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 if (mutex_lock_interruptible(&ipvs->sync_mutex)) {
2278 if (cmd == IP_VS_SO_SET_STARTDAEMON)
2279 ret = start_sync_thread(net, dm->state, dm->mcast_ifn,
2282 ret = stop_sync_thread(net, dm->state);
2283 mutex_unlock(&ipvs->sync_mutex);
2287 if (mutex_lock_interruptible(&__ip_vs_mutex)) {
2292 if (cmd == IP_VS_SO_SET_FLUSH) {
2293 /* Flush the virtual service */
2294 ret = ip_vs_flush(net, false);
2296 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2297 /* Set timeout values for (tcp tcpfin udp) */
2298 ret = ip_vs_set_timeout(net, (struct ip_vs_timeout_user *)arg);
2302 usvc_compat = (struct ip_vs_service_user *)arg;
2303 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2305 /* We only use the new structs internally, so copy userspace compat
2306 * structs to extended internal versions */
2307 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2308 ip_vs_copy_udest_compat(&udest, udest_compat);
2310 if (cmd == IP_VS_SO_SET_ZERO) {
2311 /* if no service address is set, zero counters in all */
2312 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2313 ret = ip_vs_zero_all(net);
2318 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2319 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2320 usvc.protocol != IPPROTO_SCTP) {
2321 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2322 usvc.protocol, &usvc.addr.ip,
2323 ntohs(usvc.port), usvc.sched_name);
2328 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2330 if (usvc.fwmark == 0)
2331 svc = __ip_vs_service_find(net, usvc.af, usvc.protocol,
2332 &usvc.addr, usvc.port);
2334 svc = __ip_vs_svc_fwm_find(net, usvc.af, usvc.fwmark);
2337 if (cmd != IP_VS_SO_SET_ADD
2338 && (svc == NULL || svc->protocol != usvc.protocol)) {
2344 case IP_VS_SO_SET_ADD:
2348 ret = ip_vs_add_service(net, &usvc, &svc);
2350 case IP_VS_SO_SET_EDIT:
2351 ret = ip_vs_edit_service(svc, &usvc);
2353 case IP_VS_SO_SET_DEL:
2354 ret = ip_vs_del_service(svc);
2358 case IP_VS_SO_SET_ZERO:
2359 ret = ip_vs_zero_service(svc);
2361 case IP_VS_SO_SET_ADDDEST:
2362 ret = ip_vs_add_dest(svc, &udest);
2364 case IP_VS_SO_SET_EDITDEST:
2365 ret = ip_vs_edit_dest(svc, &udest);
2367 case IP_VS_SO_SET_DELDEST:
2368 ret = ip_vs_del_dest(svc, &udest);
2375 mutex_unlock(&__ip_vs_mutex);
2377 /* decrease the module use count */
2378 ip_vs_use_count_dec();
2385 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2387 struct ip_vs_scheduler *sched;
2389 sched = rcu_dereference_protected(src->scheduler, 1);
2390 dst->protocol = src->protocol;
2391 dst->addr = src->addr.ip;
2392 dst->port = src->port;
2393 dst->fwmark = src->fwmark;
2394 strlcpy(dst->sched_name, sched->name, sizeof(dst->sched_name));
2395 dst->flags = src->flags;
2396 dst->timeout = src->timeout / HZ;
2397 dst->netmask = src->netmask;
2398 dst->num_dests = src->num_dests;
2399 ip_vs_copy_stats(&dst->stats, &src->stats);
2403 __ip_vs_get_service_entries(struct net *net,
2404 const struct ip_vs_get_services *get,
2405 struct ip_vs_get_services __user *uptr)
2408 struct ip_vs_service *svc;
2409 struct ip_vs_service_entry entry;
2412 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2413 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2414 /* Only expose IPv4 entries to old interface */
2415 if (svc->af != AF_INET || !net_eq(svc->net, net))
2418 if (count >= get->num_services)
2420 memset(&entry, 0, sizeof(entry));
2421 ip_vs_copy_service(&entry, svc);
2422 if (copy_to_user(&uptr->entrytable[count],
2423 &entry, sizeof(entry))) {
2431 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2432 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2433 /* Only expose IPv4 entries to old interface */
2434 if (svc->af != AF_INET || !net_eq(svc->net, net))
2437 if (count >= get->num_services)
2439 memset(&entry, 0, sizeof(entry));
2440 ip_vs_copy_service(&entry, svc);
2441 if (copy_to_user(&uptr->entrytable[count],
2442 &entry, sizeof(entry))) {
2454 __ip_vs_get_dest_entries(struct net *net, const struct ip_vs_get_dests *get,
2455 struct ip_vs_get_dests __user *uptr)
2457 struct ip_vs_service *svc;
2458 union nf_inet_addr addr = { .ip = get->addr };
2463 svc = __ip_vs_svc_fwm_find(net, AF_INET, get->fwmark);
2465 svc = __ip_vs_service_find(net, AF_INET, get->protocol, &addr,
2471 struct ip_vs_dest *dest;
2472 struct ip_vs_dest_entry entry;
2474 memset(&entry, 0, sizeof(entry));
2475 list_for_each_entry(dest, &svc->destinations, n_list) {
2476 if (count >= get->num_dests)
2479 entry.addr = dest->addr.ip;
