projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
ipvs: replace ip_vs_fill_ip4hdr with ip_vs_fill_iph_skb_off
[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 /* Generic access of ipvs struct */
33 static inline struct netns_ipvs *net_ipvs(struct net* net)
34 {
35 return net->ipvs;
36 }
37
38 /* Get net ptr from skb in traffic cases
39 * use skb_sknet when call is from userland (ioctl or netlink)
40 */
41 static inline struct net *skb_net(const struct sk_buff *skb)
42 {
43 #ifdef CONFIG_NET_NS
44 #ifdef CONFIG_IP_VS_DEBUG
45 /*
46 * This is used for debug only.
47 * Start with the most likely hit
48 * End with BUG
49 */
50 if (likely(skb->dev && dev_net(skb->dev)))
51 return dev_net(skb->dev);
52 if (skb_dst(skb) && skb_dst(skb)->dev)
53 return dev_net(skb_dst(skb)->dev);
54 WARN(skb->sk, "Maybe skb_sknet should be used in %s() at line:%d\n",
55 __func__, __LINE__);
56 if (likely(skb->sk && sock_net(skb->sk)))
57 return sock_net(skb->sk);
58 pr_err("There is no net ptr to find in the skb in %s() line:%d\n",
59 __func__, __LINE__);
60 BUG();
61 #else
62 return dev_net(skb->dev ? : skb_dst(skb)->dev);
63 #endif
64 #else
65 return &init_net;
66 #endif
67 }
68
69 static inline struct net *skb_sknet(const struct sk_buff *skb)
70 {
71 #ifdef CONFIG_NET_NS
72 #ifdef CONFIG_IP_VS_DEBUG
73 /* Start with the most likely hit */
74 if (likely(skb->sk && sock_net(skb->sk)))
75 return sock_net(skb->sk);
76 WARN(skb->dev, "Maybe skb_net should be used instead in %s() line:%d\n",
77 __func__, __LINE__);
78 if (likely(skb->dev && dev_net(skb->dev)))
79 return dev_net(skb->dev);
80 pr_err("There is no net ptr to find in the skb in %s() line:%d\n",
81 __func__, __LINE__);
82 BUG();
83 #else
84 return sock_net(skb->sk);
85 #endif
86 #else
87 return &init_net;
88 #endif
89 }
90
91 /* This one needed for single_open_net since net is stored directly in
92 * private not as a struct i.e. seq_file_net can't be used.
93 */
94 static inline struct net *seq_file_single_net(struct seq_file *seq)
95 {
96 #ifdef CONFIG_NET_NS
97 return (struct net *)seq->private;
98 #else
99 return &init_net;
100 #endif
101 }
102
103 /* Connections' size value needed by ip_vs_ctl.c */
104 extern int ip_vs_conn_tab_size;
105
106 struct ip_vs_iphdr {
107 __u32 off; /* Where IP or IPv4 header starts */
108 __u32 len; /* IPv4 simply where L4 starts
109 * IPv6 where L4 Transport Header starts */
110 __u16 fragoffs; /* IPv6 fragment offset, 0 if first frag (or not frag)*/
111 __s16 protocol;
112 __s32 flags;
113 union nf_inet_addr saddr;
114 union nf_inet_addr daddr;
115 };
116
117 static inline void *frag_safe_skb_hp(const struct sk_buff *skb, int offset,
118 int len, void *buffer,
119 const struct ip_vs_iphdr *ipvsh)
120 {
121 return skb_header_pointer(skb, offset, len, buffer);
122 }
123
124 /* This function handles filling *ip_vs_iphdr, both for IPv4 and IPv6.
125 * IPv6 requires some extra work, as finding proper header position,
126 * depend on the IPv6 extension headers.
127 */
128 static inline int
129 ip_vs_fill_iph_skb_off(int af, const struct sk_buff *skb, int offset,
130 struct ip_vs_iphdr *iphdr)
131 {
132 iphdr->off = offset;
133 #ifdef CONFIG_IP_VS_IPV6
134 if (af == AF_INET6) {
135 struct ipv6hdr _iph;
136 const struct ipv6hdr *iph = skb_header_pointer(
137 skb, offset, sizeof(_iph), &_iph);
138 if (!iph)
139 return 0;
140
141 iphdr->saddr.in6 = iph->saddr;
142 iphdr->daddr.in6 = iph->daddr;
143 /* ipv6_find_hdr() updates len, flags */
144 iphdr->len = offset;
145 iphdr->flags = 0;
146 iphdr->protocol = ipv6_find_hdr(skb, &iphdr->len, -1,
147 &iphdr->fragoffs,
148 &iphdr->flags);
149 if (iphdr->protocol < 0)
150 return 0;
151 } else
152 #endif
153 {
154 struct iphdr _iph;
155 const struct iphdr *iph = skb_header_pointer(
156 skb, offset, sizeof(_iph), &_iph);
157 if (!iph)
158 return 0;
159
160 iphdr->len = offset + iph->ihl * 4;
161 iphdr->fragoffs = 0;
162 iphdr->protocol = iph->protocol;
163 iphdr->saddr.ip = iph->saddr;
164 iphdr->daddr.ip = iph->daddr;
165 }
166
167 return 1;
168 }
169
170 static inline int
171 ip_vs_fill_iph_skb(int af, const struct sk_buff *skb, struct ip_vs_iphdr *iphdr)
172 {
173 return ip_vs_fill_iph_skb_off(af, skb, skb_network_offset(skb), iphdr);
174 }
175
176 static inline void ip_vs_addr_copy(int af, union nf_inet_addr *dst,
177 const union nf_inet_addr *src)
178 {
179 #ifdef CONFIG_IP_VS_IPV6
180 if (af == AF_INET6)
181 dst->in6 = src->in6;
182 else
183 #endif
184 dst->ip = src->ip;
185 }
186
187 static inline void ip_vs_addr_set(int af, union nf_inet_addr *dst,
188 const union nf_inet_addr *src)
189 {
190 #ifdef CONFIG_IP_VS_IPV6
191 if (af == AF_INET6) {
192 dst->in6 = src->in6;
