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 if (svc->stats.cpustats)
469 free_percpu(svc->stats.cpustats);
473 static void ip_vs_service_rcu_free(struct rcu_head *head)
475 struct ip_vs_service *svc;
477 svc = container_of(head, struct ip_vs_service, rcu_head);
478 ip_vs_service_free(svc);
481 static void __ip_vs_svc_put(struct ip_vs_service *svc, bool do_delay)
483 if (atomic_dec_and_test(&svc->refcnt)) {
484 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n",
486 IP_VS_DBG_ADDR(svc->af, &svc->addr),
489 call_rcu(&svc->rcu_head, ip_vs_service_rcu_free);
491 ip_vs_service_free(svc);
497 * Returns hash value for real service
499 static inline unsigned int ip_vs_rs_hashkey(int af,
500 const union nf_inet_addr *addr,
503 register unsigned int porth = ntohs(port);
504 __be32 addr_fold = addr->ip;
506 #ifdef CONFIG_IP_VS_IPV6
508 addr_fold = addr->ip6[0]^addr->ip6[1]^
509 addr->ip6[2]^addr->ip6[3];
512 return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
516 /* Hash ip_vs_dest in rs_table by <proto,addr,port>. */
517 static void ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
521 if (dest->in_rs_table)
525 * Hash by proto,addr,port,
526 * which are the parameters of the real service.
528 hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
530 hlist_add_head_rcu(&dest->d_list, &ipvs->rs_table[hash]);
531 dest->in_rs_table = 1;
534 /* Unhash ip_vs_dest from rs_table. */
535 static void ip_vs_rs_unhash(struct ip_vs_dest *dest)
538 * Remove it from the rs_table table.
540 if (dest->in_rs_table) {
541 hlist_del_rcu(&dest->d_list);
542 dest->in_rs_table = 0;
546 /* Check if real service by <proto,addr,port> is present */
547 bool ip_vs_has_real_service(struct net *net, int af, __u16 protocol,
548 const union nf_inet_addr *daddr, __be16 dport)
550 struct netns_ipvs *ipvs = net_ipvs(net);
552 struct ip_vs_dest *dest;
554 /* Check for "full" addressed entries */
555 hash = ip_vs_rs_hashkey(af, daddr, dport);
558 hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
559 if (dest->port == dport &&
561 ip_vs_addr_equal(af, &dest->addr, daddr) &&
562 (dest->protocol == protocol || dest->vfwmark)) {
573 /* Lookup destination by {addr,port} in the given service
574 * Called under RCU lock.
576 static struct ip_vs_dest *
577 ip_vs_lookup_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
580 struct ip_vs_dest *dest;
583 * Find the destination for the given service
585 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
586 if ((dest->af == svc->af)
587 && ip_vs_addr_equal(svc->af, &dest->addr, daddr)
588 && (dest->port == dport)) {
598 * Find destination by {daddr,dport,vaddr,protocol}
599 * Created to be used in ip_vs_process_message() in
600 * the backup synchronization daemon. It finds the
601 * destination to be bound to the received connection
603 * Called under RCU lock, no refcnt is returned.
605 struct ip_vs_dest *ip_vs_find_dest(struct net *net, int af,
606 const union nf_inet_addr *daddr,
608 const union nf_inet_addr *vaddr,
609 __be16 vport, __u16 protocol, __u32 fwmark,
612 struct ip_vs_dest *dest;
613 struct ip_vs_service *svc;
616 svc = ip_vs_service_find(net, af, fwmark, protocol, vaddr, vport);
619 if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
621 dest = ip_vs_lookup_dest(svc, daddr, port);
623 dest = ip_vs_lookup_dest(svc, daddr, port ^ dport);
627 void ip_vs_dest_dst_rcu_free(struct rcu_head *head)
629 struct ip_vs_dest_dst *dest_dst = container_of(head,
630 struct ip_vs_dest_dst,
633 dst_release(dest_dst->dst_cache);
637 /* Release dest_dst and dst_cache for dest in user context */
638 static void __ip_vs_dst_cache_reset(struct ip_vs_dest *dest)
640 struct ip_vs_dest_dst *old;
642 old = rcu_dereference_protected(dest->dest_dst, 1);
644 RCU_INIT_POINTER(dest->dest_dst, NULL);
645 call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
650 * Lookup dest by {svc,addr,port} in the destination trash.
651 * The destination trash is used to hold the destinations that are removed
652 * from the service table but are still referenced by some conn entries.
653 * The reason to add the destination trash is when the dest is temporary
654 * down (either by administrator or by monitor program), the dest can be
655 * picked back from the trash, the remaining connections to the dest can
656 * continue, and the counting information of the dest is also useful for
659 static struct ip_vs_dest *
660 ip_vs_trash_get_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
663 struct ip_vs_dest *dest;
664 struct netns_ipvs *ipvs = net_ipvs(svc->net);
667 * Find the destination in trash
669 spin_lock_bh(&ipvs->dest_trash_lock);
670 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
671 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
674 IP_VS_DBG_ADDR(svc->af, &dest->addr),
676 atomic_read(&dest->refcnt));
677 if (dest->af == svc->af &&
678 ip_vs_addr_equal(svc->af, &dest->addr, daddr) &&
679 dest->port == dport &&
680 dest->vfwmark == svc->fwmark &&
681 dest->protocol == svc->protocol &&
683 (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
684 dest->vport == svc->port))) {
686 list_del(&dest->t_list);
687 ip_vs_dest_hold(dest);
695 spin_unlock_bh(&ipvs->dest_trash_lock);
700 static void ip_vs_dest_free(struct ip_vs_dest *dest)
702 struct ip_vs_service *svc = rcu_dereference_protected(dest->svc, 1);
704 __ip_vs_dst_cache_reset(dest);
705 __ip_vs_svc_put(svc, false);
706 free_percpu(dest->stats.cpustats);
711 * Clean up all the destinations in the trash
712 * Called by the ip_vs_control_cleanup()
714 * When the ip_vs_control_clearup is activated by ipvs module exit,
715 * the service tables must have been flushed and all the connections
716 * are expired, and the refcnt of each destination in the trash must
717 * be 0, so we simply release them here.
719 static void ip_vs_trash_cleanup(struct net *net)
721 struct ip_vs_dest *dest, *nxt;
722 struct netns_ipvs *ipvs = net_ipvs(net);
724 del_timer_sync(&ipvs->dest_trash_timer);
725 /* No need to use dest_trash_lock */
726 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, t_list) {
727 list_del(&dest->t_list);
728 ip_vs_dest_free(dest);
733 ip_vs_copy_stats(struct ip_vs_stats_user *dst, struct ip_vs_stats *src)
735 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->ustats.c - src->ustats0.c
737 spin_lock_bh(&src->lock);
739 IP_VS_SHOW_STATS_COUNTER(conns);
740 IP_VS_SHOW_STATS_COUNTER(inpkts);
741 IP_VS_SHOW_STATS_COUNTER(outpkts);
742 IP_VS_SHOW_STATS_COUNTER(inbytes);
743 IP_VS_SHOW_STATS_COUNTER(outbytes);
745 ip_vs_read_estimator(dst, src);
747 spin_unlock_bh(&src->lock);
751 ip_vs_zero_stats(struct ip_vs_stats *stats)
753 spin_lock_bh(&stats->lock);
755 /* get current counters as zero point, rates are zeroed */
757 #define IP_VS_ZERO_STATS_COUNTER(c) stats->ustats0.c = stats->ustats.c
759 IP_VS_ZERO_STATS_COUNTER(conns);
760 IP_VS_ZERO_STATS_COUNTER(inpkts);
761 IP_VS_ZERO_STATS_COUNTER(outpkts);
762 IP_VS_ZERO_STATS_COUNTER(inbytes);
763 IP_VS_ZERO_STATS_COUNTER(outbytes);
765 ip_vs_zero_estimator(stats);
767 spin_unlock_bh(&stats->lock);
771 * Update a destination in the given service
774 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
775 struct ip_vs_dest_user_kern *udest, int add)
777 struct netns_ipvs *ipvs = net_ipvs(svc->net);
778 struct ip_vs_service *old_svc;
779 struct ip_vs_scheduler *sched;
782 /* set the weight and the flags */
783 atomic_set(&dest->weight, udest->weight);
784 conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
785 conn_flags |= IP_VS_CONN_F_INACTIVE;
787 /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
788 if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
789 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
792 * Put the real service in rs_table if not present.
793 * For now only for NAT!
795 ip_vs_rs_hash(ipvs, dest);
797 atomic_set(&dest->conn_flags, conn_flags);
799 /* bind the service */
800 old_svc = rcu_dereference_protected(dest->svc, 1);
802 __ip_vs_bind_svc(dest, svc);
804 if (old_svc != svc) {
805 ip_vs_zero_stats(&dest->stats);
806 __ip_vs_bind_svc(dest, svc);
807 __ip_vs_svc_put(old_svc, true);
811 /* set the dest status flags */
812 dest->flags |= IP_VS_DEST_F_AVAILABLE;
814 if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
815 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
816 dest->u_threshold = udest->u_threshold;
817 dest->l_threshold = udest->l_threshold;
819 spin_lock_bh(&dest->dst_lock);
820 __ip_vs_dst_cache_reset(dest);
821 spin_unlock_bh(&dest->dst_lock);
823 sched = rcu_dereference_protected(svc->scheduler, 1);
825 ip_vs_start_estimator(svc->net, &dest->stats);
826 list_add_rcu(&dest->n_list, &svc->destinations);
829 sched->add_dest(svc, dest);
832 sched->upd_dest(svc, dest);
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_HLIST_NODE(&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);
927 /* We use function that requires RCU lock */
929 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),
952 __ip_vs_update_dest(svc, dest, udest, 1);
956 * Allocate and initialize the dest structure
958 ret = ip_vs_new_dest(svc, udest, &dest);
967 * Edit a destination in the given service
970 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
972 struct ip_vs_dest *dest;
973 union nf_inet_addr daddr;
974 __be16 dport = udest->port;
978 if (udest->weight < 0) {
979 pr_err("%s(): server weight less than zero\n", __func__);
983 if (udest->l_threshold > udest->u_threshold) {
984 pr_err("%s(): lower threshold is higher than upper threshold\n",
989 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
991 /* We use function that requires RCU lock */
993 dest = ip_vs_lookup_dest(svc, &daddr, dport);
997 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1001 __ip_vs_update_dest(svc, dest, udest, 0);
1008 * Delete a destination (must be already unlinked from the service)
1010 static void __ip_vs_del_dest(struct net *net, struct ip_vs_dest *dest,
1013 struct netns_ipvs *ipvs = net_ipvs(net);
1015 ip_vs_stop_estimator(net, &dest->stats);
1018 * Remove it from the d-linked list with the real services.
