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/interrupt.h>
29 #include <linux/net.h>
30 #include <linux/kernel.h>
31 #include <linux/module.h>
32 #include <linux/vmalloc.h>
33 #include <linux/proc_fs.h> /* for proc_net_* */
34 #include <linux/seq_file.h>
35 #include <linux/jhash.h>
36 #include <linux/random.h>
38 #include <net/net_namespace.h>
39 #include <net/ip_vs.h>
43 * Connection hash table: for input and output packets lookups of IPVS
45 static struct list_head *ip_vs_conn_tab;
47 /* SLAB cache for IPVS connections */
48 static struct kmem_cache *ip_vs_conn_cachep __read_mostly;
50 /* counter for current IPVS connections */
51 static atomic_t ip_vs_conn_count = ATOMIC_INIT(0);
53 /* counter for no client port connections */
54 static atomic_t ip_vs_conn_no_cport_cnt = ATOMIC_INIT(0);
56 /* random value for IPVS connection hash */
57 static unsigned int ip_vs_conn_rnd;
60 * Fine locking granularity for big connection hash table
62 #define CT_LOCKARRAY_BITS 4
63 #define CT_LOCKARRAY_SIZE (1<<CT_LOCKARRAY_BITS)
64 #define CT_LOCKARRAY_MASK (CT_LOCKARRAY_SIZE-1)
66 struct ip_vs_aligned_lock
69 } __attribute__((__aligned__(SMP_CACHE_BYTES)));
71 /* lock array for conn table */
72 static struct ip_vs_aligned_lock
73 __ip_vs_conntbl_lock_array[CT_LOCKARRAY_SIZE] __cacheline_aligned;
75 static inline void ct_read_lock(unsigned key)
77 read_lock(&__ip_vs_conntbl_lock_array[key&CT_LOCKARRAY_MASK].l);
80 static inline void ct_read_unlock(unsigned key)
82 read_unlock(&__ip_vs_conntbl_lock_array[key&CT_LOCKARRAY_MASK].l);
85 static inline void ct_write_lock(unsigned key)
87 write_lock(&__ip_vs_conntbl_lock_array[key&CT_LOCKARRAY_MASK].l);
90 static inline void ct_write_unlock(unsigned key)
92 write_unlock(&__ip_vs_conntbl_lock_array[key&CT_LOCKARRAY_MASK].l);
95 static inline void ct_read_lock_bh(unsigned key)
97 read_lock_bh(&__ip_vs_conntbl_lock_array[key&CT_LOCKARRAY_MASK].l);
100 static inline void ct_read_unlock_bh(unsigned key)
102 read_unlock_bh(&__ip_vs_conntbl_lock_array[key&CT_LOCKARRAY_MASK].l);
105 static inline void ct_write_lock_bh(unsigned key)
107 write_lock_bh(&__ip_vs_conntbl_lock_array[key&CT_LOCKARRAY_MASK].l);
110 static inline void ct_write_unlock_bh(unsigned key)
112 write_unlock_bh(&__ip_vs_conntbl_lock_array[key&CT_LOCKARRAY_MASK].l);
117 * Returns hash value for IPVS connection entry
119 static unsigned int ip_vs_conn_hashkey(unsigned proto, __be32 addr, __be16 port)
121 return jhash_3words((__force u32)addr, (__force u32)port, proto, ip_vs_conn_rnd)
122 & IP_VS_CONN_TAB_MASK;
127 * Hashes ip_vs_conn in ip_vs_conn_tab by proto,addr,port.
128 * returns bool success.
130 static inline int ip_vs_conn_hash(struct ip_vs_conn *cp)
135 /* Hash by protocol, client address and port */
136 hash = ip_vs_conn_hashkey(cp->protocol, cp->caddr, cp->cport);
140 if (!(cp->flags & IP_VS_CONN_F_HASHED)) {
141 list_add(&cp->c_list, &ip_vs_conn_tab[hash]);
142 cp->flags |= IP_VS_CONN_F_HASHED;
143 atomic_inc(&cp->refcnt);
146 IP_VS_ERR("ip_vs_conn_hash(): request for already hashed, "
147 "called from %p\n", __builtin_return_address(0));
151 ct_write_unlock(hash);
158 * UNhashes ip_vs_conn from ip_vs_conn_tab.
