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>
41 #include <linux/nsproxy.h>
43 #ifdef CONFIG_IP_VS_IPV6
45 #include <net/ip6_route.h>
47 #include <net/route.h>
49 #include <net/genetlink.h>
51 #include <asm/uaccess.h>
53 #include <net/ip_vs.h>
55 /* semaphore for IPVS sockopts. And, [gs]etsockopt may sleep. */
56 static DEFINE_MUTEX(__ip_vs_mutex);
58 /* lock for service table */
59 static DEFINE_RWLOCK(__ip_vs_svc_lock);
61 /* sysctl variables */
63 #ifdef CONFIG_IP_VS_DEBUG
64 static int sysctl_ip_vs_debug_level = 0;
66 int ip_vs_get_debug_level(void)
68 return sysctl_ip_vs_debug_level;
74 static void __ip_vs_del_service(struct ip_vs_service *svc);
77 #ifdef CONFIG_IP_VS_IPV6
78 /* Taken from rt6_fill_node() in net/ipv6/route.c, is there a better way? */
79 static int __ip_vs_addr_is_local_v6(struct net *net,
80 const struct in6_addr *addr)
87 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
88 if (rt && rt->rt6i_dev && (rt->rt6i_dev->flags & IFF_LOOPBACK))
97 * update_defense_level is called from keventd and from sysctl,
98 * so it needs to protect itself from softirqs
100 static void update_defense_level(struct netns_ipvs *ipvs)
103 static int old_secure_tcp = 0;
108 /* we only count free and buffered memory (in pages) */
110 availmem = i.freeram + i.bufferram;
111 /* however in linux 2.5 the i.bufferram is total page cache size,
113 /* si_swapinfo(&i); */
114 /* availmem = availmem - (i.totalswap - i.freeswap); */
116 nomem = (availmem < ipvs->sysctl_amemthresh);
121 spin_lock(&ipvs->dropentry_lock);
122 switch (ipvs->sysctl_drop_entry) {
124 atomic_set(&ipvs->dropentry, 0);
128 atomic_set(&ipvs->dropentry, 1);
129 ipvs->sysctl_drop_entry = 2;
131 atomic_set(&ipvs->dropentry, 0);
136 atomic_set(&ipvs->dropentry, 1);
138 atomic_set(&ipvs->dropentry, 0);
139 ipvs->sysctl_drop_entry = 1;
143 atomic_set(&ipvs->dropentry, 1);
146 spin_unlock(&ipvs->dropentry_lock);
149 spin_lock(&ipvs->droppacket_lock);
150 switch (ipvs->sysctl_drop_packet) {
156 ipvs->drop_rate = ipvs->drop_counter
157 = ipvs->sysctl_amemthresh /
158 (ipvs->sysctl_amemthresh-availmem);
159 ipvs->sysctl_drop_packet = 2;
166 ipvs->drop_rate = ipvs->drop_counter
167 = ipvs->sysctl_amemthresh /
168 (ipvs->sysctl_amemthresh-availmem);
171 ipvs->sysctl_drop_packet = 1;
175 ipvs->drop_rate = ipvs->sysctl_am_droprate;
178 spin_unlock(&ipvs->droppacket_lock);
181 spin_lock(&ipvs->securetcp_lock);
182 switch (ipvs->sysctl_secure_tcp) {
184 if (old_secure_tcp >= 2)
189 if (old_secure_tcp < 2)
191 ipvs->sysctl_secure_tcp = 2;
193 if (old_secure_tcp >= 2)
199 if (old_secure_tcp < 2)
202 if (old_secure_tcp >= 2)
204 ipvs->sysctl_secure_tcp = 1;
208 if (old_secure_tcp < 2)
212 old_secure_tcp = ipvs->sysctl_secure_tcp;
214 ip_vs_protocol_timeout_change(ipvs,
215 ipvs->sysctl_secure_tcp > 1);
216 spin_unlock(&ipvs->securetcp_lock);
223 * Timer for checking the defense
225 #define DEFENSE_TIMER_PERIOD 1*HZ
227 static void defense_work_handler(struct work_struct *work)
229 struct netns_ipvs *ipvs =
230 container_of(work, struct netns_ipvs, defense_work.work);
232 update_defense_level(ipvs);
233 if (atomic_read(&ipvs->dropentry))
234 ip_vs_random_dropentry(ipvs->net);
235 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
240 ip_vs_use_count_inc(void)
242 return try_module_get(THIS_MODULE);
246 ip_vs_use_count_dec(void)
248 module_put(THIS_MODULE);
253 * Hash table: for virtual service lookups
255 #define IP_VS_SVC_TAB_BITS 8
256 #define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
257 #define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)
259 /* the service table hashed by <protocol, addr, port> */
260 static struct list_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE];
261 /* the service table hashed by fwmark */
262 static struct list_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE];
266 * Returns hash value for virtual service
268 static inline unsigned
269 ip_vs_svc_hashkey(struct net *net, int af, unsigned proto,
270 const union nf_inet_addr *addr, __be16 port)
272 register unsigned porth = ntohs(port);
273 __be32 addr_fold = addr->ip;
275 #ifdef CONFIG_IP_VS_IPV6
277 addr_fold = addr->ip6[0]^addr->ip6[1]^
278 addr->ip6[2]^addr->ip6[3];
280 addr_fold ^= ((size_t)net>>8);
282 return (proto^ntohl(addr_fold)^(porth>>IP_VS_SVC_TAB_BITS)^porth)
283 & IP_VS_SVC_TAB_MASK;
287 * Returns hash value of fwmark for virtual service lookup
289 static inline unsigned ip_vs_svc_fwm_hashkey(struct net *net, __u32 fwmark)
291 return (((size_t)net>>8) ^ fwmark) & IP_VS_SVC_TAB_MASK;
295 * Hashes a service in the ip_vs_svc_table by <netns,proto,addr,port>
296 * or in the ip_vs_svc_fwm_table by fwmark.
297 * Should be called with locked tables.
299 static int ip_vs_svc_hash(struct ip_vs_service *svc)
303 if (svc->flags & IP_VS_SVC_F_HASHED) {
304 pr_err("%s(): request for already hashed, called from %pF\n",
305 __func__, __builtin_return_address(0));
309 if (svc->fwmark == 0) {
311 * Hash it by <netns,protocol,addr,port> in ip_vs_svc_table
313 hash = ip_vs_svc_hashkey(svc->net, svc->af, svc->protocol,
314 &svc->addr, svc->port);
315 list_add(&svc->s_list, &ip_vs_svc_table[hash]);
318 * Hash it by fwmark in svc_fwm_table
320 hash = ip_vs_svc_fwm_hashkey(svc->net, svc->fwmark);
321 list_add(&svc->f_list, &ip_vs_svc_fwm_table[hash]);
324 svc->flags |= IP_VS_SVC_F_HASHED;
325 /* increase its refcnt because it is referenced by the svc table */
326 atomic_inc(&svc->refcnt);
332 * Unhashes a service from svc_table / svc_fwm_table.
333 * Should be called with locked tables.
335 static int ip_vs_svc_unhash(struct ip_vs_service *svc)
337 if (!(svc->flags & IP_VS_SVC_F_HASHED)) {
338 pr_err("%s(): request for unhash flagged, called from %pF\n",
339 __func__, __builtin_return_address(0));
343 if (svc->fwmark == 0) {
344 /* Remove it from the svc_table table */
345 list_del(&svc->s_list);
347 /* Remove it from the svc_fwm_table table */
348 list_del(&svc->f_list);
351 svc->flags &= ~IP_VS_SVC_F_HASHED;
352 atomic_dec(&svc->refcnt);
358 * Get service by {netns, proto,addr,port} in the service table.
360 static inline struct ip_vs_service *
361 __ip_vs_service_find(struct net *net, int af, __u16 protocol,
362 const union nf_inet_addr *vaddr, __be16 vport)
365 struct ip_vs_service *svc;
367 /* Check for "full" addressed entries */
368 hash = ip_vs_svc_hashkey(net, af, protocol, vaddr, vport);
370 list_for_each_entry(svc, &ip_vs_svc_table[hash], s_list){
372 && ip_vs_addr_equal(af, &svc->addr, vaddr)
373 && (svc->port == vport)
374 && (svc->protocol == protocol)
375 && net_eq(svc->net, net)) {
386 * Get service by {fwmark} in the service table.
388 static inline struct ip_vs_service *
389 __ip_vs_svc_fwm_find(struct net *net, int af, __u32 fwmark)
392 struct ip_vs_service *svc;
394 /* Check for fwmark addressed entries */
395 hash = ip_vs_svc_fwm_hashkey(net, fwmark);
397 list_for_each_entry(svc, &ip_vs_svc_fwm_table[hash], f_list) {
398 if (svc->fwmark == fwmark && svc->af == af
399 && net_eq(svc->net, net)) {
408 struct ip_vs_service *
409 ip_vs_service_get(struct net *net, int af, __u32 fwmark, __u16 protocol,
410 const union nf_inet_addr *vaddr, __be16 vport)
412 struct ip_vs_service *svc;
413 struct netns_ipvs *ipvs = net_ipvs(net);
415 read_lock(&__ip_vs_svc_lock);
418 * Check the table hashed by fwmark first
421 svc = __ip_vs_svc_fwm_find(net, af, fwmark);
427 * Check the table hashed by <protocol,addr,port>
428 * for "full" addressed entries
430 svc = __ip_vs_service_find(net, af, protocol, vaddr, vport);
433 && protocol == IPPROTO_TCP
434 && atomic_read(&ipvs->ftpsvc_counter)
435 && (vport == FTPDATA || ntohs(vport) >= PROT_SOCK)) {
437 * Check if ftp service entry exists, the packet
438 * might belong to FTP data connections.
440 svc = __ip_vs_service_find(net, af, protocol, vaddr, FTPPORT);
444 && atomic_read(&ipvs->nullsvc_counter)) {
446 * Check if the catch-all port (port zero) exists
448 svc = __ip_vs_service_find(net, af, protocol, vaddr, 0);
453 atomic_inc(&svc->usecnt);
454 read_unlock(&__ip_vs_svc_lock);
456 IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
457 fwmark, ip_vs_proto_name(protocol),
458 IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
459 svc ? "hit" : "not hit");
466 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
468 atomic_inc(&svc->refcnt);
473 __ip_vs_unbind_svc(struct ip_vs_dest *dest)
475 struct ip_vs_service *svc = dest->svc;
478 if (atomic_dec_and_test(&svc->refcnt)) {
479 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
481 IP_VS_DBG_ADDR(svc->af, &svc->addr),
482 ntohs(svc->port), atomic_read(&svc->usecnt));
483 free_percpu(svc->stats.cpustats);
490 * Returns hash value for real service
492 static inline unsigned ip_vs_rs_hashkey(int af,
493 const union nf_inet_addr *addr,
496 register unsigned porth = ntohs(port);
497 __be32 addr_fold = addr->ip;
499 #ifdef CONFIG_IP_VS_IPV6
501 addr_fold = addr->ip6[0]^addr->ip6[1]^
502 addr->ip6[2]^addr->ip6[3];
505 return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
510 * Hashes ip_vs_dest in rs_table by <proto,addr,port>.
