]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - include/net/ip_vs.h
projects / mirror_ubuntu-jammy-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge tag 'nfsd-5.1' of git://linux-nfs.org/~bfields/linux
[mirror_ubuntu-jammy-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 /* Connections' size value needed by ip_vs_ctl.c */
45 extern int ip_vs_conn_tab_size;
46
47 struct ip_vs_iphdr {
48 int hdr_flags; /* ipvs flags */
49 __u32 off; /* Where IP or IPv4 header starts */
50 __u32 len; /* IPv4 simply where L4 starts
51 * IPv6 where L4 Transport Header starts */
52 __u16 fragoffs; /* IPv6 fragment offset, 0 if first frag (or not frag)*/
53 __s16 protocol;
54 __s32 flags;
55 union nf_inet_addr saddr;
56 union nf_inet_addr daddr;
57 };
58
59 static inline void *frag_safe_skb_hp(const struct sk_buff *skb, int offset,
60 int len, void *buffer)
61 {
62 return skb_header_pointer(skb, offset, len, buffer);
63 }
64
65 /* This function handles filling *ip_vs_iphdr, both for IPv4 and IPv6.
66 * IPv6 requires some extra work, as finding proper header position,
67 * depend on the IPv6 extension headers.
68 */
69 static inline int
70 ip_vs_fill_iph_skb_off(int af, const struct sk_buff *skb, int offset,
71 int hdr_flags, struct ip_vs_iphdr *iphdr)
72 {
73 iphdr->hdr_flags = hdr_flags;
74 iphdr->off = offset;
75
76 #ifdef CONFIG_IP_VS_IPV6
77 if (af == AF_INET6) {
78 struct ipv6hdr _iph;
79 const struct ipv6hdr *iph = skb_header_pointer(
80 skb, offset, sizeof(_iph), &_iph);
81 if (!iph)
82 return 0;
83
84 iphdr->saddr.in6 = iph->saddr;
85 iphdr->daddr.in6 = iph->daddr;
86 /* ipv6_find_hdr() updates len, flags */
87 iphdr->len = offset;
88 iphdr->flags = 0;
89 iphdr->protocol = ipv6_find_hdr(skb, &iphdr->len, -1,
90 &iphdr->fragoffs,
91 &iphdr->flags);
92 if (iphdr->protocol < 0)
93 return 0;
94 } else
95 #endif
96 {
97 struct iphdr _iph;
98 const struct iphdr *iph = skb_header_pointer(
99 skb, offset, sizeof(_iph), &_iph);
100 if (!iph)
101 return 0;
102
103 iphdr->len = offset + iph->ihl * 4;
104 iphdr->fragoffs = 0;
105 iphdr->protocol = iph->protocol;
106 iphdr->saddr.ip = iph->saddr;
107 iphdr->daddr.ip = iph->daddr;
108 }
109
110 return 1;
111 }
112
113 static inline int
114 ip_vs_fill_iph_skb_icmp(int af, const struct sk_buff *skb, int offset,
115 bool inverse, struct ip_vs_iphdr *iphdr)
116 {
117 int hdr_flags = IP_VS_HDR_ICMP;
118
119 if (inverse)
120 hdr_flags |= IP_VS_HDR_INVERSE;
121
122 return ip_vs_fill_iph_skb_off(af, skb, offset, hdr_flags, iphdr);
123 }
124
125 static inline int
126 ip_vs_fill_iph_skb(int af, const struct sk_buff *skb, bool inverse,
127 struct ip_vs_iphdr *iphdr)
128 {
129 int hdr_flags = 0;
130
131 if (inverse)
132 hdr_flags |= IP_VS_HDR_INVERSE;
133
134 return ip_vs_fill_iph_skb_off(af, skb, skb_network_offset(skb),
135 hdr_flags, iphdr);
136 }
137
138 static inline bool
139 ip_vs_iph_inverse(const struct ip_vs_iphdr *iph)
140 {
141 return !!(iph->hdr_flags & IP_VS_HDR_INVERSE);
142 }
143
144 static inline bool
145 ip_vs_iph_icmp(const struct ip_vs_iphdr *iph)
146 {
147 return !!(iph->hdr_flags & IP_VS_HDR_ICMP);
148 }
149
150 static inline void ip_vs_addr_copy(int af, union nf_inet_addr *dst,
151 const union nf_inet_addr *src)
152 {
153 #ifdef CONFIG_IP_VS_IPV6
154 if (af == AF_INET6)
155 dst->in6 = src->in6;
156 else
157 #endif
158 dst->ip = src->ip;
159 }
160
161 static inline void ip_vs_addr_set(int af, union nf_inet_addr *dst,
162 const union nf_inet_addr *src)
163 {
164 #ifdef CONFIG_IP_VS_IPV6
165 if (af == AF_INET6) {
166 dst->in6 = src->in6;
167 return;
168 }
169 #endif
170 dst->ip = src->ip;
171 dst->all[1] = 0;
172 dst->all[2] = 0;
173 dst->all[3] = 0;
174 }
175
176 static inline int ip_vs_addr_equal(int af, const union nf_inet_addr *a,
177 const union nf_inet_addr *b)
178 {
179 #ifdef CONFIG_IP_VS_IPV6
180 if (af == AF_INET6)
181 return ipv6_addr_equal(&a->in6, &b->in6);
182 #endif
183 return a->ip == b->ip;
184 }
185
186 #ifdef CONFIG_IP_VS_DEBUG
187 #include <linux/net.h>
188
189 int ip_vs_get_debug_level(void);
190
191 static inline const char *ip_vs_dbg_addr(int af, char *buf, size_t buf_len,
192 const union nf_inet_addr *addr,
193 int *idx)
194 {
195 int len;
196 #ifdef CONFIG_IP_VS_IPV6
197 if (af == AF_INET6)
198 len = snprintf(&buf[*idx], buf_len - *idx, "[%pI6c]",
199 &addr->in6) + 1;
200 else
201 #endif
202 len = snprintf(&buf[*idx], buf_len - *idx, "%pI4",
203 &addr->ip) + 1;
204
205 *idx += len;
206 BUG_ON(*idx > buf_len + 1);
207 return &buf[*idx - len];
208 }
209
210 #define IP_VS_DBG_BUF(level, msg, ...) \
211 do { \
212 char ip_vs_dbg_buf[160]; \
213 int ip_vs_dbg_idx = 0; \
214 if (level <= ip_vs_get_debug_level()) \
215 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
216 } while (0)
217 #define IP_VS_ERR_BUF(msg...) \
218 do { \
219 char ip_vs_dbg_buf[160]; \
220 int ip_vs_dbg_idx = 0; \
221 pr_err(msg); \
222 } while (0)
223
224 /* Only use from within IP_VS_DBG_BUF() or IP_VS_ERR_BUF macros */
225 #define IP_VS_DBG_ADDR(af, addr) \
226 ip_vs_dbg_addr(af, ip_vs_dbg_buf, \
227 sizeof(ip_vs_dbg_buf), addr, \
228 &ip_vs_dbg_idx)
229
230 #define IP_VS_DBG(level, msg, ...) \
231 do { \
232 if (level <= ip_vs_get_debug_level()) \
233 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
234 } while (0)
235 #define IP_VS_DBG_RL(msg, ...) \
236 do { \
237 if (net_ratelimit()) \
238 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
239 } while (0)
240 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) \
241 do { \
