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,
625 struct ip_vs_dest *dest;
626 struct ip_vs_service *svc;
629 svc = ip_vs_service_get(net, af, fwmark, protocol, vaddr, vport);
632 if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
634 dest = ip_vs_lookup_dest(svc, daddr, port);
636 dest = ip_vs_lookup_dest(svc, daddr, port ^ dport);
638 atomic_inc(&dest->refcnt);
639 ip_vs_service_put(svc);
644 * Lookup dest by {svc,addr,port} in the destination trash.
645 * The destination trash is used to hold the destinations that are removed
646 * from the service table but are still referenced by some conn entries.
647 * The reason to add the destination trash is when the dest is temporary
648 * down (either by administrator or by monitor program), the dest can be
649 * picked back from the trash, the remaining connections to the dest can
650 * continue, and the counting information of the dest is also useful for
653 static struct ip_vs_dest *
654 ip_vs_trash_get_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
657 struct ip_vs_dest *dest, *nxt;
658 struct netns_ipvs *ipvs = net_ipvs(svc->net);
661 * Find the destination in trash
663 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, n_list) {
664 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
667 IP_VS_DBG_ADDR(svc->af, &dest->addr),
669 atomic_read(&dest->refcnt));
670 if (dest->af == svc->af &&
671 ip_vs_addr_equal(svc->af, &dest->addr, daddr) &&
672 dest->port == dport &&
673 dest->vfwmark == svc->fwmark &&
674 dest->protocol == svc->protocol &&
676 (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
677 dest->vport == svc->port))) {
683 * Try to purge the destination from trash if not referenced
685 if (atomic_read(&dest->refcnt) == 1) {
686 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u "
689 IP_VS_DBG_ADDR(svc->af, &dest->addr),
691 list_del(&dest->n_list);
692 ip_vs_dst_reset(dest);
693 __ip_vs_unbind_svc(dest);
694 free_percpu(dest->stats.cpustats);
704 * Clean up all the destinations in the trash
705 * Called by the ip_vs_control_cleanup()
707 * When the ip_vs_control_clearup is activated by ipvs module exit,
708 * the service tables must have been flushed and all the connections
709 * are expired, and the refcnt of each destination in the trash must
710 * be 1, so we simply release them here.
712 static void ip_vs_trash_cleanup(struct net *net)
714 struct ip_vs_dest *dest, *nxt;
715 struct netns_ipvs *ipvs = net_ipvs(net);
717 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, n_list) {
718 list_del(&dest->n_list);
719 ip_vs_dst_reset(dest);
720 __ip_vs_unbind_svc(dest);
721 free_percpu(dest->stats.cpustats);
727 ip_vs_copy_stats(struct ip_vs_stats_user *dst, struct ip_vs_stats *src)
729 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->ustats.c - src->ustats0.c
731 spin_lock_bh(&src->lock);
733 IP_VS_SHOW_STATS_COUNTER(conns);
734 IP_VS_SHOW_STATS_COUNTER(inpkts);
735 IP_VS_SHOW_STATS_COUNTER(outpkts);
736 IP_VS_SHOW_STATS_COUNTER(inbytes);
737 IP_VS_SHOW_STATS_COUNTER(outbytes);
739 ip_vs_read_estimator(dst, src);
741 spin_unlock_bh(&src->lock);
745 ip_vs_zero_stats(struct ip_vs_stats *stats)
747 spin_lock_bh(&stats->lock);
749 /* get current counters as zero point, rates are zeroed */
751 #define IP_VS_ZERO_STATS_COUNTER(c) stats->ustats0.c = stats->ustats.c
753 IP_VS_ZERO_STATS_COUNTER(conns);
754 IP_VS_ZERO_STATS_COUNTER(inpkts);
755 IP_VS_ZERO_STATS_COUNTER(outpkts);
756 IP_VS_ZERO_STATS_COUNTER(inbytes);
757 IP_VS_ZERO_STATS_COUNTER(outbytes);
759 ip_vs_zero_estimator(stats);
761 spin_unlock_bh(&stats->lock);
765 * Update a destination in the given service
768 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
769 struct ip_vs_dest_user_kern *udest, int add)
771 struct netns_ipvs *ipvs = net_ipvs(svc->net);
774 /* set the weight and the flags */
775 atomic_set(&dest->weight, udest->weight);
776 conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
777 conn_flags |= IP_VS_CONN_F_INACTIVE;
779 /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
780 if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
781 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
784 * Put the real service in rs_table if not present.
785 * For now only for NAT!
787 write_lock_bh(&ipvs->rs_lock);
788 ip_vs_rs_hash(ipvs, dest);
789 write_unlock_bh(&ipvs->rs_lock);
791 atomic_set(&dest->conn_flags, conn_flags);
793 /* bind the service */
795 __ip_vs_bind_svc(dest, svc);
797 if (dest->svc != svc) {
798 __ip_vs_unbind_svc(dest);
799 ip_vs_zero_stats(&dest->stats);
800 __ip_vs_bind_svc(dest, svc);
804 /* set the dest status flags */
805 dest->flags |= IP_VS_DEST_F_AVAILABLE;
807 if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
808 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
809 dest->u_threshold = udest->u_threshold;
810 dest->l_threshold = udest->l_threshold;
812 spin_lock_bh(&dest->dst_lock);
813 ip_vs_dst_reset(dest);
814 spin_unlock_bh(&dest->dst_lock);
817 ip_vs_start_estimator(svc->net, &dest->stats);
819 write_lock_bh(&__ip_vs_svc_lock);
821 /* Wait until all other svc users go away */
822 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
825 list_add(&dest->n_list, &svc->destinations);
829 /* call the update_service, because server weight may be changed */
830 if (svc->scheduler->update_service)
831 svc->scheduler->update_service(svc);
833 write_unlock_bh(&__ip_vs_svc_lock);
838 * Create a destination for the given service
841 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
842 struct ip_vs_dest **dest_p)
844 struct ip_vs_dest *dest;
849 #ifdef CONFIG_IP_VS_IPV6
850 if (svc->af == AF_INET6) {
851 atype = ipv6_addr_type(&udest->addr.in6);
852 if ((!(atype & IPV6_ADDR_UNICAST) ||
853 atype & IPV6_ADDR_LINKLOCAL) &&
854 !__ip_vs_addr_is_local_v6(svc->net, &udest->addr.in6))
859 atype = inet_addr_type(svc->net, udest->addr.ip);
860 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
864 dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
868 dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
869 if (!dest->stats.cpustats)
873 dest->protocol = svc->protocol;
874 dest->vaddr = svc->addr;
875 dest->vport = svc->port;
876 dest->vfwmark = svc->fwmark;
877 ip_vs_addr_copy(svc->af, &dest->addr, &udest->addr);
878 dest->port = udest->port;
880 atomic_set(&dest->activeconns, 0);
881 atomic_set(&dest->inactconns, 0);
882 atomic_set(&dest->persistconns, 0);
883 atomic_set(&dest->refcnt, 1);
885 INIT_LIST_HEAD(&dest->d_list);
886 spin_lock_init(&dest->dst_lock);
887 spin_lock_init(&dest->stats.lock);
888 __ip_vs_update_dest(svc, dest, udest, 1);
902 * Add a destination into an existing service
905 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
907 struct ip_vs_dest *dest;
908 union nf_inet_addr daddr;
909 __be16 dport = udest->port;
914 if (udest->weight < 0) {
915 pr_err("%s(): server weight less than zero\n", __func__);
919 if (udest->l_threshold > udest->u_threshold) {
920 pr_err("%s(): lower threshold is higher than upper threshold\n",
925 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
928 * Check if the dest already exists in the list
930 dest = ip_vs_lookup_dest(svc, &daddr, dport);
933 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
938 * Check if the dest already exists in the trash and
939 * is from the same service
941 dest = ip_vs_trash_get_dest(svc, &daddr, dport);
944 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
945 "dest->refcnt=%d, service %u/%s:%u\n",
946 IP_VS_DBG_ADDR(svc->af, &daddr), ntohs(dport),
947 atomic_read(&dest->refcnt),
949 IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
953 * Get the destination from the trash
955 list_del(&dest->n_list);
957 __ip_vs_update_dest(svc, dest, udest, 1);
961 * Allocate and initialize the dest structure
963 ret = ip_vs_new_dest(svc, udest, &dest);
972 * Edit a destination in the given service
975 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
977 struct ip_vs_dest *dest;
978 union nf_inet_addr daddr;
979 __be16 dport = udest->port;
983 if (udest->weight < 0) {
984 pr_err("%s(): server weight less than zero\n", __func__);
988 if (udest->l_threshold > udest->u_threshold) {
989 pr_err("%s(): lower threshold is higher than upper threshold\n",
994 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
997 * Lookup the destination list
999 dest = ip_vs_lookup_dest(svc, &daddr, dport);
1002 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1006 __ip_vs_update_dest(svc, dest, udest, 0);
1014 * Delete a destination (must be already unlinked from the service)
1016 static void __ip_vs_del_dest(struct net *net, struct ip_vs_dest *dest)
1018 struct netns_ipvs *ipvs = net_ipvs(net);
1020 ip_vs_stop_estimator(net, &dest->stats);
1023 * Remove it from the d-linked list with the real services.
1025 write_lock_bh(&ipvs->rs_lock);
1026 ip_vs_rs_unhash(dest);
1027 write_unlock_bh(&ipvs->rs_lock);
1030 * Decrease the refcnt of the dest, and free the dest
1031 * if nobody refers to it (refcnt=0). Otherwise, throw
1032 * the destination into the trash.