511 * should be called with locked tables.
513 static int ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
517 if (!list_empty(&dest->d_list)) {
522 * Hash by proto,addr,port,
523 * which are the parameters of the real service.
525 hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
527 list_add(&dest->d_list, &ipvs->rs_table[hash]);
533 * UNhashes ip_vs_dest from rs_table.
534 * should be called with locked tables.
536 static int ip_vs_rs_unhash(struct ip_vs_dest *dest)
539 * Remove it from the rs_table table.
541 if (!list_empty(&dest->d_list)) {
542 list_del(&dest->d_list);
543 INIT_LIST_HEAD(&dest->d_list);
550 * Lookup real service by <proto,addr,port> in the real service table.
553 ip_vs_lookup_real_service(struct net *net, int af, __u16 protocol,
554 const union nf_inet_addr *daddr,
557 struct netns_ipvs *ipvs = net_ipvs(net);
559 struct ip_vs_dest *dest;
562 * Check for "full" addressed entries
563 * Return the first found entry
565 hash = ip_vs_rs_hashkey(af, daddr, dport);
567 read_lock(&ipvs->rs_lock);
568 list_for_each_entry(dest, &ipvs->rs_table[hash], d_list) {
570 && ip_vs_addr_equal(af, &dest->addr, daddr)
571 && (dest->port == dport)
572 && ((dest->protocol == protocol) ||
575 read_unlock(&ipvs->rs_lock);
579 read_unlock(&ipvs->rs_lock);
585 * Lookup destination by {addr,port} in the given service
587 static struct ip_vs_dest *
588 ip_vs_lookup_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
591 struct ip_vs_dest *dest;
594 * Find the destination for the given service
596 list_for_each_entry(dest, &svc->destinations, n_list) {
597 if ((dest->af == svc->af)
598 && ip_vs_addr_equal(svc->af, &dest->addr, daddr)
599 && (dest->port == dport)) {
609 * Find destination by {daddr,dport,vaddr,protocol}
610 * Cretaed to be used in ip_vs_process_message() in
611 * the backup synchronization daemon. It finds the
612 * destination to be bound to the received connection
615 * ip_vs_lookup_real_service() looked promissing, but
616 * seems not working as expected.
618 struct ip_vs_dest *ip_vs_find_dest(struct net *net, int af,
619 const union nf_inet_addr *daddr,
621 const union nf_inet_addr *vaddr,
622 __be16 vport, __u16 protocol, __u32 fwmark)
624 struct ip_vs_dest *dest;
625 struct ip_vs_service *svc;
627 svc = ip_vs_service_get(net, af, fwmark, protocol, vaddr, vport);
630 dest = ip_vs_lookup_dest(svc, daddr, dport);
632 atomic_inc(&dest->refcnt);
633 ip_vs_service_put(svc);
638 * Lookup dest by {svc,addr,port} in the destination trash.
639 * The destination trash is used to hold the destinations that are removed
640 * from the service table but are still referenced by some conn entries.
641 * The reason to add the destination trash is when the dest is temporary
642 * down (either by administrator or by monitor program), the dest can be
643 * picked back from the trash, the remaining connections to the dest can
644 * continue, and the counting information of the dest is also useful for
647 static struct ip_vs_dest *
648 ip_vs_trash_get_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
651 struct ip_vs_dest *dest, *nxt;
652 struct netns_ipvs *ipvs = net_ipvs(svc->net);
655 * Find the destination in trash
657 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, n_list) {
658 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
661 IP_VS_DBG_ADDR(svc->af, &dest->addr),
663 atomic_read(&dest->refcnt));
664 if (dest->af == svc->af &&
665 ip_vs_addr_equal(svc->af, &dest->addr, daddr) &&
666 dest->port == dport &&
667 dest->vfwmark == svc->fwmark &&
668 dest->protocol == svc->protocol &&
670 (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
671 dest->vport == svc->port))) {
677 * Try to purge the destination from trash if not referenced
679 if (atomic_read(&dest->refcnt) == 1) {
680 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u "
683 IP_VS_DBG_ADDR(svc->af, &dest->addr),
685 list_del(&dest->n_list);
686 ip_vs_dst_reset(dest);
687 __ip_vs_unbind_svc(dest);
688 free_percpu(dest->stats.cpustats);
698 * Clean up all the destinations in the trash
699 * Called by the ip_vs_control_cleanup()
701 * When the ip_vs_control_clearup is activated by ipvs module exit,
702 * the service tables must have been flushed and all the connections
703 * are expired, and the refcnt of each destination in the trash must
704 * be 1, so we simply release them here.
706 static void ip_vs_trash_cleanup(struct net *net)
708 struct ip_vs_dest *dest, *nxt;
709 struct netns_ipvs *ipvs = net_ipvs(net);
711 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, n_list) {
712 list_del(&dest->n_list);
713 ip_vs_dst_reset(dest);
714 __ip_vs_unbind_svc(dest);
715 free_percpu(dest->stats.cpustats);
721 ip_vs_copy_stats(struct ip_vs_stats_user *dst, struct ip_vs_stats *src)
723 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->ustats.c - src->ustats0.c
725 spin_lock_bh(&src->lock);
727 IP_VS_SHOW_STATS_COUNTER(conns);
728 IP_VS_SHOW_STATS_COUNTER(inpkts);
729 IP_VS_SHOW_STATS_COUNTER(outpkts);
730 IP_VS_SHOW_STATS_COUNTER(inbytes);
731 IP_VS_SHOW_STATS_COUNTER(outbytes);
733 ip_vs_read_estimator(dst, src);
735 spin_unlock_bh(&src->lock);
739 ip_vs_zero_stats(struct ip_vs_stats *stats)
741 spin_lock_bh(&stats->lock);
743 /* get current counters as zero point, rates are zeroed */
745 #define IP_VS_ZERO_STATS_COUNTER(c) stats->ustats0.c = stats->ustats.c
747 IP_VS_ZERO_STATS_COUNTER(conns);
748 IP_VS_ZERO_STATS_COUNTER(inpkts);
749 IP_VS_ZERO_STATS_COUNTER(outpkts);
750 IP_VS_ZERO_STATS_COUNTER(inbytes);
751 IP_VS_ZERO_STATS_COUNTER(outbytes);
753 ip_vs_zero_estimator(stats);
755 spin_unlock_bh(&stats->lock);
759 * Update a destination in the given service
762 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
763 struct ip_vs_dest_user_kern *udest, int add)
765 struct netns_ipvs *ipvs = net_ipvs(svc->net);
768 /* set the weight and the flags */
769 atomic_set(&dest->weight, udest->weight);
770 conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
771 conn_flags |= IP_VS_CONN_F_INACTIVE;
773 /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
774 if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
775 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
778 * Put the real service in rs_table if not present.
779 * For now only for NAT!
781 write_lock_bh(&ipvs->rs_lock);
782 ip_vs_rs_hash(ipvs, dest);
783 write_unlock_bh(&ipvs->rs_lock);
785 atomic_set(&dest->conn_flags, conn_flags);
787 /* bind the service */
789 __ip_vs_bind_svc(dest, svc);
791 if (dest->svc != svc) {
792 __ip_vs_unbind_svc(dest);
793 ip_vs_zero_stats(&dest->stats);
794 __ip_vs_bind_svc(dest, svc);
798 /* set the dest status flags */
799 dest->flags |= IP_VS_DEST_F_AVAILABLE;
801 if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
802 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
803 dest->u_threshold = udest->u_threshold;
804 dest->l_threshold = udest->l_threshold;
806 spin_lock_bh(&dest->dst_lock);
807 ip_vs_dst_reset(dest);
808 spin_unlock_bh(&dest->dst_lock);
811 ip_vs_start_estimator(svc->net, &dest->stats);
813 write_lock_bh(&__ip_vs_svc_lock);
815 /* Wait until all other svc users go away */
816 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
819 list_add(&dest->n_list, &svc->destinations);
823 /* call the update_service, because server weight may be changed */
824 if (svc->scheduler->update_service)
825 svc->scheduler->update_service(svc);
827 write_unlock_bh(&__ip_vs_svc_lock);
832 * Create a destination for the given service
835 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
836 struct ip_vs_dest **dest_p)
838 struct ip_vs_dest *dest;
843 #ifdef CONFIG_IP_VS_IPV6
844 if (svc->af == AF_INET6) {
845 atype = ipv6_addr_type(&udest->addr.in6);
846 if ((!(atype & IPV6_ADDR_UNICAST) ||
847 atype & IPV6_ADDR_LINKLOCAL) &&
848 !__ip_vs_addr_is_local_v6(svc->net, &udest->addr.in6))
853 atype = inet_addr_type(svc->net, udest->addr.ip);
854 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
858 dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
860 pr_err("%s(): no memory.\n", __func__);
863 dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
864 if (!dest->stats.cpustats) {
865 pr_err("%s() alloc_percpu failed\n", __func__);
870 dest->protocol = svc->protocol;
871 dest->vaddr = svc->addr;
872 dest->vport = svc->port;
873 dest->vfwmark = svc->fwmark;
874 ip_vs_addr_copy(svc->af, &dest->addr, &udest->addr);
875 dest->port = udest->port;
877 atomic_set(&dest->activeconns, 0);
878 atomic_set(&dest->inactconns, 0);
879 atomic_set(&dest->persistconns, 0);
880 atomic_set(&dest->refcnt, 1);
882 INIT_LIST_HEAD(&dest->d_list);
883 spin_lock_init(&dest->dst_lock);
884 spin_lock_init(&dest->stats.lock);
885 __ip_vs_update_dest(svc, dest, udest, 1);
899 * Add a destination into an existing service
902 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
904 struct ip_vs_dest *dest;
905 union nf_inet_addr daddr;
906 __be16 dport = udest->port;
911 if (udest->weight < 0) {
912 pr_err("%s(): server weight less than zero\n", __func__);
916 if (udest->l_threshold > udest->u_threshold) {
917 pr_err("%s(): lower threshold is higher than upper threshold\n",
922 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
925 * Check if the dest already exists in the list
927 dest = ip_vs_lookup_dest(svc, &daddr, dport);
930 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
935 * Check if the dest already exists in the trash and
936 * is from the same service
938 dest = ip_vs_trash_get_dest(svc, &daddr, dport);
941 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
942 "dest->refcnt=%d, service %u/%s:%u\n",
943 IP_VS_DBG_ADDR(svc->af, &daddr), ntohs(dport),
944 atomic_read(&dest->refcnt),
946 IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
950 * Get the destination from the trash
952 list_del(&dest->n_list);
954 __ip_vs_update_dest(svc, dest, udest, 1);
958 * Allocate and initialize the dest structure
960 ret = ip_vs_new_dest(svc, udest, &dest);
969 * Edit a destination in the given service
972 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
974 struct ip_vs_dest *dest;
975 union nf_inet_addr daddr;
976 __be16 dport = udest->port;
980 if (udest->weight < 0) {
981 pr_err("%s(): server weight less than zero\n", __func__);
985 if (udest->l_threshold > udest->u_threshold) {
986 pr_err("%s(): lower threshold is higher than upper threshold\n",
991 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
994 * Lookup the destination list
996 dest = ip_vs_lookup_dest(svc, &daddr, dport);
999 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1003 __ip_vs_update_dest(svc, dest, udest, 0);
1011 * Delete a destination (must be already unlinked from the service)
1013 static void __ip_vs_del_dest(struct net *net, struct ip_vs_dest *dest)
1015 struct netns_ipvs *ipvs = net_ipvs(net);
1017 ip_vs_stop_estimator(net, &dest->stats);
1020 * Remove it from the d-linked list with the real services.
