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 /* lock for service table */
59 static DEFINE_RWLOCK(__ip_vs_svc_lock);
61 /* sysctl variables */
63 #ifdef CONFIG_IP_VS_DEBUG
64 static int sysctl_ip_vs_debug_level = 0;
66 int ip_vs_get_debug_level(void)
68 return sysctl_ip_vs_debug_level;
74 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup);
77 #ifdef CONFIG_IP_VS_IPV6
78 /* Taken from rt6_fill_node() in net/ipv6/route.c, is there a better way? */
79 static bool __ip_vs_addr_is_local_v6(struct net *net,
80 const struct in6_addr *addr)
85 struct dst_entry *dst = ip6_route_output(net, NULL, &fl6);
88 is_local = !dst->error && dst->dev && (dst->dev->flags & IFF_LOOPBACK);
97 * update_defense_level is called from keventd and from sysctl,
98 * so it needs to protect itself from softirqs
100 static void update_defense_level(struct netns_ipvs *ipvs)
103 static int old_secure_tcp = 0;
108 /* we only count free and buffered memory (in pages) */
110 availmem = i.freeram + i.bufferram;
111 /* however in linux 2.5 the i.bufferram is total page cache size,
113 /* si_swapinfo(&i); */
114 /* availmem = availmem - (i.totalswap - i.freeswap); */
116 nomem = (availmem < ipvs->sysctl_amemthresh);
121 spin_lock(&ipvs->dropentry_lock);
122 switch (ipvs->sysctl_drop_entry) {
124 atomic_set(&ipvs->dropentry, 0);
128 atomic_set(&ipvs->dropentry, 1);
129 ipvs->sysctl_drop_entry = 2;
131 atomic_set(&ipvs->dropentry, 0);
136 atomic_set(&ipvs->dropentry, 1);
138 atomic_set(&ipvs->dropentry, 0);
139 ipvs->sysctl_drop_entry = 1;
143 atomic_set(&ipvs->dropentry, 1);
146 spin_unlock(&ipvs->dropentry_lock);
149 spin_lock(&ipvs->droppacket_lock);
150 switch (ipvs->sysctl_drop_packet) {
156 ipvs->drop_rate = ipvs->drop_counter
157 = ipvs->sysctl_amemthresh /
158 (ipvs->sysctl_amemthresh-availmem);
159 ipvs->sysctl_drop_packet = 2;
166 ipvs->drop_rate = ipvs->drop_counter
167 = ipvs->sysctl_amemthresh /
168 (ipvs->sysctl_amemthresh-availmem);
171 ipvs->sysctl_drop_packet = 1;
175 ipvs->drop_rate = ipvs->sysctl_am_droprate;
178 spin_unlock(&ipvs->droppacket_lock);
181 spin_lock(&ipvs->securetcp_lock);
182 switch (ipvs->sysctl_secure_tcp) {
184 if (old_secure_tcp >= 2)
189 if (old_secure_tcp < 2)
191 ipvs->sysctl_secure_tcp = 2;
193 if (old_secure_tcp >= 2)
199 if (old_secure_tcp < 2)
202 if (old_secure_tcp >= 2)
204 ipvs->sysctl_secure_tcp = 1;
208 if (old_secure_tcp < 2)
212 old_secure_tcp = ipvs->sysctl_secure_tcp;
214 ip_vs_protocol_timeout_change(ipvs,
215 ipvs->sysctl_secure_tcp > 1);
216 spin_unlock(&ipvs->securetcp_lock);
223 * Timer for checking the defense
225 #define DEFENSE_TIMER_PERIOD 1*HZ
227 static void defense_work_handler(struct work_struct *work)
229 struct netns_ipvs *ipvs =
230 container_of(work, struct netns_ipvs, defense_work.work);
232 update_defense_level(ipvs);
233 if (atomic_read(&ipvs->dropentry))
234 ip_vs_random_dropentry(ipvs->net);
235 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
240 ip_vs_use_count_inc(void)
242 return try_module_get(THIS_MODULE);
246 ip_vs_use_count_dec(void)
248 module_put(THIS_MODULE);
253 * Hash table: for virtual service lookups
255 #define IP_VS_SVC_TAB_BITS 8
256 #define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
257 #define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)
259 /* the service table hashed by <protocol, addr, port> */
260 static struct list_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE];
261 /* the service table hashed by fwmark */
262 static struct list_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE];
266 * Returns hash value for virtual service
268 static inline unsigned int
269 ip_vs_svc_hashkey(struct net *net, int af, unsigned int proto,
270 const union nf_inet_addr *addr, __be16 port)
272 register unsigned int porth = ntohs(port);
273 __be32 addr_fold = addr->ip;
276 #ifdef CONFIG_IP_VS_IPV6
278 addr_fold = addr->ip6[0]^addr->ip6[1]^
279 addr->ip6[2]^addr->ip6[3];
281 ahash = ntohl(addr_fold);
282 ahash ^= ((size_t) net >> 8);
284 return (proto ^ ahash ^ (porth >> IP_VS_SVC_TAB_BITS) ^ porth) &
289 * Returns hash value of fwmark for virtual service lookup
291 static inline unsigned int ip_vs_svc_fwm_hashkey(struct net *net, __u32 fwmark)
293 return (((size_t)net>>8) ^ fwmark) & IP_VS_SVC_TAB_MASK;
297 * Hashes a service in the ip_vs_svc_table by <netns,proto,addr,port>
298 * or in the ip_vs_svc_fwm_table by fwmark.
299 * Should be called with locked tables.
301 static int ip_vs_svc_hash(struct ip_vs_service *svc)
305 if (svc->flags & IP_VS_SVC_F_HASHED) {
306 pr_err("%s(): request for already hashed, called from %pF\n",
307 __func__, __builtin_return_address(0));
311 if (svc->fwmark == 0) {
313 * Hash it by <netns,protocol,addr,port> in ip_vs_svc_table
315 hash = ip_vs_svc_hashkey(svc->net, svc->af, svc->protocol,
316 &svc->addr, svc->port);
317 list_add(&svc->s_list, &ip_vs_svc_table[hash]);
320 * Hash it by fwmark in svc_fwm_table
322 hash = ip_vs_svc_fwm_hashkey(svc->net, svc->fwmark);
323 list_add(&svc->f_list, &ip_vs_svc_fwm_table[hash]);
326 svc->flags |= IP_VS_SVC_F_HASHED;
327 /* increase its refcnt because it is referenced by the svc table */
328 atomic_inc(&svc->refcnt);
334 * Unhashes a service from svc_table / svc_fwm_table.
335 * Should be called with locked tables.
337 static int ip_vs_svc_unhash(struct ip_vs_service *svc)
339 if (!(svc->flags & IP_VS_SVC_F_HASHED)) {
340 pr_err("%s(): request for unhash flagged, called from %pF\n",
341 __func__, __builtin_return_address(0));
345 if (svc->fwmark == 0) {
346 /* Remove it from the svc_table table */
347 list_del(&svc->s_list);
349 /* Remove it from the svc_fwm_table table */
350 list_del(&svc->f_list);
353 svc->flags &= ~IP_VS_SVC_F_HASHED;
354 atomic_dec(&svc->refcnt);
360 * Get service by {netns, proto,addr,port} in the service table.
362 static inline struct ip_vs_service *
363 __ip_vs_service_find(struct net *net, int af, __u16 protocol,
364 const union nf_inet_addr *vaddr, __be16 vport)
367 struct ip_vs_service *svc;
369 /* Check for "full" addressed entries */
370 hash = ip_vs_svc_hashkey(net, af, protocol, vaddr, vport);
372 list_for_each_entry(svc, &ip_vs_svc_table[hash], s_list){
374 && ip_vs_addr_equal(af, &svc->addr, vaddr)
375 && (svc->port == vport)
376 && (svc->protocol == protocol)
377 && net_eq(svc->net, net)) {
388 * Get service by {fwmark} in the service table.
390 static inline struct ip_vs_service *
391 __ip_vs_svc_fwm_find(struct net *net, int af, __u32 fwmark)
394 struct ip_vs_service *svc;
396 /* Check for fwmark addressed entries */
397 hash = ip_vs_svc_fwm_hashkey(net, fwmark);
399 list_for_each_entry(svc, &ip_vs_svc_fwm_table[hash], f_list) {
400 if (svc->fwmark == fwmark && svc->af == af
401 && net_eq(svc->net, net)) {
410 struct ip_vs_service *
411 ip_vs_service_get(struct net *net, int af, __u32 fwmark, __u16 protocol,
412 const union nf_inet_addr *vaddr, __be16 vport)
414 struct ip_vs_service *svc;
415 struct netns_ipvs *ipvs = net_ipvs(net);
417 read_lock(&__ip_vs_svc_lock);
420 * Check the table hashed by fwmark first
423 svc = __ip_vs_svc_fwm_find(net, af, fwmark);
429 * Check the table hashed by <protocol,addr,port>
430 * for "full" addressed entries
432 svc = __ip_vs_service_find(net, af, protocol, vaddr, vport);
435 && protocol == IPPROTO_TCP
436 && atomic_read(&ipvs->ftpsvc_counter)
437 && (vport == FTPDATA || ntohs(vport) >= PROT_SOCK)) {
439 * Check if ftp service entry exists, the packet
440 * might belong to FTP data connections.
442 svc = __ip_vs_service_find(net, af, protocol, vaddr, FTPPORT);
446 && atomic_read(&ipvs->nullsvc_counter)) {
448 * Check if the catch-all port (port zero) exists
450 svc = __ip_vs_service_find(net, af, protocol, vaddr, 0);
455 atomic_inc(&svc->usecnt);
456 read_unlock(&__ip_vs_svc_lock);
458 IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
459 fwmark, ip_vs_proto_name(protocol),
460 IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
461 svc ? "hit" : "not hit");
468 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
470 atomic_inc(&svc->refcnt);
475 __ip_vs_unbind_svc(struct ip_vs_dest *dest)
477 struct ip_vs_service *svc = dest->svc;
480 if (atomic_dec_and_test(&svc->refcnt)) {
481 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
483 IP_VS_DBG_ADDR(svc->af, &svc->addr),
484 ntohs(svc->port), atomic_read(&svc->usecnt));
485 free_percpu(svc->stats.cpustats);
492 * Returns hash value for real service
494 static inline unsigned int ip_vs_rs_hashkey(int af,
495 const union nf_inet_addr *addr,
498 register unsigned int porth = ntohs(port);
499 __be32 addr_fold = addr->ip;
501 #ifdef CONFIG_IP_VS_IPV6
503 addr_fold = addr->ip6[0]^addr->ip6[1]^
504 addr->ip6[2]^addr->ip6[3];
507 return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
511 /* Hash ip_vs_dest in rs_table by <proto,addr,port>. */
512 static void ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
516 if (dest->in_rs_table)
520 * Hash by proto,addr,port,
521 * which are the parameters of the real service.
523 hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
525 hlist_add_head_rcu(&dest->d_list, &ipvs->rs_table[hash]);
526 dest->in_rs_table = 1;
529 /* Unhash ip_vs_dest from rs_table. */
530 static void ip_vs_rs_unhash(struct ip_vs_dest *dest)
533 * Remove it from the rs_table table.
