Loading...
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#include <linux/workqueue.h>
18
19#include <linux/compiler.h>
20#include <linux/timer.h>
21#include <linux/bug.h>
22
23#include <net/checksum.h>
24#include <linux/netfilter.h> /* for union nf_inet_addr */
25#include <linux/ip.h>
26#include <linux/ipv6.h> /* for struct ipv6hdr */
27#include <net/ipv6.h>
28#if IS_ENABLED(CONFIG_NF_CONNTRACK)
29#include <net/netfilter/nf_conntrack.h>
30#endif
31#include <net/net_namespace.h> /* Netw namespace */
32#include <linux/sched/isolation.h>
33
34#define IP_VS_HDR_INVERSE 1
35#define IP_VS_HDR_ICMP 2
36
37/* Generic access of ipvs struct */
38static inline struct netns_ipvs *net_ipvs(struct net* net)
39{
40 return net->ipvs;
41}
42
43/* Connections' size value needed by ip_vs_ctl.c */
44extern int ip_vs_conn_tab_size;
45
46extern struct mutex __ip_vs_mutex;
47
48struct ip_vs_iphdr {
49 int hdr_flags; /* ipvs flags */
50 __u32 off; /* Where IP or IPv4 header starts */
51 __u32 len; /* IPv4 simply where L4 starts
52 * IPv6 where L4 Transport Header starts */
53 __u16 fragoffs; /* IPv6 fragment offset, 0 if first frag (or not frag)*/
54 __s16 protocol;
55 __s32 flags;
56 union nf_inet_addr saddr;
57 union nf_inet_addr daddr;
58};
59
60static inline void *frag_safe_skb_hp(const struct sk_buff *skb, int offset,
61 int len, void *buffer)
62{
63 return skb_header_pointer(skb, offset, len, buffer);
64}
65
66/* This function handles filling *ip_vs_iphdr, both for IPv4 and IPv6.
67 * IPv6 requires some extra work, as finding proper header position,
68 * depend on the IPv6 extension headers.
69 */
70static inline int
71ip_vs_fill_iph_skb_off(int af, const struct sk_buff *skb, int offset,
72 int hdr_flags, struct ip_vs_iphdr *iphdr)
73{
74 iphdr->hdr_flags = hdr_flags;
75 iphdr->off = offset;
76
77#ifdef CONFIG_IP_VS_IPV6
78 if (af == AF_INET6) {
79 struct ipv6hdr _iph;
80 const struct ipv6hdr *iph = skb_header_pointer(
81 skb, offset, sizeof(_iph), &_iph);
82 if (!iph)
83 return 0;
84
85 iphdr->saddr.in6 = iph->saddr;
86 iphdr->daddr.in6 = iph->daddr;
87 /* ipv6_find_hdr() updates len, flags */
88 iphdr->len = offset;
89 iphdr->flags = 0;
90 iphdr->protocol = ipv6_find_hdr(skb, &iphdr->len, -1,
91 &iphdr->fragoffs,
92 &iphdr->flags);
93 if (iphdr->protocol < 0)
94 return 0;
95 } else
96#endif
97 {
98 struct iphdr _iph;
99 const struct iphdr *iph = skb_header_pointer(
100 skb, offset, sizeof(_iph), &_iph);
101 if (!iph)
102 return 0;
103
104 iphdr->len = offset + iph->ihl * 4;
105 iphdr->fragoffs = 0;
106 iphdr->protocol = iph->protocol;
107 iphdr->saddr.ip = iph->saddr;
108 iphdr->daddr.ip = iph->daddr;
109 }
110
111 return 1;
112}
113
114static inline int
115ip_vs_fill_iph_skb_icmp(int af, const struct sk_buff *skb, int offset,
116 bool inverse, struct ip_vs_iphdr *iphdr)
117{
118 int hdr_flags = IP_VS_HDR_ICMP;
119
120 if (inverse)
121 hdr_flags |= IP_VS_HDR_INVERSE;
122
123 return ip_vs_fill_iph_skb_off(af, skb, offset, hdr_flags, iphdr);
124}
125
126static inline int
127ip_vs_fill_iph_skb(int af, const struct sk_buff *skb, bool inverse,
128 struct ip_vs_iphdr *iphdr)
129{
130 int hdr_flags = 0;
131
132 if (inverse)
133 hdr_flags |= IP_VS_HDR_INVERSE;
134
135 return ip_vs_fill_iph_skb_off(af, skb, skb_network_offset(skb),
136 hdr_flags, iphdr);
137}
138
139static inline bool
140ip_vs_iph_inverse(const struct ip_vs_iphdr *iph)
141{
142 return !!(iph->hdr_flags & IP_VS_HDR_INVERSE);
143}
144
145static inline bool
146ip_vs_iph_icmp(const struct ip_vs_iphdr *iph)
147{
148 return !!(iph->hdr_flags & IP_VS_HDR_ICMP);
149}
150
151static inline void ip_vs_addr_copy(int af, union nf_inet_addr *dst,
152 const union nf_inet_addr *src)
153{
154#ifdef CONFIG_IP_VS_IPV6
155 if (af == AF_INET6)
156 dst->in6 = src->in6;
157 else
158#endif
159 dst->ip = src->ip;
160}
161
162static inline void ip_vs_addr_set(int af, union nf_inet_addr *dst,
163 const union nf_inet_addr *src)
164{
165#ifdef CONFIG_IP_VS_IPV6
166 if (af == AF_INET6) {
167 dst->in6 = src->in6;
168 return;
169 }
170#endif
171 dst->ip = src->ip;
172 dst->all[1] = 0;
173 dst->all[2] = 0;
174 dst->all[3] = 0;
175}
176
177static inline int ip_vs_addr_equal(int af, const union nf_inet_addr *a,
178 const union nf_inet_addr *b)
179{
180#ifdef CONFIG_IP_VS_IPV6
181 if (af == AF_INET6)
182 return ipv6_addr_equal(&a->in6, &b->in6);
183#endif
184 return a->ip == b->ip;
185}
186
187#ifdef CONFIG_IP_VS_DEBUG
188#include <linux/net.h>
189
190int ip_vs_get_debug_level(void);
191
192static inline const char *ip_vs_dbg_addr(int af, char *buf, size_t buf_len,
193 const union nf_inet_addr *addr,
194 int *idx)
195{
196 int len;
197#ifdef CONFIG_IP_VS_IPV6
198 if (af == AF_INET6)
199 len = snprintf(&buf[*idx], buf_len - *idx, "[%pI6c]",
200 &addr->in6) + 1;
201 else
202#endif
203 len = snprintf(&buf[*idx], buf_len - *idx, "%pI4",
204 &addr->ip) + 1;
205
206 *idx += len;
207 BUG_ON(*idx > buf_len + 1);
208 return &buf[*idx - len];
209}
210
211#define IP_VS_DBG_BUF(level, msg, ...) \
212 do { \
213 char ip_vs_dbg_buf[160]; \
214 int ip_vs_dbg_idx = 0; \
215 if (level <= ip_vs_get_debug_level()) \
216 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
217 } while (0)
218#define IP_VS_ERR_BUF(msg...) \
219 do { \
220 char ip_vs_dbg_buf[160]; \
221 int ip_vs_dbg_idx = 0; \
222 pr_err(msg); \
223 } while (0)
224
225/* Only use from within IP_VS_DBG_BUF() or IP_VS_ERR_BUF macros */
226#define IP_VS_DBG_ADDR(af, addr) \
227 ip_vs_dbg_addr(af, ip_vs_dbg_buf, \
228 sizeof(ip_vs_dbg_buf), addr, \
229 &ip_vs_dbg_idx)
230
231#define IP_VS_DBG(level, msg, ...) \
232 do { \
233 if (level <= ip_vs_get_debug_level()) \
234 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
235 } while (0)
236#define IP_VS_DBG_RL(msg, ...) \
237 do { \
238 if (net_ratelimit()) \
239 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
240 } while (0)
241#define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) \
242 do { \
243 if (level <= ip_vs_get_debug_level()) \
244 pp->debug_packet(af, pp, skb, ofs, msg); \
245 } while (0)
246#define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) \
247 do { \
248 if (level <= ip_vs_get_debug_level() && \
249 net_ratelimit()) \
250 pp->debug_packet(af, pp, skb, ofs, msg); \
251 } while (0)
252#else /* NO DEBUGGING at ALL */
253#define IP_VS_DBG_BUF(level, msg...) do {} while (0)
254#define IP_VS_ERR_BUF(msg...) do {} while (0)
255#define IP_VS_DBG(level, msg...) do {} while (0)
256#define IP_VS_DBG_RL(msg...) do {} while (0)
257#define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
258#define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
259#endif
260
261#define IP_VS_BUG() BUG()
262#define IP_VS_ERR_RL(msg, ...) \
263 do { \
264 if (net_ratelimit()) \
265 pr_err(msg, ##__VA_ARGS__); \
266 } while (0)
267
268#ifdef CONFIG_IP_VS_DEBUG
269#define EnterFunction(level) \
270 do { \
271 if (level <= ip_vs_get_debug_level()) \
272 printk(KERN_DEBUG \
273 pr_fmt("Enter: %s, %s line %i\n"), \
274 __func__, __FILE__, __LINE__); \
275 } while (0)
276#define LeaveFunction(level) \
277 do { \
278 if (level <= ip_vs_get_debug_level()) \
279 printk(KERN_DEBUG \
280 pr_fmt("Leave: %s, %s line %i\n"), \
281 __func__, __FILE__, __LINE__); \
282 } while (0)
283#else
284#define EnterFunction(level) do {} while (0)
285#define LeaveFunction(level) do {} while (0)
286#endif
287
288/* The port number of FTP service (in network order). */
289#define FTPPORT cpu_to_be16(21)
290#define FTPDATA cpu_to_be16(20)
291
292/* TCP State Values */
293enum {
294 IP_VS_TCP_S_NONE = 0,
295 IP_VS_TCP_S_ESTABLISHED,
296 IP_VS_TCP_S_SYN_SENT,
297 IP_VS_TCP_S_SYN_RECV,
298 IP_VS_TCP_S_FIN_WAIT,
299 IP_VS_TCP_S_TIME_WAIT,
300 IP_VS_TCP_S_CLOSE,
301 IP_VS_TCP_S_CLOSE_WAIT,
302 IP_VS_TCP_S_LAST_ACK,
303 IP_VS_TCP_S_LISTEN,
304 IP_VS_TCP_S_SYNACK,
305 IP_VS_TCP_S_LAST
306};
307
308/* UDP State Values */
309enum {
310 IP_VS_UDP_S_NORMAL,
311 IP_VS_UDP_S_LAST,
312};
313
314/* ICMP State Values */
315enum {
316 IP_VS_ICMP_S_NORMAL,
317 IP_VS_ICMP_S_LAST,
318};
319
320/* SCTP State Values */
321enum ip_vs_sctp_states {
322 IP_VS_SCTP_S_NONE,
323 IP_VS_SCTP_S_INIT1,
324 IP_VS_SCTP_S_INIT,
325 IP_VS_SCTP_S_COOKIE_SENT,
326 IP_VS_SCTP_S_COOKIE_REPLIED,
327 IP_VS_SCTP_S_COOKIE_WAIT,
328 IP_VS_SCTP_S_COOKIE,
329 IP_VS_SCTP_S_COOKIE_ECHOED,
330 IP_VS_SCTP_S_ESTABLISHED,
331 IP_VS_SCTP_S_SHUTDOWN_SENT,
332 IP_VS_SCTP_S_SHUTDOWN_RECEIVED,
333 IP_VS_SCTP_S_SHUTDOWN_ACK_SENT,
334 IP_VS_SCTP_S_REJECTED,
335 IP_VS_SCTP_S_CLOSED,
336 IP_VS_SCTP_S_LAST
337};
338
339/* Connection templates use bits from state */
340#define IP_VS_CTPL_S_NONE 0x0000
341#define IP_VS_CTPL_S_ASSURED 0x0001
342#define IP_VS_CTPL_S_LAST 0x0002
343
344/* Delta sequence info structure
345 * Each ip_vs_conn has 2 (output AND input seq. changes).
346 * Only used in the VS/NAT.
347 */
348struct ip_vs_seq {
349 __u32 init_seq; /* Add delta from this seq */
350 __u32 delta; /* Delta in sequence numbers */
351 __u32 previous_delta; /* Delta in sequence numbers
352 * before last resized pkt */
353};
354
355/* counters per cpu */
356struct ip_vs_counters {
357 u64_stats_t conns; /* connections scheduled */
358 u64_stats_t inpkts; /* incoming packets */
359 u64_stats_t outpkts; /* outgoing packets */
360 u64_stats_t inbytes; /* incoming bytes */
361 u64_stats_t outbytes; /* outgoing bytes */
362};
363/* Stats per cpu */
364struct ip_vs_cpu_stats {
365 struct ip_vs_counters cnt;
366 struct u64_stats_sync syncp;
367};
368
369/* Default nice for estimator kthreads */
370#define IPVS_EST_NICE 0
371
372/* IPVS statistics objects */
373struct ip_vs_estimator {
374 struct hlist_node list;
375
376 u64 last_inbytes;
377 u64 last_outbytes;
378 u64 last_conns;
379 u64 last_inpkts;
380 u64 last_outpkts;
381
382 u64 cps;
383 u64 inpps;
384 u64 outpps;
385 u64 inbps;
386 u64 outbps;
387
388 s32 ktid:16, /* kthread ID, -1=temp list */
389 ktrow:8, /* row/tick ID for kthread */
390 ktcid:8; /* chain ID for kthread tick */
391};
392
393/*
394 * IPVS statistics object, 64-bit kernel version of struct ip_vs_stats_user
395 */
396struct 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
410struct 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
418struct ip_vs_stats_rcu {
419 struct ip_vs_stats s;
420 struct rcu_head rcu_head;
421};
422
423int ip_vs_stats_init_alloc(struct ip_vs_stats *s);
424struct ip_vs_stats *ip_vs_stats_alloc(void);
425void ip_vs_stats_release(struct ip_vs_stats *stats);
426void ip_vs_stats_free(struct ip_vs_stats *stats);
427
428/* Process estimators in multiple timer ticks (20/50/100, see ktrow) */
429#define IPVS_EST_NTICKS 50
430/* Estimation uses a 2-second period containing ticks (in jiffies) */
431#define IPVS_EST_TICK ((2 * HZ) / IPVS_EST_NTICKS)
432
433/* Limit of CPU load per kthread (8 for 12.5%), ratio of CPU capacity (1/C).
