2 * IPVS An implementation of the IP virtual server support for the
3 * LINUX operating system. IPVS is now implemented as a module
4 * over the Netfilter framework. IPVS can be used to build a
5 * high-performance and highly available server based on a
8 * Version: $Id: ip_vs_conn.c,v 1.31 2003/04/18 09:03:16 wensong Exp $
10 * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
11 * Peter Kese <peter.kese@ijs.si>
12 * Julian Anastasov <ja@ssi.bg>
14 * This program is free software; you can redistribute it and/or
15 * modify it under the terms of the GNU General Public License
16 * as published by the Free Software Foundation; either version
17 * 2 of the License, or (at your option) any later version.
19 * The IPVS code for kernel 2.2 was done by Wensong Zhang and Peter Kese,
20 * with changes/fixes from Julian Anastasov, Lars Marowsky-Bree, Horms
21 * and others. Many code here is taken from IP MASQ code of kernel 2.2.
27 #include <linux/kernel.h>
28 #include <linux/vmalloc.h>
29 #include <linux/proc_fs.h> /* for proc_net_* */
30 #include <linux/seq_file.h>
31 #include <linux/jhash.h>
32 #include <linux/random.h>
34 #include <net/ip_vs.h>
38 * Connection hash table: for input and output packets lookups of IPVS
40 static struct list_head
*ip_vs_conn_tab
;
42 /* SLAB cache for IPVS connections */
43 static kmem_cache_t
*ip_vs_conn_cachep
;
45 /* counter for current IPVS connections */
46 static atomic_t ip_vs_conn_count
= ATOMIC_INIT(0);
48 /* counter for no client port connections */
49 static atomic_t ip_vs_conn_no_cport_cnt
= ATOMIC_INIT(0);
51 /* random value for IPVS connection hash */
52 static unsigned int ip_vs_conn_rnd
;
55 * Fine locking granularity for big connection hash table
57 #define CT_LOCKARRAY_BITS 4
58 #define CT_LOCKARRAY_SIZE (1<<CT_LOCKARRAY_BITS)
59 #define CT_LOCKARRAY_MASK (CT_LOCKARRAY_SIZE-1)
61 struct ip_vs_aligned_lock
64 } __attribute__((__aligned__(SMP_CACHE_BYTES
)));
66 /* lock array for conn table */
67 static struct ip_vs_aligned_lock
68 __ip_vs_conntbl_lock_array
[CT_LOCKARRAY_SIZE
] __cacheline_aligned
;
70 static inline void ct_read_lock(unsigned key
)
72 read_lock(&__ip_vs_conntbl_lock_array
[key
&CT_LOCKARRAY_MASK
].l
);
75 static inline void ct_read_unlock(unsigned key
)
77 read_unlock(&__ip_vs_conntbl_lock_array
[key
&CT_LOCKARRAY_MASK
].l
);
80 static inline void ct_write_lock(unsigned key
)
82 write_lock(&__ip_vs_conntbl_lock_array
[key
&CT_LOCKARRAY_MASK
].l
);
85 static inline void ct_write_unlock(unsigned key
)
87 write_unlock(&__ip_vs_conntbl_lock_array
[key
&CT_LOCKARRAY_MASK
].l
);
90 static inline void ct_read_lock_bh(unsigned key
)
92 read_lock_bh(&__ip_vs_conntbl_lock_array
[key
&CT_LOCKARRAY_MASK
].l
);
95 static inline void ct_read_unlock_bh(unsigned key
)
97 read_unlock_bh(&__ip_vs_conntbl_lock_array
[key
&CT_LOCKARRAY_MASK
].l
);
100 static inline void ct_write_lock_bh(unsigned key
)
102 write_lock_bh(&__ip_vs_conntbl_lock_array
[key
&CT_LOCKARRAY_MASK
].l
);
105 static inline void ct_write_unlock_bh(unsigned key
)
107 write_unlock_bh(&__ip_vs_conntbl_lock_array
[key
&CT_LOCKARRAY_MASK
].l
);
112 * Returns hash value for IPVS connection entry
114 static unsigned int ip_vs_conn_hashkey(unsigned proto
, __u32 addr
, __u16 port
)
116 return jhash_3words(addr
, port
, proto
, ip_vs_conn_rnd
)
117 & IP_VS_CONN_TAB_MASK
;
122 * Hashes ip_vs_conn in ip_vs_conn_tab by proto,addr,port.
