]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - include/net/ip_vs.h
projects / mirror_ubuntu-artful-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
ipvs: Wrap sysctl_cache_bypass and remove ifdefs in ip_vs_leave
[mirror_ubuntu-artful-kernel.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 netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
487
488 void (*exit_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
489
490 int (*conn_schedule)(struct netns_ipvs *ipvs,
491 int af, struct sk_buff *skb,
492 struct ip_vs_proto_data *pd,
493 int *verdict, struct ip_vs_conn **cpp,
494 struct ip_vs_iphdr *iph);
495
496 struct ip_vs_conn *
497 (*conn_in_get)(struct netns_ipvs *ipvs,
498 int af,
499 const struct sk_buff *skb,
500 const struct ip_vs_iphdr *iph);
501
502 struct ip_vs_conn *
503 (*conn_out_get)(struct netns_ipvs *ipvs,
504 int af,
505 const struct sk_buff *skb,
506 const struct ip_vs_iphdr *iph);
507
508 int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
509 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
510
511 int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
512 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
513
514 int (*csum_check)(int af, struct sk_buff *skb,
515 struct ip_vs_protocol *pp);
516
517 const char *(*state_name)(int state);
518
519 void (*state_transition)(struct ip_vs_conn *cp, int direction,
520 const struct sk_buff *skb,
521 struct ip_vs_proto_data *pd);
522
523 int (*register_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
524
525 void (*unregister_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
526
527 int (*app_conn_bind)(struct ip_vs_conn *cp);
528
529 void (*debug_packet)(int af, struct ip_vs_protocol *pp,
530 const struct sk_buff *skb,
531 int offset,
532 const char *msg);
533
534 void (*timeout_change)(struct ip_vs_proto_data *pd, int flags);
535 };
536
537 /* protocol data per netns */
538 struct ip_vs_proto_data {
539 struct ip_vs_proto_data *next;
540 struct ip_vs_protocol *pp;
541 int *timeout_table; /* protocol timeout table */
542 atomic_t appcnt; /* counter of proto app incs. */
543 struct tcp_states_t *tcp_state_table;
544 };
545
546 struct ip_vs_protocol *ip_vs_proto_get(unsigned short proto);
547 struct ip_vs_proto_data *ip_vs_proto_data_get(struct netns_ipvs *ipvs,
548 unsigned short proto);
549
550 struct ip_vs_conn_param {
551 struct netns_ipvs *ipvs;
552 const union nf_inet_addr *caddr;
553 const union nf_inet_addr *vaddr;
554 __be16 cport;
555 __be16 vport;
556 __u16 protocol;
557 u16 af;
558
559 const struct ip_vs_pe *pe;
560 char *pe_data;
561 __u8 pe_data_len;
562 };
563
564 /* IP_VS structure allocated for each dynamically scheduled connection */
565 struct ip_vs_conn {
566 struct hlist_node c_list; /* hashed list heads */
567 /* Protocol, addresses and port numbers */
568 __be16 cport;
569 __be16 dport;
570 __be16 vport;
571 u16 af; /* address family */
572 union nf_inet_addr caddr; /* client address */
573 union nf_inet_addr vaddr; /* virtual address */
574 union nf_inet_addr daddr; /* destination address */
575 volatile __u32 flags; /* status flags */
576 __u16 protocol; /* Which protocol (TCP/UDP) */
577 __u16 daf; /* Address family of the dest */
578 struct netns_ipvs *ipvs;
579
580 /* counter and timer */
581 atomic_t refcnt; /* reference count */
582 struct timer_list timer; /* Expiration timer */
583 volatile unsigned long timeout; /* timeout */
584
585 /* Flags and state transition */
586 spinlock_t lock; /* lock for state transition */
587 volatile __u16 state; /* state info */
588 volatile __u16 old_state; /* old state, to be used for
589 * state transition triggerd
590 * synchronization
591 */
592 __u32 fwmark; /* Fire wall mark from skb */
593 unsigned long sync_endtime; /* jiffies + sent_retries */
594
595 /* Control members */
596 struct ip_vs_conn *control; /* Master control connection */
597 atomic_t n_control; /* Number of controlled ones */
598 struct ip_vs_dest *dest; /* real server */
599 atomic_t in_pkts; /* incoming packet counter */
600
601 /* Packet transmitter for different forwarding methods. If it
602 * mangles the packet, it must return NF_DROP or better NF_STOLEN,
603 * otherwise this must be changed to a sk_buff **.
604 * NF_ACCEPT can be returned when destination is local.
605 */
606 int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp,
607 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
608
609 /* Note: we can group the following members into a structure,
610 * in order to save more space, and the following members are
611 * only used in VS/NAT anyway
612 */
613 struct ip_vs_app *app; /* bound ip_vs_app object */
614 void *app_data; /* Application private data */
615 struct ip_vs_seq in_seq; /* incoming seq. struct */
616 struct ip_vs_seq out_seq; /* outgoing seq. struct */
617
618 const struct ip_vs_pe *pe;
619 char *pe_data;
620 __u8 pe_data_len;
621
622 struct rcu_head rcu_head;
623 };
624
625 /* Extended internal versions of struct ip_vs_service_user and ip_vs_dest_user
626 * for IPv6 support.
