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 * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
9 * Peter Kese <peter.kese@ijs.si>
10 * Julian Anastasov <ja@ssi.bg>
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
21 #include <linux/module.h>
22 #include <linux/init.h>
23 #include <linux/types.h>
24 #include <linux/capability.h>
26 #include <linux/sysctl.h>
27 #include <linux/proc_fs.h>
28 #include <linux/workqueue.h>
29 #include <linux/swap.h>
30 #include <linux/seq_file.h>
32 #include <linux/netfilter.h>
33 #include <linux/netfilter_ipv4.h>
34 #include <linux/mutex.h>
36 #include <net/net_namespace.h>
38 #ifdef CONFIG_IP_VS_IPV6
40 #include <net/ip6_route.h>
42 #include <net/route.h>
44 #include <net/genetlink.h>
46 #include <asm/uaccess.h>
48 #include <net/ip_vs.h>
50 /* semaphore for IPVS sockopts. And, [gs]etsockopt may sleep. */
51 static DEFINE_MUTEX(__ip_vs_mutex);
53 /* lock for service table */
54 static DEFINE_RWLOCK(__ip_vs_svc_lock);
56 /* lock for table with the real services */
57 static DEFINE_RWLOCK(__ip_vs_rs_lock);
59 /* lock for state and timeout tables */
60 static DEFINE_RWLOCK(__ip_vs_securetcp_lock);
62 /* lock for drop entry handling */
63 static DEFINE_SPINLOCK(__ip_vs_dropentry_lock);
65 /* lock for drop packet handling */
66 static DEFINE_SPINLOCK(__ip_vs_droppacket_lock);
68 /* 1/rate drop and drop-entry variables */
69 int ip_vs_drop_rate = 0;
70 int ip_vs_drop_counter = 0;
71 static atomic_t ip_vs_dropentry = ATOMIC_INIT(0);
73 /* number of virtual services */
74 static int ip_vs_num_services = 0;
76 /* sysctl variables */
77 static int sysctl_ip_vs_drop_entry = 0;
78 static int sysctl_ip_vs_drop_packet = 0;
79 static int sysctl_ip_vs_secure_tcp = 0;
80 static int sysctl_ip_vs_amemthresh = 1024;
81 static int sysctl_ip_vs_am_droprate = 10;
82 int sysctl_ip_vs_cache_bypass = 0;
83 int sysctl_ip_vs_expire_nodest_conn = 0;
84 int sysctl_ip_vs_expire_quiescent_template = 0;
85 int sysctl_ip_vs_sync_threshold[2] = { 3, 50 };
86 int sysctl_ip_vs_nat_icmp_send = 0;
89 #ifdef CONFIG_IP_VS_DEBUG
90 static int sysctl_ip_vs_debug_level = 0;
92 int ip_vs_get_debug_level(void)
94 return sysctl_ip_vs_debug_level;
98 #ifdef CONFIG_IP_VS_IPV6
99 /* Taken from rt6_fill_node() in net/ipv6/route.c, is there a better way? */
100 static int __ip_vs_addr_is_local_v6(const struct in6_addr *addr)
108 .saddr = { .s6_addr32 = {0, 0, 0, 0} }, } },
111 rt = (struct rt6_info *)ip6_route_output(&init_net, NULL, &fl);
112 if (rt && rt->rt6i_dev && (rt->rt6i_dev->flags & IFF_LOOPBACK))
119 * update_defense_level is called from keventd and from sysctl,
120 * so it needs to protect itself from softirqs
122 static void update_defense_level(void)
125 static int old_secure_tcp = 0;
130 /* we only count free and buffered memory (in pages) */
132 availmem = i.freeram + i.bufferram;
133 /* however in linux 2.5 the i.bufferram is total page cache size,
135 /* si_swapinfo(&i); */
136 /* availmem = availmem - (i.totalswap - i.freeswap); */
138 nomem = (availmem < sysctl_ip_vs_amemthresh);
143 spin_lock(&__ip_vs_dropentry_lock);
144 switch (sysctl_ip_vs_drop_entry) {
146 atomic_set(&ip_vs_dropentry, 0);
150 atomic_set(&ip_vs_dropentry, 1);
151 sysctl_ip_vs_drop_entry = 2;
153 atomic_set(&ip_vs_dropentry, 0);
158 atomic_set(&ip_vs_dropentry, 1);
160 atomic_set(&ip_vs_dropentry, 0);
161 sysctl_ip_vs_drop_entry = 1;
165 atomic_set(&ip_vs_dropentry, 1);
168 spin_unlock(&__ip_vs_dropentry_lock);
171 spin_lock(&__ip_vs_droppacket_lock);
172 switch (sysctl_ip_vs_drop_packet) {
178 ip_vs_drop_rate = ip_vs_drop_counter
179 = sysctl_ip_vs_amemthresh /
180 (sysctl_ip_vs_amemthresh-availmem);
181 sysctl_ip_vs_drop_packet = 2;
188 ip_vs_drop_rate = ip_vs_drop_counter
189 = sysctl_ip_vs_amemthresh /
190 (sysctl_ip_vs_amemthresh-availmem);
193 sysctl_ip_vs_drop_packet = 1;
197 ip_vs_drop_rate = sysctl_ip_vs_am_droprate;
200 spin_unlock(&__ip_vs_droppacket_lock);
203 write_lock(&__ip_vs_securetcp_lock);
204 switch (sysctl_ip_vs_secure_tcp) {
206 if (old_secure_tcp >= 2)
211 if (old_secure_tcp < 2)
213 sysctl_ip_vs_secure_tcp = 2;
215 if (old_secure_tcp >= 2)
221 if (old_secure_tcp < 2)
224 if (old_secure_tcp >= 2)
226 sysctl_ip_vs_secure_tcp = 1;
230 if (old_secure_tcp < 2)
234 old_secure_tcp = sysctl_ip_vs_secure_tcp;
236 ip_vs_protocol_timeout_change(sysctl_ip_vs_secure_tcp>1);
237 write_unlock(&__ip_vs_securetcp_lock);
244 * Timer for checking the defense
246 #define DEFENSE_TIMER_PERIOD 1*HZ
247 static void defense_work_handler(struct work_struct *work);
248 static DECLARE_DELAYED_WORK(defense_work, defense_work_handler);
250 static void defense_work_handler(struct work_struct *work)
252 update_defense_level();
253 if (atomic_read(&ip_vs_dropentry))
254 ip_vs_random_dropentry();
256 schedule_delayed_work(&defense_work, DEFENSE_TIMER_PERIOD);
260 ip_vs_use_count_inc(void)
262 return try_module_get(THIS_MODULE);
266 ip_vs_use_count_dec(void)
268 module_put(THIS_MODULE);
273 * Hash table: for virtual service lookups
275 #define IP_VS_SVC_TAB_BITS 8
276 #define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
277 #define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)
279 /* the service table hashed by <protocol, addr, port> */
280 static struct list_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE];
281 /* the service table hashed by fwmark */
282 static struct list_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE];
285 * Hash table: for real service lookups
287 #define IP_VS_RTAB_BITS 4
288 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
289 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
291 static struct list_head ip_vs_rtable[IP_VS_RTAB_SIZE];
294 * Trash for destinations
296 static LIST_HEAD(ip_vs_dest_trash);
299 * FTP & NULL virtual service counters
301 static atomic_t ip_vs_ftpsvc_counter = ATOMIC_INIT(0);
302 static atomic_t ip_vs_nullsvc_counter = ATOMIC_INIT(0);
306 * Returns hash value for virtual service
308 static __inline__ unsigned
309 ip_vs_svc_hashkey(int af, unsigned proto, const union nf_inet_addr *addr,
312 register unsigned porth = ntohs(port);
313 __be32 addr_fold = addr->ip;
315 #ifdef CONFIG_IP_VS_IPV6
317 addr_fold = addr->ip6[0]^addr->ip6[1]^
318 addr->ip6[2]^addr->ip6[3];
321 return (proto^ntohl(addr_fold)^(porth>>IP_VS_SVC_TAB_BITS)^porth)
322 & IP_VS_SVC_TAB_MASK;
326 * Returns hash value of fwmark for virtual service lookup
328 static __inline__ unsigned ip_vs_svc_fwm_hashkey(__u32 fwmark)
330 return fwmark & IP_VS_SVC_TAB_MASK;
334 * Hashes a service in the ip_vs_svc_table by <proto,addr,port>
335 * or in the ip_vs_svc_fwm_table by fwmark.
336 * Should be called with locked tables.
338 static int ip_vs_svc_hash(struct ip_vs_service *svc)
342 if (svc->flags & IP_VS_SVC_F_HASHED) {
343 IP_VS_ERR("ip_vs_svc_hash(): request for already hashed, "
344 "called from %p\n", __builtin_return_address(0));
348 if (svc->fwmark == 0) {
350 * Hash it by <protocol,addr,port> in ip_vs_svc_table
352 hash = ip_vs_svc_hashkey(svc->af, svc->protocol, &svc->addr,
354 list_add(&svc->s_list, &ip_vs_svc_table[hash]);
357 * Hash it by fwmark in ip_vs_svc_fwm_table
359 hash = ip_vs_svc_fwm_hashkey(svc->fwmark);
360 list_add(&svc->f_list, &ip_vs_svc_fwm_table[hash]);
363 svc->flags |= IP_VS_SVC_F_HASHED;
364 /* increase its refcnt because it is referenced by the svc table */
365 atomic_inc(&svc->refcnt);
371 * Unhashes a service from ip_vs_svc_table/ip_vs_svc_fwm_table.
372 * Should be called with locked tables.
374 static int ip_vs_svc_unhash(struct ip_vs_service *svc)
376 if (!(svc->flags & IP_VS_SVC_F_HASHED)) {
377 IP_VS_ERR("ip_vs_svc_unhash(): request for unhash flagged, "
378 "called from %p\n", __builtin_return_address(0));
382 if (svc->fwmark == 0) {
383 /* Remove it from the ip_vs_svc_table table */
384 list_del(&svc->s_list);
386 /* Remove it from the ip_vs_svc_fwm_table table */
387 list_del(&svc->f_list);
390 svc->flags &= ~IP_VS_SVC_F_HASHED;
391 atomic_dec(&svc->refcnt);
397 * Get service by {proto,addr,port} in the service table.
399 static inline struct ip_vs_service *
400 __ip_vs_service_get(int af, __u16 protocol, const union nf_inet_addr *vaddr,
404 struct ip_vs_service *svc;
406 /* Check for "full" addressed entries */
407 hash = ip_vs_svc_hashkey(af, protocol, vaddr, vport);
409 list_for_each_entry(svc, &ip_vs_svc_table[hash], s_list){
411 && ip_vs_addr_equal(af, &svc->addr, vaddr)
412 && (svc->port == vport)
413 && (svc->protocol == protocol)) {
415 atomic_inc(&svc->usecnt);
425 * Get service by {fwmark} in the service table.
