projects / deliverable / linux.git / blob
? search:


728a670d69fd027e7b7d5c91228b695241e4ab60
[deliverable/linux.git] / include / net / ip_vs.h
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 /* Get net ptr from skb in traffic cases
42 * use skb_sknet when call is from userland (ioctl or netlink)
43 */
44 static inline struct net *skb_net(const struct sk_buff *skb)
45 {
46 #ifdef CONFIG_NET_NS
47 #ifdef CONFIG_IP_VS_DEBUG
48 /*
49 * This is used for debug only.
50 * Start with the most likely hit
51 * End with BUG
52 */
53 if (likely(skb->dev && dev_net(skb->dev)))
54 return dev_net(skb->dev);
55 if (skb_dst(skb) && skb_dst(skb)->dev)
56 return dev_net(skb_dst(skb)->dev);
57 WARN(skb->sk, "Maybe skb_sknet should be used in %s() at line:%d\n",
58 __func__, __LINE__);
59 if (likely(skb->sk && sock_net(skb->sk)))
60 return sock_net(skb->sk);
61 pr_err("There is no net ptr to find in the skb in %s() line:%d\n",
62 __func__, __LINE__);
63 BUG();
64 #else
65 return dev_net(skb->dev ? : skb_dst(skb)->dev);
66 #endif
67 #else
68 return &init_net;
69 #endif
70 }
71
72 static inline struct net *skb_sknet(const struct sk_buff *skb)
73 {
74 #ifdef CONFIG_NET_NS
75 #ifdef CONFIG_IP_VS_DEBUG
76 /* Start with the most likely hit */
77 if (likely(skb->sk && sock_net(skb->sk)))
78 return sock_net(skb->sk);
79 WARN(skb->dev, "Maybe skb_net should be used instead in %s() line:%d\n",
80 __func__, __LINE__);
81 if (likely(skb->dev && dev_net(skb->dev)))
82 return dev_net(skb->dev);
83 pr_err("There is no net ptr to find in the skb in %s() line:%d\n",
84 __func__, __LINE__);
85 BUG();
86 #else
87 return sock_net(skb->sk);
88 #endif
89 #else
90 return &init_net;
91 #endif
92 }
93
94 /* This one needed for single_open_net since net is stored directly in
95 * private not as a struct i.e. seq_file_net can't be used.
96 */
97 static inline struct net *seq_file_single_net(struct seq_file *seq)
98 {
99 #ifdef CONFIG_NET_NS
100 return (struct net *)seq->private;
101 #else
102 return &init_net;
103 #endif
104 }
105
106 /* Connections' size value needed by ip_vs_ctl.c */
107 extern int ip_vs_conn_tab_size;
108
109 struct ip_vs_iphdr {
110 int hdr_flags; /* ipvs flags */
111 __u32 off; /* Where IP or IPv4 header starts */
112 __u32 len; /* IPv4 simply where L4 starts
113 * IPv6 where L4 Transport Header starts */
114 __u16 fragoffs; /* IPv6 fragment offset, 0 if first frag (or not frag)*/
115 __s16 protocol;
116 __s32 flags;
117 union nf_inet_addr saddr;
118 union nf_inet_addr daddr;
119 };
120
121 static inline void *frag_safe_skb_hp(const struct sk_buff *skb, int offset,
122 int len, void *buffer,
123 const struct ip_vs_iphdr *ipvsh)
124 {
125 return skb_header_pointer(skb, offset, len, buffer);
126 }
127
128 /* This function handles filling *ip_vs_iphdr, both for IPv4 and IPv6.
129 * IPv6 requires some extra work, as finding proper header position,
130 * depend on the IPv6 extension headers.
131 */
132 static inline int
133 ip_vs_fill_iph_skb_off(int af, const struct sk_buff *skb, int offset,
134 int hdr_flags, struct ip_vs_iphdr *iphdr)
135 {
136 iphdr->hdr_flags = hdr_flags;
137 iphdr->off = offset;
138
139 #ifdef CONFIG_IP_VS_IPV6
140 if (af == AF_INET6) {
141 struct ipv6hdr _iph;
142 const struct ipv6hdr *iph = skb_header_pointer(
143 skb, offset, sizeof(_iph), &_iph);
144 if (!iph)
145 return 0;
146
147 iphdr->saddr.in6 = iph->saddr;
148 iphdr->daddr.in6 = iph->daddr;
149 /* ipv6_find_hdr() updates len, flags */
150 iphdr->len = offset;
151 iphdr->flags = 0;
152 iphdr->protocol = ipv6_find_hdr(skb, &iphdr->len, -1,
153 &iphdr->fragoffs,
154 &iphdr->flags);
155 if (iphdr->protocol < 0)
156 return 0;
157 } else
158 #endif
159 {
160 struct iphdr _iph;
161 const struct iphdr *iph = skb_header_pointer(
162 skb, offset, sizeof(_iph), &_iph);
163 if (!iph)
164 return 0;
165
166 iphdr->len = offset + iph->ihl * 4;
167 iphdr->fragoffs = 0;
168 iphdr->protocol = iph->protocol;
169 iphdr->saddr.ip = iph->saddr;
170 iphdr->daddr.ip = iph->daddr;
171 }
172
173 return 1;
174 }
175
176 static inline int
177 ip_vs_fill_iph_skb_icmp(int af, const struct sk_buff *skb, int offset,
178 bool inverse, struct ip_vs_iphdr *iphdr)
179 {
180 int hdr_flags = IP_VS_HDR_ICMP;
181
182 if (inverse)
183 hdr_flags |= IP_VS_HDR_INVERSE;
184
185 return ip_vs_fill_iph_skb_off(af, skb, offset, hdr_flags, iphdr);
186 }
187
188 static inline int
189 ip_vs_fill_iph_skb(int af, const struct sk_buff *skb, bool inverse,
190 struct ip_vs_iphdr *iphdr)
191 {
192 int hdr_flags = 0;
193
194 if (inverse)
195 hdr_flags |= IP_VS_HDR_INVERSE;
196
197 return ip_vs_fill_iph_skb_off(af, skb, skb_network_offset(skb),
198 hdr_flags, iphdr);
199 }
200
201 static inline bool
202 ip_vs_iph_inverse(const struct ip_vs_iphdr *iph)
203 {
204 return !!(iph->hdr_flags & IP_VS_HDR_INVERSE);
205 }
206
207 static inline bool
208 ip_vs_iph_icmp(const struct ip_vs_iphdr *iph)
209 {
210 return !!(iph->hdr_flags & IP_VS_HDR_ICMP);
211 }
212
213 static inline void ip_vs_addr_copy(int af, union nf_inet_addr *dst,