242 if (level <= ip_vs_get_debug_level()) \
243 pp->debug_packet(af, pp, skb, ofs, msg); \
244 } while (0)
245 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) \
246 do { \
247 if (level <= ip_vs_get_debug_level() && \
248 net_ratelimit()) \
249 pp->debug_packet(af, pp, skb, ofs, msg); \
250 } while (0)
251 #else /* NO DEBUGGING at ALL */
252 #define IP_VS_DBG_BUF(level, msg...) do {} while (0)
253 #define IP_VS_ERR_BUF(msg...) do {} while (0)
254 #define IP_VS_DBG(level, msg...) do {} while (0)
255 #define IP_VS_DBG_RL(msg...) do {} while (0)
256 #define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
257 #define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
258 #endif
259
260 #define IP_VS_BUG() BUG()
261 #define IP_VS_ERR_RL(msg, ...) \
262 do { \
263 if (net_ratelimit()) \
264 pr_err(msg, ##__VA_ARGS__); \
265 } while (0)
266
267 #ifdef CONFIG_IP_VS_DEBUG
268 #define EnterFunction(level) \
269 do { \
270 if (level <= ip_vs_get_debug_level()) \
271 printk(KERN_DEBUG \
272 pr_fmt("Enter: %s, %s line %i\n"), \
273 __func__, __FILE__, __LINE__); \
274 } while (0)
275 #define LeaveFunction(level) \
276 do { \
277 if (level <= ip_vs_get_debug_level()) \
278 printk(KERN_DEBUG \
279 pr_fmt("Leave: %s, %s line %i\n"), \
280 __func__, __FILE__, __LINE__); \
281 } while (0)
282 #else
283 #define EnterFunction(level) do {} while (0)
284 #define LeaveFunction(level) do {} while (0)
285 #endif
286
287 /* The port number of FTP service (in network order). */
288 #define FTPPORT cpu_to_be16(21)
289 #define FTPDATA cpu_to_be16(20)
290
291 /* TCP State Values */
292 enum {
293 IP_VS_TCP_S_NONE = 0,
294 IP_VS_TCP_S_ESTABLISHED,
295 IP_VS_TCP_S_SYN_SENT,
296 IP_VS_TCP_S_SYN_RECV,
297 IP_VS_TCP_S_FIN_WAIT,
298 IP_VS_TCP_S_TIME_WAIT,
299 IP_VS_TCP_S_CLOSE,
300 IP_VS_TCP_S_CLOSE_WAIT,
301 IP_VS_TCP_S_LAST_ACK,
302 IP_VS_TCP_S_LISTEN,
303 IP_VS_TCP_S_SYNACK,
304 IP_VS_TCP_S_LAST
305 };
306
307 /* UDP State Values */
308 enum {
309 IP_VS_UDP_S_NORMAL,
310 IP_VS_UDP_S_LAST,
311 };
312
313 /* ICMP State Values */
314 enum {
315 IP_VS_ICMP_S_NORMAL,
316 IP_VS_ICMP_S_LAST,
317 };
318
319 /* SCTP State Values */
320 enum ip_vs_sctp_states {
321 IP_VS_SCTP_S_NONE,
322 IP_VS_SCTP_S_INIT1,
323 IP_VS_SCTP_S_INIT,
324 IP_VS_SCTP_S_COOKIE_SENT,
325 IP_VS_SCTP_S_COOKIE_REPLIED,
326 IP_VS_SCTP_S_COOKIE_WAIT,
327 IP_VS_SCTP_S_COOKIE,
328 IP_VS_SCTP_S_COOKIE_ECHOED,
329 IP_VS_SCTP_S_ESTABLISHED,
330 IP_VS_SCTP_S_SHUTDOWN_SENT,
331 IP_VS_SCTP_S_SHUTDOWN_RECEIVED,
332 IP_VS_SCTP_S_SHUTDOWN_ACK_SENT,
333 IP_VS_SCTP_S_REJECTED,
334 IP_VS_SCTP_S_CLOSED,
335 IP_VS_SCTP_S_LAST
336 };
337
338 /* Connection templates use bits from state */
339 #define IP_VS_CTPL_S_NONE 0x0000
340 #define IP_VS_CTPL_S_ASSURED 0x0001
341 #define IP_VS_CTPL_S_LAST 0x0002
342
343 /* Delta sequence info structure
344 * Each ip_vs_conn has 2 (output AND input seq. changes).
345 * Only used in the VS/NAT.
346 */
347 struct ip_vs_seq {
348 __u32 init_seq; /* Add delta from this seq */
349 __u32 delta; /* Delta in sequence numbers */
350 __u32 previous_delta; /* Delta in sequence numbers
351 * before last resized pkt */
352 };
353
354 /* counters per cpu */
355 struct ip_vs_counters {
356 __u64 conns; /* connections scheduled */
357 __u64 inpkts; /* incoming packets */
358 __u64 outpkts; /* outgoing packets */
359 __u64 inbytes; /* incoming bytes */
360 __u64 outbytes; /* outgoing bytes */
361 };
362 /* Stats per cpu */
363 struct ip_vs_cpu_stats {
364 struct ip_vs_counters cnt;
365 struct u64_stats_sync syncp;
366 };
367
368 /* IPVS statistics objects */
369 struct ip_vs_estimator {
370 struct list_head list;
371
372 u64 last_inbytes;
373 u64 last_outbytes;
374 u64 last_conns;
375 u64 last_inpkts;
376 u64 last_outpkts;
377
378 u64 cps;
379 u64 inpps;
380 u64 outpps;
381 u64 inbps;
382 u64 outbps;
383 };
384
385 /*
386 * IPVS statistics object, 64-bit kernel version of struct ip_vs_stats_user
387 */
388 struct ip_vs_kstats {
389 u64 conns; /* connections scheduled */
390 u64 inpkts; /* incoming packets */
391 u64 outpkts; /* outgoing packets */
392 u64 inbytes; /* incoming bytes */
393 u64 outbytes; /* outgoing bytes */
394
395 u64 cps; /* current connection rate */
396 u64 inpps; /* current in packet rate */
397 u64 outpps; /* current out packet rate */
398 u64 inbps; /* current in byte rate */
399 u64 outbps; /* current out byte rate */
400 };
401
402 struct ip_vs_stats {
403 struct ip_vs_kstats kstats; /* kernel statistics */
404 struct ip_vs_estimator est; /* estimator */
405 struct ip_vs_cpu_stats __percpu *cpustats; /* per cpu counters */
406 spinlock_t lock; /* spin lock */
407 struct ip_vs_kstats kstats0; /* reset values */
408 };
409
410 struct dst_entry;
411 struct iphdr;
412 struct ip_vs_conn;
413 struct ip_vs_app;
414 struct sk_buff;
415 struct ip_vs_proto_data;
416
417 struct ip_vs_protocol {
418 struct ip_vs_protocol *next;
419 char *name;
420 u16 protocol;
421 u16 num_states;
422 int dont_defrag;
423
424 void (*init)(struct ip_vs_protocol *pp);
425
426 void (*exit)(struct ip_vs_protocol *pp);
427
428 int (*init_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
429
430 void (*exit_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
431
432 int (*conn_schedule)(struct netns_ipvs *ipvs,
433 int af, struct sk_buff *skb,
434 struct ip_vs_proto_data *pd,
435 int *verdict, struct ip_vs_conn **cpp,
436 struct ip_vs_iphdr *iph);
437
438 struct ip_vs_conn *
439 (*conn_in_get)(struct netns_ipvs *ipvs,
440 int af,
441 const struct sk_buff *skb,
