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);
707 ip_vs_dest_put_and_free(dest);
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;
845 unsigned int atype, i;
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)
872 for_each_possible_cpu(i) {
873 struct ip_vs_cpu_stats *ip_vs_dest_stats;
874 ip_vs_dest_stats = per_cpu_ptr(dest->stats.cpustats, i);
875 u64_stats_init(&ip_vs_dest_stats->syncp);
879 dest->protocol = svc->protocol;
880 dest->vaddr = svc->addr;
881 dest->vport = svc->port;
882 dest->vfwmark = svc->fwmark;
883 ip_vs_addr_copy(svc->af, &dest->addr, &udest->addr);
884 dest->port = udest->port;
886 atomic_set(&dest->activeconns, 0);
887 atomic_set(&dest->inactconns, 0);
888 atomic_set(&dest->persistconns, 0);
889 atomic_set(&dest->refcnt, 1);
891 INIT_HLIST_NODE(&dest->d_list);
892 spin_lock_init(&dest->dst_lock);
893 spin_lock_init(&dest->stats.lock);
894 __ip_vs_update_dest(svc, dest, udest, 1);
908 * Add a destination into an existing service
911 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
913 struct ip_vs_dest *dest;
914 union nf_inet_addr daddr;
915 __be16 dport = udest->port;
920 if (udest->weight < 0) {
921 pr_err("%s(): server weight less than zero\n", __func__);
925 if (udest->l_threshold > udest->u_threshold) {
926 pr_err("%s(): lower threshold is higher than upper threshold\n",
931 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
933 /* We use function that requires RCU lock */
935 dest = ip_vs_lookup_dest(svc, &daddr, dport);
939 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
944 * Check if the dest already exists in the trash and
945 * is from the same service
947 dest = ip_vs_trash_get_dest(svc, &daddr, dport);
950 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
951 "dest->refcnt=%d, service %u/%s:%u\n",
952 IP_VS_DBG_ADDR(svc->af, &daddr), ntohs(dport),
953 atomic_read(&dest->refcnt),
955 IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
958 __ip_vs_update_dest(svc, dest, udest, 1);
962 * Allocate and initialize the dest structure
964 ret = ip_vs_new_dest(svc, udest, &dest);
973 * Edit a destination in the given service
976 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
978 struct ip_vs_dest *dest;
979 union nf_inet_addr daddr;
980 __be16 dport = udest->port;
984 if (udest->weight < 0) {
985 pr_err("%s(): server weight less than zero\n", __func__);
989 if (udest->l_threshold > udest->u_threshold) {
990 pr_err("%s(): lower threshold is higher than upper threshold\n",
995 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
997 /* We use function that requires RCU lock */
999 dest = ip_vs_lookup_dest(svc, &daddr, dport);
1003 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1007 __ip_vs_update_dest(svc, dest, udest, 0);
1014 * Delete a destination (must be already unlinked from the service)
1016 static void __ip_vs_del_dest(struct net *net, struct ip_vs_dest *dest,
1019 struct netns_ipvs *ipvs = net_ipvs(net);
1021 ip_vs_stop_estimator(net, &dest->stats);
1024 * Remove it from the d-linked list with the real services.
1026 ip_vs_rs_unhash(dest);
1028 spin_lock_bh(&ipvs->dest_trash_lock);
1029 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1030 IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port),
1031 atomic_read(&dest->refcnt));
1032 if (list_empty(&ipvs->dest_trash) && !cleanup)
1033 mod_timer(&ipvs->dest_trash_timer,
1034 jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1035 /* dest lives in trash without reference */
1036 list_add(&dest->t_list, &ipvs->dest_trash);
1037 dest->idle_start = 0;
1038 spin_unlock_bh(&ipvs->dest_trash_lock);
1039 ip_vs_dest_put(dest);
1044 * Unlink a destination from the given service
1046 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1047 struct ip_vs_dest *dest,
1050 dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1053 * Remove it from the d-linked destination list.
1055 list_del_rcu(&dest->n_list);
1059 struct ip_vs_scheduler *sched;
1061 sched = rcu_dereference_protected(svc->scheduler, 1);
1062 if (sched->del_dest)
1063 sched->del_dest(svc, dest);
1069 * Delete a destination server in the given service
1072 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1074 struct ip_vs_dest *dest;
1075 __be16 dport = udest->port;
1079 /* We use function that requires RCU lock */
1081 dest = ip_vs_lookup_dest(svc, &udest->addr, dport);
1085 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1090 * Unlink dest from the service
1092 __ip_vs_unlink_dest(svc, dest, 1);
1095 * Delete the destination
1097 __ip_vs_del_dest(svc->net, dest, false);
1104 static void ip_vs_dest_trash_expire(unsigned long data)
1106 struct net *net = (struct net *) data;
1107 struct netns_ipvs *ipvs = net_ipvs(net);
1108 struct ip_vs_dest *dest, *next;
1109 unsigned long now = jiffies;
1111 spin_lock(&ipvs->dest_trash_lock);
1112 list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) {
1113 if (atomic_read(&dest->refcnt) > 0)
1115 if (dest->idle_start) {
1116 if (time_before(now, dest->idle_start +
1117 IP_VS_DEST_TRASH_PERIOD))
1120 dest->idle_start = max(1UL, now);
1123 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1125 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1127 list_del(&dest->t_list);
1128 ip_vs_dest_free(dest);
1130 if (!list_empty(&ipvs->dest_trash))
1131 mod_timer(&ipvs->dest_trash_timer,
1132 jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1133 spin_unlock(&ipvs->dest_trash_lock);
1137 * Add a service into the service hash table
1140 ip_vs_add_service(struct net *net, struct ip_vs_service_user_kern *u,
1141 struct ip_vs_service **svc_p)
1144 struct ip_vs_scheduler *sched = NULL;
1145 struct ip_vs_pe *pe = NULL;
1146 struct ip_vs_service *svc = NULL;
1147 struct netns_ipvs *ipvs = net_ipvs(net);
1149 /* increase the module use count */
1150 ip_vs_use_count_inc();
1152 /* Lookup the scheduler by 'u->sched_name' */
1153 sched = ip_vs_scheduler_get(u->sched_name);
1154 if (sched == NULL) {
1155 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1160 if (u->pe_name && *u->pe_name) {
1161 pe = ip_vs_pe_getbyname(u->pe_name);
1163 pr_info("persistence engine module ip_vs_pe_%s "
1164 "not found\n", u->pe_name);
1170 #ifdef CONFIG_IP_VS_IPV6
1171 if (u->af == AF_INET6) {
1172 __u32 plen = (__force __u32) u->netmask;
1174 if (plen < 1 || plen > 128) {
1181 svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1183 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1187 svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1188 if (!svc->stats.cpustats) {
1193 for_each_possible_cpu(i) {
1194 struct ip_vs_cpu_stats *ip_vs_stats;
1195 ip_vs_stats = per_cpu_ptr(svc->stats.cpustats, i);
1196 u64_stats_init(&ip_vs_stats->syncp);
1200 /* I'm the first user of the service */
1201 atomic_set(&svc->refcnt, 0);
1204 svc->protocol = u->protocol;
1205 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1206 svc->port = u->port;
1207 svc->fwmark = u->fwmark;
1208 svc->flags = u->flags;
1209 svc->timeout = u->timeout * HZ;
1210 svc->netmask = u->netmask;
1213 INIT_LIST_HEAD(&svc->destinations);
1214 spin_lock_init(&svc->sched_lock);
1215 spin_lock_init(&svc->stats.lock);
1217 /* Bind the scheduler */
1218 ret = ip_vs_bind_scheduler(svc, sched);
1223 /* Bind the ct retriever */
1224 RCU_INIT_POINTER(svc->pe, pe);
1227 /* Update the virtual service counters */
1228 if (svc->port == FTPPORT)
1229 atomic_inc(&ipvs->ftpsvc_counter);
1230 else if (svc->port == 0)
1231 atomic_inc(&ipvs->nullsvc_counter);
1233 ip_vs_start_estimator(net, &svc->stats);
1235 /* Count only IPv4 services for old get/setsockopt interface */
1236 if (svc->af == AF_INET)
1237 ipvs->num_services++;
1239 /* Hash the service into the service table */
1240 ip_vs_svc_hash(svc);
1243 /* Now there is a service - full throttle */
1250 ip_vs_unbind_scheduler(svc, sched);
1251 ip_vs_service_free(svc);
1253 ip_vs_scheduler_put(sched);
1256 /* decrease the module use count */
1257 ip_vs_use_count_dec();
1264 * Edit a service and bind it with a new scheduler
1267 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1269 struct ip_vs_scheduler *sched, *old_sched;
1270 struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1274 * Lookup the scheduler, by 'u->sched_name'
1276 sched = ip_vs_scheduler_get(u->sched_name);
1277 if (sched == NULL) {
1278 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1283 if (u->pe_name && *u->pe_name) {
1284 pe = ip_vs_pe_getbyname(u->pe_name);
1286 pr_info("persistence engine module ip_vs_pe_%s "
1287 "not found\n", u->pe_name);
1294 #ifdef CONFIG_IP_VS_IPV6
1295 if (u->af == AF_INET6) {
1296 __u32 plen = (__force __u32) u->netmask;
1298 if (plen < 1 || plen > 128) {
1305 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1306 if (sched != old_sched) {
1307 /* Bind the new scheduler */
1308 ret = ip_vs_bind_scheduler(svc, sched);
1313 /* Unbind the old scheduler on success */
1314 ip_vs_unbind_scheduler(svc, old_sched);
1318 * Set the flags and timeout value
1320 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1321 svc->timeout = u->timeout * HZ;
1322 svc->netmask = u->netmask;
1324 old_pe = rcu_dereference_protected(svc->pe, 1);
1326 rcu_assign_pointer(svc->pe, pe);
1329 ip_vs_scheduler_put(old_sched);
1330 ip_vs_pe_put(old_pe);
1335 * Delete a service from the service list
1336 * - The service must be unlinked, unlocked and not referenced!