535 if (dest->in_rs_table) {
536 hlist_del_rcu(&dest->d_list);
537 dest->in_rs_table = 0;
541 /* Check if real service by <proto,addr,port> is present */
542 bool ip_vs_has_real_service(struct net *net, int af, __u16 protocol,
543 const union nf_inet_addr *daddr, __be16 dport)
545 struct netns_ipvs *ipvs = net_ipvs(net);
547 struct ip_vs_dest *dest;
549 /* Check for "full" addressed entries */
550 hash = ip_vs_rs_hashkey(af, daddr, dport);
553 hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
554 if (dest->port == dport &&
556 ip_vs_addr_equal(af, &dest->addr, daddr) &&
557 (dest->protocol == protocol || dest->vfwmark)) {
569 * Lookup destination by {addr,port} in the given service
571 static struct ip_vs_dest *
572 ip_vs_lookup_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
575 struct ip_vs_dest *dest;
578 * Find the destination for the given service
580 list_for_each_entry(dest, &svc->destinations, n_list) {
581 if ((dest->af == svc->af)
582 && ip_vs_addr_equal(svc->af, &dest->addr, daddr)
583 && (dest->port == dport)) {
593 * Find destination by {daddr,dport,vaddr,protocol}
594 * Cretaed to be used in ip_vs_process_message() in
595 * the backup synchronization daemon. It finds the
596 * destination to be bound to the received connection
599 struct ip_vs_dest *ip_vs_find_dest(struct net *net, int af,
600 const union nf_inet_addr *daddr,
602 const union nf_inet_addr *vaddr,
603 __be16 vport, __u16 protocol, __u32 fwmark,
606 struct ip_vs_dest *dest;
607 struct ip_vs_service *svc;
610 svc = ip_vs_service_get(net, af, fwmark, protocol, vaddr, vport);
613 if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
615 dest = ip_vs_lookup_dest(svc, daddr, port);
617 dest = ip_vs_lookup_dest(svc, daddr, port ^ dport);
619 ip_vs_dest_hold(dest);
620 ip_vs_service_put(svc);
624 void ip_vs_dest_dst_rcu_free(struct rcu_head *head)
626 struct ip_vs_dest_dst *dest_dst = container_of(head,
627 struct ip_vs_dest_dst,
630 dst_release(dest_dst->dst_cache);
634 /* Release dest_dst and dst_cache for dest in user context */
635 static void __ip_vs_dst_cache_reset(struct ip_vs_dest *dest)
637 struct ip_vs_dest_dst *old;
639 old = rcu_dereference_protected(dest->dest_dst, 1);
641 RCU_INIT_POINTER(dest->dest_dst, NULL);
642 call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
647 * Lookup dest by {svc,addr,port} in the destination trash.
648 * The destination trash is used to hold the destinations that are removed
649 * from the service table but are still referenced by some conn entries.
650 * The reason to add the destination trash is when the dest is temporary
651 * down (either by administrator or by monitor program), the dest can be
652 * picked back from the trash, the remaining connections to the dest can
653 * continue, and the counting information of the dest is also useful for
656 static struct ip_vs_dest *
657 ip_vs_trash_get_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
660 struct ip_vs_dest *dest;
661 struct netns_ipvs *ipvs = net_ipvs(svc->net);
664 * Find the destination in trash
666 spin_lock_bh(&ipvs->dest_trash_lock);
667 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
668 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
671 IP_VS_DBG_ADDR(svc->af, &dest->addr),
673 atomic_read(&dest->refcnt));
674 /* We can not reuse dest while in grace period
675 * because conns still can use dest->svc
677 if (test_bit(IP_VS_DEST_STATE_REMOVING, &dest->state))
679 if (dest->af == svc->af &&
680 ip_vs_addr_equal(svc->af, &dest->addr, daddr) &&
681 dest->port == dport &&
682 dest->vfwmark == svc->fwmark &&
683 dest->protocol == svc->protocol &&
685 (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
686 dest->vport == svc->port))) {
688 list_del(&dest->t_list);
689 ip_vs_dest_hold(dest);
697 spin_unlock_bh(&ipvs->dest_trash_lock);
702 static void ip_vs_dest_free(struct ip_vs_dest *dest)
704 __ip_vs_dst_cache_reset(dest);
705 __ip_vs_unbind_svc(dest);
706 free_percpu(dest->stats.cpustats);
711 * Clean up all the destinations in the trash
712 * Called by the ip_vs_control_cleanup()
714 * When the ip_vs_control_clearup is activated by ipvs module exit,
715 * the service tables must have been flushed and all the connections
716 * are expired, and the refcnt of each destination in the trash must
717 * be 0, so we simply release them here.
719 static void ip_vs_trash_cleanup(struct net *net)
721 struct ip_vs_dest *dest, *nxt;
722 struct netns_ipvs *ipvs = net_ipvs(net);
724 del_timer_sync(&ipvs->dest_trash_timer);
725 /* No need to use dest_trash_lock */
726 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, t_list) {
727 list_del(&dest->t_list);
728 ip_vs_dest_free(dest);
733 ip_vs_copy_stats(struct ip_vs_stats_user *dst, struct ip_vs_stats *src)
735 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->ustats.c - src->ustats0.c
737 spin_lock_bh(&src->lock);
739 IP_VS_SHOW_STATS_COUNTER(conns);
740 IP_VS_SHOW_STATS_COUNTER(inpkts);
741 IP_VS_SHOW_STATS_COUNTER(outpkts);
742 IP_VS_SHOW_STATS_COUNTER(inbytes);
743 IP_VS_SHOW_STATS_COUNTER(outbytes);
745 ip_vs_read_estimator(dst, src);
747 spin_unlock_bh(&src->lock);
751 ip_vs_zero_stats(struct ip_vs_stats *stats)
753 spin_lock_bh(&stats->lock);
755 /* get current counters as zero point, rates are zeroed */
757 #define IP_VS_ZERO_STATS_COUNTER(c) stats->ustats0.c = stats->ustats.c
759 IP_VS_ZERO_STATS_COUNTER(conns);
760 IP_VS_ZERO_STATS_COUNTER(inpkts);
761 IP_VS_ZERO_STATS_COUNTER(outpkts);
762 IP_VS_ZERO_STATS_COUNTER(inbytes);
763 IP_VS_ZERO_STATS_COUNTER(outbytes);
765 ip_vs_zero_estimator(stats);
767 spin_unlock_bh(&stats->lock);
771 * Update a destination in the given service
774 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
775 struct ip_vs_dest_user_kern *udest, int add)
777 struct netns_ipvs *ipvs = net_ipvs(svc->net);
780 /* set the weight and the flags */
781 atomic_set(&dest->weight, udest->weight);
782 conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
783 conn_flags |= IP_VS_CONN_F_INACTIVE;
785 /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
786 if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
787 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
790 * Put the real service in rs_table if not present.
791 * For now only for NAT!
793 ip_vs_rs_hash(ipvs, dest);
795 atomic_set(&dest->conn_flags, conn_flags);
797 /* bind the service */
799 __ip_vs_bind_svc(dest, svc);
801 if (dest->svc != svc) {
802 __ip_vs_unbind_svc(dest);
803 ip_vs_zero_stats(&dest->stats);
804 __ip_vs_bind_svc(dest, svc);
808 /* set the dest status flags */
809 dest->flags |= IP_VS_DEST_F_AVAILABLE;
811 if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
812 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
813 dest->u_threshold = udest->u_threshold;
814 dest->l_threshold = udest->l_threshold;
816 spin_lock_bh(&dest->dst_lock);
817 __ip_vs_dst_cache_reset(dest);
818 spin_unlock_bh(&dest->dst_lock);
821 ip_vs_start_estimator(svc->net, &dest->stats);
823 write_lock_bh(&__ip_vs_svc_lock);
825 /* Wait until all other svc users go away */
826 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
829 list_add(&dest->n_list, &svc->destinations);
831 if (svc->scheduler->add_dest)
832 svc->scheduler->add_dest(svc, dest);
834 if (svc->scheduler->upd_dest)
835 svc->scheduler->upd_dest(svc, dest);
838 /* call the update_service, because server weight may be changed */
839 if (svc->scheduler->update_service)
840 svc->scheduler->update_service(svc);
842 write_unlock_bh(&__ip_vs_svc_lock);
847 * Create a destination for the given service
850 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
851 struct ip_vs_dest **dest_p)
853 struct ip_vs_dest *dest;
858 #ifdef CONFIG_IP_VS_IPV6
859 if (svc->af == AF_INET6) {
860 atype = ipv6_addr_type(&udest->addr.in6);
861 if ((!(atype & IPV6_ADDR_UNICAST) ||
862 atype & IPV6_ADDR_LINKLOCAL) &&
863 !__ip_vs_addr_is_local_v6(svc->net, &udest->addr.in6))
868 atype = inet_addr_type(svc->net, udest->addr.ip);
869 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
873 dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
877 dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
878 if (!dest->stats.cpustats)
882 dest->protocol = svc->protocol;
883 dest->vaddr = svc->addr;
884 dest->vport = svc->port;
885 dest->vfwmark = svc->fwmark;
886 ip_vs_addr_copy(svc->af, &dest->addr, &udest->addr);
887 dest->port = udest->port;
889 atomic_set(&dest->activeconns, 0);
890 atomic_set(&dest->inactconns, 0);
891 atomic_set(&dest->persistconns, 0);
892 atomic_set(&dest->refcnt, 1);
894 INIT_HLIST_NODE(&dest->d_list);
895 spin_lock_init(&dest->dst_lock);
896 spin_lock_init(&dest->stats.lock);
897 __ip_vs_update_dest(svc, dest, udest, 1);
911 * Add a destination into an existing service
914 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
916 struct ip_vs_dest *dest;
917 union nf_inet_addr daddr;
918 __be16 dport = udest->port;
923 if (udest->weight < 0) {
924 pr_err("%s(): server weight less than zero\n", __func__);
928 if (udest->l_threshold > udest->u_threshold) {
929 pr_err("%s(): lower threshold is higher than upper threshold\n",
934 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
937 * Check if the dest already exists in the list
939 dest = ip_vs_lookup_dest(svc, &daddr, dport);
942 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
947 * Check if the dest already exists in the trash and
948 * is from the same service
950 dest = ip_vs_trash_get_dest(svc, &daddr, dport);
953 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
954 "dest->refcnt=%d, service %u/%s:%u\n",
955 IP_VS_DBG_ADDR(svc->af, &daddr), ntohs(dport),
956 atomic_read(&dest->refcnt),
958 IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
961 __ip_vs_update_dest(svc, dest, udest, 1);
965 * Allocate and initialize the dest structure
967 ret = ip_vs_new_dest(svc, udest, &dest);
976 * Edit a destination in the given service
979 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
981 struct ip_vs_dest *dest;
982 union nf_inet_addr daddr;
983 __be16 dport = udest->port;
987 if (udest->weight < 0) {
988 pr_err("%s(): server weight less than zero\n", __func__);
992 if (udest->l_threshold > udest->u_threshold) {
993 pr_err("%s(): lower threshold is higher than upper threshold\n",
998 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
1001 * Lookup the destination list
1003 dest = ip_vs_lookup_dest(svc, &daddr, dport);
1006 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1010 __ip_vs_update_dest(svc, dest, udest, 0);
1016 static void ip_vs_dest_wait_readers(struct rcu_head *head)
1018 struct ip_vs_dest *dest = container_of(head, struct ip_vs_dest,
1021 /* End of grace period after unlinking */
1022 clear_bit(IP_VS_DEST_STATE_REMOVING, &dest->state);
1027 * Delete a destination (must be already unlinked from the service)
1029 static void __ip_vs_del_dest(struct net *net, struct ip_vs_dest *dest,
1032 struct netns_ipvs *ipvs = net_ipvs(net);
1034 ip_vs_stop_estimator(net, &dest->stats);
1037 * Remove it from the d-linked list with the real services.
