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);
466 static void ip_vs_service_free(struct ip_vs_service *svc)
468 if (svc->stats.cpustats)
469 free_percpu(svc->stats.cpustats);
474 __ip_vs_unbind_svc(struct ip_vs_dest *dest)
476 struct ip_vs_service *svc = dest->svc;
479 if (atomic_dec_and_test(&svc->refcnt)) {
480 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n",
482 IP_VS_DBG_ADDR(svc->af, &svc->addr),
484 ip_vs_service_free(svc);
490 * Returns hash value for real service
492 static inline unsigned int ip_vs_rs_hashkey(int af,
493 const union nf_inet_addr *addr,
496 register unsigned int porth = ntohs(port);
497 __be32 addr_fold = addr->ip;
499 #ifdef CONFIG_IP_VS_IPV6
501 addr_fold = addr->ip6[0]^addr->ip6[1]^
502 addr->ip6[2]^addr->ip6[3];
505 return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
509 /* Hash ip_vs_dest in rs_table by <proto,addr,port>. */
510 static void ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
514 if (dest->in_rs_table)
518 * Hash by proto,addr,port,
519 * which are the parameters of the real service.
521 hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
523 hlist_add_head_rcu(&dest->d_list, &ipvs->rs_table[hash]);
524 dest->in_rs_table = 1;
527 /* Unhash ip_vs_dest from rs_table. */
528 static void ip_vs_rs_unhash(struct ip_vs_dest *dest)
531 * Remove it from the rs_table table.
533 if (dest->in_rs_table) {
534 hlist_del_rcu(&dest->d_list);
535 dest->in_rs_table = 0;
539 /* Check if real service by <proto,addr,port> is present */
540 bool ip_vs_has_real_service(struct net *net, int af, __u16 protocol,
541 const union nf_inet_addr *daddr, __be16 dport)
543 struct netns_ipvs *ipvs = net_ipvs(net);
545 struct ip_vs_dest *dest;
547 /* Check for "full" addressed entries */
548 hash = ip_vs_rs_hashkey(af, daddr, dport);
551 hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
552 if (dest->port == dport &&
554 ip_vs_addr_equal(af, &dest->addr, daddr) &&
555 (dest->protocol == protocol || dest->vfwmark)) {
566 /* Lookup destination by {addr,port} in the given service
567 * Called under RCU lock.
569 static struct ip_vs_dest *
570 ip_vs_lookup_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
573 struct ip_vs_dest *dest;
576 * Find the destination for the given service
578 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
579 if ((dest->af == svc->af)
580 && ip_vs_addr_equal(svc->af, &dest->addr, daddr)
581 && (dest->port == dport)) {
591 * Find destination by {daddr,dport,vaddr,protocol}
592 * Created to be used in ip_vs_process_message() in
593 * the backup synchronization daemon. It finds the
594 * destination to be bound to the received connection
596 * Called under RCU lock, no refcnt is returned.
598 struct ip_vs_dest *ip_vs_find_dest(struct net *net, int af,
599 const union nf_inet_addr *daddr,
601 const union nf_inet_addr *vaddr,
602 __be16 vport, __u16 protocol, __u32 fwmark,
605 struct ip_vs_dest *dest;
606 struct ip_vs_service *svc;
609 svc = ip_vs_service_find(net, af, fwmark, protocol, vaddr, vport);
612 if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
614 dest = ip_vs_lookup_dest(svc, daddr, port);
616 dest = ip_vs_lookup_dest(svc, daddr, port ^ dport);
620 void ip_vs_dest_dst_rcu_free(struct rcu_head *head)
622 struct ip_vs_dest_dst *dest_dst = container_of(head,
623 struct ip_vs_dest_dst,
626 dst_release(dest_dst->dst_cache);
630 /* Release dest_dst and dst_cache for dest in user context */
631 static void __ip_vs_dst_cache_reset(struct ip_vs_dest *dest)
633 struct ip_vs_dest_dst *old;
635 old = rcu_dereference_protected(dest->dest_dst, 1);
637 RCU_INIT_POINTER(dest->dest_dst, NULL);
638 call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
643 * Lookup dest by {svc,addr,port} in the destination trash.
644 * The destination trash is used to hold the destinations that are removed
645 * from the service table but are still referenced by some conn entries.
646 * The reason to add the destination trash is when the dest is temporary
647 * down (either by administrator or by monitor program), the dest can be
648 * picked back from the trash, the remaining connections to the dest can
649 * continue, and the counting information of the dest is also useful for
652 static struct ip_vs_dest *
653 ip_vs_trash_get_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
656 struct ip_vs_dest *dest;
657 struct netns_ipvs *ipvs = net_ipvs(svc->net);
660 * Find the destination in trash
662 spin_lock_bh(&ipvs->dest_trash_lock);
663 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
664 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
667 IP_VS_DBG_ADDR(svc->af, &dest->addr),
669 atomic_read(&dest->refcnt));
670 /* We can not reuse dest while in grace period
671 * because conns still can use dest->svc
673 if (test_bit(IP_VS_DEST_STATE_REMOVING, &dest->state))
675 if (dest->af == svc->af &&
676 ip_vs_addr_equal(svc->af, &dest->addr, daddr) &&
677 dest->port == dport &&
678 dest->vfwmark == svc->fwmark &&
679 dest->protocol == svc->protocol &&
681 (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
682 dest->vport == svc->port))) {
684 list_del(&dest->t_list);
685 ip_vs_dest_hold(dest);
693 spin_unlock_bh(&ipvs->dest_trash_lock);
698 static void ip_vs_dest_free(struct ip_vs_dest *dest)
700 __ip_vs_dst_cache_reset(dest);
701 __ip_vs_unbind_svc(dest);
702 free_percpu(dest->stats.cpustats);
707 * Clean up all the destinations in the trash
708 * Called by the ip_vs_control_cleanup()
710 * When the ip_vs_control_clearup is activated by ipvs module exit,
711 * the service tables must have been flushed and all the connections
712 * are expired, and the refcnt of each destination in the trash must
713 * be 0, so we simply release them here.
715 static void ip_vs_trash_cleanup(struct net *net)
717 struct ip_vs_dest *dest, *nxt;
718 struct netns_ipvs *ipvs = net_ipvs(net);
720 del_timer_sync(&ipvs->dest_trash_timer);
721 /* No need to use dest_trash_lock */
722 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, t_list) {
723 list_del(&dest->t_list);
724 ip_vs_dest_free(dest);
729 ip_vs_copy_stats(struct ip_vs_stats_user *dst, struct ip_vs_stats *src)
731 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->ustats.c - src->ustats0.c
733 spin_lock_bh(&src->lock);
735 IP_VS_SHOW_STATS_COUNTER(conns);
736 IP_VS_SHOW_STATS_COUNTER(inpkts);
737 IP_VS_SHOW_STATS_COUNTER(outpkts);
738 IP_VS_SHOW_STATS_COUNTER(inbytes);
739 IP_VS_SHOW_STATS_COUNTER(outbytes);
741 ip_vs_read_estimator(dst, src);
743 spin_unlock_bh(&src->lock);
747 ip_vs_zero_stats(struct ip_vs_stats *stats)
749 spin_lock_bh(&stats->lock);
751 /* get current counters as zero point, rates are zeroed */
753 #define IP_VS_ZERO_STATS_COUNTER(c) stats->ustats0.c = stats->ustats.c
755 IP_VS_ZERO_STATS_COUNTER(conns);
756 IP_VS_ZERO_STATS_COUNTER(inpkts);
757 IP_VS_ZERO_STATS_COUNTER(outpkts);
758 IP_VS_ZERO_STATS_COUNTER(inbytes);
759 IP_VS_ZERO_STATS_COUNTER(outbytes);
761 ip_vs_zero_estimator(stats);
763 spin_unlock_bh(&stats->lock);
767 * Update a destination in the given service
770 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
771 struct ip_vs_dest_user_kern *udest, int add)
773 struct netns_ipvs *ipvs = net_ipvs(svc->net);
774 struct ip_vs_scheduler *sched;
777 /* set the weight and the flags */
778 atomic_set(&dest->weight, udest->weight);
779 conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
780 conn_flags |= IP_VS_CONN_F_INACTIVE;
782 /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
783 if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
784 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
787 * Put the real service in rs_table if not present.
788 * For now only for NAT!