1020 ip_vs_rs_unhash(dest);
1022 spin_lock_bh(&ipvs->dest_trash_lock);
1023 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1024 IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port),
1025 atomic_read(&dest->refcnt));
1026 if (list_empty(&ipvs->dest_trash) && !cleanup)
1027 mod_timer(&ipvs->dest_trash_timer,
1028 jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1029 /* dest lives in trash without reference */
1030 list_add(&dest->t_list, &ipvs->dest_trash);
1031 dest->idle_start = 0;
1032 spin_unlock_bh(&ipvs->dest_trash_lock);
1033 ip_vs_dest_put(dest);
1038 * Unlink a destination from the given service
1040 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1041 struct ip_vs_dest *dest,
1044 dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1047 * Remove it from the d-linked destination list.
1049 list_del_rcu(&dest->n_list);
1053 struct ip_vs_scheduler *sched;
1055 sched = rcu_dereference_protected(svc->scheduler, 1);
1056 if (sched->del_dest)
1057 sched->del_dest(svc, dest);
1063 * Delete a destination server in the given service
1066 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1068 struct ip_vs_dest *dest;
1069 __be16 dport = udest->port;
1073 /* We use function that requires RCU lock */
1075 dest = ip_vs_lookup_dest(svc, &udest->addr, dport);
1079 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1084 * Unlink dest from the service
1086 __ip_vs_unlink_dest(svc, dest, 1);
1089 * Delete the destination
1091 __ip_vs_del_dest(svc->net, dest, false);
1098 static void ip_vs_dest_trash_expire(unsigned long data)
1100 struct net *net = (struct net *) data;
1101 struct netns_ipvs *ipvs = net_ipvs(net);
1102 struct ip_vs_dest *dest, *next;
1103 unsigned long now = jiffies;
1105 spin_lock(&ipvs->dest_trash_lock);
1106 list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) {
1107 if (atomic_read(&dest->refcnt) > 0)
1109 if (dest->idle_start) {
1110 if (time_before(now, dest->idle_start +
1111 IP_VS_DEST_TRASH_PERIOD))
1114 dest->idle_start = max(1UL, now);
1117 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1119 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1121 list_del(&dest->t_list);
1122 ip_vs_dest_free(dest);
1124 if (!list_empty(&ipvs->dest_trash))
1125 mod_timer(&ipvs->dest_trash_timer,
1126 jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1127 spin_unlock(&ipvs->dest_trash_lock);
1131 * Add a service into the service hash table
1134 ip_vs_add_service(struct net *net, struct ip_vs_service_user_kern *u,
1135 struct ip_vs_service **svc_p)
1138 struct ip_vs_scheduler *sched = NULL;
1139 struct ip_vs_pe *pe = NULL;
1140 struct ip_vs_service *svc = NULL;
1141 struct netns_ipvs *ipvs = net_ipvs(net);
1143 /* increase the module use count */
1144 ip_vs_use_count_inc();
1146 /* Lookup the scheduler by 'u->sched_name' */
1147 sched = ip_vs_scheduler_get(u->sched_name);
1148 if (sched == NULL) {
1149 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1154 if (u->pe_name && *u->pe_name) {
1155 pe = ip_vs_pe_getbyname(u->pe_name);
1157 pr_info("persistence engine module ip_vs_pe_%s "
1158 "not found\n", u->pe_name);
1164 #ifdef CONFIG_IP_VS_IPV6
1165 if (u->af == AF_INET6) {
1166 __u32 plen = (__force __u32) u->netmask;
1168 if (plen < 1 || plen > 128) {
1175 svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1177 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1181 svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1182 if (!svc->stats.cpustats) {
1187 /* I'm the first user of the service */
1188 atomic_set(&svc->refcnt, 0);
1191 svc->protocol = u->protocol;
1192 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1193 svc->port = u->port;
1194 svc->fwmark = u->fwmark;
1195 svc->flags = u->flags;
1196 svc->timeout = u->timeout * HZ;
1197 svc->netmask = u->netmask;
1200 INIT_LIST_HEAD(&svc->destinations);
1201 spin_lock_init(&svc->sched_lock);
1202 spin_lock_init(&svc->stats.lock);
1204 /* Bind the scheduler */
1205 ret = ip_vs_bind_scheduler(svc, sched);
1210 /* Bind the ct retriever */
1211 RCU_INIT_POINTER(svc->pe, pe);
1214 /* Update the virtual service counters */
1215 if (svc->port == FTPPORT)
1216 atomic_inc(&ipvs->ftpsvc_counter);
1217 else if (svc->port == 0)
1218 atomic_inc(&ipvs->nullsvc_counter);
1220 ip_vs_start_estimator(net, &svc->stats);
1222 /* Count only IPv4 services for old get/setsockopt interface */
1223 if (svc->af == AF_INET)
1224 ipvs->num_services++;
1226 /* Hash the service into the service table */
1227 ip_vs_svc_hash(svc);
1230 /* Now there is a service - full throttle */
1237 ip_vs_unbind_scheduler(svc, sched);
1238 ip_vs_service_free(svc);
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) {
1283 __u32 plen = (__force __u32) u->netmask;
1285 if (plen < 1 || plen > 128) {
1292 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1293 if (sched != old_sched) {
1294 /* Bind the new scheduler */
1295 ret = ip_vs_bind_scheduler(svc, sched);
1300 /* Unbind the old scheduler on success */
1301 ip_vs_unbind_scheduler(svc, old_sched);
1305 * Set the flags and timeout value
1307 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1308 svc->timeout = u->timeout * HZ;
1309 svc->netmask = u->netmask;
1311 old_pe = rcu_dereference_protected(svc->pe, 1);
1313 rcu_assign_pointer(svc->pe, pe);
1316 ip_vs_scheduler_put(old_sched);
1317 ip_vs_pe_put(old_pe);
1322 * Delete a service from the service list
1323 * - The service must be unlinked, unlocked and not referenced!
1324 * - We are called under _bh lock
1326 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1328 struct ip_vs_dest *dest, *nxt;
1329 struct ip_vs_scheduler *old_sched;
1330 struct ip_vs_pe *old_pe;
1331 struct netns_ipvs *ipvs = net_ipvs(svc->net);
1333 pr_info("%s: enter\n", __func__);
1335 /* Count only IPv4 services for old get/setsockopt interface */
1336 if (svc->af == AF_INET)
1337 ipvs->num_services--;
1339 ip_vs_stop_estimator(svc->net, &svc->stats);
1341 /* Unbind scheduler */
1342 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1343 ip_vs_unbind_scheduler(svc, old_sched);
1344 ip_vs_scheduler_put(old_sched);
1346 /* Unbind persistence engine, keep svc->pe */
1347 old_pe = rcu_dereference_protected(svc->pe, 1);
1348 ip_vs_pe_put(old_pe);
1351 * Unlink the whole destination list
1353 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1354 __ip_vs_unlink_dest(svc, dest, 0);
1355 __ip_vs_del_dest(svc->net, dest, cleanup);
1359 * Update the virtual service counters
1361 if (svc->port == FTPPORT)
1362 atomic_dec(&ipvs->ftpsvc_counter);
1363 else if (svc->port == 0)
1364 atomic_dec(&ipvs->nullsvc_counter);
1367 * Free the service if nobody refers to it
1369 __ip_vs_svc_put(svc, true);
1371 /* decrease the module use count */
1372 ip_vs_use_count_dec();
1376 * Unlink a service from list and try to delete it if its refcnt reached 0
1378 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1380 /* Hold svc to avoid double release from dest_trash */
1381 atomic_inc(&svc->refcnt);
1383 * Unhash it from the service table
1385 ip_vs_svc_unhash(svc);
1387 __ip_vs_del_service(svc, cleanup);
1391 * Delete a service from the service list
1393 static int ip_vs_del_service(struct ip_vs_service *svc)
1397 ip_vs_unlink_service(svc, false);
1404 * Flush all the virtual services
1406 static int ip_vs_flush(struct net *net, bool cleanup)
1409 struct ip_vs_service *svc;
1410 struct hlist_node *n;
1413 * Flush the service table hashed by <netns,protocol,addr,port>
1415 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1416 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1418 if (net_eq(svc->net, net))
1419 ip_vs_unlink_service(svc, cleanup);
1424 * Flush the service table hashed by fwmark
1426 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1427 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1429 if (net_eq(svc->net, net))
1430 ip_vs_unlink_service(svc, cleanup);
1438 * Delete service by {netns} in the service table.
