?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
vhost_net: disable zerocopy by default
[linux/fpc-iii.git] / include / net / ip_vs.h
blobcd6018a9ee2467cec1fbe4d621fc294bc843a05e
1 /* IP Virtual Server
2 * data structure and functionality definitions
3 */
4
5 #ifndef _NET_IP_VS_H
6 #define _NET_IP_VS_H
7
8 #include <linux/ip_vs.h> /* definitions shared with userland */
9
10 #include <asm/types.h> /* for __uXX types */
11
12 #include <linux/list.h> /* for struct list_head */
13 #include <linux/spinlock.h> /* for struct rwlock_t */
14 #include <linux/atomic.h> /* for struct atomic_t */
15 #include <linux/compiler.h>
16 #include <linux/timer.h>
17 #include <linux/bug.h>
18
19 #include <net/checksum.h>
20 #include <linux/netfilter.h> /* for union nf_inet_addr */
21 #include <linux/ip.h>
22 #include <linux/ipv6.h> /* for struct ipv6hdr */
23 #include <net/ipv6.h>
24 #if IS_ENABLED(CONFIG_IP_VS_IPV6)
25 #include <linux/netfilter_ipv6/ip6_tables.h>
26 #endif
27 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
28 #include <net/netfilter/nf_conntrack.h>
29 #endif
30 #include <net/net_namespace.h> /* Netw namespace */
31
32 #define IP_VS_HDR_INVERSE 1
33 #define IP_VS_HDR_ICMP 2
34
35 /* Generic access of ipvs struct */
36 static inline struct netns_ipvs *net_ipvs(struct net* net)
37 {
38 return net->ipvs;
39 }
40
41 /* This one needed for single_open_net since net is stored directly in
42 * private not as a struct i.e. seq_file_net can't be used.
43 */
44 static inline struct net *seq_file_single_net(struct seq_file *seq)
45 {
46 #ifdef CONFIG_NET_NS
47 return (struct net *)seq->private;
48 #else
49 return &init_net;
50 #endif
51 }
52
53 /* Connections' size value needed by ip_vs_ctl.c */
54 extern int ip_vs_conn_tab_size;
55
56 struct ip_vs_iphdr {
57 int hdr_flags; /* ipvs flags */
58 __u32 off; /* Where IP or IPv4 header starts */
59 __u32 len; /* IPv4 simply where L4 starts
60 * IPv6 where L4 Transport Header starts */
61 __u16 fragoffs; /* IPv6 fragment offset, 0 if first frag (or not frag)*/
62 __s16 protocol;
63 __s32 flags;
64 union nf_inet_addr saddr;
65 union nf_inet_addr daddr;
66 };
67
68 static inline void *frag_safe_skb_hp(const struct sk_buff *skb, int offset,
69 int len, void *buffer,
70 const struct ip_vs_iphdr *ipvsh)
71 {
72 return skb_header_pointer(skb, offset, len, buffer);
73 }
74
75 /* This function handles filling *ip_vs_iphdr, both for IPv4 and IPv6.
76 * IPv6 requires some extra work, as finding proper header position,
77 * depend on the IPv6 extension headers.
78 */
79 static inline int
80 ip_vs_fill_iph_skb_off(int af, const struct sk_buff *skb, int offset,
81 int hdr_flags, struct ip_vs_iphdr *iphdr)
82 {
83 iphdr->hdr_flags = hdr_flags;
84 iphdr->off = offset;
85
86 #ifdef CONFIG_IP_VS_IPV6
87 if (af == AF_INET6) {
88 struct ipv6hdr _iph;
89 const struct ipv6hdr *iph = skb_header_pointer(
90 skb, offset, sizeof(_iph), &_iph);
91 if (!iph)
92 return 0;
93
94 iphdr->saddr.in6 = iph->saddr;
95 iphdr->daddr.in6 = iph->daddr;
96 /* ipv6_find_hdr() updates len, flags */
97 iphdr->len = offset;
98 iphdr->flags = 0;
99 iphdr->protocol = ipv6_find_hdr(skb, &iphdr->len, -1,
100 &iphdr->fragoffs,
101 &iphdr->flags);
102 if (iphdr->protocol < 0)
103 return 0;
104 } else
105 #endif
106 {
107 struct iphdr _iph;
108 const struct iphdr *iph = skb_header_pointer(
109 skb, offset, sizeof(_iph), &_iph);
110 if (!iph)
111 return 0;
112
113 iphdr->len = offset + iph->ihl * 4;
114 iphdr->fragoffs = 0;
115 iphdr->protocol = iph->protocol;
116 iphdr->saddr.ip = iph->saddr;
117 iphdr->daddr.ip = iph->daddr;
118 }
119
120 return 1;
121 }
122
123 static inline int
124 ip_vs_fill_iph_skb_icmp(int af, const struct sk_buff *skb, int offset,
125 bool inverse, struct ip_vs_iphdr *iphdr)
126 {
127 int hdr_flags = IP_VS_HDR_ICMP;
128
129 if (inverse)
130 hdr_flags |= IP_VS_HDR_INVERSE;
131
132 return ip_vs_fill_iph_skb_off(af, skb, offset, hdr_flags, iphdr);
133 }
134
135 static inline int
136 ip_vs_fill_iph_skb(int af, const struct sk_buff *skb, bool inverse,
137 struct ip_vs_iphdr *iphdr)
138 {
139 int hdr_flags = 0;
140
141 if (inverse)
142 hdr_flags |= IP_VS_HDR_INVERSE;
143
144 return ip_vs_fill_iph_skb_off(af, skb, skb_network_offset(skb),
145 hdr_flags, iphdr);
146 }
147
148 static inline bool
149 ip_vs_iph_inverse(const struct ip_vs_iphdr *iph)
150 {
151 return !!(iph->hdr_flags & IP_VS_HDR_INVERSE);
152 }
153
154 static inline bool
155 ip_vs_iph_icmp(const struct ip_vs_iphdr *iph)
156 {
157 return !!(iph->hdr_flags & IP_VS_HDR_ICMP);
158 }
159
160 static inline void ip_vs_addr_copy(int af, union nf_inet_addr *dst,
161 const union nf_inet_addr *src)
162 {
163 #ifdef CONFIG_IP_VS_IPV6
164 if (af == AF_INET6)
165 dst->in6 = src->in6;
166 else
167 #endif
168 dst->ip = src->ip;
169 }
170
171 static inline void ip_vs_addr_set(int af, union nf_inet_addr *dst,
