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 #define KMSG_COMPONENT "IPVS"
22 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
24 #include <linux/module.h>
25 #include <linux/init.h>
26 #include <linux/types.h>
27 #include <linux/capability.h>
29 #include <linux/sysctl.h>
30 #include <linux/proc_fs.h>
31 #include <linux/workqueue.h>
32 #include <linux/swap.h>
33 #include <linux/seq_file.h>
34 #include <linux/slab.h>
36 #include <linux/netfilter.h>
37 #include <linux/netfilter_ipv4.h>
38 #include <linux/mutex.h>
40 #include <net/net_namespace.h>
42 #ifdef CONFIG_IP_VS_IPV6
44 #include <net/ip6_route.h>
46 #include <net/route.h>
48 #include <net/genetlink.h>
50 #include <asm/uaccess.h>
52 #include <net/ip_vs.h>
54 /* semaphore for IPVS sockopts. And, [gs]etsockopt may sleep. */
55 static DEFINE_MUTEX(__ip_vs_mutex);
57 /* lock for service table */
58 static DEFINE_RWLOCK(__ip_vs_svc_lock);
60 /* lock for table with the real services */
61 static DEFINE_RWLOCK(__ip_vs_rs_lock);
63 /* lock for state and timeout tables */
64 static DEFINE_SPINLOCK(ip_vs_securetcp_lock);
66 /* lock for drop entry handling */
67 static DEFINE_SPINLOCK(__ip_vs_dropentry_lock);
69 /* lock for drop packet handling */
70 static DEFINE_SPINLOCK(__ip_vs_droppacket_lock);
72 /* 1/rate drop and drop-entry variables */
73 int ip_vs_drop_rate = 0;
74 int ip_vs_drop_counter = 0;
75 static atomic_t ip_vs_dropentry = ATOMIC_INIT(0);
77 /* number of virtual services */
78 static int ip_vs_num_services = 0;
80 /* sysctl variables */
81 static int sysctl_ip_vs_drop_entry = 0;
82 static int sysctl_ip_vs_drop_packet = 0;
83 static int sysctl_ip_vs_secure_tcp = 0;
84 static int sysctl_ip_vs_amemthresh = 1024;
85 static int sysctl_ip_vs_am_droprate = 10;
86 int sysctl_ip_vs_cache_bypass = 0;
87 int sysctl_ip_vs_expire_nodest_conn = 0;
88 int sysctl_ip_vs_expire_quiescent_template = 0;
89 int sysctl_ip_vs_sync_threshold[2] = { 3, 50 };
90 int sysctl_ip_vs_nat_icmp_send = 0;
91 #ifdef CONFIG_IP_VS_NFCT
92 int sysctl_ip_vs_conntrack;
94 int sysctl_ip_vs_snat_reroute = 1;
97 #ifdef CONFIG_IP_VS_DEBUG
98 static int sysctl_ip_vs_debug_level = 0;
100 int ip_vs_get_debug_level(void)
102 return sysctl_ip_vs_debug_level;
106 #ifdef CONFIG_IP_VS_IPV6
107 /* Taken from rt6_fill_node() in net/ipv6/route.c, is there a better way? */
108 static int __ip_vs_addr_is_local_v6(const struct in6_addr *addr)
116 .saddr = { .s6_addr32 = {0, 0, 0, 0} }, } },
119 rt = (struct rt6_info *)ip6_route_output(&init_net, NULL, &fl);
120 if (rt && rt->rt6i_dev && (rt->rt6i_dev->flags & IFF_LOOPBACK))
127 * update_defense_level is called from keventd and from sysctl,
128 * so it needs to protect itself from softirqs
130 static void update_defense_level(void)
133 static int old_secure_tcp = 0;
138 /* we only count free and buffered memory (in pages) */
140 availmem = i.freeram + i.bufferram;
141 /* however in linux 2.5 the i.bufferram is total page cache size,
143 /* si_swapinfo(&i); */
144 /* availmem = availmem - (i.totalswap - i.freeswap); */
146 nomem = (availmem < sysctl_ip_vs_amemthresh);
151 spin_lock(&__ip_vs_dropentry_lock);
152 switch (sysctl_ip_vs_drop_entry) {
154 atomic_set(&ip_vs_dropentry, 0);
158 atomic_set(&ip_vs_dropentry, 1);
159 sysctl_ip_vs_drop_entry = 2;
161 atomic_set(&ip_vs_dropentry, 0);
166 atomic_set(&ip_vs_dropentry, 1);
168 atomic_set(&ip_vs_dropentry, 0);
169 sysctl_ip_vs_drop_entry = 1;
173 atomic_set(&ip_vs_dropentry, 1);
176 spin_unlock(&__ip_vs_dropentry_lock);
179 spin_lock(&__ip_vs_droppacket_lock);
180 switch (sysctl_ip_vs_drop_packet) {
186 ip_vs_drop_rate = ip_vs_drop_counter
187 = sysctl_ip_vs_amemthresh /
188 (sysctl_ip_vs_amemthresh-availmem);
189 sysctl_ip_vs_drop_packet = 2;
196 ip_vs_drop_rate = ip_vs_drop_counter
197 = sysctl_ip_vs_amemthresh /
198 (sysctl_ip_vs_amemthresh-availmem);
201 sysctl_ip_vs_drop_packet = 1;
205 ip_vs_drop_rate = sysctl_ip_vs_am_droprate;
208 spin_unlock(&__ip_vs_droppacket_lock);
211 spin_lock(&ip_vs_securetcp_lock);
212 switch (sysctl_ip_vs_secure_tcp) {
214 if (old_secure_tcp >= 2)
219 if (old_secure_tcp < 2)
221 sysctl_ip_vs_secure_tcp = 2;
223 if (old_secure_tcp >= 2)
229 if (old_secure_tcp < 2)
232 if (old_secure_tcp >= 2)
234 sysctl_ip_vs_secure_tcp = 1;
238 if (old_secure_tcp < 2)
242 old_secure_tcp = sysctl_ip_vs_secure_tcp;
244 ip_vs_protocol_timeout_change(sysctl_ip_vs_secure_tcp>1);
245 spin_unlock(&ip_vs_securetcp_lock);
252 * Timer for checking the defense
254 #define DEFENSE_TIMER_PERIOD 1*HZ
255 static void defense_work_handler(struct work_struct *work);
256 static DECLARE_DELAYED_WORK(defense_work, defense_work_handler);
258 static void defense_work_handler(struct work_struct *work)
260 update_defense_level();
261 if (atomic_read(&ip_vs_dropentry))
262 ip_vs_random_dropentry();
264 schedule_delayed_work(&defense_work, DEFENSE_TIMER_PERIOD);
268 ip_vs_use_count_inc(void)
270 return try_module_get(THIS_MODULE);
274 ip_vs_use_count_dec(void)
276 module_put(THIS_MODULE);
281 * Hash table: for virtual service lookups
283 #define IP_VS_SVC_TAB_BITS 8
284 #define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
285 #define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)
287 /* the service table hashed by <protocol, addr, port> */
288 static struct list_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE];
289 /* the service table hashed by fwmark */
290 static struct list_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE];
293 * Hash table: for real service lookups
295 #define IP_VS_RTAB_BITS 4
296 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
297 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
299 static struct list_head ip_vs_rtable[IP_VS_RTAB_SIZE];
302 * Trash for destinations
304 static LIST_HEAD(ip_vs_dest_trash);
307 * FTP & NULL virtual service counters
309 static atomic_t ip_vs_ftpsvc_counter = ATOMIC_INIT(0);
310 static atomic_t ip_vs_nullsvc_counter = ATOMIC_INIT(0);
314 * Returns hash value for virtual service
316 static __inline__ unsigned
317 ip_vs_svc_hashkey(int af, unsigned proto, const union nf_inet_addr *addr,
320 register unsigned porth = ntohs(port);
321 __be32 addr_fold = addr->ip;
323 #ifdef CONFIG_IP_VS_IPV6
325 addr_fold = addr->ip6[0]^addr->ip6[1]^
326 addr->ip6[2]^addr->ip6[3];
329 return (proto^ntohl(addr_fold)^(porth>>IP_VS_SVC_TAB_BITS)^porth)
330 & IP_VS_SVC_TAB_MASK;
334 * Returns hash value of fwmark for virtual service lookup
336 static __inline__ unsigned ip_vs_svc_fwm_hashkey(__u32 fwmark)
338 return fwmark & IP_VS_SVC_TAB_MASK;
342 * Hashes a service in the ip_vs_svc_table by <proto,addr,port>
343 * or in the ip_vs_svc_fwm_table by fwmark.
344 * Should be called with locked tables.
346 static int ip_vs_svc_hash(struct ip_vs_service *svc)
350 if (svc->flags & IP_VS_SVC_F_HASHED) {
351 pr_err("%s(): request for already hashed, called from %pF\n",
352 __func__, __builtin_return_address(0));
356 if (svc->fwmark == 0) {
358 * Hash it by <protocol,addr,port> in ip_vs_svc_table
360 hash = ip_vs_svc_hashkey(svc->af, svc->protocol, &svc->addr,
362 list_add(&svc->s_list, &ip_vs_svc_table[hash]);
365 * Hash it by fwmark in ip_vs_svc_fwm_table
367 hash = ip_vs_svc_fwm_hashkey(svc->fwmark);
368 list_add(&svc->f_list, &ip_vs_svc_fwm_table[hash]);
371 svc->flags |= IP_VS_SVC_F_HASHED;
372 /* increase its refcnt because it is referenced by the svc table */
373 atomic_inc(&svc->refcnt);
379 * Unhashes a service from ip_vs_svc_table/ip_vs_svc_fwm_table.
380 * Should be called with locked tables.
382 static int ip_vs_svc_unhash(struct ip_vs_service *svc)
384 if (!(svc->flags & IP_VS_SVC_F_HASHED)) {
385 pr_err("%s(): request for unhash flagged, called from %pF\n",
386 __func__, __builtin_return_address(0));
390 if (svc->fwmark == 0) {
391 /* Remove it from the ip_vs_svc_table table */
392 list_del(&svc->s_list);
394 /* Remove it from the ip_vs_svc_fwm_table table */
395 list_del(&svc->f_list);
398 svc->flags &= ~IP_VS_SVC_F_HASHED;
399 atomic_dec(&svc->refcnt);
405 * Get service by {proto,addr,port} in the service table.