1439 * Called by __ip_vs_cleanup()
1441 void ip_vs_service_net_cleanup(struct net *net)
1444 /* Check for "full" addressed entries */
1445 mutex_lock(&__ip_vs_mutex);
1446 ip_vs_flush(net, true);
1447 mutex_unlock(&__ip_vs_mutex);
1451 /* Put all references for device (dst_cache) */
1453 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1455 struct ip_vs_dest_dst *dest_dst;
1457 spin_lock_bh(&dest->dst_lock);
1458 dest_dst = rcu_dereference_protected(dest->dest_dst, 1);
1459 if (dest_dst && dest_dst->dst_cache->dev == dev) {
1460 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1462 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1464 atomic_read(&dest->refcnt));
1465 __ip_vs_dst_cache_reset(dest);
1467 spin_unlock_bh(&dest->dst_lock);
1470 /* Netdev event receiver
1471 * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1473 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1476 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1477 struct net *net = dev_net(dev);
1478 struct netns_ipvs *ipvs = net_ipvs(net);
1479 struct ip_vs_service *svc;
1480 struct ip_vs_dest *dest;
1483 if (event != NETDEV_DOWN || !ipvs)
1485 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1487 mutex_lock(&__ip_vs_mutex);
1488 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1489 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1490 if (net_eq(svc->net, net)) {
1491 list_for_each_entry(dest, &svc->destinations,
1493 ip_vs_forget_dev(dest, dev);
1498 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1499 if (net_eq(svc->net, net)) {
1500 list_for_each_entry(dest, &svc->destinations,
1502 ip_vs_forget_dev(dest, dev);
1509 spin_lock_bh(&ipvs->dest_trash_lock);
1510 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1511 ip_vs_forget_dev(dest, dev);
1513 spin_unlock_bh(&ipvs->dest_trash_lock);
1514 mutex_unlock(&__ip_vs_mutex);
1520 * Zero counters in a service or all services
1522 static int ip_vs_zero_service(struct ip_vs_service *svc)
1524 struct ip_vs_dest *dest;
1526 list_for_each_entry(dest, &svc->destinations, n_list) {
1527 ip_vs_zero_stats(&dest->stats);
1529 ip_vs_zero_stats(&svc->stats);
1533 static int ip_vs_zero_all(struct net *net)
1536 struct ip_vs_service *svc;
1538 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1539 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1540 if (net_eq(svc->net, net))
1541 ip_vs_zero_service(svc);
1545 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1546 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1547 if (net_eq(svc->net, net))
1548 ip_vs_zero_service(svc);
1552 ip_vs_zero_stats(&net_ipvs(net)->tot_stats);
1556 #ifdef CONFIG_SYSCTL
1559 static int three = 3;
1562 proc_do_defense_mode(struct ctl_table *table, int write,
1563 void __user *buffer, size_t *lenp, loff_t *ppos)
1565 struct net *net = current->nsproxy->net_ns;
1566 int *valp = table->data;
1570 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1571 if (write && (*valp != val)) {
1572 if ((*valp < 0) || (*valp > 3)) {
1573 /* Restore the correct value */
1576 update_defense_level(net_ipvs(net));
1583 proc_do_sync_threshold(struct ctl_table *table, int write,
1584 void __user *buffer, size_t *lenp, loff_t *ppos)
1586 int *valp = table->data;
1590 /* backup the value first */
1591 memcpy(val, valp, sizeof(val));
1593 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1594 if (write && (valp[0] < 0 || valp[1] < 0 ||
1595 (valp[0] >= valp[1] && valp[1]))) {
1596 /* Restore the correct value */
1597 memcpy(valp, val, sizeof(val));
1603 proc_do_sync_mode(struct ctl_table *table, int write,
1604 void __user *buffer, size_t *lenp, loff_t *ppos)
1606 int *valp = table->data;
1610 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1611 if (write && (*valp != val)) {
1612 if ((*valp < 0) || (*valp > 1)) {
1613 /* Restore the correct value */
1621 proc_do_sync_ports(struct ctl_table *table, int write,
1622 void __user *buffer, size_t *lenp, loff_t *ppos)
1624 int *valp = table->data;
1628 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1629 if (write && (*valp != val)) {
1630 if (*valp < 1 || !is_power_of_2(*valp)) {
1631 /* Restore the correct value */
1639 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1640 * Do not change order or insert new entries without
1641 * align with netns init in ip_vs_control_net_init()
1644 static struct ctl_table vs_vars[] = {
1646 .procname = "amemthresh",
1647 .maxlen = sizeof(int),
1649 .proc_handler = proc_dointvec,
1652 .procname = "am_droprate",
1653 .maxlen = sizeof(int),
1655 .proc_handler = proc_dointvec,
1658 .procname = "drop_entry",
1659 .maxlen = sizeof(int),
1661 .proc_handler = proc_do_defense_mode,
1664 .procname = "drop_packet",
1665 .maxlen = sizeof(int),
1667 .proc_handler = proc_do_defense_mode,
1669 #ifdef CONFIG_IP_VS_NFCT
1671 .procname = "conntrack",
1672 .maxlen = sizeof(int),
1674 .proc_handler = &proc_dointvec,
1678 .procname = "secure_tcp",
1679 .maxlen = sizeof(int),
1681 .proc_handler = proc_do_defense_mode,
1684 .procname = "snat_reroute",
1685 .maxlen = sizeof(int),
1687 .proc_handler = &proc_dointvec,
1690 .procname = "sync_version",
1691 .maxlen = sizeof(int),
1693 .proc_handler = &proc_do_sync_mode,
1696 .procname = "sync_ports",
1697 .maxlen = sizeof(int),
1699 .proc_handler = &proc_do_sync_ports,
1702 .procname = "sync_persist_mode",
1703 .maxlen = sizeof(int),
1705 .proc_handler = proc_dointvec,
1708 .procname = "sync_qlen_max",
1709 .maxlen = sizeof(unsigned long),
1711 .proc_handler = proc_doulongvec_minmax,
1714 .procname = "sync_sock_size",
1715 .maxlen = sizeof(int),
1717 .proc_handler = proc_dointvec,
1720 .procname = "cache_bypass",
1721 .maxlen = sizeof(int),
1723 .proc_handler = proc_dointvec,
1726 .procname = "expire_nodest_conn",
1727 .maxlen = sizeof(int),
1729 .proc_handler = proc_dointvec,
1732 .procname = "sloppy_tcp",
1733 .maxlen = sizeof(int),
1735 .proc_handler = proc_dointvec,
1738 .procname = "sloppy_sctp",
1739 .maxlen = sizeof(int),
1741 .proc_handler = proc_dointvec,
1744 .procname = "expire_quiescent_template",
1745 .maxlen = sizeof(int),
1747 .proc_handler = proc_dointvec,
1750 .procname = "sync_threshold",
1752 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1754 .proc_handler = proc_do_sync_threshold,
1757 .procname = "sync_refresh_period",
1758 .maxlen = sizeof(int),
1760 .proc_handler = proc_dointvec_jiffies,
1763 .procname = "sync_retries",
1764 .maxlen = sizeof(int),
1766 .proc_handler = proc_dointvec_minmax,
1771 .procname = "nat_icmp_send",
1772 .maxlen = sizeof(int),
1774 .proc_handler = proc_dointvec,
1777 .procname = "pmtu_disc",
1778 .maxlen = sizeof(int),
1780 .proc_handler = proc_dointvec,
1783 .procname = "backup_only",
1784 .maxlen = sizeof(int),
1786 .proc_handler = proc_dointvec,
1788 #ifdef CONFIG_IP_VS_DEBUG
1790 .procname = "debug_level",
1791 .data = &sysctl_ip_vs_debug_level,
1792 .maxlen = sizeof(int),
1794 .proc_handler = proc_dointvec,
1799 .procname = "timeout_established",
1800 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1801 .maxlen = sizeof(int),
1803 .proc_handler = proc_dointvec_jiffies,
1806 .procname = "timeout_synsent",
1807 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1808 .maxlen = sizeof(int),
1810 .proc_handler = proc_dointvec_jiffies,
1813 .procname = "timeout_synrecv",
1814 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1815 .maxlen = sizeof(int),
1817 .proc_handler = proc_dointvec_jiffies,
1820 .procname = "timeout_finwait",
1821 .data = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1822 .maxlen = sizeof(int),
1824 .proc_handler = proc_dointvec_jiffies,
1827 .procname = "timeout_timewait",
1828 .data = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1829 .maxlen = sizeof(int),
1831 .proc_handler = proc_dointvec_jiffies,
1834 .procname = "timeout_close",
1835 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1836 .maxlen = sizeof(int),
1838 .proc_handler = proc_dointvec_jiffies,
1841 .procname = "timeout_closewait",
1842 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1843 .maxlen = sizeof(int),
1845 .proc_handler = proc_dointvec_jiffies,
1848 .procname = "timeout_lastack",
1849 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1850 .maxlen = sizeof(int),
1852 .proc_handler = proc_dointvec_jiffies,
1855 .procname = "timeout_listen",
1856 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1857 .maxlen = sizeof(int),
1859 .proc_handler = proc_dointvec_jiffies,
1862 .procname = "timeout_synack",
1863 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1864 .maxlen = sizeof(int),
1866 .proc_handler = proc_dointvec_jiffies,
1869 .procname = "timeout_udp",
1870 .data = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1871 .maxlen = sizeof(int),
1873 .proc_handler = proc_dointvec_jiffies,
1876 .procname = "timeout_icmp",
1877 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1878 .maxlen = sizeof(int),
1880 .proc_handler = proc_dointvec_jiffies,
1888 #ifdef CONFIG_PROC_FS
1891 struct seq_net_private p; /* Do not move this, netns depends upon it*/
1892 struct hlist_head *table;
1897 * Write the contents of the VS rule table to a PROCfs file.