172 const union nf_inet_addr *src)
173 {
174 #ifdef CONFIG_IP_VS_IPV6
175 if (af == AF_INET6) {
176 dst->in6 = src->in6;
177 return;
178 }
179 #endif
180 dst->ip = src->ip;
181 dst->all[1] = 0;
182 dst->all[2] = 0;
183 dst->all[3] = 0;
184 }
185
186 static inline int ip_vs_addr_equal(int af, const union nf_inet_addr *a,
187 const union nf_inet_addr *b)
188 {
189 #ifdef CONFIG_IP_VS_IPV6
190 if (af == AF_INET6)
191 return ipv6_addr_equal(&a->in6, &b->in6);
192 #endif
193 return a->ip == b->ip;
194 }
195
196 #ifdef CONFIG_IP_VS_DEBUG
197 #include <linux/net.h>
198
199 int ip_vs_get_debug_level(void);
200
201 static inline const char *ip_vs_dbg_addr(int af, char *buf, size_t buf_len,
202 const union nf_inet_addr *addr,
203 int *idx)
204 {
205 int len;
206 #ifdef CONFIG_IP_VS_IPV6
207 if (af == AF_INET6)
208 len = snprintf(&buf[*idx], buf_len - *idx, "[%pI6c]",
209 &addr->in6) + 1;
210 else
211 #endif
212 len = snprintf(&buf[*idx], buf_len - *idx, "%pI4",
213 &addr->ip) + 1;
214
215 *idx += len;
216 BUG_ON(*idx > buf_len + 1);
217 return &buf[*idx - len];
218 }
219
220 #define IP_VS_DBG_BUF(level, msg, ...) \
221 do { \
222 char ip_vs_dbg_buf[160]; \
223 int ip_vs_dbg_idx = 0; \
224 if (level <= ip_vs_get_debug_level()) \
225 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
226 } while (0)
227 #define IP_VS_ERR_BUF(msg...) \
228 do { \
229 char ip_vs_dbg_buf[160]; \
230 int ip_vs_dbg_idx = 0; \
231 pr_err(msg); \
232 } while (0)
233
234 /* Only use from within IP_VS_DBG_BUF() or IP_VS_ERR_BUF macros */
235 #define IP_VS_DBG_ADDR(af, addr) \
236 ip_vs_dbg_addr(af, ip_vs_dbg_buf, \
237 sizeof(ip_vs_dbg_buf), addr, \
238 &ip_vs_dbg_idx)
239
240 #define IP_VS_DBG(level, msg, ...) \
241 do { \
242 if (level <= ip_vs_get_debug_level()) \
243 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
244 } while (0)
245 #define IP_VS_DBG_RL(msg, ...) \
246 do { \
247 if (net_ratelimit()) \
248 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
249 } while (0)
250 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) \
251 do { \
252 if (level <= ip_vs_get_debug_level()) \
253 pp->debug_packet(af, pp, skb, ofs, msg); \
254 } while (0)
255 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) \
256 do { \
257 if (level <= ip_vs_get_debug_level() && \
258 net_ratelimit()) \
259 pp->debug_packet(af, pp, skb, ofs, msg); \
260 } while (0)
261 #else /* NO DEBUGGING at ALL */
262 #define IP_VS_DBG_BUF(level, msg...) do {} while (0)
263 #define IP_VS_ERR_BUF(msg...) do {} while (0)
264 #define IP_VS_DBG(level, msg...) do {} while (0)
265 #define IP_VS_DBG_RL(msg...) do {} while (0)
266 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
267 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
268 #endif
269
270 #define IP_VS_BUG() BUG()
271 #define IP_VS_ERR_RL(msg, ...) \
272 do { \
273 if (net_ratelimit()) \
274 pr_err(msg, ##__VA_ARGS__); \
275 } while (0)
276
277 #ifdef CONFIG_IP_VS_DEBUG
278 #define EnterFunction(level) \
279 do { \
280 if (level <= ip_vs_get_debug_level()) \
281 printk(KERN_DEBUG \
282 pr_fmt("Enter: %s, %s line %i\n"), \
283 __func__, __FILE__, __LINE__); \
284 } while (0)
285 #define LeaveFunction(level) \
286 do { \
287 if (level <= ip_vs_get_debug_level()) \
288 printk(KERN_DEBUG \
289 pr_fmt("Leave: %s, %s line %i\n"), \
290 __func__, __FILE__, __LINE__); \
291 } while (0)
292 #else
293 #define EnterFunction(level) do {} while (0)
294 #define LeaveFunction(level) do {} while (0)
295 #endif
296
297 /* The port number of FTP service (in network order). */
298 #define FTPPORT cpu_to_be16(21)
299 #define FTPDATA cpu_to_be16(20)
300
301 /* TCP State Values */
302 enum {
303 IP_VS_TCP_S_NONE = 0,
304 IP_VS_TCP_S_ESTABLISHED,
305 IP_VS_TCP_S_SYN_SENT,
306 IP_VS_TCP_S_SYN_RECV,
307 IP_VS_TCP_S_FIN_WAIT,
308 IP_VS_TCP_S_TIME_WAIT,
309 IP_VS_TCP_S_CLOSE,
310 IP_VS_TCP_S_CLOSE_WAIT,
311 IP_VS_TCP_S_LAST_ACK,
312 IP_VS_TCP_S_LISTEN,
313 IP_VS_TCP_S_SYNACK,
314 IP_VS_TCP_S_LAST
315 };
316
317 /* UDP State Values */
318 enum {
319 IP_VS_UDP_S_NORMAL,
320 IP_VS_UDP_S_LAST,
321 };
322
323 /* ICMP State Values */
324 enum {
325 IP_VS_ICMP_S_NORMAL,
326 IP_VS_ICMP_S_LAST,
327 };
328
329 /* SCTP State Values */
330 enum ip_vs_sctp_states {
331 IP_VS_SCTP_S_NONE,
332 IP_VS_SCTP_S_INIT1,
333 IP_VS_SCTP_S_INIT,
334 IP_VS_SCTP_S_COOKIE_SENT,
335 IP_VS_SCTP_S_COOKIE_REPLIED,
336 IP_VS_SCTP_S_COOKIE_WAIT,
337 IP_VS_SCTP_S_COOKIE,
338 IP_VS_SCTP_S_COOKIE_ECHOED,
339 IP_VS_SCTP_S_ESTABLISHED,
340 IP_VS_SCTP_S_SHUTDOWN_SENT,
341 IP_VS_SCTP_S_SHUTDOWN_RECEIVED,
342 IP_VS_SCTP_S_SHUTDOWN_ACK_SENT,
343 IP_VS_SCTP_S_REJECTED,
344 IP_VS_SCTP_S_CLOSED,
345 IP_VS_SCTP_S_LAST
346 };
347
348 /* Delta sequence info structure
349 * Each ip_vs_conn has 2 (output AND input seq. changes).
350 * Only used in the VS/NAT.