790 ip_vs_rs_hash(ipvs, dest);
792 atomic_set(&dest->conn_flags, conn_flags);
794 /* bind the service */
796 __ip_vs_bind_svc(dest, svc);
798 if (dest->svc != svc) {
799 __ip_vs_unbind_svc(dest);
800 ip_vs_zero_stats(&dest->stats);
801 __ip_vs_bind_svc(dest, svc);
805 /* set the dest status flags */
806 dest->flags |= IP_VS_DEST_F_AVAILABLE;
808 if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
809 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
810 dest->u_threshold = udest->u_threshold;
811 dest->l_threshold = udest->l_threshold;
813 spin_lock_bh(&dest->dst_lock);
814 __ip_vs_dst_cache_reset(dest);
815 spin_unlock_bh(&dest->dst_lock);
817 sched = rcu_dereference_protected(svc->scheduler, 1);
819 ip_vs_start_estimator(svc->net, &dest->stats);
820 list_add_rcu(&dest->n_list, &svc->destinations);
823 sched->add_dest(svc, dest);
826 sched->upd_dest(svc, dest);
832 * Create a destination for the given service
835 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
836 struct ip_vs_dest **dest_p)
838 struct ip_vs_dest *dest;
843 #ifdef CONFIG_IP_VS_IPV6
844 if (svc->af == AF_INET6) {
845 atype = ipv6_addr_type(&udest->addr.in6);
846 if ((!(atype & IPV6_ADDR_UNICAST) ||
847 atype & IPV6_ADDR_LINKLOCAL) &&
848 !__ip_vs_addr_is_local_v6(svc->net, &udest->addr.in6))
853 atype = inet_addr_type(svc->net, udest->addr.ip);
854 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
858 dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
862 dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
863 if (!dest->stats.cpustats)
867 dest->protocol = svc->protocol;
868 dest->vaddr = svc->addr;
869 dest->vport = svc->port;
870 dest->vfwmark = svc->fwmark;
871 ip_vs_addr_copy(svc->af, &dest->addr, &udest->addr);
872 dest->port = udest->port;
874 atomic_set(&dest->activeconns, 0);
875 atomic_set(&dest->inactconns, 0);
876 atomic_set(&dest->persistconns, 0);
877 atomic_set(&dest->refcnt, 1);
879 INIT_HLIST_NODE(&dest->d_list);
880 spin_lock_init(&dest->dst_lock);
881 spin_lock_init(&dest->stats.lock);
882 __ip_vs_update_dest(svc, dest, udest, 1);
896 * Add a destination into an existing service
899 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
901 struct ip_vs_dest *dest;
902 union nf_inet_addr daddr;
903 __be16 dport = udest->port;
908 if (udest->weight < 0) {
909 pr_err("%s(): server weight less than zero\n", __func__);
913 if (udest->l_threshold > udest->u_threshold) {
914 pr_err("%s(): lower threshold is higher than upper threshold\n",
919 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
921 /* We use function that requires RCU lock */
923 dest = ip_vs_lookup_dest(svc, &daddr, dport);
927 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
932 * Check if the dest already exists in the trash and
933 * is from the same service
935 dest = ip_vs_trash_get_dest(svc, &daddr, dport);
938 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
939 "dest->refcnt=%d, service %u/%s:%u\n",
940 IP_VS_DBG_ADDR(svc->af, &daddr), ntohs(dport),
941 atomic_read(&dest->refcnt),
943 IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
946 __ip_vs_update_dest(svc, dest, udest, 1);
950 * Allocate and initialize the dest structure
952 ret = ip_vs_new_dest(svc, udest, &dest);
961 * Edit a destination in the given service
964 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
966 struct ip_vs_dest *dest;
967 union nf_inet_addr daddr;
968 __be16 dport = udest->port;
972 if (udest->weight < 0) {
973 pr_err("%s(): server weight less than zero\n", __func__);
977 if (udest->l_threshold > udest->u_threshold) {
978 pr_err("%s(): lower threshold is higher than upper threshold\n",
983 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
985 /* We use function that requires RCU lock */
987 dest = ip_vs_lookup_dest(svc, &daddr, dport);
991 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
995 __ip_vs_update_dest(svc, dest, udest, 0);
1001 static void ip_vs_dest_wait_readers(struct rcu_head *head)
1003 struct ip_vs_dest *dest = container_of(head, struct ip_vs_dest,
1006 /* End of grace period after unlinking */
1007 clear_bit(IP_VS_DEST_STATE_REMOVING, &dest->state);
1012 * Delete a destination (must be already unlinked from the service)
1014 static void __ip_vs_del_dest(struct net *net, struct ip_vs_dest *dest,
1017 struct netns_ipvs *ipvs = net_ipvs(net);
1019 ip_vs_stop_estimator(net, &dest->stats);
1022 * Remove it from the d-linked list with the real services.
1024 ip_vs_rs_unhash(dest);
1027 set_bit(IP_VS_DEST_STATE_REMOVING, &dest->state);
1028 call_rcu(&dest->rcu_head, ip_vs_dest_wait_readers);
1031 spin_lock_bh(&ipvs->dest_trash_lock);
1032 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1033 IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port),
1034 atomic_read(&dest->refcnt));
1035 if (list_empty(&ipvs->dest_trash) && !cleanup)
1036 mod_timer(&ipvs->dest_trash_timer,
1037 jiffies + IP_VS_DEST_TRASH_PERIOD);
1038 /* dest lives in trash without reference */
1039 list_add(&dest->t_list, &ipvs->dest_trash);
1040 spin_unlock_bh(&ipvs->dest_trash_lock);
1041 ip_vs_dest_put(dest);
1046 * Unlink a destination from the given service
1048 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1049 struct ip_vs_dest *dest,
1052 dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1055 * Remove it from the d-linked destination list.
1057 list_del_rcu(&dest->n_list);
1061 struct ip_vs_scheduler *sched;
1063 sched = rcu_dereference_protected(svc->scheduler, 1);
1064 if (sched->del_dest)
1065 sched->del_dest(svc, dest);
1071 * Delete a destination server in the given service
1074 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1076 struct ip_vs_dest *dest;
1077 __be16 dport = udest->port;
1081 /* We use function that requires RCU lock */
1083 dest = ip_vs_lookup_dest(svc, &udest->addr, dport);
1087 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1092 * Unlink dest from the service
1094 __ip_vs_unlink_dest(svc, dest, 1);
1097 * Delete the destination
1099 __ip_vs_del_dest(svc->net, dest, false);
1106 static void ip_vs_dest_trash_expire(unsigned long data)
1108 struct net *net = (struct net *) data;
1109 struct netns_ipvs *ipvs = net_ipvs(net);
1110 struct ip_vs_dest *dest, *next;
1112 spin_lock(&ipvs->dest_trash_lock);
1113 list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) {
1114 /* Skip if dest is in grace period */
1115 if (test_bit(IP_VS_DEST_STATE_REMOVING, &dest->state))
1117 if (atomic_read(&dest->refcnt) > 0)
1119 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1121 IP_VS_DBG_ADDR(dest->svc->af, &dest->addr),
1123 list_del(&dest->t_list);
1124 ip_vs_dest_free(dest);
1126 if (!list_empty(&ipvs->dest_trash))
1127 mod_timer(&ipvs->dest_trash_timer,
1128 jiffies + IP_VS_DEST_TRASH_PERIOD);
1129 spin_unlock(&ipvs->dest_trash_lock);
1133 * Add a service into the service hash table
1136 ip_vs_add_service(struct net *net, struct ip_vs_service_user_kern *u,
1137 struct ip_vs_service **svc_p)
1140 struct ip_vs_scheduler *sched = NULL;
1141 struct ip_vs_pe *pe = NULL;
1142 struct ip_vs_service *svc = NULL;
1143 struct netns_ipvs *ipvs = net_ipvs(net);
1145 /* increase the module use count */
1146 ip_vs_use_count_inc();
1148 /* Lookup the scheduler by 'u->sched_name' */
1149 sched = ip_vs_scheduler_get(u->sched_name);
1150 if (sched == NULL) {
1151 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1156 if (u->pe_name && *u->pe_name) {
1157 pe = ip_vs_pe_getbyname(u->pe_name);
1159 pr_info("persistence engine module ip_vs_pe_%s "
1160 "not found\n", u->pe_name);
1166 #ifdef CONFIG_IP_VS_IPV6
1167 if (u->af == AF_INET6) {
1168 __u32 plen = (__force __u32) u->netmask;
1170 if (plen < 1 || plen > 128) {
1177 svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1179 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1183 svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1184 if (!svc->stats.cpustats) {
1189 /* I'm the first user of the service */
1190 atomic_set(&svc->refcnt, 0);
1193 svc->protocol = u->protocol;
1194 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1195 svc->port = u->port;
1196 svc->fwmark = u->fwmark;
1197 svc->flags = u->flags;
1198 svc->timeout = u->timeout * HZ;
1199 svc->netmask = u->netmask;
1202 INIT_LIST_HEAD(&svc->destinations);
1203 spin_lock_init(&svc->sched_lock);
1204 spin_lock_init(&svc->stats.lock);
1206 /* Bind the scheduler */
1207 ret = ip_vs_bind_scheduler(svc, sched);
1212 /* Bind the ct retriever */
1213 RCU_INIT_POINTER(svc->pe, pe);
1216 /* Update the virtual service counters */
1217 if (svc->port == FTPPORT)
1218 atomic_inc(&ipvs->ftpsvc_counter);
1219 else if (svc->port == 0)
1220 atomic_inc(&ipvs->nullsvc_counter);
1222 ip_vs_start_estimator(net, &svc->stats);
1224 /* Count only IPv4 services for old get/setsockopt interface */
1225 if (svc->af == AF_INET)
1226 ipvs->num_services++;
1228 /* Hash the service into the service table */
1229 ip_vs_svc_hash(svc);
1232 /* Now there is a service - full throttle */
1239 ip_vs_unbind_scheduler(svc, sched);
1240 ip_vs_service_free(svc);
1242 ip_vs_scheduler_put(sched);
1245 /* decrease the module use count */
1246 ip_vs_use_count_dec();
1253 * Edit a service and bind it with a new scheduler
1256 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1258 struct ip_vs_scheduler *sched, *old_sched;
1259 struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1263 * Lookup the scheduler, by 'u->sched_name'
1265 sched = ip_vs_scheduler_get(u->sched_name);
1266 if (sched == NULL) {
1267 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1272 if (u->pe_name && *u->pe_name) {
1273 pe = ip_vs_pe_getbyname(u->pe_name);
1275 pr_info("persistence engine module ip_vs_pe_%s "
1276 "not found\n", u->pe_name);
1283 #ifdef CONFIG_IP_VS_IPV6
1284 if (u->af == AF_INET6) {
1285 __u32 plen = (__force __u32) u->netmask;
1287 if (plen < 1 || plen > 128) {
1294 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1295 if (sched != old_sched) {
1296 /* Bind the new scheduler */
1297 ret = ip_vs_bind_scheduler(svc, sched);
1302 /* Unbind the old scheduler on success */
1303 ip_vs_unbind_scheduler(svc, old_sched);
1307 * Set the flags and timeout value
1309 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1310 svc->timeout = u->timeout * HZ;
1311 svc->netmask = u->netmask;
1313 old_pe = rcu_dereference_protected(svc->pe, 1);
1315 rcu_assign_pointer(svc->pe, pe);
1318 ip_vs_scheduler_put(old_sched);
1319 ip_vs_pe_put(old_pe);
1323 static void ip_vs_service_rcu_free(struct rcu_head *head)
1325 struct ip_vs_service *svc;
1327 svc = container_of(head, struct ip_vs_service, rcu_head);
1328 ip_vs_service_free(svc);
1332 * Delete a service from the service list
1333 * - The service must be unlinked, unlocked and not referenced!