627 *
628 * We need these to conveniently pass around service and destination
629 * options, but unfortunately, we also need to keep the old definitions to
630 * maintain userspace backwards compatibility for the setsockopt interface.
631 */
632 struct ip_vs_service_user_kern {
633 /* virtual service addresses */
634 u16 af;
635 u16 protocol;
636 union nf_inet_addr addr; /* virtual ip address */
637 __be16 port;
638 u32 fwmark; /* firwall mark of service */
639
640 /* virtual service options */
641 char *sched_name;
642 char *pe_name;
643 unsigned int flags; /* virtual service flags */
644 unsigned int timeout; /* persistent timeout in sec */
645 __be32 netmask; /* persistent netmask or plen */
646 };
647
648
649 struct ip_vs_dest_user_kern {
650 /* destination server address */
651 union nf_inet_addr addr;
652 __be16 port;
653
654 /* real server options */
655 unsigned int conn_flags; /* connection flags */
656 int weight; /* destination weight */
657
658 /* thresholds for active connections */
659 u32 u_threshold; /* upper threshold */
660 u32 l_threshold; /* lower threshold */
661
662 /* Address family of addr */
663 u16 af;
664 };
665
666
667 /*
668 * The information about the virtual service offered to the net and the
669 * forwarding entries.
670 */
671 struct ip_vs_service {
672 struct hlist_node s_list; /* for normal service table */
673 struct hlist_node f_list; /* for fwmark-based service table */
674 atomic_t refcnt; /* reference counter */
675
676 u16 af; /* address family */
677 __u16 protocol; /* which protocol (TCP/UDP) */
678 union nf_inet_addr addr; /* IP address for virtual service */
679 __be16 port; /* port number for the service */
680 __u32 fwmark; /* firewall mark of the service */
681 unsigned int flags; /* service status flags */
682 unsigned int timeout; /* persistent timeout in ticks */
683 __be32 netmask; /* grouping granularity, mask/plen */
684 struct netns_ipvs *ipvs;
685
686 struct list_head destinations; /* real server d-linked list */
687 __u32 num_dests; /* number of servers */
688 struct ip_vs_stats stats; /* statistics for the service */
689
690 /* for scheduling */
691 struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */
692 spinlock_t sched_lock; /* lock sched_data */
693 void *sched_data; /* scheduler application data */
694
695 /* alternate persistence engine */
696 struct ip_vs_pe __rcu *pe;
697
698 struct rcu_head rcu_head;
699 };
700
701 /* Information for cached dst */
702 struct ip_vs_dest_dst {
703 struct dst_entry *dst_cache; /* destination cache entry */
704 u32 dst_cookie;
705 union nf_inet_addr dst_saddr;
706 struct rcu_head rcu_head;
707 };
708
709 /* The real server destination forwarding entry with ip address, port number,
710 * and so on.
711 */
712 struct ip_vs_dest {
713 struct list_head n_list; /* for the dests in the service */
714 struct hlist_node d_list; /* for table with all the dests */
715
716 u16 af; /* address family */
717 __be16 port; /* port number of the server */
718 union nf_inet_addr addr; /* IP address of the server */
719 volatile unsigned int flags; /* dest status flags */
720 atomic_t conn_flags; /* flags to copy to conn */
721 atomic_t weight; /* server weight */
722
723 atomic_t refcnt; /* reference counter */
724 struct ip_vs_stats stats; /* statistics */
725 unsigned long idle_start; /* start time, jiffies */
726
727 /* connection counters and thresholds */
728 atomic_t activeconns; /* active connections */
729 atomic_t inactconns; /* inactive connections */
730 atomic_t persistconns; /* persistent connections */
731 __u32 u_threshold; /* upper threshold */
732 __u32 l_threshold; /* lower threshold */
733
734 /* for destination cache */
735 spinlock_t dst_lock; /* lock of dst_cache */
736 struct ip_vs_dest_dst __rcu *dest_dst; /* cached dst info */
737
738 /* for virtual service */
739 struct ip_vs_service __rcu *svc; /* service it belongs to */
740 __u16 protocol; /* which protocol (TCP/UDP) */
741 __be16 vport; /* virtual port number */
742 union nf_inet_addr vaddr; /* virtual IP address */
743 __u32 vfwmark; /* firewall mark of service */
744
745 struct list_head t_list; /* in dest_trash */
746 unsigned int in_rs_table:1; /* we are in rs_table */
747 };
748
749 /* The scheduler object */
750 struct ip_vs_scheduler {