214 const union nf_inet_addr *src)
215 {
216 #ifdef CONFIG_IP_VS_IPV6
217 if (af == AF_INET6)
218 dst->in6 = src->in6;
219 else
220 #endif
221 dst->ip = src->ip;
222 }
223
224 static inline void ip_vs_addr_set(int af, union nf_inet_addr *dst,
225 const union nf_inet_addr *src)
226 {
227 #ifdef CONFIG_IP_VS_IPV6
228 if (af == AF_INET6) {
229 dst->in6 = src->in6;
230 return;
231 }
232 #endif
233 dst->ip = src->ip;
234 dst->all[1] = 0;
235 dst->all[2] = 0;
236 dst->all[3] = 0;
237 }
238
239 static inline int ip_vs_addr_equal(int af, const union nf_inet_addr *a,
240 const union nf_inet_addr *b)
241 {
242 #ifdef CONFIG_IP_VS_IPV6
243 if (af == AF_INET6)
244 return ipv6_addr_equal(&a->in6, &b->in6);
245 #endif
246 return a->ip == b->ip;
247 }
248
249 #ifdef CONFIG_IP_VS_DEBUG
250 #include <linux/net.h>
251
252 int ip_vs_get_debug_level(void);
253
254 static inline const char *ip_vs_dbg_addr(int af, char *buf, size_t buf_len,
255 const union nf_inet_addr *addr,
256 int *idx)
257 {
258 int len;
259 #ifdef CONFIG_IP_VS_IPV6
260 if (af == AF_INET6)
261 len = snprintf(&buf[*idx], buf_len - *idx, "[%pI6c]",
262 &addr->in6) + 1;
263 else
264 #endif
265 len = snprintf(&buf[*idx], buf_len - *idx, "%pI4",
266 &addr->ip) + 1;
267
268 *idx += len;
269 BUG_ON(*idx > buf_len + 1);
270 return &buf[*idx - len];
271 }
272
273 #define IP_VS_DBG_BUF(level, msg, ...) \
274 do { \
275 char ip_vs_dbg_buf[160]; \
276 int ip_vs_dbg_idx = 0; \
277 if (level <= ip_vs_get_debug_level()) \
278 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
279 } while (0)
280 #define IP_VS_ERR_BUF(msg...) \
281 do { \
282 char ip_vs_dbg_buf[160]; \
283 int ip_vs_dbg_idx = 0; \
284 pr_err(msg); \
285 } while (0)
286
287 /* Only use from within IP_VS_DBG_BUF() or IP_VS_ERR_BUF macros */
288 #define IP_VS_DBG_ADDR(af, addr) \
289 ip_vs_dbg_addr(af, ip_vs_dbg_buf, \
290 sizeof(ip_vs_dbg_buf), addr, \
291 &ip_vs_dbg_idx)
292
293 #define IP_VS_DBG(level, msg, ...) \
294 do { \
295 if (level <= ip_vs_get_debug_level()) \
296 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
297 } while (0)
298 #define IP_VS_DBG_RL(msg, ...) \
299 do { \
300 if (net_ratelimit()) \
301 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
302 } while (0)
303 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) \
304 do { \
305 if (level <= ip_vs_get_debug_level()) \
306 pp->debug_packet(af, pp, skb, ofs, msg); \
307 } while (0)
308 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) \
309 do { \
310 if (level <= ip_vs_get_debug_level() && \
311 net_ratelimit()) \
312 pp->debug_packet(af, pp, skb, ofs, msg); \
313 } while (0)
314 #else /* NO DEBUGGING at ALL */
315 #define IP_VS_DBG_BUF(level, msg...) do {} while (0)
316 #define IP_VS_ERR_BUF(msg...) do {} while (0)
317 #define IP_VS_DBG(level, msg...) do {} while (0)
318 #define IP_VS_DBG_RL(msg...) do {} while (0)
319 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
320 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
321 #endif
322
323 #define IP_VS_BUG() BUG()
324 #define IP_VS_ERR_RL(msg, ...) \
325 do { \
326 if (net_ratelimit()) \
327 pr_err(msg, ##__VA_ARGS__); \
328 } while (0)
329
330 #ifdef CONFIG_IP_VS_DEBUG
331 #define EnterFunction(level) \
332 do { \
333 if (level <= ip_vs_get_debug_level()) \
334 printk(KERN_DEBUG \
335 pr_fmt("Enter: %s, %s line %i\n"), \
336 __func__, __FILE__, __LINE__); \
337 } while (0)
338 #define LeaveFunction(level) \
339 do { \
340 if (level <= ip_vs_get_debug_level()) \
341 printk(KERN_DEBUG \
342 pr_fmt("Leave: %s, %s line %i\n"), \
343 __func__, __FILE__, __LINE__); \
344 } while (0)
345 #else
346 #define EnterFunction(level) do {} while (0)
347 #define LeaveFunction(level) do {} while (0)
348 #endif
349
350 /* The port number of FTP service (in network order). */
351 #define FTPPORT cpu_to_be16(21)
352 #define FTPDATA cpu_to_be16(20)
353
354 /* TCP State Values */
355 enum {
356 IP_VS_TCP_S_NONE = 0,
357 IP_VS_TCP_S_ESTABLISHED,
358 IP_VS_TCP_S_SYN_SENT,
359 IP_VS_TCP_S_SYN_RECV,
360 IP_VS_TCP_S_FIN_WAIT,
361 IP_VS_TCP_S_TIME_WAIT,
362 IP_VS_TCP_S_CLOSE,
363 IP_VS_TCP_S_CLOSE_WAIT,
364 IP_VS_TCP_S_LAST_ACK,
365 IP_VS_TCP_S_LISTEN,
366 IP_VS_TCP_S_SYNACK,
367 IP_VS_TCP_S_LAST
368 };
369
370 /* UDP State Values */
371 enum {
372 IP_VS_UDP_S_NORMAL,
373 IP_VS_UDP_S_LAST,
374 };
375
376 /* ICMP State Values */
377 enum {
378 IP_VS_ICMP_S_NORMAL,
379 IP_VS_ICMP_S_LAST,
380 };
381
382 /* SCTP State Values */
383 enum ip_vs_sctp_states {
384 IP_VS_SCTP_S_NONE,
385 IP_VS_SCTP_S_INIT1,
386 IP_VS_SCTP_S_INIT,
387 IP_VS_SCTP_S_COOKIE_SENT,
388 IP_VS_SCTP_S_COOKIE_REPLIED,
389 IP_VS_SCTP_S_COOKIE_WAIT,
390 IP_VS_SCTP_S_COOKIE,
391 IP_VS_SCTP_S_COOKIE_ECHOED,
392 IP_VS_SCTP_S_ESTABLISHED,
393 IP_VS_SCTP_S_SHUTDOWN_SENT,
394 IP_VS_SCTP_S_SHUTDOWN_RECEIVED,
395 IP_VS_SCTP_S_SHUTDOWN_ACK_SENT,
396 IP_VS_SCTP_S_REJECTED,
397 IP_VS_SCTP_S_CLOSED,
398 IP_VS_SCTP_S_LAST
399 };
400
401 /* Delta sequence info structure
402 * Each ip_vs_conn has 2 (output AND input seq. changes).
403 * Only used in the VS/NAT.