123 * returns bool success.
125 static inline int ip_vs_conn_hash(struct ip_vs_conn
*cp
)
130 /* Hash by protocol, client address and port */
131 hash
= ip_vs_conn_hashkey(cp
->protocol
, cp
->caddr
, cp
->cport
);
135 if (!(cp
->flags
& IP_VS_CONN_F_HASHED
)) {
136 list_add(&cp
->c_list
, &ip_vs_conn_tab
[hash
]);
137 cp
->flags
|= IP_VS_CONN_F_HASHED
;
138 atomic_inc(&cp
->refcnt
);
141 IP_VS_ERR("ip_vs_conn_hash(): request for already hashed, "
142 "called from %p\n", __builtin_return_address(0));
146 ct_write_unlock(hash
);
153 * UNhashes ip_vs_conn from ip_vs_conn_tab.
154 * returns bool success.
156 static inline int ip_vs_conn_unhash(struct ip_vs_conn
*cp
)
161 /* unhash it and decrease its reference counter */
162 hash
= ip_vs_conn_hashkey(cp
->protocol
, cp
->caddr
, cp
->cport
);
166 if (cp
->flags
& IP_VS_CONN_F_HASHED
) {
167 list_del(&cp
->c_list
);
168 cp
->flags
&= ~IP_VS_CONN_F_HASHED
;
169 atomic_dec(&cp
->refcnt
);
174 ct_write_unlock(hash
);
181 * Gets ip_vs_conn associated with supplied parameters in the ip_vs_conn_tab.
182 * Called for pkts coming from OUTside-to-INside.
183 * s_addr, s_port: pkt source address (foreign host)
184 * d_addr, d_port: pkt dest address (load balancer)
186 static inline struct ip_vs_conn
*__ip_vs_conn_in_get
187 (int protocol
, __u32 s_addr
, __u16 s_port
, __u32 d_addr
, __u16 d_port
)
190 struct ip_vs_conn
*cp
;
192 hash
= ip_vs_conn_hashkey(protocol
, s_addr
, s_port
);
196 list_for_each_entry(cp
, &ip_vs_conn_tab
[hash
], c_list
) {
197 if (s_addr
==cp
->caddr
&& s_port
==cp
->cport
&&
198 d_port
==cp
->vport
&& d_addr
==cp
->vaddr
&&
199 protocol
==cp
->protocol
) {
201 atomic_inc(&cp
->refcnt
);
202 ct_read_unlock(hash
);
207 ct_read_unlock(hash
);
212 struct ip_vs_conn
*ip_vs_conn_in_get
213 (int protocol
, __u32 s_addr
, __u16 s_port
, __u32 d_addr
, __u16 d_port
)
215 struct ip_vs_conn
*cp
;
217 cp
= __ip_vs_conn_in_get(protocol
, s_addr
, s_port
, d_addr
, d_port
);
218 if (!cp
&& atomic_read(&ip_vs_conn_no_cport_cnt
))
219 cp
= __ip_vs_conn_in_get(protocol
, s_addr
, 0, d_addr
, d_port
);
221 IP_VS_DBG(7, "lookup/in %s %u.%u.%u.%u:%d->%u.%u.%u.%u:%d %s\n",
222 ip_vs_proto_name(protocol
),
223 NIPQUAD(s_addr
), ntohs(s_port
),
224 NIPQUAD(d_addr
), ntohs(d_port
),
232 * Gets ip_vs_conn associated with supplied parameters in the ip_vs_conn_tab.
233 * Called for pkts coming from inside-to-OUTside.
