projects/linux·blob master · refs

2261 lines · 66.1 KiB

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/rculist_bl.h> /* for struct hlist_bl_head */
15#include <linux/spinlock.h> /* for struct rwlock_t */
16#include <linux/atomic.h> /* for struct atomic_t */
17#include <linux/refcount.h> /* for struct refcount_t */
18#include <linux/workqueue.h>
19
20#include <linux/compiler.h>
21#include <linux/timer.h>
22#include <linux/bug.h>
23
24#include <net/checksum.h>
25#include <linux/netfilter.h> /* for union nf_inet_addr */
26#include <linux/ip.h>
27#include <linux/ipv6.h> /* for struct ipv6hdr */
28#include <net/ipv6.h>
29#if IS_ENABLED(CONFIG_NF_CONNTRACK)
30#include <net/netfilter/nf_conntrack.h>
31#endif
32#include <net/net_namespace.h> /* Netw namespace */
33#include <linux/sched/isolation.h>
34#include <linux/siphash.h>
35
36#define IP_VS_HDR_INVERSE 1
37#define IP_VS_HDR_ICMP 2
38
39/* Destination Server Flags */
40#define IP_VS_DEST_F_OVERLOAD 0x0002 /* server is overloaded */
41
42/* Destination Server Config Flags */
43#define IP_VS_DEST_CF_AVAILABLE 0x0001 /* server is available */
44
45/* conn_tab limits (as per Kconfig) */
46#define IP_VS_CONN_TAB_MIN_BITS 8
47#if BITS_PER_LONG > 32
48#define IP_VS_CONN_TAB_MAX_BITS 27
49#else
50#define IP_VS_CONN_TAB_MAX_BITS 20
51#endif
52
53/* conn_max limits */
54#if BITS_PER_LONG > 32
55/* Limit of atomic_t but restricted by roundup_pow_of_two() in ip_vs_core.c */
56#define IP_VS_CONN_MAX (1 << 30)
57#else
58#define IP_VS_CONN_MAX (1 << 24)
59#endif
60
61/* svc_table limits */
62#define IP_VS_SVC_TAB_MIN_BITS 4
63#define IP_VS_SVC_TAB_MAX_BITS 20
64
65/* Generic access of ipvs struct */
66static inline struct netns_ipvs *net_ipvs(struct net* net)
67{
68 return net->ipvs;
69}
70
71/* Connections' size value needed by ip_vs_ctl.c */
72extern int ip_vs_conn_tab_size;
73
74struct ip_vs_iphdr {
75 int hdr_flags; /* ipvs flags */
76 __u32 off; /* Where IP or IPv4 header starts */
77 __u32 len; /* IPv4 simply where L4 starts
78 * IPv6 where L4 Transport Header starts */
79 __u16 fragoffs; /* IPv6 fragment offset, 0 if first frag (or not frag)*/
80 __s16 protocol;
81 __s32 flags;
82 union nf_inet_addr saddr;
83 union nf_inet_addr daddr;
84};
85
86static inline void *frag_safe_skb_hp(const struct sk_buff *skb, int offset,
87 int len, void *buffer)
88{
89 return skb_header_pointer(skb, offset, len, buffer);
90}
91
92/* This function handles filling *ip_vs_iphdr, both for IPv4 and IPv6.
93 * IPv6 requires some extra work, as finding proper header position,
94 * depend on the IPv6 extension headers.
95 */
96static inline int
97ip_vs_fill_iph_skb_off(int af, const struct sk_buff *skb, int offset,
98 int hdr_flags, struct ip_vs_iphdr *iphdr)
99{
100 iphdr->hdr_flags = hdr_flags;
101 iphdr->off = offset;
102
103#ifdef CONFIG_IP_VS_IPV6
104 if (af == AF_INET6) {
105 struct ipv6hdr _iph;
106 const struct ipv6hdr *iph = skb_header_pointer(
107 skb, offset, sizeof(_iph), &_iph);
108 if (!iph)
109 return 0;
110
111 iphdr->saddr.in6 = iph->saddr;
112 iphdr->daddr.in6 = iph->daddr;
113 /* ipv6_find_hdr() updates len, flags */
114 iphdr->len = offset;
115 iphdr->flags = 0;
116 iphdr->protocol = ipv6_find_hdr(skb, &iphdr->len, -1,
117 &iphdr->fragoffs,
118 &iphdr->flags);
119 if (iphdr->protocol < 0)
120 return 0;
121 } else
122#endif
123 {
124 struct iphdr _iph;
125 const struct iphdr *iph = skb_header_pointer(
126 skb, offset, sizeof(_iph), &_iph);
127 if (!iph)
128 return 0;
129
130 iphdr->len = offset + iph->ihl * 4;
131 iphdr->fragoffs = 0;
132 iphdr->protocol = iph->protocol;
133 iphdr->saddr.ip = iph->saddr;
134 iphdr->daddr.ip = iph->daddr;
135 }
136
137 return 1;
138}
139
140static inline int
141ip_vs_fill_iph_skb_icmp(int af, const struct sk_buff *skb, int offset,
142 bool inverse, struct ip_vs_iphdr *iphdr)
143{
144 int hdr_flags = IP_VS_HDR_ICMP;
145
146 if (inverse)
147 hdr_flags |= IP_VS_HDR_INVERSE;
148
149 return ip_vs_fill_iph_skb_off(af, skb, offset, hdr_flags, iphdr);
150}
151
152static inline int
153ip_vs_fill_iph_skb(int af, const struct sk_buff *skb, bool inverse,
154 struct ip_vs_iphdr *iphdr)
155{
156 int hdr_flags = 0;
157
158 if (inverse)
159 hdr_flags |= IP_VS_HDR_INVERSE;
160
161 return ip_vs_fill_iph_skb_off(af, skb, skb_network_offset(skb),
162 hdr_flags, iphdr);
163}
164
165static inline bool
166ip_vs_iph_inverse(const struct ip_vs_iphdr *iph)
167{
168 return !!(iph->hdr_flags & IP_VS_HDR_INVERSE);
169}
170
171static inline bool
172ip_vs_iph_icmp(const struct ip_vs_iphdr *iph)
173{
174 return !!(iph->hdr_flags & IP_VS_HDR_ICMP);
175}
176
177static inline void ip_vs_addr_copy(int af, union nf_inet_addr *dst,
178 const union nf_inet_addr *src)
179{
180#ifdef CONFIG_IP_VS_IPV6
181 if (af == AF_INET6)
182 dst->in6 = src->in6;
183 else
184#endif
185 dst->ip = src->ip;
186}
187
188static inline void ip_vs_addr_set(int af, union nf_inet_addr *dst,
189 const union nf_inet_addr *src)
190{
191#ifdef CONFIG_IP_VS_IPV6
192 if (af == AF_INET6) {
193 dst->in6 = src->in6;
194 return;
195 }
196#endif
197 dst->ip = src->ip;
198 dst->all[1] = 0;
199 dst->all[2] = 0;
200 dst->all[3] = 0;
201}
202
203static inline int ip_vs_addr_equal(int af, const union nf_inet_addr *a,
204 const union nf_inet_addr *b)
205{
206#ifdef CONFIG_IP_VS_IPV6
207 if (af == AF_INET6)
208 return ipv6_addr_equal(&a->in6, &b->in6);
209#endif
210 return a->ip == b->ip;
211}
212
213#ifdef CONFIG_IP_VS_DEBUG
214#include <linux/net.h>
215
216int ip_vs_get_debug_level(void);
217
218static inline const char *ip_vs_dbg_addr(int af, char *buf, size_t buf_len,
219 const union nf_inet_addr *addr,
220 int *idx)
221{
222 int len;
223#ifdef CONFIG_IP_VS_IPV6
224 if (af == AF_INET6)
225 len = snprintf(&buf[*idx], buf_len - *idx, "[%pI6c]",
226 &addr->in6) + 1;
227 else
228#endif
229 len = snprintf(&buf[*idx], buf_len - *idx, "%pI4",
230 &addr->ip) + 1;
231
232 *idx += len;
233 BUG_ON(*idx > buf_len + 1);
234 return &buf[*idx - len];
235}
236
237#define IP_VS_DBG_BUF(level, msg, ...) \
238 do { \
239 char ip_vs_dbg_buf[160]; \
240 int ip_vs_dbg_idx = 0; \
241 if (level <= ip_vs_get_debug_level()) \
242 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
243 } while (0)
244#define IP_VS_ERR_BUF(msg...) \
245 do { \
246 char ip_vs_dbg_buf[160]; \
247 int ip_vs_dbg_idx = 0; \
248 pr_err(msg); \
249 } while (0)
250
251/* Only use from within IP_VS_DBG_BUF() or IP_VS_ERR_BUF macros */
252#define IP_VS_DBG_ADDR(af, addr) \
253 ip_vs_dbg_addr(af, ip_vs_dbg_buf, \
254 sizeof(ip_vs_dbg_buf), addr, \
255 &ip_vs_dbg_idx)
256
257#define IP_VS_DBG(level, msg, ...) \
258 do { \
259 if (level <= ip_vs_get_debug_level()) \
260 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
261 } while (0)
262#define IP_VS_DBG_RL(msg, ...) \
263 do { \
264 if (net_ratelimit()) \
265 printk(KERN_DEBUG pr_fmt(msg), ##__VA_ARGS__); \
266 } while (0)
267#define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) \
268 do { \
269 if (level <= ip_vs_get_debug_level()) \
270 pp->debug_packet(af, pp, skb, ofs, msg); \
271 } while (0)
272#define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) \
273 do { \
274 if (level <= ip_vs_get_debug_level() && \
275 net_ratelimit()) \
276 pp->debug_packet(af, pp, skb, ofs, msg); \
277 } while (0)
278#else /* NO DEBUGGING at ALL */
279#define IP_VS_DBG_BUF(level, msg...) do {} while (0)
280#define IP_VS_ERR_BUF(msg...) do {} while (0)
281#define IP_VS_DBG(level, msg...) do {} while (0)
282#define IP_VS_DBG_RL(msg...) do {} while (0)
283#define IP_VS_DBG_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
284#define IP_VS_DBG_RL_PKT(level, af, pp, skb, ofs, msg) do {} while (0)
285#endif
286
287#define IP_VS_BUG() BUG()
288#define IP_VS_ERR_RL(msg, ...) \
289 do { \
290 if (net_ratelimit()) \
291 pr_err(msg, ##__VA_ARGS__); \
292 } while (0)
293
294struct ip_vs_aligned_lock {
295 spinlock_t l; /* Protect buckets */
296} ____cacheline_aligned_in_smp;
297
298/* For arrays per family */
299enum {
300 IP_VS_AF_INET,
301 IP_VS_AF_INET6,
302 IP_VS_AF_MAX
303};
304
305static inline int ip_vs_af_index(int af)
306{
307 return af == AF_INET6 ? IP_VS_AF_INET6 : IP_VS_AF_INET;
308}
309
310/* work_flags */
311enum {
312 IP_VS_WORK_SVC_RESIZE, /* Schedule svc_resize_work */
313 IP_VS_WORK_SVC_NORESIZE, /* Stopping svc_resize_work */
314 IP_VS_WORK_CONN_RESIZE, /* Schedule conn_resize_work */
315};
316
317/* The port number of FTP service (in network order). */
318#define FTPPORT cpu_to_be16(21)
319#define FTPDATA cpu_to_be16(20)
320
321/* TCP State Values */
322enum {
323 IP_VS_TCP_S_NONE = 0,
324 IP_VS_TCP_S_ESTABLISHED,
325 IP_VS_TCP_S_SYN_SENT,
326 IP_VS_TCP_S_SYN_RECV,
327 IP_VS_TCP_S_FIN_WAIT,
328 IP_VS_TCP_S_TIME_WAIT,
329 IP_VS_TCP_S_CLOSE,
330 IP_VS_TCP_S_CLOSE_WAIT,
331 IP_VS_TCP_S_LAST_ACK,
332 IP_VS_TCP_S_LISTEN,
333 IP_VS_TCP_S_SYNACK,
334 IP_VS_TCP_S_LAST
335};
336
337/* UDP State Values */
338enum {
339 IP_VS_UDP_S_NORMAL,
340 IP_VS_UDP_S_LAST,
341};
342
343/* ICMP State Values */
344enum {
345 IP_VS_ICMP_S_NORMAL,
346 IP_VS_ICMP_S_LAST,
347};
348
349/* SCTP State Values */
350enum ip_vs_sctp_states {
351 IP_VS_SCTP_S_NONE,
352 IP_VS_SCTP_S_INIT1,
353 IP_VS_SCTP_S_INIT,
354 IP_VS_SCTP_S_COOKIE_SENT,
355 IP_VS_SCTP_S_COOKIE_REPLIED,
356 IP_VS_SCTP_S_COOKIE_WAIT,
357 IP_VS_SCTP_S_COOKIE,
358 IP_VS_SCTP_S_COOKIE_ECHOED,
359 IP_VS_SCTP_S_ESTABLISHED,
360 IP_VS_SCTP_S_SHUTDOWN_SENT,
361 IP_VS_SCTP_S_SHUTDOWN_RECEIVED,
362 IP_VS_SCTP_S_SHUTDOWN_ACK_SENT,
363 IP_VS_SCTP_S_REJECTED,
364 IP_VS_SCTP_S_CLOSED,
365 IP_VS_SCTP_S_LAST
366};
367
368/* Connection templates use bits from state */
369#define IP_VS_CTPL_S_NONE 0x0000
370#define IP_VS_CTPL_S_ASSURED 0x0001
371#define IP_VS_CTPL_S_LAST 0x0002
372
373/* Delta sequence info structure
374 * Each ip_vs_conn has 2 (output AND input seq. changes).
