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 /* sysctl variables */
60 #ifdef CONFIG_IP_VS_DEBUG
61 static int sysctl_ip_vs_debug_level = 0;
63 int ip_vs_get_debug_level(void)
65 return sysctl_ip_vs_debug_level;
71 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup);
74 #ifdef CONFIG_IP_VS_IPV6
75 /* Taken from rt6_fill_node() in net/ipv6/route.c, is there a better way? */
76 static bool __ip_vs_addr_is_local_v6(struct net *net,
77 const struct in6_addr *addr)
82 struct dst_entry *dst = ip6_route_output(net, NULL, &fl6);
85 is_local = !dst->error && dst->dev && (dst->dev->flags & IFF_LOOPBACK);
94 * update_defense_level is called from keventd and from sysctl,
95 * so it needs to protect itself from softirqs
97 static void update_defense_level(struct netns_ipvs *ipvs)
100 static int old_secure_tcp = 0;
105 /* we only count free and buffered memory (in pages) */
107 availmem = i.freeram + i.bufferram;
108 /* however in linux 2.5 the i.bufferram is total page cache size,
110 /* si_swapinfo(&i); */
111 /* availmem = availmem - (i.totalswap - i.freeswap); */
113 nomem = (availmem < ipvs->sysctl_amemthresh);
118 spin_lock(&ipvs->dropentry_lock);
119 switch (ipvs->sysctl_drop_entry) {
121 atomic_set(&ipvs->dropentry, 0);
125 atomic_set(&ipvs->dropentry, 1);
126 ipvs->sysctl_drop_entry = 2;
128 atomic_set(&ipvs->dropentry, 0);
133 atomic_set(&ipvs->dropentry, 1);
135 atomic_set(&ipvs->dropentry, 0);
136 ipvs->sysctl_drop_entry = 1;
140 atomic_set(&ipvs->dropentry, 1);
143 spin_unlock(&ipvs->dropentry_lock);
146 spin_lock(&ipvs->droppacket_lock);
147 switch (ipvs->sysctl_drop_packet) {
153 ipvs->drop_rate = ipvs->drop_counter
154 = ipvs->sysctl_amemthresh /
155 (ipvs->sysctl_amemthresh-availmem);
156 ipvs->sysctl_drop_packet = 2;
163 ipvs->drop_rate = ipvs->drop_counter
164 = ipvs->sysctl_amemthresh /
165 (ipvs->sysctl_amemthresh-availmem);
168 ipvs->sysctl_drop_packet = 1;
172 ipvs->drop_rate = ipvs->sysctl_am_droprate;
175 spin_unlock(&ipvs->droppacket_lock);
178 spin_lock(&ipvs->securetcp_lock);
179 switch (ipvs->sysctl_secure_tcp) {
181 if (old_secure_tcp >= 2)
186 if (old_secure_tcp < 2)
188 ipvs->sysctl_secure_tcp = 2;
190 if (old_secure_tcp >= 2)
196 if (old_secure_tcp < 2)
199 if (old_secure_tcp >= 2)
201 ipvs->sysctl_secure_tcp = 1;
205 if (old_secure_tcp < 2)
209 old_secure_tcp = ipvs->sysctl_secure_tcp;
211 ip_vs_protocol_timeout_change(ipvs,
212 ipvs->sysctl_secure_tcp > 1);
213 spin_unlock(&ipvs->securetcp_lock);
220 * Timer for checking the defense
222 #define DEFENSE_TIMER_PERIOD 1*HZ
224 static void defense_work_handler(struct work_struct *work)
226 struct netns_ipvs *ipvs =
227 container_of(work, struct netns_ipvs, defense_work.work);
229 update_defense_level(ipvs);
230 if (atomic_read(&ipvs->dropentry))
231 ip_vs_random_dropentry(ipvs->net);
232 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
237 ip_vs_use_count_inc(void)
239 return try_module_get(THIS_MODULE);
243 ip_vs_use_count_dec(void)
245 module_put(THIS_MODULE);
250 * Hash table: for virtual service lookups
252 #define IP_VS_SVC_TAB_BITS 8
253 #define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
254 #define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)
256 /* the service table hashed by <protocol, addr, port> */
257 static struct hlist_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE];
258 /* the service table hashed by fwmark */
259 static struct hlist_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE];
263 * Returns hash value for virtual service
265 static inline unsigned int
266 ip_vs_svc_hashkey(struct net *net, int af, unsigned int proto,
267 const union nf_inet_addr *addr, __be16 port)
269 register unsigned int porth = ntohs(port);
270 __be32 addr_fold = addr->ip;
273 #ifdef CONFIG_IP_VS_IPV6
275 addr_fold = addr->ip6[0]^addr->ip6[1]^
276 addr->ip6[2]^addr->ip6[3];
278 ahash = ntohl(addr_fold);
279 ahash ^= ((size_t) net >> 8);
281 return (proto ^ ahash ^ (porth >> IP_VS_SVC_TAB_BITS) ^ porth) &
286 * Returns hash value of fwmark for virtual service lookup
288 static inline unsigned int ip_vs_svc_fwm_hashkey(struct net *net, __u32 fwmark)
290 return (((size_t)net>>8) ^ fwmark) & IP_VS_SVC_TAB_MASK;
294 * Hashes a service in the ip_vs_svc_table by <netns,proto,addr,port>
295 * or in the ip_vs_svc_fwm_table by fwmark.
296 * Should be called with locked tables.
298 static int ip_vs_svc_hash(struct ip_vs_service *svc)
302 if (svc->flags & IP_VS_SVC_F_HASHED) {
303 pr_err("%s(): request for already hashed, called from %pF\n",
304 __func__, __builtin_return_address(0));
308 if (svc->fwmark == 0) {
310 * Hash it by <netns,protocol,addr,port> in ip_vs_svc_table
312 hash = ip_vs_svc_hashkey(svc->net, svc->af, svc->protocol,
313 &svc->addr, svc->port);
314 hlist_add_head_rcu(&svc->s_list, &ip_vs_svc_table[hash]);
317 * Hash it by fwmark in svc_fwm_table
319 hash = ip_vs_svc_fwm_hashkey(svc->net, svc->fwmark);
320 hlist_add_head_rcu(&svc->f_list, &ip_vs_svc_fwm_table[hash]);
323 svc->flags |= IP_VS_SVC_F_HASHED;
324 /* increase its refcnt because it is referenced by the svc table */
325 atomic_inc(&svc->refcnt);
331 * Unhashes a service from svc_table / svc_fwm_table.
332 * Should be called with locked tables.
334 static int ip_vs_svc_unhash(struct ip_vs_service *svc)
336 if (!(svc->flags & IP_VS_SVC_F_HASHED)) {
337 pr_err("%s(): request for unhash flagged, called from %pF\n",
338 __func__, __builtin_return_address(0));
342 if (svc->fwmark == 0) {
343 /* Remove it from the svc_table table */
344 hlist_del_rcu(&svc->s_list);
346 /* Remove it from the svc_fwm_table table */
347 hlist_del_rcu(&svc->f_list);
350 svc->flags &= ~IP_VS_SVC_F_HASHED;
351 atomic_dec(&svc->refcnt);
357 * Get service by {netns, proto,addr,port} in the service table.
359 static inline struct ip_vs_service *
360 __ip_vs_service_find(struct net *net, int af, __u16 protocol,
361 const union nf_inet_addr *vaddr, __be16 vport)
364 struct ip_vs_service *svc;
366 /* Check for "full" addressed entries */
367 hash = ip_vs_svc_hashkey(net, af, protocol, vaddr, vport);
369 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[hash], s_list) {
371 && ip_vs_addr_equal(af, &svc->addr, vaddr)
372 && (svc->port == vport)
373 && (svc->protocol == protocol)
374 && net_eq(svc->net, net)) {
385 * Get service by {fwmark} in the service table.
387 static inline struct ip_vs_service *
388 __ip_vs_svc_fwm_find(struct net *net, int af, __u32 fwmark)
391 struct ip_vs_service *svc;
393 /* Check for fwmark addressed entries */
394 hash = ip_vs_svc_fwm_hashkey(net, fwmark);
396 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[hash], f_list) {
397 if (svc->fwmark == fwmark && svc->af == af
398 && net_eq(svc->net, net)) {
407 /* Find service, called under RCU lock */
408 struct ip_vs_service *
409 ip_vs_service_find(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);
416 * Check the table hashed by fwmark first
419 svc = __ip_vs_svc_fwm_find(net, af, fwmark);
425 * Check the table hashed by <protocol,addr,port>
426 * for "full" addressed entries
428 svc = __ip_vs_service_find(net, af, protocol, vaddr, vport);
431 && protocol == IPPROTO_TCP
432 && atomic_read(&ipvs->ftpsvc_counter)
433 && (vport == FTPDATA || ntohs(vport) >= PROT_SOCK)) {
435 * Check if ftp service entry exists, the packet
436 * might belong to FTP data connections.
438 svc = __ip_vs_service_find(net, af, protocol, vaddr, FTPPORT);
442 && atomic_read(&ipvs->nullsvc_counter)) {
444 * Check if the catch-all port (port zero) exists
446 svc = __ip_vs_service_find(net, af, protocol, vaddr, 0);
450 IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
451 fwmark, ip_vs_proto_name(protocol),
452 IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
453 svc ? "hit" : "not hit");
460 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
462 atomic_inc(&svc->refcnt);
463 rcu_assign_pointer(dest->svc, svc);
466 static void ip_vs_service_free(struct ip_vs_service *svc)
468 free_percpu(svc->stats.cpustats);
472 static void ip_vs_service_rcu_free(struct rcu_head *head)
474 struct ip_vs_service *svc;
476 svc = container_of(head, struct ip_vs_service, rcu_head);
477 ip_vs_service_free(svc);
480 static void __ip_vs_svc_put(struct ip_vs_service *svc, bool do_delay)
482 if (atomic_dec_and_test(&svc->refcnt)) {
483 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n",
485 IP_VS_DBG_ADDR(svc->af, &svc->addr),
488 call_rcu(&svc->rcu_head, ip_vs_service_rcu_free);
490 ip_vs_service_free(svc);
496 * Returns hash value for real service
498 static inline unsigned int ip_vs_rs_hashkey(int af,
499 const union nf_inet_addr *addr,
502 register unsigned int porth = ntohs(port);
503 __be32 addr_fold = addr->ip;
505 #ifdef CONFIG_IP_VS_IPV6
507 addr_fold = addr->ip6[0]^addr->ip6[1]^
508 addr->ip6[2]^addr->ip6[3];
511 return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
515 /* Hash ip_vs_dest in rs_table by <proto,addr,port>. */
516 static void ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
520 if (dest->in_rs_table)
524 * Hash by proto,addr,port,
525 * which are the parameters of the real service.
527 hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
529 hlist_add_head_rcu(&dest->d_list, &ipvs->rs_table[hash]);
530 dest->in_rs_table = 1;
533 /* Unhash ip_vs_dest from rs_table. */
534 static void ip_vs_rs_unhash(struct ip_vs_dest *dest)
537 * Remove it from the rs_table table.
539 if (dest->in_rs_table) {
540 hlist_del_rcu(&dest->d_list);
541 dest->in_rs_table = 0;
545 /* Check if real service by <proto,addr,port> is present */
546 bool ip_vs_has_real_service(struct net *net, int af, __u16 protocol,
547 const union nf_inet_addr *daddr, __be16 dport)
549 struct netns_ipvs *ipvs = net_ipvs(net);
551 struct ip_vs_dest *dest;
553 /* Check for "full" addressed entries */
554 hash = ip_vs_rs_hashkey(af, daddr, dport);
557 hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
558 if (dest->port == dport &&
560 ip_vs_addr_equal(af, &dest->addr, daddr) &&
561 (dest->protocol == protocol || dest->vfwmark)) {
572 /* Lookup destination by {addr,port} in the given service
573 * Called under RCU lock.