1039 ip_vs_rs_unhash(dest);
1042 set_bit(IP_VS_DEST_STATE_REMOVING, &dest->state);
1043 call_rcu(&dest->rcu_head, ip_vs_dest_wait_readers);
1046 spin_lock_bh(&ipvs->dest_trash_lock);
1047 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1048 IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port),
1049 atomic_read(&dest->refcnt));
1050 if (list_empty(&ipvs->dest_trash) && !cleanup)
1051 mod_timer(&ipvs->dest_trash_timer,
1052 jiffies + IP_VS_DEST_TRASH_PERIOD);
1053 /* dest lives in trash without reference */
1054 list_add(&dest->t_list, &ipvs->dest_trash);
1055 spin_unlock_bh(&ipvs->dest_trash_lock);
1056 ip_vs_dest_put(dest);
1061 * Unlink a destination from the given service
1063 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1064 struct ip_vs_dest *dest,
1067 dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1070 * Remove it from the d-linked destination list.
1072 list_del(&dest->n_list);
1075 if (svcupd && svc->scheduler->del_dest)
1076 svc->scheduler->del_dest(svc, dest);
1079 * Call the update_service function of its scheduler
1081 if (svcupd && svc->scheduler->update_service)
1082 svc->scheduler->update_service(svc);
1087 * Delete a destination server in the given service
1090 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1092 struct ip_vs_dest *dest;
1093 __be16 dport = udest->port;
1097 dest = ip_vs_lookup_dest(svc, &udest->addr, dport);
1100 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1104 write_lock_bh(&__ip_vs_svc_lock);
1107 * Wait until all other svc users go away.
1109 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1112 * Unlink dest from the service
1114 __ip_vs_unlink_dest(svc, dest, 1);
1116 write_unlock_bh(&__ip_vs_svc_lock);
1119 * Delete the destination
1121 __ip_vs_del_dest(svc->net, dest, false);
1128 static void ip_vs_dest_trash_expire(unsigned long data)
1130 struct net *net = (struct net *) data;
1131 struct netns_ipvs *ipvs = net_ipvs(net);
1132 struct ip_vs_dest *dest, *next;
1134 spin_lock(&ipvs->dest_trash_lock);
1135 list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) {
1136 /* Skip if dest is in grace period */
1137 if (test_bit(IP_VS_DEST_STATE_REMOVING, &dest->state))
1139 if (atomic_read(&dest->refcnt) > 0)
1141 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1143 IP_VS_DBG_ADDR(dest->svc->af, &dest->addr),
1145 list_del(&dest->t_list);
1146 ip_vs_dest_free(dest);
1148 if (!list_empty(&ipvs->dest_trash))
1149 mod_timer(&ipvs->dest_trash_timer,
1150 jiffies + IP_VS_DEST_TRASH_PERIOD);
1151 spin_unlock(&ipvs->dest_trash_lock);
1155 * Add a service into the service hash table
1158 ip_vs_add_service(struct net *net, struct ip_vs_service_user_kern *u,
1159 struct ip_vs_service **svc_p)
1162 struct ip_vs_scheduler *sched = NULL;
1163 struct ip_vs_pe *pe = NULL;
1164 struct ip_vs_service *svc = NULL;
1165 struct netns_ipvs *ipvs = net_ipvs(net);
1167 /* increase the module use count */
1168 ip_vs_use_count_inc();
1170 /* Lookup the scheduler by 'u->sched_name' */
1171 sched = ip_vs_scheduler_get(u->sched_name);
1172 if (sched == NULL) {
1173 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1178 if (u->pe_name && *u->pe_name) {
1179 pe = ip_vs_pe_getbyname(u->pe_name);
1181 pr_info("persistence engine module ip_vs_pe_%s "
1182 "not found\n", u->pe_name);
1188 #ifdef CONFIG_IP_VS_IPV6
1189 if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1195 svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1197 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1201 svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1202 if (!svc->stats.cpustats) {
1207 /* I'm the first user of the service */
1208 atomic_set(&svc->usecnt, 0);
1209 atomic_set(&svc->refcnt, 0);
1212 svc->protocol = u->protocol;
1213 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1214 svc->port = u->port;
1215 svc->fwmark = u->fwmark;
1216 svc->flags = u->flags;
1217 svc->timeout = u->timeout * HZ;
1218 svc->netmask = u->netmask;
1221 INIT_LIST_HEAD(&svc->destinations);
1222 spin_lock_init(&svc->sched_lock);
1223 spin_lock_init(&svc->stats.lock);
1225 /* Bind the scheduler */
1226 ret = ip_vs_bind_scheduler(svc, sched);
1231 /* Bind the ct retriever */
1232 ip_vs_bind_pe(svc, pe);
1235 /* Update the virtual service counters */
1236 if (svc->port == FTPPORT)
1237 atomic_inc(&ipvs->ftpsvc_counter);
1238 else if (svc->port == 0)
1239 atomic_inc(&ipvs->nullsvc_counter);
1241 ip_vs_start_estimator(net, &svc->stats);
1243 /* Count only IPv4 services for old get/setsockopt interface */
1244 if (svc->af == AF_INET)
1245 ipvs->num_services++;
1247 /* Hash the service into the service table */
1248 write_lock_bh(&__ip_vs_svc_lock);
1249 ip_vs_svc_hash(svc);
1250 write_unlock_bh(&__ip_vs_svc_lock);
1253 /* Now there is a service - full throttle */
1260 ip_vs_unbind_scheduler(svc);
1263 ip_vs_app_inc_put(svc->inc);
1266 if (svc->stats.cpustats)
1267 free_percpu(svc->stats.cpustats);
1270 ip_vs_scheduler_put(sched);
1273 /* decrease the module use count */
1274 ip_vs_use_count_dec();
1281 * Edit a service and bind it with a new scheduler
1284 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1286 struct ip_vs_scheduler *sched, *old_sched;
1287 struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1291 * Lookup the scheduler, by 'u->sched_name'
1293 sched = ip_vs_scheduler_get(u->sched_name);
1294 if (sched == NULL) {
1295 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1300 if (u->pe_name && *u->pe_name) {
1301 pe = ip_vs_pe_getbyname(u->pe_name);
1303 pr_info("persistence engine module ip_vs_pe_%s "
1304 "not found\n", u->pe_name);
1311 #ifdef CONFIG_IP_VS_IPV6
1312 if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1318 write_lock_bh(&__ip_vs_svc_lock);
1321 * Wait until all other svc users go away.
1323 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1326 * Set the flags and timeout value
1328 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1329 svc->timeout = u->timeout * HZ;
1330 svc->netmask = u->netmask;
1332 old_sched = svc->scheduler;
1333 if (sched != old_sched) {
1335 * Unbind the old scheduler
1337 ip_vs_unbind_scheduler(svc);
1340 * Bind the new scheduler
1342 if ((ret = ip_vs_bind_scheduler(svc, sched))) {
1344 * If ip_vs_bind_scheduler fails, restore the old
1346 * The main reason of failure is out of memory.
1348 * The question is if the old scheduler can be
1349 * restored all the time. TODO: if it cannot be
1350 * restored some time, we must delete the service,
1351 * otherwise the system may crash.
1353 ip_vs_bind_scheduler(svc, old_sched);
1361 ip_vs_unbind_pe(svc);
1362 ip_vs_bind_pe(svc, pe);
1366 write_unlock_bh(&__ip_vs_svc_lock);
1368 ip_vs_scheduler_put(old_sched);
1369 ip_vs_pe_put(old_pe);
1375 * Delete a service from the service list
1376 * - The service must be unlinked, unlocked and not referenced!
1377 * - We are called under _bh lock
1379 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1381 struct ip_vs_dest *dest, *nxt;
1382 struct ip_vs_scheduler *old_sched;
1383 struct ip_vs_pe *old_pe;
1384 struct netns_ipvs *ipvs = net_ipvs(svc->net);
1386 pr_info("%s: enter\n", __func__);
1388 /* Count only IPv4 services for old get/setsockopt interface */
1389 if (svc->af == AF_INET)
1390 ipvs->num_services--;
1392 ip_vs_stop_estimator(svc->net, &svc->stats);
1394 /* Unbind scheduler */
1395 old_sched = svc->scheduler;
1396 ip_vs_unbind_scheduler(svc);
1397 ip_vs_scheduler_put(old_sched);
1399 /* Unbind persistence engine */
1401 ip_vs_unbind_pe(svc);
1402 ip_vs_pe_put(old_pe);
1404 /* Unbind app inc */
1406 ip_vs_app_inc_put(svc->inc);
1411 * Unlink the whole destination list
1413 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1414 __ip_vs_unlink_dest(svc, dest, 0);
1415 __ip_vs_del_dest(svc->net, dest, cleanup);
1419 * Update the virtual service counters
1421 if (svc->port == FTPPORT)
1422 atomic_dec(&ipvs->ftpsvc_counter);
1423 else if (svc->port == 0)
1424 atomic_dec(&ipvs->nullsvc_counter);
1427 * Free the service if nobody refers to it
1429 if (atomic_read(&svc->refcnt) == 0) {
1430 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
1432 IP_VS_DBG_ADDR(svc->af, &svc->addr),
1433 ntohs(svc->port), atomic_read(&svc->usecnt));
1434 free_percpu(svc->stats.cpustats);
1438 /* decrease the module use count */
1439 ip_vs_use_count_dec();
1443 * Unlink a service from list and try to delete it if its refcnt reached 0
1445 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1448 * Unhash it from the service table
1450 write_lock_bh(&__ip_vs_svc_lock);
1452 ip_vs_svc_unhash(svc);
1455 * Wait until all the svc users go away.
1457 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1459 __ip_vs_del_service(svc, cleanup);
1461 write_unlock_bh(&__ip_vs_svc_lock);
1465 * Delete a service from the service list
1467 static int ip_vs_del_service(struct ip_vs_service *svc)
1471 ip_vs_unlink_service(svc, false);
1478 * Flush all the virtual services
1480 static int ip_vs_flush(struct net *net, bool cleanup)
1483 struct ip_vs_service *svc, *nxt;
1486 * Flush the service table hashed by <netns,protocol,addr,port>
1488 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1489 list_for_each_entry_safe(svc, nxt, &ip_vs_svc_table[idx],
1491 if (net_eq(svc->net, net))
1492 ip_vs_unlink_service(svc, cleanup);
1497 * Flush the service table hashed by fwmark
1499 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1500 list_for_each_entry_safe(svc, nxt,
1501 &ip_vs_svc_fwm_table[idx], f_list) {
1502 if (net_eq(svc->net, net))
1503 ip_vs_unlink_service(svc, cleanup);
1511 * Delete service by {netns} in the service table.
