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 && (u->netmask < 1 || u->netmask > 128)) {
1173 svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1175 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1179 svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1180 if (!svc->stats.cpustats) {
1185 /* I'm the first user of the service */
1186 atomic_set(&svc->refcnt, 0);
1189 svc->protocol = u->protocol;
1190 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1191 svc->port = u->port;
1192 svc->fwmark = u->fwmark;
1193 svc->flags = u->flags;
1194 svc->timeout = u->timeout * HZ;
1195 svc->netmask = u->netmask;
1198 INIT_LIST_HEAD(&svc->destinations);
1199 spin_lock_init(&svc->sched_lock);
1200 spin_lock_init(&svc->stats.lock);
1202 /* Bind the scheduler */
1203 ret = ip_vs_bind_scheduler(svc, sched);
1208 /* Bind the ct retriever */
1209 RCU_INIT_POINTER(svc->pe, pe);
1212 /* Update the virtual service counters */
1213 if (svc->port == FTPPORT)
1214 atomic_inc(&ipvs->ftpsvc_counter);
1215 else if (svc->port == 0)
1216 atomic_inc(&ipvs->nullsvc_counter);
1218 ip_vs_start_estimator(net, &svc->stats);
1220 /* Count only IPv4 services for old get/setsockopt interface */
1221 if (svc->af == AF_INET)
1222 ipvs->num_services++;
1224 /* Hash the service into the service table */
1225 ip_vs_svc_hash(svc);
1228 /* Now there is a service - full throttle */
1235 ip_vs_unbind_scheduler(svc, sched);
1236 ip_vs_service_free(svc);
1238 ip_vs_scheduler_put(sched);
1241 /* decrease the module use count */
1242 ip_vs_use_count_dec();
1249 * Edit a service and bind it with a new scheduler
1252 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1254 struct ip_vs_scheduler *sched, *old_sched;
1255 struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1259 * Lookup the scheduler, by 'u->sched_name'
1261 sched = ip_vs_scheduler_get(u->sched_name);
1262 if (sched == NULL) {
1263 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1268 if (u->pe_name && *u->pe_name) {
1269 pe = ip_vs_pe_getbyname(u->pe_name);
1271 pr_info("persistence engine module ip_vs_pe_%s "
1272 "not found\n", u->pe_name);
1279 #ifdef CONFIG_IP_VS_IPV6
1280 if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1286 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1287 if (sched != old_sched) {
1288 /* Bind the new scheduler */
1289 ret = ip_vs_bind_scheduler(svc, sched);
1294 /* Unbind the old scheduler on success */
1295 ip_vs_unbind_scheduler(svc, old_sched);
1299 * Set the flags and timeout value
1301 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1302 svc->timeout = u->timeout * HZ;
1303 svc->netmask = u->netmask;
1305 old_pe = rcu_dereference_protected(svc->pe, 1);
1307 rcu_assign_pointer(svc->pe, pe);
1310 ip_vs_scheduler_put(old_sched);
1311 ip_vs_pe_put(old_pe);
1315 static void ip_vs_service_rcu_free(struct rcu_head *head)
1317 struct ip_vs_service *svc;
1319 svc = container_of(head, struct ip_vs_service, rcu_head);
1320 ip_vs_service_free(svc);
1324 * Delete a service from the service list
1325 * - The service must be unlinked, unlocked and not referenced!
1326 * - We are called under _bh lock
1328 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1330 struct ip_vs_dest *dest, *nxt;
1331 struct ip_vs_scheduler *old_sched;
1332 struct ip_vs_pe *old_pe;
1333 struct netns_ipvs *ipvs = net_ipvs(svc->net);
1335 pr_info("%s: enter\n", __func__);
1337 /* Count only IPv4 services for old get/setsockopt interface */
1338 if (svc->af == AF_INET)
1339 ipvs->num_services--;
1341 ip_vs_stop_estimator(svc->net, &svc->stats);
1343 /* Unbind scheduler */
1344 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1345 ip_vs_unbind_scheduler(svc, old_sched);
1346 ip_vs_scheduler_put(old_sched);
1348 /* Unbind persistence engine, keep svc->pe */
1349 old_pe = rcu_dereference_protected(svc->pe, 1);
1350 ip_vs_pe_put(old_pe);
1353 * Unlink the whole destination list
1355 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1356 __ip_vs_unlink_dest(svc, dest, 0);
1357 __ip_vs_del_dest(svc->net, dest, cleanup);
1361 * Update the virtual service counters
1363 if (svc->port == FTPPORT)
1364 atomic_dec(&ipvs->ftpsvc_counter);
1365 else if (svc->port == 0)
1366 atomic_dec(&ipvs->nullsvc_counter);
1369 * Free the service if nobody refers to it
1371 if (atomic_dec_and_test(&svc->refcnt)) {
1372 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n",
1374 IP_VS_DBG_ADDR(svc->af, &svc->addr),
1376 call_rcu(&svc->rcu_head, ip_vs_service_rcu_free);
1379 /* decrease the module use count */
1380 ip_vs_use_count_dec();
1384 * Unlink a service from list and try to delete it if its refcnt reached 0
1386 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1388 /* Hold svc to avoid double release from dest_trash */
1389 atomic_inc(&svc->refcnt);
1391 * Unhash it from the service table
1393 ip_vs_svc_unhash(svc);
1395 __ip_vs_del_service(svc, cleanup);
1399 * Delete a service from the service list
1401 static int ip_vs_del_service(struct ip_vs_service *svc)
1405 ip_vs_unlink_service(svc, false);
1412 * Flush all the virtual services
1414 static int ip_vs_flush(struct net *net, bool cleanup)
1417 struct ip_vs_service *svc;
1418 struct hlist_node *n;
1421 * Flush the service table hashed by <netns,protocol,addr,port>
1423 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1424 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1426 if (net_eq(svc->net, net))
1427 ip_vs_unlink_service(svc, cleanup);
1432 * Flush the service table hashed by fwmark
1434 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1435 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1437 if (net_eq(svc->net, net))
1438 ip_vs_unlink_service(svc, cleanup);
1446 * Delete service by {netns} in the service table.
