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 && (u->netmask < 1 || u->netmask > 128)) {
1177 svc = kzalloc(sizeof(struct ip_vs_service), GFP_ATOMIC);
1179 IP_VS_DBG(1, "ip_vs_add_service: kmalloc failed.\n");
1184 /* I'm the first user of the service */
1185 atomic_set(&svc->usecnt, 1);
1186 atomic_set(&svc->refcnt, 0);
1189 svc->protocol = u->protocol;
1190 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1191 svc->port = u->port;
1192 svc->fwmark = u->fwmark;
1193 svc->flags = u->flags;
1194 svc->timeout = u->timeout * HZ;
1195 svc->netmask = u->netmask;
1197 INIT_LIST_HEAD(&svc->destinations);
1198 rwlock_init(&svc->sched_lock);
1199 spin_lock_init(&svc->stats.lock);
1201 /* Bind the scheduler */
1202 ret = ip_vs_bind_scheduler(svc, sched);
1207 /* Update the virtual service counters */
1208 if (svc->port == FTPPORT)
1209 atomic_inc(&ip_vs_ftpsvc_counter);
1210 else if (svc->port == 0)
1211 atomic_inc(&ip_vs_nullsvc_counter);
1213 ip_vs_new_estimator(&svc->stats);
1215 /* Count only IPv4 services for old get/setsockopt interface */
1216 if (svc->af == AF_INET)
1217 ip_vs_num_services++;
1219 /* Hash the service into the service table */
1220 write_lock_bh(&__ip_vs_svc_lock);
1221 ip_vs_svc_hash(svc);
1222 write_unlock_bh(&__ip_vs_svc_lock);
1230 ip_vs_unbind_scheduler(svc);
1233 ip_vs_app_inc_put(svc->inc);
1238 ip_vs_scheduler_put(sched);
1241 /* decrease the module use count */
1242 ip_vs_use_count_dec();
1249 * Edit a service and bind it with a new scheduler
1252 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1254 struct ip_vs_scheduler *sched, *old_sched;
1258 * Lookup the scheduler, by 'u->sched_name'
1260 sched = ip_vs_scheduler_get(u->sched_name);
1261 if (sched == NULL) {
1262 IP_VS_INFO("Scheduler module ip_vs_%s not found\n",
1268 #ifdef CONFIG_IP_VS_IPV6
1269 if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1275 write_lock_bh(&__ip_vs_svc_lock);
1278 * Wait until all other svc users go away.
1280 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 1);
1283 * Set the flags and timeout value
1285 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1286 svc->timeout = u->timeout * HZ;
1287 svc->netmask = u->netmask;
1289 old_sched = svc->scheduler;
1290 if (sched != old_sched) {
1292 * Unbind the old scheduler
1294 if ((ret = ip_vs_unbind_scheduler(svc))) {
1300 * Bind the new scheduler
1302 if ((ret = ip_vs_bind_scheduler(svc, sched))) {
1304 * If ip_vs_bind_scheduler fails, restore the old
1306 * The main reason of failure is out of memory.
1308 * The question is if the old scheduler can be
1309 * restored all the time. TODO: if it cannot be
1310 * restored some time, we must delete the service,
1311 * otherwise the system may crash.
1313 ip_vs_bind_scheduler(svc, old_sched);
1320 write_unlock_bh(&__ip_vs_svc_lock);
1321 #ifdef CONFIG_IP_VS_IPV6
1326 ip_vs_scheduler_put(old_sched);
1333 * Delete a service from the service list
1334 * - The service must be unlinked, unlocked and not referenced!
1335 * - We are called under _bh lock
1337 static void __ip_vs_del_service(struct ip_vs_service *svc)
1339 struct ip_vs_dest *dest, *nxt;
1340 struct ip_vs_scheduler *old_sched;
1342 /* Count only IPv4 services for old get/setsockopt interface */
1343 if (svc->af == AF_INET)
1344 ip_vs_num_services--;
1346 ip_vs_kill_estimator(&svc->stats);
1348 /* Unbind scheduler */
1349 old_sched = svc->scheduler;
1350 ip_vs_unbind_scheduler(svc);
1352 ip_vs_scheduler_put(old_sched);
1354 /* Unbind app inc */
1356 ip_vs_app_inc_put(svc->inc);
1361 * Unlink the whole destination list
1363 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1364 __ip_vs_unlink_dest(svc, dest, 0);
1365 __ip_vs_del_dest(dest);
1369 * Update the virtual service counters
1371 if (svc->port == FTPPORT)
1372 atomic_dec(&ip_vs_ftpsvc_counter);
1373 else if (svc->port == 0)
1374 atomic_dec(&ip_vs_nullsvc_counter);
1377 * Free the service if nobody refers to it
1379 if (atomic_read(&svc->refcnt) == 0)
1382 /* decrease the module use count */
1383 ip_vs_use_count_dec();
1387 * Delete a service from the service list
1389 static int ip_vs_del_service(struct ip_vs_service *svc)
1395 * Unhash it from the service table
1397 write_lock_bh(&__ip_vs_svc_lock);
1399 ip_vs_svc_unhash(svc);
1402 * Wait until all the svc users go away.
1404 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 1);
1406 __ip_vs_del_service(svc);
1408 write_unlock_bh(&__ip_vs_svc_lock);
1415 * Flush all the virtual services
1417 static int ip_vs_flush(void)
1420 struct ip_vs_service *svc, *nxt;
1423 * Flush the service table hashed by <protocol,addr,port>
1425 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1426 list_for_each_entry_safe(svc, nxt, &ip_vs_svc_table[idx], s_list) {
1427 write_lock_bh(&__ip_vs_svc_lock);
1428 ip_vs_svc_unhash(svc);
1430 * Wait until all the svc users go away.
1432 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1433 __ip_vs_del_service(svc);
1434 write_unlock_bh(&__ip_vs_svc_lock);
1439 * Flush the service table hashed by fwmark
1441 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1442 list_for_each_entry_safe(svc, nxt,
1443 &ip_vs_svc_fwm_table[idx], f_list) {
1444 write_lock_bh(&__ip_vs_svc_lock);
1445 ip_vs_svc_unhash(svc);
1447 * Wait until all the svc users go away.