193 return;
194 }
195 #endif
196 dst->ip = src->ip;
197 dst->all[1] = 0;
198 dst->all[2] = 0;
199 dst->all[3] = 0;
200 }
201
202 static inline int ip_vs_addr_equal(int af, const union nf_inet_addr *a,
203 const union nf_inet_addr *b)
204 {
205 #ifdef CONFIG_IP_VS_IPV6
206 if (af == AF_INET6)
207 return ipv6_addr_equal(&a->in6, &b->in6);
208 #endif
209 return a->ip == b->ip;
210 }
211
212 #ifdef CONFIG_IP_VS_DEBUG
213 #include <linux/net.h>
214
215 int ip_vs_get_debug_level(void);
216
217 static inline const char *ip_vs_dbg_addr(int af, char *buf, size_t buf_len,
218 const union nf_inet_addr *addr,
219 int *idx)
220 {
221 int len;
222 #ifdef CONFIG_IP_VS_IPV6
223 if (af == AF_INET6)
224 len = snprintf(&buf[*idx], buf_len - *idx, "[%pI6c]",
225 &addr->in6) + 1;
226 else
227 #endif
228 len = snprintf(&buf[*idx], buf_len - *idx, "%pI4",
229 &addr->ip) + 1;
230
231 *idx += len;
232 BUG_ON(*idx > buf_len + 1);
233 return &buf[*idx - len];
234 }
235
236 #define IP_VS_DBG_BUF(level, msg, ...) \
237 do { \
238 char ip_vs_dbg_buf[160]; \
239 int ip_vs_dbg_idx = 0; \
240 if (level <= ip_vs_get_debug_level()) \
241 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
242 } while (0)
243 #define IP_VS_ERR_BUF(msg...) \
244 do { \
245 char ip_vs_dbg_buf[160]; \
246 int ip_vs_dbg_idx = 0; \
247 pr_err(msg); \
248 } while (0)
249
250 /* Only use from within IP_VS_DBG_BUF() or IP_VS_ERR_BUF macros */
251 #define IP_VS_DBG_ADDR(af, addr) \
252 ip_vs_dbg_addr(af, ip_vs_dbg_buf, \
253 sizeof(ip_vs_dbg_buf), addr, \
254 &ip_vs_dbg_idx)
255
256 #define IP_VS_DBG(level, msg, ...) \
257 do { \
258 if (level <= ip_vs_get_debug_level()) \
259 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
260 } while (0)
261 #define IP_VS_DBG_RL(msg, ...) \
262 do { \
263 if (net_ratelimit()) \
264 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
265 } while (0)
266 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) \
267 do { \
268 if (level <= ip_vs_get_debug_level()) \
269 pp->debug_packet(af, pp, skb, ofs, msg); \
270 } while (0)
271 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) \
272 do { \
273 if (level <= ip_vs_get_debug_level() && \
274 net_ratelimit()) \
275 pp->debug_packet(af, pp, skb, ofs, msg); \
276 } while (0)
277 #else /* NO DEBUGGING at ALL */
278 #define IP_VS_DBG_BUF(level, msg...) do {} while (0)
279 #define IP_VS_ERR_BUF(msg...) do {} while (0)
280 #define IP_VS_DBG(level, msg...) do {} while (0)
281 #define IP_VS_DBG_RL(msg...) do {} while (0)
282 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
283 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
284 #endif
285
286 #define IP_VS_BUG() BUG()
287 #define IP_VS_ERR_RL(msg, ...) \
288 do { \
289 if (net_ratelimit()) \
290 pr_err(msg, ##__VA_ARGS__); \
291 } while (0)
292
293 #ifdef CONFIG_IP_VS_DEBUG
294 #define EnterFunction(level) \
295 do { \
296 if (level <= ip_vs_get_debug_level()) \
297 printk(KERN_DEBUG \
298 pr_fmt("Enter: %s, %s line %i\n"), \
299 __func__, __FILE__, __LINE__); \
300 } while (0)
301 #define LeaveFunction(level) \
302 do { \
303 if (level <= ip_vs_get_debug_level()) \
304 printk(KERN_DEBUG \
305 pr_fmt("Leave: %s, %s line %i\n"), \
306 __func__, __FILE__, __LINE__); \
307 } while (0)
308 #else
309 #define EnterFunction(level) do {} while (0)
310 #define LeaveFunction(level) do {} while (0)
311 #endif
312
313 /* The port number of FTP service (in network order). */
314 #define FTPPORT cpu_to_be16(21)
315 #define FTPDATA cpu_to_be16(20)
316
317 /* TCP State Values */
318 enum {
319 IP_VS_TCP_S_NONE = 0,
320 IP_VS_TCP_S_ESTABLISHED,
321 IP_VS_TCP_S_SYN_SENT,
322 IP_VS_TCP_S_SYN_RECV,
323 IP_VS_TCP_S_FIN_WAIT,
324 IP_VS_TCP_S_TIME_WAIT,
325 IP_VS_TCP_S_CLOSE,
326 IP_VS_TCP_S_CLOSE_WAIT,
327 IP_VS_TCP_S_LAST_ACK,
328 IP_VS_TCP_S_LISTEN,
329 IP_VS_TCP_S_SYNACK,
330 IP_VS_TCP_S_LAST
331 };
332
333 /* UDP State Values */
334 enum {
335 IP_VS_UDP_S_NORMAL,
336 IP_VS_UDP_S_LAST,
337 };
338
339 /* ICMP State Values */
340 enum {
341 IP_VS_ICMP_S_NORMAL,
342 IP_VS_ICMP_S_LAST,
343 };
344
345 /* SCTP State Values */
346 enum ip_vs_sctp_states {
347 IP_VS_SCTP_S_NONE,
348 IP_VS_SCTP_S_INIT1,
349 IP_VS_SCTP_S_INIT,
350 IP_VS_SCTP_S_COOKIE_SENT,
351 IP_VS_SCTP_S_COOKIE_REPLIED,
352 IP_VS_SCTP_S_COOKIE_WAIT,
353 IP_VS_SCTP_S_COOKIE,
354 IP_VS_SCTP_S_COOKIE_ECHOED,
355 IP_VS_SCTP_S_ESTABLISHED,
356 IP_VS_SCTP_S_SHUTDOWN_SENT,
357 IP_VS_SCTP_S_SHUTDOWN_RECEIVED,
358 IP_VS_SCTP_S_SHUTDOWN_ACK_SENT,
359 IP_VS_SCTP_S_REJECTED,
360 IP_VS_SCTP_S_CLOSED,
361 IP_VS_SCTP_S_LAST
362 };
363
364 /* Delta sequence info structure
365 * Each ip_vs_conn has 2 (output AND input seq. changes).
366 * Only used in the VS/NAT.