2480 entry.port = dest->port;
2481 entry.conn_flags = atomic_read(&dest->conn_flags);
2482 entry.weight = atomic_read(&dest->weight);
2483 entry.u_threshold = dest->u_threshold;
2484 entry.l_threshold = dest->l_threshold;
2485 entry.activeconns = atomic_read(&dest->activeconns);
2486 entry.inactconns = atomic_read(&dest->inactconns);
2487 entry.persistconns = atomic_read(&dest->persistconns);
2488 ip_vs_copy_stats(&entry.stats, &dest->stats);
2489 if (copy_to_user(&uptr->entrytable[count],
2490 &entry, sizeof(entry))) {
2502 __ip_vs_get_timeouts(struct net *net, struct ip_vs_timeout_user *u)
2504 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2505 struct ip_vs_proto_data *pd;
2508 memset(u, 0, sizeof (*u));
2510 #ifdef CONFIG_IP_VS_PROTO_TCP
2511 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2512 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2513 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2515 #ifdef CONFIG_IP_VS_PROTO_UDP
2516 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2518 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2523 #define GET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2524 #define GET_INFO_ARG_LEN (sizeof(struct ip_vs_getinfo))
2525 #define GET_SERVICES_ARG_LEN (sizeof(struct ip_vs_get_services))
2526 #define GET_SERVICE_ARG_LEN (sizeof(struct ip_vs_service_entry))
2527 #define GET_DESTS_ARG_LEN (sizeof(struct ip_vs_get_dests))
2528 #define GET_TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2529 #define GET_DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user) * 2)
2531 static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
2532 [GET_CMDID(IP_VS_SO_GET_VERSION)] = 64,
2533 [GET_CMDID(IP_VS_SO_GET_INFO)] = GET_INFO_ARG_LEN,
2534 [GET_CMDID(IP_VS_SO_GET_SERVICES)] = GET_SERVICES_ARG_LEN,
2535 [GET_CMDID(IP_VS_SO_GET_SERVICE)] = GET_SERVICE_ARG_LEN,
2536 [GET_CMDID(IP_VS_SO_GET_DESTS)] = GET_DESTS_ARG_LEN,
2537 [GET_CMDID(IP_VS_SO_GET_TIMEOUT)] = GET_TIMEOUT_ARG_LEN,
2538 [GET_CMDID(IP_VS_SO_GET_DAEMON)] = GET_DAEMON_ARG_LEN,
2542 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2544 unsigned char arg[128];
2546 unsigned int copylen;
2547 struct net *net = sock_net(sk);
2548 struct netns_ipvs *ipvs = net_ipvs(net);
2551 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2554 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2557 if (*len < get_arglen[GET_CMDID(cmd)]) {
2558 pr_err("get_ctl: len %u < %u\n",
2559 *len, get_arglen[GET_CMDID(cmd)]);
2563 copylen = get_arglen[GET_CMDID(cmd)];
2567 if (copy_from_user(arg, user, copylen) != 0)
2570 * Handle daemons first since it has its own locking
2572 if (cmd == IP_VS_SO_GET_DAEMON) {
2573 struct ip_vs_daemon_user d[2];
2575 memset(&d, 0, sizeof(d));
2576 if (mutex_lock_interruptible(&ipvs->sync_mutex))
2577 return -ERESTARTSYS;
2579 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2580 d[0].state = IP_VS_STATE_MASTER;
2581 strlcpy(d[0].mcast_ifn, ipvs->master_mcast_ifn,
2582 sizeof(d[0].mcast_ifn));
2583 d[0].syncid = ipvs->master_syncid;
2585 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2586 d[1].state = IP_VS_STATE_BACKUP;
2587 strlcpy(d[1].mcast_ifn, ipvs->backup_mcast_ifn,
2588 sizeof(d[1].mcast_ifn));
2589 d[1].syncid = ipvs->backup_syncid;
2591 if (copy_to_user(user, &d, sizeof(d)) != 0)
2593 mutex_unlock(&ipvs->sync_mutex);
2597 if (mutex_lock_interruptible(&__ip_vs_mutex))
2598 return -ERESTARTSYS;
2601 case IP_VS_SO_GET_VERSION:
2605 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2606 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2607 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2611 *len = strlen(buf)+1;
2615 case IP_VS_SO_GET_INFO:
2617 struct ip_vs_getinfo info;
2618 info.version = IP_VS_VERSION_CODE;
2619 info.size = ip_vs_conn_tab_size;
2620 info.num_services = ipvs->num_services;
2621 if (copy_to_user(user, &info, sizeof(info)) != 0)
2626 case IP_VS_SO_GET_SERVICES:
2628 struct ip_vs_get_services *get;
2631 get = (struct ip_vs_get_services *)arg;
2632 size = sizeof(*get) +
2633 sizeof(struct ip_vs_service_entry) * get->num_services;
2635 pr_err("length: %u != %u\n", *len, size);
2639 ret = __ip_vs_get_service_entries(net, get, user);
2643 case IP_VS_SO_GET_SERVICE:
2645 struct ip_vs_service_entry *entry;
2646 struct ip_vs_service *svc;
2647 union nf_inet_addr addr;
2649 entry = (struct ip_vs_service_entry *)arg;
2650 addr.ip = entry->addr;
2653 svc = __ip_vs_svc_fwm_find(net, AF_INET, entry->fwmark);
2655 svc = __ip_vs_service_find(net, AF_INET,
2656 entry->protocol, &addr,