1449 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1450 __ip_vs_del_service(svc);
1451 write_unlock_bh(&__ip_vs_svc_lock);
1460 * Zero counters in a service or all services
1462 static int ip_vs_zero_service(struct ip_vs_service *svc)
1464 struct ip_vs_dest *dest;
1466 write_lock_bh(&__ip_vs_svc_lock);
1467 list_for_each_entry(dest, &svc->destinations, n_list) {
1468 ip_vs_zero_stats(&dest->stats);
1470 ip_vs_zero_stats(&svc->stats);
1471 write_unlock_bh(&__ip_vs_svc_lock);
1475 static int ip_vs_zero_all(void)
1478 struct ip_vs_service *svc;
1480 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1481 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1482 ip_vs_zero_service(svc);
1486 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1487 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1488 ip_vs_zero_service(svc);
1492 ip_vs_zero_stats(&ip_vs_stats);
1498 proc_do_defense_mode(ctl_table *table, int write, struct file * filp,
1499 void __user *buffer, size_t *lenp, loff_t *ppos)
1501 int *valp = table->data;
1505 rc = proc_dointvec(table, write, filp, buffer, lenp, ppos);
1506 if (write && (*valp != val)) {
1507 if ((*valp < 0) || (*valp > 3)) {
1508 /* Restore the correct value */
1511 update_defense_level();
1519 proc_do_sync_threshold(ctl_table *table, int write, struct file *filp,
1520 void __user *buffer, size_t *lenp, loff_t *ppos)
1522 int *valp = table->data;
1526 /* backup the value first */
1527 memcpy(val, valp, sizeof(val));
1529 rc = proc_dointvec(table, write, filp, buffer, lenp, ppos);
1530 if (write && (valp[0] < 0 || valp[1] < 0 || valp[0] >= valp[1])) {
1531 /* Restore the correct value */
1532 memcpy(valp, val, sizeof(val));
1539 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1542 static struct ctl_table vs_vars[] = {
1544 .procname = "amemthresh",
1545 .data = &sysctl_ip_vs_amemthresh,
1546 .maxlen = sizeof(int),
1548 .proc_handler = proc_dointvec,
1550 #ifdef CONFIG_IP_VS_DEBUG
1552 .procname = "debug_level",
1553 .data = &sysctl_ip_vs_debug_level,
1554 .maxlen = sizeof(int),
1556 .proc_handler = proc_dointvec,
1560 .procname = "am_droprate",
1561 .data = &sysctl_ip_vs_am_droprate,
1562 .maxlen = sizeof(int),
1564 .proc_handler = proc_dointvec,
1567 .procname = "drop_entry",
1568 .data = &sysctl_ip_vs_drop_entry,
1569 .maxlen = sizeof(int),
1571 .proc_handler = proc_do_defense_mode,
1574 .procname = "drop_packet",
1575 .data = &sysctl_ip_vs_drop_packet,
1576 .maxlen = sizeof(int),
1578 .proc_handler = proc_do_defense_mode,
1581 .procname = "secure_tcp",
1582 .data = &sysctl_ip_vs_secure_tcp,
1583 .maxlen = sizeof(int),
1585 .proc_handler = proc_do_defense_mode,
1589 .procname = "timeout_established",
1590 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1591 .maxlen = sizeof(int),
1593 .proc_handler = proc_dointvec_jiffies,
1596 .procname = "timeout_synsent",
1597 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1598 .maxlen = sizeof(int),
1600 .proc_handler = proc_dointvec_jiffies,
1603 .procname = "timeout_synrecv",
1604 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1605 .maxlen = sizeof(int),
1607 .proc_handler = proc_dointvec_jiffies,
1610 .procname = "timeout_finwait",
1611 .data = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1612 .maxlen = sizeof(int),
1614 .proc_handler = proc_dointvec_jiffies,
1617 .procname = "timeout_timewait",
1618 .data = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1619 .maxlen = sizeof(int),
1621 .proc_handler = proc_dointvec_jiffies,
1624 .procname = "timeout_close",
1625 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1626 .maxlen = sizeof(int),
1628 .proc_handler = proc_dointvec_jiffies,
1631 .procname = "timeout_closewait",
1632 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1633 .maxlen = sizeof(int),
1635 .proc_handler = proc_dointvec_jiffies,
1638 .procname = "timeout_lastack",
1639 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1640 .maxlen = sizeof(int),
1642 .proc_handler = proc_dointvec_jiffies,
1645 .procname = "timeout_listen",
1646 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1647 .maxlen = sizeof(int),
1649 .proc_handler = proc_dointvec_jiffies,
1652 .procname = "timeout_synack",
1653 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1654 .maxlen = sizeof(int),
1656 .proc_handler = proc_dointvec_jiffies,
1659 .procname = "timeout_udp",
1660 .data = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1661 .maxlen = sizeof(int),
1663 .proc_handler = proc_dointvec_jiffies,
1666 .procname = "timeout_icmp",
1667 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1668 .maxlen = sizeof(int),
1670 .proc_handler = proc_dointvec_jiffies,
1674 .procname = "cache_bypass",
1675 .data = &sysctl_ip_vs_cache_bypass,
1676 .maxlen = sizeof(int),
1678 .proc_handler = proc_dointvec,
1681 .procname = "expire_nodest_conn",
1682 .data = &sysctl_ip_vs_expire_nodest_conn,
1683 .maxlen = sizeof(int),
1685 .proc_handler = proc_dointvec,
1688 .procname = "expire_quiescent_template",
1689 .data = &sysctl_ip_vs_expire_quiescent_template,
1690 .maxlen = sizeof(int),
1692 .proc_handler = proc_dointvec,
1695 .procname = "sync_threshold",
1696 .data = &sysctl_ip_vs_sync_threshold,
1697 .maxlen = sizeof(sysctl_ip_vs_sync_threshold),
1699 .proc_handler = proc_do_sync_threshold,
1702 .procname = "nat_icmp_send",
1703 .data = &sysctl_ip_vs_nat_icmp_send,
1704 .maxlen = sizeof(int),
1706 .proc_handler = proc_dointvec,
1711 const struct ctl_path net_vs_ctl_path[] = {
1712 { .procname = "net", .ctl_name = CTL_NET, },
1713 { .procname = "ipv4", .ctl_name = NET_IPV4, },
1714 { .procname = "vs", },
1717 EXPORT_SYMBOL_GPL(net_vs_ctl_path);
1719 static struct ctl_table_header * sysctl_header;
1721 #ifdef CONFIG_PROC_FS
1724 struct list_head *table;
1729 * Write the contents of the VS rule table to a PROCfs file.