1334 * - We are called under _bh lock
1336 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1338 struct ip_vs_dest *dest, *nxt;
1339 struct ip_vs_scheduler *old_sched;
1340 struct ip_vs_pe *old_pe;
1341 struct netns_ipvs *ipvs = net_ipvs(svc->net);
1343 pr_info("%s: enter\n", __func__);
1345 /* Count only IPv4 services for old get/setsockopt interface */
1346 if (svc->af == AF_INET)
1347 ipvs->num_services--;
1349 ip_vs_stop_estimator(svc->net, &svc->stats);
1351 /* Unbind scheduler */
1352 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1353 ip_vs_unbind_scheduler(svc, old_sched);
1354 ip_vs_scheduler_put(old_sched);
1356 /* Unbind persistence engine, keep svc->pe */
1357 old_pe = rcu_dereference_protected(svc->pe, 1);
1358 ip_vs_pe_put(old_pe);
1361 * Unlink the whole destination list
1363 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1364 __ip_vs_unlink_dest(svc, dest, 0);
1365 __ip_vs_del_dest(svc->net, dest, cleanup);
1369 * Update the virtual service counters
1371 if (svc->port == FTPPORT)
1372 atomic_dec(&ipvs->ftpsvc_counter);
1373 else if (svc->port == 0)
1374 atomic_dec(&ipvs->nullsvc_counter);
1377 * Free the service if nobody refers to it
1379 if (atomic_dec_and_test(&svc->refcnt)) {
1380 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n",
1382 IP_VS_DBG_ADDR(svc->af, &svc->addr),
1384 call_rcu(&svc->rcu_head, ip_vs_service_rcu_free);
1387 /* decrease the module use count */
1388 ip_vs_use_count_dec();
1392 * Unlink a service from list and try to delete it if its refcnt reached 0
1394 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1396 /* Hold svc to avoid double release from dest_trash */
1397 atomic_inc(&svc->refcnt);
1399 * Unhash it from the service table
1401 ip_vs_svc_unhash(svc);
1403 __ip_vs_del_service(svc, cleanup);
1407 * Delete a service from the service list
1409 static int ip_vs_del_service(struct ip_vs_service *svc)
1413 ip_vs_unlink_service(svc, false);
1420 * Flush all the virtual services
1422 static int ip_vs_flush(struct net *net, bool cleanup)
1425 struct ip_vs_service *svc;
1426 struct hlist_node *n;
1429 * Flush the service table hashed by <netns,protocol,addr,port>
1431 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1432 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1434 if (net_eq(svc->net, net))
1435 ip_vs_unlink_service(svc, cleanup);
1440 * Flush the service table hashed by fwmark
1442 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1443 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1445 if (net_eq(svc->net, net))
1446 ip_vs_unlink_service(svc, cleanup);
1454 * Delete service by {netns} in the service table.
1455 * Called by __ip_vs_cleanup()
1457 void ip_vs_service_net_cleanup(struct net *net)
1460 /* Check for "full" addressed entries */
1461 mutex_lock(&__ip_vs_mutex);
1462 ip_vs_flush(net, true);
1463 mutex_unlock(&__ip_vs_mutex);
1467 /* Put all references for device (dst_cache) */
1469 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1471 struct ip_vs_dest_dst *dest_dst;
1473 spin_lock_bh(&dest->dst_lock);
1474 dest_dst = rcu_dereference_protected(dest->dest_dst, 1);
1475 if (dest_dst && dest_dst->dst_cache->dev == dev) {
1476 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1478 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1480 atomic_read(&dest->refcnt));
1481 __ip_vs_dst_cache_reset(dest);
1483 spin_unlock_bh(&dest->dst_lock);
1486 /* Netdev event receiver
1487 * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1489 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1492 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1493 struct net *net = dev_net(dev);
1494 struct netns_ipvs *ipvs = net_ipvs(net);
1495 struct ip_vs_service *svc;
1496 struct ip_vs_dest *dest;
1499 if (event != NETDEV_DOWN || !ipvs)
1501 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1503 mutex_lock(&__ip_vs_mutex);
1504 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1505 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1506 if (net_eq(svc->net, net)) {
1507 list_for_each_entry(dest, &svc->destinations,
1509 ip_vs_forget_dev(dest, dev);
1514 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1515 if (net_eq(svc->net, net)) {
1516 list_for_each_entry(dest, &svc->destinations,
1518 ip_vs_forget_dev(dest, dev);
1525 spin_lock_bh(&ipvs->dest_trash_lock);
1526 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1527 ip_vs_forget_dev(dest, dev);
1529 spin_unlock_bh(&ipvs->dest_trash_lock);
1530 mutex_unlock(&__ip_vs_mutex);
1536 * Zero counters in a service or all services
1538 static int ip_vs_zero_service(struct ip_vs_service *svc)
1540 struct ip_vs_dest *dest;
1542 list_for_each_entry(dest, &svc->destinations, n_list) {
1543 ip_vs_zero_stats(&dest->stats);
1545 ip_vs_zero_stats(&svc->stats);
1549 static int ip_vs_zero_all(struct net *net)
1552 struct ip_vs_service *svc;
1554 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1555 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1556 if (net_eq(svc->net, net))
1557 ip_vs_zero_service(svc);
1561 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1562 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1563 if (net_eq(svc->net, net))
1564 ip_vs_zero_service(svc);
1568 ip_vs_zero_stats(&net_ipvs(net)->tot_stats);
1572 #ifdef CONFIG_SYSCTL
1575 static int three = 3;
1578 proc_do_defense_mode(struct ctl_table *table, int write,
1579 void __user *buffer, size_t *lenp, loff_t *ppos)
1581 struct net *net = current->nsproxy->net_ns;
1582 int *valp = table->data;
1586 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1587 if (write && (*valp != val)) {
1588 if ((*valp < 0) || (*valp > 3)) {
1589 /* Restore the correct value */
1592 update_defense_level(net_ipvs(net));
1599 proc_do_sync_threshold(struct ctl_table *table, int write,
1600 void __user *buffer, size_t *lenp, loff_t *ppos)
1602 int *valp = table->data;
1606 /* backup the value first */
1607 memcpy(val, valp, sizeof(val));
1609 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1610 if (write && (valp[0] < 0 || valp[1] < 0 ||
1611 (valp[0] >= valp[1] && valp[1]))) {
1612 /* Restore the correct value */
1613 memcpy(valp, val, sizeof(val));
1619 proc_do_sync_mode(struct ctl_table *table, int write,
1620 void __user *buffer, size_t *lenp, loff_t *ppos)
1622 int *valp = table->data;
1626 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1627 if (write && (*valp != val)) {
1628 if ((*valp < 0) || (*valp > 1)) {
1629 /* Restore the correct value */
1637 proc_do_sync_ports(struct ctl_table *table, int write,
1638 void __user *buffer, size_t *lenp, loff_t *ppos)
1640 int *valp = table->data;
1644 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1645 if (write && (*valp != val)) {
1646 if (*valp < 1 || !is_power_of_2(*valp)) {
1647 /* Restore the correct value */
1655 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1656 * Do not change order or insert new entries without
1657 * align with netns init in ip_vs_control_net_init()
1660 static struct ctl_table vs_vars[] = {
1662 .procname = "amemthresh",
1663 .maxlen = sizeof(int),
1665 .proc_handler = proc_dointvec,
1668 .procname = "am_droprate",
1669 .maxlen = sizeof(int),
1671 .proc_handler = proc_dointvec,
1674 .procname = "drop_entry",
1675 .maxlen = sizeof(int),
1677 .proc_handler = proc_do_defense_mode,
1680 .procname = "drop_packet",
1681 .maxlen = sizeof(int),
1683 .proc_handler = proc_do_defense_mode,
1685 #ifdef CONFIG_IP_VS_NFCT
1687 .procname = "conntrack",
1688 .maxlen = sizeof(int),
1690 .proc_handler = &proc_dointvec,
1694 .procname = "secure_tcp",
1695 .maxlen = sizeof(int),
1697 .proc_handler = proc_do_defense_mode,
1700 .procname = "snat_reroute",
1701 .maxlen = sizeof(int),
1703 .proc_handler = &proc_dointvec,
1706 .procname = "sync_version",
1707 .maxlen = sizeof(int),
1709 .proc_handler = &proc_do_sync_mode,
1712 .procname = "sync_ports",
1713 .maxlen = sizeof(int),
1715 .proc_handler = &proc_do_sync_ports,
1718 .procname = "sync_persist_mode",
1719 .maxlen = sizeof(int),
1721 .proc_handler = proc_dointvec,
1724 .procname = "sync_qlen_max",
1725 .maxlen = sizeof(unsigned long),
1727 .proc_handler = proc_doulongvec_minmax,
1730 .procname = "sync_sock_size",
1731 .maxlen = sizeof(int),
1733 .proc_handler = proc_dointvec,
1736 .procname = "cache_bypass",
1737 .maxlen = sizeof(int),
1739 .proc_handler = proc_dointvec,
1742 .procname = "expire_nodest_conn",
1743 .maxlen = sizeof(int),
1745 .proc_handler = proc_dointvec,
1748 .procname = "sloppy_tcp",
1749 .maxlen = sizeof(int),
1751 .proc_handler = proc_dointvec,
1754 .procname = "sloppy_sctp",
1755 .maxlen = sizeof(int),
1757 .proc_handler = proc_dointvec,
1760 .procname = "expire_quiescent_template",
1761 .maxlen = sizeof(int),
1763 .proc_handler = proc_dointvec,
1766 .procname = "sync_threshold",
1768 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1770 .proc_handler = proc_do_sync_threshold,
1773 .procname = "sync_refresh_period",
1774 .maxlen = sizeof(int),
1776 .proc_handler = proc_dointvec_jiffies,
1779 .procname = "sync_retries",
1780 .maxlen = sizeof(int),
1782 .proc_handler = proc_dointvec_minmax,
1787 .procname = "nat_icmp_send",
1788 .maxlen = sizeof(int),
1790 .proc_handler = proc_dointvec,
1793 .procname = "pmtu_disc",
1794 .maxlen = sizeof(int),
1796 .proc_handler = proc_dointvec,
1799 .procname = "backup_only",
1800 .maxlen = sizeof(int),
1802 .proc_handler = proc_dointvec,
1804 #ifdef CONFIG_IP_VS_DEBUG
1806 .procname = "debug_level",
1807 .data = &sysctl_ip_vs_debug_level,
1808 .maxlen = sizeof(int),
1810 .proc_handler = proc_dointvec,
1815 .procname = "timeout_established",
1816 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1817 .maxlen = sizeof(int),
1819 .proc_handler = proc_dointvec_jiffies,
1822 .procname = "timeout_synsent",
1823 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1824 .maxlen = sizeof(int),
1826 .proc_handler = proc_dointvec_jiffies,
1829 .procname = "timeout_synrecv",
1830 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1831 .maxlen = sizeof(int),
1833 .proc_handler = proc_dointvec_jiffies,
1836 .procname = "timeout_finwait",
1837 .data = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1838 .maxlen = sizeof(int),
1840 .proc_handler = proc_dointvec_jiffies,
1843 .procname = "timeout_timewait",
1844 .data = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1845 .maxlen = sizeof(int),
1847 .proc_handler = proc_dointvec_jiffies,
1850 .procname = "timeout_close",
1851 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1852 .maxlen = sizeof(int),
1854 .proc_handler = proc_dointvec_jiffies,
1857 .procname = "timeout_closewait",
1858 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1859 .maxlen = sizeof(int),
1861 .proc_handler = proc_dointvec_jiffies,
1864 .procname = "timeout_lastack",
1865 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1866 .maxlen = sizeof(int),
1868 .proc_handler = proc_dointvec_jiffies,
1871 .procname = "timeout_listen",
1872 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1873 .maxlen = sizeof(int),
1875 .proc_handler = proc_dointvec_jiffies,
1878 .procname = "timeout_synack",
1879 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1880 .maxlen = sizeof(int),
1882 .proc_handler = proc_dointvec_jiffies,
1885 .procname = "timeout_udp",
1886 .data = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1887 .maxlen = sizeof(int),
1889 .proc_handler = proc_dointvec_jiffies,
1892 .procname = "timeout_icmp",
1893 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1894 .maxlen = sizeof(int),
1896 .proc_handler = proc_dointvec_jiffies,
1904 #ifdef CONFIG_PROC_FS
1907 struct seq_net_private p; /* Do not move this, netns depends upon it*/
1908 struct hlist_head *table;
1913 * Write the contents of the VS rule table to a PROCfs file.