1512 * Called by __ip_vs_cleanup()
1514 void ip_vs_service_net_cleanup(struct net *net)
1517 /* Check for "full" addressed entries */
1518 mutex_lock(&__ip_vs_mutex);
1519 ip_vs_flush(net, true);
1520 mutex_unlock(&__ip_vs_mutex);
1524 /* Put all references for device (dst_cache) */
1526 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1528 spin_lock_bh(&dest->dst_lock);
1529 if (dest->dest_dst && dest->dest_dst->dst_cache->dev == dev) {
1530 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1532 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1534 atomic_read(&dest->refcnt));
1535 __ip_vs_dst_cache_reset(dest);
1537 spin_unlock_bh(&dest->dst_lock);
1540 /* Netdev event receiver
1541 * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1543 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1546 struct net_device *dev = ptr;
1547 struct net *net = dev_net(dev);
1548 struct netns_ipvs *ipvs = net_ipvs(net);
1549 struct ip_vs_service *svc;
1550 struct ip_vs_dest *dest;
1553 if (event != NETDEV_DOWN || !ipvs)
1555 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1557 mutex_lock(&__ip_vs_mutex);
1558 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1559 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1560 if (net_eq(svc->net, net)) {
1561 list_for_each_entry(dest, &svc->destinations,
1563 ip_vs_forget_dev(dest, dev);
1568 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1569 if (net_eq(svc->net, net)) {
1570 list_for_each_entry(dest, &svc->destinations,
1572 ip_vs_forget_dev(dest, dev);
1579 spin_lock_bh(&ipvs->dest_trash_lock);
1580 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1581 ip_vs_forget_dev(dest, dev);
1583 spin_unlock_bh(&ipvs->dest_trash_lock);
1584 mutex_unlock(&__ip_vs_mutex);
1590 * Zero counters in a service or all services
1592 static int ip_vs_zero_service(struct ip_vs_service *svc)
1594 struct ip_vs_dest *dest;
1596 write_lock_bh(&__ip_vs_svc_lock);
1597 list_for_each_entry(dest, &svc->destinations, n_list) {
1598 ip_vs_zero_stats(&dest->stats);
1600 ip_vs_zero_stats(&svc->stats);
1601 write_unlock_bh(&__ip_vs_svc_lock);
1605 static int ip_vs_zero_all(struct net *net)
1608 struct ip_vs_service *svc;
1610 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1611 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1612 if (net_eq(svc->net, net))
1613 ip_vs_zero_service(svc);
1617 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1618 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1619 if (net_eq(svc->net, net))
1620 ip_vs_zero_service(svc);
1624 ip_vs_zero_stats(&net_ipvs(net)->tot_stats);
1628 #ifdef CONFIG_SYSCTL
1631 static int three = 3;
1634 proc_do_defense_mode(ctl_table *table, int write,
1635 void __user *buffer, size_t *lenp, loff_t *ppos)
1637 struct net *net = current->nsproxy->net_ns;
1638 int *valp = table->data;
1642 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1643 if (write && (*valp != val)) {
1644 if ((*valp < 0) || (*valp > 3)) {
1645 /* Restore the correct value */
1648 update_defense_level(net_ipvs(net));
1655 proc_do_sync_threshold(ctl_table *table, int write,
1656 void __user *buffer, size_t *lenp, loff_t *ppos)
1658 int *valp = table->data;
1662 /* backup the value first */
1663 memcpy(val, valp, sizeof(val));
1665 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1666 if (write && (valp[0] < 0 || valp[1] < 0 ||
1667 (valp[0] >= valp[1] && valp[1]))) {
1668 /* Restore the correct value */
1669 memcpy(valp, val, sizeof(val));
1675 proc_do_sync_mode(ctl_table *table, int write,
1676 void __user *buffer, size_t *lenp, loff_t *ppos)
1678 int *valp = table->data;
1682 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1683 if (write && (*valp != val)) {
1684 if ((*valp < 0) || (*valp > 1)) {
1685 /* Restore the correct value */
1693 proc_do_sync_ports(ctl_table *table, int write,
1694 void __user *buffer, size_t *lenp, loff_t *ppos)
1696 int *valp = table->data;
1700 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1701 if (write && (*valp != val)) {
1702 if (*valp < 1 || !is_power_of_2(*valp)) {
1703 /* Restore the correct value */
1711 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1712 * Do not change order or insert new entries without
1713 * align with netns init in ip_vs_control_net_init()
1716 static struct ctl_table vs_vars[] = {
1718 .procname = "amemthresh",
1719 .maxlen = sizeof(int),
1721 .proc_handler = proc_dointvec,
1724 .procname = "am_droprate",
1725 .maxlen = sizeof(int),
1727 .proc_handler = proc_dointvec,
1730 .procname = "drop_entry",
1731 .maxlen = sizeof(int),
1733 .proc_handler = proc_do_defense_mode,
1736 .procname = "drop_packet",
1737 .maxlen = sizeof(int),
1739 .proc_handler = proc_do_defense_mode,
1741 #ifdef CONFIG_IP_VS_NFCT
1743 .procname = "conntrack",
1744 .maxlen = sizeof(int),
1746 .proc_handler = &proc_dointvec,
1750 .procname = "secure_tcp",
1751 .maxlen = sizeof(int),
1753 .proc_handler = proc_do_defense_mode,
1756 .procname = "snat_reroute",
1757 .maxlen = sizeof(int),
1759 .proc_handler = &proc_dointvec,
1762 .procname = "sync_version",
1763 .maxlen = sizeof(int),
1765 .proc_handler = &proc_do_sync_mode,
1768 .procname = "sync_ports",
1769 .maxlen = sizeof(int),
1771 .proc_handler = &proc_do_sync_ports,
1774 .procname = "sync_qlen_max",
1775 .maxlen = sizeof(int),
1777 .proc_handler = proc_dointvec,
1780 .procname = "sync_sock_size",
1781 .maxlen = sizeof(int),
1783 .proc_handler = proc_dointvec,
1786 .procname = "cache_bypass",
1787 .maxlen = sizeof(int),
1789 .proc_handler = proc_dointvec,
1792 .procname = "expire_nodest_conn",
1793 .maxlen = sizeof(int),
1795 .proc_handler = proc_dointvec,
1798 .procname = "expire_quiescent_template",
1799 .maxlen = sizeof(int),
1801 .proc_handler = proc_dointvec,
1804 .procname = "sync_threshold",
1806 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1808 .proc_handler = proc_do_sync_threshold,
1811 .procname = "sync_refresh_period",
1812 .maxlen = sizeof(int),
1814 .proc_handler = proc_dointvec_jiffies,
1817 .procname = "sync_retries",
1818 .maxlen = sizeof(int),
1820 .proc_handler = proc_dointvec_minmax,
1825 .procname = "nat_icmp_send",
1826 .maxlen = sizeof(int),
1828 .proc_handler = proc_dointvec,
1831 .procname = "pmtu_disc",
1832 .maxlen = sizeof(int),
1834 .proc_handler = proc_dointvec,
1837 .procname = "backup_only",
1838 .maxlen = sizeof(int),
1840 .proc_handler = proc_dointvec,
1842 #ifdef CONFIG_IP_VS_DEBUG
1844 .procname = "debug_level",
1845 .data = &sysctl_ip_vs_debug_level,
1846 .maxlen = sizeof(int),
1848 .proc_handler = proc_dointvec,
1853 .procname = "timeout_established",
1854 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1855 .maxlen = sizeof(int),
1857 .proc_handler = proc_dointvec_jiffies,
1860 .procname = "timeout_synsent",
1861 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1862 .maxlen = sizeof(int),
1864 .proc_handler = proc_dointvec_jiffies,
1867 .procname = "timeout_synrecv",
1868 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1869 .maxlen = sizeof(int),
1871 .proc_handler = proc_dointvec_jiffies,
1874 .procname = "timeout_finwait",
1875 .data = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1876 .maxlen = sizeof(int),
1878 .proc_handler = proc_dointvec_jiffies,
1881 .procname = "timeout_timewait",
1882 .data = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1883 .maxlen = sizeof(int),
1885 .proc_handler = proc_dointvec_jiffies,
1888 .procname = "timeout_close",
1889 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1890 .maxlen = sizeof(int),
1892 .proc_handler = proc_dointvec_jiffies,
1895 .procname = "timeout_closewait",
1896 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1897 .maxlen = sizeof(int),
1899 .proc_handler = proc_dointvec_jiffies,
1902 .procname = "timeout_lastack",
1903 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1904 .maxlen = sizeof(int),
1906 .proc_handler = proc_dointvec_jiffies,
1909 .procname = "timeout_listen",
1910 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1911 .maxlen = sizeof(int),
1913 .proc_handler = proc_dointvec_jiffies,
1916 .procname = "timeout_synack",
1917 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1918 .maxlen = sizeof(int),
1920 .proc_handler = proc_dointvec_jiffies,
1923 .procname = "timeout_udp",
1924 .data = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1925 .maxlen = sizeof(int),
1927 .proc_handler = proc_dointvec_jiffies,
1930 .procname = "timeout_icmp",
1931 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1932 .maxlen = sizeof(int),
1934 .proc_handler = proc_dointvec_jiffies,
1942 #ifdef CONFIG_PROC_FS
1945 struct seq_net_private p; /* Do not move this, netns depends upon it*/
1946 struct list_head *table;
1951 * Write the contents of the VS rule table to a PROCfs file.