159 * returns bool success.
161 static inline int ip_vs_conn_unhash(struct ip_vs_conn *cp)
166 /* unhash it and decrease its reference counter */
167 hash = ip_vs_conn_hashkey(cp->protocol, cp->caddr, cp->cport);
171 if (cp->flags & IP_VS_CONN_F_HASHED) {
172 list_del(&cp->c_list);
173 cp->flags &= ~IP_VS_CONN_F_HASHED;
174 atomic_dec(&cp->refcnt);
179 ct_write_unlock(hash);
186 * Gets ip_vs_conn associated with supplied parameters in the ip_vs_conn_tab.
187 * Called for pkts coming from OUTside-to-INside.
188 * s_addr, s_port: pkt source address (foreign host)
189 * d_addr, d_port: pkt dest address (load balancer)
191 static inline struct ip_vs_conn *__ip_vs_conn_in_get
192 (int protocol, __be32 s_addr, __be16 s_port, __be32 d_addr, __be16 d_port)
195 struct ip_vs_conn *cp;
197 hash = ip_vs_conn_hashkey(protocol, s_addr, s_port);
201 list_for_each_entry(cp, &ip_vs_conn_tab[hash], c_list) {
202 if (s_addr==cp->caddr && s_port==cp->cport &&
203 d_port==cp->vport && d_addr==cp->vaddr &&
204 ((!s_port) ^ (!(cp->flags & IP_VS_CONN_F_NO_CPORT))) &&
205 protocol==cp->protocol) {
207 atomic_inc(&cp->refcnt);
208 ct_read_unlock(hash);
213 ct_read_unlock(hash);
218 struct ip_vs_conn *ip_vs_conn_in_get
219 (int protocol, __be32 s_addr, __be16 s_port, __be32 d_addr, __be16 d_port)
221 struct ip_vs_conn *cp;
223 cp = __ip_vs_conn_in_get(protocol, s_addr, s_port, d_addr, d_port);
224 if (!cp && atomic_read(&ip_vs_conn_no_cport_cnt))
225 cp = __ip_vs_conn_in_get(protocol, s_addr, 0, d_addr, d_port);
227 IP_VS_DBG(9, "lookup/in %s %u.%u.%u.%u:%d->%u.%u.%u.%u:%d %s\n",
228 ip_vs_proto_name(protocol),
229 NIPQUAD(s_addr), ntohs(s_port),
230 NIPQUAD(d_addr), ntohs(d_port),
236 /* Get reference to connection template */
237 struct ip_vs_conn *ip_vs_ct_in_get
238 (int protocol, __be32 s_addr, __be16 s_port, __be32 d_addr, __be16 d_port)
241 struct ip_vs_conn *cp;
243 hash = ip_vs_conn_hashkey(protocol, s_addr, s_port);
247 list_for_each_entry(cp, &ip_vs_conn_tab[hash], c_list) {
248 if (s_addr==cp->caddr && s_port==cp->cport &&
249 d_port==cp->vport && d_addr==cp->vaddr &&
250 cp->flags & IP_VS_CONN_F_TEMPLATE &&
251 protocol==cp->protocol) {
253 atomic_inc(&cp->refcnt);
260 ct_read_unlock(hash);
262 IP_VS_DBG(9, "template lookup/in %s %u.%u.%u.%u:%d->%u.%u.%u.%u:%d %s\n",
263 ip_vs_proto_name(protocol),
264 NIPQUAD(s_addr), ntohs(s_port),
265 NIPQUAD(d_addr), ntohs(d_port),
272 * Gets ip_vs_conn associated with supplied parameters in the ip_vs_conn_tab.
273 * Called for pkts coming from inside-to-OUTside.