427 static inline struct ip_vs_service *
428 __ip_vs_svc_fwm_get(int af, __u32 fwmark)
431 struct ip_vs_service *svc;
433 /* Check for fwmark addressed entries */
434 hash = ip_vs_svc_fwm_hashkey(fwmark);
436 list_for_each_entry(svc, &ip_vs_svc_fwm_table[hash], f_list) {
437 if (svc->fwmark == fwmark && svc->af == af) {
439 atomic_inc(&svc->usecnt);
447 struct ip_vs_service *
448 ip_vs_service_get(int af, __u32 fwmark, __u16 protocol,
449 const union nf_inet_addr *vaddr, __be16 vport)
451 struct ip_vs_service *svc;
453 read_lock(&__ip_vs_svc_lock);
456 * Check the table hashed by fwmark first
458 if (fwmark && (svc = __ip_vs_svc_fwm_get(af, fwmark)))
462 * Check the table hashed by <protocol,addr,port>
463 * for "full" addressed entries
465 svc = __ip_vs_service_get(af, protocol, vaddr, vport);
468 && protocol == IPPROTO_TCP
469 && atomic_read(&ip_vs_ftpsvc_counter)
470 && (vport == FTPDATA || ntohs(vport) >= PROT_SOCK)) {
472 * Check if ftp service entry exists, the packet
473 * might belong to FTP data connections.
475 svc = __ip_vs_service_get(af, protocol, vaddr, FTPPORT);
479 && atomic_read(&ip_vs_nullsvc_counter)) {
481 * Check if the catch-all port (port zero) exists
483 svc = __ip_vs_service_get(af, protocol, vaddr, 0);
487 read_unlock(&__ip_vs_svc_lock);
489 IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
490 fwmark, ip_vs_proto_name(protocol),
491 IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
492 svc ? "hit" : "not hit");
499 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
501 atomic_inc(&svc->refcnt);
506 __ip_vs_unbind_svc(struct ip_vs_dest *dest)
508 struct ip_vs_service *svc = dest->svc;
511 if (atomic_dec_and_test(&svc->refcnt))
517 * Returns hash value for real service
519 static inline unsigned ip_vs_rs_hashkey(int af,
520 const union nf_inet_addr *addr,
523 register unsigned porth = ntohs(port);
524 __be32 addr_fold = addr->ip;
526 #ifdef CONFIG_IP_VS_IPV6
528 addr_fold = addr->ip6[0]^addr->ip6[1]^
529 addr->ip6[2]^addr->ip6[3];
532 return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
537 * Hashes ip_vs_dest in ip_vs_rtable by <proto,addr,port>.
538 * should be called with locked tables.
540 static int ip_vs_rs_hash(struct ip_vs_dest *dest)
544 if (!list_empty(&dest->d_list)) {
549 * Hash by proto,addr,port,
550 * which are the parameters of the real service.
552 hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
554 list_add(&dest->d_list, &ip_vs_rtable[hash]);
560 * UNhashes ip_vs_dest from ip_vs_rtable.
561 * should be called with locked tables.
563 static int ip_vs_rs_unhash(struct ip_vs_dest *dest)
566 * Remove it from the ip_vs_rtable table.
568 if (!list_empty(&dest->d_list)) {
569 list_del(&dest->d_list);
570 INIT_LIST_HEAD(&dest->d_list);
577 * Lookup real service by <proto,addr,port> in the real service table.
580 ip_vs_lookup_real_service(int af, __u16 protocol,
581 const union nf_inet_addr *daddr,
585 struct ip_vs_dest *dest;
588 * Check for "full" addressed entries
589 * Return the first found entry
591 hash = ip_vs_rs_hashkey(af, daddr, dport);
593 read_lock(&__ip_vs_rs_lock);
594 list_for_each_entry(dest, &ip_vs_rtable[hash], d_list) {
596 && ip_vs_addr_equal(af, &dest->addr, daddr)
597 && (dest->port == dport)
598 && ((dest->protocol == protocol) ||
601 read_unlock(&__ip_vs_rs_lock);
605 read_unlock(&__ip_vs_rs_lock);
611 * Lookup destination by {addr,port} in the given service
613 static struct ip_vs_dest *
614 ip_vs_lookup_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
617 struct ip_vs_dest *dest;
620 * Find the destination for the given service
622 list_for_each_entry(dest, &svc->destinations, n_list) {
623 if ((dest->af == svc->af)
624 && ip_vs_addr_equal(svc->af, &dest->addr, daddr)
625 && (dest->port == dport)) {
635 * Find destination by {daddr,dport,vaddr,protocol}
636 * Cretaed to be used in ip_vs_process_message() in
637 * the backup synchronization daemon. It finds the
638 * destination to be bound to the received connection
641 * ip_vs_lookup_real_service() looked promissing, but
642 * seems not working as expected.
644 struct ip_vs_dest *ip_vs_find_dest(int af, const union nf_inet_addr *daddr,
646 const union nf_inet_addr *vaddr,
647 __be16 vport, __u16 protocol)
649 struct ip_vs_dest *dest;
650 struct ip_vs_service *svc;
652 svc = ip_vs_service_get(af, 0, protocol, vaddr, vport);
655 dest = ip_vs_lookup_dest(svc, daddr, dport);
657 atomic_inc(&dest->refcnt);
658 ip_vs_service_put(svc);
663 * Lookup dest by {svc,addr,port} in the destination trash.
664 * The destination trash is used to hold the destinations that are removed
665 * from the service table but are still referenced by some conn entries.
666 * The reason to add the destination trash is when the dest is temporary
667 * down (either by administrator or by monitor program), the dest can be
668 * picked back from the trash, the remaining connections to the dest can
669 * continue, and the counting information of the dest is also useful for
672 static struct ip_vs_dest *
673 ip_vs_trash_get_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
676 struct ip_vs_dest *dest, *nxt;
679 * Find the destination in trash
681 list_for_each_entry_safe(dest, nxt, &ip_vs_dest_trash, n_list) {
682 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
685 IP_VS_DBG_ADDR(svc->af, &dest->addr),
687 atomic_read(&dest->refcnt));
688 if (dest->af == svc->af &&
689 ip_vs_addr_equal(svc->af, &dest->addr, daddr) &&
690 dest->port == dport &&
691 dest->vfwmark == svc->fwmark &&
692 dest->protocol == svc->protocol &&
694 (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
695 dest->vport == svc->port))) {
701 * Try to purge the destination from trash if not referenced
703 if (atomic_read(&dest->refcnt) == 1) {
704 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u "
707 IP_VS_DBG_ADDR(svc->af, &dest->addr),
709 list_del(&dest->n_list);
710 ip_vs_dst_reset(dest);
711 __ip_vs_unbind_svc(dest);
721 * Clean up all the destinations in the trash
722 * Called by the ip_vs_control_cleanup()
724 * When the ip_vs_control_clearup is activated by ipvs module exit,
725 * the service tables must have been flushed and all the connections
726 * are expired, and the refcnt of each destination in the trash must
727 * be 1, so we simply release them here.
729 static void ip_vs_trash_cleanup(void)
731 struct ip_vs_dest *dest, *nxt;
733 list_for_each_entry_safe(dest, nxt, &ip_vs_dest_trash, n_list) {
734 list_del(&dest->n_list);
735 ip_vs_dst_reset(dest);
736 __ip_vs_unbind_svc(dest);
743 ip_vs_zero_stats(struct ip_vs_stats *stats)
745 spin_lock_bh(&stats->lock);
747 memset(&stats->ustats, 0, sizeof(stats->ustats));
748 ip_vs_zero_estimator(stats);
750 spin_unlock_bh(&stats->lock);
754 * Update a destination in the given service
757 __ip_vs_update_dest(struct ip_vs_service *svc,
758 struct ip_vs_dest *dest, struct ip_vs_dest_user_kern *udest)
762 /* set the weight and the flags */
763 atomic_set(&dest->weight, udest->weight);
764 conn_flags = udest->conn_flags | IP_VS_CONN_F_INACTIVE;
766 /* check if local node and update the flags */
767 #ifdef CONFIG_IP_VS_IPV6
768 if (svc->af == AF_INET6) {
769 if (__ip_vs_addr_is_local_v6(&udest->addr.in6)) {
770 conn_flags = (conn_flags & ~IP_VS_CONN_F_FWD_MASK)
771 | IP_VS_CONN_F_LOCALNODE;
775 if (inet_addr_type(&init_net, udest->addr.ip) == RTN_LOCAL) {
776 conn_flags = (conn_flags & ~IP_VS_CONN_F_FWD_MASK)
777 | IP_VS_CONN_F_LOCALNODE;
780 /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
781 if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != 0) {
782 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
785 * Put the real service in ip_vs_rtable if not present.
786 * For now only for NAT!