1730 * (It is kept just for backward compatibility)
1732 static inline const char *ip_vs_fwd_name(unsigned flags)
1734 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1735 case IP_VS_CONN_F_LOCALNODE:
1737 case IP_VS_CONN_F_TUNNEL:
1739 case IP_VS_CONN_F_DROUTE:
1747 /* Get the Nth entry in the two lists */
1748 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1750 struct ip_vs_iter *iter = seq->private;
1752 struct ip_vs_service *svc;
1754 /* look in hash by protocol */
1755 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1756 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1758 iter->table = ip_vs_svc_table;
1765 /* keep looking in fwmark */
1766 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1767 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1769 iter->table = ip_vs_svc_fwm_table;
1779 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1780 __acquires(__ip_vs_svc_lock)
1783 read_lock_bh(&__ip_vs_svc_lock);
1784 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1788 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1790 struct list_head *e;
1791 struct ip_vs_iter *iter;
1792 struct ip_vs_service *svc;
1795 if (v == SEQ_START_TOKEN)
1796 return ip_vs_info_array(seq,0);
1799 iter = seq->private;
1801 if (iter->table == ip_vs_svc_table) {
1802 /* next service in table hashed by protocol */
1803 if ((e = svc->s_list.next) != &ip_vs_svc_table[iter->bucket])
1804 return list_entry(e, struct ip_vs_service, s_list);
1807 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1808 list_for_each_entry(svc,&ip_vs_svc_table[iter->bucket],
1814 iter->table = ip_vs_svc_fwm_table;
1819 /* next service in hashed by fwmark */
1820 if ((e = svc->f_list.next) != &ip_vs_svc_fwm_table[iter->bucket])
1821 return list_entry(e, struct ip_vs_service, f_list);
1824 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1825 list_for_each_entry(svc, &ip_vs_svc_fwm_table[iter->bucket],
1833 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
1834 __releases(__ip_vs_svc_lock)
1836 read_unlock_bh(&__ip_vs_svc_lock);
1840 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
1842 if (v == SEQ_START_TOKEN) {
1844 "IP Virtual Server version %d.%d.%d (size=%d)\n",
1845 NVERSION(IP_VS_VERSION_CODE), IP_VS_CONN_TAB_SIZE);
1847 "Prot LocalAddress:Port Scheduler Flags\n");
1849 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
1851 const struct ip_vs_service *svc = v;
1852 const struct ip_vs_iter *iter = seq->private;
1853 const struct ip_vs_dest *dest;
1855 if (iter->table == ip_vs_svc_table) {
1856 #ifdef CONFIG_IP_VS_IPV6
1857 if (svc->af == AF_INET6)
1858 seq_printf(seq, "%s [%pI6]:%04X %s ",
1859 ip_vs_proto_name(svc->protocol),
1862 svc->scheduler->name);
1865 seq_printf(seq, "%s %08X:%04X %s ",
1866 ip_vs_proto_name(svc->protocol),
1867 ntohl(svc->addr.ip),
1869 svc->scheduler->name);
1871 seq_printf(seq, "FWM %08X %s ",
1872 svc->fwmark, svc->scheduler->name);
1875 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
1876 seq_printf(seq, "persistent %d %08X\n",
1878 ntohl(svc->netmask));
1880 seq_putc(seq, '\n');
1882 list_for_each_entry(dest, &svc->destinations, n_list) {
1883 #ifdef CONFIG_IP_VS_IPV6
1884 if (dest->af == AF_INET6)
1887 " %-7s %-6d %-10d %-10d\n",
1890 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
1891 atomic_read(&dest->weight),
1892 atomic_read(&dest->activeconns),
1893 atomic_read(&dest->inactconns));
1898 "%-7s %-6d %-10d %-10d\n",
1899 ntohl(dest->addr.ip),
1901 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
1902 atomic_read(&dest->weight),
1903 atomic_read(&dest->activeconns),
1904 atomic_read(&dest->inactconns));
1911 static const struct seq_operations ip_vs_info_seq_ops = {
1912 .start = ip_vs_info_seq_start,
1913 .next = ip_vs_info_seq_next,
1914 .stop = ip_vs_info_seq_stop,
1915 .show = ip_vs_info_seq_show,
1918 static int ip_vs_info_open(struct inode *inode, struct file *file)
1920 return seq_open_private(file, &ip_vs_info_seq_ops,
1921 sizeof(struct ip_vs_iter));
1924 static const struct file_operations ip_vs_info_fops = {
1925 .owner = THIS_MODULE,
1926 .open = ip_vs_info_open,
1928 .llseek = seq_lseek,
1929 .release = seq_release_private,
1934 struct ip_vs_stats ip_vs_stats = {
1935 .lock = __SPIN_LOCK_UNLOCKED(ip_vs_stats.lock),
1938 #ifdef CONFIG_PROC_FS
1939 static int ip_vs_stats_show(struct seq_file *seq, void *v)
1942 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
1944 " Total Incoming Outgoing Incoming Outgoing\n");
1946 " Conns Packets Packets Bytes Bytes\n");
1948 spin_lock_bh(&ip_vs_stats.lock);
1949 seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", ip_vs_stats.ustats.conns,
1950 ip_vs_stats.ustats.inpkts, ip_vs_stats.ustats.outpkts,
1951 (unsigned long long) ip_vs_stats.ustats.inbytes,
1952 (unsigned long long) ip_vs_stats.ustats.outbytes);
1954 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
1956 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
1957 seq_printf(seq,"%8X %8X %8X %16X %16X\n",
1958 ip_vs_stats.ustats.cps,
1959 ip_vs_stats.ustats.inpps,
1960 ip_vs_stats.ustats.outpps,
1961 ip_vs_stats.ustats.inbps,
1962 ip_vs_stats.ustats.outbps);
1963 spin_unlock_bh(&ip_vs_stats.lock);
1968 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
1970 return single_open(file, ip_vs_stats_show, NULL);
1973 static const struct file_operations ip_vs_stats_fops = {
1974 .owner = THIS_MODULE,
1975 .open = ip_vs_stats_seq_open,
1977 .llseek = seq_lseek,
1978 .release = single_release,
1984 * Set timeout values for tcp tcpfin udp in the timeout_table.
1986 static int ip_vs_set_timeout(struct ip_vs_timeout_user *u)
1988 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
1993 #ifdef CONFIG_IP_VS_PROTO_TCP
1994 if (u->tcp_timeout) {
1995 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_ESTABLISHED]
1996 = u->tcp_timeout * HZ;
1999 if (u->tcp_fin_timeout) {
2000 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_FIN_WAIT]
2001 = u->tcp_fin_timeout * HZ;
2005 #ifdef CONFIG_IP_VS_PROTO_UDP
2006 if (u->udp_timeout) {
2007 ip_vs_protocol_udp.timeout_table[IP_VS_UDP_S_NORMAL]
2008 = u->udp_timeout * HZ;
2015 #define SET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2016 #define SERVICE_ARG_LEN (sizeof(struct ip_vs_service_user))
2017 #define SVCDEST_ARG_LEN (sizeof(struct ip_vs_service_user) + \
2018 sizeof(struct ip_vs_dest_user))
2019 #define TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2020 #define DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user))
2021 #define MAX_ARG_LEN SVCDEST_ARG_LEN
2023 static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
2024 [SET_CMDID(IP_VS_SO_SET_ADD)] = SERVICE_ARG_LEN,
2025 [SET_CMDID(IP_VS_SO_SET_EDIT)] = SERVICE_ARG_LEN,
2026 [SET_CMDID(IP_VS_SO_SET_DEL)] = SERVICE_ARG_LEN,