404 */
405 struct ip_vs_seq {
406 __u32 init_seq; /* Add delta from this seq */
407 __u32 delta; /* Delta in sequence numbers */
408 __u32 previous_delta; /* Delta in sequence numbers
409 * before last resized pkt */
410 };
411
412 /* counters per cpu */
413 struct ip_vs_counters {
414 __u64 conns; /* connections scheduled */
415 __u64 inpkts; /* incoming packets */
416 __u64 outpkts; /* outgoing packets */
417 __u64 inbytes; /* incoming bytes */
418 __u64 outbytes; /* outgoing bytes */
419 };
420 /* Stats per cpu */
421 struct ip_vs_cpu_stats {
422 struct ip_vs_counters cnt;
423 struct u64_stats_sync syncp;
424 };
425
426 /* IPVS statistics objects */
427 struct ip_vs_estimator {
428 struct list_head list;
429
430 u64 last_inbytes;
431 u64 last_outbytes;
432 u64 last_conns;
433 u64 last_inpkts;
434 u64 last_outpkts;
435
436 u64 cps;
437 u64 inpps;
438 u64 outpps;
439 u64 inbps;
440 u64 outbps;
441 };
442
443 /*
444 * IPVS statistics object, 64-bit kernel version of struct ip_vs_stats_user
445 */
446 struct ip_vs_kstats {
447 u64 conns; /* connections scheduled */
448 u64 inpkts; /* incoming packets */
449 u64 outpkts; /* outgoing packets */
450 u64 inbytes; /* incoming bytes */
451 u64 outbytes; /* outgoing bytes */
452
453 u64 cps; /* current connection rate */
454 u64 inpps; /* current in packet rate */
455 u64 outpps; /* current out packet rate */
456 u64 inbps; /* current in byte rate */
457 u64 outbps; /* current out byte rate */
458 };
459
460 struct ip_vs_stats {
461 struct ip_vs_kstats kstats; /* kernel statistics */
462 struct ip_vs_estimator est; /* estimator */
463 struct ip_vs_cpu_stats __percpu *cpustats; /* per cpu counters */
464 spinlock_t lock; /* spin lock */
465 struct ip_vs_kstats kstats0; /* reset values */
466 };
467
468 struct dst_entry;
469 struct iphdr;
470 struct ip_vs_conn;
471 struct ip_vs_app;
472 struct sk_buff;
473 struct ip_vs_proto_data;
474
475 struct ip_vs_protocol {
476 struct ip_vs_protocol *next;
477 char *name;
478 u16 protocol;
479 u16 num_states;
480 int dont_defrag;
481
482 void (*init)(struct ip_vs_protocol *pp);
483
484 void (*exit)(struct ip_vs_protocol *pp);
485
486 int (*init_netns)(struct net *net, struct ip_vs_proto_data *pd);
487
488 void (*exit_netns)(struct net *net, struct ip_vs_proto_data *pd);
489
490 int (*conn_schedule)(int af, struct sk_buff *skb,
491 struct ip_vs_proto_data *pd,
492 int *verdict, struct ip_vs_conn **cpp,
493 struct ip_vs_iphdr *iph);
494
495 struct ip_vs_conn *
496 (*conn_in_get)(int af,
497 const struct sk_buff *skb,
498 const struct ip_vs_iphdr *iph);
499
500 struct ip_vs_conn *
501 (*conn_out_get)(int af,
502 const struct sk_buff *skb,
503 const struct ip_vs_iphdr *iph);
504
505 int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
506 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
507
508 int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
509 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
510
511 int (*csum_check)(int af, struct sk_buff *skb,
512 struct ip_vs_protocol *pp);
513
514 const char *(*state_name)(int state);
515
516 void (*state_transition)(struct ip_vs_conn *cp, int direction,
517 const struct sk_buff *skb,
518 struct ip_vs_proto_data *pd);
519
520 int (*register_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
521
522 void (*unregister_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
523
524 int (*app_conn_bind)(struct ip_vs_conn *cp);
525
526 void (*debug_packet)(int af, struct ip_vs_protocol *pp,
527 const struct sk_buff *skb,
528 int offset,
529 const char *msg);
530
531 void (*timeout_change)(struct ip_vs_proto_data *pd, int flags);
532 };
533
534 /* protocol data per netns */
535 struct ip_vs_proto_data {
536 struct ip_vs_proto_data *next;
537 struct ip_vs_protocol *pp;
538 int *timeout_table; /* protocol timeout table */
539 atomic_t appcnt; /* counter of proto app incs. */
540 struct tcp_states_t *tcp_state_table;
541 };
542
543 struct ip_vs_protocol *ip_vs_proto_get(unsigned short proto);
544 struct ip_vs_proto_data *ip_vs_proto_data_get(struct netns_ipvs *ipvs,
545 unsigned short proto);
546
547 struct ip_vs_conn_param {
548 struct netns_ipvs *ipvs;
549 const union nf_inet_addr *caddr;
550 const union nf_inet_addr *vaddr;
551 __be16 cport;
552 __be16 vport;
553 __u16 protocol;
554 u16 af;
555
556 const struct ip_vs_pe *pe;
557 char *pe_data;
558 __u8 pe_data_len;
559 };
560
561 /* IP_VS structure allocated for each dynamically scheduled connection */
562 struct ip_vs_conn {
563 struct hlist_node c_list; /* hashed list heads */
564 /* Protocol, addresses and port numbers */
565 __be16 cport;
566 __be16 dport;
567 __be16 vport;
568 u16 af; /* address family */
569 union nf_inet_addr caddr; /* client address */
570 union nf_inet_addr vaddr; /* virtual address */
571 union nf_inet_addr daddr; /* destination address */
572 volatile __u32 flags; /* status flags */
573 __u16 protocol; /* Which protocol (TCP/UDP) */
574 __u16 daf; /* Address family of the dest */
575 struct netns_ipvs *ipvs;
576
577 /* counter and timer */
578 atomic_t refcnt; /* reference count */
579 struct timer_list timer; /* Expiration timer */
580 volatile unsigned long timeout; /* timeout */
581
582 /* Flags and state transition */
583 spinlock_t lock; /* lock for state transition */
584 volatile __u16 state; /* state info */
585 volatile __u16 old_state; /* old state, to be used for
586 * state transition triggerd
587 * synchronization
588 */
589 __u32 fwmark; /* Fire wall mark from skb */
590 unsigned long sync_endtime; /* jiffies + sent_retries */
591
592 /* Control members */
593 struct ip_vs_conn *control; /* Master control connection */
594 atomic_t n_control; /* Number of controlled ones */
595 struct ip_vs_dest *dest; /* real server */
596 atomic_t in_pkts; /* incoming packet counter */
597
598 /* Packet transmitter for different forwarding methods. If it
599 * mangles the packet, it must return NF_DROP or better NF_STOLEN,
600 * otherwise this must be changed to a sk_buff **.
601 * NF_ACCEPT can be returned when destination is local.