1447 * Called by __ip_vs_cleanup()
1449 void ip_vs_service_net_cleanup(struct net *net)
1452 /* Check for "full" addressed entries */
1453 mutex_lock(&__ip_vs_mutex);
1454 ip_vs_flush(net, true);
1455 mutex_unlock(&__ip_vs_mutex);
1459 /* Put all references for device (dst_cache) */
1461 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1463 spin_lock_bh(&dest->dst_lock);
1464 if (dest->dest_dst && dest->dest_dst->dst_cache->dev == dev) {
1465 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1467 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1469 atomic_read(&dest->refcnt));
1470 __ip_vs_dst_cache_reset(dest);
1472 spin_unlock_bh(&dest->dst_lock);
1475 /* Netdev event receiver
1476 * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1478 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1481 struct net_device *dev = ptr;
1482 struct net *net = dev_net(dev);
1483 struct netns_ipvs *ipvs = net_ipvs(net);
1484 struct ip_vs_service *svc;
1485 struct ip_vs_dest *dest;
1488 if (event != NETDEV_DOWN || !ipvs)
1490 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1492 mutex_lock(&__ip_vs_mutex);
1493 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1494 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1495 if (net_eq(svc->net, net)) {
1496 list_for_each_entry(dest, &svc->destinations,
1498 ip_vs_forget_dev(dest, dev);
1503 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1504 if (net_eq(svc->net, net)) {
1505 list_for_each_entry(dest, &svc->destinations,
1507 ip_vs_forget_dev(dest, dev);
1514 spin_lock_bh(&ipvs->dest_trash_lock);
1515 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1516 ip_vs_forget_dev(dest, dev);
1518 spin_unlock_bh(&ipvs->dest_trash_lock);
1519 mutex_unlock(&__ip_vs_mutex);
1525 * Zero counters in a service or all services
1527 static int ip_vs_zero_service(struct ip_vs_service *svc)
1529 struct ip_vs_dest *dest;
1531 list_for_each_entry(dest, &svc->destinations, n_list) {
1532 ip_vs_zero_stats(&dest->stats);
1534 ip_vs_zero_stats(&svc->stats);
1538 static int ip_vs_zero_all(struct net *net)
1541 struct ip_vs_service *svc;
1543 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1544 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1545 if (net_eq(svc->net, net))
1546 ip_vs_zero_service(svc);
1550 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1551 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1552 if (net_eq(svc->net, net))
1553 ip_vs_zero_service(svc);
1557 ip_vs_zero_stats(&net_ipvs(net)->tot_stats);
1561 #ifdef CONFIG_SYSCTL
1564 static int three = 3;
1567 proc_do_defense_mode(ctl_table *table, int write,
1568 void __user *buffer, size_t *lenp, loff_t *ppos)
1570 struct net *net = current->nsproxy->net_ns;
1571 int *valp = table->data;
1575 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1576 if (write && (*valp != val)) {
1577 if ((*valp < 0) || (*valp > 3)) {
1578 /* Restore the correct value */
1581 update_defense_level(net_ipvs(net));
1588 proc_do_sync_threshold(ctl_table *table, int write,
1589 void __user *buffer, size_t *lenp, loff_t *ppos)
1591 int *valp = table->data;
1595 /* backup the value first */
1596 memcpy(val, valp, sizeof(val));
1598 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1599 if (write && (valp[0] < 0 || valp[1] < 0 ||
1600 (valp[0] >= valp[1] && valp[1]))) {
1601 /* Restore the correct value */
1602 memcpy(valp, val, sizeof(val));
1608 proc_do_sync_mode(ctl_table *table, int write,
1609 void __user *buffer, size_t *lenp, loff_t *ppos)
1611 int *valp = table->data;
1615 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1616 if (write && (*valp != val)) {
1617 if ((*valp < 0) || (*valp > 1)) {
1618 /* Restore the correct value */
1626 proc_do_sync_ports(ctl_table *table, int write,
1627 void __user *buffer, size_t *lenp, loff_t *ppos)
1629 int *valp = table->data;
1633 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1634 if (write && (*valp != val)) {
1635 if (*valp < 1 || !is_power_of_2(*valp)) {
1636 /* Restore the correct value */
1644 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1645 * Do not change order or insert new entries without
1646 * align with netns init in ip_vs_control_net_init()
1649 static struct ctl_table vs_vars[] = {
1651 .procname = "amemthresh",
1652 .maxlen = sizeof(int),
1654 .proc_handler = proc_dointvec,
1657 .procname = "am_droprate",
1658 .maxlen = sizeof(int),
1660 .proc_handler = proc_dointvec,
1663 .procname = "drop_entry",
1664 .maxlen = sizeof(int),
1666 .proc_handler = proc_do_defense_mode,
1669 .procname = "drop_packet",
1670 .maxlen = sizeof(int),
1672 .proc_handler = proc_do_defense_mode,
1674 #ifdef CONFIG_IP_VS_NFCT
1676 .procname = "conntrack",
1677 .maxlen = sizeof(int),
1679 .proc_handler = &proc_dointvec,
1683 .procname = "secure_tcp",
1684 .maxlen = sizeof(int),
1686 .proc_handler = proc_do_defense_mode,
1689 .procname = "snat_reroute",
1690 .maxlen = sizeof(int),
1692 .proc_handler = &proc_dointvec,
1695 .procname = "sync_version",
1696 .maxlen = sizeof(int),
1698 .proc_handler = &proc_do_sync_mode,
1701 .procname = "sync_ports",
1702 .maxlen = sizeof(int),
1704 .proc_handler = &proc_do_sync_ports,
1707 .procname = "sync_qlen_max",
1708 .maxlen = sizeof(int),
1710 .proc_handler = proc_dointvec,
1713 .procname = "sync_sock_size",
1714 .maxlen = sizeof(int),
1716 .proc_handler = proc_dointvec,
1719 .procname = "cache_bypass",
1720 .maxlen = sizeof(int),
1722 .proc_handler = proc_dointvec,
1725 .procname = "expire_nodest_conn",
1726 .maxlen = sizeof(int),
1728 .proc_handler = proc_dointvec,
1731 .procname = "expire_quiescent_template",
1732 .maxlen = sizeof(int),
1734 .proc_handler = proc_dointvec,
1737 .procname = "sync_threshold",
1739 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1741 .proc_handler = proc_do_sync_threshold,
1744 .procname = "sync_refresh_period",
1745 .maxlen = sizeof(int),
1747 .proc_handler = proc_dointvec_jiffies,
1750 .procname = "sync_retries",
1751 .maxlen = sizeof(int),
1753 .proc_handler = proc_dointvec_minmax,
1758 .procname = "nat_icmp_send",
1759 .maxlen = sizeof(int),
1761 .proc_handler = proc_dointvec,
1764 .procname = "pmtu_disc",
1765 .maxlen = sizeof(int),
1767 .proc_handler = proc_dointvec,
1770 .procname = "backup_only",
1771 .maxlen = sizeof(int),
1773 .proc_handler = proc_dointvec,
1775 #ifdef CONFIG_IP_VS_DEBUG
1777 .procname = "debug_level",
1778 .data = &sysctl_ip_vs_debug_level,
1779 .maxlen = sizeof(int),
1781 .proc_handler = proc_dointvec,
1786 .procname = "timeout_established",
1787 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1788 .maxlen = sizeof(int),
1790 .proc_handler = proc_dointvec_jiffies,
1793 .procname = "timeout_synsent",
1794 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1795 .maxlen = sizeof(int),
1797 .proc_handler = proc_dointvec_jiffies,
1800 .procname = "timeout_synrecv",
1801 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1802 .maxlen = sizeof(int),
1804 .proc_handler = proc_dointvec_jiffies,
1807 .procname = "timeout_finwait",
1808 .data = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1809 .maxlen = sizeof(int),
1811 .proc_handler = proc_dointvec_jiffies,
1814 .procname = "timeout_timewait",
1815 .data = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1816 .maxlen = sizeof(int),
1818 .proc_handler = proc_dointvec_jiffies,
1821 .procname = "timeout_close",
1822 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1823 .maxlen = sizeof(int),
1825 .proc_handler = proc_dointvec_jiffies,
1828 .procname = "timeout_closewait",
1829 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1830 .maxlen = sizeof(int),
1832 .proc_handler = proc_dointvec_jiffies,
1835 .procname = "timeout_lastack",
1836 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1837 .maxlen = sizeof(int),
1839 .proc_handler = proc_dointvec_jiffies,
1842 .procname = "timeout_listen",
1843 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1844 .maxlen = sizeof(int),
1846 .proc_handler = proc_dointvec_jiffies,
1849 .procname = "timeout_synack",
1850 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1851 .maxlen = sizeof(int),
1853 .proc_handler = proc_dointvec_jiffies,
1856 .procname = "timeout_udp",
1857 .data = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1858 .maxlen = sizeof(int),
1860 .proc_handler = proc_dointvec_jiffies,
1863 .procname = "timeout_icmp",
1864 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1865 .maxlen = sizeof(int),
1867 .proc_handler = proc_dointvec_jiffies,
1875 #ifdef CONFIG_PROC_FS
1878 struct seq_net_private p; /* Do not move this, netns depends upon it*/
1879 struct hlist_head *table;
1884 * Write the contents of the VS rule table to a PROCfs file.
