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>
46 #include <net/netfilter/ipv6/nf_defrag_ipv6.h>
48 #include <net/route.h>
50 #include <net/genetlink.h>
52 #include <linux/uaccess.h>
54 #include <net/ip_vs.h>
56 /* semaphore for IPVS sockopts. And, [gs]etsockopt may sleep. */
57 static DEFINE_MUTEX(__ip_vs_mutex);
59 /* sysctl variables */
61 #ifdef CONFIG_IP_VS_DEBUG
62 static int sysctl_ip_vs_debug_level = 0;
64 int ip_vs_get_debug_level(void)
66 return sysctl_ip_vs_debug_level;
72 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup);
75 #ifdef CONFIG_IP_VS_IPV6
76 /* Taken from rt6_fill_node() in net/ipv6/route.c, is there a better way? */
77 static bool __ip_vs_addr_is_local_v6(struct net *net,
78 const struct in6_addr *addr)
83 struct dst_entry *dst = ip6_route_output(net, NULL, &fl6);
86 is_local = !dst->error && dst->dev && (dst->dev->flags & IFF_LOOPBACK);
95 * update_defense_level is called from keventd and from sysctl,
96 * so it needs to protect itself from softirqs
98 static void update_defense_level(struct netns_ipvs *ipvs)
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 (ipvs->old_secure_tcp >= 2)
186 if (ipvs->old_secure_tcp < 2)
188 ipvs->sysctl_secure_tcp = 2;
190 if (ipvs->old_secure_tcp >= 2)
196 if (ipvs->old_secure_tcp < 2)
199 if (ipvs->old_secure_tcp >= 2)
201 ipvs->sysctl_secure_tcp = 1;
205 if (ipvs->old_secure_tcp < 2)
209 ipvs->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);
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 netns_ipvs *ipvs, 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) ipvs >> 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 netns_ipvs *ipvs, __u32 fwmark)
290 return (((size_t)ipvs>>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->ipvs, 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->ipvs, 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 netns_ipvs *ipvs, 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(ipvs, 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 && (svc->ipvs == ipvs)) {
385 * Get service by {fwmark} in the service table.
387 static inline struct ip_vs_service *
388 __ip_vs_svc_fwm_find(struct netns_ipvs *ipvs, int af, __u32 fwmark)
391 struct ip_vs_service *svc;
393 /* Check for fwmark addressed entries */
394 hash = ip_vs_svc_fwm_hashkey(ipvs, 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 && (svc->ipvs == ipvs)) {
407 /* Find service, called under RCU lock */
408 struct ip_vs_service *
409 ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
410 const union nf_inet_addr *vaddr, __be16 vport)
412 struct ip_vs_service *svc;
415 * Check the table hashed by fwmark first
418 svc = __ip_vs_svc_fwm_find(ipvs, af, fwmark);
424 * Check the table hashed by <protocol,addr,port>
425 * for "full" addressed entries
427 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, vport);
429 if (!svc && protocol == IPPROTO_TCP &&
430 atomic_read(&ipvs->ftpsvc_counter) &&
431 (vport == FTPDATA || ntohs(vport) >= inet_prot_sock(ipvs->net))) {
433 * Check if ftp service entry exists, the packet
434 * might belong to FTP data connections.
436 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, FTPPORT);
440 && atomic_read(&ipvs->nullsvc_counter)) {
442 * Check if the catch-all port (port zero) exists
444 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, 0);
448 IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
449 fwmark, ip_vs_proto_name(protocol),
450 IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
451 svc ? "hit" : "not hit");
458 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
460 atomic_inc(&svc->refcnt);
461 rcu_assign_pointer(dest->svc, svc);
464 static void ip_vs_service_free(struct ip_vs_service *svc)
466 free_percpu(svc->stats.cpustats);
470 static void ip_vs_service_rcu_free(struct rcu_head *head)
472 struct ip_vs_service *svc;
474 svc = container_of(head, struct ip_vs_service, rcu_head);
475 ip_vs_service_free(svc);
478 static void __ip_vs_svc_put(struct ip_vs_service *svc, bool do_delay)
480 if (atomic_dec_and_test(&svc->refcnt)) {
481 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n",
483 IP_VS_DBG_ADDR(svc->af, &svc->addr),
486 call_rcu(&svc->rcu_head, ip_vs_service_rcu_free);
488 ip_vs_service_free(svc);
494 * Returns hash value for real service
496 static inline unsigned int ip_vs_rs_hashkey(int af,
497 const union nf_inet_addr *addr,
500 register unsigned int porth = ntohs(port);
501 __be32 addr_fold = addr->ip;
503 #ifdef CONFIG_IP_VS_IPV6
505 addr_fold = addr->ip6[0]^addr->ip6[1]^
506 addr->ip6[2]^addr->ip6[3];
509 return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
513 /* Hash ip_vs_dest in rs_table by <proto,addr,port>. */
514 static void ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
518 if (dest->in_rs_table)
522 * Hash by proto,addr,port,
523 * which are the parameters of the real service.
525 hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
527 hlist_add_head_rcu(&dest->d_list, &ipvs->rs_table[hash]);
528 dest->in_rs_table = 1;
531 /* Unhash ip_vs_dest from rs_table. */
532 static void ip_vs_rs_unhash(struct ip_vs_dest *dest)
535 * Remove it from the rs_table table.
537 if (dest->in_rs_table) {
538 hlist_del_rcu(&dest->d_list);
539 dest->in_rs_table = 0;
543 /* Check if real service by <proto,addr,port> is present */
544 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
545 const union nf_inet_addr *daddr, __be16 dport)
548 struct ip_vs_dest *dest;
550 /* Check for "full" addressed entries */
551 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)) {
566 /* Find real service record by <proto,addr,port>.
567 * In case of multiple records with the same <proto,addr,port>, only
568 * the first found record is returned.
570 * To be called under RCU lock.
572 struct ip_vs_dest *ip_vs_find_real_service(struct netns_ipvs *ipvs, int af,
574 const union nf_inet_addr *daddr,
578 struct ip_vs_dest *dest;
580 /* Check for "full" addressed entries */
581 hash = ip_vs_rs_hashkey(af, daddr, dport);
583 hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
584 if (dest->port == dport &&
586 ip_vs_addr_equal(af, &dest->addr, daddr) &&
587 (dest->protocol == protocol || dest->vfwmark)) {
596 /* Lookup destination by {addr,port} in the given service
597 * Called under RCU lock.
599 static struct ip_vs_dest *
600 ip_vs_lookup_dest(struct ip_vs_service *svc, int dest_af,
601 const union nf_inet_addr *daddr, __be16 dport)
603 struct ip_vs_dest *dest;
606 * Find the destination for the given service
608 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
609 if ((dest->af == dest_af) &&
610 ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
611 (dest->port == dport)) {
621 * Find destination by {daddr,dport,vaddr,protocol}
622 * Created to be used in ip_vs_process_message() in
623 * the backup synchronization daemon. It finds the
624 * destination to be bound to the received connection
626 * Called under RCU lock, no refcnt is returned.
628 struct ip_vs_dest *ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
629 const union nf_inet_addr *daddr,
631 const union nf_inet_addr *vaddr,
632 __be16 vport, __u16 protocol, __u32 fwmark,
635 struct ip_vs_dest *dest;
636 struct ip_vs_service *svc;
639 svc = ip_vs_service_find(ipvs, svc_af, fwmark, protocol, vaddr, vport);
642 if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
644 dest = ip_vs_lookup_dest(svc, dest_af, daddr, port);
646 dest = ip_vs_lookup_dest(svc, dest_af, daddr, port ^ dport);
650 void ip_vs_dest_dst_rcu_free(struct rcu_head *head)
652 struct ip_vs_dest_dst *dest_dst = container_of(head,
653 struct ip_vs_dest_dst,
656 dst_release(dest_dst->dst_cache);
660 /* Release dest_dst and dst_cache for dest in user context */
661 static void __ip_vs_dst_cache_reset(struct ip_vs_dest *dest)
663 struct ip_vs_dest_dst *old;
665 old = rcu_dereference_protected(dest->dest_dst, 1);
667 RCU_INIT_POINTER(dest->dest_dst, NULL);
668 call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
673 * Lookup dest by {svc,addr,port} in the destination trash.