351 */
352 struct ip_vs_seq {
353 __u32 init_seq; /* Add delta from this seq */
354 __u32 delta; /* Delta in sequence numbers */
355 __u32 previous_delta; /* Delta in sequence numbers
356 * before last resized pkt */
357 };
358
359 /* counters per cpu */
360 struct ip_vs_counters {
361 __u64 conns; /* connections scheduled */
362 __u64 inpkts; /* incoming packets */
363 __u64 outpkts; /* outgoing packets */
364 __u64 inbytes; /* incoming bytes */
365 __u64 outbytes; /* outgoing bytes */
366 };
367 /* Stats per cpu */
368 struct ip_vs_cpu_stats {
369 struct ip_vs_counters cnt;
370 struct u64_stats_sync syncp;
371 };
372
373 /* IPVS statistics objects */
374 struct ip_vs_estimator {
375 struct list_head list;
376
377 u64 last_inbytes;
378 u64 last_outbytes;
379 u64 last_conns;
380 u64 last_inpkts;
381 u64 last_outpkts;
382
383 u64 cps;
384 u64 inpps;
385 u64 outpps;
386 u64 inbps;
387 u64 outbps;
388 };
389
390 /*
391 * IPVS statistics object, 64-bit kernel version of struct ip_vs_stats_user
392 */
393 struct ip_vs_kstats {
394 u64 conns; /* connections scheduled */
395 u64 inpkts; /* incoming packets */
396 u64 outpkts; /* outgoing packets */
397 u64 inbytes; /* incoming bytes */
398 u64 outbytes; /* outgoing bytes */
399
400 u64 cps; /* current connection rate */
401 u64 inpps; /* current in packet rate */
402 u64 outpps; /* current out packet rate */
403 u64 inbps; /* current in byte rate */
404 u64 outbps; /* current out byte rate */
405 };
406
407 struct ip_vs_stats {
408 struct ip_vs_kstats kstats; /* kernel statistics */
409 struct ip_vs_estimator est; /* estimator */
410 struct ip_vs_cpu_stats __percpu *cpustats; /* per cpu counters */
411 spinlock_t lock; /* spin lock */
412 struct ip_vs_kstats kstats0; /* reset values */
413 };
414
415 struct dst_entry;
416 struct iphdr;
417 struct ip_vs_conn;
418 struct ip_vs_app;
419 struct sk_buff;
420 struct ip_vs_proto_data;
421
422 struct ip_vs_protocol {
423 struct ip_vs_protocol *next;
424 char *name;
425 u16 protocol;
426 u16 num_states;
427 int dont_defrag;
428
429 void (*init)(struct ip_vs_protocol *pp);
430
431 void (*exit)(struct ip_vs_protocol *pp);
432
433 int (*init_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
434
435 void (*exit_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
436
437 int (*conn_schedule)(struct netns_ipvs *ipvs,
438 int af, struct sk_buff *skb,
439 struct ip_vs_proto_data *pd,
440 int *verdict, struct ip_vs_conn **cpp,
441 struct ip_vs_iphdr *iph);
442
443 struct ip_vs_conn *
444 (*conn_in_get)(struct netns_ipvs *ipvs,
445 int af,
446 const struct sk_buff *skb,
447 const struct ip_vs_iphdr *iph);
448
449 struct ip_vs_conn *
450 (*conn_out_get)(struct netns_ipvs *ipvs,
451 int af,
452 const struct sk_buff *skb,
453 const struct ip_vs_iphdr *iph);
454
455 int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
456 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
457
458 int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
459 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
460
461 int (*csum_check)(int af, struct sk_buff *skb,
462 struct ip_vs_protocol *pp);
463
464 const char *(*state_name)(int state);
465
466 void (*state_transition)(struct ip_vs_conn *cp, int direction,
467 const struct sk_buff *skb,
468 struct ip_vs_proto_data *pd);
469
470 int (*register_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
471
472 void (*unregister_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
473
474 int (*app_conn_bind)(struct ip_vs_conn *cp);
475
476 void (*debug_packet)(int af, struct ip_vs_protocol *pp,
477 const struct sk_buff *skb,
478 int offset,
479 const char *msg);
480
481 void (*timeout_change)(struct ip_vs_proto_data *pd, int flags);
482 };
483
484 /* protocol data per netns */
485 struct ip_vs_proto_data {
486 struct ip_vs_proto_data *next;
487 struct ip_vs_protocol *pp;
488 int *timeout_table; /* protocol timeout table */
489 atomic_t appcnt; /* counter of proto app incs. */
490 struct tcp_states_t *tcp_state_table;
491 };
492
493 struct ip_vs_protocol *ip_vs_proto_get(unsigned short proto);
494 struct ip_vs_proto_data *ip_vs_proto_data_get(struct netns_ipvs *ipvs,
495 unsigned short proto);
496
497 struct ip_vs_conn_param {
498 struct netns_ipvs *ipvs;
499 const union nf_inet_addr *caddr;
500 const union nf_inet_addr *vaddr;
501 __be16 cport;
502 __be16 vport;
503 __u16 protocol;
504 u16 af;
505
506 const struct ip_vs_pe *pe;
507 char *pe_data;
508 __u8 pe_data_len;
509 };
510
511 /* IP_VS structure allocated for each dynamically scheduled connection */
512 struct ip_vs_conn {
513 struct hlist_node c_list; /* hashed list heads */
514 /* Protocol, addresses and port numbers */
515 __be16 cport;
516 __be16 dport;
517 __be16 vport;
518 u16 af; /* address family */
519 union nf_inet_addr caddr; /* client address */
520 union nf_inet_addr vaddr; /* virtual address */
521 union nf_inet_addr daddr; /* destination address */
522 volatile __u32 flags; /* status flags */
523 __u16 protocol; /* Which protocol (TCP/UDP) */
524 __u16 daf; /* Address family of the dest */
525 struct netns_ipvs *ipvs;
526
527 /* counter and timer */
528 atomic_t refcnt; /* reference count */
529 struct timer_list timer; /* Expiration timer */
530 volatile unsigned long timeout; /* timeout */
531
532 /* Flags and state transition */
533 spinlock_t lock; /* lock for state transition */
534 volatile __u16 state; /* state info */
535 volatile __u16 old_state; /* old state, to be used for
536 * state transition triggerd
537 * synchronization
538 */
539 __u32 fwmark; /* Fire wall mark from skb */
540 unsigned long sync_endtime; /* jiffies + sent_retries */
541
542 /* Control members */
543 struct ip_vs_conn *control; /* Master control connection */
544 atomic_t n_control; /* Number of controlled ones */
545 struct ip_vs_dest *dest; /* real server */
546 atomic_t in_pkts; /* incoming packet counter */
547
548 /* Packet transmitter for different forwarding methods. If it
549 * mangles the packet, it must return NF_DROP or better NF_STOLEN,
550 * otherwise this must be changed to a sk_buff **.
551 * NF_ACCEPT can be returned when destination is local.
552 */
553 int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp,
554 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
555
556 /* Note: we can group the following members into a structure,
557 * in order to save more space, and the following members are
558 * only used in VS/NAT anyway
559 */
560 struct ip_vs_app *app; /* bound ip_vs_app object */
561 void *app_data; /* Application private data */
562 struct ip_vs_seq in_seq; /* incoming seq. struct */
563 struct ip_vs_seq out_seq; /* outgoing seq. struct */
564
565 const struct ip_vs_pe *pe;
566 char *pe_data;
567 __u8 pe_data_len;
568
569 struct rcu_head rcu_head;
570 };
571
572 /* Extended internal versions of struct ip_vs_service_user and ip_vs_dest_user
573 * for IPv6 support.
574 *
575 * We need these to conveniently pass around service and destination
576 * options, but unfortunately, we also need to keep the old definitions to
577 * maintain userspace backwards compatibility for the setsockopt interface.