751 struct list_head n_list; /* d-linked list head */
752 char *name; /* scheduler name */
753 atomic_t refcnt; /* reference counter */
754 struct module *module; /* THIS_MODULE/NULL */
755
756 /* scheduler initializing service */
757 int (*init_service)(struct ip_vs_service *svc);
758 /* scheduling service finish */
759 void (*done_service)(struct ip_vs_service *svc);
760 /* dest is linked */
761 int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
762 /* dest is unlinked */
763 int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
764 /* dest is updated */
765 int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
766
767 /* selecting a server from the given service */
768 struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc,
769 const struct sk_buff *skb,
770 struct ip_vs_iphdr *iph);
771 };
772
773 /* The persistence engine object */
774 struct ip_vs_pe {
775 struct list_head n_list; /* d-linked list head */
776 char *name; /* scheduler name */
777 atomic_t refcnt; /* reference counter */
778 struct module *module; /* THIS_MODULE/NULL */
779
780 /* get the connection template, if any */
781 int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb);
782 bool (*ct_match)(const struct ip_vs_conn_param *p,
783 struct ip_vs_conn *ct);
784 u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval,
785 bool inverse);
786 int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf);
787 };
788
789 /* The application module object (a.k.a. app incarnation) */
790 struct ip_vs_app {
791 struct list_head a_list; /* member in app list */
792 int type; /* IP_VS_APP_TYPE_xxx */
793 char *name; /* application module name */
794 __u16 protocol;
795 struct module *module; /* THIS_MODULE/NULL */
796 struct list_head incs_list; /* list of incarnations */
797
798 /* members for application incarnations */
799 struct list_head p_list; /* member in proto app list */
800 struct ip_vs_app *app; /* its real application */
801 __be16 port; /* port number in net order */
802 atomic_t usecnt; /* usage counter */
803 struct rcu_head rcu_head;
804
805 /* output hook: Process packet in inout direction, diff set for TCP.
806 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
807 * 2=Mangled but checksum was not updated
808 */
809 int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *,
810 struct sk_buff *, int *diff);
811
812 /* input hook: Process packet in outin direction, diff set for TCP.
813 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
814 * 2=Mangled but checksum was not updated
815 */
816 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *,
817 struct sk_buff *, int *diff);
818
819 /* ip_vs_app initializer */
820 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *);
821
822 /* ip_vs_app finish */
823 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *);
824
825
826 /* not used now */
827 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *,
828 struct ip_vs_protocol *);
829
830 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *);
831
832 int * timeout_table;
833 int * timeouts;
834 int timeouts_size;
835
836 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app,
837 int *verdict, struct ip_vs_conn **cpp);
838
839 struct ip_vs_conn *
840 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app,
841 const struct iphdr *iph, int inverse);
842
843 struct ip_vs_conn *
844 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app,
845 const struct iphdr *iph, int inverse);
846
847 int (*state_transition)(struct ip_vs_conn *cp, int direction,
848 const struct sk_buff *skb,
849 struct ip_vs_app *app);
850
851 void (*timeout_change)(struct ip_vs_app *app, int flags);
852 };
853
854 struct ipvs_master_sync_state {
855 struct list_head sync_queue;
856 struct ip_vs_sync_buff *sync_buff;
857 unsigned long sync_queue_len;
858 unsigned int sync_queue_delay;
859 struct task_struct *master_thread;
860 struct delayed_work master_wakeup_work;
861 struct netns_ipvs *ipvs;
862 };
863
864 /* How much time to keep dests in trash */
865 #define IP_VS_DEST_TRASH_PERIOD (120 * HZ)
866
867 struct ipvs_sync_daemon_cfg {
868 union nf_inet_addr mcast_group;
869 int syncid;
870 u16 sync_maxlen;
871 u16 mcast_port;
872 u8 mcast_af;
873 u8 mcast_ttl;
874 /* multicast interface name */
875 char mcast_ifn[IP_VS_IFNAME_MAXLEN];
876 };
877