407 static inline struct ip_vs_service *
408 __ip_vs_service_find(int af, __u16 protocol, const union nf_inet_addr *vaddr,
412 struct ip_vs_service *svc;
414 /* Check for "full" addressed entries */
415 hash = ip_vs_svc_hashkey(af, protocol, vaddr, vport);
417 list_for_each_entry(svc, &ip_vs_svc_table[hash], s_list){
419 && ip_vs_addr_equal(af, &svc->addr, vaddr)
420 && (svc->port == vport)
421 && (svc->protocol == protocol)) {
432 * Get service by {fwmark} in the service table.
434 static inline struct ip_vs_service *
435 __ip_vs_svc_fwm_find(int af, __u32 fwmark)
438 struct ip_vs_service *svc;
440 /* Check for fwmark addressed entries */
441 hash = ip_vs_svc_fwm_hashkey(fwmark);
443 list_for_each_entry(svc, &ip_vs_svc_fwm_table[hash], f_list) {
444 if (svc->fwmark == fwmark && svc->af == af) {
453 struct ip_vs_service *
454 ip_vs_service_get(int af, __u32 fwmark, __u16 protocol,
455 const union nf_inet_addr *vaddr, __be16 vport)
457 struct ip_vs_service *svc;
459 read_lock(&__ip_vs_svc_lock);
462 * Check the table hashed by fwmark first
464 if (fwmark && (svc = __ip_vs_svc_fwm_find(af, fwmark)))
468 * Check the table hashed by <protocol,addr,port>
469 * for "full" addressed entries
471 svc = __ip_vs_service_find(af, protocol, vaddr, vport);
474 && protocol == IPPROTO_TCP
475 && atomic_read(&ip_vs_ftpsvc_counter)
476 && (vport == FTPDATA || ntohs(vport) >= PROT_SOCK)) {
478 * Check if ftp service entry exists, the packet
479 * might belong to FTP data connections.
481 svc = __ip_vs_service_find(af, protocol, vaddr, FTPPORT);
485 && atomic_read(&ip_vs_nullsvc_counter)) {
487 * Check if the catch-all port (port zero) exists
489 svc = __ip_vs_service_find(af, protocol, vaddr, 0);
494 atomic_inc(&svc->usecnt);
495 read_unlock(&__ip_vs_svc_lock);
497 IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
498 fwmark, ip_vs_proto_name(protocol),
499 IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
500 svc ? "hit" : "not hit");
507 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
509 atomic_inc(&svc->refcnt);
514 __ip_vs_unbind_svc(struct ip_vs_dest *dest)
516 struct ip_vs_service *svc = dest->svc;
519 if (atomic_dec_and_test(&svc->refcnt)) {
520 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
522 IP_VS_DBG_ADDR(svc->af, &svc->addr),
523 ntohs(svc->port), atomic_read(&svc->usecnt));
530 * Returns hash value for real service
532 static inline unsigned ip_vs_rs_hashkey(int af,
533 const union nf_inet_addr *addr,
536 register unsigned porth = ntohs(port);
537 __be32 addr_fold = addr->ip;
539 #ifdef CONFIG_IP_VS_IPV6
541 addr_fold = addr->ip6[0]^addr->ip6[1]^
542 addr->ip6[2]^addr->ip6[3];
545 return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
550 * Hashes ip_vs_dest in ip_vs_rtable by <proto,addr,port>.
551 * should be called with locked tables.
553 static int ip_vs_rs_hash(struct ip_vs_dest *dest)
557 if (!list_empty(&dest->d_list)) {
562 * Hash by proto,addr,port,
563 * which are the parameters of the real service.
565 hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
567 list_add(&dest->d_list, &ip_vs_rtable[hash]);
573 * UNhashes ip_vs_dest from ip_vs_rtable.
574 * should be called with locked tables.
576 static int ip_vs_rs_unhash(struct ip_vs_dest *dest)
579 * Remove it from the ip_vs_rtable table.
581 if (!list_empty(&dest->d_list)) {
582 list_del(&dest->d_list);
583 INIT_LIST_HEAD(&dest->d_list);
590 * Lookup real service by <proto,addr,port> in the real service table.
593 ip_vs_lookup_real_service(int af, __u16 protocol,
594 const union nf_inet_addr *daddr,
598 struct ip_vs_dest *dest;
601 * Check for "full" addressed entries
602 * Return the first found entry
604 hash = ip_vs_rs_hashkey(af, daddr, dport);
606 read_lock(&__ip_vs_rs_lock);
607 list_for_each_entry(dest, &ip_vs_rtable[hash], d_list) {
609 && ip_vs_addr_equal(af, &dest->addr, daddr)
610 && (dest->port == dport)
611 && ((dest->protocol == protocol) ||
614 read_unlock(&__ip_vs_rs_lock);
618 read_unlock(&__ip_vs_rs_lock);
624 * Lookup destination by {addr,port} in the given service
626 static struct ip_vs_dest *
627 ip_vs_lookup_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
630 struct ip_vs_dest *dest;
633 * Find the destination for the given service
635 list_for_each_entry(dest, &svc->destinations, n_list) {
636 if ((dest->af == svc->af)
637 && ip_vs_addr_equal(svc->af, &dest->addr, daddr)
638 && (dest->port == dport)) {
648 * Find destination by {daddr,dport,vaddr,protocol}
649 * Cretaed to be used in ip_vs_process_message() in
650 * the backup synchronization daemon. It finds the
651 * destination to be bound to the received connection
654 * ip_vs_lookup_real_service() looked promissing, but
655 * seems not working as expected.
657 struct ip_vs_dest *ip_vs_find_dest(int af, const union nf_inet_addr *daddr,
659 const union nf_inet_addr *vaddr,
660 __be16 vport, __u16 protocol)
662 struct ip_vs_dest *dest;
663 struct ip_vs_service *svc;
665 svc = ip_vs_service_get(af, 0, protocol, vaddr, vport);
668 dest = ip_vs_lookup_dest(svc, daddr, dport);
670 atomic_inc(&dest->refcnt);
671 ip_vs_service_put(svc);
676 * Lookup dest by {svc,addr,port} in the destination trash.
677 * The destination trash is used to hold the destinations that are removed
678 * from the service table but are still referenced by some conn entries.
679 * The reason to add the destination trash is when the dest is temporary
680 * down (either by administrator or by monitor program), the dest can be
681 * picked back from the trash, the remaining connections to the dest can
682 * continue, and the counting information of the dest is also useful for
685 static struct ip_vs_dest *
686 ip_vs_trash_get_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
689 struct ip_vs_dest *dest, *nxt;
692 * Find the destination in trash
694 list_for_each_entry_safe(dest, nxt, &ip_vs_dest_trash, n_list) {
695 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
698 IP_VS_DBG_ADDR(svc->af, &dest->addr),
700 atomic_read(&dest->refcnt));
701 if (dest->af == svc->af &&
702 ip_vs_addr_equal(svc->af, &dest->addr, daddr) &&
703 dest->port == dport &&
704 dest->vfwmark == svc->fwmark &&
705 dest->protocol == svc->protocol &&
707 (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
708 dest->vport == svc->port))) {
714 * Try to purge the destination from trash if not referenced
716 if (atomic_read(&dest->refcnt) == 1) {
717 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u "
720 IP_VS_DBG_ADDR(svc->af, &dest->addr),
722 list_del(&dest->n_list);
723 ip_vs_dst_reset(dest);
724 __ip_vs_unbind_svc(dest);
734 * Clean up all the destinations in the trash
735 * Called by the ip_vs_control_cleanup()
737 * When the ip_vs_control_clearup is activated by ipvs module exit,
738 * the service tables must have been flushed and all the connections
739 * are expired, and the refcnt of each destination in the trash must
740 * be 1, so we simply release them here.
742 static void ip_vs_trash_cleanup(void)
744 struct ip_vs_dest *dest, *nxt;
746 list_for_each_entry_safe(dest, nxt, &ip_vs_dest_trash, n_list) {
747 list_del(&dest->n_list);
748 ip_vs_dst_reset(dest);
749 __ip_vs_unbind_svc(dest);
756 ip_vs_zero_stats(struct ip_vs_stats *stats)
758 spin_lock_bh(&stats->lock);
760 memset(&stats->ustats, 0, sizeof(stats->ustats));
761 ip_vs_zero_estimator(stats);
763 spin_unlock_bh(&stats->lock);
767 * Update a destination in the given service
770 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
771 struct ip_vs_dest_user_kern *udest, int add)
775 /* set the weight and the flags */
776 atomic_set(&dest->weight, udest->weight);
777 conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
778 conn_flags |= IP_VS_CONN_F_INACTIVE;
780 /* check if local node and update the flags */
781 #ifdef CONFIG_IP_VS_IPV6
782 if (svc->af == AF_INET6) {
783 if (__ip_vs_addr_is_local_v6(&udest->addr.in6)) {
784 conn_flags = (conn_flags & ~IP_VS_CONN_F_FWD_MASK)
785 | IP_VS_CONN_F_LOCALNODE;
789 if (inet_addr_type(&init_net, udest->addr.ip) == RTN_LOCAL) {
790 conn_flags = (conn_flags & ~IP_VS_CONN_F_FWD_MASK)
791 | IP_VS_CONN_F_LOCALNODE;
794 /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
795 if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
796 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
799 * Put the real service in ip_vs_rtable if not present.
800 * For now only for NAT!