1337 * - We are called under _bh lock
1339 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1341 struct ip_vs_dest *dest, *nxt;
1342 struct ip_vs_scheduler *old_sched;
1343 struct ip_vs_pe *old_pe;
1344 struct netns_ipvs *ipvs = net_ipvs(svc->net);
1346 pr_info("%s: enter\n", __func__);
1348 /* Count only IPv4 services for old get/setsockopt interface */
1349 if (svc->af == AF_INET)
1350 ipvs->num_services--;
1352 ip_vs_stop_estimator(svc->net, &svc->stats);
1354 /* Unbind scheduler */
1355 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1356 ip_vs_unbind_scheduler(svc, old_sched);
1357 ip_vs_scheduler_put(old_sched);
1359 /* Unbind persistence engine, keep svc->pe */
1360 old_pe = rcu_dereference_protected(svc->pe, 1);
1361 ip_vs_pe_put(old_pe);
1364 * Unlink the whole destination list
1366 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1367 __ip_vs_unlink_dest(svc, dest, 0);
1368 __ip_vs_del_dest(svc->net, dest, cleanup);
1372 * Update the virtual service counters
1374 if (svc->port == FTPPORT)
1375 atomic_dec(&ipvs->ftpsvc_counter);
1376 else if (svc->port == 0)
1377 atomic_dec(&ipvs->nullsvc_counter);
1380 * Free the service if nobody refers to it
1382 __ip_vs_svc_put(svc, true);
1384 /* decrease the module use count */
1385 ip_vs_use_count_dec();
1389 * Unlink a service from list and try to delete it if its refcnt reached 0
1391 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1393 /* Hold svc to avoid double release from dest_trash */
1394 atomic_inc(&svc->refcnt);
1396 * Unhash it from the service table
1398 ip_vs_svc_unhash(svc);
1400 __ip_vs_del_service(svc, cleanup);
1404 * Delete a service from the service list
1406 static int ip_vs_del_service(struct ip_vs_service *svc)
1410 ip_vs_unlink_service(svc, false);
1417 * Flush all the virtual services
1419 static int ip_vs_flush(struct net *net, bool cleanup)
1422 struct ip_vs_service *svc;
1423 struct hlist_node *n;
1426 * Flush the service table hashed by <netns,protocol,addr,port>
1428 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1429 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1431 if (net_eq(svc->net, net))
1432 ip_vs_unlink_service(svc, cleanup);
1437 * Flush the service table hashed by fwmark
1439 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1440 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1442 if (net_eq(svc->net, net))
1443 ip_vs_unlink_service(svc, cleanup);
1451 * Delete service by {netns} in the service table.
1452 * Called by __ip_vs_cleanup()
1454 void ip_vs_service_net_cleanup(struct net *net)
1457 /* Check for "full" addressed entries */
1458 mutex_lock(&__ip_vs_mutex);
1459 ip_vs_flush(net, true);
1460 mutex_unlock(&__ip_vs_mutex);
1464 /* Put all references for device (dst_cache) */
1466 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1468 struct ip_vs_dest_dst *dest_dst;
1470 spin_lock_bh(&dest->dst_lock);
1471 dest_dst = rcu_dereference_protected(dest->dest_dst, 1);
1472 if (dest_dst && dest_dst->dst_cache->dev == dev) {
1473 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1475 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1477 atomic_read(&dest->refcnt));
1478 __ip_vs_dst_cache_reset(dest);
1480 spin_unlock_bh(&dest->dst_lock);
1483 /* Netdev event receiver
1484 * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1486 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1489 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1490 struct net *net = dev_net(dev);
1491 struct netns_ipvs *ipvs = net_ipvs(net);
1492 struct ip_vs_service *svc;
1493 struct ip_vs_dest *dest;
1496 if (event != NETDEV_DOWN || !ipvs)
1498 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1500 mutex_lock(&__ip_vs_mutex);
1501 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1502 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1503 if (net_eq(svc->net, net)) {
1504 list_for_each_entry(dest, &svc->destinations,
1506 ip_vs_forget_dev(dest, dev);
1511 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1512 if (net_eq(svc->net, net)) {
1513 list_for_each_entry(dest, &svc->destinations,
1515 ip_vs_forget_dev(dest, dev);
1522 spin_lock_bh(&ipvs->dest_trash_lock);
1523 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1524 ip_vs_forget_dev(dest, dev);
1526 spin_unlock_bh(&ipvs->dest_trash_lock);
1527 mutex_unlock(&__ip_vs_mutex);
1533 * Zero counters in a service or all services
1535 static int ip_vs_zero_service(struct ip_vs_service *svc)
1537 struct ip_vs_dest *dest;
1539 list_for_each_entry(dest, &svc->destinations, n_list) {
1540 ip_vs_zero_stats(&dest->stats);
1542 ip_vs_zero_stats(&svc->stats);
1546 static int ip_vs_zero_all(struct net *net)
1549 struct ip_vs_service *svc;
1551 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1552 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1553 if (net_eq(svc->net, net))
1554 ip_vs_zero_service(svc);
1558 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1559 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1560 if (net_eq(svc->net, net))
1561 ip_vs_zero_service(svc);
1565 ip_vs_zero_stats(&net_ipvs(net)->tot_stats);
1569 #ifdef CONFIG_SYSCTL
1572 static int three = 3;
1575 proc_do_defense_mode(struct ctl_table *table, int write,
1576 void __user *buffer, size_t *lenp, loff_t *ppos)
1578 struct net *net = current->nsproxy->net_ns;
1579 int *valp = table->data;
1583 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1584 if (write && (*valp != val)) {
1585 if ((*valp < 0) || (*valp > 3)) {
1586 /* Restore the correct value */
1589 update_defense_level(net_ipvs(net));
1596 proc_do_sync_threshold(struct ctl_table *table, int write,
1597 void __user *buffer, size_t *lenp, loff_t *ppos)
1599 int *valp = table->data;
1603 /* backup the value first */
1604 memcpy(val, valp, sizeof(val));
1606 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1607 if (write && (valp[0] < 0 || valp[1] < 0 ||
1608 (valp[0] >= valp[1] && valp[1]))) {
1609 /* Restore the correct value */
1610 memcpy(valp, val, sizeof(val));
1616 proc_do_sync_mode(struct ctl_table *table, int write,
1617 void __user *buffer, size_t *lenp, loff_t *ppos)
1619 int *valp = table->data;
1623 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1624 if (write && (*valp != val)) {
1625 if ((*valp < 0) || (*valp > 1)) {
1626 /* Restore the correct value */
1634 proc_do_sync_ports(struct ctl_table *table, int write,
1635 void __user *buffer, size_t *lenp, loff_t *ppos)
1637 int *valp = table->data;
1641 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1642 if (write && (*valp != val)) {
1643 if (*valp < 1 || !is_power_of_2(*valp)) {
1644 /* Restore the correct value */
1652 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1653 * Do not change order or insert new entries without
1654 * align with netns init in ip_vs_control_net_init()
1657 static struct ctl_table vs_vars[] = {
1659 .procname = "amemthresh",
1660 .maxlen = sizeof(int),
1662 .proc_handler = proc_dointvec,
1665 .procname = "am_droprate",
1666 .maxlen = sizeof(int),
1668 .proc_handler = proc_dointvec,
1671 .procname = "drop_entry",
1672 .maxlen = sizeof(int),
1674 .proc_handler = proc_do_defense_mode,
1677 .procname = "drop_packet",
1678 .maxlen = sizeof(int),
1680 .proc_handler = proc_do_defense_mode,
1682 #ifdef CONFIG_IP_VS_NFCT
1684 .procname = "conntrack",
1685 .maxlen = sizeof(int),
1687 .proc_handler = &proc_dointvec,
1691 .procname = "secure_tcp",
1692 .maxlen = sizeof(int),
1694 .proc_handler = proc_do_defense_mode,
1697 .procname = "snat_reroute",
1698 .maxlen = sizeof(int),
1700 .proc_handler = &proc_dointvec,
1703 .procname = "sync_version",
1704 .maxlen = sizeof(int),
1706 .proc_handler = &proc_do_sync_mode,
1709 .procname = "sync_ports",
1710 .maxlen = sizeof(int),
1712 .proc_handler = &proc_do_sync_ports,
1715 .procname = "sync_persist_mode",
1716 .maxlen = sizeof(int),
1718 .proc_handler = proc_dointvec,
1721 .procname = "sync_qlen_max",
1722 .maxlen = sizeof(unsigned long),
1724 .proc_handler = proc_doulongvec_minmax,
1727 .procname = "sync_sock_size",
1728 .maxlen = sizeof(int),
1730 .proc_handler = proc_dointvec,
1733 .procname = "cache_bypass",
1734 .maxlen = sizeof(int),
1736 .proc_handler = proc_dointvec,
1739 .procname = "expire_nodest_conn",
1740 .maxlen = sizeof(int),
1742 .proc_handler = proc_dointvec,
1745 .procname = "sloppy_tcp",
1746 .maxlen = sizeof(int),
1748 .proc_handler = proc_dointvec,