1885 * (It is kept just for backward compatibility)
1887 static inline const char *ip_vs_fwd_name(unsigned int flags)
1889 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1890 case IP_VS_CONN_F_LOCALNODE:
1892 case IP_VS_CONN_F_TUNNEL:
1894 case IP_VS_CONN_F_DROUTE:
1902 /* Get the Nth entry in the two lists */
1903 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1905 struct net *net = seq_file_net(seq);
1906 struct ip_vs_iter *iter = seq->private;
1908 struct ip_vs_service *svc;
1910 /* look in hash by protocol */
1911 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1912 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
1913 if (net_eq(svc->net, net) && pos-- == 0) {
1914 iter->table = ip_vs_svc_table;
1921 /* keep looking in fwmark */
1922 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1923 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
1925 if (net_eq(svc->net, net) && pos-- == 0) {
1926 iter->table = ip_vs_svc_fwm_table;
1936 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1940 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1944 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1946 struct hlist_node *e;
1947 struct ip_vs_iter *iter;
1948 struct ip_vs_service *svc;
1951 if (v == SEQ_START_TOKEN)
1952 return ip_vs_info_array(seq,0);
1955 iter = seq->private;
1957 if (iter->table == ip_vs_svc_table) {
1958 /* next service in table hashed by protocol */
1959 e = rcu_dereference(hlist_next_rcu(&svc->s_list));
1961 return hlist_entry(e, struct ip_vs_service, s_list);
1963 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1964 hlist_for_each_entry_rcu(svc,
1965 &ip_vs_svc_table[iter->bucket],
1971 iter->table = ip_vs_svc_fwm_table;
1976 /* next service in hashed by fwmark */
1977 e = rcu_dereference(hlist_next_rcu(&svc->f_list));
1979 return hlist_entry(e, struct ip_vs_service, f_list);
1982 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1983 hlist_for_each_entry_rcu(svc,
1984 &ip_vs_svc_fwm_table[iter->bucket],
1992 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
1998 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
2000 if (v == SEQ_START_TOKEN) {
2002 "IP Virtual Server version %d.%d.%d (size=%d)\n",
2003 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2005 "Prot LocalAddress:Port Scheduler Flags\n");
2007 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2009 const struct ip_vs_service *svc = v;
2010 const struct ip_vs_iter *iter = seq->private;
2011 const struct ip_vs_dest *dest;
2012 struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler);
2014 if (iter->table == ip_vs_svc_table) {
2015 #ifdef CONFIG_IP_VS_IPV6
2016 if (svc->af == AF_INET6)
2017 seq_printf(seq, "%s [%pI6]:%04X %s ",
2018 ip_vs_proto_name(svc->protocol),
2024 seq_printf(seq, "%s %08X:%04X %s %s ",
2025 ip_vs_proto_name(svc->protocol),
2026 ntohl(svc->addr.ip),
2029 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2031 seq_printf(seq, "FWM %08X %s %s",
2032 svc->fwmark, sched->name,
2033 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2036 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2037 seq_printf(seq, "persistent %d %08X\n",
2039 ntohl(svc->netmask));
2041 seq_putc(seq, '\n');
2043 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
2044 #ifdef CONFIG_IP_VS_IPV6
2045 if (dest->af == AF_INET6)
2048 " %-7s %-6d %-10d %-10d\n",
2051 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2052 atomic_read(&dest->weight),
2053 atomic_read(&dest->activeconns),
2054 atomic_read(&dest->inactconns));
2059 "%-7s %-6d %-10d %-10d\n",
2060 ntohl(dest->addr.ip),
2062 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2063 atomic_read(&dest->weight),
2064 atomic_read(&dest->activeconns),
2065 atomic_read(&dest->inactconns));
2072 static const struct seq_operations ip_vs_info_seq_ops = {
2073 .start = ip_vs_info_seq_start,
2074 .next = ip_vs_info_seq_next,
2075 .stop = ip_vs_info_seq_stop,
2076 .show = ip_vs_info_seq_show,
2079 static int ip_vs_info_open(struct inode *inode, struct file *file)
2081 return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2082 sizeof(struct ip_vs_iter));
2085 static const struct file_operations ip_vs_info_fops = {
2086 .owner = THIS_MODULE,
2087 .open = ip_vs_info_open,
2089 .llseek = seq_lseek,
2090 .release = seq_release_net,
2093 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2095 struct net *net = seq_file_single_net(seq);
2096 struct ip_vs_stats_user show;
2098 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2100 " Total Incoming Outgoing Incoming Outgoing\n");
2102 " Conns Packets Packets Bytes Bytes\n");
2104 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2105 seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", show.conns,
2106 show.inpkts, show.outpkts,
2107 (unsigned long long) show.inbytes,
2108 (unsigned long long) show.outbytes);
2110 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2112 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2113 seq_printf(seq, "%8X %8X %8X %16X %16X\n",
2114 show.cps, show.inpps, show.outpps,
2115 show.inbps, show.outbps);
2120 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2122 return single_open_net(inode, file, ip_vs_stats_show);
2125 static const struct file_operations ip_vs_stats_fops = {
2126 .owner = THIS_MODULE,
2127 .open = ip_vs_stats_seq_open,
2129 .llseek = seq_lseek,
2130 .release = single_release_net,
2133 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2135 struct net *net = seq_file_single_net(seq);
2136 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2137 struct ip_vs_cpu_stats *cpustats = tot_stats->cpustats;
2138 struct ip_vs_stats_user rates;
2141 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2143 " Total Incoming Outgoing Incoming Outgoing\n");
2145 "CPU Conns Packets Packets Bytes Bytes\n");
2147 for_each_possible_cpu(i) {
2148 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2150 __u64 inbytes, outbytes;
2153 start = u64_stats_fetch_begin_bh(&u->syncp);
2154 inbytes = u->ustats.inbytes;
2155 outbytes = u->ustats.outbytes;
2156 } while (u64_stats_fetch_retry_bh(&u->syncp, start));
2158 seq_printf(seq, "%3X %8X %8X %8X %16LX %16LX\n",
2159 i, u->ustats.conns, u->ustats.inpkts,
2160 u->ustats.outpkts, (__u64)inbytes,
2164 spin_lock_bh(&tot_stats->lock);
2166 seq_printf(seq, " ~ %8X %8X %8X %16LX %16LX\n\n",
2167 tot_stats->ustats.conns, tot_stats->ustats.inpkts,
2168 tot_stats->ustats.outpkts,
2169 (unsigned long long) tot_stats->ustats.inbytes,
2170 (unsigned long long) tot_stats->ustats.outbytes);
2172 ip_vs_read_estimator(&rates, tot_stats);
2174 spin_unlock_bh(&tot_stats->lock);
2176 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2178 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2179 seq_printf(seq, " %8X %8X %8X %16X %16X\n",
2189 static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2191 return single_open_net(inode, file, ip_vs_stats_percpu_show);
2194 static const struct file_operations ip_vs_stats_percpu_fops = {
2195 .owner = THIS_MODULE,
2196 .open = ip_vs_stats_percpu_seq_open,
2198 .llseek = seq_lseek,
2199 .release = single_release_net,
2204 * Set timeout values for tcp tcpfin udp in the timeout_table.