1914 * (It is kept just for backward compatibility)
1916 static inline const char *ip_vs_fwd_name(unsigned int flags)
1918 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1919 case IP_VS_CONN_F_LOCALNODE:
1921 case IP_VS_CONN_F_TUNNEL:
1923 case IP_VS_CONN_F_DROUTE:
1931 /* Get the Nth entry in the two lists */
1932 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1934 struct net *net = seq_file_net(seq);
1935 struct ip_vs_iter *iter = seq->private;
1937 struct ip_vs_service *svc;
1939 /* look in hash by protocol */
1940 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1941 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
1942 if (net_eq(svc->net, net) && pos-- == 0) {
1943 iter->table = ip_vs_svc_table;
1950 /* keep looking in fwmark */
1951 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1952 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
1954 if (net_eq(svc->net, net) && pos-- == 0) {
1955 iter->table = ip_vs_svc_fwm_table;
1965 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1969 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1973 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1975 struct hlist_node *e;
1976 struct ip_vs_iter *iter;
1977 struct ip_vs_service *svc;
1980 if (v == SEQ_START_TOKEN)
1981 return ip_vs_info_array(seq,0);
1984 iter = seq->private;
1986 if (iter->table == ip_vs_svc_table) {
1987 /* next service in table hashed by protocol */
1988 e = rcu_dereference(hlist_next_rcu(&svc->s_list));
1990 return hlist_entry(e, struct ip_vs_service, s_list);
1992 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1993 hlist_for_each_entry_rcu(svc,
1994 &ip_vs_svc_table[iter->bucket],
2000 iter->table = ip_vs_svc_fwm_table;
2005 /* next service in hashed by fwmark */
2006 e = rcu_dereference(hlist_next_rcu(&svc->f_list));
2008 return hlist_entry(e, struct ip_vs_service, f_list);
2011 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2012 hlist_for_each_entry_rcu(svc,
2013 &ip_vs_svc_fwm_table[iter->bucket],
2021 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
2028 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
2030 if (v == SEQ_START_TOKEN) {
2032 "IP Virtual Server version %d.%d.%d (size=%d)\n",
2033 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2035 "Prot LocalAddress:Port Scheduler Flags\n");
2037 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2039 const struct ip_vs_service *svc = v;
2040 const struct ip_vs_iter *iter = seq->private;
2041 const struct ip_vs_dest *dest;
2042 struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler);
2044 if (iter->table == ip_vs_svc_table) {
2045 #ifdef CONFIG_IP_VS_IPV6
2046 if (svc->af == AF_INET6)
2047 seq_printf(seq, "%s [%pI6]:%04X %s ",
2048 ip_vs_proto_name(svc->protocol),
2054 seq_printf(seq, "%s %08X:%04X %s %s ",
2055 ip_vs_proto_name(svc->protocol),
2056 ntohl(svc->addr.ip),
2059 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2061 seq_printf(seq, "FWM %08X %s %s",
2062 svc->fwmark, sched->name,
2063 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2066 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2067 seq_printf(seq, "persistent %d %08X\n",
2069 ntohl(svc->netmask));
2071 seq_putc(seq, '\n');
2073 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
2074 #ifdef CONFIG_IP_VS_IPV6
2075 if (dest->af == AF_INET6)
2078 " %-7s %-6d %-10d %-10d\n",
2081 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2082 atomic_read(&dest->weight),
2083 atomic_read(&dest->activeconns),
2084 atomic_read(&dest->inactconns));
2089 "%-7s %-6d %-10d %-10d\n",
2090 ntohl(dest->addr.ip),
2092 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2093 atomic_read(&dest->weight),
2094 atomic_read(&dest->activeconns),
2095 atomic_read(&dest->inactconns));
2102 static const struct seq_operations ip_vs_info_seq_ops = {
2103 .start = ip_vs_info_seq_start,
2104 .next = ip_vs_info_seq_next,
2105 .stop = ip_vs_info_seq_stop,
2106 .show = ip_vs_info_seq_show,
2109 static int ip_vs_info_open(struct inode *inode, struct file *file)
2111 return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2112 sizeof(struct ip_vs_iter));
2115 static const struct file_operations ip_vs_info_fops = {
2116 .owner = THIS_MODULE,
2117 .open = ip_vs_info_open,
2119 .llseek = seq_lseek,
2120 .release = seq_release_net,
2123 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2125 struct net *net = seq_file_single_net(seq);
2126 struct ip_vs_stats_user show;
2128 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2130 " Total Incoming Outgoing Incoming Outgoing\n");
2132 " Conns Packets Packets Bytes Bytes\n");
2134 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2135 seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", show.conns,
2136 show.inpkts, show.outpkts,
2137 (unsigned long long) show.inbytes,
2138 (unsigned long long) show.outbytes);
2140 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2142 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2143 seq_printf(seq, "%8X %8X %8X %16X %16X\n",
2144 show.cps, show.inpps, show.outpps,
2145 show.inbps, show.outbps);
2150 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2152 return single_open_net(inode, file, ip_vs_stats_show);
2155 static const struct file_operations ip_vs_stats_fops = {
2156 .owner = THIS_MODULE,
2157 .open = ip_vs_stats_seq_open,
2159 .llseek = seq_lseek,
2160 .release = single_release_net,
2163 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2165 struct net *net = seq_file_single_net(seq);
2166 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2167 struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats;
2168 struct ip_vs_stats_user rates;
2171 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2173 " Total Incoming Outgoing Incoming Outgoing\n");
2175 "CPU Conns Packets Packets Bytes Bytes\n");
2177 for_each_possible_cpu(i) {
2178 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2180 __u64 inbytes, outbytes;
2183 start = u64_stats_fetch_begin_bh(&u->syncp);
2184 inbytes = u->ustats.inbytes;
2185 outbytes = u->ustats.outbytes;
2186 } while (u64_stats_fetch_retry_bh(&u->syncp, start));
2188 seq_printf(seq, "%3X %8X %8X %8X %16LX %16LX\n",
2189 i, u->ustats.conns, u->ustats.inpkts,
2190 u->ustats.outpkts, (__u64)inbytes,
2194 spin_lock_bh(&tot_stats->lock);
2196 seq_printf(seq, " ~ %8X %8X %8X %16LX %16LX\n\n",
2197 tot_stats->ustats.conns, tot_stats->ustats.inpkts,
2198 tot_stats->ustats.outpkts,
2199 (unsigned long long) tot_stats->ustats.inbytes,
2200 (unsigned long long) tot_stats->ustats.outbytes);
2202 ip_vs_read_estimator(&rates, tot_stats);
2204 spin_unlock_bh(&tot_stats->lock);
2206 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2208 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2209 seq_printf(seq, " %8X %8X %8X %16X %16X\n",
2219 static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2221 return single_open_net(inode, file, ip_vs_stats_percpu_show);
2224 static const struct file_operations ip_vs_stats_percpu_fops = {
2225 .owner = THIS_MODULE,
2226 .open = ip_vs_stats_percpu_seq_open,
2228 .llseek = seq_lseek,
2229 .release = single_release_net,
2234 * Set timeout values for tcp tcpfin udp in the timeout_table.