602 */
603 int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp,
604 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
605
606 /* Note: we can group the following members into a structure,
607 * in order to save more space, and the following members are
608 * only used in VS/NAT anyway
609 */
610 struct ip_vs_app *app; /* bound ip_vs_app object */
611 void *app_data; /* Application private data */
612 struct ip_vs_seq in_seq; /* incoming seq. struct */
613 struct ip_vs_seq out_seq; /* outgoing seq. struct */
614
615 const struct ip_vs_pe *pe;
616 char *pe_data;
617 __u8 pe_data_len;
618
619 struct rcu_head rcu_head;
620 };
621
622 /* Extended internal versions of struct ip_vs_service_user and ip_vs_dest_user
623 * for IPv6 support.
624 *
625 * We need these to conveniently pass around service and destination
626 * options, but unfortunately, we also need to keep the old definitions to
627 * maintain userspace backwards compatibility for the setsockopt interface.
628 */
629 struct ip_vs_service_user_kern {
630 /* virtual service addresses */
631 u16 af;
632 u16 protocol;
633 union nf_inet_addr addr; /* virtual ip address */
634 __be16 port;
635 u32 fwmark; /* firwall mark of service */
636
637 /* virtual service options */
638 char *sched_name;
639 char *pe_name;
640 unsigned int flags; /* virtual service flags */
641 unsigned int timeout; /* persistent timeout in sec */
642 __be32 netmask; /* persistent netmask or plen */
643 };
644
645
646 struct ip_vs_dest_user_kern {
647 /* destination server address */
648 union nf_inet_addr addr;
649 __be16 port;
650
651 /* real server options */
652 unsigned int conn_flags; /* connection flags */
653 int weight; /* destination weight */
654
655 /* thresholds for active connections */
656 u32 u_threshold; /* upper threshold */
657 u32 l_threshold; /* lower threshold */
658
659 /* Address family of addr */
660 u16 af;
661 };
662
663
664 /*
665 * The information about the virtual service offered to the net and the
666 * forwarding entries.
667 */
668 struct ip_vs_service {
669 struct hlist_node s_list; /* for normal service table */
670 struct hlist_node f_list; /* for fwmark-based service table */
671 atomic_t refcnt; /* reference counter */
672
673 u16 af; /* address family */
674 __u16 protocol; /* which protocol (TCP/UDP) */
675 union nf_inet_addr addr; /* IP address for virtual service */
676 __be16 port; /* port number for the service */
677 __u32 fwmark; /* firewall mark of the service */
678 unsigned int flags; /* service status flags */
679 unsigned int timeout; /* persistent timeout in ticks */
680 __be32 netmask; /* grouping granularity, mask/plen */
681 struct netns_ipvs *ipvs;
682
683 struct list_head destinations; /* real server d-linked list */
684 __u32 num_dests; /* number of servers */
685 struct ip_vs_stats stats; /* statistics for the service */
686
687 /* for scheduling */
688 struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */
689 spinlock_t sched_lock; /* lock sched_data */
690 void *sched_data; /* scheduler application data */
691
692 /* alternate persistence engine */
693 struct ip_vs_pe __rcu *pe;
694
695 struct rcu_head rcu_head;
696 };
697
698 /* Information for cached dst */
699 struct ip_vs_dest_dst {
700 struct dst_entry *dst_cache; /* destination cache entry */
701 u32 dst_cookie;
702 union nf_inet_addr dst_saddr;
703 struct rcu_head rcu_head;
704 };
705
706 /* The real server destination forwarding entry with ip address, port number,
707 * and so on.
708 */
709 struct ip_vs_dest {
710 struct list_head n_list; /* for the dests in the service */
711 struct hlist_node d_list; /* for table with all the dests */
712
713 u16 af; /* address family */
714 __be16 port; /* port number of the server */
715 union nf_inet_addr addr; /* IP address of the server */
716 volatile unsigned int flags; /* dest status flags */
717 atomic_t conn_flags; /* flags to copy to conn */
718 atomic_t weight; /* server weight */
719
720 atomic_t refcnt; /* reference counter */
721 struct ip_vs_stats stats; /* statistics */
722 unsigned long idle_start; /* start time, jiffies */
723
724 /* connection counters and thresholds */
725 atomic_t activeconns; /* active connections */
726 atomic_t inactconns; /* inactive connections */
727 atomic_t persistconns; /* persistent connections */
728 __u32 u_threshold; /* upper threshold */
729 __u32 l_threshold; /* lower threshold */
730
731 /* for destination cache */
732 spinlock_t dst_lock; /* lock of dst_cache */
733 struct ip_vs_dest_dst __rcu *dest_dst; /* cached dst info */
734
735 /* for virtual service */
736 struct ip_vs_service __rcu *svc; /* service it belongs to */
737 __u16 protocol; /* which protocol (TCP/UDP) */
738 __be16 vport; /* virtual port number */
739 union nf_inet_addr vaddr; /* virtual IP address */
740 __u32 vfwmark; /* firewall mark of service */
741
742 struct list_head t_list; /* in dest_trash */
743 unsigned int in_rs_table:1; /* we are in rs_table */
744 };
745
746 /* The scheduler object */
747 struct ip_vs_scheduler {
748 struct list_head n_list; /* d-linked list head */
749 char *name; /* scheduler name */
750 atomic_t refcnt; /* reference counter */
751 struct module *module; /* THIS_MODULE/NULL */
752
753 /* scheduler initializing service */
754 int (*init_service)(struct ip_vs_service *svc);
755 /* scheduling service finish */
756 void (*done_service)(struct ip_vs_service *svc);
757 /* dest is linked */
758 int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
759 /* dest is unlinked */
760 int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
761 /* dest is updated */
762 int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
763
764 /* selecting a server from the given service */
765 struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc,
766 const struct sk_buff *skb,
767 struct ip_vs_iphdr *iph);
768 };
769
770 /* The persistence engine object */
771 struct ip_vs_pe {
772 struct list_head n_list; /* d-linked list head */
773 char *name; /* scheduler name */
774 atomic_t refcnt; /* reference counter */
775 struct module *module; /* THIS_MODULE/NULL */
776
777 /* get the connection template, if any */
778 int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb);
779 bool (*ct_match)(const struct ip_vs_conn_param *p,
780 struct ip_vs_conn *ct);
781 u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval,
782 bool inverse);
783 int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf);
784 };
785
786 /* The application module object (a.k.a. app incarnation) */
787 struct ip_vs_app {
788 struct list_head a_list; /* member in app list */
789 int type; /* IP_VS_APP_TYPE_xxx */
790 char *name; /* application module name */
791 __u16 protocol;
792 struct module *module; /* THIS_MODULE/NULL */
793 struct list_head incs_list; /* list of incarnations */
794
795 /* members for application incarnations */
796 struct list_head p_list; /* member in proto app list */
797 struct ip_vs_app *app; /* its real application */
798 __be16 port; /* port number in net order */
799 atomic_t usecnt; /* usage counter */
800 struct rcu_head rcu_head;
801
802 /* output hook: Process packet in inout direction, diff set for TCP.
803 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
804 * 2=Mangled but checksum was not updated
805 */
806 int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *,
807 struct sk_buff *, int *diff);
808
809 /* input hook: Process packet in outin direction, diff set for TCP.