2206 static int ip_vs_set_timeout(struct net *net, struct ip_vs_timeout_user *u)
2208 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2209 struct ip_vs_proto_data *pd;
2212 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2217 #ifdef CONFIG_IP_VS_PROTO_TCP
2218 if (u->tcp_timeout) {
2219 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2220 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2221 = u->tcp_timeout * HZ;
2224 if (u->tcp_fin_timeout) {
2225 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2226 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2227 = u->tcp_fin_timeout * HZ;
2231 #ifdef CONFIG_IP_VS_PROTO_UDP
2232 if (u->udp_timeout) {
2233 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2234 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2235 = u->udp_timeout * HZ;
2242 #define SET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2243 #define SERVICE_ARG_LEN (sizeof(struct ip_vs_service_user))
2244 #define SVCDEST_ARG_LEN (sizeof(struct ip_vs_service_user) + \
2245 sizeof(struct ip_vs_dest_user))
2246 #define TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2247 #define DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user))
2248 #define MAX_ARG_LEN SVCDEST_ARG_LEN
2250 static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
2251 [SET_CMDID(IP_VS_SO_SET_ADD)] = SERVICE_ARG_LEN,
2252 [SET_CMDID(IP_VS_SO_SET_EDIT)] = SERVICE_ARG_LEN,
2253 [SET_CMDID(IP_VS_SO_SET_DEL)] = SERVICE_ARG_LEN,
2254 [SET_CMDID(IP_VS_SO_SET_FLUSH)] = 0,
2255 [SET_CMDID(IP_VS_SO_SET_ADDDEST)] = SVCDEST_ARG_LEN,
2256 [SET_CMDID(IP_VS_SO_SET_DELDEST)] = SVCDEST_ARG_LEN,
2257 [SET_CMDID(IP_VS_SO_SET_EDITDEST)] = SVCDEST_ARG_LEN,
2258 [SET_CMDID(IP_VS_SO_SET_TIMEOUT)] = TIMEOUT_ARG_LEN,
2259 [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)] = DAEMON_ARG_LEN,
2260 [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)] = DAEMON_ARG_LEN,
2261 [SET_CMDID(IP_VS_SO_SET_ZERO)] = SERVICE_ARG_LEN,
2264 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2265 struct ip_vs_service_user *usvc_compat)
2267 memset(usvc, 0, sizeof(*usvc));
2270 usvc->protocol = usvc_compat->protocol;
2271 usvc->addr.ip = usvc_compat->addr;
2272 usvc->port = usvc_compat->port;
2273 usvc->fwmark = usvc_compat->fwmark;
2275 /* Deep copy of sched_name is not needed here */
2276 usvc->sched_name = usvc_compat->sched_name;
2278 usvc->flags = usvc_compat->flags;
2279 usvc->timeout = usvc_compat->timeout;
2280 usvc->netmask = usvc_compat->netmask;
2283 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2284 struct ip_vs_dest_user *udest_compat)
2286 memset(udest, 0, sizeof(*udest));
2288 udest->addr.ip = udest_compat->addr;
2289 udest->port = udest_compat->port;
2290 udest->conn_flags = udest_compat->conn_flags;
2291 udest->weight = udest_compat->weight;
2292 udest->u_threshold = udest_compat->u_threshold;
2293 udest->l_threshold = udest_compat->l_threshold;
2297 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2299 struct net *net = sock_net(sk);
2301 unsigned char arg[MAX_ARG_LEN];
2302 struct ip_vs_service_user *usvc_compat;
2303 struct ip_vs_service_user_kern usvc;
2304 struct ip_vs_service *svc;
2305 struct ip_vs_dest_user *udest_compat;
2306 struct ip_vs_dest_user_kern udest;
2307 struct netns_ipvs *ipvs = net_ipvs(net);
2309 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2312 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2314 if (len < 0 || len > MAX_ARG_LEN)
2316 if (len != set_arglen[SET_CMDID(cmd)]) {
2317 pr_err("set_ctl: len %u != %u\n",
2318 len, set_arglen[SET_CMDID(cmd)]);
2322 if (copy_from_user(arg, user, len) != 0)
2325 /* increase the module use count */
2326 ip_vs_use_count_inc();
2328 /* Handle daemons since they have another lock */
2329 if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2330 cmd == IP_VS_SO_SET_STOPDAEMON) {
2331 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2333 if (mutex_lock_interruptible(&ipvs->sync_mutex)) {
2337 if (cmd == IP_VS_SO_SET_STARTDAEMON)
2338 ret = start_sync_thread(net, dm->state, dm->mcast_ifn,
2341 ret = stop_sync_thread(net, dm->state);
2342 mutex_unlock(&ipvs->sync_mutex);
2346 if (mutex_lock_interruptible(&__ip_vs_mutex)) {
2351 if (cmd == IP_VS_SO_SET_FLUSH) {
2352 /* Flush the virtual service */
2353 ret = ip_vs_flush(net, false);
2355 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2356 /* Set timeout values for (tcp tcpfin udp) */
2357 ret = ip_vs_set_timeout(net, (struct ip_vs_timeout_user *)arg);
2361 usvc_compat = (struct ip_vs_service_user *)arg;
2362 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2364 /* We only use the new structs internally, so copy userspace compat
2365 * structs to extended internal versions */
2366 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2367 ip_vs_copy_udest_compat(&udest, udest_compat);
2369 if (cmd == IP_VS_SO_SET_ZERO) {
2370 /* if no service address is set, zero counters in all */
2371 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2372 ret = ip_vs_zero_all(net);
2377 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2378 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2379 usvc.protocol != IPPROTO_SCTP) {
2380 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2381 usvc.protocol, &usvc.addr.ip,
2382 ntohs(usvc.port), usvc.sched_name);
2387 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2389 if (usvc.fwmark == 0)
2390 svc = __ip_vs_service_find(net, usvc.af, usvc.protocol,
2391 &usvc.addr, usvc.port);
2393 svc = __ip_vs_svc_fwm_find(net, usvc.af, usvc.fwmark);
2396 if (cmd != IP_VS_SO_SET_ADD
2397 && (svc == NULL || svc->protocol != usvc.protocol)) {
2403 case IP_VS_SO_SET_ADD:
2407 ret = ip_vs_add_service(net, &usvc, &svc);
2409 case IP_VS_SO_SET_EDIT:
2410 ret = ip_vs_edit_service(svc, &usvc);
2412 case IP_VS_SO_SET_DEL:
2413 ret = ip_vs_del_service(svc);
2417 case IP_VS_SO_SET_ZERO:
2418 ret = ip_vs_zero_service(svc);
2420 case IP_VS_SO_SET_ADDDEST:
2421 ret = ip_vs_add_dest(svc, &udest);
2423 case IP_VS_SO_SET_EDITDEST:
2424 ret = ip_vs_edit_dest(svc, &udest);
2426 case IP_VS_SO_SET_DELDEST:
2427 ret = ip_vs_del_dest(svc, &udest);
2434 mutex_unlock(&__ip_vs_mutex);
2436 /* decrease the module use count */
2437 ip_vs_use_count_dec();
2444 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2446 struct ip_vs_scheduler *sched;
2448 sched = rcu_dereference_protected(src->scheduler, 1);
2449 dst->protocol = src->protocol;
2450 dst->addr = src->addr.ip;
2451 dst->port = src->port;
2452 dst->fwmark = src->fwmark;
2453 strlcpy(dst->sched_name, sched->name, sizeof(dst->sched_name));
2454 dst->flags = src->flags;
2455 dst->timeout = src->timeout / HZ;
2456 dst->netmask = src->netmask;
2457 dst->num_dests = src->num_dests;
2458 ip_vs_copy_stats(&dst->stats, &src->stats);
2462 __ip_vs_get_service_entries(struct net *net,
2463 const struct ip_vs_get_services *get,
2464 struct ip_vs_get_services __user *uptr)
2467 struct ip_vs_service *svc;
2468 struct ip_vs_service_entry entry;
2471 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2472 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2473 /* Only expose IPv4 entries to old interface */
2474 if (svc->af != AF_INET || !net_eq(svc->net, net))
2477 if (count >= get->num_services)
2479 memset(&entry, 0, sizeof(entry));
2480 ip_vs_copy_service(&entry, svc);
2481 if (copy_to_user(&uptr->entrytable[count],
2482 &entry, sizeof(entry))) {
2490 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2491 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2492 /* Only expose IPv4 entries to old interface */
2493 if (svc->af != AF_INET || !net_eq(svc->net, net))
2496 if (count >= get->num_services)
2498 memset(&entry, 0, sizeof(entry));
2499 ip_vs_copy_service(&entry, svc);
2500 if (copy_to_user(&uptr->entrytable[count],
2501 &entry, sizeof(entry))) {
2513 __ip_vs_get_dest_entries(struct net *net, const struct ip_vs_get_dests *get,
2514 struct ip_vs_get_dests __user *uptr)
2516 struct ip_vs_service *svc;
2517 union nf_inet_addr addr = { .ip = get->addr };
2522 svc = __ip_vs_svc_fwm_find(net, AF_INET, get->fwmark);
2524 svc = __ip_vs_service_find(net, AF_INET, get->protocol, &addr,
2530 struct ip_vs_dest *dest;
2531 struct ip_vs_dest_entry entry;
2533 list_for_each_entry(dest, &svc->destinations, n_list) {
2534 if (count >= get->num_dests)
2537 entry.addr = dest->addr.ip;
2538 entry.port = dest->port;
2539 entry.conn_flags = atomic_read(&dest->conn_flags);
2540 entry.weight = atomic_read(&dest->weight);
2541 entry.u_threshold = dest->u_threshold;
2542 entry.l_threshold = dest->l_threshold;
2543 entry.activeconns = atomic_read(&dest->activeconns);
2544 entry.inactconns = atomic_read(&dest->inactconns);
2545 entry.persistconns = atomic_read(&dest->persistconns);
2546 ip_vs_copy_stats(&entry.stats, &dest->stats);
2547 if (copy_to_user(&uptr->entrytable[count],
2548 &entry, sizeof(entry))) {
2560 __ip_vs_get_timeouts(struct net *net, struct ip_vs_timeout_user *u)