878 /* IPVS in network namespace */
879 struct netns_ipvs {
880 int gen; /* Generation */
881 int enable; /* enable like nf_hooks do */
882 /* Hash table: for real service lookups */
883 #define IP_VS_RTAB_BITS 4
884 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
885 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
886
887 struct hlist_head rs_table[IP_VS_RTAB_SIZE];
888 /* ip_vs_app */
889 struct list_head app_list;
890 /* ip_vs_proto */
891 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */
892 struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE];
893 /* ip_vs_proto_tcp */
894 #ifdef CONFIG_IP_VS_PROTO_TCP
895 #define TCP_APP_TAB_BITS 4
896 #define TCP_APP_TAB_SIZE (1 << TCP_APP_TAB_BITS)
897 #define TCP_APP_TAB_MASK (TCP_APP_TAB_SIZE - 1)
898 struct list_head tcp_apps[TCP_APP_TAB_SIZE];
899 #endif
900 /* ip_vs_proto_udp */
901 #ifdef CONFIG_IP_VS_PROTO_UDP
902 #define UDP_APP_TAB_BITS 4
903 #define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS)
904 #define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1)
905 struct list_head udp_apps[UDP_APP_TAB_SIZE];
906 #endif
907 /* ip_vs_proto_sctp */
908 #ifdef CONFIG_IP_VS_PROTO_SCTP
909 #define SCTP_APP_TAB_BITS 4
910 #define SCTP_APP_TAB_SIZE (1 << SCTP_APP_TAB_BITS)
911 #define SCTP_APP_TAB_MASK (SCTP_APP_TAB_SIZE - 1)
912 /* Hash table for SCTP application incarnations */
913 struct list_head sctp_apps[SCTP_APP_TAB_SIZE];
914 #endif
915 /* ip_vs_conn */
916 atomic_t conn_count; /* connection counter */
917
918 /* ip_vs_ctl */
919 struct ip_vs_stats tot_stats; /* Statistics & est. */
920
921 int num_services; /* no of virtual services */
922
923 /* Trash for destinations */
924 struct list_head dest_trash;
925 spinlock_t dest_trash_lock;
926 struct timer_list dest_trash_timer; /* expiration timer */
927 /* Service counters */
928 atomic_t ftpsvc_counter;
929 atomic_t nullsvc_counter;
930
931 #ifdef CONFIG_SYSCTL
932 /* 1/rate drop and drop-entry variables */
933 struct delayed_work defense_work; /* Work handler */
934 int drop_rate;
935 int drop_counter;
936 atomic_t dropentry;
937 /* locks in ctl.c */
938 spinlock_t dropentry_lock; /* drop entry handling */
939 spinlock_t droppacket_lock; /* drop packet handling */
940 spinlock_t securetcp_lock; /* state and timeout tables */
941
942 /* sys-ctl struct */
943 struct ctl_table_header *sysctl_hdr;
944 struct ctl_table *sysctl_tbl;
945 #endif
946
947 /* sysctl variables */
948 int sysctl_amemthresh;
949 int sysctl_am_droprate;
950 int sysctl_drop_entry;
951 int sysctl_drop_packet;
952 int sysctl_secure_tcp;
953 #ifdef CONFIG_IP_VS_NFCT
954 int sysctl_conntrack;
955 #endif
956 int sysctl_snat_reroute;
957 int sysctl_sync_ver;
958 int sysctl_sync_ports;
959 int sysctl_sync_persist_mode;
960 unsigned long sysctl_sync_qlen_max;
961 int sysctl_sync_sock_size;
962 int sysctl_cache_bypass;
963 int sysctl_expire_nodest_conn;
964 int sysctl_sloppy_tcp;
965 int sysctl_sloppy_sctp;
966 int sysctl_expire_quiescent_template;
967 int sysctl_sync_threshold[2];
968 unsigned int sysctl_sync_refresh_period;
969 int sysctl_sync_retries;
970 int sysctl_nat_icmp_send;
971 int sysctl_pmtu_disc;
972 int sysctl_backup_only;
973 int sysctl_conn_reuse_mode;
974 int sysctl_schedule_icmp;
975 int sysctl_ignore_tunneled;
976
977 /* ip_vs_lblc */
978 int sysctl_lblc_expiration;
979 struct ctl_table_header *lblc_ctl_header;
980 struct ctl_table *lblc_ctl_table;
981 /* ip_vs_lblcr */
982 int sysctl_lblcr_expiration;
983 struct ctl_table_header *lblcr_ctl_header;
984 struct ctl_table *lblcr_ctl_table;
985 /* ip_vs_est */
986 struct list_head est_list; /* estimator list */
987 spinlock_t est_lock;
988 struct timer_list est_timer; /* Estimation timer */
989 /* ip_vs_sync */
990 spinlock_t sync_lock;
991 struct ipvs_master_sync_state *ms;
992 spinlock_t sync_buff_lock;
993 struct task_struct **backup_threads;
994 int threads_mask;
995 volatile int sync_state;
996 struct mutex sync_mutex;
997 struct ipvs_sync_daemon_cfg mcfg; /* Master Configuration */
998 struct ipvs_sync_daemon_cfg bcfg; /* Backup Configuration */
999 /* net name space ptr */
1000 struct net *net; /* Needed by timer routines */
1001 /* Number of heterogeneous destinations, needed becaus heterogeneous
1002 * are not supported when synchronization is enabled.