674 * The destination trash is used to hold the destinations that are removed
675 * from the service table but are still referenced by some conn entries.
676 * The reason to add the destination trash is when the dest is temporary
677 * down (either by administrator or by monitor program), the dest can be
678 * picked back from the trash, the remaining connections to the dest can
679 * continue, and the counting information of the dest is also useful for
682 static struct ip_vs_dest *
683 ip_vs_trash_get_dest(struct ip_vs_service *svc, int dest_af,
684 const union nf_inet_addr *daddr, __be16 dport)
686 struct ip_vs_dest *dest;
687 struct netns_ipvs *ipvs = svc->ipvs;
690 * Find the destination in trash
692 spin_lock_bh(&ipvs->dest_trash_lock);
693 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
694 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
697 IP_VS_DBG_ADDR(dest->af, &dest->addr),
699 refcount_read(&dest->refcnt));
700 if (dest->af == dest_af &&
701 ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
702 dest->port == dport &&
703 dest->vfwmark == svc->fwmark &&
704 dest->protocol == svc->protocol &&
706 (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
707 dest->vport == svc->port))) {
709 list_del(&dest->t_list);
717 spin_unlock_bh(&ipvs->dest_trash_lock);
722 static void ip_vs_dest_free(struct ip_vs_dest *dest)
724 struct ip_vs_service *svc = rcu_dereference_protected(dest->svc, 1);
726 __ip_vs_dst_cache_reset(dest);
727 __ip_vs_svc_put(svc, false);
728 free_percpu(dest->stats.cpustats);
729 ip_vs_dest_put_and_free(dest);
733 * Clean up all the destinations in the trash
734 * Called by the ip_vs_control_cleanup()
736 * When the ip_vs_control_clearup is activated by ipvs module exit,
737 * the service tables must have been flushed and all the connections
738 * are expired, and the refcnt of each destination in the trash must
739 * be 1, so we simply release them here.
741 static void ip_vs_trash_cleanup(struct netns_ipvs *ipvs)
743 struct ip_vs_dest *dest, *nxt;
745 del_timer_sync(&ipvs->dest_trash_timer);
746 /* No need to use dest_trash_lock */
747 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, t_list) {
748 list_del(&dest->t_list);
749 ip_vs_dest_free(dest);
754 ip_vs_copy_stats(struct ip_vs_kstats *dst, struct ip_vs_stats *src)
756 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->kstats.c - src->kstats0.c
758 spin_lock_bh(&src->lock);
760 IP_VS_SHOW_STATS_COUNTER(conns);
761 IP_VS_SHOW_STATS_COUNTER(inpkts);
762 IP_VS_SHOW_STATS_COUNTER(outpkts);
763 IP_VS_SHOW_STATS_COUNTER(inbytes);
764 IP_VS_SHOW_STATS_COUNTER(outbytes);
766 ip_vs_read_estimator(dst, src);
768 spin_unlock_bh(&src->lock);
772 ip_vs_export_stats_user(struct ip_vs_stats_user *dst, struct ip_vs_kstats *src)
774 dst->conns = (u32)src->conns;
775 dst->inpkts = (u32)src->inpkts;
776 dst->outpkts = (u32)src->outpkts;
777 dst->inbytes = src->inbytes;
778 dst->outbytes = src->outbytes;
779 dst->cps = (u32)src->cps;
780 dst->inpps = (u32)src->inpps;
781 dst->outpps = (u32)src->outpps;
782 dst->inbps = (u32)src->inbps;
783 dst->outbps = (u32)src->outbps;
787 ip_vs_zero_stats(struct ip_vs_stats *stats)
789 spin_lock_bh(&stats->lock);
791 /* get current counters as zero point, rates are zeroed */
793 #define IP_VS_ZERO_STATS_COUNTER(c) stats->kstats0.c = stats->kstats.c
795 IP_VS_ZERO_STATS_COUNTER(conns);
796 IP_VS_ZERO_STATS_COUNTER(inpkts);
797 IP_VS_ZERO_STATS_COUNTER(outpkts);
798 IP_VS_ZERO_STATS_COUNTER(inbytes);
799 IP_VS_ZERO_STATS_COUNTER(outbytes);
801 ip_vs_zero_estimator(stats);
803 spin_unlock_bh(&stats->lock);
807 * Update a destination in the given service
810 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
811 struct ip_vs_dest_user_kern *udest, int add)
813 struct netns_ipvs *ipvs = svc->ipvs;
814 struct ip_vs_service *old_svc;
815 struct ip_vs_scheduler *sched;
818 /* We cannot modify an address and change the address family */
819 BUG_ON(!add && udest->af != dest->af);
821 if (add && udest->af != svc->af)
822 ipvs->mixed_address_family_dests++;
824 /* set the weight and the flags */
825 atomic_set(&dest->weight, udest->weight);
826 conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
827 conn_flags |= IP_VS_CONN_F_INACTIVE;
829 /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
830 if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
831 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
834 * Put the real service in rs_table if not present.
835 * For now only for NAT!
837 ip_vs_rs_hash(ipvs, dest);
839 atomic_set(&dest->conn_flags, conn_flags);
841 /* bind the service */
842 old_svc = rcu_dereference_protected(dest->svc, 1);
844 __ip_vs_bind_svc(dest, svc);
846 if (old_svc != svc) {
847 ip_vs_zero_stats(&dest->stats);
848 __ip_vs_bind_svc(dest, svc);
849 __ip_vs_svc_put(old_svc, true);
853 /* set the dest status flags */
854 dest->flags |= IP_VS_DEST_F_AVAILABLE;
856 if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
857 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
858 dest->u_threshold = udest->u_threshold;
859 dest->l_threshold = udest->l_threshold;
861 dest->af = udest->af;
863 spin_lock_bh(&dest->dst_lock);
864 __ip_vs_dst_cache_reset(dest);
865 spin_unlock_bh(&dest->dst_lock);
868 ip_vs_start_estimator(svc->ipvs, &dest->stats);
869 list_add_rcu(&dest->n_list, &svc->destinations);
871 sched = rcu_dereference_protected(svc->scheduler, 1);
872 if (sched && sched->add_dest)
873 sched->add_dest(svc, dest);
875 sched = rcu_dereference_protected(svc->scheduler, 1);
876 if (sched && sched->upd_dest)
877 sched->upd_dest(svc, dest);
883 * Create a destination for the given service
886 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
887 struct ip_vs_dest **dest_p)
889 struct ip_vs_dest *dest;
890 unsigned int atype, i;
894 #ifdef CONFIG_IP_VS_IPV6
895 if (udest->af == AF_INET6) {
898 atype = ipv6_addr_type(&udest->addr.in6);
899 if ((!(atype & IPV6_ADDR_UNICAST) ||
900 atype & IPV6_ADDR_LINKLOCAL) &&
901 !__ip_vs_addr_is_local_v6(svc->ipvs->net, &udest->addr.in6))
904 ret = nf_defrag_ipv6_enable(svc->ipvs->net);
910 atype = inet_addr_type(svc->ipvs->net, udest->addr.ip);
911 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
915 dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
919 dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
920 if (!dest->stats.cpustats)
923 for_each_possible_cpu(i) {
924 struct ip_vs_cpu_stats *ip_vs_dest_stats;
925 ip_vs_dest_stats = per_cpu_ptr(dest->stats.cpustats, i);
926 u64_stats_init(&ip_vs_dest_stats->syncp);
929 dest->af = udest->af;
930 dest->protocol = svc->protocol;
931 dest->vaddr = svc->addr;
932 dest->vport = svc->port;
933 dest->vfwmark = svc->fwmark;
934 ip_vs_addr_copy(udest->af, &dest->addr, &udest->addr);
935 dest->port = udest->port;
937 atomic_set(&dest->activeconns, 0);
938 atomic_set(&dest->inactconns, 0);
939 atomic_set(&dest->persistconns, 0);
940 refcount_set(&dest->refcnt, 1);
942 INIT_HLIST_NODE(&dest->d_list);
943 spin_lock_init(&dest->dst_lock);
944 spin_lock_init(&dest->stats.lock);
945 __ip_vs_update_dest(svc, dest, udest, 1);
959 * Add a destination into an existing service
962 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
964 struct ip_vs_dest *dest;
965 union nf_inet_addr daddr;
966 __be16 dport = udest->port;
971 if (udest->weight < 0) {
972 pr_err("%s(): server weight less than zero\n", __func__);
976 if (udest->l_threshold > udest->u_threshold) {
977 pr_err("%s(): lower threshold is higher than upper threshold\n",
982 ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
984 /* We use function that requires RCU lock */
986 dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
990 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
995 * Check if the dest already exists in the trash and
996 * is from the same service
998 dest = ip_vs_trash_get_dest(svc, udest->af, &daddr, dport);
1001 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
1002 "dest->refcnt=%d, service %u/%s:%u\n",
1003 IP_VS_DBG_ADDR(udest->af, &daddr), ntohs(dport),
1004 refcount_read(&dest->refcnt),
1006 IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
1007 ntohs(dest->vport));
1009 __ip_vs_update_dest(svc, dest, udest, 1);
1013 * Allocate and initialize the dest structure
1015 ret = ip_vs_new_dest(svc, udest, &dest);
1024 * Edit a destination in the given service
1027 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1029 struct ip_vs_dest *dest;
1030 union nf_inet_addr daddr;
1031 __be16 dport = udest->port;
1035 if (udest->weight < 0) {
1036 pr_err("%s(): server weight less than zero\n", __func__);
1040 if (udest->l_threshold > udest->u_threshold) {
1041 pr_err("%s(): lower threshold is higher than upper threshold\n",
1046 ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
1048 /* We use function that requires RCU lock */
1050 dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
1054 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1058 __ip_vs_update_dest(svc, dest, udest, 0);
1065 * Delete a destination (must be already unlinked from the service)
1067 static void __ip_vs_del_dest(struct netns_ipvs *ipvs, struct ip_vs_dest *dest,
1070 ip_vs_stop_estimator(ipvs, &dest->stats);
1073 * Remove it from the d-linked list with the real services.