434 * Value of 4 and above ensures kthreads will take work without exceeding
435 * the CPU capacity under different circumstances.
436 */
437#define IPVS_EST_LOAD_DIVISOR 8
438
439/* Kthreads should not have work that exceeds the CPU load above 50% */
440#define IPVS_EST_CPU_KTHREADS (IPVS_EST_LOAD_DIVISOR / 2)
441
442/* Desired number of chains per timer tick (chain load factor in 100us units),
443 * 48=4.8ms of 40ms tick (12% CPU usage):
444 * 2 sec * 1000 ms in sec * 10 (100us in ms) / 8 (12.5%) / 50
445 */
446#define IPVS_EST_CHAIN_FACTOR \
447 ALIGN_DOWN(2 * 1000 * 10 / IPVS_EST_LOAD_DIVISOR / IPVS_EST_NTICKS, 8)
448
449/* Compiled number of chains per tick
450 * The defines should match cond_resched_rcu
451 */
452#if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
453#define IPVS_EST_TICK_CHAINS IPVS_EST_CHAIN_FACTOR
454#else
455#define IPVS_EST_TICK_CHAINS 1
456#endif
457
458#if IPVS_EST_NTICKS > 127
459#error Too many timer ticks for ktrow
460#endif
461
462/* Multiple chains processed in same tick */
463struct ip_vs_est_tick_data {
464 struct hlist_head chains[IPVS_EST_TICK_CHAINS];
465 DECLARE_BITMAP(present, IPVS_EST_TICK_CHAINS);
466 DECLARE_BITMAP(full, IPVS_EST_TICK_CHAINS);
467 int chain_len[IPVS_EST_TICK_CHAINS];
468};
469
470/* Context for estimation kthread */
471struct ip_vs_est_kt_data {
472 struct netns_ipvs *ipvs;
473 struct task_struct *task; /* task if running */
474 struct ip_vs_est_tick_data __rcu *ticks[IPVS_EST_NTICKS];
475 DECLARE_BITMAP(avail, IPVS_EST_NTICKS); /* tick has space for ests */
476 unsigned long est_timer; /* estimation timer (jiffies) */
477 struct ip_vs_stats *calc_stats; /* Used for calculation */
478 int tick_len[IPVS_EST_NTICKS]; /* est count */
479 int id; /* ktid per netns */
480 int chain_max; /* max ests per tick chain */
481 int tick_max; /* max ests per tick */
482 int est_count; /* attached ests to kthread */
483 int est_max_count; /* max ests per kthread */
484 int add_row; /* row for new ests */
485 int est_row; /* estimated row */
486};
487
488struct dst_entry;
489struct iphdr;
490struct ip_vs_conn;
491struct ip_vs_app;
492struct sk_buff;
493struct ip_vs_proto_data;
494
495struct ip_vs_protocol {
496 struct ip_vs_protocol *next;
497 char *name;
498 u16 protocol;
499 u16 num_states;
500 int dont_defrag;
501
502 void (*init)(struct ip_vs_protocol *pp);
503
504 void (*exit)(struct ip_vs_protocol *pp);
505
506 int (*init_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
507
508 void (*exit_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
509
510 int (*conn_schedule)(struct netns_ipvs *ipvs,
511 int af, struct sk_buff *skb,
512 struct ip_vs_proto_data *pd,
513 int *verdict, struct ip_vs_conn **cpp,
514 struct ip_vs_iphdr *iph);
515
516 struct ip_vs_conn *
517 (*conn_in_get)(struct netns_ipvs *ipvs,
518 int af,
519 const struct sk_buff *skb,
520 const struct ip_vs_iphdr *iph);
521
522 struct ip_vs_conn *
523 (*conn_out_get)(struct netns_ipvs *ipvs,
524 int af,
525 const struct sk_buff *skb,
526 const struct ip_vs_iphdr *iph);
527
528 int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
529 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
530
531 int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
532 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
533
534 const char *(*state_name)(int state);
535
536 void (*state_transition)(struct ip_vs_conn *cp, int direction,
537 const struct sk_buff *skb,
538 struct ip_vs_proto_data *pd);
539
540 int (*register_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
541
542 void (*unregister_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
543
544 int (*app_conn_bind)(struct ip_vs_conn *cp);
545
546 void (*debug_packet)(int af, struct ip_vs_protocol *pp,
547 const struct sk_buff *skb,
548 int offset,
549 const char *msg);
550
551 void (*timeout_change)(struct ip_vs_proto_data *pd, int flags);
552};
553
554/* protocol data per netns */
555struct ip_vs_proto_data {
556 struct ip_vs_proto_data *next;
557 struct ip_vs_protocol *pp;
558 int *timeout_table; /* protocol timeout table */
559 atomic_t appcnt; /* counter of proto app incs. */
560 struct tcp_states_t *tcp_state_table;
561};
562
563struct ip_vs_protocol *ip_vs_proto_get(unsigned short proto);
564struct ip_vs_proto_data *ip_vs_proto_data_get(struct netns_ipvs *ipvs,
565 unsigned short proto);
566
567struct ip_vs_conn_param {
568 struct netns_ipvs *ipvs;
569 const union nf_inet_addr *caddr;
570 const union nf_inet_addr *vaddr;
571 __be16 cport;
572 __be16 vport;
573 __u16 protocol;
574 u16 af;
575
576 const struct ip_vs_pe *pe;
577 char *pe_data;
578 __u8 pe_data_len;
579};
580
581/* IP_VS structure allocated for each dynamically scheduled connection */
582struct ip_vs_conn {
583 struct hlist_node c_list; /* hashed list heads */
584 /* Protocol, addresses and port numbers */
585 __be16 cport;
586 __be16 dport;
587 __be16 vport;
588 u16 af; /* address family */
589 union nf_inet_addr caddr; /* client address */
590 union nf_inet_addr vaddr; /* virtual address */
591 union nf_inet_addr daddr; /* destination address */
592 volatile __u32 flags; /* status flags */
593 __u16 protocol; /* Which protocol (TCP/UDP) */
594 __u16 daf; /* Address family of the dest */
595 struct netns_ipvs *ipvs;
596
597 /* counter and timer */
598 refcount_t refcnt; /* reference count */
599 struct timer_list timer; /* Expiration timer */
600 volatile unsigned long timeout; /* timeout */
601
602 /* Flags and state transition */
603 spinlock_t lock; /* lock for state transition */
604 volatile __u16 state; /* state info */
605 volatile __u16 old_state; /* old state, to be used for
606 * state transition triggerd
607 * synchronization
608 */
609 __u32 fwmark; /* Fire wall mark from skb */
610 unsigned long sync_endtime; /* jiffies + sent_retries */
611
612 /* Control members */
613 struct ip_vs_conn *control; /* Master control connection */
614 atomic_t n_control; /* Number of controlled ones */
615 struct ip_vs_dest *dest; /* real server */
616 atomic_t in_pkts; /* incoming packet counter */
617
618 /* Packet transmitter for different forwarding methods. If it
619 * mangles the packet, it must return NF_DROP or better NF_STOLEN,
620 * otherwise this must be changed to a sk_buff **.
621 * NF_ACCEPT can be returned when destination is local.
622 */
623 int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp,
624 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
625
626 /* Note: we can group the following members into a structure,
627 * in order to save more space, and the following members are
628 * only used in VS/NAT anyway
629 */
630 struct ip_vs_app *app; /* bound ip_vs_app object */
631 void *app_data; /* Application private data */
632 struct ip_vs_seq in_seq; /* incoming seq. struct */
633 struct ip_vs_seq out_seq; /* outgoing seq. struct */
634
635 const struct ip_vs_pe *pe;
636 char *pe_data;
637 __u8 pe_data_len;
638
639 struct rcu_head rcu_head;
640};
641
642/* Extended internal versions of struct ip_vs_service_user and ip_vs_dest_user
643 * for IPv6 support.
644 *
645 * We need these to conveniently pass around service and destination
646 * options, but unfortunately, we also need to keep the old definitions to
647 * maintain userspace backwards compatibility for the setsockopt interface.
648 */
649struct ip_vs_service_user_kern {
650 /* virtual service addresses */
651 u16 af;
652 u16 protocol;
653 union nf_inet_addr addr; /* virtual ip address */
654 __be16 port;
655 u32 fwmark; /* firwall mark of service */
656
657 /* virtual service options */
658 char *sched_name;
659 char *pe_name;
660 unsigned int flags; /* virtual service flags */
661 unsigned int timeout; /* persistent timeout in sec */
662 __be32 netmask; /* persistent netmask or plen */
663};
664
665
666struct ip_vs_dest_user_kern {
667 /* destination server address */
668 union nf_inet_addr addr;
669 __be16 port;
670
671 /* real server options */
672 unsigned int conn_flags; /* connection flags */
673 int weight; /* destination weight */
674
675 /* thresholds for active connections */
676 u32 u_threshold; /* upper threshold */
677 u32 l_threshold; /* lower threshold */
678
679 /* Address family of addr */
680 u16 af;
681
682 u16 tun_type; /* tunnel type */
683 __be16 tun_port; /* tunnel port */
684 u16 tun_flags; /* tunnel flags */
685};
686
687
688/*
689 * The information about the virtual service offered to the net and the
690 * forwarding entries.
691 */
692struct ip_vs_service {
693 struct hlist_node s_list; /* for normal service table */
694 struct hlist_node f_list; /* for fwmark-based service table */
695 atomic_t refcnt; /* reference counter */
696
697 u16 af; /* address family */
698 __u16 protocol; /* which protocol (TCP/UDP) */
699 union nf_inet_addr addr; /* IP address for virtual service */
700 __be16 port; /* port number for the service */
701 __u32 fwmark; /* firewall mark of the service */
702 unsigned int flags; /* service status flags */
703 unsigned int timeout; /* persistent timeout in ticks */
704 __be32 netmask; /* grouping granularity, mask/plen */
705 struct netns_ipvs *ipvs;
706
707 struct list_head destinations; /* real server d-linked list */
708 __u32 num_dests; /* number of servers */
709 struct ip_vs_stats stats; /* statistics for the service */
710
711 /* for scheduling */
712 struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */
713 spinlock_t sched_lock; /* lock sched_data */
714 void *sched_data; /* scheduler application data */
715
716 /* alternate persistence engine */
717 struct ip_vs_pe __rcu *pe;
718 int conntrack_afmask;
719
720 struct rcu_head rcu_head;
721};
722
723/* Information for cached dst */
724struct ip_vs_dest_dst {
725 struct dst_entry *dst_cache; /* destination cache entry */
726 u32 dst_cookie;
727 union nf_inet_addr dst_saddr;
728 struct rcu_head rcu_head;
729};
730
731/* The real server destination forwarding entry with ip address, port number,
732 * and so on.
733 */
734struct ip_vs_dest {
735 struct list_head n_list; /* for the dests in the service */
736 struct hlist_node d_list; /* for table with all the dests */
737
738 u16 af; /* address family */
739 __be16 port; /* port number of the server */
740 union nf_inet_addr addr; /* IP address of the server */
741 volatile unsigned int flags; /* dest status flags */
742 atomic_t conn_flags; /* flags to copy to conn */
743 atomic_t weight; /* server weight */
744 atomic_t last_weight; /* server latest weight */
745 __u16 tun_type; /* tunnel type */
746 __be16 tun_port; /* tunnel port */
747 __u16 tun_flags; /* tunnel flags */
748
749 refcount_t refcnt; /* reference counter */
750 struct ip_vs_stats stats; /* statistics */
751 unsigned long idle_start; /* start time, jiffies */
752
753 /* connection counters and thresholds */
754 atomic_t activeconns; /* active connections */
755 atomic_t inactconns; /* inactive connections */
756 atomic_t persistconns; /* persistent connections */
757 __u32 u_threshold; /* upper threshold */
758 __u32 l_threshold; /* lower threshold */
759
760 /* for destination cache */
761 spinlock_t dst_lock; /* lock of dst_cache */
762 struct ip_vs_dest_dst __rcu *dest_dst; /* cached dst info */
763
764 /* for virtual service */
765 struct ip_vs_service __rcu *svc; /* service it belongs to */
766 __u16 protocol; /* which protocol (TCP/UDP) */
767 __be16 vport; /* virtual port number */
768 union nf_inet_addr vaddr; /* virtual IP address */
769 __u32 vfwmark; /* firewall mark of service */
770
771 struct rcu_head rcu_head;
772 struct list_head t_list; /* in dest_trash */
773 unsigned int in_rs_table:1; /* we are in rs_table */
774};
775
776/* The scheduler object */
777struct ip_vs_scheduler {
778 struct list_head n_list; /* d-linked list head */
779 char *name; /* scheduler name */
780 atomic_t refcnt; /* reference counter */
781 struct module *module; /* THIS_MODULE/NULL */
782
783 /* scheduler initializing service */
784 int (*init_service)(struct ip_vs_service *svc);
785 /* scheduling service finish */
786 void (*done_service)(struct ip_vs_service *svc);
787 /* dest is linked */
788 int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
789 /* dest is unlinked */
790 int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
791 /* dest is updated */
792 int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
793
794 /* selecting a server from the given service */
795 struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc,
796 const struct sk_buff *skb,
797 struct ip_vs_iphdr *iph);
798};
799
800/* The persistence engine object */
801struct ip_vs_pe {
802 struct list_head n_list; /* d-linked list head */
803 char *name; /* scheduler name */
804 atomic_t refcnt; /* reference counter */
805 struct module *module; /* THIS_MODULE/NULL */
806
807 /* get the connection template, if any */
808 int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb);
809 bool (*ct_match)(const struct ip_vs_conn_param *p,
810 struct ip_vs_conn *ct);
811 u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval,
812 bool inverse);
813 int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf);
814 /* create connections for real-server outgoing packets */
815 struct ip_vs_conn* (*conn_out)(struct ip_vs_service *svc,
816 struct ip_vs_dest *dest,
817 struct sk_buff *skb,
818 const struct ip_vs_iphdr *iph,
819 __be16 dport, __be16 cport);
820};
821
822/* The application module object (a.k.a. app incarnation) */
823struct ip_vs_app {
824 struct list_head a_list; /* member in app list */
825 int type; /* IP_VS_APP_TYPE_xxx */
826 char *name; /* application module name */
827 __u16 protocol;
828 struct module *module; /* THIS_MODULE/NULL */
829 struct list_head incs_list; /* list of incarnations */
830
831 /* members for application incarnations */
832 struct list_head p_list; /* member in proto app list */
833 struct ip_vs_app *app; /* its real application */
834 __be16 port; /* port number in net order */
835 atomic_t usecnt; /* usage counter */
836 struct rcu_head rcu_head;
837
838 /* output hook: Process packet in inout direction, diff set for TCP.