1003 */
1004 unsigned int mixed_address_family_dests;
1005 };
1006
1007 #define DEFAULT_SYNC_THRESHOLD 3
1008 #define DEFAULT_SYNC_PERIOD 50
1009 #define DEFAULT_SYNC_VER 1
1010 #define DEFAULT_SLOPPY_TCP 0
1011 #define DEFAULT_SLOPPY_SCTP 0
1012 #define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ)
1013 #define DEFAULT_SYNC_RETRIES 0
1014 #define IPVS_SYNC_WAKEUP_RATE 8
1015 #define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4)
1016 #define IPVS_SYNC_SEND_DELAY (HZ / 50)
1017 #define IPVS_SYNC_CHECK_PERIOD HZ
1018 #define IPVS_SYNC_FLUSH_TIME (HZ * 2)
1019 #define IPVS_SYNC_PORTS_MAX (1 << 6)
1020
1021 #ifdef CONFIG_SYSCTL
1022
1023 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1024 {
1025 return ipvs->sysctl_sync_threshold[0];
1026 }
1027
1028 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1029 {
1030 return ACCESS_ONCE(ipvs->sysctl_sync_threshold[1]);
1031 }
1032
1033 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1034 {
1035 return ACCESS_ONCE(ipvs->sysctl_sync_refresh_period);
1036 }
1037
1038 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1039 {
1040 return ipvs->sysctl_sync_retries;
1041 }
1042
1043 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1044 {
1045 return ipvs->sysctl_sync_ver;
1046 }
1047
1048 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1049 {
1050 return ipvs->sysctl_sloppy_tcp;
1051 }
1052
1053 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1054 {
1055 return ipvs->sysctl_sloppy_sctp;
1056 }
1057
1058 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1059 {
1060 return ACCESS_ONCE(ipvs->sysctl_sync_ports);
1061 }
1062
1063 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1064 {
1065 return ipvs->sysctl_sync_persist_mode;
1066 }
1067
1068 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1069 {
1070 return ipvs->sysctl_sync_qlen_max;
1071 }
1072
1073 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1074 {
1075 return ipvs->sysctl_sync_sock_size;
1076 }
1077
1078 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1079 {
1080 return ipvs->sysctl_pmtu_disc;
1081 }
1082
1083 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1084 {
1085 return ipvs->sync_state & IP_VS_STATE_BACKUP &&
1086 ipvs->sysctl_backup_only;
1087 }
1088
1089 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1090 {
1091 return ipvs->sysctl_conn_reuse_mode;
1092 }
1093
1094 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1095 {
1096 return ipvs->sysctl_schedule_icmp;
1097 }
1098
1099 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1100 {
1101 return ipvs->sysctl_ignore_tunneled;
1102 }
1103
1104 static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1105 {
1106 return ipvs->sysctl_cache_bypass;
1107 }
1108
1109 #else
1110
1111 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1112 {
1113 return DEFAULT_SYNC_THRESHOLD;
1114 }
1115
1116 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1117 {
1118 return DEFAULT_SYNC_PERIOD;
1119 }
1120
1121 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1122 {
1123 return DEFAULT_SYNC_REFRESH_PERIOD;
1124 }
1125
1126 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1127 {
1128 return DEFAULT_SYNC_RETRIES & 3;
1129 }
1130
1131 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1132 {
1133 return DEFAULT_SYNC_VER;
1134 }
1135
1136 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1137 {
1138 return DEFAULT_SLOPPY_TCP;
1139 }
1140
1141 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1142 {
1143 return DEFAULT_SLOPPY_SCTP;
1144 }
1145
1146 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1147 {
1148 return 1;
1149 }
1150
1151 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1152 {
1153 return 0;
1154 }
1155
1156 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1157 {
1158 return IPVS_SYNC_QLEN_MAX;
1159 }
1160
1161 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1162 {
1163 return 0;
1164 }
1165
1166 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1167 {
1168 return 1;
1169 }
1170
1171 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1172 {
1173 return 0;
1174 }
1175
1176 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1177 {
1178 return 1;
1179 }
1180
1181 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1182 {
1183 return 0;
1184 }
1185
1186 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1187 {
1188 return 0;
1189 }
1190
1191 static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1192 {
1193 return 0;
1194 }
1195
1196 #endif
1197
1198 /* IPVS core functions
1199 * (from ip_vs_core.c)
1200 */
1201 const char *ip_vs_proto_name(unsigned int proto);
1202 void ip_vs_init_hash_table(struct list_head *table, int rows);
1203 #define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t)))
1204
1205 #define IP_VS_APP_TYPE_FTP 1
1206
1207 /* ip_vs_conn handling functions
1208 * (from ip_vs_conn.c)
1209 */
1210 enum {
1211 IP_VS_DIR_INPUT = 0,
1212 IP_VS_DIR_OUTPUT,
1213 IP_VS_DIR_INPUT_ONLY,
1214 IP_VS_DIR_LAST,
1215 };
1216
1217 static inline void ip_vs_conn_fill_param(struct netns_ipvs *ipvs, int af, int protocol,
1218 const union nf_inet_addr *caddr,
1219 __be16 cport,
1220 const union nf_inet_addr *vaddr,
1221 __be16 vport,
1222 struct ip_vs_conn_param *p)
1223 {
1224 p->ipvs = ipvs;
1225 p->af = af;
1226 p->protocol = protocol;
1227 p->caddr = caddr;
1228 p->cport = cport;
1229 p->vaddr = vaddr;
1230 p->vport = vport;
1231 p->pe = NULL;
1232 p->pe_data = NULL;
1233 }
1234
1235 struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p);
1236 struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p);
1237
1238 struct ip_vs_conn * ip_vs_conn_in_get_proto(struct netns_ipvs *ipvs, int af,
1239 const struct sk_buff *skb,
1240 const struct ip_vs_iphdr *iph);
1241
1242 struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p);
1243
1244 struct ip_vs_conn * ip_vs_conn_out_get_proto(struct netns_ipvs *ipvs, int af,
1245 const struct sk_buff *skb,
1246 const struct ip_vs_iphdr *iph);
1247
1248 /* Get reference to gain full access to conn.