2562 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2563 struct ip_vs_proto_data *pd;
2566 memset(u, 0, sizeof (*u));
2568 #ifdef CONFIG_IP_VS_PROTO_TCP
2569 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2570 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2571 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2573 #ifdef CONFIG_IP_VS_PROTO_UDP
2574 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2576 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2581 #define GET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2582 #define GET_INFO_ARG_LEN (sizeof(struct ip_vs_getinfo))
2583 #define GET_SERVICES_ARG_LEN (sizeof(struct ip_vs_get_services))
2584 #define GET_SERVICE_ARG_LEN (sizeof(struct ip_vs_service_entry))
2585 #define GET_DESTS_ARG_LEN (sizeof(struct ip_vs_get_dests))
2586 #define GET_TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2587 #define GET_DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user) * 2)
2589 static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
2590 [GET_CMDID(IP_VS_SO_GET_VERSION)] = 64,
2591 [GET_CMDID(IP_VS_SO_GET_INFO)] = GET_INFO_ARG_LEN,
2592 [GET_CMDID(IP_VS_SO_GET_SERVICES)] = GET_SERVICES_ARG_LEN,
2593 [GET_CMDID(IP_VS_SO_GET_SERVICE)] = GET_SERVICE_ARG_LEN,
2594 [GET_CMDID(IP_VS_SO_GET_DESTS)] = GET_DESTS_ARG_LEN,
2595 [GET_CMDID(IP_VS_SO_GET_TIMEOUT)] = GET_TIMEOUT_ARG_LEN,
2596 [GET_CMDID(IP_VS_SO_GET_DAEMON)] = GET_DAEMON_ARG_LEN,
2600 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2602 unsigned char arg[128];
2604 unsigned int copylen;
2605 struct net *net = sock_net(sk);
2606 struct netns_ipvs *ipvs = net_ipvs(net);
2609 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2612 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2615 if (*len < get_arglen[GET_CMDID(cmd)]) {
2616 pr_err("get_ctl: len %u < %u\n",
2617 *len, get_arglen[GET_CMDID(cmd)]);
2621 copylen = get_arglen[GET_CMDID(cmd)];
2625 if (copy_from_user(arg, user, copylen) != 0)
2628 * Handle daemons first since it has its own locking
2630 if (cmd == IP_VS_SO_GET_DAEMON) {
2631 struct ip_vs_daemon_user d[2];
2633 memset(&d, 0, sizeof(d));
2634 if (mutex_lock_interruptible(&ipvs->sync_mutex))
2635 return -ERESTARTSYS;
2637 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2638 d[0].state = IP_VS_STATE_MASTER;
2639 strlcpy(d[0].mcast_ifn, ipvs->master_mcast_ifn,
2640 sizeof(d[0].mcast_ifn));
2641 d[0].syncid = ipvs->master_syncid;
2643 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2644 d[1].state = IP_VS_STATE_BACKUP;
2645 strlcpy(d[1].mcast_ifn, ipvs->backup_mcast_ifn,
2646 sizeof(d[1].mcast_ifn));
2647 d[1].syncid = ipvs->backup_syncid;
2649 if (copy_to_user(user, &d, sizeof(d)) != 0)
2651 mutex_unlock(&ipvs->sync_mutex);
2655 if (mutex_lock_interruptible(&__ip_vs_mutex))
2656 return -ERESTARTSYS;
2659 case IP_VS_SO_GET_VERSION:
2663 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2664 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2665 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2669 *len = strlen(buf)+1;
2673 case IP_VS_SO_GET_INFO:
2675 struct ip_vs_getinfo info;
2676 info.version = IP_VS_VERSION_CODE;
2677 info.size = ip_vs_conn_tab_size;
2678 info.num_services = ipvs->num_services;
2679 if (copy_to_user(user, &info, sizeof(info)) != 0)
2684 case IP_VS_SO_GET_SERVICES:
2686 struct ip_vs_get_services *get;
2689 get = (struct ip_vs_get_services *)arg;
2690 size = sizeof(*get) +
2691 sizeof(struct ip_vs_service_entry) * get->num_services;
2693 pr_err("length: %u != %u\n", *len, size);
2697 ret = __ip_vs_get_service_entries(net, get, user);
2701 case IP_VS_SO_GET_SERVICE:
2703 struct ip_vs_service_entry *entry;
2704 struct ip_vs_service *svc;
2705 union nf_inet_addr addr;
2707 entry = (struct ip_vs_service_entry *)arg;
2708 addr.ip = entry->addr;
2711 svc = __ip_vs_svc_fwm_find(net, AF_INET, entry->fwmark);
2713 svc = __ip_vs_service_find(net, AF_INET,
2714 entry->protocol, &addr,
2718 ip_vs_copy_service(entry, svc);
2719 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2726 case IP_VS_SO_GET_DESTS:
2728 struct ip_vs_get_dests *get;
2731 get = (struct ip_vs_get_dests *)arg;
2732 size = sizeof(*get) +
2733 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2735 pr_err("length: %u != %u\n", *len, size);
2739 ret = __ip_vs_get_dest_entries(net, get, user);
2743 case IP_VS_SO_GET_TIMEOUT:
2745 struct ip_vs_timeout_user t;
2747 __ip_vs_get_timeouts(net, &t);
2748 if (copy_to_user(user, &t, sizeof(t)) != 0)
2758 mutex_unlock(&__ip_vs_mutex);
2763 static struct nf_sockopt_ops ip_vs_sockopts = {
2765 .set_optmin = IP_VS_BASE_CTL,
2766 .set_optmax = IP_VS_SO_SET_MAX+1,
2767 .set = do_ip_vs_set_ctl,
2768 .get_optmin = IP_VS_BASE_CTL,
2769 .get_optmax = IP_VS_SO_GET_MAX+1,
2770 .get = do_ip_vs_get_ctl,
2771 .owner = THIS_MODULE,
2775 * Generic Netlink interface
2778 /* IPVS genetlink family */
2779 static struct genl_family ip_vs_genl_family = {
2780 .id = GENL_ID_GENERATE,
2782 .name = IPVS_GENL_NAME,
2783 .version = IPVS_GENL_VERSION,
2784 .maxattr = IPVS_CMD_MAX,
2785 .netnsok = true, /* Make ipvsadm to work on netns */
2788 /* Policy used for first-level command attributes */
2789 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2790 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2791 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2792 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2793 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2794 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2795 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2798 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2799 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2800 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2801 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2802 .len = IP_VS_IFNAME_MAXLEN },
2803 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2806 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2807 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2808 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2809 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2810 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2811 .len = sizeof(union nf_inet_addr) },
2812 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2813 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2814 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2815 .len = IP_VS_SCHEDNAME_MAXLEN },
2816 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2817 .len = IP_VS_PENAME_MAXLEN },
2818 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2819 .len = sizeof(struct ip_vs_flags) },
2820 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2821 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2822 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2825 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2826 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2827 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2828 .len = sizeof(union nf_inet_addr) },
2829 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2830 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2831 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2832 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2833 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2834 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2835 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2836 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2837 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2840 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2841 struct ip_vs_stats *stats)
2843 struct ip_vs_stats_user ustats;
2844 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2848 ip_vs_copy_stats(&ustats, stats);
2850 if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, ustats.conns) ||
2851 nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, ustats.inpkts) ||
2852 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, ustats.outpkts) ||
2853 nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, ustats.inbytes) ||
2854 nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, ustats.outbytes) ||
2855 nla_put_u32(skb, IPVS_STATS_ATTR_CPS, ustats.cps) ||
2856 nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, ustats.inpps) ||
2857 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, ustats.outpps) ||
2858 nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, ustats.inbps) ||
2859 nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, ustats.outbps))
2860 goto nla_put_failure;
2861 nla_nest_end(skb, nl_stats);
2866 nla_nest_cancel(skb, nl_stats);
2870 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2871 struct ip_vs_service *svc)
2873 struct ip_vs_scheduler *sched;
2874 struct nlattr *nl_service;
2875 struct ip_vs_flags flags = { .flags = svc->flags,