274 * s_addr, s_port: pkt source address (inside host)
275 * d_addr, d_port: pkt dest address (foreign host)
277 struct ip_vs_conn *ip_vs_conn_out_get
278 (int protocol, __be32 s_addr, __be16 s_port, __be32 d_addr, __be16 d_port)
281 struct ip_vs_conn *cp, *ret=NULL;
284 * Check for "full" addressed entries
286 hash = ip_vs_conn_hashkey(protocol, d_addr, d_port);
290 list_for_each_entry(cp, &ip_vs_conn_tab[hash], c_list) {
291 if (d_addr == cp->caddr && d_port == cp->cport &&
292 s_port == cp->dport && s_addr == cp->daddr &&
293 protocol == cp->protocol) {
295 atomic_inc(&cp->refcnt);
301 ct_read_unlock(hash);
303 IP_VS_DBG(9, "lookup/out %s %u.%u.%u.%u:%d->%u.%u.%u.%u:%d %s\n",
304 ip_vs_proto_name(protocol),
305 NIPQUAD(s_addr), ntohs(s_port),
306 NIPQUAD(d_addr), ntohs(d_port),
307 ret?"hit":"not hit");
314 * Put back the conn and restart its timer with its timeout
316 void ip_vs_conn_put(struct ip_vs_conn *cp)
318 /* reset it expire in its timeout */
319 mod_timer(&cp->timer, jiffies+cp->timeout);
321 __ip_vs_conn_put(cp);
326 * Fill a no_client_port connection with a client port number
328 void ip_vs_conn_fill_cport(struct ip_vs_conn *cp, __be16 cport)
330 if (ip_vs_conn_unhash(cp)) {
331 spin_lock(&cp->lock);
332 if (cp->flags & IP_VS_CONN_F_NO_CPORT) {
333 atomic_dec(&ip_vs_conn_no_cport_cnt);
334 cp->flags &= ~IP_VS_CONN_F_NO_CPORT;
337 spin_unlock(&cp->lock);
339 /* hash on new dport */
346 * Bind a connection entry with the corresponding packet_xmit.
347 * Called by ip_vs_conn_new.
349 static inline void ip_vs_bind_xmit(struct ip_vs_conn *cp)
351 switch (IP_VS_FWD_METHOD(cp)) {
352 case IP_VS_CONN_F_MASQ:
353 cp->packet_xmit = ip_vs_nat_xmit;
356 case IP_VS_CONN_F_TUNNEL:
357 cp->packet_xmit = ip_vs_tunnel_xmit;
360 case IP_VS_CONN_F_DROUTE:
361 cp->packet_xmit = ip_vs_dr_xmit;
364 case IP_VS_CONN_F_LOCALNODE:
365 cp->packet_xmit = ip_vs_null_xmit;
368 case IP_VS_CONN_F_BYPASS:
369 cp->packet_xmit = ip_vs_bypass_xmit;
375 static inline int ip_vs_dest_totalconns(struct ip_vs_dest *dest)
377 return atomic_read(&dest->activeconns)
378 + atomic_read(&dest->inactconns);
382 * Bind a connection entry with a virtual service destination
383 * Called just after a new connection entry is created.
386 ip_vs_bind_dest(struct ip_vs_conn *cp, struct ip_vs_dest *dest)
388 /* if dest is NULL, then return directly */
392 /* Increase the refcnt counter of the dest */
393 atomic_inc(&dest->refcnt);
395 /* Bind with the destination and its corresponding transmitter */
396 cp->flags |= atomic_read(&dest->conn_flags);
399 IP_VS_DBG(7, "Bind-dest %s c:%u.%u.%u.%u:%d v:%u.%u.%u.%u:%d "
400 "d:%u.%u.%u.%u:%d fwd:%c s:%u conn->flags:%X conn->refcnt:%d "
402 ip_vs_proto_name(cp->protocol),
403 NIPQUAD(cp->caddr), ntohs(cp->cport),
404 NIPQUAD(cp->vaddr), ntohs(cp->vport),
405 NIPQUAD(cp->daddr), ntohs(cp->dport),
406 ip_vs_fwd_tag(cp), cp->state,
407 cp->flags, atomic_read(&cp->refcnt),
408 atomic_read(&dest->refcnt));
410 /* Update the connection counters */
411 if (!(cp->flags & IP_VS_CONN_F_TEMPLATE)) {
412 /* It is a normal connection, so increase the inactive
413 connection counter because it is in TCP SYNRECV
414 state (inactive) or other protocol inacive state */
415 atomic_inc(&dest->inactconns);
417 /* It is a persistent connection/template, so increase
418 the peristent connection counter */
419 atomic_inc(&dest->persistconns);
422 if (dest->u_threshold != 0 &&
423 ip_vs_dest_totalconns(dest) >= dest->u_threshold)
424 dest->flags |= IP_VS_DEST_F_OVERLOAD;
429 * Check if there is a destination for the connection, if so
430 * bind the connection to the destination.