810 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
811 * 2=Mangled but checksum was not updated
812 */
813 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *,
814 struct sk_buff *, int *diff);
815
816 /* ip_vs_app initializer */
817 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *);
818
819 /* ip_vs_app finish */
820 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *);
821
822
823 /* not used now */
824 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *,
825 struct ip_vs_protocol *);
826
827 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *);
828
829 int * timeout_table;
830 int * timeouts;
831 int timeouts_size;
832
833 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app,
834 int *verdict, struct ip_vs_conn **cpp);
835
836 struct ip_vs_conn *
837 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app,
838 const struct iphdr *iph, int inverse);
839
840 struct ip_vs_conn *
841 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app,
842 const struct iphdr *iph, int inverse);
843
844 int (*state_transition)(struct ip_vs_conn *cp, int direction,
845 const struct sk_buff *skb,
846 struct ip_vs_app *app);
847
848 void (*timeout_change)(struct ip_vs_app *app, int flags);
849 };
850
851 struct ipvs_master_sync_state {
852 struct list_head sync_queue;
853 struct ip_vs_sync_buff *sync_buff;
854 unsigned long sync_queue_len;
855 unsigned int sync_queue_delay;
856 struct task_struct *master_thread;
857 struct delayed_work master_wakeup_work;
858 struct netns_ipvs *ipvs;
859 };
860
861 /* How much time to keep dests in trash */
862 #define IP_VS_DEST_TRASH_PERIOD (120 * HZ)
863
864 struct ipvs_sync_daemon_cfg {
865 union nf_inet_addr mcast_group;
866 int syncid;
867 u16 sync_maxlen;
868 u16 mcast_port;
869 u8 mcast_af;
870 u8 mcast_ttl;
871 /* multicast interface name */
872 char mcast_ifn[IP_VS_IFNAME_MAXLEN];
873 };
874
875 /* IPVS in network namespace */
876 struct netns_ipvs {
877 int gen; /* Generation */
878 int enable; /* enable like nf_hooks do */
879 /* Hash table: for real service lookups */
880 #define IP_VS_RTAB_BITS 4
881 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
882 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
883
884 struct hlist_head rs_table[IP_VS_RTAB_SIZE];
885 /* ip_vs_app */
886 struct list_head app_list;
887 /* ip_vs_proto */
888 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */
889 struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE];
890 /* ip_vs_proto_tcp */
891 #ifdef CONFIG_IP_VS_PROTO_TCP
892 #define TCP_APP_TAB_BITS 4
893 #define TCP_APP_TAB_SIZE (1 << TCP_APP_TAB_BITS)
894 #define TCP_APP_TAB_MASK (TCP_APP_TAB_SIZE - 1)
895 struct list_head tcp_apps[TCP_APP_TAB_SIZE];
896 #endif
897 /* ip_vs_proto_udp */
898 #ifdef CONFIG_IP_VS_PROTO_UDP
899 #define UDP_APP_TAB_BITS 4
900 #define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS)
901 #define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1)
902 struct list_head udp_apps[UDP_APP_TAB_SIZE];
903 #endif
904 /* ip_vs_proto_sctp */
905 #ifdef CONFIG_IP_VS_PROTO_SCTP
906 #define SCTP_APP_TAB_BITS 4
907 #define SCTP_APP_TAB_SIZE (1 << SCTP_APP_TAB_BITS)
908 #define SCTP_APP_TAB_MASK (SCTP_APP_TAB_SIZE - 1)
909 /* Hash table for SCTP application incarnations */
910 struct list_head sctp_apps[SCTP_APP_TAB_SIZE];
911 #endif
912 /* ip_vs_conn */
913 atomic_t conn_count; /* connection counter */
914
915 /* ip_vs_ctl */
916 struct ip_vs_stats tot_stats; /* Statistics & est. */
917
918 int num_services; /* no of virtual services */
919
920 /* Trash for destinations */
921 struct list_head dest_trash;
922 spinlock_t dest_trash_lock;
923 struct timer_list dest_trash_timer; /* expiration timer */
924 /* Service counters */
925 atomic_t ftpsvc_counter;
926 atomic_t nullsvc_counter;
927
928 #ifdef CONFIG_SYSCTL
929 /* 1/rate drop and drop-entry variables */
930 struct delayed_work defense_work; /* Work handler */
931 int drop_rate;
932 int drop_counter;
933 atomic_t dropentry;
934 /* locks in ctl.c */
935 spinlock_t dropentry_lock; /* drop entry handling */
936 spinlock_t droppacket_lock; /* drop packet handling */
937 spinlock_t securetcp_lock; /* state and timeout tables */
938
939 /* sys-ctl struct */
940 struct ctl_table_header *sysctl_hdr;
941 struct ctl_table *sysctl_tbl;
942 #endif
943
944 /* sysctl variables */
945 int sysctl_amemthresh;
946 int sysctl_am_droprate;
947 int sysctl_drop_entry;
948 int sysctl_drop_packet;
949 int sysctl_secure_tcp;
950 #ifdef CONFIG_IP_VS_NFCT
951 int sysctl_conntrack;
952 #endif
953 int sysctl_snat_reroute;
954 int sysctl_sync_ver;
955 int sysctl_sync_ports;
956 int sysctl_sync_persist_mode;
957 unsigned long sysctl_sync_qlen_max;
958 int sysctl_sync_sock_size;
959 int sysctl_cache_bypass;
960 int sysctl_expire_nodest_conn;
961 int sysctl_sloppy_tcp;
962 int sysctl_sloppy_sctp;
963 int sysctl_expire_quiescent_template;
964 int sysctl_sync_threshold[2];
965 unsigned int sysctl_sync_refresh_period;
966 int sysctl_sync_retries;
967 int sysctl_nat_icmp_send;
968 int sysctl_pmtu_disc;
969 int sysctl_backup_only;
970 int sysctl_conn_reuse_mode;
971 int sysctl_schedule_icmp;
972 int sysctl_ignore_tunneled;
973
974 /* ip_vs_lblc */
975 int sysctl_lblc_expiration;
976 struct ctl_table_header *lblc_ctl_header;
977 struct ctl_table *lblc_ctl_table;
978 /* ip_vs_lblcr */
979 int sysctl_lblcr_expiration;
980 struct ctl_table_header *lblcr_ctl_header;
981 struct ctl_table *lblcr_ctl_table;
982 /* ip_vs_est */
983 struct list_head est_list; /* estimator list */
984 spinlock_t est_lock;
985 struct timer_list est_timer; /* Estimation timer */
986 /* ip_vs_sync */
987 spinlock_t sync_lock;
988 struct ipvs_master_sync_state *ms;
989 spinlock_t sync_buff_lock;
990 struct task_struct **backup_threads;
991 int threads_mask;
992 volatile int sync_state;
993 struct mutex sync_mutex;
994 struct ipvs_sync_daemon_cfg mcfg; /* Master Configuration */
995 struct ipvs_sync_daemon_cfg bcfg; /* Backup Configuration */
996 /* net name space ptr */
997 struct net *net; /* Needed by timer routines */
998 /* Number of heterogeneous destinations, needed becaus heterogeneous
999 * are not supported when synchronization is enabled.