2660 ip_vs_copy_service(entry, svc);
2661 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2668 case IP_VS_SO_GET_DESTS:
2670 struct ip_vs_get_dests *get;
2673 get = (struct ip_vs_get_dests *)arg;
2674 size = sizeof(*get) +
2675 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2677 pr_err("length: %u != %u\n", *len, size);
2681 ret = __ip_vs_get_dest_entries(net, get, user);
2685 case IP_VS_SO_GET_TIMEOUT:
2687 struct ip_vs_timeout_user t;
2689 __ip_vs_get_timeouts(net, &t);
2690 if (copy_to_user(user, &t, sizeof(t)) != 0)
2700 mutex_unlock(&__ip_vs_mutex);
2705 static struct nf_sockopt_ops ip_vs_sockopts = {
2707 .set_optmin = IP_VS_BASE_CTL,
2708 .set_optmax = IP_VS_SO_SET_MAX+1,
2709 .set = do_ip_vs_set_ctl,
2710 .get_optmin = IP_VS_BASE_CTL,
2711 .get_optmax = IP_VS_SO_GET_MAX+1,
2712 .get = do_ip_vs_get_ctl,
2713 .owner = THIS_MODULE,
2717 * Generic Netlink interface
2720 /* IPVS genetlink family */
2721 static struct genl_family ip_vs_genl_family = {
2722 .id = GENL_ID_GENERATE,
2724 .name = IPVS_GENL_NAME,
2725 .version = IPVS_GENL_VERSION,
2726 .maxattr = IPVS_CMD_MAX,
2727 .netnsok = true, /* Make ipvsadm to work on netns */
2730 /* Policy used for first-level command attributes */
2731 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2732 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2733 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2734 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2735 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2736 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2737 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2740 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2741 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2742 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2743 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2744 .len = IP_VS_IFNAME_MAXLEN },
2745 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2748 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2749 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2750 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2751 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2752 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2753 .len = sizeof(union nf_inet_addr) },
2754 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2755 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2756 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2757 .len = IP_VS_SCHEDNAME_MAXLEN },
2758 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2759 .len = IP_VS_PENAME_MAXLEN },
2760 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2761 .len = sizeof(struct ip_vs_flags) },
2762 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2763 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2764 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2767 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2768 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2769 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2770 .len = sizeof(union nf_inet_addr) },
2771 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2772 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2773 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2774 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2775 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2776 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2777 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2778 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2779 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2782 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2783 struct ip_vs_stats *stats)
2785 struct ip_vs_stats_user ustats;
2786 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2790 ip_vs_copy_stats(&ustats, stats);
2792 if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, ustats.conns) ||
2793 nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, ustats.inpkts) ||
2794 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, ustats.outpkts) ||
2795 nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, ustats.inbytes) ||
2796 nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, ustats.outbytes) ||
2797 nla_put_u32(skb, IPVS_STATS_ATTR_CPS, ustats.cps) ||
2798 nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, ustats.inpps) ||
2799 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, ustats.outpps) ||
2800 nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, ustats.inbps) ||