1034 if (atomic_dec_and_test(&dest->refcnt)) {
1035 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u\n",
1037 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1039 ip_vs_dst_reset(dest);
1040 /* simply decrease svc->refcnt here, let the caller check
1041 and release the service if nobody refers to it.
1042 Only user context can release destination and service,
1043 and only one user context can update virtual service at a
1044 time, so the operation here is OK */
1045 atomic_dec(&dest->svc->refcnt);
1046 free_percpu(dest->stats.cpustats);
1049 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, "
1050 "dest->refcnt=%d\n",
1051 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1053 atomic_read(&dest->refcnt));
1054 list_add(&dest->n_list, &ipvs->dest_trash);
1055 atomic_inc(&dest->refcnt);
1061 * Unlink a destination from the given service
1063 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1064 struct ip_vs_dest *dest,
1067 dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1070 * Remove it from the d-linked destination list.
1072 list_del(&dest->n_list);
1076 * Call the update_service function of its scheduler
1078 if (svcupd && svc->scheduler->update_service)
1079 svc->scheduler->update_service(svc);
1084 * Delete a destination server in the given service
1087 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1089 struct ip_vs_dest *dest;
1090 __be16 dport = udest->port;
1094 dest = ip_vs_lookup_dest(svc, &udest->addr, dport);
1097 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1101 write_lock_bh(&__ip_vs_svc_lock);
1104 * Wait until all other svc users go away.
1106 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1109 * Unlink dest from the service
1111 __ip_vs_unlink_dest(svc, dest, 1);
1113 write_unlock_bh(&__ip_vs_svc_lock);
1116 * Delete the destination
1118 __ip_vs_del_dest(svc->net, dest);
1127 * Add a service into the service hash table
1130 ip_vs_add_service(struct net *net, struct ip_vs_service_user_kern *u,
1131 struct ip_vs_service **svc_p)
1134 struct ip_vs_scheduler *sched = NULL;
1135 struct ip_vs_pe *pe = NULL;
1136 struct ip_vs_service *svc = NULL;
1137 struct netns_ipvs *ipvs = net_ipvs(net);
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 if (u->pe_name && *u->pe_name) {
1151 pe = ip_vs_pe_getbyname(u->pe_name);
1153 pr_info("persistence engine module ip_vs_pe_%s "
1154 "not found\n", u->pe_name);
1160 #ifdef CONFIG_IP_VS_IPV6
1161 if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1167 svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1169 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1173 svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1174 if (!svc->stats.cpustats)
1177 /* I'm the first user of the service */
1178 atomic_set(&svc->usecnt, 0);
1179 atomic_set(&svc->refcnt, 0);
1182 svc->protocol = u->protocol;
1183 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1184 svc->port = u->port;
1185 svc->fwmark = u->fwmark;
1186 svc->flags = u->flags;
1187 svc->timeout = u->timeout * HZ;
1188 svc->netmask = u->netmask;
1191 INIT_LIST_HEAD(&svc->destinations);
1192 rwlock_init(&svc->sched_lock);
1193 spin_lock_init(&svc->stats.lock);
1195 /* Bind the scheduler */
1196 ret = ip_vs_bind_scheduler(svc, sched);
1201 /* Bind the ct retriever */
1202 ip_vs_bind_pe(svc, pe);
1205 /* Update the virtual service counters */
1206 if (svc->port == FTPPORT)
1207 atomic_inc(&ipvs->ftpsvc_counter);
1208 else if (svc->port == 0)
1209 atomic_inc(&ipvs->nullsvc_counter);
1211 ip_vs_start_estimator(net, &svc->stats);
1213 /* Count only IPv4 services for old get/setsockopt interface */
1214 if (svc->af == AF_INET)
1215 ipvs->num_services++;
1217 /* Hash the service into the service table */
1218 write_lock_bh(&__ip_vs_svc_lock);
1219 ip_vs_svc_hash(svc);
1220 write_unlock_bh(&__ip_vs_svc_lock);
1223 /* Now there is a service - full throttle */
1230 ip_vs_unbind_scheduler(svc);
1233 ip_vs_app_inc_put(svc->inc);
1236 if (svc->stats.cpustats)
1237 free_percpu(svc->stats.cpustats);
1240 ip_vs_scheduler_put(sched);
1243 /* decrease the module use count */
1244 ip_vs_use_count_dec();
1251 * Edit a service and bind it with a new scheduler
1254 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1256 struct ip_vs_scheduler *sched, *old_sched;
1257 struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1261 * Lookup the scheduler, by 'u->sched_name'
1263 sched = ip_vs_scheduler_get(u->sched_name);
1264 if (sched == NULL) {
1265 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1270 if (u->pe_name && *u->pe_name) {
1271 pe = ip_vs_pe_getbyname(u->pe_name);
1273 pr_info("persistence engine module ip_vs_pe_%s "
1274 "not found\n", u->pe_name);
1281 #ifdef CONFIG_IP_VS_IPV6
1282 if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1288 write_lock_bh(&__ip_vs_svc_lock);
1291 * Wait until all other svc users go away.
1293 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1296 * Set the flags and timeout value
1298 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1299 svc->timeout = u->timeout * HZ;
1300 svc->netmask = u->netmask;
1302 old_sched = svc->scheduler;
1303 if (sched != old_sched) {
1305 * Unbind the old scheduler
1307 if ((ret = ip_vs_unbind_scheduler(svc))) {
1313 * Bind the new scheduler
1315 if ((ret = ip_vs_bind_scheduler(svc, sched))) {
1317 * If ip_vs_bind_scheduler fails, restore the old
1319 * The main reason of failure is out of memory.
1321 * The question is if the old scheduler can be
1322 * restored all the time. TODO: if it cannot be
1323 * restored some time, we must delete the service,
1324 * otherwise the system may crash.
1326 ip_vs_bind_scheduler(svc, old_sched);
1334 ip_vs_unbind_pe(svc);
1335 ip_vs_bind_pe(svc, pe);
1339 write_unlock_bh(&__ip_vs_svc_lock);
1341 ip_vs_scheduler_put(old_sched);
1342 ip_vs_pe_put(old_pe);
1348 * Delete a service from the service list
1349 * - The service must be unlinked, unlocked and not referenced!
1350 * - We are called under _bh lock
1352 static void __ip_vs_del_service(struct ip_vs_service *svc)
1354 struct ip_vs_dest *dest, *nxt;
1355 struct ip_vs_scheduler *old_sched;
1356 struct ip_vs_pe *old_pe;
1357 struct netns_ipvs *ipvs = net_ipvs(svc->net);
1359 pr_info("%s: enter\n", __func__);
1361 /* Count only IPv4 services for old get/setsockopt interface */
1362 if (svc->af == AF_INET)
1363 ipvs->num_services--;
1365 ip_vs_stop_estimator(svc->net, &svc->stats);
1367 /* Unbind scheduler */
1368 old_sched = svc->scheduler;
1369 ip_vs_unbind_scheduler(svc);
1370 ip_vs_scheduler_put(old_sched);
1372 /* Unbind persistence engine */
1374 ip_vs_unbind_pe(svc);
1375 ip_vs_pe_put(old_pe);
1377 /* Unbind app inc */
1379 ip_vs_app_inc_put(svc->inc);
1384 * Unlink the whole destination list
1386 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1387 __ip_vs_unlink_dest(svc, dest, 0);
1388 __ip_vs_del_dest(svc->net, dest);
1392 * Update the virtual service counters
1394 if (svc->port == FTPPORT)
1395 atomic_dec(&ipvs->ftpsvc_counter);
1396 else if (svc->port == 0)
1397 atomic_dec(&ipvs->nullsvc_counter);
1400 * Free the service if nobody refers to it
1402 if (atomic_read(&svc->refcnt) == 0) {
1403 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
1405 IP_VS_DBG_ADDR(svc->af, &svc->addr),
1406 ntohs(svc->port), atomic_read(&svc->usecnt));
1407 free_percpu(svc->stats.cpustats);
1411 /* decrease the module use count */
1412 ip_vs_use_count_dec();
1416 * Unlink a service from list and try to delete it if its refcnt reached 0
1418 static void ip_vs_unlink_service(struct ip_vs_service *svc)
1421 * Unhash it from the service table
1423 write_lock_bh(&__ip_vs_svc_lock);
1425 ip_vs_svc_unhash(svc);
1428 * Wait until all the svc users go away.
1430 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1432 __ip_vs_del_service(svc);
1434 write_unlock_bh(&__ip_vs_svc_lock);
1438 * Delete a service from the service list
1440 static int ip_vs_del_service(struct ip_vs_service *svc)
1444 ip_vs_unlink_service(svc);
1451 * Flush all the virtual services
1453 static int ip_vs_flush(struct net *net)
1456 struct ip_vs_service *svc, *nxt;
1459 * Flush the service table hashed by <netns,protocol,addr,port>
1461 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1462 list_for_each_entry_safe(svc, nxt, &ip_vs_svc_table[idx],
1464 if (net_eq(svc->net, net))
1465 ip_vs_unlink_service(svc);
1470 * Flush the service table hashed by fwmark
1472 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1473 list_for_each_entry_safe(svc, nxt,
1474 &ip_vs_svc_fwm_table[idx], f_list) {
1475 if (net_eq(svc->net, net))
1476 ip_vs_unlink_service(svc);
1484 * Delete service by {netns} in the service table.