1952 * (It is kept just for backward compatibility)
1954 static inline const char *ip_vs_fwd_name(unsigned int flags)
1956 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1957 case IP_VS_CONN_F_LOCALNODE:
1959 case IP_VS_CONN_F_TUNNEL:
1961 case IP_VS_CONN_F_DROUTE:
1969 /* Get the Nth entry in the two lists */
1970 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1972 struct net *net = seq_file_net(seq);
1973 struct ip_vs_iter *iter = seq->private;
1975 struct ip_vs_service *svc;
1977 /* look in hash by protocol */
1978 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1979 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1980 if (net_eq(svc->net, net) && pos-- == 0) {
1981 iter->table = ip_vs_svc_table;
1988 /* keep looking in fwmark */
1989 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1990 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1991 if (net_eq(svc->net, net) && pos-- == 0) {
1992 iter->table = ip_vs_svc_fwm_table;
2002 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
2003 __acquires(__ip_vs_svc_lock)
2006 read_lock_bh(&__ip_vs_svc_lock);
2007 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
2011 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2013 struct list_head *e;
2014 struct ip_vs_iter *iter;
2015 struct ip_vs_service *svc;
2018 if (v == SEQ_START_TOKEN)
2019 return ip_vs_info_array(seq,0);
2022 iter = seq->private;
2024 if (iter->table == ip_vs_svc_table) {
2025 /* next service in table hashed by protocol */
2026 if ((e = svc->s_list.next) != &ip_vs_svc_table[iter->bucket])
2027 return list_entry(e, struct ip_vs_service, s_list);
2030 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2031 list_for_each_entry(svc,&ip_vs_svc_table[iter->bucket],
2037 iter->table = ip_vs_svc_fwm_table;
2042 /* next service in hashed by fwmark */
2043 if ((e = svc->f_list.next) != &ip_vs_svc_fwm_table[iter->bucket])
2044 return list_entry(e, struct ip_vs_service, f_list);
2047 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2048 list_for_each_entry(svc, &ip_vs_svc_fwm_table[iter->bucket],
2056 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
2057 __releases(__ip_vs_svc_lock)
2059 read_unlock_bh(&__ip_vs_svc_lock);
2063 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
2065 if (v == SEQ_START_TOKEN) {
2067 "IP Virtual Server version %d.%d.%d (size=%d)\n",
2068 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2070 "Prot LocalAddress:Port Scheduler Flags\n");
2072 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2074 const struct ip_vs_service *svc = v;
2075 const struct ip_vs_iter *iter = seq->private;
2076 const struct ip_vs_dest *dest;
2078 if (iter->table == ip_vs_svc_table) {
2079 #ifdef CONFIG_IP_VS_IPV6
2080 if (svc->af == AF_INET6)
2081 seq_printf(seq, "%s [%pI6]:%04X %s ",
2082 ip_vs_proto_name(svc->protocol),
2085 svc->scheduler->name);
2088 seq_printf(seq, "%s %08X:%04X %s %s ",
2089 ip_vs_proto_name(svc->protocol),
2090 ntohl(svc->addr.ip),
2092 svc->scheduler->name,
2093 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2095 seq_printf(seq, "FWM %08X %s %s",
2096 svc->fwmark, svc->scheduler->name,
2097 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2100 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2101 seq_printf(seq, "persistent %d %08X\n",
2103 ntohl(svc->netmask));
2105 seq_putc(seq, '\n');
2107 list_for_each_entry(dest, &svc->destinations, n_list) {
2108 #ifdef CONFIG_IP_VS_IPV6
2109 if (dest->af == AF_INET6)
2112 " %-7s %-6d %-10d %-10d\n",
2115 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2116 atomic_read(&dest->weight),
2117 atomic_read(&dest->activeconns),
2118 atomic_read(&dest->inactconns));
2123 "%-7s %-6d %-10d %-10d\n",
2124 ntohl(dest->addr.ip),
2126 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2127 atomic_read(&dest->weight),
2128 atomic_read(&dest->activeconns),
2129 atomic_read(&dest->inactconns));
2136 static const struct seq_operations ip_vs_info_seq_ops = {
2137 .start = ip_vs_info_seq_start,
2138 .next = ip_vs_info_seq_next,
2139 .stop = ip_vs_info_seq_stop,
2140 .show = ip_vs_info_seq_show,
2143 static int ip_vs_info_open(struct inode *inode, struct file *file)
2145 return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2146 sizeof(struct ip_vs_iter));
2149 static const struct file_operations ip_vs_info_fops = {
2150 .owner = THIS_MODULE,
2151 .open = ip_vs_info_open,
2153 .llseek = seq_lseek,
2154 .release = seq_release_net,
2157 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2159 struct net *net = seq_file_single_net(seq);
2160 struct ip_vs_stats_user show;
2162 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2164 " Total Incoming Outgoing Incoming Outgoing\n");
2166 " Conns Packets Packets Bytes Bytes\n");
2168 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2169 seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", show.conns,
2170 show.inpkts, show.outpkts,
2171 (unsigned long long) show.inbytes,
2172 (unsigned long long) show.outbytes);
2174 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2176 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2177 seq_printf(seq, "%8X %8X %8X %16X %16X\n",
2178 show.cps, show.inpps, show.outpps,
2179 show.inbps, show.outbps);
2184 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2186 return single_open_net(inode, file, ip_vs_stats_show);
2189 static const struct file_operations ip_vs_stats_fops = {
2190 .owner = THIS_MODULE,
2191 .open = ip_vs_stats_seq_open,
2193 .llseek = seq_lseek,
2194 .release = single_release_net,
2197 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2199 struct net *net = seq_file_single_net(seq);
2200 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2201 struct ip_vs_cpu_stats *cpustats = tot_stats->cpustats;
2202 struct ip_vs_stats_user rates;
2205 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2207 " Total Incoming Outgoing Incoming Outgoing\n");
2209 "CPU Conns Packets Packets Bytes Bytes\n");
2211 for_each_possible_cpu(i) {
2212 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2214 __u64 inbytes, outbytes;
2217 start = u64_stats_fetch_begin_bh(&u->syncp);
2218 inbytes = u->ustats.inbytes;
2219 outbytes = u->ustats.outbytes;
2220 } while (u64_stats_fetch_retry_bh(&u->syncp, start));
2222 seq_printf(seq, "%3X %8X %8X %8X %16LX %16LX\n",
2223 i, u->ustats.conns, u->ustats.inpkts,
2224 u->ustats.outpkts, (__u64)inbytes,
2228 spin_lock_bh(&tot_stats->lock);
2230 seq_printf(seq, " ~ %8X %8X %8X %16LX %16LX\n\n",
2231 tot_stats->ustats.conns, tot_stats->ustats.inpkts,
2232 tot_stats->ustats.outpkts,
2233 (unsigned long long) tot_stats->ustats.inbytes,
2234 (unsigned long long) tot_stats->ustats.outbytes);
2236 ip_vs_read_estimator(&rates, tot_stats);
2238 spin_unlock_bh(&tot_stats->lock);
2240 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2242 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2243 seq_printf(seq, " %8X %8X %8X %16X %16X\n",
2253 static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2255 return single_open_net(inode, file, ip_vs_stats_percpu_show);
2258 static const struct file_operations ip_vs_stats_percpu_fops = {
2259 .owner = THIS_MODULE,
2260 .open = ip_vs_stats_percpu_seq_open,
2262 .llseek = seq_lseek,
2263 .release = single_release_net,
2268 * Set timeout values for tcp tcpfin udp in the timeout_table.
2270 static int ip_vs_set_timeout(struct net *net, struct ip_vs_timeout_user *u)
2272 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2273 struct ip_vs_proto_data *pd;
2276 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2281 #ifdef CONFIG_IP_VS_PROTO_TCP
2282 if (u->tcp_timeout) {
2283 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2284 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2285 = u->tcp_timeout * HZ;
2288 if (u->tcp_fin_timeout) {
2289 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2290 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2291 = u->tcp_fin_timeout * HZ;
2295 #ifdef CONFIG_IP_VS_PROTO_UDP
2296 if (u->udp_timeout) {
2297 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2298 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2299 = u->udp_timeout * HZ;
2306 #define SET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2307 #define SERVICE_ARG_LEN (sizeof(struct ip_vs_service_user))
2308 #define SVCDEST_ARG_LEN (sizeof(struct ip_vs_service_user) + \
2309 sizeof(struct ip_vs_dest_user))
2310 #define TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2311 #define DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user))
2312 #define MAX_ARG_LEN SVCDEST_ARG_LEN
2314 static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
2315 [SET_CMDID(IP_VS_SO_SET_ADD)] = SERVICE_ARG_LEN,
2316 [SET_CMDID(IP_VS_SO_SET_EDIT)] = SERVICE_ARG_LEN,
2317 [SET_CMDID(IP_VS_SO_SET_DEL)] = SERVICE_ARG_LEN,
2318 [SET_CMDID(IP_VS_SO_SET_FLUSH)] = 0,
2319 [SET_CMDID(IP_VS_SO_SET_ADDDEST)] = SVCDEST_ARG_LEN,
2320 [SET_CMDID(IP_VS_SO_SET_DELDEST)] = SVCDEST_ARG_LEN,
2321 [SET_CMDID(IP_VS_SO_SET_EDITDEST)] = SVCDEST_ARG_LEN,
2322 [SET_CMDID(IP_VS_SO_SET_TIMEOUT)] = TIMEOUT_ARG_LEN,
2323 [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)] = DAEMON_ARG_LEN,
2324 [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)] = DAEMON_ARG_LEN,
2325 [SET_CMDID(IP_VS_SO_SET_ZERO)] = SERVICE_ARG_LEN,
2328 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2329 struct ip_vs_service_user *usvc_compat)
2331 memset(usvc, 0, sizeof(*usvc));
2334 usvc->protocol = usvc_compat->protocol;
2335 usvc->addr.ip = usvc_compat->addr;
2336 usvc->port = usvc_compat->port;
2337 usvc->fwmark = usvc_compat->fwmark;
2339 /* Deep copy of sched_name is not needed here */
2340 usvc->sched_name = usvc_compat->sched_name;
2342 usvc->flags = usvc_compat->flags;
2343 usvc->timeout = usvc_compat->timeout;
2344 usvc->netmask = usvc_compat->netmask;
2347 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2348 struct ip_vs_dest_user *udest_compat)
2350 memset(udest, 0, sizeof(*udest));
2352 udest->addr.ip = udest_compat->addr;
2353 udest->port = udest_compat->port;
2354 udest->conn_flags = udest_compat->conn_flags;
2355 udest->weight = udest_compat->weight;
2356 udest->u_threshold = udest_compat->u_threshold;
2357 udest->l_threshold = udest_compat->l_threshold;
2361 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2363 struct net *net = sock_net(sk);
2365 unsigned char arg[MAX_ARG_LEN];
2366 struct ip_vs_service_user *usvc_compat;
2367 struct ip_vs_service_user_kern usvc;
2368 struct ip_vs_service *svc;
2369 struct ip_vs_dest_user *udest_compat;
2370 struct ip_vs_dest_user_kern udest;
2371 struct netns_ipvs *ipvs = net_ipvs(net);
2373 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2376 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2378 if (len < 0 || len > MAX_ARG_LEN)
2380 if (len != set_arglen[SET_CMDID(cmd)]) {
2381 pr_err("set_ctl: len %u != %u\n",
2382 len, set_arglen[SET_CMDID(cmd)]);
2386 if (copy_from_user(arg, user, len) != 0)
2389 /* increase the module use count */
2390 ip_vs_use_count_inc();
2392 /* Handle daemons since they have another lock */
2393 if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2394 cmd == IP_VS_SO_SET_STOPDAEMON) {
2395 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2397 if (mutex_lock_interruptible(&ipvs->sync_mutex)) {
2401 if (cmd == IP_VS_SO_SET_STARTDAEMON)
2402 ret = start_sync_thread(net, dm->state, dm->mcast_ifn,
2405 ret = stop_sync_thread(net, dm->state);
2406 mutex_unlock(&ipvs->sync_mutex);
2410 if (mutex_lock_interruptible(&__ip_vs_mutex)) {
2415 if (cmd == IP_VS_SO_SET_FLUSH) {
2416 /* Flush the virtual service */
2417 ret = ip_vs_flush(net, false);
2419 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2420 /* Set timeout values for (tcp tcpfin udp) */
2421 ret = ip_vs_set_timeout(net, (struct ip_vs_timeout_user *)arg);
2425 usvc_compat = (struct ip_vs_service_user *)arg;
2426 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2428 /* We only use the new structs internally, so copy userspace compat
2429 * structs to extended internal versions */
2430 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2431 ip_vs_copy_udest_compat(&udest, udest_compat);