1022 write_lock_bh(&ipvs->rs_lock);
1023 ip_vs_rs_unhash(dest);
1024 write_unlock_bh(&ipvs->rs_lock);
1027 * Decrease the refcnt of the dest, and free the dest
1028 * if nobody refers to it (refcnt=0). Otherwise, throw
1029 * the destination into the trash.
1031 if (atomic_dec_and_test(&dest->refcnt)) {
1032 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u\n",
1034 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1036 ip_vs_dst_reset(dest);
1037 /* simply decrease svc->refcnt here, let the caller check
1038 and release the service if nobody refers to it.
1039 Only user context can release destination and service,
1040 and only one user context can update virtual service at a
1041 time, so the operation here is OK */
1042 atomic_dec(&dest->svc->refcnt);
1043 free_percpu(dest->stats.cpustats);
1046 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, "
1047 "dest->refcnt=%d\n",
1048 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1050 atomic_read(&dest->refcnt));
1051 list_add(&dest->n_list, &ipvs->dest_trash);
1052 atomic_inc(&dest->refcnt);
1058 * Unlink a destination from the given service
1060 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1061 struct ip_vs_dest *dest,
1064 dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1067 * Remove it from the d-linked destination list.
1069 list_del(&dest->n_list);
1073 * Call the update_service function of its scheduler
1075 if (svcupd && svc->scheduler->update_service)
1076 svc->scheduler->update_service(svc);
1081 * Delete a destination server in the given service
1084 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1086 struct ip_vs_dest *dest;
1087 __be16 dport = udest->port;
1091 dest = ip_vs_lookup_dest(svc, &udest->addr, dport);
1094 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1098 write_lock_bh(&__ip_vs_svc_lock);
1101 * Wait until all other svc users go away.
1103 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1106 * Unlink dest from the service
1108 __ip_vs_unlink_dest(svc, dest, 1);
1110 write_unlock_bh(&__ip_vs_svc_lock);
1113 * Delete the destination
1115 __ip_vs_del_dest(svc->net, dest);
1124 * Add a service into the service hash table
1127 ip_vs_add_service(struct net *net, struct ip_vs_service_user_kern *u,
1128 struct ip_vs_service **svc_p)
1131 struct ip_vs_scheduler *sched = NULL;
1132 struct ip_vs_pe *pe = NULL;
1133 struct ip_vs_service *svc = NULL;
1134 struct netns_ipvs *ipvs = net_ipvs(net);
1136 /* increase the module use count */
1137 ip_vs_use_count_inc();
1139 /* Lookup the scheduler by 'u->sched_name' */
1140 sched = ip_vs_scheduler_get(u->sched_name);
1141 if (sched == NULL) {
1142 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1147 if (u->pe_name && *u->pe_name) {
1148 pe = ip_vs_pe_getbyname(u->pe_name);
1150 pr_info("persistence engine module ip_vs_pe_%s "
1151 "not found\n", u->pe_name);
1157 #ifdef CONFIG_IP_VS_IPV6
1158 if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1164 svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1166 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1170 svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1171 if (!svc->stats.cpustats) {
1172 pr_err("%s() alloc_percpu failed\n", __func__);
1176 /* I'm the first user of the service */
1177 atomic_set(&svc->usecnt, 0);
1178 atomic_set(&svc->refcnt, 0);
1181 svc->protocol = u->protocol;
1182 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1183 svc->port = u->port;
1184 svc->fwmark = u->fwmark;
1185 svc->flags = u->flags;
1186 svc->timeout = u->timeout * HZ;
1187 svc->netmask = u->netmask;
1190 INIT_LIST_HEAD(&svc->destinations);
1191 rwlock_init(&svc->sched_lock);
1192 spin_lock_init(&svc->stats.lock);
1194 /* Bind the scheduler */
1195 ret = ip_vs_bind_scheduler(svc, sched);
1200 /* Bind the ct retriever */
1201 ip_vs_bind_pe(svc, pe);
1204 /* Update the virtual service counters */
1205 if (svc->port == FTPPORT)
1206 atomic_inc(&ipvs->ftpsvc_counter);
1207 else if (svc->port == 0)
1208 atomic_inc(&ipvs->nullsvc_counter);
1210 ip_vs_start_estimator(net, &svc->stats);
1212 /* Count only IPv4 services for old get/setsockopt interface */
1213 if (svc->af == AF_INET)
1214 ipvs->num_services++;
1216 /* Hash the service into the service table */
1217 write_lock_bh(&__ip_vs_svc_lock);
1218 ip_vs_svc_hash(svc);
1219 write_unlock_bh(&__ip_vs_svc_lock);
1222 /* Now there is a service - full throttle */
1229 ip_vs_unbind_scheduler(svc);
1232 ip_vs_app_inc_put(svc->inc);
1235 if (svc->stats.cpustats)
1236 free_percpu(svc->stats.cpustats);
1239 ip_vs_scheduler_put(sched);
1242 /* decrease the module use count */
1243 ip_vs_use_count_dec();
1250 * Edit a service and bind it with a new scheduler
1253 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1255 struct ip_vs_scheduler *sched, *old_sched;
1256 struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1260 * Lookup the scheduler, by 'u->sched_name'
1262 sched = ip_vs_scheduler_get(u->sched_name);
1263 if (sched == NULL) {
1264 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1269 if (u->pe_name && *u->pe_name) {
1270 pe = ip_vs_pe_getbyname(u->pe_name);
1272 pr_info("persistence engine module ip_vs_pe_%s "
1273 "not found\n", u->pe_name);
1280 #ifdef CONFIG_IP_VS_IPV6
1281 if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1287 write_lock_bh(&__ip_vs_svc_lock);
1290 * Wait until all other svc users go away.
1292 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1295 * Set the flags and timeout value
1297 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1298 svc->timeout = u->timeout * HZ;
1299 svc->netmask = u->netmask;
1301 old_sched = svc->scheduler;
1302 if (sched != old_sched) {
1304 * Unbind the old scheduler
1306 if ((ret = ip_vs_unbind_scheduler(svc))) {
1312 * Bind the new scheduler
1314 if ((ret = ip_vs_bind_scheduler(svc, sched))) {
1316 * If ip_vs_bind_scheduler fails, restore the old
1318 * The main reason of failure is out of memory.
1320 * The question is if the old scheduler can be
1321 * restored all the time. TODO: if it cannot be
1322 * restored some time, we must delete the service,
1323 * otherwise the system may crash.
1325 ip_vs_bind_scheduler(svc, old_sched);
1333 ip_vs_unbind_pe(svc);
1334 ip_vs_bind_pe(svc, pe);
1338 write_unlock_bh(&__ip_vs_svc_lock);
1340 ip_vs_scheduler_put(old_sched);
1341 ip_vs_pe_put(old_pe);
1347 * Delete a service from the service list
1348 * - The service must be unlinked, unlocked and not referenced!
1349 * - We are called under _bh lock
1351 static void __ip_vs_del_service(struct ip_vs_service *svc)
1353 struct ip_vs_dest *dest, *nxt;
1354 struct ip_vs_scheduler *old_sched;
1355 struct ip_vs_pe *old_pe;
1356 struct netns_ipvs *ipvs = net_ipvs(svc->net);
1358 pr_info("%s: enter\n", __func__);
1360 /* Count only IPv4 services for old get/setsockopt interface */
1361 if (svc->af == AF_INET)
1362 ipvs->num_services--;
1364 ip_vs_stop_estimator(svc->net, &svc->stats);
1366 /* Unbind scheduler */
1367 old_sched = svc->scheduler;
1368 ip_vs_unbind_scheduler(svc);
1369 ip_vs_scheduler_put(old_sched);
1371 /* Unbind persistence engine */
1373 ip_vs_unbind_pe(svc);
1374 ip_vs_pe_put(old_pe);
1376 /* Unbind app inc */
1378 ip_vs_app_inc_put(svc->inc);
1383 * Unlink the whole destination list
1385 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1386 __ip_vs_unlink_dest(svc, dest, 0);
1387 __ip_vs_del_dest(svc->net, dest);
1391 * Update the virtual service counters
1393 if (svc->port == FTPPORT)
1394 atomic_dec(&ipvs->ftpsvc_counter);
1395 else if (svc->port == 0)
1396 atomic_dec(&ipvs->nullsvc_counter);
1399 * Free the service if nobody refers to it
1401 if (atomic_read(&svc->refcnt) == 0) {
1402 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
1404 IP_VS_DBG_ADDR(svc->af, &svc->addr),
1405 ntohs(svc->port), atomic_read(&svc->usecnt));
1406 free_percpu(svc->stats.cpustats);
1410 /* decrease the module use count */
1411 ip_vs_use_count_dec();
1415 * Unlink a service from list and try to delete it if its refcnt reached 0
1417 static void ip_vs_unlink_service(struct ip_vs_service *svc)
1420 * Unhash it from the service table
1422 write_lock_bh(&__ip_vs_svc_lock);
1424 ip_vs_svc_unhash(svc);
1427 * Wait until all the svc users go away.
1429 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1431 __ip_vs_del_service(svc);
1433 write_unlock_bh(&__ip_vs_svc_lock);
1437 * Delete a service from the service list
1439 static int ip_vs_del_service(struct ip_vs_service *svc)
1443 ip_vs_unlink_service(svc);
1450 * Flush all the virtual services
1452 static int ip_vs_flush(struct net *net)
1455 struct ip_vs_service *svc, *nxt;
1458 * Flush the service table hashed by <netns,protocol,addr,port>
1460 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1461 list_for_each_entry_safe(svc, nxt, &ip_vs_svc_table[idx],
1463 if (net_eq(svc->net, net))
1464 ip_vs_unlink_service(svc);
1469 * Flush the service table hashed by fwmark
1471 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1472 list_for_each_entry_safe(svc, nxt,
1473 &ip_vs_svc_fwm_table[idx], f_list) {
1474 if (net_eq(svc->net, net))
1475 ip_vs_unlink_service(svc);
1483 * Delete service by {netns} in the service table.
1484 * Called by __ip_vs_cleanup()
1486 void __ip_vs_service_cleanup(struct net *net)
1489 /* Check for "full" addressed entries */
1490 mutex_lock(&__ip_vs_mutex);
1492 mutex_unlock(&__ip_vs_mutex);
1496 * Release dst hold by dst_cache
1499 __ip_vs_dev_reset(struct ip_vs_dest *dest, struct net_device *dev)
1501 spin_lock_bh(&dest->dst_lock);
1502 if (dest->dst_cache && dest->dst_cache->dev == dev) {
1503 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1505 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1507 atomic_read(&dest->refcnt));
1508 ip_vs_dst_reset(dest);
1510 spin_unlock_bh(&dest->dst_lock);
1514 * Netdev event receiver
1515 * Currently only NETDEV_UNREGISTER is handled, i.e. if we hold a reference to
1516 * a device that is "unregister" it must be released.