1075 ip_vs_rs_unhash(dest);
1077 spin_lock_bh(&ipvs->dest_trash_lock);
1078 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1079 IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port),
1080 refcount_read(&dest->refcnt));
1081 if (list_empty(&ipvs->dest_trash) && !cleanup)
1082 mod_timer(&ipvs->dest_trash_timer,
1083 jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1084 /* dest lives in trash with reference */
1085 list_add(&dest->t_list, &ipvs->dest_trash);
1086 dest->idle_start = 0;
1087 spin_unlock_bh(&ipvs->dest_trash_lock);
1092 * Unlink a destination from the given service
1094 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1095 struct ip_vs_dest *dest,
1098 dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1101 * Remove it from the d-linked destination list.
1103 list_del_rcu(&dest->n_list);
1106 if (dest->af != svc->af)
1107 svc->ipvs->mixed_address_family_dests--;
1110 struct ip_vs_scheduler *sched;
1112 sched = rcu_dereference_protected(svc->scheduler, 1);
1113 if (sched && sched->del_dest)
1114 sched->del_dest(svc, dest);
1120 * Delete a destination server in the given service
1123 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1125 struct ip_vs_dest *dest;
1126 __be16 dport = udest->port;
1130 /* We use function that requires RCU lock */
1132 dest = ip_vs_lookup_dest(svc, udest->af, &udest->addr, dport);
1136 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1141 * Unlink dest from the service
1143 __ip_vs_unlink_dest(svc, dest, 1);
1146 * Delete the destination
1148 __ip_vs_del_dest(svc->ipvs, dest, false);
1155 static void ip_vs_dest_trash_expire(unsigned long data)
1157 struct netns_ipvs *ipvs = (struct netns_ipvs *)data;
1158 struct ip_vs_dest *dest, *next;
1159 unsigned long now = jiffies;
1161 spin_lock(&ipvs->dest_trash_lock);
1162 list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) {
1163 if (refcount_read(&dest->refcnt) > 1)
1165 if (dest->idle_start) {
1166 if (time_before(now, dest->idle_start +
1167 IP_VS_DEST_TRASH_PERIOD))
1170 dest->idle_start = max(1UL, now);
1173 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1175 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1177 list_del(&dest->t_list);
1178 ip_vs_dest_free(dest);
1180 if (!list_empty(&ipvs->dest_trash))
1181 mod_timer(&ipvs->dest_trash_timer,
1182 jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1183 spin_unlock(&ipvs->dest_trash_lock);
1187 * Add a service into the service hash table
1190 ip_vs_add_service(struct netns_ipvs *ipvs, struct ip_vs_service_user_kern *u,
1191 struct ip_vs_service **svc_p)
1194 struct ip_vs_scheduler *sched = NULL;
1195 struct ip_vs_pe *pe = NULL;
1196 struct ip_vs_service *svc = NULL;
1198 /* increase the module use count */
1199 if (!ip_vs_use_count_inc())
1200 return -ENOPROTOOPT;
1202 /* Lookup the scheduler by 'u->sched_name' */
1203 if (strcmp(u->sched_name, "none")) {
1204 sched = ip_vs_scheduler_get(u->sched_name);
1206 pr_info("Scheduler module ip_vs_%s not found\n",
1213 if (u->pe_name && *u->pe_name) {
1214 pe = ip_vs_pe_getbyname(u->pe_name);
1216 pr_info("persistence engine module ip_vs_pe_%s "
1217 "not found\n", u->pe_name);
1223 #ifdef CONFIG_IP_VS_IPV6
1224 if (u->af == AF_INET6) {
1225 __u32 plen = (__force __u32) u->netmask;
1227 if (plen < 1 || plen > 128) {
1232 ret = nf_defrag_ipv6_enable(ipvs->net);
1238 svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1240 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1244 svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1245 if (!svc->stats.cpustats) {
1250 for_each_possible_cpu(i) {
1251 struct ip_vs_cpu_stats *ip_vs_stats;
1252 ip_vs_stats = per_cpu_ptr(svc->stats.cpustats, i);
1253 u64_stats_init(&ip_vs_stats->syncp);
1257 /* I'm the first user of the service */
1258 atomic_set(&svc->refcnt, 0);
1261 svc->protocol = u->protocol;
1262 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1263 svc->port = u->port;
1264 svc->fwmark = u->fwmark;
1265 svc->flags = u->flags & ~IP_VS_SVC_F_HASHED;
1266 svc->timeout = u->timeout * HZ;
1267 svc->netmask = u->netmask;
1270 INIT_LIST_HEAD(&svc->destinations);
1271 spin_lock_init(&svc->sched_lock);
1272 spin_lock_init(&svc->stats.lock);
1274 /* Bind the scheduler */
1276 ret = ip_vs_bind_scheduler(svc, sched);
1282 /* Bind the ct retriever */
1283 RCU_INIT_POINTER(svc->pe, pe);
1286 /* Update the virtual service counters */
1287 if (svc->port == FTPPORT)
1288 atomic_inc(&ipvs->ftpsvc_counter);
1289 else if (svc->port == 0)
1290 atomic_inc(&ipvs->nullsvc_counter);
1291 if (svc->pe && svc->pe->conn_out)
1292 atomic_inc(&ipvs->conn_out_counter);
1294 ip_vs_start_estimator(ipvs, &svc->stats);
1296 /* Count only IPv4 services for old get/setsockopt interface */
1297 if (svc->af == AF_INET)
1298 ipvs->num_services++;
1300 /* Hash the service into the service table */
1301 ip_vs_svc_hash(svc);
1304 /* Now there is a service - full throttle */
1311 ip_vs_unbind_scheduler(svc, sched);
1312 ip_vs_service_free(svc);
1314 ip_vs_scheduler_put(sched);
1317 /* decrease the module use count */
1318 ip_vs_use_count_dec();
1325 * Edit a service and bind it with a new scheduler
1328 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1330 struct ip_vs_scheduler *sched = NULL, *old_sched;
1331 struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1333 bool new_pe_conn_out, old_pe_conn_out;
1336 * Lookup the scheduler, by 'u->sched_name'
1338 if (strcmp(u->sched_name, "none")) {
1339 sched = ip_vs_scheduler_get(u->sched_name);
1341 pr_info("Scheduler module ip_vs_%s not found\n",
1348 if (u->pe_name && *u->pe_name) {
1349 pe = ip_vs_pe_getbyname(u->pe_name);
1351 pr_info("persistence engine module ip_vs_pe_%s "
1352 "not found\n", u->pe_name);
1359 #ifdef CONFIG_IP_VS_IPV6
1360 if (u->af == AF_INET6) {
1361 __u32 plen = (__force __u32) u->netmask;
1363 if (plen < 1 || plen > 128) {
1370 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1371 if (sched != old_sched) {
1373 ip_vs_unbind_scheduler(svc, old_sched);
1374 RCU_INIT_POINTER(svc->scheduler, NULL);
1375 /* Wait all svc->sched_data users */
1378 /* Bind the new scheduler */
1380 ret = ip_vs_bind_scheduler(svc, sched);
1382 ip_vs_scheduler_put(sched);
1389 * Set the flags and timeout value
1391 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1392 svc->timeout = u->timeout * HZ;
1393 svc->netmask = u->netmask;
1395 old_pe = rcu_dereference_protected(svc->pe, 1);
1397 rcu_assign_pointer(svc->pe, pe);
1398 /* check for optional methods in new pe */
1399 new_pe_conn_out = (pe && pe->conn_out) ? true : false;
1400 old_pe_conn_out = (old_pe && old_pe->conn_out) ? true : false;
1401 if (new_pe_conn_out && !old_pe_conn_out)
1402 atomic_inc(&svc->ipvs->conn_out_counter);
1403 if (old_pe_conn_out && !new_pe_conn_out)
1404 atomic_dec(&svc->ipvs->conn_out_counter);
1408 ip_vs_scheduler_put(old_sched);
1409 ip_vs_pe_put(old_pe);
1414 * Delete a service from the service list
1415 * - The service must be unlinked, unlocked and not referenced!