2236 static int ip_vs_set_timeout(struct net *net, struct ip_vs_timeout_user *u)
2238 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2239 struct ip_vs_proto_data *pd;
2242 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2247 #ifdef CONFIG_IP_VS_PROTO_TCP
2248 if (u->tcp_timeout) {
2249 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2250 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2251 = u->tcp_timeout * HZ;
2254 if (u->tcp_fin_timeout) {
2255 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2256 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2257 = u->tcp_fin_timeout * HZ;
2261 #ifdef CONFIG_IP_VS_PROTO_UDP
2262 if (u->udp_timeout) {
2263 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2264 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2265 = u->udp_timeout * HZ;
2272 #define SET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2273 #define SERVICE_ARG_LEN (sizeof(struct ip_vs_service_user))
2274 #define SVCDEST_ARG_LEN (sizeof(struct ip_vs_service_user) + \
2275 sizeof(struct ip_vs_dest_user))
2276 #define TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2277 #define DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user))
2278 #define MAX_ARG_LEN SVCDEST_ARG_LEN
2280 static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
2281 [SET_CMDID(IP_VS_SO_SET_ADD)] = SERVICE_ARG_LEN,
2282 [SET_CMDID(IP_VS_SO_SET_EDIT)] = SERVICE_ARG_LEN,
2283 [SET_CMDID(IP_VS_SO_SET_DEL)] = SERVICE_ARG_LEN,
2284 [SET_CMDID(IP_VS_SO_SET_FLUSH)] = 0,
2285 [SET_CMDID(IP_VS_SO_SET_ADDDEST)] = SVCDEST_ARG_LEN,
2286 [SET_CMDID(IP_VS_SO_SET_DELDEST)] = SVCDEST_ARG_LEN,
2287 [SET_CMDID(IP_VS_SO_SET_EDITDEST)] = SVCDEST_ARG_LEN,
2288 [SET_CMDID(IP_VS_SO_SET_TIMEOUT)] = TIMEOUT_ARG_LEN,
2289 [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)] = DAEMON_ARG_LEN,
2290 [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)] = DAEMON_ARG_LEN,
2291 [SET_CMDID(IP_VS_SO_SET_ZERO)] = SERVICE_ARG_LEN,
2294 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2295 struct ip_vs_service_user *usvc_compat)
2297 memset(usvc, 0, sizeof(*usvc));
2300 usvc->protocol = usvc_compat->protocol;
2301 usvc->addr.ip = usvc_compat->addr;
2302 usvc->port = usvc_compat->port;
2303 usvc->fwmark = usvc_compat->fwmark;
2305 /* Deep copy of sched_name is not needed here */
2306 usvc->sched_name = usvc_compat->sched_name;
2308 usvc->flags = usvc_compat->flags;
2309 usvc->timeout = usvc_compat->timeout;
2310 usvc->netmask = usvc_compat->netmask;
2313 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2314 struct ip_vs_dest_user *udest_compat)
2316 memset(udest, 0, sizeof(*udest));
2318 udest->addr.ip = udest_compat->addr;
2319 udest->port = udest_compat->port;
2320 udest->conn_flags = udest_compat->conn_flags;
2321 udest->weight = udest_compat->weight;
2322 udest->u_threshold = udest_compat->u_threshold;
2323 udest->l_threshold = udest_compat->l_threshold;
2327 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2329 struct net *net = sock_net(sk);
2331 unsigned char arg[MAX_ARG_LEN];
2332 struct ip_vs_service_user *usvc_compat;
2333 struct ip_vs_service_user_kern usvc;
2334 struct ip_vs_service *svc;
2335 struct ip_vs_dest_user *udest_compat;
2336 struct ip_vs_dest_user_kern udest;
2337 struct netns_ipvs *ipvs = net_ipvs(net);
2339 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2342 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2344 if (len < 0 || len > MAX_ARG_LEN)
2346 if (len != set_arglen[SET_CMDID(cmd)]) {
2347 pr_err("set_ctl: len %u != %u\n",
2348 len, set_arglen[SET_CMDID(cmd)]);
2352 if (copy_from_user(arg, user, len) != 0)
2355 /* increase the module use count */
2356 ip_vs_use_count_inc();
2358 /* Handle daemons since they have another lock */
2359 if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2360 cmd == IP_VS_SO_SET_STOPDAEMON) {
2361 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2363 if (mutex_lock_interruptible(&ipvs->sync_mutex)) {
2367 if (cmd == IP_VS_SO_SET_STARTDAEMON)
2368 ret = start_sync_thread(net, dm->state, dm->mcast_ifn,
2371 ret = stop_sync_thread(net, dm->state);
2372 mutex_unlock(&ipvs->sync_mutex);
2376 if (mutex_lock_interruptible(&__ip_vs_mutex)) {
2381 if (cmd == IP_VS_SO_SET_FLUSH) {
2382 /* Flush the virtual service */
2383 ret = ip_vs_flush(net, false);
2385 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2386 /* Set timeout values for (tcp tcpfin udp) */
2387 ret = ip_vs_set_timeout(net, (struct ip_vs_timeout_user *)arg);
2391 usvc_compat = (struct ip_vs_service_user *)arg;
2392 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2394 /* We only use the new structs internally, so copy userspace compat
2395 * structs to extended internal versions */
2396 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2397 ip_vs_copy_udest_compat(&udest, udest_compat);
2399 if (cmd == IP_VS_SO_SET_ZERO) {
2400 /* if no service address is set, zero counters in all */
2401 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2402 ret = ip_vs_zero_all(net);
2407 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2408 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2409 usvc.protocol != IPPROTO_SCTP) {
2410 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2411 usvc.protocol, &usvc.addr.ip,
2412 ntohs(usvc.port), usvc.sched_name);
2417 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2419 if (usvc.fwmark == 0)
2420 svc = __ip_vs_service_find(net, usvc.af, usvc.protocol,
2421 &usvc.addr, usvc.port);
2423 svc = __ip_vs_svc_fwm_find(net, usvc.af, usvc.fwmark);
2426 if (cmd != IP_VS_SO_SET_ADD
2427 && (svc == NULL || svc->protocol != usvc.protocol)) {
2433 case IP_VS_SO_SET_ADD:
2437 ret = ip_vs_add_service(net, &usvc, &svc);
2439 case IP_VS_SO_SET_EDIT:
2440 ret = ip_vs_edit_service(svc, &usvc);
2442 case IP_VS_SO_SET_DEL:
2443 ret = ip_vs_del_service(svc);
2447 case IP_VS_SO_SET_ZERO:
2448 ret = ip_vs_zero_service(svc);
2450 case IP_VS_SO_SET_ADDDEST:
2451 ret = ip_vs_add_dest(svc, &udest);
2453 case IP_VS_SO_SET_EDITDEST:
2454 ret = ip_vs_edit_dest(svc, &udest);
2456 case IP_VS_SO_SET_DELDEST:
2457 ret = ip_vs_del_dest(svc, &udest);
2464 mutex_unlock(&__ip_vs_mutex);
2466 /* decrease the module use count */
2467 ip_vs_use_count_dec();
2474 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2476 struct ip_vs_scheduler *sched;
2478 sched = rcu_dereference_protected(src->scheduler, 1);
2479 dst->protocol = src->protocol;
2480 dst->addr = src->addr.ip;
2481 dst->port = src->port;
2482 dst->fwmark = src->fwmark;
2483 strlcpy(dst->sched_name, sched->name, sizeof(dst->sched_name));
2484 dst->flags = src->flags;
2485 dst->timeout = src->timeout / HZ;
2486 dst->netmask = src->netmask;
2487 dst->num_dests = src->num_dests;
2488 ip_vs_copy_stats(&dst->stats, &src->stats);
2492 __ip_vs_get_service_entries(struct net *net,
2493 const struct ip_vs_get_services *get,
2494 struct ip_vs_get_services __user *uptr)
2497 struct ip_vs_service *svc;
2498 struct ip_vs_service_entry entry;
2501 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2502 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2503 /* Only expose IPv4 entries to old interface */
2504 if (svc->af != AF_INET || !net_eq(svc->net, net))
2507 if (count >= get->num_services)
2509 memset(&entry, 0, sizeof(entry));
2510 ip_vs_copy_service(&entry, svc);
2511 if (copy_to_user(&uptr->entrytable[count],
2512 &entry, sizeof(entry))) {
2520 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2521 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2522 /* Only expose IPv4 entries to old interface */
2523 if (svc->af != AF_INET || !net_eq(svc->net, net))
2526 if (count >= get->num_services)
2528 memset(&entry, 0, sizeof(entry));
2529 ip_vs_copy_service(&entry, svc);
2530 if (copy_to_user(&uptr->entrytable[count],
2531 &entry, sizeof(entry))) {
2543 __ip_vs_get_dest_entries(struct net *net, const struct ip_vs_get_dests *get,
2544 struct ip_vs_get_dests __user *uptr)
2546 struct ip_vs_service *svc;
2547 union nf_inet_addr addr = { .ip = get->addr };
2552 svc = __ip_vs_svc_fwm_find(net, AF_INET, get->fwmark);
2554 svc = __ip_vs_service_find(net, AF_INET, get->protocol, &addr,
2560 struct ip_vs_dest *dest;
2561 struct ip_vs_dest_entry entry;
2563 memset(&entry, 0, sizeof(entry));
2564 list_for_each_entry(dest, &svc->destinations, n_list) {
2565 if (count >= get->num_dests)
2568 entry.addr = dest->addr.ip;
2569 entry.port = dest->port;
2570 entry.conn_flags = atomic_read(&dest->conn_flags);
2571 entry.weight = atomic_read(&dest->weight);
2572 entry.u_threshold = dest->u_threshold;
2573 entry.l_threshold = dest->l_threshold;
2574 entry.activeconns = atomic_read(&dest->activeconns);
2575 entry.inactconns = atomic_read(&dest->inactconns);
2576 entry.persistconns = atomic_read(&dest->persistconns);
2577 ip_vs_copy_stats(&entry.stats, &dest->stats);
2578 if (copy_to_user(&uptr->entrytable[count],
2579 &entry, sizeof(entry))) {
2591 __ip_vs_get_timeouts(struct net *net, struct ip_vs_timeout_user *u)
2593 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2594 struct ip_vs_proto_data *pd;
2597 memset(u, 0, sizeof (*u));
2599 #ifdef CONFIG_IP_VS_PROTO_TCP
2600 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2601 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2602 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2604 #ifdef CONFIG_IP_VS_PROTO_UDP
2605 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2607 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2612 #define GET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2613 #define GET_INFO_ARG_LEN (sizeof(struct ip_vs_getinfo))
2614 #define GET_SERVICES_ARG_LEN (sizeof(struct ip_vs_get_services))
2615 #define GET_SERVICE_ARG_LEN (sizeof(struct ip_vs_service_entry))
2616 #define GET_DESTS_ARG_LEN (sizeof(struct ip_vs_get_dests))
2617 #define GET_TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2618 #define GET_DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user) * 2)
2620 static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
2621 [GET_CMDID(IP_VS_SO_GET_VERSION)] = 64,
2622 [GET_CMDID(IP_VS_SO_GET_INFO)] = GET_INFO_ARG_LEN,
2623 [GET_CMDID(IP_VS_SO_GET_SERVICES)] = GET_SERVICES_ARG_LEN,
2624 [GET_CMDID(IP_VS_SO_GET_SERVICE)] = GET_SERVICE_ARG_LEN,
2625 [GET_CMDID(IP_VS_SO_GET_DESTS)] = GET_DESTS_ARG_LEN,
2626 [GET_CMDID(IP_VS_SO_GET_TIMEOUT)] = GET_TIMEOUT_ARG_LEN,