1485 * Called by __ip_vs_cleanup()
1487 void ip_vs_service_net_cleanup(struct net *net)
1490 /* Check for "full" addressed entries */
1491 mutex_lock(&__ip_vs_mutex);
1493 mutex_unlock(&__ip_vs_mutex);
1497 * Release dst hold by dst_cache
1500 __ip_vs_dev_reset(struct ip_vs_dest *dest, struct net_device *dev)
1502 spin_lock_bh(&dest->dst_lock);
1503 if (dest->dst_cache && dest->dst_cache->dev == dev) {
1504 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1506 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1508 atomic_read(&dest->refcnt));
1509 ip_vs_dst_reset(dest);
1511 spin_unlock_bh(&dest->dst_lock);
1515 * Netdev event receiver
1516 * Currently only NETDEV_UNREGISTER is handled, i.e. if we hold a reference to
1517 * a device that is "unregister" it must be released.
1519 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1522 struct net_device *dev = ptr;
1523 struct net *net = dev_net(dev);
1524 struct ip_vs_service *svc;
1525 struct ip_vs_dest *dest;
1528 if (event != NETDEV_UNREGISTER)
1530 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1532 mutex_lock(&__ip_vs_mutex);
1533 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1534 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1535 if (net_eq(svc->net, net)) {
1536 list_for_each_entry(dest, &svc->destinations,
1538 __ip_vs_dev_reset(dest, dev);
1543 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1544 if (net_eq(svc->net, net)) {
1545 list_for_each_entry(dest, &svc->destinations,
1547 __ip_vs_dev_reset(dest, dev);
1554 list_for_each_entry(dest, &net_ipvs(net)->dest_trash, n_list) {
1555 __ip_vs_dev_reset(dest, dev);
1557 mutex_unlock(&__ip_vs_mutex);
1563 * Zero counters in a service or all services
1565 static int ip_vs_zero_service(struct ip_vs_service *svc)
1567 struct ip_vs_dest *dest;
1569 write_lock_bh(&__ip_vs_svc_lock);
1570 list_for_each_entry(dest, &svc->destinations, n_list) {
1571 ip_vs_zero_stats(&dest->stats);
1573 ip_vs_zero_stats(&svc->stats);
1574 write_unlock_bh(&__ip_vs_svc_lock);
1578 static int ip_vs_zero_all(struct net *net)
1581 struct ip_vs_service *svc;
1583 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1584 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1585 if (net_eq(svc->net, net))
1586 ip_vs_zero_service(svc);
1590 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1591 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1592 if (net_eq(svc->net, net))
1593 ip_vs_zero_service(svc);
1597 ip_vs_zero_stats(&net_ipvs(net)->tot_stats);
1601 #ifdef CONFIG_SYSCTL
1603 proc_do_defense_mode(ctl_table *table, int write,
1604 void __user *buffer, size_t *lenp, loff_t *ppos)
1606 struct net *net = current->nsproxy->net_ns;
1607 int *valp = table->data;
1611 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1612 if (write && (*valp != val)) {
1613 if ((*valp < 0) || (*valp > 3)) {
1614 /* Restore the correct value */
1617 update_defense_level(net_ipvs(net));
1624 proc_do_sync_threshold(ctl_table *table, int write,
1625 void __user *buffer, size_t *lenp, loff_t *ppos)
1627 int *valp = table->data;
1631 /* backup the value first */
1632 memcpy(val, valp, sizeof(val));
1634 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1635 if (write && (valp[0] < 0 || valp[1] < 0 || valp[0] >= valp[1])) {
1636 /* Restore the correct value */
1637 memcpy(valp, val, sizeof(val));
1643 proc_do_sync_mode(ctl_table *table, int write,
1644 void __user *buffer, size_t *lenp, loff_t *ppos)
1646 int *valp = table->data;
1650 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1651 if (write && (*valp != val)) {
1652 if ((*valp < 0) || (*valp > 1)) {
1653 /* Restore the correct value */
1656 struct net *net = current->nsproxy->net_ns;
1657 ip_vs_sync_switch_mode(net, val);
1664 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1665 * Do not change order or insert new entries without
1666 * align with netns init in ip_vs_control_net_init()
1669 static struct ctl_table vs_vars[] = {
1671 .procname = "amemthresh",
1672 .maxlen = sizeof(int),
1674 .proc_handler = proc_dointvec,
1677 .procname = "am_droprate",
1678 .maxlen = sizeof(int),
1680 .proc_handler = proc_dointvec,
1683 .procname = "drop_entry",
1684 .maxlen = sizeof(int),
1686 .proc_handler = proc_do_defense_mode,
1689 .procname = "drop_packet",
1690 .maxlen = sizeof(int),
1692 .proc_handler = proc_do_defense_mode,
1694 #ifdef CONFIG_IP_VS_NFCT
1696 .procname = "conntrack",
1697 .maxlen = sizeof(int),
1699 .proc_handler = &proc_dointvec,
1703 .procname = "secure_tcp",
1704 .maxlen = sizeof(int),
1706 .proc_handler = proc_do_defense_mode,
1709 .procname = "snat_reroute",
1710 .maxlen = sizeof(int),
1712 .proc_handler = &proc_dointvec,
1715 .procname = "sync_version",
1716 .maxlen = sizeof(int),
1718 .proc_handler = &proc_do_sync_mode,
1721 .procname = "cache_bypass",
1722 .maxlen = sizeof(int),
1724 .proc_handler = proc_dointvec,
1727 .procname = "expire_nodest_conn",
1728 .maxlen = sizeof(int),
1730 .proc_handler = proc_dointvec,
1733 .procname = "expire_quiescent_template",
1734 .maxlen = sizeof(int),
1736 .proc_handler = proc_dointvec,
1739 .procname = "sync_threshold",
1741 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1743 .proc_handler = proc_do_sync_threshold,
1746 .procname = "nat_icmp_send",
1747 .maxlen = sizeof(int),
1749 .proc_handler = proc_dointvec,
1751 #ifdef CONFIG_IP_VS_DEBUG
1753 .procname = "debug_level",
1754 .data = &sysctl_ip_vs_debug_level,
1755 .maxlen = sizeof(int),
1757 .proc_handler = proc_dointvec,
1762 .procname = "timeout_established",
1763 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1764 .maxlen = sizeof(int),
1766 .proc_handler = proc_dointvec_jiffies,
1769 .procname = "timeout_synsent",
1770 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1771 .maxlen = sizeof(int),
1773 .proc_handler = proc_dointvec_jiffies,
1776 .procname = "timeout_synrecv",
1777 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1778 .maxlen = sizeof(int),
1780 .proc_handler = proc_dointvec_jiffies,
1783 .procname = "timeout_finwait",
1784 .data = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1785 .maxlen = sizeof(int),
1787 .proc_handler = proc_dointvec_jiffies,
1790 .procname = "timeout_timewait",
1791 .data = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1792 .maxlen = sizeof(int),
1794 .proc_handler = proc_dointvec_jiffies,
1797 .procname = "timeout_close",
1798 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1799 .maxlen = sizeof(int),
1801 .proc_handler = proc_dointvec_jiffies,
1804 .procname = "timeout_closewait",
1805 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1806 .maxlen = sizeof(int),
1808 .proc_handler = proc_dointvec_jiffies,
1811 .procname = "timeout_lastack",
1812 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1813 .maxlen = sizeof(int),
1815 .proc_handler = proc_dointvec_jiffies,
1818 .procname = "timeout_listen",
1819 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1820 .maxlen = sizeof(int),
1822 .proc_handler = proc_dointvec_jiffies,
1825 .procname = "timeout_synack",
1826 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1827 .maxlen = sizeof(int),
1829 .proc_handler = proc_dointvec_jiffies,
1832 .procname = "timeout_udp",
1833 .data = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1834 .maxlen = sizeof(int),
1836 .proc_handler = proc_dointvec_jiffies,
1839 .procname = "timeout_icmp",
1840 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1841 .maxlen = sizeof(int),
1843 .proc_handler = proc_dointvec_jiffies,
1849 const struct ctl_path net_vs_ctl_path[] = {
1850 { .procname = "net", },
1851 { .procname = "ipv4", },
1852 { .procname = "vs", },
1855 EXPORT_SYMBOL_GPL(net_vs_ctl_path);
1858 #ifdef CONFIG_PROC_FS
1861 struct seq_net_private p; /* Do not move this, netns depends upon it*/
1862 struct list_head *table;
1867 * Write the contents of the VS rule table to a PROCfs file.