575 static struct ip_vs_dest *
576 ip_vs_lookup_dest(struct ip_vs_service *svc, int dest_af,
577 const union nf_inet_addr *daddr, __be16 dport)
579 struct ip_vs_dest *dest;
582 * Find the destination for the given service
584 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
585 if ((dest->af == dest_af) &&
586 ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
587 (dest->port == dport)) {
597 * Find destination by {daddr,dport,vaddr,protocol}
598 * Created to be used in ip_vs_process_message() in
599 * the backup synchronization daemon. It finds the
600 * destination to be bound to the received connection
602 * Called under RCU lock, no refcnt is returned.
604 struct ip_vs_dest *ip_vs_find_dest(struct net *net, int svc_af, int dest_af,
605 const union nf_inet_addr *daddr,
607 const union nf_inet_addr *vaddr,
608 __be16 vport, __u16 protocol, __u32 fwmark,
611 struct ip_vs_dest *dest;
612 struct ip_vs_service *svc;
615 svc = ip_vs_service_find(net, svc_af, fwmark, protocol, vaddr, vport);
618 if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
620 dest = ip_vs_lookup_dest(svc, dest_af, daddr, port);
622 dest = ip_vs_lookup_dest(svc, dest_af, daddr, port ^ dport);
626 void ip_vs_dest_dst_rcu_free(struct rcu_head *head)
628 struct ip_vs_dest_dst *dest_dst = container_of(head,
629 struct ip_vs_dest_dst,
632 dst_release(dest_dst->dst_cache);
636 /* Release dest_dst and dst_cache for dest in user context */
637 static void __ip_vs_dst_cache_reset(struct ip_vs_dest *dest)
639 struct ip_vs_dest_dst *old;
641 old = rcu_dereference_protected(dest->dest_dst, 1);
643 RCU_INIT_POINTER(dest->dest_dst, NULL);
644 call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
649 * Lookup dest by {svc,addr,port} in the destination trash.
650 * The destination trash is used to hold the destinations that are removed
651 * from the service table but are still referenced by some conn entries.
652 * The reason to add the destination trash is when the dest is temporary
653 * down (either by administrator or by monitor program), the dest can be
654 * picked back from the trash, the remaining connections to the dest can
655 * continue, and the counting information of the dest is also useful for
658 static struct ip_vs_dest *
659 ip_vs_trash_get_dest(struct ip_vs_service *svc, int dest_af,
660 const union nf_inet_addr *daddr, __be16 dport)
662 struct ip_vs_dest *dest;
663 struct netns_ipvs *ipvs = net_ipvs(svc->net);
666 * Find the destination in trash
668 spin_lock_bh(&ipvs->dest_trash_lock);
669 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
670 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
673 IP_VS_DBG_ADDR(dest->af, &dest->addr),
675 atomic_read(&dest->refcnt));
676 if (dest->af == dest_af &&
677 ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
678 dest->port == dport &&
679 dest->vfwmark == svc->fwmark &&
680 dest->protocol == svc->protocol &&
682 (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
683 dest->vport == svc->port))) {
685 list_del(&dest->t_list);
686 ip_vs_dest_hold(dest);
694 spin_unlock_bh(&ipvs->dest_trash_lock);
699 static void ip_vs_dest_free(struct ip_vs_dest *dest)
701 struct ip_vs_service *svc = rcu_dereference_protected(dest->svc, 1);
703 __ip_vs_dst_cache_reset(dest);
704 __ip_vs_svc_put(svc, false);
705 free_percpu(dest->stats.cpustats);
706 ip_vs_dest_put_and_free(dest);
710 * Clean up all the destinations in the trash
711 * Called by the ip_vs_control_cleanup()
713 * When the ip_vs_control_clearup is activated by ipvs module exit,
714 * the service tables must have been flushed and all the connections
715 * are expired, and the refcnt of each destination in the trash must
716 * be 0, so we simply release them here.
718 static void ip_vs_trash_cleanup(struct net *net)
720 struct ip_vs_dest *dest, *nxt;
721 struct netns_ipvs *ipvs = net_ipvs(net);
723 del_timer_sync(&ipvs->dest_trash_timer);
724 /* No need to use dest_trash_lock */
725 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, t_list) {
726 list_del(&dest->t_list);
727 ip_vs_dest_free(dest);
732 ip_vs_copy_stats(struct ip_vs_kstats *dst, struct ip_vs_stats *src)
734 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->kstats.c - src->kstats0.c
736 spin_lock_bh(&src->lock);
738 IP_VS_SHOW_STATS_COUNTER(conns);
739 IP_VS_SHOW_STATS_COUNTER(inpkts);
740 IP_VS_SHOW_STATS_COUNTER(outpkts);
741 IP_VS_SHOW_STATS_COUNTER(inbytes);
742 IP_VS_SHOW_STATS_COUNTER(outbytes);
744 ip_vs_read_estimator(dst, src);
746 spin_unlock_bh(&src->lock);
750 ip_vs_export_stats_user(struct ip_vs_stats_user *dst, struct ip_vs_kstats *src)
752 dst->conns = (u32)src->conns;
753 dst->inpkts = (u32)src->inpkts;
754 dst->outpkts = (u32)src->outpkts;
755 dst->inbytes = src->inbytes;
756 dst->outbytes = src->outbytes;
757 dst->cps = (u32)src->cps;
758 dst->inpps = (u32)src->inpps;
759 dst->outpps = (u32)src->outpps;
760 dst->inbps = (u32)src->inbps;
761 dst->outbps = (u32)src->outbps;
765 ip_vs_zero_stats(struct ip_vs_stats *stats)
767 spin_lock_bh(&stats->lock);
769 /* get current counters as zero point, rates are zeroed */
771 #define IP_VS_ZERO_STATS_COUNTER(c) stats->kstats0.c = stats->kstats.c
773 IP_VS_ZERO_STATS_COUNTER(conns);
774 IP_VS_ZERO_STATS_COUNTER(inpkts);
775 IP_VS_ZERO_STATS_COUNTER(outpkts);
776 IP_VS_ZERO_STATS_COUNTER(inbytes);
777 IP_VS_ZERO_STATS_COUNTER(outbytes);
779 ip_vs_zero_estimator(stats);
781 spin_unlock_bh(&stats->lock);
785 * Update a destination in the given service
788 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
789 struct ip_vs_dest_user_kern *udest, int add)
791 struct netns_ipvs *ipvs = net_ipvs(svc->net);
792 struct ip_vs_service *old_svc;
793 struct ip_vs_scheduler *sched;
796 /* We cannot modify an address and change the address family */
797 BUG_ON(!add && udest->af != dest->af);
799 if (add && udest->af != svc->af)
800 ipvs->mixed_address_family_dests++;
802 /* set the weight and the flags */
803 atomic_set(&dest->weight, udest->weight);
804 conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
805 conn_flags |= IP_VS_CONN_F_INACTIVE;
807 /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
808 if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
809 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
812 * Put the real service in rs_table if not present.
813 * For now only for NAT!
815 ip_vs_rs_hash(ipvs, dest);
817 atomic_set(&dest->conn_flags, conn_flags);
819 /* bind the service */
820 old_svc = rcu_dereference_protected(dest->svc, 1);
822 __ip_vs_bind_svc(dest, svc);
824 if (old_svc != svc) {
825 ip_vs_zero_stats(&dest->stats);
826 __ip_vs_bind_svc(dest, svc);
827 __ip_vs_svc_put(old_svc, true);
831 /* set the dest status flags */
832 dest->flags |= IP_VS_DEST_F_AVAILABLE;
834 if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
835 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
836 dest->u_threshold = udest->u_threshold;
837 dest->l_threshold = udest->l_threshold;
839 dest->af = udest->af;
841 spin_lock_bh(&dest->dst_lock);
842 __ip_vs_dst_cache_reset(dest);
843 spin_unlock_bh(&dest->dst_lock);
846 ip_vs_start_estimator(svc->net, &dest->stats);
847 list_add_rcu(&dest->n_list, &svc->destinations);
849 sched = rcu_dereference_protected(svc->scheduler, 1);
850 if (sched && sched->add_dest)
851 sched->add_dest(svc, dest);
853 sched = rcu_dereference_protected(svc->scheduler, 1);
854 if (sched && sched->upd_dest)
855 sched->upd_dest(svc, dest);
861 * Create a destination for the given service
864 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
865 struct ip_vs_dest **dest_p)
867 struct ip_vs_dest *dest;
868 unsigned int atype, i;
872 #ifdef CONFIG_IP_VS_IPV6
873 if (udest->af == AF_INET6) {
874 atype = ipv6_addr_type(&udest->addr.in6);
875 if ((!(atype & IPV6_ADDR_UNICAST) ||
876 atype & IPV6_ADDR_LINKLOCAL) &&
877 !__ip_vs_addr_is_local_v6(svc->net, &udest->addr.in6))
882 atype = inet_addr_type(svc->net, udest->addr.ip);
883 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
887 dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
891 dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
892 if (!dest->stats.cpustats)
895 for_each_possible_cpu(i) {
896 struct ip_vs_cpu_stats *ip_vs_dest_stats;
897 ip_vs_dest_stats = per_cpu_ptr(dest->stats.cpustats, i);
898 u64_stats_init(&ip_vs_dest_stats->syncp);
901 dest->af = udest->af;
902 dest->protocol = svc->protocol;
903 dest->vaddr = svc->addr;
904 dest->vport = svc->port;
905 dest->vfwmark = svc->fwmark;
906 ip_vs_addr_copy(udest->af, &dest->addr, &udest->addr);
907 dest->port = udest->port;
909 atomic_set(&dest->activeconns, 0);
910 atomic_set(&dest->inactconns, 0);
911 atomic_set(&dest->persistconns, 0);
912 atomic_set(&dest->refcnt, 1);
914 INIT_HLIST_NODE(&dest->d_list);
915 spin_lock_init(&dest->dst_lock);
916 spin_lock_init(&dest->stats.lock);
917 __ip_vs_update_dest(svc, dest, udest, 1);
931 * Add a destination into an existing service
934 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
936 struct ip_vs_dest *dest;
937 union nf_inet_addr daddr;
938 __be16 dport = udest->port;
943 if (udest->weight < 0) {
944 pr_err("%s(): server weight less than zero\n", __func__);
948 if (udest->l_threshold > udest->u_threshold) {
949 pr_err("%s(): lower threshold is higher than upper threshold\n",
954 ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
956 /* We use function that requires RCU lock */
958 dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
962 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
967 * Check if the dest already exists in the trash and
968 * is from the same service
970 dest = ip_vs_trash_get_dest(svc, udest->af, &daddr, dport);
973 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
974 "dest->refcnt=%d, service %u/%s:%u\n",
975 IP_VS_DBG_ADDR(udest->af, &daddr), ntohs(dport),
976 atomic_read(&dest->refcnt),
978 IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
981 __ip_vs_update_dest(svc, dest, udest, 1);
985 * Allocate and initialize the dest structure
987 ret = ip_vs_new_dest(svc, udest, &dest);
996 * Edit a destination in the given service
999 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1001 struct ip_vs_dest *dest;
1002 union nf_inet_addr daddr;
1003 __be16 dport = udest->port;
1007 if (udest->weight < 0) {
1008 pr_err("%s(): server weight less than zero\n", __func__);
1012 if (udest->l_threshold > udest->u_threshold) {
1013 pr_err("%s(): lower threshold is higher than upper threshold\n",
1018 ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
1020 /* We use function that requires RCU lock */
1022 dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
1026 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1030 __ip_vs_update_dest(svc, dest, udest, 0);
1037 * Delete a destination (must be already unlinked from the service)
1039 static void __ip_vs_del_dest(struct net *net, struct ip_vs_dest *dest,
1042 struct netns_ipvs *ipvs = net_ipvs(net);
1044 ip_vs_stop_estimator(net, &dest->stats);
1047 * Remove it from the d-linked list with the real services.