1416 * - We are called under _bh lock
1418 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1420 struct ip_vs_dest *dest, *nxt;
1421 struct ip_vs_scheduler *old_sched;
1422 struct ip_vs_pe *old_pe;
1423 struct netns_ipvs *ipvs = svc->ipvs;
1425 /* Count only IPv4 services for old get/setsockopt interface */
1426 if (svc->af == AF_INET)
1427 ipvs->num_services--;
1429 ip_vs_stop_estimator(svc->ipvs, &svc->stats);
1431 /* Unbind scheduler */
1432 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1433 ip_vs_unbind_scheduler(svc, old_sched);
1434 ip_vs_scheduler_put(old_sched);
1436 /* Unbind persistence engine, keep svc->pe */
1437 old_pe = rcu_dereference_protected(svc->pe, 1);
1438 if (old_pe && old_pe->conn_out)
1439 atomic_dec(&ipvs->conn_out_counter);
1440 ip_vs_pe_put(old_pe);
1443 * Unlink the whole destination list
1445 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1446 __ip_vs_unlink_dest(svc, dest, 0);
1447 __ip_vs_del_dest(svc->ipvs, dest, cleanup);
1451 * Update the virtual service counters
1453 if (svc->port == FTPPORT)
1454 atomic_dec(&ipvs->ftpsvc_counter);
1455 else if (svc->port == 0)
1456 atomic_dec(&ipvs->nullsvc_counter);
1459 * Free the service if nobody refers to it
1461 __ip_vs_svc_put(svc, true);
1463 /* decrease the module use count */
1464 ip_vs_use_count_dec();
1468 * Unlink a service from list and try to delete it if its refcnt reached 0
1470 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1472 /* Hold svc to avoid double release from dest_trash */
1473 atomic_inc(&svc->refcnt);
1475 * Unhash it from the service table
1477 ip_vs_svc_unhash(svc);
1479 __ip_vs_del_service(svc, cleanup);
1483 * Delete a service from the service list
1485 static int ip_vs_del_service(struct ip_vs_service *svc)
1489 ip_vs_unlink_service(svc, false);
1496 * Flush all the virtual services
1498 static int ip_vs_flush(struct netns_ipvs *ipvs, bool cleanup)
1501 struct ip_vs_service *svc;
1502 struct hlist_node *n;
1505 * Flush the service table hashed by <netns,protocol,addr,port>
1507 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1508 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1510 if (svc->ipvs == ipvs)
1511 ip_vs_unlink_service(svc, cleanup);
1516 * Flush the service table hashed by fwmark
1518 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1519 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1521 if (svc->ipvs == ipvs)
1522 ip_vs_unlink_service(svc, cleanup);
1530 * Delete service by {netns} in the service table.
1531 * Called by __ip_vs_cleanup()
1533 void ip_vs_service_net_cleanup(struct netns_ipvs *ipvs)
1536 /* Check for "full" addressed entries */
1537 mutex_lock(&__ip_vs_mutex);
1538 ip_vs_flush(ipvs, true);
1539 mutex_unlock(&__ip_vs_mutex);
1543 /* Put all references for device (dst_cache) */
1545 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1547 struct ip_vs_dest_dst *dest_dst;
1549 spin_lock_bh(&dest->dst_lock);
1550 dest_dst = rcu_dereference_protected(dest->dest_dst, 1);
1551 if (dest_dst && dest_dst->dst_cache->dev == dev) {
1552 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1554 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1556 refcount_read(&dest->refcnt));
1557 __ip_vs_dst_cache_reset(dest);
1559 spin_unlock_bh(&dest->dst_lock);
1562 /* Netdev event receiver
1563 * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1565 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1568 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1569 struct net *net = dev_net(dev);
1570 struct netns_ipvs *ipvs = net_ipvs(net);
1571 struct ip_vs_service *svc;
1572 struct ip_vs_dest *dest;
1575 if (event != NETDEV_DOWN || !ipvs)
1577 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1579 mutex_lock(&__ip_vs_mutex);
1580 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1581 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1582 if (svc->ipvs == ipvs) {
1583 list_for_each_entry(dest, &svc->destinations,
1585 ip_vs_forget_dev(dest, dev);
1590 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1591 if (svc->ipvs == ipvs) {
1592 list_for_each_entry(dest, &svc->destinations,
1594 ip_vs_forget_dev(dest, dev);
1601 spin_lock_bh(&ipvs->dest_trash_lock);
1602 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1603 ip_vs_forget_dev(dest, dev);
1605 spin_unlock_bh(&ipvs->dest_trash_lock);
1606 mutex_unlock(&__ip_vs_mutex);
1612 * Zero counters in a service or all services
1614 static int ip_vs_zero_service(struct ip_vs_service *svc)
1616 struct ip_vs_dest *dest;
1618 list_for_each_entry(dest, &svc->destinations, n_list) {
1619 ip_vs_zero_stats(&dest->stats);
1621 ip_vs_zero_stats(&svc->stats);
1625 static int ip_vs_zero_all(struct netns_ipvs *ipvs)
1628 struct ip_vs_service *svc;
1630 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1631 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1632 if (svc->ipvs == ipvs)
1633 ip_vs_zero_service(svc);
1637 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1638 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1639 if (svc->ipvs == ipvs)
1640 ip_vs_zero_service(svc);
1644 ip_vs_zero_stats(&ipvs->tot_stats);
1648 #ifdef CONFIG_SYSCTL
1652 static int three = 3;
1655 proc_do_defense_mode(struct ctl_table *table, int write,
1656 void __user *buffer, size_t *lenp, loff_t *ppos)
1658 struct netns_ipvs *ipvs = table->extra2;
1659 int *valp = table->data;
1663 struct ctl_table tmp = {
1665 .maxlen = sizeof(int),
1666 .mode = table->mode,
1669 rc = proc_dointvec(&tmp, write, buffer, lenp, ppos);
1670 if (write && (*valp != val)) {
1671 if (val < 0 || val > 3) {
1675 update_defense_level(ipvs);
1682 proc_do_sync_threshold(struct ctl_table *table, int write,
1683 void __user *buffer, size_t *lenp, loff_t *ppos)
1685 struct netns_ipvs *ipvs = table->extra2;
1686 int *valp = table->data;
1689 struct ctl_table tmp = {
1691 .maxlen = table->maxlen,
1692 .mode = table->mode,
1695 mutex_lock(&ipvs->sync_mutex);
1696 memcpy(val, valp, sizeof(val));
1697 rc = proc_dointvec(&tmp, write, buffer, lenp, ppos);
1699 if (val[0] < 0 || val[1] < 0 ||
1700 (val[0] >= val[1] && val[1]))
1703 memcpy(valp, val, sizeof(val));
1705 mutex_unlock(&ipvs->sync_mutex);
1710 proc_do_sync_ports(struct ctl_table *table, int write,
1711 void __user *buffer, size_t *lenp, loff_t *ppos)
1713 int *valp = table->data;
1717 struct ctl_table tmp = {
1719 .maxlen = sizeof(int),
1720 .mode = table->mode,
1723 rc = proc_dointvec(&tmp, write, buffer, lenp, ppos);
1724 if (write && (*valp != val)) {
1725 if (val < 1 || !is_power_of_2(val))
1734 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1735 * Do not change order or insert new entries without