442 const struct ip_vs_iphdr *iph);
443
444 struct ip_vs_conn *
445 (*conn_out_get)(struct netns_ipvs *ipvs,
446 int af,
447 const struct sk_buff *skb,
448 const struct ip_vs_iphdr *iph);
449
450 int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
451 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
452
453 int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
454 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
455
456 const char *(*state_name)(int state);
457
458 void (*state_transition)(struct ip_vs_conn *cp, int direction,
459 const struct sk_buff *skb,
460 struct ip_vs_proto_data *pd);
461
462 int (*register_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
463
464 void (*unregister_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
465
466 int (*app_conn_bind)(struct ip_vs_conn *cp);
467
468 void (*debug_packet)(int af, struct ip_vs_protocol *pp,
469 const struct sk_buff *skb,
470 int offset,
471 const char *msg);
472
473 void (*timeout_change)(struct ip_vs_proto_data *pd, int flags);
474 };
475
476 /* protocol data per netns */
477 struct ip_vs_proto_data {
478 struct ip_vs_proto_data *next;
479 struct ip_vs_protocol *pp;
480 int *timeout_table; /* protocol timeout table */
481 atomic_t appcnt; /* counter of proto app incs. */
482 struct tcp_states_t *tcp_state_table;
483 };
484
485 struct ip_vs_protocol *ip_vs_proto_get(unsigned short proto);
486 struct ip_vs_proto_data *ip_vs_proto_data_get(struct netns_ipvs *ipvs,
487 unsigned short proto);
488
489 struct ip_vs_conn_param {
490 struct netns_ipvs *ipvs;
491 const union nf_inet_addr *caddr;
492 const union nf_inet_addr *vaddr;
493 __be16 cport;
494 __be16 vport;
495 __u16 protocol;
496 u16 af;
497
498 const struct ip_vs_pe *pe;
499 char *pe_data;
500 __u8 pe_data_len;
501 };
502
503 /* IP_VS structure allocated for each dynamically scheduled connection */
504 struct ip_vs_conn {
505 struct hlist_node c_list; /* hashed list heads */
506 /* Protocol, addresses and port numbers */
507 __be16 cport;
508 __be16 dport;
509 __be16 vport;
510 u16 af; /* address family */
511 union nf_inet_addr caddr; /* client address */
512 union nf_inet_addr vaddr; /* virtual address */
513 union nf_inet_addr daddr; /* destination address */
514 volatile __u32 flags; /* status flags */
515 __u16 protocol; /* Which protocol (TCP/UDP) */
516 __u16 daf; /* Address family of the dest */
517 struct netns_ipvs *ipvs;
518
519 /* counter and timer */
520 refcount_t refcnt; /* reference count */
521 struct timer_list timer; /* Expiration timer */
522 volatile unsigned long timeout; /* timeout */
523
524 /* Flags and state transition */
525 spinlock_t lock; /* lock for state transition */
526 volatile __u16 state; /* state info */
527 volatile __u16 old_state; /* old state, to be used for
528 * state transition triggerd
529 * synchronization
530 */
531 __u32 fwmark; /* Fire wall mark from skb */
532 unsigned long sync_endtime; /* jiffies + sent_retries */
533
534 /* Control members */
535 struct ip_vs_conn *control; /* Master control connection */
536 atomic_t n_control; /* Number of controlled ones */
537 struct ip_vs_dest *dest; /* real server */
538 atomic_t in_pkts; /* incoming packet counter */
539
540 /* Packet transmitter for different forwarding methods. If it
541 * mangles the packet, it must return NF_DROP or better NF_STOLEN,
542 * otherwise this must be changed to a sk_buff **.
543 * NF_ACCEPT can be returned when destination is local.
544 */
545 int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp,
546 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
547
548 /* Note: we can group the following members into a structure,
549 * in order to save more space, and the following members are
550 * only used in VS/NAT anyway
551 */
552 struct ip_vs_app *app; /* bound ip_vs_app object */
553 void *app_data; /* Application private data */
554 struct ip_vs_seq in_seq; /* incoming seq. struct */
555 struct ip_vs_seq out_seq; /* outgoing seq. struct */
556
557 const struct ip_vs_pe *pe;
558 char *pe_data;
559 __u8 pe_data_len;
560
561 struct rcu_head rcu_head;
562 };
563
564 /* Extended internal versions of struct ip_vs_service_user and ip_vs_dest_user
565 * for IPv6 support.
566 *
567 * We need these to conveniently pass around service and destination
568 * options, but unfortunately, we also need to keep the old definitions to
569 * maintain userspace backwards compatibility for the setsockopt interface.
570 */
571 struct ip_vs_service_user_kern {
572 /* virtual service addresses */
573 u16 af;
574 u16 protocol;
575 union nf_inet_addr addr; /* virtual ip address */
576 __be16 port;
577 u32 fwmark; /* firwall mark of service */
578
579 /* virtual service options */
580 char *sched_name;
581 char *pe_name;
582 unsigned int flags; /* virtual service flags */
583 unsigned int timeout; /* persistent timeout in sec */
584 __be32 netmask; /* persistent netmask or plen */
585 };
586
587
588 struct ip_vs_dest_user_kern {
589 /* destination server address */
590 union nf_inet_addr addr;
591 __be16 port;
592
593 /* real server options */
594 unsigned int conn_flags; /* connection flags */
595 int weight; /* destination weight */
596
597 /* thresholds for active connections */
598 u32 u_threshold; /* upper threshold */
599 u32 l_threshold; /* lower threshold */
600
601 /* Address family of addr */
602 u16 af;
603 };
604
605
606 /*
607 * The information about the virtual service offered to the net and the
608 * forwarding entries.