1049 ip_vs_rs_unhash(dest);
1051 spin_lock_bh(&ipvs->dest_trash_lock);
1052 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1053 IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port),
1054 atomic_read(&dest->refcnt));
1055 if (list_empty(&ipvs->dest_trash) && !cleanup)
1056 mod_timer(&ipvs->dest_trash_timer,
1057 jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1058 /* dest lives in trash without reference */
1059 list_add(&dest->t_list, &ipvs->dest_trash);
1060 dest->idle_start = 0;
1061 spin_unlock_bh(&ipvs->dest_trash_lock);
1062 ip_vs_dest_put(dest);
1067 * Unlink a destination from the given service
1069 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1070 struct ip_vs_dest *dest,
1073 dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1076 * Remove it from the d-linked destination list.
1078 list_del_rcu(&dest->n_list);
1081 if (dest->af != svc->af)
1082 net_ipvs(svc->net)->mixed_address_family_dests--;
1085 struct ip_vs_scheduler *sched;
1087 sched = rcu_dereference_protected(svc->scheduler, 1);
1088 if (sched && sched->del_dest)
1089 sched->del_dest(svc, dest);
1095 * Delete a destination server in the given service
1098 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1100 struct ip_vs_dest *dest;
1101 __be16 dport = udest->port;
1105 /* We use function that requires RCU lock */
1107 dest = ip_vs_lookup_dest(svc, udest->af, &udest->addr, dport);
1111 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1116 * Unlink dest from the service
1118 __ip_vs_unlink_dest(svc, dest, 1);
1121 * Delete the destination
1123 __ip_vs_del_dest(svc->net, dest, false);
1130 static void ip_vs_dest_trash_expire(unsigned long data)
1132 struct net *net = (struct net *) data;
1133 struct netns_ipvs *ipvs = net_ipvs(net);
1134 struct ip_vs_dest *dest, *next;
1135 unsigned long now = jiffies;
1137 spin_lock(&ipvs->dest_trash_lock);
1138 list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) {
1139 if (atomic_read(&dest->refcnt) > 0)
1141 if (dest->idle_start) {
1142 if (time_before(now, dest->idle_start +
1143 IP_VS_DEST_TRASH_PERIOD))
1146 dest->idle_start = max(1UL, now);
1149 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1151 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1153 list_del(&dest->t_list);
1154 ip_vs_dest_free(dest);
1156 if (!list_empty(&ipvs->dest_trash))
1157 mod_timer(&ipvs->dest_trash_timer,
1158 jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1159 spin_unlock(&ipvs->dest_trash_lock);
1163 * Add a service into the service hash table
1166 ip_vs_add_service(struct net *net, struct ip_vs_service_user_kern *u,
1167 struct ip_vs_service **svc_p)
1170 struct ip_vs_scheduler *sched = NULL;
1171 struct ip_vs_pe *pe = NULL;
1172 struct ip_vs_service *svc = NULL;
1173 struct netns_ipvs *ipvs = net_ipvs(net);
1175 /* increase the module use count */
1176 ip_vs_use_count_inc();
1178 /* Lookup the scheduler by 'u->sched_name' */
1179 if (strcmp(u->sched_name, "none")) {
1180 sched = ip_vs_scheduler_get(u->sched_name);
1182 pr_info("Scheduler module ip_vs_%s not found\n",
1189 if (u->pe_name && *u->pe_name) {
1190 pe = ip_vs_pe_getbyname(u->pe_name);
1192 pr_info("persistence engine module ip_vs_pe_%s "
1193 "not found\n", u->pe_name);
1199 #ifdef CONFIG_IP_VS_IPV6
1200 if (u->af == AF_INET6) {
1201 __u32 plen = (__force __u32) u->netmask;
1203 if (plen < 1 || plen > 128) {
1210 svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1212 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1216 svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1217 if (!svc->stats.cpustats) {
1222 for_each_possible_cpu(i) {
1223 struct ip_vs_cpu_stats *ip_vs_stats;
1224 ip_vs_stats = per_cpu_ptr(svc->stats.cpustats, i);
1225 u64_stats_init(&ip_vs_stats->syncp);
1229 /* I'm the first user of the service */
1230 atomic_set(&svc->refcnt, 0);
1233 svc->protocol = u->protocol;
1234 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1235 svc->port = u->port;
1236 svc->fwmark = u->fwmark;
1237 svc->flags = u->flags;
1238 svc->timeout = u->timeout * HZ;
1239 svc->netmask = u->netmask;
1242 INIT_LIST_HEAD(&svc->destinations);
1243 spin_lock_init(&svc->sched_lock);
1244 spin_lock_init(&svc->stats.lock);
1246 /* Bind the scheduler */
1248 ret = ip_vs_bind_scheduler(svc, sched);
1254 /* Bind the ct retriever */
1255 RCU_INIT_POINTER(svc->pe, pe);
1258 /* Update the virtual service counters */
1259 if (svc->port == FTPPORT)
1260 atomic_inc(&ipvs->ftpsvc_counter);
1261 else if (svc->port == 0)
1262 atomic_inc(&ipvs->nullsvc_counter);
1264 ip_vs_start_estimator(net, &svc->stats);
1266 /* Count only IPv4 services for old get/setsockopt interface */
1267 if (svc->af == AF_INET)
1268 ipvs->num_services++;
1270 /* Hash the service into the service table */
1271 ip_vs_svc_hash(svc);
1274 /* Now there is a service - full throttle */
1281 ip_vs_unbind_scheduler(svc, sched);
1282 ip_vs_service_free(svc);
1284 ip_vs_scheduler_put(sched);
1287 /* decrease the module use count */
1288 ip_vs_use_count_dec();
1295 * Edit a service and bind it with a new scheduler
1298 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1300 struct ip_vs_scheduler *sched = NULL, *old_sched;
1301 struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1305 * Lookup the scheduler, by 'u->sched_name'
1307 if (strcmp(u->sched_name, "none")) {
1308 sched = ip_vs_scheduler_get(u->sched_name);
1310 pr_info("Scheduler module ip_vs_%s not found\n",
1317 if (u->pe_name && *u->pe_name) {
1318 pe = ip_vs_pe_getbyname(u->pe_name);
1320 pr_info("persistence engine module ip_vs_pe_%s "
1321 "not found\n", u->pe_name);
1328 #ifdef CONFIG_IP_VS_IPV6
1329 if (u->af == AF_INET6) {
1330 __u32 plen = (__force __u32) u->netmask;
1332 if (plen < 1 || plen > 128) {
1339 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1340 if (sched != old_sched) {
1342 ip_vs_unbind_scheduler(svc, old_sched);
1343 RCU_INIT_POINTER(svc->scheduler, NULL);
1344 /* Wait all svc->sched_data users */
1347 /* Bind the new scheduler */
1349 ret = ip_vs_bind_scheduler(svc, sched);
1351 ip_vs_scheduler_put(sched);
1358 * Set the flags and timeout value
1360 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1361 svc->timeout = u->timeout * HZ;
1362 svc->netmask = u->netmask;
1364 old_pe = rcu_dereference_protected(svc->pe, 1);
1366 rcu_assign_pointer(svc->pe, pe);
1369 ip_vs_scheduler_put(old_sched);
1370 ip_vs_pe_put(old_pe);
1375 * Delete a service from the service list
1376 * - The service must be unlinked, unlocked and not referenced!
1377 * - We are called under _bh lock
1379 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1381 struct ip_vs_dest *dest, *nxt;
1382 struct ip_vs_scheduler *old_sched;
1383 struct ip_vs_pe *old_pe;
1384 struct netns_ipvs *ipvs = net_ipvs(svc->net);
1386 pr_info("%s: enter\n", __func__);
1388 /* Count only IPv4 services for old get/setsockopt interface */
1389 if (svc->af == AF_INET)
1390 ipvs->num_services--;
1392 ip_vs_stop_estimator(svc->net, &svc->stats);
1394 /* Unbind scheduler */
1395 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1396 ip_vs_unbind_scheduler(svc, old_sched);
1397 ip_vs_scheduler_put(old_sched);
1399 /* Unbind persistence engine, keep svc->pe */
1400 old_pe = rcu_dereference_protected(svc->pe, 1);
1401 ip_vs_pe_put(old_pe);
1404 * Unlink the whole destination list
1406 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1407 __ip_vs_unlink_dest(svc, dest, 0);
1408 __ip_vs_del_dest(svc->net, dest, cleanup);
1412 * Update the virtual service counters
1414 if (svc->port == FTPPORT)
1415 atomic_dec(&ipvs->ftpsvc_counter);
1416 else if (svc->port == 0)
1417 atomic_dec(&ipvs->nullsvc_counter);
1420 * Free the service if nobody refers to it
1422 __ip_vs_svc_put(svc, true);
1424 /* decrease the module use count */
1425 ip_vs_use_count_dec();
1429 * Unlink a service from list and try to delete it if its refcnt reached 0
1431 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1433 /* Hold svc to avoid double release from dest_trash */
1434 atomic_inc(&svc->refcnt);
1436 * Unhash it from the service table
1438 ip_vs_svc_unhash(svc);
1440 __ip_vs_del_service(svc, cleanup);
1444 * Delete a service from the service list
1446 static int ip_vs_del_service(struct ip_vs_service *svc)
1450 ip_vs_unlink_service(svc, false);
1457 * Flush all the virtual services
1459 static int ip_vs_flush(struct net *net, bool cleanup)
1462 struct ip_vs_service *svc;
1463 struct hlist_node *n;
1466 * Flush the service table hashed by <netns,protocol,addr,port>
1468 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1469 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1471 if (net_eq(svc->net, net))
1472 ip_vs_unlink_service(svc, cleanup);
1477 * Flush the service table hashed by fwmark
1479 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1480 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1482 if (net_eq(svc->net, net))
1483 ip_vs_unlink_service(svc, cleanup);
1491 * Delete service by {netns} in the service table.