1868 * (It is kept just for backward compatibility)
1870 static inline const char *ip_vs_fwd_name(unsigned flags)
1872 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1873 case IP_VS_CONN_F_LOCALNODE:
1875 case IP_VS_CONN_F_TUNNEL:
1877 case IP_VS_CONN_F_DROUTE:
1885 /* Get the Nth entry in the two lists */
1886 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1888 struct net *net = seq_file_net(seq);
1889 struct ip_vs_iter *iter = seq->private;
1891 struct ip_vs_service *svc;
1893 /* look in hash by protocol */
1894 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1895 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1896 if (net_eq(svc->net, net) && pos-- == 0) {
1897 iter->table = ip_vs_svc_table;
1904 /* keep looking in fwmark */
1905 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1906 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1907 if (net_eq(svc->net, net) && pos-- == 0) {
1908 iter->table = ip_vs_svc_fwm_table;
1918 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1919 __acquires(__ip_vs_svc_lock)
1922 read_lock_bh(&__ip_vs_svc_lock);
1923 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1927 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1929 struct list_head *e;
1930 struct ip_vs_iter *iter;
1931 struct ip_vs_service *svc;
1934 if (v == SEQ_START_TOKEN)
1935 return ip_vs_info_array(seq,0);
1938 iter = seq->private;
1940 if (iter->table == ip_vs_svc_table) {
1941 /* next service in table hashed by protocol */
1942 if ((e = svc->s_list.next) != &ip_vs_svc_table[iter->bucket])
1943 return list_entry(e, struct ip_vs_service, s_list);
1946 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1947 list_for_each_entry(svc,&ip_vs_svc_table[iter->bucket],
1953 iter->table = ip_vs_svc_fwm_table;
1958 /* next service in hashed by fwmark */
1959 if ((e = svc->f_list.next) != &ip_vs_svc_fwm_table[iter->bucket])
1960 return list_entry(e, struct ip_vs_service, f_list);
1963 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1964 list_for_each_entry(svc, &ip_vs_svc_fwm_table[iter->bucket],
1972 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
1973 __releases(__ip_vs_svc_lock)
1975 read_unlock_bh(&__ip_vs_svc_lock);
1979 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
1981 if (v == SEQ_START_TOKEN) {
1983 "IP Virtual Server version %d.%d.%d (size=%d)\n",
1984 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
1986 "Prot LocalAddress:Port Scheduler Flags\n");
1988 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
1990 const struct ip_vs_service *svc = v;
1991 const struct ip_vs_iter *iter = seq->private;
1992 const struct ip_vs_dest *dest;
1994 if (iter->table == ip_vs_svc_table) {
1995 #ifdef CONFIG_IP_VS_IPV6
1996 if (svc->af == AF_INET6)
1997 seq_printf(seq, "%s [%pI6]:%04X %s ",
1998 ip_vs_proto_name(svc->protocol),
2001 svc->scheduler->name);
2004 seq_printf(seq, "%s %08X:%04X %s %s ",
2005 ip_vs_proto_name(svc->protocol),
2006 ntohl(svc->addr.ip),
2008 svc->scheduler->name,
2009 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2011 seq_printf(seq, "FWM %08X %s %s",
2012 svc->fwmark, svc->scheduler->name,
2013 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2016 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2017 seq_printf(seq, "persistent %d %08X\n",
2019 ntohl(svc->netmask));
2021 seq_putc(seq, '\n');
2023 list_for_each_entry(dest, &svc->destinations, n_list) {
2024 #ifdef CONFIG_IP_VS_IPV6
2025 if (dest->af == AF_INET6)
2028 " %-7s %-6d %-10d %-10d\n",
2031 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2032 atomic_read(&dest->weight),
2033 atomic_read(&dest->activeconns),
2034 atomic_read(&dest->inactconns));
2039 "%-7s %-6d %-10d %-10d\n",
2040 ntohl(dest->addr.ip),
2042 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2043 atomic_read(&dest->weight),
2044 atomic_read(&dest->activeconns),
2045 atomic_read(&dest->inactconns));
2052 static const struct seq_operations ip_vs_info_seq_ops = {
2053 .start = ip_vs_info_seq_start,
2054 .next = ip_vs_info_seq_next,
2055 .stop = ip_vs_info_seq_stop,
2056 .show = ip_vs_info_seq_show,
2059 static int ip_vs_info_open(struct inode *inode, struct file *file)
2061 return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2062 sizeof(struct ip_vs_iter));
2065 static const struct file_operations ip_vs_info_fops = {
2066 .owner = THIS_MODULE,
2067 .open = ip_vs_info_open,
2069 .llseek = seq_lseek,
2070 .release = seq_release_net,
2073 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2075 struct net *net = seq_file_single_net(seq);
2076 struct ip_vs_stats_user show;
2078 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2080 " Total Incoming Outgoing Incoming Outgoing\n");
2082 " Conns Packets Packets Bytes Bytes\n");
2084 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2085 seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", show.conns,
2086 show.inpkts, show.outpkts,
2087 (unsigned long long) show.inbytes,
2088 (unsigned long long) show.outbytes);
2090 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2092 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2093 seq_printf(seq, "%8X %8X %8X %16X %16X\n",
2094 show.cps, show.inpps, show.outpps,
2095 show.inbps, show.outbps);
2100 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2102 return single_open_net(inode, file, ip_vs_stats_show);
2105 static const struct file_operations ip_vs_stats_fops = {
2106 .owner = THIS_MODULE,
2107 .open = ip_vs_stats_seq_open,
2109 .llseek = seq_lseek,
2110 .release = single_release_net,
2113 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2115 struct net *net = seq_file_single_net(seq);
2116 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2117 struct ip_vs_cpu_stats *cpustats = tot_stats->cpustats;
2118 struct ip_vs_stats_user rates;
2121 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2123 " Total Incoming Outgoing Incoming Outgoing\n");
2125 "CPU Conns Packets Packets Bytes Bytes\n");
2127 for_each_possible_cpu(i) {
2128 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2130 __u64 inbytes, outbytes;
2133 start = u64_stats_fetch_begin_bh(&u->syncp);
2134 inbytes = u->ustats.inbytes;
2135 outbytes = u->ustats.outbytes;
2136 } while (u64_stats_fetch_retry_bh(&u->syncp, start));
2138 seq_printf(seq, "%3X %8X %8X %8X %16LX %16LX\n",
2139 i, u->ustats.conns, u->ustats.inpkts,
2140 u->ustats.outpkts, (__u64)inbytes,
2144 spin_lock_bh(&tot_stats->lock);
2146 seq_printf(seq, " ~ %8X %8X %8X %16LX %16LX\n\n",
2147 tot_stats->ustats.conns, tot_stats->ustats.inpkts,
2148 tot_stats->ustats.outpkts,
2149 (unsigned long long) tot_stats->ustats.inbytes,
2150 (unsigned long long) tot_stats->ustats.outbytes);
2152 ip_vs_read_estimator(&rates, tot_stats);
2154 spin_unlock_bh(&tot_stats->lock);
2156 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2158 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2159 seq_printf(seq, " %8X %8X %8X %16X %16X\n",
2169 static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2171 return single_open_net(inode, file, ip_vs_stats_percpu_show);
2174 static const struct file_operations ip_vs_stats_percpu_fops = {
2175 .owner = THIS_MODULE,
2176 .open = ip_vs_stats_percpu_seq_open,
2178 .llseek = seq_lseek,
2179 .release = single_release_net,
2184 * Set timeout values for tcp tcpfin udp in the timeout_table.
2186 static int ip_vs_set_timeout(struct net *net, struct ip_vs_timeout_user *u)
2188 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2189 struct ip_vs_proto_data *pd;
2192 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2197 #ifdef CONFIG_IP_VS_PROTO_TCP
2198 if (u->tcp_timeout) {
2199 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2200 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2201 = u->tcp_timeout * HZ;
2204 if (u->tcp_fin_timeout) {
2205 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2206 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2207 = u->tcp_fin_timeout * HZ;
2211 #ifdef CONFIG_IP_VS_PROTO_UDP
2212 if (u->udp_timeout) {
2213 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2214 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2215 = u->udp_timeout * HZ;
2222 #define SET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2223 #define SERVICE_ARG_LEN (sizeof(struct ip_vs_service_user))
2224 #define SVCDEST_ARG_LEN (sizeof(struct ip_vs_service_user) + \
2225 sizeof(struct ip_vs_dest_user))
2226 #define TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2227 #define DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user))
2228 #define MAX_ARG_LEN SVCDEST_ARG_LEN
2230 static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
2231 [SET_CMDID(IP_VS_SO_SET_ADD)] = SERVICE_ARG_LEN,
2232 [SET_CMDID(IP_VS_SO_SET_EDIT)] = SERVICE_ARG_LEN,
2233 [SET_CMDID(IP_VS_SO_SET_DEL)] = SERVICE_ARG_LEN,
2234 [SET_CMDID(IP_VS_SO_SET_FLUSH)] = 0,
2235 [SET_CMDID(IP_VS_SO_SET_ADDDEST)] = SVCDEST_ARG_LEN,
2236 [SET_CMDID(IP_VS_SO_SET_DELDEST)] = SVCDEST_ARG_LEN,
2237 [SET_CMDID(IP_VS_SO_SET_EDITDEST)] = SVCDEST_ARG_LEN,
2238 [SET_CMDID(IP_VS_SO_SET_TIMEOUT)] = TIMEOUT_ARG_LEN,
2239 [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)] = DAEMON_ARG_LEN,
2240 [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)] = DAEMON_ARG_LEN,
2241 [SET_CMDID(IP_VS_SO_SET_ZERO)] = SERVICE_ARG_LEN,
2244 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2245 struct ip_vs_service_user *usvc_compat)