609 */
610 struct ip_vs_service {
611 struct hlist_node s_list; /* for normal service table */
612 struct hlist_node f_list; /* for fwmark-based service table */
613 atomic_t refcnt; /* reference counter */
614
615 u16 af; /* address family */
616 __u16 protocol; /* which protocol (TCP/UDP) */
617 union nf_inet_addr addr; /* IP address for virtual service */
618 __be16 port; /* port number for the service */
619 __u32 fwmark; /* firewall mark of the service */
620 unsigned int flags; /* service status flags */
621 unsigned int timeout; /* persistent timeout in ticks */
622 __be32 netmask; /* grouping granularity, mask/plen */
623 struct netns_ipvs *ipvs;
624
625 struct list_head destinations; /* real server d-linked list */
626 __u32 num_dests; /* number of servers */
627 struct ip_vs_stats stats; /* statistics for the service */
628
629 /* for scheduling */
630 struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */
631 spinlock_t sched_lock; /* lock sched_data */
632 void *sched_data; /* scheduler application data */
633
634 /* alternate persistence engine */
635 struct ip_vs_pe __rcu *pe;
636 int conntrack_afmask;
637
638 struct rcu_head rcu_head;
639 };
640
641 /* Information for cached dst */
642 struct ip_vs_dest_dst {
643 struct dst_entry *dst_cache; /* destination cache entry */
644 u32 dst_cookie;
645 union nf_inet_addr dst_saddr;
646 struct rcu_head rcu_head;
647 };
648
649 /* The real server destination forwarding entry with ip address, port number,
650 * and so on.
651 */
652 struct ip_vs_dest {
653 struct list_head n_list; /* for the dests in the service */
654 struct hlist_node d_list; /* for table with all the dests */
655
656 u16 af; /* address family */
657 __be16 port; /* port number of the server */
658 union nf_inet_addr addr; /* IP address of the server */
659 volatile unsigned int flags; /* dest status flags */
660 atomic_t conn_flags; /* flags to copy to conn */
661 atomic_t weight; /* server weight */
662 atomic_t last_weight; /* server latest weight */
663
664 refcount_t refcnt; /* reference counter */
665 struct ip_vs_stats stats; /* statistics */
666 unsigned long idle_start; /* start time, jiffies */
667
668 /* connection counters and thresholds */
669 atomic_t activeconns; /* active connections */
670 atomic_t inactconns; /* inactive connections */
671 atomic_t persistconns; /* persistent connections */
672 __u32 u_threshold; /* upper threshold */
673 __u32 l_threshold; /* lower threshold */
674
675 /* for destination cache */
676 spinlock_t dst_lock; /* lock of dst_cache */
677 struct ip_vs_dest_dst __rcu *dest_dst; /* cached dst info */
678
679 /* for virtual service */
680 struct ip_vs_service __rcu *svc; /* service it belongs to */
681 __u16 protocol; /* which protocol (TCP/UDP) */
682 __be16 vport; /* virtual port number */
683 union nf_inet_addr vaddr; /* virtual IP address */
684 __u32 vfwmark; /* firewall mark of service */
685
686 struct list_head t_list; /* in dest_trash */
687 unsigned int in_rs_table:1; /* we are in rs_table */
688 };
689
690 /* The scheduler object */
691 struct ip_vs_scheduler {
692 struct list_head n_list; /* d-linked list head */
693 char *name; /* scheduler name */
694 atomic_t refcnt; /* reference counter */
695 struct module *module; /* THIS_MODULE/NULL */
696
697 /* scheduler initializing service */
698 int (*init_service)(struct ip_vs_service *svc);
699 /* scheduling service finish */
700 void (*done_service)(struct ip_vs_service *svc);
701 /* dest is linked */
702 int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
703 /* dest is unlinked */
704 int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
705 /* dest is updated */
706 int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
707
708 /* selecting a server from the given service */
709 struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc,
710 const struct sk_buff *skb,
711 struct ip_vs_iphdr *iph);
712 };
713
714 /* The persistence engine object */
715 struct ip_vs_pe {
716 struct list_head n_list; /* d-linked list head */
717 char *name; /* scheduler name */
718 atomic_t refcnt; /* reference counter */
719 struct module *module; /* THIS_MODULE/NULL */
720
721 /* get the connection template, if any */
722 int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb);
723 bool (*ct_match)(const struct ip_vs_conn_param *p,
724 struct ip_vs_conn *ct);
725 u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval,
726 bool inverse);
727 int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf);
728 /* create connections for real-server outgoing packets */
729 struct ip_vs_conn* (*conn_out)(struct ip_vs_service *svc,
730 struct ip_vs_dest *dest,
731 struct sk_buff *skb,
732 const struct ip_vs_iphdr *iph,
733 __be16 dport, __be16 cport);
734 };
735
736 /* The application module object (a.k.a. app incarnation) */
737 struct ip_vs_app {
738 struct list_head a_list; /* member in app list */
739 int type; /* IP_VS_APP_TYPE_xxx */
740 char *name; /* application module name */
741 __u16 protocol;
742 struct module *module; /* THIS_MODULE/NULL */
743 struct list_head incs_list; /* list of incarnations */
744
745 /* members for application incarnations */
746 struct list_head p_list; /* member in proto app list */
747 struct ip_vs_app *app; /* its real application */
748 __be16 port; /* port number in net order */
749 atomic_t usecnt; /* usage counter */
750 struct rcu_head rcu_head;
751
752 /* output hook: Process packet in inout direction, diff set for TCP.
753 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
754 * 2=Mangled but checksum was not updated
755 */
756 int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *,
757 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh);
758
759 /* input hook: Process packet in outin direction, diff set for TCP.
760 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
761 * 2=Mangled but checksum was not updated
762 */
763 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *,
764 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh);
765
766 /* ip_vs_app initializer */
767 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *);
768
769 /* ip_vs_app finish */
770 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *);
771
772
773 /* not used now */
774 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *,
775 struct ip_vs_protocol *);
776
777 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *);
778
779 int * timeout_table;
780 int * timeouts;
781 int timeouts_size;
782
783 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app,
784 int *verdict, struct ip_vs_conn **cpp);
785
786 struct ip_vs_conn *
787 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app,
788 const struct iphdr *iph, int inverse);
789
790 struct ip_vs_conn *
791 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app,
792 const struct iphdr *iph, int inverse);
793
794 int (*state_transition)(struct ip_vs_conn *cp, int direction,