2878 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2882 if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2883 goto nla_put_failure;
2885 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2886 goto nla_put_failure;
2888 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2889 nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2890 nla_put_u16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2891 goto nla_put_failure;
2894 sched = rcu_dereference_protected(svc->scheduler, 1);
2895 if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched->name) ||
2897 nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, svc->pe->name)) ||
2898 nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2899 nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2900 nla_put_u32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
2901 goto nla_put_failure;
2902 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2903 goto nla_put_failure;
2905 nla_nest_end(skb, nl_service);
2910 nla_nest_cancel(skb, nl_service);
2914 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2915 struct ip_vs_service *svc,
2916 struct netlink_callback *cb)
2920 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2921 &ip_vs_genl_family, NLM_F_MULTI,
2922 IPVS_CMD_NEW_SERVICE);
2926 if (ip_vs_genl_fill_service(skb, svc) < 0)
2927 goto nla_put_failure;
2929 return genlmsg_end(skb, hdr);
2932 genlmsg_cancel(skb, hdr);
2936 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2937 struct netlink_callback *cb)
2940 int start = cb->args[0];
2941 struct ip_vs_service *svc;
2942 struct net *net = skb_sknet(skb);
2944 mutex_lock(&__ip_vs_mutex);
2945 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2946 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2947 if (++idx <= start || !net_eq(svc->net, net))
2949 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2951 goto nla_put_failure;
2956 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2957 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
2958 if (++idx <= start || !net_eq(svc->net, net))
2960 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2962 goto nla_put_failure;
2968 mutex_unlock(&__ip_vs_mutex);
2974 static int ip_vs_genl_parse_service(struct net *net,
2975 struct ip_vs_service_user_kern *usvc,
2976 struct nlattr *nla, int full_entry,
2977 struct ip_vs_service **ret_svc)
2979 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
2980 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
2981 struct ip_vs_service *svc;
2983 /* Parse mandatory identifying service fields first */
2985 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
2988 nla_af = attrs[IPVS_SVC_ATTR_AF];
2989 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
2990 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
2991 nla_port = attrs[IPVS_SVC_ATTR_PORT];
2992 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
2994 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
2997 memset(usvc, 0, sizeof(*usvc));
2999 usvc->af = nla_get_u16(nla_af);
3000 #ifdef CONFIG_IP_VS_IPV6
3001 if (usvc->af != AF_INET && usvc->af != AF_INET6)
3003 if (usvc->af != AF_INET)
3005 return -EAFNOSUPPORT;
3008 usvc->protocol = IPPROTO_TCP;
3009 usvc->fwmark = nla_get_u32(nla_fwmark);
3011 usvc->protocol = nla_get_u16(nla_protocol);
3012 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3013 usvc->port = nla_get_u16(nla_port);
3019 svc = __ip_vs_svc_fwm_find(net, usvc->af, usvc->fwmark);
3021 svc = __ip_vs_service_find(net, usvc->af, usvc->protocol,
3022 &usvc->addr, usvc->port);
3026 /* If a full entry was requested, check for the additional fields */
3028 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3030 struct ip_vs_flags flags;
3032 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3033 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3034 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3035 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3036 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3038 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3041 nla_memcpy(&flags, nla_flags, sizeof(flags));
3043 /* prefill flags from service if it already exists */
3045 usvc->flags = svc->flags;
3047 /* set new flags from userland */
3048 usvc->flags = (usvc->flags & ~flags.mask) |
3049 (flags.flags & flags.mask);
3050 usvc->sched_name = nla_data(nla_sched);
3051 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3052 usvc->timeout = nla_get_u32(nla_timeout);
3053 usvc->netmask = nla_get_u32(nla_netmask);
3059 static struct ip_vs_service *ip_vs_genl_find_service(struct net *net,
3062 struct ip_vs_service_user_kern usvc;
3063 struct ip_vs_service *svc;
3066 ret = ip_vs_genl_parse_service(net, &usvc, nla, 0, &svc);
3067 return ret ? ERR_PTR(ret) : svc;
3070 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3072 struct nlattr *nl_dest;
3074 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3078 if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3079 nla_put_u16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3080 nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3081 (atomic_read(&dest->conn_flags) &
3082 IP_VS_CONN_F_FWD_MASK)) ||
3083 nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3084 atomic_read(&dest->weight)) ||
3085 nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3086 nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3087 nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3088 atomic_read(&dest->activeconns)) ||
3089 nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3090 atomic_read(&dest->inactconns)) ||
3091 nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3092 atomic_read(&dest->persistconns)))
3093 goto nla_put_failure;
3094 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
3095 goto nla_put_failure;
3097 nla_nest_end(skb, nl_dest);
3102 nla_nest_cancel(skb, nl_dest);
3106 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3107 struct netlink_callback *cb)
3111 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3112 &ip_vs_genl_family, NLM_F_MULTI,
3117 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3118 goto nla_put_failure;
3120 return genlmsg_end(skb, hdr);
3123 genlmsg_cancel(skb, hdr);
3127 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3128 struct netlink_callback *cb)
3131 int start = cb->args[0];
3132 struct ip_vs_service *svc;
3133 struct ip_vs_dest *dest;
3134 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3135 struct net *net = skb_sknet(skb);
3137 mutex_lock(&__ip_vs_mutex);
3139 /* Try to find the service for which to dump destinations */
3140 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3141 IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3145 svc = ip_vs_genl_find_service(net, attrs[IPVS_CMD_ATTR_SERVICE]);
3146 if (IS_ERR(svc) || svc == NULL)
3149 /* Dump the destinations */
3150 list_for_each_entry(dest, &svc->destinations, n_list) {
3153 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3155 goto nla_put_failure;
3163 mutex_unlock(&__ip_vs_mutex);
3168 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3169 struct nlattr *nla, int full_entry)
3171 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3172 struct nlattr *nla_addr, *nla_port;
3174 /* Parse mandatory identifying destination fields first */
3176 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3179 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3180 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3182 if (!(nla_addr && nla_port))
3185 memset(udest, 0, sizeof(*udest));
3187 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3188 udest->port = nla_get_u16(nla_port);
3190 /* If a full entry was requested, check for the additional fields */
3192 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3195 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3196 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3197 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3198 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3200 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3203 udest->conn_flags = nla_get_u32(nla_fwd)
3204 & IP_VS_CONN_F_FWD_MASK;
3205 udest->weight = nla_get_u32(nla_weight);
3206 udest->u_threshold = nla_get_u32(nla_u_thresh);
3207 udest->l_threshold = nla_get_u32(nla_l_thresh);
3213 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __be32 state,
3214 const char *mcast_ifn, __be32 syncid)
3216 struct nlattr *nl_daemon;
3218 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3222 if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3223 nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn) ||
3224 nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid))
3225 goto nla_put_failure;
3226 nla_nest_end(skb, nl_daemon);
3231 nla_nest_cancel(skb, nl_daemon);
3235 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __be32 state,
3236 const char *mcast_ifn, __be32 syncid,
3237 struct netlink_callback *cb)