578 */
579 struct ip_vs_service_user_kern {
580 /* virtual service addresses */
581 u16 af;
582 u16 protocol;
583 union nf_inet_addr addr; /* virtual ip address */
584 __be16 port;
585 u32 fwmark; /* firwall mark of service */
586
587 /* virtual service options */
588 char *sched_name;
589 char *pe_name;
590 unsigned int flags; /* virtual service flags */
591 unsigned int timeout; /* persistent timeout in sec */
592 __be32 netmask; /* persistent netmask or plen */
593 };
594
595
596 struct ip_vs_dest_user_kern {
597 /* destination server address */
598 union nf_inet_addr addr;
599 __be16 port;
600
601 /* real server options */
602 unsigned int conn_flags; /* connection flags */
603 int weight; /* destination weight */
604
605 /* thresholds for active connections */
606 u32 u_threshold; /* upper threshold */
607 u32 l_threshold; /* lower threshold */
608
609 /* Address family of addr */
610 u16 af;
611 };
612
613
614 /*
615 * The information about the virtual service offered to the net and the
616 * forwarding entries.
617 */
618 struct ip_vs_service {
619 struct hlist_node s_list; /* for normal service table */
620 struct hlist_node f_list; /* for fwmark-based service table */
621 atomic_t refcnt; /* reference counter */
622
623 u16 af; /* address family */
624 __u16 protocol; /* which protocol (TCP/UDP) */
625 union nf_inet_addr addr; /* IP address for virtual service */
626 __be16 port; /* port number for the service */
627 __u32 fwmark; /* firewall mark of the service */
628 unsigned int flags; /* service status flags */
629 unsigned int timeout; /* persistent timeout in ticks */
630 __be32 netmask; /* grouping granularity, mask/plen */
631 struct netns_ipvs *ipvs;
632
633 struct list_head destinations; /* real server d-linked list */
634 __u32 num_dests; /* number of servers */
635 struct ip_vs_stats stats; /* statistics for the service */
636
637 /* for scheduling */
638 struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */
639 spinlock_t sched_lock; /* lock sched_data */
640 void *sched_data; /* scheduler application data */
641
642 /* alternate persistence engine */
643 struct ip_vs_pe __rcu *pe;
644
645 struct rcu_head rcu_head;
646 };
647
648 /* Information for cached dst */
649 struct ip_vs_dest_dst {
650 struct dst_entry *dst_cache; /* destination cache entry */
651 u32 dst_cookie;
652 union nf_inet_addr dst_saddr;
653 struct rcu_head rcu_head;
654 };
655
656 /* The real server destination forwarding entry with ip address, port number,
657 * and so on.
658 */
659 struct ip_vs_dest {
660 struct list_head n_list; /* for the dests in the service */
661 struct hlist_node d_list; /* for table with all the dests */
662
663 u16 af; /* address family */
664 __be16 port; /* port number of the server */
665 union nf_inet_addr addr; /* IP address of the server */
666 volatile unsigned int flags; /* dest status flags */
667 atomic_t conn_flags; /* flags to copy to conn */
668 atomic_t weight; /* server weight */
669
670 atomic_t refcnt; /* reference counter */
671 struct ip_vs_stats stats; /* statistics */
672 unsigned long idle_start; /* start time, jiffies */
673
674 /* connection counters and thresholds */
675 atomic_t activeconns; /* active connections */
676 atomic_t inactconns; /* inactive connections */
677 atomic_t persistconns; /* persistent connections */
678 __u32 u_threshold; /* upper threshold */
679 __u32 l_threshold; /* lower threshold */
680
681 /* for destination cache */
682 spinlock_t dst_lock; /* lock of dst_cache */
683 struct ip_vs_dest_dst __rcu *dest_dst; /* cached dst info */
684
685 /* for virtual service */
686 struct ip_vs_service __rcu *svc; /* service it belongs to */
687 __u16 protocol; /* which protocol (TCP/UDP) */
688 __be16 vport; /* virtual port number */
689 union nf_inet_addr vaddr; /* virtual IP address */
690 __u32 vfwmark; /* firewall mark of service */
691
692 struct list_head t_list; /* in dest_trash */
693 unsigned int in_rs_table:1; /* we are in rs_table */
694 };
695
696 /* The scheduler object */
697 struct ip_vs_scheduler {
698 struct list_head n_list; /* d-linked list head */
699 char *name; /* scheduler name */
700 atomic_t refcnt; /* reference counter */
701 struct module *module; /* THIS_MODULE/NULL */
702
703 /* scheduler initializing service */
704 int (*init_service)(struct ip_vs_service *svc);
705 /* scheduling service finish */
706 void (*done_service)(struct ip_vs_service *svc);
707 /* dest is linked */
708 int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
709 /* dest is unlinked */
710 int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
711 /* dest is updated */
712 int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
713
714 /* selecting a server from the given service */
715 struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc,
716 const struct sk_buff *skb,
717 struct ip_vs_iphdr *iph);
718 };
719
720 /* The persistence engine object */
721 struct ip_vs_pe {
722 struct list_head n_list; /* d-linked list head */
723 char *name; /* scheduler name */
724 atomic_t refcnt; /* reference counter */
725 struct module *module; /* THIS_MODULE/NULL */
726
727 /* get the connection template, if any */
728 int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb);
729 bool (*ct_match)(const struct ip_vs_conn_param *p,
730 struct ip_vs_conn *ct);
731 u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval,
732 bool inverse);
733 int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf);
734 /* create connections for real-server outgoing packets */
735 struct ip_vs_conn* (*conn_out)(struct ip_vs_service *svc,
736 struct ip_vs_dest *dest,
737 struct sk_buff *skb,
738 const struct ip_vs_iphdr *iph,
739 __be16 dport, __be16 cport);
740 };
741
742 /* The application module object (a.k.a. app incarnation) */
743 struct ip_vs_app {
744 struct list_head a_list; /* member in app list */
745 int type; /* IP_VS_APP_TYPE_xxx */
746 char *name; /* application module name */
747 __u16 protocol;
748 struct module *module; /* THIS_MODULE/NULL */
749 struct list_head incs_list; /* list of incarnations */
750
751 /* members for application incarnations */
752 struct list_head p_list; /* member in proto app list */
753 struct ip_vs_app *app; /* its real application */
754 __be16 port; /* port number in net order */
755 atomic_t usecnt; /* usage counter */
756 struct rcu_head rcu_head;
757
758 /* output hook: Process packet in inout direction, diff set for TCP.
759 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
760 * 2=Mangled but checksum was not updated
761 */
762 int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *,
763 struct sk_buff *, int *diff);
764
765 /* input hook: Process packet in outin direction, diff set for TCP.