839 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
840 * 2=Mangled but checksum was not updated
841 */
842 int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *,
843 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh);
844
845 /* input hook: Process packet in outin direction, diff set for TCP.
846 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
847 * 2=Mangled but checksum was not updated
848 */
849 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *,
850 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh);
851
852 /* ip_vs_app initializer */
853 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *);
854
855 /* ip_vs_app finish */
856 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *);
857
858
859 /* not used now */
860 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *,
861 struct ip_vs_protocol *);
862
863 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *);
864
865 int * timeout_table;
866 int * timeouts;
867 int timeouts_size;
868
869 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app,
870 int *verdict, struct ip_vs_conn **cpp);
871
872 struct ip_vs_conn *
873 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app,
874 const struct iphdr *iph, int inverse);
875
876 struct ip_vs_conn *
877 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app,
878 const struct iphdr *iph, int inverse);
879
880 int (*state_transition)(struct ip_vs_conn *cp, int direction,
881 const struct sk_buff *skb,
882 struct ip_vs_app *app);
883
884 void (*timeout_change)(struct ip_vs_app *app, int flags);
885};
886
887struct ipvs_master_sync_state {
888 struct list_head sync_queue;
889 struct ip_vs_sync_buff *sync_buff;
890 unsigned long sync_queue_len;
891 unsigned int sync_queue_delay;
892 struct delayed_work master_wakeup_work;
893 struct netns_ipvs *ipvs;
894};
895
896struct ip_vs_sync_thread_data;
897
898/* How much time to keep dests in trash */
899#define IP_VS_DEST_TRASH_PERIOD (120 * HZ)
900
901struct ipvs_sync_daemon_cfg {
902 union nf_inet_addr mcast_group;
903 int syncid;
904 u16 sync_maxlen;
905 u16 mcast_port;
906 u8 mcast_af;
907 u8 mcast_ttl;
908 /* multicast interface name */
909 char mcast_ifn[IP_VS_IFNAME_MAXLEN];
910};
911
912/* IPVS in network namespace */
913struct netns_ipvs {
914 int gen; /* Generation */
915 int enable; /* enable like nf_hooks do */
916 /* Hash table: for real service lookups */
917 #define IP_VS_RTAB_BITS 4
918 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
919 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
920
921 struct hlist_head rs_table[IP_VS_RTAB_SIZE];
922 /* ip_vs_app */
923 struct list_head app_list;
924 /* ip_vs_proto */
925 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */
926 struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE];
927 /* ip_vs_proto_tcp */
928#ifdef CONFIG_IP_VS_PROTO_TCP
929 #define TCP_APP_TAB_BITS 4
930 #define TCP_APP_TAB_SIZE (1 << TCP_APP_TAB_BITS)
931 #define TCP_APP_TAB_MASK (TCP_APP_TAB_SIZE - 1)
932 struct list_head tcp_apps[TCP_APP_TAB_SIZE];
933#endif
934 /* ip_vs_proto_udp */
935#ifdef CONFIG_IP_VS_PROTO_UDP
936 #define UDP_APP_TAB_BITS 4
937 #define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS)
938 #define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1)
939 struct list_head udp_apps[UDP_APP_TAB_SIZE];
940#endif
941 /* ip_vs_proto_sctp */
942#ifdef CONFIG_IP_VS_PROTO_SCTP
943 #define SCTP_APP_TAB_BITS 4
944 #define SCTP_APP_TAB_SIZE (1 << SCTP_APP_TAB_BITS)
945 #define SCTP_APP_TAB_MASK (SCTP_APP_TAB_SIZE - 1)
946 /* Hash table for SCTP application incarnations */
947 struct list_head sctp_apps[SCTP_APP_TAB_SIZE];
948#endif
949 /* ip_vs_conn */
950 atomic_t conn_count; /* connection counter */
951
952 /* ip_vs_ctl */
953 struct ip_vs_stats_rcu *tot_stats; /* Statistics & est. */
954
955 int num_services; /* no of virtual services */
956 int num_services6; /* IPv6 virtual services */
957
958 /* Trash for destinations */
959 struct list_head dest_trash;
960 spinlock_t dest_trash_lock;
961 struct timer_list dest_trash_timer; /* expiration timer */
962 /* Service counters */
963 atomic_t ftpsvc_counter;
964 atomic_t nullsvc_counter;
965 atomic_t conn_out_counter;
966
967#ifdef CONFIG_SYSCTL
968 /* delayed work for expiring no dest connections */
969 struct delayed_work expire_nodest_conn_work;
970 /* 1/rate drop and drop-entry variables */
971 struct delayed_work defense_work; /* Work handler */
972 int drop_rate;
973 int drop_counter;
974 int old_secure_tcp;
975 atomic_t dropentry;
976 /* locks in ctl.c */
977 spinlock_t dropentry_lock; /* drop entry handling */
978 spinlock_t droppacket_lock; /* drop packet handling */
979 spinlock_t securetcp_lock; /* state and timeout tables */
980
981 /* sys-ctl struct */
982 struct ctl_table_header *sysctl_hdr;
983 struct ctl_table *sysctl_tbl;
984#endif
985
986 /* sysctl variables */
987 int sysctl_amemthresh;
988 int sysctl_am_droprate;
989 int sysctl_drop_entry;
990 int sysctl_drop_packet;
991 int sysctl_secure_tcp;
992#ifdef CONFIG_IP_VS_NFCT
993 int sysctl_conntrack;
994#endif
995 int sysctl_snat_reroute;
996 int sysctl_sync_ver;
997 int sysctl_sync_ports;
998 int sysctl_sync_persist_mode;
999 unsigned long sysctl_sync_qlen_max;
1000 int sysctl_sync_sock_size;
1001 int sysctl_cache_bypass;
1002 int sysctl_expire_nodest_conn;
1003 int sysctl_sloppy_tcp;
1004 int sysctl_sloppy_sctp;
1005 int sysctl_expire_quiescent_template;
1006 int sysctl_sync_threshold[2];
1007 unsigned int sysctl_sync_refresh_period;
1008 int sysctl_sync_retries;
1009 int sysctl_nat_icmp_send;
1010 int sysctl_pmtu_disc;
1011 int sysctl_backup_only;
1012 int sysctl_conn_reuse_mode;
1013 int sysctl_schedule_icmp;
1014 int sysctl_ignore_tunneled;
1015 int sysctl_run_estimation;
1016#ifdef CONFIG_SYSCTL
1017 cpumask_var_t sysctl_est_cpulist; /* kthread cpumask */
1018 int est_cpulist_valid; /* cpulist set */
1019 int sysctl_est_nice; /* kthread nice */
1020 int est_stopped; /* stop tasks */
1021#endif
1022
1023 /* ip_vs_lblc */
1024 int sysctl_lblc_expiration;
1025 struct ctl_table_header *lblc_ctl_header;
1026 struct ctl_table *lblc_ctl_table;
1027 /* ip_vs_lblcr */
1028 int sysctl_lblcr_expiration;
1029 struct ctl_table_header *lblcr_ctl_header;
1030 struct ctl_table *lblcr_ctl_table;
1031 /* ip_vs_est */
1032 struct delayed_work est_reload_work;/* Reload kthread tasks */
1033 struct mutex est_mutex; /* protect kthread tasks */
1034 struct hlist_head est_temp_list; /* Ests during calc phase */
1035 struct ip_vs_est_kt_data **est_kt_arr; /* Array of kthread data ptrs */
1036 unsigned long est_max_threads;/* Hard limit of kthreads */
1037 int est_calc_phase; /* Calculation phase */
1038 int est_chain_max; /* Calculated chain_max */
1039 int est_kt_count; /* Allocated ptrs */
1040 int est_add_ktid; /* ktid where to add ests */
1041 atomic_t est_genid; /* kthreads reload genid */
1042 atomic_t est_genid_done; /* applied genid */
1043 /* ip_vs_sync */
1044 spinlock_t sync_lock;
1045 struct ipvs_master_sync_state *ms;
1046 spinlock_t sync_buff_lock;
1047 struct ip_vs_sync_thread_data *master_tinfo;
1048 struct ip_vs_sync_thread_data *backup_tinfo;
1049 int threads_mask;
1050 volatile int sync_state;
1051 struct mutex sync_mutex;
1052 struct ipvs_sync_daemon_cfg mcfg; /* Master Configuration */
1053 struct ipvs_sync_daemon_cfg bcfg; /* Backup Configuration */
1054 /* net name space ptr */
1055 struct net *net; /* Needed by timer routines */
1056 /* Number of heterogeneous destinations, needed becaus heterogeneous
1057 * are not supported when synchronization is enabled.
1058 */
1059 unsigned int mixed_address_family_dests;
1060 unsigned int hooks_afmask; /* &1=AF_INET, &2=AF_INET6 */
1061};
1062
1063#define DEFAULT_SYNC_THRESHOLD 3
1064#define DEFAULT_SYNC_PERIOD 50
1065#define DEFAULT_SYNC_VER 1
1066#define DEFAULT_SLOPPY_TCP 0
1067#define DEFAULT_SLOPPY_SCTP 0
1068#define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ)
1069#define DEFAULT_SYNC_RETRIES 0
1070#define IPVS_SYNC_WAKEUP_RATE 8
1071#define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4)
1072#define IPVS_SYNC_SEND_DELAY (HZ / 50)
1073#define IPVS_SYNC_CHECK_PERIOD HZ
1074#define IPVS_SYNC_FLUSH_TIME (HZ * 2)
1075#define IPVS_SYNC_PORTS_MAX (1 << 6)
1076
1077#ifdef CONFIG_SYSCTL
1078
1079static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1080{
1081 return ipvs->sysctl_sync_threshold[0];
1082}
1083
1084static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1085{
1086 return READ_ONCE(ipvs->sysctl_sync_threshold[1]);
1087}
1088
1089static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1090{
1091 return READ_ONCE(ipvs->sysctl_sync_refresh_period);
1092}
1093
1094static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1095{
1096 return ipvs->sysctl_sync_retries;
1097}
1098
1099static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1100{
1101 return ipvs->sysctl_sync_ver;
1102}
1103
1104static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1105{
1106 return ipvs->sysctl_sloppy_tcp;
1107}
1108
1109static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1110{
1111 return ipvs->sysctl_sloppy_sctp;
1112}
1113
1114static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1115{
1116 return READ_ONCE(ipvs->sysctl_sync_ports);
1117}
1118
1119static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1120{
1121 return ipvs->sysctl_sync_persist_mode;
1122}
1123
1124static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1125{
1126 return ipvs->sysctl_sync_qlen_max;
1127}
1128
1129static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1130{
1131 return ipvs->sysctl_sync_sock_size;
1132}
1133
1134static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1135{
1136 return ipvs->sysctl_pmtu_disc;
1137}
1138
1139static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1140{
1141 return ipvs->sync_state & IP_VS_STATE_BACKUP &&
1142 ipvs->sysctl_backup_only;
1143}
1144
1145static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1146{
1147 return ipvs->sysctl_conn_reuse_mode;
1148}
1149
1150static inline int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs)
1151{
1152 return ipvs->sysctl_expire_nodest_conn;
1153}
1154
1155static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1156{
1157 return ipvs->sysctl_schedule_icmp;
1158}
1159
1160static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1161{
1162 return ipvs->sysctl_ignore_tunneled;
1163}
1164
1165static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1166{
1167 return ipvs->sysctl_cache_bypass;
1168}
1169
1170static inline int sysctl_run_estimation(struct netns_ipvs *ipvs)
1171{
1172 return ipvs->sysctl_run_estimation;
1173}
1174
1175static inline const struct cpumask *sysctl_est_cpulist(struct netns_ipvs *ipvs)
1176{
1177 if (ipvs->est_cpulist_valid)
1178 return ipvs->sysctl_est_cpulist;
1179 else
1180 return housekeeping_cpumask(HK_TYPE_KTHREAD);
1181}
1182
1183static inline int sysctl_est_nice(struct netns_ipvs *ipvs)
1184{
1185 return ipvs->sysctl_est_nice;
1186}
1187
1188#else
1189
1190static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1191{
1192 return DEFAULT_SYNC_THRESHOLD;
1193}
1194
1195static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1196{
1197 return DEFAULT_SYNC_PERIOD;
1198}
1199
1200static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1201{
1202 return DEFAULT_SYNC_REFRESH_PERIOD;
1203}
1204
1205static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1206{
1207 return DEFAULT_SYNC_RETRIES & 3;
1208}
1209
1210static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1211{
1212 return DEFAULT_SYNC_VER;
1213}
1214
1215static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1216{
1217 return DEFAULT_SLOPPY_TCP;
1218}
1219
1220static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1221{
1222 return DEFAULT_SLOPPY_SCTP;
1223}
1224
1225static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1226{
1227 return 1;
1228}
1229
1230static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1231{
1232 return 0;
1233}
1234
1235static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1236{
1237 return IPVS_SYNC_QLEN_MAX;
1238}
1239
1240static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1241{
1242 return 0;
1243}
1244
1245static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1246{
1247 return 1;
1248}
1249
1250static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1251{
1252 return 0;
1253}
1254
1255static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1256{
1257 return 1;
1258}
1259
1260static inline int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs)
1261{
1262 return 0;
1263}
1264
1265static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1266{
1267 return 0;
1268}
1269
1270static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1271{
1272 return 0;
1273}
1274
1275static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1276{
1277 return 0;
1278}
1279
1280static inline int sysctl_run_estimation(struct netns_ipvs *ipvs)
1281{
1282 return 1;
1283}
1284
1285static inline const struct cpumask *sysctl_est_cpulist(struct netns_ipvs *ipvs)
1286{
1287 return housekeeping_cpumask(HK_TYPE_KTHREAD);
1288}
1289
1290static inline int sysctl_est_nice(struct netns_ipvs *ipvs)
1291{
1292 return IPVS_EST_NICE;
1293}
1294
1295#endif
1296
1297/* IPVS core functions
1298 * (from ip_vs_core.c)
1299 */
1300const char *ip_vs_proto_name(unsigned int proto);
1301void ip_vs_init_hash_table(struct list_head *table, int rows);
1302struct ip_vs_conn *ip_vs_new_conn_out(struct ip_vs_service *svc,
1303 struct ip_vs_dest *dest,
1304 struct sk_buff *skb,
1305 const struct ip_vs_iphdr *iph,
1306 __be16 dport,
1307 __be16 cport);
1308#define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t)))
1309
1310#define IP_VS_APP_TYPE_FTP 1
1311
1312/* ip_vs_conn handling functions
1313 * (from ip_vs_conn.c)
1314 */
1315enum {
1316 IP_VS_DIR_INPUT = 0,
1317 IP_VS_DIR_OUTPUT,
1318 IP_VS_DIR_INPUT_ONLY,
1319 IP_VS_DIR_LAST,
1320};
1321
1322static inline void ip_vs_conn_fill_param(struct netns_ipvs *ipvs, int af, int protocol,
1323 const union nf_inet_addr *caddr,
1324 __be16 cport,
1325 const union nf_inet_addr *vaddr,
1326 __be16 vport,
1327 struct ip_vs_conn_param *p)
1328{
1329 p->ipvs = ipvs;
1330 p->af = af;
1331 p->protocol = protocol;
1332 p->caddr = caddr;
1333 p->cport = cport;
1334 p->vaddr = vaddr;
1335 p->vport = vport;
1336 p->pe = NULL;
1337 p->pe_data = NULL;
1338}
1339
1340struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p);
1341struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p);
1342
1343struct ip_vs_conn * ip_vs_conn_in_get_proto(struct netns_ipvs *ipvs, int af,
1344 const struct sk_buff *skb,
1345 const struct ip_vs_iphdr *iph);
1346
1347struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p);
1348
1349struct ip_vs_conn * ip_vs_conn_out_get_proto(struct netns_ipvs *ipvs, int af,
1350 const struct sk_buff *skb,
1351 const struct ip_vs_iphdr *iph);
1352
1353/* Get reference to gain full access to conn.