3240 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3241 &ip_vs_genl_family, NLM_F_MULTI,
3242 IPVS_CMD_NEW_DAEMON);
3246 if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
3247 goto nla_put_failure;
3249 return genlmsg_end(skb, hdr);
3252 genlmsg_cancel(skb, hdr);
3256 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3257 struct netlink_callback *cb)
3259 struct net *net = skb_sknet(skb);
3260 struct netns_ipvs *ipvs = net_ipvs(net);
3262 mutex_lock(&ipvs->sync_mutex);
3263 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3264 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3265 ipvs->master_mcast_ifn,
3266 ipvs->master_syncid, cb) < 0)
3267 goto nla_put_failure;
3272 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3273 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3274 ipvs->backup_mcast_ifn,
3275 ipvs->backup_syncid, cb) < 0)
3276 goto nla_put_failure;
3282 mutex_unlock(&ipvs->sync_mutex);
3287 static int ip_vs_genl_new_daemon(struct net *net, struct nlattr **attrs)
3289 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3290 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3291 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3294 return start_sync_thread(net,
3295 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3296 nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3297 nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3300 static int ip_vs_genl_del_daemon(struct net *net, struct nlattr **attrs)
3302 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3305 return stop_sync_thread(net,
3306 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3309 static int ip_vs_genl_set_config(struct net *net, struct nlattr **attrs)
3311 struct ip_vs_timeout_user t;
3313 __ip_vs_get_timeouts(net, &t);
3315 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3316 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3318 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3320 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3322 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3323 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3325 return ip_vs_set_timeout(net, &t);
3328 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3332 struct netns_ipvs *ipvs;
3334 net = skb_sknet(skb);
3335 ipvs = net_ipvs(net);
3336 cmd = info->genlhdr->cmd;
3338 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3339 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3341 mutex_lock(&ipvs->sync_mutex);
3342 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3343 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3344 info->attrs[IPVS_CMD_ATTR_DAEMON],
3345 ip_vs_daemon_policy)) {
3350 if (cmd == IPVS_CMD_NEW_DAEMON)
3351 ret = ip_vs_genl_new_daemon(net, daemon_attrs);
3353 ret = ip_vs_genl_del_daemon(net, daemon_attrs);
3355 mutex_unlock(&ipvs->sync_mutex);
3360 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3362 struct ip_vs_service *svc = NULL;
3363 struct ip_vs_service_user_kern usvc;
3364 struct ip_vs_dest_user_kern udest;
3366 int need_full_svc = 0, need_full_dest = 0;
3369 net = skb_sknet(skb);
3370 cmd = info->genlhdr->cmd;
3372 mutex_lock(&__ip_vs_mutex);
3374 if (cmd == IPVS_CMD_FLUSH) {
3375 ret = ip_vs_flush(net, false);
3377 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3378 ret = ip_vs_genl_set_config(net, info->attrs);
3380 } else if (cmd == IPVS_CMD_ZERO &&
3381 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3382 ret = ip_vs_zero_all(net);
3386 /* All following commands require a service argument, so check if we
3387 * received a valid one. We need a full service specification when
3388 * adding / editing a service. Only identifying members otherwise. */
3389 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3392 ret = ip_vs_genl_parse_service(net, &usvc,
3393 info->attrs[IPVS_CMD_ATTR_SERVICE],
3394 need_full_svc, &svc);
3398 /* Unless we're adding a new service, the service must already exist */
3399 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3404 /* Destination commands require a valid destination argument. For
3405 * adding / editing a destination, we need a full destination
3407 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3408 cmd == IPVS_CMD_DEL_DEST) {
3409 if (cmd != IPVS_CMD_DEL_DEST)
3412 ret = ip_vs_genl_parse_dest(&udest,
3413 info->attrs[IPVS_CMD_ATTR_DEST],
3420 case IPVS_CMD_NEW_SERVICE:
3422 ret = ip_vs_add_service(net, &usvc, &svc);
3426 case IPVS_CMD_SET_SERVICE:
3427 ret = ip_vs_edit_service(svc, &usvc);
3429 case IPVS_CMD_DEL_SERVICE:
3430 ret = ip_vs_del_service(svc);
3431 /* do not use svc, it can be freed */
3433 case IPVS_CMD_NEW_DEST:
3434 ret = ip_vs_add_dest(svc, &udest);
3436 case IPVS_CMD_SET_DEST:
3437 ret = ip_vs_edit_dest(svc, &udest);
3439 case IPVS_CMD_DEL_DEST:
3440 ret = ip_vs_del_dest(svc, &udest);
3443 ret = ip_vs_zero_service(svc);
3450 mutex_unlock(&__ip_vs_mutex);
3455 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3457 struct sk_buff *msg;
3459 int ret, cmd, reply_cmd;
3462 net = skb_sknet(skb);
3463 cmd = info->genlhdr->cmd;
3465 if (cmd == IPVS_CMD_GET_SERVICE)
3466 reply_cmd = IPVS_CMD_NEW_SERVICE;
3467 else if (cmd == IPVS_CMD_GET_INFO)
3468 reply_cmd = IPVS_CMD_SET_INFO;
3469 else if (cmd == IPVS_CMD_GET_CONFIG)
3470 reply_cmd = IPVS_CMD_SET_CONFIG;
3472 pr_err("unknown Generic Netlink command\n");
3476 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3480 mutex_lock(&__ip_vs_mutex);
3482 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3484 goto nla_put_failure;
3487 case IPVS_CMD_GET_SERVICE:
3489 struct ip_vs_service *svc;
3491 svc = ip_vs_genl_find_service(net,
3492 info->attrs[IPVS_CMD_ATTR_SERVICE]);
3497 ret = ip_vs_genl_fill_service(msg, svc);
3499 goto nla_put_failure;
3508 case IPVS_CMD_GET_CONFIG:
3510 struct ip_vs_timeout_user t;
3512 __ip_vs_get_timeouts(net, &t);
3513 #ifdef CONFIG_IP_VS_PROTO_TCP
3514 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3516 nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3518 goto nla_put_failure;
3520 #ifdef CONFIG_IP_VS_PROTO_UDP
3521 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3522 goto nla_put_failure;
3528 case IPVS_CMD_GET_INFO:
3529 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3530 IP_VS_VERSION_CODE) ||
3531 nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3532 ip_vs_conn_tab_size))
3533 goto nla_put_failure;
3537 genlmsg_end(msg, reply);
3538 ret = genlmsg_reply(msg, info);
3542 pr_err("not enough space in Netlink message\n");
3548 mutex_unlock(&__ip_vs_mutex);
3554 static struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3556 .cmd = IPVS_CMD_NEW_SERVICE,
3557 .flags = GENL_ADMIN_PERM,
3558 .policy = ip_vs_cmd_policy,
3559 .doit = ip_vs_genl_set_cmd,
3562 .cmd = IPVS_CMD_SET_SERVICE,
3563 .flags = GENL_ADMIN_PERM,
3564 .policy = ip_vs_cmd_policy,
3565 .doit = ip_vs_genl_set_cmd,
3568 .cmd = IPVS_CMD_DEL_SERVICE,
3569 .flags = GENL_ADMIN_PERM,
3570 .policy = ip_vs_cmd_policy,
3571 .doit = ip_vs_genl_set_cmd,
3574 .cmd = IPVS_CMD_GET_SERVICE,
3575 .flags = GENL_ADMIN_PERM,
3576 .doit = ip_vs_genl_get_cmd,
3577 .dumpit = ip_vs_genl_dump_services,
3578 .policy = ip_vs_cmd_policy,
3581 .cmd = IPVS_CMD_NEW_DEST,
3582 .flags = GENL_ADMIN_PERM,
3583 .policy = ip_vs_cmd_policy,
3584 .doit = ip_vs_genl_set_cmd,
3587 .cmd = IPVS_CMD_SET_DEST,
3588 .flags = GENL_ADMIN_PERM,
3589 .policy = ip_vs_cmd_policy,
3590 .doit = ip_vs_genl_set_cmd,
3593 .cmd = IPVS_CMD_DEL_DEST,
3594 .flags = GENL_ADMIN_PERM,
3595 .policy = ip_vs_cmd_policy,
3596 .doit = ip_vs_genl_set_cmd,
3599 .cmd = IPVS_CMD_GET_DEST,
3600 .flags = GENL_ADMIN_PERM,
3601 .policy = ip_vs_cmd_policy,
3602 .dumpit = ip_vs_genl_dump_dests,
3605 .cmd = IPVS_CMD_NEW_DAEMON,
3606 .flags = GENL_ADMIN_PERM,
3607 .policy = ip_vs_cmd_policy,
3608 .doit = ip_vs_genl_set_daemon,
3611 .cmd = IPVS_CMD_DEL_DAEMON,
3612 .flags = GENL_ADMIN_PERM,
3613 .policy = ip_vs_cmd_policy,
3614 .doit = ip_vs_genl_set_daemon,
3617 .cmd = IPVS_CMD_GET_DAEMON,
3618 .flags = GENL_ADMIN_PERM,
3619 .dumpit = ip_vs_genl_dump_daemons,
3622 .cmd = IPVS_CMD_SET_CONFIG,
3623 .flags = GENL_ADMIN_PERM,
3624 .policy = ip_vs_cmd_policy,
3625 .doit = ip_vs_genl_set_cmd,
3628 .cmd = IPVS_CMD_GET_CONFIG,
3629 .flags = GENL_ADMIN_PERM,
3630 .doit = ip_vs_genl_get_cmd,
3633 .cmd = IPVS_CMD_GET_INFO,
3634 .flags = GENL_ADMIN_PERM,
3635 .doit = ip_vs_genl_get_cmd,
3638 .cmd = IPVS_CMD_ZERO,
3639 .flags = GENL_ADMIN_PERM,
3640 .policy = ip_vs_cmd_policy,
3641 .doit = ip_vs_genl_set_cmd,
3644 .cmd = IPVS_CMD_FLUSH,
3645 .flags = GENL_ADMIN_PERM,
3646 .doit = ip_vs_genl_set_cmd,
3650 static int __init ip_vs_genl_register(void)
3652 return genl_register_family_with_ops(&ip_vs_genl_family,
3653 ip_vs_genl_ops, ARRAY_SIZE(ip_vs_genl_ops));
3656 static void ip_vs_genl_unregister(void)
3658 genl_unregister_family(&ip_vs_genl_family);
3661 /* End of Generic Netlink interface definitions */
3664 * per netns intit/exit func.