2433 if (cmd == IP_VS_SO_SET_ZERO) {
2434 /* if no service address is set, zero counters in all */
2435 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2436 ret = ip_vs_zero_all(net);
2441 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2442 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2443 usvc.protocol != IPPROTO_SCTP) {
2444 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2445 usvc.protocol, &usvc.addr.ip,
2446 ntohs(usvc.port), usvc.sched_name);
2451 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2452 if (usvc.fwmark == 0)
2453 svc = __ip_vs_service_find(net, usvc.af, usvc.protocol,
2454 &usvc.addr, usvc.port);
2456 svc = __ip_vs_svc_fwm_find(net, usvc.af, usvc.fwmark);
2458 if (cmd != IP_VS_SO_SET_ADD
2459 && (svc == NULL || svc->protocol != usvc.protocol)) {
2465 case IP_VS_SO_SET_ADD:
2469 ret = ip_vs_add_service(net, &usvc, &svc);
2471 case IP_VS_SO_SET_EDIT:
2472 ret = ip_vs_edit_service(svc, &usvc);
2474 case IP_VS_SO_SET_DEL:
2475 ret = ip_vs_del_service(svc);
2479 case IP_VS_SO_SET_ZERO:
2480 ret = ip_vs_zero_service(svc);
2482 case IP_VS_SO_SET_ADDDEST:
2483 ret = ip_vs_add_dest(svc, &udest);
2485 case IP_VS_SO_SET_EDITDEST:
2486 ret = ip_vs_edit_dest(svc, &udest);
2488 case IP_VS_SO_SET_DELDEST:
2489 ret = ip_vs_del_dest(svc, &udest);
2496 mutex_unlock(&__ip_vs_mutex);
2498 /* decrease the module use count */
2499 ip_vs_use_count_dec();
2506 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2508 dst->protocol = src->protocol;
2509 dst->addr = src->addr.ip;
2510 dst->port = src->port;
2511 dst->fwmark = src->fwmark;
2512 strlcpy(dst->sched_name, src->scheduler->name, sizeof(dst->sched_name));
2513 dst->flags = src->flags;
2514 dst->timeout = src->timeout / HZ;
2515 dst->netmask = src->netmask;
2516 dst->num_dests = src->num_dests;
2517 ip_vs_copy_stats(&dst->stats, &src->stats);
2521 __ip_vs_get_service_entries(struct net *net,
2522 const struct ip_vs_get_services *get,
2523 struct ip_vs_get_services __user *uptr)
2526 struct ip_vs_service *svc;
2527 struct ip_vs_service_entry entry;
2530 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2531 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2532 /* Only expose IPv4 entries to old interface */
2533 if (svc->af != AF_INET || !net_eq(svc->net, net))
2536 if (count >= get->num_services)
2538 memset(&entry, 0, sizeof(entry));
2539 ip_vs_copy_service(&entry, svc);
2540 if (copy_to_user(&uptr->entrytable[count],
2541 &entry, sizeof(entry))) {
2549 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2550 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2551 /* Only expose IPv4 entries to old interface */
2552 if (svc->af != AF_INET || !net_eq(svc->net, net))
2555 if (count >= get->num_services)
2557 memset(&entry, 0, sizeof(entry));
2558 ip_vs_copy_service(&entry, svc);
2559 if (copy_to_user(&uptr->entrytable[count],
2560 &entry, sizeof(entry))) {
2572 __ip_vs_get_dest_entries(struct net *net, const struct ip_vs_get_dests *get,
2573 struct ip_vs_get_dests __user *uptr)
2575 struct ip_vs_service *svc;
2576 union nf_inet_addr addr = { .ip = get->addr };
2580 svc = __ip_vs_svc_fwm_find(net, AF_INET, get->fwmark);
2582 svc = __ip_vs_service_find(net, AF_INET, get->protocol, &addr,
2587 struct ip_vs_dest *dest;
2588 struct ip_vs_dest_entry entry;
2590 list_for_each_entry(dest, &svc->destinations, n_list) {
2591 if (count >= get->num_dests)
2594 entry.addr = dest->addr.ip;
2595 entry.port = dest->port;
2596 entry.conn_flags = atomic_read(&dest->conn_flags);
2597 entry.weight = atomic_read(&dest->weight);
2598 entry.u_threshold = dest->u_threshold;
2599 entry.l_threshold = dest->l_threshold;
2600 entry.activeconns = atomic_read(&dest->activeconns);
2601 entry.inactconns = atomic_read(&dest->inactconns);
2602 entry.persistconns = atomic_read(&dest->persistconns);
2603 ip_vs_copy_stats(&entry.stats, &dest->stats);
2604 if (copy_to_user(&uptr->entrytable[count],
2605 &entry, sizeof(entry))) {
2617 __ip_vs_get_timeouts(struct net *net, struct ip_vs_timeout_user *u)
2619 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2620 struct ip_vs_proto_data *pd;
2623 memset(u, 0, sizeof (*u));
2625 #ifdef CONFIG_IP_VS_PROTO_TCP
2626 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2627 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2628 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2630 #ifdef CONFIG_IP_VS_PROTO_UDP
2631 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2633 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2638 #define GET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2639 #define GET_INFO_ARG_LEN (sizeof(struct ip_vs_getinfo))
2640 #define GET_SERVICES_ARG_LEN (sizeof(struct ip_vs_get_services))
2641 #define GET_SERVICE_ARG_LEN (sizeof(struct ip_vs_service_entry))
2642 #define GET_DESTS_ARG_LEN (sizeof(struct ip_vs_get_dests))
2643 #define GET_TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2644 #define GET_DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user) * 2)
2646 static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
2647 [GET_CMDID(IP_VS_SO_GET_VERSION)] = 64,
2648 [GET_CMDID(IP_VS_SO_GET_INFO)] = GET_INFO_ARG_LEN,
2649 [GET_CMDID(IP_VS_SO_GET_SERVICES)] = GET_SERVICES_ARG_LEN,
2650 [GET_CMDID(IP_VS_SO_GET_SERVICE)] = GET_SERVICE_ARG_LEN,
2651 [GET_CMDID(IP_VS_SO_GET_DESTS)] = GET_DESTS_ARG_LEN,
2652 [GET_CMDID(IP_VS_SO_GET_TIMEOUT)] = GET_TIMEOUT_ARG_LEN,
2653 [GET_CMDID(IP_VS_SO_GET_DAEMON)] = GET_DAEMON_ARG_LEN,
2657 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2659 unsigned char arg[128];
2661 unsigned int copylen;
2662 struct net *net = sock_net(sk);
2663 struct netns_ipvs *ipvs = net_ipvs(net);
2666 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2669 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2672 if (*len < get_arglen[GET_CMDID(cmd)]) {
2673 pr_err("get_ctl: len %u < %u\n",
2674 *len, get_arglen[GET_CMDID(cmd)]);
2678 copylen = get_arglen[GET_CMDID(cmd)];
2682 if (copy_from_user(arg, user, copylen) != 0)
2685 * Handle daemons first since it has its own locking
2687 if (cmd == IP_VS_SO_GET_DAEMON) {
2688 struct ip_vs_daemon_user d[2];
2690 memset(&d, 0, sizeof(d));
2691 if (mutex_lock_interruptible(&ipvs->sync_mutex))
2692 return -ERESTARTSYS;
2694 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2695 d[0].state = IP_VS_STATE_MASTER;
2696 strlcpy(d[0].mcast_ifn, ipvs->master_mcast_ifn,
2697 sizeof(d[0].mcast_ifn));
2698 d[0].syncid = ipvs->master_syncid;
2700 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2701 d[1].state = IP_VS_STATE_BACKUP;
2702 strlcpy(d[1].mcast_ifn, ipvs->backup_mcast_ifn,
2703 sizeof(d[1].mcast_ifn));
2704 d[1].syncid = ipvs->backup_syncid;
2706 if (copy_to_user(user, &d, sizeof(d)) != 0)
2708 mutex_unlock(&ipvs->sync_mutex);
2712 if (mutex_lock_interruptible(&__ip_vs_mutex))
2713 return -ERESTARTSYS;
2716 case IP_VS_SO_GET_VERSION:
2720 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2721 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2722 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2726 *len = strlen(buf)+1;
2730 case IP_VS_SO_GET_INFO:
2732 struct ip_vs_getinfo info;
2733 info.version = IP_VS_VERSION_CODE;
2734 info.size = ip_vs_conn_tab_size;
2735 info.num_services = ipvs->num_services;
2736 if (copy_to_user(user, &info, sizeof(info)) != 0)
2741 case IP_VS_SO_GET_SERVICES:
2743 struct ip_vs_get_services *get;
2746 get = (struct ip_vs_get_services *)arg;
2747 size = sizeof(*get) +
2748 sizeof(struct ip_vs_service_entry) * get->num_services;
2750 pr_err("length: %u != %u\n", *len, size);
2754 ret = __ip_vs_get_service_entries(net, get, user);
2758 case IP_VS_SO_GET_SERVICE:
2760 struct ip_vs_service_entry *entry;
2761 struct ip_vs_service *svc;
2762 union nf_inet_addr addr;
2764 entry = (struct ip_vs_service_entry *)arg;
2765 addr.ip = entry->addr;
2767 svc = __ip_vs_svc_fwm_find(net, AF_INET, entry->fwmark);
2769 svc = __ip_vs_service_find(net, AF_INET,
2770 entry->protocol, &addr,
2773 ip_vs_copy_service(entry, svc);
2774 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2781 case IP_VS_SO_GET_DESTS:
2783 struct ip_vs_get_dests *get;
2786 get = (struct ip_vs_get_dests *)arg;
2787 size = sizeof(*get) +
2788 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2790 pr_err("length: %u != %u\n", *len, size);
2794 ret = __ip_vs_get_dest_entries(net, get, user);
2798 case IP_VS_SO_GET_TIMEOUT:
2800 struct ip_vs_timeout_user t;
2802 __ip_vs_get_timeouts(net, &t);
2803 if (copy_to_user(user, &t, sizeof(t)) != 0)
2813 mutex_unlock(&__ip_vs_mutex);
2818 static struct nf_sockopt_ops ip_vs_sockopts = {
2820 .set_optmin = IP_VS_BASE_CTL,
2821 .set_optmax = IP_VS_SO_SET_MAX+1,
2822 .set = do_ip_vs_set_ctl,
2823 .get_optmin = IP_VS_BASE_CTL,
2824 .get_optmax = IP_VS_SO_GET_MAX+1,
2825 .get = do_ip_vs_get_ctl,
2826 .owner = THIS_MODULE,
2830 * Generic Netlink interface
2833 /* IPVS genetlink family */
2834 static struct genl_family ip_vs_genl_family = {
2835 .id = GENL_ID_GENERATE,
2837 .name = IPVS_GENL_NAME,
2838 .version = IPVS_GENL_VERSION,
2839 .maxattr = IPVS_CMD_MAX,
2840 .netnsok = true, /* Make ipvsadm to work on netns */
2843 /* Policy used for first-level command attributes */
2844 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2845 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2846 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2847 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2848 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2849 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2850 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2853 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2854 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2855 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2856 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2857 .len = IP_VS_IFNAME_MAXLEN },
2858 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2861 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2862 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2863 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2864 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2865 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2866 .len = sizeof(union nf_inet_addr) },
2867 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2868 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2869 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2870 .len = IP_VS_SCHEDNAME_MAXLEN },
2871 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2872 .len = IP_VS_PENAME_MAXLEN },
2873 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2874 .len = sizeof(struct ip_vs_flags) },
2875 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2876 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2877 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2880 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2881 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2882 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2883 .len = sizeof(union nf_inet_addr) },
2884 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2885 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2886 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2887 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2888 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2889 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2890 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2891 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2892 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2895 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2896 struct ip_vs_stats *stats)
2898 struct ip_vs_stats_user ustats;
2899 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2903 ip_vs_copy_stats(&ustats, stats);
2905 if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, ustats.conns) ||
2906 nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, ustats.inpkts) ||
2907 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, ustats.outpkts) ||