1354 * By default, RCU read-side critical sections have access only to
1355 * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference.
1356 */
1357static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp)
1358{
1359 return refcount_inc_not_zero(&cp->refcnt);
1360}
1361
1362/* put back the conn without restarting its timer */
1363static inline void __ip_vs_conn_put(struct ip_vs_conn *cp)
1364{
1365 smp_mb__before_atomic();
1366 refcount_dec(&cp->refcnt);
1367}
1368void ip_vs_conn_put(struct ip_vs_conn *cp);
1369void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport);
1370
1371struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, int dest_af,
1372 const union nf_inet_addr *daddr,
1373 __be16 dport, unsigned int flags,
1374 struct ip_vs_dest *dest, __u32 fwmark);
1375void ip_vs_conn_expire_now(struct ip_vs_conn *cp);
1376
1377const char *ip_vs_state_name(const struct ip_vs_conn *cp);
1378
1379void ip_vs_tcp_conn_listen(struct ip_vs_conn *cp);
1380int ip_vs_check_template(struct ip_vs_conn *ct, struct ip_vs_dest *cdest);
1381void ip_vs_random_dropentry(struct netns_ipvs *ipvs);
1382int ip_vs_conn_init(void);
1383void ip_vs_conn_cleanup(void);
1384
1385static inline void ip_vs_control_del(struct ip_vs_conn *cp)
1386{
1387 struct ip_vs_conn *ctl_cp = cp->control;
1388 if (!ctl_cp) {
1389 IP_VS_ERR_BUF("request control DEL for uncontrolled: "
1390 "%s:%d to %s:%d\n",
1391 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1392 ntohs(cp->cport),
1393 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1394 ntohs(cp->vport));
1395
1396 return;
1397 }
1398
1399 IP_VS_DBG_BUF(7, "DELeting control for: "
1400 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1401 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1402 ntohs(cp->cport),
1403 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1404 ntohs(ctl_cp->cport));
1405
1406 cp->control = NULL;
1407 if (atomic_read(&ctl_cp->n_control) == 0) {
1408 IP_VS_ERR_BUF("BUG control DEL with n=0 : "
1409 "%s:%d to %s:%d\n",
1410 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1411 ntohs(cp->cport),
1412 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1413 ntohs(cp->vport));
1414
1415 return;
1416 }
1417 atomic_dec(&ctl_cp->n_control);
1418}
1419
1420static inline void
1421ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp)
1422{
1423 if (cp->control) {
1424 IP_VS_ERR_BUF("request control ADD for already controlled: "
1425 "%s:%d to %s:%d\n",
1426 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1427 ntohs(cp->cport),
1428 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1429 ntohs(cp->vport));
1430
1431 ip_vs_control_del(cp);
1432 }
1433
1434 IP_VS_DBG_BUF(7, "ADDing control for: "
1435 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1436 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1437 ntohs(cp->cport),
1438 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1439 ntohs(ctl_cp->cport));
1440
1441 cp->control = ctl_cp;
1442 atomic_inc(&ctl_cp->n_control);
1443}
1444
1445/* Mark our template as assured */
1446static inline void
1447ip_vs_control_assure_ct(struct ip_vs_conn *cp)
1448{
1449 struct ip_vs_conn *ct = cp->control;
1450
1451 if (ct && !(ct->state & IP_VS_CTPL_S_ASSURED) &&
1452 (ct->flags & IP_VS_CONN_F_TEMPLATE))
1453 ct->state |= IP_VS_CTPL_S_ASSURED;
1454}
1455
1456/* IPVS netns init & cleanup functions */
1457int ip_vs_estimator_net_init(struct netns_ipvs *ipvs);
1458int ip_vs_control_net_init(struct netns_ipvs *ipvs);
1459int ip_vs_protocol_net_init(struct netns_ipvs *ipvs);
1460int ip_vs_app_net_init(struct netns_ipvs *ipvs);
1461int ip_vs_conn_net_init(struct netns_ipvs *ipvs);
1462int ip_vs_sync_net_init(struct netns_ipvs *ipvs);
1463void ip_vs_conn_net_cleanup(struct netns_ipvs *ipvs);
1464void ip_vs_app_net_cleanup(struct netns_ipvs *ipvs);
1465void ip_vs_protocol_net_cleanup(struct netns_ipvs *ipvs);
1466void ip_vs_control_net_cleanup(struct netns_ipvs *ipvs);
1467void ip_vs_estimator_net_cleanup(struct netns_ipvs *ipvs);
1468void ip_vs_sync_net_cleanup(struct netns_ipvs *ipvs);
1469void ip_vs_service_nets_cleanup(struct list_head *net_list);
1470
1471/* IPVS application functions
1472 * (from ip_vs_app.c)
1473 */
1474#define IP_VS_APP_MAX_PORTS 8
1475struct ip_vs_app *register_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1476void unregister_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1477int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1478void ip_vs_unbind_app(struct ip_vs_conn *cp);
1479int register_ip_vs_app_inc(struct netns_ipvs *ipvs, struct ip_vs_app *app, __u16 proto,
1480 __u16 port);
1481int ip_vs_app_inc_get(struct ip_vs_app *inc);
1482void ip_vs_app_inc_put(struct ip_vs_app *inc);
1483
1484int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb,
1485 struct ip_vs_iphdr *ipvsh);
1486int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb,
1487 struct ip_vs_iphdr *ipvsh);
1488
1489int register_ip_vs_pe(struct ip_vs_pe *pe);
1490int unregister_ip_vs_pe(struct ip_vs_pe *pe);
1491struct ip_vs_pe *ip_vs_pe_getbyname(const char *name);
1492struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name);
1493
1494/* Use a #define to avoid all of module.h just for these trivial ops */
1495#define ip_vs_pe_get(pe) \
1496 if (pe && pe->module) \
1497 __module_get(pe->module);
1498
1499#define ip_vs_pe_put(pe) \
1500 if (pe && pe->module) \
1501 module_put(pe->module);
1502
1503/* IPVS protocol functions (from ip_vs_proto.c) */
1504int ip_vs_protocol_init(void);
1505void ip_vs_protocol_cleanup(void);
1506void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags);
1507int *ip_vs_create_timeout_table(int *table, int size);
1508void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
1509 const struct sk_buff *skb, int offset,
1510 const char *msg);
1511
1512extern struct ip_vs_protocol ip_vs_protocol_tcp;
1513extern struct ip_vs_protocol ip_vs_protocol_udp;
1514extern struct ip_vs_protocol ip_vs_protocol_icmp;
1515extern struct ip_vs_protocol ip_vs_protocol_esp;
1516extern struct ip_vs_protocol ip_vs_protocol_ah;
1517extern struct ip_vs_protocol ip_vs_protocol_sctp;
1518
1519/* Registering/unregistering scheduler functions
1520 * (from ip_vs_sched.c)
1521 */
1522int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1523int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1524int ip_vs_bind_scheduler(struct ip_vs_service *svc,
1525 struct ip_vs_scheduler *scheduler);
1526void ip_vs_unbind_scheduler(struct ip_vs_service *svc,
1527 struct ip_vs_scheduler *sched);
1528struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name);
1529void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler);
1530struct ip_vs_conn *
1531ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb,
1532 struct ip_vs_proto_data *pd, int *ignored,
1533 struct ip_vs_iphdr *iph);
1534int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb,
1535 struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph);
1536
1537void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg);
1538
1539/* IPVS control data and functions (from ip_vs_ctl.c) */
1540extern struct ip_vs_stats ip_vs_stats;
1541extern int sysctl_ip_vs_sync_ver;
1542
1543struct ip_vs_service *
1544ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
1545 const union nf_inet_addr *vaddr, __be16 vport);
1546
1547bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1548 const union nf_inet_addr *daddr, __be16 dport);
1549
1550struct ip_vs_dest *
1551ip_vs_find_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1552 const union nf_inet_addr *daddr, __be16 dport);
1553struct ip_vs_dest *ip_vs_find_tunnel(struct netns_ipvs *ipvs, int af,
1554 const union nf_inet_addr *daddr,
1555 __be16 tun_port);
1556
1557int ip_vs_use_count_inc(void);
1558void ip_vs_use_count_dec(void);
1559int ip_vs_register_nl_ioctl(void);
1560void ip_vs_unregister_nl_ioctl(void);
1561int ip_vs_control_init(void);
1562void ip_vs_control_cleanup(void);
1563struct ip_vs_dest *
1564ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
1565 const union nf_inet_addr *daddr, __be16 dport,
1566 const union nf_inet_addr *vaddr, __be16 vport,
1567 __u16 protocol, __u32 fwmark, __u32 flags);
1568void ip_vs_try_bind_dest(struct ip_vs_conn *cp);
1569
1570static inline void ip_vs_dest_hold(struct ip_vs_dest *dest)
1571{
1572 refcount_inc(&dest->refcnt);
1573}
1574
1575static inline void ip_vs_dest_put(struct ip_vs_dest *dest)
1576{
1577 smp_mb__before_atomic();
1578 refcount_dec(&dest->refcnt);
1579}
1580
1581static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest)
1582{
1583 if (refcount_dec_and_test(&dest->refcnt))
1584 kfree(dest);
1585}
1586
1587/* IPVS sync daemon data and function prototypes
1588 * (from ip_vs_sync.c)
1589 */
1590int start_sync_thread(struct netns_ipvs *ipvs, struct ipvs_sync_daemon_cfg *cfg,
1591 int state);
1592int stop_sync_thread(struct netns_ipvs *ipvs, int state);
1593void ip_vs_sync_conn(struct netns_ipvs *ipvs, struct ip_vs_conn *cp, int pkts);
1594
1595/* IPVS rate estimator prototypes (from ip_vs_est.c) */
1596int ip_vs_start_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1597void ip_vs_stop_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1598void ip_vs_zero_estimator(struct ip_vs_stats *stats);
1599void ip_vs_read_estimator(struct ip_vs_kstats *dst, struct ip_vs_stats *stats);
1600void ip_vs_est_reload_start(struct netns_ipvs *ipvs);
1601int ip_vs_est_kthread_start(struct netns_ipvs *ipvs,
1602 struct ip_vs_est_kt_data *kd);
1603void ip_vs_est_kthread_stop(struct ip_vs_est_kt_data *kd);
1604
1605static inline void ip_vs_est_stopped_recalc(struct netns_ipvs *ipvs)
1606{
1607#ifdef CONFIG_SYSCTL
1608 /* Stop tasks while cpulist is empty or if disabled with flag */
1609 ipvs->est_stopped = !sysctl_run_estimation(ipvs) ||
1610 (ipvs->est_cpulist_valid &&
1611 cpumask_empty(sysctl_est_cpulist(ipvs)));
1612#endif
1613}
1614
1615static inline bool ip_vs_est_stopped(struct netns_ipvs *ipvs)
1616{
1617#ifdef CONFIG_SYSCTL
1618 return ipvs->est_stopped;
1619#else
1620 return false;
1621#endif
1622}
1623
1624static inline int ip_vs_est_max_threads(struct netns_ipvs *ipvs)
1625{
1626 unsigned int limit = IPVS_EST_CPU_KTHREADS *
1627 cpumask_weight(sysctl_est_cpulist(ipvs));
1628
1629 return max(1U, limit);
1630}
1631
1632/* Various IPVS packet transmitters (from ip_vs_xmit.c) */
1633int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1634 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1635int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1636 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1637int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1638 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1639int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1640 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1641int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1642 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1643int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1644 struct ip_vs_protocol *pp, int offset,
1645 unsigned int hooknum, struct ip_vs_iphdr *iph);
1646void ip_vs_dest_dst_rcu_free(struct rcu_head *head);
1647
1648#ifdef CONFIG_IP_VS_IPV6
1649int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1650 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1651int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1652 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1653int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1654 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1655int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1656 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1657int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1658 struct ip_vs_protocol *pp, int offset,
1659 unsigned int hooknum, struct ip_vs_iphdr *iph);
1660#endif
1661
1662#ifdef CONFIG_SYSCTL
1663/* This is a simple mechanism to ignore packets when
1664 * we are loaded. Just set ip_vs_drop_rate to 'n' and
1665 * we start to drop 1/rate of the packets
1666 */
1667static inline int ip_vs_todrop(struct netns_ipvs *ipvs)
1668{
1669 if (!ipvs->drop_rate)
1670 return 0;
1671 if (--ipvs->drop_counter > 0)
1672 return 0;
1673 ipvs->drop_counter = ipvs->drop_rate;
1674 return 1;
1675}
1676#else
1677static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; }
1678#endif
1679
1680#ifdef CONFIG_SYSCTL
1681/* Enqueue delayed work for expiring no dest connections
1682 * Only run when sysctl_expire_nodest=1
1683 */
1684static inline void ip_vs_enqueue_expire_nodest_conns(struct netns_ipvs *ipvs)
1685{