802 write_lock_bh(&__ip_vs_rs_lock);
804 write_unlock_bh(&__ip_vs_rs_lock);
806 atomic_set(&dest->conn_flags, conn_flags);
808 /* bind the service */
810 __ip_vs_bind_svc(dest, svc);
812 if (dest->svc != svc) {
813 __ip_vs_unbind_svc(dest);
814 ip_vs_zero_stats(&dest->stats);
815 __ip_vs_bind_svc(dest, svc);
819 /* set the dest status flags */
820 dest->flags |= IP_VS_DEST_F_AVAILABLE;
822 if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
823 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
824 dest->u_threshold = udest->u_threshold;
825 dest->l_threshold = udest->l_threshold;
828 ip_vs_new_estimator(&dest->stats);
830 write_lock_bh(&__ip_vs_svc_lock);
832 /* Wait until all other svc users go away */
833 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
836 list_add(&dest->n_list, &svc->destinations);
840 /* call the update_service, because server weight may be changed */
841 if (svc->scheduler->update_service)
842 svc->scheduler->update_service(svc);
844 write_unlock_bh(&__ip_vs_svc_lock);
849 * Create a destination for the given service
852 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
853 struct ip_vs_dest **dest_p)
855 struct ip_vs_dest *dest;
860 #ifdef CONFIG_IP_VS_IPV6
861 if (svc->af == AF_INET6) {
862 atype = ipv6_addr_type(&udest->addr.in6);
863 if ((!(atype & IPV6_ADDR_UNICAST) ||
864 atype & IPV6_ADDR_LINKLOCAL) &&
865 !__ip_vs_addr_is_local_v6(&udest->addr.in6))
870 atype = inet_addr_type(&init_net, udest->addr.ip);
871 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
875 dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
877 pr_err("%s(): no memory.\n", __func__);
882 dest->protocol = svc->protocol;
883 dest->vaddr = svc->addr;
884 dest->vport = svc->port;
885 dest->vfwmark = svc->fwmark;
886 ip_vs_addr_copy(svc->af, &dest->addr, &udest->addr);
887 dest->port = udest->port;
889 atomic_set(&dest->activeconns, 0);
890 atomic_set(&dest->inactconns, 0);
891 atomic_set(&dest->persistconns, 0);
892 atomic_set(&dest->refcnt, 1);
894 INIT_LIST_HEAD(&dest->d_list);
895 spin_lock_init(&dest->dst_lock);
896 spin_lock_init(&dest->stats.lock);
897 __ip_vs_update_dest(svc, dest, udest, 1);
907 * Add a destination into an existing service
910 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
912 struct ip_vs_dest *dest;
913 union nf_inet_addr daddr;
914 __be16 dport = udest->port;
919 if (udest->weight < 0) {
920 pr_err("%s(): server weight less than zero\n", __func__);
924 if (udest->l_threshold > udest->u_threshold) {
925 pr_err("%s(): lower threshold is higher than upper threshold\n",
930 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
933 * Check if the dest already exists in the list
935 dest = ip_vs_lookup_dest(svc, &daddr, dport);
938 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
943 * Check if the dest already exists in the trash and
944 * is from the same service
946 dest = ip_vs_trash_get_dest(svc, &daddr, dport);
949 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
950 "dest->refcnt=%d, service %u/%s:%u\n",
951 IP_VS_DBG_ADDR(svc->af, &daddr), ntohs(dport),
952 atomic_read(&dest->refcnt),
954 IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
958 * Get the destination from the trash
960 list_del(&dest->n_list);
962 __ip_vs_update_dest(svc, dest, udest, 1);
966 * Allocate and initialize the dest structure
968 ret = ip_vs_new_dest(svc, udest, &dest);
977 * Edit a destination in the given service
980 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
982 struct ip_vs_dest *dest;
983 union nf_inet_addr daddr;
984 __be16 dport = udest->port;
988 if (udest->weight < 0) {
989 pr_err("%s(): server weight less than zero\n", __func__);
993 if (udest->l_threshold > udest->u_threshold) {
994 pr_err("%s(): lower threshold is higher than upper threshold\n",
999 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
1002 * Lookup the destination list
1004 dest = ip_vs_lookup_dest(svc, &daddr, dport);
1007 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1011 __ip_vs_update_dest(svc, dest, udest, 0);
1019 * Delete a destination (must be already unlinked from the service)
1021 static void __ip_vs_del_dest(struct ip_vs_dest *dest)
1023 ip_vs_kill_estimator(&dest->stats);
1026 * Remove it from the d-linked list with the real services.
1028 write_lock_bh(&__ip_vs_rs_lock);
1029 ip_vs_rs_unhash(dest);
1030 write_unlock_bh(&__ip_vs_rs_lock);
1033 * Decrease the refcnt of the dest, and free the dest
1034 * if nobody refers to it (refcnt=0). Otherwise, throw
1035 * the destination into the trash.
1037 if (atomic_dec_and_test(&dest->refcnt)) {
1038 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u\n",
1040 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1042 ip_vs_dst_reset(dest);
1043 /* simply decrease svc->refcnt here, let the caller check
1044 and release the service if nobody refers to it.
1045 Only user context can release destination and service,
1046 and only one user context can update virtual service at a
1047 time, so the operation here is OK */
1048 atomic_dec(&dest->svc->refcnt);
1051 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, "
1052 "dest->refcnt=%d\n",
1053 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1055 atomic_read(&dest->refcnt));
1056 list_add(&dest->n_list, &ip_vs_dest_trash);
1057 atomic_inc(&dest->refcnt);
1063 * Unlink a destination from the given service
1065 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1066 struct ip_vs_dest *dest,
1069 dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1072 * Remove it from the d-linked destination list.
1074 list_del(&dest->n_list);
1078 * Call the update_service function of its scheduler
1080 if (svcupd && svc->scheduler->update_service)
1081 svc->scheduler->update_service(svc);
1086 * Delete a destination server in the given service
1089 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1091 struct ip_vs_dest *dest;
1092 __be16 dport = udest->port;
1096 dest = ip_vs_lookup_dest(svc, &udest->addr, dport);
1099 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1103 write_lock_bh(&__ip_vs_svc_lock);
1106 * Wait until all other svc users go away.
1108 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1111 * Unlink dest from the service
1113 __ip_vs_unlink_dest(svc, dest, 1);
1115 write_unlock_bh(&__ip_vs_svc_lock);
1118 * Delete the destination
1120 __ip_vs_del_dest(dest);
1129 * Add a service into the service hash table
1132 ip_vs_add_service(struct ip_vs_service_user_kern *u,
1133 struct ip_vs_service **svc_p)
1136 struct ip_vs_scheduler *sched = NULL;
1137 struct ip_vs_service *svc = NULL;
1139 /* increase the module use count */
1140 ip_vs_use_count_inc();
1142 /* Lookup the scheduler by 'u->sched_name' */
1143 sched = ip_vs_scheduler_get(u->sched_name);
1144 if (sched == NULL) {
1145 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1150 #ifdef CONFIG_IP_VS_IPV6
1151 if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1157 svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1159 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1164 /* I'm the first user of the service */
1165 atomic_set(&svc->usecnt, 0);
1166 atomic_set(&svc->refcnt, 0);
1169 svc->protocol = u->protocol;
1170 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1171 svc->port = u->port;
1172 svc->fwmark = u->fwmark;
1173 svc->flags = u->flags;
1174 svc->timeout = u->timeout * HZ;
1175 svc->netmask = u->netmask;
1177 INIT_LIST_HEAD(&svc->destinations);
1178 rwlock_init(&svc->sched_lock);
1179 spin_lock_init(&svc->stats.lock);
1181 /* Bind the scheduler */
1182 ret = ip_vs_bind_scheduler(svc, sched);
1187 /* Update the virtual service counters */
1188 if (svc->port == FTPPORT)
1189 atomic_inc(&ip_vs_ftpsvc_counter);
1190 else if (svc->port == 0)
1191 atomic_inc(&ip_vs_nullsvc_counter);
1193 ip_vs_new_estimator(&svc->stats);
1195 /* Count only IPv4 services for old get/setsockopt interface */
1196 if (svc->af == AF_INET)
1197 ip_vs_num_services++;
1199 /* Hash the service into the service table */
1200 write_lock_bh(&__ip_vs_svc_lock);
1201 ip_vs_svc_hash(svc);
1202 write_unlock_bh(&__ip_vs_svc_lock);
1209 ip_vs_unbind_scheduler(svc);
1212 ip_vs_app_inc_put(svc->inc);
1217 ip_vs_scheduler_put(sched);
1219 /* decrease the module use count */
1220 ip_vs_use_count_dec();
1227 * Edit a service and bind it with a new scheduler
1230 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1232 struct ip_vs_scheduler *sched, *old_sched;
1236 * Lookup the scheduler, by 'u->sched_name'
1238 sched = ip_vs_scheduler_get(u->sched_name);
1239 if (sched == NULL) {
1240 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1245 #ifdef CONFIG_IP_VS_IPV6
1246 if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1252 write_lock_bh(&__ip_vs_svc_lock);
1255 * Wait until all other svc users go away.
1257 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1260 * Set the flags and timeout value
1262 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1263 svc->timeout = u->timeout * HZ;
1264 svc->netmask = u->netmask;
1266 old_sched = svc->scheduler;
1267 if (sched != old_sched) {
1269 * Unbind the old scheduler
1271 if ((ret = ip_vs_unbind_scheduler(svc))) {
1277 * Bind the new scheduler
1279 if ((ret = ip_vs_bind_scheduler(svc, sched))) {
1281 * If ip_vs_bind_scheduler fails, restore the old
1283 * The main reason of failure is out of memory.
1285 * The question is if the old scheduler can be
1286 * restored all the time. TODO: if it cannot be
1287 * restored some time, we must delete the service,
1288 * otherwise the system may crash.
1290 ip_vs_bind_scheduler(svc, old_sched);
1297 write_unlock_bh(&__ip_vs_svc_lock);
1298 #ifdef CONFIG_IP_VS_IPV6
1301 ip_vs_scheduler_put(old_sched);
1307 * Delete a service from the service list
1308 * - The service must be unlinked, unlocked and not referenced!