2027 [SET_CMDID(IP_VS_SO_SET_FLUSH)] = 0,
2028 [SET_CMDID(IP_VS_SO_SET_ADDDEST)] = SVCDEST_ARG_LEN,
2029 [SET_CMDID(IP_VS_SO_SET_DELDEST)] = SVCDEST_ARG_LEN,
2030 [SET_CMDID(IP_VS_SO_SET_EDITDEST)] = SVCDEST_ARG_LEN,
2031 [SET_CMDID(IP_VS_SO_SET_TIMEOUT)] = TIMEOUT_ARG_LEN,
2032 [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)] = DAEMON_ARG_LEN,
2033 [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)] = DAEMON_ARG_LEN,
2034 [SET_CMDID(IP_VS_SO_SET_ZERO)] = SERVICE_ARG_LEN,
2037 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2038 struct ip_vs_service_user *usvc_compat)
2041 usvc->protocol = usvc_compat->protocol;
2042 usvc->addr.ip = usvc_compat->addr;
2043 usvc->port = usvc_compat->port;
2044 usvc->fwmark = usvc_compat->fwmark;
2046 /* Deep copy of sched_name is not needed here */
2047 usvc->sched_name = usvc_compat->sched_name;
2049 usvc->flags = usvc_compat->flags;
2050 usvc->timeout = usvc_compat->timeout;
2051 usvc->netmask = usvc_compat->netmask;
2054 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2055 struct ip_vs_dest_user *udest_compat)
2057 udest->addr.ip = udest_compat->addr;
2058 udest->port = udest_compat->port;
2059 udest->conn_flags = udest_compat->conn_flags;
2060 udest->weight = udest_compat->weight;
2061 udest->u_threshold = udest_compat->u_threshold;
2062 udest->l_threshold = udest_compat->l_threshold;
2066 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2069 unsigned char arg[MAX_ARG_LEN];
2070 struct ip_vs_service_user *usvc_compat;
2071 struct ip_vs_service_user_kern usvc;
2072 struct ip_vs_service *svc;
2073 struct ip_vs_dest_user *udest_compat;
2074 struct ip_vs_dest_user_kern udest;
2076 if (!capable(CAP_NET_ADMIN))
2079 if (len != set_arglen[SET_CMDID(cmd)]) {
2080 IP_VS_ERR("set_ctl: len %u != %u\n",
2081 len, set_arglen[SET_CMDID(cmd)]);
2085 if (copy_from_user(arg, user, len) != 0)
2088 /* increase the module use count */
2089 ip_vs_use_count_inc();
2091 if (mutex_lock_interruptible(&__ip_vs_mutex)) {
2096 if (cmd == IP_VS_SO_SET_FLUSH) {
2097 /* Flush the virtual service */
2098 ret = ip_vs_flush();
2100 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2101 /* Set timeout values for (tcp tcpfin udp) */
2102 ret = ip_vs_set_timeout((struct ip_vs_timeout_user *)arg);
2104 } else if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2105 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2106 ret = start_sync_thread(dm->state, dm->mcast_ifn, dm->syncid);
2108 } else if (cmd == IP_VS_SO_SET_STOPDAEMON) {
2109 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2110 ret = stop_sync_thread(dm->state);
2114 usvc_compat = (struct ip_vs_service_user *)arg;
2115 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2117 /* We only use the new structs internally, so copy userspace compat
2118 * structs to extended internal versions */
2119 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2120 ip_vs_copy_udest_compat(&udest, udest_compat);
2122 if (cmd == IP_VS_SO_SET_ZERO) {
2123 /* if no service address is set, zero counters in all */
2124 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2125 ret = ip_vs_zero_all();
2130 /* Check for valid protocol: TCP or UDP, even for fwmark!=0 */
2131 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP) {
2132 IP_VS_ERR("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2133 usvc.protocol, &usvc.addr.ip,
2134 ntohs(usvc.port), usvc.sched_name);
2139 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2140 if (usvc.fwmark == 0)
2141 svc = __ip_vs_service_get(usvc.af, usvc.protocol,
2142 &usvc.addr, usvc.port);
2144 svc = __ip_vs_svc_fwm_get(usvc.af, usvc.fwmark);
2146 if (cmd != IP_VS_SO_SET_ADD
2147 && (svc == NULL || svc->protocol != usvc.protocol)) {
2153 case IP_VS_SO_SET_ADD:
2157 ret = ip_vs_add_service(&usvc, &svc);
2159 case IP_VS_SO_SET_EDIT:
2160 ret = ip_vs_edit_service(svc, &usvc);
2162 case IP_VS_SO_SET_DEL:
2163 ret = ip_vs_del_service(svc);
2167 case IP_VS_SO_SET_ZERO:
2168 ret = ip_vs_zero_service(svc);
2170 case IP_VS_SO_SET_ADDDEST:
2171 ret = ip_vs_add_dest(svc, &udest);
2173 case IP_VS_SO_SET_EDITDEST:
2174 ret = ip_vs_edit_dest(svc, &udest);
2176 case IP_VS_SO_SET_DELDEST:
2177 ret = ip_vs_del_dest(svc, &udest);
2184 ip_vs_service_put(svc);
2187 mutex_unlock(&__ip_vs_mutex);
2189 /* decrease the module use count */
2190 ip_vs_use_count_dec();
2197 ip_vs_copy_stats(struct ip_vs_stats_user *dst, struct ip_vs_stats *src)
2199 spin_lock_bh(&src->lock);
2200 memcpy(dst, &src->ustats, sizeof(*dst));
2201 spin_unlock_bh(&src->lock);
2205 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2207 dst->protocol = src->protocol;
2208 dst->addr = src->addr.ip;
2209 dst->port = src->port;
2210 dst->fwmark = src->fwmark;
2211 strlcpy(dst->sched_name, src->scheduler->name, sizeof(dst->sched_name));
2212 dst->flags = src->flags;
2213 dst->timeout = src->timeout / HZ;
2214 dst->netmask = src->netmask;
2215 dst->num_dests = src->num_dests;
2216 ip_vs_copy_stats(&dst->stats, &src->stats);
2220 __ip_vs_get_service_entries(const struct ip_vs_get_services *get,
2221 struct ip_vs_get_services __user *uptr)
2224 struct ip_vs_service *svc;
2225 struct ip_vs_service_entry entry;
2228 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2229 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2230 /* Only expose IPv4 entries to old interface */
2231 if (svc->af != AF_INET)
2234 if (count >= get->num_services)
2236 memset(&entry, 0, sizeof(entry));
2237 ip_vs_copy_service(&entry, svc);
2238 if (copy_to_user(&uptr->entrytable[count],
2239 &entry, sizeof(entry))) {
2247 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2248 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2249 /* Only expose IPv4 entries to old interface */
2250 if (svc->af != AF_INET)
2253 if (count >= get->num_services)
2255 memset(&entry, 0, sizeof(entry));
2256 ip_vs_copy_service(&entry, svc);
2257 if (copy_to_user(&uptr->entrytable[count],
2258 &entry, sizeof(entry))) {
2270 __ip_vs_get_dest_entries(const struct ip_vs_get_dests *get,
2271 struct ip_vs_get_dests __user *uptr)
2273 struct ip_vs_service *svc;
2274 union nf_inet_addr addr = { .ip = get->addr };
2278 svc = __ip_vs_svc_fwm_get(AF_INET, get->fwmark);
2280 svc = __ip_vs_service_get(AF_INET, get->protocol, &addr,
2285 struct ip_vs_dest *dest;
2286 struct ip_vs_dest_entry entry;
2288 list_for_each_entry(dest, &svc->destinations, n_list) {
2289 if (count >= get->num_dests)
2292 entry.addr = dest->addr.ip;
2293 entry.port = dest->port;
2294 entry.conn_flags = atomic_read(&dest->conn_flags);
2295 entry.weight = atomic_read(&dest->weight);
2296 entry.u_threshold = dest->u_threshold;
2297 entry.l_threshold = dest->l_threshold;
2298 entry.activeconns = atomic_read(&dest->activeconns);