2627 [GET_CMDID(IP_VS_SO_GET_DAEMON)] = GET_DAEMON_ARG_LEN,
2631 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2633 unsigned char arg[128];
2635 unsigned int copylen;
2636 struct net *net = sock_net(sk);
2637 struct netns_ipvs *ipvs = net_ipvs(net);
2640 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2643 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2646 if (*len < get_arglen[GET_CMDID(cmd)]) {
2647 pr_err("get_ctl: len %u < %u\n",
2648 *len, get_arglen[GET_CMDID(cmd)]);
2652 copylen = get_arglen[GET_CMDID(cmd)];
2656 if (copy_from_user(arg, user, copylen) != 0)
2659 * Handle daemons first since it has its own locking
2661 if (cmd == IP_VS_SO_GET_DAEMON) {
2662 struct ip_vs_daemon_user d[2];
2664 memset(&d, 0, sizeof(d));
2665 if (mutex_lock_interruptible(&ipvs->sync_mutex))
2666 return -ERESTARTSYS;
2668 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2669 d[0].state = IP_VS_STATE_MASTER;
2670 strlcpy(d[0].mcast_ifn, ipvs->master_mcast_ifn,
2671 sizeof(d[0].mcast_ifn));
2672 d[0].syncid = ipvs->master_syncid;
2674 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2675 d[1].state = IP_VS_STATE_BACKUP;
2676 strlcpy(d[1].mcast_ifn, ipvs->backup_mcast_ifn,
2677 sizeof(d[1].mcast_ifn));
2678 d[1].syncid = ipvs->backup_syncid;
2680 if (copy_to_user(user, &d, sizeof(d)) != 0)
2682 mutex_unlock(&ipvs->sync_mutex);
2686 if (mutex_lock_interruptible(&__ip_vs_mutex))
2687 return -ERESTARTSYS;
2690 case IP_VS_SO_GET_VERSION:
2694 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2695 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2696 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2700 *len = strlen(buf)+1;
2704 case IP_VS_SO_GET_INFO:
2706 struct ip_vs_getinfo info;
2707 info.version = IP_VS_VERSION_CODE;
2708 info.size = ip_vs_conn_tab_size;
2709 info.num_services = ipvs->num_services;
2710 if (copy_to_user(user, &info, sizeof(info)) != 0)
2715 case IP_VS_SO_GET_SERVICES:
2717 struct ip_vs_get_services *get;
2720 get = (struct ip_vs_get_services *)arg;
2721 size = sizeof(*get) +
2722 sizeof(struct ip_vs_service_entry) * get->num_services;
2724 pr_err("length: %u != %u\n", *len, size);
2728 ret = __ip_vs_get_service_entries(net, get, user);
2732 case IP_VS_SO_GET_SERVICE:
2734 struct ip_vs_service_entry *entry;
2735 struct ip_vs_service *svc;
2736 union nf_inet_addr addr;
2738 entry = (struct ip_vs_service_entry *)arg;
2739 addr.ip = entry->addr;
2742 svc = __ip_vs_svc_fwm_find(net, AF_INET, entry->fwmark);
2744 svc = __ip_vs_service_find(net, AF_INET,
2745 entry->protocol, &addr,
2749 ip_vs_copy_service(entry, svc);
2750 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2757 case IP_VS_SO_GET_DESTS:
2759 struct ip_vs_get_dests *get;
2762 get = (struct ip_vs_get_dests *)arg;
2763 size = sizeof(*get) +
2764 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2766 pr_err("length: %u != %u\n", *len, size);
2770 ret = __ip_vs_get_dest_entries(net, get, user);
2774 case IP_VS_SO_GET_TIMEOUT:
2776 struct ip_vs_timeout_user t;
2778 __ip_vs_get_timeouts(net, &t);
2779 if (copy_to_user(user, &t, sizeof(t)) != 0)
2789 mutex_unlock(&__ip_vs_mutex);
2794 static struct nf_sockopt_ops ip_vs_sockopts = {
2796 .set_optmin = IP_VS_BASE_CTL,
2797 .set_optmax = IP_VS_SO_SET_MAX+1,
2798 .set = do_ip_vs_set_ctl,
2799 .get_optmin = IP_VS_BASE_CTL,
2800 .get_optmax = IP_VS_SO_GET_MAX+1,
2801 .get = do_ip_vs_get_ctl,
2802 .owner = THIS_MODULE,
2806 * Generic Netlink interface
2809 /* IPVS genetlink family */
2810 static struct genl_family ip_vs_genl_family = {
2811 .id = GENL_ID_GENERATE,
2813 .name = IPVS_GENL_NAME,
2814 .version = IPVS_GENL_VERSION,
2815 .maxattr = IPVS_CMD_MAX,
2816 .netnsok = true, /* Make ipvsadm to work on netns */
2819 /* Policy used for first-level command attributes */
2820 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2821 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2822 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2823 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2824 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2825 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2826 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2829 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2830 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2831 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2832 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2833 .len = IP_VS_IFNAME_MAXLEN },
2834 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2837 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2838 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2839 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2840 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2841 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2842 .len = sizeof(union nf_inet_addr) },
2843 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2844 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2845 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2846 .len = IP_VS_SCHEDNAME_MAXLEN },
2847 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2848 .len = IP_VS_PENAME_MAXLEN },
2849 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2850 .len = sizeof(struct ip_vs_flags) },
2851 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2852 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2853 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2856 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2857 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2858 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2859 .len = sizeof(union nf_inet_addr) },
2860 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2861 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2862 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2863 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2864 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2865 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2866 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2867 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2868 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2871 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2872 struct ip_vs_stats *stats)
2874 struct ip_vs_stats_user ustats;
2875 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2879 ip_vs_copy_stats(&ustats, stats);
2881 if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, ustats.conns) ||
2882 nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, ustats.inpkts) ||
2883 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, ustats.outpkts) ||
2884 nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, ustats.inbytes) ||
2885 nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, ustats.outbytes) ||
2886 nla_put_u32(skb, IPVS_STATS_ATTR_CPS, ustats.cps) ||
2887 nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, ustats.inpps) ||
2888 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, ustats.outpps) ||
2889 nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, ustats.inbps) ||
2890 nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, ustats.outbps))
2891 goto nla_put_failure;
2892 nla_nest_end(skb, nl_stats);
2897 nla_nest_cancel(skb, nl_stats);
2901 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2902 struct ip_vs_service *svc)
2904 struct ip_vs_scheduler *sched;
2905 struct ip_vs_pe *pe;
2906 struct nlattr *nl_service;
2907 struct ip_vs_flags flags = { .flags = svc->flags,
2910 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2914 if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2915 goto nla_put_failure;
2917 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2918 goto nla_put_failure;
2920 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2921 nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2922 nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2923 goto nla_put_failure;
2926 sched = rcu_dereference_protected(svc->scheduler, 1);
2927 pe = rcu_dereference_protected(svc->pe, 1);
2928 if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched->name) ||
2929 (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
2930 nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2931 nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2932 nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
2933 goto nla_put_failure;
2934 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2935 goto nla_put_failure;
2937 nla_nest_end(skb, nl_service);
2942 nla_nest_cancel(skb, nl_service);
2946 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2947 struct ip_vs_service *svc,
2948 struct netlink_callback *cb)
2952 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2953 &ip_vs_genl_family, NLM_F_MULTI,
2954 IPVS_CMD_NEW_SERVICE);
2958 if (ip_vs_genl_fill_service(skb, svc) < 0)
2959 goto nla_put_failure;
2961 return genlmsg_end(skb, hdr);
2964 genlmsg_cancel(skb, hdr);
2968 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2969 struct netlink_callback *cb)
2972 int start = cb->args[0];
2973 struct ip_vs_service *svc;
2974 struct net *net = skb_sknet(skb);
2976 mutex_lock(&__ip_vs_mutex);
2977 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2978 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2979 if (++idx <= start || !net_eq(svc->net, net))
2981 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2983 goto nla_put_failure;
2988 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2989 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
2990 if (++idx <= start || !net_eq(svc->net, net))
2992 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2994 goto nla_put_failure;
3000 mutex_unlock(&__ip_vs_mutex);
3006 static int ip_vs_genl_parse_service(struct net *net,
3007 struct ip_vs_service_user_kern *usvc,
3008 struct nlattr *nla, int full_entry,
3009 struct ip_vs_service **ret_svc)
3011 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
3012 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
3013 struct ip_vs_service *svc;
3015 /* Parse mandatory identifying service fields first */
3017 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
3020 nla_af = attrs[IPVS_SVC_ATTR_AF];
3021 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
3022 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
3023 nla_port = attrs[IPVS_SVC_ATTR_PORT];
3024 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