2247 memset(usvc, 0, sizeof(*usvc));
2250 usvc->protocol = usvc_compat->protocol;
2251 usvc->addr.ip = usvc_compat->addr;
2252 usvc->port = usvc_compat->port;
2253 usvc->fwmark = usvc_compat->fwmark;
2255 /* Deep copy of sched_name is not needed here */
2256 usvc->sched_name = usvc_compat->sched_name;
2258 usvc->flags = usvc_compat->flags;
2259 usvc->timeout = usvc_compat->timeout;
2260 usvc->netmask = usvc_compat->netmask;
2263 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2264 struct ip_vs_dest_user *udest_compat)
2266 memset(udest, 0, sizeof(*udest));
2268 udest->addr.ip = udest_compat->addr;
2269 udest->port = udest_compat->port;
2270 udest->conn_flags = udest_compat->conn_flags;
2271 udest->weight = udest_compat->weight;
2272 udest->u_threshold = udest_compat->u_threshold;
2273 udest->l_threshold = udest_compat->l_threshold;
2277 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2279 struct net *net = sock_net(sk);
2281 unsigned char arg[MAX_ARG_LEN];
2282 struct ip_vs_service_user *usvc_compat;
2283 struct ip_vs_service_user_kern usvc;
2284 struct ip_vs_service *svc;
2285 struct ip_vs_dest_user *udest_compat;
2286 struct ip_vs_dest_user_kern udest;
2287 struct netns_ipvs *ipvs = net_ipvs(net);
2289 if (!capable(CAP_NET_ADMIN))
2292 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2294 if (len < 0 || len > MAX_ARG_LEN)
2296 if (len != set_arglen[SET_CMDID(cmd)]) {
2297 pr_err("set_ctl: len %u != %u\n",
2298 len, set_arglen[SET_CMDID(cmd)]);
2302 if (copy_from_user(arg, user, len) != 0)
2305 /* increase the module use count */
2306 ip_vs_use_count_inc();
2308 /* Handle daemons since they have another lock */
2309 if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2310 cmd == IP_VS_SO_SET_STOPDAEMON) {
2311 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2313 if (mutex_lock_interruptible(&ipvs->sync_mutex)) {
2317 if (cmd == IP_VS_SO_SET_STARTDAEMON)
2318 ret = start_sync_thread(net, dm->state, dm->mcast_ifn,
2321 ret = stop_sync_thread(net, dm->state);
2322 mutex_unlock(&ipvs->sync_mutex);
2326 if (mutex_lock_interruptible(&__ip_vs_mutex)) {
2331 if (cmd == IP_VS_SO_SET_FLUSH) {
2332 /* Flush the virtual service */
2333 ret = ip_vs_flush(net);
2335 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2336 /* Set timeout values for (tcp tcpfin udp) */
2337 ret = ip_vs_set_timeout(net, (struct ip_vs_timeout_user *)arg);
2341 usvc_compat = (struct ip_vs_service_user *)arg;
2342 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2344 /* We only use the new structs internally, so copy userspace compat
2345 * structs to extended internal versions */
2346 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2347 ip_vs_copy_udest_compat(&udest, udest_compat);
2349 if (cmd == IP_VS_SO_SET_ZERO) {
2350 /* if no service address is set, zero counters in all */
2351 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2352 ret = ip_vs_zero_all(net);
2357 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2358 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2359 usvc.protocol != IPPROTO_SCTP) {
2360 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2361 usvc.protocol, &usvc.addr.ip,
2362 ntohs(usvc.port), usvc.sched_name);
2367 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2368 if (usvc.fwmark == 0)
2369 svc = __ip_vs_service_find(net, usvc.af, usvc.protocol,
2370 &usvc.addr, usvc.port);
2372 svc = __ip_vs_svc_fwm_find(net, usvc.af, usvc.fwmark);
2374 if (cmd != IP_VS_SO_SET_ADD
2375 && (svc == NULL || svc->protocol != usvc.protocol)) {
2381 case IP_VS_SO_SET_ADD:
2385 ret = ip_vs_add_service(net, &usvc, &svc);
2387 case IP_VS_SO_SET_EDIT:
2388 ret = ip_vs_edit_service(svc, &usvc);
2390 case IP_VS_SO_SET_DEL:
2391 ret = ip_vs_del_service(svc);
2395 case IP_VS_SO_SET_ZERO:
2396 ret = ip_vs_zero_service(svc);
2398 case IP_VS_SO_SET_ADDDEST:
2399 ret = ip_vs_add_dest(svc, &udest);
2401 case IP_VS_SO_SET_EDITDEST:
2402 ret = ip_vs_edit_dest(svc, &udest);
2404 case IP_VS_SO_SET_DELDEST:
2405 ret = ip_vs_del_dest(svc, &udest);
2412 mutex_unlock(&__ip_vs_mutex);
2414 /* decrease the module use count */
2415 ip_vs_use_count_dec();
2422 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2424 dst->protocol = src->protocol;
2425 dst->addr = src->addr.ip;
2426 dst->port = src->port;
2427 dst->fwmark = src->fwmark;
2428 strlcpy(dst->sched_name, src->scheduler->name, sizeof(dst->sched_name));
2429 dst->flags = src->flags;
2430 dst->timeout = src->timeout / HZ;
2431 dst->netmask = src->netmask;
2432 dst->num_dests = src->num_dests;
2433 ip_vs_copy_stats(&dst->stats, &src->stats);
2437 __ip_vs_get_service_entries(struct net *net,
2438 const struct ip_vs_get_services *get,
2439 struct ip_vs_get_services __user *uptr)
2442 struct ip_vs_service *svc;
2443 struct ip_vs_service_entry entry;
2446 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2447 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2448 /* Only expose IPv4 entries to old interface */
2449 if (svc->af != AF_INET || !net_eq(svc->net, net))
2452 if (count >= get->num_services)
2454 memset(&entry, 0, sizeof(entry));
2455 ip_vs_copy_service(&entry, svc);
2456 if (copy_to_user(&uptr->entrytable[count],
2457 &entry, sizeof(entry))) {
2465 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2466 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2467 /* Only expose IPv4 entries to old interface */
2468 if (svc->af != AF_INET || !net_eq(svc->net, net))
2471 if (count >= get->num_services)
2473 memset(&entry, 0, sizeof(entry));
2474 ip_vs_copy_service(&entry, svc);
2475 if (copy_to_user(&uptr->entrytable[count],
2476 &entry, sizeof(entry))) {
2488 __ip_vs_get_dest_entries(struct net *net, const struct ip_vs_get_dests *get,
2489 struct ip_vs_get_dests __user *uptr)
2491 struct ip_vs_service *svc;
2492 union nf_inet_addr addr = { .ip = get->addr };
2496 svc = __ip_vs_svc_fwm_find(net, AF_INET, get->fwmark);
2498 svc = __ip_vs_service_find(net, AF_INET, get->protocol, &addr,
2503 struct ip_vs_dest *dest;
2504 struct ip_vs_dest_entry entry;
2506 list_for_each_entry(dest, &svc->destinations, n_list) {
2507 if (count >= get->num_dests)
2510 entry.addr = dest->addr.ip;
2511 entry.port = dest->port;
2512 entry.conn_flags = atomic_read(&dest->conn_flags);
2513 entry.weight = atomic_read(&dest->weight);
2514 entry.u_threshold = dest->u_threshold;
2515 entry.l_threshold = dest->l_threshold;
2516 entry.activeconns = atomic_read(&dest->activeconns);
2517 entry.inactconns = atomic_read(&dest->inactconns);
2518 entry.persistconns = atomic_read(&dest->persistconns);
2519 ip_vs_copy_stats(&entry.stats, &dest->stats);
2520 if (copy_to_user(&uptr->entrytable[count],
2521 &entry, sizeof(entry))) {
2533 __ip_vs_get_timeouts(struct net *net, struct ip_vs_timeout_user *u)
2535 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2536 struct ip_vs_proto_data *pd;
2539 #ifdef CONFIG_IP_VS_PROTO_TCP
2540 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2541 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2542 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2544 #ifdef CONFIG_IP_VS_PROTO_UDP
2545 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2547 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2552 #define GET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2553 #define GET_INFO_ARG_LEN (sizeof(struct ip_vs_getinfo))
2554 #define GET_SERVICES_ARG_LEN (sizeof(struct ip_vs_get_services))
2555 #define GET_SERVICE_ARG_LEN (sizeof(struct ip_vs_service_entry))
2556 #define GET_DESTS_ARG_LEN (sizeof(struct ip_vs_get_dests))
2557 #define GET_TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2558 #define GET_DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user) * 2)
2560 static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
2561 [GET_CMDID(IP_VS_SO_GET_VERSION)] = 64,
2562 [GET_CMDID(IP_VS_SO_GET_INFO)] = GET_INFO_ARG_LEN,
2563 [GET_CMDID(IP_VS_SO_GET_SERVICES)] = GET_SERVICES_ARG_LEN,
2564 [GET_CMDID(IP_VS_SO_GET_SERVICE)] = GET_SERVICE_ARG_LEN,
2565 [GET_CMDID(IP_VS_SO_GET_DESTS)] = GET_DESTS_ARG_LEN,
2566 [GET_CMDID(IP_VS_SO_GET_TIMEOUT)] = GET_TIMEOUT_ARG_LEN,
2567 [GET_CMDID(IP_VS_SO_GET_DAEMON)] = GET_DAEMON_ARG_LEN,
2571 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2573 unsigned char arg[128];
2575 unsigned int copylen;
2576 struct net *net = sock_net(sk);
2577 struct netns_ipvs *ipvs = net_ipvs(net);
2580 if (!capable(CAP_NET_ADMIN))
2583 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2586 if (*len < get_arglen[GET_CMDID(cmd)]) {
2587 pr_err("get_ctl: len %u < %u\n",
2588 *len, get_arglen[GET_CMDID(cmd)]);
2592 copylen = get_arglen[GET_CMDID(cmd)];
2596 if (copy_from_user(arg, user, copylen) != 0)
2599 * Handle daemons first since it has its own locking
2601 if (cmd == IP_VS_SO_GET_DAEMON) {
2602 struct ip_vs_daemon_user d[2];
2604 memset(&d, 0, sizeof(d));
2605 if (mutex_lock_interruptible(&ipvs->sync_mutex))
2606 return -ERESTARTSYS;
2608 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2609 d[0].state = IP_VS_STATE_MASTER;
2610 strlcpy(d[0].mcast_ifn, ipvs->master_mcast_ifn,
2611 sizeof(d[0].mcast_ifn));
2612 d[0].syncid = ipvs->master_syncid;
2614 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2615 d[1].state = IP_VS_STATE_BACKUP;
2616 strlcpy(d[1].mcast_ifn, ipvs->backup_mcast_ifn,
2617 sizeof(d[1].mcast_ifn));
2618 d[1].syncid = ipvs->backup_syncid;
2620 if (copy_to_user(user, &d, sizeof(d)) != 0)