1898 * (It is kept just for backward compatibility)
1900 static inline const char *ip_vs_fwd_name(unsigned int flags)
1902 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1903 case IP_VS_CONN_F_LOCALNODE:
1905 case IP_VS_CONN_F_TUNNEL:
1907 case IP_VS_CONN_F_DROUTE:
1915 /* Get the Nth entry in the two lists */
1916 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1918 struct net *net = seq_file_net(seq);
1919 struct ip_vs_iter *iter = seq->private;
1921 struct ip_vs_service *svc;
1923 /* look in hash by protocol */
1924 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1925 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
1926 if (net_eq(svc->net, net) && pos-- == 0) {
1927 iter->table = ip_vs_svc_table;
1934 /* keep looking in fwmark */
1935 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1936 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
1938 if (net_eq(svc->net, net) && pos-- == 0) {
1939 iter->table = ip_vs_svc_fwm_table;
1949 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1953 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1957 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1959 struct hlist_node *e;
1960 struct ip_vs_iter *iter;
1961 struct ip_vs_service *svc;
1964 if (v == SEQ_START_TOKEN)
1965 return ip_vs_info_array(seq,0);
1968 iter = seq->private;
1970 if (iter->table == ip_vs_svc_table) {
1971 /* next service in table hashed by protocol */
1972 e = rcu_dereference(hlist_next_rcu(&svc->s_list));
1974 return hlist_entry(e, struct ip_vs_service, s_list);
1976 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1977 hlist_for_each_entry_rcu(svc,
1978 &ip_vs_svc_table[iter->bucket],
1984 iter->table = ip_vs_svc_fwm_table;
1989 /* next service in hashed by fwmark */
1990 e = rcu_dereference(hlist_next_rcu(&svc->f_list));
1992 return hlist_entry(e, struct ip_vs_service, f_list);
1995 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1996 hlist_for_each_entry_rcu(svc,
1997 &ip_vs_svc_fwm_table[iter->bucket],
2005 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
2012 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
2014 if (v == SEQ_START_TOKEN) {
2016 "IP Virtual Server version %d.%d.%d (size=%d)\n",
2017 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2019 "Prot LocalAddress:Port Scheduler Flags\n");
2021 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2023 const struct ip_vs_service *svc = v;
2024 const struct ip_vs_iter *iter = seq->private;
2025 const struct ip_vs_dest *dest;
2026 struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler);
2028 if (iter->table == ip_vs_svc_table) {
2029 #ifdef CONFIG_IP_VS_IPV6
2030 if (svc->af == AF_INET6)
2031 seq_printf(seq, "%s [%pI6]:%04X %s ",
2032 ip_vs_proto_name(svc->protocol),
2038 seq_printf(seq, "%s %08X:%04X %s %s ",
2039 ip_vs_proto_name(svc->protocol),
2040 ntohl(svc->addr.ip),
2043 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2045 seq_printf(seq, "FWM %08X %s %s",
2046 svc->fwmark, sched->name,
2047 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2050 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2051 seq_printf(seq, "persistent %d %08X\n",
2053 ntohl(svc->netmask));
2055 seq_putc(seq, '\n');
2057 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
2058 #ifdef CONFIG_IP_VS_IPV6
2059 if (dest->af == AF_INET6)
2062 " %-7s %-6d %-10d %-10d\n",
2065 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2066 atomic_read(&dest->weight),
2067 atomic_read(&dest->activeconns),
2068 atomic_read(&dest->inactconns));
2073 "%-7s %-6d %-10d %-10d\n",
2074 ntohl(dest->addr.ip),
2076 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2077 atomic_read(&dest->weight),
2078 atomic_read(&dest->activeconns),
2079 atomic_read(&dest->inactconns));
2086 static const struct seq_operations ip_vs_info_seq_ops = {
2087 .start = ip_vs_info_seq_start,
2088 .next = ip_vs_info_seq_next,
2089 .stop = ip_vs_info_seq_stop,
2090 .show = ip_vs_info_seq_show,
2093 static int ip_vs_info_open(struct inode *inode, struct file *file)
2095 return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2096 sizeof(struct ip_vs_iter));
2099 static const struct file_operations ip_vs_info_fops = {
2100 .owner = THIS_MODULE,
2101 .open = ip_vs_info_open,
2103 .llseek = seq_lseek,
2104 .release = seq_release_net,
2107 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2109 struct net *net = seq_file_single_net(seq);
2110 struct ip_vs_stats_user show;
2112 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2114 " Total Incoming Outgoing Incoming Outgoing\n");
2116 " Conns Packets Packets Bytes Bytes\n");
2118 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2119 seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", show.conns,
2120 show.inpkts, show.outpkts,
2121 (unsigned long long) show.inbytes,
2122 (unsigned long long) show.outbytes);
2124 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2126 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2127 seq_printf(seq, "%8X %8X %8X %16X %16X\n",
2128 show.cps, show.inpps, show.outpps,
2129 show.inbps, show.outbps);
2134 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2136 return single_open_net(inode, file, ip_vs_stats_show);
2139 static const struct file_operations ip_vs_stats_fops = {
2140 .owner = THIS_MODULE,
2141 .open = ip_vs_stats_seq_open,
2143 .llseek = seq_lseek,
2144 .release = single_release_net,
2147 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2149 struct net *net = seq_file_single_net(seq);
2150 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2151 struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats;
2152 struct ip_vs_stats_user rates;
2155 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2157 " Total Incoming Outgoing Incoming Outgoing\n");
2159 "CPU Conns Packets Packets Bytes Bytes\n");
2161 for_each_possible_cpu(i) {
2162 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2164 __u64 inbytes, outbytes;
2167 start = u64_stats_fetch_begin_bh(&u->syncp);
2168 inbytes = u->ustats.inbytes;
2169 outbytes = u->ustats.outbytes;
2170 } while (u64_stats_fetch_retry_bh(&u->syncp, start));
2172 seq_printf(seq, "%3X %8X %8X %8X %16LX %16LX\n",
2173 i, u->ustats.conns, u->ustats.inpkts,
2174 u->ustats.outpkts, (__u64)inbytes,
2178 spin_lock_bh(&tot_stats->lock);
2180 seq_printf(seq, " ~ %8X %8X %8X %16LX %16LX\n\n",
2181 tot_stats->ustats.conns, tot_stats->ustats.inpkts,
2182 tot_stats->ustats.outpkts,
2183 (unsigned long long) tot_stats->ustats.inbytes,
2184 (unsigned long long) tot_stats->ustats.outbytes);
2186 ip_vs_read_estimator(&rates, tot_stats);
2188 spin_unlock_bh(&tot_stats->lock);
2190 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2192 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2193 seq_printf(seq, " %8X %8X %8X %16X %16X\n",
2203 static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2205 return single_open_net(inode, file, ip_vs_stats_percpu_show);
2208 static const struct file_operations ip_vs_stats_percpu_fops = {
2209 .owner = THIS_MODULE,
2210 .open = ip_vs_stats_percpu_seq_open,
2212 .llseek = seq_lseek,
2213 .release = single_release_net,
2218 * Set timeout values for tcp tcpfin udp in the timeout_table.