1249 * By default, RCU read-side critical sections have access only to
1250 * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference.
1251 */
1252 static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp)
1253 {
1254 return atomic_inc_not_zero(&cp->refcnt);
1255 }
1256
1257 /* put back the conn without restarting its timer */
1258 static inline void __ip_vs_conn_put(struct ip_vs_conn *cp)
1259 {
1260 smp_mb__before_atomic();
1261 atomic_dec(&cp->refcnt);
1262 }
1263 void ip_vs_conn_put(struct ip_vs_conn *cp);
1264 void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport);
1265
1266 struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, int dest_af,
1267 const union nf_inet_addr *daddr,
1268 __be16 dport, unsigned int flags,
1269 struct ip_vs_dest *dest, __u32 fwmark);
1270 void ip_vs_conn_expire_now(struct ip_vs_conn *cp);
1271
1272 const char *ip_vs_state_name(__u16 proto, int state);
1273
1274 void ip_vs_tcp_conn_listen(struct net *net, struct ip_vs_conn *cp);
1275 int ip_vs_check_template(struct ip_vs_conn *ct);
1276 void ip_vs_random_dropentry(struct netns_ipvs *ipvs);
1277 int ip_vs_conn_init(void);
1278 void ip_vs_conn_cleanup(void);
1279
1280 static inline void ip_vs_control_del(struct ip_vs_conn *cp)
1281 {
1282 struct ip_vs_conn *ctl_cp = cp->control;
1283 if (!ctl_cp) {
1284 IP_VS_ERR_BUF("request control DEL for uncontrolled: "
1285 "%s:%d to %s:%d\n",
1286 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1287 ntohs(cp->cport),
1288 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1289 ntohs(cp->vport));
1290
1291 return;
1292 }
1293
1294 IP_VS_DBG_BUF(7, "DELeting control for: "
1295 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1296 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1297 ntohs(cp->cport),
1298 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1299 ntohs(ctl_cp->cport));
1300
1301 cp->control = NULL;
1302 if (atomic_read(&ctl_cp->n_control) == 0) {
1303 IP_VS_ERR_BUF("BUG control DEL with n=0 : "
1304 "%s:%d to %s:%d\n",
1305 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1306 ntohs(cp->cport),
1307 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1308 ntohs(cp->vport));
1309
1310 return;
1311 }
1312 atomic_dec(&ctl_cp->n_control);
1313 }
1314
1315 static inline void
1316 ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp)
1317 {
1318 if (cp->control) {
1319 IP_VS_ERR_BUF("request control ADD for already controlled: "
1320 "%s:%d to %s:%d\n",
1321 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1322 ntohs(cp->cport),
1323 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1324 ntohs(cp->vport));
1325
1326 ip_vs_control_del(cp);
1327 }
1328
1329 IP_VS_DBG_BUF(7, "ADDing control for: "
1330 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1331 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1332 ntohs(cp->cport),
1333 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1334 ntohs(ctl_cp->cport));
1335
1336 cp->control = ctl_cp;
1337 atomic_inc(&ctl_cp->n_control);
1338 }
1339
1340 /* IPVS netns init & cleanup functions */
1341 int ip_vs_estimator_net_init(struct netns_ipvs *ipvs);
1342 int ip_vs_control_net_init(struct netns_ipvs *ipvs);
1343 int ip_vs_protocol_net_init(struct net *net);
1344 int ip_vs_app_net_init(struct netns_ipvs *ipvs);
1345 int ip_vs_conn_net_init(struct netns_ipvs *ipvs);
1346 int ip_vs_sync_net_init(struct netns_ipvs *ipvs);
1347 void ip_vs_conn_net_cleanup(struct netns_ipvs *ipvs);
1348 void ip_vs_app_net_cleanup(struct netns_ipvs *ipvs);
1349 void ip_vs_protocol_net_cleanup(struct net *net);
1350 void ip_vs_control_net_cleanup(struct netns_ipvs *ipvs);
1351 void ip_vs_estimator_net_cleanup(struct netns_ipvs *ipvs);
1352 void ip_vs_sync_net_cleanup(struct netns_ipvs *ipvs);
1353 void ip_vs_service_net_cleanup(struct netns_ipvs *ipvs);
1354
1355 /* IPVS application functions
1356 * (from ip_vs_app.c)
1357 */
1358 #define IP_VS_APP_MAX_PORTS 8
1359 struct ip_vs_app *register_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1360 void unregister_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1361 int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1362 void ip_vs_unbind_app(struct ip_vs_conn *cp);
1363 int register_ip_vs_app_inc(struct netns_ipvs *ipvs, struct ip_vs_app *app, __u16 proto,
1364 __u16 port);
1365 int ip_vs_app_inc_get(struct ip_vs_app *inc);
1366 void ip_vs_app_inc_put(struct ip_vs_app *inc);
1367
1368 int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb);
1369 int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb);
1370
1371 int register_ip_vs_pe(struct ip_vs_pe *pe);
1372 int unregister_ip_vs_pe(struct ip_vs_pe *pe);
1373 struct ip_vs_pe *ip_vs_pe_getbyname(const char *name);
1374 struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name);
1375
1376 /* Use a #define to avoid all of module.h just for these trivial ops */
1377 #define ip_vs_pe_get(pe) \
1378 if (pe && pe->module) \