234 * s_addr, s_port: pkt source address (inside host)
235 * d_addr, d_port: pkt dest address (foreign host)
237 struct ip_vs_conn
*ip_vs_conn_out_get
238 (int protocol
, __u32 s_addr
, __u16 s_port
, __u32 d_addr
, __u16 d_port
)
241 struct ip_vs_conn
*cp
, *ret
=NULL
;
244 * Check for "full" addressed entries
246 hash
= ip_vs_conn_hashkey(protocol
, d_addr
, d_port
);
250 list_for_each_entry(cp
, &ip_vs_conn_tab
[hash
], c_list
) {
251 if (d_addr
== cp
->caddr
&& d_port
== cp
->cport
&&
252 s_port
== cp
->dport
&& s_addr
== cp
->daddr
&&
253 protocol
== cp
->protocol
) {
255 atomic_inc(&cp
->refcnt
);
261 ct_read_unlock(hash
);
263 IP_VS_DBG(7, "lookup/out %s %u.%u.%u.%u:%d->%u.%u.%u.%u:%d %s\n",
264 ip_vs_proto_name(protocol
),
265 NIPQUAD(s_addr
), ntohs(s_port
),
266 NIPQUAD(d_addr
), ntohs(d_port
),
267 ret
?"hit":"not hit");
274 * Put back the conn and restart its timer with its timeout
276 void ip_vs_conn_put(struct ip_vs_conn
*cp
)
278 /* reset it expire in its timeout */
279 mod_timer(&cp
->timer
, jiffies
+cp
->timeout
);
281 __ip_vs_conn_put(cp
);
286 * Fill a no_client_port connection with a client port number
288 void ip_vs_conn_fill_cport(struct ip_vs_conn
*cp
, __u16 cport
)
290 if (ip_vs_conn_unhash(cp
)) {
291 spin_lock(&cp
->lock
);
292 if (cp
->flags
& IP_VS_CONN_F_NO_CPORT
) {
293 atomic_dec(&ip_vs_conn_no_cport_cnt
);
294 cp
->flags
&= ~IP_VS_CONN_F_NO_CPORT
;
297 spin_unlock(&cp
->lock
);
299 /* hash on new dport */
306 * Bind a connection entry with the corresponding packet_xmit.
307 * Called by ip_vs_conn_new.
309 static inline void ip_vs_bind_xmit(struct ip_vs_conn
*cp
)
311 switch (IP_VS_FWD_METHOD(cp
)) {
312 case IP_VS_CONN_F_MASQ
:
313 cp
->packet_xmit
= ip_vs_nat_xmit
;
316 case IP_VS_CONN_F_TUNNEL
:
317 cp
->packet_xmit
= ip_vs_tunnel_xmit
;
320 case IP_VS_CONN_F_DROUTE
:
321 cp
->packet_xmit
= ip_vs_dr_xmit
;
324 case IP_VS_CONN_F_LOCALNODE
:
325 cp
->packet_xmit
= ip_vs_null_xmit
;
328 case IP_VS_CONN_F_BYPASS
:
329 cp
->packet_xmit
= ip_vs_bypass_xmit
;
335 static inline int ip_vs_dest_totalconns(struct ip_vs_dest
*dest
)
337 return atomic_read(&dest
->activeconns
)
338 + atomic_read(&dest
->inactconns
);
342 * Bind a connection entry with a virtual service destination
343 * Called just after a new connection entry is created.
346 ip_vs_bind_dest(struct ip_vs_conn
*cp
, struct ip_vs_dest
*dest
)
348 /* if dest is NULL, then return directly */
352 /* Increase the refcnt counter of the dest */
353 atomic_inc(&dest
->refcnt
);
355 /* Bind with the destination and its corresponding transmitter */
356 cp
->flags
|= atomic_read(&dest
->conn_flags
);
359 IP_VS_DBG(9, "Bind-dest %s c:%u.%u.%u.%u:%d v:%u.%u.%u.%u:%d "
360 "d:%u.%u.%u.%u:%d fwd:%c s:%u flg:%X cnt:%d destcnt:%d\n",
361 ip_vs_proto_name(cp
->protocol
),
362 NIPQUAD(cp
->caddr
), ntohs(cp
->cport
),
363 NIPQUAD(cp
->vaddr
), ntohs(cp
->vport
),
364 NIPQUAD(cp
->daddr
), ntohs(cp
->dport
),
365 ip_vs_fwd_tag(cp
), cp
->state
,
366 cp
->flags
, atomic_read(&cp
->refcnt
),
367 atomic_read(&dest
->refcnt
));
369 /* Update the connection counters */
370 if (cp
->cport
|| (cp
->flags
& IP_VS_CONN_F_NO_CPORT
)) {
371 /* It is a normal connection, so increase the inactive
372 connection counter because it is in TCP SYNRECV
373 state (inactive) or other protocol inacive state */
374 atomic_inc(&dest
->inactconns
);
376 /* It is a persistent connection/template, so increase
377 the peristent connection counter */
378 atomic_inc(&dest
->persistconns
);
381 if (dest
->u_threshold
!= 0 &&
382 ip_vs_dest_totalconns(dest
) >= dest
->u_threshold
)
383 dest
->flags
|= IP_VS_DEST_F_OVERLOAD
;
388 * Unbind a connection entry with its VS destination
389 * Called by the ip_vs_conn_expire function.