1751 .procname = "sloppy_sctp",
1752 .maxlen = sizeof(int),
1754 .proc_handler = proc_dointvec,
1757 .procname = "expire_quiescent_template",
1758 .maxlen = sizeof(int),
1760 .proc_handler = proc_dointvec,
1763 .procname = "sync_threshold",
1765 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1767 .proc_handler = proc_do_sync_threshold,
1770 .procname = "sync_refresh_period",
1771 .maxlen = sizeof(int),
1773 .proc_handler = proc_dointvec_jiffies,
1776 .procname = "sync_retries",
1777 .maxlen = sizeof(int),
1779 .proc_handler = proc_dointvec_minmax,
1784 .procname = "nat_icmp_send",
1785 .maxlen = sizeof(int),
1787 .proc_handler = proc_dointvec,
1790 .procname = "pmtu_disc",
1791 .maxlen = sizeof(int),
1793 .proc_handler = proc_dointvec,
1796 .procname = "backup_only",
1797 .maxlen = sizeof(int),
1799 .proc_handler = proc_dointvec,
1801 #ifdef CONFIG_IP_VS_DEBUG
1803 .procname = "debug_level",
1804 .data = &sysctl_ip_vs_debug_level,
1805 .maxlen = sizeof(int),
1807 .proc_handler = proc_dointvec,
1815 #ifdef CONFIG_PROC_FS
1818 struct seq_net_private p; /* Do not move this, netns depends upon it*/
1819 struct hlist_head *table;
1824 * Write the contents of the VS rule table to a PROCfs file.
1825 * (It is kept just for backward compatibility)
1827 static inline const char *ip_vs_fwd_name(unsigned int flags)
1829 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1830 case IP_VS_CONN_F_LOCALNODE:
1832 case IP_VS_CONN_F_TUNNEL:
1834 case IP_VS_CONN_F_DROUTE:
1842 /* Get the Nth entry in the two lists */
1843 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1845 struct net *net = seq_file_net(seq);
1846 struct ip_vs_iter *iter = seq->private;
1848 struct ip_vs_service *svc;
1850 /* look in hash by protocol */
1851 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1852 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
1853 if (net_eq(svc->net, net) && pos-- == 0) {
1854 iter->table = ip_vs_svc_table;
1861 /* keep looking in fwmark */
1862 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1863 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
1865 if (net_eq(svc->net, net) && pos-- == 0) {
1866 iter->table = ip_vs_svc_fwm_table;
1876 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1880 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1884 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1886 struct hlist_node *e;
1887 struct ip_vs_iter *iter;
1888 struct ip_vs_service *svc;
1891 if (v == SEQ_START_TOKEN)
1892 return ip_vs_info_array(seq,0);
1895 iter = seq->private;
1897 if (iter->table == ip_vs_svc_table) {
1898 /* next service in table hashed by protocol */
1899 e = rcu_dereference(hlist_next_rcu(&svc->s_list));
1901 return hlist_entry(e, struct ip_vs_service, s_list);
1903 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1904 hlist_for_each_entry_rcu(svc,
1905 &ip_vs_svc_table[iter->bucket],
1911 iter->table = ip_vs_svc_fwm_table;
1916 /* next service in hashed by fwmark */
1917 e = rcu_dereference(hlist_next_rcu(&svc->f_list));
1919 return hlist_entry(e, struct ip_vs_service, f_list);
1922 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1923 hlist_for_each_entry_rcu(svc,
1924 &ip_vs_svc_fwm_table[iter->bucket],
1932 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
1939 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
1941 if (v == SEQ_START_TOKEN) {
1943 "IP Virtual Server version %d.%d.%d (size=%d)\n",
1944 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
1946 "Prot LocalAddress:Port Scheduler Flags\n");
1948 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
1950 const struct ip_vs_service *svc = v;
1951 const struct ip_vs_iter *iter = seq->private;
1952 const struct ip_vs_dest *dest;
1953 struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler);
1955 if (iter->table == ip_vs_svc_table) {
1956 #ifdef CONFIG_IP_VS_IPV6
1957 if (svc->af == AF_INET6)
1958 seq_printf(seq, "%s [%pI6]:%04X %s ",
1959 ip_vs_proto_name(svc->protocol),
1965 seq_printf(seq, "%s %08X:%04X %s %s ",
1966 ip_vs_proto_name(svc->protocol),
1967 ntohl(svc->addr.ip),
1970 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
1972 seq_printf(seq, "FWM %08X %s %s",
1973 svc->fwmark, sched->name,
1974 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
1977 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
1978 seq_printf(seq, "persistent %d %08X\n",
1980 ntohl(svc->netmask));
1982 seq_putc(seq, '\n');
1984 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
1985 #ifdef CONFIG_IP_VS_IPV6
1986 if (dest->af == AF_INET6)
1989 " %-7s %-6d %-10d %-10d\n",
1992 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
1993 atomic_read(&dest->weight),
1994 atomic_read(&dest->activeconns),
1995 atomic_read(&dest->inactconns));
2000 "%-7s %-6d %-10d %-10d\n",
2001 ntohl(dest->addr.ip),
2003 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2004 atomic_read(&dest->weight),
2005 atomic_read(&dest->activeconns),
2006 atomic_read(&dest->inactconns));
2013 static const struct seq_operations ip_vs_info_seq_ops = {
2014 .start = ip_vs_info_seq_start,
2015 .next = ip_vs_info_seq_next,
2016 .stop = ip_vs_info_seq_stop,
2017 .show = ip_vs_info_seq_show,
2020 static int ip_vs_info_open(struct inode *inode, struct file *file)
2022 return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2023 sizeof(struct ip_vs_iter));
2026 static const struct file_operations ip_vs_info_fops = {
2027 .owner = THIS_MODULE,
2028 .open = ip_vs_info_open,
2030 .llseek = seq_lseek,
2031 .release = seq_release_net,
2034 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2036 struct net *net = seq_file_single_net(seq);
2037 struct ip_vs_stats_user show;
2039 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2041 " Total Incoming Outgoing Incoming Outgoing\n");
2043 " Conns Packets Packets Bytes Bytes\n");
2045 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2046 seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", show.conns,
2047 show.inpkts, show.outpkts,
2048 (unsigned long long) show.inbytes,
2049 (unsigned long long) show.outbytes);
2051 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2053 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2054 seq_printf(seq, "%8X %8X %8X %16X %16X\n",
2055 show.cps, show.inpps, show.outpps,
2056 show.inbps, show.outbps);
2061 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2063 return single_open_net(inode, file, ip_vs_stats_show);
2066 static const struct file_operations ip_vs_stats_fops = {
2067 .owner = THIS_MODULE,
2068 .open = ip_vs_stats_seq_open,
2070 .llseek = seq_lseek,
2071 .release = single_release_net,
2074 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2076 struct net *net = seq_file_single_net(seq);
2077 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2078 struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats;
2079 struct ip_vs_stats_user rates;
2082 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2084 " Total Incoming Outgoing Incoming Outgoing\n");
2086 "CPU Conns Packets Packets Bytes Bytes\n");
2088 for_each_possible_cpu(i) {
2089 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2091 __u64 inbytes, outbytes;
2094 start = u64_stats_fetch_begin_irq(&u->syncp);
2095 inbytes = u->ustats.inbytes;
2096 outbytes = u->ustats.outbytes;
2097 } while (u64_stats_fetch_retry_irq(&u->syncp, start));
2099 seq_printf(seq, "%3X %8X %8X %8X %16LX %16LX\n",
2100 i, u->ustats.conns, u->ustats.inpkts,
2101 u->ustats.outpkts, (__u64)inbytes,
2105 spin_lock_bh(&tot_stats->lock);
2107 seq_printf(seq, " ~ %8X %8X %8X %16LX %16LX\n\n",
2108 tot_stats->ustats.conns, tot_stats->ustats.inpkts,
2109 tot_stats->ustats.outpkts,
2110 (unsigned long long) tot_stats->ustats.inbytes,
2111 (unsigned long long) tot_stats->ustats.outbytes);
2113 ip_vs_read_estimator(&rates, tot_stats);
2115 spin_unlock_bh(&tot_stats->lock);
2117 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2119 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2120 seq_printf(seq, " %8X %8X %8X %16X %16X\n",
2130 static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2132 return single_open_net(inode, file, ip_vs_stats_percpu_show);
2135 static const struct file_operations ip_vs_stats_percpu_fops = {
2136 .owner = THIS_MODULE,
2137 .open = ip_vs_stats_percpu_seq_open,
2139 .llseek = seq_lseek,
2140 .release = single_release_net,
2145 * Set timeout values for tcp tcpfin udp in the timeout_table.