367 */
368 struct ip_vs_seq {
369 __u32 init_seq; /* Add delta from this seq */
370 __u32 delta; /* Delta in sequence numbers */
371 __u32 previous_delta; /* Delta in sequence numbers
372 * before last resized pkt */
373 };
374
375 /* counters per cpu */
376 struct ip_vs_counters {
377 __u64 conns; /* connections scheduled */
378 __u64 inpkts; /* incoming packets */
379 __u64 outpkts; /* outgoing packets */
380 __u64 inbytes; /* incoming bytes */
381 __u64 outbytes; /* outgoing bytes */
382 };
383 /* Stats per cpu */
384 struct ip_vs_cpu_stats {
385 struct ip_vs_counters cnt;
386 struct u64_stats_sync syncp;
387 };
388
389 /* IPVS statistics objects */
390 struct ip_vs_estimator {
391 struct list_head list;
392
393 u64 last_inbytes;
394 u64 last_outbytes;
395 u64 last_conns;
396 u64 last_inpkts;
397 u64 last_outpkts;
398
399 u64 cps;
400 u64 inpps;
401 u64 outpps;
402 u64 inbps;
403 u64 outbps;
404 };
405
406 /*
407 * IPVS statistics object, 64-bit kernel version of struct ip_vs_stats_user
408 */
409 struct ip_vs_kstats {
410 u64 conns; /* connections scheduled */
411 u64 inpkts; /* incoming packets */
412 u64 outpkts; /* outgoing packets */
413 u64 inbytes; /* incoming bytes */
414 u64 outbytes; /* outgoing bytes */
415
416 u64 cps; /* current connection rate */
417 u64 inpps; /* current in packet rate */
418 u64 outpps; /* current out packet rate */
419 u64 inbps; /* current in byte rate */
420 u64 outbps; /* current out byte rate */
421 };
422
423 struct ip_vs_stats {
424 struct ip_vs_kstats kstats; /* kernel statistics */
425 struct ip_vs_estimator est; /* estimator */
426 struct ip_vs_cpu_stats __percpu *cpustats; /* per cpu counters */
427 spinlock_t lock; /* spin lock */
428 struct ip_vs_kstats kstats0; /* reset values */
429 };
430
431 struct dst_entry;
432 struct iphdr;
433 struct ip_vs_conn;
434 struct ip_vs_app;
435 struct sk_buff;
436 struct ip_vs_proto_data;
437
438 struct ip_vs_protocol {
439 struct ip_vs_protocol *next;
440 char *name;
441 u16 protocol;
442 u16 num_states;
443 int dont_defrag;
444
445 void (*init)(struct ip_vs_protocol *pp);
446
447 void (*exit)(struct ip_vs_protocol *pp);
448
449 int (*init_netns)(struct net *net, struct ip_vs_proto_data *pd);
450
451 void (*exit_netns)(struct net *net, struct ip_vs_proto_data *pd);
452
453 int (*conn_schedule)(int af, struct sk_buff *skb,
454 struct ip_vs_proto_data *pd,
455 int *verdict, struct ip_vs_conn **cpp,
456 struct ip_vs_iphdr *iph);
457
458 struct ip_vs_conn *
459 (*conn_in_get)(int af,
460 const struct sk_buff *skb,
461 const struct ip_vs_iphdr *iph,
462 int inverse);
463
464 struct ip_vs_conn *
465 (*conn_out_get)(int af,
466 const struct sk_buff *skb,
467 const struct ip_vs_iphdr *iph,
468 int inverse);
469
470 int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
471 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
472
473 int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
474 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
475
476 int (*csum_check)(int af, struct sk_buff *skb,
477 struct ip_vs_protocol *pp);
478
479 const char *(*state_name)(int state);
480
481 void (*state_transition)(struct ip_vs_conn *cp, int direction,
482 const struct sk_buff *skb,
483 struct ip_vs_proto_data *pd);
484
485 int (*register_app)(struct net *net, struct ip_vs_app *inc);
486
487 void (*unregister_app)(struct net *net, struct ip_vs_app *inc);
488
489 int (*app_conn_bind)(struct ip_vs_conn *cp);
490
491 void (*debug_packet)(int af, struct ip_vs_protocol *pp,
492 const struct sk_buff *skb,
493 int offset,
494 const char *msg);
495
496 void (*timeout_change)(struct ip_vs_proto_data *pd, int flags);
497 };
498
499 /* protocol data per netns */
500 struct ip_vs_proto_data {
501 struct ip_vs_proto_data *next;
502 struct ip_vs_protocol *pp;
503 int *timeout_table; /* protocol timeout table */
504 atomic_t appcnt; /* counter of proto app incs. */
505 struct tcp_states_t *tcp_state_table;
506 };
507
508 struct ip_vs_protocol *ip_vs_proto_get(unsigned short proto);
509 struct ip_vs_proto_data *ip_vs_proto_data_get(struct net *net,
510 unsigned short proto);
511
512 struct ip_vs_conn_param {
513 struct net *net;
514 const union nf_inet_addr *caddr;
515 const union nf_inet_addr *vaddr;
516 __be16 cport;
517 __be16 vport;
518 __u16 protocol;
519 u16 af;
520
521 const struct ip_vs_pe *pe;
522 char *pe_data;
523 __u8 pe_data_len;
524 };
525
526 /* IP_VS structure allocated for each dynamically scheduled connection */
527 struct ip_vs_conn {
528 struct hlist_node c_list; /* hashed list heads */
529 /* Protocol, addresses and port numbers */
530 __be16 cport;
531 __be16 dport;
532 __be16 vport;
533 u16 af; /* address family */
534 union nf_inet_addr caddr; /* client address */
535 union nf_inet_addr vaddr; /* virtual address */
536 union nf_inet_addr daddr; /* destination address */
537 volatile __u32 flags; /* status flags */
538 __u16 protocol; /* Which protocol (TCP/UDP) */
539 __u16 daf; /* Address family of the dest */
540 #ifdef CONFIG_NET_NS
541 struct net *net; /* Name space */
542 #endif
543
544 /* counter and timer */
545 atomic_t refcnt; /* reference count */
546 struct timer_list timer; /* Expiration timer */
547 volatile unsigned long timeout; /* timeout */
548
549 /* Flags and state transition */
550 spinlock_t lock; /* lock for state transition */
551 volatile __u16 state; /* state info */
552 volatile __u16 old_state; /* old state, to be used for
553 * state transition triggerd
554 * synchronization
555 */
556 __u32 fwmark; /* Fire wall mark from skb */
557 unsigned long sync_endtime; /* jiffies + sent_retries */
558
559 /* Control members */
560 struct ip_vs_conn *control; /* Master control connection */
561 atomic_t n_control; /* Number of controlled ones */
562 struct ip_vs_dest *dest; /* real server */
563 atomic_t in_pkts; /* incoming packet counter */
564
565 /* Packet transmitter for different forwarding methods. If it
566 * mangles the packet, it must return NF_DROP or better NF_STOLEN,
567 * otherwise this must be changed to a sk_buff **.
568 * NF_ACCEPT can be returned when destination is local.
569 */
570 int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp,
571 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
572
573 /* Note: we can group the following members into a structure,
574 * in order to save more space, and the following members are
575 * only used in VS/NAT anyway
576 */
577 struct ip_vs_app *app; /* bound ip_vs_app object */
578 void *app_data; /* Application private data */
579 struct ip_vs_seq in_seq; /* incoming seq. struct */
580 struct ip_vs_seq out_seq; /* outgoing seq. struct */
581
582 const struct ip_vs_pe *pe;
583 char *pe_data;
584 __u8 pe_data_len;
585
586 struct rcu_head rcu_head;
587 };
588
589 /* To save some memory in conn table when name space is disabled. */
590 static inline struct net *ip_vs_conn_net(const struct ip_vs_conn *cp)
591 {
592 #ifdef CONFIG_NET_NS
593 return cp->net;
594 #else
595 return &init_net;
596 #endif
597 }
598
599 static inline void ip_vs_conn_net_set(struct ip_vs_conn *cp, struct net *net)
600 {
601 #ifdef CONFIG_NET_NS
602 cp->net = net;
603 #endif
604 }
605
606 static inline int ip_vs_conn_net_eq(const struct ip_vs_conn *cp,
607 struct net *net)
608 {
609 #ifdef CONFIG_NET_NS
610 return cp->net == net;
611 #else
612 return 1;
613 #endif
614 }
615
616 /* Extended internal versions of struct ip_vs_service_user and ip_vs_dest_user
617 * for IPv6 support.