1492 * Called by __ip_vs_cleanup()
1494 void ip_vs_service_net_cleanup(struct net *net)
1497 /* Check for "full" addressed entries */
1498 mutex_lock(&__ip_vs_mutex);
1499 ip_vs_flush(net, true);
1500 mutex_unlock(&__ip_vs_mutex);
1504 /* Put all references for device (dst_cache) */
1506 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1508 struct ip_vs_dest_dst *dest_dst;
1510 spin_lock_bh(&dest->dst_lock);
1511 dest_dst = rcu_dereference_protected(dest->dest_dst, 1);
1512 if (dest_dst && dest_dst->dst_cache->dev == dev) {
1513 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1515 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1517 atomic_read(&dest->refcnt));
1518 __ip_vs_dst_cache_reset(dest);
1520 spin_unlock_bh(&dest->dst_lock);
1523 /* Netdev event receiver
1524 * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1526 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1529 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1530 struct net *net = dev_net(dev);
1531 struct netns_ipvs *ipvs = net_ipvs(net);
1532 struct ip_vs_service *svc;
1533 struct ip_vs_dest *dest;
1536 if (event != NETDEV_DOWN || !ipvs)
1538 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1540 mutex_lock(&__ip_vs_mutex);
1541 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1542 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1543 if (net_eq(svc->net, net)) {
1544 list_for_each_entry(dest, &svc->destinations,
1546 ip_vs_forget_dev(dest, dev);
1551 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1552 if (net_eq(svc->net, net)) {
1553 list_for_each_entry(dest, &svc->destinations,
1555 ip_vs_forget_dev(dest, dev);
1562 spin_lock_bh(&ipvs->dest_trash_lock);
1563 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1564 ip_vs_forget_dev(dest, dev);
1566 spin_unlock_bh(&ipvs->dest_trash_lock);
1567 mutex_unlock(&__ip_vs_mutex);
1573 * Zero counters in a service or all services
1575 static int ip_vs_zero_service(struct ip_vs_service *svc)
1577 struct ip_vs_dest *dest;
1579 list_for_each_entry(dest, &svc->destinations, n_list) {
1580 ip_vs_zero_stats(&dest->stats);
1582 ip_vs_zero_stats(&svc->stats);
1586 static int ip_vs_zero_all(struct net *net)
1589 struct ip_vs_service *svc;
1591 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1592 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1593 if (net_eq(svc->net, net))
1594 ip_vs_zero_service(svc);
1598 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1599 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1600 if (net_eq(svc->net, net))
1601 ip_vs_zero_service(svc);
1605 ip_vs_zero_stats(&net_ipvs(net)->tot_stats);
1609 #ifdef CONFIG_SYSCTL
1612 static int three = 3;
1615 proc_do_defense_mode(struct ctl_table *table, int write,
1616 void __user *buffer, size_t *lenp, loff_t *ppos)
1618 struct net *net = current->nsproxy->net_ns;
1619 int *valp = table->data;
1623 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1624 if (write && (*valp != val)) {
1625 if ((*valp < 0) || (*valp > 3)) {
1626 /* Restore the correct value */
1629 update_defense_level(net_ipvs(net));
1636 proc_do_sync_threshold(struct ctl_table *table, int write,
1637 void __user *buffer, size_t *lenp, loff_t *ppos)
1639 int *valp = table->data;
1643 /* backup the value first */
1644 memcpy(val, valp, sizeof(val));
1646 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1647 if (write && (valp[0] < 0 || valp[1] < 0 ||
1648 (valp[0] >= valp[1] && valp[1]))) {
1649 /* Restore the correct value */
1650 memcpy(valp, val, sizeof(val));
1656 proc_do_sync_mode(struct ctl_table *table, int write,
1657 void __user *buffer, size_t *lenp, loff_t *ppos)
1659 int *valp = table->data;
1663 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1664 if (write && (*valp != val)) {
1665 if ((*valp < 0) || (*valp > 1)) {
1666 /* Restore the correct value */
1674 proc_do_sync_ports(struct ctl_table *table, int write,
1675 void __user *buffer, size_t *lenp, loff_t *ppos)
1677 int *valp = table->data;
1681 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1682 if (write && (*valp != val)) {
1683 if (*valp < 1 || !is_power_of_2(*valp)) {
1684 /* Restore the correct value */
1692 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1693 * Do not change order or insert new entries without
1694 * align with netns init in ip_vs_control_net_init()
1697 static struct ctl_table vs_vars[] = {
1699 .procname = "amemthresh",
1700 .maxlen = sizeof(int),
1702 .proc_handler = proc_dointvec,
1705 .procname = "am_droprate",
1706 .maxlen = sizeof(int),
1708 .proc_handler = proc_dointvec,
1711 .procname = "drop_entry",
1712 .maxlen = sizeof(int),
1714 .proc_handler = proc_do_defense_mode,
1717 .procname = "drop_packet",
1718 .maxlen = sizeof(int),
1720 .proc_handler = proc_do_defense_mode,
1722 #ifdef CONFIG_IP_VS_NFCT
1724 .procname = "conntrack",
1725 .maxlen = sizeof(int),
1727 .proc_handler = &proc_dointvec,
1731 .procname = "secure_tcp",
1732 .maxlen = sizeof(int),
1734 .proc_handler = proc_do_defense_mode,
1737 .procname = "snat_reroute",
1738 .maxlen = sizeof(int),
1740 .proc_handler = &proc_dointvec,
1743 .procname = "sync_version",
1744 .maxlen = sizeof(int),
1746 .proc_handler = &proc_do_sync_mode,
1749 .procname = "sync_ports",
1750 .maxlen = sizeof(int),
1752 .proc_handler = &proc_do_sync_ports,
1755 .procname = "sync_persist_mode",
1756 .maxlen = sizeof(int),
1758 .proc_handler = proc_dointvec,
1761 .procname = "sync_qlen_max",
1762 .maxlen = sizeof(unsigned long),
1764 .proc_handler = proc_doulongvec_minmax,
1767 .procname = "sync_sock_size",
1768 .maxlen = sizeof(int),
1770 .proc_handler = proc_dointvec,
1773 .procname = "cache_bypass",
1774 .maxlen = sizeof(int),
1776 .proc_handler = proc_dointvec,
1779 .procname = "expire_nodest_conn",
1780 .maxlen = sizeof(int),
1782 .proc_handler = proc_dointvec,
1785 .procname = "sloppy_tcp",
1786 .maxlen = sizeof(int),
1788 .proc_handler = proc_dointvec,
1791 .procname = "sloppy_sctp",
1792 .maxlen = sizeof(int),
1794 .proc_handler = proc_dointvec,
1797 .procname = "expire_quiescent_template",
1798 .maxlen = sizeof(int),
1800 .proc_handler = proc_dointvec,
1803 .procname = "sync_threshold",
1805 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1807 .proc_handler = proc_do_sync_threshold,
1810 .procname = "sync_refresh_period",
1811 .maxlen = sizeof(int),
1813 .proc_handler = proc_dointvec_jiffies,
1816 .procname = "sync_retries",
1817 .maxlen = sizeof(int),
1819 .proc_handler = proc_dointvec_minmax,
1824 .procname = "nat_icmp_send",
1825 .maxlen = sizeof(int),
1827 .proc_handler = proc_dointvec,
1830 .procname = "pmtu_disc",
1831 .maxlen = sizeof(int),
1833 .proc_handler = proc_dointvec,
1836 .procname = "backup_only",
1837 .maxlen = sizeof(int),
1839 .proc_handler = proc_dointvec,
1842 .procname = "conn_reuse_mode",
1843 .maxlen = sizeof(int),
1845 .proc_handler = proc_dointvec,
1847 #ifdef CONFIG_IP_VS_DEBUG
1849 .procname = "debug_level",
1850 .data = &sysctl_ip_vs_debug_level,
1851 .maxlen = sizeof(int),
1853 .proc_handler = proc_dointvec,
1861 #ifdef CONFIG_PROC_FS
1864 struct seq_net_private p; /* Do not move this, netns depends upon it*/
1865 struct hlist_head *table;
1870 * Write the contents of the VS rule table to a PROCfs file.
1871 * (It is kept just for backward compatibility)
1873 static inline const char *ip_vs_fwd_name(unsigned int flags)
1875 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1876 case IP_VS_CONN_F_LOCALNODE:
1878 case IP_VS_CONN_F_TUNNEL:
1880 case IP_VS_CONN_F_DROUTE:
1888 /* Get the Nth entry in the two lists */
1889 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1891 struct net *net = seq_file_net(seq);
1892 struct ip_vs_iter *iter = seq->private;
1894 struct ip_vs_service *svc;
1896 /* look in hash by protocol */
1897 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1898 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
1899 if (net_eq(svc->net, net) && pos-- == 0) {
1900 iter->table = ip_vs_svc_table;
1907 /* keep looking in fwmark */
1908 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1909 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
1911 if (net_eq(svc->net, net) && pos-- == 0) {
1912 iter->table = ip_vs_svc_fwm_table;
1922 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1926 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1930 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1932 struct hlist_node *e;
1933 struct ip_vs_iter *iter;
1934 struct ip_vs_service *svc;
1937 if (v == SEQ_START_TOKEN)
1938 return ip_vs_info_array(seq,0);
1941 iter = seq->private;
1943 if (iter->table == ip_vs_svc_table) {
1944 /* next service in table hashed by protocol */
1945 e = rcu_dereference(hlist_next_rcu(&svc->s_list));
1947 return hlist_entry(e, struct ip_vs_service, s_list);
1949 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1950 hlist_for_each_entry_rcu(svc,
1951 &ip_vs_svc_table[iter->bucket],
1957 iter->table = ip_vs_svc_fwm_table;
1962 /* next service in hashed by fwmark */
1963 e = rcu_dereference(hlist_next_rcu(&svc->f_list));
1965 return hlist_entry(e, struct ip_vs_service, f_list);
1968 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1969 hlist_for_each_entry_rcu(svc,
1970 &ip_vs_svc_fwm_table[iter->bucket],
1978 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
1985 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
1987 if (v == SEQ_START_TOKEN) {
1989 "IP Virtual Server version %d.%d.%d (size=%d)\n",
1990 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
1992 "Prot LocalAddress:Port Scheduler Flags\n");
1994 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
1996 const struct ip_vs_service *svc = v;
1997 const struct ip_vs_iter *iter = seq->private;
1998 const struct ip_vs_dest *dest;
1999 struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler);
2000 char *sched_name = sched ? sched->name : "none";
2002 if (iter->table == ip_vs_svc_table) {
2003 #ifdef CONFIG_IP_VS_IPV6
2004 if (svc->af == AF_INET6)
2005 seq_printf(seq, "%s [%pI6]:%04X %s ",
2006 ip_vs_proto_name(svc->protocol),
2012 seq_printf(seq, "%s %08X:%04X %s %s ",
2013 ip_vs_proto_name(svc->protocol),
2014 ntohl(svc->addr.ip),
2017 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2019 seq_printf(seq, "FWM %08X %s %s",
2020 svc->fwmark, sched_name,
2021 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2024 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2025 seq_printf(seq, "persistent %d %08X\n",
2027 ntohl(svc->netmask));
2029 seq_putc(seq, '\n');
2031 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
2032 #ifdef CONFIG_IP_VS_IPV6
2033 if (dest->af == AF_INET6)
2036 " %-7s %-6d %-10d %-10d\n",
2039 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2040 atomic_read(&dest->weight),
2041 atomic_read(&dest->activeconns),
2042 atomic_read(&dest->inactconns));
2047 "%-7s %-6d %-10d %-10d\n",
2048 ntohl(dest->addr.ip),
2050 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2051 atomic_read(&dest->weight),
2052 atomic_read(&dest->activeconns),
2053 atomic_read(&dest->inactconns));
2060 static const struct seq_operations ip_vs_info_seq_ops = {
2061 .start = ip_vs_info_seq_start,
2062 .next = ip_vs_info_seq_next,
2063 .stop = ip_vs_info_seq_stop,
2064 .show = ip_vs_info_seq_show,
2067 static int ip_vs_info_open(struct inode *inode, struct file *file)
2069 return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2070 sizeof(struct ip_vs_iter));
2073 static const struct file_operations ip_vs_info_fops = {
2074 .owner = THIS_MODULE,
2075 .open = ip_vs_info_open,
2077 .llseek = seq_lseek,
2078 .release = seq_release_net,
2081 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2083 struct net *net = seq_file_single_net(seq);
2084 struct ip_vs_kstats show;
2086 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2088 " Total Incoming Outgoing Incoming Outgoing\n");
2090 " Conns Packets Packets Bytes Bytes\n");
2092 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2093 seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n\n",
2094 (unsigned long long)show.conns,
2095 (unsigned long long)show.inpkts,
2096 (unsigned long long)show.outpkts,
2097 (unsigned long long)show.inbytes,
2098 (unsigned long long)show.outbytes);
2100 /* 01234567 01234567 01234567 0123456701234567 0123456701234567*/
2102 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2103 seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n",
2104 (unsigned long long)show.cps,
2105 (unsigned long long)show.inpps,
2106 (unsigned long long)show.outpps,
2107 (unsigned long long)show.inbps,
2108 (unsigned long long)show.outbps);
2113 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2115 return single_open_net(inode, file, ip_vs_stats_show);
2118 static const struct file_operations ip_vs_stats_fops = {
2119 .owner = THIS_MODULE,
2120 .open = ip_vs_stats_seq_open,
2122 .llseek = seq_lseek,
2123 .release = single_release_net,
2126 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2128 struct net *net = seq_file_single_net(seq);
2129 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2130 struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats;
2131 struct ip_vs_kstats kstats;
2134 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2136 " Total Incoming Outgoing Incoming Outgoing\n");
2138 "CPU Conns Packets Packets Bytes Bytes\n");
2140 for_each_possible_cpu(i) {
2141 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2143 u64 conns, inpkts, outpkts, inbytes, outbytes;
2146 start = u64_stats_fetch_begin_irq(&u->syncp);
2147 conns = u->cnt.conns;
2148 inpkts = u->cnt.inpkts;
2149 outpkts = u->cnt.outpkts;
2150 inbytes = u->cnt.inbytes;
2151 outbytes = u->cnt.outbytes;
2152 } while (u64_stats_fetch_retry_irq(&u->syncp, start));
2154 seq_printf(seq, "%3X %8LX %8LX %8LX %16LX %16LX\n",
2155 i, (u64)conns, (u64)inpkts,
2156 (u64)outpkts, (u64)inbytes,
2160 ip_vs_copy_stats(&kstats, tot_stats);
2162 seq_printf(seq, " ~ %8LX %8LX %8LX %16LX %16LX\n\n",
2163 (unsigned long long)kstats.conns,
2164 (unsigned long long)kstats.inpkts,
2165 (unsigned long long)kstats.outpkts,
2166 (unsigned long long)kstats.inbytes,
2167 (unsigned long long)kstats.outbytes);
2169 /* ... 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2171 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2172 seq_printf(seq, " %8LX %8LX %8LX %16LX %16LX\n",
2182 static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2184 return single_open_net(inode, file, ip_vs_stats_percpu_show);
2187 static const struct file_operations ip_vs_stats_percpu_fops = {
2188 .owner = THIS_MODULE,
2189 .open = ip_vs_stats_percpu_seq_open,
2191 .llseek = seq_lseek,
2192 .release = single_release_net,
2197 * Set timeout values for tcp tcpfin udp in the timeout_table.