375 * Only used in the VS/NAT.
376 */
377struct ip_vs_seq {
378 __u32 init_seq; /* Add delta from this seq */
379 __u32 delta; /* Delta in sequence numbers */
380 __u32 previous_delta; /* Delta in sequence numbers
381 * before last resized pkt */
382};
383
384/* counters per cpu */
385struct ip_vs_counters {
386 u64_stats_t conns; /* connections scheduled */
387 u64_stats_t inpkts; /* incoming packets */
388 u64_stats_t outpkts; /* outgoing packets */
389 u64_stats_t inbytes; /* incoming bytes */
390 u64_stats_t outbytes; /* outgoing bytes */
391};
392/* Stats per cpu */
393struct ip_vs_cpu_stats {
394 struct ip_vs_counters cnt;
395 struct u64_stats_sync syncp;
396};
397
398/* Default nice for estimator kthreads */
399#define IPVS_EST_NICE 0
400
401/* IPVS statistics objects */
402struct ip_vs_estimator {
403 struct hlist_node list;
404
405 u64 last_inbytes;
406 u64 last_outbytes;
407 u64 last_conns;
408 u64 last_inpkts;
409 u64 last_outpkts;
410
411 u64 cps;
412 u64 inpps;
413 u64 outpps;
414 u64 inbps;
415 u64 outbps;
416
417 s32 ktid:16, /* kthread ID, -1=temp list */
418 ktrow:8, /* row/tick ID for kthread */
419 ktcid:8; /* chain ID for kthread tick */
420};
421
422/*
423 * IPVS statistics object, 64-bit kernel version of struct ip_vs_stats_user
424 */
425struct ip_vs_kstats {
426 u64 conns; /* connections scheduled */
427 u64 inpkts; /* incoming packets */
428 u64 outpkts; /* outgoing packets */
429 u64 inbytes; /* incoming bytes */
430 u64 outbytes; /* outgoing bytes */
431
432 u64 cps; /* current connection rate */
433 u64 inpps; /* current in packet rate */
434 u64 outpps; /* current out packet rate */
435 u64 inbps; /* current in byte rate */
436 u64 outbps; /* current out byte rate */
437};
438
439struct ip_vs_stats {
440 struct ip_vs_kstats kstats; /* kernel statistics */
441 struct ip_vs_estimator est; /* estimator */
442 struct ip_vs_cpu_stats __percpu *cpustats; /* per cpu counters */
443 spinlock_t lock; /* spin lock */
444 struct ip_vs_kstats kstats0; /* reset values */
445};
446
447struct ip_vs_stats_rcu {
448 struct ip_vs_stats s;
449 struct rcu_head rcu_head;
450};
451
452int ip_vs_stats_init_alloc(struct ip_vs_stats *s);
453struct ip_vs_stats *ip_vs_stats_alloc(void);
454void ip_vs_stats_release(struct ip_vs_stats *stats);
455void ip_vs_stats_free(struct ip_vs_stats *stats);
456
457/* Process estimators in multiple timer ticks (20/50/100, see ktrow) */
458#define IPVS_EST_NTICKS 50
459/* Estimation uses a 2-second period containing ticks (in jiffies) */
460#define IPVS_EST_TICK ((2 * HZ) / IPVS_EST_NTICKS)
461
462/* Limit of CPU load per kthread (8 for 12.5%), ratio of CPU capacity (1/C).
463 * Value of 4 and above ensures kthreads will take work without exceeding
464 * the CPU capacity under different circumstances.
465 */
466#define IPVS_EST_LOAD_DIVISOR 8
467
468/* Kthreads should not have work that exceeds the CPU load above 50% */
469#define IPVS_EST_CPU_KTHREADS (IPVS_EST_LOAD_DIVISOR / 2)
470
471/* Desired number of chains per timer tick (chain load factor in 100us units),
472 * 48=4.8ms of 40ms tick (12% CPU usage):
473 * 2 sec * 1000 ms in sec * 10 (100us in ms) / 8 (12.5%) / 50
474 */
475#define IPVS_EST_CHAIN_FACTOR \
476 ALIGN_DOWN(2 * 1000 * 10 / IPVS_EST_LOAD_DIVISOR / IPVS_EST_NTICKS, 8)
477
478/* Compiled number of chains per tick
479 * The defines should match cond_resched_rcu
480 */
481#if defined(CONFIG_DEBUG_ATOMIC_SLEEP) || !defined(CONFIG_PREEMPT_RCU)
482#define IPVS_EST_TICK_CHAINS IPVS_EST_CHAIN_FACTOR
483#else
484#define IPVS_EST_TICK_CHAINS 1
485#endif
486
487#if IPVS_EST_NTICKS > 127
488#error Too many timer ticks for ktrow
489#endif
490
491/* Multiple chains processed in same tick */
492struct ip_vs_est_tick_data {
493 struct rcu_head rcu_head;
494 struct hlist_head chains[IPVS_EST_TICK_CHAINS];
495 DECLARE_BITMAP(present, IPVS_EST_TICK_CHAINS);
496 DECLARE_BITMAP(full, IPVS_EST_TICK_CHAINS);
497 int chain_len[IPVS_EST_TICK_CHAINS];
498};
499
500/* Context for estimation kthread */
501struct ip_vs_est_kt_data {
502 struct netns_ipvs *ipvs;
503 struct task_struct *task; /* task if running */
504 struct ip_vs_est_tick_data __rcu *ticks[IPVS_EST_NTICKS];
505 DECLARE_BITMAP(avail, IPVS_EST_NTICKS); /* tick has space for ests */
506 unsigned long est_timer; /* estimation timer (jiffies) */
507 struct ip_vs_stats *calc_stats; /* Used for calculation */
508 int needed; /* task is needed */
509 int tick_len[IPVS_EST_NTICKS]; /* est count */
510 int id; /* ktid per netns */
511 int chain_max; /* max ests per tick chain */
512 int tick_max; /* max ests per tick */
513 int est_count; /* attached ests to kthread */
514 int est_max_count; /* max ests per kthread */
515 int add_row; /* row for new ests */
516 int est_row; /* estimated row */
517};
518
519/* IPVS resizable hash tables */
520struct ip_vs_rht {
521 struct hlist_bl_head *buckets;
522 struct ip_vs_rht __rcu *new_tbl; /* New/Same table */
523 seqcount_t *seqc; /* Protects moves */
524 struct ip_vs_aligned_lock *lock; /* Protect seqc */
525 int mask; /* Buckets mask */
526 int size; /* Buckets */
527 int seqc_mask; /* seqc mask */
528 int lock_mask; /* lock mask */
529 u32 table_id;
530 int u_thresh; /* upper threshold */
531 int l_thresh; /* lower threshold */
532 int lfactor; /* Load Factor (shift)*/
533 int bits; /* size = 1 << bits */
534 siphash_key_t hash_key;
535 struct rcu_head rcu_head;
536};
537
538/**
539 * ip_vs_rht_for_each_table() - Walk the hash tables
540 * @table: struct ip_vs_rht __rcu *table
541 * @t: current table, used as cursor, struct ip_vs_rht *var
542 * @p: previous table, temp struct ip_vs_rht *var
543 *
544 * Walk tables assuming others can not change the installed tables
545 */
546#define ip_vs_rht_for_each_table(table, t, p) \
547 for (p = NULL, t = rcu_dereference_protected(table, 1); \
548 t != p; \
549 p = t, t = rcu_dereference_protected(t->new_tbl, 1))
550
551/**
552 * ip_vs_rht_for_each_table_rcu() - Walk the hash tables under RCU reader lock
553 * @table: struct ip_vs_rht __rcu *table
554 * @t: current table, used as cursor, struct ip_vs_rht *var
555 * @p: previous table, temp struct ip_vs_rht *var
556 *
557 * We usually search in one table and also in second table on resizing
558 */
559#define ip_vs_rht_for_each_table_rcu(table, t, p) \
560 for (p = NULL, t = rcu_dereference(table); \
561 t != p; \
562 p = t, t = rcu_dereference(t->new_tbl))
563
564/**
565 * ip_vs_rht_for_each_bucket() - Walk all table buckets
566 * @t: current table, used as cursor, struct ip_vs_rht *var
567 * @bucket: bucket index, used as cursor, u32 var
568 * @head: bucket address, used as cursor, struct hlist_bl_head *var
569 */
570#define ip_vs_rht_for_each_bucket(t, bucket, head) \
571 for (bucket = 0, head = (t)->buckets; \
572 bucket < t->size; bucket++, head++)
573
574/**
575 * ip_vs_rht_for_bucket_retry() - Retry bucket if entries are moved
576 * @t: current table, used as cursor, struct ip_vs_rht *var
577 * @bucket: index of current bucket or hash key
578 * @sc: temp seqcount_t *var
579 * @seq: temp unsigned int var for sequence count
580 * @retry: temp int var
581 */
582#define ip_vs_rht_for_bucket_retry(t, bucket, sc, seq, retry) \
583 for (retry = 1, sc = &(t)->seqc[(bucket) & (t)->seqc_mask]; \
584 retry && ({ seq = read_seqcount_begin(sc); 1; }); \
585 retry = read_seqcount_retry(sc, seq))
586
587/**
588 * DECLARE_IP_VS_RHT_WALK_BUCKETS_RCU() - Declare variables
589 *
590 * Variables for ip_vs_rht_walk_buckets_rcu
591 */
592#define DECLARE_IP_VS_RHT_WALK_BUCKETS_RCU() \
593 struct ip_vs_rht *_t, *_p; \
594 unsigned int _seq; \
595 seqcount_t *_sc; \
596 u32 _bucket; \
597 int _retry
598/**
599 * ip_vs_rht_walk_buckets_rcu() - Walk all buckets under RCU read lock
600 * @table: struct ip_vs_rht __rcu *table
601 * @head: bucket address, used as cursor, struct hlist_bl_head *var
602 *
603 * Can be used while others add/delete/move entries
604 * Not suitable if duplicates are not desired
605 * Possible cases for reader that uses cond_resched_rcu() in the loop:
606 * - new table can not be installed, no need to repeat
607 * - new table can be installed => check and repeat if new table is
608 * installed, needed for !PREEMPT_RCU
609 */
610#define ip_vs_rht_walk_buckets_rcu(table, head) \
611 ip_vs_rht_for_each_table_rcu(table, _t, _p) \
612 ip_vs_rht_for_each_bucket(_t, _bucket, head) \
613 ip_vs_rht_for_bucket_retry(_t, _bucket, _sc, \
614 _seq, _retry)
615
616/**
617 * DECLARE_IP_VS_RHT_WALK_BUCKET_RCU() - Declare variables
618 *
619 * Variables for ip_vs_rht_walk_bucket_rcu