1518 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1521 struct net_device *dev = ptr;
1522 struct net *net = dev_net(dev);
1523 struct ip_vs_service *svc;
1524 struct ip_vs_dest *dest;
1527 if (event != NETDEV_UNREGISTER)
1529 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1531 mutex_lock(&__ip_vs_mutex);
1532 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1533 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1534 if (net_eq(svc->net, net)) {
1535 list_for_each_entry(dest, &svc->destinations,
1537 __ip_vs_dev_reset(dest, dev);
1542 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1543 if (net_eq(svc->net, net)) {
1544 list_for_each_entry(dest, &svc->destinations,
1546 __ip_vs_dev_reset(dest, dev);
1553 list_for_each_entry(dest, &net_ipvs(net)->dest_trash, n_list) {
1554 __ip_vs_dev_reset(dest, dev);
1556 mutex_unlock(&__ip_vs_mutex);
1562 * Zero counters in a service or all services
1564 static int ip_vs_zero_service(struct ip_vs_service *svc)
1566 struct ip_vs_dest *dest;
1568 write_lock_bh(&__ip_vs_svc_lock);
1569 list_for_each_entry(dest, &svc->destinations, n_list) {
1570 ip_vs_zero_stats(&dest->stats);
1572 ip_vs_zero_stats(&svc->stats);
1573 write_unlock_bh(&__ip_vs_svc_lock);
1577 static int ip_vs_zero_all(struct net *net)
1580 struct ip_vs_service *svc;
1582 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1583 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1584 if (net_eq(svc->net, net))
1585 ip_vs_zero_service(svc);
1589 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1590 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1591 if (net_eq(svc->net, net))
1592 ip_vs_zero_service(svc);
1596 ip_vs_zero_stats(&net_ipvs(net)->tot_stats);
1600 #ifdef CONFIG_SYSCTL
1602 proc_do_defense_mode(ctl_table *table, int write,
1603 void __user *buffer, size_t *lenp, loff_t *ppos)
1605 struct net *net = current->nsproxy->net_ns;
1606 int *valp = table->data;
1610 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1611 if (write && (*valp != val)) {
1612 if ((*valp < 0) || (*valp > 3)) {
1613 /* Restore the correct value */
1616 update_defense_level(net_ipvs(net));
1623 proc_do_sync_threshold(ctl_table *table, int write,
1624 void __user *buffer, size_t *lenp, loff_t *ppos)
1626 int *valp = table->data;
1630 /* backup the value first */
1631 memcpy(val, valp, sizeof(val));
1633 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1634 if (write && (valp[0] < 0 || valp[1] < 0 || valp[0] >= valp[1])) {
1635 /* Restore the correct value */
1636 memcpy(valp, val, sizeof(val));
1642 proc_do_sync_mode(ctl_table *table, int write,
1643 void __user *buffer, size_t *lenp, loff_t *ppos)
1645 int *valp = table->data;
1649 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1650 if (write && (*valp != val)) {
1651 if ((*valp < 0) || (*valp > 1)) {
1652 /* Restore the correct value */
1655 struct net *net = current->nsproxy->net_ns;
1656 ip_vs_sync_switch_mode(net, val);
1663 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1664 * Do not change order or insert new entries without
1665 * align with netns init in __ip_vs_control_init()
1668 static struct ctl_table vs_vars[] = {
1670 .procname = "amemthresh",
1671 .maxlen = sizeof(int),
1673 .proc_handler = proc_dointvec,
1676 .procname = "am_droprate",
1677 .maxlen = sizeof(int),
1679 .proc_handler = proc_dointvec,
1682 .procname = "drop_entry",
1683 .maxlen = sizeof(int),
1685 .proc_handler = proc_do_defense_mode,
1688 .procname = "drop_packet",
1689 .maxlen = sizeof(int),
1691 .proc_handler = proc_do_defense_mode,
1693 #ifdef CONFIG_IP_VS_NFCT
1695 .procname = "conntrack",
1696 .maxlen = sizeof(int),
1698 .proc_handler = &proc_dointvec,
1702 .procname = "secure_tcp",
1703 .maxlen = sizeof(int),
1705 .proc_handler = proc_do_defense_mode,
1708 .procname = "snat_reroute",
1709 .maxlen = sizeof(int),
1711 .proc_handler = &proc_dointvec,
1714 .procname = "sync_version",
1715 .maxlen = sizeof(int),
1717 .proc_handler = &proc_do_sync_mode,
1720 .procname = "cache_bypass",
1721 .maxlen = sizeof(int),
1723 .proc_handler = proc_dointvec,
1726 .procname = "expire_nodest_conn",
1727 .maxlen = sizeof(int),
1729 .proc_handler = proc_dointvec,
1732 .procname = "expire_quiescent_template",
1733 .maxlen = sizeof(int),
1735 .proc_handler = proc_dointvec,
1738 .procname = "sync_threshold",
1740 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1742 .proc_handler = proc_do_sync_threshold,
1745 .procname = "nat_icmp_send",
1746 .maxlen = sizeof(int),
1748 .proc_handler = proc_dointvec,
1750 #ifdef CONFIG_IP_VS_DEBUG
1752 .procname = "debug_level",
1753 .data = &sysctl_ip_vs_debug_level,
1754 .maxlen = sizeof(int),
1756 .proc_handler = proc_dointvec,
1761 .procname = "timeout_established",
1762 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1763 .maxlen = sizeof(int),
1765 .proc_handler = proc_dointvec_jiffies,
1768 .procname = "timeout_synsent",
1769 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1770 .maxlen = sizeof(int),
1772 .proc_handler = proc_dointvec_jiffies,
1775 .procname = "timeout_synrecv",
1776 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1777 .maxlen = sizeof(int),
1779 .proc_handler = proc_dointvec_jiffies,
1782 .procname = "timeout_finwait",
1783 .data = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1784 .maxlen = sizeof(int),
1786 .proc_handler = proc_dointvec_jiffies,
1789 .procname = "timeout_timewait",
1790 .data = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1791 .maxlen = sizeof(int),
1793 .proc_handler = proc_dointvec_jiffies,
1796 .procname = "timeout_close",
1797 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1798 .maxlen = sizeof(int),
1800 .proc_handler = proc_dointvec_jiffies,
1803 .procname = "timeout_closewait",
1804 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1805 .maxlen = sizeof(int),
1807 .proc_handler = proc_dointvec_jiffies,
1810 .procname = "timeout_lastack",
1811 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1812 .maxlen = sizeof(int),
1814 .proc_handler = proc_dointvec_jiffies,
1817 .procname = "timeout_listen",
1818 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1819 .maxlen = sizeof(int),
1821 .proc_handler = proc_dointvec_jiffies,
1824 .procname = "timeout_synack",
1825 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1826 .maxlen = sizeof(int),
1828 .proc_handler = proc_dointvec_jiffies,
1831 .procname = "timeout_udp",
1832 .data = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1833 .maxlen = sizeof(int),
1835 .proc_handler = proc_dointvec_jiffies,
1838 .procname = "timeout_icmp",
1839 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1840 .maxlen = sizeof(int),
1842 .proc_handler = proc_dointvec_jiffies,
1848 const struct ctl_path net_vs_ctl_path[] = {
1849 { .procname = "net", },
1850 { .procname = "ipv4", },
1851 { .procname = "vs", },
1854 EXPORT_SYMBOL_GPL(net_vs_ctl_path);
1857 #ifdef CONFIG_PROC_FS
1860 struct seq_net_private p; /* Do not move this, netns depends upon it*/
1861 struct list_head *table;
1866 * Write the contents of the VS rule table to a PROCfs file.
1867 * (It is kept just for backward compatibility)
1869 static inline const char *ip_vs_fwd_name(unsigned flags)
1871 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1872 case IP_VS_CONN_F_LOCALNODE:
1874 case IP_VS_CONN_F_TUNNEL:
1876 case IP_VS_CONN_F_DROUTE:
1884 /* Get the Nth entry in the two lists */
1885 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1887 struct net *net = seq_file_net(seq);
1888 struct ip_vs_iter *iter = seq->private;
1890 struct ip_vs_service *svc;
1892 /* look in hash by protocol */
1893 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1894 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1895 if (net_eq(svc->net, net) && pos-- == 0) {
1896 iter->table = ip_vs_svc_table;
1903 /* keep looking in fwmark */
1904 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1905 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1906 if (net_eq(svc->net, net) && pos-- == 0) {
1907 iter->table = ip_vs_svc_fwm_table;
1917 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1918 __acquires(__ip_vs_svc_lock)
1921 read_lock_bh(&__ip_vs_svc_lock);
1922 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1926 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1928 struct list_head *e;
1929 struct ip_vs_iter *iter;
1930 struct ip_vs_service *svc;
1933 if (v == SEQ_START_TOKEN)
1934 return ip_vs_info_array(seq,0);
1937 iter = seq->private;
1939 if (iter->table == ip_vs_svc_table) {
1940 /* next service in table hashed by protocol */
1941 if ((e = svc->s_list.next) != &ip_vs_svc_table[iter->bucket])
1942 return list_entry(e, struct ip_vs_service, s_list);
1945 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1946 list_for_each_entry(svc,&ip_vs_svc_table[iter->bucket],
1952 iter->table = ip_vs_svc_fwm_table;
1957 /* next service in hashed by fwmark */
1958 if ((e = svc->f_list.next) != &ip_vs_svc_fwm_table[iter->bucket])
1959 return list_entry(e, struct ip_vs_service, f_list);
1962 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1963 list_for_each_entry(svc, &ip_vs_svc_fwm_table[iter->bucket],
1971 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
1972 __releases(__ip_vs_svc_lock)
1974 read_unlock_bh(&__ip_vs_svc_lock);
1978 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
1980 if (v == SEQ_START_TOKEN) {
1982 "IP Virtual Server version %d.%d.%d (size=%d)\n",
1983 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
1985 "Prot LocalAddress:Port Scheduler Flags\n");
1987 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
1989 const struct ip_vs_service *svc = v;
1990 const struct ip_vs_iter *iter = seq->private;
1991 const struct ip_vs_dest *dest;
1993 if (iter->table == ip_vs_svc_table) {
1994 #ifdef CONFIG_IP_VS_IPV6
1995 if (svc->af == AF_INET6)
1996 seq_printf(seq, "%s [%pI6]:%04X %s ",
1997 ip_vs_proto_name(svc->protocol),
2000 svc->scheduler->name);
2003 seq_printf(seq, "%s %08X:%04X %s %s ",
2004 ip_vs_proto_name(svc->protocol),
2005 ntohl(svc->addr.ip),
2007 svc->scheduler->name,
2008 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2010 seq_printf(seq, "FWM %08X %s %s",
2011 svc->fwmark, svc->scheduler->name,
2012 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2015 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2016 seq_printf(seq, "persistent %d %08X\n",
2018 ntohl(svc->netmask));
2020 seq_putc(seq, '\n');
2022 list_for_each_entry(dest, &svc->destinations, n_list) {
2023 #ifdef CONFIG_IP_VS_IPV6
2024 if (dest->af == AF_INET6)
2027 " %-7s %-6d %-10d %-10d\n",
2030 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2031 atomic_read(&dest->weight),
2032 atomic_read(&dest->activeconns),