1379 __module_get(pe->module);
1380
1381 #define ip_vs_pe_put(pe) \
1382 if (pe && pe->module) \
1383 module_put(pe->module);
1384
1385 /* IPVS protocol functions (from ip_vs_proto.c) */
1386 int ip_vs_protocol_init(void);
1387 void ip_vs_protocol_cleanup(void);
1388 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags);
1389 int *ip_vs_create_timeout_table(int *table, int size);
1390 int ip_vs_set_state_timeout(int *table, int num, const char *const *names,
1391 const char *name, int to);
1392 void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
1393 const struct sk_buff *skb, int offset,
1394 const char *msg);
1395
1396 extern struct ip_vs_protocol ip_vs_protocol_tcp;
1397 extern struct ip_vs_protocol ip_vs_protocol_udp;
1398 extern struct ip_vs_protocol ip_vs_protocol_icmp;
1399 extern struct ip_vs_protocol ip_vs_protocol_esp;
1400 extern struct ip_vs_protocol ip_vs_protocol_ah;
1401 extern struct ip_vs_protocol ip_vs_protocol_sctp;
1402
1403 /* Registering/unregistering scheduler functions
1404 * (from ip_vs_sched.c)
1405 */
1406 int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1407 int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1408 int ip_vs_bind_scheduler(struct ip_vs_service *svc,
1409 struct ip_vs_scheduler *scheduler);
1410 void ip_vs_unbind_scheduler(struct ip_vs_service *svc,
1411 struct ip_vs_scheduler *sched);
1412 struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name);
1413 void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler);
1414 struct ip_vs_conn *
1415 ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb,
1416 struct ip_vs_proto_data *pd, int *ignored,
1417 struct ip_vs_iphdr *iph);
1418 int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb,
1419 struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph);
1420
1421 void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg);
1422
1423 /* IPVS control data and functions (from ip_vs_ctl.c) */
1424 extern struct ip_vs_stats ip_vs_stats;
1425 extern int sysctl_ip_vs_sync_ver;
1426
1427 struct ip_vs_service *
1428 ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
1429 const union nf_inet_addr *vaddr, __be16 vport);
1430
1431 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1432 const union nf_inet_addr *daddr, __be16 dport);
1433
1434 int ip_vs_use_count_inc(void);
1435 void ip_vs_use_count_dec(void);
1436 int ip_vs_register_nl_ioctl(void);
1437 void ip_vs_unregister_nl_ioctl(void);
1438 int ip_vs_control_init(void);
1439 void ip_vs_control_cleanup(void);
1440 struct ip_vs_dest *
1441 ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
1442 const union nf_inet_addr *daddr, __be16 dport,
1443 const union nf_inet_addr *vaddr, __be16 vport,
1444 __u16 protocol, __u32 fwmark, __u32 flags);
1445 void ip_vs_try_bind_dest(struct ip_vs_conn *cp);
1446
1447 static inline void ip_vs_dest_hold(struct ip_vs_dest *dest)
1448 {
1449 atomic_inc(&dest->refcnt);
1450 }
1451
1452 static inline void ip_vs_dest_put(struct ip_vs_dest *dest)
1453 {
1454 smp_mb__before_atomic();
1455 atomic_dec(&dest->refcnt);
1456 }
1457
1458 static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest)
1459 {
1460 if (atomic_dec_return(&dest->refcnt) < 0)
1461 kfree(dest);
1462 }
1463
1464 /* IPVS sync daemon data and function prototypes
1465 * (from ip_vs_sync.c)
1466 */
1467 int start_sync_thread(struct netns_ipvs *ipvs, struct ipvs_sync_daemon_cfg *cfg,
1468 int state);
1469 int stop_sync_thread(struct netns_ipvs *ipvs, int state);
1470 void ip_vs_sync_conn(struct netns_ipvs *ipvs, struct ip_vs_conn *cp, int pkts);
1471
1472 /* IPVS rate estimator prototypes (from ip_vs_est.c) */
1473 void ip_vs_start_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1474 void ip_vs_stop_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1475 void ip_vs_zero_estimator(struct ip_vs_stats *stats);
1476 void ip_vs_read_estimator(struct ip_vs_kstats *dst, struct ip_vs_stats *stats);
1477
1478 /* Various IPVS packet transmitters (from ip_vs_xmit.c) */
1479 int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1480 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1481 int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1482 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1483 int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1484 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1485 int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1486 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1487 int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1488 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1489 int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1490 struct ip_vs_protocol *pp, int offset,
1491 unsigned int hooknum, struct ip_vs_iphdr *iph);
1492 void ip_vs_dest_dst_rcu_free(struct rcu_head *head);
1493
1494 #ifdef CONFIG_IP_VS_IPV6