618 *
619 * We need these to conveniently pass around service and destination
620 * options, but unfortunately, we also need to keep the old definitions to
621 * maintain userspace backwards compatibility for the setsockopt interface.
622 */
623 struct ip_vs_service_user_kern {
624 /* virtual service addresses */
625 u16 af;
626 u16 protocol;
627 union nf_inet_addr addr; /* virtual ip address */
628 __be16 port;
629 u32 fwmark; /* firwall mark of service */
630
631 /* virtual service options */
632 char *sched_name;
633 char *pe_name;
634 unsigned int flags; /* virtual service flags */
635 unsigned int timeout; /* persistent timeout in sec */
636 __be32 netmask; /* persistent netmask or plen */
637 };
638
639
640 struct ip_vs_dest_user_kern {
641 /* destination server address */
642 union nf_inet_addr addr;
643 __be16 port;
644
645 /* real server options */
646 unsigned int conn_flags; /* connection flags */
647 int weight; /* destination weight */
648
649 /* thresholds for active connections */
650 u32 u_threshold; /* upper threshold */
651 u32 l_threshold; /* lower threshold */
652
653 /* Address family of addr */
654 u16 af;
655 };
656
657
658 /*
659 * The information about the virtual service offered to the net and the
660 * forwarding entries.
661 */
662 struct ip_vs_service {
663 struct hlist_node s_list; /* for normal service table */
664 struct hlist_node f_list; /* for fwmark-based service table */
665 atomic_t refcnt; /* reference counter */
666
667 u16 af; /* address family */
668 __u16 protocol; /* which protocol (TCP/UDP) */
669 union nf_inet_addr addr; /* IP address for virtual service */
670 __be16 port; /* port number for the service */
671 __u32 fwmark; /* firewall mark of the service */
672 unsigned int flags; /* service status flags */
673 unsigned int timeout; /* persistent timeout in ticks */
674 __be32 netmask; /* grouping granularity, mask/plen */
675 struct net *net;
676
677 struct list_head destinations; /* real server d-linked list */
678 __u32 num_dests; /* number of servers */
679 struct ip_vs_stats stats; /* statistics for the service */
680
681 /* for scheduling */
682 struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */
683 spinlock_t sched_lock; /* lock sched_data */
684 void *sched_data; /* scheduler application data */
685
686 /* alternate persistence engine */
687 struct ip_vs_pe __rcu *pe;
688
689 struct rcu_head rcu_head;
690 };
691
692 /* Information for cached dst */
693 struct ip_vs_dest_dst {
694 struct dst_entry *dst_cache; /* destination cache entry */
695 u32 dst_cookie;
696 union nf_inet_addr dst_saddr;
697 struct rcu_head rcu_head;
698 };
699
700 /* The real server destination forwarding entry with ip address, port number,
701 * and so on.
702 */
703 struct ip_vs_dest {
704 struct list_head n_list; /* for the dests in the service */
705 struct hlist_node d_list; /* for table with all the dests */
706
707 u16 af; /* address family */
708 __be16 port; /* port number of the server */
709 union nf_inet_addr addr; /* IP address of the server */
710 volatile unsigned int flags; /* dest status flags */
711 atomic_t conn_flags; /* flags to copy to conn */
712 atomic_t weight; /* server weight */
713
714 atomic_t refcnt; /* reference counter */
715 struct ip_vs_stats stats; /* statistics */
716 unsigned long idle_start; /* start time, jiffies */
717
718 /* connection counters and thresholds */
719 atomic_t activeconns; /* active connections */
720 atomic_t inactconns; /* inactive connections */
721 atomic_t persistconns; /* persistent connections */
722 __u32 u_threshold; /* upper threshold */
723 __u32 l_threshold; /* lower threshold */
724
725 /* for destination cache */
726 spinlock_t dst_lock; /* lock of dst_cache */
727 struct ip_vs_dest_dst __rcu *dest_dst; /* cached dst info */
728
729 /* for virtual service */
730 struct ip_vs_service __rcu *svc; /* service it belongs to */
731 __u16 protocol; /* which protocol (TCP/UDP) */
732 __be16 vport; /* virtual port number */
733 union nf_inet_addr vaddr; /* virtual IP address */
734 __u32 vfwmark; /* firewall mark of service */
735
736 struct list_head t_list; /* in dest_trash */
737 unsigned int in_rs_table:1; /* we are in rs_table */
738 };
739
740 /* The scheduler object */
741 struct ip_vs_scheduler {
742 struct list_head n_list; /* d-linked list head */
743 char *name; /* scheduler name */
744 atomic_t refcnt; /* reference counter */
745 struct module *module; /* THIS_MODULE/NULL */
746
747 /* scheduler initializing service */
748 int (*init_service)(struct ip_vs_service *svc);
749 /* scheduling service finish */
750 void (*done_service)(struct ip_vs_service *svc);
751 /* dest is linked */
752 int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
753 /* dest is unlinked */
754 int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
755 /* dest is updated */
756 int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
757
758 /* selecting a server from the given service */
759 struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc,
760 const struct sk_buff *skb,
761 struct ip_vs_iphdr *iph);
762 };
763
764 /* The persistence engine object */
765 struct ip_vs_pe {
766 struct list_head n_list; /* d-linked list head */
767 char *name; /* scheduler name */
768 atomic_t refcnt; /* reference counter */
769 struct module *module; /* THIS_MODULE/NULL */
770
771 /* get the connection template, if any */
772 int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb);
773 bool (*ct_match)(const struct ip_vs_conn_param *p,
774 struct ip_vs_conn *ct);
775 u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval,
776 bool inverse);
777 int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf);
778 };
779
780 /* The application module object (a.k.a. app incarnation) */
781 struct ip_vs_app {
782 struct list_head a_list; /* member in app list */
783 int type; /* IP_VS_APP_TYPE_xxx */
784 char *name; /* application module name */
785 __u16 protocol;
786 struct module *module; /* THIS_MODULE/NULL */
787 struct list_head incs_list; /* list of incarnations */
788
789 /* members for application incarnations */
790 struct list_head p_list; /* member in proto app list */
791 struct ip_vs_app *app; /* its real application */
792 __be16 port; /* port number in net order */
793 atomic_t usecnt; /* usage counter */
794 struct rcu_head rcu_head;
795
796 /* output hook: Process packet in inout direction, diff set for TCP.
797 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
798 * 2=Mangled but checksum was not updated
799 */
800 int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *,
801 struct sk_buff *, int *diff);
802
803 /* input hook: Process packet in outin direction, diff set for TCP.
