]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - include/net/ip_vs.h
projects / mirror_ubuntu-bionic-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
net/mlx5e: Rx, Fix checksum calculation for new hardware
[mirror_ubuntu-bionic-kernel.git] / include / net / ip_vs.h
1 /* SPDX-License-Identifier: GPL-2.0 */
2 /* IP Virtual Server
3 * data structure and functionality definitions
4 */
5
6 #ifndef _NET_IP_VS_H
7 #define _NET_IP_VS_H
8
9 #include <linux/ip_vs.h> /* definitions shared with userland */
10
11 #include <asm/types.h> /* for __uXX types */
12
13 #include <linux/list.h> /* for struct list_head */
14 #include <linux/spinlock.h> /* for struct rwlock_t */
15 #include <linux/atomic.h> /* for struct atomic_t */
16 #include <linux/refcount.h> /* for struct refcount_t */
17
18 #include <linux/compiler.h>
19 #include <linux/timer.h>
20 #include <linux/bug.h>
21
22 #include <net/checksum.h>
23 #include <linux/netfilter.h> /* for union nf_inet_addr */
24 #include <linux/ip.h>
25 #include <linux/ipv6.h> /* for struct ipv6hdr */
26 #include <net/ipv6.h>
27 #if IS_ENABLED(CONFIG_IP_VS_IPV6)
28 #include <linux/netfilter_ipv6/ip6_tables.h>
29 #endif
30 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
31 #include <net/netfilter/nf_conntrack.h>
32 #endif
33 #include <net/net_namespace.h> /* Netw namespace */
34
35 #define IP_VS_HDR_INVERSE 1
36 #define IP_VS_HDR_ICMP 2
37
38 /* Generic access of ipvs struct */
39 static inline struct netns_ipvs *net_ipvs(struct net* net)
40 {
41 return net->ipvs;
42 }
43
44 /* This one needed for single_open_net since net is stored directly in
45 * private not as a struct i.e. seq_file_net can't be used.
46 */
47 static inline struct net *seq_file_single_net(struct seq_file *seq)
48 {
49 #ifdef CONFIG_NET_NS
50 return (struct net *)seq->private;
51 #else
52 return &init_net;
53 #endif
54 }
55
56 /* Connections' size value needed by ip_vs_ctl.c */
57 extern int ip_vs_conn_tab_size;
58
59 struct ip_vs_iphdr {
60 int hdr_flags; /* ipvs flags */
61 __u32 off; /* Where IP or IPv4 header starts */
62 __u32 len; /* IPv4 simply where L4 starts
63 * IPv6 where L4 Transport Header starts */
64 __u16 fragoffs; /* IPv6 fragment offset, 0 if first frag (or not frag)*/
65 __s16 protocol;
66 __s32 flags;
67 union nf_inet_addr saddr;
68 union nf_inet_addr daddr;
69 };
70
71 static inline void *frag_safe_skb_hp(const struct sk_buff *skb, int offset,
72 int len, void *buffer,
73 const struct ip_vs_iphdr *ipvsh)
74 {
75 return skb_header_pointer(skb, offset, len, buffer);
76 }
77
78 /* This function handles filling *ip_vs_iphdr, both for IPv4 and IPv6.
79 * IPv6 requires some extra work, as finding proper header position,
80 * depend on the IPv6 extension headers.
81 */
82 static inline int
83 ip_vs_fill_iph_skb_off(int af, const struct sk_buff *skb, int offset,
84 int hdr_flags, struct ip_vs_iphdr *iphdr)
85 {
86 iphdr->hdr_flags = hdr_flags;
87 iphdr->off = offset;
88
89 #ifdef CONFIG_IP_VS_IPV6
90 if (af == AF_INET6) {
91 struct ipv6hdr _iph;
92 const struct ipv6hdr *iph = skb_header_pointer(
93 skb, offset, sizeof(_iph), &_iph);
94 if (!iph)
95 return 0;
96
97 iphdr->saddr.in6 = iph->saddr;
98 iphdr->daddr.in6 = iph->daddr;
99 /* ipv6_find_hdr() updates len, flags */
100 iphdr->len = offset;
101 iphdr->flags = 0;
102 iphdr->protocol = ipv6_find_hdr(skb, &iphdr->len, -1,
103 &iphdr->fragoffs,
104 &iphdr->flags);
105 if (iphdr->protocol < 0)
106 return 0;
107 } else
108 #endif
109 {
110 struct iphdr _iph;
111 const struct iphdr *iph = skb_header_pointer(
112 skb, offset, sizeof(_iph), &_iph);
113 if (!iph)
114 return 0;
115
116 iphdr->len = offset + iph->ihl * 4;
117 iphdr->fragoffs = 0;
118 iphdr->protocol = iph->protocol;
119 iphdr->saddr.ip = iph->saddr;
120 iphdr->daddr.ip = iph->daddr;
121 }
122
123 return 1;
124 }
125
126 static inline int
127 ip_vs_fill_iph_skb_icmp(int af, const struct sk_buff *skb, int offset,
128 bool inverse, struct ip_vs_iphdr *iphdr)
129 {
130 int hdr_flags = IP_VS_HDR_ICMP;
131
132 if (inverse)
133 hdr_flags |= IP_VS_HDR_INVERSE;
134
135 return ip_vs_fill_iph_skb_off(af, skb, offset, hdr_flags, iphdr);
136 }
137
138 static inline int
139 ip_vs_fill_iph_skb(int af, const struct sk_buff *skb, bool inverse,
140 struct ip_vs_iphdr *iphdr)
141 {
142 int hdr_flags = 0;
143
144 if (inverse)
145 hdr_flags |= IP_VS_HDR_INVERSE;
146
147 return ip_vs_fill_iph_skb_off(af, skb, skb_network_offset(skb),
148 hdr_flags, iphdr);
149 }
150
151 static inline bool
152 ip_vs_iph_inverse(const struct ip_vs_iphdr *iph)
153 {
154 return !!(iph->hdr_flags & IP_VS_HDR_INVERSE);
155 }
156
157 static inline bool
158 ip_vs_iph_icmp(const struct ip_vs_iphdr *iph)
159 {
160 return !!(iph->hdr_flags & IP_VS_HDR_ICMP);
161 }
162
163 static inline void ip_vs_addr_copy(int af, union nf_inet_addr *dst,
164 const union nf_inet_addr *src)
165 {
166 #ifdef CONFIG_IP_VS_IPV6
167 if (af == AF_INET6)
168 dst->in6 = src->in6;