391 static inline void ip_vs_unbind_dest(struct ip_vs_conn
*cp
)
393 struct ip_vs_dest
*dest
= cp
->dest
;
398 IP_VS_DBG(9, "Unbind-dest %s c:%u.%u.%u.%u:%d v:%u.%u.%u.%u:%d "
399 "d:%u.%u.%u.%u:%d fwd:%c s:%u flg:%X cnt:%d destcnt:%d\n",
400 ip_vs_proto_name(cp
->protocol
),
401 NIPQUAD(cp
->caddr
), ntohs(cp
->cport
),
402 NIPQUAD(cp
->vaddr
), ntohs(cp
->vport
),
403 NIPQUAD(cp
->daddr
), ntohs(cp
->dport
),
404 ip_vs_fwd_tag(cp
), cp
->state
,
405 cp
->flags
, atomic_read(&cp
->refcnt
),
406 atomic_read(&dest
->refcnt
));
408 /* Update the connection counters */
409 if (cp
->cport
|| (cp
->flags
& IP_VS_CONN_F_NO_CPORT
)) {
410 /* It is a normal connection, so decrease the inactconns
411 or activeconns counter */
412 if (cp
->flags
& IP_VS_CONN_F_INACTIVE
) {
413 atomic_dec(&dest
->inactconns
);
415 atomic_dec(&dest
->activeconns
);
418 /* It is a persistent connection/template, so decrease
419 the peristent connection counter */
420 atomic_dec(&dest
->persistconns
);
423 if (dest
->l_threshold
!= 0) {
424 if (ip_vs_dest_totalconns(dest
) < dest
->l_threshold
)
425 dest
->flags
&= ~IP_VS_DEST_F_OVERLOAD
;
426 } else if (dest
->u_threshold
!= 0) {
427 if (ip_vs_dest_totalconns(dest
) * 4 < dest
->u_threshold
* 3)
428 dest
->flags
&= ~IP_VS_DEST_F_OVERLOAD
;
430 if (dest
->flags
& IP_VS_DEST_F_OVERLOAD
)
431 dest
->flags
&= ~IP_VS_DEST_F_OVERLOAD
;
435 * Simply decrease the refcnt of the dest, because the
436 * dest will be either in service's destination list
439 atomic_dec(&dest
->refcnt
);
444 * Checking if the destination of a connection template is available.
445 * If available, return 1, otherwise invalidate this connection
446 * template and return 0.
448 int ip_vs_check_template(struct ip_vs_conn
*ct
)
450 struct ip_vs_dest
*dest
= ct
->dest
;
453 * Checking the dest server status.
455 if ((dest
== NULL
) ||
456 !(dest
->flags
& IP_VS_DEST_F_AVAILABLE
) ||
457 (sysctl_ip_vs_expire_quiescent_template
&&
458 (atomic_read(&dest
->weight
) == 0))) {
459 IP_VS_DBG(9, "check_template: dest not available for "
460 "protocol %s s:%u.%u.%u.%u:%d v:%u.%u.%u.%u:%d "
461 "-> d:%u.%u.%u.%u:%d\n",
462 ip_vs_proto_name(ct
->protocol
),
463 NIPQUAD(ct
->caddr
), ntohs(ct
->cport
),
464 NIPQUAD(ct
->vaddr
), ntohs(ct
->vport
),
465 NIPQUAD(ct
->daddr
), ntohs(ct
->dport
));
468 * Invalidate the connection template
471 if (ip_vs_conn_unhash(ct
)) {
480 * Simply decrease the refcnt of the template,
481 * don't restart its timer.