766 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
767 * 2=Mangled but checksum was not updated
768 */
769 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *,
770 struct sk_buff *, int *diff);
771
772 /* ip_vs_app initializer */
773 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *);
774
775 /* ip_vs_app finish */
776 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *);
777
778
779 /* not used now */
780 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *,
781 struct ip_vs_protocol *);
782
783 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *);
784
785 int * timeout_table;
786 int * timeouts;
787 int timeouts_size;
788
789 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app,
790 int *verdict, struct ip_vs_conn **cpp);
791
792 struct ip_vs_conn *
793 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app,
794 const struct iphdr *iph, int inverse);
795
796 struct ip_vs_conn *
797 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app,
798 const struct iphdr *iph, int inverse);
799
800 int (*state_transition)(struct ip_vs_conn *cp, int direction,
801 const struct sk_buff *skb,
802 struct ip_vs_app *app);
803
804 void (*timeout_change)(struct ip_vs_app *app, int flags);
805 };
806
807 struct ipvs_master_sync_state {
808 struct list_head sync_queue;
809 struct ip_vs_sync_buff *sync_buff;
810 unsigned long sync_queue_len;
811 unsigned int sync_queue_delay;
812 struct task_struct *master_thread;
813 struct delayed_work master_wakeup_work;
814 struct netns_ipvs *ipvs;
815 };
816
817 /* How much time to keep dests in trash */
818 #define IP_VS_DEST_TRASH_PERIOD (120 * HZ)
819
820 struct ipvs_sync_daemon_cfg {
821 union nf_inet_addr mcast_group;
822 int syncid;
823 u16 sync_maxlen;
824 u16 mcast_port;
825 u8 mcast_af;
826 u8 mcast_ttl;
827 /* multicast interface name */
828 char mcast_ifn[IP_VS_IFNAME_MAXLEN];
829 };
830
831 /* IPVS in network namespace */
832 struct netns_ipvs {
833 int gen; /* Generation */
834 int enable; /* enable like nf_hooks do */
835 /* Hash table: for real service lookups */
836 #define IP_VS_RTAB_BITS 4
837 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
838 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
839
840 struct hlist_head rs_table[IP_VS_RTAB_SIZE];
841 /* ip_vs_app */
842 struct list_head app_list;
843 /* ip_vs_proto */
844 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */
845 struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE];
846 /* ip_vs_proto_tcp */
847 #ifdef CONFIG_IP_VS_PROTO_TCP
848 #define TCP_APP_TAB_BITS 4
849 #define TCP_APP_TAB_SIZE (1 << TCP_APP_TAB_BITS)
850 #define TCP_APP_TAB_MASK (TCP_APP_TAB_SIZE - 1)
851 struct list_head tcp_apps[TCP_APP_TAB_SIZE];
852 #endif
853 /* ip_vs_proto_udp */
854 #ifdef CONFIG_IP_VS_PROTO_UDP
855 #define UDP_APP_TAB_BITS 4
856 #define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS)
857 #define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1)
858 struct list_head udp_apps[UDP_APP_TAB_SIZE];
859 #endif
860 /* ip_vs_proto_sctp */
861 #ifdef CONFIG_IP_VS_PROTO_SCTP
862 #define SCTP_APP_TAB_BITS 4
863 #define SCTP_APP_TAB_SIZE (1 << SCTP_APP_TAB_BITS)
864 #define SCTP_APP_TAB_MASK (SCTP_APP_TAB_SIZE - 1)
865 /* Hash table for SCTP application incarnations */
866 struct list_head sctp_apps[SCTP_APP_TAB_SIZE];
867 #endif
868 /* ip_vs_conn */
869 atomic_t conn_count; /* connection counter */
870
871 /* ip_vs_ctl */
872 struct ip_vs_stats tot_stats; /* Statistics & est. */
873
874 int num_services; /* no of virtual services */
875
876 /* Trash for destinations */
877 struct list_head dest_trash;
878 spinlock_t dest_trash_lock;
879 struct timer_list dest_trash_timer; /* expiration timer */
880 /* Service counters */
881 atomic_t ftpsvc_counter;
882 atomic_t nullsvc_counter;
883 atomic_t conn_out_counter;
884
885 #ifdef CONFIG_SYSCTL
886 /* 1/rate drop and drop-entry variables */
887 struct delayed_work defense_work; /* Work handler */
888 int drop_rate;
889 int drop_counter;
890 atomic_t dropentry;
891 /* locks in ctl.c */
892 spinlock_t dropentry_lock; /* drop entry handling */
893 spinlock_t droppacket_lock; /* drop packet handling */
894 spinlock_t securetcp_lock; /* state and timeout tables */
895
896 /* sys-ctl struct */
897 struct ctl_table_header *sysctl_hdr;
898 struct ctl_table *sysctl_tbl;
899 #endif
900
901 /* sysctl variables */
902 int sysctl_amemthresh;
903 int sysctl_am_droprate;
904 int sysctl_drop_entry;
905 int sysctl_drop_packet;
906 int sysctl_secure_tcp;
907 #ifdef CONFIG_IP_VS_NFCT
908 int sysctl_conntrack;
909 #endif
910 int sysctl_snat_reroute;
911 int sysctl_sync_ver;
912 int sysctl_sync_ports;
913 int sysctl_sync_persist_mode;
914 unsigned long sysctl_sync_qlen_max;
915 int sysctl_sync_sock_size;
916 int sysctl_cache_bypass;
917 int sysctl_expire_nodest_conn;
918 int sysctl_sloppy_tcp;
919 int sysctl_sloppy_sctp;
920 int sysctl_expire_quiescent_template;
921 int sysctl_sync_threshold[2];
922 unsigned int sysctl_sync_refresh_period;
923 int sysctl_sync_retries;
924 int sysctl_nat_icmp_send;
925 int sysctl_pmtu_disc;
926 int sysctl_backup_only;
927 int sysctl_conn_reuse_mode;
928 int sysctl_schedule_icmp;
929 int sysctl_ignore_tunneled;
930
931 /* ip_vs_lblc */
932 int sysctl_lblc_expiration;
933 struct ctl_table_header *lblc_ctl_header;
934 struct ctl_table *lblc_ctl_table;
935 /* ip_vs_lblcr */
936 int sysctl_lblcr_expiration;
937 struct ctl_table_header *lblcr_ctl_header;
938 struct ctl_table *lblcr_ctl_table;
939 /* ip_vs_est */
940 struct list_head est_list; /* estimator list */
941 spinlock_t est_lock;
942 struct timer_list est_timer; /* Estimation timer */
943 /* ip_vs_sync */
944 spinlock_t sync_lock;
945 struct ipvs_master_sync_state *ms;
946 spinlock_t sync_buff_lock;
947 struct task_struct **backup_threads;
948 int threads_mask;
949 volatile int sync_state;
950 struct mutex sync_mutex;
951 struct ipvs_sync_daemon_cfg mcfg; /* Master Configuration */
952 struct ipvs_sync_daemon_cfg bcfg; /* Backup Configuration */
953 /* net name space ptr */
954 struct net *net; /* Needed by timer routines */
955 /* Number of heterogeneous destinations, needed becaus heterogeneous
956 * are not supported when synchronization is enabled.