620 */
621#define DECLARE_IP_VS_RHT_WALK_BUCKET_RCU() \
622 unsigned int _seq; \
623 seqcount_t *_sc; \
624 int _retry
625/**
626 * ip_vs_rht_walk_bucket_rcu() - Walk bucket under RCU read lock
627 * @t: current table, struct ip_vs_rht *var
628 * @bucket: index of current bucket or hash key
629 * @head: bucket address, used as cursor, struct hlist_bl_head *var
630 *
631 * Can be used while others add/delete/move entries
632 * Not suitable if duplicates are not desired
633 * Possible cases for reader that uses cond_resched_rcu() in the loop:
634 * - new table can not be installed, no need to repeat
635 * - new table can be installed => check and repeat if new table is
636 * installed, needed for !PREEMPT_RCU
637 */
638#define ip_vs_rht_walk_bucket_rcu(t, bucket, head) \
639 if (({ head = (t)->buckets + ((bucket) & (t)->mask); 0; })) \
640 {} \
641 else \
642 ip_vs_rht_for_bucket_retry(t, (bucket), _sc, _seq, _retry)
643
644/**
645 * DECLARE_IP_VS_RHT_WALK_BUCKETS_SAFE_RCU() - Declare variables
646 *
647 * Variables for ip_vs_rht_walk_buckets_safe_rcu
648 */
649#define DECLARE_IP_VS_RHT_WALK_BUCKETS_SAFE_RCU() \
650 struct ip_vs_rht *_t, *_p; \
651 u32 _bucket
652/**
653 * ip_vs_rht_walk_buckets_safe_rcu() - Walk all buckets under RCU read lock
654 * @table: struct ip_vs_rht __rcu *table
655 * @head: bucket address, used as cursor, struct hlist_bl_head *var
656 *
657 * Can be used while others add/delete entries but moving is disabled
658 * Using cond_resched_rcu() should be safe if tables do not change
659 */
660#define ip_vs_rht_walk_buckets_safe_rcu(table, head) \
661 ip_vs_rht_for_each_table_rcu(table, _t, _p) \
662 ip_vs_rht_for_each_bucket(_t, _bucket, head)
663
664/**
665 * DECLARE_IP_VS_RHT_WALK_BUCKETS() - Declare variables
666 *
667 * Variables for ip_vs_rht_walk_buckets
668 */
669#define DECLARE_IP_VS_RHT_WALK_BUCKETS() \
670 struct ip_vs_rht *_t, *_p; \
671 u32 _bucket
672
673/**
674 * ip_vs_rht_walk_buckets() - Walk all buckets
675 * @table: struct ip_vs_rht __rcu *table
676 * @head: bucket address, used as cursor, struct hlist_bl_head *var
677 *
678 * Use if others can not add/delete/move entries
679 */
680#define ip_vs_rht_walk_buckets(table, head) \
681 ip_vs_rht_for_each_table(table, _t, _p) \
682 ip_vs_rht_for_each_bucket(_t, _bucket, head)
683
684/* Entries can be in one of two tables, so we flip bit when new table is
685 * created and store it as highest bit in hash keys
686 */
687#define IP_VS_RHT_TABLE_ID_MASK BIT(31)
688
689/* Check if hash key is from this table */
690static inline bool ip_vs_rht_same_table(struct ip_vs_rht *t, u32 hash_key)
691{
692 return !((t->table_id ^ hash_key) & IP_VS_RHT_TABLE_ID_MASK);
693}
694
695/* Build per-table hash key from hash value */
696static inline u32 ip_vs_rht_build_hash_key(struct ip_vs_rht *t, u32 hash)
697{
698 return t->table_id | (hash & ~IP_VS_RHT_TABLE_ID_MASK);
699}
700
701void ip_vs_rht_free(struct ip_vs_rht *t);
702void ip_vs_rht_rcu_free(struct rcu_head *head);
703struct ip_vs_rht *ip_vs_rht_alloc(int buckets, int scounts, int locks);
704int ip_vs_rht_desired_size(struct netns_ipvs *ipvs, struct ip_vs_rht *t, int n,
705 int lfactor, int min_bits, int max_bits);
706void ip_vs_rht_set_thresholds(struct ip_vs_rht *t, int size, int lfactor,
707 int min_bits, int max_bits);
708u32 ip_vs_rht_hash_linfo(struct ip_vs_rht *t, int af,
709 const union nf_inet_addr *addr, u32 v1, u32 v2);
710
711struct dst_entry;
712struct iphdr;
713struct ip_vs_conn;
714struct ip_vs_app;
715struct sk_buff;
716struct ip_vs_proto_data;
717
718struct ip_vs_protocol {
719 struct ip_vs_protocol *next;
720 char *name;
721 u16 protocol;
722 u16 num_states;
723 int dont_defrag;
724
725 void (*init)(struct ip_vs_protocol *pp);
726
727 void (*exit)(struct ip_vs_protocol *pp);
728
729 int (*init_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
730
731 void (*exit_netns)(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd);
732
733 int (*conn_schedule)(struct netns_ipvs *ipvs,
734 int af, struct sk_buff *skb,
735 struct ip_vs_proto_data *pd,
736 int *verdict, struct ip_vs_conn **cpp,
737 struct ip_vs_iphdr *iph);
738
739 struct ip_vs_conn *
740 (*conn_in_get)(struct netns_ipvs *ipvs,
741 int af,
742 const struct sk_buff *skb,
743 const struct ip_vs_iphdr *iph);
744
745 struct ip_vs_conn *
746 (*conn_out_get)(struct netns_ipvs *ipvs,
747 int af,
748 const struct sk_buff *skb,
749 const struct ip_vs_iphdr *iph);
750
751 int (*snat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
752 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
753
754 int (*dnat_handler)(struct sk_buff *skb, struct ip_vs_protocol *pp,
755 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph);
756
757 const char *(*state_name)(int state);
758
759 void (*state_transition)(struct ip_vs_conn *cp, int direction,
760 const struct sk_buff *skb,
761 struct ip_vs_proto_data *pd,
762 unsigned int iph_len);
763
764 int (*register_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
765
766 void (*unregister_app)(struct netns_ipvs *ipvs, struct ip_vs_app *inc);
767
768 int (*app_conn_bind)(struct ip_vs_conn *cp);
769
770 void (*debug_packet)(int af, struct ip_vs_protocol *pp,
771 const struct sk_buff *skb,
772 int offset,
773 const char *msg);
774
775 void (*timeout_change)(struct ip_vs_proto_data *pd, int flags);
776};
777
778/* protocol data per netns */
779struct ip_vs_proto_data {
780 struct ip_vs_proto_data *next;
781 struct ip_vs_protocol *pp;
782 int *timeout_table; /* protocol timeout table */
783 atomic_t appcnt; /* counter of proto app incs. */
784 struct tcp_states_t *tcp_state_table;
785};
786
787struct ip_vs_protocol *ip_vs_proto_get(unsigned short proto);
788struct ip_vs_proto_data *ip_vs_proto_data_get(struct netns_ipvs *ipvs,
789 unsigned short proto);
790
791struct ip_vs_conn_param {
792 struct netns_ipvs *ipvs;
793 const union nf_inet_addr *caddr;
794 const union nf_inet_addr *vaddr;
795 __be16 cport;
796 __be16 vport;
797 __u16 protocol;
798 u16 af;
799
800 const struct ip_vs_pe *pe;
801 char *pe_data;
802 __u8 pe_data_len;
803};
804
805/* Hash node in conn_tab */
806struct ip_vs_conn_hnode {
807 struct hlist_bl_node node; /* node in conn_tab */
808 u32 hash_key; /* Key for the hash table */
809 u8 dir; /* 0=out->in, 1=in->out */
810} __packed;
811
812/* IP_VS structure allocated for each dynamically scheduled connection */
813struct ip_vs_conn {
814 /* Cacheline for hash table nodes - rarely modified */
815
816 struct ip_vs_conn_hnode hn0; /* Original direction */
817 u8 af; /* address family */
818 __be16 cport;
819 struct ip_vs_conn_hnode hn1; /* Reply direction */
820 u8 daf; /* Address family of the dest */
821 __be16 dport;
822 struct ip_vs_dest *dest; /* real server */
823 atomic_t n_control; /* Number of controlled ones */
824 volatile __u32 flags; /* status flags */
825 /* 44/64 */
826
827 struct ip_vs_conn *control; /* Master control connection */
828 const struct ip_vs_pe *pe;
829 char *pe_data;
830 __u8 pe_data_len;
831 volatile __u16 state; /* state info */
832 volatile __u16 old_state; /* old state, to be used for
833 * state transition triggered
834 * synchronization
835 */
836 /* 2-byte hole */
837 /* 64/96 */
838
839 union nf_inet_addr caddr; /* client address */
840 union nf_inet_addr vaddr; /* virtual address */
841 /* 96/128 */
842
843 union nf_inet_addr daddr; /* destination address */
844 __u32 fwmark; /* Fire wall mark from skb */
845 __be16 vport;
846 __u16 protocol; /* Which protocol (TCP/UDP) */
847
848 /* Note: we can group the following members into a structure,
849 * in order to save more space, and the following members are
850 * only used in VS/NAT anyway
851 */
852 struct ip_vs_app *app; /* bound ip_vs_app object */
853 void *app_data; /* Application private data */
854 /* 128/168 */
855 struct_group(sync_conn_opt,
856 struct ip_vs_seq in_seq; /* incoming seq. struct */
857 struct ip_vs_seq out_seq; /* outgoing seq. struct */
858 );
859 /* 152/192 */
860
861 struct timer_list timer; /* Expiration timer */
862 volatile unsigned long timeout; /* timeout */
863 spinlock_t lock; /* lock for state transition */
864 refcount_t refcnt; /* reference count */
865 atomic_t in_pkts; /* incoming packet counter */
866 /* 64-bit: 4-byte gap */
867
868 /* 188/256 */
869 unsigned long sync_endtime; /* jiffies + sent_retries */
870 struct netns_ipvs *ipvs;
871
872 /* Packet transmitter for different forwarding methods. If it
873 * mangles the packet, it must return NF_DROP or better NF_STOLEN,
874 * otherwise this must be changed to a sk_buff **.