169 else
170 #endif
171 dst->ip = src->ip;
172 }
173
174 static inline void ip_vs_addr_set(int af, union nf_inet_addr *dst,
175 const union nf_inet_addr *src)
176 {
177 #ifdef CONFIG_IP_VS_IPV6
178 if (af == AF_INET6) {
179 dst->in6 = src->in6;
180 return;
181 }
182 #endif
183 dst->ip = src->ip;
184 dst->all[1] = 0;
185 dst->all[2] = 0;
186 dst->all[3] = 0;
187 }
188
189 static inline int ip_vs_addr_equal(int af, const union nf_inet_addr *a,
190 const union nf_inet_addr *b)
191 {
192 #ifdef CONFIG_IP_VS_IPV6
193 if (af == AF_INET6)
194 return ipv6_addr_equal(&a->in6, &b->in6);
195 #endif
196 return a->ip == b->ip;
197 }
198
199 #ifdef CONFIG_IP_VS_DEBUG
200 #include <linux/net.h>
201
202 int ip_vs_get_debug_level(void);
203
204 static inline const char *ip_vs_dbg_addr(int af, char *buf, size_t buf_len,
205 const union nf_inet_addr *addr,
206 int *idx)
207 {
208 int len;
209 #ifdef CONFIG_IP_VS_IPV6
210 if (af == AF_INET6)
211 len = snprintf(&buf[*idx], buf_len - *idx, "[%pI6c]",
212 &addr->in6) + 1;
213 else
214 #endif
215 len = snprintf(&buf[*idx], buf_len - *idx, "%pI4",
216 &addr->ip) + 1;
217
218 *idx += len;
219 BUG_ON(*idx > buf_len + 1);
220 return &buf[*idx - len];
221 }
222
223 #define IP_VS_DBG_BUF(level, msg, ...) \
224 do { \
225 char ip_vs_dbg_buf[160]; \
226 int ip_vs_dbg_idx = 0; \
227 if (level <= ip_vs_get_debug_level()) \
228 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
229 } while (0)
230 #define IP_VS_ERR_BUF(msg...) \
231 do { \
232 char ip_vs_dbg_buf[160]; \
233 int ip_vs_dbg_idx = 0; \
234 pr_err(msg); \
235 } while (0)
236
237 /* Only use from within IP_VS_DBG_BUF() or IP_VS_ERR_BUF macros */
238 #define IP_VS_DBG_ADDR(af, addr) \
239 ip_vs_dbg_addr(af, ip_vs_dbg_buf, \
240 sizeof(ip_vs_dbg_buf), addr, \
241 &ip_vs_dbg_idx)
242
243 #define IP_VS_DBG(level, msg, ...) \
244 do { \
245 if (level <= ip_vs_get_debug_level()) \
246 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
247 } while (0)
248 #define IP_VS_DBG_RL(msg, ...) \
249 do { \
250 if (net_ratelimit()) \
251 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
252 } while (0)
253 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) \
254 do { \
255 if (level <= ip_vs_get_debug_level()) \
256 pp->debug_packet(af, pp, skb, ofs, msg); \
257 } while (0)
258 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) \
259 do { \
260 if (level <= ip_vs_get_debug_level() && \
261 net_ratelimit()) \
262 pp->debug_packet(af, pp, skb, ofs, msg); \
263 } while (0)
264 #else /* NO DEBUGGING at ALL */
265 #define IP_VS_DBG_BUF(level, msg...) do {} while (0)
266 #define IP_VS_ERR_BUF(msg...) do {} while (0)
267 #define IP_VS_DBG(level, msg...) do {} while (0)
268 #define IP_VS_DBG_RL(msg...) do {} while (0)
269 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
270 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
271 #endif
272
273 #define IP_VS_BUG() BUG()
274 #define IP_VS_ERR_RL(msg, ...) \
275 do { \
276 if (net_ratelimit()) \
277 pr_err(msg, ##__VA_ARGS__); \
278 } while (0)
279
280 #ifdef CONFIG_IP_VS_DEBUG
281 #define EnterFunction(level) \
282 do { \
283 if (level <= ip_vs_get_debug_level()) \
284 printk(KERN_DEBUG \
285 pr_fmt("Enter: %s, %s line %i\n"), \
286 __func__, __FILE__, __LINE__); \
287 } while (0)
288 #define LeaveFunction(level) \
289 do { \
290 if (level <= ip_vs_get_debug_level()) \
291 printk(KERN_DEBUG \
292 pr_fmt("Leave: %s, %s line %i\n"), \
293 __func__, __FILE__, __LINE__); \
294 } while (0)
295 #else
296 #define EnterFunction(level) do {} while (0)
297 #define LeaveFunction(level) do {} while (0)
298 #endif
299
300 /* The port number of FTP service (in network order). */
301 #define FTPPORT cpu_to_be16(21)
302 #define FTPDATA cpu_to_be16(20)
303
304 /* TCP State Values */
305 enum {
306 IP_VS_TCP_S_NONE = 0,
307 IP_VS_TCP_S_ESTABLISHED,
308 IP_VS_TCP_S_SYN_SENT,
309 IP_VS_TCP_S_SYN_RECV,
310 IP_VS_TCP_S_FIN_WAIT,
311 IP_VS_TCP_S_TIME_WAIT,
312 IP_VS_TCP_S_CLOSE,
313 IP_VS_TCP_S_CLOSE_WAIT,
314 IP_VS_TCP_S_LAST_ACK,
315 IP_VS_TCP_S_LISTEN,
316 IP_VS_TCP_S_SYNACK,
317 IP_VS_TCP_S_LAST
318 };
319
320 /* UDP State Values */
321 enum {
322 IP_VS_UDP_S_NORMAL,
323 IP_VS_UDP_S_LAST,
324 };
325
326 /* ICMP State Values */
327 enum {
328 IP_VS_ICMP_S_NORMAL,
329 IP_VS_ICMP_S_LAST,
330 };
331
332 /* SCTP State Values */
333 enum ip_vs_sctp_states {
334 IP_VS_SCTP_S_NONE,
335 IP_VS_SCTP_S_INIT1,
336 IP_VS_SCTP_S_INIT,
337 IP_VS_SCTP_S_COOKIE_SENT,
338 IP_VS_SCTP_S_COOKIE_REPLIED,
339 IP_VS_SCTP_S_COOKIE_WAIT,
340 IP_VS_SCTP_S_COOKIE,
341 IP_VS_SCTP_S_COOKIE_ECHOED,
342 IP_VS_SCTP_S_ESTABLISHED,
343 IP_VS_SCTP_S_SHUTDOWN_SENT,
344 IP_VS_SCTP_S_SHUTDOWN_RECEIVED,
345 IP_VS_SCTP_S_SHUTDOWN_ACK_SENT,
346 IP_VS_SCTP_S_REJECTED,
347 IP_VS_SCTP_S_CLOSED,
348 IP_VS_SCTP_S_LAST
349 };
350
351 /* Delta sequence info structure
352 * Each ip_vs_conn has 2 (output AND input seq. changes).
353 * Only used in the VS/NAT.