2801 nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, ustats.outbps))
2802 goto nla_put_failure;
2803 nla_nest_end(skb, nl_stats);
2808 nla_nest_cancel(skb, nl_stats);
2812 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2813 struct ip_vs_service *svc)
2815 struct ip_vs_scheduler *sched;
2816 struct ip_vs_pe *pe;
2817 struct nlattr *nl_service;
2818 struct ip_vs_flags flags = { .flags = svc->flags,
2821 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2825 if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2826 goto nla_put_failure;
2828 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2829 goto nla_put_failure;
2831 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2832 nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2833 nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2834 goto nla_put_failure;
2837 sched = rcu_dereference_protected(svc->scheduler, 1);
2838 pe = rcu_dereference_protected(svc->pe, 1);
2839 if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched->name) ||
2840 (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
2841 nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2842 nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2843 nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
2844 goto nla_put_failure;
2845 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2846 goto nla_put_failure;
2848 nla_nest_end(skb, nl_service);
2853 nla_nest_cancel(skb, nl_service);
2857 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2858 struct ip_vs_service *svc,
2859 struct netlink_callback *cb)
2863 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2864 &ip_vs_genl_family, NLM_F_MULTI,
2865 IPVS_CMD_NEW_SERVICE);
2869 if (ip_vs_genl_fill_service(skb, svc) < 0)
2870 goto nla_put_failure;
2872 return genlmsg_end(skb, hdr);
2875 genlmsg_cancel(skb, hdr);
2879 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2880 struct netlink_callback *cb)
2883 int start = cb->args[0];
2884 struct ip_vs_service *svc;
2885 struct net *net = skb_sknet(skb);
2887 mutex_lock(&__ip_vs_mutex);
2888 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2889 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_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;
2899 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2900 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
2901 if (++idx <= start || !net_eq(svc->net, net))
2903 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2905 goto nla_put_failure;
2911 mutex_unlock(&__ip_vs_mutex);
2917 static int ip_vs_genl_parse_service(struct net *net,
2918 struct ip_vs_service_user_kern *usvc,
2919 struct nlattr *nla, int full_entry,
2920 struct ip_vs_service **ret_svc)
2922 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
2923 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
2924 struct ip_vs_service *svc;
2926 /* Parse mandatory identifying service fields first */
2928 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
2931 nla_af = attrs[IPVS_SVC_ATTR_AF];
2932 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
2933 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
2934 nla_port = attrs[IPVS_SVC_ATTR_PORT];
2935 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
2937 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
2940 memset(usvc, 0, sizeof(*usvc));
2942 usvc->af = nla_get_u16(nla_af);
2943 #ifdef CONFIG_IP_VS_IPV6
2944 if (usvc->af != AF_INET && usvc->af != AF_INET6)
2946 if (usvc->af != AF_INET)
2948 return -EAFNOSUPPORT;
2951 usvc->protocol = IPPROTO_TCP;
2952 usvc->fwmark = nla_get_u32(nla_fwmark);
2954 usvc->protocol = nla_get_u16(nla_protocol);
2955 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
2956 usvc->port = nla_get_be16(nla_port);
2962 svc = __ip_vs_svc_fwm_find(net, usvc->af, usvc->fwmark);
2964 svc = __ip_vs_service_find(net, usvc->af, usvc->protocol,
2965 &usvc->addr, usvc->port);
2969 /* If a full entry was requested, check for the additional fields */
2971 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
2973 struct ip_vs_flags flags;
2975 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
2976 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
2977 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
2978 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
2979 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
2981 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
2984 nla_memcpy(&flags, nla_flags, sizeof(flags));
2986 /* prefill flags from service if it already exists */