875 * NF_ACCEPT can be returned when destination is local.
876 */
877 int (*packet_xmit)(struct sk_buff *skb, struct ip_vs_conn *cp,
878 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
879
880 struct rcu_head rcu_head;
881};
882
883/* Extended internal versions of struct ip_vs_service_user and ip_vs_dest_user
884 * for IPv6 support.
885 *
886 * We need these to conveniently pass around service and destination
887 * options, but unfortunately, we also need to keep the old definitions to
888 * maintain userspace backwards compatibility for the setsockopt interface.
889 */
890struct ip_vs_service_user_kern {
891 /* virtual service addresses */
892 u16 af;
893 u16 protocol;
894 union nf_inet_addr addr; /* virtual ip address */
895 __be16 port;
896 u32 fwmark; /* firewall mark of service */
897
898 /* virtual service options */
899 char *sched_name;
900 char *pe_name;
901 unsigned int flags; /* virtual service flags */
902 unsigned int timeout; /* persistent timeout in sec */
903 __be32 netmask; /* persistent netmask or plen */
904};
905
906
907struct ip_vs_dest_user_kern {
908 /* destination server address */
909 union nf_inet_addr addr;
910 __be16 port;
911
912 /* real server options */
913 unsigned int conn_flags; /* connection flags */
914 int weight; /* destination weight */
915
916 /* thresholds for active connections */
917 u32 u_threshold; /* upper threshold */
918 u32 l_threshold; /* lower threshold */
919
920 /* Address family of addr */
921 u16 af;
922
923 u16 tun_type; /* tunnel type */
924 __be16 tun_port; /* tunnel port */
925 u16 tun_flags; /* tunnel flags */
926};
927
928
929/*
930 * The information about the virtual service offered to the net and the
931 * forwarding entries.
932 */
933struct ip_vs_service {
934 struct hlist_bl_node s_list; /* node in service table */
935 u32 hash_key; /* Key for the hash table */
936 u16 af; /* address family */
937 __u16 protocol; /* which protocol (TCP/UDP) */
938
939 union nf_inet_addr addr; /* IP address for virtual service */
940 __u32 fwmark; /* firewall mark of the service */
941 atomic_t refcnt; /* reference counter */
942 __be16 port; /* port number for the service */
943 unsigned int flags; /* service status flags */
944 unsigned int timeout; /* persistent timeout in ticks */
945 __be32 netmask; /* grouping granularity, mask/plen */
946 struct netns_ipvs *ipvs;
947
948 struct list_head destinations; /* real server d-linked list */
949 __u32 num_dests; /* number of servers */
950 struct ip_vs_stats stats; /* statistics for the service */
951
952 /* for scheduling */
953 struct ip_vs_scheduler __rcu *scheduler; /* bound scheduler object */
954 spinlock_t sched_lock; /* lock sched_data */
955 void *sched_data; /* scheduler application data */
956
957 /* alternate persistence engine */
958 struct ip_vs_pe __rcu *pe;
959 int conntrack_afmask;
960
961 struct rcu_head rcu_head;
962};
963
964/* Information for cached dst */
965struct ip_vs_dest_dst {
966 struct dst_entry *dst_cache; /* destination cache entry */
967 u32 dst_cookie;
968 union nf_inet_addr dst_saddr;
969 struct rcu_head rcu_head;
970};
971
972/* The real server destination forwarding entry with ip address, port number,
973 * and so on.
974 */
975struct ip_vs_dest {
976 struct list_head n_list; /* for the dests in the service */
977 struct hlist_node d_list; /* for table with all the dests */
978
979 u16 af; /* address family */
980 __be16 port; /* port number of the server */
981 union nf_inet_addr addr; /* IP address of the server */
982 volatile unsigned int flags; /* dest status flags */
983 atomic_t conn_flags; /* flags to copy to conn */
984 atomic_t weight; /* server weight */
985 unsigned long cflags; /* config flags */
986 atomic_t last_weight; /* server latest weight */
987 __u16 tun_type; /* tunnel type */
988 __be16 tun_port; /* tunnel port */
989 __u16 tun_flags; /* tunnel flags */
990
991 refcount_t refcnt; /* reference counter */
992 struct ip_vs_stats stats; /* statistics */
993 unsigned long idle_start; /* start time, jiffies */
994
995 /* connection counters and thresholds */
996 atomic_t activeconns; /* active connections */
997 atomic_t totalconns; /* total connections */
998 atomic_t persistconns; /* persistent connections */
999 __u32 u_threshold; /* upper threshold */
1000 __u32 l_threshold; /* lower threshold */
1001 __u32 l_threshold_val;/* used lower threshold */
1002
1003 /* for destination cache */
1004 spinlock_t dst_lock; /* lock of dst_cache */
1005 struct ip_vs_dest_dst __rcu *dest_dst; /* cached dst info */
1006
1007 /* for virtual service */
1008 struct ip_vs_service __rcu *svc; /* service it belongs to */
1009 __u16 protocol; /* which protocol (TCP/UDP) */
1010 __be16 vport; /* virtual port number */
1011 union nf_inet_addr vaddr; /* virtual IP address */
1012 __u32 vfwmark; /* firewall mark of service */
1013
1014 struct rcu_head rcu_head;
1015 struct list_head t_list; /* in dest_trash */
1016 unsigned int in_rs_table:1; /* we are in rs_table */
1017};
1018
1019/* The scheduler object */
1020struct ip_vs_scheduler {
1021 struct list_head n_list; /* d-linked list head */
1022 char *name; /* scheduler name */
1023 atomic_t refcnt; /* reference counter */
1024 struct module *module; /* THIS_MODULE/NULL */
1025
1026 /* scheduler initializing service */
1027 int (*init_service)(struct ip_vs_service *svc);
1028 /* scheduling service finish */
1029 void (*done_service)(struct ip_vs_service *svc);
1030 /* dest is linked */
1031 int (*add_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
1032 /* dest is unlinked */
1033 int (*del_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
1034 /* dest is updated */
1035 int (*upd_dest)(struct ip_vs_service *svc, struct ip_vs_dest *dest);
1036
1037 /* selecting a server from the given service */
1038 struct ip_vs_dest* (*schedule)(struct ip_vs_service *svc,
1039 const struct sk_buff *skb,
1040 struct ip_vs_iphdr *iph);
1041};
1042
1043/* The persistence engine object */
1044struct ip_vs_pe {
1045 struct list_head n_list; /* d-linked list head */
1046 char *name; /* scheduler name */
1047 atomic_t refcnt; /* reference counter */
1048 struct module *module; /* THIS_MODULE/NULL */
1049
1050 /* get the connection template, if any */
1051 int (*fill_param)(struct ip_vs_conn_param *p, struct sk_buff *skb);
1052 bool (*ct_match)(const struct ip_vs_conn_param *p,
1053 struct ip_vs_conn *ct);
1054 u32 (*hashkey_raw)(const struct ip_vs_conn_param *p,
1055 struct ip_vs_rht *t, bool inverse);
1056 int (*show_pe_data)(const struct ip_vs_conn *cp, char *buf);
1057 /* create connections for real-server outgoing packets */
1058 struct ip_vs_conn* (*conn_out)(struct ip_vs_service *svc,
1059 struct ip_vs_dest *dest,
1060 struct sk_buff *skb,
1061 const struct ip_vs_iphdr *iph,
1062 __be16 dport, __be16 cport);
1063};
1064
1065/* The application module object (a.k.a. app incarnation) */
1066struct ip_vs_app {
1067 struct list_head a_list; /* member in app list */
1068 int type; /* IP_VS_APP_TYPE_xxx */
1069 char *name; /* application module name */
1070 __u16 protocol;
1071 struct module *module; /* THIS_MODULE/NULL */
1072 struct list_head incs_list; /* list of incarnations */
1073
1074 /* members for application incarnations */
1075 struct list_head p_list; /* member in proto app list */
1076 struct ip_vs_app *app; /* its real application */
1077 __be16 port; /* port number in net order */
1078 atomic_t usecnt; /* usage counter */
1079 struct rcu_head rcu_head;
1080
1081 /* output hook: Process packet in inout direction, diff set for TCP.
1082 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
1083 * 2=Mangled but checksum was not updated
1084 */
1085 int (*pkt_out)(struct ip_vs_app *, struct ip_vs_conn *,
1086 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh);
1087
1088 /* input hook: Process packet in outin direction, diff set for TCP.