2199 static int ip_vs_set_timeout(struct net *net, struct ip_vs_timeout_user *u)
2201 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2202 struct ip_vs_proto_data *pd;
2205 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2210 #ifdef CONFIG_IP_VS_PROTO_TCP
2211 if (u->tcp_timeout) {
2212 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2213 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2214 = u->tcp_timeout * HZ;
2217 if (u->tcp_fin_timeout) {
2218 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2219 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2220 = u->tcp_fin_timeout * HZ;
2224 #ifdef CONFIG_IP_VS_PROTO_UDP
2225 if (u->udp_timeout) {
2226 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2227 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2228 = u->udp_timeout * HZ;
2234 #define CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2236 struct ip_vs_svcdest_user {
2237 struct ip_vs_service_user s;
2238 struct ip_vs_dest_user d;
2241 static const unsigned char set_arglen[CMDID(IP_VS_SO_SET_MAX) + 1] = {
2242 [CMDID(IP_VS_SO_SET_ADD)] = sizeof(struct ip_vs_service_user),
2243 [CMDID(IP_VS_SO_SET_EDIT)] = sizeof(struct ip_vs_service_user),
2244 [CMDID(IP_VS_SO_SET_DEL)] = sizeof(struct ip_vs_service_user),
2245 [CMDID(IP_VS_SO_SET_ADDDEST)] = sizeof(struct ip_vs_svcdest_user),
2246 [CMDID(IP_VS_SO_SET_DELDEST)] = sizeof(struct ip_vs_svcdest_user),
2247 [CMDID(IP_VS_SO_SET_EDITDEST)] = sizeof(struct ip_vs_svcdest_user),
2248 [CMDID(IP_VS_SO_SET_TIMEOUT)] = sizeof(struct ip_vs_timeout_user),
2249 [CMDID(IP_VS_SO_SET_STARTDAEMON)] = sizeof(struct ip_vs_daemon_user),
2250 [CMDID(IP_VS_SO_SET_STOPDAEMON)] = sizeof(struct ip_vs_daemon_user),
2251 [CMDID(IP_VS_SO_SET_ZERO)] = sizeof(struct ip_vs_service_user),
2254 union ip_vs_set_arglen {
2255 struct ip_vs_service_user field_IP_VS_SO_SET_ADD;
2256 struct ip_vs_service_user field_IP_VS_SO_SET_EDIT;
2257 struct ip_vs_service_user field_IP_VS_SO_SET_DEL;
2258 struct ip_vs_svcdest_user field_IP_VS_SO_SET_ADDDEST;
2259 struct ip_vs_svcdest_user field_IP_VS_SO_SET_DELDEST;
2260 struct ip_vs_svcdest_user field_IP_VS_SO_SET_EDITDEST;
2261 struct ip_vs_timeout_user field_IP_VS_SO_SET_TIMEOUT;
2262 struct ip_vs_daemon_user field_IP_VS_SO_SET_STARTDAEMON;
2263 struct ip_vs_daemon_user field_IP_VS_SO_SET_STOPDAEMON;
2264 struct ip_vs_service_user field_IP_VS_SO_SET_ZERO;
2267 #define MAX_SET_ARGLEN sizeof(union ip_vs_set_arglen)
2269 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2270 struct ip_vs_service_user *usvc_compat)
2272 memset(usvc, 0, sizeof(*usvc));
2275 usvc->protocol = usvc_compat->protocol;
2276 usvc->addr.ip = usvc_compat->addr;
2277 usvc->port = usvc_compat->port;
2278 usvc->fwmark = usvc_compat->fwmark;
2280 /* Deep copy of sched_name is not needed here */
2281 usvc->sched_name = usvc_compat->sched_name;
2283 usvc->flags = usvc_compat->flags;
2284 usvc->timeout = usvc_compat->timeout;
2285 usvc->netmask = usvc_compat->netmask;
2288 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2289 struct ip_vs_dest_user *udest_compat)
2291 memset(udest, 0, sizeof(*udest));
2293 udest->addr.ip = udest_compat->addr;
2294 udest->port = udest_compat->port;
2295 udest->conn_flags = udest_compat->conn_flags;
2296 udest->weight = udest_compat->weight;
2297 udest->u_threshold = udest_compat->u_threshold;
2298 udest->l_threshold = udest_compat->l_threshold;
2299 udest->af = AF_INET;
2303 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2305 struct net *net = sock_net(sk);
2307 unsigned char arg[MAX_SET_ARGLEN];
2308 struct ip_vs_service_user *usvc_compat;
2309 struct ip_vs_service_user_kern usvc;
2310 struct ip_vs_service *svc;
2311 struct ip_vs_dest_user *udest_compat;
2312 struct ip_vs_dest_user_kern udest;
2313 struct netns_ipvs *ipvs = net_ipvs(net);
2315 BUILD_BUG_ON(sizeof(arg) > 255);
2316 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2319 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2321 if (len != set_arglen[CMDID(cmd)]) {
2322 IP_VS_DBG(1, "set_ctl: len %u != %u\n",
2323 len, set_arglen[CMDID(cmd)]);
2327 if (copy_from_user(arg, user, len) != 0)
2330 /* increase the module use count */
2331 ip_vs_use_count_inc();
2333 /* Handle daemons since they have another lock */
2334 if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2335 cmd == IP_VS_SO_SET_STOPDAEMON) {
2336 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2338 mutex_lock(&ipvs->sync_mutex);
2339 if (cmd == IP_VS_SO_SET_STARTDAEMON)
2340 ret = start_sync_thread(net, dm->state, dm->mcast_ifn,
2343 ret = stop_sync_thread(net, dm->state);
2344 mutex_unlock(&ipvs->sync_mutex);
2348 mutex_lock(&__ip_vs_mutex);
2349 if (cmd == IP_VS_SO_SET_FLUSH) {
2350 /* Flush the virtual service */
2351 ret = ip_vs_flush(net, false);
2353 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2354 /* Set timeout values for (tcp tcpfin udp) */
2355 ret = ip_vs_set_timeout(net, (struct ip_vs_timeout_user *)arg);
2359 usvc_compat = (struct ip_vs_service_user *)arg;
2360 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2362 /* We only use the new structs internally, so copy userspace compat
2363 * structs to extended internal versions */
2364 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2365 ip_vs_copy_udest_compat(&udest, udest_compat);
2367 if (cmd == IP_VS_SO_SET_ZERO) {
2368 /* if no service address is set, zero counters in all */
2369 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2370 ret = ip_vs_zero_all(net);
2375 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2376 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2377 usvc.protocol != IPPROTO_SCTP) {
2378 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2379 usvc.protocol, &usvc.addr.ip,
2380 ntohs(usvc.port), usvc.sched_name);
2385 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2387 if (usvc.fwmark == 0)
2388 svc = __ip_vs_service_find(net, usvc.af, usvc.protocol,
2389 &usvc.addr, usvc.port);
2391 svc = __ip_vs_svc_fwm_find(net, usvc.af, usvc.fwmark);
2394 if (cmd != IP_VS_SO_SET_ADD
2395 && (svc == NULL || svc->protocol != usvc.protocol)) {
2401 case IP_VS_SO_SET_ADD:
2405 ret = ip_vs_add_service(net, &usvc, &svc);
2407 case IP_VS_SO_SET_EDIT:
2408 ret = ip_vs_edit_service(svc, &usvc);
2410 case IP_VS_SO_SET_DEL:
2411 ret = ip_vs_del_service(svc);
2415 case IP_VS_SO_SET_ZERO:
2416 ret = ip_vs_zero_service(svc);
2418 case IP_VS_SO_SET_ADDDEST:
2419 ret = ip_vs_add_dest(svc, &udest);
2421 case IP_VS_SO_SET_EDITDEST:
2422 ret = ip_vs_edit_dest(svc, &udest);
2424 case IP_VS_SO_SET_DELDEST:
2425 ret = ip_vs_del_dest(svc, &udest);
2432 mutex_unlock(&__ip_vs_mutex);
2434 /* decrease the module use count */
2435 ip_vs_use_count_dec();
2442 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2444 struct ip_vs_scheduler *sched;
2445 struct ip_vs_kstats kstats;
2448 sched = rcu_dereference_protected(src->scheduler, 1);
2449 sched_name = sched ? sched->name : "none";
2450 dst->protocol = src->protocol;
2451 dst->addr = src->addr.ip;
2452 dst->port = src->port;
2453 dst->fwmark = src->fwmark;
2454 strlcpy(dst->sched_name, sched_name, sizeof(dst->sched_name));
2455 dst->flags = src->flags;
2456 dst->timeout = src->timeout / HZ;
2457 dst->netmask = src->netmask;
2458 dst->num_dests = src->num_dests;
2459 ip_vs_copy_stats(&kstats, &src->stats);
2460 ip_vs_export_stats_user(&dst->stats, &kstats);
2464 __ip_vs_get_service_entries(struct net *net,
2465 const struct ip_vs_get_services *get,
2466 struct ip_vs_get_services __user *uptr)
2469 struct ip_vs_service *svc;
2470 struct ip_vs_service_entry entry;
2473 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2474 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2475 /* Only expose IPv4 entries to old interface */
2476 if (svc->af != AF_INET || !net_eq(svc->net, net))
2479 if (count >= get->num_services)
2481 memset(&entry, 0, sizeof(entry));
2482 ip_vs_copy_service(&entry, svc);
2483 if (copy_to_user(&uptr->entrytable[count],
2484 &entry, sizeof(entry))) {
2492 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2493 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2494 /* Only expose IPv4 entries to old interface */
2495 if (svc->af != AF_INET || !net_eq(svc->net, net))
2498 if (count >= get->num_services)
2500 memset(&entry, 0, sizeof(entry));
2501 ip_vs_copy_service(&entry, svc);
2502 if (copy_to_user(&uptr->entrytable[count],
2503 &entry, sizeof(entry))) {
2515 __ip_vs_get_dest_entries(struct net *net, const struct ip_vs_get_dests *get,
2516 struct ip_vs_get_dests __user *uptr)
2518 struct ip_vs_service *svc;
2519 union nf_inet_addr addr = { .ip = get->addr };
2524 svc = __ip_vs_svc_fwm_find(net, AF_INET, get->fwmark);
2526 svc = __ip_vs_service_find(net, AF_INET, get->protocol, &addr,
2532 struct ip_vs_dest *dest;
2533 struct ip_vs_dest_entry entry;
2534 struct ip_vs_kstats kstats;
2536 memset(&entry, 0, sizeof(entry));
2537 list_for_each_entry(dest, &svc->destinations, n_list) {
2538 if (count >= get->num_dests)
2541 /* Cannot expose heterogeneous members via sockopt
2544 if (dest->af != svc->af)
2547 entry.addr = dest->addr.ip;
2548 entry.port = dest->port;
2549 entry.conn_flags = atomic_read(&dest->conn_flags);
2550 entry.weight = atomic_read(&dest->weight);
2551 entry.u_threshold = dest->u_threshold;
2552 entry.l_threshold = dest->l_threshold;
2553 entry.activeconns = atomic_read(&dest->activeconns);
2554 entry.inactconns = atomic_read(&dest->inactconns);
2555 entry.persistconns = atomic_read(&dest->persistconns);
2556 ip_vs_copy_stats(&kstats, &dest->stats);
2557 ip_vs_export_stats_user(&entry.stats, &kstats);
2558 if (copy_to_user(&uptr->entrytable[count],
2559 &entry, sizeof(entry))) {
2571 __ip_vs_get_timeouts(struct net *net, struct ip_vs_timeout_user *u)
2573 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2574 struct ip_vs_proto_data *pd;
2577 memset(u, 0, sizeof (*u));
2579 #ifdef CONFIG_IP_VS_PROTO_TCP
2580 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2581 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2582 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2584 #ifdef CONFIG_IP_VS_PROTO_UDP
2585 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2587 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2591 static const unsigned char get_arglen[CMDID(IP_VS_SO_GET_MAX) + 1] = {
2592 [CMDID(IP_VS_SO_GET_VERSION)] = 64,
2593 [CMDID(IP_VS_SO_GET_INFO)] = sizeof(struct ip_vs_getinfo),
2594 [CMDID(IP_VS_SO_GET_SERVICES)] = sizeof(struct ip_vs_get_services),
2595 [CMDID(IP_VS_SO_GET_SERVICE)] = sizeof(struct ip_vs_service_entry),
2596 [CMDID(IP_VS_SO_GET_DESTS)] = sizeof(struct ip_vs_get_dests),
2597 [CMDID(IP_VS_SO_GET_TIMEOUT)] = sizeof(struct ip_vs_timeout_user),
2598 [CMDID(IP_VS_SO_GET_DAEMON)] = 2 * sizeof(struct ip_vs_daemon_user),
2601 union ip_vs_get_arglen {
2602 char field_IP_VS_SO_GET_VERSION[64];
2603 struct ip_vs_getinfo field_IP_VS_SO_GET_INFO;
2604 struct ip_vs_get_services field_IP_VS_SO_GET_SERVICES;
2605 struct ip_vs_service_entry field_IP_VS_SO_GET_SERVICE;
2606 struct ip_vs_get_dests field_IP_VS_SO_GET_DESTS;
2607 struct ip_vs_timeout_user field_IP_VS_SO_GET_TIMEOUT;
2608 struct ip_vs_daemon_user field_IP_VS_SO_GET_DAEMON[2];