2033 atomic_read(&dest->inactconns));
2038 "%-7s %-6d %-10d %-10d\n",
2039 ntohl(dest->addr.ip),
2041 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2042 atomic_read(&dest->weight),
2043 atomic_read(&dest->activeconns),
2044 atomic_read(&dest->inactconns));
2051 static const struct seq_operations ip_vs_info_seq_ops = {
2052 .start = ip_vs_info_seq_start,
2053 .next = ip_vs_info_seq_next,
2054 .stop = ip_vs_info_seq_stop,
2055 .show = ip_vs_info_seq_show,
2058 static int ip_vs_info_open(struct inode *inode, struct file *file)
2060 return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2061 sizeof(struct ip_vs_iter));
2064 static const struct file_operations ip_vs_info_fops = {
2065 .owner = THIS_MODULE,
2066 .open = ip_vs_info_open,
2068 .llseek = seq_lseek,
2069 .release = seq_release_private,
2074 #ifdef CONFIG_PROC_FS
2075 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2077 struct net *net = seq_file_single_net(seq);
2078 struct ip_vs_stats_user show;
2080 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2082 " Total Incoming Outgoing Incoming Outgoing\n");
2084 " Conns Packets Packets Bytes Bytes\n");
2086 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2087 seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", show.conns,
2088 show.inpkts, show.outpkts,
2089 (unsigned long long) show.inbytes,
2090 (unsigned long long) show.outbytes);
2092 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2094 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2095 seq_printf(seq, "%8X %8X %8X %16X %16X\n",
2096 show.cps, show.inpps, show.outpps,
2097 show.inbps, show.outbps);
2102 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2104 return single_open_net(inode, file, ip_vs_stats_show);
2107 static const struct file_operations ip_vs_stats_fops = {
2108 .owner = THIS_MODULE,
2109 .open = ip_vs_stats_seq_open,
2111 .llseek = seq_lseek,
2112 .release = single_release,
2115 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2117 struct net *net = seq_file_single_net(seq);
2118 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2119 struct ip_vs_cpu_stats *cpustats = tot_stats->cpustats;
2120 struct ip_vs_stats_user rates;
2123 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2125 " Total Incoming Outgoing Incoming Outgoing\n");
2127 "CPU Conns Packets Packets Bytes Bytes\n");
2129 for_each_possible_cpu(i) {
2130 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2132 __u64 inbytes, outbytes;
2135 start = u64_stats_fetch_begin_bh(&u->syncp);
2136 inbytes = u->ustats.inbytes;
2137 outbytes = u->ustats.outbytes;
2138 } while (u64_stats_fetch_retry_bh(&u->syncp, start));
2140 seq_printf(seq, "%3X %8X %8X %8X %16LX %16LX\n",
2141 i, u->ustats.conns, u->ustats.inpkts,
2142 u->ustats.outpkts, (__u64)inbytes,
2146 spin_lock_bh(&tot_stats->lock);
2148 seq_printf(seq, " ~ %8X %8X %8X %16LX %16LX\n\n",
2149 tot_stats->ustats.conns, tot_stats->ustats.inpkts,
2150 tot_stats->ustats.outpkts,
2151 (unsigned long long) tot_stats->ustats.inbytes,
2152 (unsigned long long) tot_stats->ustats.outbytes);
2154 ip_vs_read_estimator(&rates, tot_stats);
2156 spin_unlock_bh(&tot_stats->lock);
2158 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2160 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2161 seq_printf(seq, " %8X %8X %8X %16X %16X\n",
2171 static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2173 return single_open_net(inode, file, ip_vs_stats_percpu_show);
2176 static const struct file_operations ip_vs_stats_percpu_fops = {
2177 .owner = THIS_MODULE,
2178 .open = ip_vs_stats_percpu_seq_open,
2180 .llseek = seq_lseek,
2181 .release = single_release,
2186 * Set timeout values for tcp tcpfin udp in the timeout_table.
2188 static int ip_vs_set_timeout(struct net *net, struct ip_vs_timeout_user *u)
2190 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2191 struct ip_vs_proto_data *pd;
2194 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2199 #ifdef CONFIG_IP_VS_PROTO_TCP
2200 if (u->tcp_timeout) {
2201 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2202 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2203 = u->tcp_timeout * HZ;
2206 if (u->tcp_fin_timeout) {
2207 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2208 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2209 = u->tcp_fin_timeout * HZ;
2213 #ifdef CONFIG_IP_VS_PROTO_UDP
2214 if (u->udp_timeout) {
2215 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2216 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2217 = u->udp_timeout * HZ;
2224 #define SET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2225 #define SERVICE_ARG_LEN (sizeof(struct ip_vs_service_user))
2226 #define SVCDEST_ARG_LEN (sizeof(struct ip_vs_service_user) + \
2227 sizeof(struct ip_vs_dest_user))
2228 #define TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2229 #define DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user))
2230 #define MAX_ARG_LEN SVCDEST_ARG_LEN
2232 static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
2233 [SET_CMDID(IP_VS_SO_SET_ADD)] = SERVICE_ARG_LEN,
2234 [SET_CMDID(IP_VS_SO_SET_EDIT)] = SERVICE_ARG_LEN,
2235 [SET_CMDID(IP_VS_SO_SET_DEL)] = SERVICE_ARG_LEN,
2236 [SET_CMDID(IP_VS_SO_SET_FLUSH)] = 0,
2237 [SET_CMDID(IP_VS_SO_SET_ADDDEST)] = SVCDEST_ARG_LEN,
2238 [SET_CMDID(IP_VS_SO_SET_DELDEST)] = SVCDEST_ARG_LEN,
2239 [SET_CMDID(IP_VS_SO_SET_EDITDEST)] = SVCDEST_ARG_LEN,
2240 [SET_CMDID(IP_VS_SO_SET_TIMEOUT)] = TIMEOUT_ARG_LEN,
2241 [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)] = DAEMON_ARG_LEN,
2242 [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)] = DAEMON_ARG_LEN,
2243 [SET_CMDID(IP_VS_SO_SET_ZERO)] = SERVICE_ARG_LEN,
2246 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2247 struct ip_vs_service_user *usvc_compat)
2249 memset(usvc, 0, sizeof(*usvc));
2252 usvc->protocol = usvc_compat->protocol;
2253 usvc->addr.ip = usvc_compat->addr;
2254 usvc->port = usvc_compat->port;
2255 usvc->fwmark = usvc_compat->fwmark;
2257 /* Deep copy of sched_name is not needed here */
2258 usvc->sched_name = usvc_compat->sched_name;
2260 usvc->flags = usvc_compat->flags;
2261 usvc->timeout = usvc_compat->timeout;
2262 usvc->netmask = usvc_compat->netmask;
2265 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2266 struct ip_vs_dest_user *udest_compat)
2268 memset(udest, 0, sizeof(*udest));
2270 udest->addr.ip = udest_compat->addr;
2271 udest->port = udest_compat->port;
2272 udest->conn_flags = udest_compat->conn_flags;
2273 udest->weight = udest_compat->weight;
2274 udest->u_threshold = udest_compat->u_threshold;
2275 udest->l_threshold = udest_compat->l_threshold;
2279 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2281 struct net *net = sock_net(sk);
2283 unsigned char arg[MAX_ARG_LEN];
2284 struct ip_vs_service_user *usvc_compat;
2285 struct ip_vs_service_user_kern usvc;
2286 struct ip_vs_service *svc;
2287 struct ip_vs_dest_user *udest_compat;
2288 struct ip_vs_dest_user_kern udest;
2290 if (!capable(CAP_NET_ADMIN))
2293 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2295 if (len < 0 || len > MAX_ARG_LEN)
2297 if (len != set_arglen[SET_CMDID(cmd)]) {
2298 pr_err("set_ctl: len %u != %u\n",
2299 len, set_arglen[SET_CMDID(cmd)]);
2303 if (copy_from_user(arg, user, len) != 0)
2306 /* increase the module use count */
2307 ip_vs_use_count_inc();
2309 if (mutex_lock_interruptible(&__ip_vs_mutex)) {
2314 if (cmd == IP_VS_SO_SET_FLUSH) {
2315 /* Flush the virtual service */
2316 ret = ip_vs_flush(net);
2318 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2319 /* Set timeout values for (tcp tcpfin udp) */
2320 ret = ip_vs_set_timeout(net, (struct ip_vs_timeout_user *)arg);
2322 } else if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2323 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2324 ret = start_sync_thread(net, dm->state, dm->mcast_ifn,
2327 } else if (cmd == IP_VS_SO_SET_STOPDAEMON) {
2328 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2329 ret = stop_sync_thread(net, dm->state);
2333 usvc_compat = (struct ip_vs_service_user *)arg;
2334 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2336 /* We only use the new structs internally, so copy userspace compat
2337 * structs to extended internal versions */
2338 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2339 ip_vs_copy_udest_compat(&udest, udest_compat);
2341 if (cmd == IP_VS_SO_SET_ZERO) {
2342 /* if no service address is set, zero counters in all */
2343 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2344 ret = ip_vs_zero_all(net);
2349 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2350 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2351 usvc.protocol != IPPROTO_SCTP) {
2352 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2353 usvc.protocol, &usvc.addr.ip,
2354 ntohs(usvc.port), usvc.sched_name);
2359 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2360 if (usvc.fwmark == 0)
2361 svc = __ip_vs_service_find(net, usvc.af, usvc.protocol,
2362 &usvc.addr, usvc.port);
2364 svc = __ip_vs_svc_fwm_find(net, usvc.af, usvc.fwmark);
2366 if (cmd != IP_VS_SO_SET_ADD
2367 && (svc == NULL || svc->protocol != usvc.protocol)) {
2373 case IP_VS_SO_SET_ADD:
2377 ret = ip_vs_add_service(net, &usvc, &svc);
2379 case IP_VS_SO_SET_EDIT:
2380 ret = ip_vs_edit_service(svc, &usvc);
2382 case IP_VS_SO_SET_DEL:
2383 ret = ip_vs_del_service(svc);
2387 case IP_VS_SO_SET_ZERO:
2388 ret = ip_vs_zero_service(svc);
2390 case IP_VS_SO_SET_ADDDEST:
2391 ret = ip_vs_add_dest(svc, &udest);
2393 case IP_VS_SO_SET_EDITDEST:
2394 ret = ip_vs_edit_dest(svc, &udest);
2396 case IP_VS_SO_SET_DELDEST:
2397 ret = ip_vs_del_dest(svc, &udest);
2404 mutex_unlock(&__ip_vs_mutex);
2406 /* decrease the module use count */
2407 ip_vs_use_count_dec();
2414 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2416 dst->protocol = src->protocol;
2417 dst->addr = src->addr.ip;
2418 dst->port = src->port;
2419 dst->fwmark = src->fwmark;
2420 strlcpy(dst->sched_name, src->scheduler->name, sizeof(dst->sched_name));
2421 dst->flags = src->flags;
2422 dst->timeout = src->timeout / HZ;
2423 dst->netmask = src->netmask;
2424 dst->num_dests = src->num_dests;
2425 ip_vs_copy_stats(&dst->stats, &src->stats);
2429 __ip_vs_get_service_entries(struct net *net,
2430 const struct ip_vs_get_services *get,
2431 struct ip_vs_get_services __user *uptr)
2434 struct ip_vs_service *svc;
2435 struct ip_vs_service_entry entry;
2438 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2439 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2440 /* Only expose IPv4 entries to old interface */
2441 if (svc->af != AF_INET || !net_eq(svc->net, net))
2444 if (count >= get->num_services)
2446 memset(&entry, 0, sizeof(entry));
2447 ip_vs_copy_service(&entry, svc);