354 */
355 struct ip_vs_seq {
356 __u32 init_seq; /* Add delta from this seq */
357 __u32 delta; /* Delta in sequence numbers */
358 __u32 previous_delta; /* Delta in sequence numbers
359 * before last resized pkt */
360 };
361
362 /* counters per cpu */
363 struct ip_vs_counters {
364 __u64 conns; /* connections scheduled */
365 __u64 inpkts; /* incoming packets */
366 __u64 outpkts; /* outgoing packets */
367 __u64 inbytes; /* incoming bytes */
368 __u64 outbytes; /* outgoing bytes */
369 };
370 /* Stats per cpu */
371 struct ip_vs_cpu_stats {
372 struct ip_vs_counters cnt;
373 struct u64_stats_sync syncp;
374 };
375
376 /* IPVS statistics objects */
377 struct ip_vs_estimator {
378 struct list_head list;
379
380 u64 last_inbytes;
381 u64 last_outbytes;
382 u64 last_conns;
383 u64 last_inpkts;
384 u64 last_outpkts;
385
386 u64 cps;
387 u64 inpps;
388 u64 outpps;
389 u64 inbps;
390 u64 outbps;
391 };
392
393 /*
394 * IPVS statistics object, 64-bit kernel version of struct ip_vs_stats_user
395 */
396 struct ip_vs_kstats {
397 u64 conns; /* connections scheduled */
398 u64 inpkts; /* incoming packets */
399 u64 outpkts; /* outgoing packets */
400 u64 inbytes; /* incoming bytes */
401 u64 outbytes; /* outgoing bytes */
402
403 u64 cps; /* current connection rate */
404 u64 inpps; /* current in packet rate */
405 u64 outpps; /* current out packet rate */
406 u64 inbps; /* current in byte rate */
407 u64 outbps; /* current out byte rate */
408 };
409
410 struct ip_vs_stats {
411 struct ip_vs_kstats kstats; /* kernel statistics */
412 struct ip_vs_estimator est; /* estimator */
413 struct ip_vs_cpu_stats __percpu *cpustats; /* per cpu counters */
414 spinlock_t lock; /* spin lock */
415 struct ip_vs_kstats kstats0; /* reset values */
416 };
417
418 struct dst_entry;
419 struct iphdr;
420 struct ip_vs_conn;
421 struct ip_vs_app;
422 struct sk_buff;
423 struct ip_vs_proto_data;
424
425 struct ip_vs_protocol {
426 struct ip_vs_protocol *next;
427 char *name;
428 u16 protocol;
429 u16 num_states;
430 int dont_defrag;
431
432 void (*init)(struct ip_vs_protocol *pp);
433
434 void (*exit)(struct ip_vs_protocol *pp);
435
436 int (*init_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
437
438 void (*exit_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
439
440 int (*conn_schedule)(struct netns_ipvs *ipvs,
441 int af, struct sk_buff *skb,
442 struct ip_vs_proto_data *pd,
443 int *verdict, struct ip_vs_conn **cpp,
444 struct ip_vs_iphdr *iph);
445
446 struct ip_vs_conn *
447 (*conn_in_get)(struct netns_ipvs *ipvs,
448 int af,
449 const struct sk_buff *skb,
450 const struct ip_vs_iphdr *iph);
451
452 struct ip_vs_conn *
453 (*conn_out_get)(struct netns_ipvs *ipvs,
454 int af,
455 const struct sk_buff *skb,
456 const struct ip_vs_iphdr *iph);
457
458 int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
459 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
460
461 int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
462 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
463
464 int (*csum_check)(int af, struct sk_buff *skb,
465 struct ip_vs_protocol *pp);
466
467 const char *(*state_name)(int state);
468
469 void (*state_transition)(struct ip_vs_conn *cp, int direction,
470 const struct sk_buff *skb,
471 struct ip_vs_proto_data *pd);
472
473 int (*register_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
474
475 void (*unregister_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
476
477 int (*app_conn_bind)(struct ip_vs_conn *cp);
478
479 void (*debug_packet)(int af, struct ip_vs_protocol *pp,
480 const struct sk_buff *skb,
481 int offset,
482 const char *msg);
483
484 void (*timeout_change)(struct ip_vs_proto_data *pd, int flags);
485 };
486
487 /* protocol data per netns */
488 struct ip_vs_proto_data {
489 struct ip_vs_proto_data *next;
490 struct ip_vs_protocol *pp;
491 int *timeout_table; /* protocol timeout table */
492 atomic_t appcnt; /* counter of proto app incs. */
493 struct tcp_states_t *tcp_state_table;
494 };
495
496 struct ip_vs_protocol *ip_vs_proto_get(unsigned short proto);
497 struct ip_vs_proto_data *ip_vs_proto_data_get(struct netns_ipvs *ipvs,
498 unsigned short proto);
499
500 struct ip_vs_conn_param {
501 struct netns_ipvs *ipvs;
502 const union nf_inet_addr *caddr;
503 const union nf_inet_addr *vaddr;
504 __be16 cport;
505 __be16 vport;
506 __u16 protocol;
507 u16 af;
508
509 const struct ip_vs_pe *pe;
510 char *pe_data;
511 __u8 pe_data_len;
512 };
513
514 /* IP_VS structure allocated for each dynamically scheduled connection */
515 struct ip_vs_conn {
516 struct hlist_node c_list; /* hashed list heads */
517 /* Protocol, addresses and port numbers */
518 __be16 cport;
519 __be16 dport;
520 __be16 vport;
521 u16 af; /* address family */
522 union nf_inet_addr caddr; /* client address */
523 union nf_inet_addr vaddr; /* virtual address */
524 union nf_inet_addr daddr; /* destination address */
525 volatile __u32 flags; /* status flags */
526 __u16 protocol; /* Which protocol (TCP/UDP) */
527 __u16 daf; /* Address family of the dest */
528 struct netns_ipvs *ipvs;
529
530 /* counter and timer */
531 refcount_t refcnt; /* reference count */
532 struct timer_list timer; /* Expiration timer */
533 volatile unsigned long timeout; /* timeout */
534
535 /* Flags and state transition */
536 spinlock_t lock; /* lock for state transition */
537 volatile __u16 state; /* state info */
538 volatile __u16 old_state; /* old state, to be used for
539 * state transition triggerd
540 * synchronization
541 */
542 __u32 fwmark; /* Fire wall mark from skb */
543 unsigned long sync_endtime; /* jiffies + sent_retries */
544
545 /* Control members */
546 struct ip_vs_conn *control; /* Master control connection */
547 atomic_t n_control; /* Number of controlled ones */
548 struct ip_vs_dest *dest; /* real server */
549 atomic_t in_pkts; /* incoming packet counter */
550
551 /* Packet transmitter for different forwarding methods. If it
552 * mangles the packet, it must return NF_DROP or better NF_STOLEN,
553 * otherwise this must be changed to a sk_buff **.
554 * NF_ACCEPT can be returned when destination is local.
555 */
556 int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp,
557 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
558
559 /* Note: we can group the following members into a structure,
560 * in order to save more space, and the following members are
561 * only used in VS/NAT anyway
562 */
563 struct ip_vs_app *app; /* bound ip_vs_app object */
564 void *app_data; /* Application private data */
565 struct ip_vs_seq in_seq; /* incoming seq. struct */
566 struct ip_vs_seq out_seq; /* outgoing seq. struct */
567
568 const struct ip_vs_pe *pe;
569 char *pe_data;
570 __u8 pe_data_len;
571
572 struct rcu_head rcu_head;
573 };
574
575 /* Extended internal versions of struct ip_vs_service_user and ip_vs_dest_user
576 * for IPv6 support.
577 *
578 * We need these to conveniently pass around service and destination
579 * options, but unfortunately, we also need to keep the old definitions to
580 * maintain userspace backwards compatibility for the setsockopt interface.