2147 static int ip_vs_set_timeout(struct net *net, struct ip_vs_timeout_user *u)
2149 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2150 struct ip_vs_proto_data *pd;
2153 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2158 #ifdef CONFIG_IP_VS_PROTO_TCP
2159 if (u->tcp_timeout) {
2160 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2161 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2162 = u->tcp_timeout * HZ;
2165 if (u->tcp_fin_timeout) {
2166 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2167 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2168 = u->tcp_fin_timeout * HZ;
2172 #ifdef CONFIG_IP_VS_PROTO_UDP
2173 if (u->udp_timeout) {
2174 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2175 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2176 = u->udp_timeout * HZ;
2183 #define SET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2184 #define SERVICE_ARG_LEN (sizeof(struct ip_vs_service_user))
2185 #define SVCDEST_ARG_LEN (sizeof(struct ip_vs_service_user) + \
2186 sizeof(struct ip_vs_dest_user))
2187 #define TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2188 #define DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user))
2189 #define MAX_ARG_LEN SVCDEST_ARG_LEN
2191 static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
2192 [SET_CMDID(IP_VS_SO_SET_ADD)] = SERVICE_ARG_LEN,
2193 [SET_CMDID(IP_VS_SO_SET_EDIT)] = SERVICE_ARG_LEN,
2194 [SET_CMDID(IP_VS_SO_SET_DEL)] = SERVICE_ARG_LEN,
2195 [SET_CMDID(IP_VS_SO_SET_FLUSH)] = 0,
2196 [SET_CMDID(IP_VS_SO_SET_ADDDEST)] = SVCDEST_ARG_LEN,
2197 [SET_CMDID(IP_VS_SO_SET_DELDEST)] = SVCDEST_ARG_LEN,
2198 [SET_CMDID(IP_VS_SO_SET_EDITDEST)] = SVCDEST_ARG_LEN,
2199 [SET_CMDID(IP_VS_SO_SET_TIMEOUT)] = TIMEOUT_ARG_LEN,
2200 [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)] = DAEMON_ARG_LEN,
2201 [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)] = DAEMON_ARG_LEN,
2202 [SET_CMDID(IP_VS_SO_SET_ZERO)] = SERVICE_ARG_LEN,
2205 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2206 struct ip_vs_service_user *usvc_compat)
2208 memset(usvc, 0, sizeof(*usvc));
2211 usvc->protocol = usvc_compat->protocol;
2212 usvc->addr.ip = usvc_compat->addr;
2213 usvc->port = usvc_compat->port;
2214 usvc->fwmark = usvc_compat->fwmark;
2216 /* Deep copy of sched_name is not needed here */
2217 usvc->sched_name = usvc_compat->sched_name;
2219 usvc->flags = usvc_compat->flags;
2220 usvc->timeout = usvc_compat->timeout;
2221 usvc->netmask = usvc_compat->netmask;
2224 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2225 struct ip_vs_dest_user *udest_compat)
2227 memset(udest, 0, sizeof(*udest));
2229 udest->addr.ip = udest_compat->addr;
2230 udest->port = udest_compat->port;
2231 udest->conn_flags = udest_compat->conn_flags;
2232 udest->weight = udest_compat->weight;
2233 udest->u_threshold = udest_compat->u_threshold;
2234 udest->l_threshold = udest_compat->l_threshold;
2238 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2240 struct net *net = sock_net(sk);
2242 unsigned char arg[MAX_ARG_LEN];
2243 struct ip_vs_service_user *usvc_compat;
2244 struct ip_vs_service_user_kern usvc;
2245 struct ip_vs_service *svc;
2246 struct ip_vs_dest_user *udest_compat;
2247 struct ip_vs_dest_user_kern udest;
2248 struct netns_ipvs *ipvs = net_ipvs(net);
2250 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2253 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2255 if (len < 0 || len > MAX_ARG_LEN)
2257 if (len != set_arglen[SET_CMDID(cmd)]) {
2258 pr_err("set_ctl: len %u != %u\n",
2259 len, set_arglen[SET_CMDID(cmd)]);
2263 if (copy_from_user(arg, user, len) != 0)
2266 /* increase the module use count */
2267 ip_vs_use_count_inc();
2269 /* Handle daemons since they have another lock */
2270 if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2271 cmd == IP_VS_SO_SET_STOPDAEMON) {
2272 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2274 mutex_lock(&ipvs->sync_mutex);
2275 if (cmd == IP_VS_SO_SET_STARTDAEMON)
2276 ret = start_sync_thread(net, dm->state, dm->mcast_ifn,
2279 ret = stop_sync_thread(net, dm->state);
2280 mutex_unlock(&ipvs->sync_mutex);
2284 mutex_lock(&__ip_vs_mutex);
2285 if (cmd == IP_VS_SO_SET_FLUSH) {
2286 /* Flush the virtual service */
2287 ret = ip_vs_flush(net, false);
2289 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2290 /* Set timeout values for (tcp tcpfin udp) */
2291 ret = ip_vs_set_timeout(net, (struct ip_vs_timeout_user *)arg);
2295 usvc_compat = (struct ip_vs_service_user *)arg;
2296 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2298 /* We only use the new structs internally, so copy userspace compat
2299 * structs to extended internal versions */
2300 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2301 ip_vs_copy_udest_compat(&udest, udest_compat);
2303 if (cmd == IP_VS_SO_SET_ZERO) {
2304 /* if no service address is set, zero counters in all */
2305 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2306 ret = ip_vs_zero_all(net);
2311 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2312 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2313 usvc.protocol != IPPROTO_SCTP) {
2314 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2315 usvc.protocol, &usvc.addr.ip,
2316 ntohs(usvc.port), usvc.sched_name);
2321 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2323 if (usvc.fwmark == 0)
2324 svc = __ip_vs_service_find(net, usvc.af, usvc.protocol,
2325 &usvc.addr, usvc.port);
2327 svc = __ip_vs_svc_fwm_find(net, usvc.af, usvc.fwmark);
2330 if (cmd != IP_VS_SO_SET_ADD
2331 && (svc == NULL || svc->protocol != usvc.protocol)) {
2337 case IP_VS_SO_SET_ADD:
2341 ret = ip_vs_add_service(net, &usvc, &svc);
2343 case IP_VS_SO_SET_EDIT:
2344 ret = ip_vs_edit_service(svc, &usvc);
2346 case IP_VS_SO_SET_DEL:
2347 ret = ip_vs_del_service(svc);
2351 case IP_VS_SO_SET_ZERO:
2352 ret = ip_vs_zero_service(svc);
2354 case IP_VS_SO_SET_ADDDEST:
2355 ret = ip_vs_add_dest(svc, &udest);
2357 case IP_VS_SO_SET_EDITDEST:
2358 ret = ip_vs_edit_dest(svc, &udest);
2360 case IP_VS_SO_SET_DELDEST:
2361 ret = ip_vs_del_dest(svc, &udest);
2368 mutex_unlock(&__ip_vs_mutex);
2370 /* decrease the module use count */
2371 ip_vs_use_count_dec();
2378 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2380 struct ip_vs_scheduler *sched;
2382 sched = rcu_dereference_protected(src->scheduler, 1);
2383 dst->protocol = src->protocol;
2384 dst->addr = src->addr.ip;
2385 dst->port = src->port;
2386 dst->fwmark = src->fwmark;
2387 strlcpy(dst->sched_name, sched->name, sizeof(dst->sched_name));
2388 dst->flags = src->flags;
2389 dst->timeout = src->timeout / HZ;
2390 dst->netmask = src->netmask;
2391 dst->num_dests = src->num_dests;
2392 ip_vs_copy_stats(&dst->stats, &src->stats);
2396 __ip_vs_get_service_entries(struct net *net,
2397 const struct ip_vs_get_services *get,
2398 struct ip_vs_get_services __user *uptr)
2401 struct ip_vs_service *svc;
2402 struct ip_vs_service_entry entry;
2405 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2406 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2407 /* Only expose IPv4 entries to old interface */
2408 if (svc->af != AF_INET || !net_eq(svc->net, net))
2411 if (count >= get->num_services)
2413 memset(&entry, 0, sizeof(entry));
2414 ip_vs_copy_service(&entry, svc);
2415 if (copy_to_user(&uptr->entrytable[count],
2416 &entry, sizeof(entry))) {
2424 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2425 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2426 /* Only expose IPv4 entries to old interface */
2427 if (svc->af != AF_INET || !net_eq(svc->net, net))
2430 if (count >= get->num_services)
2432 memset(&entry, 0, sizeof(entry));
2433 ip_vs_copy_service(&entry, svc);
2434 if (copy_to_user(&uptr->entrytable[count],
2435 &entry, sizeof(entry))) {
2447 __ip_vs_get_dest_entries(struct net *net, const struct ip_vs_get_dests *get,
2448 struct ip_vs_get_dests __user *uptr)
2450 struct ip_vs_service *svc;
2451 union nf_inet_addr addr = { .ip = get->addr };
2456 svc = __ip_vs_svc_fwm_find(net, AF_INET, get->fwmark);
2458 svc = __ip_vs_service_find(net, AF_INET, get->protocol, &addr,
2464 struct ip_vs_dest *dest;
2465 struct ip_vs_dest_entry entry;
2467 memset(&entry, 0, sizeof(entry));
2468 list_for_each_entry(dest, &svc->destinations, n_list) {