804 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
805 * 2=Mangled but checksum was not updated
806 */
807 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *,
808 struct sk_buff *, int *diff);
809
810 /* ip_vs_app initializer */
811 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *);
812
813 /* ip_vs_app finish */
814 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *);
815
816
817 /* not used now */
818 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *,
819 struct ip_vs_protocol *);
820
821 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *);
822
823 int * timeout_table;
824 int * timeouts;
825 int timeouts_size;
826
827 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app,
828 int *verdict, struct ip_vs_conn **cpp);
829
830 struct ip_vs_conn *
831 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app,
832 const struct iphdr *iph, int inverse);
833
834 struct ip_vs_conn *
835 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app,
836 const struct iphdr *iph, int inverse);
837
838 int (*state_transition)(struct ip_vs_conn *cp, int direction,
839 const struct sk_buff *skb,
840 struct ip_vs_app *app);
841
842 void (*timeout_change)(struct ip_vs_app *app, int flags);
843 };
844
845 struct ipvs_master_sync_state {
846 struct list_head sync_queue;
847 struct ip_vs_sync_buff *sync_buff;
848 unsigned long sync_queue_len;
849 unsigned int sync_queue_delay;
850 struct task_struct *master_thread;
851 struct delayed_work master_wakeup_work;
852 struct netns_ipvs *ipvs;
853 };
854
855 /* How much time to keep dests in trash */
856 #define IP_VS_DEST_TRASH_PERIOD (120 * HZ)
857
858 struct ipvs_sync_daemon_cfg {
859 union nf_inet_addr mcast_group;
860 int syncid;
861 u16 sync_maxlen;
862 u16 mcast_port;
863 u8 mcast_af;
864 u8 mcast_ttl;
865 /* multicast interface name */
866 char mcast_ifn[IP_VS_IFNAME_MAXLEN];
867 };
868
869 /* IPVS in network namespace */
870 struct netns_ipvs {
871 int gen; /* Generation */
872 int enable; /* enable like nf_hooks do */
873 /* Hash table: for real service lookups */
874 #define IP_VS_RTAB_BITS 4
875 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
876 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
877
878 struct hlist_head rs_table[IP_VS_RTAB_SIZE];
879 /* ip_vs_app */
880 struct list_head app_list;
881 /* ip_vs_proto */
882 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */
883 struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE];
884 /* ip_vs_proto_tcp */
885 #ifdef CONFIG_IP_VS_PROTO_TCP
886 #define TCP_APP_TAB_BITS 4
887 #define TCP_APP_TAB_SIZE (1 << TCP_APP_TAB_BITS)
888 #define TCP_APP_TAB_MASK (TCP_APP_TAB_SIZE - 1)
889 struct list_head tcp_apps[TCP_APP_TAB_SIZE];
890 #endif
891 /* ip_vs_proto_udp */
892 #ifdef CONFIG_IP_VS_PROTO_UDP
893 #define UDP_APP_TAB_BITS 4
894 #define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS)
895 #define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1)
896 struct list_head udp_apps[UDP_APP_TAB_SIZE];
897 #endif
898 /* ip_vs_proto_sctp */
899 #ifdef CONFIG_IP_VS_PROTO_SCTP
900 #define SCTP_APP_TAB_BITS 4
901 #define SCTP_APP_TAB_SIZE (1 << SCTP_APP_TAB_BITS)
902 #define SCTP_APP_TAB_MASK (SCTP_APP_TAB_SIZE - 1)
903 /* Hash table for SCTP application incarnations */
904 struct list_head sctp_apps[SCTP_APP_TAB_SIZE];
905 #endif
906 /* ip_vs_conn */
907 atomic_t conn_count; /* connection counter */
908
909 /* ip_vs_ctl */
910 struct ip_vs_stats tot_stats; /* Statistics & est. */
911
912 int num_services; /* no of virtual services */
913
914 /* Trash for destinations */
915 struct list_head dest_trash;
916 spinlock_t dest_trash_lock;
917 struct timer_list dest_trash_timer; /* expiration timer */
918 /* Service counters */
919 atomic_t ftpsvc_counter;
920 atomic_t nullsvc_counter;
921
922 #ifdef CONFIG_SYSCTL
923 /* 1/rate drop and drop-entry variables */
924 struct delayed_work defense_work; /* Work handler */
925 int drop_rate;
926 int drop_counter;
927 atomic_t dropentry;
928 /* locks in ctl.c */
929 spinlock_t dropentry_lock; /* drop entry handling */
930 spinlock_t droppacket_lock; /* drop packet handling */
931 spinlock_t securetcp_lock; /* state and timeout tables */
932
933 /* sys-ctl struct */
934 struct ctl_table_header *sysctl_hdr;
935 struct ctl_table *sysctl_tbl;
936 #endif
937
938 /* sysctl variables */
939 int sysctl_amemthresh;
940 int sysctl_am_droprate;
941 int sysctl_drop_entry;
942 int sysctl_drop_packet;
943 int sysctl_secure_tcp;
944 #ifdef CONFIG_IP_VS_NFCT
945 int sysctl_conntrack;
946 #endif
947 int sysctl_snat_reroute;
948 int sysctl_sync_ver;
949 int sysctl_sync_ports;
950 int sysctl_sync_persist_mode;
951 unsigned long sysctl_sync_qlen_max;
952 int sysctl_sync_sock_size;
953 int sysctl_cache_bypass;
954 int sysctl_expire_nodest_conn;
955 int sysctl_sloppy_tcp;
956 int sysctl_sloppy_sctp;
957 int sysctl_expire_quiescent_template;
958 int sysctl_sync_threshold[2];
959 unsigned int sysctl_sync_refresh_period;
960 int sysctl_sync_retries;
961 int sysctl_nat_icmp_send;
962 int sysctl_pmtu_disc;
963 int sysctl_backup_only;
964 int sysctl_conn_reuse_mode;
965
966 /* ip_vs_lblc */
967 int sysctl_lblc_expiration;
968 struct ctl_table_header *lblc_ctl_header;
969 struct ctl_table *lblc_ctl_table;
970 /* ip_vs_lblcr */
971 int sysctl_lblcr_expiration;
972 struct ctl_table_header *lblcr_ctl_header;
973 struct ctl_table *lblcr_ctl_table;
974 /* ip_vs_est */
975 struct list_head est_list; /* estimator list */
976 spinlock_t est_lock;
977 struct timer_list est_timer; /* Estimation timer */
978 /* ip_vs_sync */
979 spinlock_t sync_lock;
980 struct ipvs_master_sync_state *ms;
981 spinlock_t sync_buff_lock;
982 struct task_struct **backup_threads;
983 int threads_mask;
984 volatile int sync_state;
985 struct mutex sync_mutex;
986 struct ipvs_sync_daemon_cfg mcfg; /* Master Configuration */
987 struct ipvs_sync_daemon_cfg bcfg; /* Backup Configuration */
988 /* net name space ptr */
989 struct net *net; /* Needed by timer routines */
990 /* Number of heterogeneous destinations, needed becaus heterogeneous
991 * are not supported when synchronization is enabled.