3666 #ifdef CONFIG_SYSCTL
3667 static int __net_init ip_vs_control_net_init_sysctl(struct net *net)
3670 struct netns_ipvs *ipvs = net_ipvs(net);
3671 struct ctl_table *tbl;
3673 atomic_set(&ipvs->dropentry, 0);
3674 spin_lock_init(&ipvs->dropentry_lock);
3675 spin_lock_init(&ipvs->droppacket_lock);
3676 spin_lock_init(&ipvs->securetcp_lock);
3678 if (!net_eq(net, &init_net)) {
3679 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3683 /* Don't export sysctls to unprivileged users */
3684 if (net->user_ns != &init_user_ns)
3685 tbl[0].procname = NULL;
3688 /* Initialize sysctl defaults */
3690 ipvs->sysctl_amemthresh = 1024;
3691 tbl[idx++].data = &ipvs->sysctl_amemthresh;
3692 ipvs->sysctl_am_droprate = 10;
3693 tbl[idx++].data = &ipvs->sysctl_am_droprate;
3694 tbl[idx++].data = &ipvs->sysctl_drop_entry;
3695 tbl[idx++].data = &ipvs->sysctl_drop_packet;
3696 #ifdef CONFIG_IP_VS_NFCT
3697 tbl[idx++].data = &ipvs->sysctl_conntrack;
3699 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3700 ipvs->sysctl_snat_reroute = 1;
3701 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3702 ipvs->sysctl_sync_ver = 1;
3703 tbl[idx++].data = &ipvs->sysctl_sync_ver;
3704 ipvs->sysctl_sync_ports = 1;
3705 tbl[idx++].data = &ipvs->sysctl_sync_ports;
3706 ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3707 tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3708 ipvs->sysctl_sync_sock_size = 0;
3709 tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3710 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3711 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3712 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3713 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3714 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3715 tbl[idx].data = &ipvs->sysctl_sync_threshold;
3716 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3717 ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3718 tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3719 ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3720 tbl[idx++].data = &ipvs->sysctl_sync_retries;
3721 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3722 ipvs->sysctl_pmtu_disc = 1;
3723 tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
3724 tbl[idx++].data = &ipvs->sysctl_backup_only;
3727 ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3728 if (ipvs->sysctl_hdr == NULL) {
3729 if (!net_eq(net, &init_net))
3733 ip_vs_start_estimator(net, &ipvs->tot_stats);
3734 ipvs->sysctl_tbl = tbl;
3735 /* Schedule defense work */
3736 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3737 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3742 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net)
3744 struct netns_ipvs *ipvs = net_ipvs(net);
3746 cancel_delayed_work_sync(&ipvs->defense_work);
3747 cancel_work_sync(&ipvs->defense_work.work);
3748 unregister_net_sysctl_table(ipvs->sysctl_hdr);
3753 static int __net_init ip_vs_control_net_init_sysctl(struct net *net) { return 0; }
3754 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net) { }
3758 static struct notifier_block ip_vs_dst_notifier = {
3759 .notifier_call = ip_vs_dst_event,
3762 int __net_init ip_vs_control_net_init(struct net *net)
3765 struct netns_ipvs *ipvs = net_ipvs(net);
3767 /* Initialize rs_table */
3768 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
3769 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
3771 INIT_LIST_HEAD(&ipvs->dest_trash);
3772 spin_lock_init(&ipvs->dest_trash_lock);
3773 setup_timer(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire,
3774 (unsigned long) net);
3775 atomic_set(&ipvs->ftpsvc_counter, 0);
3776 atomic_set(&ipvs->nullsvc_counter, 0);
3779 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
3780 if (!ipvs->tot_stats.cpustats)
3783 spin_lock_init(&ipvs->tot_stats.lock);
3785 proc_create("ip_vs", 0, net->proc_net, &ip_vs_info_fops);
3786 proc_create("ip_vs_stats", 0, net->proc_net, &ip_vs_stats_fops);
3787 proc_create("ip_vs_stats_percpu", 0, net->proc_net,
3788 &ip_vs_stats_percpu_fops);
3790 if (ip_vs_control_net_init_sysctl(net))
3796 free_percpu(ipvs->tot_stats.cpustats);
3800 void __net_exit ip_vs_control_net_cleanup(struct net *net)
3802 struct netns_ipvs *ipvs = net_ipvs(net);
3804 /* Some dest can be in grace period even before cleanup, we have to
3805 * defer ip_vs_trash_cleanup until ip_vs_dest_wait_readers is called.
3808 ip_vs_trash_cleanup(net);
3809 ip_vs_stop_estimator(net, &ipvs->tot_stats);
3810 ip_vs_control_net_cleanup_sysctl(net);
3811 remove_proc_entry("ip_vs_stats_percpu", net->proc_net);
3812 remove_proc_entry("ip_vs_stats", net->proc_net);
3813 remove_proc_entry("ip_vs", net->proc_net);
3814 free_percpu(ipvs->tot_stats.cpustats);
3817 int __init ip_vs_register_nl_ioctl(void)
3821 ret = nf_register_sockopt(&ip_vs_sockopts);
3823 pr_err("cannot register sockopt.\n");
3827 ret = ip_vs_genl_register();
3829 pr_err("cannot register Generic Netlink interface.\n");
3835 nf_unregister_sockopt(&ip_vs_sockopts);
3840 void ip_vs_unregister_nl_ioctl(void)
3842 ip_vs_genl_unregister();
3843 nf_unregister_sockopt(&ip_vs_sockopts);
3846 int __init ip_vs_control_init(void)
3853 /* Initialize svc_table, ip_vs_svc_fwm_table */
3854 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
3855 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
3856 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3859 smp_wmb(); /* Do we really need it now ? */
3861 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
3870 void ip_vs_control_cleanup(void)
3873 unregister_netdevice_notifier(&ip_vs_dst_notifier);