795 const struct sk_buff *skb,
796 struct ip_vs_app *app);
797
798 void (*timeout_change)(struct ip_vs_app *app, int flags);
799 };
800
801 struct ipvs_master_sync_state {
802 struct list_head sync_queue;
803 struct ip_vs_sync_buff *sync_buff;
804 unsigned long sync_queue_len;
805 unsigned int sync_queue_delay;
806 struct task_struct *master_thread;
807 struct delayed_work master_wakeup_work;
808 struct netns_ipvs *ipvs;
809 };
810
811 /* How much time to keep dests in trash */
812 #define IP_VS_DEST_TRASH_PERIOD (120 * HZ)
813
814 struct ipvs_sync_daemon_cfg {
815 union nf_inet_addr mcast_group;
816 int syncid;
817 u16 sync_maxlen;
818 u16 mcast_port;
819 u8 mcast_af;
820 u8 mcast_ttl;
821 /* multicast interface name */
822 char mcast_ifn[IP_VS_IFNAME_MAXLEN];
823 };
824
825 /* IPVS in network namespace */
826 struct netns_ipvs {
827 int gen; /* Generation */
828 int enable; /* enable like nf_hooks do */
829 /* Hash table: for real service lookups */
830 #define IP_VS_RTAB_BITS 4
831 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
832 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
833
834 struct hlist_head rs_table[IP_VS_RTAB_SIZE];
835 /* ip_vs_app */
836 struct list_head app_list;
837 /* ip_vs_proto */
838 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */
839 struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE];
840 /* ip_vs_proto_tcp */
841 #ifdef CONFIG_IP_VS_PROTO_TCP
842 #define TCP_APP_TAB_BITS 4
843 #define TCP_APP_TAB_SIZE (1 << TCP_APP_TAB_BITS)
844 #define TCP_APP_TAB_MASK (TCP_APP_TAB_SIZE - 1)
845 struct list_head tcp_apps[TCP_APP_TAB_SIZE];
846 #endif
847 /* ip_vs_proto_udp */
848 #ifdef CONFIG_IP_VS_PROTO_UDP
849 #define UDP_APP_TAB_BITS 4
850 #define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS)
851 #define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1)
852 struct list_head udp_apps[UDP_APP_TAB_SIZE];
853 #endif
854 /* ip_vs_proto_sctp */
855 #ifdef CONFIG_IP_VS_PROTO_SCTP
856 #define SCTP_APP_TAB_BITS 4
857 #define SCTP_APP_TAB_SIZE (1 << SCTP_APP_TAB_BITS)
858 #define SCTP_APP_TAB_MASK (SCTP_APP_TAB_SIZE - 1)
859 /* Hash table for SCTP application incarnations */
860 struct list_head sctp_apps[SCTP_APP_TAB_SIZE];
861 #endif
862 /* ip_vs_conn */
863 atomic_t conn_count; /* connection counter */
864
865 /* ip_vs_ctl */
866 struct ip_vs_stats tot_stats; /* Statistics & est. */
867
868 int num_services; /* no of virtual services */
869
870 /* Trash for destinations */
871 struct list_head dest_trash;
872 spinlock_t dest_trash_lock;
873 struct timer_list dest_trash_timer; /* expiration timer */
874 /* Service counters */
875 atomic_t ftpsvc_counter;
876 atomic_t nullsvc_counter;
877 atomic_t conn_out_counter;
878
879 #ifdef CONFIG_SYSCTL
880 /* 1/rate drop and drop-entry variables */
881 struct delayed_work defense_work; /* Work handler */
882 int drop_rate;
883 int drop_counter;
884 atomic_t dropentry;
885 /* locks in ctl.c */
886 spinlock_t dropentry_lock; /* drop entry handling */
887 spinlock_t droppacket_lock; /* drop packet handling */
888 spinlock_t securetcp_lock; /* state and timeout tables */
889
890 /* sys-ctl struct */
891 struct ctl_table_header *sysctl_hdr;
892 struct ctl_table *sysctl_tbl;
893 #endif
894
895 /* sysctl variables */
896 int sysctl_amemthresh;
897 int sysctl_am_droprate;
898 int sysctl_drop_entry;
899 int sysctl_drop_packet;
900 int sysctl_secure_tcp;
901 #ifdef CONFIG_IP_VS_NFCT
902 int sysctl_conntrack;
903 #endif
904 int sysctl_snat_reroute;
905 int sysctl_sync_ver;
906 int sysctl_sync_ports;
907 int sysctl_sync_persist_mode;
908 unsigned long sysctl_sync_qlen_max;
909 int sysctl_sync_sock_size;
910 int sysctl_cache_bypass;
911 int sysctl_expire_nodest_conn;
912 int sysctl_sloppy_tcp;
913 int sysctl_sloppy_sctp;
914 int sysctl_expire_quiescent_template;
915 int sysctl_sync_threshold[2];
916 unsigned int sysctl_sync_refresh_period;
917 int sysctl_sync_retries;
918 int sysctl_nat_icmp_send;
919 int sysctl_pmtu_disc;
920 int sysctl_backup_only;
921 int sysctl_conn_reuse_mode;
922 int sysctl_schedule_icmp;
923 int sysctl_ignore_tunneled;
924
925 /* ip_vs_lblc */
926 int sysctl_lblc_expiration;
927 struct ctl_table_header *lblc_ctl_header;
928 struct ctl_table *lblc_ctl_table;
929 /* ip_vs_lblcr */
930 int sysctl_lblcr_expiration;
931 struct ctl_table_header *lblcr_ctl_header;
932 struct ctl_table *lblcr_ctl_table;
933 /* ip_vs_est */
934 struct list_head est_list; /* estimator list */
935 spinlock_t est_lock;
936 struct timer_list est_timer; /* Estimation timer */
937 /* ip_vs_sync */
938 spinlock_t sync_lock;
939 struct ipvs_master_sync_state *ms;
940 spinlock_t sync_buff_lock;
941 struct task_struct **backup_threads;
942 int threads_mask;
943 volatile int sync_state;
944 struct mutex sync_mutex;
945 struct ipvs_sync_daemon_cfg mcfg; /* Master Configuration */
946 struct ipvs_sync_daemon_cfg bcfg; /* Backup Configuration */
947 /* net name space ptr */
948 struct net *net; /* Needed by timer routines */
949 /* Number of heterogeneous destinations, needed becaus heterogeneous
950 * are not supported when synchronization is enabled.
951 */
952 unsigned int mixed_address_family_dests;
953 };
954
955 #define DEFAULT_SYNC_THRESHOLD 3
956 #define DEFAULT_SYNC_PERIOD 50
957 #define DEFAULT_SYNC_VER 1
958 #define DEFAULT_SLOPPY_TCP 0
959 #define DEFAULT_SLOPPY_SCTP 0
960 #define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ)
961 #define DEFAULT_SYNC_RETRIES 0
962 #define IPVS_SYNC_WAKEUP_RATE 8
963 #define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4)
964 #define IPVS_SYNC_SEND_DELAY (HZ / 50)
965 #define IPVS_SYNC_CHECK_PERIOD HZ
966 #define IPVS_SYNC_FLUSH_TIME (HZ * 2)
967 #define IPVS_SYNC_PORTS_MAX (1 << 6)
968
969 #ifdef CONFIG_SYSCTL
970
971 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
972 {
973 return ipvs->sysctl_sync_threshold[0];
974 }
975
976 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
977 {
978 return READ_ONCE(ipvs->sysctl_sync_threshold[1]);
979 }
980
981 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
982 {
983 return READ_ONCE(ipvs->sysctl_sync_refresh_period);
984 }
985
986 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
987 {
988 return ipvs->sysctl_sync_retries;
989 }
990
991 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
992 {
993 return ipvs->sysctl_sync_ver;
994 }
995
996 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
997 {
998 return ipvs->sysctl_sloppy_tcp;
999 }
1000
1001 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1002 {
1003 return ipvs->sysctl_sloppy_sctp;
1004 }
1005
1006 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1007 {
1008 return READ_ONCE(ipvs->sysctl_sync_ports);
1009 }
1010
1011 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1012 {
1013 return ipvs->sysctl_sync_persist_mode;
1014 }
1015
1016 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1017 {
1018 return ipvs->sysctl_sync_qlen_max;
1019 }
1020