1309 * - We are called under _bh lock
1311 static void __ip_vs_del_service(struct ip_vs_service *svc)
1313 struct ip_vs_dest *dest, *nxt;
1314 struct ip_vs_scheduler *old_sched;
1316 /* Count only IPv4 services for old get/setsockopt interface */
1317 if (svc->af == AF_INET)
1318 ip_vs_num_services--;
1320 ip_vs_kill_estimator(&svc->stats);
1322 /* Unbind scheduler */
1323 old_sched = svc->scheduler;
1324 ip_vs_unbind_scheduler(svc);
1325 ip_vs_scheduler_put(old_sched);
1327 /* Unbind app inc */
1329 ip_vs_app_inc_put(svc->inc);
1334 * Unlink the whole destination list
1336 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1337 __ip_vs_unlink_dest(svc, dest, 0);
1338 __ip_vs_del_dest(dest);
1342 * Update the virtual service counters
1344 if (svc->port == FTPPORT)
1345 atomic_dec(&ip_vs_ftpsvc_counter);
1346 else if (svc->port == 0)
1347 atomic_dec(&ip_vs_nullsvc_counter);
1350 * Free the service if nobody refers to it
1352 if (atomic_read(&svc->refcnt) == 0) {
1353 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
1355 IP_VS_DBG_ADDR(svc->af, &svc->addr),
1356 ntohs(svc->port), atomic_read(&svc->usecnt));
1360 /* decrease the module use count */
1361 ip_vs_use_count_dec();
1365 * Unlink a service from list and try to delete it if its refcnt reached 0
1367 static void ip_vs_unlink_service(struct ip_vs_service *svc)
1370 * Unhash it from the service table
1372 write_lock_bh(&__ip_vs_svc_lock);
1374 ip_vs_svc_unhash(svc);
1377 * Wait until all the svc users go away.
1379 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1381 __ip_vs_del_service(svc);
1383 write_unlock_bh(&__ip_vs_svc_lock);
1387 * Delete a service from the service list
1389 static int ip_vs_del_service(struct ip_vs_service *svc)
1393 ip_vs_unlink_service(svc);
1400 * Flush all the virtual services
1402 static int ip_vs_flush(void)
1405 struct ip_vs_service *svc, *nxt;
1408 * Flush the service table hashed by <protocol,addr,port>
1410 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1411 list_for_each_entry_safe(svc, nxt, &ip_vs_svc_table[idx], s_list) {
1412 ip_vs_unlink_service(svc);
1417 * Flush the service table hashed by fwmark
1419 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1420 list_for_each_entry_safe(svc, nxt,
1421 &ip_vs_svc_fwm_table[idx], f_list) {
1422 ip_vs_unlink_service(svc);
1431 * Zero counters in a service or all services
1433 static int ip_vs_zero_service(struct ip_vs_service *svc)
1435 struct ip_vs_dest *dest;
1437 write_lock_bh(&__ip_vs_svc_lock);
1438 list_for_each_entry(dest, &svc->destinations, n_list) {
1439 ip_vs_zero_stats(&dest->stats);
1441 ip_vs_zero_stats(&svc->stats);
1442 write_unlock_bh(&__ip_vs_svc_lock);
1446 static int ip_vs_zero_all(void)
1449 struct ip_vs_service *svc;
1451 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1452 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1453 ip_vs_zero_service(svc);
1457 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1458 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1459 ip_vs_zero_service(svc);
1463 ip_vs_zero_stats(&ip_vs_stats);
1469 proc_do_defense_mode(ctl_table *table, int write,
1470 void __user *buffer, size_t *lenp, loff_t *ppos)
1472 int *valp = table->data;
1476 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1477 if (write && (*valp != val)) {
1478 if ((*valp < 0) || (*valp > 3)) {
1479 /* Restore the correct value */
1482 update_defense_level();
1490 proc_do_sync_threshold(ctl_table *table, int write,
1491 void __user *buffer, size_t *lenp, loff_t *ppos)
1493 int *valp = table->data;
1497 /* backup the value first */
1498 memcpy(val, valp, sizeof(val));
1500 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1501 if (write && (valp[0] < 0 || valp[1] < 0 || valp[0] >= valp[1])) {
1502 /* Restore the correct value */
1503 memcpy(valp, val, sizeof(val));
1510 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1513 static struct ctl_table vs_vars[] = {
1515 .procname = "amemthresh",
1516 .data = &sysctl_ip_vs_amemthresh,
1517 .maxlen = sizeof(int),
1519 .proc_handler = proc_dointvec,
1521 #ifdef CONFIG_IP_VS_DEBUG
1523 .procname = "debug_level",
1524 .data = &sysctl_ip_vs_debug_level,
1525 .maxlen = sizeof(int),
1527 .proc_handler = proc_dointvec,
1531 .procname = "am_droprate",
1532 .data = &sysctl_ip_vs_am_droprate,
1533 .maxlen = sizeof(int),
1535 .proc_handler = proc_dointvec,
1538 .procname = "drop_entry",
1539 .data = &sysctl_ip_vs_drop_entry,
1540 .maxlen = sizeof(int),
1542 .proc_handler = proc_do_defense_mode,
1545 .procname = "drop_packet",
1546 .data = &sysctl_ip_vs_drop_packet,
1547 .maxlen = sizeof(int),
1549 .proc_handler = proc_do_defense_mode,
1551 #ifdef CONFIG_IP_VS_NFCT
1553 .procname = "conntrack",
1554 .data = &sysctl_ip_vs_conntrack,
1555 .maxlen = sizeof(int),
1557 .proc_handler = &proc_dointvec,
1561 .procname = "secure_tcp",
1562 .data = &sysctl_ip_vs_secure_tcp,
1563 .maxlen = sizeof(int),
1565 .proc_handler = proc_do_defense_mode,
1568 .procname = "snat_reroute",
1569 .data = &sysctl_ip_vs_snat_reroute,
1570 .maxlen = sizeof(int),
1572 .proc_handler = &proc_dointvec,
1576 .procname = "timeout_established",
1577 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1578 .maxlen = sizeof(int),
1580 .proc_handler = proc_dointvec_jiffies,
1583 .procname = "timeout_synsent",
1584 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1585 .maxlen = sizeof(int),
1587 .proc_handler = proc_dointvec_jiffies,
1590 .procname = "timeout_synrecv",
1591 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1592 .maxlen = sizeof(int),
1594 .proc_handler = proc_dointvec_jiffies,
1597 .procname = "timeout_finwait",
1598 .data = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1599 .maxlen = sizeof(int),
1601 .proc_handler = proc_dointvec_jiffies,
1604 .procname = "timeout_timewait",
1605 .data = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1606 .maxlen = sizeof(int),
1608 .proc_handler = proc_dointvec_jiffies,
1611 .procname = "timeout_close",
1612 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1613 .maxlen = sizeof(int),
1615 .proc_handler = proc_dointvec_jiffies,
1618 .procname = "timeout_closewait",
1619 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1620 .maxlen = sizeof(int),
1622 .proc_handler = proc_dointvec_jiffies,
1625 .procname = "timeout_lastack",
1626 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1627 .maxlen = sizeof(int),
1629 .proc_handler = proc_dointvec_jiffies,
1632 .procname = "timeout_listen",
1633 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1634 .maxlen = sizeof(int),
1636 .proc_handler = proc_dointvec_jiffies,
1639 .procname = "timeout_synack",
1640 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1641 .maxlen = sizeof(int),
1643 .proc_handler = proc_dointvec_jiffies,
1646 .procname = "timeout_udp",
1647 .data = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1648 .maxlen = sizeof(int),
1650 .proc_handler = proc_dointvec_jiffies,
1653 .procname = "timeout_icmp",
1654 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1655 .maxlen = sizeof(int),
1657 .proc_handler = proc_dointvec_jiffies,
1661 .procname = "cache_bypass",
1662 .data = &sysctl_ip_vs_cache_bypass,
1663 .maxlen = sizeof(int),
1665 .proc_handler = proc_dointvec,
1668 .procname = "expire_nodest_conn",
1669 .data = &sysctl_ip_vs_expire_nodest_conn,
1670 .maxlen = sizeof(int),
1672 .proc_handler = proc_dointvec,
1675 .procname = "expire_quiescent_template",
1676 .data = &sysctl_ip_vs_expire_quiescent_template,
1677 .maxlen = sizeof(int),
1679 .proc_handler = proc_dointvec,
1682 .procname = "sync_threshold",
1683 .data = &sysctl_ip_vs_sync_threshold,
1684 .maxlen = sizeof(sysctl_ip_vs_sync_threshold),
1686 .proc_handler = proc_do_sync_threshold,
1689 .procname = "nat_icmp_send",
1690 .data = &sysctl_ip_vs_nat_icmp_send,
1691 .maxlen = sizeof(int),
1693 .proc_handler = proc_dointvec,
1698 const struct ctl_path net_vs_ctl_path[] = {
1699 { .procname = "net", },
1700 { .procname = "ipv4", },
1701 { .procname = "vs", },
1704 EXPORT_SYMBOL_GPL(net_vs_ctl_path);
1706 static struct ctl_table_header * sysctl_header;
1708 #ifdef CONFIG_PROC_FS
1711 struct list_head *table;
1716 * Write the contents of the VS rule table to a PROCfs file.