2611 #define MAX_GET_ARGLEN sizeof(union ip_vs_get_arglen)
2614 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2616 unsigned char arg[MAX_GET_ARGLEN];
2618 unsigned int copylen;
2619 struct net *net = sock_net(sk);
2620 struct netns_ipvs *ipvs = net_ipvs(net);
2623 BUILD_BUG_ON(sizeof(arg) > 255);
2624 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2627 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2630 copylen = get_arglen[CMDID(cmd)];
2631 if (*len < (int) copylen) {
2632 IP_VS_DBG(1, "get_ctl: len %d < %u\n", *len, copylen);
2636 if (copy_from_user(arg, user, copylen) != 0)
2639 * Handle daemons first since it has its own locking
2641 if (cmd == IP_VS_SO_GET_DAEMON) {
2642 struct ip_vs_daemon_user d[2];
2644 memset(&d, 0, sizeof(d));
2645 mutex_lock(&ipvs->sync_mutex);
2646 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2647 d[0].state = IP_VS_STATE_MASTER;
2648 strlcpy(d[0].mcast_ifn, ipvs->master_mcast_ifn,
2649 sizeof(d[0].mcast_ifn));
2650 d[0].syncid = ipvs->master_syncid;
2652 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2653 d[1].state = IP_VS_STATE_BACKUP;
2654 strlcpy(d[1].mcast_ifn, ipvs->backup_mcast_ifn,
2655 sizeof(d[1].mcast_ifn));
2656 d[1].syncid = ipvs->backup_syncid;
2658 if (copy_to_user(user, &d, sizeof(d)) != 0)
2660 mutex_unlock(&ipvs->sync_mutex);
2664 mutex_lock(&__ip_vs_mutex);
2666 case IP_VS_SO_GET_VERSION:
2670 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2671 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2672 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2676 *len = strlen(buf)+1;
2680 case IP_VS_SO_GET_INFO:
2682 struct ip_vs_getinfo info;
2683 info.version = IP_VS_VERSION_CODE;
2684 info.size = ip_vs_conn_tab_size;
2685 info.num_services = ipvs->num_services;
2686 if (copy_to_user(user, &info, sizeof(info)) != 0)
2691 case IP_VS_SO_GET_SERVICES:
2693 struct ip_vs_get_services *get;
2696 get = (struct ip_vs_get_services *)arg;
2697 size = sizeof(*get) +
2698 sizeof(struct ip_vs_service_entry) * get->num_services;
2700 pr_err("length: %u != %u\n", *len, size);
2704 ret = __ip_vs_get_service_entries(net, get, user);
2708 case IP_VS_SO_GET_SERVICE:
2710 struct ip_vs_service_entry *entry;
2711 struct ip_vs_service *svc;
2712 union nf_inet_addr addr;
2714 entry = (struct ip_vs_service_entry *)arg;
2715 addr.ip = entry->addr;
2718 svc = __ip_vs_svc_fwm_find(net, AF_INET, entry->fwmark);
2720 svc = __ip_vs_service_find(net, AF_INET,
2721 entry->protocol, &addr,
2725 ip_vs_copy_service(entry, svc);
2726 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2733 case IP_VS_SO_GET_DESTS:
2735 struct ip_vs_get_dests *get;
2738 get = (struct ip_vs_get_dests *)arg;
2739 size = sizeof(*get) +
2740 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2742 pr_err("length: %u != %u\n", *len, size);
2746 ret = __ip_vs_get_dest_entries(net, get, user);
2750 case IP_VS_SO_GET_TIMEOUT:
2752 struct ip_vs_timeout_user t;
2754 __ip_vs_get_timeouts(net, &t);
2755 if (copy_to_user(user, &t, sizeof(t)) != 0)
2765 mutex_unlock(&__ip_vs_mutex);
2770 static struct nf_sockopt_ops ip_vs_sockopts = {
2772 .set_optmin = IP_VS_BASE_CTL,
2773 .set_optmax = IP_VS_SO_SET_MAX+1,
2774 .set = do_ip_vs_set_ctl,
2775 .get_optmin = IP_VS_BASE_CTL,
2776 .get_optmax = IP_VS_SO_GET_MAX+1,
2777 .get = do_ip_vs_get_ctl,
2778 .owner = THIS_MODULE,
2782 * Generic Netlink interface
2785 /* IPVS genetlink family */
2786 static struct genl_family ip_vs_genl_family = {
2787 .id = GENL_ID_GENERATE,
2789 .name = IPVS_GENL_NAME,
2790 .version = IPVS_GENL_VERSION,
2791 .maxattr = IPVS_CMD_MAX,
2792 .netnsok = true, /* Make ipvsadm to work on netns */
2795 /* Policy used for first-level command attributes */
2796 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2797 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2798 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2799 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2800 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2801 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2802 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2805 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2806 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2807 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2808 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2809 .len = IP_VS_IFNAME_MAXLEN },
2810 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2813 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2814 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2815 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2816 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2817 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2818 .len = sizeof(union nf_inet_addr) },
2819 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2820 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2821 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2822 .len = IP_VS_SCHEDNAME_MAXLEN },
2823 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2824 .len = IP_VS_PENAME_MAXLEN },
2825 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2826 .len = sizeof(struct ip_vs_flags) },
2827 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2828 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2829 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2832 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2833 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2834 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2835 .len = sizeof(union nf_inet_addr) },
2836 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2837 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2838 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2839 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2840 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2841 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2842 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2843 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2844 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2845 [IPVS_DEST_ATTR_ADDR_FAMILY] = { .type = NLA_U16 },
2848 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2849 struct ip_vs_kstats *kstats)
2851 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2856 if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, (u32)kstats->conns) ||
2857 nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, (u32)kstats->inpkts) ||
2858 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, (u32)kstats->outpkts) ||
2859 nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes) ||
2860 nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes) ||
2861 nla_put_u32(skb, IPVS_STATS_ATTR_CPS, (u32)kstats->cps) ||
2862 nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, (u32)kstats->inpps) ||
2863 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, (u32)kstats->outpps) ||
2864 nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, (u32)kstats->inbps) ||
2865 nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, (u32)kstats->outbps))
2866 goto nla_put_failure;
2867 nla_nest_end(skb, nl_stats);
2872 nla_nest_cancel(skb, nl_stats);
2876 static int ip_vs_genl_fill_stats64(struct sk_buff *skb, int container_type,
2877 struct ip_vs_kstats *kstats)
2879 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2884 if (nla_put_u64(skb, IPVS_STATS_ATTR_CONNS, kstats->conns) ||
2885 nla_put_u64(skb, IPVS_STATS_ATTR_INPKTS, kstats->inpkts) ||
2886 nla_put_u64(skb, IPVS_STATS_ATTR_OUTPKTS, kstats->outpkts) ||
2887 nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes) ||
2888 nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes) ||
2889 nla_put_u64(skb, IPVS_STATS_ATTR_CPS, kstats->cps) ||
2890 nla_put_u64(skb, IPVS_STATS_ATTR_INPPS, kstats->inpps) ||
2891 nla_put_u64(skb, IPVS_STATS_ATTR_OUTPPS, kstats->outpps) ||
2892 nla_put_u64(skb, IPVS_STATS_ATTR_INBPS, kstats->inbps) ||
2893 nla_put_u64(skb, IPVS_STATS_ATTR_OUTBPS, kstats->outbps))
2894 goto nla_put_failure;
2895 nla_nest_end(skb, nl_stats);
2900 nla_nest_cancel(skb, nl_stats);
2904 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2905 struct ip_vs_service *svc)
2907 struct ip_vs_scheduler *sched;
2908 struct ip_vs_pe *pe;
2909 struct nlattr *nl_service;
2910 struct ip_vs_flags flags = { .flags = svc->flags,
2912 struct ip_vs_kstats kstats;
2915 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2919 if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2920 goto nla_put_failure;
2922 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2923 goto nla_put_failure;
2925 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2926 nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2927 nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2928 goto nla_put_failure;
2931 sched = rcu_dereference_protected(svc->scheduler, 1);
2932 sched_name = sched ? sched->name : "none";
2933 pe = rcu_dereference_protected(svc->pe, 1);
2934 if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched_name) ||
2935 (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
2936 nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2937 nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2938 nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
2939 goto nla_put_failure;
2940 ip_vs_copy_stats(&kstats, &svc->stats);
2941 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &kstats))
2942 goto nla_put_failure;
2943 if (ip_vs_genl_fill_stats64(skb, IPVS_SVC_ATTR_STATS64, &kstats))
2944 goto nla_put_failure;
2946 nla_nest_end(skb, nl_service);
2951 nla_nest_cancel(skb, nl_service);
2955 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2956 struct ip_vs_service *svc,
2957 struct netlink_callback *cb)
2961 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2962 &ip_vs_genl_family, NLM_F_MULTI,
2963 IPVS_CMD_NEW_SERVICE);
2967 if (ip_vs_genl_fill_service(skb, svc) < 0)
2968 goto nla_put_failure;
2970 genlmsg_end(skb, hdr);
2974 genlmsg_cancel(skb, hdr);
2978 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2979 struct netlink_callback *cb)
2982 int start = cb->args[0];
2983 struct ip_vs_service *svc;
2984 struct net *net = skb_sknet(skb);
2986 mutex_lock(&__ip_vs_mutex);
2987 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2988 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2989 if (++idx <= start || !net_eq(svc->net, net))
2991 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2993 goto nla_put_failure;
2998 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2999 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
3000 if (++idx <= start || !net_eq(svc->net, net))
3002 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3004 goto nla_put_failure;
3010 mutex_unlock(&__ip_vs_mutex);
3016 static int ip_vs_genl_parse_service(struct net *net,
3017 struct ip_vs_service_user_kern *usvc,
3018 struct nlattr *nla, int full_entry,
3019 struct ip_vs_service **ret_svc)
3021 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
3022 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
3023 struct ip_vs_service *svc;
3025 /* Parse mandatory identifying service fields first */