1000 */
1001 unsigned int mixed_address_family_dests;
1002 };
1003
1004 #define DEFAULT_SYNC_THRESHOLD 3
1005 #define DEFAULT_SYNC_PERIOD 50
1006 #define DEFAULT_SYNC_VER 1
1007 #define DEFAULT_SLOPPY_TCP 0
1008 #define DEFAULT_SLOPPY_SCTP 0
1009 #define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ)
1010 #define DEFAULT_SYNC_RETRIES 0
1011 #define IPVS_SYNC_WAKEUP_RATE 8
1012 #define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4)
1013 #define IPVS_SYNC_SEND_DELAY (HZ / 50)
1014 #define IPVS_SYNC_CHECK_PERIOD HZ
1015 #define IPVS_SYNC_FLUSH_TIME (HZ * 2)
1016 #define IPVS_SYNC_PORTS_MAX (1 << 6)
1017
1018 #ifdef CONFIG_SYSCTL
1019
1020 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1021 {
1022 return ipvs->sysctl_sync_threshold[0];
1023 }
1024
1025 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1026 {
1027 return ACCESS_ONCE(ipvs->sysctl_sync_threshold[1]);
1028 }
1029
1030 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1031 {
1032 return ACCESS_ONCE(ipvs->sysctl_sync_refresh_period);
1033 }
1034
1035 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1036 {
1037 return ipvs->sysctl_sync_retries;
1038 }
1039
1040 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1041 {
1042 return ipvs->sysctl_sync_ver;
1043 }
1044
1045 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1046 {
1047 return ipvs->sysctl_sloppy_tcp;
1048 }
1049
1050 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1051 {
1052 return ipvs->sysctl_sloppy_sctp;
1053 }
1054
1055 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1056 {
1057 return ACCESS_ONCE(ipvs->sysctl_sync_ports);
1058 }
1059
1060 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1061 {
1062 return ipvs->sysctl_sync_persist_mode;
1063 }
1064
1065 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1066 {
1067 return ipvs->sysctl_sync_qlen_max;
1068 }
1069
1070 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1071 {
1072 return ipvs->sysctl_sync_sock_size;
1073 }
1074
1075 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1076 {
1077 return ipvs->sysctl_pmtu_disc;
1078 }
1079
1080 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1081 {
1082 return ipvs->sync_state & IP_VS_STATE_BACKUP &&
1083 ipvs->sysctl_backup_only;
1084 }
1085
1086 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1087 {
1088 return ipvs->sysctl_conn_reuse_mode;
1089 }
1090
1091 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1092 {
1093 return ipvs->sysctl_schedule_icmp;
1094 }
1095
1096 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1097 {
1098 return ipvs->sysctl_ignore_tunneled;
1099 }
1100
1101 #else
1102
1103 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1104 {
1105 return DEFAULT_SYNC_THRESHOLD;
1106 }
1107
1108 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1109 {
1110 return DEFAULT_SYNC_PERIOD;
1111 }
1112
1113 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1114 {
1115 return DEFAULT_SYNC_REFRESH_PERIOD;
1116 }
1117
1118 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1119 {
1120 return DEFAULT_SYNC_RETRIES & 3;
1121 }
1122
1123 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1124 {
1125 return DEFAULT_SYNC_VER;
1126 }
1127
1128 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1129 {
1130 return DEFAULT_SLOPPY_TCP;
1131 }
1132
1133 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1134 {
1135 return DEFAULT_SLOPPY_SCTP;
1136 }
1137
1138 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1139 {
1140 return 1;
1141 }
1142
1143 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1144 {
1145 return 0;
1146 }
1147
1148 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1149 {
1150 return IPVS_SYNC_QLEN_MAX;
1151 }
1152
1153 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1154 {
1155 return 0;
1156 }
1157
1158 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1159 {
1160 return 1;
1161 }
1162
1163 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1164 {
1165 return 0;
1166 }
1167
1168 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1169 {
1170 return 1;
1171 }
1172
1173 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1174 {
1175 return 0;
1176 }
1177
1178 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1179 {
1180 return 0;
1181 }
1182
1183 #endif
1184
1185 /* IPVS core functions
1186 * (from ip_vs_core.c)
1187 */
1188 const char *ip_vs_proto_name(unsigned int proto);
1189 void ip_vs_init_hash_table(struct list_head *table, int rows);
1190 #define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t)))
1191
1192 #define IP_VS_APP_TYPE_FTP 1
1193
1194 /* ip_vs_conn handling functions
1195 * (from ip_vs_conn.c)
1196 */
1197 enum {
1198 IP_VS_DIR_INPUT = 0,
1199 IP_VS_DIR_OUTPUT,
1200 IP_VS_DIR_INPUT_ONLY,
1201 IP_VS_DIR_LAST,
1202 };
1203
1204 static inline void ip_vs_conn_fill_param(struct netns_ipvs *ipvs, int af, int protocol,
1205 const union nf_inet_addr *caddr,
1206 __be16 cport,
1207 const union nf_inet_addr *vaddr,
1208 __be16 vport,
1209 struct ip_vs_conn_param *p)
1210 {
1211 p->ipvs = ipvs;
1212 p->af = af;
1213 p->protocol = protocol;
1214 p->caddr = caddr;
1215 p->cport = cport;
1216 p->vaddr = vaddr;
1217 p->vport = vport;
1218 p->pe = NULL;
1219 p->pe_data = NULL;
1220 }
1221
1222 struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p);
1223 struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p);
1224
1225 struct ip_vs_conn * ip_vs_conn_in_get_proto(int af, const struct sk_buff *skb,
1226 const struct ip_vs_iphdr *iph);
1227
1228 struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p);
1229
1230 struct ip_vs_conn * ip_vs_conn_out_get_proto(int af, const struct sk_buff *skb,
1231 const struct ip_vs_iphdr *iph);
1232
1233 /* Get reference to gain full access to conn.
1234 * By default, RCU read-side critical sections have access only to
1235 * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference.