992 */
993 unsigned int mixed_address_family_dests;
994 };
995
996 #define DEFAULT_SYNC_THRESHOLD 3
997 #define DEFAULT_SYNC_PERIOD 50
998 #define DEFAULT_SYNC_VER 1
999 #define DEFAULT_SLOPPY_TCP 0
1000 #define DEFAULT_SLOPPY_SCTP 0
1001 #define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ)
1002 #define DEFAULT_SYNC_RETRIES 0
1003 #define IPVS_SYNC_WAKEUP_RATE 8
1004 #define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4)
1005 #define IPVS_SYNC_SEND_DELAY (HZ / 50)
1006 #define IPVS_SYNC_CHECK_PERIOD HZ
1007 #define IPVS_SYNC_FLUSH_TIME (HZ * 2)
1008 #define IPVS_SYNC_PORTS_MAX (1 << 6)
1009
1010 #ifdef CONFIG_SYSCTL
1011
1012 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1013 {
1014 return ipvs->sysctl_sync_threshold[0];
1015 }
1016
1017 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1018 {
1019 return ACCESS_ONCE(ipvs->sysctl_sync_threshold[1]);
1020 }
1021
1022 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1023 {
1024 return ACCESS_ONCE(ipvs->sysctl_sync_refresh_period);
1025 }
1026
1027 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1028 {
1029 return ipvs->sysctl_sync_retries;
1030 }
1031
1032 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1033 {
1034 return ipvs->sysctl_sync_ver;
1035 }
1036
1037 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1038 {
1039 return ipvs->sysctl_sloppy_tcp;
1040 }
1041
1042 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1043 {
1044 return ipvs->sysctl_sloppy_sctp;
1045 }
1046
1047 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1048 {
1049 return ACCESS_ONCE(ipvs->sysctl_sync_ports);
1050 }
1051
1052 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1053 {
1054 return ipvs->sysctl_sync_persist_mode;
1055 }
1056
1057 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1058 {
1059 return ipvs->sysctl_sync_qlen_max;
1060 }
1061
1062 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1063 {
1064 return ipvs->sysctl_sync_sock_size;
1065 }
1066
1067 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1068 {
1069 return ipvs->sysctl_pmtu_disc;
1070 }
1071
1072 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1073 {
1074 return ipvs->sync_state & IP_VS_STATE_BACKUP &&
1075 ipvs->sysctl_backup_only;
1076 }
1077
1078 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1079 {
1080 return ipvs->sysctl_conn_reuse_mode;
1081 }
1082
1083 #else
1084
1085 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1086 {
1087 return DEFAULT_SYNC_THRESHOLD;
1088 }
1089
1090 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1091 {
1092 return DEFAULT_SYNC_PERIOD;
1093 }
1094
1095 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1096 {
1097 return DEFAULT_SYNC_REFRESH_PERIOD;
1098 }
1099
1100 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1101 {
1102 return DEFAULT_SYNC_RETRIES & 3;
1103 }
1104
1105 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1106 {
1107 return DEFAULT_SYNC_VER;
1108 }
1109
1110 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1111 {
1112 return DEFAULT_SLOPPY_TCP;
1113 }
1114
1115 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1116 {
1117 return DEFAULT_SLOPPY_SCTP;
1118 }
1119
1120 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1121 {
1122 return 1;
1123 }
1124
1125 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1126 {
1127 return 0;
1128 }
1129
1130 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1131 {
1132 return IPVS_SYNC_QLEN_MAX;
1133 }
1134
1135 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1136 {
1137 return 0;
1138 }
1139
1140 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1141 {
1142 return 1;
1143 }
1144
1145 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1146 {
1147 return 0;
1148 }
1149
1150 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1151 {
1152 return 1;
1153 }
1154
1155 #endif
1156
1157 /* IPVS core functions
1158 * (from ip_vs_core.c)
1159 */
1160 const char *ip_vs_proto_name(unsigned int proto);
1161 void ip_vs_init_hash_table(struct list_head *table, int rows);
1162 #define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t)))
1163
1164 #define IP_VS_APP_TYPE_FTP 1
1165
1166 /* ip_vs_conn handling functions
1167 * (from ip_vs_conn.c)
1168 */
1169 enum {
1170 IP_VS_DIR_INPUT = 0,
1171 IP_VS_DIR_OUTPUT,
1172 IP_VS_DIR_INPUT_ONLY,
1173 IP_VS_DIR_LAST,
1174 };
1175
1176 static inline void ip_vs_conn_fill_param(struct net *net, int af, int protocol,
1177 const union nf_inet_addr *caddr,
1178 __be16 cport,
1179 const union nf_inet_addr *vaddr,
1180 __be16 vport,
1181 struct ip_vs_conn_param *p)
1182 {
1183 p->net = net;
1184 p->af = af;
1185 p->protocol = protocol;
1186 p->caddr = caddr;
1187 p->cport = cport;
1188 p->vaddr = vaddr;
1189 p->vport = vport;
1190 p->pe = NULL;
1191 p->pe_data = NULL;
1192 }
1193
1194 struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p);
1195 struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p);
1196
1197 struct ip_vs_conn * ip_vs_conn_in_get_proto(int af, const struct sk_buff *skb,
1198 const struct ip_vs_iphdr *iph,
1199 int inverse);
1200
1201 struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p);
1202
1203 struct ip_vs_conn * ip_vs_conn_out_get_proto(int af, const struct sk_buff *skb,
1204 const struct ip_vs_iphdr *iph,
1205 int inverse);
1206
1207 /* Get reference to gain full access to conn.
1208 * By default, RCU read-side critical sections have access only to
1209 * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference.