2220 static int ip_vs_set_timeout(struct net *net, struct ip_vs_timeout_user *u)
2222 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2223 struct ip_vs_proto_data *pd;
2226 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2231 #ifdef CONFIG_IP_VS_PROTO_TCP
2232 if (u->tcp_timeout) {
2233 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2234 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2235 = u->tcp_timeout * HZ;
2238 if (u->tcp_fin_timeout) {
2239 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2240 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2241 = u->tcp_fin_timeout * HZ;
2245 #ifdef CONFIG_IP_VS_PROTO_UDP
2246 if (u->udp_timeout) {
2247 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2248 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2249 = u->udp_timeout * HZ;
2256 #define SET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2257 #define SERVICE_ARG_LEN (sizeof(struct ip_vs_service_user))
2258 #define SVCDEST_ARG_LEN (sizeof(struct ip_vs_service_user) + \
2259 sizeof(struct ip_vs_dest_user))
2260 #define TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2261 #define DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user))
2262 #define MAX_ARG_LEN SVCDEST_ARG_LEN
2264 static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
2265 [SET_CMDID(IP_VS_SO_SET_ADD)] = SERVICE_ARG_LEN,
2266 [SET_CMDID(IP_VS_SO_SET_EDIT)] = SERVICE_ARG_LEN,
2267 [SET_CMDID(IP_VS_SO_SET_DEL)] = SERVICE_ARG_LEN,
2268 [SET_CMDID(IP_VS_SO_SET_FLUSH)] = 0,
2269 [SET_CMDID(IP_VS_SO_SET_ADDDEST)] = SVCDEST_ARG_LEN,
2270 [SET_CMDID(IP_VS_SO_SET_DELDEST)] = SVCDEST_ARG_LEN,
2271 [SET_CMDID(IP_VS_SO_SET_EDITDEST)] = SVCDEST_ARG_LEN,
2272 [SET_CMDID(IP_VS_SO_SET_TIMEOUT)] = TIMEOUT_ARG_LEN,
2273 [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)] = DAEMON_ARG_LEN,
2274 [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)] = DAEMON_ARG_LEN,
2275 [SET_CMDID(IP_VS_SO_SET_ZERO)] = SERVICE_ARG_LEN,
2278 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2279 struct ip_vs_service_user *usvc_compat)
2281 memset(usvc, 0, sizeof(*usvc));
2284 usvc->protocol = usvc_compat->protocol;
2285 usvc->addr.ip = usvc_compat->addr;
2286 usvc->port = usvc_compat->port;
2287 usvc->fwmark = usvc_compat->fwmark;
2289 /* Deep copy of sched_name is not needed here */
2290 usvc->sched_name = usvc_compat->sched_name;
2292 usvc->flags = usvc_compat->flags;
2293 usvc->timeout = usvc_compat->timeout;
2294 usvc->netmask = usvc_compat->netmask;
2297 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2298 struct ip_vs_dest_user *udest_compat)
2300 memset(udest, 0, sizeof(*udest));
2302 udest->addr.ip = udest_compat->addr;
2303 udest->port = udest_compat->port;
2304 udest->conn_flags = udest_compat->conn_flags;
2305 udest->weight = udest_compat->weight;
2306 udest->u_threshold = udest_compat->u_threshold;
2307 udest->l_threshold = udest_compat->l_threshold;
2311 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2313 struct net *net = sock_net(sk);
2315 unsigned char arg[MAX_ARG_LEN];
2316 struct ip_vs_service_user *usvc_compat;
2317 struct ip_vs_service_user_kern usvc;
2318 struct ip_vs_service *svc;
2319 struct ip_vs_dest_user *udest_compat;
2320 struct ip_vs_dest_user_kern udest;
2321 struct netns_ipvs *ipvs = net_ipvs(net);
2323 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2326 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2328 if (len < 0 || len > MAX_ARG_LEN)
2330 if (len != set_arglen[SET_CMDID(cmd)]) {
2331 pr_err("set_ctl: len %u != %u\n",
2332 len, set_arglen[SET_CMDID(cmd)]);
2336 if (copy_from_user(arg, user, len) != 0)
2339 /* increase the module use count */
2340 ip_vs_use_count_inc();
2342 /* Handle daemons since they have another lock */
2343 if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2344 cmd == IP_VS_SO_SET_STOPDAEMON) {
2345 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2347 if (mutex_lock_interruptible(&ipvs->sync_mutex)) {
2351 if (cmd == IP_VS_SO_SET_STARTDAEMON)
2352 ret = start_sync_thread(net, dm->state, dm->mcast_ifn,
2355 ret = stop_sync_thread(net, dm->state);
2356 mutex_unlock(&ipvs->sync_mutex);
2360 if (mutex_lock_interruptible(&__ip_vs_mutex)) {
2365 if (cmd == IP_VS_SO_SET_FLUSH) {
2366 /* Flush the virtual service */
2367 ret = ip_vs_flush(net, false);
2369 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2370 /* Set timeout values for (tcp tcpfin udp) */
2371 ret = ip_vs_set_timeout(net, (struct ip_vs_timeout_user *)arg);
2375 usvc_compat = (struct ip_vs_service_user *)arg;
2376 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2378 /* We only use the new structs internally, so copy userspace compat
2379 * structs to extended internal versions */
2380 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2381 ip_vs_copy_udest_compat(&udest, udest_compat);
2383 if (cmd == IP_VS_SO_SET_ZERO) {
2384 /* if no service address is set, zero counters in all */
2385 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2386 ret = ip_vs_zero_all(net);
2391 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2392 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2393 usvc.protocol != IPPROTO_SCTP) {
2394 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2395 usvc.protocol, &usvc.addr.ip,
2396 ntohs(usvc.port), usvc.sched_name);
2401 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2403 if (usvc.fwmark == 0)
2404 svc = __ip_vs_service_find(net, usvc.af, usvc.protocol,
2405 &usvc.addr, usvc.port);
2407 svc = __ip_vs_svc_fwm_find(net, usvc.af, usvc.fwmark);
2410 if (cmd != IP_VS_SO_SET_ADD
2411 && (svc == NULL || svc->protocol != usvc.protocol)) {
2417 case IP_VS_SO_SET_ADD:
2421 ret = ip_vs_add_service(net, &usvc, &svc);
2423 case IP_VS_SO_SET_EDIT:
2424 ret = ip_vs_edit_service(svc, &usvc);
2426 case IP_VS_SO_SET_DEL:
2427 ret = ip_vs_del_service(svc);
2431 case IP_VS_SO_SET_ZERO:
2432 ret = ip_vs_zero_service(svc);
2434 case IP_VS_SO_SET_ADDDEST:
2435 ret = ip_vs_add_dest(svc, &udest);
2437 case IP_VS_SO_SET_EDITDEST:
2438 ret = ip_vs_edit_dest(svc, &udest);
2440 case IP_VS_SO_SET_DELDEST:
2441 ret = ip_vs_del_dest(svc, &udest);
2448 mutex_unlock(&__ip_vs_mutex);
2450 /* decrease the module use count */
2451 ip_vs_use_count_dec();
2458 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2460 struct ip_vs_scheduler *sched;
2462 sched = rcu_dereference_protected(src->scheduler, 1);
2463 dst->protocol = src->protocol;
2464 dst->addr = src->addr.ip;
2465 dst->port = src->port;
2466 dst->fwmark = src->fwmark;
2467 strlcpy(dst->sched_name, sched->name, sizeof(dst->sched_name));
2468 dst->flags = src->flags;
2469 dst->timeout = src->timeout / HZ;
2470 dst->netmask = src->netmask;
2471 dst->num_dests = src->num_dests;
2472 ip_vs_copy_stats(&dst->stats, &src->stats);
2476 __ip_vs_get_service_entries(struct net *net,
2477 const struct ip_vs_get_services *get,
2478 struct ip_vs_get_services __user *uptr)
2481 struct ip_vs_service *svc;
2482 struct ip_vs_service_entry entry;
2485 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2486 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2487 /* Only expose IPv4 entries to old interface */
2488 if (svc->af != AF_INET || !net_eq(svc->net, net))
2491 if (count >= get->num_services)
2493 memset(&entry, 0, sizeof(entry));
2494 ip_vs_copy_service(&entry, svc);
2495 if (copy_to_user(&uptr->entrytable[count],
2496 &entry, sizeof(entry))) {
2504 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2505 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2506 /* Only expose IPv4 entries to old interface */
2507 if (svc->af != AF_INET || !net_eq(svc->net, net))
2510 if (count >= get->num_services)
2512 memset(&entry, 0, sizeof(entry));
2513 ip_vs_copy_service(&entry, svc);
2514 if (copy_to_user(&uptr->entrytable[count],
2515 &entry, sizeof(entry))) {
2527 __ip_vs_get_dest_entries(struct net *net, const struct ip_vs_get_dests *get,
2528 struct ip_vs_get_dests __user *uptr)
2530 struct ip_vs_service *svc;
2531 union nf_inet_addr addr = { .ip = get->addr };
2536 svc = __ip_vs_svc_fwm_find(net, AF_INET, get->fwmark);
2538 svc = __ip_vs_service_find(net, AF_INET, get->protocol, &addr,
2544 struct ip_vs_dest *dest;
2545 struct ip_vs_dest_entry entry;
2547 memset(&entry, 0, sizeof(entry));
2548 list_for_each_entry(dest, &svc->destinations, n_list) {
2549 if (count >= get->num_dests)
2552 entry.addr = dest->addr.ip;
2553 entry.port = dest->port;
2554 entry.conn_flags = atomic_read(&dest->conn_flags);
2555 entry.weight = atomic_read(&dest->weight);
2556 entry.u_threshold = dest->u_threshold;
2557 entry.l_threshold = dest->l_threshold;
2558 entry.activeconns = atomic_read(&dest->activeconns);
2559 entry.inactconns = atomic_read(&dest->inactconns);
2560 entry.persistconns = atomic_read(&dest->persistconns);
2561 ip_vs_copy_stats(&entry.stats, &dest->stats);
2562 if (copy_to_user(&uptr->entrytable[count],
2563 &entry, sizeof(entry))) {
2575 __ip_vs_get_timeouts(struct net *net, struct ip_vs_timeout_user *u)
2577 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2578 struct ip_vs_proto_data *pd;
2581 memset(u, 0, sizeof (*u));
2583 #ifdef CONFIG_IP_VS_PROTO_TCP
2584 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2585 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2586 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2588 #ifdef CONFIG_IP_VS_PROTO_UDP
2589 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2591 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2596 #define GET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2597 #define GET_INFO_ARG_LEN (sizeof(struct ip_vs_getinfo))
2598 #define GET_SERVICES_ARG_LEN (sizeof(struct ip_vs_get_services))
2599 #define GET_SERVICE_ARG_LEN (sizeof(struct ip_vs_service_entry))
2600 #define GET_DESTS_ARG_LEN (sizeof(struct ip_vs_get_dests))
2601 #define GET_TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2602 #define GET_DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user) * 2)
2604 static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
2605 [GET_CMDID(IP_VS_SO_GET_VERSION)] = 64,
2606 [GET_CMDID(IP_VS_SO_GET_INFO)] = GET_INFO_ARG_LEN,
2607 [GET_CMDID(IP_VS_SO_GET_SERVICES)] = GET_SERVICES_ARG_LEN,
2608 [GET_CMDID(IP_VS_SO_GET_SERVICE)] = GET_SERVICE_ARG_LEN,
2609 [GET_CMDID(IP_VS_SO_GET_DESTS)] = GET_DESTS_ARG_LEN,
2610 [GET_CMDID(IP_VS_SO_GET_TIMEOUT)] = GET_TIMEOUT_ARG_LEN,