432 struct ip_vs_dest *ip_vs_try_bind_dest(struct ip_vs_conn *cp)
434 struct ip_vs_dest *dest;
436 if ((cp) && (!cp->dest)) {
437 dest = ip_vs_find_dest(cp->daddr, cp->dport,
438 cp->vaddr, cp->vport, cp->protocol);
439 ip_vs_bind_dest(cp, dest);
444 EXPORT_SYMBOL(ip_vs_try_bind_dest);
448 * Unbind a connection entry with its VS destination
449 * Called by the ip_vs_conn_expire function.
451 static inline void ip_vs_unbind_dest(struct ip_vs_conn *cp)
453 struct ip_vs_dest *dest = cp->dest;
458 IP_VS_DBG(7, "Unbind-dest %s c:%u.%u.%u.%u:%d v:%u.%u.%u.%u:%d "
459 "d:%u.%u.%u.%u:%d fwd:%c s:%u conn->flags:%X conn->refcnt:%d "
461 ip_vs_proto_name(cp->protocol),
462 NIPQUAD(cp->caddr), ntohs(cp->cport),
463 NIPQUAD(cp->vaddr), ntohs(cp->vport),
464 NIPQUAD(cp->daddr), ntohs(cp->dport),
465 ip_vs_fwd_tag(cp), cp->state,
466 cp->flags, atomic_read(&cp->refcnt),
467 atomic_read(&dest->refcnt));
469 /* Update the connection counters */
470 if (!(cp->flags & IP_VS_CONN_F_TEMPLATE)) {
471 /* It is a normal connection, so decrease the inactconns
472 or activeconns counter */
473 if (cp->flags & IP_VS_CONN_F_INACTIVE) {
474 atomic_dec(&dest->inactconns);
476 atomic_dec(&dest->activeconns);
479 /* It is a persistent connection/template, so decrease
480 the peristent connection counter */
481 atomic_dec(&dest->persistconns);
484 if (dest->l_threshold != 0) {
485 if (ip_vs_dest_totalconns(dest) < dest->l_threshold)
486 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
487 } else if (dest->u_threshold != 0) {
488 if (ip_vs_dest_totalconns(dest) * 4 < dest->u_threshold * 3)
489 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
491 if (dest->flags & IP_VS_DEST_F_OVERLOAD)
492 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
496 * Simply decrease the refcnt of the dest, because the
497 * dest will be either in service's destination list
500 atomic_dec(&dest->refcnt);
505 * Checking if the destination of a connection template is available.
506 * If available, return 1, otherwise invalidate this connection
507 * template and return 0.
509 int ip_vs_check_template(struct ip_vs_conn *ct)
511 struct ip_vs_dest *dest = ct->dest;
514 * Checking the dest server status.