581 */
582 struct ip_vs_service_user_kern {
583 /* virtual service addresses */
584 u16 af;
585 u16 protocol;
586 union nf_inet_addr addr; /* virtual ip address */
587 __be16 port;
588 u32 fwmark; /* firwall mark of service */
589
590 /* virtual service options */
591 char *sched_name;
592 char *pe_name;
593 unsigned int flags; /* virtual service flags */
594 unsigned int timeout; /* persistent timeout in sec */
595 __be32 netmask; /* persistent netmask or plen */
596 };
597
598
599 struct ip_vs_dest_user_kern {
600 /* destination server address */
601 union nf_inet_addr addr;
602 __be16 port;
603
604 /* real server options */
605 unsigned int conn_flags; /* connection flags */
606 int weight; /* destination weight */
607
608 /* thresholds for active connections */
609 u32 u_threshold; /* upper threshold */
610 u32 l_threshold; /* lower threshold */
611
612 /* Address family of addr */
613 u16 af;
614 };
615
616
617 /*
618 * The information about the virtual service offered to the net and the
619 * forwarding entries.
620 */
621 struct ip_vs_service {
622 struct hlist_node s_list; /* for normal service table */
623 struct hlist_node f_list; /* for fwmark-based service table */
624 atomic_t refcnt; /* reference counter */
625
626 u16 af; /* address family */
627 __u16 protocol; /* which protocol (TCP/UDP) */
628 union nf_inet_addr addr; /* IP address for virtual service */
629 __be16 port; /* port number for the service */
630 __u32 fwmark; /* firewall mark of the service */
631 unsigned int flags; /* service status flags */
632 unsigned int timeout; /* persistent timeout in ticks */
633 __be32 netmask; /* grouping granularity, mask/plen */
634 struct netns_ipvs *ipvs;
635
636 struct list_head destinations; /* real server d-linked list */
637 __u32 num_dests; /* number of servers */
638 struct ip_vs_stats stats; /* statistics for the service */
639
640 /* for scheduling */
641 struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */
642 spinlock_t sched_lock; /* lock sched_data */
643 void *sched_data; /* scheduler application data */
644
645 /* alternate persistence engine */
646 struct ip_vs_pe __rcu *pe;
647
648 struct rcu_head rcu_head;
649 };
650
651 /* Information for cached dst */
652 struct ip_vs_dest_dst {
653 struct dst_entry *dst_cache; /* destination cache entry */
654 u32 dst_cookie;
655 union nf_inet_addr dst_saddr;
656 struct rcu_head rcu_head;
657 };
658
659 /* The real server destination forwarding entry with ip address, port number,
660 * and so on.
661 */
662 struct ip_vs_dest {
663 struct list_head n_list; /* for the dests in the service */
664 struct hlist_node d_list; /* for table with all the dests */
665
666 u16 af; /* address family */
667 __be16 port; /* port number of the server */
668 union nf_inet_addr addr; /* IP address of the server */
669 volatile unsigned int flags; /* dest status flags */
670 atomic_t conn_flags; /* flags to copy to conn */
671 atomic_t weight; /* server weight */
672
673 refcount_t refcnt; /* reference counter */
674 struct ip_vs_stats stats; /* statistics */
675 unsigned long idle_start; /* start time, jiffies */
676
677 /* connection counters and thresholds */
678 atomic_t activeconns; /* active connections */
679 atomic_t inactconns; /* inactive connections */
680 atomic_t persistconns; /* persistent connections */
681 __u32 u_threshold; /* upper threshold */
682 __u32 l_threshold; /* lower threshold */
683
684 /* for destination cache */
685 spinlock_t dst_lock; /* lock of dst_cache */
686 struct ip_vs_dest_dst __rcu *dest_dst; /* cached dst info */
687
688 /* for virtual service */
689 struct ip_vs_service __rcu *svc; /* service it belongs to */
690 __u16 protocol; /* which protocol (TCP/UDP) */
691 __be16 vport; /* virtual port number */
692 union nf_inet_addr vaddr; /* virtual IP address */
693 __u32 vfwmark; /* firewall mark of service */
694
695 struct list_head t_list; /* in dest_trash */
696 unsigned int in_rs_table:1; /* we are in rs_table */
697 };
698
699 /* The scheduler object */
700 struct ip_vs_scheduler {
701 struct list_head n_list; /* d-linked list head */
702 char *name; /* scheduler name */
703 atomic_t refcnt; /* reference counter */
704 struct module *module; /* THIS_MODULE/NULL */
705
706 /* scheduler initializing service */
707 int (*init_service)(struct ip_vs_service *svc);
708 /* scheduling service finish */
709 void (*done_service)(struct ip_vs_service *svc);
710 /* dest is linked */
711 int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
712 /* dest is unlinked */
713 int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
714 /* dest is updated */
715 int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
716
717 /* selecting a server from the given service */
718 struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc,
719 const struct sk_buff *skb,
720 struct ip_vs_iphdr *iph);
721 };
722
723 /* The persistence engine object */
724 struct ip_vs_pe {
725 struct list_head n_list; /* d-linked list head */
726 char *name; /* scheduler name */
727 atomic_t refcnt; /* reference counter */
728 struct module *module; /* THIS_MODULE/NULL */
729
730 /* get the connection template, if any */
731 int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb);
732 bool (*ct_match)(const struct ip_vs_conn_param *p,
733 struct ip_vs_conn *ct);
734 u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval,
735 bool inverse);
736 int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf);
737 /* create connections for real-server outgoing packets */
738 struct ip_vs_conn* (*conn_out)(struct ip_vs_service *svc,
739 struct ip_vs_dest *dest,
740 struct sk_buff *skb,
741 const struct ip_vs_iphdr *iph,
742 __be16 dport, __be16 cport);
743 };
744
745 /* The application module object (a.k.a. app incarnation) */
746 struct ip_vs_app {
747 struct list_head a_list; /* member in app list */
748 int type; /* IP_VS_APP_TYPE_xxx */
749 char *name; /* application module name */
750 __u16 protocol;
751 struct module *module; /* THIS_MODULE/NULL */
752 struct list_head incs_list; /* list of incarnations */
753
754 /* members for application incarnations */
755 struct list_head p_list; /* member in proto app list */
756 struct ip_vs_app *app; /* its real application */
757 __be16 port; /* port number in net order */
758 atomic_t usecnt; /* usage counter */
759 struct rcu_head rcu_head;
760
761 /* output hook: Process packet in inout direction, diff set for TCP.
762 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
763 * 2=Mangled but checksum was not updated
764 */
765 int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *,
766 struct sk_buff *, int *diff);
767
768 /* input hook: Process packet in outin direction, diff set for TCP.