1021 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1022 {
1023 return ipvs->sysctl_sync_sock_size;
1024 }
1025
1026 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1027 {
1028 return ipvs->sysctl_pmtu_disc;
1029 }
1030
1031 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1032 {
1033 return ipvs->sync_state & IP_VS_STATE_BACKUP &&
1034 ipvs->sysctl_backup_only;
1035 }
1036
1037 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1038 {
1039 return ipvs->sysctl_conn_reuse_mode;
1040 }
1041
1042 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1043 {
1044 return ipvs->sysctl_schedule_icmp;
1045 }
1046
1047 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1048 {
1049 return ipvs->sysctl_ignore_tunneled;
1050 }
1051
1052 static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1053 {
1054 return ipvs->sysctl_cache_bypass;
1055 }
1056
1057 #else
1058
1059 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1060 {
1061 return DEFAULT_SYNC_THRESHOLD;
1062 }
1063
1064 static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1065 {
1066 return DEFAULT_SYNC_PERIOD;
1067 }
1068
1069 static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1070 {
1071 return DEFAULT_SYNC_REFRESH_PERIOD;
1072 }
1073
1074 static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1075 {
1076 return DEFAULT_SYNC_RETRIES & 3;
1077 }
1078
1079 static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1080 {
1081 return DEFAULT_SYNC_VER;
1082 }
1083
1084 static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1085 {
1086 return DEFAULT_SLOPPY_TCP;
1087 }
1088
1089 static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1090 {
1091 return DEFAULT_SLOPPY_SCTP;
1092 }
1093
1094 static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1095 {
1096 return 1;
1097 }
1098
1099 static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1100 {
1101 return 0;
1102 }
1103
1104 static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1105 {
1106 return IPVS_SYNC_QLEN_MAX;
1107 }
1108
1109 static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1110 {
1111 return 0;
1112 }
1113
1114 static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1115 {
1116 return 1;
1117 }
1118
1119 static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1120 {
1121 return 0;
1122 }
1123
1124 static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1125 {
1126 return 1;
1127 }
1128
1129 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1130 {
1131 return 0;
1132 }
1133
1134 static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1135 {
1136 return 0;
1137 }
1138
1139 static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1140 {
1141 return 0;
1142 }
1143
1144 #endif
1145
1146 /* IPVS core functions
1147 * (from ip_vs_core.c)
1148 */
1149 const char *ip_vs_proto_name(unsigned int proto);
1150 void ip_vs_init_hash_table(struct list_head *table, int rows);
1151 struct ip_vs_conn *ip_vs_new_conn_out(struct ip_vs_service *svc,
1152 struct ip_vs_dest *dest,
1153 struct sk_buff *skb,
1154 const struct ip_vs_iphdr *iph,
1155 __be16 dport,
1156 __be16 cport);
1157 #define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t)))
1158
1159 #define IP_VS_APP_TYPE_FTP 1
1160
1161 /* ip_vs_conn handling functions
1162 * (from ip_vs_conn.c)
1163 */
1164 enum {
1165 IP_VS_DIR_INPUT = 0,
1166 IP_VS_DIR_OUTPUT,
1167 IP_VS_DIR_INPUT_ONLY,
1168 IP_VS_DIR_LAST,
1169 };
1170
1171 static inline void ip_vs_conn_fill_param(struct netns_ipvs *ipvs, int af, int protocol,
1172 const union nf_inet_addr *caddr,
1173 __be16 cport,
1174 const union nf_inet_addr *vaddr,
1175 __be16 vport,
1176 struct ip_vs_conn_param *p)
1177 {
1178 p->ipvs = ipvs;
1179 p->af = af;
1180 p->protocol = protocol;
1181 p->caddr = caddr;
1182 p->cport = cport;
1183 p->vaddr = vaddr;
1184 p->vport = vport;
1185 p->pe = NULL;
1186 p->pe_data = NULL;
1187 }
1188
1189 struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p);
1190 struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p);
1191
1192 struct ip_vs_conn * ip_vs_conn_in_get_proto(struct netns_ipvs *ipvs, int af,
1193 const struct sk_buff *skb,
1194 const struct ip_vs_iphdr *iph);
1195
1196 struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p);
1197
1198 struct ip_vs_conn * ip_vs_conn_out_get_proto(struct netns_ipvs *ipvs, int af,
1199 const struct sk_buff *skb,
1200 const struct ip_vs_iphdr *iph);
1201
1202 /* Get reference to gain full access to conn.
1203 * By default, RCU read-side critical sections have access only to
1204 * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference.
1205 */
1206 static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp)
1207 {
1208 return refcount_inc_not_zero(&cp->refcnt);
1209 }
1210
1211 /* put back the conn without restarting its timer */
1212 static inline void __ip_vs_conn_put(struct ip_vs_conn *cp)
1213 {
1214 smp_mb__before_atomic();
1215 refcount_dec(&cp->refcnt);
1216 }
1217 void ip_vs_conn_put(struct ip_vs_conn *cp);
1218 void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport);
1219
1220 struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, int dest_af,
1221 const union nf_inet_addr *daddr,
1222 __be16 dport, unsigned int flags,
1223 struct ip_vs_dest *dest, __u32 fwmark);
1224 void ip_vs_conn_expire_now(struct ip_vs_conn *cp);
1225
1226 const char *ip_vs_state_name(const struct ip_vs_conn *cp);
1227
1228 void ip_vs_tcp_conn_listen(struct ip_vs_conn *cp);
1229 int ip_vs_check_template(struct ip_vs_conn *ct, struct ip_vs_dest *cdest);
1230 void ip_vs_random_dropentry(struct netns_ipvs *ipvs);
1231 int ip_vs_conn_init(void);
1232 void ip_vs_conn_cleanup(void);
1233
1234 static inline void ip_vs_control_del(struct ip_vs_conn *cp)
1235 {
1236 struct ip_vs_conn *ctl_cp = cp->control;
1237 if (!ctl_cp) {
1238 IP_VS_ERR_BUF("request control DEL for uncontrolled: "
1239 "%s:%d to %s:%d\n",
1240 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1241 ntohs(cp->cport),
1242 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1243 ntohs(cp->vport));
1244
1245 return;
1246 }
1247
1248 IP_VS_DBG_BUF(7, "DELeting control for: "
1249 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1250 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1251 ntohs(cp->cport),
1252 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1253 ntohs(ctl_cp->cport));
1254
1255 cp->control = NULL;
1256 if (atomic_read(&ctl_cp->n_control) == 0) {
1257 IP_VS_ERR_BUF("BUG control DEL with n=0 : "
1258 "%s:%d to %s:%d\n",
1259 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1260 ntohs(cp->cport),
1261 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1262 ntohs(cp->vport));
1263
1264 return;
1265 }
1266 atomic_dec(&ctl_cp->n_control);
1267 }
1268
1269 static inline void
1270 ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp)
1271 {
1272 if (cp->control) {
1273 IP_VS_ERR_BUF("request control ADD for already controlled: "
1274 "%s:%d to %s:%d\n",