2448 if (copy_to_user(&uptr->entrytable[count],
2449 &entry, sizeof(entry))) {
2457 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2458 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2459 /* Only expose IPv4 entries to old interface */
2460 if (svc->af != AF_INET || !net_eq(svc->net, net))
2463 if (count >= get->num_services)
2465 memset(&entry, 0, sizeof(entry));
2466 ip_vs_copy_service(&entry, svc);
2467 if (copy_to_user(&uptr->entrytable[count],
2468 &entry, sizeof(entry))) {
2480 __ip_vs_get_dest_entries(struct net *net, const struct ip_vs_get_dests *get,
2481 struct ip_vs_get_dests __user *uptr)
2483 struct ip_vs_service *svc;
2484 union nf_inet_addr addr = { .ip = get->addr };
2488 svc = __ip_vs_svc_fwm_find(net, AF_INET, get->fwmark);
2490 svc = __ip_vs_service_find(net, AF_INET, get->protocol, &addr,
2495 struct ip_vs_dest *dest;
2496 struct ip_vs_dest_entry entry;
2498 list_for_each_entry(dest, &svc->destinations, n_list) {
2499 if (count >= get->num_dests)
2502 entry.addr = dest->addr.ip;
2503 entry.port = dest->port;
2504 entry.conn_flags = atomic_read(&dest->conn_flags);
2505 entry.weight = atomic_read(&dest->weight);
2506 entry.u_threshold = dest->u_threshold;
2507 entry.l_threshold = dest->l_threshold;
2508 entry.activeconns = atomic_read(&dest->activeconns);
2509 entry.inactconns = atomic_read(&dest->inactconns);
2510 entry.persistconns = atomic_read(&dest->persistconns);
2511 ip_vs_copy_stats(&entry.stats, &dest->stats);
2512 if (copy_to_user(&uptr->entrytable[count],
2513 &entry, sizeof(entry))) {
2525 __ip_vs_get_timeouts(struct net *net, struct ip_vs_timeout_user *u)
2527 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2528 struct ip_vs_proto_data *pd;
2531 #ifdef CONFIG_IP_VS_PROTO_TCP
2532 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2533 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2534 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2536 #ifdef CONFIG_IP_VS_PROTO_UDP
2537 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2539 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2544 #define GET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2545 #define GET_INFO_ARG_LEN (sizeof(struct ip_vs_getinfo))
2546 #define GET_SERVICES_ARG_LEN (sizeof(struct ip_vs_get_services))
2547 #define GET_SERVICE_ARG_LEN (sizeof(struct ip_vs_service_entry))
2548 #define GET_DESTS_ARG_LEN (sizeof(struct ip_vs_get_dests))
2549 #define GET_TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2550 #define GET_DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user) * 2)
2552 static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
2553 [GET_CMDID(IP_VS_SO_GET_VERSION)] = 64,
2554 [GET_CMDID(IP_VS_SO_GET_INFO)] = GET_INFO_ARG_LEN,
2555 [GET_CMDID(IP_VS_SO_GET_SERVICES)] = GET_SERVICES_ARG_LEN,
2556 [GET_CMDID(IP_VS_SO_GET_SERVICE)] = GET_SERVICE_ARG_LEN,
2557 [GET_CMDID(IP_VS_SO_GET_DESTS)] = GET_DESTS_ARG_LEN,
2558 [GET_CMDID(IP_VS_SO_GET_TIMEOUT)] = GET_TIMEOUT_ARG_LEN,
2559 [GET_CMDID(IP_VS_SO_GET_DAEMON)] = GET_DAEMON_ARG_LEN,
2563 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2565 unsigned char arg[128];
2567 unsigned int copylen;
2568 struct net *net = sock_net(sk);
2569 struct netns_ipvs *ipvs = net_ipvs(net);
2572 if (!capable(CAP_NET_ADMIN))
2575 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2578 if (*len < get_arglen[GET_CMDID(cmd)]) {
2579 pr_err("get_ctl: len %u < %u\n",
2580 *len, get_arglen[GET_CMDID(cmd)]);
2584 copylen = get_arglen[GET_CMDID(cmd)];
2588 if (copy_from_user(arg, user, copylen) != 0)
2591 if (mutex_lock_interruptible(&__ip_vs_mutex))
2592 return -ERESTARTSYS;
2595 case IP_VS_SO_GET_VERSION:
2599 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2600 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2601 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2605 *len = strlen(buf)+1;
2609 case IP_VS_SO_GET_INFO:
2611 struct ip_vs_getinfo info;
2612 info.version = IP_VS_VERSION_CODE;
2613 info.size = ip_vs_conn_tab_size;
2614 info.num_services = ipvs->num_services;
2615 if (copy_to_user(user, &info, sizeof(info)) != 0)
2620 case IP_VS_SO_GET_SERVICES:
2622 struct ip_vs_get_services *get;
2625 get = (struct ip_vs_get_services *)arg;
2626 size = sizeof(*get) +
2627 sizeof(struct ip_vs_service_entry) * get->num_services;
2629 pr_err("length: %u != %u\n", *len, size);
2633 ret = __ip_vs_get_service_entries(net, get, user);
2637 case IP_VS_SO_GET_SERVICE:
2639 struct ip_vs_service_entry *entry;
2640 struct ip_vs_service *svc;
2641 union nf_inet_addr addr;
2643 entry = (struct ip_vs_service_entry *)arg;
2644 addr.ip = entry->addr;
2646 svc = __ip_vs_svc_fwm_find(net, AF_INET, entry->fwmark);
2648 svc = __ip_vs_service_find(net, AF_INET,
2649 entry->protocol, &addr,
2652 ip_vs_copy_service(entry, svc);
2653 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2660 case IP_VS_SO_GET_DESTS:
2662 struct ip_vs_get_dests *get;
2665 get = (struct ip_vs_get_dests *)arg;
2666 size = sizeof(*get) +
2667 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2669 pr_err("length: %u != %u\n", *len, size);
2673 ret = __ip_vs_get_dest_entries(net, get, user);
2677 case IP_VS_SO_GET_TIMEOUT:
2679 struct ip_vs_timeout_user t;
2681 __ip_vs_get_timeouts(net, &t);
2682 if (copy_to_user(user, &t, sizeof(t)) != 0)
2687 case IP_VS_SO_GET_DAEMON:
2689 struct ip_vs_daemon_user d[2];
2691 memset(&d, 0, sizeof(d));
2692 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2693 d[0].state = IP_VS_STATE_MASTER;
2694 strlcpy(d[0].mcast_ifn, ipvs->master_mcast_ifn,
2695 sizeof(d[0].mcast_ifn));
2696 d[0].syncid = ipvs->master_syncid;
2698 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2699 d[1].state = IP_VS_STATE_BACKUP;
2700 strlcpy(d[1].mcast_ifn, ipvs->backup_mcast_ifn,
2701 sizeof(d[1].mcast_ifn));
2702 d[1].syncid = ipvs->backup_syncid;
2704 if (copy_to_user(user, &d, sizeof(d)) != 0)
2714 mutex_unlock(&__ip_vs_mutex);
2719 static struct nf_sockopt_ops ip_vs_sockopts = {
2721 .set_optmin = IP_VS_BASE_CTL,
2722 .set_optmax = IP_VS_SO_SET_MAX+1,
2723 .set = do_ip_vs_set_ctl,
2724 .get_optmin = IP_VS_BASE_CTL,
2725 .get_optmax = IP_VS_SO_GET_MAX+1,
2726 .get = do_ip_vs_get_ctl,
2727 .owner = THIS_MODULE,
2731 * Generic Netlink interface
2734 /* IPVS genetlink family */
2735 static struct genl_family ip_vs_genl_family = {
2736 .id = GENL_ID_GENERATE,
2738 .name = IPVS_GENL_NAME,
2739 .version = IPVS_GENL_VERSION,
2740 .maxattr = IPVS_CMD_MAX,
2741 .netnsok = true, /* Make ipvsadm to work on netns */
2744 /* Policy used for first-level command attributes */
2745 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2746 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2747 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2748 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2749 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2750 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2751 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2754 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2755 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2756 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2757 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2758 .len = IP_VS_IFNAME_MAXLEN },
2759 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2762 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2763 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2764 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2765 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2766 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2767 .len = sizeof(union nf_inet_addr) },
2768 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2769 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2770 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2771 .len = IP_VS_SCHEDNAME_MAXLEN },
2772 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2773 .len = IP_VS_PENAME_MAXLEN },
2774 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2775 .len = sizeof(struct ip_vs_flags) },
2776 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2777 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2778 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2781 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2782 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2783 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2784 .len = sizeof(union nf_inet_addr) },
2785 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2786 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2787 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2788 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2789 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2790 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2791 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2792 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2793 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2796 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2797 struct ip_vs_stats *stats)
2799 struct ip_vs_stats_user ustats;
2800 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2804 ip_vs_copy_stats(&ustats, stats);
2806 NLA_PUT_U32(skb, IPVS_STATS_ATTR_CONNS, ustats.conns);
2807 NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPKTS, ustats.inpkts);
2808 NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPKTS, ustats.outpkts);
2809 NLA_PUT_U64(skb, IPVS_STATS_ATTR_INBYTES, ustats.inbytes);
2810 NLA_PUT_U64(skb, IPVS_STATS_ATTR_OUTBYTES, ustats.outbytes);
2811 NLA_PUT_U32(skb, IPVS_STATS_ATTR_CPS, ustats.cps);
2812 NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPPS, ustats.inpps);
2813 NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPPS, ustats.outpps);
2814 NLA_PUT_U32(skb, IPVS_STATS_ATTR_INBPS, ustats.inbps);
2815 NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTBPS, ustats.outbps);
2817 nla_nest_end(skb, nl_stats);
2822 nla_nest_cancel(skb, nl_stats);
2826 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2827 struct ip_vs_service *svc)
2829 struct nlattr *nl_service;
2830 struct ip_vs_flags flags = { .flags = svc->flags,
2833 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2837 NLA_PUT_U16(skb, IPVS_SVC_ATTR_AF, svc->af);
2840 NLA_PUT_U32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark);
2842 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol);
2843 NLA_PUT(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr);
2844 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PORT, svc->port);
2847 NLA_PUT_STRING(skb, IPVS_SVC_ATTR_SCHED_NAME, svc->scheduler->name);
2849 NLA_PUT_STRING(skb, IPVS_SVC_ATTR_PE_NAME, svc->pe->name);
2850 NLA_PUT(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags);