1736 * align with netns init in ip_vs_control_net_init()
1739 static struct ctl_table vs_vars[] = {
1741 .procname = "amemthresh",
1742 .maxlen = sizeof(int),
1744 .proc_handler = proc_dointvec,
1747 .procname = "am_droprate",
1748 .maxlen = sizeof(int),
1750 .proc_handler = proc_dointvec,
1753 .procname = "drop_entry",
1754 .maxlen = sizeof(int),
1756 .proc_handler = proc_do_defense_mode,
1759 .procname = "drop_packet",
1760 .maxlen = sizeof(int),
1762 .proc_handler = proc_do_defense_mode,
1764 #ifdef CONFIG_IP_VS_NFCT
1766 .procname = "conntrack",
1767 .maxlen = sizeof(int),
1769 .proc_handler = &proc_dointvec,
1773 .procname = "secure_tcp",
1774 .maxlen = sizeof(int),
1776 .proc_handler = proc_do_defense_mode,
1779 .procname = "snat_reroute",
1780 .maxlen = sizeof(int),
1782 .proc_handler = &proc_dointvec,
1785 .procname = "sync_version",
1786 .maxlen = sizeof(int),
1788 .proc_handler = proc_dointvec_minmax,
1793 .procname = "sync_ports",
1794 .maxlen = sizeof(int),
1796 .proc_handler = proc_do_sync_ports,
1799 .procname = "sync_persist_mode",
1800 .maxlen = sizeof(int),
1802 .proc_handler = proc_dointvec,
1805 .procname = "sync_qlen_max",
1806 .maxlen = sizeof(unsigned long),
1808 .proc_handler = proc_doulongvec_minmax,
1811 .procname = "sync_sock_size",
1812 .maxlen = sizeof(int),
1814 .proc_handler = proc_dointvec,
1817 .procname = "cache_bypass",
1818 .maxlen = sizeof(int),
1820 .proc_handler = proc_dointvec,
1823 .procname = "expire_nodest_conn",
1824 .maxlen = sizeof(int),
1826 .proc_handler = proc_dointvec,
1829 .procname = "sloppy_tcp",
1830 .maxlen = sizeof(int),
1832 .proc_handler = proc_dointvec,
1835 .procname = "sloppy_sctp",
1836 .maxlen = sizeof(int),
1838 .proc_handler = proc_dointvec,
1841 .procname = "expire_quiescent_template",
1842 .maxlen = sizeof(int),
1844 .proc_handler = proc_dointvec,
1847 .procname = "sync_threshold",
1849 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1851 .proc_handler = proc_do_sync_threshold,
1854 .procname = "sync_refresh_period",
1855 .maxlen = sizeof(int),
1857 .proc_handler = proc_dointvec_jiffies,
1860 .procname = "sync_retries",
1861 .maxlen = sizeof(int),
1863 .proc_handler = proc_dointvec_minmax,
1868 .procname = "nat_icmp_send",
1869 .maxlen = sizeof(int),
1871 .proc_handler = proc_dointvec,
1874 .procname = "pmtu_disc",
1875 .maxlen = sizeof(int),
1877 .proc_handler = proc_dointvec,
1880 .procname = "backup_only",
1881 .maxlen = sizeof(int),
1883 .proc_handler = proc_dointvec,
1886 .procname = "conn_reuse_mode",
1887 .maxlen = sizeof(int),
1889 .proc_handler = proc_dointvec,
1892 .procname = "schedule_icmp",
1893 .maxlen = sizeof(int),
1895 .proc_handler = proc_dointvec,
1898 .procname = "ignore_tunneled",
1899 .maxlen = sizeof(int),
1901 .proc_handler = proc_dointvec,
1903 #ifdef CONFIG_IP_VS_DEBUG
1905 .procname = "debug_level",
1906 .data = &sysctl_ip_vs_debug_level,
1907 .maxlen = sizeof(int),
1909 .proc_handler = proc_dointvec,
1917 #ifdef CONFIG_PROC_FS
1920 struct seq_net_private p; /* Do not move this, netns depends upon it*/
1921 struct hlist_head *table;
1926 * Write the contents of the VS rule table to a PROCfs file.
1927 * (It is kept just for backward compatibility)
1929 static inline const char *ip_vs_fwd_name(unsigned int flags)
1931 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1932 case IP_VS_CONN_F_LOCALNODE:
1934 case IP_VS_CONN_F_TUNNEL:
1936 case IP_VS_CONN_F_DROUTE:
1944 /* Get the Nth entry in the two lists */
1945 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1947 struct net *net = seq_file_net(seq);
1948 struct netns_ipvs *ipvs = net_ipvs(net);
1949 struct ip_vs_iter *iter = seq->private;
1951 struct ip_vs_service *svc;
1953 /* look in hash by protocol */
1954 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1955 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
1956 if ((svc->ipvs == ipvs) && pos-- == 0) {
1957 iter->table = ip_vs_svc_table;
1964 /* keep looking in fwmark */
1965 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1966 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
1968 if ((svc->ipvs == ipvs) && pos-- == 0) {
1969 iter->table = ip_vs_svc_fwm_table;
1979 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1983 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1987 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1989 struct hlist_node *e;
1990 struct ip_vs_iter *iter;
1991 struct ip_vs_service *svc;
1994 if (v == SEQ_START_TOKEN)
1995 return ip_vs_info_array(seq,0);
1998 iter = seq->private;
2000 if (iter->table == ip_vs_svc_table) {
2001 /* next service in table hashed by protocol */
2002 e = rcu_dereference(hlist_next_rcu(&svc->s_list));
2004 return hlist_entry(e, struct ip_vs_service, s_list);
2006 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2007 hlist_for_each_entry_rcu(svc,
2008 &ip_vs_svc_table[iter->bucket],
2014 iter->table = ip_vs_svc_fwm_table;
2019 /* next service in hashed by fwmark */
2020 e = rcu_dereference(hlist_next_rcu(&svc->f_list));
2022 return hlist_entry(e, struct ip_vs_service, f_list);
2025 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2026 hlist_for_each_entry_rcu(svc,
2027 &ip_vs_svc_fwm_table[iter->bucket],
2035 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
2042 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
2044 if (v == SEQ_START_TOKEN) {
2046 "IP Virtual Server version %d.%d.%d (size=%d)\n",
2047 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2049 "Prot LocalAddress:Port Scheduler Flags\n");
2051 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2053 struct net *net = seq_file_net(seq);
2054 struct netns_ipvs *ipvs = net_ipvs(net);
2055 const struct ip_vs_service *svc = v;
2056 const struct ip_vs_iter *iter = seq->private;
2057 const struct ip_vs_dest *dest;
2058 struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler);
2059 char *sched_name = sched ? sched->name : "none";
2061 if (svc->ipvs != ipvs)
2063 if (iter->table == ip_vs_svc_table) {
2064 #ifdef CONFIG_IP_VS_IPV6
2065 if (svc->af == AF_INET6)
2066 seq_printf(seq, "%s [%pI6]:%04X %s ",
2067 ip_vs_proto_name(svc->protocol),
2073 seq_printf(seq, "%s %08X:%04X %s %s ",
2074 ip_vs_proto_name(svc->protocol),
2075 ntohl(svc->addr.ip),
2078 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2080 seq_printf(seq, "FWM %08X %s %s",
2081 svc->fwmark, sched_name,
2082 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2085 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2086 seq_printf(seq, "persistent %d %08X\n",
2088 ntohl(svc->netmask));
2090 seq_putc(seq, '\n');
2092 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
2093 #ifdef CONFIG_IP_VS_IPV6
2094 if (dest->af == AF_INET6)