2299 entry.inactconns = atomic_read(&dest->inactconns);
2300 entry.persistconns = atomic_read(&dest->persistconns);
2301 ip_vs_copy_stats(&entry.stats, &dest->stats);
2302 if (copy_to_user(&uptr->entrytable[count],
2303 &entry, sizeof(entry))) {
2309 ip_vs_service_put(svc);
2316 __ip_vs_get_timeouts(struct ip_vs_timeout_user *u)
2318 #ifdef CONFIG_IP_VS_PROTO_TCP
2320 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2321 u->tcp_fin_timeout =
2322 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2324 #ifdef CONFIG_IP_VS_PROTO_UDP
2326 ip_vs_protocol_udp.timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2331 #define GET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2332 #define GET_INFO_ARG_LEN (sizeof(struct ip_vs_getinfo))
2333 #define GET_SERVICES_ARG_LEN (sizeof(struct ip_vs_get_services))
2334 #define GET_SERVICE_ARG_LEN (sizeof(struct ip_vs_service_entry))
2335 #define GET_DESTS_ARG_LEN (sizeof(struct ip_vs_get_dests))
2336 #define GET_TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2337 #define GET_DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user) * 2)
2339 static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
2340 [GET_CMDID(IP_VS_SO_GET_VERSION)] = 64,
2341 [GET_CMDID(IP_VS_SO_GET_INFO)] = GET_INFO_ARG_LEN,
2342 [GET_CMDID(IP_VS_SO_GET_SERVICES)] = GET_SERVICES_ARG_LEN,
2343 [GET_CMDID(IP_VS_SO_GET_SERVICE)] = GET_SERVICE_ARG_LEN,
2344 [GET_CMDID(IP_VS_SO_GET_DESTS)] = GET_DESTS_ARG_LEN,
2345 [GET_CMDID(IP_VS_SO_GET_TIMEOUT)] = GET_TIMEOUT_ARG_LEN,
2346 [GET_CMDID(IP_VS_SO_GET_DAEMON)] = GET_DAEMON_ARG_LEN,
2350 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2352 unsigned char arg[128];
2355 if (!capable(CAP_NET_ADMIN))
2358 if (*len < get_arglen[GET_CMDID(cmd)]) {
2359 IP_VS_ERR("get_ctl: len %u < %u\n",
2360 *len, get_arglen[GET_CMDID(cmd)]);
2364 if (copy_from_user(arg, user, get_arglen[GET_CMDID(cmd)]) != 0)
2367 if (mutex_lock_interruptible(&__ip_vs_mutex))
2368 return -ERESTARTSYS;
2371 case IP_VS_SO_GET_VERSION:
2375 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2376 NVERSION(IP_VS_VERSION_CODE), IP_VS_CONN_TAB_SIZE);
2377 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2381 *len = strlen(buf)+1;
2385 case IP_VS_SO_GET_INFO:
2387 struct ip_vs_getinfo info;
2388 info.version = IP_VS_VERSION_CODE;
2389 info.size = IP_VS_CONN_TAB_SIZE;
2390 info.num_services = ip_vs_num_services;
2391 if (copy_to_user(user, &info, sizeof(info)) != 0)
2396 case IP_VS_SO_GET_SERVICES:
2398 struct ip_vs_get_services *get;
2401 get = (struct ip_vs_get_services *)arg;
2402 size = sizeof(*get) +
2403 sizeof(struct ip_vs_service_entry) * get->num_services;
2405 IP_VS_ERR("length: %u != %u\n", *len, size);
2409 ret = __ip_vs_get_service_entries(get, user);
2413 case IP_VS_SO_GET_SERVICE:
2415 struct ip_vs_service_entry *entry;
2416 struct ip_vs_service *svc;
2417 union nf_inet_addr addr;
2419 entry = (struct ip_vs_service_entry *)arg;
2420 addr.ip = entry->addr;
2422 svc = __ip_vs_svc_fwm_get(AF_INET, entry->fwmark);
2424 svc = __ip_vs_service_get(AF_INET, entry->protocol,
2425 &addr, entry->port);
2427 ip_vs_copy_service(entry, svc);
2428 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2430 ip_vs_service_put(svc);
2436 case IP_VS_SO_GET_DESTS:
2438 struct ip_vs_get_dests *get;
2441 get = (struct ip_vs_get_dests *)arg;
2442 size = sizeof(*get) +
2443 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2445 IP_VS_ERR("length: %u != %u\n", *len, size);
2449 ret = __ip_vs_get_dest_entries(get, user);
2453 case IP_VS_SO_GET_TIMEOUT:
2455 struct ip_vs_timeout_user t;
2457 __ip_vs_get_timeouts(&t);
2458 if (copy_to_user(user, &t, sizeof(t)) != 0)
2463 case IP_VS_SO_GET_DAEMON:
2465 struct ip_vs_daemon_user d[2];
2467 memset(&d, 0, sizeof(d));
2468 if (ip_vs_sync_state & IP_VS_STATE_MASTER) {
2469 d[0].state = IP_VS_STATE_MASTER;
2470 strlcpy(d[0].mcast_ifn, ip_vs_master_mcast_ifn, sizeof(d[0].mcast_ifn));
2471 d[0].syncid = ip_vs_master_syncid;
2473 if (ip_vs_sync_state & IP_VS_STATE_BACKUP) {
2474 d[1].state = IP_VS_STATE_BACKUP;
2475 strlcpy(d[1].mcast_ifn, ip_vs_backup_mcast_ifn, sizeof(d[1].mcast_ifn));
2476 d[1].syncid = ip_vs_backup_syncid;
2478 if (copy_to_user(user, &d, sizeof(d)) != 0)
2488 mutex_unlock(&__ip_vs_mutex);
2493 static struct nf_sockopt_ops ip_vs_sockopts = {
2495 .set_optmin = IP_VS_BASE_CTL,
2496 .set_optmax = IP_VS_SO_SET_MAX+1,
2497 .set = do_ip_vs_set_ctl,
2498 .get_optmin = IP_VS_BASE_CTL,
2499 .get_optmax = IP_VS_SO_GET_MAX+1,
2500 .get = do_ip_vs_get_ctl,
2501 .owner = THIS_MODULE,
2505 * Generic Netlink interface
2508 /* IPVS genetlink family */
2509 static struct genl_family ip_vs_genl_family = {
2510 .id = GENL_ID_GENERATE,
2512 .name = IPVS_GENL_NAME,
2513 .version = IPVS_GENL_VERSION,
2514 .maxattr = IPVS_CMD_MAX,
2517 /* Policy used for first-level command attributes */
2518 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2519 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2520 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2521 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2522 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2523 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2524 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2527 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2528 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2529 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2530 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2531 .len = IP_VS_IFNAME_MAXLEN },
2532 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2535 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2536 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2537 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2538 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2539 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2540 .len = sizeof(union nf_inet_addr) },
2541 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2542 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2543 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2544 .len = IP_VS_SCHEDNAME_MAXLEN },
2545 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2546 .len = sizeof(struct ip_vs_flags) },
2547 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2548 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2549 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2552 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2553 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2554 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2555 .len = sizeof(union nf_inet_addr) },