2469 if (count >= get->num_dests)
2472 entry.addr = dest->addr.ip;
2473 entry.port = dest->port;
2474 entry.conn_flags = atomic_read(&dest->conn_flags);
2475 entry.weight = atomic_read(&dest->weight);
2476 entry.u_threshold = dest->u_threshold;
2477 entry.l_threshold = dest->l_threshold;
2478 entry.activeconns = atomic_read(&dest->activeconns);
2479 entry.inactconns = atomic_read(&dest->inactconns);
2480 entry.persistconns = atomic_read(&dest->persistconns);
2481 ip_vs_copy_stats(&entry.stats, &dest->stats);
2482 if (copy_to_user(&uptr->entrytable[count],
2483 &entry, sizeof(entry))) {
2495 __ip_vs_get_timeouts(struct net *net, struct ip_vs_timeout_user *u)
2497 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2498 struct ip_vs_proto_data *pd;
2501 memset(u, 0, sizeof (*u));
2503 #ifdef CONFIG_IP_VS_PROTO_TCP
2504 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2505 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2506 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2508 #ifdef CONFIG_IP_VS_PROTO_UDP
2509 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2511 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2516 #define GET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2517 #define GET_INFO_ARG_LEN (sizeof(struct ip_vs_getinfo))
2518 #define GET_SERVICES_ARG_LEN (sizeof(struct ip_vs_get_services))
2519 #define GET_SERVICE_ARG_LEN (sizeof(struct ip_vs_service_entry))
2520 #define GET_DESTS_ARG_LEN (sizeof(struct ip_vs_get_dests))
2521 #define GET_TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2522 #define GET_DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user) * 2)
2524 static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
2525 [GET_CMDID(IP_VS_SO_GET_VERSION)] = 64,
2526 [GET_CMDID(IP_VS_SO_GET_INFO)] = GET_INFO_ARG_LEN,
2527 [GET_CMDID(IP_VS_SO_GET_SERVICES)] = GET_SERVICES_ARG_LEN,
2528 [GET_CMDID(IP_VS_SO_GET_SERVICE)] = GET_SERVICE_ARG_LEN,
2529 [GET_CMDID(IP_VS_SO_GET_DESTS)] = GET_DESTS_ARG_LEN,
2530 [GET_CMDID(IP_VS_SO_GET_TIMEOUT)] = GET_TIMEOUT_ARG_LEN,
2531 [GET_CMDID(IP_VS_SO_GET_DAEMON)] = GET_DAEMON_ARG_LEN,
2535 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2537 unsigned char arg[128];
2539 unsigned int copylen;
2540 struct net *net = sock_net(sk);
2541 struct netns_ipvs *ipvs = net_ipvs(net);
2544 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2547 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2550 if (*len < get_arglen[GET_CMDID(cmd)]) {
2551 pr_err("get_ctl: len %u < %u\n",
2552 *len, get_arglen[GET_CMDID(cmd)]);
2556 copylen = get_arglen[GET_CMDID(cmd)];
2560 if (copy_from_user(arg, user, copylen) != 0)
2563 * Handle daemons first since it has its own locking
2565 if (cmd == IP_VS_SO_GET_DAEMON) {
2566 struct ip_vs_daemon_user d[2];
2568 memset(&d, 0, sizeof(d));
2569 mutex_lock(&ipvs->sync_mutex);
2570 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2571 d[0].state = IP_VS_STATE_MASTER;
2572 strlcpy(d[0].mcast_ifn, ipvs->master_mcast_ifn,
2573 sizeof(d[0].mcast_ifn));
2574 d[0].syncid = ipvs->master_syncid;
2576 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2577 d[1].state = IP_VS_STATE_BACKUP;
2578 strlcpy(d[1].mcast_ifn, ipvs->backup_mcast_ifn,
2579 sizeof(d[1].mcast_ifn));
2580 d[1].syncid = ipvs->backup_syncid;
2582 if (copy_to_user(user, &d, sizeof(d)) != 0)
2584 mutex_unlock(&ipvs->sync_mutex);
2588 mutex_lock(&__ip_vs_mutex);
2590 case IP_VS_SO_GET_VERSION:
2594 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2595 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2596 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2600 *len = strlen(buf)+1;
2604 case IP_VS_SO_GET_INFO:
2606 struct ip_vs_getinfo info;
2607 info.version = IP_VS_VERSION_CODE;
2608 info.size = ip_vs_conn_tab_size;
2609 info.num_services = ipvs->num_services;
2610 if (copy_to_user(user, &info, sizeof(info)) != 0)
2615 case IP_VS_SO_GET_SERVICES:
2617 struct ip_vs_get_services *get;
2620 get = (struct ip_vs_get_services *)arg;
2621 size = sizeof(*get) +
2622 sizeof(struct ip_vs_service_entry) * get->num_services;
2624 pr_err("length: %u != %u\n", *len, size);
2628 ret = __ip_vs_get_service_entries(net, get, user);
2632 case IP_VS_SO_GET_SERVICE:
2634 struct ip_vs_service_entry *entry;
2635 struct ip_vs_service *svc;
2636 union nf_inet_addr addr;
2638 entry = (struct ip_vs_service_entry *)arg;
2639 addr.ip = entry->addr;
2642 svc = __ip_vs_svc_fwm_find(net, AF_INET, entry->fwmark);
2644 svc = __ip_vs_service_find(net, AF_INET,
2645 entry->protocol, &addr,
2649 ip_vs_copy_service(entry, svc);
2650 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2657 case IP_VS_SO_GET_DESTS:
2659 struct ip_vs_get_dests *get;
2662 get = (struct ip_vs_get_dests *)arg;
2663 size = sizeof(*get) +
2664 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2666 pr_err("length: %u != %u\n", *len, size);
2670 ret = __ip_vs_get_dest_entries(net, get, user);
2674 case IP_VS_SO_GET_TIMEOUT:
2676 struct ip_vs_timeout_user t;
2678 __ip_vs_get_timeouts(net, &t);
2679 if (copy_to_user(user, &t, sizeof(t)) != 0)
2689 mutex_unlock(&__ip_vs_mutex);
2694 static struct nf_sockopt_ops ip_vs_sockopts = {
2696 .set_optmin = IP_VS_BASE_CTL,
2697 .set_optmax = IP_VS_SO_SET_MAX+1,
2698 .set = do_ip_vs_set_ctl,
2699 .get_optmin = IP_VS_BASE_CTL,
2700 .get_optmax = IP_VS_SO_GET_MAX+1,
2701 .get = do_ip_vs_get_ctl,
2702 .owner = THIS_MODULE,
2706 * Generic Netlink interface
2709 /* IPVS genetlink family */
2710 static struct genl_family ip_vs_genl_family = {
2711 .id = GENL_ID_GENERATE,
2713 .name = IPVS_GENL_NAME,
2714 .version = IPVS_GENL_VERSION,
2715 .maxattr = IPVS_CMD_MAX,
2716 .netnsok = true, /* Make ipvsadm to work on netns */
2719 /* Policy used for first-level command attributes */
2720 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2721 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2722 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2723 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2724 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2725 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2726 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2729 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2730 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2731 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2732 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2733 .len = IP_VS_IFNAME_MAXLEN },
2734 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2737 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2738 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2739 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2740 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2741 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2742 .len = sizeof(union nf_inet_addr) },
2743 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2744 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2745 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2746 .len = IP_VS_SCHEDNAME_MAXLEN },
2747 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2748 .len = IP_VS_PENAME_MAXLEN },
2749 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2750 .len = sizeof(struct ip_vs_flags) },
2751 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2752 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2753 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2756 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2757 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2758 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2759 .len = sizeof(union nf_inet_addr) },
2760 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2761 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2762 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2763 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2764 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2765 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2766 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2767 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2768 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2771 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2772 struct ip_vs_stats *stats)
2774 struct ip_vs_stats_user ustats;