2097 " %-7s %-6d %-10d %-10d\n",
2100 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2101 atomic_read(&dest->weight),
2102 atomic_read(&dest->activeconns),
2103 atomic_read(&dest->inactconns));
2108 "%-7s %-6d %-10d %-10d\n",
2109 ntohl(dest->addr.ip),
2111 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2112 atomic_read(&dest->weight),
2113 atomic_read(&dest->activeconns),
2114 atomic_read(&dest->inactconns));
2121 static const struct seq_operations ip_vs_info_seq_ops = {
2122 .start = ip_vs_info_seq_start,
2123 .next = ip_vs_info_seq_next,
2124 .stop = ip_vs_info_seq_stop,
2125 .show = ip_vs_info_seq_show,
2128 static int ip_vs_info_open(struct inode *inode, struct file *file)
2130 return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2131 sizeof(struct ip_vs_iter));
2134 static const struct file_operations ip_vs_info_fops = {
2135 .owner = THIS_MODULE,
2136 .open = ip_vs_info_open,
2138 .llseek = seq_lseek,
2139 .release = seq_release_net,
2142 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2144 struct net *net = seq_file_single_net(seq);
2145 struct ip_vs_kstats show;
2147 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2149 " Total Incoming Outgoing Incoming Outgoing\n");
2151 " Conns Packets Packets Bytes Bytes\n");
2153 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2154 seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n\n",
2155 (unsigned long long)show.conns,
2156 (unsigned long long)show.inpkts,
2157 (unsigned long long)show.outpkts,
2158 (unsigned long long)show.inbytes,
2159 (unsigned long long)show.outbytes);
2161 /* 01234567 01234567 01234567 0123456701234567 0123456701234567*/
2163 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2164 seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n",
2165 (unsigned long long)show.cps,
2166 (unsigned long long)show.inpps,
2167 (unsigned long long)show.outpps,
2168 (unsigned long long)show.inbps,
2169 (unsigned long long)show.outbps);
2174 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2176 return single_open_net(inode, file, ip_vs_stats_show);
2179 static const struct file_operations ip_vs_stats_fops = {
2180 .owner = THIS_MODULE,
2181 .open = ip_vs_stats_seq_open,
2183 .llseek = seq_lseek,
2184 .release = single_release_net,
2187 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2189 struct net *net = seq_file_single_net(seq);
2190 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2191 struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats;
2192 struct ip_vs_kstats kstats;
2195 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2197 " Total Incoming Outgoing Incoming Outgoing\n");
2199 "CPU Conns Packets Packets Bytes Bytes\n");
2201 for_each_possible_cpu(i) {
2202 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2204 u64 conns, inpkts, outpkts, inbytes, outbytes;
2207 start = u64_stats_fetch_begin_irq(&u->syncp);
2208 conns = u->cnt.conns;
2209 inpkts = u->cnt.inpkts;
2210 outpkts = u->cnt.outpkts;
2211 inbytes = u->cnt.inbytes;
2212 outbytes = u->cnt.outbytes;
2213 } while (u64_stats_fetch_retry_irq(&u->syncp, start));
2215 seq_printf(seq, "%3X %8LX %8LX %8LX %16LX %16LX\n",
2216 i, (u64)conns, (u64)inpkts,
2217 (u64)outpkts, (u64)inbytes,
2221 ip_vs_copy_stats(&kstats, tot_stats);
2223 seq_printf(seq, " ~ %8LX %8LX %8LX %16LX %16LX\n\n",
2224 (unsigned long long)kstats.conns,
2225 (unsigned long long)kstats.inpkts,
2226 (unsigned long long)kstats.outpkts,
2227 (unsigned long long)kstats.inbytes,
2228 (unsigned long long)kstats.outbytes);
2230 /* ... 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2232 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2233 seq_printf(seq, " %8LX %8LX %8LX %16LX %16LX\n",
2243 static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2245 return single_open_net(inode, file, ip_vs_stats_percpu_show);
2248 static const struct file_operations ip_vs_stats_percpu_fops = {
2249 .owner = THIS_MODULE,
2250 .open = ip_vs_stats_percpu_seq_open,
2252 .llseek = seq_lseek,
2253 .release = single_release_net,
2258 * Set timeout values for tcp tcpfin udp in the timeout_table.
2260 static int ip_vs_set_timeout(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2262 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2263 struct ip_vs_proto_data *pd;
2266 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2271 #ifdef CONFIG_IP_VS_PROTO_TCP
2272 if (u->tcp_timeout < 0 || u->tcp_timeout > (INT_MAX / HZ) ||
2273 u->tcp_fin_timeout < 0 || u->tcp_fin_timeout > (INT_MAX / HZ)) {
2278 #ifdef CONFIG_IP_VS_PROTO_UDP
2279 if (u->udp_timeout < 0 || u->udp_timeout > (INT_MAX / HZ))
2283 #ifdef CONFIG_IP_VS_PROTO_TCP
2284 if (u->tcp_timeout) {
2285 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2286 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2287 = u->tcp_timeout * HZ;
2290 if (u->tcp_fin_timeout) {
2291 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2292 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2293 = u->tcp_fin_timeout * HZ;
2297 #ifdef CONFIG_IP_VS_PROTO_UDP
2298 if (u->udp_timeout) {
2299 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2300 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2301 = u->udp_timeout * HZ;
2307 #define CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2309 struct ip_vs_svcdest_user {
2310 struct ip_vs_service_user s;
2311 struct ip_vs_dest_user d;
2314 static const unsigned char set_arglen[CMDID(IP_VS_SO_SET_MAX) + 1] = {
2315 [CMDID(IP_VS_SO_SET_ADD)] = sizeof(struct ip_vs_service_user),
2316 [CMDID(IP_VS_SO_SET_EDIT)] = sizeof(struct ip_vs_service_user),
2317 [CMDID(IP_VS_SO_SET_DEL)] = sizeof(struct ip_vs_service_user),
2318 [CMDID(IP_VS_SO_SET_ADDDEST)] = sizeof(struct ip_vs_svcdest_user),
2319 [CMDID(IP_VS_SO_SET_DELDEST)] = sizeof(struct ip_vs_svcdest_user),
2320 [CMDID(IP_VS_SO_SET_EDITDEST)] = sizeof(struct ip_vs_svcdest_user),
2321 [CMDID(IP_VS_SO_SET_TIMEOUT)] = sizeof(struct ip_vs_timeout_user),
2322 [CMDID(IP_VS_SO_SET_STARTDAEMON)] = sizeof(struct ip_vs_daemon_user),
2323 [CMDID(IP_VS_SO_SET_STOPDAEMON)] = sizeof(struct ip_vs_daemon_user),
2324 [CMDID(IP_VS_SO_SET_ZERO)] = sizeof(struct ip_vs_service_user),
2327 union ip_vs_set_arglen {
2328 struct ip_vs_service_user field_IP_VS_SO_SET_ADD;
2329 struct ip_vs_service_user field_IP_VS_SO_SET_EDIT;
2330 struct ip_vs_service_user field_IP_VS_SO_SET_DEL;
2331 struct ip_vs_svcdest_user field_IP_VS_SO_SET_ADDDEST;
2332 struct ip_vs_svcdest_user field_IP_VS_SO_SET_DELDEST;
2333 struct ip_vs_svcdest_user field_IP_VS_SO_SET_EDITDEST;
2334 struct ip_vs_timeout_user field_IP_VS_SO_SET_TIMEOUT;
2335 struct ip_vs_daemon_user field_IP_VS_SO_SET_STARTDAEMON;
2336 struct ip_vs_daemon_user field_IP_VS_SO_SET_STOPDAEMON;