788 write_lock_bh(&__ip_vs_rs_lock);
790 write_unlock_bh(&__ip_vs_rs_lock);
792 atomic_set(&dest->conn_flags, conn_flags);
794 /* bind the service */
796 __ip_vs_bind_svc(dest, svc);
798 if (dest->svc != svc) {
799 __ip_vs_unbind_svc(dest);
800 ip_vs_zero_stats(&dest->stats);
801 __ip_vs_bind_svc(dest, svc);
805 /* set the dest status flags */
806 dest->flags |= IP_VS_DEST_F_AVAILABLE;
808 if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
809 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
810 dest->u_threshold = udest->u_threshold;
811 dest->l_threshold = udest->l_threshold;
816 * Create a destination for the given service
819 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
820 struct ip_vs_dest **dest_p)
822 struct ip_vs_dest *dest;
827 #ifdef CONFIG_IP_VS_IPV6
828 if (svc->af == AF_INET6) {
829 atype = ipv6_addr_type(&udest->addr.in6);
830 if ((!(atype & IPV6_ADDR_UNICAST) ||
831 atype & IPV6_ADDR_LINKLOCAL) &&
832 !__ip_vs_addr_is_local_v6(&udest->addr.in6))
837 atype = inet_addr_type(&init_net, udest->addr.ip);
838 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
842 dest = kzalloc(sizeof(struct ip_vs_dest), GFP_ATOMIC);
844 IP_VS_ERR("ip_vs_new_dest: kmalloc failed.\n");
849 dest->protocol = svc->protocol;
850 dest->vaddr = svc->addr;
851 dest->vport = svc->port;
852 dest->vfwmark = svc->fwmark;
853 ip_vs_addr_copy(svc->af, &dest->addr, &udest->addr);
854 dest->port = udest->port;
856 atomic_set(&dest->activeconns, 0);
857 atomic_set(&dest->inactconns, 0);
858 atomic_set(&dest->persistconns, 0);
859 atomic_set(&dest->refcnt, 0);
861 INIT_LIST_HEAD(&dest->d_list);
862 spin_lock_init(&dest->dst_lock);
863 spin_lock_init(&dest->stats.lock);
864 __ip_vs_update_dest(svc, dest, udest);
865 ip_vs_new_estimator(&dest->stats);
875 * Add a destination into an existing service
878 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
880 struct ip_vs_dest *dest;
881 union nf_inet_addr daddr;
882 __be16 dport = udest->port;
887 if (udest->weight < 0) {
888 IP_VS_ERR("ip_vs_add_dest(): server weight less than zero\n");
892 if (udest->l_threshold > udest->u_threshold) {
893 IP_VS_ERR("ip_vs_add_dest(): lower threshold is higher than "
894 "upper threshold\n");
898 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
901 * Check if the dest already exists in the list
903 dest = ip_vs_lookup_dest(svc, &daddr, dport);
906 IP_VS_DBG(1, "ip_vs_add_dest(): dest already exists\n");
911 * Check if the dest already exists in the trash and
912 * is from the same service
914 dest = ip_vs_trash_get_dest(svc, &daddr, dport);
917 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
918 "dest->refcnt=%d, service %u/%s:%u\n",
919 IP_VS_DBG_ADDR(svc->af, &daddr), ntohs(dport),
920 atomic_read(&dest->refcnt),
922 IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
925 __ip_vs_update_dest(svc, dest, udest);
928 * Get the destination from the trash
930 list_del(&dest->n_list);
932 ip_vs_new_estimator(&dest->stats);
934 write_lock_bh(&__ip_vs_svc_lock);
937 * Wait until all other svc users go away.
939 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 1);
941 list_add(&dest->n_list, &svc->destinations);
944 /* call the update_service function of its scheduler */
945 if (svc->scheduler->update_service)
946 svc->scheduler->update_service(svc);
948 write_unlock_bh(&__ip_vs_svc_lock);
953 * Allocate and initialize the dest structure
955 ret = ip_vs_new_dest(svc, udest, &dest);
961 * Add the dest entry into the list
963 atomic_inc(&dest->refcnt);
965 write_lock_bh(&__ip_vs_svc_lock);
968 * Wait until all other svc users go away.
970 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 1);
972 list_add(&dest->n_list, &svc->destinations);
975 /* call the update_service function of its scheduler */
976 if (svc->scheduler->update_service)
977 svc->scheduler->update_service(svc);
979 write_unlock_bh(&__ip_vs_svc_lock);
988 * Edit a destination in the given service
991 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
993 struct ip_vs_dest *dest;
994 union nf_inet_addr daddr;
995 __be16 dport = udest->port;
999 if (udest->weight < 0) {
1000 IP_VS_ERR("ip_vs_edit_dest(): server weight less than zero\n");
1004 if (udest->l_threshold > udest->u_threshold) {
1005 IP_VS_ERR("ip_vs_edit_dest(): lower threshold is higher than "
1006 "upper threshold\n");
1010 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
1013 * Lookup the destination list
1015 dest = ip_vs_lookup_dest(svc, &daddr, dport);
1018 IP_VS_DBG(1, "ip_vs_edit_dest(): dest doesn't exist\n");
1022 __ip_vs_update_dest(svc, dest, udest);
1024 write_lock_bh(&__ip_vs_svc_lock);
1026 /* Wait until all other svc users go away */
1027 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 1);
1029 /* call the update_service, because server weight may be changed */
1030 if (svc->scheduler->update_service)
1031 svc->scheduler->update_service(svc);
1033 write_unlock_bh(&__ip_vs_svc_lock);
1042 * Delete a destination (must be already unlinked from the service)
1044 static void __ip_vs_del_dest(struct ip_vs_dest *dest)
1046 ip_vs_kill_estimator(&dest->stats);
1049 * Remove it from the d-linked list with the real services.
1051 write_lock_bh(&__ip_vs_rs_lock);
1052 ip_vs_rs_unhash(dest);
1053 write_unlock_bh(&__ip_vs_rs_lock);
1056 * Decrease the refcnt of the dest, and free the dest
1057 * if nobody refers to it (refcnt=0). Otherwise, throw
1058 * the destination into the trash.
1060 if (atomic_dec_and_test(&dest->refcnt)) {
1061 ip_vs_dst_reset(dest);
1062 /* simply decrease svc->refcnt here, let the caller check
1063 and release the service if nobody refers to it.
1064 Only user context can release destination and service,
1065 and only one user context can update virtual service at a
1066 time, so the operation here is OK */
1067 atomic_dec(&dest->svc->refcnt);
1070 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, "
1071 "dest->refcnt=%d\n",
1072 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1074 atomic_read(&dest->refcnt));
1075 list_add(&dest->n_list, &ip_vs_dest_trash);
1076 atomic_inc(&dest->refcnt);
1082 * Unlink a destination from the given service
1084 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1085 struct ip_vs_dest *dest,
1088 dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1091 * Remove it from the d-linked destination list.
1093 list_del(&dest->n_list);
1097 * Call the update_service function of its scheduler
1099 if (svcupd && svc->scheduler->update_service)
1100 svc->scheduler->update_service(svc);
1105 * Delete a destination server in the given service
1108 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1110 struct ip_vs_dest *dest;
1111 __be16 dport = udest->port;
1115 dest = ip_vs_lookup_dest(svc, &udest->addr, dport);
1118 IP_VS_DBG(1, "ip_vs_del_dest(): destination not found!\n");
1122 write_lock_bh(&__ip_vs_svc_lock);
1125 * Wait until all other svc users go away.
1127 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 1);
1130 * Unlink dest from the service
1132 __ip_vs_unlink_dest(svc, dest, 1);
1134 write_unlock_bh(&__ip_vs_svc_lock);
1137 * Delete the destination
1139 __ip_vs_del_dest(dest);
1148 * Add a service into the service hash table
1151 ip_vs_add_service(struct ip_vs_service_user_kern *u,
1152 struct ip_vs_service **svc_p)
1155 struct ip_vs_scheduler *sched = NULL;
1156 struct ip_vs_service *svc = NULL;
1158 /* increase the module use count */
1159 ip_vs_use_count_inc();
1161 /* Lookup the scheduler by 'u->sched_name' */
1162 sched = ip_vs_scheduler_get(u->sched_name);
1163 if (sched == NULL) {
1164 IP_VS_INFO("Scheduler module ip_vs_%s not found\n",
1170 #ifdef CONFIG_IP_VS_IPV6
1171 if (u->af == AF_INET6) {
1172 if (!sched->supports_ipv6) {
1173 ret = -EAFNOSUPPORT;
1176 if ((u->netmask < 1) || (u->netmask > 128)) {
1183 svc = kzalloc(sizeof(struct ip_vs_service), GFP_ATOMIC);
1185 IP_VS_DBG(1, "ip_vs_add_service: kmalloc failed.\n");
1190 /* I'm the first user of the service */
1191 atomic_set(&svc->usecnt, 1);
1192 atomic_set(&svc->refcnt, 0);
1195 svc->protocol = u->protocol;
1196 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1197 svc->port = u->port;
1198 svc->fwmark = u->fwmark;
1199 svc->flags = u->flags;
1200 svc->timeout = u->timeout * HZ;
1201 svc->netmask = u->netmask;
1203 INIT_LIST_HEAD(&svc->destinations);
1204 rwlock_init(&svc->sched_lock);
1205 spin_lock_init(&svc->stats.lock);
1207 /* Bind the scheduler */
1208 ret = ip_vs_bind_scheduler(svc, sched);
1213 /* Update the virtual service counters */
1214 if (svc->port == FTPPORT)
1215 atomic_inc(&ip_vs_ftpsvc_counter);
1216 else if (svc->port == 0)
1217 atomic_inc(&ip_vs_nullsvc_counter);
1219 ip_vs_new_estimator(&svc->stats);
1221 /* Count only IPv4 services for old get/setsockopt interface */
1222 if (svc->af == AF_INET)
1223 ip_vs_num_services++;
1225 /* Hash the service into the service table */
1226 write_lock_bh(&__ip_vs_svc_lock);
1227 ip_vs_svc_hash(svc);
1228 write_unlock_bh(&__ip_vs_svc_lock);
1236 ip_vs_unbind_scheduler(svc);
1239 ip_vs_app_inc_put(svc->inc);
1244 ip_vs_scheduler_put(sched);
1247 /* decrease the module use count */
1248 ip_vs_use_count_dec();
1255 * Edit a service and bind it with a new scheduler
1258 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1260 struct ip_vs_scheduler *sched, *old_sched;
1264 * Lookup the scheduler, by 'u->sched_name'
1266 sched = ip_vs_scheduler_get(u->sched_name);
1267 if (sched == NULL) {
1268 IP_VS_INFO("Scheduler module ip_vs_%s not found\n",
1274 #ifdef CONFIG_IP_VS_IPV6
1275 if (u->af == AF_INET6) {
1276 if (!sched->supports_ipv6) {
1277 ret = -EAFNOSUPPORT;
1280 if ((u->netmask < 1) || (u->netmask > 128)) {
1287 write_lock_bh(&__ip_vs_svc_lock);
1290 * Wait until all other svc users go away.
1292 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 1);
1295 * Set the flags and timeout value
1297 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1298 svc->timeout = u->timeout * HZ;
1299 svc->netmask = u->netmask;
1301 old_sched = svc->scheduler;
1302 if (sched != old_sched) {
1304 * Unbind the old scheduler
1306 if ((ret = ip_vs_unbind_scheduler(svc))) {
1312 * Bind the new scheduler
1314 if ((ret = ip_vs_bind_scheduler(svc, sched))) {
1316 * If ip_vs_bind_scheduler fails, restore the old
1318 * The main reason of failure is out of memory.
1320 * The question is if the old scheduler can be
1321 * restored all the time. TODO: if it cannot be
1322 * restored some time, we must delete the service,
1323 * otherwise the system may crash.
1325 ip_vs_bind_scheduler(svc, old_sched);
1332 write_unlock_bh(&__ip_vs_svc_lock);
1333 #ifdef CONFIG_IP_VS_IPV6
1338 ip_vs_scheduler_put(old_sched);
1345 * Delete a service from the service list
1346 * - The service must be unlinked, unlocked and not referenced!