2775 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2779 ip_vs_copy_stats(&ustats, stats);
2781 if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, ustats.conns) ||
2782 nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, ustats.inpkts) ||
2783 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, ustats.outpkts) ||
2784 nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, ustats.inbytes) ||
2785 nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, ustats.outbytes) ||
2786 nla_put_u32(skb, IPVS_STATS_ATTR_CPS, ustats.cps) ||
2787 nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, ustats.inpps) ||
2788 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, ustats.outpps) ||
2789 nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, ustats.inbps) ||
2790 nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, ustats.outbps))
2791 goto nla_put_failure;
2792 nla_nest_end(skb, nl_stats);
2797 nla_nest_cancel(skb, nl_stats);
2801 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2802 struct ip_vs_service *svc)
2804 struct ip_vs_scheduler *sched;
2805 struct ip_vs_pe *pe;
2806 struct nlattr *nl_service;
2807 struct ip_vs_flags flags = { .flags = svc->flags,
2810 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2814 if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2815 goto nla_put_failure;
2817 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2818 goto nla_put_failure;
2820 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2821 nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2822 nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2823 goto nla_put_failure;
2826 sched = rcu_dereference_protected(svc->scheduler, 1);
2827 pe = rcu_dereference_protected(svc->pe, 1);
2828 if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched->name) ||
2829 (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
2830 nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2831 nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2832 nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
2833 goto nla_put_failure;
2834 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2835 goto nla_put_failure;
2837 nla_nest_end(skb, nl_service);
2842 nla_nest_cancel(skb, nl_service);
2846 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2847 struct ip_vs_service *svc,
2848 struct netlink_callback *cb)
2852 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2853 &ip_vs_genl_family, NLM_F_MULTI,
2854 IPVS_CMD_NEW_SERVICE);
2858 if (ip_vs_genl_fill_service(skb, svc) < 0)
2859 goto nla_put_failure;
2861 return genlmsg_end(skb, hdr);
2864 genlmsg_cancel(skb, hdr);
2868 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2869 struct netlink_callback *cb)
2872 int start = cb->args[0];
2873 struct ip_vs_service *svc;
2874 struct net *net = skb_sknet(skb);
2876 mutex_lock(&__ip_vs_mutex);
2877 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2878 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2879 if (++idx <= start || !net_eq(svc->net, net))
2881 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2883 goto nla_put_failure;
2888 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2889 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
2890 if (++idx <= start || !net_eq(svc->net, net))
2892 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2894 goto nla_put_failure;
2900 mutex_unlock(&__ip_vs_mutex);
2906 static int ip_vs_genl_parse_service(struct net *net,
2907 struct ip_vs_service_user_kern *usvc,
2908 struct nlattr *nla, int full_entry,
2909 struct ip_vs_service **ret_svc)
2911 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
2912 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
2913 struct ip_vs_service *svc;
2915 /* Parse mandatory identifying service fields first */
2917 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
2920 nla_af = attrs[IPVS_SVC_ATTR_AF];
2921 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
2922 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
2923 nla_port = attrs[IPVS_SVC_ATTR_PORT];
2924 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
2926 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
2929 memset(usvc, 0, sizeof(*usvc));
2931 usvc->af = nla_get_u16(nla_af);
2932 #ifdef CONFIG_IP_VS_IPV6
2933 if (usvc->af != AF_INET && usvc->af != AF_INET6)
2935 if (usvc->af != AF_INET)
2937 return -EAFNOSUPPORT;
2940 usvc->protocol = IPPROTO_TCP;
2941 usvc->fwmark = nla_get_u32(nla_fwmark);
2943 usvc->protocol = nla_get_u16(nla_protocol);
2944 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
2945 usvc->port = nla_get_be16(nla_port);
2951 svc = __ip_vs_svc_fwm_find(net, usvc->af, usvc->fwmark);
2953 svc = __ip_vs_service_find(net, usvc->af, usvc->protocol,
2954 &usvc->addr, usvc->port);
2958 /* If a full entry was requested, check for the additional fields */
2960 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
2962 struct ip_vs_flags flags;
2964 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
2965 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
2966 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
2967 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
2968 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
2970 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
2973 nla_memcpy(&flags, nla_flags, sizeof(flags));
2975 /* prefill flags from service if it already exists */
2977 usvc->flags = svc->flags;
2979 /* set new flags from userland */
2980 usvc->flags = (usvc->flags & ~flags.mask) |
2981 (flags.flags & flags.mask);
2982 usvc->sched_name = nla_data(nla_sched);
2983 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
2984 usvc->timeout = nla_get_u32(nla_timeout);
2985 usvc->netmask = nla_get_be32(nla_netmask);
2991 static struct ip_vs_service *ip_vs_genl_find_service(struct net *net,
2994 struct ip_vs_service_user_kern usvc;
2995 struct ip_vs_service *svc;
2998 ret = ip_vs_genl_parse_service(net, &usvc, nla, 0, &svc);
2999 return ret ? ERR_PTR(ret) : svc;
3002 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3004 struct nlattr *nl_dest;
3006 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3010 if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3011 nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3012 nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3013 (atomic_read(&dest->conn_flags) &
3014 IP_VS_CONN_F_FWD_MASK)) ||
3015 nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3016 atomic_read(&dest->weight)) ||
3017 nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3018 nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3019 nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3020 atomic_read(&dest->activeconns)) ||
3021 nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3022 atomic_read(&dest->inactconns)) ||
3023 nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3024 atomic_read(&dest->persistconns)))
3025 goto nla_put_failure;
3026 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
3027 goto nla_put_failure;
3029 nla_nest_end(skb, nl_dest);
3034 nla_nest_cancel(skb, nl_dest);
3038 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3039 struct netlink_callback *cb)
3043 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3044 &ip_vs_genl_family, NLM_F_MULTI,
3049 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3050 goto nla_put_failure;
3052 return genlmsg_end(skb, hdr);
3055 genlmsg_cancel(skb, hdr);
3059 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3060 struct netlink_callback *cb)
3063 int start = cb->args[0];
3064 struct ip_vs_service *svc;
3065 struct ip_vs_dest *dest;
3066 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3067 struct net *net = skb_sknet(skb);
3069 mutex_lock(&__ip_vs_mutex);
3071 /* Try to find the service for which to dump destinations */
3072 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3073 IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3077 svc = ip_vs_genl_find_service(net, attrs[IPVS_CMD_ATTR_SERVICE]);
3078 if (IS_ERR(svc) || svc == NULL)
3081 /* Dump the destinations */
3082 list_for_each_entry(dest, &svc->destinations, n_list) {
3085 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3087 goto nla_put_failure;
3095 mutex_unlock(&__ip_vs_mutex);