1717 * (It is kept just for backward compatibility)
1719 static inline const char *ip_vs_fwd_name(unsigned flags)
1721 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1722 case IP_VS_CONN_F_LOCALNODE:
1724 case IP_VS_CONN_F_TUNNEL:
1726 case IP_VS_CONN_F_DROUTE:
1734 /* Get the Nth entry in the two lists */
1735 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1737 struct ip_vs_iter *iter = seq->private;
1739 struct ip_vs_service *svc;
1741 /* look in hash by protocol */
1742 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1743 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1745 iter->table = ip_vs_svc_table;
1752 /* keep looking in fwmark */
1753 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1754 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1756 iter->table = ip_vs_svc_fwm_table;
1766 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1767 __acquires(__ip_vs_svc_lock)
1770 read_lock_bh(&__ip_vs_svc_lock);
1771 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1775 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1777 struct list_head *e;
1778 struct ip_vs_iter *iter;
1779 struct ip_vs_service *svc;
1782 if (v == SEQ_START_TOKEN)
1783 return ip_vs_info_array(seq,0);
1786 iter = seq->private;
1788 if (iter->table == ip_vs_svc_table) {
1789 /* next service in table hashed by protocol */
1790 if ((e = svc->s_list.next) != &ip_vs_svc_table[iter->bucket])
1791 return list_entry(e, struct ip_vs_service, s_list);
1794 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1795 list_for_each_entry(svc,&ip_vs_svc_table[iter->bucket],
1801 iter->table = ip_vs_svc_fwm_table;
1806 /* next service in hashed by fwmark */
1807 if ((e = svc->f_list.next) != &ip_vs_svc_fwm_table[iter->bucket])
1808 return list_entry(e, struct ip_vs_service, f_list);
1811 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1812 list_for_each_entry(svc, &ip_vs_svc_fwm_table[iter->bucket],
1820 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
1821 __releases(__ip_vs_svc_lock)
1823 read_unlock_bh(&__ip_vs_svc_lock);
1827 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
1829 if (v == SEQ_START_TOKEN) {
1831 "IP Virtual Server version %d.%d.%d (size=%d)\n",
1832 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
1834 "Prot LocalAddress:Port Scheduler Flags\n");
1836 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
1838 const struct ip_vs_service *svc = v;
1839 const struct ip_vs_iter *iter = seq->private;
1840 const struct ip_vs_dest *dest;
1842 if (iter->table == ip_vs_svc_table) {
1843 #ifdef CONFIG_IP_VS_IPV6
1844 if (svc->af == AF_INET6)
1845 seq_printf(seq, "%s [%pI6]:%04X %s ",
1846 ip_vs_proto_name(svc->protocol),
1849 svc->scheduler->name);
1852 seq_printf(seq, "%s %08X:%04X %s %s ",
1853 ip_vs_proto_name(svc->protocol),
1854 ntohl(svc->addr.ip),
1856 svc->scheduler->name,
1857 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
1859 seq_printf(seq, "FWM %08X %s %s",
1860 svc->fwmark, svc->scheduler->name,
1861 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
1864 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
1865 seq_printf(seq, "persistent %d %08X\n",
1867 ntohl(svc->netmask));
1869 seq_putc(seq, '\n');
1871 list_for_each_entry(dest, &svc->destinations, n_list) {
1872 #ifdef CONFIG_IP_VS_IPV6
1873 if (dest->af == AF_INET6)
1876 " %-7s %-6d %-10d %-10d\n",
1879 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
1880 atomic_read(&dest->weight),
1881 atomic_read(&dest->activeconns),
1882 atomic_read(&dest->inactconns));
1887 "%-7s %-6d %-10d %-10d\n",
1888 ntohl(dest->addr.ip),
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));
1900 static const struct seq_operations ip_vs_info_seq_ops = {
1901 .start = ip_vs_info_seq_start,
1902 .next = ip_vs_info_seq_next,
1903 .stop = ip_vs_info_seq_stop,
1904 .show = ip_vs_info_seq_show,
1907 static int ip_vs_info_open(struct inode *inode, struct file *file)
1909 return seq_open_private(file, &ip_vs_info_seq_ops,
1910 sizeof(struct ip_vs_iter));
1913 static const struct file_operations ip_vs_info_fops = {
1914 .owner = THIS_MODULE,
1915 .open = ip_vs_info_open,
1917 .llseek = seq_lseek,
1918 .release = seq_release_private,
1923 struct ip_vs_stats ip_vs_stats = {
1924 .lock = __SPIN_LOCK_UNLOCKED(ip_vs_stats.lock),
1927 #ifdef CONFIG_PROC_FS
1928 static int ip_vs_stats_show(struct seq_file *seq, void *v)
1931 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
1933 " Total Incoming Outgoing Incoming Outgoing\n");
1935 " Conns Packets Packets Bytes Bytes\n");
1937 spin_lock_bh(&ip_vs_stats.lock);
1938 seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", ip_vs_stats.ustats.conns,
1939 ip_vs_stats.ustats.inpkts, ip_vs_stats.ustats.outpkts,
1940 (unsigned long long) ip_vs_stats.ustats.inbytes,
1941 (unsigned long long) ip_vs_stats.ustats.outbytes);
1943 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
1945 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
1946 seq_printf(seq,"%8X %8X %8X %16X %16X\n",
1947 ip_vs_stats.ustats.cps,
1948 ip_vs_stats.ustats.inpps,
1949 ip_vs_stats.ustats.outpps,
1950 ip_vs_stats.ustats.inbps,
1951 ip_vs_stats.ustats.outbps);
1952 spin_unlock_bh(&ip_vs_stats.lock);
1957 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
1959 return single_open(file, ip_vs_stats_show, NULL);
1962 static const struct file_operations ip_vs_stats_fops = {
1963 .owner = THIS_MODULE,
1964 .open = ip_vs_stats_seq_open,
1966 .llseek = seq_lseek,
1967 .release = single_release,
1973 * Set timeout values for tcp tcpfin udp in the timeout_table.
1975 static int ip_vs_set_timeout(struct ip_vs_timeout_user *u)
1977 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
1982 #ifdef CONFIG_IP_VS_PROTO_TCP
1983 if (u->tcp_timeout) {
1984 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_ESTABLISHED]
1985 = u->tcp_timeout * HZ;
1988 if (u->tcp_fin_timeout) {
1989 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_FIN_WAIT]
1990 = u->tcp_fin_timeout * HZ;
1994 #ifdef CONFIG_IP_VS_PROTO_UDP
1995 if (u->udp_timeout) {
1996 ip_vs_protocol_udp.timeout_table[IP_VS_UDP_S_NORMAL]
1997 = u->udp_timeout * HZ;
2004 #define SET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2005 #define SERVICE_ARG_LEN (sizeof(struct ip_vs_service_user))
2006 #define SVCDEST_ARG_LEN (sizeof(struct ip_vs_service_user) + \
2007 sizeof(struct ip_vs_dest_user))
2008 #define TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2009 #define DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user))
2010 #define MAX_ARG_LEN SVCDEST_ARG_LEN
2012 static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
2013 [SET_CMDID(IP_VS_SO_SET_ADD)] = SERVICE_ARG_LEN,
2014 [SET_CMDID(IP_VS_SO_SET_EDIT)] = SERVICE_ARG_LEN,
2015 [SET_CMDID(IP_VS_SO_SET_DEL)] = SERVICE_ARG_LEN,
2016 [SET_CMDID(IP_VS_SO_SET_FLUSH)] = 0,
2017 [SET_CMDID(IP_VS_SO_SET_ADDDEST)] = SVCDEST_ARG_LEN,
2018 [SET_CMDID(IP_VS_SO_SET_DELDEST)] = SVCDEST_ARG_LEN,
2019 [SET_CMDID(IP_VS_SO_SET_EDITDEST)] = SVCDEST_ARG_LEN,
2020 [SET_CMDID(IP_VS_SO_SET_TIMEOUT)] = TIMEOUT_ARG_LEN,
2021 [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)] = DAEMON_ARG_LEN,
2022 [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)] = DAEMON_ARG_LEN,
2023 [SET_CMDID(IP_VS_SO_SET_ZERO)] = SERVICE_ARG_LEN,
2026 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2027 struct ip_vs_service_user *usvc_compat)
2030 usvc->protocol = usvc_compat->protocol;
2031 usvc->addr.ip = usvc_compat->addr;
2032 usvc->port = usvc_compat->port;
2033 usvc->fwmark = usvc_compat->fwmark;
2035 /* Deep copy of sched_name is not needed here */
2036 usvc->sched_name = usvc_compat->sched_name;
2038 usvc->flags = usvc_compat->flags;
2039 usvc->timeout = usvc_compat->timeout;
2040 usvc->netmask = usvc_compat->netmask;
2043 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2044 struct ip_vs_dest_user *udest_compat)
2046 udest->addr.ip = udest_compat->addr;
2047 udest->port = udest_compat->port;
2048 udest->conn_flags = udest_compat->conn_flags;
2049 udest->weight = udest_compat->weight;
2050 udest->u_threshold = udest_compat->u_threshold;
2051 udest->l_threshold = udest_compat->l_threshold;
2055 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2058 unsigned char arg[MAX_ARG_LEN];
2059 struct ip_vs_service_user *usvc_compat;
2060 struct ip_vs_service_user_kern usvc;
2061 struct ip_vs_service *svc;
2062 struct ip_vs_dest_user *udest_compat;
2063 struct ip_vs_dest_user_kern udest;
2065 if (!capable(CAP_NET_ADMIN))
2068 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2070 if (len < 0 || len > MAX_ARG_LEN)
2072 if (len != set_arglen[SET_CMDID(cmd)]) {
2073 pr_err("set_ctl: len %u != %u\n",
2074 len, set_arglen[SET_CMDID(cmd)]);
2078 if (copy_from_user(arg, user, len) != 0)
2081 /* increase the module use count */
2082 ip_vs_use_count_inc();
2084 if (mutex_lock_interruptible(&__ip_vs_mutex)) {
2089 if (cmd == IP_VS_SO_SET_FLUSH) {
2090 /* Flush the virtual service */
2091 ret = ip_vs_flush();
2093 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2094 /* Set timeout values for (tcp tcpfin udp) */
2095 ret = ip_vs_set_timeout((struct ip_vs_timeout_user *)arg);
2097 } else if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2098 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2099 ret = start_sync_thread(dm->state, dm->mcast_ifn, dm->syncid);
2101 } else if (cmd == IP_VS_SO_SET_STOPDAEMON) {
2102 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2103 ret = stop_sync_thread(dm->state);
2107 usvc_compat = (struct ip_vs_service_user *)arg;
2108 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2110 /* We only use the new structs internally, so copy userspace compat
2111 * structs to extended internal versions */
2112 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2113 ip_vs_copy_udest_compat(&udest, udest_compat);
2115 if (cmd == IP_VS_SO_SET_ZERO) {
2116 /* if no service address is set, zero counters in all */
2117 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2118 ret = ip_vs_zero_all();