1347 * - We are called under _bh lock
1349 static void __ip_vs_del_service(struct ip_vs_service *svc)
1351 struct ip_vs_dest *dest, *nxt;
1352 struct ip_vs_scheduler *old_sched;
1354 /* Count only IPv4 services for old get/setsockopt interface */
1355 if (svc->af == AF_INET)
1356 ip_vs_num_services--;
1358 ip_vs_kill_estimator(&svc->stats);
1360 /* Unbind scheduler */
1361 old_sched = svc->scheduler;
1362 ip_vs_unbind_scheduler(svc);
1364 ip_vs_scheduler_put(old_sched);
1366 /* Unbind app inc */
1368 ip_vs_app_inc_put(svc->inc);
1373 * Unlink the whole destination list
1375 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1376 __ip_vs_unlink_dest(svc, dest, 0);
1377 __ip_vs_del_dest(dest);
1381 * Update the virtual service counters
1383 if (svc->port == FTPPORT)
1384 atomic_dec(&ip_vs_ftpsvc_counter);
1385 else if (svc->port == 0)
1386 atomic_dec(&ip_vs_nullsvc_counter);
1389 * Free the service if nobody refers to it
1391 if (atomic_read(&svc->refcnt) == 0)
1394 /* decrease the module use count */
1395 ip_vs_use_count_dec();
1399 * Delete a service from the service list
1401 static int ip_vs_del_service(struct ip_vs_service *svc)
1407 * Unhash it from the service table
1409 write_lock_bh(&__ip_vs_svc_lock);
1411 ip_vs_svc_unhash(svc);
1414 * Wait until all the svc users go away.
1416 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 1);
1418 __ip_vs_del_service(svc);
1420 write_unlock_bh(&__ip_vs_svc_lock);
1427 * Flush all the virtual services
1429 static int ip_vs_flush(void)
1432 struct ip_vs_service *svc, *nxt;
1435 * Flush the service table hashed by <protocol,addr,port>
1437 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1438 list_for_each_entry_safe(svc, nxt, &ip_vs_svc_table[idx], s_list) {
1439 write_lock_bh(&__ip_vs_svc_lock);
1440 ip_vs_svc_unhash(svc);
1442 * Wait until all the svc users go away.
1444 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1445 __ip_vs_del_service(svc);
1446 write_unlock_bh(&__ip_vs_svc_lock);
1451 * Flush the service table hashed by fwmark
1453 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1454 list_for_each_entry_safe(svc, nxt,
1455 &ip_vs_svc_fwm_table[idx], f_list) {
1456 write_lock_bh(&__ip_vs_svc_lock);
1457 ip_vs_svc_unhash(svc);
1459 * Wait until all the svc users go away.
1461 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1462 __ip_vs_del_service(svc);
1463 write_unlock_bh(&__ip_vs_svc_lock);
1472 * Zero counters in a service or all services
1474 static int ip_vs_zero_service(struct ip_vs_service *svc)
1476 struct ip_vs_dest *dest;
1478 write_lock_bh(&__ip_vs_svc_lock);
1479 list_for_each_entry(dest, &svc->destinations, n_list) {
1480 ip_vs_zero_stats(&dest->stats);
1482 ip_vs_zero_stats(&svc->stats);
1483 write_unlock_bh(&__ip_vs_svc_lock);
1487 static int ip_vs_zero_all(void)
1490 struct ip_vs_service *svc;
1492 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1493 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1494 ip_vs_zero_service(svc);
1498 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1499 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1500 ip_vs_zero_service(svc);
1504 ip_vs_zero_stats(&ip_vs_stats);
1510 proc_do_defense_mode(ctl_table *table, int write, struct file * filp,
1511 void __user *buffer, size_t *lenp, loff_t *ppos)
1513 int *valp = table->data;
1517 rc = proc_dointvec(table, write, filp, buffer, lenp, ppos);
1518 if (write && (*valp != val)) {
1519 if ((*valp < 0) || (*valp > 3)) {
1520 /* Restore the correct value */
1523 update_defense_level();
1531 proc_do_sync_threshold(ctl_table *table, int write, struct file *filp,
1532 void __user *buffer, size_t *lenp, loff_t *ppos)
1534 int *valp = table->data;
1538 /* backup the value first */
1539 memcpy(val, valp, sizeof(val));
1541 rc = proc_dointvec(table, write, filp, buffer, lenp, ppos);
1542 if (write && (valp[0] < 0 || valp[1] < 0 || valp[0] >= valp[1])) {
1543 /* Restore the correct value */
1544 memcpy(valp, val, sizeof(val));
1551 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1554 static struct ctl_table vs_vars[] = {
1556 .procname = "amemthresh",
1557 .data = &sysctl_ip_vs_amemthresh,
1558 .maxlen = sizeof(int),
1560 .proc_handler = &proc_dointvec,
1562 #ifdef CONFIG_IP_VS_DEBUG
1564 .procname = "debug_level",
1565 .data = &sysctl_ip_vs_debug_level,
1566 .maxlen = sizeof(int),
1568 .proc_handler = &proc_dointvec,
1572 .procname = "am_droprate",
1573 .data = &sysctl_ip_vs_am_droprate,
1574 .maxlen = sizeof(int),
1576 .proc_handler = &proc_dointvec,
1579 .procname = "drop_entry",
1580 .data = &sysctl_ip_vs_drop_entry,
1581 .maxlen = sizeof(int),
1583 .proc_handler = &proc_do_defense_mode,
1586 .procname = "drop_packet",
1587 .data = &sysctl_ip_vs_drop_packet,
1588 .maxlen = sizeof(int),
1590 .proc_handler = &proc_do_defense_mode,
1593 .procname = "secure_tcp",
1594 .data = &sysctl_ip_vs_secure_tcp,
1595 .maxlen = sizeof(int),
1597 .proc_handler = &proc_do_defense_mode,
1601 .procname = "timeout_established",
1602 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1603 .maxlen = sizeof(int),
1605 .proc_handler = &proc_dointvec_jiffies,
1608 .procname = "timeout_synsent",
1609 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1610 .maxlen = sizeof(int),
1612 .proc_handler = &proc_dointvec_jiffies,
1615 .procname = "timeout_synrecv",
1616 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1617 .maxlen = sizeof(int),
1619 .proc_handler = &proc_dointvec_jiffies,
1622 .procname = "timeout_finwait",
1623 .data = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1624 .maxlen = sizeof(int),
1626 .proc_handler = &proc_dointvec_jiffies,
1629 .procname = "timeout_timewait",
1630 .data = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1631 .maxlen = sizeof(int),
1633 .proc_handler = &proc_dointvec_jiffies,
1636 .procname = "timeout_close",
1637 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1638 .maxlen = sizeof(int),
1640 .proc_handler = &proc_dointvec_jiffies,
1643 .procname = "timeout_closewait",
1644 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1645 .maxlen = sizeof(int),
1647 .proc_handler = &proc_dointvec_jiffies,
1650 .procname = "timeout_lastack",
1651 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1652 .maxlen = sizeof(int),
1654 .proc_handler = &proc_dointvec_jiffies,
1657 .procname = "timeout_listen",
1658 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1659 .maxlen = sizeof(int),
1661 .proc_handler = &proc_dointvec_jiffies,
1664 .procname = "timeout_synack",
1665 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1666 .maxlen = sizeof(int),
1668 .proc_handler = &proc_dointvec_jiffies,
1671 .procname = "timeout_udp",
1672 .data = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1673 .maxlen = sizeof(int),
1675 .proc_handler = &proc_dointvec_jiffies,
1678 .procname = "timeout_icmp",
1679 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1680 .maxlen = sizeof(int),
1682 .proc_handler = &proc_dointvec_jiffies,
1686 .procname = "cache_bypass",
1687 .data = &sysctl_ip_vs_cache_bypass,
1688 .maxlen = sizeof(int),
1690 .proc_handler = &proc_dointvec,
1693 .procname = "expire_nodest_conn",
1694 .data = &sysctl_ip_vs_expire_nodest_conn,
1695 .maxlen = sizeof(int),
1697 .proc_handler = &proc_dointvec,
1700 .procname = "expire_quiescent_template",
1701 .data = &sysctl_ip_vs_expire_quiescent_template,
1702 .maxlen = sizeof(int),
1704 .proc_handler = &proc_dointvec,
1707 .procname = "sync_threshold",
1708 .data = &sysctl_ip_vs_sync_threshold,
1709 .maxlen = sizeof(sysctl_ip_vs_sync_threshold),
1711 .proc_handler = &proc_do_sync_threshold,
1714 .procname = "nat_icmp_send",
1715 .data = &sysctl_ip_vs_nat_icmp_send,
1716 .maxlen = sizeof(int),
1718 .proc_handler = &proc_dointvec,
1723 const struct ctl_path net_vs_ctl_path[] = {
1724 { .procname = "net", .ctl_name = CTL_NET, },
1725 { .procname = "ipv4", .ctl_name = NET_IPV4, },
1726 { .procname = "vs", },
1729 EXPORT_SYMBOL_GPL(net_vs_ctl_path);
1731 static struct ctl_table_header * sysctl_header;
1733 #ifdef CONFIG_PROC_FS
1736 struct list_head *table;
1741 * Write the contents of the VS rule table to a PROCfs file.