769 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
770 * 2=Mangled but checksum was not updated
771 */
772 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *,
773 struct sk_buff *, int *diff);
774
775 /* ip_vs_app initializer */
776 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *);
777
778 /* ip_vs_app finish */
779 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *);
780
781
782 /* not used now */
783 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *,
784 struct ip_vs_protocol *);
785
786 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *);
787
788 int * timeout_table;
789 int * timeouts;
790 int timeouts_size;
791
792 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app,
793 int *verdict, struct ip_vs_conn **cpp);
794
795 struct ip_vs_conn *
796 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app,
797 const struct iphdr *iph, int inverse);
798
799 struct ip_vs_conn *
800 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app,
801 const struct iphdr *iph, int inverse);
802
803 int (*state_transition)(struct ip_vs_conn *cp, int direction,
804 const struct sk_buff *skb,
805 struct ip_vs_app *app);
806
807 void (*timeout_change)(struct ip_vs_app *app, int flags);
808 };
809
810 struct ipvs_master_sync_state {
811 struct list_head sync_queue;
812 struct ip_vs_sync_buff *sync_buff;
813 unsigned long sync_queue_len;
814 unsigned int sync_queue_delay;
815 struct delayed_work master_wakeup_work;
816 struct netns_ipvs *ipvs;
817 };
818
819 struct ip_vs_sync_thread_data;
820
821 /* How much time to keep dests in trash */
822 #define IP_VS_DEST_TRASH_PERIOD (120 * HZ)
823
824 struct ipvs_sync_daemon_cfg {
825 union nf_inet_addr mcast_group;
826 int syncid;
827 u16 sync_maxlen;
828 u16 mcast_port;
829 u8 mcast_af;
830 u8 mcast_ttl;
831 /* multicast interface name */
832 char mcast_ifn[IP_VS_IFNAME_MAXLEN];
833 };
834
835 /* IPVS in network namespace */
836 struct netns_ipvs {
837 int gen; /* Generation */
838 int enable; /* enable like nf_hooks do */
839 /* Hash table: for real service lookups */
840 #define IP_VS_RTAB_BITS 4
841 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
842 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
843
844 struct hlist_head rs_table[IP_VS_RTAB_SIZE];
845 /* ip_vs_app */
846 struct list_head app_list;
847 /* ip_vs_proto */
848 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */
849 struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE];
850 /* ip_vs_proto_tcp */
851 #ifdef CONFIG_IP_VS_PROTO_TCP
852 #define TCP_APP_TAB_BITS 4
853 #define TCP_APP_TAB_SIZE (1 << TCP_APP_TAB_BITS)
854 #define TCP_APP_TAB_MASK (TCP_APP_TAB_SIZE - 1)
855 struct list_head tcp_apps[TCP_APP_TAB_SIZE];
856 #endif
857 /* ip_vs_proto_udp */
858 #ifdef CONFIG_IP_VS_PROTO_UDP
859 #define UDP_APP_TAB_BITS 4
860 #define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS)
861 #define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1)
862 struct list_head udp_apps[UDP_APP_TAB_SIZE];
863 #endif
864 /* ip_vs_proto_sctp */
865 #ifdef CONFIG_IP_VS_PROTO_SCTP
866 #define SCTP_APP_TAB_BITS 4
867 #define SCTP_APP_TAB_SIZE (1 << SCTP_APP_TAB_BITS)
868 #define SCTP_APP_TAB_MASK (SCTP_APP_TAB_SIZE - 1)
869 /* Hash table for SCTP application incarnations */
870 struct list_head sctp_apps[SCTP_APP_TAB_SIZE];
871 #endif
872 /* ip_vs_conn */
873 atomic_t conn_count; /* connection counter */
874
875 /* ip_vs_ctl */
876 struct ip_vs_stats tot_stats; /* Statistics & est. */
877
878 int num_services; /* no of virtual services */
879
880 /* Trash for destinations */
881 struct list_head dest_trash;
882 spinlock_t dest_trash_lock;
883 struct timer_list dest_trash_timer; /* expiration timer */
884 /* Service counters */
885 atomic_t ftpsvc_counter;
886 atomic_t nullsvc_counter;
887 atomic_t conn_out_counter;
888
889 #ifdef CONFIG_SYSCTL
890 /* 1/rate drop and drop-entry variables */
891 struct delayed_work defense_work; /* Work handler */
892 int drop_rate;
893 int drop_counter;
894 int old_secure_tcp;
895 atomic_t dropentry;
896 /* locks in ctl.c */
897 spinlock_t dropentry_lock; /* drop entry handling */
898 spinlock_t droppacket_lock; /* drop packet handling */
899 spinlock_t securetcp_lock; /* state and timeout tables */
900
901 /* sys-ctl struct */
902 struct ctl_table_header *sysctl_hdr;
903 struct ctl_table *sysctl_tbl;
904 #endif
905
906 /* sysctl variables */
907 int sysctl_amemthresh;
908 int sysctl_am_droprate;
909 int sysctl_drop_entry;
910 int sysctl_drop_packet;
911 int sysctl_secure_tcp;
912 #ifdef CONFIG_IP_VS_NFCT
913 int sysctl_conntrack;
914 #endif
915 int sysctl_snat_reroute;
916 int sysctl_sync_ver;
917 int sysctl_sync_ports;
918 int sysctl_sync_persist_mode;
919 unsigned long sysctl_sync_qlen_max;
920 int sysctl_sync_sock_size;
921 int sysctl_cache_bypass;
922 int sysctl_expire_nodest_conn;
923 int sysctl_sloppy_tcp;
924 int sysctl_sloppy_sctp;
925 int sysctl_expire_quiescent_template;
926 int sysctl_sync_threshold[2];
927 unsigned int sysctl_sync_refresh_period;
928 int sysctl_sync_retries;
929 int sysctl_nat_icmp_send;
930 int sysctl_pmtu_disc;
931 int sysctl_backup_only;
932 int sysctl_conn_reuse_mode;
933 int sysctl_schedule_icmp;
934 int sysctl_ignore_tunneled;
935
936 /* ip_vs_lblc */
937 int sysctl_lblc_expiration;
938 struct ctl_table_header *lblc_ctl_header;
939 struct ctl_table *lblc_ctl_table;
940 /* ip_vs_lblcr */
941 int sysctl_lblcr_expiration;
942 struct ctl_table_header *lblcr_ctl_header;
943 struct ctl_table *lblcr_ctl_table;
944 /* ip_vs_est */
945 struct list_head est_list; /* estimator list */
946 spinlock_t est_lock;
947 struct timer_list est_timer; /* Estimation timer */
948 /* ip_vs_sync */
949 spinlock_t sync_lock;
950 struct ipvs_master_sync_state *ms;
951 spinlock_t sync_buff_lock;
952 struct ip_vs_sync_thread_data *master_tinfo;
953 struct ip_vs_sync_thread_data *backup_tinfo;
954 int threads_mask;
955 volatile int sync_state;
956 struct mutex sync_mutex;
957 struct ipvs_sync_daemon_cfg mcfg; /* Master Configuration */
958 struct ipvs_sync_daemon_cfg bcfg; /* Backup Configuration */
959 /* net name space ptr */
960 struct net *net; /* Needed by timer routines */
961 /* Number of heterogeneous destinations, needed becaus heterogeneous
962 * are not supported when synchronization is enabled.