1495 int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1496 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1497 int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1498 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1499 int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1500 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1501 int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1502 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1503 int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1504 struct ip_vs_protocol *pp, int offset,
1505 unsigned int hooknum, struct ip_vs_iphdr *iph);
1506 #endif
1507
1508 #ifdef CONFIG_SYSCTL
1509 /* This is a simple mechanism to ignore packets when
1510 * we are loaded. Just set ip_vs_drop_rate to 'n' and
1511 * we start to drop 1/rate of the packets
1512 */
1513 static inline int ip_vs_todrop(struct netns_ipvs *ipvs)
1514 {
1515 if (!ipvs->drop_rate)
1516 return 0;
1517 if (--ipvs->drop_counter > 0)
1518 return 0;
1519 ipvs->drop_counter = ipvs->drop_rate;
1520 return 1;
1521 }
1522 #else
1523 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; }
1524 #endif
1525
1526 /* ip_vs_fwd_tag returns the forwarding tag of the connection */
1527 #define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK)
1528
1529 static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp)
1530 {
1531 char fwd;
1532
1533 switch (IP_VS_FWD_METHOD(cp)) {
1534 case IP_VS_CONN_F_MASQ:
1535 fwd = 'M'; break;
1536 case IP_VS_CONN_F_LOCALNODE:
1537 fwd = 'L'; break;
1538 case IP_VS_CONN_F_TUNNEL:
1539 fwd = 'T'; break;
1540 case IP_VS_CONN_F_DROUTE:
1541 fwd = 'R'; break;
1542 case IP_VS_CONN_F_BYPASS:
1543 fwd = 'B'; break;
1544 default:
1545 fwd = '?'; break;
1546 }
1547 return fwd;
1548 }
1549
1550 void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp,
1551 struct ip_vs_conn *cp, int dir);
1552
1553 #ifdef CONFIG_IP_VS_IPV6
1554 void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp,
1555 struct ip_vs_conn *cp, int dir);
1556 #endif
1557
1558 __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset);
1559
1560 static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum)
1561 {
1562 __be32 diff[2] = { ~old, new };
1563
1564 return csum_partial(diff, sizeof(diff), oldsum);
1565 }
1566
1567 #ifdef CONFIG_IP_VS_IPV6
1568 static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new,
1569 __wsum oldsum)
1570 {
1571 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0],
1572 new[3], new[2], new[1], new[0] };
1573
1574 return csum_partial(diff, sizeof(diff), oldsum);
1575 }
1576 #endif
1577
1578 static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum)
1579 {
1580 __be16 diff[2] = { ~old, new };
1581
1582 return csum_partial(diff, sizeof(diff), oldsum);
1583 }
1584
1585 /* Forget current conntrack (unconfirmed) and attach notrack entry */
1586 static inline void ip_vs_notrack(struct sk_buff *skb)
1587 {
1588 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1589 enum ip_conntrack_info ctinfo;
1590 struct nf_conn *ct = nf_ct_get(skb, &ctinfo);
1591
1592 if (!ct || !nf_ct_is_untracked(ct)) {
1593 nf_conntrack_put(skb->nfct);
1594 skb->nfct = &nf_ct_untracked_get()->ct_general;
1595 skb->nfctinfo = IP_CT_NEW;
1596 nf_conntrack_get(skb->nfct);
1597 }
1598 #endif
1599 }
1600
1601 #ifdef CONFIG_IP_VS_NFCT
1602 /* Netfilter connection tracking
1603 * (from ip_vs_nfct.c)
1604 */
1605 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1606 {
1607 #ifdef CONFIG_SYSCTL
1608 return ipvs->sysctl_conntrack;
1609 #else
1610 return 0;
1611 #endif
1612 }
1613
1614 void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp,
1615 int outin);
1616 int ip_vs_confirm_conntrack(struct sk_buff *skb);
1617 void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct,
1618 struct ip_vs_conn *cp, u_int8_t proto,
1619 const __be16 port, int from_rs);
1620 void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp);
1621
1622 #else
1623
1624 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1625 {
1626 return 0;
1627 }
1628
1629 static inline void ip_vs_update_conntrack(struct sk_buff *skb,
1630 struct ip_vs_conn *cp, int outin)
1631 {
1632 }
1633
1634 static inline int ip_vs_confirm_conntrack(struct sk_buff *skb)
1635 {
1636 return NF_ACCEPT;
1637 }
1638
1639 static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp)
1640 {
1641 }
1642 #endif /* CONFIG_IP_VS_NFCT */
1643
1644 static inline int
1645 ip_vs_dest_conn_overhead(struct ip_vs_dest *dest)
1646 {
1647 /* We think the overhead of processing active connections is 256
1648 * times higher than that of inactive connections in average. (This
1649 * 256 times might not be accurate, we will change it later) We
1650 * use the following formula to estimate the overhead now:
1651 * dest->activeconns*256 + dest->inactconns
1652 */
1653 return (atomic_read(&dest->activeconns) << 8) +
1654 atomic_read(&dest->inactconns);
1655 }
1656
1657 #endif /* _NET_IP_VS_H */
Ubuntu Artful kernel mirror