1236 */
1237 static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp)
1238 {
1239 return atomic_inc_not_zero(&cp->refcnt);
1240 }
1241
1242 /* put back the conn without restarting its timer */
1243 static inline void __ip_vs_conn_put(struct ip_vs_conn *cp)
1244 {
1245 smp_mb__before_atomic();
1246 atomic_dec(&cp->refcnt);
1247 }
1248 void ip_vs_conn_put(struct ip_vs_conn *cp);
1249 void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport);
1250
1251 struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, int dest_af,
1252 const union nf_inet_addr *daddr,
1253 __be16 dport, unsigned int flags,
1254 struct ip_vs_dest *dest, __u32 fwmark);
1255 void ip_vs_conn_expire_now(struct ip_vs_conn *cp);
1256
1257 const char *ip_vs_state_name(__u16 proto, int state);
1258
1259 void ip_vs_tcp_conn_listen(struct net *net, struct ip_vs_conn *cp);
1260 int ip_vs_check_template(struct ip_vs_conn *ct);
1261 void ip_vs_random_dropentry(struct netns_ipvs *ipvs);
1262 int ip_vs_conn_init(void);
1263 void ip_vs_conn_cleanup(void);
1264
1265 static inline void ip_vs_control_del(struct ip_vs_conn *cp)
1266 {
1267 struct ip_vs_conn *ctl_cp = cp->control;
1268 if (!ctl_cp) {
1269 IP_VS_ERR_BUF("request control DEL for uncontrolled: "
1270 "%s:%d to %s:%d\n",
1271 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1272 ntohs(cp->cport),
1273 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1274 ntohs(cp->vport));
1275
1276 return;
1277 }
1278
1279 IP_VS_DBG_BUF(7, "DELeting control for: "
1280 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1281 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1282 ntohs(cp->cport),
1283 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1284 ntohs(ctl_cp->cport));
1285
1286 cp->control = NULL;
1287 if (atomic_read(&ctl_cp->n_control) == 0) {
1288 IP_VS_ERR_BUF("BUG control DEL with n=0 : "
1289 "%s:%d to %s:%d\n",
1290 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1291 ntohs(cp->cport),
1292 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1293 ntohs(cp->vport));
1294
1295 return;
1296 }
1297 atomic_dec(&ctl_cp->n_control);
1298 }
1299
1300 static inline void
1301 ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp)
1302 {
1303 if (cp->control) {
1304 IP_VS_ERR_BUF("request control ADD for already controlled: "
1305 "%s:%d to %s:%d\n",
1306 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1307 ntohs(cp->cport),
1308 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1309 ntohs(cp->vport));
1310
1311 ip_vs_control_del(cp);
1312 }
1313
1314 IP_VS_DBG_BUF(7, "ADDing control for: "
1315 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1316 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1317 ntohs(cp->cport),
1318 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1319 ntohs(ctl_cp->cport));
1320
1321 cp->control = ctl_cp;
1322 atomic_inc(&ctl_cp->n_control);
1323 }
1324
1325 /* IPVS netns init & cleanup functions */
1326 int ip_vs_estimator_net_init(struct netns_ipvs *ipvs);
1327 int ip_vs_control_net_init(struct netns_ipvs *ipvs);
1328 int ip_vs_protocol_net_init(struct net *net);
1329 int ip_vs_app_net_init(struct netns_ipvs *ipvs);
1330 int ip_vs_conn_net_init(struct net *net);
1331 int ip_vs_sync_net_init(struct netns_ipvs *ipvs);
1332 void ip_vs_conn_net_cleanup(struct net *net);
1333 void ip_vs_app_net_cleanup(struct netns_ipvs *ipvs);
1334 void ip_vs_protocol_net_cleanup(struct net *net);
1335 void ip_vs_control_net_cleanup(struct netns_ipvs *ipvs);
1336 void ip_vs_estimator_net_cleanup(struct netns_ipvs *ipvs);
1337 void ip_vs_sync_net_cleanup(struct netns_ipvs *ipvs);
1338 void ip_vs_service_net_cleanup(struct netns_ipvs *ipvs);
1339
1340 /* IPVS application functions
1341 * (from ip_vs_app.c)
1342 */
1343 #define IP_VS_APP_MAX_PORTS 8
1344 struct ip_vs_app *register_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1345 void unregister_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1346 int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1347 void ip_vs_unbind_app(struct ip_vs_conn *cp);
1348 int register_ip_vs_app_inc(struct netns_ipvs *ipvs, struct ip_vs_app *app, __u16 proto,
1349 __u16 port);
1350 int ip_vs_app_inc_get(struct ip_vs_app *inc);
1351 void ip_vs_app_inc_put(struct ip_vs_app *inc);
1352
1353 int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb);
1354 int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb);
1355
1356 int register_ip_vs_pe(struct ip_vs_pe *pe);
1357 int unregister_ip_vs_pe(struct ip_vs_pe *pe);
1358 struct ip_vs_pe *ip_vs_pe_getbyname(const char *name);
1359 struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name);
1360
1361 /* Use a #define to avoid all of module.h just for these trivial ops */
1362 #define ip_vs_pe_get(pe) \
1363 if (pe && pe->module) \
1364 __module_get(pe->module);
1365
1366 #define ip_vs_pe_put(pe) \
1367 if (pe && pe->module) \
1368 module_put(pe->module);
1369
1370 /* IPVS protocol functions (from ip_vs_proto.c) */
1371 int ip_vs_protocol_init(void);
1372 void ip_vs_protocol_cleanup(void);
1373 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags);
1374 int *ip_vs_create_timeout_table(int *table, int size);
1375 int ip_vs_set_state_timeout(int *table, int num, const char *const *names,
1376 const char *name, int to);
1377 void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
1378 const struct sk_buff *skb, int offset,
1379 const char *msg);
1380
1381 extern struct ip_vs_protocol ip_vs_protocol_tcp;
1382 extern struct ip_vs_protocol ip_vs_protocol_udp;
1383 extern struct ip_vs_protocol ip_vs_protocol_icmp;
1384 extern struct ip_vs_protocol ip_vs_protocol_esp;
1385 extern struct ip_vs_protocol ip_vs_protocol_ah;
1386 extern struct ip_vs_protocol ip_vs_protocol_sctp;
1387
1388 /* Registering/unregistering scheduler functions
1389 * (from ip_vs_sched.c)
1390 */
1391 int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1392 int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1393 int ip_vs_bind_scheduler(struct ip_vs_service *svc,
1394 struct ip_vs_scheduler *scheduler);
1395 void ip_vs_unbind_scheduler(struct ip_vs_service *svc,
1396 struct ip_vs_scheduler *sched);
1397 struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name);
1398 void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler);
1399 struct ip_vs_conn *
1400 ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb,
1401 struct ip_vs_proto_data *pd, int *ignored,
1402 struct ip_vs_iphdr *iph);
1403 int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb,
1404 struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph);
1405
1406 void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg);
1407
1408 /* IPVS control data and functions (from ip_vs_ctl.c) */
1409 extern struct ip_vs_stats ip_vs_stats;
1410 extern int sysctl_ip_vs_sync_ver;
1411
1412 struct ip_vs_service *
1413 ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
1414 const union nf_inet_addr *vaddr, __be16 vport);
1415
1416 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1417 const union nf_inet_addr *daddr, __be16 dport);
1418
1419 int ip_vs_use_count_inc(void);
1420 void ip_vs_use_count_dec(void);
1421 int ip_vs_register_nl_ioctl(void);
1422 void ip_vs_unregister_nl_ioctl(void);
1423 int ip_vs_control_init(void);
1424 void ip_vs_control_cleanup(void);
1425 struct ip_vs_dest *