2337 struct ip_vs_service_user field_IP_VS_SO_SET_ZERO;
2340 #define MAX_SET_ARGLEN sizeof(union ip_vs_set_arglen)
2342 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2343 struct ip_vs_service_user *usvc_compat)
2345 memset(usvc, 0, sizeof(*usvc));
2348 usvc->protocol = usvc_compat->protocol;
2349 usvc->addr.ip = usvc_compat->addr;
2350 usvc->port = usvc_compat->port;
2351 usvc->fwmark = usvc_compat->fwmark;
2353 /* Deep copy of sched_name is not needed here */
2354 usvc->sched_name = usvc_compat->sched_name;
2356 usvc->flags = usvc_compat->flags;
2357 usvc->timeout = usvc_compat->timeout;
2358 usvc->netmask = usvc_compat->netmask;
2361 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2362 struct ip_vs_dest_user *udest_compat)
2364 memset(udest, 0, sizeof(*udest));
2366 udest->addr.ip = udest_compat->addr;
2367 udest->port = udest_compat->port;
2368 udest->conn_flags = udest_compat->conn_flags;
2369 udest->weight = udest_compat->weight;
2370 udest->u_threshold = udest_compat->u_threshold;
2371 udest->l_threshold = udest_compat->l_threshold;
2372 udest->af = AF_INET;
2376 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2378 struct net *net = sock_net(sk);
2380 unsigned char arg[MAX_SET_ARGLEN];
2381 struct ip_vs_service_user *usvc_compat;
2382 struct ip_vs_service_user_kern usvc;
2383 struct ip_vs_service *svc;
2384 struct ip_vs_dest_user *udest_compat;
2385 struct ip_vs_dest_user_kern udest;
2386 struct netns_ipvs *ipvs = net_ipvs(net);
2388 BUILD_BUG_ON(sizeof(arg) > 255);
2389 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2392 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2394 if (len != set_arglen[CMDID(cmd)]) {
2395 IP_VS_DBG(1, "set_ctl: len %u != %u\n",
2396 len, set_arglen[CMDID(cmd)]);
2400 if (copy_from_user(arg, user, len) != 0)
2403 /* Handle daemons since they have another lock */
2404 if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2405 cmd == IP_VS_SO_SET_STOPDAEMON) {
2406 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2408 if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2409 struct ipvs_sync_daemon_cfg cfg;
2411 memset(&cfg, 0, sizeof(cfg));
2413 if (strscpy(cfg.mcast_ifn, dm->mcast_ifn,
2414 sizeof(cfg.mcast_ifn)) <= 0)
2416 cfg.syncid = dm->syncid;
2417 ret = start_sync_thread(ipvs, &cfg, dm->state);
2419 ret = stop_sync_thread(ipvs, dm->state);
2424 mutex_lock(&__ip_vs_mutex);
2425 if (cmd == IP_VS_SO_SET_FLUSH) {
2426 /* Flush the virtual service */
2427 ret = ip_vs_flush(ipvs, false);
2429 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2430 /* Set timeout values for (tcp tcpfin udp) */
2431 ret = ip_vs_set_timeout(ipvs, (struct ip_vs_timeout_user *)arg);
2434 /* No more commands with len == 0 below */
2439 usvc_compat = (struct ip_vs_service_user *)arg;
2440 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2442 /* We only use the new structs internally, so copy userspace compat
2443 * structs to extended internal versions */
2444 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2445 ip_vs_copy_udest_compat(&udest, udest_compat);
2447 if (cmd == IP_VS_SO_SET_ZERO) {
2448 /* if no service address is set, zero counters in all */
2449 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2450 ret = ip_vs_zero_all(ipvs);
2455 if ((cmd == IP_VS_SO_SET_ADD || cmd == IP_VS_SO_SET_EDIT) &&
2456 strnlen(usvc.sched_name, IP_VS_SCHEDNAME_MAXLEN) ==
2457 IP_VS_SCHEDNAME_MAXLEN) {
2462 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2463 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2464 usvc.protocol != IPPROTO_SCTP) {
2465 pr_err("set_ctl: invalid protocol: %d %pI4:%d\n",
2466 usvc.protocol, &usvc.addr.ip,
2472 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2474 if (usvc.fwmark == 0)
2475 svc = __ip_vs_service_find(ipvs, usvc.af, usvc.protocol,
2476 &usvc.addr, usvc.port);
2478 svc = __ip_vs_svc_fwm_find(ipvs, usvc.af, usvc.fwmark);
2481 if (cmd != IP_VS_SO_SET_ADD
2482 && (svc == NULL || svc->protocol != usvc.protocol)) {
2488 case IP_VS_SO_SET_ADD:
2492 ret = ip_vs_add_service(ipvs, &usvc, &svc);
2494 case IP_VS_SO_SET_EDIT:
2495 ret = ip_vs_edit_service(svc, &usvc);
2497 case IP_VS_SO_SET_DEL:
2498 ret = ip_vs_del_service(svc);
2502 case IP_VS_SO_SET_ZERO:
2503 ret = ip_vs_zero_service(svc);
2505 case IP_VS_SO_SET_ADDDEST:
2506 ret = ip_vs_add_dest(svc, &udest);
2508 case IP_VS_SO_SET_EDITDEST:
2509 ret = ip_vs_edit_dest(svc, &udest);
2511 case IP_VS_SO_SET_DELDEST:
2512 ret = ip_vs_del_dest(svc, &udest);
2516 mutex_unlock(&__ip_vs_mutex);
2522 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2524 struct ip_vs_scheduler *sched;
2525 struct ip_vs_kstats kstats;
2528 sched = rcu_dereference_protected(src->scheduler, 1);
2529 sched_name = sched ? sched->name : "none";
2530 dst->protocol = src->protocol;
2531 dst->addr = src->addr.ip;
2532 dst->port = src->port;
2533 dst->fwmark = src->fwmark;
2534 strlcpy(dst->sched_name, sched_name, sizeof(dst->sched_name));
2535 dst->flags = src->flags;
2536 dst->timeout = src->timeout / HZ;
2537 dst->netmask = src->netmask;
2538 dst->num_dests = src->num_dests;
2539 ip_vs_copy_stats(&kstats, &src->stats);
2540 ip_vs_export_stats_user(&dst->stats, &kstats);
2544 __ip_vs_get_service_entries(struct netns_ipvs *ipvs,
2545 const struct ip_vs_get_services *get,
2546 struct ip_vs_get_services __user *uptr)
2549 struct ip_vs_service *svc;
2550 struct ip_vs_service_entry entry;
2553 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2554 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2555 /* Only expose IPv4 entries to old interface */
2556 if (svc->af != AF_INET || (svc->ipvs != ipvs))
2559 if (count >= get->num_services)
2561 memset(&entry, 0, sizeof(entry));
2562 ip_vs_copy_service(&entry, svc);
2563 if (copy_to_user(&uptr->entrytable[count],
2564 &entry, sizeof(entry))) {
2572 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2573 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2574 /* Only expose IPv4 entries to old interface */
2575 if (svc->af != AF_INET || (svc->ipvs != ipvs))
2578 if (count >= get->num_services)
2580 memset(&entry, 0, sizeof(entry));
2581 ip_vs_copy_service(&entry, svc);
2582 if (copy_to_user(&uptr->entrytable[count],
2583 &entry, sizeof(entry))) {
2595 __ip_vs_get_dest_entries(struct netns_ipvs *ipvs, const struct ip_vs_get_dests *get,
2596 struct ip_vs_get_dests __user *uptr)
2598 struct ip_vs_service *svc;
2599 union nf_inet_addr addr = { .ip = get->addr };
2604 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, get->fwmark);
2606 svc = __ip_vs_service_find(ipvs, AF_INET, get->protocol, &addr,
2612 struct ip_vs_dest *dest;
2613 struct ip_vs_dest_entry entry;