3026 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3029 memset(usvc, 0, sizeof(*usvc));
3031 usvc->af = nla_get_u16(nla_af);
3032 #ifdef CONFIG_IP_VS_IPV6
3033 if (usvc->af != AF_INET && usvc->af != AF_INET6)
3035 if (usvc->af != AF_INET)
3037 return -EAFNOSUPPORT;
3040 usvc->protocol = IPPROTO_TCP;
3041 usvc->fwmark = nla_get_u32(nla_fwmark);
3043 usvc->protocol = nla_get_u16(nla_protocol);
3044 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3045 usvc->port = nla_get_be16(nla_port);
3051 svc = __ip_vs_svc_fwm_find(net, usvc->af, usvc->fwmark);
3053 svc = __ip_vs_service_find(net, usvc->af, usvc->protocol,
3054 &usvc->addr, usvc->port);
3058 /* If a full entry was requested, check for the additional fields */
3060 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3062 struct ip_vs_flags flags;
3064 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3065 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3066 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3067 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3068 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3070 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3073 nla_memcpy(&flags, nla_flags, sizeof(flags));
3075 /* prefill flags from service if it already exists */
3077 usvc->flags = svc->flags;
3079 /* set new flags from userland */
3080 usvc->flags = (usvc->flags & ~flags.mask) |
3081 (flags.flags & flags.mask);
3082 usvc->sched_name = nla_data(nla_sched);
3083 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3084 usvc->timeout = nla_get_u32(nla_timeout);
3085 usvc->netmask = nla_get_be32(nla_netmask);
3091 static struct ip_vs_service *ip_vs_genl_find_service(struct net *net,
3094 struct ip_vs_service_user_kern usvc;
3095 struct ip_vs_service *svc;
3098 ret = ip_vs_genl_parse_service(net, &usvc, nla, 0, &svc);
3099 return ret ? ERR_PTR(ret) : svc;
3102 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3104 struct nlattr *nl_dest;
3106 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3110 if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3111 nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3112 nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3113 (atomic_read(&dest->conn_flags) &
3114 IP_VS_CONN_F_FWD_MASK)) ||
3115 nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3116 atomic_read(&dest->weight)) ||
3117 nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3118 nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3119 nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3120 atomic_read(&dest->activeconns)) ||
3121 nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3122 atomic_read(&dest->inactconns)) ||
3123 nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3124 atomic_read(&dest->persistconns)))
3125 goto nla_put_failure;
3126 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
3127 goto nla_put_failure;
3129 nla_nest_end(skb, nl_dest);
3134 nla_nest_cancel(skb, nl_dest);
3138 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3139 struct netlink_callback *cb)
3143 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3144 &ip_vs_genl_family, NLM_F_MULTI,
3149 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3150 goto nla_put_failure;
3152 return genlmsg_end(skb, hdr);
3155 genlmsg_cancel(skb, hdr);
3159 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3160 struct netlink_callback *cb)
3163 int start = cb->args[0];
3164 struct ip_vs_service *svc;
3165 struct ip_vs_dest *dest;
3166 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3167 struct net *net = skb_sknet(skb);
3169 mutex_lock(&__ip_vs_mutex);
3171 /* Try to find the service for which to dump destinations */
3172 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3173 IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3177 svc = ip_vs_genl_find_service(net, attrs[IPVS_CMD_ATTR_SERVICE]);
3178 if (IS_ERR(svc) || svc == NULL)
3181 /* Dump the destinations */
3182 list_for_each_entry(dest, &svc->destinations, n_list) {
3185 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3187 goto nla_put_failure;
3195 mutex_unlock(&__ip_vs_mutex);
3200 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3201 struct nlattr *nla, int full_entry)
3203 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3204 struct nlattr *nla_addr, *nla_port;
3206 /* Parse mandatory identifying destination fields first */
3208 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3211 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3212 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3214 if (!(nla_addr && nla_port))
3217 memset(udest, 0, sizeof(*udest));
3219 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3220 udest->port = nla_get_be16(nla_port);
3222 /* If a full entry was requested, check for the additional fields */
3224 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3227 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3228 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3229 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3230 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3232 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3235 udest->conn_flags = nla_get_u32(nla_fwd)
3236 & IP_VS_CONN_F_FWD_MASK;
3237 udest->weight = nla_get_u32(nla_weight);
3238 udest->u_threshold = nla_get_u32(nla_u_thresh);
3239 udest->l_threshold = nla_get_u32(nla_l_thresh);
3245 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3246 const char *mcast_ifn, __u32 syncid)
3248 struct nlattr *nl_daemon;
3250 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3254 if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3255 nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn) ||
3256 nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid))
3257 goto nla_put_failure;
3258 nla_nest_end(skb, nl_daemon);
3263 nla_nest_cancel(skb, nl_daemon);
3267 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3268 const char *mcast_ifn, __u32 syncid,
3269 struct netlink_callback *cb)
3272 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3273 &ip_vs_genl_family, NLM_F_MULTI,
3274 IPVS_CMD_NEW_DAEMON);
3278 if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
3279 goto nla_put_failure;
3281 return genlmsg_end(skb, hdr);
3284 genlmsg_cancel(skb, hdr);
3288 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3289 struct netlink_callback *cb)
3291 struct net *net = skb_sknet(skb);
3292 struct netns_ipvs *ipvs = net_ipvs(net);
3294 mutex_lock(&ipvs->sync_mutex);
3295 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3296 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3297 ipvs->master_mcast_ifn,
3298 ipvs->master_syncid, cb) < 0)
3299 goto nla_put_failure;
3304 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3305 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3306 ipvs->backup_mcast_ifn,
3307 ipvs->backup_syncid, cb) < 0)
3308 goto nla_put_failure;
3314 mutex_unlock(&ipvs->sync_mutex);
3319 static int ip_vs_genl_new_daemon(struct net *net, struct nlattr **attrs)
3321 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3322 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3323 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3326 return start_sync_thread(net,
3327 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3328 nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3329 nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3332 static int ip_vs_genl_del_daemon(struct net *net, struct nlattr **attrs)
3334 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3337 return stop_sync_thread(net,
3338 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3341 static int ip_vs_genl_set_config(struct net *net, struct nlattr **attrs)
3343 struct ip_vs_timeout_user t;
3345 __ip_vs_get_timeouts(net, &t);
3347 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3348 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3350 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3352 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3354 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3355 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3357 return ip_vs_set_timeout(net, &t);
3360 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3364 struct netns_ipvs *ipvs;
3366 net = skb_sknet(skb);
3367 ipvs = net_ipvs(net);
3368 cmd = info->genlhdr->cmd;
3370 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3371 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3373 mutex_lock(&ipvs->sync_mutex);
3374 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3375 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3376 info->attrs[IPVS_CMD_ATTR_DAEMON],
3377 ip_vs_daemon_policy)) {
3382 if (cmd == IPVS_CMD_NEW_DAEMON)
3383 ret = ip_vs_genl_new_daemon(net, daemon_attrs);
3385 ret = ip_vs_genl_del_daemon(net, daemon_attrs);
3387 mutex_unlock(&ipvs->sync_mutex);
3392 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3394 struct ip_vs_service *svc = NULL;
3395 struct ip_vs_service_user_kern usvc;
3396 struct ip_vs_dest_user_kern udest;
3398 int need_full_svc = 0, need_full_dest = 0;
3401 net = skb_sknet(skb);
3402 cmd = info->genlhdr->cmd;
3404 mutex_lock(&__ip_vs_mutex);
3406 if (cmd == IPVS_CMD_FLUSH) {
3407 ret = ip_vs_flush(net, false);
3409 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3410 ret = ip_vs_genl_set_config(net, info->attrs);
3412 } else if (cmd == IPVS_CMD_ZERO &&
3413 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3414 ret = ip_vs_zero_all(net);
3418 /* All following commands require a service argument, so check if we
3419 * received a valid one. We need a full service specification when
3420 * adding / editing a service. Only identifying members otherwise. */
3421 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3424 ret = ip_vs_genl_parse_service(net, &usvc,
3425 info->attrs[IPVS_CMD_ATTR_SERVICE],
3426 need_full_svc, &svc);
3430 /* Unless we're adding a new service, the service must already exist */
3431 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3436 /* Destination commands require a valid destination argument. For
3437 * adding / editing a destination, we need a full destination
3439 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3440 cmd == IPVS_CMD_DEL_DEST) {
3441 if (cmd != IPVS_CMD_DEL_DEST)
3444 ret = ip_vs_genl_parse_dest(&udest,
3445 info->attrs[IPVS_CMD_ATTR_DEST],