2622 mutex_unlock(&ipvs->sync_mutex);
2626 if (mutex_lock_interruptible(&__ip_vs_mutex))
2627 return -ERESTARTSYS;
2630 case IP_VS_SO_GET_VERSION:
2634 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2635 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2636 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2640 *len = strlen(buf)+1;
2644 case IP_VS_SO_GET_INFO:
2646 struct ip_vs_getinfo info;
2647 info.version = IP_VS_VERSION_CODE;
2648 info.size = ip_vs_conn_tab_size;
2649 info.num_services = ipvs->num_services;
2650 if (copy_to_user(user, &info, sizeof(info)) != 0)
2655 case IP_VS_SO_GET_SERVICES:
2657 struct ip_vs_get_services *get;
2660 get = (struct ip_vs_get_services *)arg;
2661 size = sizeof(*get) +
2662 sizeof(struct ip_vs_service_entry) * get->num_services;
2664 pr_err("length: %u != %u\n", *len, size);
2668 ret = __ip_vs_get_service_entries(net, get, user);
2672 case IP_VS_SO_GET_SERVICE:
2674 struct ip_vs_service_entry *entry;
2675 struct ip_vs_service *svc;
2676 union nf_inet_addr addr;
2678 entry = (struct ip_vs_service_entry *)arg;
2679 addr.ip = entry->addr;
2681 svc = __ip_vs_svc_fwm_find(net, AF_INET, entry->fwmark);
2683 svc = __ip_vs_service_find(net, AF_INET,
2684 entry->protocol, &addr,
2687 ip_vs_copy_service(entry, svc);
2688 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2695 case IP_VS_SO_GET_DESTS:
2697 struct ip_vs_get_dests *get;
2700 get = (struct ip_vs_get_dests *)arg;
2701 size = sizeof(*get) +
2702 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2704 pr_err("length: %u != %u\n", *len, size);
2708 ret = __ip_vs_get_dest_entries(net, get, user);
2712 case IP_VS_SO_GET_TIMEOUT:
2714 struct ip_vs_timeout_user t;
2716 memset(&t, 0, sizeof(t));
2717 __ip_vs_get_timeouts(net, &t);
2718 if (copy_to_user(user, &t, sizeof(t)) != 0)
2728 mutex_unlock(&__ip_vs_mutex);
2733 static struct nf_sockopt_ops ip_vs_sockopts = {
2735 .set_optmin = IP_VS_BASE_CTL,
2736 .set_optmax = IP_VS_SO_SET_MAX+1,
2737 .set = do_ip_vs_set_ctl,
2738 .get_optmin = IP_VS_BASE_CTL,
2739 .get_optmax = IP_VS_SO_GET_MAX+1,
2740 .get = do_ip_vs_get_ctl,
2741 .owner = THIS_MODULE,
2745 * Generic Netlink interface
2748 /* IPVS genetlink family */
2749 static struct genl_family ip_vs_genl_family = {
2750 .id = GENL_ID_GENERATE,
2752 .name = IPVS_GENL_NAME,
2753 .version = IPVS_GENL_VERSION,
2754 .maxattr = IPVS_CMD_MAX,
2755 .netnsok = true, /* Make ipvsadm to work on netns */
2758 /* Policy used for first-level command attributes */
2759 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2760 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2761 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2762 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2763 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2764 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2765 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2768 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2769 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2770 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2771 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2772 .len = IP_VS_IFNAME_MAXLEN },
2773 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2776 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2777 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2778 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2779 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2780 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2781 .len = sizeof(union nf_inet_addr) },
2782 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2783 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2784 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2785 .len = IP_VS_SCHEDNAME_MAXLEN },
2786 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2787 .len = IP_VS_PENAME_MAXLEN },
2788 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2789 .len = sizeof(struct ip_vs_flags) },
2790 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2791 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2792 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2795 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2796 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2797 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2798 .len = sizeof(union nf_inet_addr) },
2799 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2800 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2801 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2802 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2803 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2804 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2805 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2806 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2807 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2810 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2811 struct ip_vs_stats *stats)
2813 struct ip_vs_stats_user ustats;
2814 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2818 ip_vs_copy_stats(&ustats, stats);
2820 NLA_PUT_U32(skb, IPVS_STATS_ATTR_CONNS, ustats.conns);
2821 NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPKTS, ustats.inpkts);
2822 NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPKTS, ustats.outpkts);
2823 NLA_PUT_U64(skb, IPVS_STATS_ATTR_INBYTES, ustats.inbytes);
2824 NLA_PUT_U64(skb, IPVS_STATS_ATTR_OUTBYTES, ustats.outbytes);
2825 NLA_PUT_U32(skb, IPVS_STATS_ATTR_CPS, ustats.cps);
2826 NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPPS, ustats.inpps);
2827 NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPPS, ustats.outpps);
2828 NLA_PUT_U32(skb, IPVS_STATS_ATTR_INBPS, ustats.inbps);
2829 NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTBPS, ustats.outbps);
2831 nla_nest_end(skb, nl_stats);
2836 nla_nest_cancel(skb, nl_stats);
2840 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2841 struct ip_vs_service *svc)
2843 struct nlattr *nl_service;
2844 struct ip_vs_flags flags = { .flags = svc->flags,
2847 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2851 NLA_PUT_U16(skb, IPVS_SVC_ATTR_AF, svc->af);
2854 NLA_PUT_U32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark);
2856 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol);
2857 NLA_PUT(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr);
2858 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PORT, svc->port);
2861 NLA_PUT_STRING(skb, IPVS_SVC_ATTR_SCHED_NAME, svc->scheduler->name);
2863 NLA_PUT_STRING(skb, IPVS_SVC_ATTR_PE_NAME, svc->pe->name);
2864 NLA_PUT(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags);
2865 NLA_PUT_U32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ);
2866 NLA_PUT_U32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask);
2868 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2869 goto nla_put_failure;
2871 nla_nest_end(skb, nl_service);
2876 nla_nest_cancel(skb, nl_service);
2880 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2881 struct ip_vs_service *svc,
2882 struct netlink_callback *cb)
2886 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2887 &ip_vs_genl_family, NLM_F_MULTI,
2888 IPVS_CMD_NEW_SERVICE);
2892 if (ip_vs_genl_fill_service(skb, svc) < 0)
2893 goto nla_put_failure;
2895 return genlmsg_end(skb, hdr);
2898 genlmsg_cancel(skb, hdr);
2902 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2903 struct netlink_callback *cb)
2906 int start = cb->args[0];
2907 struct ip_vs_service *svc;
2908 struct net *net = skb_sknet(skb);
2910 mutex_lock(&__ip_vs_mutex);
2911 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2912 list_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2913 if (++idx <= start || !net_eq(svc->net, net))
2915 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2917 goto nla_put_failure;
2922 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2923 list_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
2924 if (++idx <= start || !net_eq(svc->net, net))
2926 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2928 goto nla_put_failure;
2934 mutex_unlock(&__ip_vs_mutex);
2940 static int ip_vs_genl_parse_service(struct net *net,
2941 struct ip_vs_service_user_kern *usvc,
2942 struct nlattr *nla, int full_entry,
2943 struct ip_vs_service **ret_svc)
2945 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
2946 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
2947 struct ip_vs_service *svc;
2949 /* Parse mandatory identifying service fields first */
2951 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
2954 nla_af = attrs[IPVS_SVC_ATTR_AF];
2955 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
2956 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
2957 nla_port = attrs[IPVS_SVC_ATTR_PORT];
2958 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
2960 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
2963 memset(usvc, 0, sizeof(*usvc));
2965 usvc->af = nla_get_u16(nla_af);
2966 #ifdef CONFIG_IP_VS_IPV6
2967 if (usvc->af != AF_INET && usvc->af != AF_INET6)
2969 if (usvc->af != AF_INET)
2971 return -EAFNOSUPPORT;
2974 usvc->protocol = IPPROTO_TCP;
2975 usvc->fwmark = nla_get_u32(nla_fwmark);
2977 usvc->protocol = nla_get_u16(nla_protocol);
2978 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
2979 usvc->port = nla_get_u16(nla_port);
2984 svc = __ip_vs_svc_fwm_find(net, usvc->af, usvc->fwmark);