2611 [GET_CMDID(IP_VS_SO_GET_DAEMON)] = GET_DAEMON_ARG_LEN,
2615 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2617 unsigned char arg[128];
2619 unsigned int copylen;
2620 struct net *net = sock_net(sk);
2621 struct netns_ipvs *ipvs = net_ipvs(net);
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 if (*len < get_arglen[GET_CMDID(cmd)]) {
2631 pr_err("get_ctl: len %u < %u\n",
2632 *len, get_arglen[GET_CMDID(cmd)]);
2636 copylen = get_arglen[GET_CMDID(cmd)];
2640 if (copy_from_user(arg, user, copylen) != 0)
2643 * Handle daemons first since it has its own locking
2645 if (cmd == IP_VS_SO_GET_DAEMON) {
2646 struct ip_vs_daemon_user d[2];
2648 memset(&d, 0, sizeof(d));
2649 if (mutex_lock_interruptible(&ipvs->sync_mutex))
2650 return -ERESTARTSYS;
2652 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2653 d[0].state = IP_VS_STATE_MASTER;
2654 strlcpy(d[0].mcast_ifn, ipvs->master_mcast_ifn,
2655 sizeof(d[0].mcast_ifn));
2656 d[0].syncid = ipvs->master_syncid;
2658 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2659 d[1].state = IP_VS_STATE_BACKUP;
2660 strlcpy(d[1].mcast_ifn, ipvs->backup_mcast_ifn,
2661 sizeof(d[1].mcast_ifn));
2662 d[1].syncid = ipvs->backup_syncid;
2664 if (copy_to_user(user, &d, sizeof(d)) != 0)
2666 mutex_unlock(&ipvs->sync_mutex);
2670 if (mutex_lock_interruptible(&__ip_vs_mutex))
2671 return -ERESTARTSYS;
2674 case IP_VS_SO_GET_VERSION:
2678 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2679 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2680 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2684 *len = strlen(buf)+1;
2688 case IP_VS_SO_GET_INFO:
2690 struct ip_vs_getinfo info;
2691 info.version = IP_VS_VERSION_CODE;
2692 info.size = ip_vs_conn_tab_size;
2693 info.num_services = ipvs->num_services;
2694 if (copy_to_user(user, &info, sizeof(info)) != 0)
2699 case IP_VS_SO_GET_SERVICES:
2701 struct ip_vs_get_services *get;
2704 get = (struct ip_vs_get_services *)arg;
2705 size = sizeof(*get) +
2706 sizeof(struct ip_vs_service_entry) * get->num_services;
2708 pr_err("length: %u != %u\n", *len, size);
2712 ret = __ip_vs_get_service_entries(net, get, user);
2716 case IP_VS_SO_GET_SERVICE:
2718 struct ip_vs_service_entry *entry;
2719 struct ip_vs_service *svc;
2720 union nf_inet_addr addr;
2722 entry = (struct ip_vs_service_entry *)arg;
2723 addr.ip = entry->addr;
2726 svc = __ip_vs_svc_fwm_find(net, AF_INET, entry->fwmark);
2728 svc = __ip_vs_service_find(net, AF_INET,
2729 entry->protocol, &addr,
2733 ip_vs_copy_service(entry, svc);
2734 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2741 case IP_VS_SO_GET_DESTS:
2743 struct ip_vs_get_dests *get;
2746 get = (struct ip_vs_get_dests *)arg;
2747 size = sizeof(*get) +
2748 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2750 pr_err("length: %u != %u\n", *len, size);
2754 ret = __ip_vs_get_dest_entries(net, get, user);
2758 case IP_VS_SO_GET_TIMEOUT:
2760 struct ip_vs_timeout_user t;
2762 __ip_vs_get_timeouts(net, &t);
2763 if (copy_to_user(user, &t, sizeof(t)) != 0)
2773 mutex_unlock(&__ip_vs_mutex);
2778 static struct nf_sockopt_ops ip_vs_sockopts = {
2780 .set_optmin = IP_VS_BASE_CTL,
2781 .set_optmax = IP_VS_SO_SET_MAX+1,
2782 .set = do_ip_vs_set_ctl,
2783 .get_optmin = IP_VS_BASE_CTL,
2784 .get_optmax = IP_VS_SO_GET_MAX+1,
2785 .get = do_ip_vs_get_ctl,
2786 .owner = THIS_MODULE,
2790 * Generic Netlink interface
2793 /* IPVS genetlink family */
2794 static struct genl_family ip_vs_genl_family = {
2795 .id = GENL_ID_GENERATE,
2797 .name = IPVS_GENL_NAME,
2798 .version = IPVS_GENL_VERSION,
2799 .maxattr = IPVS_CMD_MAX,
2800 .netnsok = true, /* Make ipvsadm to work on netns */
2803 /* Policy used for first-level command attributes */
2804 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2805 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2806 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2807 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2808 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2809 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2810 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2813 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2814 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2815 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2816 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2817 .len = IP_VS_IFNAME_MAXLEN },
2818 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2821 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2822 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2823 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2824 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2825 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2826 .len = sizeof(union nf_inet_addr) },
2827 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2828 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2829 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2830 .len = IP_VS_SCHEDNAME_MAXLEN },
2831 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2832 .len = IP_VS_PENAME_MAXLEN },
2833 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2834 .len = sizeof(struct ip_vs_flags) },
2835 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2836 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2837 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2840 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2841 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2842 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2843 .len = sizeof(union nf_inet_addr) },
2844 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2845 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2846 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2847 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2848 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2849 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2850 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2851 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2852 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2855 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2856 struct ip_vs_stats *stats)
2858 struct ip_vs_stats_user ustats;
2859 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2863 ip_vs_copy_stats(&ustats, stats);
2865 if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, ustats.conns) ||
2866 nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, ustats.inpkts) ||
2867 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, ustats.outpkts) ||
2868 nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, ustats.inbytes) ||
2869 nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, ustats.outbytes) ||
2870 nla_put_u32(skb, IPVS_STATS_ATTR_CPS, ustats.cps) ||
2871 nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, ustats.inpps) ||
2872 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, ustats.outpps) ||
2873 nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, ustats.inbps) ||
2874 nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, ustats.outbps))
2875 goto nla_put_failure;
2876 nla_nest_end(skb, nl_stats);
2881 nla_nest_cancel(skb, nl_stats);
2885 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2886 struct ip_vs_service *svc)
2888 struct ip_vs_scheduler *sched;
2889 struct ip_vs_pe *pe;
2890 struct nlattr *nl_service;
2891 struct ip_vs_flags flags = { .flags = svc->flags,
2894 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2898 if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2899 goto nla_put_failure;
2901 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2902 goto nla_put_failure;
2904 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2905 nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2906 nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2907 goto nla_put_failure;
2910 sched = rcu_dereference_protected(svc->scheduler, 1);
2911 pe = rcu_dereference_protected(svc->pe, 1);
2912 if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched->name) ||
2913 (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
2914 nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2915 nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2916 nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
2917 goto nla_put_failure;
2918 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2919 goto nla_put_failure;
2921 nla_nest_end(skb, nl_service);
2926 nla_nest_cancel(skb, nl_service);
2930 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2931 struct ip_vs_service *svc,
2932 struct netlink_callback *cb)
2936 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2937 &ip_vs_genl_family, NLM_F_MULTI,
2938 IPVS_CMD_NEW_SERVICE);
2942 if (ip_vs_genl_fill_service(skb, svc) < 0)
2943 goto nla_put_failure;
2945 return genlmsg_end(skb, hdr);
2948 genlmsg_cancel(skb, hdr);
2952 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2953 struct netlink_callback *cb)
2956 int start = cb->args[0];
2957 struct ip_vs_service *svc;
2958 struct net *net = skb_sknet(skb);
2960 mutex_lock(&__ip_vs_mutex);
2961 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2962 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2963 if (++idx <= start || !net_eq(svc->net, net))
2965 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2967 goto nla_put_failure;
2972 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2973 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
2974 if (++idx <= start || !net_eq(svc->net, net))
2976 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2978 goto nla_put_failure;
2984 mutex_unlock(&__ip_vs_mutex);
2990 static int ip_vs_genl_parse_service(struct net *net,
2991 struct ip_vs_service_user_kern *usvc,
2992 struct nlattr *nla, int full_entry,
2993 struct ip_vs_service **ret_svc)
2995 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
2996 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
2997 struct ip_vs_service *svc;
2999 /* Parse mandatory identifying service fields first */
3001 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
3004 nla_af = attrs[IPVS_SVC_ATTR_AF];
3005 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
3006 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
3007 nla_port = attrs[IPVS_SVC_ATTR_PORT];
3008 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