1089 * Return: 0=Error, 1=Payload Not Mangled/Mangled but checksum is ok,
1090 * 2=Mangled but checksum was not updated
1091 */
1092 int (*pkt_in)(struct ip_vs_app *, struct ip_vs_conn *,
1093 struct sk_buff *, int *diff, struct ip_vs_iphdr *ipvsh);
1094
1095 /* ip_vs_app initializer */
1096 int (*init_conn)(struct ip_vs_app *, struct ip_vs_conn *);
1097
1098 /* ip_vs_app finish */
1099 int (*done_conn)(struct ip_vs_app *, struct ip_vs_conn *);
1100
1101
1102 /* not used now */
1103 int (*bind_conn)(struct ip_vs_app *, struct ip_vs_conn *,
1104 struct ip_vs_protocol *);
1105
1106 void (*unbind_conn)(struct ip_vs_app *, struct ip_vs_conn *);
1107
1108 int * timeout_table;
1109 int * timeouts;
1110 int timeouts_size;
1111
1112 int (*conn_schedule)(struct sk_buff *skb, struct ip_vs_app *app,
1113 int *verdict, struct ip_vs_conn **cpp);
1114
1115 struct ip_vs_conn *
1116 (*conn_in_get)(const struct sk_buff *skb, struct ip_vs_app *app,
1117 const struct iphdr *iph, int inverse);
1118
1119 struct ip_vs_conn *
1120 (*conn_out_get)(const struct sk_buff *skb, struct ip_vs_app *app,
1121 const struct iphdr *iph, int inverse);
1122
1123 int (*state_transition)(struct ip_vs_conn *cp, int direction,
1124 const struct sk_buff *skb,
1125 struct ip_vs_app *app);
1126
1127 void (*timeout_change)(struct ip_vs_app *app, int flags);
1128};
1129
1130struct ipvs_master_sync_state {
1131 struct list_head sync_queue;
1132 struct ip_vs_sync_buff *sync_buff;
1133 unsigned long sync_queue_len;
1134 unsigned int sync_queue_delay;
1135 struct delayed_work master_wakeup_work;
1136 struct netns_ipvs *ipvs;
1137};
1138
1139struct ip_vs_sync_thread_data;
1140
1141/* How much time to keep dests in trash */
1142#define IP_VS_DEST_TRASH_PERIOD (120 * HZ)
1143
1144struct ipvs_sync_daemon_cfg {
1145 union nf_inet_addr mcast_group;
1146 int syncid;
1147 u16 sync_maxlen;
1148 u16 mcast_port;
1149 u8 mcast_af;
1150 u8 mcast_ttl;
1151 /* multicast interface name */
1152 char mcast_ifn[IP_VS_IFNAME_MAXLEN];
1153};
1154
1155/* IPVS in network namespace */
1156struct netns_ipvs {
1157 int gen; /* Generation */
1158 int enable; /* enable like nf_hooks do */
1159 /* Hash table: for real service lookups */
1160 #define IP_VS_RTAB_BITS 4
1161 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
1162 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
1163
1164 struct hlist_head rs_table[IP_VS_RTAB_SIZE];
1165 /* ip_vs_app */
1166 struct list_head app_list;
1167 /* ip_vs_proto */
1168 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */
1169 struct ip_vs_proto_data *proto_data_table[IP_VS_PROTO_TAB_SIZE];
1170 /* ip_vs_proto_tcp */
1171#ifdef CONFIG_IP_VS_PROTO_TCP
1172 #define TCP_APP_TAB_BITS 4
1173 #define TCP_APP_TAB_SIZE (1 << TCP_APP_TAB_BITS)
1174 #define TCP_APP_TAB_MASK (TCP_APP_TAB_SIZE - 1)
1175 struct list_head tcp_apps[TCP_APP_TAB_SIZE];
1176#endif
1177 /* ip_vs_proto_udp */
1178#ifdef CONFIG_IP_VS_PROTO_UDP
1179 #define UDP_APP_TAB_BITS 4
1180 #define UDP_APP_TAB_SIZE (1 << UDP_APP_TAB_BITS)
1181 #define UDP_APP_TAB_MASK (UDP_APP_TAB_SIZE - 1)
1182 struct list_head udp_apps[UDP_APP_TAB_SIZE];
1183#endif
1184 /* ip_vs_proto_sctp */
1185#ifdef CONFIG_IP_VS_PROTO_SCTP
1186 #define SCTP_APP_TAB_BITS 4
1187 #define SCTP_APP_TAB_SIZE (1 << SCTP_APP_TAB_BITS)
1188 #define SCTP_APP_TAB_MASK (SCTP_APP_TAB_SIZE - 1)
1189 /* Hash table for SCTP application incarnations */
1190 struct list_head sctp_apps[SCTP_APP_TAB_SIZE];
1191#endif
1192 /* ip_vs_conn */
1193 atomic_t conn_count; /* connection counter */
1194 atomic_t no_cport_conns[IP_VS_AF_MAX];
1195 struct delayed_work conn_resize_work;/* resize conn_tab */
1196
1197 /* ip_vs_ctl */
1198 struct ip_vs_stats_rcu *tot_stats; /* Statistics & est. */
1199
1200 /* Trash for destinations */
1201 struct list_head dest_trash;
1202 spinlock_t dest_trash_lock;
1203 struct timer_list dest_trash_timer; /* expiration timer */
1204 struct mutex service_mutex; /* service reconfig */
1205 struct rw_semaphore svc_resize_sem; /* svc_table resizing */
1206 struct rw_semaphore svc_replace_sem; /* svc_table replace */
1207 struct delayed_work svc_resize_work; /* resize svc_table */
1208 atomic_t svc_table_changes;/* ++ on table changes */
1209 /* Service counters */
1210 atomic_t num_services[IP_VS_AF_MAX]; /* Services */
1211 atomic_t fwm_services[IP_VS_AF_MAX]; /* Services */
1212 atomic_t nonfwm_services[IP_VS_AF_MAX];/* Services */
1213 atomic_t ftpsvc_counter[IP_VS_AF_MAX]; /* FTPPORT */
1214 atomic_t nullsvc_counter[IP_VS_AF_MAX];/* Zero port */
1215 atomic_t conn_out_counter[IP_VS_AF_MAX];/* out conn */
1216
1217#ifdef CONFIG_SYSCTL
1218 /* delayed work for expiring no dest connections */
1219 struct delayed_work expire_nodest_conn_work;
1220 /* 1/rate drop and drop-entry variables */
1221 struct delayed_work defense_work; /* Work handler */
1222 int drop_rate;
1223 int drop_counter;
1224 int old_secure_tcp;
1225 atomic_t dropentry;
1226 s8 dropentry_counters[8];
1227 /* locks in ctl.c */
1228 spinlock_t dropentry_lock; /* drop entry handling */
1229 spinlock_t droppacket_lock; /* drop packet handling */
1230 spinlock_t securetcp_lock; /* state and timeout tables */
1231
1232 /* sys-ctl struct */
1233 struct ctl_table_header *sysctl_hdr;
1234 struct ctl_table *sysctl_tbl;
1235#endif
1236
1237 /* sysctl variables */
1238 int sysctl_amemthresh;
1239 int sysctl_am_droprate;
1240#ifdef CONFIG_SYSCTL
1241 int sysctl_conn_max;/* soft limit for conns */
1242 int conn_max_limit; /* hard limit for conn_max */
1243#endif
1244 int sysctl_drop_entry;
1245 int sysctl_drop_packet;
1246 int sysctl_secure_tcp;
1247#ifdef CONFIG_IP_VS_NFCT
1248 int sysctl_conntrack;
1249#endif
1250 int sysctl_snat_reroute;
1251 int sysctl_sync_ver;
1252 int sysctl_sync_ports;
1253 int sysctl_sync_persist_mode;
1254 unsigned long sysctl_sync_qlen_max;
1255 int sysctl_sync_sock_size;
1256 int sysctl_cache_bypass;
1257 int sysctl_expire_nodest_conn;
1258 int sysctl_sloppy_tcp;
1259 int sysctl_sloppy_sctp;
1260 int sysctl_expire_quiescent_template;
1261 int sysctl_sync_threshold[2];
1262 unsigned int sysctl_sync_refresh_period;
1263 int sysctl_sync_retries;
1264 int sysctl_nat_icmp_send;
1265 int sysctl_pmtu_disc;
1266 int sysctl_backup_only;
1267 int sysctl_conn_reuse_mode;
1268 int sysctl_schedule_icmp;
1269 int sysctl_ignore_tunneled;
1270 int sysctl_run_estimation;
1271#ifdef CONFIG_SYSCTL
1272 cpumask_var_t sysctl_est_cpulist; /* kthread cpumask */
1273 int est_cpulist_valid; /* cpulist set */
1274 int sysctl_est_nice; /* kthread nice */
1275 int est_stopped; /* stop tasks */
1276#endif
1277 int sysctl_conn_lfactor;
1278 int sysctl_svc_lfactor;
1279
1280 /* ip_vs_lblc */
1281 int sysctl_lblc_expiration;
1282 struct ctl_table_header *lblc_ctl_header;
1283 struct ctl_table *lblc_ctl_table;
1284 /* ip_vs_lblcr */
1285 int sysctl_lblcr_expiration;
1286 struct ctl_table_header *lblcr_ctl_header;
1287 struct ctl_table *lblcr_ctl_table;
1288 unsigned long work_flags; /* IP_VS_WORK_* flags */
1289 /* ip_vs_est */
1290 struct delayed_work est_reload_work;/* Reload kthread tasks */
1291 struct mutex est_mutex; /* protect kthread tasks */
1292 struct hlist_head est_temp_list; /* Ests during calc phase */
1293 struct ip_vs_est_kt_data **est_kt_arr; /* Array of kthread data ptrs */
1294 unsigned long est_max_threads;/* Hard limit of kthreads */
1295 int est_calc_phase; /* Calculation phase */
1296 int est_chain_max; /* Calculated chain_max */
1297 int est_kt_count; /* Allocated ptrs */
1298 int est_add_ktid; /* ktid where to add ests */
1299 atomic_t est_genid; /* kthreads reload genid */
1300 atomic_t est_genid_done; /* applied genid */
1301 /* ip_vs_sync */
1302 spinlock_t sync_lock;
1303 struct ipvs_master_sync_state *ms;
1304 spinlock_t sync_buff_lock;
1305 struct ip_vs_sync_thread_data *master_tinfo;
1306 struct ip_vs_sync_thread_data *backup_tinfo;
1307 int threads_mask;
1308 volatile int sync_state;
1309 struct mutex sync_mutex;
1310 struct ipvs_sync_daemon_cfg mcfg; /* Master Configuration */
1311 struct ipvs_sync_daemon_cfg bcfg; /* Backup Configuration */
1312 /* net name space ptr */
1313 struct net *net; /* Needed by timer routines */
1314 /* Number of heterogeneous destinations, needed because heterogeneous
1315 * are not supported when synchronization is enabled.