2986 svc = __ip_vs_service_find(net, usvc->af, usvc->protocol,
2987 &usvc->addr, usvc->port);
2990 /* If a full entry was requested, check for the additional fields */
2992 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
2994 struct ip_vs_flags flags;
2996 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
2997 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
2998 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
2999 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3000 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3002 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3005 nla_memcpy(&flags, nla_flags, sizeof(flags));
3007 /* prefill flags from service if it already exists */
3009 usvc->flags = svc->flags;
3011 /* set new flags from userland */
3012 usvc->flags = (usvc->flags & ~flags.mask) |
3013 (flags.flags & flags.mask);
3014 usvc->sched_name = nla_data(nla_sched);
3015 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3016 usvc->timeout = nla_get_u32(nla_timeout);
3017 usvc->netmask = nla_get_u32(nla_netmask);
3023 static struct ip_vs_service *ip_vs_genl_find_service(struct net *net,
3026 struct ip_vs_service_user_kern usvc;
3027 struct ip_vs_service *svc;
3030 ret = ip_vs_genl_parse_service(net, &usvc, nla, 0, &svc);
3031 return ret ? ERR_PTR(ret) : svc;
3034 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3036 struct nlattr *nl_dest;
3038 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3042 NLA_PUT(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr);
3043 NLA_PUT_U16(skb, IPVS_DEST_ATTR_PORT, dest->port);
3045 NLA_PUT_U32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3046 atomic_read(&dest->conn_flags) & IP_VS_CONN_F_FWD_MASK);
3047 NLA_PUT_U32(skb, IPVS_DEST_ATTR_WEIGHT, atomic_read(&dest->weight));
3048 NLA_PUT_U32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold);
3049 NLA_PUT_U32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold);
3050 NLA_PUT_U32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3051 atomic_read(&dest->activeconns));
3052 NLA_PUT_U32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3053 atomic_read(&dest->inactconns));
3054 NLA_PUT_U32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3055 atomic_read(&dest->persistconns));
3057 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
3058 goto nla_put_failure;
3060 nla_nest_end(skb, nl_dest);
3065 nla_nest_cancel(skb, nl_dest);
3069 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3070 struct netlink_callback *cb)
3074 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
3075 &ip_vs_genl_family, NLM_F_MULTI,
3080 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3081 goto nla_put_failure;
3083 return genlmsg_end(skb, hdr);
3086 genlmsg_cancel(skb, hdr);
3090 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3091 struct netlink_callback *cb)
3094 int start = cb->args[0];
3095 struct ip_vs_service *svc;
3096 struct ip_vs_dest *dest;
3097 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3098 struct net *net = skb_sknet(skb);
3100 mutex_lock(&__ip_vs_mutex);
3102 /* Try to find the service for which to dump destinations */
3103 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3104 IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3108 svc = ip_vs_genl_find_service(net, attrs[IPVS_CMD_ATTR_SERVICE]);
3109 if (IS_ERR(svc) || svc == NULL)
3112 /* Dump the destinations */
3113 list_for_each_entry(dest, &svc->destinations, n_list) {
3116 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3118 goto nla_put_failure;
3126 mutex_unlock(&__ip_vs_mutex);
3131 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3132 struct nlattr *nla, int full_entry)
3134 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3135 struct nlattr *nla_addr, *nla_port;
3137 /* Parse mandatory identifying destination fields first */
3139 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3142 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3143 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3145 if (!(nla_addr && nla_port))
3148 memset(udest, 0, sizeof(*udest));
3150 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3151 udest->port = nla_get_u16(nla_port);
3153 /* If a full entry was requested, check for the additional fields */
3155 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3158 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3159 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3160 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3161 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3163 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3166 udest->conn_flags = nla_get_u32(nla_fwd)
3167 & IP_VS_CONN_F_FWD_MASK;
3168 udest->weight = nla_get_u32(nla_weight);
3169 udest->u_threshold = nla_get_u32(nla_u_thresh);
3170 udest->l_threshold = nla_get_u32(nla_l_thresh);
3176 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __be32 state,
3177 const char *mcast_ifn, __be32 syncid)
3179 struct nlattr *nl_daemon;
3181 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3185 NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_STATE, state);
3186 NLA_PUT_STRING(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn);
3187 NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid);
3189 nla_nest_end(skb, nl_daemon);
3194 nla_nest_cancel(skb, nl_daemon);
3198 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __be32 state,
3199 const char *mcast_ifn, __be32 syncid,
3200 struct netlink_callback *cb)
3203 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
3204 &ip_vs_genl_family, NLM_F_MULTI,
3205 IPVS_CMD_NEW_DAEMON);
3209 if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
3210 goto nla_put_failure;
3212 return genlmsg_end(skb, hdr);
3215 genlmsg_cancel(skb, hdr);
3219 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3220 struct netlink_callback *cb)
3222 struct net *net = skb_sknet(skb);
3223 struct netns_ipvs *ipvs = net_ipvs(net);
3225 mutex_lock(&ipvs->sync_mutex);
3226 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3227 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3228 ipvs->master_mcast_ifn,
3229 ipvs->master_syncid, cb) < 0)
3230 goto nla_put_failure;
3235 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3236 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3237 ipvs->backup_mcast_ifn,
3238 ipvs->backup_syncid, cb) < 0)
3239 goto nla_put_failure;
3245 mutex_unlock(&ipvs->sync_mutex);
3250 static int ip_vs_genl_new_daemon(struct net *net, struct nlattr **attrs)
3252 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3253 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3254 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3257 return start_sync_thread(net,
3258 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3259 nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3260 nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3263 static int ip_vs_genl_del_daemon(struct net *net, struct nlattr **attrs)
3265 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3268 return stop_sync_thread(net,
3269 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3272 static int ip_vs_genl_set_config(struct net *net, struct nlattr **attrs)
3274 struct ip_vs_timeout_user t;
3276 __ip_vs_get_timeouts(net, &t);
3278 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3279 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3281 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3283 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3285 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3286 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3288 return ip_vs_set_timeout(net, &t);
3291 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3295 struct netns_ipvs *ipvs;
3297 net = skb_sknet(skb);
3298 ipvs = net_ipvs(net);
3299 cmd = info->genlhdr->cmd;
3301 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3302 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3304 mutex_lock(&ipvs->sync_mutex);
3305 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3306 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3307 info->attrs[IPVS_CMD_ATTR_DAEMON],
3308 ip_vs_daemon_policy)) {
3313 if (cmd == IPVS_CMD_NEW_DAEMON)
3314 ret = ip_vs_genl_new_daemon(net, daemon_attrs);
3316 ret = ip_vs_genl_del_daemon(net, daemon_attrs);
3318 mutex_unlock(&ipvs->sync_mutex);
3323 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3325 struct ip_vs_service *svc = NULL;
3326 struct ip_vs_service_user_kern usvc;
3327 struct ip_vs_dest_user_kern udest;
3329 int need_full_svc = 0, need_full_dest = 0;
3332 net = skb_sknet(skb);
3333 cmd = info->genlhdr->cmd;
3335 mutex_lock(&__ip_vs_mutex);
3337 if (cmd == IPVS_CMD_FLUSH) {
3338 ret = ip_vs_flush(net);
3340 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3341 ret = ip_vs_genl_set_config(net, info->attrs);
3343 } else if (cmd == IPVS_CMD_ZERO &&
3344 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3345 ret = ip_vs_zero_all(net);
3349 /* All following commands require a service argument, so check if we
3350 * received a valid one. We need a full service specification when
3351 * adding / editing a service. Only identifying members otherwise. */
3352 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3355 ret = ip_vs_genl_parse_service(net, &usvc,
3356 info->attrs[IPVS_CMD_ATTR_SERVICE],
3357 need_full_svc, &svc);
3361 /* Unless we're adding a new service, the service must already exist */