1210 */
1211 static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp)
1212 {
1213 return atomic_inc_not_zero(&cp->refcnt);
1214 }
1215
1216 /* put back the conn without restarting its timer */
1217 static inline void __ip_vs_conn_put(struct ip_vs_conn *cp)
1218 {
1219 smp_mb__before_atomic();
1220 atomic_dec(&cp->refcnt);
1221 }
1222 void ip_vs_conn_put(struct ip_vs_conn *cp);
1223 void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport);
1224
1225 struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, int dest_af,
1226 const union nf_inet_addr *daddr,
1227 __be16 dport, unsigned int flags,
1228 struct ip_vs_dest *dest, __u32 fwmark);
1229 void ip_vs_conn_expire_now(struct ip_vs_conn *cp);
1230
1231 const char *ip_vs_state_name(__u16 proto, int state);
1232
1233 void ip_vs_tcp_conn_listen(struct net *net, struct ip_vs_conn *cp);
1234 int ip_vs_check_template(struct ip_vs_conn *ct);
1235 void ip_vs_random_dropentry(struct net *net);
1236 int ip_vs_conn_init(void);
1237 void ip_vs_conn_cleanup(void);
1238
1239 static inline void ip_vs_control_del(struct ip_vs_conn *cp)
1240 {
1241 struct ip_vs_conn *ctl_cp = cp->control;
1242 if (!ctl_cp) {
1243 IP_VS_ERR_BUF("request control DEL for uncontrolled: "
1244 "%s:%d to %s:%d\n",
1245 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1246 ntohs(cp->cport),
1247 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1248 ntohs(cp->vport));
1249
1250 return;
1251 }
1252
1253 IP_VS_DBG_BUF(7, "DELeting control for: "
1254 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1255 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1256 ntohs(cp->cport),
1257 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1258 ntohs(ctl_cp->cport));
1259
1260 cp->control = NULL;
1261 if (atomic_read(&ctl_cp->n_control) == 0) {
1262 IP_VS_ERR_BUF("BUG control DEL with n=0 : "
1263 "%s:%d to %s:%d\n",
1264 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1265 ntohs(cp->cport),
1266 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1267 ntohs(cp->vport));
1268
1269 return;
1270 }
1271 atomic_dec(&ctl_cp->n_control);
1272 }
1273
1274 static inline void
1275 ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp)
1276 {
1277 if (cp->control) {
1278 IP_VS_ERR_BUF("request control ADD for already controlled: "
1279 "%s:%d to %s:%d\n",
1280 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1281 ntohs(cp->cport),
1282 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1283 ntohs(cp->vport));
1284
1285 ip_vs_control_del(cp);
1286 }
1287
1288 IP_VS_DBG_BUF(7, "ADDing control for: "
1289 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1290 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1291 ntohs(cp->cport),
1292 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1293 ntohs(ctl_cp->cport));
1294
1295 cp->control = ctl_cp;
1296 atomic_inc(&ctl_cp->n_control);
1297 }
1298
1299 /* IPVS netns init & cleanup functions */
1300 int ip_vs_estimator_net_init(struct net *net);
1301 int ip_vs_control_net_init(struct net *net);
1302 int ip_vs_protocol_net_init(struct net *net);
1303 int ip_vs_app_net_init(struct net *net);
1304 int ip_vs_conn_net_init(struct net *net);
1305 int ip_vs_sync_net_init(struct net *net);
1306 void ip_vs_conn_net_cleanup(struct net *net);
1307 void ip_vs_app_net_cleanup(struct net *net);
1308 void ip_vs_protocol_net_cleanup(struct net *net);
1309 void ip_vs_control_net_cleanup(struct net *net);
1310 void ip_vs_estimator_net_cleanup(struct net *net);
1311 void ip_vs_sync_net_cleanup(struct net *net);
1312 void ip_vs_service_net_cleanup(struct net *net);
1313
1314 /* IPVS application functions
1315 * (from ip_vs_app.c)
1316 */
1317 #define IP_VS_APP_MAX_PORTS 8
1318 struct ip_vs_app *register_ip_vs_app(struct net *net, struct ip_vs_app *app);
1319 void unregister_ip_vs_app(struct net *net, struct ip_vs_app *app);
1320 int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1321 void ip_vs_unbind_app(struct ip_vs_conn *cp);
1322 int register_ip_vs_app_inc(struct net *net, struct ip_vs_app *app, __u16 proto,
1323 __u16 port);
1324 int ip_vs_app_inc_get(struct ip_vs_app *inc);
1325 void ip_vs_app_inc_put(struct ip_vs_app *inc);
1326
1327 int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb);
1328 int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb);
1329
1330 int register_ip_vs_pe(struct ip_vs_pe *pe);
1331 int unregister_ip_vs_pe(struct ip_vs_pe *pe);
1332 struct ip_vs_pe *ip_vs_pe_getbyname(const char *name);
1333 struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name);
1334
1335 /* Use a #define to avoid all of module.h just for these trivial ops */
1336 #define ip_vs_pe_get(pe) \
1337 if (pe && pe->module) \
1338 __module_get(pe->module);
1339
1340 #define ip_vs_pe_put(pe) \
1341 if (pe && pe->module) \
1342 module_put(pe->module);
1343
1344 /* IPVS protocol functions (from ip_vs_proto.c) */
1345 int ip_vs_protocol_init(void);
1346 void ip_vs_protocol_cleanup(void);
1347 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags);
1348 int *ip_vs_create_timeout_table(int *table, int size);
1349 int ip_vs_set_state_timeout(int *table, int num, const char *const *names,
1350 const char *name, int to);
1351 void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
1352 const struct sk_buff *skb, int offset,
1353 const char *msg);
1354
1355 extern struct ip_vs_protocol ip_vs_protocol_tcp;
1356 extern struct ip_vs_protocol ip_vs_protocol_udp;
1357 extern struct ip_vs_protocol ip_vs_protocol_icmp;
1358 extern struct ip_vs_protocol ip_vs_protocol_esp;
1359 extern struct ip_vs_protocol ip_vs_protocol_ah;
1360 extern struct ip_vs_protocol ip_vs_protocol_sctp;
1361
1362 /* Registering/unregistering scheduler functions
1363 * (from ip_vs_sched.c)
1364 */
1365 int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1366 int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1367 int ip_vs_bind_scheduler(struct ip_vs_service *svc,
1368 struct ip_vs_scheduler *scheduler);
1369 void ip_vs_unbind_scheduler(struct ip_vs_service *svc,
1370 struct ip_vs_scheduler *sched);
1371 struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name);
1372 void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler);
1373 struct ip_vs_conn *
1374 ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb,
1375 struct ip_vs_proto_data *pd, int *ignored,
1376 struct ip_vs_iphdr *iph);
1377 int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb,
1378 struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph);
1379
1380 void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg);
1381
1382 /* IPVS control data and functions (from ip_vs_ctl.c) */
1383 extern struct ip_vs_stats ip_vs_stats;
1384 extern int sysctl_ip_vs_sync_ver;
1385
1386 struct ip_vs_service *
1387 ip_vs_service_find(struct net *net, int af, __u32 fwmark, __u16 protocol,
1388 const union nf_inet_addr *vaddr, __be16 vport);
1389
1390 bool ip_vs_has_real_service(struct net *net, int af, __u16 protocol,
1391 const union nf_inet_addr *daddr, __be16 dport);
1392
1393 int ip_vs_use_count_inc(void);
1394 void ip_vs_use_count_dec(void);
1395 int ip_vs_register_nl_ioctl(void);
1396 void ip_vs_unregister_nl_ioctl(void);
1397 int ip_vs_control_init(void);
1398 void ip_vs_control_cleanup(void);
1399 struct ip_vs_dest *