1316 */
1317 unsigned int mixed_address_family_dests;
1318 unsigned int hooks_afmask; /* &1=AF_INET, &2=AF_INET6 */
1319
1320 struct ip_vs_rht __rcu *svc_table; /* Services */
1321 struct ip_vs_rht __rcu *conn_tab; /* Connections */
1322 atomic_t conn_tab_changes;/* ++ on new table */
1323};
1324
1325#define DEFAULT_SYNC_THRESHOLD 3
1326#define DEFAULT_SYNC_PERIOD 50
1327#define DEFAULT_SYNC_VER 1
1328#define DEFAULT_SLOPPY_TCP 0
1329#define DEFAULT_SLOPPY_SCTP 0
1330#define DEFAULT_SYNC_REFRESH_PERIOD (0U * HZ)
1331#define DEFAULT_SYNC_RETRIES 0
1332#define IPVS_SYNC_WAKEUP_RATE 8
1333#define IPVS_SYNC_QLEN_MAX (IPVS_SYNC_WAKEUP_RATE * 4)
1334#define IPVS_SYNC_SEND_DELAY (HZ / 50)
1335#define IPVS_SYNC_CHECK_PERIOD HZ
1336#define IPVS_SYNC_FLUSH_TIME (HZ * 2)
1337#define IPVS_SYNC_PORTS_MAX (1 << 6)
1338
1339#ifdef CONFIG_SYSCTL
1340
1341static inline int sysctl_conn_max(struct netns_ipvs *ipvs)
1342{
1343 return READ_ONCE(ipvs->sysctl_conn_max);
1344}
1345
1346static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1347{
1348 return ipvs->sysctl_sync_threshold[0];
1349}
1350
1351static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1352{
1353 return READ_ONCE(ipvs->sysctl_sync_threshold[1]);
1354}
1355
1356static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1357{
1358 return READ_ONCE(ipvs->sysctl_sync_refresh_period);
1359}
1360
1361static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1362{
1363 return ipvs->sysctl_sync_retries;
1364}
1365
1366static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1367{
1368 return ipvs->sysctl_sync_ver;
1369}
1370
1371static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1372{
1373 return ipvs->sysctl_sloppy_tcp;
1374}
1375
1376static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1377{
1378 return ipvs->sysctl_sloppy_sctp;
1379}
1380
1381static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1382{
1383 return READ_ONCE(ipvs->sysctl_sync_ports);
1384}
1385
1386static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1387{
1388 return ipvs->sysctl_sync_persist_mode;
1389}
1390
1391static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1392{
1393 return ipvs->sysctl_sync_qlen_max;
1394}
1395
1396static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1397{
1398 return ipvs->sysctl_sync_sock_size;
1399}
1400
1401static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1402{
1403 return ipvs->sysctl_pmtu_disc;
1404}
1405
1406static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1407{
1408 return ipvs->sync_state & IP_VS_STATE_BACKUP &&
1409 ipvs->sysctl_backup_only;
1410}
1411
1412static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1413{
1414 return ipvs->sysctl_conn_reuse_mode;
1415}
1416
1417static inline int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs)
1418{
1419 return ipvs->sysctl_expire_nodest_conn;
1420}
1421
1422static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1423{
1424 return ipvs->sysctl_schedule_icmp;
1425}
1426
1427static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1428{
1429 return ipvs->sysctl_ignore_tunneled;
1430}
1431
1432static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1433{
1434 return ipvs->sysctl_cache_bypass;
1435}
1436
1437static inline int sysctl_run_estimation(struct netns_ipvs *ipvs)
1438{
1439 return ipvs->sysctl_run_estimation;
1440}
1441
1442static inline const struct cpumask *__sysctl_est_cpulist(struct netns_ipvs *ipvs)
1443{
1444 if (ipvs->est_cpulist_valid)
1445 return ipvs->sysctl_est_cpulist;
1446 else
1447 return housekeeping_cpumask(HK_TYPE_KTHREAD);
1448}
1449
1450static inline const struct cpumask *sysctl_est_preferred_cpulist(struct netns_ipvs *ipvs)
1451{
1452 if (ipvs->est_cpulist_valid)
1453 return ipvs->sysctl_est_cpulist;
1454 else
1455 return NULL;
1456}
1457
1458static inline int sysctl_est_nice(struct netns_ipvs *ipvs)
1459{
1460 return ipvs->sysctl_est_nice;
1461}
1462
1463#else
1464
1465static inline int sysctl_conn_max(struct netns_ipvs *ipvs)
1466{
1467 return IP_VS_CONN_MAX;
1468}
1469
1470static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
1471{
1472 return DEFAULT_SYNC_THRESHOLD;
1473}
1474
1475static inline int sysctl_sync_period(struct netns_ipvs *ipvs)
1476{
1477 return DEFAULT_SYNC_PERIOD;
1478}
1479
1480static inline unsigned int sysctl_sync_refresh_period(struct netns_ipvs *ipvs)
1481{
1482 return DEFAULT_SYNC_REFRESH_PERIOD;
1483}
1484
1485static inline int sysctl_sync_retries(struct netns_ipvs *ipvs)
1486{
1487 return DEFAULT_SYNC_RETRIES & 3;
1488}
1489
1490static inline int sysctl_sync_ver(struct netns_ipvs *ipvs)
1491{
1492 return DEFAULT_SYNC_VER;
1493}
1494
1495static inline int sysctl_sloppy_tcp(struct netns_ipvs *ipvs)
1496{
1497 return DEFAULT_SLOPPY_TCP;
1498}
1499
1500static inline int sysctl_sloppy_sctp(struct netns_ipvs *ipvs)
1501{
1502 return DEFAULT_SLOPPY_SCTP;
1503}
1504
1505static inline int sysctl_sync_ports(struct netns_ipvs *ipvs)
1506{
1507 return 1;
1508}
1509
1510static inline int sysctl_sync_persist_mode(struct netns_ipvs *ipvs)
1511{
1512 return 0;
1513}
1514
1515static inline unsigned long sysctl_sync_qlen_max(struct netns_ipvs *ipvs)
1516{
1517 return IPVS_SYNC_QLEN_MAX;
1518}
1519
1520static inline int sysctl_sync_sock_size(struct netns_ipvs *ipvs)
1521{
1522 return 0;
1523}
1524
1525static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
1526{
1527 return 1;
1528}
1529
1530static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
1531{
1532 return 0;
1533}
1534
1535static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
1536{
1537 return 1;
1538}
1539
1540static inline int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs)
1541{
1542 return 0;
1543}
1544
1545static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
1546{
1547 return 0;
1548}
1549
1550static inline int sysctl_ignore_tunneled(struct netns_ipvs *ipvs)
1551{
1552 return 0;
1553}
1554
1555static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs)
1556{
1557 return 0;
1558}
1559
1560static inline int sysctl_run_estimation(struct netns_ipvs *ipvs)
1561{
1562 return 1;
1563}
1564
1565static inline const struct cpumask *__sysctl_est_cpulist(struct netns_ipvs *ipvs)
1566{
1567 return housekeeping_cpumask(HK_TYPE_KTHREAD);
1568}
1569
1570static inline const struct cpumask *sysctl_est_preferred_cpulist(struct netns_ipvs *ipvs)
1571{
1572 return NULL;
1573}
1574
1575static inline int sysctl_est_nice(struct netns_ipvs *ipvs)
1576{
1577 return IPVS_EST_NICE;
1578}
1579
1580#endif
1581
1582/* Get load factor to map conn_count/u_thresh to t->size */
1583static inline int sysctl_conn_lfactor(struct netns_ipvs *ipvs)
1584{
1585 return READ_ONCE(ipvs->sysctl_conn_lfactor);
1586}
1587
1588/* Get load factor to map num_services/u_thresh to t->size
1589 * Smaller value decreases u_thresh to reduce collisions but increases
1590 * the table size
1591 * Returns factor where:
1592 * - <0: u_thresh = size >> -factor, eg. lfactor -2 = 25% load
1593 * - >=0: u_thresh = size << factor, eg. lfactor 1 = 200% load
1594 */
1595static inline int sysctl_svc_lfactor(struct netns_ipvs *ipvs)
1596{
1597 return READ_ONCE(ipvs->sysctl_svc_lfactor);
1598}
1599
1600static inline bool sysctl_est_cpulist_empty(struct netns_ipvs *ipvs)
1601{
1602 guard(rcu)();
1603 return cpumask_empty(__sysctl_est_cpulist(ipvs));
1604}
1605
1606static inline unsigned int sysctl_est_cpulist_weight(struct netns_ipvs *ipvs)
1607{
1608 guard(rcu)();
1609 return cpumask_weight(__sysctl_est_cpulist(ipvs));
1610}
1611
1612/* IPVS core functions
1613 * (from ip_vs_core.c)
1614 */
1615const char *ip_vs_proto_name(unsigned int proto);
1616void ip_vs_init_hash_table(struct list_head *table, int rows);
1617struct ip_vs_conn *ip_vs_new_conn_out(struct ip_vs_service *svc,
1618 struct ip_vs_dest *dest,
1619 struct sk_buff *skb,
1620 const struct ip_vs_iphdr *iph,
1621 __be16 dport,
1622 __be16 cport);
1623#define IP_VS_INIT_HASH_TABLE(t) ip_vs_init_hash_table((t), ARRAY_SIZE((t)))
1624
1625#define IP_VS_APP_TYPE_FTP 1
1626
1627/* ip_vs_conn handling functions
1628 * (from ip_vs_conn.c)
1629 */
1630enum {
1631 IP_VS_DIR_INPUT = 0,
1632 IP_VS_DIR_OUTPUT,
1633 IP_VS_DIR_INPUT_ONLY,
1634 IP_VS_DIR_LAST,
1635};
1636
1637static inline void ip_vs_conn_fill_param(struct netns_ipvs *ipvs, int af, int protocol,
1638 const union nf_inet_addr *caddr,
1639 __be16 cport,
1640 const union nf_inet_addr *vaddr,
1641 __be16 vport,
1642 struct ip_vs_conn_param *p)
1643{
1644 p->ipvs = ipvs;
1645 p->af = af;
1646 p->protocol = protocol;
1647 p->caddr = caddr;
1648 p->cport = cport;
1649 p->vaddr = vaddr;
1650 p->vport = vport;
1651 p->pe = NULL;
1652 p->pe_data = NULL;
1653}
1654
1655struct ip_vs_conn *ip_vs_conn_in_get(const struct ip_vs_conn_param *p);
1656struct ip_vs_conn *ip_vs_ct_in_get(const struct ip_vs_conn_param *p);
1657
1658struct ip_vs_conn * ip_vs_conn_in_get_proto(struct netns_ipvs *ipvs, int af,
1659 const struct sk_buff *skb,
1660 const struct ip_vs_iphdr *iph);
1661
1662struct ip_vs_conn *ip_vs_conn_out_get(const struct ip_vs_conn_param *p);
1663
1664struct ip_vs_conn * ip_vs_conn_out_get_proto(struct netns_ipvs *ipvs, int af,
1665 const struct sk_buff *skb,
1666 const struct ip_vs_iphdr *iph);
1667
1668/* Get reference to gain full access to conn.
1669 * By default, RCU read-side critical sections have access only to
1670 * conn fields and its PE data, see ip_vs_conn_rcu_free() for reference.