1686 if (sysctl_expire_nodest_conn(ipvs))
1687 queue_delayed_work(system_long_wq,
1688 &ipvs->expire_nodest_conn_work, 1);
1689}
1690
1691void ip_vs_expire_nodest_conn_flush(struct netns_ipvs *ipvs);
1692#else
1693static inline void ip_vs_enqueue_expire_nodest_conns(struct netns_ipvs *ipvs) {}
1694#endif
1695
1696#define IP_VS_DFWD_METHOD(dest) (atomic_read(&(dest)->conn_flags) & \
1697 IP_VS_CONN_F_FWD_MASK)
1698
1699/* ip_vs_fwd_tag returns the forwarding tag of the connection */
1700#define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK)
1701
1702static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp)
1703{
1704 char fwd;
1705
1706 switch (IP_VS_FWD_METHOD(cp)) {
1707 case IP_VS_CONN_F_MASQ:
1708 fwd = 'M'; break;
1709 case IP_VS_CONN_F_LOCALNODE:
1710 fwd = 'L'; break;
1711 case IP_VS_CONN_F_TUNNEL:
1712 fwd = 'T'; break;
1713 case IP_VS_CONN_F_DROUTE:
1714 fwd = 'R'; break;
1715 case IP_VS_CONN_F_BYPASS:
1716 fwd = 'B'; break;
1717 default:
1718 fwd = '?'; break;
1719 }
1720 return fwd;
1721}
1722
1723void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp,
1724 struct ip_vs_conn *cp, int dir);
1725
1726#ifdef CONFIG_IP_VS_IPV6
1727void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp,
1728 struct ip_vs_conn *cp, int dir);
1729#endif
1730
1731__sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset);
1732
1733static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum)
1734{
1735 __be32 diff[2] = { ~old, new };
1736
1737 return csum_partial(diff, sizeof(diff), oldsum);
1738}
1739
1740#ifdef CONFIG_IP_VS_IPV6
1741static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new,
1742 __wsum oldsum)
1743{
1744 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0],
1745 new[3], new[2], new[1], new[0] };
1746
1747 return csum_partial(diff, sizeof(diff), oldsum);
1748}
1749#endif
1750
1751static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum)
1752{
1753 __be16 diff[2] = { ~old, new };
1754
1755 return csum_partial(diff, sizeof(diff), oldsum);
1756}
1757
1758/* Forget current conntrack (unconfirmed) and attach notrack entry */
1759static inline void ip_vs_notrack(struct sk_buff *skb)
1760{
1761#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1762 enum ip_conntrack_info ctinfo;
1763 struct nf_conn *ct = nf_ct_get(skb, &ctinfo);
1764
1765 if (ct) {
1766 nf_conntrack_put(&ct->ct_general);
1767 nf_ct_set(skb, NULL, IP_CT_UNTRACKED);
1768 }
1769#endif
1770}
1771
1772#ifdef CONFIG_IP_VS_NFCT
1773/* Netfilter connection tracking
1774 * (from ip_vs_nfct.c)
1775 */
1776static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1777{
1778#ifdef CONFIG_SYSCTL
1779 return ipvs->sysctl_conntrack;
1780#else
1781 return 0;
1782#endif
1783}
1784
1785void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp,
1786 int outin);
1787int ip_vs_confirm_conntrack(struct sk_buff *skb);
1788void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct,
1789 struct ip_vs_conn *cp, u_int8_t proto,
1790 const __be16 port, int from_rs);
1791void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp);
1792
1793#else
1794
1795static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1796{
1797 return 0;
1798}
1799
1800static inline void ip_vs_update_conntrack(struct sk_buff *skb,
1801 struct ip_vs_conn *cp, int outin)
1802{
1803}
1804
1805static inline int ip_vs_confirm_conntrack(struct sk_buff *skb)
1806{
1807 return NF_ACCEPT;
1808}
1809
1810static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp)
1811{
1812}
1813#endif /* CONFIG_IP_VS_NFCT */
1814
1815/* Using old conntrack that can not be redirected to another real server? */
1816static inline bool ip_vs_conn_uses_old_conntrack(struct ip_vs_conn *cp,
1817 struct sk_buff *skb)
1818{
1819#ifdef CONFIG_IP_VS_NFCT
1820 enum ip_conntrack_info ctinfo;
1821 struct nf_conn *ct;
1822
1823 ct = nf_ct_get(skb, &ctinfo);
1824 if (ct && nf_ct_is_confirmed(ct))
1825 return true;
1826#endif
1827 return false;
1828}
1829
1830static inline int ip_vs_register_conntrack(struct ip_vs_service *svc)
1831{
1832#if IS_ENABLED(CONFIG_NF_CONNTRACK)
1833 int afmask = (svc->af == AF_INET6) ? 2 : 1;
1834 int ret = 0;
1835
1836 if (!(svc->conntrack_afmask & afmask)) {
1837 ret = nf_ct_netns_get(svc->ipvs->net, svc->af);
1838 if (ret >= 0)
1839 svc->conntrack_afmask |= afmask;
1840 }
1841 return ret;
1842#else
1843 return 0;
1844#endif
1845}
1846
1847static inline void ip_vs_unregister_conntrack(struct ip_vs_service *svc)
1848{
1849#if IS_ENABLED(CONFIG_NF_CONNTRACK)
1850 int afmask = (svc->af == AF_INET6) ? 2 : 1;
1851
1852 if (svc->conntrack_afmask & afmask) {
1853 nf_ct_netns_put(svc->ipvs->net, svc->af);
1854 svc->conntrack_afmask &= ~afmask;
1855 }
1856#endif
1857}
1858
1859int ip_vs_register_hooks(struct netns_ipvs *ipvs, unsigned int af);
1860void ip_vs_unregister_hooks(struct netns_ipvs *ipvs, unsigned int af);
1861
1862static inline int
1863ip_vs_dest_conn_overhead(struct ip_vs_dest *dest)
1864{
1865 /* We think the overhead of processing active connections is 256
1866 * times higher than that of inactive connections in average. (This
1867 * 256 times might not be accurate, we will change it later) We
1868 * use the following formula to estimate the overhead now:
1869 * dest->activeconns*256 + dest->inactconns
1870 */
1871 return (atomic_read(&dest->activeconns) << 8) +
1872 atomic_read(&dest->inactconns);
1873}
1874
1875#ifdef CONFIG_IP_VS_PROTO_TCP
1876INDIRECT_CALLABLE_DECLARE(int
1877 tcp_snat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
1878 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph));
1879#endif
1880
1881#ifdef CONFIG_IP_VS_PROTO_UDP
1882INDIRECT_CALLABLE_DECLARE(int
1883 udp_snat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
1884 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph));
1885#endif
1886#endif /* _NET_IP_VS_H */
1/*
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/compiler.h>
17#include <linux/timer.h>
18#include <linux/bug.h>
19
20#include <net/checksum.h>
21#include <linux/netfilter.h> /* for union nf_inet_addr */
22#include <linux/ip.h>
23#include <linux/ipv6.h> /* for struct ipv6hdr */
24#include <net/ipv6.h>
25#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
26#include <net/netfilter/nf_conntrack.h>
27#endif
28#include <net/net_namespace.h> /* Netw namespace */
29
30/*
31 * Generic access of ipvs struct
32 */
33static inline struct netns_ipvs *net_ipvs(struct net* net)
34{
35 return net->ipvs;
36}
37/*
38 * Get net ptr from skb in traffic cases
39 * use skb_sknet when call is from userland (ioctl or netlink)
40 */
41static inline struct net *skb_net(const struct sk_buff *skb)
42{
43#ifdef CONFIG_NET_NS
44#ifdef CONFIG_IP_VS_DEBUG
45 /*
46 * This is used for debug only.
47 * Start with the most likely hit
48 * End with BUG
49 */
50 if (likely(skb->dev && skb->dev->nd_net))
51 return dev_net(skb->dev);
52 if (skb_dst(skb) && skb_dst(skb)->dev)
53 return dev_net(skb_dst(skb)->dev);
54 WARN(skb->sk, "Maybe skb_sknet should be used in %s() at line:%d\n",
55 __func__, __LINE__);
56 if (likely(skb->sk && skb->sk->sk_net))
57 return sock_net(skb->sk);
58 pr_err("There is no net ptr to find in the skb in %s() line:%d\n",
59 __func__, __LINE__);
60 BUG();
61#else
62 return dev_net(skb->dev ? : skb_dst(skb)->dev);
63#endif
64#else
65 return &init_net;
66#endif
67}
68
69static inline struct net *skb_sknet(const struct sk_buff *skb)
70{
71#ifdef CONFIG_NET_NS
72#ifdef CONFIG_IP_VS_DEBUG
73 /* Start with the most likely hit */
74 if (likely(skb->sk && skb->sk->sk_net))
75 return sock_net(skb->sk);
76 WARN(skb->dev, "Maybe skb_net should be used instead in %s() line:%d\n",
77 __func__, __LINE__);
78 if (likely(skb->dev && skb->dev->nd_net))
79 return dev_net(skb->dev);
80 pr_err("There is no net ptr to find in the skb in %s() line:%d\n",
81 __func__, __LINE__);
82 BUG();
83#else
84 return sock_net(skb->sk);
85#endif
86#else
87 return &init_net;
88#endif
89}
90/*
91 * This one needed for single_open_net since net is stored directly in
92 * private not as a struct i.e. seq_file_net can't be used.
93 */
94static inline struct net *seq_file_single_net(struct seq_file *seq)
95{
96#ifdef CONFIG_NET_NS
97 return (struct net *)seq->private;
98#else
99 return &init_net;
100#endif
101}
102
103/* Connections' size value needed by ip_vs_ctl.c */
104extern int ip_vs_conn_tab_size;
105
106
107struct ip_vs_iphdr {
108 int len;
109 __u8 protocol;
110 union nf_inet_addr saddr;
111 union nf_inet_addr daddr;
112};
113
114static inline void
115ip_vs_fill_iphdr(int af, const void *nh, struct ip_vs_iphdr *iphdr)
116{
117#ifdef CONFIG_IP_VS_IPV6
118 if (af == AF_INET6) {
119 const struct ipv6hdr *iph = nh;
120 iphdr->len = sizeof(struct ipv6hdr);
121 iphdr->protocol = iph->nexthdr;
122 iphdr->saddr.in6 = iph->saddr;
123 iphdr->daddr.in6 = iph->daddr;
124 } else
125#endif
126 {
127 const struct iphdr *iph = nh;
128 iphdr->len = iph->ihl * 4;
129 iphdr->protocol = iph->protocol;
130 iphdr->saddr.ip = iph->saddr;
131 iphdr->daddr.ip = iph->daddr;
132 }
133}
134
135static inline void ip_vs_addr_copy(int af, union nf_inet_addr *dst,
136 const union nf_inet_addr *src)
137{
138#ifdef CONFIG_IP_VS_IPV6
139 if (af == AF_INET6)
140 dst->in6 = src->in6;
141 else
142#endif
143 dst->ip = src->ip;
144}
145
146static inline int ip_vs_addr_equal(int af, const union nf_inet_addr *a,
147 const union nf_inet_addr *b)
148{
149#ifdef CONFIG_IP_VS_IPV6
150 if (af == AF_INET6)
151 return ipv6_addr_equal(&a->in6, &b->in6);
152#endif
153 return a->ip == b->ip;
154}
155
156#ifdef CONFIG_IP_VS_DEBUG
157#include <linux/net.h>
158
159extern int ip_vs_get_debug_level(void);
160
161static inline const char *ip_vs_dbg_addr(int af, char *buf, size_t buf_len,
162 const union nf_inet_addr *addr,
163 int *idx)
164{
165 int len;
166#ifdef CONFIG_IP_VS_IPV6
167 if (af == AF_INET6)
168 len = snprintf(&buf[*idx], buf_len - *idx, "[%pI6]",
169 &addr->in6) + 1;
170 else
171#endif
172 len = snprintf(&buf[*idx], buf_len - *idx, "%pI4",
173 &addr->ip) + 1;
174
175 *idx += len;
176 BUG_ON(*idx > buf_len + 1);
177 return &buf[*idx - len];
178}
179
180#define IP_VS_DBG_BUF(level, msg, ...) \
181 do { \
182 char ip_vs_dbg_buf[160]; \
183 int ip_vs_dbg_idx = 0; \
184 if (level <= ip_vs_get_debug_level()) \
185 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
186 } while (0)
187#define IP_VS_ERR_BUF(msg...) \
188 do { \
189 char ip_vs_dbg_buf[160]; \
190 int ip_vs_dbg_idx = 0; \
191 pr_err(msg); \
192 } while (0)
193
194/* Only use from within IP_VS_DBG_BUF() or IP_VS_ERR_BUF macros */
195#define IP_VS_DBG_ADDR(af, addr) \
196 ip_vs_dbg_addr(af, ip_vs_dbg_buf, \
197 sizeof(ip_vs_dbg_buf), addr, \
198 &ip_vs_dbg_idx)
199
200#define IP_VS_DBG(level, msg, ...) \
201 do { \
202 if (level <= ip_vs_get_debug_level()) \
203 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
204 } while (0)
205#define IP_VS_DBG_RL(msg, ...) \
206 do { \
207 if (net_ratelimit()) \
208 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
209 } while (0)
210#define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) \
211 do { \
212 if (level <= ip_vs_get_debug_level()) \
213 pp->debug_packet(af, pp, skb, ofs, msg); \
214 } while (0)
215#define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) \
216 do { \
217 if (level <= ip_vs_get_debug_level() && \
218 net_ratelimit()) \
219 pp->debug_packet(af, pp, skb, ofs, msg); \
220 } while (0)
221#else /* NO DEBUGGING at ALL */
222#define IP_VS_DBG_BUF(level, msg...) do {} while (0)
223#define IP_VS_ERR_BUF(msg...) do {} while (0)
224#define IP_VS_DBG(level, msg...) do {} while (0)
225#define IP_VS_DBG_RL(msg...) do {} while (0)
226#define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
227#define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
228#endif
229
230#define IP_VS_BUG() BUG()
231#define IP_VS_ERR_RL(msg, ...) \
232 do { \
233 if (net_ratelimit()) \
234 pr_err(msg, ##__VA_ARGS__); \
235 } while (0)
236
237#ifdef CONFIG_IP_VS_DEBUG
238#define EnterFunction(level) \
239 do { \
240 if (level <= ip_vs_get_debug_level()) \
241 printk(KERN_DEBUG \
242 pr_fmt("Enter: %s, %s line %i\n"), \
243 __func__, __FILE__, __LINE__); \
244 } while (0)
245#define LeaveFunction(level) \
246 do { \
247 if (level <= ip_vs_get_debug_level()) \
248 printk(KERN_DEBUG \
249 pr_fmt("Leave: %s, %s line %i\n"), \
250 __func__, __FILE__, __LINE__); \
251 } while (0)
252#else
253#define EnterFunction(level) do {} while (0)
254#define LeaveFunction(level) do {} while (0)
255#endif
256
257#define IP_VS_WAIT_WHILE(expr) while (expr) { cpu_relax(); }
258
259
260/*
261 * The port number of FTP service (in network order).