2908 nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, ustats.inbytes) ||
2909 nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, ustats.outbytes) ||
2910 nla_put_u32(skb, IPVS_STATS_ATTR_CPS, ustats.cps) ||
2911 nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, ustats.inpps) ||
2912 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, ustats.outpps) ||
2913 nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, ustats.inbps) ||
2914 nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, ustats.outbps))
2915 goto nla_put_failure;
2916 nla_nest_end(skb, nl_stats);
2921 nla_nest_cancel(skb, nl_stats);
2925 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2926 struct ip_vs_service *svc)
2928 struct nlattr *nl_service;
2929 struct ip_vs_flags flags = { .flags = svc->flags,
2932 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2936 if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2937 goto nla_put_failure;
2939 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2940 goto nla_put_failure;
2942 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2943 nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2944 nla_put_u16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2945 goto nla_put_failure;
2948 if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, svc->scheduler->name) ||
2950 nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, svc->pe->name)) ||
2951 nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2952 nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2953 nla_put_u32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
2954 goto nla_put_failure;
2955 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2956 goto nla_put_failure;
2958 nla_nest_end(skb, nl_service);
2963 nla_nest_cancel(skb, nl_service);
2967 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2968 struct ip_vs_service *svc,
2969 struct netlink_callback *cb)
2973 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2974 &ip_vs_genl_family, NLM_F_MULTI,
2975 IPVS_CMD_NEW_SERVICE);
2979 if (ip_vs_genl_fill_service(skb, svc) < 0)
2980 goto nla_put_failure;
2982 return genlmsg_end(skb, hdr);
2985 genlmsg_cancel(skb, hdr);
2989 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2990 struct netlink_callback *cb)
2993 int start = cb->args[0];
2994 struct ip_vs_service *svc;
2995 struct net *net = skb_sknet(skb);
2997 mutex_lock(&__ip_vs_mutex);
2998 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2999 list_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
3000 if (++idx <= start || !net_eq(svc->net, net))
3002 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3004 goto nla_put_failure;
3009 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3010 list_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
3011 if (++idx <= start || !net_eq(svc->net, net))
3013 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3015 goto nla_put_failure;
3021 mutex_unlock(&__ip_vs_mutex);
3027 static int ip_vs_genl_parse_service(struct net *net,
3028 struct ip_vs_service_user_kern *usvc,
3029 struct nlattr *nla, int full_entry,
3030 struct ip_vs_service **ret_svc)
3032 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
3033 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
3034 struct ip_vs_service *svc;
3036 /* Parse mandatory identifying service fields first */
3038 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
3041 nla_af = attrs[IPVS_SVC_ATTR_AF];
3042 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
3043 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
3044 nla_port = attrs[IPVS_SVC_ATTR_PORT];
3045 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
3047 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3050 memset(usvc, 0, sizeof(*usvc));
3052 usvc->af = nla_get_u16(nla_af);
3053 #ifdef CONFIG_IP_VS_IPV6
3054 if (usvc->af != AF_INET && usvc->af != AF_INET6)
3056 if (usvc->af != AF_INET)
3058 return -EAFNOSUPPORT;
3061 usvc->protocol = IPPROTO_TCP;
3062 usvc->fwmark = nla_get_u32(nla_fwmark);
3064 usvc->protocol = nla_get_u16(nla_protocol);
3065 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3066 usvc->port = nla_get_u16(nla_port);
3071 svc = __ip_vs_svc_fwm_find(net, usvc->af, usvc->fwmark);
3073 svc = __ip_vs_service_find(net, usvc->af, usvc->protocol,
3074 &usvc->addr, usvc->port);
3077 /* If a full entry was requested, check for the additional fields */
3079 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3081 struct ip_vs_flags flags;
3083 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3084 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3085 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3086 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3087 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3089 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3092 nla_memcpy(&flags, nla_flags, sizeof(flags));
3094 /* prefill flags from service if it already exists */
3096 usvc->flags = svc->flags;
3098 /* set new flags from userland */
3099 usvc->flags = (usvc->flags & ~flags.mask) |
3100 (flags.flags & flags.mask);
3101 usvc->sched_name = nla_data(nla_sched);
3102 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3103 usvc->timeout = nla_get_u32(nla_timeout);
3104 usvc->netmask = nla_get_u32(nla_netmask);
3110 static struct ip_vs_service *ip_vs_genl_find_service(struct net *net,
3113 struct ip_vs_service_user_kern usvc;
3114 struct ip_vs_service *svc;
3117 ret = ip_vs_genl_parse_service(net, &usvc, nla, 0, &svc);
3118 return ret ? ERR_PTR(ret) : svc;
3121 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3123 struct nlattr *nl_dest;
3125 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3129 if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3130 nla_put_u16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3131 nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3132 (atomic_read(&dest->conn_flags) &
3133 IP_VS_CONN_F_FWD_MASK)) ||
3134 nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3135 atomic_read(&dest->weight)) ||
3136 nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3137 nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3138 nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3139 atomic_read(&dest->activeconns)) ||
3140 nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3141 atomic_read(&dest->inactconns)) ||
3142 nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3143 atomic_read(&dest->persistconns)))
3144 goto nla_put_failure;
3145 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
3146 goto nla_put_failure;
3148 nla_nest_end(skb, nl_dest);
3153 nla_nest_cancel(skb, nl_dest);
3157 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3158 struct netlink_callback *cb)
3162 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3163 &ip_vs_genl_family, NLM_F_MULTI,
3168 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3169 goto nla_put_failure;
3171 return genlmsg_end(skb, hdr);
3174 genlmsg_cancel(skb, hdr);
3178 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3179 struct netlink_callback *cb)
3182 int start = cb->args[0];
3183 struct ip_vs_service *svc;
3184 struct ip_vs_dest *dest;
3185 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3186 struct net *net = skb_sknet(skb);
3188 mutex_lock(&__ip_vs_mutex);
3190 /* Try to find the service for which to dump destinations */
3191 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3192 IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3196 svc = ip_vs_genl_find_service(net, attrs[IPVS_CMD_ATTR_SERVICE]);
3197 if (IS_ERR(svc) || svc == NULL)
3200 /* Dump the destinations */
3201 list_for_each_entry(dest, &svc->destinations, n_list) {
3204 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3206 goto nla_put_failure;
3214 mutex_unlock(&__ip_vs_mutex);
3219 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3220 struct nlattr *nla, int full_entry)
3222 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3223 struct nlattr *nla_addr, *nla_port;
3225 /* Parse mandatory identifying destination fields first */
3227 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3230 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3231 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3233 if (!(nla_addr && nla_port))
3236 memset(udest, 0, sizeof(*udest));
3238 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3239 udest->port = nla_get_u16(nla_port);
3241 /* If a full entry was requested, check for the additional fields */
3243 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3246 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3247 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3248 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3249 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3251 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3254 udest->conn_flags = nla_get_u32(nla_fwd)
3255 & IP_VS_CONN_F_FWD_MASK;
3256 udest->weight = nla_get_u32(nla_weight);
3257 udest->u_threshold = nla_get_u32(nla_u_thresh);
3258 udest->l_threshold = nla_get_u32(nla_l_thresh);
3264 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __be32 state,
3265 const char *mcast_ifn, __be32 syncid)
3267 struct nlattr *nl_daemon;
3269 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3273 if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3274 nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn) ||
3275 nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid))
3276 goto nla_put_failure;
3277 nla_nest_end(skb, nl_daemon);
3282 nla_nest_cancel(skb, nl_daemon);
3286 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __be32 state,
3287 const char *mcast_ifn, __be32 syncid,
3288 struct netlink_callback *cb)
3291 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3292 &ip_vs_genl_family, NLM_F_MULTI,
3293 IPVS_CMD_NEW_DAEMON);
3297 if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
3298 goto nla_put_failure;
3300 return genlmsg_end(skb, hdr);
3303 genlmsg_cancel(skb, hdr);
3307 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3308 struct netlink_callback *cb)
3310 struct net *net = skb_sknet(skb);
3311 struct netns_ipvs *ipvs = net_ipvs(net);
3313 mutex_lock(&ipvs->sync_mutex);
3314 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3315 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3316 ipvs->master_mcast_ifn,
3317 ipvs->master_syncid, cb) < 0)
3318 goto nla_put_failure;
3323 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3324 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3325 ipvs->backup_mcast_ifn,
3326 ipvs->backup_syncid, cb) < 0)
3327 goto nla_put_failure;
3333 mutex_unlock(&ipvs->sync_mutex);
3338 static int ip_vs_genl_new_daemon(struct net *net, struct nlattr **attrs)
3340 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3341 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3342 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3345 return start_sync_thread(net,
3346 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3347 nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3348 nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3351 static int ip_vs_genl_del_daemon(struct net *net, struct nlattr **attrs)
3353 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3356 return stop_sync_thread(net,
3357 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3360 static int ip_vs_genl_set_config(struct net *net, struct nlattr **attrs)
3362 struct ip_vs_timeout_user t;
3364 __ip_vs_get_timeouts(net, &t);
3366 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3367 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3369 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3371 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3373 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3374 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3376 return ip_vs_set_timeout(net, &t);
3379 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3383 struct netns_ipvs *ipvs;
3385 net = skb_sknet(skb);
3386 ipvs = net_ipvs(net);
3387 cmd = info->genlhdr->cmd;
3389 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3390 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3392 mutex_lock(&ipvs->sync_mutex);