1742 * (It is kept just for backward compatibility)
1744 static inline const char *ip_vs_fwd_name(unsigned flags)
1746 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1747 case IP_VS_CONN_F_LOCALNODE:
1749 case IP_VS_CONN_F_TUNNEL:
1751 case IP_VS_CONN_F_DROUTE:
1759 /* Get the Nth entry in the two lists */
1760 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1762 struct ip_vs_iter *iter = seq->private;
1764 struct ip_vs_service *svc;
1766 /* look in hash by protocol */
1767 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1768 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1770 iter->table = ip_vs_svc_table;
1777 /* keep looking in fwmark */
1778 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1779 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1781 iter->table = ip_vs_svc_fwm_table;
1791 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1792 __acquires(__ip_vs_svc_lock)
1795 read_lock_bh(&__ip_vs_svc_lock);
1796 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1800 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1802 struct list_head *e;
1803 struct ip_vs_iter *iter;
1804 struct ip_vs_service *svc;
1807 if (v == SEQ_START_TOKEN)
1808 return ip_vs_info_array(seq,0);
1811 iter = seq->private;
1813 if (iter->table == ip_vs_svc_table) {
1814 /* next service in table hashed by protocol */
1815 if ((e = svc->s_list.next) != &ip_vs_svc_table[iter->bucket])
1816 return list_entry(e, struct ip_vs_service, s_list);
1819 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1820 list_for_each_entry(svc,&ip_vs_svc_table[iter->bucket],
1826 iter->table = ip_vs_svc_fwm_table;
1831 /* next service in hashed by fwmark */
1832 if ((e = svc->f_list.next) != &ip_vs_svc_fwm_table[iter->bucket])
1833 return list_entry(e, struct ip_vs_service, f_list);
1836 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1837 list_for_each_entry(svc, &ip_vs_svc_fwm_table[iter->bucket],
1845 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
1846 __releases(__ip_vs_svc_lock)
1848 read_unlock_bh(&__ip_vs_svc_lock);
1852 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
1854 if (v == SEQ_START_TOKEN) {
1856 "IP Virtual Server version %d.%d.%d (size=%d)\n",
1857 NVERSION(IP_VS_VERSION_CODE), IP_VS_CONN_TAB_SIZE);
1859 "Prot LocalAddress:Port Scheduler Flags\n");
1861 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
1863 const struct ip_vs_service *svc = v;
1864 const struct ip_vs_iter *iter = seq->private;
1865 const struct ip_vs_dest *dest;
1867 if (iter->table == ip_vs_svc_table) {
1868 #ifdef CONFIG_IP_VS_IPV6
1869 if (svc->af == AF_INET6)
1870 seq_printf(seq, "%s [%pI6]:%04X %s ",
1871 ip_vs_proto_name(svc->protocol),
1874 svc->scheduler->name);
1877 seq_printf(seq, "%s %08X:%04X %s ",
1878 ip_vs_proto_name(svc->protocol),
1879 ntohl(svc->addr.ip),
1881 svc->scheduler->name);
1883 seq_printf(seq, "FWM %08X %s ",
1884 svc->fwmark, svc->scheduler->name);
1887 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
1888 seq_printf(seq, "persistent %d %08X\n",
1890 ntohl(svc->netmask));
1892 seq_putc(seq, '\n');
1894 list_for_each_entry(dest, &svc->destinations, n_list) {
1895 #ifdef CONFIG_IP_VS_IPV6
1896 if (dest->af == AF_INET6)
1899 " %-7s %-6d %-10d %-10d\n",
1902 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
1903 atomic_read(&dest->weight),
1904 atomic_read(&dest->activeconns),
1905 atomic_read(&dest->inactconns));
1910 "%-7s %-6d %-10d %-10d\n",
1911 ntohl(dest->addr.ip),
1913 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
1914 atomic_read(&dest->weight),
1915 atomic_read(&dest->activeconns),
1916 atomic_read(&dest->inactconns));
1923 static const struct seq_operations ip_vs_info_seq_ops = {
1924 .start = ip_vs_info_seq_start,
1925 .next = ip_vs_info_seq_next,
1926 .stop = ip_vs_info_seq_stop,
1927 .show = ip_vs_info_seq_show,
1930 static int ip_vs_info_open(struct inode *inode, struct file *file)
1932 return seq_open_private(file, &ip_vs_info_seq_ops,
1933 sizeof(struct ip_vs_iter));
1936 static const struct file_operations ip_vs_info_fops = {
1937 .owner = THIS_MODULE,
1938 .open = ip_vs_info_open,
1940 .llseek = seq_lseek,
1941 .release = seq_release_private,
1946 struct ip_vs_stats ip_vs_stats = {
1947 .lock = __SPIN_LOCK_UNLOCKED(ip_vs_stats.lock),
1950 #ifdef CONFIG_PROC_FS
1951 static int ip_vs_stats_show(struct seq_file *seq, void *v)
1954 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
1956 " Total Incoming Outgoing Incoming Outgoing\n");
1958 " Conns Packets Packets Bytes Bytes\n");
1960 spin_lock_bh(&ip_vs_stats.lock);
1961 seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", ip_vs_stats.ustats.conns,
1962 ip_vs_stats.ustats.inpkts, ip_vs_stats.ustats.outpkts,
1963 (unsigned long long) ip_vs_stats.ustats.inbytes,
1964 (unsigned long long) ip_vs_stats.ustats.outbytes);
1966 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
1968 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
1969 seq_printf(seq,"%8X %8X %8X %16X %16X\n",
1970 ip_vs_stats.ustats.cps,
1971 ip_vs_stats.ustats.inpps,
1972 ip_vs_stats.ustats.outpps,
1973 ip_vs_stats.ustats.inbps,
1974 ip_vs_stats.ustats.outbps);
1975 spin_unlock_bh(&ip_vs_stats.lock);
1980 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
1982 return single_open(file, ip_vs_stats_show, NULL);
1985 static const struct file_operations ip_vs_stats_fops = {
1986 .owner = THIS_MODULE,
1987 .open = ip_vs_stats_seq_open,
1989 .llseek = seq_lseek,
1990 .release = single_release,
1996 * Set timeout values for tcp tcpfin udp in the timeout_table.
1998 static int ip_vs_set_timeout(struct ip_vs_timeout_user *u)
2000 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2005 #ifdef CONFIG_IP_VS_PROTO_TCP
2006 if (u->tcp_timeout) {
2007 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_ESTABLISHED]
2008 = u->tcp_timeout * HZ;
2011 if (u->tcp_fin_timeout) {
2012 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_FIN_WAIT]
2013 = u->tcp_fin_timeout * HZ;
2017 #ifdef CONFIG_IP_VS_PROTO_UDP
2018 if (u->udp_timeout) {
2019 ip_vs_protocol_udp.timeout_table[IP_VS_UDP_S_NORMAL]
2020 = u->udp_timeout * HZ;
2027 #define SET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2028 #define SERVICE_ARG_LEN (sizeof(struct ip_vs_service_user))
2029 #define SVCDEST_ARG_LEN (sizeof(struct ip_vs_service_user) + \
2030 sizeof(struct ip_vs_dest_user))
2031 #define TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2032 #define DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user))
2033 #define MAX_ARG_LEN SVCDEST_ARG_LEN
2035 static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
2036 [SET_CMDID(IP_VS_SO_SET_ADD)] = SERVICE_ARG_LEN,
2037 [SET_CMDID(IP_VS_SO_SET_EDIT)] = SERVICE_ARG_LEN,
2038 [SET_CMDID(IP_VS_SO_SET_DEL)] = SERVICE_ARG_LEN,
2039 [SET_CMDID(IP_VS_SO_SET_FLUSH)] = 0,
2040 [SET_CMDID(IP_VS_SO_SET_ADDDEST)] = SVCDEST_ARG_LEN,
2041 [SET_CMDID(IP_VS_SO_SET_DELDEST)] = SVCDEST_ARG_LEN,
2042 [SET_CMDID(IP_VS_SO_SET_EDITDEST)] = SVCDEST_ARG_LEN,
2043 [SET_CMDID(IP_VS_SO_SET_TIMEOUT)] = TIMEOUT_ARG_LEN,
2044 [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)] = DAEMON_ARG_LEN,
2045 [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)] = DAEMON_ARG_LEN,
2046 [SET_CMDID(IP_VS_SO_SET_ZERO)] = SERVICE_ARG_LEN,
2049 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2050 struct ip_vs_service_user *usvc_compat)
2053 usvc->protocol = usvc_compat->protocol;
2054 usvc->addr.ip = usvc_compat->addr;
2055 usvc->port = usvc_compat->port;
2056 usvc->fwmark = usvc_compat->fwmark;
2058 /* Deep copy of sched_name is not needed here */
2059 usvc->sched_name = usvc_compat->sched_name;
2061 usvc->flags = usvc_compat->flags;
2062 usvc->timeout = usvc_compat->timeout;
2063 usvc->netmask = usvc_compat->netmask;
2066 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2067 struct ip_vs_dest_user *udest_compat)
2069 udest->addr.ip = udest_compat->addr;
2070 udest->port = udest_compat->port;
2071 udest->conn_flags = udest_compat->conn_flags;
2072 udest->weight = udest_compat->weight;
2073 udest->u_threshold = udest_compat->u_threshold;
2074 udest->l_threshold = udest_compat->l_threshold;
2078 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2081 unsigned char arg[MAX_ARG_LEN];
2082 struct ip_vs_service_user *usvc_compat;
2083 struct ip_vs_service_user_kern usvc;
2084 struct ip_vs_service *svc;
2085 struct ip_vs_dest_user *udest_compat;
2086 struct ip_vs_dest_user_kern udest;
2088 if (!capable(CAP_NET_ADMIN))
2091 if (len != set_arglen[SET_CMDID(cmd)]) {
2092 IP_VS_ERR("set_ctl: len %u != %u\n",
2093 len, set_arglen[SET_CMDID(cmd)]);
2097 if (copy_from_user(arg, user, len) != 0)
2100 /* increase the module use count */
2101 ip_vs_use_count_inc();
2103 if (mutex_lock_interruptible(&__ip_vs_mutex)) {
2108 if (cmd == IP_VS_SO_SET_FLUSH) {
2109 /* Flush the virtual service */
2110 ret = ip_vs_flush();
2112 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2113 /* Set timeout values for (tcp tcpfin udp) */
2114 ret = ip_vs_set_timeout((struct ip_vs_timeout_user *)arg);
2116 } else if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2117 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2118 ret = start_sync_thread(dm->state, dm->mcast_ifn, dm->syncid);
2120 } else if (cmd == IP_VS_SO_SET_STOPDAEMON) {
2121 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2122 ret = stop_sync_thread(dm->state);
2126 usvc_compat = (struct ip_vs_service_user *)arg;
2127 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2129 /* We only use the new structs internally, so copy userspace compat
2130 * structs to extended internal versions */
2131 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2132 ip_vs_copy_udest_compat(&udest, udest_compat);
2134 if (cmd == IP_VS_SO_SET_ZERO) {
2135 /* if no service address is set, zero counters in all */
2136 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2137 ret = ip_vs_zero_all();
2142 /* Check for valid protocol: TCP or UDP, even for fwmark!=0 */
2143 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP) {
2144 IP_VS_ERR("set_ctl: invalid protocol: %d %d.%d.%d.%d:%d %s\n",