483 atomic_dec(&ct
->refcnt
);
489 static void ip_vs_conn_expire(unsigned long data
)
491 struct ip_vs_conn
*cp
= (struct ip_vs_conn
*)data
;
498 atomic_inc(&cp
->refcnt
);
501 * do I control anybody?
503 if (atomic_read(&cp
->n_control
))
507 * unhash it if it is hashed in the conn table
509 if (!ip_vs_conn_unhash(cp
))
513 * refcnt==1 implies I'm the only one referrer
515 if (likely(atomic_read(&cp
->refcnt
) == 1)) {
516 /* delete the timer if it is activated by other users */
517 if (timer_pending(&cp
->timer
))
518 del_timer(&cp
->timer
);
520 /* does anybody control me? */
522 ip_vs_control_del(cp
);
524 if (unlikely(cp
->app
!= NULL
))
525 ip_vs_unbind_app(cp
);
526 ip_vs_unbind_dest(cp
);
527 if (cp
->flags
& IP_VS_CONN_F_NO_CPORT
)
528 atomic_dec(&ip_vs_conn_no_cport_cnt
);
529 atomic_dec(&ip_vs_conn_count
);
531 kmem_cache_free(ip_vs_conn_cachep
, cp
);
535 /* hash it back to the table */
539 IP_VS_DBG(7, "delayed: refcnt-1=%d conn.n_control=%d\n",
540 atomic_read(&cp
->refcnt
)-1,
541 atomic_read(&cp
->n_control
));
547 void ip_vs_conn_expire_now(struct ip_vs_conn
*cp
)
549 if (del_timer(&cp
->timer
))
550 mod_timer(&cp
->timer
, jiffies
);
555 * Create a new connection entry and hash it into the ip_vs_conn_tab
558 ip_vs_conn_new(int proto
, __u32 caddr
, __u16 cport
, __u32 vaddr
, __u16 vport
,
559 __u32 daddr
, __u16 dport
, unsigned flags
,
560 struct ip_vs_dest
*dest
)
562 struct ip_vs_conn
*cp
;
563 struct ip_vs_protocol
*pp
= ip_vs_proto_get(proto
);
565 cp
= kmem_cache_alloc(ip_vs_conn_cachep
, GFP_ATOMIC
);
567 IP_VS_ERR_RL("ip_vs_conn_new: no memory available.\n");
571 memset(cp
, 0, sizeof(*cp
));
572 INIT_LIST_HEAD(&cp
->c_list
);
573 init_timer(&cp
->timer
);
574 cp
->timer
.data
= (unsigned long)cp
;
575 cp
->timer
.function
= ip_vs_conn_expire
;
576 cp
->protocol
= proto
;
584 spin_lock_init(&cp
->lock
);
587 * Set the entry is referenced by the current thread before hashing
588 * it in the table, so that other thread run ip_vs_random_dropentry
589 * but cannot drop this entry.