957 */
958 unsigned int mixed_address_family_dests;
959 };
960
961 #define DEFAULT_SYNC_THRESHOLD 3
962 #define DEFAULT_SYNC_PERIOD 50
963 #define DEFAULT_SYNC_VER 1
964 #define DEFAULT_SLOPPY_TCP 0
965 #define DEFAULT_SLOPPY_SCTP 0
966 #define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ)
967 #define DEFAULT_SYNC_RETRIES 0
968 #define IPVS_SYNC_WAKEUP_RATE 8
969 #define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4)
970 #define IPVS_SYNC_SEND_DELAY (HZ / 50)
971 #define IPVS_SYNC_CHECK_PERIOD HZ
972 #define IPVS_SYNC_FLUSH_TIME (HZ * 2)
973 #define IPVS_SYNC_PORTS_MAX (1 << 6)
974
975 #ifdef CONFIG_SYSCTL
976
977 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
978 {
979 return ipvs->sysctl_sync_threshold[0];
980 }
981
982 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
983 {
984 return ACCESS_ONCE(ipvs->sysctl_sync_threshold[1]);
985 }
986
987 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
988 {
989 return ACCESS_ONCE(ipvs->sysctl_sync_refresh_period);
990 }
991
992 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
993 {
994 return ipvs->sysctl_sync_retries;
995 }
996
997 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
998 {
999 return ipvs->sysctl_sync_ver;
1000 }
1001
1002 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1003 {
1004 return ipvs->sysctl_sloppy_tcp;
1005 }
1006
1007 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1008 {
1009 return ipvs->sysctl_sloppy_sctp;
1010 }
1011
1012 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1013 {
1014 return ACCESS_ONCE(ipvs->sysctl_sync_ports);
1015 }
1016
1017 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1018 {
1019 return ipvs->sysctl_sync_persist_mode;
1020 }
1021
1022 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1023 {
1024 return ipvs->sysctl_sync_qlen_max;
1025 }
1026
1027 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1028 {
1029 return ipvs->sysctl_sync_sock_size;
1030 }
1031
1032 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1033 {
1034 return ipvs->sysctl_pmtu_disc;
1035 }
1036
1037 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1038 {
1039 return ipvs->sync_state & IP_VS_STATE_BACKUP &&
1040 ipvs->sysctl_backup_only;
1041 }
1042
1043 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1044 {
1045 return ipvs->sysctl_conn_reuse_mode;
1046 }
1047
1048 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1049 {
1050 return ipvs->sysctl_schedule_icmp;
1051 }
1052
1053 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1054 {
1055 return ipvs->sysctl_ignore_tunneled;
1056 }
1057
1058 static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1059 {
1060 return ipvs->sysctl_cache_bypass;
1061 }
1062
1063 #else
1064
1065 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1066 {
1067 return DEFAULT_SYNC_THRESHOLD;
1068 }
1069
1070 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1071 {
1072 return DEFAULT_SYNC_PERIOD;
1073 }
1074
1075 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1076 {
1077 return DEFAULT_SYNC_REFRESH_PERIOD;
1078 }
1079
1080 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1081 {
1082 return DEFAULT_SYNC_RETRIES & 3;
1083 }
1084
1085 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1086 {
1087 return DEFAULT_SYNC_VER;
1088 }
1089
1090 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1091 {
1092 return DEFAULT_SLOPPY_TCP;
1093 }
1094
1095 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1096 {
1097 return DEFAULT_SLOPPY_SCTP;
1098 }
1099
1100 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1101 {
1102 return 1;
1103 }
1104
1105 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1106 {
1107 return 0;
1108 }
1109
1110 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1111 {
1112 return IPVS_SYNC_QLEN_MAX;
1113 }
1114
1115 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1116 {
1117 return 0;
1118 }
1119
1120 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1121 {
1122 return 1;
1123 }
1124
1125 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1126 {
1127 return 0;
1128 }
1129
1130 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1131 {
1132 return 1;
1133 }
1134
1135 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1136 {
1137 return 0;
1138 }
1139
1140 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1141 {
1142 return 0;
1143 }
1144
1145 static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1146 {
1147 return 0;
1148 }
1149
1150 #endif
1151
1152 /* IPVS core functions
1153 * (from ip_vs_core.c)
1154 */
1155 const char *ip_vs_proto_name(unsigned int proto);
1156 void ip_vs_init_hash_table(struct list_head *table, int rows);
1157 struct ip_vs_conn *ip_vs_new_conn_out(struct ip_vs_service *svc,
1158 struct ip_vs_dest *dest,
1159 struct sk_buff *skb,
1160 const struct ip_vs_iphdr *iph,
1161 __be16 dport,
1162 __be16 cport);
1163 #define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t)))
1164
1165 #define IP_VS_APP_TYPE_FTP 1
1166
1167 /* ip_vs_conn handling functions
1168 * (from ip_vs_conn.c)
1169 */
1170 enum {
1171 IP_VS_DIR_INPUT = 0,
1172 IP_VS_DIR_OUTPUT,
1173 IP_VS_DIR_INPUT_ONLY,
1174 IP_VS_DIR_LAST,
1175 };
1176
1177 static inline void ip_vs_conn_fill_param(struct netns_ipvs *ipvs, int af, int protocol,
1178 const union nf_inet_addr *caddr,
1179 __be16 cport,
1180 const union nf_inet_addr *vaddr,
1181 __be16 vport,
1182 struct ip_vs_conn_param *p)
1183 {
1184 p->ipvs = ipvs;
1185 p->af = af;
1186 p->protocol = protocol;
1187 p->caddr = caddr;
1188 p->cport = cport;
1189 p->vaddr = vaddr;
1190 p->vport = vport;
1191 p->pe = NULL;
1192 p->pe_data = NULL;
1193 }
1194
1195 struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p);
1196 struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p);
1197
1198 struct ip_vs_conn * ip_vs_conn_in_get_proto(struct netns_ipvs *ipvs, int af,
1199 const struct sk_buff *skb,
1200 const struct ip_vs_iphdr *iph);
1201
1202 struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p);
1203
1204 struct ip_vs_conn * ip_vs_conn_out_get_proto(struct netns_ipvs *ipvs, int af,
1205 const struct sk_buff *skb,
1206 const struct ip_vs_iphdr *iph);
1207
1208 /* Get reference to gain full access to conn.
1209 * By default, RCU read-side critical sections have access only to
1210 * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference.