2145 usvc.protocol, NIPQUAD(usvc.addr.ip),
2146 ntohs(usvc.port), usvc.sched_name);
2151 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2152 if (usvc.fwmark == 0)
2153 svc = __ip_vs_service_get(usvc.af, usvc.protocol,
2154 &usvc.addr, usvc.port);
2156 svc = __ip_vs_svc_fwm_get(usvc.af, usvc.fwmark);
2158 if (cmd != IP_VS_SO_SET_ADD
2159 && (svc == NULL || svc->protocol != usvc.protocol)) {
2165 case IP_VS_SO_SET_ADD:
2169 ret = ip_vs_add_service(&usvc, &svc);
2171 case IP_VS_SO_SET_EDIT:
2172 ret = ip_vs_edit_service(svc, &usvc);
2174 case IP_VS_SO_SET_DEL:
2175 ret = ip_vs_del_service(svc);
2179 case IP_VS_SO_SET_ZERO:
2180 ret = ip_vs_zero_service(svc);
2182 case IP_VS_SO_SET_ADDDEST:
2183 ret = ip_vs_add_dest(svc, &udest);
2185 case IP_VS_SO_SET_EDITDEST:
2186 ret = ip_vs_edit_dest(svc, &udest);
2188 case IP_VS_SO_SET_DELDEST:
2189 ret = ip_vs_del_dest(svc, &udest);
2196 ip_vs_service_put(svc);
2199 mutex_unlock(&__ip_vs_mutex);
2201 /* decrease the module use count */
2202 ip_vs_use_count_dec();
2209 ip_vs_copy_stats(struct ip_vs_stats_user *dst, struct ip_vs_stats *src)
2211 spin_lock_bh(&src->lock);
2212 memcpy(dst, &src->ustats, sizeof(*dst));
2213 spin_unlock_bh(&src->lock);
2217 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2219 dst->protocol = src->protocol;
2220 dst->addr = src->addr.ip;
2221 dst->port = src->port;
2222 dst->fwmark = src->fwmark;
2223 strlcpy(dst->sched_name, src->scheduler->name, sizeof(dst->sched_name));
2224 dst->flags = src->flags;
2225 dst->timeout = src->timeout / HZ;
2226 dst->netmask = src->netmask;
2227 dst->num_dests = src->num_dests;
2228 ip_vs_copy_stats(&dst->stats, &src->stats);
2232 __ip_vs_get_service_entries(const struct ip_vs_get_services *get,
2233 struct ip_vs_get_services __user *uptr)
2236 struct ip_vs_service *svc;
2237 struct ip_vs_service_entry entry;
2240 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2241 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2242 /* Only expose IPv4 entries to old interface */
2243 if (svc->af != AF_INET)
2246 if (count >= get->num_services)
2248 memset(&entry, 0, sizeof(entry));
2249 ip_vs_copy_service(&entry, svc);
2250 if (copy_to_user(&uptr->entrytable[count],
2251 &entry, sizeof(entry))) {
2259 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2260 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2261 /* Only expose IPv4 entries to old interface */
2262 if (svc->af != AF_INET)
2265 if (count >= get->num_services)
2267 memset(&entry, 0, sizeof(entry));
2268 ip_vs_copy_service(&entry, svc);
2269 if (copy_to_user(&uptr->entrytable[count],
2270 &entry, sizeof(entry))) {
2282 __ip_vs_get_dest_entries(const struct ip_vs_get_dests *get,
2283 struct ip_vs_get_dests __user *uptr)
2285 struct ip_vs_service *svc;
2286 union nf_inet_addr addr = { .ip = get->addr };
2290 svc = __ip_vs_svc_fwm_get(AF_INET, get->fwmark);
2292 svc = __ip_vs_service_get(AF_INET, get->protocol, &addr,
2297 struct ip_vs_dest *dest;
2298 struct ip_vs_dest_entry entry;
2300 list_for_each_entry(dest, &svc->destinations, n_list) {
2301 if (count >= get->num_dests)
2304 entry.addr = dest->addr.ip;
2305 entry.port = dest->port;
2306 entry.conn_flags = atomic_read(&dest->conn_flags);
2307 entry.weight = atomic_read(&dest->weight);
2308 entry.u_threshold = dest->u_threshold;
2309 entry.l_threshold = dest->l_threshold;
2310 entry.activeconns = atomic_read(&dest->activeconns);
2311 entry.inactconns = atomic_read(&dest->inactconns);
2312 entry.persistconns = atomic_read(&dest->persistconns);
2313 ip_vs_copy_stats(&entry.stats, &dest->stats);
2314 if (copy_to_user(&uptr->entrytable[count],
2315 &entry, sizeof(entry))) {
2321 ip_vs_service_put(svc);
2328 __ip_vs_get_timeouts(struct ip_vs_timeout_user *u)
2330 #ifdef CONFIG_IP_VS_PROTO_TCP
2332 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2333 u->tcp_fin_timeout =
2334 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2336 #ifdef CONFIG_IP_VS_PROTO_UDP
2338 ip_vs_protocol_udp.timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2343 #define GET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2344 #define GET_INFO_ARG_LEN (sizeof(struct ip_vs_getinfo))
2345 #define GET_SERVICES_ARG_LEN (sizeof(struct ip_vs_get_services))
2346 #define GET_SERVICE_ARG_LEN (sizeof(struct ip_vs_service_entry))
2347 #define GET_DESTS_ARG_LEN (sizeof(struct ip_vs_get_dests))
2348 #define GET_TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2349 #define GET_DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user) * 2)
2351 static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
2352 [GET_CMDID(IP_VS_SO_GET_VERSION)] = 64,
2353 [GET_CMDID(IP_VS_SO_GET_INFO)] = GET_INFO_ARG_LEN,
2354 [GET_CMDID(IP_VS_SO_GET_SERVICES)] = GET_SERVICES_ARG_LEN,
2355 [GET_CMDID(IP_VS_SO_GET_SERVICE)] = GET_SERVICE_ARG_LEN,
2356 [GET_CMDID(IP_VS_SO_GET_DESTS)] = GET_DESTS_ARG_LEN,
2357 [GET_CMDID(IP_VS_SO_GET_TIMEOUT)] = GET_TIMEOUT_ARG_LEN,
2358 [GET_CMDID(IP_VS_SO_GET_DAEMON)] = GET_DAEMON_ARG_LEN,
2362 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2364 unsigned char arg[128];
2367 if (!capable(CAP_NET_ADMIN))
2370 if (*len < get_arglen[GET_CMDID(cmd)]) {
2371 IP_VS_ERR("get_ctl: len %u < %u\n",
2372 *len, get_arglen[GET_CMDID(cmd)]);
2376 if (copy_from_user(arg, user, get_arglen[GET_CMDID(cmd)]) != 0)
2379 if (mutex_lock_interruptible(&__ip_vs_mutex))
2380 return -ERESTARTSYS;
2383 case IP_VS_SO_GET_VERSION:
2387 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2388 NVERSION(IP_VS_VERSION_CODE), IP_VS_CONN_TAB_SIZE);
2389 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2393 *len = strlen(buf)+1;
2397 case IP_VS_SO_GET_INFO:
2399 struct ip_vs_getinfo info;
2400 info.version = IP_VS_VERSION_CODE;
2401 info.size = IP_VS_CONN_TAB_SIZE;
2402 info.num_services = ip_vs_num_services;
2403 if (copy_to_user(user, &info, sizeof(info)) != 0)
2408 case IP_VS_SO_GET_SERVICES:
2410 struct ip_vs_get_services *get;
2413 get = (struct ip_vs_get_services *)arg;
2414 size = sizeof(*get) +
2415 sizeof(struct ip_vs_service_entry) * get->num_services;
2417 IP_VS_ERR("length: %u != %u\n", *len, size);
2421 ret = __ip_vs_get_service_entries(get, user);
2425 case IP_VS_SO_GET_SERVICE:
2427 struct ip_vs_service_entry *entry;
2428 struct ip_vs_service *svc;
2429 union nf_inet_addr addr;
2431 entry = (struct ip_vs_service_entry *)arg;
2432 addr.ip = entry->addr;
2434 svc = __ip_vs_svc_fwm_get(AF_INET, entry->fwmark);
2436 svc = __ip_vs_service_get(AF_INET, entry->protocol,
2437 &addr, entry->port);
2439 ip_vs_copy_service(entry, svc);
2440 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2442 ip_vs_service_put(svc);
2448 case IP_VS_SO_GET_DESTS:
2450 struct ip_vs_get_dests *get;
2453 get = (struct ip_vs_get_dests *)arg;
2454 size = sizeof(*get) +
2455 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2457 IP_VS_ERR("length: %u != %u\n", *len, size);
2461 ret = __ip_vs_get_dest_entries(get, user);
2465 case IP_VS_SO_GET_TIMEOUT:
2467 struct ip_vs_timeout_user t;
2469 __ip_vs_get_timeouts(&t);
2470 if (copy_to_user(user, &t, sizeof(t)) != 0)
2475 case IP_VS_SO_GET_DAEMON:
2477 struct ip_vs_daemon_user d[2];
2479 memset(&d, 0, sizeof(d));
2480 if (ip_vs_sync_state & IP_VS_STATE_MASTER) {
2481 d[0].state = IP_VS_STATE_MASTER;
2482 strlcpy(d[0].mcast_ifn, ip_vs_master_mcast_ifn, sizeof(d[0].mcast_ifn));
2483 d[0].syncid = ip_vs_master_syncid;
2485 if (ip_vs_sync_state & IP_VS_STATE_BACKUP) {
2486 d[1].state = IP_VS_STATE_BACKUP;
2487 strlcpy(d[1].mcast_ifn, ip_vs_backup_mcast_ifn, sizeof(d[1].mcast_ifn));
2488 d[1].syncid = ip_vs_backup_syncid;
2490 if (copy_to_user(user, &d, sizeof(d)) != 0)
2500 mutex_unlock(&__ip_vs_mutex);
2505 static struct nf_sockopt_ops ip_vs_sockopts = {
2507 .set_optmin = IP_VS_BASE_CTL,
2508 .set_optmax = IP_VS_SO_SET_MAX+1,
2509 .set = do_ip_vs_set_ctl,
2510 .get_optmin = IP_VS_BASE_CTL,
2511 .get_optmax = IP_VS_SO_GET_MAX+1,
2512 .get = do_ip_vs_get_ctl,
2513 .owner = THIS_MODULE,
2517 * Generic Netlink interface
2520 /* IPVS genetlink family */
2521 static struct genl_family ip_vs_genl_family = {
2522 .id = GENL_ID_GENERATE,
2524 .name = IPVS_GENL_NAME,
2525 .version = IPVS_GENL_VERSION,
2526 .maxattr = IPVS_CMD_MAX,
2529 /* Policy used for first-level command attributes */
2530 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2531 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2532 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2533 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2534 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2535 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2536 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2539 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2540 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2541 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2542 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2543 .len = IP_VS_IFNAME_MAXLEN },
2544 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2547 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2548 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2549 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2550 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2551 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2552 .len = sizeof(union nf_inet_addr) },
2553 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2554 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2555 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2556 .len = IP_VS_SCHEDNAME_MAXLEN },
2557 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2558 .len = sizeof(struct ip_vs_flags) },
2559 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2560 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2561 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2564 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2565 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2566 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2567 .len = sizeof(union nf_inet_addr) },