3010 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3013 memset(usvc, 0, sizeof(*usvc));
3015 usvc->af = nla_get_u16(nla_af);
3016 #ifdef CONFIG_IP_VS_IPV6
3017 if (usvc->af != AF_INET && usvc->af != AF_INET6)
3019 if (usvc->af != AF_INET)
3021 return -EAFNOSUPPORT;
3024 usvc->protocol = IPPROTO_TCP;
3025 usvc->fwmark = nla_get_u32(nla_fwmark);
3027 usvc->protocol = nla_get_u16(nla_protocol);
3028 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3029 usvc->port = nla_get_be16(nla_port);
3035 svc = __ip_vs_svc_fwm_find(net, usvc->af, usvc->fwmark);
3037 svc = __ip_vs_service_find(net, usvc->af, usvc->protocol,
3038 &usvc->addr, usvc->port);
3042 /* If a full entry was requested, check for the additional fields */
3044 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3046 struct ip_vs_flags flags;
3048 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3049 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3050 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3051 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3052 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3054 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3057 nla_memcpy(&flags, nla_flags, sizeof(flags));
3059 /* prefill flags from service if it already exists */
3061 usvc->flags = svc->flags;
3063 /* set new flags from userland */
3064 usvc->flags = (usvc->flags & ~flags.mask) |
3065 (flags.flags & flags.mask);
3066 usvc->sched_name = nla_data(nla_sched);
3067 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3068 usvc->timeout = nla_get_u32(nla_timeout);
3069 usvc->netmask = nla_get_be32(nla_netmask);
3075 static struct ip_vs_service *ip_vs_genl_find_service(struct net *net,
3078 struct ip_vs_service_user_kern usvc;
3079 struct ip_vs_service *svc;
3082 ret = ip_vs_genl_parse_service(net, &usvc, nla, 0, &svc);
3083 return ret ? ERR_PTR(ret) : svc;
3086 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3088 struct nlattr *nl_dest;
3090 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3094 if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3095 nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3096 nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3097 (atomic_read(&dest->conn_flags) &
3098 IP_VS_CONN_F_FWD_MASK)) ||
3099 nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3100 atomic_read(&dest->weight)) ||
3101 nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3102 nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3103 nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3104 atomic_read(&dest->activeconns)) ||
3105 nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3106 atomic_read(&dest->inactconns)) ||
3107 nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3108 atomic_read(&dest->persistconns)))
3109 goto nla_put_failure;
3110 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
3111 goto nla_put_failure;
3113 nla_nest_end(skb, nl_dest);
3118 nla_nest_cancel(skb, nl_dest);
3122 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3123 struct netlink_callback *cb)
3127 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3128 &ip_vs_genl_family, NLM_F_MULTI,
3133 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3134 goto nla_put_failure;
3136 return genlmsg_end(skb, hdr);
3139 genlmsg_cancel(skb, hdr);
3143 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3144 struct netlink_callback *cb)
3147 int start = cb->args[0];
3148 struct ip_vs_service *svc;
3149 struct ip_vs_dest *dest;
3150 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3151 struct net *net = skb_sknet(skb);
3153 mutex_lock(&__ip_vs_mutex);
3155 /* Try to find the service for which to dump destinations */
3156 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3157 IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3161 svc = ip_vs_genl_find_service(net, attrs[IPVS_CMD_ATTR_SERVICE]);
3162 if (IS_ERR(svc) || svc == NULL)
3165 /* Dump the destinations */
3166 list_for_each_entry(dest, &svc->destinations, n_list) {
3169 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3171 goto nla_put_failure;
3179 mutex_unlock(&__ip_vs_mutex);
3184 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3185 struct nlattr *nla, int full_entry)
3187 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3188 struct nlattr *nla_addr, *nla_port;
3190 /* Parse mandatory identifying destination fields first */
3192 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3195 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3196 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3198 if (!(nla_addr && nla_port))
3201 memset(udest, 0, sizeof(*udest));
3203 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3204 udest->port = nla_get_be16(nla_port);
3206 /* If a full entry was requested, check for the additional fields */
3208 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3211 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3212 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3213 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3214 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3216 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3219 udest->conn_flags = nla_get_u32(nla_fwd)
3220 & IP_VS_CONN_F_FWD_MASK;
3221 udest->weight = nla_get_u32(nla_weight);
3222 udest->u_threshold = nla_get_u32(nla_u_thresh);
3223 udest->l_threshold = nla_get_u32(nla_l_thresh);
3229 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3230 const char *mcast_ifn, __u32 syncid)
3232 struct nlattr *nl_daemon;
3234 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3238 if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3239 nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn) ||
3240 nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid))
3241 goto nla_put_failure;
3242 nla_nest_end(skb, nl_daemon);
3247 nla_nest_cancel(skb, nl_daemon);
3251 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3252 const char *mcast_ifn, __u32 syncid,
3253 struct netlink_callback *cb)
3256 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3257 &ip_vs_genl_family, NLM_F_MULTI,
3258 IPVS_CMD_NEW_DAEMON);
3262 if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
3263 goto nla_put_failure;
3265 return genlmsg_end(skb, hdr);
3268 genlmsg_cancel(skb, hdr);
3272 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3273 struct netlink_callback *cb)
3275 struct net *net = skb_sknet(skb);
3276 struct netns_ipvs *ipvs = net_ipvs(net);
3278 mutex_lock(&ipvs->sync_mutex);
3279 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3280 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3281 ipvs->master_mcast_ifn,
3282 ipvs->master_syncid, cb) < 0)
3283 goto nla_put_failure;
3288 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3289 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3290 ipvs->backup_mcast_ifn,
3291 ipvs->backup_syncid, cb) < 0)
3292 goto nla_put_failure;
3298 mutex_unlock(&ipvs->sync_mutex);
3303 static int ip_vs_genl_new_daemon(struct net *net, struct nlattr **attrs)
3305 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3306 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3307 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3310 return start_sync_thread(net,
3311 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3312 nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3313 nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3316 static int ip_vs_genl_del_daemon(struct net *net, struct nlattr **attrs)
3318 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3321 return stop_sync_thread(net,
3322 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3325 static int ip_vs_genl_set_config(struct net *net, struct nlattr **attrs)
3327 struct ip_vs_timeout_user t;
3329 __ip_vs_get_timeouts(net, &t);
3331 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3332 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3334 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3336 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3338 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3339 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3341 return ip_vs_set_timeout(net, &t);
3344 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3348 struct netns_ipvs *ipvs;
3350 net = skb_sknet(skb);
3351 ipvs = net_ipvs(net);
3352 cmd = info->genlhdr->cmd;
3354 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3355 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3357 mutex_lock(&ipvs->sync_mutex);
3358 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3359 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3360 info->attrs[IPVS_CMD_ATTR_DAEMON],
3361 ip_vs_daemon_policy)) {
3366 if (cmd == IPVS_CMD_NEW_DAEMON)
3367 ret = ip_vs_genl_new_daemon(net, daemon_attrs);
3369 ret = ip_vs_genl_del_daemon(net, daemon_attrs);
3371 mutex_unlock(&ipvs->sync_mutex);
3376 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3378 struct ip_vs_service *svc = NULL;
3379 struct ip_vs_service_user_kern usvc;
3380 struct ip_vs_dest_user_kern udest;
3382 int need_full_svc = 0, need_full_dest = 0;
3385 net = skb_sknet(skb);
3386 cmd = info->genlhdr->cmd;
3388 mutex_lock(&__ip_vs_mutex);
3390 if (cmd == IPVS_CMD_FLUSH) {
3391 ret = ip_vs_flush(net, false);
3393 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3394 ret = ip_vs_genl_set_config(net, info->attrs);
3396 } else if (cmd == IPVS_CMD_ZERO &&
3397 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3398 ret = ip_vs_zero_all(net);
3402 /* All following commands require a service argument, so check if we
3403 * received a valid one. We need a full service specification when
3404 * adding / editing a service. Only identifying members otherwise. */
3405 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3408 ret = ip_vs_genl_parse_service(net, &usvc,
3409 info->attrs[IPVS_CMD_ATTR_SERVICE],
3410 need_full_svc, &svc);
3414 /* Unless we're adding a new service, the service must already exist */
3415 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3420 /* Destination commands require a valid destination argument. For
3421 * adding / editing a destination, we need a full destination
3423 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3424 cmd == IPVS_CMD_DEL_DEST) {
3425 if (cmd != IPVS_CMD_DEL_DEST)
3428 ret = ip_vs_genl_parse_dest(&udest,
3429 info->attrs[IPVS_CMD_ATTR_DEST],
3436 case IPVS_CMD_NEW_SERVICE:
3438 ret = ip_vs_add_service(net, &usvc, &svc);
3442 case IPVS_CMD_SET_SERVICE:
3443 ret = ip_vs_edit_service(svc, &usvc);
3445 case IPVS_CMD_DEL_SERVICE:
3446 ret = ip_vs_del_service(svc);
3447 /* do not use svc, it can be freed */
3449 case IPVS_CMD_NEW_DEST:
3450 ret = ip_vs_add_dest(svc, &udest);
3452 case IPVS_CMD_SET_DEST:
3453 ret = ip_vs_edit_dest(svc, &udest);
3455 case IPVS_CMD_DEL_DEST:
3456 ret = ip_vs_del_dest(svc, &udest);
3459 ret = ip_vs_zero_service(svc);
3466 mutex_unlock(&__ip_vs_mutex);
3471 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3473 struct sk_buff *msg;
3475 int ret, cmd, reply_cmd;
3478 net = skb_sknet(skb);
3479 cmd = info->genlhdr->cmd;
3481 if (cmd == IPVS_CMD_GET_SERVICE)
3482 reply_cmd = IPVS_CMD_NEW_SERVICE;
3483 else if (cmd == IPVS_CMD_GET_INFO)
3484 reply_cmd = IPVS_CMD_SET_INFO;
3485 else if (cmd == IPVS_CMD_GET_CONFIG)
3486 reply_cmd = IPVS_CMD_SET_CONFIG;
3488 pr_err("unknown Generic Netlink command\n");
3492 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3496 mutex_lock(&__ip_vs_mutex);
3498 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3500 goto nla_put_failure;
3503 case IPVS_CMD_GET_SERVICE:
3505 struct ip_vs_service *svc;
3507 svc = ip_vs_genl_find_service(net,
3508 info->attrs[IPVS_CMD_ATTR_SERVICE]);
3513 ret = ip_vs_genl_fill_service(msg, svc);
3515 goto nla_put_failure;
3524 case IPVS_CMD_GET_CONFIG:
3526 struct ip_vs_timeout_user t;
3528 __ip_vs_get_timeouts(net, &t);
3529 #ifdef CONFIG_IP_VS_PROTO_TCP
3530 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3532 nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3534 goto nla_put_failure;
3536 #ifdef CONFIG_IP_VS_PROTO_UDP
3537 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3538 goto nla_put_failure;
3544 case IPVS_CMD_GET_INFO:
3545 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3546 IP_VS_VERSION_CODE) ||
3547 nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3548 ip_vs_conn_tab_size))
3549 goto nla_put_failure;
3553 genlmsg_end(msg, reply);
3554 ret = genlmsg_reply(msg, info);
3558 pr_err("not enough space in Netlink message\n");
3564 mutex_unlock(&__ip_vs_mutex);
3570 static struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3572 .cmd = IPVS_CMD_NEW_SERVICE,
3573 .flags = GENL_ADMIN_PERM,
3574 .policy = ip_vs_cmd_policy,
3575 .doit = ip_vs_genl_set_cmd,
3578 .cmd = IPVS_CMD_SET_SERVICE,
3579 .flags = GENL_ADMIN_PERM,
3580 .policy = ip_vs_cmd_policy,
3581 .doit = ip_vs_genl_set_cmd,
3584 .cmd = IPVS_CMD_DEL_SERVICE,
3585 .flags = GENL_ADMIN_PERM,
3586 .policy = ip_vs_cmd_policy,
3587 .doit = ip_vs_genl_set_cmd,
3590 .cmd = IPVS_CMD_GET_SERVICE,
3591 .flags = GENL_ADMIN_PERM,
3592 .doit = ip_vs_genl_get_cmd,
3593 .dumpit = ip_vs_genl_dump_services,
3594 .policy = ip_vs_cmd_policy,
3597 .cmd = IPVS_CMD_NEW_DEST,
3598 .flags = GENL_ADMIN_PERM,
3599 .policy = ip_vs_cmd_policy,
3600 .doit = ip_vs_genl_set_cmd,
3603 .cmd = IPVS_CMD_SET_DEST,
3604 .flags = GENL_ADMIN_PERM,
3605 .policy = ip_vs_cmd_policy,
3606 .doit = ip_vs_genl_set_cmd,
3609 .cmd = IPVS_CMD_DEL_DEST,
3610 .flags = GENL_ADMIN_PERM,
3611 .policy = ip_vs_cmd_policy,
3612 .doit = ip_vs_genl_set_cmd,
3615 .cmd = IPVS_CMD_GET_DEST,
3616 .flags = GENL_ADMIN_PERM,
3617 .policy = ip_vs_cmd_policy,
3618 .dumpit = ip_vs_genl_dump_dests,
3621 .cmd = IPVS_CMD_NEW_DAEMON,
3622 .flags = GENL_ADMIN_PERM,
3623 .policy = ip_vs_cmd_policy,
3624 .doit = ip_vs_genl_set_daemon,
3627 .cmd = IPVS_CMD_DEL_DAEMON,
3628 .flags = GENL_ADMIN_PERM,
3629 .policy = ip_vs_cmd_policy,
3630 .doit = ip_vs_genl_set_daemon,
3633 .cmd = IPVS_CMD_GET_DAEMON,
3634 .flags = GENL_ADMIN_PERM,
3635 .dumpit = ip_vs_genl_dump_daemons,
3638 .cmd = IPVS_CMD_SET_CONFIG,
3639 .flags = GENL_ADMIN_PERM,
3640 .policy = ip_vs_cmd_policy,
3641 .doit = ip_vs_genl_set_cmd,
3644 .cmd = IPVS_CMD_GET_CONFIG,
3645 .flags = GENL_ADMIN_PERM,
3646 .doit = ip_vs_genl_get_cmd,
3649 .cmd = IPVS_CMD_GET_INFO,
3650 .flags = GENL_ADMIN_PERM,
3651 .doit = ip_vs_genl_get_cmd,
3654 .cmd = IPVS_CMD_ZERO,
3655 .flags = GENL_ADMIN_PERM,
3656 .policy = ip_vs_cmd_policy,
3657 .doit = ip_vs_genl_set_cmd,
3660 .cmd = IPVS_CMD_FLUSH,
3661 .flags = GENL_ADMIN_PERM,
3662 .doit = ip_vs_genl_set_cmd,
3666 static int __init ip_vs_genl_register(void)
3668 return genl_register_family_with_ops(&ip_vs_genl_family,
3669 ip_vs_genl_ops, ARRAY_SIZE(ip_vs_genl_ops));
3672 static void ip_vs_genl_unregister(void)
3674 genl_unregister_family(&ip_vs_genl_family);
3677 /* End of Generic Netlink interface definitions */
3680 * per netns intit/exit func.
3682 #ifdef CONFIG_SYSCTL
3683 static int __net_init ip_vs_control_net_init_sysctl(struct net *net)
3686 struct netns_ipvs *ipvs = net_ipvs(net);
3687 struct ctl_table *tbl;
3689 atomic_set(&ipvs->dropentry, 0);
3690 spin_lock_init(&ipvs->dropentry_lock);
3691 spin_lock_init(&ipvs->droppacket_lock);
3692 spin_lock_init(&ipvs->securetcp_lock);
3694 if (!net_eq(net, &init_net)) {
3695 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3699 /* Don't export sysctls to unprivileged users */
3700 if (net->user_ns != &init_user_ns)
3701 tbl[0].procname = NULL;
3704 /* Initialize sysctl defaults */
3706 ipvs->sysctl_amemthresh = 1024;
3707 tbl[idx++].data = &ipvs->sysctl_amemthresh;
3708 ipvs->sysctl_am_droprate = 10;
3709 tbl[idx++].data = &ipvs->sysctl_am_droprate;
3710 tbl[idx++].data = &ipvs->sysctl_drop_entry;
3711 tbl[idx++].data = &ipvs->sysctl_drop_packet;
3712 #ifdef CONFIG_IP_VS_NFCT
3713 tbl[idx++].data = &ipvs->sysctl_conntrack;
3715 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3716 ipvs->sysctl_snat_reroute = 1;
3717 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3718 ipvs->sysctl_sync_ver = 1;
3719 tbl[idx++].data = &ipvs->sysctl_sync_ver;
3720 ipvs->sysctl_sync_ports = 1;
3721 tbl[idx++].data = &ipvs->sysctl_sync_ports;
3722 tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
3723 ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3724 tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3725 ipvs->sysctl_sync_sock_size = 0;
3726 tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3727 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3728 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3729 tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
3730 tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
3731 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3732 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3733 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3734 tbl[idx].data = &ipvs->sysctl_sync_threshold;
3735 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3736 ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3737 tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3738 ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3739 tbl[idx++].data = &ipvs->sysctl_sync_retries;
3740 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3741 ipvs->sysctl_pmtu_disc = 1;
3742 tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
3743 tbl[idx++].data = &ipvs->sysctl_backup_only;
3746 ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3747 if (ipvs->sysctl_hdr == NULL) {
3748 if (!net_eq(net, &init_net))
3752 ip_vs_start_estimator(net, &ipvs->tot_stats);
3753 ipvs->sysctl_tbl = tbl;
3754 /* Schedule defense work */
3755 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3756 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3761 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net)
3763 struct netns_ipvs *ipvs = net_ipvs(net);
3765 cancel_delayed_work_sync(&ipvs->defense_work);
3766 cancel_work_sync(&ipvs->defense_work.work);
3767 unregister_net_sysctl_table(ipvs->sysctl_hdr);
3772 static int __net_init ip_vs_control_net_init_sysctl(struct net *net) { return 0; }
3773 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net) { }
3777 static struct notifier_block ip_vs_dst_notifier = {
3778 .notifier_call = ip_vs_dst_event,
3781 int __net_init ip_vs_control_net_init(struct net *net)
3784 struct netns_ipvs *ipvs = net_ipvs(net);
3786 /* Initialize rs_table */
3787 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
3788 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
3790 INIT_LIST_HEAD(&ipvs->dest_trash);
3791 spin_lock_init(&ipvs->dest_trash_lock);
3792 setup_timer(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire,
3793 (unsigned long) net);
3794 atomic_set(&ipvs->ftpsvc_counter, 0);
3795 atomic_set(&ipvs->nullsvc_counter, 0);
3798 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
3799 if (!ipvs->tot_stats.cpustats)
3802 spin_lock_init(&ipvs->tot_stats.lock);
3804 proc_create("ip_vs", 0, net->proc_net, &ip_vs_info_fops);
3805 proc_create("ip_vs_stats", 0, net->proc_net, &ip_vs_stats_fops);
3806 proc_create("ip_vs_stats_percpu", 0, net->proc_net,
3807 &ip_vs_stats_percpu_fops);
3809 if (ip_vs_control_net_init_sysctl(net))
3815 free_percpu(ipvs->tot_stats.cpustats);
3819 void __net_exit ip_vs_control_net_cleanup(struct net *net)
3821 struct netns_ipvs *ipvs = net_ipvs(net);
3823 /* Some dest can be in grace period even before cleanup, we have to
3824 * defer ip_vs_trash_cleanup until ip_vs_dest_wait_readers is called.
3827 ip_vs_trash_cleanup(net);
3828 ip_vs_stop_estimator(net, &ipvs->tot_stats);
3829 ip_vs_control_net_cleanup_sysctl(net);
3830 remove_proc_entry("ip_vs_stats_percpu", net->proc_net);
3831 remove_proc_entry("ip_vs_stats", net->proc_net);
3832 remove_proc_entry("ip_vs", net->proc_net);
3833 free_percpu(ipvs->tot_stats.cpustats);
3836 int __init ip_vs_register_nl_ioctl(void)
3840 ret = nf_register_sockopt(&ip_vs_sockopts);
3842 pr_err("cannot register sockopt.\n");
3846 ret = ip_vs_genl_register();
3848 pr_err("cannot register Generic Netlink interface.\n");
3854 nf_unregister_sockopt(&ip_vs_sockopts);
3859 void ip_vs_unregister_nl_ioctl(void)
3861 ip_vs_genl_unregister();
3862 nf_unregister_sockopt(&ip_vs_sockopts);
3865 int __init ip_vs_control_init(void)
3872 /* Initialize svc_table, ip_vs_svc_fwm_table */
3873 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
3874 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
3875 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3878 smp_wmb(); /* Do we really need it now ? */
3880 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
3889 void ip_vs_control_cleanup(void)
3892 unregister_netdevice_notifier(&ip_vs_dst_notifier);