2556 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2557 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2558 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2559 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2560 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2561 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2562 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2563 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2564 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2567 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2568 struct ip_vs_stats *stats)
2570 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2574 spin_lock_bh(&stats->lock);
2576 NLA_PUT_U32(skb, IPVS_STATS_ATTR_CONNS, stats->ustats.conns);
2577 NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPKTS, stats->ustats.inpkts);
2578 NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPKTS, stats->ustats.outpkts);
2579 NLA_PUT_U64(skb, IPVS_STATS_ATTR_INBYTES, stats->ustats.inbytes);
2580 NLA_PUT_U64(skb, IPVS_STATS_ATTR_OUTBYTES, stats->ustats.outbytes);
2581 NLA_PUT_U32(skb, IPVS_STATS_ATTR_CPS, stats->ustats.cps);
2582 NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPPS, stats->ustats.inpps);
2583 NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPPS, stats->ustats.outpps);
2584 NLA_PUT_U32(skb, IPVS_STATS_ATTR_INBPS, stats->ustats.inbps);
2585 NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTBPS, stats->ustats.outbps);
2587 spin_unlock_bh(&stats->lock);
2589 nla_nest_end(skb, nl_stats);
2594 spin_unlock_bh(&stats->lock);
2595 nla_nest_cancel(skb, nl_stats);
2599 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2600 struct ip_vs_service *svc)
2602 struct nlattr *nl_service;
2603 struct ip_vs_flags flags = { .flags = svc->flags,
2606 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2610 NLA_PUT_U16(skb, IPVS_SVC_ATTR_AF, svc->af);
2613 NLA_PUT_U32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark);
2615 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol);
2616 NLA_PUT(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr);
2617 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PORT, svc->port);
2620 NLA_PUT_STRING(skb, IPVS_SVC_ATTR_SCHED_NAME, svc->scheduler->name);
2621 NLA_PUT(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags);
2622 NLA_PUT_U32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ);
2623 NLA_PUT_U32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask);
2625 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2626 goto nla_put_failure;
2628 nla_nest_end(skb, nl_service);
2633 nla_nest_cancel(skb, nl_service);
2637 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2638 struct ip_vs_service *svc,
2639 struct netlink_callback *cb)
2643 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2644 &ip_vs_genl_family, NLM_F_MULTI,
2645 IPVS_CMD_NEW_SERVICE);
2649 if (ip_vs_genl_fill_service(skb, svc) < 0)
2650 goto nla_put_failure;
2652 return genlmsg_end(skb, hdr);
2655 genlmsg_cancel(skb, hdr);
2659 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2660 struct netlink_callback *cb)
2663 int start = cb->args[0];
2664 struct ip_vs_service *svc;
2666 mutex_lock(&__ip_vs_mutex);
2667 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2668 list_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2671 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2673 goto nla_put_failure;
2678 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2679 list_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
2682 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2684 goto nla_put_failure;
2690 mutex_unlock(&__ip_vs_mutex);
2696 static int ip_vs_genl_parse_service(struct ip_vs_service_user_kern *usvc,
2697 struct nlattr *nla, int full_entry)
2699 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
2700 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
2702 /* Parse mandatory identifying service fields first */
2704 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
2707 nla_af = attrs[IPVS_SVC_ATTR_AF];
2708 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
2709 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
2710 nla_port = attrs[IPVS_SVC_ATTR_PORT];
2711 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
2713 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
2716 usvc->af = nla_get_u16(nla_af);
2717 #ifdef CONFIG_IP_VS_IPV6
2718 if (usvc->af != AF_INET && usvc->af != AF_INET6)
2720 if (usvc->af != AF_INET)
2722 return -EAFNOSUPPORT;
2725 usvc->protocol = IPPROTO_TCP;
2726 usvc->fwmark = nla_get_u32(nla_fwmark);
2728 usvc->protocol = nla_get_u16(nla_protocol);
2729 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
2730 usvc->port = nla_get_u16(nla_port);
2734 /* If a full entry was requested, check for the additional fields */
2736 struct nlattr *nla_sched, *nla_flags, *nla_timeout,
2738 struct ip_vs_flags flags;
2739 struct ip_vs_service *svc;
2741 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
2742 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
2743 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
2744 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
2746 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
2749 nla_memcpy(&flags, nla_flags, sizeof(flags));
2751 /* prefill flags from service if it already exists */
2753 svc = __ip_vs_svc_fwm_get(usvc->af, usvc->fwmark);
2755 svc = __ip_vs_service_get(usvc->af, usvc->protocol,
2756 &usvc->addr, usvc->port);
2758 usvc->flags = svc->flags;
2759 ip_vs_service_put(svc);
2763 /* set new flags from userland */
2764 usvc->flags = (usvc->flags & ~flags.mask) |
2765 (flags.flags & flags.mask);
2766 usvc->sched_name = nla_data(nla_sched);
2767 usvc->timeout = nla_get_u32(nla_timeout);
2768 usvc->netmask = nla_get_u32(nla_netmask);
2774 static struct ip_vs_service *ip_vs_genl_find_service(struct nlattr *nla)
2776 struct ip_vs_service_user_kern usvc;
2779 ret = ip_vs_genl_parse_service(&usvc, nla, 0);
2781 return ERR_PTR(ret);
2784 return __ip_vs_svc_fwm_get(usvc.af, usvc.fwmark);
2786 return __ip_vs_service_get(usvc.af, usvc.protocol,
2787 &usvc.addr, usvc.port);
2790 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
2792 struct nlattr *nl_dest;
2794 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
2798 NLA_PUT(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr);
2799 NLA_PUT_U16(skb, IPVS_DEST_ATTR_PORT, dest->port);
2801 NLA_PUT_U32(skb, IPVS_DEST_ATTR_FWD_METHOD,
2802 atomic_read(&dest->conn_flags) & IP_VS_CONN_F_FWD_MASK);
2803 NLA_PUT_U32(skb, IPVS_DEST_ATTR_WEIGHT, atomic_read(&dest->weight));
2804 NLA_PUT_U32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold);
2805 NLA_PUT_U32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold);
2806 NLA_PUT_U32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