2988 usvc->flags = svc->flags;
2990 /* set new flags from userland */
2991 usvc->flags = (usvc->flags & ~flags.mask) |
2992 (flags.flags & flags.mask);
2993 usvc->sched_name = nla_data(nla_sched);
2994 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
2995 usvc->timeout = nla_get_u32(nla_timeout);
2996 usvc->netmask = nla_get_be32(nla_netmask);
3002 static struct ip_vs_service *ip_vs_genl_find_service(struct net *net,
3005 struct ip_vs_service_user_kern usvc;
3006 struct ip_vs_service *svc;
3009 ret = ip_vs_genl_parse_service(net, &usvc, nla, 0, &svc);
3010 return ret ? ERR_PTR(ret) : svc;
3013 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3015 struct nlattr *nl_dest;
3017 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3021 if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3022 nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3023 nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3024 (atomic_read(&dest->conn_flags) &
3025 IP_VS_CONN_F_FWD_MASK)) ||
3026 nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3027 atomic_read(&dest->weight)) ||
3028 nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3029 nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3030 nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3031 atomic_read(&dest->activeconns)) ||
3032 nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3033 atomic_read(&dest->inactconns)) ||
3034 nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3035 atomic_read(&dest->persistconns)))
3036 goto nla_put_failure;
3037 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
3038 goto nla_put_failure;
3040 nla_nest_end(skb, nl_dest);
3045 nla_nest_cancel(skb, nl_dest);
3049 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3050 struct netlink_callback *cb)
3054 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3055 &ip_vs_genl_family, NLM_F_MULTI,
3060 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3061 goto nla_put_failure;
3063 return genlmsg_end(skb, hdr);
3066 genlmsg_cancel(skb, hdr);
3070 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3071 struct netlink_callback *cb)
3074 int start = cb->args[0];
3075 struct ip_vs_service *svc;
3076 struct ip_vs_dest *dest;
3077 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3078 struct net *net = skb_sknet(skb);
3080 mutex_lock(&__ip_vs_mutex);
3082 /* Try to find the service for which to dump destinations */
3083 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3084 IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3088 svc = ip_vs_genl_find_service(net, attrs[IPVS_CMD_ATTR_SERVICE]);
3089 if (IS_ERR(svc) || svc == NULL)
3092 /* Dump the destinations */
3093 list_for_each_entry(dest, &svc->destinations, n_list) {
3096 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3098 goto nla_put_failure;
3106 mutex_unlock(&__ip_vs_mutex);
3111 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3112 struct nlattr *nla, int full_entry)
3114 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3115 struct nlattr *nla_addr, *nla_port;
3117 /* Parse mandatory identifying destination fields first */
3119 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3122 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3123 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3125 if (!(nla_addr && nla_port))
3128 memset(udest, 0, sizeof(*udest));
3130 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3131 udest->port = nla_get_be16(nla_port);
3133 /* If a full entry was requested, check for the additional fields */
3135 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3138 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3139 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3140 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3141 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3143 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3146 udest->conn_flags = nla_get_u32(nla_fwd)
3147 & IP_VS_CONN_F_FWD_MASK;
3148 udest->weight = nla_get_u32(nla_weight);
3149 udest->u_threshold = nla_get_u32(nla_u_thresh);
3150 udest->l_threshold = nla_get_u32(nla_l_thresh);
3156 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3157 const char *mcast_ifn, __u32 syncid)
3159 struct nlattr *nl_daemon;
3161 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3165 if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3166 nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn) ||
3167 nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid))
3168 goto nla_put_failure;
3169 nla_nest_end(skb, nl_daemon);