2123 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2124 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2125 usvc.protocol != IPPROTO_SCTP) {
2126 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2127 usvc.protocol, &usvc.addr.ip,
2128 ntohs(usvc.port), usvc.sched_name);
2133 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2134 if (usvc.fwmark == 0)
2135 svc = __ip_vs_service_find(usvc.af, usvc.protocol,
2136 &usvc.addr, usvc.port);
2138 svc = __ip_vs_svc_fwm_find(usvc.af, usvc.fwmark);
2140 if (cmd != IP_VS_SO_SET_ADD
2141 && (svc == NULL || svc->protocol != usvc.protocol)) {
2147 case IP_VS_SO_SET_ADD:
2151 ret = ip_vs_add_service(&usvc, &svc);
2153 case IP_VS_SO_SET_EDIT:
2154 ret = ip_vs_edit_service(svc, &usvc);
2156 case IP_VS_SO_SET_DEL:
2157 ret = ip_vs_del_service(svc);
2161 case IP_VS_SO_SET_ZERO:
2162 ret = ip_vs_zero_service(svc);
2164 case IP_VS_SO_SET_ADDDEST:
2165 ret = ip_vs_add_dest(svc, &udest);
2167 case IP_VS_SO_SET_EDITDEST:
2168 ret = ip_vs_edit_dest(svc, &udest);
2170 case IP_VS_SO_SET_DELDEST:
2171 ret = ip_vs_del_dest(svc, &udest);
2178 mutex_unlock(&__ip_vs_mutex);
2180 /* decrease the module use count */
2181 ip_vs_use_count_dec();
2188 ip_vs_copy_stats(struct ip_vs_stats_user *dst, struct ip_vs_stats *src)
2190 spin_lock_bh(&src->lock);
2191 memcpy(dst, &src->ustats, sizeof(*dst));
2192 spin_unlock_bh(&src->lock);
2196 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2198 dst->protocol = src->protocol;
2199 dst->addr = src->addr.ip;
2200 dst->port = src->port;
2201 dst->fwmark = src->fwmark;
2202 strlcpy(dst->sched_name, src->scheduler->name, sizeof(dst->sched_name));
2203 dst->flags = src->flags;
2204 dst->timeout = src->timeout / HZ;
2205 dst->netmask = src->netmask;
2206 dst->num_dests = src->num_dests;
2207 ip_vs_copy_stats(&dst->stats, &src->stats);
2211 __ip_vs_get_service_entries(const struct ip_vs_get_services *get,
2212 struct ip_vs_get_services __user *uptr)
2215 struct ip_vs_service *svc;
2216 struct ip_vs_service_entry entry;
2219 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2220 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2221 /* Only expose IPv4 entries to old interface */
2222 if (svc->af != AF_INET)
2225 if (count >= get->num_services)
2227 memset(&entry, 0, sizeof(entry));
2228 ip_vs_copy_service(&entry, svc);
2229 if (copy_to_user(&uptr->entrytable[count],
2230 &entry, sizeof(entry))) {
2238 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2239 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2240 /* Only expose IPv4 entries to old interface */
2241 if (svc->af != AF_INET)
2244 if (count >= get->num_services)
2246 memset(&entry, 0, sizeof(entry));
2247 ip_vs_copy_service(&entry, svc);
2248 if (copy_to_user(&uptr->entrytable[count],
2249 &entry, sizeof(entry))) {
2261 __ip_vs_get_dest_entries(const struct ip_vs_get_dests *get,
2262 struct ip_vs_get_dests __user *uptr)
2264 struct ip_vs_service *svc;
2265 union nf_inet_addr addr = { .ip = get->addr };
2269 svc = __ip_vs_svc_fwm_find(AF_INET, get->fwmark);
2271 svc = __ip_vs_service_find(AF_INET, get->protocol, &addr,
2276 struct ip_vs_dest *dest;
2277 struct ip_vs_dest_entry entry;
2279 list_for_each_entry(dest, &svc->destinations, n_list) {
2280 if (count >= get->num_dests)
2283 entry.addr = dest->addr.ip;
2284 entry.port = dest->port;
2285 entry.conn_flags = atomic_read(&dest->conn_flags);
2286 entry.weight = atomic_read(&dest->weight);
2287 entry.u_threshold = dest->u_threshold;
2288 entry.l_threshold = dest->l_threshold;
2289 entry.activeconns = atomic_read(&dest->activeconns);
2290 entry.inactconns = atomic_read(&dest->inactconns);
2291 entry.persistconns = atomic_read(&dest->persistconns);
2292 ip_vs_copy_stats(&entry.stats, &dest->stats);
2293 if (copy_to_user(&uptr->entrytable[count],
2294 &entry, sizeof(entry))) {
2306 __ip_vs_get_timeouts(struct ip_vs_timeout_user *u)
2308 #ifdef CONFIG_IP_VS_PROTO_TCP
2310 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2311 u->tcp_fin_timeout =
2312 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2314 #ifdef CONFIG_IP_VS_PROTO_UDP
2316 ip_vs_protocol_udp.timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2321 #define GET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2322 #define GET_INFO_ARG_LEN (sizeof(struct ip_vs_getinfo))
2323 #define GET_SERVICES_ARG_LEN (sizeof(struct ip_vs_get_services))
2324 #define GET_SERVICE_ARG_LEN (sizeof(struct ip_vs_service_entry))
2325 #define GET_DESTS_ARG_LEN (sizeof(struct ip_vs_get_dests))
2326 #define GET_TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2327 #define GET_DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user) * 2)
2329 static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
2330 [GET_CMDID(IP_VS_SO_GET_VERSION)] = 64,
2331 [GET_CMDID(IP_VS_SO_GET_INFO)] = GET_INFO_ARG_LEN,
2332 [GET_CMDID(IP_VS_SO_GET_SERVICES)] = GET_SERVICES_ARG_LEN,
2333 [GET_CMDID(IP_VS_SO_GET_SERVICE)] = GET_SERVICE_ARG_LEN,
2334 [GET_CMDID(IP_VS_SO_GET_DESTS)] = GET_DESTS_ARG_LEN,
2335 [GET_CMDID(IP_VS_SO_GET_TIMEOUT)] = GET_TIMEOUT_ARG_LEN,
2336 [GET_CMDID(IP_VS_SO_GET_DAEMON)] = GET_DAEMON_ARG_LEN,
2340 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2342 unsigned char arg[128];
2344 unsigned int copylen;
2346 if (!capable(CAP_NET_ADMIN))
2349 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2352 if (*len < get_arglen[GET_CMDID(cmd)]) {
2353 pr_err("get_ctl: len %u < %u\n",
2354 *len, get_arglen[GET_CMDID(cmd)]);
2358 copylen = get_arglen[GET_CMDID(cmd)];
2362 if (copy_from_user(arg, user, copylen) != 0)
2365 if (mutex_lock_interruptible(&__ip_vs_mutex))
2366 return -ERESTARTSYS;
2369 case IP_VS_SO_GET_VERSION:
2373 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2374 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2375 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2379 *len = strlen(buf)+1;
2383 case IP_VS_SO_GET_INFO:
2385 struct ip_vs_getinfo info;
2386 info.version = IP_VS_VERSION_CODE;
2387 info.size = ip_vs_conn_tab_size;
2388 info.num_services = ip_vs_num_services;
2389 if (copy_to_user(user, &info, sizeof(info)) != 0)
2394 case IP_VS_SO_GET_SERVICES:
2396 struct ip_vs_get_services *get;
2399 get = (struct ip_vs_get_services *)arg;
2400 size = sizeof(*get) +
2401 sizeof(struct ip_vs_service_entry) * get->num_services;
2403 pr_err("length: %u != %u\n", *len, size);
2407 ret = __ip_vs_get_service_entries(get, user);
2411 case IP_VS_SO_GET_SERVICE:
2413 struct ip_vs_service_entry *entry;
2414 struct ip_vs_service *svc;
2415 union nf_inet_addr addr;
2417 entry = (struct ip_vs_service_entry *)arg;
2418 addr.ip = entry->addr;
2420 svc = __ip_vs_svc_fwm_find(AF_INET, entry->fwmark);
2422 svc = __ip_vs_service_find(AF_INET, entry->protocol,
2423 &addr, entry->port);
2425 ip_vs_copy_service(entry, svc);
2426 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2433 case IP_VS_SO_GET_DESTS:
2435 struct ip_vs_get_dests *get;
2438 get = (struct ip_vs_get_dests *)arg;
2439 size = sizeof(*get) +
2440 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2442 pr_err("length: %u != %u\n", *len, size);
2446 ret = __ip_vs_get_dest_entries(get, user);
2450 case IP_VS_SO_GET_TIMEOUT:
2452 struct ip_vs_timeout_user t;
2454 __ip_vs_get_timeouts(&t);
2455 if (copy_to_user(user, &t, sizeof(t)) != 0)
2460 case IP_VS_SO_GET_DAEMON:
2462 struct ip_vs_daemon_user d[2];
2464 memset(&d, 0, sizeof(d));
2465 if (ip_vs_sync_state & IP_VS_STATE_MASTER) {
2466 d[0].state = IP_VS_STATE_MASTER;
2467 strlcpy(d[0].mcast_ifn, ip_vs_master_mcast_ifn, sizeof(d[0].mcast_ifn));
2468 d[0].syncid = ip_vs_master_syncid;
2470 if (ip_vs_sync_state & IP_VS_STATE_BACKUP) {
2471 d[1].state = IP_VS_STATE_BACKUP;
2472 strlcpy(d[1].mcast_ifn, ip_vs_backup_mcast_ifn, sizeof(d[1].mcast_ifn));
2473 d[1].syncid = ip_vs_backup_syncid;
2475 if (copy_to_user(user, &d, sizeof(d)) != 0)
2485 mutex_unlock(&__ip_vs_mutex);
2490 static struct nf_sockopt_ops ip_vs_sockopts = {
2492 .set_optmin = IP_VS_BASE_CTL,
2493 .set_optmax = IP_VS_SO_SET_MAX+1,
2494 .set = do_ip_vs_set_ctl,
2495 .get_optmin = IP_VS_BASE_CTL,
2496 .get_optmax = IP_VS_SO_GET_MAX+1,
2497 .get = do_ip_vs_get_ctl,
2498 .owner = THIS_MODULE,
2502 * Generic Netlink interface
2505 /* IPVS genetlink family */
2506 static struct genl_family ip_vs_genl_family = {
2507 .id = GENL_ID_GENERATE,
2509 .name = IPVS_GENL_NAME,
2510 .version = IPVS_GENL_VERSION,
2511 .maxattr = IPVS_CMD_MAX,
2514 /* Policy used for first-level command attributes */
2515 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2516 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2517 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2518 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2519 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2520 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2521 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2524 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2525 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2526 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2527 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2528 .len = IP_VS_IFNAME_MAXLEN },
2529 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2532 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2533 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2534 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2535 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2536 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2537 .len = sizeof(union nf_inet_addr) },
2538 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2539 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2540 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2541 .len = IP_VS_SCHEDNAME_MAXLEN },
2542 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2543 .len = sizeof(struct ip_vs_flags) },
2544 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2545 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2546 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2549 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2550 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2551 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2552 .len = sizeof(union nf_inet_addr) },