262 */
263#define FTPPORT cpu_to_be16(21)
264#define FTPDATA cpu_to_be16(20)
265
266/*
267 * TCP State Values
268 */
269enum {
270 IP_VS_TCP_S_NONE = 0,
271 IP_VS_TCP_S_ESTABLISHED,
272 IP_VS_TCP_S_SYN_SENT,
273 IP_VS_TCP_S_SYN_RECV,
274 IP_VS_TCP_S_FIN_WAIT,
275 IP_VS_TCP_S_TIME_WAIT,
276 IP_VS_TCP_S_CLOSE,
277 IP_VS_TCP_S_CLOSE_WAIT,
278 IP_VS_TCP_S_LAST_ACK,
279 IP_VS_TCP_S_LISTEN,
280 IP_VS_TCP_S_SYNACK,
281 IP_VS_TCP_S_LAST
282};
283
284/*
285 * UDP State Values
286 */
287enum {
288 IP_VS_UDP_S_NORMAL,
289 IP_VS_UDP_S_LAST,
290};
291
292/*
293 * ICMP State Values
294 */
295enum {
296 IP_VS_ICMP_S_NORMAL,
297 IP_VS_ICMP_S_LAST,
298};
299
300/*
301 * SCTP State Values
302 */
303enum ip_vs_sctp_states {
304 IP_VS_SCTP_S_NONE,
305 IP_VS_SCTP_S_INIT_CLI,
306 IP_VS_SCTP_S_INIT_SER,
307 IP_VS_SCTP_S_INIT_ACK_CLI,
308 IP_VS_SCTP_S_INIT_ACK_SER,
309 IP_VS_SCTP_S_ECHO_CLI,
310 IP_VS_SCTP_S_ECHO_SER,
311 IP_VS_SCTP_S_ESTABLISHED,
312 IP_VS_SCTP_S_SHUT_CLI,
313 IP_VS_SCTP_S_SHUT_SER,
314 IP_VS_SCTP_S_SHUT_ACK_CLI,
315 IP_VS_SCTP_S_SHUT_ACK_SER,
316 IP_VS_SCTP_S_CLOSED,
317 IP_VS_SCTP_S_LAST
318};
319
320/*
321 * Delta sequence info structure
322 * Each ip_vs_conn has 2 (output AND input seq. changes).
323 * Only used in the VS/NAT.
324 */
325struct ip_vs_seq {
326 __u32 init_seq; /* Add delta from this seq */
327 __u32 delta; /* Delta in sequence numbers */
328 __u32 previous_delta; /* Delta in sequence numbers
329 before last resized pkt */
330};
331
332/*
333 * counters per cpu
334 */
335struct ip_vs_counters {
336 __u32 conns; /* connections scheduled */
337 __u32 inpkts; /* incoming packets */
338 __u32 outpkts; /* outgoing packets */
339 __u64 inbytes; /* incoming bytes */
340 __u64 outbytes; /* outgoing bytes */
341};
342/*
343 * Stats per cpu
344 */
345struct ip_vs_cpu_stats {
346 struct ip_vs_counters ustats;
347 struct u64_stats_sync syncp;
348};
349
350/*
351 * IPVS statistics objects
352 */
353struct ip_vs_estimator {
354 struct list_head list;
355
356 u64 last_inbytes;
357 u64 last_outbytes;
358 u32 last_conns;
359 u32 last_inpkts;
360 u32 last_outpkts;
361
362 u32 cps;
363 u32 inpps;
364 u32 outpps;
365 u32 inbps;
366 u32 outbps;
367};
368
369struct ip_vs_stats {
370 struct ip_vs_stats_user ustats; /* statistics */
371 struct ip_vs_estimator est; /* estimator */
372 struct ip_vs_cpu_stats *cpustats; /* per cpu counters */
373 spinlock_t lock; /* spin lock */
374 struct ip_vs_stats_user ustats0; /* reset values */
375};
376
377struct dst_entry;
378struct iphdr;
379struct ip_vs_conn;
380struct ip_vs_app;
381struct sk_buff;
382struct ip_vs_proto_data;
383
384struct ip_vs_protocol {
385 struct ip_vs_protocol *next;
386 char *name;
387 u16 protocol;
388 u16 num_states;
389 int dont_defrag;
390
391 void (*init)(struct ip_vs_protocol *pp);
392
393 void (*exit)(struct ip_vs_protocol *pp);
394
395 int (*init_netns)(struct net *net, struct ip_vs_proto_data *pd);
396
397 void (*exit_netns)(struct net *net, struct ip_vs_proto_data *pd);
398
399 int (*conn_schedule)(int af, struct sk_buff *skb,
400 struct ip_vs_proto_data *pd,
401 int *verdict, struct ip_vs_conn **cpp);
402
403 struct ip_vs_conn *
404 (*conn_in_get)(int af,
405 const struct sk_buff *skb,
406 const struct ip_vs_iphdr *iph,
407 unsigned int proto_off,
408 int inverse);
409
410 struct ip_vs_conn *
411 (*conn_out_get)(int af,
412 const struct sk_buff *skb,
413 const struct ip_vs_iphdr *iph,
414 unsigned int proto_off,
415 int inverse);
416
417 int (*snat_handler)(struct sk_buff *skb,
418 struct ip_vs_protocol *pp, struct ip_vs_conn *cp);
419
420 int (*dnat_handler)(struct sk_buff *skb,
421 struct ip_vs_protocol *pp, struct ip_vs_conn *cp);
422
423 int (*csum_check)(int af, struct sk_buff *skb,
424 struct ip_vs_protocol *pp);
425
426 const char *(*state_name)(int state);
427
428 void (*state_transition)(struct ip_vs_conn *cp, int direction,
429 const struct sk_buff *skb,
430 struct ip_vs_proto_data *pd);
431
432 int (*register_app)(struct net *net, struct ip_vs_app *inc);
433
434 void (*unregister_app)(struct net *net, struct ip_vs_app *inc);
435
436 int (*app_conn_bind)(struct ip_vs_conn *cp);
437
438 void (*debug_packet)(int af, struct ip_vs_protocol *pp,
439 const struct sk_buff *skb,
440 int offset,
441 const char *msg);
442
443 void (*timeout_change)(struct ip_vs_proto_data *pd, int flags);
444};
445
446/*
447 * protocol data per netns
448 */
449struct ip_vs_proto_data {
450 struct ip_vs_proto_data *next;
451 struct ip_vs_protocol *pp;
452 int *timeout_table; /* protocol timeout table */
453 atomic_t appcnt; /* counter of proto app incs. */
454 struct tcp_states_t *tcp_state_table;
455};
456
457extern struct ip_vs_protocol *ip_vs_proto_get(unsigned short proto);
458extern struct ip_vs_proto_data *ip_vs_proto_data_get(struct net *net,
459 unsigned short proto);
460
461struct ip_vs_conn_param {
462 struct net *net;
463 const union nf_inet_addr *caddr;
464 const union nf_inet_addr *vaddr;
465 __be16 cport;
466 __be16 vport;
467 __u16 protocol;
468 u16 af;
469
470 const struct ip_vs_pe *pe;
471 char *pe_data;
472 __u8 pe_data_len;
473};
474
475/*
476 * IP_VS structure allocated for each dynamically scheduled connection
477 */
478struct ip_vs_conn {
479 struct hlist_node c_list; /* hashed list heads */
480#ifdef CONFIG_NET_NS
481 struct net *net; /* Name space */
482#endif
483 /* Protocol, addresses and port numbers */
484 u16 af; /* address family */
485 __be16 cport;
486 __be16 vport;
487 __be16 dport;
488 __u32 fwmark; /* Fire wall mark from skb */
489 union nf_inet_addr caddr; /* client address */
490 union nf_inet_addr vaddr; /* virtual address */
491 union nf_inet_addr daddr; /* destination address */
492 volatile __u32 flags; /* status flags */
493 __u16 protocol; /* Which protocol (TCP/UDP) */
494
495 /* counter and timer */
496 atomic_t refcnt; /* reference count */
497 struct timer_list timer; /* Expiration timer */
498 volatile unsigned long timeout; /* timeout */
499
500 /* Flags and state transition */
501 spinlock_t lock; /* lock for state transition */
502 volatile __u16 state; /* state info */
503 volatile __u16 old_state; /* old state, to be used for
504 * state transition triggerd
505 * synchronization
506 */
507 unsigned long sync_endtime; /* jiffies + sent_retries */
508
509 /* Control members */
510 struct ip_vs_conn *control; /* Master control connection */
511 atomic_t n_control; /* Number of controlled ones */
512 struct ip_vs_dest *dest; /* real server */
513 atomic_t in_pkts; /* incoming packet counter */
514
515 /* packet transmitter for different forwarding methods. If it
516 mangles the packet, it must return NF_DROP or better NF_STOLEN,
517 otherwise this must be changed to a sk_buff **.
518 NF_ACCEPT can be returned when destination is local.
519 */
520 int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp,
521 struct ip_vs_protocol *pp);
522
523 /* Note: we can group the following members into a structure,
524 in order to save more space, and the following members are
525 only used in VS/NAT anyway */
526 struct ip_vs_app *app; /* bound ip_vs_app object */
527 void *app_data; /* Application private data */
528 struct ip_vs_seq in_seq; /* incoming seq. struct */
529 struct ip_vs_seq out_seq; /* outgoing seq. struct */
530
531 const struct ip_vs_pe *pe;
532 char *pe_data;
533 __u8 pe_data_len;
534};
535
536/*
537 * To save some memory in conn table when name space is disabled.
538 */
539static inline struct net *ip_vs_conn_net(const struct ip_vs_conn *cp)
540{
541#ifdef CONFIG_NET_NS
542 return cp->net;
543#else
544 return &init_net;
545#endif
546}
547static inline void ip_vs_conn_net_set(struct ip_vs_conn *cp, struct net *net)
548{
549#ifdef CONFIG_NET_NS
550 cp->net = net;
551#endif
552}
553
554static inline int ip_vs_conn_net_eq(const struct ip_vs_conn *cp,
555 struct net *net)
556{
557#ifdef CONFIG_NET_NS
558 return cp->net == net;
559#else
560 return 1;
561#endif
562}
563
564/*
565 * Extended internal versions of struct ip_vs_service_user and
566 * ip_vs_dest_user for IPv6 support.
567 *
568 * We need these to conveniently pass around service and destination
569 * options, but unfortunately, we also need to keep the old definitions to
570 * maintain userspace backwards compatibility for the setsockopt interface.
571 */
572struct ip_vs_service_user_kern {
573 /* virtual service addresses */
574 u16 af;
575 u16 protocol;
576 union nf_inet_addr addr; /* virtual ip address */
577 u16 port;
578 u32 fwmark; /* firwall mark of service */
579
580 /* virtual service options */
581 char *sched_name;
582 char *pe_name;
583 unsigned int flags; /* virtual service flags */
584 unsigned int timeout; /* persistent timeout in sec */
585 u32 netmask; /* persistent netmask */
586};
587
588
589struct ip_vs_dest_user_kern {
590 /* destination server address */
591 union nf_inet_addr addr;
592 u16 port;
593
594 /* real server options */
595 unsigned int conn_flags; /* connection flags */
596 int weight; /* destination weight */
597
598 /* thresholds for active connections */
599 u32 u_threshold; /* upper threshold */
600 u32 l_threshold; /* lower threshold */
601};
602
603
604/*
605 * The information about the virtual service offered to the net
606 * and the forwarding entries
607 */
608struct ip_vs_service {
609 struct list_head s_list; /* for normal service table */
610 struct list_head f_list; /* for fwmark-based service table */
611 atomic_t refcnt; /* reference counter */
612 atomic_t usecnt; /* use counter */
613
614 u16 af; /* address family */
615 __u16 protocol; /* which protocol (TCP/UDP) */
616 union nf_inet_addr addr; /* IP address for virtual service */
617 __be16 port; /* port number for the service */
618 __u32 fwmark; /* firewall mark of the service */
619 unsigned int flags; /* service status flags */
620 unsigned int timeout; /* persistent timeout in ticks */
621 __be32 netmask; /* grouping granularity */
622 struct net *net;
623
624 struct list_head destinations; /* real server d-linked list */
625 __u32 num_dests; /* number of servers */
626 struct ip_vs_stats stats; /* statistics for the service */
627 struct ip_vs_app *inc; /* bind conns to this app inc */
628
629 /* for scheduling */
630 struct ip_vs_scheduler *scheduler; /* bound scheduler object */
631 rwlock_t sched_lock; /* lock sched_data */
632 void *sched_data; /* scheduler application data */
633
634 /* alternate persistence engine */
635 struct ip_vs_pe *pe;
636};
637
638
639/*
640 * The real server destination forwarding entry
641 * with ip address, port number, and so on.