3174 nla_nest_cancel(skb, nl_daemon);
3178 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3179 const char *mcast_ifn, __u32 syncid,
3180 struct netlink_callback *cb)
3183 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3184 &ip_vs_genl_family, NLM_F_MULTI,
3185 IPVS_CMD_NEW_DAEMON);
3189 if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
3190 goto nla_put_failure;
3192 return genlmsg_end(skb, hdr);
3195 genlmsg_cancel(skb, hdr);
3199 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3200 struct netlink_callback *cb)
3202 struct net *net = skb_sknet(skb);
3203 struct netns_ipvs *ipvs = net_ipvs(net);
3205 mutex_lock(&ipvs->sync_mutex);
3206 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3207 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3208 ipvs->master_mcast_ifn,
3209 ipvs->master_syncid, cb) < 0)
3210 goto nla_put_failure;
3215 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3216 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3217 ipvs->backup_mcast_ifn,
3218 ipvs->backup_syncid, cb) < 0)
3219 goto nla_put_failure;
3225 mutex_unlock(&ipvs->sync_mutex);
3230 static int ip_vs_genl_new_daemon(struct net *net, struct nlattr **attrs)
3232 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3233 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3234 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3237 return start_sync_thread(net,
3238 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3239 nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3240 nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3243 static int ip_vs_genl_del_daemon(struct net *net, struct nlattr **attrs)
3245 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3248 return stop_sync_thread(net,
3249 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3252 static int ip_vs_genl_set_config(struct net *net, struct nlattr **attrs)
3254 struct ip_vs_timeout_user t;
3256 __ip_vs_get_timeouts(net, &t);
3258 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3259 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3261 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3263 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3265 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3266 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3268 return ip_vs_set_timeout(net, &t);
3271 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3275 struct netns_ipvs *ipvs;
3277 net = skb_sknet(skb);
3278 ipvs = net_ipvs(net);
3279 cmd = info->genlhdr->cmd;
3281 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3282 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3284 mutex_lock(&ipvs->sync_mutex);
3285 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3286 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3287 info->attrs[IPVS_CMD_ATTR_DAEMON],
3288 ip_vs_daemon_policy)) {
3293 if (cmd == IPVS_CMD_NEW_DAEMON)
3294 ret = ip_vs_genl_new_daemon(net, daemon_attrs);
3296 ret = ip_vs_genl_del_daemon(net, daemon_attrs);
3298 mutex_unlock(&ipvs->sync_mutex);
3303 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3305 struct ip_vs_service *svc = NULL;
3306 struct ip_vs_service_user_kern usvc;
3307 struct ip_vs_dest_user_kern udest;
3309 int need_full_svc = 0, need_full_dest = 0;
3312 net = skb_sknet(skb);
3313 cmd = info->genlhdr->cmd;
3315 mutex_lock(&__ip_vs_mutex);
3317 if (cmd == IPVS_CMD_FLUSH) {
3318 ret = ip_vs_flush(net, false);
3320 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3321 ret = ip_vs_genl_set_config(net, info->attrs);
3323 } else if (cmd == IPVS_CMD_ZERO &&
3324 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3325 ret = ip_vs_zero_all(net);
3329 /* All following commands require a service argument, so check if we
3330 * received a valid one. We need a full service specification when
3331 * adding / editing a service. Only identifying members otherwise. */
3332 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3335 ret = ip_vs_genl_parse_service(net, &usvc,
3336 info->attrs[IPVS_CMD_ATTR_SERVICE],
3337 need_full_svc, &svc);
3341 /* Unless we're adding a new service, the service must already exist */
3342 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3347 /* Destination commands require a valid destination argument. For
3348 * adding / editing a destination, we need a full destination
3350 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3351 cmd == IPVS_CMD_DEL_DEST) {
3352 if (cmd != IPVS_CMD_DEL_DEST)
3355 ret = ip_vs_genl_parse_dest(&udest,
3356 info->attrs[IPVS_CMD_ATTR_DEST],