2553 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2554 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2555 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2556 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2557 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2558 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2559 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2560 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2561 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2564 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2565 struct ip_vs_stats *stats)
2567 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2571 spin_lock_bh(&stats->lock);
2573 NLA_PUT_U32(skb, IPVS_STATS_ATTR_CONNS, stats->ustats.conns);
2574 NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPKTS, stats->ustats.inpkts);
2575 NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPKTS, stats->ustats.outpkts);
2576 NLA_PUT_U64(skb, IPVS_STATS_ATTR_INBYTES, stats->ustats.inbytes);
2577 NLA_PUT_U64(skb, IPVS_STATS_ATTR_OUTBYTES, stats->ustats.outbytes);
2578 NLA_PUT_U32(skb, IPVS_STATS_ATTR_CPS, stats->ustats.cps);
2579 NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPPS, stats->ustats.inpps);
2580 NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPPS, stats->ustats.outpps);
2581 NLA_PUT_U32(skb, IPVS_STATS_ATTR_INBPS, stats->ustats.inbps);
2582 NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTBPS, stats->ustats.outbps);
2584 spin_unlock_bh(&stats->lock);
2586 nla_nest_end(skb, nl_stats);
2591 spin_unlock_bh(&stats->lock);
2592 nla_nest_cancel(skb, nl_stats);
2596 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2597 struct ip_vs_service *svc)
2599 struct nlattr *nl_service;
2600 struct ip_vs_flags flags = { .flags = svc->flags,
2603 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2607 NLA_PUT_U16(skb, IPVS_SVC_ATTR_AF, svc->af);
2610 NLA_PUT_U32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark);
2612 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol);
2613 NLA_PUT(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr);
2614 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PORT, svc->port);
2617 NLA_PUT_STRING(skb, IPVS_SVC_ATTR_SCHED_NAME, svc->scheduler->name);
2618 NLA_PUT(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags);
2619 NLA_PUT_U32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ);
2620 NLA_PUT_U32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask);
2622 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2623 goto nla_put_failure;
2625 nla_nest_end(skb, nl_service);
2630 nla_nest_cancel(skb, nl_service);
2634 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2635 struct ip_vs_service *svc,
2636 struct netlink_callback *cb)
2640 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2641 &ip_vs_genl_family, NLM_F_MULTI,
2642 IPVS_CMD_NEW_SERVICE);
2646 if (ip_vs_genl_fill_service(skb, svc) < 0)
2647 goto nla_put_failure;
2649 return genlmsg_end(skb, hdr);
2652 genlmsg_cancel(skb, hdr);
2656 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2657 struct netlink_callback *cb)
2660 int start = cb->args[0];
2661 struct ip_vs_service *svc;
2663 mutex_lock(&__ip_vs_mutex);
2664 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2665 list_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2668 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2670 goto nla_put_failure;
2675 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2676 list_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
2679 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2681 goto nla_put_failure;
2687 mutex_unlock(&__ip_vs_mutex);
2693 static int ip_vs_genl_parse_service(struct ip_vs_service_user_kern *usvc,
2694 struct nlattr *nla, int full_entry,
2695 struct ip_vs_service **ret_svc)
2697 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
2698 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
2699 struct ip_vs_service *svc;
2701 /* Parse mandatory identifying service fields first */
2703 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
2706 nla_af = attrs[IPVS_SVC_ATTR_AF];
2707 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
2708 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
2709 nla_port = attrs[IPVS_SVC_ATTR_PORT];
2710 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
2712 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
2715 memset(usvc, 0, sizeof(*usvc));
2717 usvc->af = nla_get_u16(nla_af);
2718 #ifdef CONFIG_IP_VS_IPV6
2719 if (usvc->af != AF_INET && usvc->af != AF_INET6)
2721 if (usvc->af != AF_INET)
2723 return -EAFNOSUPPORT;
2726 usvc->protocol = IPPROTO_TCP;
2727 usvc->fwmark = nla_get_u32(nla_fwmark);
2729 usvc->protocol = nla_get_u16(nla_protocol);
2730 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
2731 usvc->port = nla_get_u16(nla_port);
2736 svc = __ip_vs_svc_fwm_find(usvc->af, usvc->fwmark);
2738 svc = __ip_vs_service_find(usvc->af, usvc->protocol,
2739 &usvc->addr, usvc->port);
2742 /* If a full entry was requested, check for the additional fields */
2744 struct nlattr *nla_sched, *nla_flags, *nla_timeout,
2746 struct ip_vs_flags flags;
2748 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
2749 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
2750 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
2751 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
2753 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
2756 nla_memcpy(&flags, nla_flags, sizeof(flags));
2758 /* prefill flags from service if it already exists */
2760 usvc->flags = svc->flags;
2762 /* set new flags from userland */
2763 usvc->flags = (usvc->flags & ~flags.mask) |
2764 (flags.flags & flags.mask);
2765 usvc->sched_name = nla_data(nla_sched);
2766 usvc->timeout = nla_get_u32(nla_timeout);
2767 usvc->netmask = nla_get_u32(nla_netmask);
2773 static struct ip_vs_service *ip_vs_genl_find_service(struct nlattr *nla)
2775 struct ip_vs_service_user_kern usvc;
2776 struct ip_vs_service *svc;
2779 ret = ip_vs_genl_parse_service(&usvc, nla, 0, &svc);
2780 return ret ? ERR_PTR(ret) : svc;
2783 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
2785 struct nlattr *nl_dest;
2787 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
2791 NLA_PUT(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr);
2792 NLA_PUT_U16(skb, IPVS_DEST_ATTR_PORT, dest->port);
2794 NLA_PUT_U32(skb, IPVS_DEST_ATTR_FWD_METHOD,
2795 atomic_read(&dest->conn_flags) & IP_VS_CONN_F_FWD_MASK);
2796 NLA_PUT_U32(skb, IPVS_DEST_ATTR_WEIGHT, atomic_read(&dest->weight));
2797 NLA_PUT_U32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold);
2798 NLA_PUT_U32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold);
2799 NLA_PUT_U32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
2800 atomic_read(&dest->activeconns));
2801 NLA_PUT_U32(skb, IPVS_DEST_ATTR_INACT_CONNS,
2802 atomic_read(&dest->inactconns));
2803 NLA_PUT_U32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
2804 atomic_read(&dest->persistconns));
2806 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
2807 goto nla_put_failure;
2809 nla_nest_end(skb, nl_dest);
2814 nla_nest_cancel(skb, nl_dest);
2818 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
2819 struct netlink_callback *cb)
2823 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2824 &ip_vs_genl_family, NLM_F_MULTI,
2829 if (ip_vs_genl_fill_dest(skb, dest) < 0)
2830 goto nla_put_failure;
2832 return genlmsg_end(skb, hdr);
2835 genlmsg_cancel(skb, hdr);
2839 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
2840 struct netlink_callback *cb)
2843 int start = cb->args[0];
2844 struct ip_vs_service *svc;
2845 struct ip_vs_dest *dest;
2846 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
2848 mutex_lock(&__ip_vs_mutex);
2850 /* Try to find the service for which to dump destinations */
2851 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
2852 IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
2855 svc = ip_vs_genl_find_service(attrs[IPVS_CMD_ATTR_SERVICE]);
2856 if (IS_ERR(svc) || svc == NULL)
2859 /* Dump the destinations */
2860 list_for_each_entry(dest, &svc->destinations, n_list) {
2863 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
2865 goto nla_put_failure;
2873 mutex_unlock(&__ip_vs_mutex);
2878 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
2879 struct nlattr *nla, int full_entry)
2881 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
2882 struct nlattr *nla_addr, *nla_port;
2884 /* Parse mandatory identifying destination fields first */
2886 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
2889 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
2890 nla_port = attrs[IPVS_DEST_ATTR_PORT];
2892 if (!(nla_addr && nla_port))
2895 memset(udest, 0, sizeof(*udest));
2897 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
2898 udest->port = nla_get_u16(nla_port);
2900 /* If a full entry was requested, check for the additional fields */
2902 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
2905 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
2906 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
2907 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
2908 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
2910 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
2913 udest->conn_flags = nla_get_u32(nla_fwd)
2914 & IP_VS_CONN_F_FWD_MASK;
2915 udest->weight = nla_get_u32(nla_weight);
2916 udest->u_threshold = nla_get_u32(nla_u_thresh);
2917 udest->l_threshold = nla_get_u32(nla_l_thresh);
2923 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __be32 state,
2924 const char *mcast_ifn, __be32 syncid)
2926 struct nlattr *nl_daemon;
2928 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
2932 NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_STATE, state);
2933 NLA_PUT_STRING(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn);