1275 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1276 ntohs(cp->cport),
1277 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1278 ntohs(cp->vport));
1279
1280 ip_vs_control_del(cp);
1281 }
1282
1283 IP_VS_DBG_BUF(7, "ADDing control for: "
1284 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1285 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1286 ntohs(cp->cport),
1287 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1288 ntohs(ctl_cp->cport));
1289
1290 cp->control = ctl_cp;
1291 atomic_inc(&ctl_cp->n_control);
1292 }
1293
1294 /* Mark our template as assured */
1295 static inline void
1296 ip_vs_control_assure_ct(struct ip_vs_conn *cp)
1297 {
1298 struct ip_vs_conn *ct = cp->control;
1299
1300 if (ct && !(ct->state & IP_VS_CTPL_S_ASSURED) &&
1301 (ct->flags & IP_VS_CONN_F_TEMPLATE))
1302 ct->state |= IP_VS_CTPL_S_ASSURED;
1303 }
1304
1305 /* IPVS netns init & cleanup functions */
1306 int ip_vs_estimator_net_init(struct netns_ipvs *ipvs);
1307 int ip_vs_control_net_init(struct netns_ipvs *ipvs);
1308 int ip_vs_protocol_net_init(struct netns_ipvs *ipvs);
1309 int ip_vs_app_net_init(struct netns_ipvs *ipvs);
1310 int ip_vs_conn_net_init(struct netns_ipvs *ipvs);
1311 int ip_vs_sync_net_init(struct netns_ipvs *ipvs);
1312 void ip_vs_conn_net_cleanup(struct netns_ipvs *ipvs);
1313 void ip_vs_app_net_cleanup(struct netns_ipvs *ipvs);
1314 void ip_vs_protocol_net_cleanup(struct netns_ipvs *ipvs);
1315 void ip_vs_control_net_cleanup(struct netns_ipvs *ipvs);
1316 void ip_vs_estimator_net_cleanup(struct netns_ipvs *ipvs);
1317 void ip_vs_sync_net_cleanup(struct netns_ipvs *ipvs);
1318 void ip_vs_service_net_cleanup(struct netns_ipvs *ipvs);
1319
1320 /* IPVS application functions
1321 * (from ip_vs_app.c)
1322 */
1323 #define IP_VS_APP_MAX_PORTS 8
1324 struct ip_vs_app *register_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1325 void unregister_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1326 int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1327 void ip_vs_unbind_app(struct ip_vs_conn *cp);
1328 int register_ip_vs_app_inc(struct netns_ipvs *ipvs, struct ip_vs_app *app, __u16 proto,
1329 __u16 port);
1330 int ip_vs_app_inc_get(struct ip_vs_app *inc);
1331 void ip_vs_app_inc_put(struct ip_vs_app *inc);
1332
1333 int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb,
1334 struct ip_vs_iphdr *ipvsh);
1335 int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb,
1336 struct ip_vs_iphdr *ipvsh);
1337
1338 int register_ip_vs_pe(struct ip_vs_pe *pe);
1339 int unregister_ip_vs_pe(struct ip_vs_pe *pe);
1340 struct ip_vs_pe *ip_vs_pe_getbyname(const char *name);
1341 struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name);
1342
1343 /* Use a #define to avoid all of module.h just for these trivial ops */
1344 #define ip_vs_pe_get(pe) \
1345 if (pe && pe->module) \
1346 __module_get(pe->module);
1347
1348 #define ip_vs_pe_put(pe) \
1349 if (pe && pe->module) \
1350 module_put(pe->module);
1351
1352 /* IPVS protocol functions (from ip_vs_proto.c) */
1353 int ip_vs_protocol_init(void);
1354 void ip_vs_protocol_cleanup(void);
1355 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags);
1356 int *ip_vs_create_timeout_table(int *table, int size);
1357 void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
1358 const struct sk_buff *skb, int offset,
1359 const char *msg);
1360
1361 extern struct ip_vs_protocol ip_vs_protocol_tcp;
1362 extern struct ip_vs_protocol ip_vs_protocol_udp;
1363 extern struct ip_vs_protocol ip_vs_protocol_icmp;
1364 extern struct ip_vs_protocol ip_vs_protocol_esp;
1365 extern struct ip_vs_protocol ip_vs_protocol_ah;
1366 extern struct ip_vs_protocol ip_vs_protocol_sctp;
1367
1368 /* Registering/unregistering scheduler functions
1369 * (from ip_vs_sched.c)
1370 */
1371 int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1372 int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1373 int ip_vs_bind_scheduler(struct ip_vs_service *svc,
1374 struct ip_vs_scheduler *scheduler);
1375 void ip_vs_unbind_scheduler(struct ip_vs_service *svc,
1376 struct ip_vs_scheduler *sched);
1377 struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name);
1378 void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler);
1379 struct ip_vs_conn *
1380 ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb,
1381 struct ip_vs_proto_data *pd, int *ignored,
1382 struct ip_vs_iphdr *iph);
1383 int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb,
1384 struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph);
1385
1386 void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg);
1387
1388 /* IPVS control data and functions (from ip_vs_ctl.c) */
1389 extern struct ip_vs_stats ip_vs_stats;
1390 extern int sysctl_ip_vs_sync_ver;
1391
1392 struct ip_vs_service *
1393 ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
1394 const union nf_inet_addr *vaddr, __be16 vport);
1395
1396 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1397 const union nf_inet_addr *daddr, __be16 dport);
1398
1399 struct ip_vs_dest *
1400 ip_vs_find_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1401 const union nf_inet_addr *daddr, __be16 dport);
1402
1403 int ip_vs_use_count_inc(void);
1404 void ip_vs_use_count_dec(void);
1405 int ip_vs_register_nl_ioctl(void);
1406 void ip_vs_unregister_nl_ioctl(void);
1407 int ip_vs_control_init(void);
1408 void ip_vs_control_cleanup(void);
1409 struct ip_vs_dest *
1410 ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
1411 const union nf_inet_addr *daddr, __be16 dport,
1412 const union nf_inet_addr *vaddr, __be16 vport,
1413 __u16 protocol, __u32 fwmark, __u32 flags);
1414 void ip_vs_try_bind_dest(struct ip_vs_conn *cp);
1415
1416 static inline void ip_vs_dest_hold(struct ip_vs_dest *dest)
1417 {
1418 refcount_inc(&dest->refcnt);
1419 }
1420
1421 static inline void ip_vs_dest_put(struct ip_vs_dest *dest)
1422 {
1423 smp_mb__before_atomic();
1424 refcount_dec(&dest->refcnt);
1425 }
1426
1427 static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest)
1428 {
1429 if (refcount_dec_and_test(&dest->refcnt))
1430 kfree(dest);
1431 }
1432
1433 /* IPVS sync daemon data and function prototypes
1434 * (from ip_vs_sync.c)
1435 */
1436 int start_sync_thread(struct netns_ipvs *ipvs, struct ipvs_sync_daemon_cfg *cfg,
1437 int state);
1438 int stop_sync_thread(struct netns_ipvs *ipvs, int state);
1439 void ip_vs_sync_conn(struct netns_ipvs *ipvs, struct ip_vs_conn *cp, int pkts);
1440
1441 /* IPVS rate estimator prototypes (from ip_vs_est.c) */
1442 void ip_vs_start_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1443 void ip_vs_stop_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1444 void ip_vs_zero_estimator(struct ip_vs_stats *stats);
1445 void ip_vs_read_estimator(struct ip_vs_kstats *dst, struct ip_vs_stats *stats);
1446
1447 /* Various IPVS packet transmitters (from ip_vs_xmit.c) */