591 atomic_set(&cp
->refcnt
, 1);
593 atomic_set(&cp
->n_control
, 0);
594 atomic_set(&cp
->in_pkts
, 0);
596 atomic_inc(&ip_vs_conn_count
);
597 if (flags
& IP_VS_CONN_F_NO_CPORT
)
598 atomic_inc(&ip_vs_conn_no_cport_cnt
);
600 /* Bind the connection with a destination server */
601 ip_vs_bind_dest(cp
, dest
);
603 /* Set its state and timeout */
607 /* Bind its packet transmitter */
610 if (unlikely(pp
&& atomic_read(&pp
->appcnt
)))
611 ip_vs_bind_app(cp
, pp
);
613 /* Hash it in the ip_vs_conn_tab finally */
621 * /proc/net/ip_vs_conn entries
623 #ifdef CONFIG_PROC_FS
625 static void *ip_vs_conn_array(struct seq_file
*seq
, loff_t pos
)
628 struct ip_vs_conn
*cp
;
630 for(idx
= 0; idx
< IP_VS_CONN_TAB_SIZE
; idx
++) {
631 ct_read_lock_bh(idx
);
632 list_for_each_entry(cp
, &ip_vs_conn_tab
[idx
], c_list
) {
634 seq
->private = &ip_vs_conn_tab
[idx
];
638 ct_read_unlock_bh(idx
);
644 static void *ip_vs_conn_seq_start(struct seq_file
*seq
, loff_t
*pos
)
647 return *pos
? ip_vs_conn_array(seq
, *pos
- 1) :SEQ_START_TOKEN
;
650 static void *ip_vs_conn_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
652 struct ip_vs_conn
*cp
= v
;
653 struct list_head
*e
, *l
= seq
->private;
657 if (v
== SEQ_START_TOKEN
)
658 return ip_vs_conn_array(seq
, 0);
660 /* more on same hash chain? */
661 if ((e
= cp
->c_list
.next
) != l
)
662 return list_entry(e
, struct ip_vs_conn
, c_list
);
664 idx
= l
- ip_vs_conn_tab
;
665 ct_read_unlock_bh(idx
);
667 while (++idx
< IP_VS_CONN_TAB_SIZE
) {
668 ct_read_lock_bh(idx
);
669 list_for_each_entry(cp
, &ip_vs_conn_tab
[idx
], c_list
) {
670 seq
->private = &ip_vs_conn_tab
[idx
];
673 ct_read_unlock_bh(idx
);
679 static void ip_vs_conn_seq_stop(struct seq_file
*seq
, void *v
)
681 struct list_head
*l
= seq
->private;
684 ct_read_unlock_bh(l
- ip_vs_conn_tab
);
687 static int ip_vs_conn_seq_show(struct seq_file
*seq
, void *v
)
690 if (v
== SEQ_START_TOKEN
)
692 "Pro FromIP FPrt ToIP TPrt DestIP DPrt State Expires\n");
694 const struct ip_vs_conn
*cp
= v
;
697 "%-3s %08X %04X %08X %04X %08X %04X %-11s %7lu\n",
698 ip_vs_proto_name(cp
->protocol
),
699 ntohl(cp
->caddr
), ntohs(cp
->cport
),
700 ntohl(cp
->vaddr
), ntohs(cp
->vport
),
701 ntohl(cp
->daddr
), ntohs(cp
->dport
),
702 ip_vs_state_name(cp
->protocol
, cp
->state
),
703 (cp
->timer
.expires
-jiffies
)/HZ
);
708 static struct seq_operations ip_vs_conn_seq_ops
= {
709 .start
= ip_vs_conn_seq_start
,
710 .next
= ip_vs_conn_seq_next
,
711 .stop
= ip_vs_conn_seq_stop
,
712 .show
= ip_vs_conn_seq_show
,
715 static int ip_vs_conn_open(struct inode
*inode
, struct file
*file
)
717 return seq_open(file
, &ip_vs_conn_seq_ops
);
720 static struct file_operations ip_vs_conn_fops
= {
721 .owner
= THIS_MODULE
,
722 .open
= ip_vs_conn_open
,
725 .release
= seq_release
,
731 * Randomly drop connection entries before running out of memory
733 static inline int todrop_entry(struct ip_vs_conn
*cp
)
736 * The drop rate array needs tuning for real environments.
737 * Called from timer bh only => no locking
739 static char todrop_rate
[9] = {0, 1, 2, 3, 4, 5, 6, 7, 8};
740 static char todrop_counter
[9] = {0};
743 /* if the conn entry hasn't lasted for 60 seconds, don't drop it.
744 This will leave enough time for normal connection to get
746 if (time_before(cp
->timeout
+ jiffies
, cp
->timer
.expires
+ 60*HZ
))
749 /* Don't drop the entry if its number of incoming packets is not
751 i
= atomic_read(&cp
->in_pkts
);
752 if (i
> 8 || i
< 0) return 0;
754 if (!todrop_rate
[i
]) return 0;
755 if (--todrop_counter
[i
] > 0) return 0;
757 todrop_counter
[i
] = todrop_rate
[i
];
761 /* Called from keventd and must protect itself from softirqs */
762 void ip_vs_random_dropentry(void)
765 struct ip_vs_conn
*cp
;
768 * Randomly scan 1/32 of the whole table every second
770 for (idx
= 0; idx
< (IP_VS_CONN_TAB_SIZE
>>5); idx
++) {
771 unsigned hash
= net_random() & IP_VS_CONN_TAB_MASK
;
774 * Lock is actually needed in this loop.