3393 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3394 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3395 info->attrs[IPVS_CMD_ATTR_DAEMON],
3396 ip_vs_daemon_policy)) {
3401 if (cmd == IPVS_CMD_NEW_DAEMON)
3402 ret = ip_vs_genl_new_daemon(net, daemon_attrs);
3404 ret = ip_vs_genl_del_daemon(net, daemon_attrs);
3406 mutex_unlock(&ipvs->sync_mutex);
3411 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3413 struct ip_vs_service *svc = NULL;
3414 struct ip_vs_service_user_kern usvc;
3415 struct ip_vs_dest_user_kern udest;
3417 int need_full_svc = 0, need_full_dest = 0;
3420 net = skb_sknet(skb);
3421 cmd = info->genlhdr->cmd;
3423 mutex_lock(&__ip_vs_mutex);
3425 if (cmd == IPVS_CMD_FLUSH) {
3426 ret = ip_vs_flush(net, false);
3428 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3429 ret = ip_vs_genl_set_config(net, info->attrs);
3431 } else if (cmd == IPVS_CMD_ZERO &&
3432 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3433 ret = ip_vs_zero_all(net);
3437 /* All following commands require a service argument, so check if we
3438 * received a valid one. We need a full service specification when
3439 * adding / editing a service. Only identifying members otherwise. */
3440 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3443 ret = ip_vs_genl_parse_service(net, &usvc,
3444 info->attrs[IPVS_CMD_ATTR_SERVICE],
3445 need_full_svc, &svc);
3449 /* Unless we're adding a new service, the service must already exist */
3450 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3455 /* Destination commands require a valid destination argument. For
3456 * adding / editing a destination, we need a full destination
3458 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3459 cmd == IPVS_CMD_DEL_DEST) {
3460 if (cmd != IPVS_CMD_DEL_DEST)
3463 ret = ip_vs_genl_parse_dest(&udest,
3464 info->attrs[IPVS_CMD_ATTR_DEST],
3471 case IPVS_CMD_NEW_SERVICE:
3473 ret = ip_vs_add_service(net, &usvc, &svc);
3477 case IPVS_CMD_SET_SERVICE:
3478 ret = ip_vs_edit_service(svc, &usvc);
3480 case IPVS_CMD_DEL_SERVICE:
3481 ret = ip_vs_del_service(svc);
3482 /* do not use svc, it can be freed */
3484 case IPVS_CMD_NEW_DEST:
3485 ret = ip_vs_add_dest(svc, &udest);
3487 case IPVS_CMD_SET_DEST:
3488 ret = ip_vs_edit_dest(svc, &udest);
3490 case IPVS_CMD_DEL_DEST:
3491 ret = ip_vs_del_dest(svc, &udest);
3494 ret = ip_vs_zero_service(svc);
3501 mutex_unlock(&__ip_vs_mutex);
3506 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3508 struct sk_buff *msg;
3510 int ret, cmd, reply_cmd;
3513 net = skb_sknet(skb);
3514 cmd = info->genlhdr->cmd;
3516 if (cmd == IPVS_CMD_GET_SERVICE)
3517 reply_cmd = IPVS_CMD_NEW_SERVICE;
3518 else if (cmd == IPVS_CMD_GET_INFO)
3519 reply_cmd = IPVS_CMD_SET_INFO;
3520 else if (cmd == IPVS_CMD_GET_CONFIG)
3521 reply_cmd = IPVS_CMD_SET_CONFIG;
3523 pr_err("unknown Generic Netlink command\n");
3527 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3531 mutex_lock(&__ip_vs_mutex);
3533 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3535 goto nla_put_failure;
3538 case IPVS_CMD_GET_SERVICE:
3540 struct ip_vs_service *svc;
3542 svc = ip_vs_genl_find_service(net,
3543 info->attrs[IPVS_CMD_ATTR_SERVICE]);
3548 ret = ip_vs_genl_fill_service(msg, svc);
3550 goto nla_put_failure;
3559 case IPVS_CMD_GET_CONFIG:
3561 struct ip_vs_timeout_user t;
3563 __ip_vs_get_timeouts(net, &t);
3564 #ifdef CONFIG_IP_VS_PROTO_TCP
3565 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3567 nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3569 goto nla_put_failure;
3571 #ifdef CONFIG_IP_VS_PROTO_UDP
3572 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3573 goto nla_put_failure;
3579 case IPVS_CMD_GET_INFO:
3580 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3581 IP_VS_VERSION_CODE) ||
3582 nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3583 ip_vs_conn_tab_size))
3584 goto nla_put_failure;
3588 genlmsg_end(msg, reply);
3589 ret = genlmsg_reply(msg, info);
3593 pr_err("not enough space in Netlink message\n");
3599 mutex_unlock(&__ip_vs_mutex);
3605 static struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3607 .cmd = IPVS_CMD_NEW_SERVICE,
3608 .flags = GENL_ADMIN_PERM,
3609 .policy = ip_vs_cmd_policy,
3610 .doit = ip_vs_genl_set_cmd,
3613 .cmd = IPVS_CMD_SET_SERVICE,
3614 .flags = GENL_ADMIN_PERM,
3615 .policy = ip_vs_cmd_policy,
3616 .doit = ip_vs_genl_set_cmd,
3619 .cmd = IPVS_CMD_DEL_SERVICE,
3620 .flags = GENL_ADMIN_PERM,
3621 .policy = ip_vs_cmd_policy,
3622 .doit = ip_vs_genl_set_cmd,
3625 .cmd = IPVS_CMD_GET_SERVICE,
3626 .flags = GENL_ADMIN_PERM,
3627 .doit = ip_vs_genl_get_cmd,
3628 .dumpit = ip_vs_genl_dump_services,
3629 .policy = ip_vs_cmd_policy,
3632 .cmd = IPVS_CMD_NEW_DEST,
3633 .flags = GENL_ADMIN_PERM,
3634 .policy = ip_vs_cmd_policy,
3635 .doit = ip_vs_genl_set_cmd,
3638 .cmd = IPVS_CMD_SET_DEST,
3639 .flags = GENL_ADMIN_PERM,
3640 .policy = ip_vs_cmd_policy,
3641 .doit = ip_vs_genl_set_cmd,
3644 .cmd = IPVS_CMD_DEL_DEST,
3645 .flags = GENL_ADMIN_PERM,
3646 .policy = ip_vs_cmd_policy,
3647 .doit = ip_vs_genl_set_cmd,
3650 .cmd = IPVS_CMD_GET_DEST,
3651 .flags = GENL_ADMIN_PERM,
3652 .policy = ip_vs_cmd_policy,
3653 .dumpit = ip_vs_genl_dump_dests,
3656 .cmd = IPVS_CMD_NEW_DAEMON,
3657 .flags = GENL_ADMIN_PERM,
3658 .policy = ip_vs_cmd_policy,
3659 .doit = ip_vs_genl_set_daemon,
3662 .cmd = IPVS_CMD_DEL_DAEMON,
3663 .flags = GENL_ADMIN_PERM,
3664 .policy = ip_vs_cmd_policy,
3665 .doit = ip_vs_genl_set_daemon,
3668 .cmd = IPVS_CMD_GET_DAEMON,
3669 .flags = GENL_ADMIN_PERM,
3670 .dumpit = ip_vs_genl_dump_daemons,
3673 .cmd = IPVS_CMD_SET_CONFIG,
3674 .flags = GENL_ADMIN_PERM,
3675 .policy = ip_vs_cmd_policy,
3676 .doit = ip_vs_genl_set_cmd,
3679 .cmd = IPVS_CMD_GET_CONFIG,
3680 .flags = GENL_ADMIN_PERM,
3681 .doit = ip_vs_genl_get_cmd,
3684 .cmd = IPVS_CMD_GET_INFO,
3685 .flags = GENL_ADMIN_PERM,
3686 .doit = ip_vs_genl_get_cmd,
3689 .cmd = IPVS_CMD_ZERO,
3690 .flags = GENL_ADMIN_PERM,
3691 .policy = ip_vs_cmd_policy,
3692 .doit = ip_vs_genl_set_cmd,
3695 .cmd = IPVS_CMD_FLUSH,
3696 .flags = GENL_ADMIN_PERM,
3697 .doit = ip_vs_genl_set_cmd,
3701 static int __init ip_vs_genl_register(void)
3703 return genl_register_family_with_ops(&ip_vs_genl_family,
3704 ip_vs_genl_ops, ARRAY_SIZE(ip_vs_genl_ops));
3707 static void ip_vs_genl_unregister(void)
3709 genl_unregister_family(&ip_vs_genl_family);
3712 /* End of Generic Netlink interface definitions */
3715 * per netns intit/exit func.
3717 #ifdef CONFIG_SYSCTL
3718 static int __net_init ip_vs_control_net_init_sysctl(struct net *net)
3721 struct netns_ipvs *ipvs = net_ipvs(net);
3722 struct ctl_table *tbl;
3724 atomic_set(&ipvs->dropentry, 0);
3725 spin_lock_init(&ipvs->dropentry_lock);
3726 spin_lock_init(&ipvs->droppacket_lock);
3727 spin_lock_init(&ipvs->securetcp_lock);
3729 if (!net_eq(net, &init_net)) {
3730 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3734 /* Don't export sysctls to unprivileged users */
3735 if (net->user_ns != &init_user_ns)
3736 tbl[0].procname = NULL;
3739 /* Initialize sysctl defaults */
3741 ipvs->sysctl_amemthresh = 1024;
3742 tbl[idx++].data = &ipvs->sysctl_amemthresh;
3743 ipvs->sysctl_am_droprate = 10;
3744 tbl[idx++].data = &ipvs->sysctl_am_droprate;
3745 tbl[idx++].data = &ipvs->sysctl_drop_entry;
3746 tbl[idx++].data = &ipvs->sysctl_drop_packet;
3747 #ifdef CONFIG_IP_VS_NFCT
3748 tbl[idx++].data = &ipvs->sysctl_conntrack;
3750 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3751 ipvs->sysctl_snat_reroute = 1;
3752 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3753 ipvs->sysctl_sync_ver = 1;
3754 tbl[idx++].data = &ipvs->sysctl_sync_ver;
3755 ipvs->sysctl_sync_ports = 1;
3756 tbl[idx++].data = &ipvs->sysctl_sync_ports;
3757 ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3758 tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3759 ipvs->sysctl_sync_sock_size = 0;
3760 tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3761 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3762 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3763 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3764 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3765 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3766 tbl[idx].data = &ipvs->sysctl_sync_threshold;
3767 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3768 ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3769 tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3770 ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3771 tbl[idx++].data = &ipvs->sysctl_sync_retries;
3772 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3773 ipvs->sysctl_pmtu_disc = 1;
3774 tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
3775 tbl[idx++].data = &ipvs->sysctl_backup_only;
3778 ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3779 if (ipvs->sysctl_hdr == NULL) {
3780 if (!net_eq(net, &init_net))
3784 ip_vs_start_estimator(net, &ipvs->tot_stats);
3785 ipvs->sysctl_tbl = tbl;
3786 /* Schedule defense work */
3787 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3788 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3793 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net)
3795 struct netns_ipvs *ipvs = net_ipvs(net);
3797 cancel_delayed_work_sync(&ipvs->defense_work);
3798 cancel_work_sync(&ipvs->defense_work.work);
3799 unregister_net_sysctl_table(ipvs->sysctl_hdr);
3804 static int __net_init ip_vs_control_net_init_sysctl(struct net *net) { return 0; }
3805 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net) { }
3809 static struct notifier_block ip_vs_dst_notifier = {
3810 .notifier_call = ip_vs_dst_event,
3813 int __net_init ip_vs_control_net_init(struct net *net)
3816 struct netns_ipvs *ipvs = net_ipvs(net);
3818 /* Initialize rs_table */
3819 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
3820 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
3822 INIT_LIST_HEAD(&ipvs->dest_trash);
3823 spin_lock_init(&ipvs->dest_trash_lock);
3824 setup_timer(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire,
3825 (unsigned long) net);
3826 atomic_set(&ipvs->ftpsvc_counter, 0);
3827 atomic_set(&ipvs->nullsvc_counter, 0);
3830 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
3831 if (!ipvs->tot_stats.cpustats)
3834 spin_lock_init(&ipvs->tot_stats.lock);
3836 proc_create("ip_vs", 0, net->proc_net, &ip_vs_info_fops);
3837 proc_create("ip_vs_stats", 0, net->proc_net, &ip_vs_stats_fops);
3838 proc_create("ip_vs_stats_percpu", 0, net->proc_net,
3839 &ip_vs_stats_percpu_fops);
3841 if (ip_vs_control_net_init_sysctl(net))
3847 free_percpu(ipvs->tot_stats.cpustats);
3851 void __net_exit ip_vs_control_net_cleanup(struct net *net)
3853 struct netns_ipvs *ipvs = net_ipvs(net);
3855 /* Some dest can be in grace period even before cleanup, we have to
3856 * defer ip_vs_trash_cleanup until ip_vs_dest_wait_readers is called.
3859 ip_vs_trash_cleanup(net);
3860 ip_vs_stop_estimator(net, &ipvs->tot_stats);
3861 ip_vs_control_net_cleanup_sysctl(net);
3862 remove_proc_entry("ip_vs_stats_percpu", net->proc_net);
3863 remove_proc_entry("ip_vs_stats", net->proc_net);
3864 remove_proc_entry("ip_vs", net->proc_net);
3865 free_percpu(ipvs->tot_stats.cpustats);
3868 int __init ip_vs_register_nl_ioctl(void)
3872 ret = nf_register_sockopt(&ip_vs_sockopts);
3874 pr_err("cannot register sockopt.\n");
3878 ret = ip_vs_genl_register();
3880 pr_err("cannot register Generic Netlink interface.\n");
3886 nf_unregister_sockopt(&ip_vs_sockopts);
3891 void ip_vs_unregister_nl_ioctl(void)
3893 ip_vs_genl_unregister();
3894 nf_unregister_sockopt(&ip_vs_sockopts);
3897 int __init ip_vs_control_init(void)
3904 /* Initialize svc_table, ip_vs_svc_fwm_table */
3905 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
3906 INIT_LIST_HEAD(&ip_vs_svc_table[idx]);
3907 INIT_LIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3910 smp_wmb(); /* Do we really need it now ? */
3912 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
3921 void ip_vs_control_cleanup(void)
3924 unregister_netdevice_notifier(&ip_vs_dst_notifier);