vs_conn *cp, int dir);
1708
1709#ifdef CONFIG_IP_VS_IPV6
1710void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp,
1711 struct ip_vs_conn *cp, int dir);
1712#endif
1713
1714__sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset);
1715
1716static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum)
1717{
1718 __be32 diff[2] = { ~old, new };
1719
1720 return csum_partial(diff, sizeof(diff), oldsum);
1721}
1722
1723#ifdef CONFIG_IP_VS_IPV6
1724static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new,
1725 __wsum oldsum)
1726{
1727 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0],
1728 new[3], new[2], new[1], new[0] };
1729
1730 return csum_partial(diff, sizeof(diff), oldsum);
1731}
1732#endif
1733
1734static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum)
1735{
1736 __be16 diff[2] = { ~old, new };
1737
1738 return csum_partial(diff, sizeof(diff), oldsum);
1739}
1740
1741/* Forget current conntrack (unconfirmed) and attach notrack entry */
1742static inline void ip_vs_notrack(struct sk_buff *skb)
1743{
1744#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1745 enum ip_conntrack_info ctinfo;
1746 struct nf_conn *ct = nf_ct_get(skb, &ctinfo);
1747
1748 if (ct) {
1749 nf_conntrack_put(&ct->ct_general);
1750 nf_ct_set(skb, NULL, IP_CT_UNTRACKED);
1751 }
1752#endif
1753}
1754
1755#ifdef CONFIG_IP_VS_NFCT
1756/* Netfilter connection tracking
1757 * (from ip_vs_nfct.c)
1758 */
1759static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1760{
1761#ifdef CONFIG_SYSCTL
1762 return ipvs->sysctl_conntrack;
1763#else
1764 return 0;
1765#endif
1766}
1767
1768void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp,
1769 int outin);
1770int ip_vs_confirm_conntrack(struct sk_buff *skb);
1771void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct,
1772 struct ip_vs_conn *cp, u_int8_t proto,
1773 const __be16 port, int from_rs);
1774void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp);
1775
1776#else
1777
1778static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1779{
1780 return 0;
1781}
1782
1783static inline void ip_vs_update_conntrack(struct sk_buff *skb,
1784 struct ip_vs_conn *cp, int outin)
1785{
1786}
1787
1788static inline int ip_vs_confirm_conntrack(struct sk_buff *skb)
1789{
1790 return NF_ACCEPT;
1791}
1792
1793static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp)
1794{
1795}
1796#endif /* CONFIG_IP_VS_NFCT */
1797
1798/* Using old conntrack that can not be redirected to another real server? */
1799static inline bool ip_vs_conn_uses_old_conntrack(struct ip_vs_conn *cp,
1800 struct sk_buff *skb)
1801{
1802#ifdef CONFIG_IP_VS_NFCT
1803 enum ip_conntrack_info ctinfo;
1804 struct nf_conn *ct;
1805
1806 ct = nf_ct_get(skb, &ctinfo);
1807 if (ct && nf_ct_is_confirmed(ct))
1808 return true;
1809#endif
1810 return false;
1811}
1812
1813static inline int ip_vs_register_conntrack(struct ip_vs_service *svc)
1814{
1815#if IS_ENABLED(CONFIG_NF_CONNTRACK)
1816 int afmask = (svc->af == AF_INET6) ? 2 : 1;
1817 int ret = 0;
1818
1819 if (!(svc->conntrack_afmask & afmask)) {
1820 ret = nf_ct_netns_get(svc->ipvs->net, svc->af);
1821 if (ret >= 0)
1822 svc->conntrack_afmask |= afmask;
1823 }
1824 return ret;
1825#else
1826 return 0;
1827#endif
1828}
1829
1830static inline void ip_vs_unregister_conntrack(struct ip_vs_service *svc)
1831{
1832#if IS_ENABLED(CONFIG_NF_CONNTRACK)
1833 int afmask = (svc->af == AF_INET6) ? 2 : 1;
1834
1835 if (svc->conntrack_afmask & afmask) {
1836 nf_ct_netns_put(svc->ipvs->net, svc->af);
1837 svc->conntrack_afmask &= ~afmask;
1838 }
1839#endif
1840}
1841
1842int ip_vs_register_hooks(struct netns_ipvs *ipvs, unsigned int af);
1843void ip_vs_unregister_hooks(struct netns_ipvs *ipvs, unsigned int af);
1844
1845static inline int
1846ip_vs_dest_conn_overhead(struct ip_vs_dest *dest)
1847{
1848 /* We think the overhead of processing active connections is 256
1849 * times higher than that of inactive connections in average. (This
1850 * 256 times might not be accurate, we will change it later) We
1851 * use the following formula to estimate the overhead now:
1852 * dest->activeconns*256 + dest->inactconns
1853 */
1854 return (atomic_read(&dest->activeconns) << 8) +
1855 atomic_read(&dest->inactconns);
1856}
1857
1858#ifdef CONFIG_IP_VS_PROTO_TCP
1859INDIRECT_CALLABLE_DECLARE(int
1860 tcp_snat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
1861 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph));
1862#endif
1863
1864#ifdef CONFIG_IP_VS_PROTO_UDP
1865INDIRECT_CALLABLE_DECLARE(int
1866 udp_snat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
1867 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph));
1868#endif
1869#endif /* _NET_IP_VS_H */