2568 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2569 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2570 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2571 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2572 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2573 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2574 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2575 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2576 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2579 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2580 struct ip_vs_stats *stats)
2582 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2586 spin_lock_bh(&stats->lock);
2588 NLA_PUT_U32(skb, IPVS_STATS_ATTR_CONNS, stats->ustats.conns);
2589 NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPKTS, stats->ustats.inpkts);
2590 NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPKTS, stats->ustats.outpkts);
2591 NLA_PUT_U64(skb, IPVS_STATS_ATTR_INBYTES, stats->ustats.inbytes);
2592 NLA_PUT_U64(skb, IPVS_STATS_ATTR_OUTBYTES, stats->ustats.outbytes);
2593 NLA_PUT_U32(skb, IPVS_STATS_ATTR_CPS, stats->ustats.cps);
2594 NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPPS, stats->ustats.inpps);
2595 NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPPS, stats->ustats.outpps);
2596 NLA_PUT_U32(skb, IPVS_STATS_ATTR_INBPS, stats->ustats.inbps);
2597 NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTBPS, stats->ustats.outbps);
2599 spin_unlock_bh(&stats->lock);
2601 nla_nest_end(skb, nl_stats);
2606 spin_unlock_bh(&stats->lock);
2607 nla_nest_cancel(skb, nl_stats);
2611 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2612 struct ip_vs_service *svc)
2614 struct nlattr *nl_service;
2615 struct ip_vs_flags flags = { .flags = svc->flags,
2618 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2622 NLA_PUT_U16(skb, IPVS_SVC_ATTR_AF, svc->af);
2625 NLA_PUT_U32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark);
2627 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol);
2628 NLA_PUT(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr);
2629 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PORT, svc->port);
2632 NLA_PUT_STRING(skb, IPVS_SVC_ATTR_SCHED_NAME, svc->scheduler->name);
2633 NLA_PUT(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags);
2634 NLA_PUT_U32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ);
2635 NLA_PUT_U32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask);
2637 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2638 goto nla_put_failure;
2640 nla_nest_end(skb, nl_service);
2645 nla_nest_cancel(skb, nl_service);
2649 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2650 struct ip_vs_service *svc,
2651 struct netlink_callback *cb)
2655 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2656 &ip_vs_genl_family, NLM_F_MULTI,
2657 IPVS_CMD_NEW_SERVICE);
2661 if (ip_vs_genl_fill_service(skb, svc) < 0)
2662 goto nla_put_failure;
2664 return genlmsg_end(skb, hdr);
2667 genlmsg_cancel(skb, hdr);
2671 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2672 struct netlink_callback *cb)
2675 int start = cb->args[0];
2676 struct ip_vs_service *svc;
2678 mutex_lock(&__ip_vs_mutex);
2679 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2680 list_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2683 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2685 goto nla_put_failure;
2690 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2691 list_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
2694 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2696 goto nla_put_failure;
2702 mutex_unlock(&__ip_vs_mutex);
2708 static int ip_vs_genl_parse_service(struct ip_vs_service_user_kern *usvc,
2709 struct nlattr *nla, int full_entry)
2711 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
2712 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
2714 /* Parse mandatory identifying service fields first */
2716 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
2719 nla_af = attrs[IPVS_SVC_ATTR_AF];
2720 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
2721 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
2722 nla_port = attrs[IPVS_SVC_ATTR_PORT];
2723 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
2725 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
2728 usvc->af = nla_get_u16(nla_af);
2729 #ifdef CONFIG_IP_VS_IPV6
2730 if (usvc->af != AF_INET && usvc->af != AF_INET6)
2732 if (usvc->af != AF_INET)
2734 return -EAFNOSUPPORT;
2737 usvc->protocol = IPPROTO_TCP;
2738 usvc->fwmark = nla_get_u32(nla_fwmark);
2740 usvc->protocol = nla_get_u16(nla_protocol);
2741 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
2742 usvc->port = nla_get_u16(nla_port);
2746 /* If a full entry was requested, check for the additional fields */
2748 struct nlattr *nla_sched, *nla_flags, *nla_timeout,
2750 struct ip_vs_flags flags;
2751 struct ip_vs_service *svc;
2753 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
2754 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
2755 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
2756 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
2758 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
2761 nla_memcpy(&flags, nla_flags, sizeof(flags));
2763 /* prefill flags from service if it already exists */
2765 svc = __ip_vs_svc_fwm_get(usvc->af, usvc->fwmark);
2767 svc = __ip_vs_service_get(usvc->af, usvc->protocol,
2768 &usvc->addr, usvc->port);
2770 usvc->flags = svc->flags;
2771 ip_vs_service_put(svc);
2775 /* set new flags from userland */
2776 usvc->flags = (usvc->flags & ~flags.mask) |
2777 (flags.flags & flags.mask);
2778 usvc->sched_name = nla_data(nla_sched);
2779 usvc->timeout = nla_get_u32(nla_timeout);
2780 usvc->netmask = nla_get_u32(nla_netmask);
2786 static struct ip_vs_service *ip_vs_genl_find_service(struct nlattr *nla)
2788 struct ip_vs_service_user_kern usvc;
2791 ret = ip_vs_genl_parse_service(&usvc, nla, 0);
2793 return ERR_PTR(ret);
2796 return __ip_vs_svc_fwm_get(usvc.af, usvc.fwmark);
2798 return __ip_vs_service_get(usvc.af, usvc.protocol,
2799 &usvc.addr, usvc.port);
2802 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
2804 struct nlattr *nl_dest;
2806 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
2810 NLA_PUT(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr);
2811 NLA_PUT_U16(skb, IPVS_DEST_ATTR_PORT, dest->port);
2813 NLA_PUT_U32(skb, IPVS_DEST_ATTR_FWD_METHOD,
2814 atomic_read(&dest->conn_flags) & IP_VS_CONN_F_FWD_MASK);
2815 NLA_PUT_U32(skb, IPVS_DEST_ATTR_WEIGHT, atomic_read(&dest->weight));
2816 NLA_PUT_U32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold);
2817 NLA_PUT_U32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold);
2818 NLA_PUT_U32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
2819 atomic_read(&dest->activeconns));
2820 NLA_PUT_U32(skb, IPVS_DEST_ATTR_INACT_CONNS,
2821 atomic_read(&dest->inactconns));
2822 NLA_PUT_U32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
2823 atomic_read(&dest->persistconns));
2825 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
2826 goto nla_put_failure;
2828 nla_nest_end(skb, nl_dest);
2833 nla_nest_cancel(skb, nl_dest);
2837 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
2838 struct netlink_callback *cb)
2842 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2843 &ip_vs_genl_family, NLM_F_MULTI,
2848 if (ip_vs_genl_fill_dest(skb, dest) < 0)
2849 goto nla_put_failure;
2851 return genlmsg_end(skb, hdr);
2854 genlmsg_cancel(skb, hdr);
2858 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
2859 struct netlink_callback *cb)
2862 int start = cb->args[0];
2863 struct ip_vs_service *svc;
2864 struct ip_vs_dest *dest;
2865 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
2867 mutex_lock(&__ip_vs_mutex);
2869 /* Try to find the service for which to dump destinations */
2870 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
2871 IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
2874 svc = ip_vs_genl_find_service(attrs[IPVS_CMD_ATTR_SERVICE]);
2875 if (IS_ERR(svc) || svc == NULL)
2878 /* Dump the destinations */
2879 list_for_each_entry(dest, &svc->destinations, n_list) {
2882 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
2884 goto nla_put_failure;
2890 ip_vs_service_put(svc);
2893 mutex_unlock(&__ip_vs_mutex);
2898 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
2899 struct nlattr *nla, int full_entry)
2901 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
2902 struct nlattr *nla_addr, *nla_port;
2904 /* Parse mandatory identifying destination fields first */
2906 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
2909 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
2910 nla_port = attrs[IPVS_DEST_ATTR_PORT];
2912 if (!(nla_addr && nla_port))
2915 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
2916 udest->port = nla_get_u16(nla_port);
2918 /* If a full entry was requested, check for the additional fields */
2920 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
2923 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
2924 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
2925 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
2926 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
2928 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
2931 udest->conn_flags = nla_get_u32(nla_fwd)
2932 & IP_VS_CONN_F_FWD_MASK;
2933 udest->weight = nla_get_u32(nla_weight);
2934 udest->u_threshold = nla_get_u32(nla_u_thresh);
2935 udest->l_threshold = nla_get_u32(nla_l_thresh);
2941 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __be32 state,
2942 const char *mcast_ifn, __be32 syncid)
2944 struct nlattr *nl_daemon;
2946 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
2950 NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_STATE, state);
2951 NLA_PUT_STRING(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn);
2952 NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid);
2954 nla_nest_end(skb, nl_daemon);
2959 nla_nest_cancel(skb, nl_daemon);
2963 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __be32 state,
2964 const char *mcast_ifn, __be32 syncid,
2965 struct netlink_callback *cb)