963 */
964 unsigned int mixed_address_family_dests;
965 };
966
967 #define DEFAULT_SYNC_THRESHOLD 3
968 #define DEFAULT_SYNC_PERIOD 50
969 #define DEFAULT_SYNC_VER 1
970 #define DEFAULT_SLOPPY_TCP 0
971 #define DEFAULT_SLOPPY_SCTP 0
972 #define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ)
973 #define DEFAULT_SYNC_RETRIES 0
974 #define IPVS_SYNC_WAKEUP_RATE 8
975 #define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4)
976 #define IPVS_SYNC_SEND_DELAY (HZ / 50)
977 #define IPVS_SYNC_CHECK_PERIOD HZ
978 #define IPVS_SYNC_FLUSH_TIME (HZ * 2)
979 #define IPVS_SYNC_PORTS_MAX (1 << 6)
980
981 #ifdef CONFIG_SYSCTL
982
983 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
984 {
985 return ipvs->sysctl_sync_threshold[0];
986 }
987
988 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
989 {
990 return READ_ONCE(ipvs->sysctl_sync_threshold[1]);
991 }
992
993 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
994 {
995 return READ_ONCE(ipvs->sysctl_sync_refresh_period);
996 }
997
998 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
999 {
1000 return ipvs->sysctl_sync_retries;
1001 }
1002
1003 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1004 {
1005 return ipvs->sysctl_sync_ver;
1006 }
1007
1008 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1009 {
1010 return ipvs->sysctl_sloppy_tcp;
1011 }
1012
1013 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1014 {
1015 return ipvs->sysctl_sloppy_sctp;
1016 }
1017
1018 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1019 {
1020 return READ_ONCE(ipvs->sysctl_sync_ports);
1021 }
1022
1023 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1024 {
1025 return ipvs->sysctl_sync_persist_mode;
1026 }
1027
1028 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1029 {
1030 return ipvs->sysctl_sync_qlen_max;
1031 }
1032
1033 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1034 {
1035 return ipvs->sysctl_sync_sock_size;
1036 }
1037
1038 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1039 {
1040 return ipvs->sysctl_pmtu_disc;
1041 }
1042
1043 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1044 {
1045 return ipvs->sync_state & IP_VS_STATE_BACKUP &&
1046 ipvs->sysctl_backup_only;
1047 }
1048
1049 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1050 {
1051 return ipvs->sysctl_conn_reuse_mode;
1052 }
1053
1054 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1055 {
1056 return ipvs->sysctl_schedule_icmp;
1057 }
1058
1059 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1060 {
1061 return ipvs->sysctl_ignore_tunneled;
1062 }
1063
1064 static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1065 {
1066 return ipvs->sysctl_cache_bypass;
1067 }
1068
1069 #else
1070
1071 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1072 {
1073 return DEFAULT_SYNC_THRESHOLD;
1074 }
1075
1076 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1077 {
1078 return DEFAULT_SYNC_PERIOD;
1079 }
1080
1081 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1082 {
1083 return DEFAULT_SYNC_REFRESH_PERIOD;
1084 }
1085
1086 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1087 {
1088 return DEFAULT_SYNC_RETRIES & 3;
1089 }
1090
1091 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1092 {
1093 return DEFAULT_SYNC_VER;
1094 }
1095
1096 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1097 {
1098 return DEFAULT_SLOPPY_TCP;
1099 }
1100
1101 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1102 {
1103 return DEFAULT_SLOPPY_SCTP;
1104 }
1105
1106 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1107 {
1108 return 1;
1109 }
1110
1111 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1112 {
1113 return 0;
1114 }
1115
1116 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1117 {
1118 return IPVS_SYNC_QLEN_MAX;
1119 }
1120
1121 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1122 {
1123 return 0;
1124 }
1125
1126 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1127 {
1128 return 1;
1129 }
1130
1131 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1132 {
1133 return 0;
1134 }
1135
1136 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1137 {
1138 return 1;
1139 }
1140
1141 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1142 {
1143 return 0;
1144 }
1145
1146 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1147 {
1148 return 0;
1149 }
1150
1151 static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1152 {
1153 return 0;
1154 }
1155
1156 #endif
1157
1158 /* IPVS core functions
1159 * (from ip_vs_core.c)
1160 */
1161 const char *ip_vs_proto_name(unsigned int proto);
1162 void ip_vs_init_hash_table(struct list_head *table, int rows);
1163 struct ip_vs_conn *ip_vs_new_conn_out(struct ip_vs_service *svc,
1164 struct ip_vs_dest *dest,
1165 struct sk_buff *skb,
1166 const struct ip_vs_iphdr *iph,
1167 __be16 dport,
1168 __be16 cport);
1169 #define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t)))
1170
1171 #define IP_VS_APP_TYPE_FTP 1
1172
1173 /* ip_vs_conn handling functions
1174 * (from ip_vs_conn.c)
1175 */
1176 enum {
1177 IP_VS_DIR_INPUT = 0,
1178 IP_VS_DIR_OUTPUT,
1179 IP_VS_DIR_INPUT_ONLY,
1180 IP_VS_DIR_LAST,
1181 };
1182
1183 static inline void ip_vs_conn_fill_param(struct netns_ipvs *ipvs, int af, int protocol,
1184 const union nf_inet_addr *caddr,
1185 __be16 cport,
1186 const union nf_inet_addr *vaddr,
1187 __be16 vport,
1188 struct ip_vs_conn_param *p)
1189 {
1190 p->ipvs = ipvs;
1191 p->af = af;
1192 p->protocol = protocol;
1193 p->caddr = caddr;
1194 p->cport = cport;
1195 p->vaddr = vaddr;
1196 p->vport = vport;
1197 p->pe = NULL;
1198 p->pe_data = NULL;
1199 }
1200
1201 struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p);
1202 struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p);
1203
1204 struct ip_vs_conn * ip_vs_conn_in_get_proto(struct netns_ipvs *ipvs, int af,
1205 const struct sk_buff *skb,
1206 const struct ip_vs_iphdr *iph);
1207
1208 struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p);
1209
1210 struct ip_vs_conn * ip_vs_conn_out_get_proto(struct netns_ipvs *ipvs, int af,
1211 const struct sk_buff *skb,
1212 const struct ip_vs_iphdr *iph);
1213
1214 /* Get reference to gain full access to conn.
1215 * By default, RCU read-side critical sections have access only to
1216 * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference.