2807 atomic_read(&dest->activeconns));
2808 NLA_PUT_U32(skb, IPVS_DEST_ATTR_INACT_CONNS,
2809 atomic_read(&dest->inactconns));
2810 NLA_PUT_U32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
2811 atomic_read(&dest->persistconns));
2813 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
2814 goto nla_put_failure;
2816 nla_nest_end(skb, nl_dest);
2821 nla_nest_cancel(skb, nl_dest);
2825 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
2826 struct netlink_callback *cb)
2830 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2831 &ip_vs_genl_family, NLM_F_MULTI,
2836 if (ip_vs_genl_fill_dest(skb, dest) < 0)
2837 goto nla_put_failure;
2839 return genlmsg_end(skb, hdr);
2842 genlmsg_cancel(skb, hdr);
2846 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
2847 struct netlink_callback *cb)
2850 int start = cb->args[0];
2851 struct ip_vs_service *svc;
2852 struct ip_vs_dest *dest;
2853 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
2855 mutex_lock(&__ip_vs_mutex);
2857 /* Try to find the service for which to dump destinations */
2858 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
2859 IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
2862 svc = ip_vs_genl_find_service(attrs[IPVS_CMD_ATTR_SERVICE]);
2863 if (IS_ERR(svc) || svc == NULL)
2866 /* Dump the destinations */
2867 list_for_each_entry(dest, &svc->destinations, n_list) {
2870 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
2872 goto nla_put_failure;
2878 ip_vs_service_put(svc);
2881 mutex_unlock(&__ip_vs_mutex);
2886 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
2887 struct nlattr *nla, int full_entry)
2889 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
2890 struct nlattr *nla_addr, *nla_port;
2892 /* Parse mandatory identifying destination fields first */
2894 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
2897 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
2898 nla_port = attrs[IPVS_DEST_ATTR_PORT];
2900 if (!(nla_addr && nla_port))
2903 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
2904 udest->port = nla_get_u16(nla_port);
2906 /* If a full entry was requested, check for the additional fields */
2908 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
2911 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
2912 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
2913 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
2914 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
2916 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
2919 udest->conn_flags = nla_get_u32(nla_fwd)
2920 & IP_VS_CONN_F_FWD_MASK;
2921 udest->weight = nla_get_u32(nla_weight);
2922 udest->u_threshold = nla_get_u32(nla_u_thresh);
2923 udest->l_threshold = nla_get_u32(nla_l_thresh);
2929 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __be32 state,
2930 const char *mcast_ifn, __be32 syncid)
2932 struct nlattr *nl_daemon;
2934 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
2938 NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_STATE, state);
2939 NLA_PUT_STRING(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn);
2940 NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid);
2942 nla_nest_end(skb, nl_daemon);
2947 nla_nest_cancel(skb, nl_daemon);
2951 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __be32 state,
2952 const char *mcast_ifn, __be32 syncid,
2953 struct netlink_callback *cb)
2956 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2957 &ip_vs_genl_family, NLM_F_MULTI,
2958 IPVS_CMD_NEW_DAEMON);
2962 if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
2963 goto nla_put_failure;
2965 return genlmsg_end(skb, hdr);
2968 genlmsg_cancel(skb, hdr);
2972 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
2973 struct netlink_callback *cb)
2975 mutex_lock(&__ip_vs_mutex);
2976 if ((ip_vs_sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
2977 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
2978 ip_vs_master_mcast_ifn,
2979 ip_vs_master_syncid, cb) < 0)
2980 goto nla_put_failure;
2985 if ((ip_vs_sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
2986 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
2987 ip_vs_backup_mcast_ifn,
2988 ip_vs_backup_syncid, cb) < 0)
2989 goto nla_put_failure;
2995 mutex_unlock(&__ip_vs_mutex);
3000 static int ip_vs_genl_new_daemon(struct nlattr **attrs)
3002 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3003 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3004 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3007 return start_sync_thread(nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3008 nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3009 nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3012 static int ip_vs_genl_del_daemon(struct nlattr **attrs)
3014 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3017 return stop_sync_thread(nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3020 static int ip_vs_genl_set_config(struct nlattr **attrs)
3022 struct ip_vs_timeout_user t;
3024 __ip_vs_get_timeouts(&t);
3026 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3027 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3029 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3031 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3033 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3034 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3036 return ip_vs_set_timeout(&t);
3039 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3041 struct ip_vs_service *svc = NULL;
3042 struct ip_vs_service_user_kern usvc;
3043 struct ip_vs_dest_user_kern udest;
3045 int need_full_svc = 0, need_full_dest = 0;
3047 cmd = info->genlhdr->cmd;
3049 mutex_lock(&__ip_vs_mutex);
3051 if (cmd == IPVS_CMD_FLUSH) {
3052 ret = ip_vs_flush();
3054 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3055 ret = ip_vs_genl_set_config(info->attrs);
3057 } else if (cmd == IPVS_CMD_NEW_DAEMON ||
3058 cmd == IPVS_CMD_DEL_DAEMON) {
3060 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3062 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3063 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3064 info->attrs[IPVS_CMD_ATTR_DAEMON],
3065 ip_vs_daemon_policy)) {
3070 if (cmd == IPVS_CMD_NEW_DAEMON)
3071 ret = ip_vs_genl_new_daemon(daemon_attrs);
3073 ret = ip_vs_genl_del_daemon(daemon_attrs);
3075 } else if (cmd == IPVS_CMD_ZERO &&
3076 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3077 ret = ip_vs_zero_all();
3081 /* All following commands require a service argument, so check if we
3082 * received a valid one. We need a full service specification when
3083 * adding / editing a service. Only identifying members otherwise. */
3084 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3087 ret = ip_vs_genl_parse_service(&usvc,