1671 */
1672static inline bool __ip_vs_conn_get(struct ip_vs_conn *cp)
1673{
1674 return refcount_inc_not_zero(&cp->refcnt);
1675}
1676
1677/* put back the conn without restarting its timer */
1678static inline void __ip_vs_conn_put(struct ip_vs_conn *cp)
1679{
1680 smp_mb__before_atomic();
1681 refcount_dec(&cp->refcnt);
1682}
1683void ip_vs_conn_put(struct ip_vs_conn *cp);
1684void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport);
1685int ip_vs_conn_desired_size(struct netns_ipvs *ipvs, struct ip_vs_rht *t,
1686 int lfactor);
1687struct ip_vs_rht *ip_vs_conn_tab_alloc(struct netns_ipvs *ipvs, int buckets,
1688 int lfactor);
1689
1690static inline struct ip_vs_conn *
1691ip_vs_hn0_to_conn(struct ip_vs_conn_hnode *hn)
1692{
1693 return container_of(hn, struct ip_vs_conn, hn0);
1694}
1695
1696static inline struct ip_vs_conn *
1697ip_vs_hn_to_conn(struct ip_vs_conn_hnode *hn)
1698{
1699 return hn->dir ? container_of(hn, struct ip_vs_conn, hn1) :
1700 container_of(hn, struct ip_vs_conn, hn0);
1701}
1702
1703struct ip_vs_conn *ip_vs_conn_new(const struct ip_vs_conn_param *p, int dest_af,
1704 const union nf_inet_addr *daddr,
1705 __be16 dport, unsigned int flags,
1706 struct ip_vs_dest *dest, __u32 fwmark);
1707void ip_vs_conn_expire_now(struct ip_vs_conn *cp);
1708
1709const char *ip_vs_state_name(const struct ip_vs_conn *cp);
1710
1711void ip_vs_tcp_conn_listen(struct ip_vs_conn *cp);
1712int ip_vs_check_template(struct ip_vs_conn *ct, struct ip_vs_dest *cdest);
1713void ip_vs_random_dropentry(struct netns_ipvs *ipvs);
1714int ip_vs_conn_init(void);
1715void ip_vs_conn_cleanup(void);
1716
1717static inline void ip_vs_control_del(struct ip_vs_conn *cp)
1718{
1719 struct ip_vs_conn *ctl_cp = cp->control;
1720 if (!ctl_cp) {
1721 IP_VS_ERR_BUF("request control DEL for uncontrolled: "
1722 "%s:%d to %s:%d\n",
1723 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1724 ntohs(cp->cport),
1725 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1726 ntohs(cp->vport));
1727
1728 return;
1729 }
1730
1731 IP_VS_DBG_BUF(7, "DELeting control for: "
1732 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1733 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1734 ntohs(cp->cport),
1735 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1736 ntohs(ctl_cp->cport));
1737
1738 cp->control = NULL;
1739 if (atomic_read(&ctl_cp->n_control) == 0) {
1740 IP_VS_ERR_BUF("BUG control DEL with n=0 : "
1741 "%s:%d to %s:%d\n",
1742 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1743 ntohs(cp->cport),
1744 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1745 ntohs(cp->vport));
1746
1747 return;
1748 }
1749 atomic_dec(&ctl_cp->n_control);
1750}
1751
1752static inline void
1753ip_vs_control_add(struct ip_vs_conn *cp, struct ip_vs_conn *ctl_cp)
1754{
1755 if (cp->control) {
1756 IP_VS_ERR_BUF("request control ADD for already controlled: "
1757 "%s:%d to %s:%d\n",
1758 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1759 ntohs(cp->cport),
1760 IP_VS_DBG_ADDR(cp->af, &cp->vaddr),
1761 ntohs(cp->vport));
1762
1763 ip_vs_control_del(cp);
1764 }
1765
1766 IP_VS_DBG_BUF(7, "ADDing control for: "
1767 "cp.dst=%s:%d ctl_cp.dst=%s:%d\n",
1768 IP_VS_DBG_ADDR(cp->af, &cp->caddr),
1769 ntohs(cp->cport),
1770 IP_VS_DBG_ADDR(cp->af, &ctl_cp->caddr),
1771 ntohs(ctl_cp->cport));
1772
1773 cp->control = ctl_cp;
1774 atomic_inc(&ctl_cp->n_control);
1775}
1776
1777/* Mark our template as assured */
1778static inline void
1779ip_vs_control_assure_ct(struct ip_vs_conn *cp)
1780{
1781 struct ip_vs_conn *ct = cp->control;
1782
1783 if (ct && !(ct->state & IP_VS_CTPL_S_ASSURED) &&
1784 (ct->flags & IP_VS_CONN_F_TEMPLATE))
1785 ct->state |= IP_VS_CTPL_S_ASSURED;
1786}
1787
1788/* IPVS netns init & cleanup functions */
1789int ip_vs_estimator_net_init(struct netns_ipvs *ipvs);
1790int ip_vs_control_net_init(struct netns_ipvs *ipvs);
1791int ip_vs_protocol_net_init(struct netns_ipvs *ipvs);
1792int ip_vs_app_net_init(struct netns_ipvs *ipvs);
1793int ip_vs_conn_net_init(struct netns_ipvs *ipvs);
1794int ip_vs_sync_net_init(struct netns_ipvs *ipvs);
1795void ip_vs_conn_net_cleanup(struct netns_ipvs *ipvs);
1796void ip_vs_app_net_cleanup(struct netns_ipvs *ipvs);
1797void ip_vs_protocol_net_cleanup(struct netns_ipvs *ipvs);
1798void ip_vs_control_net_cleanup(struct netns_ipvs *ipvs);
1799void ip_vs_estimator_net_cleanup(struct netns_ipvs *ipvs);
1800void ip_vs_sync_net_cleanup(struct netns_ipvs *ipvs);
1801void ip_vs_service_nets_cleanup(struct list_head *net_list);
1802
1803/* IPVS application functions
1804 * (from ip_vs_app.c)
1805 */
1806#define IP_VS_APP_MAX_PORTS 8
1807struct ip_vs_app *register_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1808void unregister_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app);
1809int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp);
1810void ip_vs_unbind_app(struct ip_vs_conn *cp);
1811int register_ip_vs_app_inc(struct netns_ipvs *ipvs, struct ip_vs_app *app, __u16 proto,
1812 __u16 port);
1813int ip_vs_app_inc_get(struct ip_vs_app *inc);
1814void ip_vs_app_inc_put(struct ip_vs_app *inc);
1815
1816int ip_vs_app_pkt_out(struct ip_vs_conn *, struct sk_buff *skb,
1817 struct ip_vs_iphdr *ipvsh);
1818int ip_vs_app_pkt_in(struct ip_vs_conn *, struct sk_buff *skb,
1819 struct ip_vs_iphdr *ipvsh);
1820
1821int register_ip_vs_pe(struct ip_vs_pe *pe);
1822int unregister_ip_vs_pe(struct ip_vs_pe *pe);
1823struct ip_vs_pe *ip_vs_pe_getbyname(const char *name);
1824struct ip_vs_pe *__ip_vs_pe_getbyname(const char *pe_name);
1825
1826/* Use a #define to avoid all of module.h just for these trivial ops */
1827#define ip_vs_pe_get(pe) \
1828 if (pe && pe->module) \
1829 __module_get(pe->module);
1830
1831#define ip_vs_pe_put(pe) \
1832 if (pe && pe->module) \
1833 module_put(pe->module);
1834
1835/* IPVS protocol functions (from ip_vs_proto.c) */
1836int ip_vs_protocol_init(void);
1837void ip_vs_protocol_cleanup(void);
1838void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags);
1839int *ip_vs_create_timeout_table(int *table, int size);
1840void ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
1841 const struct sk_buff *skb, int offset,
1842 const char *msg);
1843
1844extern struct ip_vs_protocol ip_vs_protocol_tcp;
1845extern struct ip_vs_protocol ip_vs_protocol_udp;
1846extern struct ip_vs_protocol ip_vs_protocol_icmp;
1847extern struct ip_vs_protocol ip_vs_protocol_esp;
1848extern struct ip_vs_protocol ip_vs_protocol_ah;
1849extern struct ip_vs_protocol ip_vs_protocol_sctp;
1850
1851/* Registering/unregistering scheduler functions
1852 * (from ip_vs_sched.c)
1853 */
1854int register_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1855int unregister_ip_vs_scheduler(struct ip_vs_scheduler *scheduler);
1856int ip_vs_bind_scheduler(struct ip_vs_service *svc,
1857 struct ip_vs_scheduler *scheduler);
1858void ip_vs_unbind_scheduler(struct ip_vs_service *svc);
1859struct ip_vs_scheduler *ip_vs_scheduler_get(const char *sched_name);
1860void ip_vs_scheduler_put(struct ip_vs_scheduler *scheduler);
1861struct ip_vs_conn *
1862ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb,
1863 struct ip_vs_proto_data *pd, int *ignored,
1864 struct ip_vs_iphdr *iph);
1865int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb,
1866 struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph);
1867
1868void ip_vs_scheduler_err(struct ip_vs_service *svc, const char *msg);
1869
1870/* IPVS control data and functions (from ip_vs_ctl.c) */
1871extern struct ip_vs_stats ip_vs_stats;
1872extern int sysctl_ip_vs_sync_ver;
1873
1874struct ip_vs_service *
1875ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
1876 const union nf_inet_addr *vaddr, __be16 vport);
1877
1878bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1879 const union nf_inet_addr *daddr, __be16 dport);
1880
1881struct ip_vs_dest *
1882ip_vs_find_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
1883 const union nf_inet_addr *daddr, __be16 dport);
1884struct ip_vs_dest *ip_vs_find_tunnel(struct netns_ipvs *ipvs, int af,
1885 const union nf_inet_addr *daddr,
1886 __be16 tun_port);
1887
1888int ip_vs_use_count_inc(void);
1889void ip_vs_use_count_dec(void);
1890int ip_vs_register_nl_ioctl(void);
1891void ip_vs_unregister_nl_ioctl(void);
1892int ip_vs_control_init(void);
1893void ip_vs_control_cleanup(void);
1894struct ip_vs_dest *
1895ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
1896 const union nf_inet_addr *daddr, __be16 dport,
1897 const union nf_inet_addr *vaddr, __be16 vport,
1898 __u16 protocol, __u32 fwmark, __u32 flags);
1899void ip_vs_try_bind_dest(struct ip_vs_conn *cp);
1900
1901static inline void ip_vs_dest_hold(struct ip_vs_dest *dest)
1902{
1903 refcount_inc(&dest->refcnt);
1904}
1905
1906static inline void ip_vs_dest_put(struct ip_vs_dest *dest)
1907{
1908 smp_mb__before_atomic();
1909 refcount_dec(&dest->refcnt);
1910}
1911
1912static inline void ip_vs_dest_put_and_free(struct ip_vs_dest *dest)
1913{
1914 if (refcount_dec_and_test(&dest->refcnt))
1915 kfree(dest);
1916}
1917
1918void ip_vs_dest_update_overload(struct ip_vs_dest *dest, int mode);
1919
1920/* IPVS sync daemon data and function prototypes
1921 * (from ip_vs_sync.c)
1922 */
1923int start_sync_thread(struct netns_ipvs *ipvs, struct ipvs_sync_daemon_cfg *cfg,
1924 int state);
1925int stop_sync_thread(struct netns_ipvs *ipvs, int state);
1926void ip_vs_sync_conn(struct netns_ipvs *ipvs, struct ip_vs_conn *cp, int pkts);
1927
1928/* IPVS rate estimator prototypes (from ip_vs_est.c) */
1929int ip_vs_start_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1930void ip_vs_stop_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);
1931void ip_vs_zero_estimator(struct ip_vs_stats *stats);
1932void ip_vs_read_estimator(struct ip_vs_kstats *dst, struct ip_vs_stats *stats);
1933void ip_vs_est_reload_start(struct netns_ipvs *ipvs, bool restart);
1934int ip_vs_est_kthread_start(struct netns_ipvs *ipvs,
1935 struct ip_vs_est_kt_data *kd);
1936void ip_vs_est_kthread_stop(struct ip_vs_est_kt_data *kd);
1937
1938static inline void ip_vs_stop_estimator_tot_stats(struct netns_ipvs *ipvs)