2851 NLA_PUT_U32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ);
2852 NLA_PUT_U32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask);
2854 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2855 goto nla_put_failure;
2857 nla_nest_end(skb, nl_service);
2862 nla_nest_cancel(skb, nl_service);
2866 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2867 struct ip_vs_service *svc,
2868 struct netlink_callback *cb)
2872 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2873 &ip_vs_genl_family, NLM_F_MULTI,
2874 IPVS_CMD_NEW_SERVICE);
2878 if (ip_vs_genl_fill_service(skb, svc) < 0)
2879 goto nla_put_failure;
2881 return genlmsg_end(skb, hdr);
2884 genlmsg_cancel(skb, hdr);
2888 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2889 struct netlink_callback *cb)
2892 int start = cb->args[0];
2893 struct ip_vs_service *svc;
2894 struct net *net = skb_sknet(skb);
2896 mutex_lock(&__ip_vs_mutex);
2897 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2898 list_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2899 if (++idx <= start || !net_eq(svc->net, net))
2901 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2903 goto nla_put_failure;
2908 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2909 list_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
2910 if (++idx <= start || !net_eq(svc->net, net))
2912 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2914 goto nla_put_failure;
2920 mutex_unlock(&__ip_vs_mutex);
2926 static int ip_vs_genl_parse_service(struct net *net,
2927 struct ip_vs_service_user_kern *usvc,
2928 struct nlattr *nla, int full_entry,
2929 struct ip_vs_service **ret_svc)
2931 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
2932 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
2933 struct ip_vs_service *svc;
2935 /* Parse mandatory identifying service fields first */
2937 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
2940 nla_af = attrs[IPVS_SVC_ATTR_AF];
2941 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
2942 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
2943 nla_port = attrs[IPVS_SVC_ATTR_PORT];
2944 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
2946 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
2949 memset(usvc, 0, sizeof(*usvc));
2951 usvc->af = nla_get_u16(nla_af);
2952 #ifdef CONFIG_IP_VS_IPV6
2953 if (usvc->af != AF_INET && usvc->af != AF_INET6)
2955 if (usvc->af != AF_INET)
2957 return -EAFNOSUPPORT;
2960 usvc->protocol = IPPROTO_TCP;
2961 usvc->fwmark = nla_get_u32(nla_fwmark);
2963 usvc->protocol = nla_get_u16(nla_protocol);
2964 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
2965 usvc->port = nla_get_u16(nla_port);
2970 svc = __ip_vs_svc_fwm_find(net, usvc->af, usvc->fwmark);
2972 svc = __ip_vs_service_find(net, usvc->af, usvc->protocol,
2973 &usvc->addr, usvc->port);
2976 /* If a full entry was requested, check for the additional fields */
2978 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
2980 struct ip_vs_flags flags;
2982 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
2983 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
2984 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
2985 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
2986 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
2988 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
2991 nla_memcpy(&flags, nla_flags, sizeof(flags));
2993 /* prefill flags from service if it already exists */
2995 usvc->flags = svc->flags;
2997 /* set new flags from userland */
2998 usvc->flags = (usvc->flags & ~flags.mask) |
2999 (flags.flags & flags.mask);
3000 usvc->sched_name = nla_data(nla_sched);
3001 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3002 usvc->timeout = nla_get_u32(nla_timeout);
3003 usvc->netmask = nla_get_u32(nla_netmask);
3009 static struct ip_vs_service *ip_vs_genl_find_service(struct net *net,
3012 struct ip_vs_service_user_kern usvc;
3013 struct ip_vs_service *svc;
3016 ret = ip_vs_genl_parse_service(net, &usvc, nla, 0, &svc);
3017 return ret ? ERR_PTR(ret) : svc;
3020 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3022 struct nlattr *nl_dest;
3024 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3028 NLA_PUT(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr);
3029 NLA_PUT_U16(skb, IPVS_DEST_ATTR_PORT, dest->port);
3031 NLA_PUT_U32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3032 atomic_read(&dest->conn_flags) & IP_VS_CONN_F_FWD_MASK);
3033 NLA_PUT_U32(skb, IPVS_DEST_ATTR_WEIGHT, atomic_read(&dest->weight));
3034 NLA_PUT_U32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold);
3035 NLA_PUT_U32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold);
3036 NLA_PUT_U32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3037 atomic_read(&dest->activeconns));
3038 NLA_PUT_U32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3039 atomic_read(&dest->inactconns));
3040 NLA_PUT_U32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3041 atomic_read(&dest->persistconns));
3043 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
3044 goto nla_put_failure;
3046 nla_nest_end(skb, nl_dest);
3051 nla_nest_cancel(skb, nl_dest);
3055 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3056 struct netlink_callback *cb)
3060 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
3061 &ip_vs_genl_family, NLM_F_MULTI,
3066 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3067 goto nla_put_failure;
3069 return genlmsg_end(skb, hdr);
3072 genlmsg_cancel(skb, hdr);
3076 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3077 struct netlink_callback *cb)
3080 int start = cb->args[0];
3081 struct ip_vs_service *svc;
3082 struct ip_vs_dest *dest;
3083 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3084 struct net *net = skb_sknet(skb);
3086 mutex_lock(&__ip_vs_mutex);
3088 /* Try to find the service for which to dump destinations */
3089 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3090 IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3094 svc = ip_vs_genl_find_service(net, attrs[IPVS_CMD_ATTR_SERVICE]);
3095 if (IS_ERR(svc) || svc == NULL)
3098 /* Dump the destinations */
3099 list_for_each_entry(dest, &svc->destinations, n_list) {
3102 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3104 goto nla_put_failure;
3112 mutex_unlock(&__ip_vs_mutex);
3117 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3118 struct nlattr *nla, int full_entry)
3120 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3121 struct nlattr *nla_addr, *nla_port;
3123 /* Parse mandatory identifying destination fields first */
3125 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3128 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3129 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3131 if (!(nla_addr && nla_port))
3134 memset(udest, 0, sizeof(*udest));
3136 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3137 udest->port = nla_get_u16(nla_port);
3139 /* If a full entry was requested, check for the additional fields */
3141 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3144 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3145 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3146 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3147 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3149 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3152 udest->conn_flags = nla_get_u32(nla_fwd)
3153 & IP_VS_CONN_F_FWD_MASK;
3154 udest->weight = nla_get_u32(nla_weight);
3155 udest->u_threshold = nla_get_u32(nla_u_thresh);
3156 udest->l_threshold = nla_get_u32(nla_l_thresh);
3162 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __be32 state,
3163 const char *mcast_ifn, __be32 syncid)
3165 struct nlattr *nl_daemon;
3167 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3171 NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_STATE, state);
3172 NLA_PUT_STRING(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn);
3173 NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid);
3175 nla_nest_end(skb, nl_daemon);
3180 nla_nest_cancel(skb, nl_daemon);
3184 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __be32 state,
3185 const char *mcast_ifn, __be32 syncid,
3186 struct netlink_callback *cb)
3189 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
3190 &ip_vs_genl_family, NLM_F_MULTI,
3191 IPVS_CMD_NEW_DAEMON);
3195 if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
3196 goto nla_put_failure;
3198 return genlmsg_end(skb, hdr);
3201 genlmsg_cancel(skb, hdr);
3205 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3206 struct netlink_callback *cb)
3208 struct net *net = skb_sknet(skb);
3209 struct netns_ipvs *ipvs = net_ipvs(net);
3211 mutex_lock(&__ip_vs_mutex);
3212 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3213 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3214 ipvs->master_mcast_ifn,
3215 ipvs->master_syncid, cb) < 0)
3216 goto nla_put_failure;
3221 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3222 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3223 ipvs->backup_mcast_ifn,
3224 ipvs->backup_syncid, cb) < 0)
3225 goto nla_put_failure;
3231 mutex_unlock(&__ip_vs_mutex);
3236 static int ip_vs_genl_new_daemon(struct net *net, struct nlattr **attrs)
3238 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3239 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3240 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3243 return start_sync_thread(net,
3244 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3245 nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3246 nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3249 static int ip_vs_genl_del_daemon(struct net *net, struct nlattr **attrs)
3251 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3254 return stop_sync_thread(net,
3255 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3258 static int ip_vs_genl_set_config(struct net *net, struct nlattr **attrs)
3260 struct ip_vs_timeout_user t;
3262 __ip_vs_get_timeouts(net, &t);
3264 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3265 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3267 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3269 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3271 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3272 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3274 return ip_vs_set_timeout(net, &t);
3277 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3279 struct ip_vs_service *svc = NULL;
3280 struct ip_vs_service_user_kern usvc;
3281 struct ip_vs_dest_user_kern udest;
3283 int need_full_svc = 0, need_full_dest = 0;
3285 struct netns_ipvs *ipvs;
3287 net = skb_sknet(skb);
3288 ipvs = net_ipvs(net);