3088 info->attrs[IPVS_CMD_ATTR_SERVICE],
3093 /* Lookup the exact service by <protocol, addr, port> or fwmark */
3094 if (usvc.fwmark == 0)
3095 svc = __ip_vs_service_get(usvc.af, usvc.protocol,
3096 &usvc.addr, usvc.port);
3098 svc = __ip_vs_svc_fwm_get(usvc.af, usvc.fwmark);
3100 /* Unless we're adding a new service, the service must already exist */
3101 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3106 /* Destination commands require a valid destination argument. For
3107 * adding / editing a destination, we need a full destination
3109 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3110 cmd == IPVS_CMD_DEL_DEST) {
3111 if (cmd != IPVS_CMD_DEL_DEST)
3114 ret = ip_vs_genl_parse_dest(&udest,
3115 info->attrs[IPVS_CMD_ATTR_DEST],
3122 case IPVS_CMD_NEW_SERVICE:
3124 ret = ip_vs_add_service(&usvc, &svc);
3128 case IPVS_CMD_SET_SERVICE:
3129 ret = ip_vs_edit_service(svc, &usvc);
3131 case IPVS_CMD_DEL_SERVICE:
3132 ret = ip_vs_del_service(svc);
3134 case IPVS_CMD_NEW_DEST:
3135 ret = ip_vs_add_dest(svc, &udest);
3137 case IPVS_CMD_SET_DEST:
3138 ret = ip_vs_edit_dest(svc, &udest);
3140 case IPVS_CMD_DEL_DEST:
3141 ret = ip_vs_del_dest(svc, &udest);
3144 ret = ip_vs_zero_service(svc);
3152 ip_vs_service_put(svc);
3153 mutex_unlock(&__ip_vs_mutex);
3158 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3160 struct sk_buff *msg;
3162 int ret, cmd, reply_cmd;
3164 cmd = info->genlhdr->cmd;
3166 if (cmd == IPVS_CMD_GET_SERVICE)
3167 reply_cmd = IPVS_CMD_NEW_SERVICE;
3168 else if (cmd == IPVS_CMD_GET_INFO)
3169 reply_cmd = IPVS_CMD_SET_INFO;
3170 else if (cmd == IPVS_CMD_GET_CONFIG)
3171 reply_cmd = IPVS_CMD_SET_CONFIG;
3173 IP_VS_ERR("unknown Generic Netlink command\n");
3177 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3181 mutex_lock(&__ip_vs_mutex);
3183 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3185 goto nla_put_failure;
3188 case IPVS_CMD_GET_SERVICE:
3190 struct ip_vs_service *svc;
3192 svc = ip_vs_genl_find_service(info->attrs[IPVS_CMD_ATTR_SERVICE]);
3197 ret = ip_vs_genl_fill_service(msg, svc);
3198 ip_vs_service_put(svc);
3200 goto nla_put_failure;
3209 case IPVS_CMD_GET_CONFIG:
3211 struct ip_vs_timeout_user t;
3213 __ip_vs_get_timeouts(&t);
3214 #ifdef CONFIG_IP_VS_PROTO_TCP
3215 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP, t.tcp_timeout);
3216 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3219 #ifdef CONFIG_IP_VS_PROTO_UDP
3220 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout);
3226 case IPVS_CMD_GET_INFO:
3227 NLA_PUT_U32(msg, IPVS_INFO_ATTR_VERSION, IP_VS_VERSION_CODE);
3228 NLA_PUT_U32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3229 IP_VS_CONN_TAB_SIZE);
3233 genlmsg_end(msg, reply);
3234 ret = genlmsg_unicast(msg, info->snd_pid);
3238 IP_VS_ERR("not enough space in Netlink message\n");
3244 mutex_unlock(&__ip_vs_mutex);
3250 static struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3252 .cmd = IPVS_CMD_NEW_SERVICE,
3253 .flags = GENL_ADMIN_PERM,
3254 .policy = ip_vs_cmd_policy,
3255 .doit = ip_vs_genl_set_cmd,
3258 .cmd = IPVS_CMD_SET_SERVICE,
3259 .flags = GENL_ADMIN_PERM,
3260 .policy = ip_vs_cmd_policy,
3261 .doit = ip_vs_genl_set_cmd,
3264 .cmd = IPVS_CMD_DEL_SERVICE,
3265 .flags = GENL_ADMIN_PERM,
3266 .policy = ip_vs_cmd_policy,
3267 .doit = ip_vs_genl_set_cmd,
3270 .cmd = IPVS_CMD_GET_SERVICE,
3271 .flags = GENL_ADMIN_PERM,
3272 .doit = ip_vs_genl_get_cmd,
3273 .dumpit = ip_vs_genl_dump_services,
3274 .policy = ip_vs_cmd_policy,
3277 .cmd = IPVS_CMD_NEW_DEST,
3278 .flags = GENL_ADMIN_PERM,
3279 .policy = ip_vs_cmd_policy,
3280 .doit = ip_vs_genl_set_cmd,
3283 .cmd = IPVS_CMD_SET_DEST,
3284 .flags = GENL_ADMIN_PERM,
3285 .policy = ip_vs_cmd_policy,
3286 .doit = ip_vs_genl_set_cmd,
3289 .cmd = IPVS_CMD_DEL_DEST,
3290 .flags = GENL_ADMIN_PERM,
3291 .policy = ip_vs_cmd_policy,
3292 .doit = ip_vs_genl_set_cmd,
3295 .cmd = IPVS_CMD_GET_DEST,
3296 .flags = GENL_ADMIN_PERM,
3297 .policy = ip_vs_cmd_policy,
3298 .dumpit = ip_vs_genl_dump_dests,
3301 .cmd = IPVS_CMD_NEW_DAEMON,
3302 .flags = GENL_ADMIN_PERM,
3303 .policy = ip_vs_cmd_policy,
3304 .doit = ip_vs_genl_set_cmd,
3307 .cmd = IPVS_CMD_DEL_DAEMON,
3308 .flags = GENL_ADMIN_PERM,
3309 .policy = ip_vs_cmd_policy,
3310 .doit = ip_vs_genl_set_cmd,
3313 .cmd = IPVS_CMD_GET_DAEMON,
3314 .flags = GENL_ADMIN_PERM,
3315 .dumpit = ip_vs_genl_dump_daemons,
3318 .cmd = IPVS_CMD_SET_CONFIG,
3319 .flags = GENL_ADMIN_PERM,
3320 .policy = ip_vs_cmd_policy,
3321 .doit = ip_vs_genl_set_cmd,
3324 .cmd = IPVS_CMD_GET_CONFIG,
3325 .flags = GENL_ADMIN_PERM,
3326 .doit = ip_vs_genl_get_cmd,
3329 .cmd = IPVS_CMD_GET_INFO,
3330 .flags = GENL_ADMIN_PERM,
3331 .doit = ip_vs_genl_get_cmd,
3334 .cmd = IPVS_CMD_ZERO,
3335 .flags = GENL_ADMIN_PERM,
3336 .policy = ip_vs_cmd_policy,
3337 .doit = ip_vs_genl_set_cmd,
3340 .cmd = IPVS_CMD_FLUSH,
3341 .flags = GENL_ADMIN_PERM,
3342 .doit = ip_vs_genl_set_cmd,
3346 static int __init ip_vs_genl_register(void)
3350 ret = genl_register_family(&ip_vs_genl_family);
3354 for (i = 0; i < ARRAY_SIZE(ip_vs_genl_ops); i++) {
3355 ret = genl_register_ops(&ip_vs_genl_family, &ip_vs_genl_ops[i]);
3362 genl_unregister_family(&ip_vs_genl_family);
3366 static void ip_vs_genl_unregister(void)
3368 genl_unregister_family(&ip_vs_genl_family);
3371 /* End of Generic Netlink interface definitions */
3374 int __init ip_vs_control_init(void)
3381 ret = nf_register_sockopt(&ip_vs_sockopts);
3383 IP_VS_ERR("cannot register sockopt.\n");
3387 ret = ip_vs_genl_register();
3389 IP_VS_ERR("cannot register Generic Netlink interface.\n");
3390 nf_unregister_sockopt(&ip_vs_sockopts);
3394 proc_net_fops_create(&init_net, "ip_vs", 0, &ip_vs_info_fops);
3395 proc_net_fops_create(&init_net, "ip_vs_stats",0, &ip_vs_stats_fops);
3397 sysctl_header = register_sysctl_paths(net_vs_ctl_path, vs_vars);
3399 /* Initialize ip_vs_svc_table, ip_vs_svc_fwm_table, ip_vs_rtable */
3400 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
3401 INIT_LIST_HEAD(&ip_vs_svc_table[idx]);
3402 INIT_LIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3404 for(idx = 0; idx < IP_VS_RTAB_SIZE; idx++) {
3405 INIT_LIST_HEAD(&ip_vs_rtable[idx]);
3408 ip_vs_new_estimator(&ip_vs_stats);
3410 /* Hook the defense timer */
3411 schedule_delayed_work(&defense_work, DEFENSE_TIMER_PERIOD);
3418 void ip_vs_control_cleanup(void)
3421 ip_vs_trash_cleanup();
3422 cancel_rearming_delayed_work(&defense_work);
3423 cancel_work_sync(&defense_work.work);
3424 ip_vs_kill_estimator(&ip_vs_stats);
3425 unregister_sysctl_table(sysctl_header);
3426 proc_net_remove(&init_net, "ip_vs_stats");
3427 proc_net_remove(&init_net, "ip_vs");
3428 ip_vs_genl_unregister();
3429 nf_unregister_sockopt(&ip_vs_sockopts);