2968 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2969 &ip_vs_genl_family, NLM_F_MULTI,
2970 IPVS_CMD_NEW_DAEMON);
2974 if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
2975 goto nla_put_failure;
2977 return genlmsg_end(skb, hdr);
2980 genlmsg_cancel(skb, hdr);
2984 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
2985 struct netlink_callback *cb)
2987 mutex_lock(&__ip_vs_mutex);
2988 if ((ip_vs_sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
2989 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
2990 ip_vs_master_mcast_ifn,
2991 ip_vs_master_syncid, cb) < 0)
2992 goto nla_put_failure;
2997 if ((ip_vs_sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
2998 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
2999 ip_vs_backup_mcast_ifn,
3000 ip_vs_backup_syncid, cb) < 0)
3001 goto nla_put_failure;
3007 mutex_unlock(&__ip_vs_mutex);
3012 static int ip_vs_genl_new_daemon(struct nlattr **attrs)
3014 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3015 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3016 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3019 return start_sync_thread(nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3020 nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3021 nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3024 static int ip_vs_genl_del_daemon(struct nlattr **attrs)
3026 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3029 return stop_sync_thread(nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3032 static int ip_vs_genl_set_config(struct nlattr **attrs)
3034 struct ip_vs_timeout_user t;
3036 __ip_vs_get_timeouts(&t);
3038 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3039 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3041 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3043 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3045 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3046 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3048 return ip_vs_set_timeout(&t);
3051 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3053 struct ip_vs_service *svc = NULL;
3054 struct ip_vs_service_user_kern usvc;
3055 struct ip_vs_dest_user_kern udest;
3057 int need_full_svc = 0, need_full_dest = 0;
3059 cmd = info->genlhdr->cmd;
3061 mutex_lock(&__ip_vs_mutex);
3063 if (cmd == IPVS_CMD_FLUSH) {
3064 ret = ip_vs_flush();
3066 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3067 ret = ip_vs_genl_set_config(info->attrs);
3069 } else if (cmd == IPVS_CMD_NEW_DAEMON ||
3070 cmd == IPVS_CMD_DEL_DAEMON) {
3072 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3074 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3075 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3076 info->attrs[IPVS_CMD_ATTR_DAEMON],
3077 ip_vs_daemon_policy)) {
3082 if (cmd == IPVS_CMD_NEW_DAEMON)
3083 ret = ip_vs_genl_new_daemon(daemon_attrs);
3085 ret = ip_vs_genl_del_daemon(daemon_attrs);
3087 } else if (cmd == IPVS_CMD_ZERO &&
3088 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3089 ret = ip_vs_zero_all();
3093 /* All following commands require a service argument, so check if we
3094 * received a valid one. We need a full service specification when
3095 * adding / editing a service. Only identifying members otherwise. */
3096 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3099 ret = ip_vs_genl_parse_service(&usvc,
3100 info->attrs[IPVS_CMD_ATTR_SERVICE],
3105 /* Lookup the exact service by <protocol, addr, port> or fwmark */
3106 if (usvc.fwmark == 0)
3107 svc = __ip_vs_service_get(usvc.af, usvc.protocol,
3108 &usvc.addr, usvc.port);
3110 svc = __ip_vs_svc_fwm_get(usvc.af, usvc.fwmark);
3112 /* Unless we're adding a new service, the service must already exist */
3113 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3118 /* Destination commands require a valid destination argument. For
3119 * adding / editing a destination, we need a full destination
3121 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3122 cmd == IPVS_CMD_DEL_DEST) {
3123 if (cmd != IPVS_CMD_DEL_DEST)
3126 ret = ip_vs_genl_parse_dest(&udest,
3127 info->attrs[IPVS_CMD_ATTR_DEST],
3134 case IPVS_CMD_NEW_SERVICE:
3136 ret = ip_vs_add_service(&usvc, &svc);
3140 case IPVS_CMD_SET_SERVICE:
3141 ret = ip_vs_edit_service(svc, &usvc);
3143 case IPVS_CMD_DEL_SERVICE:
3144 ret = ip_vs_del_service(svc);
3146 case IPVS_CMD_NEW_DEST:
3147 ret = ip_vs_add_dest(svc, &udest);
3149 case IPVS_CMD_SET_DEST:
3150 ret = ip_vs_edit_dest(svc, &udest);
3152 case IPVS_CMD_DEL_DEST:
3153 ret = ip_vs_del_dest(svc, &udest);
3156 ret = ip_vs_zero_service(svc);
3164 ip_vs_service_put(svc);
3165 mutex_unlock(&__ip_vs_mutex);
3170 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3172 struct sk_buff *msg;
3174 int ret, cmd, reply_cmd;
3176 cmd = info->genlhdr->cmd;
3178 if (cmd == IPVS_CMD_GET_SERVICE)
3179 reply_cmd = IPVS_CMD_NEW_SERVICE;
3180 else if (cmd == IPVS_CMD_GET_INFO)
3181 reply_cmd = IPVS_CMD_SET_INFO;
3182 else if (cmd == IPVS_CMD_GET_CONFIG)
3183 reply_cmd = IPVS_CMD_SET_CONFIG;
3185 IP_VS_ERR("unknown Generic Netlink command\n");
3189 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3193 mutex_lock(&__ip_vs_mutex);
3195 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3197 goto nla_put_failure;
3200 case IPVS_CMD_GET_SERVICE:
3202 struct ip_vs_service *svc;
3204 svc = ip_vs_genl_find_service(info->attrs[IPVS_CMD_ATTR_SERVICE]);
3209 ret = ip_vs_genl_fill_service(msg, svc);
3210 ip_vs_service_put(svc);
3212 goto nla_put_failure;
3221 case IPVS_CMD_GET_CONFIG:
3223 struct ip_vs_timeout_user t;
3225 __ip_vs_get_timeouts(&t);
3226 #ifdef CONFIG_IP_VS_PROTO_TCP
3227 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP, t.tcp_timeout);
3228 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3231 #ifdef CONFIG_IP_VS_PROTO_UDP
3232 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout);
3238 case IPVS_CMD_GET_INFO:
3239 NLA_PUT_U32(msg, IPVS_INFO_ATTR_VERSION, IP_VS_VERSION_CODE);
3240 NLA_PUT_U32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3241 IP_VS_CONN_TAB_SIZE);
3245 genlmsg_end(msg, reply);
3246 ret = genlmsg_unicast(msg, info->snd_pid);
3250 IP_VS_ERR("not enough space in Netlink message\n");
3256 mutex_unlock(&__ip_vs_mutex);
3262 static struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3264 .cmd = IPVS_CMD_NEW_SERVICE,
3265 .flags = GENL_ADMIN_PERM,
3266 .policy = ip_vs_cmd_policy,
3267 .doit = ip_vs_genl_set_cmd,
3270 .cmd = IPVS_CMD_SET_SERVICE,
3271 .flags = GENL_ADMIN_PERM,
3272 .policy = ip_vs_cmd_policy,
3273 .doit = ip_vs_genl_set_cmd,
3276 .cmd = IPVS_CMD_DEL_SERVICE,
3277 .flags = GENL_ADMIN_PERM,
3278 .policy = ip_vs_cmd_policy,
3279 .doit = ip_vs_genl_set_cmd,
3282 .cmd = IPVS_CMD_GET_SERVICE,
3283 .flags = GENL_ADMIN_PERM,
3284 .doit = ip_vs_genl_get_cmd,
3285 .dumpit = ip_vs_genl_dump_services,
3286 .policy = ip_vs_cmd_policy,
3289 .cmd = IPVS_CMD_NEW_DEST,
3290 .flags = GENL_ADMIN_PERM,
3291 .policy = ip_vs_cmd_policy,
3292 .doit = ip_vs_genl_set_cmd,
3295 .cmd = IPVS_CMD_SET_DEST,
3296 .flags = GENL_ADMIN_PERM,
3297 .policy = ip_vs_cmd_policy,
3298 .doit = ip_vs_genl_set_cmd,
3301 .cmd = IPVS_CMD_DEL_DEST,
3302 .flags = GENL_ADMIN_PERM,
3303 .policy = ip_vs_cmd_policy,
3304 .doit = ip_vs_genl_set_cmd,
3307 .cmd = IPVS_CMD_GET_DEST,
3308 .flags = GENL_ADMIN_PERM,
3309 .policy = ip_vs_cmd_policy,
3310 .dumpit = ip_vs_genl_dump_dests,
3313 .cmd = IPVS_CMD_NEW_DAEMON,
3314 .flags = GENL_ADMIN_PERM,
3315 .policy = ip_vs_cmd_policy,
3316 .doit = ip_vs_genl_set_cmd,
3319 .cmd = IPVS_CMD_DEL_DAEMON,
3320 .flags = GENL_ADMIN_PERM,
3321 .policy = ip_vs_cmd_policy,
3322 .doit = ip_vs_genl_set_cmd,
3325 .cmd = IPVS_CMD_GET_DAEMON,
3326 .flags = GENL_ADMIN_PERM,
3327 .dumpit = ip_vs_genl_dump_daemons,
3330 .cmd = IPVS_CMD_SET_CONFIG,
3331 .flags = GENL_ADMIN_PERM,
3332 .policy = ip_vs_cmd_policy,
3333 .doit = ip_vs_genl_set_cmd,
3336 .cmd = IPVS_CMD_GET_CONFIG,
3337 .flags = GENL_ADMIN_PERM,
3338 .doit = ip_vs_genl_get_cmd,
3341 .cmd = IPVS_CMD_GET_INFO,
3342 .flags = GENL_ADMIN_PERM,
3343 .doit = ip_vs_genl_get_cmd,
3346 .cmd = IPVS_CMD_ZERO,
3347 .flags = GENL_ADMIN_PERM,
3348 .policy = ip_vs_cmd_policy,
3349 .doit = ip_vs_genl_set_cmd,
3352 .cmd = IPVS_CMD_FLUSH,
3353 .flags = GENL_ADMIN_PERM,
3354 .doit = ip_vs_genl_set_cmd,
3358 static int __init ip_vs_genl_register(void)
3362 ret = genl_register_family(&ip_vs_genl_family);
3366 for (i = 0; i < ARRAY_SIZE(ip_vs_genl_ops); i++) {
3367 ret = genl_register_ops(&ip_vs_genl_family, &ip_vs_genl_ops[i]);
3374 genl_unregister_family(&ip_vs_genl_family);
3378 static void ip_vs_genl_unregister(void)
3380 genl_unregister_family(&ip_vs_genl_family);
3383 /* End of Generic Netlink interface definitions */
3386 int __init ip_vs_control_init(void)
3393 ret = nf_register_sockopt(&ip_vs_sockopts);
3395 IP_VS_ERR("cannot register sockopt.\n");
3399 ret = ip_vs_genl_register();
3401 IP_VS_ERR("cannot register Generic Netlink interface.\n");
3402 nf_unregister_sockopt(&ip_vs_sockopts);
3406 proc_net_fops_create(&init_net, "ip_vs", 0, &ip_vs_info_fops);
3407 proc_net_fops_create(&init_net, "ip_vs_stats",0, &ip_vs_stats_fops);
3409 sysctl_header = register_sysctl_paths(net_vs_ctl_path, vs_vars);
3411 /* Initialize ip_vs_svc_table, ip_vs_svc_fwm_table, ip_vs_rtable */
3412 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
3413 INIT_LIST_HEAD(&ip_vs_svc_table[idx]);
3414 INIT_LIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3416 for(idx = 0; idx < IP_VS_RTAB_SIZE; idx++) {
3417 INIT_LIST_HEAD(&ip_vs_rtable[idx]);
3420 ip_vs_new_estimator(&ip_vs_stats);
3422 /* Hook the defense timer */
3423 schedule_delayed_work(&defense_work, DEFENSE_TIMER_PERIOD);
3430 void ip_vs_control_cleanup(void)
3433 ip_vs_trash_cleanup();
3434 cancel_rearming_delayed_work(&defense_work);
3435 cancel_work_sync(&defense_work.work);
3436 ip_vs_kill_estimator(&ip_vs_stats);
3437 unregister_sysctl_table(sysctl_header);
3438 proc_net_remove(&init_net, "ip_vs_stats");
3439 proc_net_remove(&init_net, "ip_vs");
3440 ip_vs_genl_unregister();
3441 nf_unregister_sockopt(&ip_vs_sockopts);