642 */
643struct ip_vs_dest {
644 struct list_head n_list; /* for the dests in the service */
645 struct list_head d_list; /* for table with all the dests */
646
647 u16 af; /* address family */
648 __be16 port; /* port number of the server */
649 union nf_inet_addr addr; /* IP address of the server */
650 volatile unsigned int flags; /* dest status flags */
651 atomic_t conn_flags; /* flags to copy to conn */
652 atomic_t weight; /* server weight */
653
654 atomic_t refcnt; /* reference counter */
655 struct ip_vs_stats stats; /* statistics */
656
657 /* connection counters and thresholds */
658 atomic_t activeconns; /* active connections */
659 atomic_t inactconns; /* inactive connections */
660 atomic_t persistconns; /* persistent connections */
661 __u32 u_threshold; /* upper threshold */
662 __u32 l_threshold; /* lower threshold */
663
664 /* for destination cache */
665 spinlock_t dst_lock; /* lock of dst_cache */
666 struct dst_entry *dst_cache; /* destination cache entry */
667 u32 dst_rtos; /* RT_TOS(tos) for dst */
668 u32 dst_cookie;
669 union nf_inet_addr dst_saddr;
670
671 /* for virtual service */
672 struct ip_vs_service *svc; /* service it belongs to */
673 __u16 protocol; /* which protocol (TCP/UDP) */
674 __be16 vport; /* virtual port number */
675 union nf_inet_addr vaddr; /* virtual IP address */
676 __u32 vfwmark; /* firewall mark of service */
677};
678
679
680/*
681 * The scheduler object
682 */
683struct ip_vs_scheduler {
684 struct list_head n_list; /* d-linked list head */
685 char *name; /* scheduler name */
686 atomic_t refcnt; /* reference counter */
687 struct module *module; /* THIS_MODULE/NULL */
688
689 /* scheduler initializing service */
690 int (*init_service)(struct ip_vs_service *svc);
691 /* scheduling service finish */
692 int (*done_service)(struct ip_vs_service *svc);
693 /* scheduler updating service */
694 int (*update_service)(struct ip_vs_service *svc);
695
696 /* selecting a server from the given service */
697 struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc,
698 const struct sk_buff *skb);
699};
700
701/* The persistence engine object */
702struct ip_vs_pe {
703 struct list_head n_list; /* d-linked list head */
704 char *name; /* scheduler name */
705 atomic_t refcnt; /* reference counter */
706 struct module *module; /* THIS_MODULE/NULL */
707
708 /* get the connection template, if any */
709 int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb);
710 bool (*ct_match)(const struct ip_vs_conn_param *p,
711 struct ip_vs_conn *ct);
712 u32 (*hashkey_raw)(const struct ip_vs_conn_param *p, u32 initval,
713 bool inverse);
714 int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf);
715};
716
717/*
718 * The application module object (a.k.a. app incarnation)
719 */
720struct ip_vs_app {
721 struct list_head a_list; /* member in app list */
722 int type; /* IP_VS_APP_TYPE_xxx */
723 char *name; /* application module name */
724 __u16 protocol;
725 struct module *module; /* THIS_MODULE/NULL */
726 struct list_head incs_list; /* list of incarnations */
727
728 /* members for application incarnations */
729 struct list_head p_list; /* member in proto app list */
730 struct ip_vs_app *app; /* its real application */
731 __be16 port; /* port number in net order */
732 atomic_t usecnt; /* usage counter */
733
734 /*
735 * output hook: Process packet in inout direction, diff set for TCP.
736 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
737 * 2=Mangled but checksum was not updated
738 */
739 int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *,
740 struct sk_buff *, int *diff);
741
742 /*
743 * input hook: Process packet in outin direction, diff set for TCP.
744 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
745 * 2=Mangled but checksum was not updated
746 */
747 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *,
748 struct sk_buff *, int *diff);
749
750 /* ip_vs_app initializer */
751 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *);
752
753 /* ip_vs_app finish */
754 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *);
755
756
757 /* not used now */
758 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *,
759 struct ip_vs_protocol *);
760
761 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *);
762
763 int * timeout_table;
764 int * timeouts;
765 int timeouts_size;
766
767 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app,
768 int *verdict, struct ip_vs_conn **cpp);
769
770 struct ip_vs_conn *
771 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app,
772 const struct iphdr *iph, unsigned int proto_off,
773 int inverse);
774
775 struct ip_vs_conn *
776 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app,
777 const struct iphdr *iph, unsigned int proto_off,
778 int inverse);
779
780 int (*state_transition)(struct ip_vs_conn *cp, int direction,
781 const struct sk_buff *skb,
782 struct ip_vs_app *app);
783
784 void (*timeout_change)(struct ip_vs_app *app, int flags);
785};
786
787struct ipvs_master_sync_state {
788 struct list_head sync_queue;
789 struct ip_vs_sync_buff *sync_buff;
790 int sync_queue_len;
791 unsigned int sync_queue_delay;
792 struct task_struct *master_thread;
793 struct delayed_work master_wakeup_work;
794 struct netns_ipvs *ipvs;
795};
796
797/* IPVS in network namespace */
798struct netns_ipvs {
799 int gen; /* Generation */
800 int enable; /* enable like nf_hooks do */
801 /*
802 * Hash table: for real service lookups
803 */
804 #define IP_VS_RTAB_BITS 4
805 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
806 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
807
808 struct list_head rs_table[IP_VS_RTAB_SIZE];
809 /* ip_vs_app */
810 struct list_head app_list;
811 /* ip_vs_ftp */
812 struct ip_vs_app *ftp_app;
813 /* ip_vs_proto */
814 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */
815 struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE];
816 /* ip_vs_proto_tcp */
817#ifdef CONFIG_IP_VS_PROTO_TCP
818 #define TCP_APP_TAB_BITS 4
819 #define TCP_APP_TAB_SIZE (1 << TCP_APP_TAB_BITS)
820 #define TCP_APP_TAB_MASK (TCP_APP_TAB_SIZE - 1)
821 struct list_head tcp_apps[TCP_APP_TAB_SIZE];
822 spinlock_t tcp_app_lock;
823#endif
824 /* ip_vs_proto_udp */
825#ifdef CONFIG_IP_VS_PROTO_UDP
826 #define UDP_APP_TAB_BITS 4
827 #define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS)
828 #define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1)
829 struct list_head udp_apps[UDP_APP_TAB_SIZE];
830 spinlock_t udp_app_lock;
831#endif
832 /* ip_vs_proto_sctp */
833#ifdef CONFIG_IP_VS_PROTO_SCTP
834 #define SCTP_APP_TAB_BITS 4
835 #define SCTP_APP_TAB_SIZE (1 << SCTP_APP_TAB_BITS)
836 #define SCTP_APP_TAB_MASK (SCTP_APP_TAB_SIZE - 1)
837 /* Hash table for SCTP application incarnations */
838 struct list_head sctp_apps[SCTP_APP_TAB_SIZE];
839 spinlock_t sctp_app_lock;
840#endif
841 /* ip_vs_conn */
842 atomic_t conn_count; /* connection counter */
843
844 /* ip_vs_ctl */
845 struct ip_vs_stats tot_stats; /* Statistics & est. */
846
847 int num_services; /* no of virtual services */
848
849 rwlock_t rs_lock; /* real services table */
850 /* Trash for destinations */
851 struct list_head dest_trash;
852 /* Service counters */
853 atomic_t ftpsvc_counter;
854 atomic_t nullsvc_counter;
855
856#ifdef CONFIG_SYSCTL
857 /* 1/rate drop and drop-entry variables */
858 struct delayed_work defense_work; /* Work handler */
859 int drop_rate;
860 int drop_counter;
861 atomic_t dropentry;
862 /* locks in ctl.c */
863 spinlock_t dropentry_lock; /* drop entry handling */
864 spinlock_t droppacket_lock; /* drop packet handling */
865 spinlock_t securetcp_lock; /* state and timeout tables */
866
867 /* sys-ctl struct */
868 struct ctl_table_header *sysctl_hdr;
869 struct ctl_table *sysctl_tbl;
870#endif
871
872 /* sysctl variables */
873 int sysctl_amemthresh;
874 int sysctl_am_droprate;
875 int sysctl_drop_entry;
876 int sysctl_drop_packet;
877 int sysctl_secure_tcp;
878#ifdef CONFIG_IP_VS_NFCT
879 int sysctl_conntrack;
880#endif
881 int sysctl_snat_reroute;
882 int sysctl_sync_ver;
883 int sysctl_sync_ports;
884 int sysctl_sync_qlen_max;
885 int sysctl_sync_sock_size;
886 int sysctl_cache_bypass;
887 int sysctl_expire_nodest_conn;
888 int sysctl_expire_quiescent_template;
889 int sysctl_sync_threshold[2];
890 unsigned int sysctl_sync_refresh_period;
891 int sysctl_sync_retries;
892 int sysctl_nat_icmp_send;
893
894 /* ip_vs_lblc */
895 int sysctl_lblc_expiration;
896 struct ctl_table_header *lblc_ctl_header;
897 struct ctl_table *lblc_ctl_table;
898 /* ip_vs_lblcr */
899 int sysctl_lblcr_expiration;
900 struct ctl_table_header *lblcr_ctl_header;
901 struct ctl_table *lblcr_ctl_table;
902 /* ip_vs_est */
903 struct list_head est_list; /* estimator list */
904 spinlock_t est_lock;
905 struct timer_list est_timer; /* Estimation timer */
906 /* ip_vs_sync */
907 spinlock_t sync_lock;
908 struct ipvs_master_sync_state *ms;
909 spinlock_t sync_buff_lock;
910 struct task_struct **backup_threads;
911 int threads_mask;
912 int send_mesg_maxlen;
913 int recv_mesg_maxlen;
914 volatile int sync_state;
915 volatile int master_syncid;
916 volatile int backup_syncid;
917 struct mutex sync_mutex;
918 /* multicast interface name */
919 char master_mcast_ifn[IP_VS_IFNAME_MAXLEN];
920 char backup_mcast_ifn[IP_VS_IFNAME_MAXLEN];
921 /* net name space ptr */
922 struct net *net; /* Needed by timer routines */
923};
924
925#define DEFAULT_SYNC_THRESHOLD 3
926#define DEFAULT_SYNC_PERIOD 50
927#define DEFAULT_SYNC_VER 1
928#define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ)
929#define DEFAULT_SYNC_RETRIES 0
930#define IPVS_SYNC_WAKEUP_RATE 8
931#define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4)
932#define IPVS_SYNC_SEND_DELAY (HZ / 50)
933#define IPVS_SYNC_CHECK_PERIOD HZ
934#define IPVS_SYNC_FLUSH_TIME (HZ * 2)
935#define IPVS_SYNC_PORTS_MAX (1 << 6)
936
937#ifdef CONFIG_SYSCTL
938
939static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
940{
941 return ipvs->sysctl_sync_threshold[0];
942}
943
944static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
945{
946 return ACCESS_ONCE(ipvs->sysctl_sync_threshold[1]);
947}
948
949static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
950{
951 return ACCESS_ONCE(ipvs->sysctl_sync_refresh_period);
952}
953
954static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
955{
956 return ipvs->sysctl_sync_retries;
957}
958
959static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
960{
961 return ipvs->sysctl_sync_ver;
962}
963
964static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
965{
966 return ACCESS_ONCE(ipvs->sysctl_sync_ports);
967}
968
969static inline int sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
970{
971 return ipvs->sysctl_sync_qlen_max;
972}
973
974static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
975{
976 return ipvs->sysctl_sync_sock_size;
977}
978
979#else
980
981static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
982{
983 return DEFAULT_SYNC_THRESHOLD;
984}
985
986static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
987{
988 return DEFAULT_SYNC_PERIOD;
989}
990
991static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
992{
993 return DEFAULT_SYNC_REFRESH_PERIOD;
994}
995
996static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
997{
998 return DEFAULT_SYNC_RETRIES & 3;
999}
1000
1001static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1002{
1003 return DEFAULT_SYNC_VER;
1004}
1005
1006static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1007{
1008 return 1;
1009}
1010
1011static inline int sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1012{
1013 return IPVS_SYNC_QLEN_MAX;
1014}
1015
1016static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1017{
1018 return 0;
1019}
1020
1021#endif
1022
1023/*
1024 * IPVS core functions
1025 * (from ip_vs_core.c)
1026 */
1027extern const char *ip_vs_proto_name(unsigned int proto);
1028extern void ip_vs_init_hash_table(struct list_head *table, int rows);
1029#define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t)))
1030
1031#define IP_VS_APP_TYPE_FTP 1
1032
1033/*
1034 * ip_vs_conn handling functions
1035 * (from ip_vs_conn.c)
1036 */
1037
1038enum {
1039 IP_VS_DIR_INPUT = 0,
1040 IP_VS_DIR_OUTPUT,
1041 IP_VS_DIR_INPUT_ONLY,
1042 IP_VS_DIR_LAST,
1043};
1044
1045static inline void ip_vs_conn_fill_param(struct net *net, int af, int protocol,
1046 const union nf_inet_addr *caddr,
1047 __be16 cport,
1048 const union nf_inet_addr *vaddr,
1049 __be16 vport,
1050 struct ip_vs_conn_param *p)
1051{
1052 p->net = net;
1053 p->af = af;
1054 p->protocol = protocol;
1055 p->caddr = caddr;
1056 p->cport = cport;
1057 p->vaddr = vaddr;
1058 p->vport = vport;
1059 p->pe = NULL;
1060 p->pe_data = NULL;
1061}
1062
1063struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p);
1064struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p);
1065
1066struct ip_vs_conn * ip_vs_conn_in_get_proto(int af, const struct sk_buff *skb,
1067 const struct ip_vs_iphdr *iph,
1068 unsigned int proto_off,
1069 int inverse);
1070
1071struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p);
1072
1073struct ip_vs_conn * ip_vs_conn_out_get_proto(int af, const struct sk_buff *skb,
1074 const struct ip_vs_iphdr *iph,
1075 unsigned int proto_off,
1076 int inverse);
1077
1078/* put back the conn without restarting its timer */
1079static inline void __ip_vs_conn_put(struct ip_vs_conn *cp)
1080{
1081 atomic_dec(&cp->refcnt);
1082}
1083extern void ip_vs_conn_put(struct ip_vs_conn *cp);
1084extern void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport);
1085
1086struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p,
1087 const union nf_inet_addr *daddr,
1088 __be16 dport, unsigned int flags,
1089 struct ip_vs_dest *dest, __u32 fwmark);
1090extern void ip_vs_conn_expire_now(struct ip_vs_conn *cp);
1091
1092extern const char * ip_vs_state_name(__u16 proto, int state);