1400 ip_vs_find_dest(struct net *net, int svc_af, int dest_af,
1401 const union nf_inet_addr *daddr, __be16 dport,
1402 const union nf_inet_addr *vaddr, __be16 vport,
1403 __u16 protocol, __u32 fwmark, __u32 flags);
1404 void ip_vs_try_bind_dest(struct ip_vs_conn *cp);
1405
1406 static inline void ip_vs_dest_hold(struct ip_vs_dest *dest)
1407 {
1408 atomic_inc(&dest->refcnt);
1409 }
1410
1411 static inline void ip_vs_dest_put(struct ip_vs_dest *dest)
1412 {
1413 smp_mb__before_atomic();
1414 atomic_dec(&dest->refcnt);
1415 }
1416
1417 static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest)
1418 {
1419 if (atomic_dec_return(&dest->refcnt) < 0)
1420 kfree(dest);
1421 }
1422
1423 /* IPVS sync daemon data and function prototypes
1424 * (from ip_vs_sync.c)
1425 */
1426 int start_sync_thread(struct net *net, struct ipvs_sync_daemon_cfg *cfg,
1427 int state);
1428 int stop_sync_thread(struct net *net, int state);
1429 void ip_vs_sync_conn(struct net *net, struct ip_vs_conn *cp, int pkts);
1430
1431 /* IPVS rate estimator prototypes (from ip_vs_est.c) */
1432 void ip_vs_start_estimator(struct net *net, struct ip_vs_stats *stats);
1433 void ip_vs_stop_estimator(struct net *net, struct ip_vs_stats *stats);
1434 void ip_vs_zero_estimator(struct ip_vs_stats *stats);
1435 void ip_vs_read_estimator(struct ip_vs_kstats *dst, struct ip_vs_stats *stats);
1436
1437 /* Various IPVS packet transmitters (from ip_vs_xmit.c) */
1438 int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1439 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1440 int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1441 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1442 int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1443 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1444 int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1445 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1446 int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1447 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1448 int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1449 struct ip_vs_protocol *pp, int offset,
1450 unsigned int hooknum, struct ip_vs_iphdr *iph);
1451 void ip_vs_dest_dst_rcu_free(struct rcu_head *head);
1452
1453 #ifdef CONFIG_IP_VS_IPV6
1454 int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1455 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1456 int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1457 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1458 int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1459 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1460 int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1461 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1462 int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1463 struct ip_vs_protocol *pp, int offset,
1464 unsigned int hooknum, struct ip_vs_iphdr *iph);
1465 #endif
1466
1467 #ifdef CONFIG_SYSCTL
1468 /* This is a simple mechanism to ignore packets when
1469 * we are loaded. Just set ip_vs_drop_rate to 'n' and
1470 * we start to drop 1/rate of the packets
1471 */
1472 static inline int ip_vs_todrop(struct netns_ipvs *ipvs)
1473 {
1474 if (!ipvs->drop_rate)
1475 return 0;
1476 if (--ipvs->drop_counter > 0)
1477 return 0;
1478 ipvs->drop_counter = ipvs->drop_rate;
1479 return 1;
1480 }
1481 #else
1482 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; }
1483 #endif
1484
1485 /* ip_vs_fwd_tag returns the forwarding tag of the connection */
1486 #define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK)
1487
1488 static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp)
1489 {
1490 char fwd;
1491
1492 switch (IP_VS_FWD_METHOD(cp)) {
1493 case IP_VS_CONN_F_MASQ:
1494 fwd = 'M'; break;
1495 case IP_VS_CONN_F_LOCALNODE:
1496 fwd = 'L'; break;
1497 case IP_VS_CONN_F_TUNNEL:
1498 fwd = 'T'; break;
1499 case IP_VS_CONN_F_DROUTE:
1500 fwd = 'R'; break;
1501 case IP_VS_CONN_F_BYPASS:
1502 fwd = 'B'; break;
1503 default:
1504 fwd = '?'; break;
1505 }
1506 return fwd;
1507 }
1508
1509 void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp,
1510 struct ip_vs_conn *cp, int dir);
1511
1512 #ifdef CONFIG_IP_VS_IPV6
1513 void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp,
1514 struct ip_vs_conn *cp, int dir);
1515 #endif
1516
1517 __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset);
1518
1519 static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum)
1520 {
1521 __be32 diff[2] = { ~old, new };
1522
1523 return csum_partial(diff, sizeof(diff), oldsum);
1524 }
1525
1526 #ifdef CONFIG_IP_VS_IPV6
1527 static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new,
1528 __wsum oldsum)
1529 {
1530 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0],
1531 new[3], new[2], new[1], new[0] };
1532
1533 return csum_partial(diff, sizeof(diff), oldsum);
1534 }
1535 #endif
1536
1537 static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum)
1538 {
1539 __be16 diff[2] = { ~old, new };
1540
1541 return csum_partial(diff, sizeof(diff), oldsum);
1542 }
1543
1544 /* Forget current conntrack (unconfirmed) and attach notrack entry */
1545 static inline void ip_vs_notrack(struct sk_buff *skb)
1546 {
1547 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1548 enum ip_conntrack_info ctinfo;
1549 struct nf_conn *ct = nf_ct_get(skb, &ctinfo);
1550
1551 if (!ct || !nf_ct_is_untracked(ct)) {
1552 nf_conntrack_put(skb->nfct);
1553 skb->nfct = &nf_ct_untracked_get()->ct_general;
1554 skb->nfctinfo = IP_CT_NEW;
1555 nf_conntrack_get(skb->nfct);
1556 }
1557 #endif
1558 }
1559
1560 #ifdef CONFIG_IP_VS_NFCT
1561 /* Netfilter connection tracking
1562 * (from ip_vs_nfct.c)
1563 */
1564 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1565 {
1566 #ifdef CONFIG_SYSCTL
1567 return ipvs->sysctl_conntrack;
1568 #else
1569 return 0;
1570 #endif
1571 }
1572
1573 void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp,
1574 int outin);
1575 int ip_vs_confirm_conntrack(struct sk_buff *skb);
1576 void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct,
1577 struct ip_vs_conn *cp, u_int8_t proto,
1578 const __be16 port, int from_rs);
1579 void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp);
1580
1581 #else
1582
1583 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1584 {
1585 return 0;
1586 }
1587
1588 static inline void ip_vs_update_conntrack(struct sk_buff *skb,
1589 struct ip_vs_conn *cp, int outin)
1590 {
1591 }
1592
1593 static inline int ip_vs_confirm_conntrack(struct sk_buff *skb)
1594 {
1595 return NF_ACCEPT;
1596 }
1597
1598 static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp)
1599 {
1600 }
1601 #endif /* CONFIG_IP_VS_NFCT */
1602
1603 static inline int
1604 ip_vs_dest_conn_overhead(struct ip_vs_dest *dest)
1605 {
1606 /* We think the overhead of processing active connections is 256
1607 * times higher than that of inactive connections in average. (This
1608 * 256 times might not be accurate, we will change it later) We
1609 * use the following formula to estimate the overhead now:
1610 * dest->activeconns*256 + dest->inactconns
1611 */
1612 return (atomic_read(&dest->activeconns) << 8) +
1613 atomic_read(&dest->inactconns);
1614 }
1615
1616 #endif /* _NET_IP_VS_H */
Linux kernel - Deliverables
This page took 0.095802 seconds and 6 git commands to generate.