776 ct_write_lock_bh(hash
);
778 list_for_each_entry(cp
, &ip_vs_conn_tab
[hash
], c_list
) {
779 if (!cp
->cport
&& !(cp
->flags
& IP_VS_CONN_F_NO_CPORT
))
780 /* connection template */
783 if (cp
->protocol
== IPPROTO_TCP
) {
785 case IP_VS_TCP_S_SYN_RECV
:
786 case IP_VS_TCP_S_SYNACK
:
789 case IP_VS_TCP_S_ESTABLISHED
:
790 if (todrop_entry(cp
))
798 if (!todrop_entry(cp
))
802 IP_VS_DBG(4, "del connection\n");
803 ip_vs_conn_expire_now(cp
);
805 IP_VS_DBG(4, "del conn template\n");
806 ip_vs_conn_expire_now(cp
->control
);
809 ct_write_unlock_bh(hash
);
815 * Flush all the connection entries in the ip_vs_conn_tab
817 static void ip_vs_conn_flush(void)
820 struct ip_vs_conn
*cp
;
823 for (idx
=0; idx
<IP_VS_CONN_TAB_SIZE
; idx
++) {
825 * Lock is actually needed in this loop.
827 ct_write_lock_bh(idx
);
829 list_for_each_entry(cp
, &ip_vs_conn_tab
[idx
], c_list
) {
831 IP_VS_DBG(4, "del connection\n");
832 ip_vs_conn_expire_now(cp
);
834 IP_VS_DBG(4, "del conn template\n");
835 ip_vs_conn_expire_now(cp
->control
);
838 ct_write_unlock_bh(idx
);
841 /* the counter may be not NULL, because maybe some conn entries
842 are run by slow timer handler or unhashed but still referred */
843 if (atomic_read(&ip_vs_conn_count
) != 0) {
850 int ip_vs_conn_init(void)
855 * Allocate the connection hash table and initialize its list heads
857 ip_vs_conn_tab
= vmalloc(IP_VS_CONN_TAB_SIZE
*sizeof(struct list_head
));
861 /* Allocate ip_vs_conn slab cache */
862 ip_vs_conn_cachep
= kmem_cache_create("ip_vs_conn",
863 sizeof(struct ip_vs_conn
), 0,
864 SLAB_HWCACHE_ALIGN
, NULL
, NULL
);
865 if (!ip_vs_conn_cachep
) {
866 vfree(ip_vs_conn_tab
);
870 IP_VS_INFO("Connection hash table configured "
871 "(size=%d, memory=%ldKbytes)\n",
873 (long)(IP_VS_CONN_TAB_SIZE
*sizeof(struct list_head
))/1024);
874 IP_VS_DBG(0, "Each connection entry needs %Zd bytes at least\n",
875 sizeof(struct ip_vs_conn
));
877 for (idx
= 0; idx
< IP_VS_CONN_TAB_SIZE
; idx
++) {
878 INIT_LIST_HEAD(&ip_vs_conn_tab
[idx
]);
881 for (idx
= 0; idx
< CT_LOCKARRAY_SIZE
; idx
++) {
882 rwlock_init(&__ip_vs_conntbl_lock_array
[idx
].l
);
885 proc_net_fops_create("ip_vs_conn", 0, &ip_vs_conn_fops
);
887 /* calculate the random value for connection hash */
888 get_random_bytes(&ip_vs_conn_rnd
, sizeof(ip_vs_conn_rnd
));
894 void ip_vs_conn_cleanup(void)
896 /* flush all the connection entries first */
899 /* Release the empty cache */
900 kmem_cache_destroy(ip_vs_conn_cachep
);
901 proc_net_remove("ip_vs_conn");
902 vfree(ip_vs_conn_tab
);