3100 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3101 struct nlattr *nla, int full_entry)
3103 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3104 struct nlattr *nla_addr, *nla_port;
3106 /* Parse mandatory identifying destination fields first */
3108 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3111 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3112 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3114 if (!(nla_addr && nla_port))
3117 memset(udest, 0, sizeof(*udest));
3119 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3120 udest->port = nla_get_be16(nla_port);
3122 /* If a full entry was requested, check for the additional fields */
3124 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3127 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3128 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3129 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3130 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3132 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3135 udest->conn_flags = nla_get_u32(nla_fwd)
3136 & IP_VS_CONN_F_FWD_MASK;
3137 udest->weight = nla_get_u32(nla_weight);
3138 udest->u_threshold = nla_get_u32(nla_u_thresh);
3139 udest->l_threshold = nla_get_u32(nla_l_thresh);
3145 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3146 const char *mcast_ifn, __u32 syncid)
3148 struct nlattr *nl_daemon;
3150 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3154 if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3155 nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn) ||
3156 nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid))
3157 goto nla_put_failure;
3158 nla_nest_end(skb, nl_daemon);
3163 nla_nest_cancel(skb, nl_daemon);
3167 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3168 const char *mcast_ifn, __u32 syncid,
3169 struct netlink_callback *cb)
3172 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3173 &ip_vs_genl_family, NLM_F_MULTI,
3174 IPVS_CMD_NEW_DAEMON);
3178 if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
3179 goto nla_put_failure;
3181 return genlmsg_end(skb, hdr);
3184 genlmsg_cancel(skb, hdr);
3188 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3189 struct netlink_callback *cb)
3191 struct net *net = skb_sknet(skb);
3192 struct netns_ipvs *ipvs = net_ipvs(net);
3194 mutex_lock(&ipvs->sync_mutex);
3195 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3196 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3197 ipvs->master_mcast_ifn,
3198 ipvs->master_syncid, cb) < 0)
3199 goto nla_put_failure;
3204 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3205 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3206 ipvs->backup_mcast_ifn,
3207 ipvs->backup_syncid, cb) < 0)
3208 goto nla_put_failure;
3214 mutex_unlock(&ipvs->sync_mutex);
3219 static int ip_vs_genl_new_daemon(struct net *net, struct nlattr **attrs)
3221 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3222 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3223 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3226 return start_sync_thread(net,
3227 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3228 nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3229 nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3232 static int ip_vs_genl_del_daemon(struct net *net, struct nlattr **attrs)
3234 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3237 return stop_sync_thread(net,
3238 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3241 static int ip_vs_genl_set_config(struct net *net, struct nlattr **attrs)
3243 struct ip_vs_timeout_user t;
3245 __ip_vs_get_timeouts(net, &t);
3247 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3248 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3250 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3252 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3254 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3255 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3257 return ip_vs_set_timeout(net, &t);
3260 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3264 struct netns_ipvs *ipvs;
3266 net = skb_sknet(skb);
3267 ipvs = net_ipvs(net);
3268 cmd = info->genlhdr->cmd;
3270 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3271 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3273 mutex_lock(&ipvs->sync_mutex);
3274 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3275 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3276 info->attrs[IPVS_CMD_ATTR_DAEMON],
3277 ip_vs_daemon_policy)) {
3282 if (cmd == IPVS_CMD_NEW_DAEMON)
3283 ret = ip_vs_genl_new_daemon(net, daemon_attrs);
3285 ret = ip_vs_genl_del_daemon(net, daemon_attrs);
3287 mutex_unlock(&ipvs->sync_mutex);
3292 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3294 struct ip_vs_service *svc = NULL;
3295 struct ip_vs_service_user_kern usvc;
3296 struct ip_vs_dest_user_kern udest;
3298 int need_full_svc = 0, need_full_dest = 0;
3301 net = skb_sknet(skb);
3302 cmd = info->genlhdr->cmd;
3304 mutex_lock(&__ip_vs_mutex);
3306 if (cmd == IPVS_CMD_FLUSH) {
3307 ret = ip_vs_flush(net, false);
3309 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3310 ret = ip_vs_genl_set_config(net, info->attrs);
3312 } else if (cmd == IPVS_CMD_ZERO &&
3313 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3314 ret = ip_vs_zero_all(net);
3318 /* All following commands require a service argument, so check if we
3319 * received a valid one. We need a full service specification when
3320 * adding / editing a service. Only identifying members otherwise. */
3321 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3324 ret = ip_vs_genl_parse_service(net, &usvc,
3325 info->attrs[IPVS_CMD_ATTR_SERVICE],
3326 need_full_svc, &svc);
3330 /* Unless we're adding a new service, the service must already exist */
3331 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3336 /* Destination commands require a valid destination argument. For
3337 * adding / editing a destination, we need a full destination
3339 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3340 cmd == IPVS_CMD_DEL_DEST) {
3341 if (cmd != IPVS_CMD_DEL_DEST)
3344 ret = ip_vs_genl_parse_dest(&udest,
3345 info->attrs[IPVS_CMD_ATTR_DEST],
3352 case IPVS_CMD_NEW_SERVICE:
3354 ret = ip_vs_add_service(net, &usvc, &svc);
3358 case IPVS_CMD_SET_SERVICE:
3359 ret = ip_vs_edit_service(svc, &usvc);
3361 case IPVS_CMD_DEL_SERVICE:
3362 ret = ip_vs_del_service(svc);
3363 /* do not use svc, it can be freed */
3365 case IPVS_CMD_NEW_DEST:
3366 ret = ip_vs_add_dest(svc, &udest);
3368 case IPVS_CMD_SET_DEST:
3369 ret = ip_vs_edit_dest(svc, &udest);
3371 case IPVS_CMD_DEL_DEST:
3372 ret = ip_vs_del_dest(svc, &udest);
3375 ret = ip_vs_zero_service(svc);
3382 mutex_unlock(&__ip_vs_mutex);
3387 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3389 struct sk_buff *msg;
3391 int ret, cmd, reply_cmd;
3394 net = skb_sknet(skb);
3395 cmd = info->genlhdr->cmd;
3397 if (cmd == IPVS_CMD_GET_SERVICE)
3398 reply_cmd = IPVS_CMD_NEW_SERVICE;
3399 else if (cmd == IPVS_CMD_GET_INFO)
3400 reply_cmd = IPVS_CMD_SET_INFO;
3401 else if (cmd == IPVS_CMD_GET_CONFIG)
3402 reply_cmd = IPVS_CMD_SET_CONFIG;
3404 pr_err("unknown Generic Netlink command\n");
3408 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3412 mutex_lock(&__ip_vs_mutex);
3414 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3416 goto nla_put_failure;
3419 case IPVS_CMD_GET_SERVICE:
3421 struct ip_vs_service *svc;
3423 svc = ip_vs_genl_find_service(net,
3424 info->attrs[IPVS_CMD_ATTR_SERVICE]);
3429 ret = ip_vs_genl_fill_service(msg, svc);
3431 goto nla_put_failure;
3440 case IPVS_CMD_GET_CONFIG:
3442 struct ip_vs_timeout_user t;
3444 __ip_vs_get_timeouts(net, &t);
3445 #ifdef CONFIG_IP_VS_PROTO_TCP
3446 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3448 nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3450 goto nla_put_failure;
3452 #ifdef CONFIG_IP_VS_PROTO_UDP
3453 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3454 goto nla_put_failure;
3460 case IPVS_CMD_GET_INFO:
3461 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3462 IP_VS_VERSION_CODE) ||
3463 nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3464 ip_vs_conn_tab_size))
3465 goto nla_put_failure;
3469 genlmsg_end(msg, reply);
3470 ret = genlmsg_reply(msg, info);