3452 case IPVS_CMD_NEW_SERVICE:
3454 ret = ip_vs_add_service(net, &usvc, &svc);
3458 case IPVS_CMD_SET_SERVICE:
3459 ret = ip_vs_edit_service(svc, &usvc);
3461 case IPVS_CMD_DEL_SERVICE:
3462 ret = ip_vs_del_service(svc);
3463 /* do not use svc, it can be freed */
3465 case IPVS_CMD_NEW_DEST:
3466 ret = ip_vs_add_dest(svc, &udest);
3468 case IPVS_CMD_SET_DEST:
3469 ret = ip_vs_edit_dest(svc, &udest);
3471 case IPVS_CMD_DEL_DEST:
3472 ret = ip_vs_del_dest(svc, &udest);
3475 ret = ip_vs_zero_service(svc);
3482 mutex_unlock(&__ip_vs_mutex);
3487 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3489 struct sk_buff *msg;
3491 int ret, cmd, reply_cmd;
3494 net = skb_sknet(skb);
3495 cmd = info->genlhdr->cmd;
3497 if (cmd == IPVS_CMD_GET_SERVICE)
3498 reply_cmd = IPVS_CMD_NEW_SERVICE;
3499 else if (cmd == IPVS_CMD_GET_INFO)
3500 reply_cmd = IPVS_CMD_SET_INFO;
3501 else if (cmd == IPVS_CMD_GET_CONFIG)
3502 reply_cmd = IPVS_CMD_SET_CONFIG;
3504 pr_err("unknown Generic Netlink command\n");
3508 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3512 mutex_lock(&__ip_vs_mutex);
3514 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3516 goto nla_put_failure;
3519 case IPVS_CMD_GET_SERVICE:
3521 struct ip_vs_service *svc;
3523 svc = ip_vs_genl_find_service(net,
3524 info->attrs[IPVS_CMD_ATTR_SERVICE]);
3529 ret = ip_vs_genl_fill_service(msg, svc);
3531 goto nla_put_failure;
3540 case IPVS_CMD_GET_CONFIG:
3542 struct ip_vs_timeout_user t;
3544 __ip_vs_get_timeouts(net, &t);
3545 #ifdef CONFIG_IP_VS_PROTO_TCP
3546 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3548 nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3550 goto nla_put_failure;
3552 #ifdef CONFIG_IP_VS_PROTO_UDP
3553 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3554 goto nla_put_failure;
3560 case IPVS_CMD_GET_INFO:
3561 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3562 IP_VS_VERSION_CODE) ||
3563 nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3564 ip_vs_conn_tab_size))
3565 goto nla_put_failure;
3569 genlmsg_end(msg, reply);
3570 ret = genlmsg_reply(msg, info);
3574 pr_err("not enough space in Netlink message\n");
3580 mutex_unlock(&__ip_vs_mutex);
3586 static struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3588 .cmd = IPVS_CMD_NEW_SERVICE,
3589 .flags = GENL_ADMIN_PERM,
3590 .policy = ip_vs_cmd_policy,
3591 .doit = ip_vs_genl_set_cmd,
3594 .cmd = IPVS_CMD_SET_SERVICE,
3595 .flags = GENL_ADMIN_PERM,
3596 .policy = ip_vs_cmd_policy,
3597 .doit = ip_vs_genl_set_cmd,
3600 .cmd = IPVS_CMD_DEL_SERVICE,
3601 .flags = GENL_ADMIN_PERM,
3602 .policy = ip_vs_cmd_policy,
3603 .doit = ip_vs_genl_set_cmd,
3606 .cmd = IPVS_CMD_GET_SERVICE,
3607 .flags = GENL_ADMIN_PERM,
3608 .doit = ip_vs_genl_get_cmd,
3609 .dumpit = ip_vs_genl_dump_services,
3610 .policy = ip_vs_cmd_policy,
3613 .cmd = IPVS_CMD_NEW_DEST,
3614 .flags = GENL_ADMIN_PERM,
3615 .policy = ip_vs_cmd_policy,
3616 .doit = ip_vs_genl_set_cmd,
3619 .cmd = IPVS_CMD_SET_DEST,
3620 .flags = GENL_ADMIN_PERM,
3621 .policy = ip_vs_cmd_policy,
3622 .doit = ip_vs_genl_set_cmd,
3625 .cmd = IPVS_CMD_DEL_DEST,
3626 .flags = GENL_ADMIN_PERM,
3627 .policy = ip_vs_cmd_policy,
3628 .doit = ip_vs_genl_set_cmd,
3631 .cmd = IPVS_CMD_GET_DEST,
3632 .flags = GENL_ADMIN_PERM,
3633 .policy = ip_vs_cmd_policy,
3634 .dumpit = ip_vs_genl_dump_dests,
3637 .cmd = IPVS_CMD_NEW_DAEMON,
3638 .flags = GENL_ADMIN_PERM,
3639 .policy = ip_vs_cmd_policy,
3640 .doit = ip_vs_genl_set_daemon,
3643 .cmd = IPVS_CMD_DEL_DAEMON,
3644 .flags = GENL_ADMIN_PERM,
3645 .policy = ip_vs_cmd_policy,
3646 .doit = ip_vs_genl_set_daemon,
3649 .cmd = IPVS_CMD_GET_DAEMON,
3650 .flags = GENL_ADMIN_PERM,
3651 .dumpit = ip_vs_genl_dump_daemons,
3654 .cmd = IPVS_CMD_SET_CONFIG,
3655 .flags = GENL_ADMIN_PERM,
3656 .policy = ip_vs_cmd_policy,
3657 .doit = ip_vs_genl_set_cmd,
3660 .cmd = IPVS_CMD_GET_CONFIG,
3661 .flags = GENL_ADMIN_PERM,
3662 .doit = ip_vs_genl_get_cmd,
3665 .cmd = IPVS_CMD_GET_INFO,
3666 .flags = GENL_ADMIN_PERM,
3667 .doit = ip_vs_genl_get_cmd,
3670 .cmd = IPVS_CMD_ZERO,
3671 .flags = GENL_ADMIN_PERM,
3672 .policy = ip_vs_cmd_policy,
3673 .doit = ip_vs_genl_set_cmd,
3676 .cmd = IPVS_CMD_FLUSH,
3677 .flags = GENL_ADMIN_PERM,
3678 .doit = ip_vs_genl_set_cmd,
3682 static int __init ip_vs_genl_register(void)
3684 return genl_register_family_with_ops(&ip_vs_genl_family,
3685 ip_vs_genl_ops, ARRAY_SIZE(ip_vs_genl_ops));
3688 static void ip_vs_genl_unregister(void)
3690 genl_unregister_family(&ip_vs_genl_family);
3693 /* End of Generic Netlink interface definitions */
3696 * per netns intit/exit func.
3698 #ifdef CONFIG_SYSCTL
3699 static int __net_init ip_vs_control_net_init_sysctl(struct net *net)
3702 struct netns_ipvs *ipvs = net_ipvs(net);
3703 struct ctl_table *tbl;
3705 atomic_set(&ipvs->dropentry, 0);
3706 spin_lock_init(&ipvs->dropentry_lock);
3707 spin_lock_init(&ipvs->droppacket_lock);
3708 spin_lock_init(&ipvs->securetcp_lock);
3710 if (!net_eq(net, &init_net)) {
3711 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3715 /* Don't export sysctls to unprivileged users */
3716 if (net->user_ns != &init_user_ns)
3717 tbl[0].procname = NULL;
3720 /* Initialize sysctl defaults */
3722 ipvs->sysctl_amemthresh = 1024;
3723 tbl[idx++].data = &ipvs->sysctl_amemthresh;
3724 ipvs->sysctl_am_droprate = 10;
3725 tbl[idx++].data = &ipvs->sysctl_am_droprate;
3726 tbl[idx++].data = &ipvs->sysctl_drop_entry;
3727 tbl[idx++].data = &ipvs->sysctl_drop_packet;
3728 #ifdef CONFIG_IP_VS_NFCT
3729 tbl[idx++].data = &ipvs->sysctl_conntrack;
3731 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3732 ipvs->sysctl_snat_reroute = 1;
3733 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3734 ipvs->sysctl_sync_ver = 1;
3735 tbl[idx++].data = &ipvs->sysctl_sync_ver;
3736 ipvs->sysctl_sync_ports = 1;
3737 tbl[idx++].data = &ipvs->sysctl_sync_ports;
3738 tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
3739 ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3740 tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3741 ipvs->sysctl_sync_sock_size = 0;
3742 tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3743 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3744 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3745 tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
3746 tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
3747 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3748 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3749 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3750 tbl[idx].data = &ipvs->sysctl_sync_threshold;
3751 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3752 ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3753 tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3754 ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3755 tbl[idx++].data = &ipvs->sysctl_sync_retries;
3756 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3757 ipvs->sysctl_pmtu_disc = 1;
3758 tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
3759 tbl[idx++].data = &ipvs->sysctl_backup_only;
3762 ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3763 if (ipvs->sysctl_hdr == NULL) {
3764 if (!net_eq(net, &init_net))
3768 ip_vs_start_estimator(net, &ipvs->tot_stats);
3769 ipvs->sysctl_tbl = tbl;
3770 /* Schedule defense work */
3771 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3772 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3777 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net)
3779 struct netns_ipvs *ipvs = net_ipvs(net);
3781 cancel_delayed_work_sync(&ipvs->defense_work);
3782 cancel_work_sync(&ipvs->defense_work.work);
3783 unregister_net_sysctl_table(ipvs->sysctl_hdr);
3788 static int __net_init ip_vs_control_net_init_sysctl(struct net *net) { return 0; }
3789 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net) { }
3793 static struct notifier_block ip_vs_dst_notifier = {
3794 .notifier_call = ip_vs_dst_event,
3797 int __net_init ip_vs_control_net_init(struct net *net)
3800 struct netns_ipvs *ipvs = net_ipvs(net);
3802 /* Initialize rs_table */
3803 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
3804 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
3806 INIT_LIST_HEAD(&ipvs->dest_trash);
3807 spin_lock_init(&ipvs->dest_trash_lock);
3808 setup_timer(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire,
3809 (unsigned long) net);
3810 atomic_set(&ipvs->ftpsvc_counter, 0);
3811 atomic_set(&ipvs->nullsvc_counter, 0);
3814 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
3815 if (!ipvs->tot_stats.cpustats)
3818 spin_lock_init(&ipvs->tot_stats.lock);
3820 proc_create("ip_vs", 0, net->proc_net, &ip_vs_info_fops);
3821 proc_create("ip_vs_stats", 0, net->proc_net, &ip_vs_stats_fops);
3822 proc_create("ip_vs_stats_percpu", 0, net->proc_net,
3823 &ip_vs_stats_percpu_fops);
3825 if (ip_vs_control_net_init_sysctl(net))
3831 free_percpu(ipvs->tot_stats.cpustats);
3835 void __net_exit ip_vs_control_net_cleanup(struct net *net)
3837 struct netns_ipvs *ipvs = net_ipvs(net);
3839 /* Some dest can be in grace period even before cleanup, we have to
3840 * defer ip_vs_trash_cleanup until ip_vs_dest_wait_readers is called.
3843 ip_vs_trash_cleanup(net);
3844 ip_vs_stop_estimator(net, &ipvs->tot_stats);
3845 ip_vs_control_net_cleanup_sysctl(net);
3846 remove_proc_entry("ip_vs_stats_percpu", net->proc_net);
3847 remove_proc_entry("ip_vs_stats", net->proc_net);
3848 remove_proc_entry("ip_vs", net->proc_net);
3849 free_percpu(ipvs->tot_stats.cpustats);
3852 int __init ip_vs_register_nl_ioctl(void)
3856 ret = nf_register_sockopt(&ip_vs_sockopts);
3858 pr_err("cannot register sockopt.\n");
3862 ret = ip_vs_genl_register();
3864 pr_err("cannot register Generic Netlink interface.\n");
3870 nf_unregister_sockopt(&ip_vs_sockopts);
3875 void ip_vs_unregister_nl_ioctl(void)
3877 ip_vs_genl_unregister();
3878 nf_unregister_sockopt(&ip_vs_sockopts);
3881 int __init ip_vs_control_init(void)
3888 /* Initialize svc_table, ip_vs_svc_fwm_table */
3889 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
3890 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
3891 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3894 smp_wmb(); /* Do we really need it now ? */
3896 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
3905 void ip_vs_control_cleanup(void)
3908 unregister_netdevice_notifier(&ip_vs_dst_notifier);