1217 */
1218 static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp)
1219 {
1220 return refcount_inc_not_zero(&cp->refcnt);
1221 }
1222
1223 /* put back the conn without restarting its timer */
1224 static inline void __ip_vs_conn_put(struct ip_vs_conn *cp)
1225 {
1226 smp_mb__before_atomic();
1227 refcount_dec(&cp->refcnt);
1228 }
1229 void ip_vs_conn_put(struct ip_vs_conn *cp);
1230 void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport);
1231
1232 struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, int dest_af,
1233 const union nf_inet_addr *daddr,
1234 __be16 dport, unsigned int flags,
1235 struct ip_vs_dest *dest, __u32 fwmark);
1236 void ip_vs_conn_expire_now(struct ip_vs_conn *cp);
1237
1238 const char *ip_vs_state_name(__u16 proto, int state);
1239
1240 void ip_vs_tcp_conn_listen(struct ip_vs_conn *cp);
1241 int ip_vs_check_template(struct ip_vs_conn *ct, struct ip_vs_dest *cdest);
1242 void ip_vs_random_dropentry(struct netns_ipvs *ipvs);
1243 int ip_vs_conn_init(void);
1244 void ip_vs_conn_cleanup(void);
1245
1246 static inline void ip_vs_control_del(struct ip_vs_conn *cp)
1247 {
1248 struct ip_vs_conn *ctl_cp = cp->control;
1249 if (!ctl_cp) {
1250 IP_VS_ERR_BUF("request control DEL for uncontrolled: "
1251 "%s:%d to %s:%d\n",
1252 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1253 ntohs(cp->cport),
1254 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1255 ntohs(cp->vport));
1256
1257 return;
1258 }
1259
1260 IP_VS_DBG_BUF(7, "DELeting control for: "
1261 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1262 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1263 ntohs(cp->cport),
1264 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1265 ntohs(ctl_cp->cport));
1266
1267 cp->control = NULL;
1268 if (atomic_read(&ctl_cp->n_control) == 0) {
1269 IP_VS_ERR_BUF("BUG control DEL with n=0 : "
1270 "%s:%d to %s:%d\n",
1271 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1272 ntohs(cp->cport),
1273 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1274 ntohs(cp->vport));
1275
1276 return;
1277 }
1278 atomic_dec(&ctl_cp->n_control);
1279 }
1280
1281 static inline void
1282 ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp)
1283 {
1284 if (cp->control) {
1285 IP_VS_ERR_BUF("request control ADD for already controlled: "
1286 "%s:%d to %s:%d\n",
1287 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1288 ntohs(cp->cport),
1289 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1290 ntohs(cp->vport));
1291
1292 ip_vs_control_del(cp);
1293 }
1294
1295 IP_VS_DBG_BUF(7, "ADDing control for: "
1296 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1297 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1298 ntohs(cp->cport),
1299 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1300 ntohs(ctl_cp->cport));
1301
1302 cp->control = ctl_cp;
1303 atomic_inc(&ctl_cp->n_control);
1304 }
1305
1306 /* IPVS netns init & cleanup functions */
1307 int ip_vs_estimator_net_init(struct netns_ipvs *ipvs);
1308 int ip_vs_control_net_init(struct netns_ipvs *ipvs);
1309 int ip_vs_protocol_net_init(struct netns_ipvs *ipvs);
1310 int ip_vs_app_net_init(struct netns_ipvs *ipvs);
1311 int ip_vs_conn_net_init(struct netns_ipvs *ipvs);
1312 int ip_vs_sync_net_init(struct netns_ipvs *ipvs);
1313 void ip_vs_conn_net_cleanup(struct netns_ipvs *ipvs);
1314 void ip_vs_app_net_cleanup(struct netns_ipvs *ipvs);
1315 void ip_vs_protocol_net_cleanup(struct netns_ipvs *ipvs);
1316 void ip_vs_control_net_cleanup(struct netns_ipvs *ipvs);
1317 void ip_vs_estimator_net_cleanup(struct netns_ipvs *ipvs);
1318 void ip_vs_sync_net_cleanup(struct netns_ipvs *ipvs);
1319 void ip_vs_service_net_cleanup(struct netns_ipvs *ipvs);
1320
1321 /* IPVS application functions
1322 * (from ip_vs_app.c)
1323 */
1324 #define IP_VS_APP_MAX_PORTS 8
1325 struct ip_vs_app *register_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1326 void unregister_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1327 int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1328 void ip_vs_unbind_app(struct ip_vs_conn *cp);
1329 int register_ip_vs_app_inc(struct netns_ipvs *ipvs, struct ip_vs_app *app, __u16 proto,
1330 __u16 port);
1331 int ip_vs_app_inc_get(struct ip_vs_app *inc);
1332 void ip_vs_app_inc_put(struct ip_vs_app *inc);
1333
1334 int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb);
1335 int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb);
1336
1337 int register_ip_vs_pe(struct ip_vs_pe *pe);
1338 int unregister_ip_vs_pe(struct ip_vs_pe *pe);
1339 struct ip_vs_pe *ip_vs_pe_getbyname(const char *name);
1340 struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name);
1341
1342 /* Use a #define to avoid all of module.h just for these trivial ops */
1343 #define ip_vs_pe_get(pe) \
1344 if (pe && pe->module) \
1345 __module_get(pe->module);
1346
1347 #define ip_vs_pe_put(pe) \
1348 if (pe && pe->module) \
1349 module_put(pe->module);
1350
1351 /* IPVS protocol functions (from ip_vs_proto.c) */
1352 int ip_vs_protocol_init(void);
1353 void ip_vs_protocol_cleanup(void);
1354 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags);
1355 int *ip_vs_create_timeout_table(int *table, int size);
1356 void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
1357 const struct sk_buff *skb, int offset,
1358 const char *msg);
1359
1360 extern struct ip_vs_protocol ip_vs_protocol_tcp;
1361 extern struct ip_vs_protocol ip_vs_protocol_udp;
1362 extern struct ip_vs_protocol ip_vs_protocol_icmp;
1363 extern struct ip_vs_protocol ip_vs_protocol_esp;
1364 extern struct ip_vs_protocol ip_vs_protocol_ah;
1365 extern struct ip_vs_protocol ip_vs_protocol_sctp;
1366
1367 /* Registering/unregistering scheduler functions
1368 * (from ip_vs_sched.c)
1369 */
1370 int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1371 int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1372 int ip_vs_bind_scheduler(struct ip_vs_service *svc,
1373 struct ip_vs_scheduler *scheduler);
1374 void ip_vs_unbind_scheduler(struct ip_vs_service *svc,
1375 struct ip_vs_scheduler *sched);
1376 struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name);
1377 void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler);
1378 struct ip_vs_conn *
1379 ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb,
1380 struct ip_vs_proto_data *pd, int *ignored,
1381 struct ip_vs_iphdr *iph);
1382 int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb,
1383 struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph);
1384
1385 void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg);
1386
1387 /* IPVS control data and functions (from ip_vs_ctl.c) */
1388 extern struct ip_vs_stats ip_vs_stats;
1389 extern int sysctl_ip_vs_sync_ver;
1390
1391 struct ip_vs_service *
1392 ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
1393 const union nf_inet_addr *vaddr, __be16 vport);
1394
1395 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1396 const union nf_inet_addr *daddr, __be16 dport);
1397
1398 struct ip_vs_dest *
1399 ip_vs_find_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1400 const union nf_inet_addr *daddr, __be16 dport);
1401
1402 int ip_vs_use_count_inc(void);
1403 void ip_vs_use_count_dec(void);
1404 int ip_vs_register_nl_ioctl(void);
1405 void ip_vs_unregister_nl_ioctl(void);
1406 int ip_vs_control_init(void);
1407 void ip_vs_control_cleanup(void);
1408 struct ip_vs_dest *
1409 ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
1410 const union nf_inet_addr *daddr, __be16 dport,