1211 */
1212 static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp)
1213 {
1214 return atomic_inc_not_zero(&cp->refcnt);
1215 }
1216
1217 /* put back the conn without restarting its timer */
1218 static inline void __ip_vs_conn_put(struct ip_vs_conn *cp)
1219 {
1220 smp_mb__before_atomic();
1221 atomic_dec(&cp->refcnt);
1222 }
1223 void ip_vs_conn_put(struct ip_vs_conn *cp);
1224 void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport);
1225
1226 struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, int dest_af,
1227 const union nf_inet_addr *daddr,
1228 __be16 dport, unsigned int flags,
1229 struct ip_vs_dest *dest, __u32 fwmark);
1230 void ip_vs_conn_expire_now(struct ip_vs_conn *cp);
1231
1232 const char *ip_vs_state_name(__u16 proto, int state);
1233
1234 void ip_vs_tcp_conn_listen(struct ip_vs_conn *cp);
1235 int ip_vs_check_template(struct ip_vs_conn *ct, struct ip_vs_dest *cdest);
1236 void ip_vs_random_dropentry(struct netns_ipvs *ipvs);
1237 int ip_vs_conn_init(void);
1238 void ip_vs_conn_cleanup(void);
1239
1240 static inline void ip_vs_control_del(struct ip_vs_conn *cp)
1241 {
1242 struct ip_vs_conn *ctl_cp = cp->control;
1243 if (!ctl_cp) {
1244 IP_VS_ERR_BUF("request control DEL for uncontrolled: "
1245 "%s:%d to %s:%d\n",
1246 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1247 ntohs(cp->cport),
1248 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1249 ntohs(cp->vport));
1250
1251 return;
1252 }
1253
1254 IP_VS_DBG_BUF(7, "DELeting control for: "
1255 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1256 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1257 ntohs(cp->cport),
1258 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1259 ntohs(ctl_cp->cport));
1260
1261 cp->control = NULL;
1262 if (atomic_read(&ctl_cp->n_control) == 0) {
1263 IP_VS_ERR_BUF("BUG control DEL with n=0 : "
1264 "%s:%d to %s:%d\n",
1265 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1266 ntohs(cp->cport),
1267 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1268 ntohs(cp->vport));
1269
1270 return;
1271 }
1272 atomic_dec(&ctl_cp->n_control);
1273 }
1274
1275 static inline void
1276 ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp)
1277 {
1278 if (cp->control) {
1279 IP_VS_ERR_BUF("request control ADD for already controlled: "
1280 "%s:%d to %s:%d\n",
1281 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1282 ntohs(cp->cport),
1283 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1284 ntohs(cp->vport));
1285
1286 ip_vs_control_del(cp);
1287 }
1288
1289 IP_VS_DBG_BUF(7, "ADDing control for: "
1290 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1291 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1292 ntohs(cp->cport),
1293 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1294 ntohs(ctl_cp->cport));
1295
1296 cp->control = ctl_cp;
1297 atomic_inc(&ctl_cp->n_control);
1298 }
1299
1300 /* IPVS netns init & cleanup functions */
1301 int ip_vs_estimator_net_init(struct netns_ipvs *ipvs);
1302 int ip_vs_control_net_init(struct netns_ipvs *ipvs);
1303 int ip_vs_protocol_net_init(struct netns_ipvs *ipvs);
1304 int ip_vs_app_net_init(struct netns_ipvs *ipvs);
1305 int ip_vs_conn_net_init(struct netns_ipvs *ipvs);
1306 int ip_vs_sync_net_init(struct netns_ipvs *ipvs);
1307 void ip_vs_conn_net_cleanup(struct netns_ipvs *ipvs);
1308 void ip_vs_app_net_cleanup(struct netns_ipvs *ipvs);
1309 void ip_vs_protocol_net_cleanup(struct netns_ipvs *ipvs);
1310 void ip_vs_control_net_cleanup(struct netns_ipvs *ipvs);
1311 void ip_vs_estimator_net_cleanup(struct netns_ipvs *ipvs);
1312 void ip_vs_sync_net_cleanup(struct netns_ipvs *ipvs);
1313 void ip_vs_service_net_cleanup(struct netns_ipvs *ipvs);
1314
1315 /* IPVS application functions
1316 * (from ip_vs_app.c)
1317 */
1318 #define IP_VS_APP_MAX_PORTS 8
1319 struct ip_vs_app *register_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1320 void unregister_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1321 int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1322 void ip_vs_unbind_app(struct ip_vs_conn *cp);
1323 int register_ip_vs_app_inc(struct netns_ipvs *ipvs, struct ip_vs_app *app, __u16 proto,
1324 __u16 port);
1325 int ip_vs_app_inc_get(struct ip_vs_app *inc);
1326 void ip_vs_app_inc_put(struct ip_vs_app *inc);
1327
1328 int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb);
1329 int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb);
1330
1331 int register_ip_vs_pe(struct ip_vs_pe *pe);
1332 int unregister_ip_vs_pe(struct ip_vs_pe *pe);
1333 struct ip_vs_pe *ip_vs_pe_getbyname(const char *name);
1334 struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name);
1335
1336 /* Use a #define to avoid all of module.h just for these trivial ops */
1337 #define ip_vs_pe_get(pe) \
1338 if (pe && pe->module) \
1339 __module_get(pe->module);
1340
1341 #define ip_vs_pe_put(pe) \
1342 if (pe && pe->module) \
1343 module_put(pe->module);
1344
1345 /* IPVS protocol functions (from ip_vs_proto.c) */
1346 int ip_vs_protocol_init(void);
1347 void ip_vs_protocol_cleanup(void);
1348 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags);
1349 int *ip_vs_create_timeout_table(int *table, int size);
1350 int ip_vs_set_state_timeout(int *table, int num, const char *const *names,
1351 const char *name, int to);
1352 void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
1353 const struct sk_buff *skb, int offset,
1354 const char *msg);
1355
1356 extern struct ip_vs_protocol ip_vs_protocol_tcp;
1357 extern struct ip_vs_protocol ip_vs_protocol_udp;
1358 extern struct ip_vs_protocol ip_vs_protocol_icmp;
1359 extern struct ip_vs_protocol ip_vs_protocol_esp;
1360 extern struct ip_vs_protocol ip_vs_protocol_ah;
1361 extern struct ip_vs_protocol ip_vs_protocol_sctp;
1362
1363 /* Registering/unregistering scheduler functions
1364 * (from ip_vs_sched.c)
1365 */
1366 int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1367 int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1368 int ip_vs_bind_scheduler(struct ip_vs_service *svc,
1369 struct ip_vs_scheduler *scheduler);
1370 void ip_vs_unbind_scheduler(struct ip_vs_service *svc,
1371 struct ip_vs_scheduler *sched);
1372 struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name);
1373 void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler);
1374 struct ip_vs_conn *
1375 ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb,
1376 struct ip_vs_proto_data *pd, int *ignored,
1377 struct ip_vs_iphdr *iph);
1378 int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb,
1379 struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph);
1380
1381 void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg);
1382
1383 /* IPVS control data and functions (from ip_vs_ctl.c) */
1384 extern struct ip_vs_stats ip_vs_stats;
1385 extern int sysctl_ip_vs_sync_ver;
1386
1387 struct ip_vs_service *
1388 ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
1389 const union nf_inet_addr *vaddr, __be16 vport);
1390
1391 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1392 const union nf_inet_addr *daddr, __be16 dport);
1393
1394 struct ip_vs_dest *
1395 ip_vs_find_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1396 const union nf_inet_addr *daddr, __be16 dport);
1397
1398 int ip_vs_use_count_inc(void);
1399 void ip_vs_use_count_dec(void);
1400 int ip_vs_register_nl_ioctl(void);
1401 void ip_vs_unregister_nl_ioctl(void);
1402 int ip_vs_control_init(void);
1403 void ip_vs_control_cleanup(void);
1404 struct ip_vs_dest *
1405 ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
1406 const union nf_inet_addr *daddr, __be16 dport,