3363 case IPVS_CMD_NEW_SERVICE:
3365 ret = ip_vs_add_service(net, &usvc, &svc);
3369 case IPVS_CMD_SET_SERVICE:
3370 ret = ip_vs_edit_service(svc, &usvc);
3372 case IPVS_CMD_DEL_SERVICE:
3373 ret = ip_vs_del_service(svc);
3374 /* do not use svc, it can be freed */
3376 case IPVS_CMD_NEW_DEST:
3377 ret = ip_vs_add_dest(svc, &udest);
3379 case IPVS_CMD_SET_DEST:
3380 ret = ip_vs_edit_dest(svc, &udest);
3382 case IPVS_CMD_DEL_DEST:
3383 ret = ip_vs_del_dest(svc, &udest);
3386 ret = ip_vs_zero_service(svc);
3393 mutex_unlock(&__ip_vs_mutex);
3398 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3400 struct sk_buff *msg;
3402 int ret, cmd, reply_cmd;
3405 net = skb_sknet(skb);
3406 cmd = info->genlhdr->cmd;
3408 if (cmd == IPVS_CMD_GET_SERVICE)
3409 reply_cmd = IPVS_CMD_NEW_SERVICE;
3410 else if (cmd == IPVS_CMD_GET_INFO)
3411 reply_cmd = IPVS_CMD_SET_INFO;
3412 else if (cmd == IPVS_CMD_GET_CONFIG)
3413 reply_cmd = IPVS_CMD_SET_CONFIG;
3415 pr_err("unknown Generic Netlink command\n");
3419 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3423 mutex_lock(&__ip_vs_mutex);
3425 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3427 goto nla_put_failure;
3430 case IPVS_CMD_GET_SERVICE:
3432 struct ip_vs_service *svc;
3434 svc = ip_vs_genl_find_service(net,
3435 info->attrs[IPVS_CMD_ATTR_SERVICE]);
3440 ret = ip_vs_genl_fill_service(msg, svc);
3442 goto nla_put_failure;
3451 case IPVS_CMD_GET_CONFIG:
3453 struct ip_vs_timeout_user t;
3455 __ip_vs_get_timeouts(net, &t);
3456 #ifdef CONFIG_IP_VS_PROTO_TCP
3457 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3459 nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3461 goto nla_put_failure;
3463 #ifdef CONFIG_IP_VS_PROTO_UDP
3464 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3465 goto nla_put_failure;
3471 case IPVS_CMD_GET_INFO:
3472 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3473 IP_VS_VERSION_CODE) ||
3474 nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3475 ip_vs_conn_tab_size))
3476 goto nla_put_failure;
3480 genlmsg_end(msg, reply);
3481 ret = genlmsg_reply(msg, info);
3485 pr_err("not enough space in Netlink message\n");
3491 mutex_unlock(&__ip_vs_mutex);
3497 static const struct genl_ops ip_vs_genl_ops[] = {
3499 .cmd = IPVS_CMD_NEW_SERVICE,
3500 .flags = GENL_ADMIN_PERM,
3501 .policy = ip_vs_cmd_policy,
3502 .doit = ip_vs_genl_set_cmd,
3505 .cmd = IPVS_CMD_SET_SERVICE,
3506 .flags = GENL_ADMIN_PERM,
3507 .policy = ip_vs_cmd_policy,
3508 .doit = ip_vs_genl_set_cmd,
3511 .cmd = IPVS_CMD_DEL_SERVICE,
3512 .flags = GENL_ADMIN_PERM,
3513 .policy = ip_vs_cmd_policy,
3514 .doit = ip_vs_genl_set_cmd,
3517 .cmd = IPVS_CMD_GET_SERVICE,
3518 .flags = GENL_ADMIN_PERM,
3519 .doit = ip_vs_genl_get_cmd,
3520 .dumpit = ip_vs_genl_dump_services,
3521 .policy = ip_vs_cmd_policy,
3524 .cmd = IPVS_CMD_NEW_DEST,
3525 .flags = GENL_ADMIN_PERM,
3526 .policy = ip_vs_cmd_policy,
3527 .doit = ip_vs_genl_set_cmd,
3530 .cmd = IPVS_CMD_SET_DEST,
3531 .flags = GENL_ADMIN_PERM,
3532 .policy = ip_vs_cmd_policy,
3533 .doit = ip_vs_genl_set_cmd,
3536 .cmd = IPVS_CMD_DEL_DEST,
3537 .flags = GENL_ADMIN_PERM,
3538 .policy = ip_vs_cmd_policy,
3539 .doit = ip_vs_genl_set_cmd,
3542 .cmd = IPVS_CMD_GET_DEST,
3543 .flags = GENL_ADMIN_PERM,
3544 .policy = ip_vs_cmd_policy,
3545 .dumpit = ip_vs_genl_dump_dests,
3548 .cmd = IPVS_CMD_NEW_DAEMON,
3549 .flags = GENL_ADMIN_PERM,
3550 .policy = ip_vs_cmd_policy,
3551 .doit = ip_vs_genl_set_daemon,
3554 .cmd = IPVS_CMD_DEL_DAEMON,
3555 .flags = GENL_ADMIN_PERM,
3556 .policy = ip_vs_cmd_policy,
3557 .doit = ip_vs_genl_set_daemon,
3560 .cmd = IPVS_CMD_GET_DAEMON,
3561 .flags = GENL_ADMIN_PERM,
3562 .dumpit = ip_vs_genl_dump_daemons,
3565 .cmd = IPVS_CMD_SET_CONFIG,
3566 .flags = GENL_ADMIN_PERM,
3567 .policy = ip_vs_cmd_policy,
3568 .doit = ip_vs_genl_set_cmd,
3571 .cmd = IPVS_CMD_GET_CONFIG,
3572 .flags = GENL_ADMIN_PERM,
3573 .doit = ip_vs_genl_get_cmd,
3576 .cmd = IPVS_CMD_GET_INFO,
3577 .flags = GENL_ADMIN_PERM,
3578 .doit = ip_vs_genl_get_cmd,
3581 .cmd = IPVS_CMD_ZERO,
3582 .flags = GENL_ADMIN_PERM,
3583 .policy = ip_vs_cmd_policy,
3584 .doit = ip_vs_genl_set_cmd,
3587 .cmd = IPVS_CMD_FLUSH,
3588 .flags = GENL_ADMIN_PERM,
3589 .doit = ip_vs_genl_set_cmd,
3593 static int __init ip_vs_genl_register(void)
3595 return genl_register_family_with_ops(&ip_vs_genl_family,
3599 static void ip_vs_genl_unregister(void)
3601 genl_unregister_family(&ip_vs_genl_family);
3604 /* End of Generic Netlink interface definitions */
3607 * per netns intit/exit func.