3027 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
3030 nla_af = attrs[IPVS_SVC_ATTR_AF];
3031 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
3032 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
3033 nla_port = attrs[IPVS_SVC_ATTR_PORT];
3034 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
3036 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3039 memset(usvc, 0, sizeof(*usvc));
3041 usvc->af = nla_get_u16(nla_af);
3042 #ifdef CONFIG_IP_VS_IPV6
3043 if (usvc->af != AF_INET && usvc->af != AF_INET6)
3045 if (usvc->af != AF_INET)
3047 return -EAFNOSUPPORT;
3050 usvc->protocol = IPPROTO_TCP;
3051 usvc->fwmark = nla_get_u32(nla_fwmark);
3053 usvc->protocol = nla_get_u16(nla_protocol);
3054 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3055 usvc->port = nla_get_be16(nla_port);
3061 svc = __ip_vs_svc_fwm_find(net, usvc->af, usvc->fwmark);
3063 svc = __ip_vs_service_find(net, usvc->af, usvc->protocol,
3064 &usvc->addr, usvc->port);
3068 /* If a full entry was requested, check for the additional fields */
3070 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3072 struct ip_vs_flags flags;
3074 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3075 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3076 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3077 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3078 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3080 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3083 nla_memcpy(&flags, nla_flags, sizeof(flags));
3085 /* prefill flags from service if it already exists */
3087 usvc->flags = svc->flags;
3089 /* set new flags from userland */
3090 usvc->flags = (usvc->flags & ~flags.mask) |
3091 (flags.flags & flags.mask);
3092 usvc->sched_name = nla_data(nla_sched);
3093 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3094 usvc->timeout = nla_get_u32(nla_timeout);
3095 usvc->netmask = nla_get_be32(nla_netmask);
3101 static struct ip_vs_service *ip_vs_genl_find_service(struct net *net,
3104 struct ip_vs_service_user_kern usvc;
3105 struct ip_vs_service *svc;
3108 ret = ip_vs_genl_parse_service(net, &usvc, nla, 0, &svc);
3109 return ret ? ERR_PTR(ret) : svc;
3112 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3114 struct nlattr *nl_dest;
3115 struct ip_vs_kstats kstats;
3117 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3121 if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3122 nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3123 nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3124 (atomic_read(&dest->conn_flags) &
3125 IP_VS_CONN_F_FWD_MASK)) ||
3126 nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3127 atomic_read(&dest->weight)) ||
3128 nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3129 nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3130 nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3131 atomic_read(&dest->activeconns)) ||
3132 nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3133 atomic_read(&dest->inactconns)) ||
3134 nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3135 atomic_read(&dest->persistconns)) ||
3136 nla_put_u16(skb, IPVS_DEST_ATTR_ADDR_FAMILY, dest->af))
3137 goto nla_put_failure;
3138 ip_vs_copy_stats(&kstats, &dest->stats);
3139 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &kstats))
3140 goto nla_put_failure;
3141 if (ip_vs_genl_fill_stats64(skb, IPVS_DEST_ATTR_STATS64, &kstats))
3142 goto nla_put_failure;
3144 nla_nest_end(skb, nl_dest);
3149 nla_nest_cancel(skb, nl_dest);
3153 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3154 struct netlink_callback *cb)
3158 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3159 &ip_vs_genl_family, NLM_F_MULTI,
3164 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3165 goto nla_put_failure;
3167 genlmsg_end(skb, hdr);
3171 genlmsg_cancel(skb, hdr);
3175 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3176 struct netlink_callback *cb)
3179 int start = cb->args[0];
3180 struct ip_vs_service *svc;
3181 struct ip_vs_dest *dest;
3182 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3183 struct net *net = skb_sknet(skb);
3185 mutex_lock(&__ip_vs_mutex);
3187 /* Try to find the service for which to dump destinations */
3188 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3189 IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3193 svc = ip_vs_genl_find_service(net, attrs[IPVS_CMD_ATTR_SERVICE]);
3194 if (IS_ERR(svc) || svc == NULL)
3197 /* Dump the destinations */
3198 list_for_each_entry(dest, &svc->destinations, n_list) {
3201 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3203 goto nla_put_failure;
3211 mutex_unlock(&__ip_vs_mutex);
3216 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3217 struct nlattr *nla, int full_entry)
3219 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3220 struct nlattr *nla_addr, *nla_port;
3221 struct nlattr *nla_addr_family;
3223 /* Parse mandatory identifying destination fields first */
3225 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3228 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3229 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3230 nla_addr_family = attrs[IPVS_DEST_ATTR_ADDR_FAMILY];
3232 if (!(nla_addr && nla_port))
3235 memset(udest, 0, sizeof(*udest));
3237 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3238 udest->port = nla_get_be16(nla_port);
3240 if (nla_addr_family)
3241 udest->af = nla_get_u16(nla_addr_family);
3245 /* If a full entry was requested, check for the additional fields */
3247 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3250 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3251 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3252 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3253 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3255 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3258 udest->conn_flags = nla_get_u32(nla_fwd)
3259 & IP_VS_CONN_F_FWD_MASK;
3260 udest->weight = nla_get_u32(nla_weight);
3261 udest->u_threshold = nla_get_u32(nla_u_thresh);
3262 udest->l_threshold = nla_get_u32(nla_l_thresh);
3268 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3269 const char *mcast_ifn, __u32 syncid)
3271 struct nlattr *nl_daemon;
3273 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3277 if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3278 nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn) ||
3279 nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid))
3280 goto nla_put_failure;
3281 nla_nest_end(skb, nl_daemon);
3286 nla_nest_cancel(skb, nl_daemon);
3290 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3291 const char *mcast_ifn, __u32 syncid,
3292 struct netlink_callback *cb)
3295 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3296 &ip_vs_genl_family, NLM_F_MULTI,
3297 IPVS_CMD_NEW_DAEMON);
3301 if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
3302 goto nla_put_failure;
3304 genlmsg_end(skb, hdr);
3308 genlmsg_cancel(skb, hdr);
3312 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3313 struct netlink_callback *cb)
3315 struct net *net = skb_sknet(skb);
3316 struct netns_ipvs *ipvs = net_ipvs(net);
3318 mutex_lock(&ipvs->sync_mutex);
3319 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3320 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3321 ipvs->master_mcast_ifn,
3322 ipvs->master_syncid, cb) < 0)
3323 goto nla_put_failure;
3328 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3329 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3330 ipvs->backup_mcast_ifn,
3331 ipvs->backup_syncid, cb) < 0)
3332 goto nla_put_failure;
3338 mutex_unlock(&ipvs->sync_mutex);
3343 static int ip_vs_genl_new_daemon(struct net *net, struct nlattr **attrs)
3345 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3346 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3347 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3350 /* The synchronization protocol is incompatible with mixed family
3353 if (net_ipvs(net)->mixed_address_family_dests > 0)
3356 return start_sync_thread(net,
3357 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3358 nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3359 nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3362 static int ip_vs_genl_del_daemon(struct net *net, struct nlattr **attrs)
3364 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3367 return stop_sync_thread(net,
3368 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3371 static int ip_vs_genl_set_config(struct net *net, struct nlattr **attrs)
3373 struct ip_vs_timeout_user t;
3375 __ip_vs_get_timeouts(net, &t);
3377 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3378 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3380 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3382 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3384 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3385 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3387 return ip_vs_set_timeout(net, &t);
3390 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3394 struct netns_ipvs *ipvs;
3396 net = skb_sknet(skb);
3397 ipvs = net_ipvs(net);
3398 cmd = info->genlhdr->cmd;
3400 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3401 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3403 mutex_lock(&ipvs->sync_mutex);
3404 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3405 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3406 info->attrs[IPVS_CMD_ATTR_DAEMON],
3407 ip_vs_daemon_policy)) {
3412 if (cmd == IPVS_CMD_NEW_DAEMON)
3413 ret = ip_vs_genl_new_daemon(net, daemon_attrs);
3415 ret = ip_vs_genl_del_daemon(net, daemon_attrs);
3417 mutex_unlock(&ipvs->sync_mutex);
3422 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3424 struct ip_vs_service *svc = NULL;
3425 struct ip_vs_service_user_kern usvc;
3426 struct ip_vs_dest_user_kern udest;
3428 int need_full_svc = 0, need_full_dest = 0;
3431 net = skb_sknet(skb);
3432 cmd = info->genlhdr->cmd;
3434 mutex_lock(&__ip_vs_mutex);
3436 if (cmd == IPVS_CMD_FLUSH) {
3437 ret = ip_vs_flush(net, false);
3439 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3440 ret = ip_vs_genl_set_config(net, info->attrs);
3442 } else if (cmd == IPVS_CMD_ZERO &&
3443 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3444 ret = ip_vs_zero_all(net);
3448 /* All following commands require a service argument, so check if we
3449 * received a valid one. We need a full service specification when
3450 * adding / editing a service. Only identifying members otherwise. */
3451 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3454 ret = ip_vs_genl_parse_service(net, &usvc,
3455 info->attrs[IPVS_CMD_ATTR_SERVICE],
3456 need_full_svc, &svc);
3460 /* Unless we're adding a new service, the service must already exist */
3461 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3466 /* Destination commands require a valid destination argument. For
3467 * adding / editing a destination, we need a full destination
3469 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3470 cmd == IPVS_CMD_DEL_DEST) {
3471 if (cmd != IPVS_CMD_DEL_DEST)