2934 NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid);
2936 nla_nest_end(skb, nl_daemon);
2941 nla_nest_cancel(skb, nl_daemon);
2945 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __be32 state,
2946 const char *mcast_ifn, __be32 syncid,
2947 struct netlink_callback *cb)
2950 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2951 &ip_vs_genl_family, NLM_F_MULTI,
2952 IPVS_CMD_NEW_DAEMON);
2956 if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
2957 goto nla_put_failure;
2959 return genlmsg_end(skb, hdr);
2962 genlmsg_cancel(skb, hdr);
2966 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
2967 struct netlink_callback *cb)
2969 mutex_lock(&__ip_vs_mutex);
2970 if ((ip_vs_sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
2971 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
2972 ip_vs_master_mcast_ifn,
2973 ip_vs_master_syncid, cb) < 0)
2974 goto nla_put_failure;
2979 if ((ip_vs_sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
2980 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
2981 ip_vs_backup_mcast_ifn,
2982 ip_vs_backup_syncid, cb) < 0)
2983 goto nla_put_failure;
2989 mutex_unlock(&__ip_vs_mutex);
2994 static int ip_vs_genl_new_daemon(struct nlattr **attrs)
2996 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
2997 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
2998 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3001 return start_sync_thread(nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3002 nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3003 nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3006 static int ip_vs_genl_del_daemon(struct nlattr **attrs)
3008 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3011 return stop_sync_thread(nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3014 static int ip_vs_genl_set_config(struct nlattr **attrs)
3016 struct ip_vs_timeout_user t;
3018 __ip_vs_get_timeouts(&t);
3020 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3021 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3023 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3025 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3027 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3028 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3030 return ip_vs_set_timeout(&t);
3033 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3035 struct ip_vs_service *svc = NULL;
3036 struct ip_vs_service_user_kern usvc;
3037 struct ip_vs_dest_user_kern udest;
3039 int need_full_svc = 0, need_full_dest = 0;
3041 cmd = info->genlhdr->cmd;
3043 mutex_lock(&__ip_vs_mutex);
3045 if (cmd == IPVS_CMD_FLUSH) {
3046 ret = ip_vs_flush();
3048 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3049 ret = ip_vs_genl_set_config(info->attrs);
3051 } else if (cmd == IPVS_CMD_NEW_DAEMON ||
3052 cmd == IPVS_CMD_DEL_DAEMON) {
3054 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3056 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3057 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3058 info->attrs[IPVS_CMD_ATTR_DAEMON],
3059 ip_vs_daemon_policy)) {
3064 if (cmd == IPVS_CMD_NEW_DAEMON)
3065 ret = ip_vs_genl_new_daemon(daemon_attrs);
3067 ret = ip_vs_genl_del_daemon(daemon_attrs);
3069 } else if (cmd == IPVS_CMD_ZERO &&
3070 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3071 ret = ip_vs_zero_all();
3075 /* All following commands require a service argument, so check if we
3076 * received a valid one. We need a full service specification when
3077 * adding / editing a service. Only identifying members otherwise. */
3078 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3081 ret = ip_vs_genl_parse_service(&usvc,
3082 info->attrs[IPVS_CMD_ATTR_SERVICE],
3083 need_full_svc, &svc);
3087 /* Unless we're adding a new service, the service must already exist */
3088 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3093 /* Destination commands require a valid destination argument. For
3094 * adding / editing a destination, we need a full destination
3096 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3097 cmd == IPVS_CMD_DEL_DEST) {
3098 if (cmd != IPVS_CMD_DEL_DEST)
3101 ret = ip_vs_genl_parse_dest(&udest,
3102 info->attrs[IPVS_CMD_ATTR_DEST],
3109 case IPVS_CMD_NEW_SERVICE:
3111 ret = ip_vs_add_service(&usvc, &svc);
3115 case IPVS_CMD_SET_SERVICE:
3116 ret = ip_vs_edit_service(svc, &usvc);
3118 case IPVS_CMD_DEL_SERVICE:
3119 ret = ip_vs_del_service(svc);
3120 /* do not use svc, it can be freed */
3122 case IPVS_CMD_NEW_DEST:
3123 ret = ip_vs_add_dest(svc, &udest);
3125 case IPVS_CMD_SET_DEST:
3126 ret = ip_vs_edit_dest(svc, &udest);
3128 case IPVS_CMD_DEL_DEST:
3129 ret = ip_vs_del_dest(svc, &udest);
3132 ret = ip_vs_zero_service(svc);
3139 mutex_unlock(&__ip_vs_mutex);
3144 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3146 struct sk_buff *msg;
3148 int ret, cmd, reply_cmd;
3150 cmd = info->genlhdr->cmd;
3152 if (cmd == IPVS_CMD_GET_SERVICE)
3153 reply_cmd = IPVS_CMD_NEW_SERVICE;
3154 else if (cmd == IPVS_CMD_GET_INFO)
3155 reply_cmd = IPVS_CMD_SET_INFO;
3156 else if (cmd == IPVS_CMD_GET_CONFIG)
3157 reply_cmd = IPVS_CMD_SET_CONFIG;
3159 pr_err("unknown Generic Netlink command\n");
3163 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3167 mutex_lock(&__ip_vs_mutex);
3169 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3171 goto nla_put_failure;
3174 case IPVS_CMD_GET_SERVICE:
3176 struct ip_vs_service *svc;
3178 svc = ip_vs_genl_find_service(info->attrs[IPVS_CMD_ATTR_SERVICE]);
3183 ret = ip_vs_genl_fill_service(msg, svc);
3185 goto nla_put_failure;
3194 case IPVS_CMD_GET_CONFIG:
3196 struct ip_vs_timeout_user t;
3198 __ip_vs_get_timeouts(&t);
3199 #ifdef CONFIG_IP_VS_PROTO_TCP
3200 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP, t.tcp_timeout);
3201 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3204 #ifdef CONFIG_IP_VS_PROTO_UDP
3205 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout);
3211 case IPVS_CMD_GET_INFO:
3212 NLA_PUT_U32(msg, IPVS_INFO_ATTR_VERSION, IP_VS_VERSION_CODE);
3213 NLA_PUT_U32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3214 ip_vs_conn_tab_size);
3218 genlmsg_end(msg, reply);
3219 ret = genlmsg_reply(msg, info);
3223 pr_err("not enough space in Netlink message\n");
3229 mutex_unlock(&__ip_vs_mutex);
3235 static struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3237 .cmd = IPVS_CMD_NEW_SERVICE,
3238 .flags = GENL_ADMIN_PERM,
3239 .policy = ip_vs_cmd_policy,
3240 .doit = ip_vs_genl_set_cmd,
3243 .cmd = IPVS_CMD_SET_SERVICE,
3244 .flags = GENL_ADMIN_PERM,
3245 .policy = ip_vs_cmd_policy,
3246 .doit = ip_vs_genl_set_cmd,
3249 .cmd = IPVS_CMD_DEL_SERVICE,
3250 .flags = GENL_ADMIN_PERM,
3251 .policy = ip_vs_cmd_policy,
3252 .doit = ip_vs_genl_set_cmd,
3255 .cmd = IPVS_CMD_GET_SERVICE,
3256 .flags = GENL_ADMIN_PERM,
3257 .doit = ip_vs_genl_get_cmd,
3258 .dumpit = ip_vs_genl_dump_services,
3259 .policy = ip_vs_cmd_policy,
3262 .cmd = IPVS_CMD_NEW_DEST,
3263 .flags = GENL_ADMIN_PERM,
3264 .policy = ip_vs_cmd_policy,
3265 .doit = ip_vs_genl_set_cmd,
3268 .cmd = IPVS_CMD_SET_DEST,
3269 .flags = GENL_ADMIN_PERM,
3270 .policy = ip_vs_cmd_policy,
3271 .doit = ip_vs_genl_set_cmd,
3274 .cmd = IPVS_CMD_DEL_DEST,
3275 .flags = GENL_ADMIN_PERM,
3276 .policy = ip_vs_cmd_policy,
3277 .doit = ip_vs_genl_set_cmd,
3280 .cmd = IPVS_CMD_GET_DEST,
3281 .flags = GENL_ADMIN_PERM,
3282 .policy = ip_vs_cmd_policy,
3283 .dumpit = ip_vs_genl_dump_dests,
3286 .cmd = IPVS_CMD_NEW_DAEMON,
3287 .flags = GENL_ADMIN_PERM,
3288 .policy = ip_vs_cmd_policy,
3289 .doit = ip_vs_genl_set_cmd,
3292 .cmd = IPVS_CMD_DEL_DAEMON,
3293 .flags = GENL_ADMIN_PERM,
3294 .policy = ip_vs_cmd_policy,
3295 .doit = ip_vs_genl_set_cmd,
3298 .cmd = IPVS_CMD_GET_DAEMON,
3299 .flags = GENL_ADMIN_PERM,
3300 .dumpit = ip_vs_genl_dump_daemons,
3303 .cmd = IPVS_CMD_SET_CONFIG,
3304 .flags = GENL_ADMIN_PERM,
3305 .policy = ip_vs_cmd_policy,
3306 .doit = ip_vs_genl_set_cmd,
3309 .cmd = IPVS_CMD_GET_CONFIG,
3310 .flags = GENL_ADMIN_PERM,
3311 .doit = ip_vs_genl_get_cmd,
3314 .cmd = IPVS_CMD_GET_INFO,
3315 .flags = GENL_ADMIN_PERM,
3316 .doit = ip_vs_genl_get_cmd,
3319 .cmd = IPVS_CMD_ZERO,
3320 .flags = GENL_ADMIN_PERM,
3321 .policy = ip_vs_cmd_policy,
3322 .doit = ip_vs_genl_set_cmd,
3325 .cmd = IPVS_CMD_FLUSH,
3326 .flags = GENL_ADMIN_PERM,
3327 .doit = ip_vs_genl_set_cmd,
3331 static int __init ip_vs_genl_register(void)
3333 return genl_register_family_with_ops(&ip_vs_genl_family,
3334 ip_vs_genl_ops, ARRAY_SIZE(ip_vs_genl_ops));
3337 static void ip_vs_genl_unregister(void)
3339 genl_unregister_family(&ip_vs_genl_family);
3342 /* End of Generic Netlink interface definitions */
3345 int __init ip_vs_control_init(void)
3352 ret = nf_register_sockopt(&ip_vs_sockopts);
3354 pr_err("cannot register sockopt.\n");
3358 ret = ip_vs_genl_register();
3360 pr_err("cannot register Generic Netlink interface.\n");
3361 nf_unregister_sockopt(&ip_vs_sockopts);
3365 proc_net_fops_create(&init_net, "ip_vs", 0, &ip_vs_info_fops);
3366 proc_net_fops_create(&init_net, "ip_vs_stats",0, &ip_vs_stats_fops);
3368 sysctl_header = register_sysctl_paths(net_vs_ctl_path, vs_vars);
3370 /* Initialize ip_vs_svc_table, ip_vs_svc_fwm_table, ip_vs_rtable */
3371 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
3372 INIT_LIST_HEAD(&ip_vs_svc_table[idx]);
3373 INIT_LIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3375 for(idx = 0; idx < IP_VS_RTAB_SIZE; idx++) {
3376 INIT_LIST_HEAD(&ip_vs_rtable[idx]);
3379 ip_vs_new_estimator(&ip_vs_stats);
3381 /* Hook the defense timer */
3382 schedule_delayed_work(&defense_work, DEFENSE_TIMER_PERIOD);
3389 void ip_vs_control_cleanup(void)
3392 ip_vs_trash_cleanup();
3393 cancel_rearming_delayed_work(&defense_work);
3394 cancel_work_sync(&defense_work.work);
3395 ip_vs_kill_estimator(&ip_vs_stats);
3396 unregister_sysctl_table(sysctl_header);
3397 proc_net_remove(&init_net, "ip_vs_stats");
3398 proc_net_remove(&init_net, "ip_vs");
3399 ip_vs_genl_unregister();
3400 nf_unregister_sockopt(&ip_vs_sockopts);