3609 #ifdef CONFIG_SYSCTL
3610 static int __net_init ip_vs_control_net_init_sysctl(struct net *net)
3613 struct netns_ipvs *ipvs = net_ipvs(net);
3614 struct ctl_table *tbl;
3616 atomic_set(&ipvs->dropentry, 0);
3617 spin_lock_init(&ipvs->dropentry_lock);
3618 spin_lock_init(&ipvs->droppacket_lock);
3619 spin_lock_init(&ipvs->securetcp_lock);
3621 if (!net_eq(net, &init_net)) {
3622 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3626 /* Don't export sysctls to unprivileged users */
3627 if (net->user_ns != &init_user_ns)
3628 tbl[0].procname = NULL;
3631 /* Initialize sysctl defaults */
3633 ipvs->sysctl_amemthresh = 1024;
3634 tbl[idx++].data = &ipvs->sysctl_amemthresh;
3635 ipvs->sysctl_am_droprate = 10;
3636 tbl[idx++].data = &ipvs->sysctl_am_droprate;
3637 tbl[idx++].data = &ipvs->sysctl_drop_entry;
3638 tbl[idx++].data = &ipvs->sysctl_drop_packet;
3639 #ifdef CONFIG_IP_VS_NFCT
3640 tbl[idx++].data = &ipvs->sysctl_conntrack;
3642 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3643 ipvs->sysctl_snat_reroute = 1;
3644 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3645 ipvs->sysctl_sync_ver = 1;
3646 tbl[idx++].data = &ipvs->sysctl_sync_ver;
3647 ipvs->sysctl_sync_ports = 1;
3648 tbl[idx++].data = &ipvs->sysctl_sync_ports;
3649 tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
3650 ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3651 tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3652 ipvs->sysctl_sync_sock_size = 0;
3653 tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3654 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3655 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3656 tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
3657 tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
3658 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3659 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3660 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3661 tbl[idx].data = &ipvs->sysctl_sync_threshold;
3662 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3663 ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3664 tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3665 ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3666 tbl[idx++].data = &ipvs->sysctl_sync_retries;
3667 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3668 ipvs->sysctl_pmtu_disc = 1;
3669 tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
3670 tbl[idx++].data = &ipvs->sysctl_backup_only;
3673 ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3674 if (ipvs->sysctl_hdr == NULL) {
3675 if (!net_eq(net, &init_net))
3679 ip_vs_start_estimator(net, &ipvs->tot_stats);
3680 ipvs->sysctl_tbl = tbl;
3681 /* Schedule defense work */
3682 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3683 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3688 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net)
3690 struct netns_ipvs *ipvs = net_ipvs(net);
3692 cancel_delayed_work_sync(&ipvs->defense_work);
3693 cancel_work_sync(&ipvs->defense_work.work);
3694 unregister_net_sysctl_table(ipvs->sysctl_hdr);
3695 ip_vs_stop_estimator(net, &ipvs->tot_stats);
3700 static int __net_init ip_vs_control_net_init_sysctl(struct net *net) { return 0; }
3701 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net) { }
3705 static struct notifier_block ip_vs_dst_notifier = {
3706 .notifier_call = ip_vs_dst_event,
3709 int __net_init ip_vs_control_net_init(struct net *net)
3712 struct netns_ipvs *ipvs = net_ipvs(net);
3714 /* Initialize rs_table */
3715 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
3716 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
3718 INIT_LIST_HEAD(&ipvs->dest_trash);
3719 spin_lock_init(&ipvs->dest_trash_lock);
3720 setup_timer(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire,
3721 (unsigned long) net);
3722 atomic_set(&ipvs->ftpsvc_counter, 0);
3723 atomic_set(&ipvs->nullsvc_counter, 0);
3726 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
3727 if (!ipvs->tot_stats.cpustats)
3730 for_each_possible_cpu(i) {
3731 struct ip_vs_cpu_stats *ipvs_tot_stats;
3732 ipvs_tot_stats = per_cpu_ptr(ipvs->tot_stats.cpustats, i);
3733 u64_stats_init(&ipvs_tot_stats->syncp);
3736 spin_lock_init(&ipvs->tot_stats.lock);
3738 proc_create("ip_vs", 0, net->proc_net, &ip_vs_info_fops);
3739 proc_create("ip_vs_stats", 0, net->proc_net, &ip_vs_stats_fops);
3740 proc_create("ip_vs_stats_percpu", 0, net->proc_net,
3741 &ip_vs_stats_percpu_fops);
3743 if (ip_vs_control_net_init_sysctl(net))
3749 free_percpu(ipvs->tot_stats.cpustats);
3753 void __net_exit ip_vs_control_net_cleanup(struct net *net)
3755 struct netns_ipvs *ipvs = net_ipvs(net);
3757 ip_vs_trash_cleanup(net);
3758 ip_vs_control_net_cleanup_sysctl(net);
3759 remove_proc_entry("ip_vs_stats_percpu", net->proc_net);
3760 remove_proc_entry("ip_vs_stats", net->proc_net);
3761 remove_proc_entry("ip_vs", net->proc_net);
3762 free_percpu(ipvs->tot_stats.cpustats);
3765 int __init ip_vs_register_nl_ioctl(void)
3769 ret = nf_register_sockopt(&ip_vs_sockopts);
3771 pr_err("cannot register sockopt.\n");
3775 ret = ip_vs_genl_register();
3777 pr_err("cannot register Generic Netlink interface.\n");
3783 nf_unregister_sockopt(&ip_vs_sockopts);
3788 void ip_vs_unregister_nl_ioctl(void)
3790 ip_vs_genl_unregister();
3791 nf_unregister_sockopt(&ip_vs_sockopts);
3794 int __init ip_vs_control_init(void)
3801 /* Initialize svc_table, ip_vs_svc_fwm_table */
3802 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
3803 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
3804 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3807 smp_wmb(); /* Do we really need it now ? */
3809 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
3818 void ip_vs_control_cleanup(void)
3821 unregister_netdevice_notifier(&ip_vs_dst_notifier);