1093
1094extern void ip_vs_tcp_conn_listen(struct net *net, struct ip_vs_conn *cp);
1095extern int ip_vs_check_template(struct ip_vs_conn *ct);
1096extern void ip_vs_random_dropentry(struct net *net);
1097extern int ip_vs_conn_init(void);
1098extern void ip_vs_conn_cleanup(void);
1099
1100static inline void ip_vs_control_del(struct ip_vs_conn *cp)
1101{
1102 struct ip_vs_conn *ctl_cp = cp->control;
1103 if (!ctl_cp) {
1104 IP_VS_ERR_BUF("request control DEL for uncontrolled: "
1105 "%s:%d to %s:%d\n",
1106 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1107 ntohs(cp->cport),
1108 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1109 ntohs(cp->vport));
1110
1111 return;
1112 }
1113
1114 IP_VS_DBG_BUF(7, "DELeting control for: "
1115 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1116 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1117 ntohs(cp->cport),
1118 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1119 ntohs(ctl_cp->cport));
1120
1121 cp->control = NULL;
1122 if (atomic_read(&ctl_cp->n_control) == 0) {
1123 IP_VS_ERR_BUF("BUG control DEL with n=0 : "
1124 "%s:%d to %s:%d\n",
1125 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1126 ntohs(cp->cport),
1127 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1128 ntohs(cp->vport));
1129
1130 return;
1131 }
1132 atomic_dec(&ctl_cp->n_control);
1133}
1134
1135static inline void
1136ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp)
1137{
1138 if (cp->control) {
1139 IP_VS_ERR_BUF("request control ADD for already controlled: "
1140 "%s:%d to %s:%d\n",
1141 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1142 ntohs(cp->cport),
1143 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1144 ntohs(cp->vport));
1145
1146 ip_vs_control_del(cp);
1147 }
1148
1149 IP_VS_DBG_BUF(7, "ADDing control for: "
1150 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1151 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1152 ntohs(cp->cport),
1153 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1154 ntohs(ctl_cp->cport));
1155
1156 cp->control = ctl_cp;
1157 atomic_inc(&ctl_cp->n_control);
1158}
1159
1160/*
1161 * IPVS netns init & cleanup functions
1162 */
1163extern int ip_vs_estimator_net_init(struct net *net);
1164extern int ip_vs_control_net_init(struct net *net);
1165extern int ip_vs_protocol_net_init(struct net *net);
1166extern int ip_vs_app_net_init(struct net *net);
1167extern int ip_vs_conn_net_init(struct net *net);
1168extern int ip_vs_sync_net_init(struct net *net);
1169extern void ip_vs_conn_net_cleanup(struct net *net);
1170extern void ip_vs_app_net_cleanup(struct net *net);
1171extern void ip_vs_protocol_net_cleanup(struct net *net);
1172extern void ip_vs_control_net_cleanup(struct net *net);
1173extern void ip_vs_estimator_net_cleanup(struct net *net);
1174extern void ip_vs_sync_net_cleanup(struct net *net);
1175extern void ip_vs_service_net_cleanup(struct net *net);
1176
1177/*
1178 * IPVS application functions
1179 * (from ip_vs_app.c)
1180 */
1181#define IP_VS_APP_MAX_PORTS 8
1182extern int register_ip_vs_app(struct net *net, struct ip_vs_app *app);
1183extern void unregister_ip_vs_app(struct net *net, struct ip_vs_app *app);
1184extern int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1185extern void ip_vs_unbind_app(struct ip_vs_conn *cp);
1186extern int register_ip_vs_app_inc(struct net *net, struct ip_vs_app *app,
1187 __u16 proto, __u16 port);
1188extern int ip_vs_app_inc_get(struct ip_vs_app *inc);
1189extern void ip_vs_app_inc_put(struct ip_vs_app *inc);
1190
1191extern int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb);
1192extern int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb);
1193
1194void ip_vs_bind_pe(struct ip_vs_service *svc, struct ip_vs_pe *pe);
1195void ip_vs_unbind_pe(struct ip_vs_service *svc);
1196int register_ip_vs_pe(struct ip_vs_pe *pe);
1197int unregister_ip_vs_pe(struct ip_vs_pe *pe);
1198struct ip_vs_pe *ip_vs_pe_getbyname(const char *name);
1199struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name);
1200
1201/*
1202 * Use a #define to avoid all of module.h just for these trivial ops
1203 */
1204#define ip_vs_pe_get(pe) \
1205 if (pe && pe->module) \
1206 __module_get(pe->module);
1207
1208#define ip_vs_pe_put(pe) \
1209 if (pe && pe->module) \
1210 module_put(pe->module);
1211
1212/*
1213 * IPVS protocol functions (from ip_vs_proto.c)
1214 */
1215extern int ip_vs_protocol_init(void);
1216extern void ip_vs_protocol_cleanup(void);
1217extern void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags);
1218extern int *ip_vs_create_timeout_table(int *table, int size);
1219extern int
1220ip_vs_set_state_timeout(int *table, int num, const char *const *names,
1221 const char *name, int to);
1222extern void
1223ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
1224 const struct sk_buff *skb,
1225 int offset, const char *msg);
1226
1227extern struct ip_vs_protocol ip_vs_protocol_tcp;
1228extern struct ip_vs_protocol ip_vs_protocol_udp;
1229extern struct ip_vs_protocol ip_vs_protocol_icmp;
1230extern struct ip_vs_protocol ip_vs_protocol_esp;
1231extern struct ip_vs_protocol ip_vs_protocol_ah;
1232extern struct ip_vs_protocol ip_vs_protocol_sctp;
1233
1234/*
1235 * Registering/unregistering scheduler functions
1236 * (from ip_vs_sched.c)
1237 */
1238extern int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1239extern int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1240extern int ip_vs_bind_scheduler(struct ip_vs_service *svc,
1241 struct ip_vs_scheduler *scheduler);
1242extern int ip_vs_unbind_scheduler(struct ip_vs_service *svc);
1243extern struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name);
1244extern void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler);
1245extern struct ip_vs_conn *
1246ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb,
1247 struct ip_vs_proto_data *pd, int *ignored);
1248extern int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb,
1249 struct ip_vs_proto_data *pd);
1250
1251extern void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg);
1252
1253
1254/*
1255 * IPVS control data and functions (from ip_vs_ctl.c)
1256 */
1257extern struct ip_vs_stats ip_vs_stats;
1258extern int sysctl_ip_vs_sync_ver;
1259
1260extern struct ip_vs_service *
1261ip_vs_service_get(struct net *net, int af, __u32 fwmark, __u16 protocol,
1262 const union nf_inet_addr *vaddr, __be16 vport);
1263
1264static inline void ip_vs_service_put(struct ip_vs_service *svc)
1265{
1266 atomic_dec(&svc->usecnt);
1267}
1268
1269extern struct ip_vs_dest *
1270ip_vs_lookup_real_service(struct net *net, int af, __u16 protocol,
1271 const union nf_inet_addr *daddr, __be16 dport);
1272
1273extern int ip_vs_use_count_inc(void);
1274extern void ip_vs_use_count_dec(void);
1275extern int ip_vs_register_nl_ioctl(void);
1276extern void ip_vs_unregister_nl_ioctl(void);
1277extern int ip_vs_control_init(void);
1278extern void ip_vs_control_cleanup(void);
1279extern struct ip_vs_dest *
1280ip_vs_find_dest(struct net *net, int af, const union nf_inet_addr *daddr,
1281 __be16 dport, const union nf_inet_addr *vaddr, __be16 vport,
1282 __u16 protocol, __u32 fwmark, __u32 flags);
1283extern struct ip_vs_dest *ip_vs_try_bind_dest(struct ip_vs_conn *cp);
1284
1285
1286/*
1287 * IPVS sync daemon data and function prototypes
1288 * (from ip_vs_sync.c)
1289 */
1290extern int start_sync_thread(struct net *net, int state, char *mcast_ifn,
1291 __u8 syncid);
1292extern int stop_sync_thread(struct net *net, int state);
1293extern void ip_vs_sync_conn(struct net *net, struct ip_vs_conn *cp, int pkts);
1294
1295
1296/*
1297 * IPVS rate estimator prototypes (from ip_vs_est.c)
1298 */
1299extern void ip_vs_start_estimator(struct net *net, struct ip_vs_stats *stats);
1300extern void ip_vs_stop_estimator(struct net *net, struct ip_vs_stats *stats);
1301extern void ip_vs_zero_estimator(struct ip_vs_stats *stats);
1302extern void ip_vs_read_estimator(struct ip_vs_stats_user *dst,
1303 struct ip_vs_stats *stats);
1304
1305/*
1306 * Various IPVS packet transmitters (from ip_vs_xmit.c)
1307 */
1308extern int ip_vs_null_xmit
1309(struct sk_buff *skb, struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1310extern int ip_vs_bypass_xmit
1311(struct sk_buff *skb, struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1312extern int ip_vs_nat_xmit
1313(struct sk_buff *skb, struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1314extern int ip_vs_tunnel_xmit
1315(struct sk_buff *skb, struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1316extern int ip_vs_dr_xmit
1317(struct sk_buff *skb, struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1318extern int ip_vs_icmp_xmit
1319(struct sk_buff *skb, struct ip_vs_conn *cp, struct ip_vs_protocol *pp,
1320 int offset, unsigned int hooknum);
1321extern void ip_vs_dst_reset(struct ip_vs_dest *dest);
1322
1323#ifdef CONFIG_IP_VS_IPV6
1324extern int ip_vs_bypass_xmit_v6
1325(struct sk_buff *skb, struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1326extern int ip_vs_nat_xmit_v6
1327(struct sk_buff *skb, struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1328extern int ip_vs_tunnel_xmit_v6
1329(struct sk_buff *skb, struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1330extern int ip_vs_dr_xmit_v6
1331(struct sk_buff *skb, struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1332extern int ip_vs_icmp_xmit_v6
1333(struct sk_buff *skb, struct ip_vs_conn *cp, struct ip_vs_protocol *pp,
1334 int offset, unsigned int hooknum);
1335#endif
1336
1337#ifdef CONFIG_SYSCTL
1338/*
1339 * This is a simple mechanism to ignore packets when
1340 * we are loaded. Just set ip_vs_drop_rate to 'n' and
1341 * we start to drop 1/rate of the packets
1342 */
1343
1344static inline int ip_vs_todrop(struct netns_ipvs *ipvs)
1345{
1346 if (!ipvs->drop_rate)
1347 return 0;
1348 if (--ipvs->drop_counter > 0)
1349 return 0;
1350 ipvs->drop_counter = ipvs->drop_rate;
1351 return 1;
1352}
1353#else
1354static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; }
1355#endif
1356
1357/*
1358 * ip_vs_fwd_tag returns the forwarding tag of the connection
1359 */
1360#define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK)
1361
1362static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp)
1363{
1364 char fwd;
1365
1366 switch (IP_VS_FWD_METHOD(cp)) {
1367 case IP_VS_CONN_F_MASQ:
1368 fwd = 'M'; break;
1369 case IP_VS_CONN_F_LOCALNODE:
1370 fwd = 'L'; break;
1371 case IP_VS_CONN_F_TUNNEL:
1372 fwd = 'T'; break;
1373 case IP_VS_CONN_F_DROUTE:
1374 fwd = 'R'; break;
1375 case IP_VS_CONN_F_BYPASS:
1376 fwd = 'B'; break;
1377 default:
1378 fwd = '?'; break;
1379 }
1380 return fwd;
1381}
1382
1383extern void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp,
1384 struct ip_vs_conn *cp, int dir);
1385
1386#ifdef CONFIG_IP_VS_IPV6
1387extern void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp,
1388 struct ip_vs_conn *cp, int dir);
1389#endif
1390
1391extern __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset);
1392
1393static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum)
1394{
1395 __be32 diff[2] = { ~old, new };
1396
1397 return csum_partial(diff, sizeof(diff), oldsum);
1398}
1399
1400#ifdef CONFIG_IP_VS_IPV6
1401static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new,
1402 __wsum oldsum)
1403{
1404 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0],
1405 new[3], new[2], new[1], new[0] };
1406
1407 return csum_partial(diff, sizeof(diff), oldsum);
1408}
1409#endif
1410
1411static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum)
1412{
1413 __be16 diff[2] = { ~old, new };
1414
1415 return csum_partial(diff, sizeof(diff), oldsum);
1416}
1417
1418/*
1419 * Forget current conntrack (unconfirmed) and attach notrack entry
1420 */
1421static inline void ip_vs_notrack(struct sk_buff *skb)
1422{
1423#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1424 enum ip_conntrack_info ctinfo;
1425 struct nf_conn *ct = nf_ct_get(skb, &ctinfo);
1426
1427 if (!ct || !nf_ct_is_untracked(ct)) {
1428 nf_conntrack_put(skb->nfct);
1429 skb->nfct = &nf_ct_untracked_get()->ct_general;
1430 skb->nfctinfo = IP_CT_NEW;
1431 nf_conntrack_get(skb->nfct);
1432 }
1433#endif
1434}
1435
1436#ifdef CONFIG_IP_VS_NFCT
1437/*
1438 * Netfilter connection tracking
1439 * (from ip_vs_nfct.c)
1440 */
1441static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1442{
1443#ifdef CONFIG_SYSCTL
1444 return ipvs->sysctl_conntrack;
1445#else
1446 return 0;
1447#endif
1448}
1449
1450extern void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp,
1451 int outin);
1452extern int ip_vs_confirm_conntrack(struct sk_buff *skb);
1453extern void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct,
1454 struct ip_vs_conn *cp, u_int8_t proto,
1455 const __be16 port, int from_rs);
1456extern void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp);
1457
1458#else
1459
1460static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
1461{
1462 return 0;
1463}
1464
1465static inline void ip_vs_update_conntrack(struct sk_buff *skb,
1466 struct ip_vs_conn *cp, int outin)
1467{
1468}
1469
1470static inline int ip_vs_confirm_conntrack(struct sk_buff *skb)
1471{
1472 return NF_ACCEPT;
1473}
1474
1475static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp)
1476{
1477}
1478/* CONFIG_IP_VS_NFCT */
1479#endif
1480
1481static inline unsigned int
1482ip_vs_dest_conn_overhead(struct ip_vs_dest *dest)
1483{
1484 /*
1485 * We think the overhead of processing active connections is 256
1486 * times higher than that of inactive connections in average. (This
1487 * 256 times might not be accurate, we will change it later) We
1488 * use the following formula to estimate the overhead now:
1489 * dest->activeconns*256 + dest->inactconns
1490 */
1491 return (atomic_read(&dest->activeconns) << 8) +
1492 atomic_read(&dest->inactconns);
1493}
1494
1495#endif /* _NET_IP_VS_H */