1407 const union nf_inet_addr *vaddr, __be16 vport,
1408 __u16 protocol, __u32 fwmark, __u32 flags);
1409 void ip_vs_try_bind_dest(struct ip_vs_conn *cp);
1410
1411 static inline void ip_vs_dest_hold(struct ip_vs_dest *dest)
1412 {
1413 atomic_inc(&dest->refcnt);
1414 }
1415
1416 static inline void ip_vs_dest_put(struct ip_vs_dest *dest)
1417 {
1418 smp_mb__before_atomic();
1419 atomic_dec(&dest->refcnt);
1420 }
1421
1422 static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest)
1423 {
1424 if (atomic_dec_return(&dest->refcnt) < 0)
1425 kfree(dest);
1426 }
1427
1428 /* IPVS sync daemon data and function prototypes
1429 * (from ip_vs_sync.c)
1430 */
1431 int start_sync_thread(struct netns_ipvs *ipvs, struct ipvs_sync_daemon_cfg *cfg,
1432 int state);
1433 int stop_sync_thread(struct netns_ipvs *ipvs, int state);
1434 void ip_vs_sync_conn(struct netns_ipvs *ipvs, struct ip_vs_conn *cp, int pkts);
1435
1436 /* IPVS rate estimator prototypes (from ip_vs_est.c) */
1437 void ip_vs_start_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1438 void ip_vs_stop_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1439 void ip_vs_zero_estimator(struct ip_vs_stats *stats);
1440 void ip_vs_read_estimator(struct ip_vs_kstats *dst, struct ip_vs_stats *stats);
1441
1442 /* Various IPVS packet transmitters (from ip_vs_xmit.c) */
1443 int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1444 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1445 int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1446 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1447 int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1448 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1449 int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1450 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1451 int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1452 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1453 int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1454 struct ip_vs_protocol *pp, int offset,
1455 unsigned int hooknum, struct ip_vs_iphdr *iph);
1456 void ip_vs_dest_dst_rcu_free(struct rcu_head *head);
1457
1458 #ifdef CONFIG_IP_VS_IPV6
1459 int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1460 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1461 int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1462 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1463 int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1464 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1465 int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1466 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1467 int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1468 struct ip_vs_protocol *pp, int offset,
1469 unsigned int hooknum, struct ip_vs_iphdr *iph);
1470 #endif
1471
1472 #ifdef CONFIG_SYSCTL
1473 /* This is a simple mechanism to ignore packets when
1474 * we are loaded. Just set ip_vs_drop_rate to 'n' and
1475 * we start to drop 1/rate of the packets
1476 */
1477 static inline int ip_vs_todrop(struct netns_ipvs *ipvs)
1478 {
1479 if (!ipvs->drop_rate)
1480 return 0;
1481 if (--ipvs->drop_counter > 0)
1482 return 0;
1483 ipvs->drop_counter = ipvs->drop_rate;
1484 return 1;
1485 }
1486 #else
1487 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; }
1488 #endif
1489
1490 /* ip_vs_fwd_tag returns the forwarding tag of the connection */
1491 #define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK)
1492
1493 static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp)
1494 {
1495 char fwd;
1496
1497 switch (IP_VS_FWD_METHOD(cp)) {
1498 case IP_VS_CONN_F_MASQ:
1499 fwd = 'M'; break;
1500 case IP_VS_CONN_F_LOCALNODE:
1501 fwd = 'L'; break;
1502 case IP_VS_CONN_F_TUNNEL:
1503 fwd = 'T'; break;
1504 case IP_VS_CONN_F_DROUTE:
1505 fwd = 'R'; break;
1506 case IP_VS_CONN_F_BYPASS:
1507 fwd = 'B'; break;
1508 default:
1509 fwd = '?'; break;
1510 }
1511 return fwd;
1512 }
1513
1514 void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp,
1515 struct ip_vs_conn *cp, int dir);
1516
1517 #ifdef CONFIG_IP_VS_IPV6
1518 void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp,
1519 struct ip_vs_conn *cp, int dir);
1520 #endif
1521
1522 __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset);
1523
1524 static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum)
1525 {
1526 __be32 diff[2] = { ~old, new };
1527
1528 return csum_partial(diff, sizeof(diff), oldsum);
1529 }
1530
1531 #ifdef CONFIG_IP_VS_IPV6
1532 static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new,
1533 __wsum oldsum)
1534 {
1535 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0],
1536 new[3], new[2], new[1], new[0] };
1537
1538 return csum_partial(diff, sizeof(diff), oldsum);
1539 }
1540 #endif
1541
1542 static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum)
1543 {
1544 __be16 diff[2] = { ~old, new };
1545
1546 return csum_partial(diff, sizeof(diff), oldsum);
1547 }
1548
1549 /* Forget current conntrack (unconfirmed) and attach notrack entry */
1550 static inline void ip_vs_notrack(struct sk_buff *skb)
1551 {
1552 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1553 enum ip_conntrack_info ctinfo;
1554 struct nf_conn *ct = nf_ct_get(skb, &ctinfo);
1555
1556 if (!ct || !nf_ct_is_untracked(ct)) {
1557 nf_conntrack_put(skb->nfct);
1558 skb->nfct = &nf_ct_untracked_get()->ct_general;
1559 skb->nfctinfo = IP_CT_NEW;
1560 nf_conntrack_get(skb->nfct);
1561 }
1562 #endif
1563 }
1564
1565 #ifdef CONFIG_IP_VS_NFCT
1566 /* Netfilter connection tracking
1567 * (from ip_vs_nfct.c)
1568 */
1569 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1570 {
1571 #ifdef CONFIG_SYSCTL
1572 return ipvs->sysctl_conntrack;
1573 #else
1574 return 0;
1575 #endif
1576 }
1577
1578 void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp,
1579 int outin);
1580 int ip_vs_confirm_conntrack(struct sk_buff *skb);
1581 void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct,
1582 struct ip_vs_conn *cp, u_int8_t proto,
1583 const __be16 port, int from_rs);
1584 void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp);
1585
1586 #else
1587
1588 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1589 {
1590 return 0;
1591 }
1592
1593 static inline void ip_vs_update_conntrack(struct sk_buff *skb,
1594 struct ip_vs_conn *cp, int outin)
1595 {
1596 }
1597
1598 static inline int ip_vs_confirm_conntrack(struct sk_buff *skb)
1599 {
1600 return NF_ACCEPT;
1601 }
1602
1603 static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp)
1604 {
1605 }
1606 #endif /* CONFIG_IP_VS_NFCT */
1607
1608 /* Really using conntrack? */
1609 static inline bool ip_vs_conn_uses_conntrack(struct ip_vs_conn *cp,
1610 struct sk_buff *skb)
1611 {
1612 #ifdef CONFIG_IP_VS_NFCT
1613 enum ip_conntrack_info ctinfo;
1614 struct nf_conn *ct;
1615
1616 if (!(cp->flags & IP_VS_CONN_F_NFCT))
1617 return false;
1618 ct = nf_ct_get(skb, &ctinfo);
1619 if (ct && !nf_ct_is_untracked(ct))
1620 return true;
1621 #endif
1622 return false;
1623 }
1624
1625 static inline int
1626 ip_vs_dest_conn_overhead(struct ip_vs_dest *dest)
1627 {
1628 /* We think the overhead of processing active connections is 256
1629 * times higher than that of inactive connections in average. (This
1630 * 256 times might not be accurate, we will change it later) We
1631 * use the following formula to estimate the overhead now:
1632 * dest->activeconns*256 + dest->inactconns
1633 */
1634 return (atomic_read(&dest->activeconns) << 8) +
1635 atomic_read(&dest->inactconns);
1636 }
1637
1638 #endif /* _NET_IP_VS_H */
Linux with added FPC-III support