1426 ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
1427 const union nf_inet_addr *daddr, __be16 dport,
1428 const union nf_inet_addr *vaddr, __be16 vport,
1429 __u16 protocol, __u32 fwmark, __u32 flags);
1430 void ip_vs_try_bind_dest(struct ip_vs_conn *cp);
1431
1432 static inline void ip_vs_dest_hold(struct ip_vs_dest *dest)
1433 {
1434 atomic_inc(&dest->refcnt);
1435 }
1436
1437 static inline void ip_vs_dest_put(struct ip_vs_dest *dest)
1438 {
1439 smp_mb__before_atomic();
1440 atomic_dec(&dest->refcnt);
1441 }
1442
1443 static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest)
1444 {
1445 if (atomic_dec_return(&dest->refcnt) < 0)
1446 kfree(dest);
1447 }
1448
1449 /* IPVS sync daemon data and function prototypes
1450 * (from ip_vs_sync.c)
1451 */
1452 int start_sync_thread(struct netns_ipvs *ipvs, struct ipvs_sync_daemon_cfg *cfg,
1453 int state);
1454 int stop_sync_thread(struct netns_ipvs *ipvs, int state);
1455 void ip_vs_sync_conn(struct netns_ipvs *ipvs, struct ip_vs_conn *cp, int pkts);
1456
1457 /* IPVS rate estimator prototypes (from ip_vs_est.c) */
1458 void ip_vs_start_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1459 void ip_vs_stop_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1460 void ip_vs_zero_estimator(struct ip_vs_stats *stats);
1461 void ip_vs_read_estimator(struct ip_vs_kstats *dst, struct ip_vs_stats *stats);
1462
1463 /* Various IPVS packet transmitters (from ip_vs_xmit.c) */
1464 int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1465 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1466 int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1467 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1468 int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1469 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1470 int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1471 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1472 int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1473 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1474 int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1475 struct ip_vs_protocol *pp, int offset,
1476 unsigned int hooknum, struct ip_vs_iphdr *iph);
1477 void ip_vs_dest_dst_rcu_free(struct rcu_head *head);
1478
1479 #ifdef CONFIG_IP_VS_IPV6
1480 int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1481 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1482 int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1483 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1484 int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1485 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1486 int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1487 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1488 int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1489 struct ip_vs_protocol *pp, int offset,
1490 unsigned int hooknum, struct ip_vs_iphdr *iph);
1491 #endif
1492
1493 #ifdef CONFIG_SYSCTL
1494 /* This is a simple mechanism to ignore packets when
1495 * we are loaded. Just set ip_vs_drop_rate to 'n' and
1496 * we start to drop 1/rate of the packets
1497 */
1498 static inline int ip_vs_todrop(struct netns_ipvs *ipvs)
1499 {
1500 if (!ipvs->drop_rate)
1501 return 0;
1502 if (--ipvs->drop_counter > 0)
1503 return 0;
1504 ipvs->drop_counter = ipvs->drop_rate;
1505 return 1;
1506 }
1507 #else
1508 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; }
1509 #endif
1510
1511 /* ip_vs_fwd_tag returns the forwarding tag of the connection */
1512 #define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK)
1513
1514 static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp)
1515 {
1516 char fwd;
1517
1518 switch (IP_VS_FWD_METHOD(cp)) {
1519 case IP_VS_CONN_F_MASQ:
1520 fwd = 'M'; break;
1521 case IP_VS_CONN_F_LOCALNODE:
1522 fwd = 'L'; break;
1523 case IP_VS_CONN_F_TUNNEL:
1524 fwd = 'T'; break;
1525 case IP_VS_CONN_F_DROUTE:
1526 fwd = 'R'; break;
1527 case IP_VS_CONN_F_BYPASS:
1528 fwd = 'B'; break;
1529 default:
1530 fwd = '?'; break;
1531 }
1532 return fwd;
1533 }
1534
1535 void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp,
1536 struct ip_vs_conn *cp, int dir);
1537
1538 #ifdef CONFIG_IP_VS_IPV6
1539 void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp,
1540 struct ip_vs_conn *cp, int dir);
1541 #endif
1542
1543 __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset);
1544
1545 static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum)
1546 {
1547 __be32 diff[2] = { ~old, new };
1548
1549 return csum_partial(diff, sizeof(diff), oldsum);
1550 }
1551
1552 #ifdef CONFIG_IP_VS_IPV6
1553 static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new,
1554 __wsum oldsum)
1555 {
1556 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0],
1557 new[3], new[2], new[1], new[0] };
1558
1559 return csum_partial(diff, sizeof(diff), oldsum);
1560 }
1561 #endif
1562
1563 static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum)
1564 {
1565 __be16 diff[2] = { ~old, new };
1566
1567 return csum_partial(diff, sizeof(diff), oldsum);
1568 }
1569
1570 /* Forget current conntrack (unconfirmed) and attach notrack entry */
1571 static inline void ip_vs_notrack(struct sk_buff *skb)
1572 {
1573 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1574 enum ip_conntrack_info ctinfo;
1575 struct nf_conn *ct = nf_ct_get(skb, &ctinfo);
1576
1577 if (!ct || !nf_ct_is_untracked(ct)) {
1578 nf_conntrack_put(skb->nfct);
1579 skb->nfct = &nf_ct_untracked_get()->ct_general;
1580 skb->nfctinfo = IP_CT_NEW;
1581 nf_conntrack_get(skb->nfct);
1582 }
1583 #endif
1584 }
1585
1586 #ifdef CONFIG_IP_VS_NFCT
1587 /* Netfilter connection tracking
1588 * (from ip_vs_nfct.c)
1589 */
1590 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1591 {
1592 #ifdef CONFIG_SYSCTL
1593 return ipvs->sysctl_conntrack;
1594 #else
1595 return 0;
1596 #endif
1597 }
1598
1599 void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp,
1600 int outin);
1601 int ip_vs_confirm_conntrack(struct sk_buff *skb);
1602 void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct,
1603 struct ip_vs_conn *cp, u_int8_t proto,
1604 const __be16 port, int from_rs);
1605 void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp);
1606
1607 #else
1608
1609 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1610 {
1611 return 0;
1612 }
1613
1614 static inline void ip_vs_update_conntrack(struct sk_buff *skb,
1615 struct ip_vs_conn *cp, int outin)
1616 {
1617 }
1618
1619 static inline int ip_vs_confirm_conntrack(struct sk_buff *skb)
1620 {
1621 return NF_ACCEPT;
1622 }
1623
1624 static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp)
1625 {
1626 }
1627 #endif /* CONFIG_IP_VS_NFCT */
1628
1629 static inline int
1630 ip_vs_dest_conn_overhead(struct ip_vs_dest *dest)
1631 {
1632 /* We think the overhead of processing active connections is 256
1633 * times higher than that of inactive connections in average. (This
1634 * 256 times might not be accurate, we will change it later) We
1635 * use the following formula to estimate the overhead now:
1636 * dest->activeconns*256 + dest->inactconns
1637 */
1638 return (atomic_read(&dest->activeconns) << 8) +
1639 atomic_read(&dest->inactconns);
1640 }
1641
1642 #endif /* _NET_IP_VS_H */
Linux kernel - Deliverables
This page took 0.062227 seconds and 4 git commands to generate.