516 if ((dest == NULL) ||
517 !(dest->flags & IP_VS_DEST_F_AVAILABLE) ||
518 (sysctl_ip_vs_expire_quiescent_template &&
519 (atomic_read(&dest->weight) == 0))) {
520 IP_VS_DBG(9, "check_template: dest not available for "
521 "protocol %s s:%u.%u.%u.%u:%d v:%u.%u.%u.%u:%d "
522 "-> d:%u.%u.%u.%u:%d\n",
523 ip_vs_proto_name(ct->protocol),
524 NIPQUAD(ct->caddr), ntohs(ct->cport),
525 NIPQUAD(ct->vaddr), ntohs(ct->vport),
526 NIPQUAD(ct->daddr), ntohs(ct->dport));
529 * Invalidate the connection template
531 if (ct->vport != htons(0xffff)) {
532 if (ip_vs_conn_unhash(ct)) {
533 ct->dport = htons(0xffff);
534 ct->vport = htons(0xffff);
541 * Simply decrease the refcnt of the template,
542 * don't restart its timer.
544 atomic_dec(&ct->refcnt);
550 static void ip_vs_conn_expire(unsigned long data)
552 struct ip_vs_conn *cp = (struct ip_vs_conn *)data;
559 atomic_inc(&cp->refcnt);
562 * do I control anybody?
564 if (atomic_read(&cp->n_control))
568 * unhash it if it is hashed in the conn table
570 if (!ip_vs_conn_unhash(cp))
574 * refcnt==1 implies I'm the only one referrer
576 if (likely(atomic_read(&cp->refcnt) == 1)) {
577 /* delete the timer if it is activated by other users */
578 if (timer_pending(&cp->timer))
579 del_timer(&cp->timer);
581 /* does anybody control me? */
583 ip_vs_control_del(cp);
585 if (unlikely(cp->app != NULL))
586 ip_vs_unbind_app(cp);
587 ip_vs_unbind_dest(cp);
588 if (cp->flags & IP_VS_CONN_F_NO_CPORT)
589 atomic_dec(&ip_vs_conn_no_cport_cnt);
590 atomic_dec(&ip_vs_conn_count);
592 kmem_cache_free(ip_vs_conn_cachep, cp);
596 /* hash it back to the table */
600 IP_VS_DBG(7, "delayed: conn->refcnt-1=%d conn->n_control=%d\n",
601 atomic_read(&cp->refcnt)-1,
602 atomic_read(&cp->n_control));
608 void ip_vs_conn_expire_now(struct ip_vs_conn *cp)
610 if (del_timer(&cp->timer))
611 mod_timer(&cp->timer, jiffies);
616 * Create a new connection entry and hash it into the ip_vs_conn_tab
619 ip_vs_conn_new(int proto, __be32 caddr, __be16 cport, __be32 vaddr, __be16 vport,
620 __be32 daddr, __be16 dport, unsigned flags,
621 struct ip_vs_dest *dest)
623 struct ip_vs_conn *cp;
624 struct ip_vs_protocol *pp = ip_vs_proto_get(proto);
626 cp = kmem_cache_zalloc(ip_vs_conn_cachep, GFP_ATOMIC);
628 IP_VS_ERR_RL("ip_vs_conn_new: no memory available.\n");
632 INIT_LIST_HEAD(&cp->c_list);
633 init_timer(&cp->timer);
634 cp->timer.data = (unsigned long)cp;
635 cp->timer.function = ip_vs_conn_expire;
636 cp->protocol = proto;
644 spin_lock_init(&cp->lock);
647 * Set the entry is referenced by the current thread before hashing
648 * it in the table, so that other thread run ip_vs_random_dropentry
649 * but cannot drop this entry.