1448 int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1449 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1450 int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1451 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1452 int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1453 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1454 int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1455 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1456 int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1457 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1458 int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1459 struct ip_vs_protocol *pp, int offset,
1460 unsigned int hooknum, struct ip_vs_iphdr *iph);
1461 void ip_vs_dest_dst_rcu_free(struct rcu_head *head);
1462
1463 #ifdef CONFIG_IP_VS_IPV6
1464 int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1465 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1466 int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1467 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1468 int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1469 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1470 int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1471 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1472 int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1473 struct ip_vs_protocol *pp, int offset,
1474 unsigned int hooknum, struct ip_vs_iphdr *iph);
1475 #endif
1476
1477 #ifdef CONFIG_SYSCTL
1478 /* This is a simple mechanism to ignore packets when
1479 * we are loaded. Just set ip_vs_drop_rate to 'n' and
1480 * we start to drop 1/rate of the packets
1481 */
1482 static inline int ip_vs_todrop(struct netns_ipvs *ipvs)
1483 {
1484 if (!ipvs->drop_rate)
1485 return 0;
1486 if (--ipvs->drop_counter > 0)
1487 return 0;
1488 ipvs->drop_counter = ipvs->drop_rate;
1489 return 1;
1490 }
1491 #else
1492 static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; }
1493 #endif
1494
1495 /* ip_vs_fwd_tag returns the forwarding tag of the connection */
1496 #define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK)
1497
1498 static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp)
1499 {
1500 char fwd;
1501
1502 switch (IP_VS_FWD_METHOD(cp)) {
1503 case IP_VS_CONN_F_MASQ:
1504 fwd = 'M'; break;
1505 case IP_VS_CONN_F_LOCALNODE:
1506 fwd = 'L'; break;
1507 case IP_VS_CONN_F_TUNNEL:
1508 fwd = 'T'; break;
1509 case IP_VS_CONN_F_DROUTE:
1510 fwd = 'R'; break;
1511 case IP_VS_CONN_F_BYPASS:
1512 fwd = 'B'; break;
1513 default:
1514 fwd = '?'; break;
1515 }
1516 return fwd;
1517 }
1518
1519 void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp,
1520 struct ip_vs_conn *cp, int dir);
1521
1522 #ifdef CONFIG_IP_VS_IPV6
1523 void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp,
1524 struct ip_vs_conn *cp, int dir);
1525 #endif
1526
1527 __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset);
1528
1529 static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum)
1530 {
1531 __be32 diff[2] = { ~old, new };
1532
1533 return csum_partial(diff, sizeof(diff), oldsum);
1534 }
1535
1536 #ifdef CONFIG_IP_VS_IPV6
1537 static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new,
1538 __wsum oldsum)
1539 {
1540 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0],
1541 new[3], new[2], new[1], new[0] };
1542
1543 return csum_partial(diff, sizeof(diff), oldsum);
1544 }
1545 #endif
1546
1547 static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum)
1548 {
1549 __be16 diff[2] = { ~old, new };
1550
1551 return csum_partial(diff, sizeof(diff), oldsum);
1552 }
1553
1554 /* Forget current conntrack (unconfirmed) and attach notrack entry */
1555 static inline void ip_vs_notrack(struct sk_buff *skb)
1556 {
1557 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1558 enum ip_conntrack_info ctinfo;
1559 struct nf_conn *ct = nf_ct_get(skb, &ctinfo);
1560
1561 if (ct) {
1562 nf_conntrack_put(&ct->ct_general);
1563 nf_ct_set(skb, NULL, IP_CT_UNTRACKED);
1564 }
1565 #endif
1566 }
1567
1568 #ifdef CONFIG_IP_VS_NFCT
1569 /* Netfilter connection tracking
1570 * (from ip_vs_nfct.c)
1571 */
1572 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1573 {
1574 #ifdef CONFIG_SYSCTL
1575 return ipvs->sysctl_conntrack;
1576 #else
1577 return 0;
1578 #endif
1579 }
1580
1581 void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp,
1582 int outin);
1583 int ip_vs_confirm_conntrack(struct sk_buff *skb);
1584 void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct,
1585 struct ip_vs_conn *cp, u_int8_t proto,
1586 const __be16 port, int from_rs);
1587 void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp);
1588
1589 #else
1590
1591 static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1592 {
1593 return 0;
1594 }
1595
1596 static inline void ip_vs_update_conntrack(struct sk_buff *skb,
1597 struct ip_vs_conn *cp, int outin)
1598 {
1599 }
1600
1601 static inline int ip_vs_confirm_conntrack(struct sk_buff *skb)
1602 {
1603 return NF_ACCEPT;
1604 }
1605
1606 static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp)
1607 {
1608 }
1609 #endif /* CONFIG_IP_VS_NFCT */
1610
1611 /* Really using conntrack? */
1612 static inline bool ip_vs_conn_uses_conntrack(struct ip_vs_conn *cp,
1613 struct sk_buff *skb)
1614 {
1615 #ifdef CONFIG_IP_VS_NFCT
1616 enum ip_conntrack_info ctinfo;
1617 struct nf_conn *ct;
1618
1619 if (!(cp->flags & IP_VS_CONN_F_NFCT))
1620 return false;
1621 ct = nf_ct_get(skb, &ctinfo);
1622 if (ct)
1623 return true;
1624 #endif
1625 return false;
1626 }
1627
1628 static inline int ip_vs_register_conntrack(struct ip_vs_service *svc)
1629 {
1630 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
1631 int afmask = (svc->af == AF_INET6) ? 2 : 1;
1632 int ret = 0;
1633
1634 if (!(svc->conntrack_afmask & afmask)) {
1635 ret = nf_ct_netns_get(svc->ipvs->net, svc->af);
1636 if (ret >= 0)
1637 svc->conntrack_afmask |= afmask;
1638 }
1639 return ret;
1640 #else
1641 return 0;
1642 #endif
1643 }
1644
1645 static inline void ip_vs_unregister_conntrack(struct ip_vs_service *svc)
1646 {
1647 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
1648 int afmask = (svc->af == AF_INET6) ? 2 : 1;
1649
1650 if (svc->conntrack_afmask & afmask) {
1651 nf_ct_netns_put(svc->ipvs->net, svc->af);
1652 svc->conntrack_afmask &= ~afmask;
1653 }
1654 #endif
1655 }
1656
1657 static inline int
1658 ip_vs_dest_conn_overhead(struct ip_vs_dest *dest)
1659 {
1660 /* We think the overhead of processing active connections is 256
1661 * times higher than that of inactive connections in average. (This
1662 * 256 times might not be accurate, we will change it later) We
1663 * use the following formula to estimate the overhead now:
1664 * dest->activeconns*256 + dest->inactconns
1665 */
1666 return (atomic_read(&dest->activeconns) << 8) +
1667 atomic_read(&dest->inactconns);
1668 }
1669
1670 #endif /* _NET_IP_VS_H */
Ubuntu Jammy Linux kernel mirror