3362 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3367 /* Destination commands require a valid destination argument. For
3368 * adding / editing a destination, we need a full destination
3370 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3371 cmd == IPVS_CMD_DEL_DEST) {
3372 if (cmd != IPVS_CMD_DEL_DEST)
3375 ret = ip_vs_genl_parse_dest(&udest,
3376 info->attrs[IPVS_CMD_ATTR_DEST],
3383 case IPVS_CMD_NEW_SERVICE:
3385 ret = ip_vs_add_service(net, &usvc, &svc);
3389 case IPVS_CMD_SET_SERVICE:
3390 ret = ip_vs_edit_service(svc, &usvc);
3392 case IPVS_CMD_DEL_SERVICE:
3393 ret = ip_vs_del_service(svc);
3394 /* do not use svc, it can be freed */
3396 case IPVS_CMD_NEW_DEST:
3397 ret = ip_vs_add_dest(svc, &udest);
3399 case IPVS_CMD_SET_DEST:
3400 ret = ip_vs_edit_dest(svc, &udest);
3402 case IPVS_CMD_DEL_DEST:
3403 ret = ip_vs_del_dest(svc, &udest);
3406 ret = ip_vs_zero_service(svc);
3413 mutex_unlock(&__ip_vs_mutex);
3418 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3420 struct sk_buff *msg;
3422 int ret, cmd, reply_cmd;
3425 net = skb_sknet(skb);
3426 cmd = info->genlhdr->cmd;
3428 if (cmd == IPVS_CMD_GET_SERVICE)
3429 reply_cmd = IPVS_CMD_NEW_SERVICE;
3430 else if (cmd == IPVS_CMD_GET_INFO)
3431 reply_cmd = IPVS_CMD_SET_INFO;
3432 else if (cmd == IPVS_CMD_GET_CONFIG)
3433 reply_cmd = IPVS_CMD_SET_CONFIG;
3435 pr_err("unknown Generic Netlink command\n");
3439 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3443 mutex_lock(&__ip_vs_mutex);
3445 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3447 goto nla_put_failure;
3450 case IPVS_CMD_GET_SERVICE:
3452 struct ip_vs_service *svc;
3454 svc = ip_vs_genl_find_service(net,
3455 info->attrs[IPVS_CMD_ATTR_SERVICE]);
3460 ret = ip_vs_genl_fill_service(msg, svc);
3462 goto nla_put_failure;
3471 case IPVS_CMD_GET_CONFIG:
3473 struct ip_vs_timeout_user t;
3475 __ip_vs_get_timeouts(net, &t);
3476 #ifdef CONFIG_IP_VS_PROTO_TCP
3477 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP, t.tcp_timeout);
3478 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3481 #ifdef CONFIG_IP_VS_PROTO_UDP
3482 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout);
3488 case IPVS_CMD_GET_INFO:
3489 NLA_PUT_U32(msg, IPVS_INFO_ATTR_VERSION, IP_VS_VERSION_CODE);
3490 NLA_PUT_U32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3491 ip_vs_conn_tab_size);
3495 genlmsg_end(msg, reply);
3496 ret = genlmsg_reply(msg, info);
3500 pr_err("not enough space in Netlink message\n");
3506 mutex_unlock(&__ip_vs_mutex);
3512 static struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3514 .cmd = IPVS_CMD_NEW_SERVICE,
3515 .flags = GENL_ADMIN_PERM,
3516 .policy = ip_vs_cmd_policy,
3517 .doit = ip_vs_genl_set_cmd,
3520 .cmd = IPVS_CMD_SET_SERVICE,
3521 .flags = GENL_ADMIN_PERM,
3522 .policy = ip_vs_cmd_policy,
3523 .doit = ip_vs_genl_set_cmd,
3526 .cmd = IPVS_CMD_DEL_SERVICE,
3527 .flags = GENL_ADMIN_PERM,
3528 .policy = ip_vs_cmd_policy,
3529 .doit = ip_vs_genl_set_cmd,
3532 .cmd = IPVS_CMD_GET_SERVICE,
3533 .flags = GENL_ADMIN_PERM,
3534 .doit = ip_vs_genl_get_cmd,
3535 .dumpit = ip_vs_genl_dump_services,
3536 .policy = ip_vs_cmd_policy,
3539 .cmd = IPVS_CMD_NEW_DEST,
3540 .flags = GENL_ADMIN_PERM,
3541 .policy = ip_vs_cmd_policy,
3542 .doit = ip_vs_genl_set_cmd,
3545 .cmd = IPVS_CMD_SET_DEST,
3546 .flags = GENL_ADMIN_PERM,
3547 .policy = ip_vs_cmd_policy,
3548 .doit = ip_vs_genl_set_cmd,
3551 .cmd = IPVS_CMD_DEL_DEST,
3552 .flags = GENL_ADMIN_PERM,
3553 .policy = ip_vs_cmd_policy,
3554 .doit = ip_vs_genl_set_cmd,
3557 .cmd = IPVS_CMD_GET_DEST,
3558 .flags = GENL_ADMIN_PERM,
3559 .policy = ip_vs_cmd_policy,
3560 .dumpit = ip_vs_genl_dump_dests,
3563 .cmd = IPVS_CMD_NEW_DAEMON,
3564 .flags = GENL_ADMIN_PERM,
3565 .policy = ip_vs_cmd_policy,
3566 .doit = ip_vs_genl_set_daemon,
3569 .cmd = IPVS_CMD_DEL_DAEMON,
3570 .flags = GENL_ADMIN_PERM,
3571 .policy = ip_vs_cmd_policy,
3572 .doit = ip_vs_genl_set_daemon,
3575 .cmd = IPVS_CMD_GET_DAEMON,
3576 .flags = GENL_ADMIN_PERM,
3577 .dumpit = ip_vs_genl_dump_daemons,
3580 .cmd = IPVS_CMD_SET_CONFIG,
3581 .flags = GENL_ADMIN_PERM,
3582 .policy = ip_vs_cmd_policy,
3583 .doit = ip_vs_genl_set_cmd,
3586 .cmd = IPVS_CMD_GET_CONFIG,
3587 .flags = GENL_ADMIN_PERM,
3588 .doit = ip_vs_genl_get_cmd,
3591 .cmd = IPVS_CMD_GET_INFO,
3592 .flags = GENL_ADMIN_PERM,
3593 .doit = ip_vs_genl_get_cmd,
3596 .cmd = IPVS_CMD_ZERO,
3597 .flags = GENL_ADMIN_PERM,
3598 .policy = ip_vs_cmd_policy,
3599 .doit = ip_vs_genl_set_cmd,
3602 .cmd = IPVS_CMD_FLUSH,
3603 .flags = GENL_ADMIN_PERM,
3604 .doit = ip_vs_genl_set_cmd,
3608 static int __init ip_vs_genl_register(void)
3610 return genl_register_family_with_ops(&ip_vs_genl_family,
3611 ip_vs_genl_ops, ARRAY_SIZE(ip_vs_genl_ops));
3614 static void ip_vs_genl_unregister(void)
3616 genl_unregister_family(&ip_vs_genl_family);
3619 /* End of Generic Netlink interface definitions */
3622 * per netns intit/exit func.
3624 #ifdef CONFIG_SYSCTL
3625 int __net_init ip_vs_control_net_init_sysctl(struct net *net)
3628 struct netns_ipvs *ipvs = net_ipvs(net);
3629 struct ctl_table *tbl;
3631 atomic_set(&ipvs->dropentry, 0);
3632 spin_lock_init(&ipvs->dropentry_lock);
3633 spin_lock_init(&ipvs->droppacket_lock);
3634 spin_lock_init(&ipvs->securetcp_lock);
3636 if (!net_eq(net, &init_net)) {
3637 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3642 /* Initialize sysctl defaults */
3644 ipvs->sysctl_amemthresh = 1024;
3645 tbl[idx++].data = &ipvs->sysctl_amemthresh;
3646 ipvs->sysctl_am_droprate = 10;
3647 tbl[idx++].data = &ipvs->sysctl_am_droprate;
3648 tbl[idx++].data = &ipvs->sysctl_drop_entry;
3649 tbl[idx++].data = &ipvs->sysctl_drop_packet;
3650 #ifdef CONFIG_IP_VS_NFCT
3651 tbl[idx++].data = &ipvs->sysctl_conntrack;
3653 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3654 ipvs->sysctl_snat_reroute = 1;
3655 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3656 ipvs->sysctl_sync_ver = 1;
3657 tbl[idx++].data = &ipvs->sysctl_sync_ver;
3658 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3659 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3660 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3661 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3662 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3663 tbl[idx].data = &ipvs->sysctl_sync_threshold;
3664 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3665 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3668 ipvs->sysctl_hdr = register_net_sysctl_table(net, net_vs_ctl_path,
3670 if (ipvs->sysctl_hdr == NULL) {
3671 if (!net_eq(net, &init_net))
3675 ip_vs_start_estimator(net, &ipvs->tot_stats);
3676 ipvs->sysctl_tbl = tbl;
3677 /* Schedule defense work */
3678 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3679 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3684 void __net_init ip_vs_control_net_cleanup_sysctl(struct net *net)
3686 struct netns_ipvs *ipvs = net_ipvs(net);
3688 cancel_delayed_work_sync(&ipvs->defense_work);
3689 cancel_work_sync(&ipvs->defense_work.work);
3690 unregister_net_sysctl_table(ipvs->sysctl_hdr);
3695 int __net_init ip_vs_control_net_init_sysctl(struct net *net) { return 0; }
3696 void __net_init ip_vs_control_net_cleanup_sysctl(struct net *net) { }
3700 static struct notifier_block ip_vs_dst_notifier = {
3701 .notifier_call = ip_vs_dst_event,
3704 int __net_init ip_vs_control_net_init(struct net *net)
3707 struct netns_ipvs *ipvs = net_ipvs(net);
3709 rwlock_init(&ipvs->rs_lock);
3711 /* Initialize rs_table */
3712 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
3713 INIT_LIST_HEAD(&ipvs->rs_table[idx]);
3715 INIT_LIST_HEAD(&ipvs->dest_trash);
3716 atomic_set(&ipvs->ftpsvc_counter, 0);
3717 atomic_set(&ipvs->nullsvc_counter, 0);
3720 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
3721 if (!ipvs->tot_stats.cpustats)
3724 spin_lock_init(&ipvs->tot_stats.lock);
3726 proc_net_fops_create(net, "ip_vs", 0, &ip_vs_info_fops);
3727 proc_net_fops_create(net, "ip_vs_stats", 0, &ip_vs_stats_fops);
3728 proc_net_fops_create(net, "ip_vs_stats_percpu", 0,
3729 &ip_vs_stats_percpu_fops);
3731 if (ip_vs_control_net_init_sysctl(net))
3737 free_percpu(ipvs->tot_stats.cpustats);
3741 void __net_exit ip_vs_control_net_cleanup(struct net *net)
3743 struct netns_ipvs *ipvs = net_ipvs(net);
3745 ip_vs_trash_cleanup(net);
3746 ip_vs_stop_estimator(net, &ipvs->tot_stats);
3747 ip_vs_control_net_cleanup_sysctl(net);
3748 proc_net_remove(net, "ip_vs_stats_percpu");
3749 proc_net_remove(net, "ip_vs_stats");
3750 proc_net_remove(net, "ip_vs");
3751 free_percpu(ipvs->tot_stats.cpustats);
3754 int __init ip_vs_control_init(void)
3761 /* Initialize svc_table, ip_vs_svc_fwm_table, rs_table */
3762 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
3763 INIT_LIST_HEAD(&ip_vs_svc_table[idx]);
3764 INIT_LIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3767 smp_wmb(); /* Do we really need it now ? */
3769 ret = nf_register_sockopt(&ip_vs_sockopts);
3771 pr_err("cannot register sockopt.\n");
3775 ret = ip_vs_genl_register();
3777 pr_err("cannot register Generic Netlink interface.\n");
3781 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
3789 ip_vs_genl_unregister();
3791 nf_unregister_sockopt(&ip_vs_sockopts);
3797 void ip_vs_control_cleanup(void)
3800 unregister_netdevice_notifier(&ip_vs_dst_notifier);
3801 ip_vs_genl_unregister();
3802 nf_unregister_sockopt(&ip_vs_sockopts);