1939{
1940#ifdef CONFIG_SYSCTL
1941 ip_vs_stop_estimator(ipvs, &ipvs->tot_stats->s);
1942 ipvs->tot_stats->s.est.ktid = -2;
1943#endif
1944}
1945
1946static inline void ip_vs_est_stopped_recalc(struct netns_ipvs *ipvs)
1947{
1948#ifdef CONFIG_SYSCTL
1949 /* Stop tasks while cpulist is empty or if disabled with flag */
1950 ipvs->est_stopped = !sysctl_run_estimation(ipvs) ||
1951 (ipvs->est_cpulist_valid &&
1952 sysctl_est_cpulist_empty(ipvs));
1953#endif
1954}
1955
1956static inline bool ip_vs_est_stopped(struct netns_ipvs *ipvs)
1957{
1958#ifdef CONFIG_SYSCTL
1959 return ipvs->est_stopped;
1960#else
1961 return false;
1962#endif
1963}
1964
1965static inline int ip_vs_est_max_threads(struct netns_ipvs *ipvs)
1966{
1967 unsigned int limit = IPVS_EST_CPU_KTHREADS *
1968 sysctl_est_cpulist_weight(ipvs);
1969
1970 return max(1U, limit);
1971}
1972
1973/* Various IPVS packet transmitters (from ip_vs_xmit.c) */
1974int ip_vs_null_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1975 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1976int ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1977 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1978int ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1979 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1980int ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1981 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1982int ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1983 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1984int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
1985 struct ip_vs_protocol *pp, unsigned int toff,
1986 unsigned int hooknum, struct ip_vs_iphdr *ciph);
1987void ip_vs_dest_dst_rcu_free(struct rcu_head *head);
1988
1989#ifdef CONFIG_IP_VS_IPV6
1990int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1991 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1992int ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1993 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1994int ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1995 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1996int ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1997 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph);
1998int ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
1999 struct ip_vs_protocol *pp, unsigned int toff,
2000 unsigned int hooknum, struct ip_vs_iphdr *ciph);
2001#endif
2002
2003#ifdef CONFIG_SYSCTL
2004/* This is a simple mechanism to ignore packets when
2005 * we are loaded. Just set ip_vs_drop_rate to 'n' and
2006 * we start to drop 1/rate of the packets
2007 */
2008static inline int ip_vs_todrop(struct netns_ipvs *ipvs)
2009{
2010 if (!ipvs->drop_rate)
2011 return 0;
2012 if (--ipvs->drop_counter > 0)
2013 return 0;
2014 ipvs->drop_counter = ipvs->drop_rate;
2015 return 1;
2016}
2017#else
2018static inline int ip_vs_todrop(struct netns_ipvs *ipvs) { return 0; }
2019#endif
2020
2021#ifdef CONFIG_SYSCTL
2022/* Enqueue delayed work for expiring no dest connections
2023 * Only run when sysctl_expire_nodest=1
2024 */
2025static inline void ip_vs_enqueue_expire_nodest_conns(struct netns_ipvs *ipvs)
2026{
2027 if (sysctl_expire_nodest_conn(ipvs))
2028 queue_delayed_work(system_long_wq,
2029 &ipvs->expire_nodest_conn_work, 1);
2030}
2031
2032void ip_vs_expire_nodest_conn_flush(struct netns_ipvs *ipvs);
2033#else
2034static inline void ip_vs_enqueue_expire_nodest_conns(struct netns_ipvs *ipvs) {}
2035#endif
2036
2037#define IP_VS_DFWD_METHOD(dest) (atomic_read(&(dest)->conn_flags) & \
2038 IP_VS_CONN_F_FWD_MASK)
2039
2040/* ip_vs_fwd_tag returns the forwarding tag of the connection */
2041#define IP_VS_FWD_METHOD(cp) (cp->flags & IP_VS_CONN_F_FWD_MASK)
2042
2043static inline char ip_vs_fwd_tag(struct ip_vs_conn *cp)
2044{
2045 char fwd;
2046
2047 switch (IP_VS_FWD_METHOD(cp)) {
2048 case IP_VS_CONN_F_MASQ:
2049 fwd = 'M'; break;
2050 case IP_VS_CONN_F_LOCALNODE:
2051 fwd = 'L'; break;
2052 case IP_VS_CONN_F_TUNNEL:
2053 fwd = 'T'; break;
2054 case IP_VS_CONN_F_DROUTE:
2055 fwd = 'R'; break;
2056 case IP_VS_CONN_F_BYPASS:
2057 fwd = 'B'; break;
2058 default:
2059 fwd = '?'; break;
2060 }
2061 return fwd;
2062}
2063
2064/* Check if connection uses double hashing */
2065static inline bool ip_vs_conn_use_hash2(struct ip_vs_conn *cp)
2066{
2067 return IP_VS_FWD_METHOD(cp) == IP_VS_CONN_F_MASQ &&
2068 !(cp->flags & IP_VS_CONN_F_TEMPLATE);
2069}
2070
2071bool ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp,
2072 struct ip_vs_conn *cp, int dir, unsigned int toff,
2073 bool has_ports, struct ip_vs_iphdr *ciph);
2074
2075#ifdef CONFIG_IP_VS_IPV6
2076void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp,
2077 struct ip_vs_conn *cp, int dir, unsigned int toff,
2078 bool has_ports, struct ip_vs_iphdr *ciph);
2079#endif
2080
2081static inline __wsum ip_vs_check_diff4(__be32 old, __be32 new, __wsum oldsum)
2082{
2083 __be32 diff[2] = { ~old, new };
2084
2085 return csum_partial(diff, sizeof(diff), oldsum);
2086}
2087
2088#ifdef CONFIG_IP_VS_IPV6
2089static inline __wsum ip_vs_check_diff16(const __be32 *old, const __be32 *new,
2090 __wsum oldsum)
2091{
2092 __be32 diff[8] = { ~old[3], ~old[2], ~old[1], ~old[0],
2093 new[3], new[2], new[1], new[0] };
2094
2095 return csum_partial(diff, sizeof(diff), oldsum);
2096}
2097#endif
2098
2099static inline __wsum ip_vs_check_diff2(__be16 old, __be16 new, __wsum oldsum)
2100{
2101 __be16 diff[2] = { ~old, new };
2102
2103 return csum_partial(diff, sizeof(diff), oldsum);
2104}
2105
2106static inline bool ip_vs_checksum_needed(struct sk_buff *skb)
2107{
2108 /* Checksum unnecessary or already validated? */
2109 if (skb_csum_unnecessary(skb))
2110 return false;
2111 /* Locally generated ? */
2112 if (!skb->dev)
2113 return false;
2114 return true;
2115}
2116
2117static inline bool ip_vs_checksum_common_check(struct sk_buff *skb,
2118 int offset, int proto, int af)
2119{
2120 if (!ip_vs_checksum_needed(skb))
2121 return true;
2122 /* Validate csum even for FORWARD */
2123 return !nf_checksum(skb, NF_INET_LOCAL_IN, offset, proto, af);
2124}
2125
2126/* Forget current conntrack (unconfirmed) and attach notrack entry */
2127static inline void ip_vs_notrack(struct sk_buff *skb)
2128{
2129#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
2130 enum ip_conntrack_info ctinfo;
2131 struct nf_conn *ct = nf_ct_get(skb, &ctinfo);
2132
2133 if (ct) {
2134 nf_reset_ct(skb);
2135 nf_ct_set(skb, NULL, IP_CT_UNTRACKED);
2136 }
2137#endif
2138}
2139
2140#ifdef CONFIG_IP_VS_NFCT
2141/* Netfilter connection tracking
2142 * (from ip_vs_nfct.c)
2143 */
2144static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
2145{
2146#ifdef CONFIG_SYSCTL
2147 return ipvs->sysctl_conntrack;
2148#else
2149 return 0;
2150#endif
2151}
2152
2153void ip_vs_update_conntrack(struct sk_buff *skb, struct ip_vs_conn *cp,
2154 int outin);
2155int ip_vs_confirm_conntrack(struct sk_buff *skb);
2156void ip_vs_nfct_expect_related(struct sk_buff *skb, struct nf_conn *ct,
2157 struct ip_vs_conn *cp, u8 proto,
2158 const __be16 port, int from_rs);
2159void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp);
2160
2161#else
2162
2163static inline int ip_vs_conntrack_enabled(struct netns_ipvs *ipvs)
2164{
2165 return 0;
2166}
2167
2168static inline void ip_vs_update_conntrack(struct sk_buff *skb,
2169 struct ip_vs_conn *cp, int outin)
2170{
2171}
2172
2173static inline int ip_vs_confirm_conntrack(struct sk_buff *skb)
2174{
2175 return NF_ACCEPT;
2176}
2177
2178static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp)
2179{
2180}
2181#endif /* CONFIG_IP_VS_NFCT */
2182
2183/* Using old conntrack that can not be redirected to another real server? */
2184static inline bool ip_vs_conn_uses_old_conntrack(struct ip_vs_conn *cp,
2185 struct sk_buff *skb)
2186{
2187#ifdef CONFIG_IP_VS_NFCT
2188 enum ip_conntrack_info ctinfo;
2189 struct nf_conn *ct;
2190
2191 ct = nf_ct_get(skb, &ctinfo);
2192 if (ct && nf_ct_is_confirmed(ct))
2193 return true;
2194#endif
2195 return false;
2196}
2197
2198static inline int ip_vs_register_conntrack(struct ip_vs_service *svc)
2199{
2200#if IS_ENABLED(CONFIG_NF_CONNTRACK)
2201 int afmask = (svc->af == AF_INET6) ? 2 : 1;
2202 int ret = 0;
2203
2204 if (!(svc->conntrack_afmask & afmask)) {
2205 ret = nf_ct_netns_get(svc->ipvs->net, svc->af);
2206 if (ret >= 0)
2207 svc->conntrack_afmask |= afmask;
2208 }
2209 return ret;
2210#else
2211 return 0;
2212#endif
2213}
2214
2215static inline void ip_vs_unregister_conntrack(struct ip_vs_service *svc)
2216{
2217#if IS_ENABLED(CONFIG_NF_CONNTRACK)
2218 int afmask = (svc->af == AF_INET6) ? 2 : 1;
2219
2220 if (svc->conntrack_afmask & afmask) {
2221 nf_ct_netns_put(svc->ipvs->net, svc->af);
2222 svc->conntrack_afmask &= ~afmask;
2223 }
2224#endif
2225}
2226
2227int ip_vs_register_hooks(struct netns_ipvs *ipvs, unsigned int af);
2228void ip_vs_unregister_hooks(struct netns_ipvs *ipvs, unsigned int af);
2229
2230static inline int
2231ip_vs_dest_conn_overhead(struct ip_vs_dest *dest)
2232{
2233 /* We think the overhead of processing active connections is 257
2234 * times higher than that of inactive connections in average. (This
2235 * 257 times might not be accurate, we will change it later) We
2236 * use the following formula to estimate the overhead now:
2237 * dest->activeconns*256 + dest->totalconns
2238 */
2239 return (atomic_read(&dest->activeconns) << 8) +
2240 atomic_read(&dest->totalconns);
2241}
2242
2243static inline int
2244ip_vs_dest_inactconns(const struct ip_vs_dest *dest)
2245{
2246 return max(atomic_read(&dest->totalconns) -
2247 atomic_read(&dest->activeconns), 0);
2248}
2249
2250#ifdef CONFIG_IP_VS_PROTO_TCP
2251INDIRECT_CALLABLE_DECLARE(int
2252 tcp_snat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
2253 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph));
2254#endif
2255
2256#ifdef CONFIG_IP_VS_PROTO_UDP
2257INDIRECT_CALLABLE_DECLARE(int
2258 udp_snat_handler(struct sk_buff *skb, struct ip_vs_protocol *pp,
2259 struct ip_vs_conn *cp, struct ip_vs_iphdr *iph));
2260#endif
2261#endif /* _NET_IP_VS_H */