3474 ret = ip_vs_genl_parse_dest(&udest,
3475 info->attrs[IPVS_CMD_ATTR_DEST],
3480 /* Old protocols did not allow the user to specify address
3481 * family, so we set it to zero instead. We also didn't
3482 * allow heterogeneous pools in the old code, so it's safe
3483 * to assume that this will have the same address family as
3489 if (udest.af != svc->af && cmd != IPVS_CMD_DEL_DEST) {
3490 /* The synchronization protocol is incompatible
3491 * with mixed family services
3493 if (net_ipvs(net)->sync_state) {
3498 /* Which connection types do we support? */
3499 switch (udest.conn_flags) {
3500 case IP_VS_CONN_F_TUNNEL:
3501 /* We are able to forward this */
3511 case IPVS_CMD_NEW_SERVICE:
3513 ret = ip_vs_add_service(net, &usvc, &svc);
3517 case IPVS_CMD_SET_SERVICE:
3518 ret = ip_vs_edit_service(svc, &usvc);
3520 case IPVS_CMD_DEL_SERVICE:
3521 ret = ip_vs_del_service(svc);
3522 /* do not use svc, it can be freed */
3524 case IPVS_CMD_NEW_DEST:
3525 ret = ip_vs_add_dest(svc, &udest);
3527 case IPVS_CMD_SET_DEST:
3528 ret = ip_vs_edit_dest(svc, &udest);
3530 case IPVS_CMD_DEL_DEST:
3531 ret = ip_vs_del_dest(svc, &udest);
3534 ret = ip_vs_zero_service(svc);
3541 mutex_unlock(&__ip_vs_mutex);
3546 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3548 struct sk_buff *msg;
3550 int ret, cmd, reply_cmd;
3553 net = skb_sknet(skb);
3554 cmd = info->genlhdr->cmd;
3556 if (cmd == IPVS_CMD_GET_SERVICE)
3557 reply_cmd = IPVS_CMD_NEW_SERVICE;
3558 else if (cmd == IPVS_CMD_GET_INFO)
3559 reply_cmd = IPVS_CMD_SET_INFO;
3560 else if (cmd == IPVS_CMD_GET_CONFIG)
3561 reply_cmd = IPVS_CMD_SET_CONFIG;
3563 pr_err("unknown Generic Netlink command\n");
3567 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3571 mutex_lock(&__ip_vs_mutex);
3573 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3575 goto nla_put_failure;
3578 case IPVS_CMD_GET_SERVICE:
3580 struct ip_vs_service *svc;
3582 svc = ip_vs_genl_find_service(net,
3583 info->attrs[IPVS_CMD_ATTR_SERVICE]);
3588 ret = ip_vs_genl_fill_service(msg, svc);
3590 goto nla_put_failure;
3599 case IPVS_CMD_GET_CONFIG:
3601 struct ip_vs_timeout_user t;
3603 __ip_vs_get_timeouts(net, &t);
3604 #ifdef CONFIG_IP_VS_PROTO_TCP
3605 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3607 nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3609 goto nla_put_failure;
3611 #ifdef CONFIG_IP_VS_PROTO_UDP
3612 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3613 goto nla_put_failure;
3619 case IPVS_CMD_GET_INFO:
3620 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3621 IP_VS_VERSION_CODE) ||
3622 nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3623 ip_vs_conn_tab_size))
3624 goto nla_put_failure;
3628 genlmsg_end(msg, reply);
3629 ret = genlmsg_reply(msg, info);
3633 pr_err("not enough space in Netlink message\n");
3639 mutex_unlock(&__ip_vs_mutex);
3645 static const struct genl_ops ip_vs_genl_ops[] = {
3647 .cmd = IPVS_CMD_NEW_SERVICE,
3648 .flags = GENL_ADMIN_PERM,
3649 .policy = ip_vs_cmd_policy,
3650 .doit = ip_vs_genl_set_cmd,
3653 .cmd = IPVS_CMD_SET_SERVICE,
3654 .flags = GENL_ADMIN_PERM,
3655 .policy = ip_vs_cmd_policy,
3656 .doit = ip_vs_genl_set_cmd,
3659 .cmd = IPVS_CMD_DEL_SERVICE,
3660 .flags = GENL_ADMIN_PERM,
3661 .policy = ip_vs_cmd_policy,
3662 .doit = ip_vs_genl_set_cmd,
3665 .cmd = IPVS_CMD_GET_SERVICE,
3666 .flags = GENL_ADMIN_PERM,
3667 .doit = ip_vs_genl_get_cmd,
3668 .dumpit = ip_vs_genl_dump_services,
3669 .policy = ip_vs_cmd_policy,
3672 .cmd = IPVS_CMD_NEW_DEST,
3673 .flags = GENL_ADMIN_PERM,
3674 .policy = ip_vs_cmd_policy,
3675 .doit = ip_vs_genl_set_cmd,
3678 .cmd = IPVS_CMD_SET_DEST,
3679 .flags = GENL_ADMIN_PERM,
3680 .policy = ip_vs_cmd_policy,
3681 .doit = ip_vs_genl_set_cmd,
3684 .cmd = IPVS_CMD_DEL_DEST,
3685 .flags = GENL_ADMIN_PERM,
3686 .policy = ip_vs_cmd_policy,
3687 .doit = ip_vs_genl_set_cmd,
3690 .cmd = IPVS_CMD_GET_DEST,
3691 .flags = GENL_ADMIN_PERM,
3692 .policy = ip_vs_cmd_policy,
3693 .dumpit = ip_vs_genl_dump_dests,
3696 .cmd = IPVS_CMD_NEW_DAEMON,
3697 .flags = GENL_ADMIN_PERM,
3698 .policy = ip_vs_cmd_policy,
3699 .doit = ip_vs_genl_set_daemon,
3702 .cmd = IPVS_CMD_DEL_DAEMON,
3703 .flags = GENL_ADMIN_PERM,
3704 .policy = ip_vs_cmd_policy,
3705 .doit = ip_vs_genl_set_daemon,
3708 .cmd = IPVS_CMD_GET_DAEMON,
3709 .flags = GENL_ADMIN_PERM,
3710 .dumpit = ip_vs_genl_dump_daemons,
3713 .cmd = IPVS_CMD_SET_CONFIG,
3714 .flags = GENL_ADMIN_PERM,
3715 .policy = ip_vs_cmd_policy,
3716 .doit = ip_vs_genl_set_cmd,
3719 .cmd = IPVS_CMD_GET_CONFIG,
3720 .flags = GENL_ADMIN_PERM,
3721 .doit = ip_vs_genl_get_cmd,
3724 .cmd = IPVS_CMD_GET_INFO,
3725 .flags = GENL_ADMIN_PERM,
3726 .doit = ip_vs_genl_get_cmd,
3729 .cmd = IPVS_CMD_ZERO,
3730 .flags = GENL_ADMIN_PERM,
3731 .policy = ip_vs_cmd_policy,
3732 .doit = ip_vs_genl_set_cmd,
3735 .cmd = IPVS_CMD_FLUSH,
3736 .flags = GENL_ADMIN_PERM,
3737 .doit = ip_vs_genl_set_cmd,
3741 static int __init ip_vs_genl_register(void)
3743 return genl_register_family_with_ops(&ip_vs_genl_family,
3747 static void ip_vs_genl_unregister(void)
3749 genl_unregister_family(&ip_vs_genl_family);
3752 /* End of Generic Netlink interface definitions */
3755 * per netns intit/exit func.
3757 #ifdef CONFIG_SYSCTL
3758 static int __net_init ip_vs_control_net_init_sysctl(struct net *net)
3761 struct netns_ipvs *ipvs = net_ipvs(net);
3762 struct ctl_table *tbl;
3764 atomic_set(&ipvs->dropentry, 0);
3765 spin_lock_init(&ipvs->dropentry_lock);
3766 spin_lock_init(&ipvs->droppacket_lock);
3767 spin_lock_init(&ipvs->securetcp_lock);
3769 if (!net_eq(net, &init_net)) {
3770 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3774 /* Don't export sysctls to unprivileged users */
3775 if (net->user_ns != &init_user_ns)
3776 tbl[0].procname = NULL;
3779 /* Initialize sysctl defaults */
3781 ipvs->sysctl_amemthresh = 1024;
3782 tbl[idx++].data = &ipvs->sysctl_amemthresh;
3783 ipvs->sysctl_am_droprate = 10;
3784 tbl[idx++].data = &ipvs->sysctl_am_droprate;
3785 tbl[idx++].data = &ipvs->sysctl_drop_entry;
3786 tbl[idx++].data = &ipvs->sysctl_drop_packet;
3787 #ifdef CONFIG_IP_VS_NFCT
3788 tbl[idx++].data = &ipvs->sysctl_conntrack;
3790 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3791 ipvs->sysctl_snat_reroute = 1;
3792 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3793 ipvs->sysctl_sync_ver = 1;
3794 tbl[idx++].data = &ipvs->sysctl_sync_ver;
3795 ipvs->sysctl_sync_ports = 1;
3796 tbl[idx++].data = &ipvs->sysctl_sync_ports;
3797 tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
3798 ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3799 tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3800 ipvs->sysctl_sync_sock_size = 0;
3801 tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3802 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3803 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3804 tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
3805 tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
3806 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3807 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3808 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3809 tbl[idx].data = &ipvs->sysctl_sync_threshold;
3810 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3811 ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3812 tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3813 ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3814 tbl[idx++].data = &ipvs->sysctl_sync_retries;
3815 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3816 ipvs->sysctl_pmtu_disc = 1;
3817 tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
3818 tbl[idx++].data = &ipvs->sysctl_backup_only;
3819 ipvs->sysctl_conn_reuse_mode = 1;
3820 tbl[idx++].data = &ipvs->sysctl_conn_reuse_mode;
3823 ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3824 if (ipvs->sysctl_hdr == NULL) {
3825 if (!net_eq(net, &init_net))
3829 ip_vs_start_estimator(net, &ipvs->tot_stats);
3830 ipvs->sysctl_tbl = tbl;
3831 /* Schedule defense work */
3832 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3833 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3838 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net)
3840 struct netns_ipvs *ipvs = net_ipvs(net);
3842 cancel_delayed_work_sync(&ipvs->defense_work);
3843 cancel_work_sync(&ipvs->defense_work.work);
3844 unregister_net_sysctl_table(ipvs->sysctl_hdr);
3845 ip_vs_stop_estimator(net, &ipvs->tot_stats);
3847 if (!net_eq(net, &init_net))
3848 kfree(ipvs->sysctl_tbl);
3853 static int __net_init ip_vs_control_net_init_sysctl(struct net *net) { return 0; }
3854 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net) { }
3858 static struct notifier_block ip_vs_dst_notifier = {
3859 .notifier_call = ip_vs_dst_event,
3862 int __net_init ip_vs_control_net_init(struct net *net)
3865 struct netns_ipvs *ipvs = net_ipvs(net);
3867 /* Initialize rs_table */
3868 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
3869 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
3871 INIT_LIST_HEAD(&ipvs->dest_trash);
3872 spin_lock_init(&ipvs->dest_trash_lock);
3873 setup_timer(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire,
3874 (unsigned long) net);
3875 atomic_set(&ipvs->ftpsvc_counter, 0);
3876 atomic_set(&ipvs->nullsvc_counter, 0);
3879 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
3880 if (!ipvs->tot_stats.cpustats)
3883 for_each_possible_cpu(i) {
3884 struct ip_vs_cpu_stats *ipvs_tot_stats;
3885 ipvs_tot_stats = per_cpu_ptr(ipvs->tot_stats.cpustats, i);
3886 u64_stats_init(&ipvs_tot_stats->syncp);
3889 spin_lock_init(&ipvs->tot_stats.lock);
3891 proc_create("ip_vs", 0, net->proc_net, &ip_vs_info_fops);
3892 proc_create("ip_vs_stats", 0, net->proc_net, &ip_vs_stats_fops);
3893 proc_create("ip_vs_stats_percpu", 0, net->proc_net,
3894 &ip_vs_stats_percpu_fops);
3896 if (ip_vs_control_net_init_sysctl(net))
3902 free_percpu(ipvs->tot_stats.cpustats);
3906 void __net_exit ip_vs_control_net_cleanup(struct net *net)
3908 struct netns_ipvs *ipvs = net_ipvs(net);
3910 ip_vs_trash_cleanup(net);
3911 ip_vs_control_net_cleanup_sysctl(net);
3912 remove_proc_entry("ip_vs_stats_percpu", net->proc_net);
3913 remove_proc_entry("ip_vs_stats", net->proc_net);
3914 remove_proc_entry("ip_vs", net->proc_net);
3915 free_percpu(ipvs->tot_stats.cpustats);
3918 int __init ip_vs_register_nl_ioctl(void)
3922 ret = nf_register_sockopt(&ip_vs_sockopts);
3924 pr_err("cannot register sockopt.\n");
3928 ret = ip_vs_genl_register();
3930 pr_err("cannot register Generic Netlink interface.\n");
3936 nf_unregister_sockopt(&ip_vs_sockopts);
3941 void ip_vs_unregister_nl_ioctl(void)
3943 ip_vs_genl_unregister();
3944 nf_unregister_sockopt(&ip_vs_sockopts);
3947 int __init ip_vs_control_init(void)
3954 /* Initialize svc_table, ip_vs_svc_fwm_table */
3955 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
3956 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
3957 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3960 smp_wmb(); /* Do we really need it now ? */
3962 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
3971 void ip_vs_control_cleanup(void)
3974 unregister_netdevice_notifier(&ip_vs_dst_notifier);