651 atomic_set(&cp->refcnt, 1);
653 atomic_set(&cp->n_control, 0);
654 atomic_set(&cp->in_pkts, 0);
656 atomic_inc(&ip_vs_conn_count);
657 if (flags & IP_VS_CONN_F_NO_CPORT)
658 atomic_inc(&ip_vs_conn_no_cport_cnt);
660 /* Bind the connection with a destination server */
661 ip_vs_bind_dest(cp, dest);
663 /* Set its state and timeout */
667 /* Bind its packet transmitter */
670 if (unlikely(pp && atomic_read(&pp->appcnt)))
671 ip_vs_bind_app(cp, pp);
673 /* Hash it in the ip_vs_conn_tab finally */
681 * /proc/net/ip_vs_conn entries
683 #ifdef CONFIG_PROC_FS
685 static void *ip_vs_conn_array(struct seq_file *seq, loff_t pos)
688 struct ip_vs_conn *cp;
690 for(idx = 0; idx < IP_VS_CONN_TAB_SIZE; idx++) {
691 ct_read_lock_bh(idx);
692 list_for_each_entry(cp, &ip_vs_conn_tab[idx], c_list) {
694 seq->private = &ip_vs_conn_tab[idx];
698 ct_read_unlock_bh(idx);
704 static void *ip_vs_conn_seq_start(struct seq_file *seq, loff_t *pos)
707 return *pos ? ip_vs_conn_array(seq, *pos - 1) :SEQ_START_TOKEN;
710 static void *ip_vs_conn_seq_next(struct seq_file *seq, void *v, loff_t *pos)
712 struct ip_vs_conn *cp = v;
713 struct list_head *e, *l = seq->private;
717 if (v == SEQ_START_TOKEN)
718 return ip_vs_conn_array(seq, 0);
720 /* more on same hash chain? */
721 if ((e = cp->c_list.next) != l)
722 return list_entry(e, struct ip_vs_conn, c_list);
724 idx = l - ip_vs_conn_tab;
725 ct_read_unlock_bh(idx);
727 while (++idx < IP_VS_CONN_TAB_SIZE) {
728 ct_read_lock_bh(idx);
729 list_for_each_entry(cp, &ip_vs_conn_tab[idx], c_list) {
730 seq->private = &ip_vs_conn_tab[idx];
733 ct_read_unlock_bh(idx);
739 static void ip_vs_conn_seq_stop(struct seq_file *seq, void *v)
741 struct list_head *l = seq->private;
744 ct_read_unlock_bh(l - ip_vs_conn_tab);
747 static int ip_vs_conn_seq_show(struct seq_file *seq, void *v)
750 if (v == SEQ_START_TOKEN)
752 "Pro FromIP FPrt ToIP TPrt DestIP DPrt State Expires\n");
754 const struct ip_vs_conn *cp = v;
757 "%-3s %08X %04X %08X %04X %08X %04X %-11s %7lu\n",
758 ip_vs_proto_name(cp->protocol),
759 ntohl(cp->caddr), ntohs(cp->cport),
760 ntohl(cp->vaddr), ntohs(cp->vport),
761 ntohl(cp->daddr), ntohs(cp->dport),
762 ip_vs_state_name(cp->protocol, cp->state),
763 (cp->timer.expires-jiffies)/HZ);
768 static const struct seq_operations ip_vs_conn_seq_ops = {
769 .start = ip_vs_conn_seq_start,
770 .next = ip_vs_conn_seq_next,
771 .stop = ip_vs_conn_seq_stop,
772 .show = ip_vs_conn_seq_show,
775 static int ip_vs_conn_open(struct inode *inode, struct file *file)
777 return seq_open(file, &ip_vs_conn_seq_ops);
780 static const struct file_operations ip_vs_conn_fops = {
781 .owner = THIS_MODULE,
782 .open = ip_vs_conn_open,
785 .release = seq_release,
791 * Randomly drop connection entries before running out of memory
793 static inline int todrop_entry(struct ip_vs_conn *cp)
796 * The drop rate array needs tuning for real environments.
797 * Called from timer bh only => no locking
799 static const char todrop_rate[9] = {0, 1, 2, 3, 4, 5, 6, 7, 8};
800 static char todrop_counter[9] = {0};
803 /* if the conn entry hasn't lasted for 60 seconds, don't drop it.