3474 pr_err("not enough space in Netlink message\n");
3480 mutex_unlock(&__ip_vs_mutex);
3486 static const struct genl_ops ip_vs_genl_ops[] = {
3488 .cmd = IPVS_CMD_NEW_SERVICE,
3489 .flags = GENL_ADMIN_PERM,
3490 .policy = ip_vs_cmd_policy,
3491 .doit = ip_vs_genl_set_cmd,
3494 .cmd = IPVS_CMD_SET_SERVICE,
3495 .flags = GENL_ADMIN_PERM,
3496 .policy = ip_vs_cmd_policy,
3497 .doit = ip_vs_genl_set_cmd,
3500 .cmd = IPVS_CMD_DEL_SERVICE,
3501 .flags = GENL_ADMIN_PERM,
3502 .policy = ip_vs_cmd_policy,
3503 .doit = ip_vs_genl_set_cmd,
3506 .cmd = IPVS_CMD_GET_SERVICE,
3507 .flags = GENL_ADMIN_PERM,
3508 .doit = ip_vs_genl_get_cmd,
3509 .dumpit = ip_vs_genl_dump_services,
3510 .policy = ip_vs_cmd_policy,
3513 .cmd = IPVS_CMD_NEW_DEST,
3514 .flags = GENL_ADMIN_PERM,
3515 .policy = ip_vs_cmd_policy,
3516 .doit = ip_vs_genl_set_cmd,
3519 .cmd = IPVS_CMD_SET_DEST,
3520 .flags = GENL_ADMIN_PERM,
3521 .policy = ip_vs_cmd_policy,
3522 .doit = ip_vs_genl_set_cmd,
3525 .cmd = IPVS_CMD_DEL_DEST,
3526 .flags = GENL_ADMIN_PERM,
3527 .policy = ip_vs_cmd_policy,
3528 .doit = ip_vs_genl_set_cmd,
3531 .cmd = IPVS_CMD_GET_DEST,
3532 .flags = GENL_ADMIN_PERM,
3533 .policy = ip_vs_cmd_policy,
3534 .dumpit = ip_vs_genl_dump_dests,
3537 .cmd = IPVS_CMD_NEW_DAEMON,
3538 .flags = GENL_ADMIN_PERM,
3539 .policy = ip_vs_cmd_policy,
3540 .doit = ip_vs_genl_set_daemon,
3543 .cmd = IPVS_CMD_DEL_DAEMON,
3544 .flags = GENL_ADMIN_PERM,
3545 .policy = ip_vs_cmd_policy,
3546 .doit = ip_vs_genl_set_daemon,
3549 .cmd = IPVS_CMD_GET_DAEMON,
3550 .flags = GENL_ADMIN_PERM,
3551 .dumpit = ip_vs_genl_dump_daemons,
3554 .cmd = IPVS_CMD_SET_CONFIG,
3555 .flags = GENL_ADMIN_PERM,
3556 .policy = ip_vs_cmd_policy,
3557 .doit = ip_vs_genl_set_cmd,
3560 .cmd = IPVS_CMD_GET_CONFIG,
3561 .flags = GENL_ADMIN_PERM,
3562 .doit = ip_vs_genl_get_cmd,
3565 .cmd = IPVS_CMD_GET_INFO,
3566 .flags = GENL_ADMIN_PERM,
3567 .doit = ip_vs_genl_get_cmd,
3570 .cmd = IPVS_CMD_ZERO,
3571 .flags = GENL_ADMIN_PERM,
3572 .policy = ip_vs_cmd_policy,
3573 .doit = ip_vs_genl_set_cmd,
3576 .cmd = IPVS_CMD_FLUSH,
3577 .flags = GENL_ADMIN_PERM,
3578 .doit = ip_vs_genl_set_cmd,
3582 static int __init ip_vs_genl_register(void)
3584 return genl_register_family_with_ops(&ip_vs_genl_family,
3588 static void ip_vs_genl_unregister(void)
3590 genl_unregister_family(&ip_vs_genl_family);
3593 /* End of Generic Netlink interface definitions */
3596 * per netns intit/exit func.
3598 #ifdef CONFIG_SYSCTL
3599 static int __net_init ip_vs_control_net_init_sysctl(struct net *net)
3602 struct netns_ipvs *ipvs = net_ipvs(net);
3603 struct ctl_table *tbl;
3605 atomic_set(&ipvs->dropentry, 0);
3606 spin_lock_init(&ipvs->dropentry_lock);
3607 spin_lock_init(&ipvs->droppacket_lock);
3608 spin_lock_init(&ipvs->securetcp_lock);
3610 if (!net_eq(net, &init_net)) {
3611 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3615 /* Don't export sysctls to unprivileged users */
3616 if (net->user_ns != &init_user_ns)
3617 tbl[0].procname = NULL;
3620 /* Initialize sysctl defaults */
3622 ipvs->sysctl_amemthresh = 1024;
3623 tbl[idx++].data = &ipvs->sysctl_amemthresh;
3624 ipvs->sysctl_am_droprate = 10;
3625 tbl[idx++].data = &ipvs->sysctl_am_droprate;
3626 tbl[idx++].data = &ipvs->sysctl_drop_entry;
3627 tbl[idx++].data = &ipvs->sysctl_drop_packet;
3628 #ifdef CONFIG_IP_VS_NFCT
3629 tbl[idx++].data = &ipvs->sysctl_conntrack;
3631 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3632 ipvs->sysctl_snat_reroute = 1;
3633 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3634 ipvs->sysctl_sync_ver = 1;
3635 tbl[idx++].data = &ipvs->sysctl_sync_ver;
3636 ipvs->sysctl_sync_ports = 1;
3637 tbl[idx++].data = &ipvs->sysctl_sync_ports;
3638 tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
3639 ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3640 tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3641 ipvs->sysctl_sync_sock_size = 0;
3642 tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3643 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3644 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3645 tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
3646 tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
3647 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3648 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3649 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3650 tbl[idx].data = &ipvs->sysctl_sync_threshold;
3651 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3652 ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3653 tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3654 ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3655 tbl[idx++].data = &ipvs->sysctl_sync_retries;
3656 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3657 ipvs->sysctl_pmtu_disc = 1;
3658 tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
3659 tbl[idx++].data = &ipvs->sysctl_backup_only;
3662 ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3663 if (ipvs->sysctl_hdr == NULL) {
3664 if (!net_eq(net, &init_net))
3668 ip_vs_start_estimator(net, &ipvs->tot_stats);
3669 ipvs->sysctl_tbl = tbl;
3670 /* Schedule defense work */
3671 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3672 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3677 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net)
3679 struct netns_ipvs *ipvs = net_ipvs(net);
3681 cancel_delayed_work_sync(&ipvs->defense_work);
3682 cancel_work_sync(&ipvs->defense_work.work);
3683 unregister_net_sysctl_table(ipvs->sysctl_hdr);
3684 ip_vs_stop_estimator(net, &ipvs->tot_stats);
3689 static int __net_init ip_vs_control_net_init_sysctl(struct net *net) { return 0; }
3690 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net) { }
3694 static struct notifier_block ip_vs_dst_notifier = {
3695 .notifier_call = ip_vs_dst_event,
3698 int __net_init ip_vs_control_net_init(struct net *net)
3701 struct netns_ipvs *ipvs = net_ipvs(net);
3703 /* Initialize rs_table */
3704 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
3705 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
3707 INIT_LIST_HEAD(&ipvs->dest_trash);
3708 spin_lock_init(&ipvs->dest_trash_lock);
3709 setup_timer(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire,
3710 (unsigned long) net);
3711 atomic_set(&ipvs->ftpsvc_counter, 0);
3712 atomic_set(&ipvs->nullsvc_counter, 0);
3715 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
3716 if (!ipvs->tot_stats.cpustats)
3719 for_each_possible_cpu(i) {
3720 struct ip_vs_cpu_stats *ipvs_tot_stats;
3721 ipvs_tot_stats = per_cpu_ptr(ipvs->tot_stats.cpustats, i);
3722 u64_stats_init(&ipvs_tot_stats->syncp);
3725 spin_lock_init(&ipvs->tot_stats.lock);
3727 proc_create("ip_vs", 0, net->proc_net, &ip_vs_info_fops);
3728 proc_create("ip_vs_stats", 0, net->proc_net, &ip_vs_stats_fops);
3729 proc_create("ip_vs_stats_percpu", 0, net->proc_net,
3730 &ip_vs_stats_percpu_fops);
3732 if (ip_vs_control_net_init_sysctl(net))
3738 free_percpu(ipvs->tot_stats.cpustats);
3742 void __net_exit ip_vs_control_net_cleanup(struct net *net)
3744 struct netns_ipvs *ipvs = net_ipvs(net);
3746 ip_vs_trash_cleanup(net);
3747 ip_vs_control_net_cleanup_sysctl(net);
3748 remove_proc_entry("ip_vs_stats_percpu", net->proc_net);
3749 remove_proc_entry("ip_vs_stats", net->proc_net);
3750 remove_proc_entry("ip_vs", net->proc_net);
3751 free_percpu(ipvs->tot_stats.cpustats);
3754 int __init ip_vs_register_nl_ioctl(void)
3758 ret = nf_register_sockopt(&ip_vs_sockopts);
3760 pr_err("cannot register sockopt.\n");
3764 ret = ip_vs_genl_register();
3766 pr_err("cannot register Generic Netlink interface.\n");
3772 nf_unregister_sockopt(&ip_vs_sockopts);
3777 void ip_vs_unregister_nl_ioctl(void)
3779 ip_vs_genl_unregister();
3780 nf_unregister_sockopt(&ip_vs_sockopts);
3783 int __init ip_vs_control_init(void)
3790 /* Initialize svc_table, ip_vs_svc_fwm_table */
3791 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
3792 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
3793 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3796 smp_wmb(); /* Do we really need it now ? */
3798 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
3807 void ip_vs_control_cleanup(void)
3810 unregister_netdevice_notifier(&ip_vs_dst_notifier);