3289 cmd = info->genlhdr->cmd;
3291 mutex_lock(&__ip_vs_mutex);
3293 if (cmd == IPVS_CMD_FLUSH) {
3294 ret = ip_vs_flush(net);
3296 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3297 ret = ip_vs_genl_set_config(net, info->attrs);
3299 } else if (cmd == IPVS_CMD_NEW_DAEMON ||
3300 cmd == IPVS_CMD_DEL_DAEMON) {
3302 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3304 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3305 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3306 info->attrs[IPVS_CMD_ATTR_DAEMON],
3307 ip_vs_daemon_policy)) {
3312 if (cmd == IPVS_CMD_NEW_DAEMON)
3313 ret = ip_vs_genl_new_daemon(net, daemon_attrs);
3315 ret = ip_vs_genl_del_daemon(net, daemon_attrs);
3317 } else if (cmd == IPVS_CMD_ZERO &&
3318 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3319 ret = ip_vs_zero_all(net);
3323 /* All following commands require a service argument, so check if we
3324 * received a valid one. We need a full service specification when
3325 * adding / editing a service. Only identifying members otherwise. */
3326 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3329 ret = ip_vs_genl_parse_service(net, &usvc,
3330 info->attrs[IPVS_CMD_ATTR_SERVICE],
3331 need_full_svc, &svc);
3335 /* Unless we're adding a new service, the service must already exist */
3336 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3341 /* Destination commands require a valid destination argument. For
3342 * adding / editing a destination, we need a full destination
3344 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3345 cmd == IPVS_CMD_DEL_DEST) {
3346 if (cmd != IPVS_CMD_DEL_DEST)
3349 ret = ip_vs_genl_parse_dest(&udest,
3350 info->attrs[IPVS_CMD_ATTR_DEST],
3357 case IPVS_CMD_NEW_SERVICE:
3359 ret = ip_vs_add_service(net, &usvc, &svc);
3363 case IPVS_CMD_SET_SERVICE:
3364 ret = ip_vs_edit_service(svc, &usvc);
3366 case IPVS_CMD_DEL_SERVICE:
3367 ret = ip_vs_del_service(svc);
3368 /* do not use svc, it can be freed */
3370 case IPVS_CMD_NEW_DEST:
3371 ret = ip_vs_add_dest(svc, &udest);
3373 case IPVS_CMD_SET_DEST:
3374 ret = ip_vs_edit_dest(svc, &udest);
3376 case IPVS_CMD_DEL_DEST:
3377 ret = ip_vs_del_dest(svc, &udest);
3380 ret = ip_vs_zero_service(svc);
3387 mutex_unlock(&__ip_vs_mutex);
3392 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3394 struct sk_buff *msg;
3396 int ret, cmd, reply_cmd;
3398 struct netns_ipvs *ipvs;
3400 net = skb_sknet(skb);
3401 ipvs = net_ipvs(net);
3402 cmd = info->genlhdr->cmd;
3404 if (cmd == IPVS_CMD_GET_SERVICE)
3405 reply_cmd = IPVS_CMD_NEW_SERVICE;
3406 else if (cmd == IPVS_CMD_GET_INFO)
3407 reply_cmd = IPVS_CMD_SET_INFO;
3408 else if (cmd == IPVS_CMD_GET_CONFIG)
3409 reply_cmd = IPVS_CMD_SET_CONFIG;
3411 pr_err("unknown Generic Netlink command\n");
3415 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3419 mutex_lock(&__ip_vs_mutex);
3421 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3423 goto nla_put_failure;
3426 case IPVS_CMD_GET_SERVICE:
3428 struct ip_vs_service *svc;
3430 svc = ip_vs_genl_find_service(net,
3431 info->attrs[IPVS_CMD_ATTR_SERVICE]);
3436 ret = ip_vs_genl_fill_service(msg, svc);
3438 goto nla_put_failure;
3447 case IPVS_CMD_GET_CONFIG:
3449 struct ip_vs_timeout_user t;
3451 __ip_vs_get_timeouts(net, &t);
3452 #ifdef CONFIG_IP_VS_PROTO_TCP
3453 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP, t.tcp_timeout);
3454 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3457 #ifdef CONFIG_IP_VS_PROTO_UDP
3458 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout);
3464 case IPVS_CMD_GET_INFO:
3465 NLA_PUT_U32(msg, IPVS_INFO_ATTR_VERSION, IP_VS_VERSION_CODE);
3466 NLA_PUT_U32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3467 ip_vs_conn_tab_size);
3471 genlmsg_end(msg, reply);
3472 ret = genlmsg_reply(msg, info);
3476 pr_err("not enough space in Netlink message\n");
3482 mutex_unlock(&__ip_vs_mutex);
3488 static struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3490 .cmd = IPVS_CMD_NEW_SERVICE,
3491 .flags = GENL_ADMIN_PERM,
3492 .policy = ip_vs_cmd_policy,
3493 .doit = ip_vs_genl_set_cmd,
3496 .cmd = IPVS_CMD_SET_SERVICE,
3497 .flags = GENL_ADMIN_PERM,
3498 .policy = ip_vs_cmd_policy,
3499 .doit = ip_vs_genl_set_cmd,
3502 .cmd = IPVS_CMD_DEL_SERVICE,
3503 .flags = GENL_ADMIN_PERM,
3504 .policy = ip_vs_cmd_policy,
3505 .doit = ip_vs_genl_set_cmd,
3508 .cmd = IPVS_CMD_GET_SERVICE,
3509 .flags = GENL_ADMIN_PERM,
3510 .doit = ip_vs_genl_get_cmd,
3511 .dumpit = ip_vs_genl_dump_services,
3512 .policy = ip_vs_cmd_policy,
3515 .cmd = IPVS_CMD_NEW_DEST,
3516 .flags = GENL_ADMIN_PERM,
3517 .policy = ip_vs_cmd_policy,
3518 .doit = ip_vs_genl_set_cmd,
3521 .cmd = IPVS_CMD_SET_DEST,
3522 .flags = GENL_ADMIN_PERM,
3523 .policy = ip_vs_cmd_policy,
3524 .doit = ip_vs_genl_set_cmd,
3527 .cmd = IPVS_CMD_DEL_DEST,
3528 .flags = GENL_ADMIN_PERM,
3529 .policy = ip_vs_cmd_policy,
3530 .doit = ip_vs_genl_set_cmd,
3533 .cmd = IPVS_CMD_GET_DEST,
3534 .flags = GENL_ADMIN_PERM,
3535 .policy = ip_vs_cmd_policy,
3536 .dumpit = ip_vs_genl_dump_dests,
3539 .cmd = IPVS_CMD_NEW_DAEMON,
3540 .flags = GENL_ADMIN_PERM,
3541 .policy = ip_vs_cmd_policy,
3542 .doit = ip_vs_genl_set_cmd,
3545 .cmd = IPVS_CMD_DEL_DAEMON,
3546 .flags = GENL_ADMIN_PERM,
3547 .policy = ip_vs_cmd_policy,
3548 .doit = ip_vs_genl_set_cmd,
3551 .cmd = IPVS_CMD_GET_DAEMON,
3552 .flags = GENL_ADMIN_PERM,
3553 .dumpit = ip_vs_genl_dump_daemons,
3556 .cmd = IPVS_CMD_SET_CONFIG,
3557 .flags = GENL_ADMIN_PERM,
3558 .policy = ip_vs_cmd_policy,
3559 .doit = ip_vs_genl_set_cmd,
3562 .cmd = IPVS_CMD_GET_CONFIG,
3563 .flags = GENL_ADMIN_PERM,
3564 .doit = ip_vs_genl_get_cmd,
3567 .cmd = IPVS_CMD_GET_INFO,
3568 .flags = GENL_ADMIN_PERM,
3569 .doit = ip_vs_genl_get_cmd,
3572 .cmd = IPVS_CMD_ZERO,
3573 .flags = GENL_ADMIN_PERM,
3574 .policy = ip_vs_cmd_policy,
3575 .doit = ip_vs_genl_set_cmd,
3578 .cmd = IPVS_CMD_FLUSH,
3579 .flags = GENL_ADMIN_PERM,
3580 .doit = ip_vs_genl_set_cmd,
3584 static int __init ip_vs_genl_register(void)
3586 return genl_register_family_with_ops(&ip_vs_genl_family,
3587 ip_vs_genl_ops, ARRAY_SIZE(ip_vs_genl_ops));
3590 static void ip_vs_genl_unregister(void)
3592 genl_unregister_family(&ip_vs_genl_family);
3595 /* End of Generic Netlink interface definitions */
3598 * per netns intit/exit func.
3600 #ifdef CONFIG_SYSCTL
3601 int __net_init __ip_vs_control_init_sysctl(struct net *net)
3604 struct netns_ipvs *ipvs = net_ipvs(net);
3605 struct ctl_table *tbl;
3607 atomic_set(&ipvs->dropentry, 0);
3608 spin_lock_init(&ipvs->dropentry_lock);
3609 spin_lock_init(&ipvs->droppacket_lock);
3610 spin_lock_init(&ipvs->securetcp_lock);
3612 if (!net_eq(net, &init_net)) {
3613 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3618 /* Initialize sysctl defaults */
3620 ipvs->sysctl_amemthresh = 1024;
3621 tbl[idx++].data = &ipvs->sysctl_amemthresh;
3622 ipvs->sysctl_am_droprate = 10;
3623 tbl[idx++].data = &ipvs->sysctl_am_droprate;
3624 tbl[idx++].data = &ipvs->sysctl_drop_entry;
3625 tbl[idx++].data = &ipvs->sysctl_drop_packet;
3626 #ifdef CONFIG_IP_VS_NFCT
3627 tbl[idx++].data = &ipvs->sysctl_conntrack;
3629 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3630 ipvs->sysctl_snat_reroute = 1;
3631 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3632 ipvs->sysctl_sync_ver = 1;
3633 tbl[idx++].data = &ipvs->sysctl_sync_ver;
3634 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3635 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3636 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3637 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3638 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3639 tbl[idx].data = &ipvs->sysctl_sync_threshold;
3640 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3641 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3644 ipvs->sysctl_hdr = register_net_sysctl_table(net, net_vs_ctl_path,
3646 if (ipvs->sysctl_hdr == NULL) {
3647 if (!net_eq(net, &init_net))
3651 ip_vs_start_estimator(net, &ipvs->tot_stats);
3652 ipvs->sysctl_tbl = tbl;
3653 /* Schedule defense work */
3654 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3655 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3660 void __net_init __ip_vs_control_cleanup_sysctl(struct net *net)
3662 struct netns_ipvs *ipvs = net_ipvs(net);
3664 cancel_delayed_work_sync(&ipvs->defense_work);
3665 cancel_work_sync(&ipvs->defense_work.work);
3666 unregister_net_sysctl_table(ipvs->sysctl_hdr);
3671 int __net_init __ip_vs_control_init_sysctl(struct net *net) { return 0; }
3672 void __net_init __ip_vs_control_cleanup_sysctl(struct net *net) { }
3676 static struct notifier_block ip_vs_dst_notifier = {
3677 .notifier_call = ip_vs_dst_event,
3680 int __net_init __ip_vs_control_init(struct net *net)
3683 struct netns_ipvs *ipvs = net_ipvs(net);
3685 ipvs->rs_lock = __RW_LOCK_UNLOCKED(ipvs->rs_lock);
3687 /* Initialize rs_table */
3688 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
3689 INIT_LIST_HEAD(&ipvs->rs_table[idx]);
3691 INIT_LIST_HEAD(&ipvs->dest_trash);
3692 atomic_set(&ipvs->ftpsvc_counter, 0);
3693 atomic_set(&ipvs->nullsvc_counter, 0);
3696 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
3697 if (!ipvs->tot_stats.cpustats) {
3698 pr_err("%s(): alloc_percpu.\n", __func__);
3701 spin_lock_init(&ipvs->tot_stats.lock);
3703 proc_net_fops_create(net, "ip_vs", 0, &ip_vs_info_fops);
3704 proc_net_fops_create(net, "ip_vs_stats", 0, &ip_vs_stats_fops);
3705 proc_net_fops_create(net, "ip_vs_stats_percpu", 0,
3706 &ip_vs_stats_percpu_fops);
3708 if (__ip_vs_control_init_sysctl(net))
3714 free_percpu(ipvs->tot_stats.cpustats);
3718 void __net_exit __ip_vs_control_cleanup(struct net *net)
3720 struct netns_ipvs *ipvs = net_ipvs(net);
3722 ip_vs_trash_cleanup(net);
3723 ip_vs_stop_estimator(net, &ipvs->tot_stats);
3724 __ip_vs_control_cleanup_sysctl(net);
3725 proc_net_remove(net, "ip_vs_stats_percpu");
3726 proc_net_remove(net, "ip_vs_stats");
3727 proc_net_remove(net, "ip_vs");
3728 free_percpu(ipvs->tot_stats.cpustats);
3731 int __init ip_vs_control_init(void)
3738 /* Initialize svc_table, ip_vs_svc_fwm_table, rs_table */
3739 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
3740 INIT_LIST_HEAD(&ip_vs_svc_table[idx]);
3741 INIT_LIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3744 smp_wmb(); /* Do we really need it now ? */
3746 ret = nf_register_sockopt(&ip_vs_sockopts);
3748 pr_err("cannot register sockopt.\n");
3752 ret = ip_vs_genl_register();
3754 pr_err("cannot register Generic Netlink interface.\n");
3758 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
3766 ip_vs_genl_unregister();
3768 nf_unregister_sockopt(&ip_vs_sockopts);
3774 void ip_vs_control_cleanup(void)
3777 ip_vs_genl_unregister();
3778 nf_unregister_sockopt(&ip_vs_sockopts);