804 This will leave enough time for normal connection to get
806 if (time_before(cp->timeout + jiffies, cp->timer.expires + 60*HZ))
809 /* Don't drop the entry if its number of incoming packets is not
811 i = atomic_read(&cp->in_pkts);
812 if (i > 8 || i < 0) return 0;
814 if (!todrop_rate[i]) return 0;
815 if (--todrop_counter[i] > 0) return 0;
817 todrop_counter[i] = todrop_rate[i];
821 /* Called from keventd and must protect itself from softirqs */
822 void ip_vs_random_dropentry(void)
825 struct ip_vs_conn *cp;
828 * Randomly scan 1/32 of the whole table every second
830 for (idx = 0; idx < (IP_VS_CONN_TAB_SIZE>>5); idx++) {
831 unsigned hash = net_random() & IP_VS_CONN_TAB_MASK;
834 * Lock is actually needed in this loop.
836 ct_write_lock_bh(hash);
838 list_for_each_entry(cp, &ip_vs_conn_tab[hash], c_list) {
839 if (cp->flags & IP_VS_CONN_F_TEMPLATE)
840 /* connection template */
843 if (cp->protocol == IPPROTO_TCP) {
845 case IP_VS_TCP_S_SYN_RECV:
846 case IP_VS_TCP_S_SYNACK:
849 case IP_VS_TCP_S_ESTABLISHED:
850 if (todrop_entry(cp))
858 if (!todrop_entry(cp))
862 IP_VS_DBG(4, "del connection\n");
863 ip_vs_conn_expire_now(cp);
865 IP_VS_DBG(4, "del conn template\n");
866 ip_vs_conn_expire_now(cp->control);
869 ct_write_unlock_bh(hash);
875 * Flush all the connection entries in the ip_vs_conn_tab
877 static void ip_vs_conn_flush(void)
880 struct ip_vs_conn *cp;
883 for (idx=0; idx<IP_VS_CONN_TAB_SIZE; idx++) {
885 * Lock is actually needed in this loop.
887 ct_write_lock_bh(idx);
889 list_for_each_entry(cp, &ip_vs_conn_tab[idx], c_list) {
891 IP_VS_DBG(4, "del connection\n");
892 ip_vs_conn_expire_now(cp);
894 IP_VS_DBG(4, "del conn template\n");
895 ip_vs_conn_expire_now(cp->control);
898 ct_write_unlock_bh(idx);
901 /* the counter may be not NULL, because maybe some conn entries
902 are run by slow timer handler or unhashed but still referred */
903 if (atomic_read(&ip_vs_conn_count) != 0) {
910 int ip_vs_conn_init(void)
915 * Allocate the connection hash table and initialize its list heads
917 ip_vs_conn_tab = vmalloc(IP_VS_CONN_TAB_SIZE*sizeof(struct list_head));
921 /* Allocate ip_vs_conn slab cache */
922 ip_vs_conn_cachep = kmem_cache_create("ip_vs_conn",
923 sizeof(struct ip_vs_conn), 0,
924 SLAB_HWCACHE_ALIGN, NULL);
925 if (!ip_vs_conn_cachep) {
926 vfree(ip_vs_conn_tab);
930 IP_VS_INFO("Connection hash table configured "
931 "(size=%d, memory=%ldKbytes)\n",
933 (long)(IP_VS_CONN_TAB_SIZE*sizeof(struct list_head))/1024);
934 IP_VS_DBG(0, "Each connection entry needs %Zd bytes at least\n",
935 sizeof(struct ip_vs_conn));
937 for (idx = 0; idx < IP_VS_CONN_TAB_SIZE; idx++) {
938 INIT_LIST_HEAD(&ip_vs_conn_tab[idx]);
941 for (idx = 0; idx < CT_LOCKARRAY_SIZE; idx++) {
942 rwlock_init(&__ip_vs_conntbl_lock_array[idx].l);
945 proc_net_fops_create(&init_net, "ip_vs_conn", 0, &ip_vs_conn_fops);
947 /* calculate the random value for connection hash */
948 get_random_bytes(&ip_vs_conn_rnd, sizeof(ip_vs_conn_rnd));
954 void ip_vs_conn_cleanup(void)
956 /* flush all the connection entries first */
959 /* Release the empty cache */
960 kmem_cache_destroy(ip_vs_conn_cachep);
961 proc_net_remove(&init_net, "ip_vs_conn");
962 vfree(ip_vs_conn_tab);