1411 const union nf_inet_addr *vaddr, __be16 vport,
1412 __u16 protocol, __u32 fwmark, __u32 flags);
1413 void ip_vs_try_bind_dest(struct ip_vs_conn *cp);
1414
1415 static inline void ip_vs_dest_hold(struct ip_vs_dest *dest)
1416 {
1417 refcount_inc(&dest->refcnt);
1418 }
1419
1420 static inline void ip_vs_dest_put(struct ip_vs_dest *dest)
1421 {
1422 smp_mb__before_atomic();
1423 refcount_dec(&dest->refcnt);
1424 }
1425
1426 static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest)
1427 {
1428 if (refcount_dec_and_test(&dest->refcnt))
1429 kfree(dest);
1430 }
1431
1432 /* IPVS sync daemon data and function prototypes
1433 * (from ip_vs_sync.c)
1434 */
1435 int start_sync_thread(struct netns_ipvs *ipvs, struct ipvs_sync_daemon_cfg *cfg,
1436 int state);
1437 int stop_sync_thread(struct netns_ipvs *ipvs, int state);
1438 void ip_vs_sync_conn(struct netns_ipvs *ipvs, struct ip_vs_conn *cp, int pkts);
1439
1440 /* IPVS rate estimator prototypes (from ip_vs_est.c) */
1441 void ip_vs_start_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1442 void ip_vs_stop_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1443 void ip_vs_zero_estimator(struct ip_vs_stats *stats);
1444 void ip_vs_read_estimator(struct ip_vs_kstats *dst, struct ip_vs_stats *stats);
1445
1446 /* Various IPVS packet transmitters (from ip_vs_xmit.c) */
1447 int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1448 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1449 int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1450 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1451 int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1452 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1453 int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1454 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1455 int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1456 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1457 int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1458 struct ip_vs_protocol *pp, int offset,
1459 unsigned int hooknum, struct ip_vs_iphdr *iph);
1460 void ip_vs_dest_dst_rcu_free(struct rcu_head *head);
1461
1462 #ifdef CONFIG_IP_VS_IPV6
1463 int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1464 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1465 int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1466 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1467 int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1468 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1469 int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1470 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1471 int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1472 struct ip_vs_protocol *pp, int offset,
1473 unsigned int hooknum, struct ip_vs_iphdr *iph);
1474 #endif
1475
1476 #ifdef CONFIG_SYSCTL
1477 /* This is a simple mechanism to ignore packets when
1478 * we are loaded. Just set ip_vs_drop_rate to 'n' and
1479 * we start to drop 1/rate of the packets
1480 */
1481 static inline int ip_vs_todrop(struct netns_ipvs *ipvs)
1482 {
1483 if (!ipvs->drop_rate)
1484 return 0;
1485 if (--ipvs->drop_counter > 0)
1486 return 0;
1487 ipvs->drop_counter = ipvs->drop_rate;
1488 return 1;
1489 }
1490 #else
1491 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; }
1492 #endif
1493
1494 /* ip_vs_fwd_tag returns the forwarding tag of the connection */
1495 #define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK)
1496
1497 static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp)
1498 {
1499 char fwd;
1500
1501 switch (IP_VS_FWD_METHOD(cp)) {
1502 case IP_VS_CONN_F_MASQ:
1503 fwd = 'M'; break;
1504 case IP_VS_CONN_F_LOCALNODE:
1505 fwd = 'L'; break;
1506 case IP_VS_CONN_F_TUNNEL:
1507 fwd = 'T'; break;
1508 case IP_VS_CONN_F_DROUTE:
1509 fwd = 'R'; break;
1510 case IP_VS_CONN_F_BYPASS:
1511 fwd = 'B'; break;
1512 default:
1513 fwd = '?'; break;
1514 }
1515 return fwd;
1516 }
1517
1518 void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp,
1519 struct ip_vs_conn *cp, int dir);
1520
1521 #ifdef CONFIG_IP_VS_IPV6
1522 void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp,
1523 struct ip_vs_conn *cp, int dir);
1524 #endif
1525
1526 __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset);
1527
1528 static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum)
1529 {
1530 __be32 diff[2] = { ~old, new };
1531
1532 return csum_partial(diff, sizeof(diff), oldsum);
1533 }
1534
1535 #ifdef CONFIG_IP_VS_IPV6
1536 static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new,
1537 __wsum oldsum)
1538 {
1539 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0],
1540 new[3], new[2], new[1], new[0] };
1541
1542 return csum_partial(diff, sizeof(diff), oldsum);
1543 }
1544 #endif
1545
1546 static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum)
1547 {
1548 __be16 diff[2] = { ~old, new };
1549
1550 return csum_partial(diff, sizeof(diff), oldsum);
1551 }
1552
1553 /* Forget current conntrack (unconfirmed) and attach notrack entry */
1554 static inline void ip_vs_notrack(struct sk_buff *skb)
1555 {
1556 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1557 enum ip_conntrack_info ctinfo;
1558 struct nf_conn *ct = nf_ct_get(skb, &ctinfo);
1559
1560 if (ct) {
1561 nf_conntrack_put(&ct->ct_general);
1562 nf_ct_set(skb, NULL, IP_CT_UNTRACKED);
1563 }
1564 #endif
1565 }
1566
1567 #ifdef CONFIG_IP_VS_NFCT
1568 /* Netfilter connection tracking
1569 * (from ip_vs_nfct.c)
1570 */
1571 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1572 {
1573 #ifdef CONFIG_SYSCTL
1574 return ipvs->sysctl_conntrack;
1575 #else
1576 return 0;
1577 #endif
1578 }
1579
1580 void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp,
1581 int outin);
1582 int ip_vs_confirm_conntrack(struct sk_buff *skb);
1583 void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct,
1584 struct ip_vs_conn *cp, u_int8_t proto,
1585 const __be16 port, int from_rs);
1586 void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp);
1587
1588 #else
1589
1590 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1591 {
1592 return 0;
1593 }
1594
1595 static inline void ip_vs_update_conntrack(struct sk_buff *skb,
1596 struct ip_vs_conn *cp, int outin)
1597 {
1598 }
1599
1600 static inline int ip_vs_confirm_conntrack(struct sk_buff *skb)
1601 {
1602 return NF_ACCEPT;
1603 }
1604
1605 static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp)
1606 {
1607 }
1608 #endif /* CONFIG_IP_VS_NFCT */
1609
1610 /* Really using conntrack? */
1611 static inline bool ip_vs_conn_uses_conntrack(struct ip_vs_conn *cp,
1612 struct sk_buff *skb)
1613 {
1614 #ifdef CONFIG_IP_VS_NFCT
1615 enum ip_conntrack_info ctinfo;
1616 struct nf_conn *ct;
1617
1618 if (!(cp->flags & IP_VS_CONN_F_NFCT))
1619 return false;
1620 ct = nf_ct_get(skb, &ctinfo);
1621 if (ct)
1622 return true;
1623 #endif
1624 return false;
1625 }
1626
1627 static inline int
1628 ip_vs_dest_conn_overhead(struct ip_vs_dest *dest)
1629 {
1630 /* We think the overhead of processing active connections is 256
1631 * times higher than that of inactive connections in average. (This
1632 * 256 times might not be accurate, we will change it later) We
1633 * use the following formula to estimate the overhead now:
1634 * dest->activeconns*256 + dest->inactconns
1635 */
1636 return (atomic_read(&dest->activeconns) << 8) +
1637 atomic_read(&dest->inactconns);
1638 }
1639
1640 #endif /* _NET_IP_VS_H */
ubuntu-bionic-kernel mirror