2614 struct ip_vs_kstats kstats;
2616 memset(&entry, 0, sizeof(entry));
2617 list_for_each_entry(dest, &svc->destinations, n_list) {
2618 if (count >= get->num_dests)
2621 /* Cannot expose heterogeneous members via sockopt
2624 if (dest->af != svc->af)
2627 entry.addr = dest->addr.ip;
2628 entry.port = dest->port;
2629 entry.conn_flags = atomic_read(&dest->conn_flags);
2630 entry.weight = atomic_read(&dest->weight);
2631 entry.u_threshold = dest->u_threshold;
2632 entry.l_threshold = dest->l_threshold;
2633 entry.activeconns = atomic_read(&dest->activeconns);
2634 entry.inactconns = atomic_read(&dest->inactconns);
2635 entry.persistconns = atomic_read(&dest->persistconns);
2636 ip_vs_copy_stats(&kstats, &dest->stats);
2637 ip_vs_export_stats_user(&entry.stats, &kstats);
2638 if (copy_to_user(&uptr->entrytable[count],
2639 &entry, sizeof(entry))) {
2651 __ip_vs_get_timeouts(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2653 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2654 struct ip_vs_proto_data *pd;
2657 memset(u, 0, sizeof (*u));
2659 #ifdef CONFIG_IP_VS_PROTO_TCP
2660 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2661 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2662 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2664 #ifdef CONFIG_IP_VS_PROTO_UDP
2665 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2667 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2671 static const unsigned char get_arglen[CMDID(IP_VS_SO_GET_MAX) + 1] = {
2672 [CMDID(IP_VS_SO_GET_VERSION)] = 64,
2673 [CMDID(IP_VS_SO_GET_INFO)] = sizeof(struct ip_vs_getinfo),
2674 [CMDID(IP_VS_SO_GET_SERVICES)] = sizeof(struct ip_vs_get_services),
2675 [CMDID(IP_VS_SO_GET_SERVICE)] = sizeof(struct ip_vs_service_entry),
2676 [CMDID(IP_VS_SO_GET_DESTS)] = sizeof(struct ip_vs_get_dests),
2677 [CMDID(IP_VS_SO_GET_TIMEOUT)] = sizeof(struct ip_vs_timeout_user),
2678 [CMDID(IP_VS_SO_GET_DAEMON)] = 2 * sizeof(struct ip_vs_daemon_user),
2681 union ip_vs_get_arglen {
2682 char field_IP_VS_SO_GET_VERSION[64];
2683 struct ip_vs_getinfo field_IP_VS_SO_GET_INFO;
2684 struct ip_vs_get_services field_IP_VS_SO_GET_SERVICES;
2685 struct ip_vs_service_entry field_IP_VS_SO_GET_SERVICE;
2686 struct ip_vs_get_dests field_IP_VS_SO_GET_DESTS;
2687 struct ip_vs_timeout_user field_IP_VS_SO_GET_TIMEOUT;
2688 struct ip_vs_daemon_user field_IP_VS_SO_GET_DAEMON[2];
2691 #define MAX_GET_ARGLEN sizeof(union ip_vs_get_arglen)
2694 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2696 unsigned char arg[MAX_GET_ARGLEN];
2698 unsigned int copylen;
2699 struct net *net = sock_net(sk);
2700 struct netns_ipvs *ipvs = net_ipvs(net);
2703 BUILD_BUG_ON(sizeof(arg) > 255);
2704 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2707 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2710 copylen = get_arglen[CMDID(cmd)];
2711 if (*len < (int) copylen) {
2712 IP_VS_DBG(1, "get_ctl: len %d < %u\n", *len, copylen);
2716 if (copy_from_user(arg, user, copylen) != 0)
2719 * Handle daemons first since it has its own locking
2721 if (cmd == IP_VS_SO_GET_DAEMON) {
2722 struct ip_vs_daemon_user d[2];
2724 memset(&d, 0, sizeof(d));
2725 mutex_lock(&ipvs->sync_mutex);
2726 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2727 d[0].state = IP_VS_STATE_MASTER;
2728 strlcpy(d[0].mcast_ifn, ipvs->mcfg.mcast_ifn,
2729 sizeof(d[0].mcast_ifn));
2730 d[0].syncid = ipvs->mcfg.syncid;
2732 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2733 d[1].state = IP_VS_STATE_BACKUP;
2734 strlcpy(d[1].mcast_ifn, ipvs->bcfg.mcast_ifn,
2735 sizeof(d[1].mcast_ifn));
2736 d[1].syncid = ipvs->bcfg.syncid;
2738 if (copy_to_user(user, &d, sizeof(d)) != 0)
2740 mutex_unlock(&ipvs->sync_mutex);
2744 mutex_lock(&__ip_vs_mutex);
2746 case IP_VS_SO_GET_VERSION:
2750 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2751 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2752 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2756 *len = strlen(buf)+1;
2760 case IP_VS_SO_GET_INFO:
2762 struct ip_vs_getinfo info;
2763 info.version = IP_VS_VERSION_CODE;
2764 info.size = ip_vs_conn_tab_size;
2765 info.num_services = ipvs->num_services;
2766 if (copy_to_user(user, &info, sizeof(info)) != 0)
2771 case IP_VS_SO_GET_SERVICES:
2773 struct ip_vs_get_services *get;
2776 get = (struct ip_vs_get_services *)arg;
2777 size = sizeof(*get) +
2778 sizeof(struct ip_vs_service_entry) * get->num_services;
2780 pr_err("length: %u != %u\n", *len, size);
2784 ret = __ip_vs_get_service_entries(ipvs, get, user);
2788 case IP_VS_SO_GET_SERVICE:
2790 struct ip_vs_service_entry *entry;
2791 struct ip_vs_service *svc;
2792 union nf_inet_addr addr;
2794 entry = (struct ip_vs_service_entry *)arg;
2795 addr.ip = entry->addr;
2798 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, entry->fwmark);
2800 svc = __ip_vs_service_find(ipvs, AF_INET,
2801 entry->protocol, &addr,
2805 ip_vs_copy_service(entry, svc);
2806 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2813 case IP_VS_SO_GET_DESTS:
2815 struct ip_vs_get_dests *get;
2818 get = (struct ip_vs_get_dests *)arg;
2819 size = sizeof(*get) +
2820 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2822 pr_err("length: %u != %u\n", *len, size);
2826 ret = __ip_vs_get_dest_entries(ipvs, get, user);
2830 case IP_VS_SO_GET_TIMEOUT:
2832 struct ip_vs_timeout_user t;
2834 __ip_vs_get_timeouts(ipvs, &t);
2835 if (copy_to_user(user, &t, sizeof(t)) != 0)
2845 mutex_unlock(&__ip_vs_mutex);
2850 static struct nf_sockopt_ops ip_vs_sockopts = {
2852 .set_optmin = IP_VS_BASE_CTL,
2853 .set_optmax = IP_VS_SO_SET_MAX+1,
2854 .set = do_ip_vs_set_ctl,
2855 .get_optmin = IP_VS_BASE_CTL,
2856 .get_optmax = IP_VS_SO_GET_MAX+1,
2857 .get = do_ip_vs_get_ctl,
2858 .owner = THIS_MODULE,
2862 * Generic Netlink interface
2865 /* IPVS genetlink family */
2866 static struct genl_family ip_vs_genl_family;
2868 /* Policy used for first-level command attributes */
2869 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2870 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2871 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2872 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2873 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2874 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2875 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2878 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2879 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2880 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2881 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2882 .len = IP_VS_IFNAME_MAXLEN - 1 },
2883 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2884 [IPVS_DAEMON_ATTR_SYNC_MAXLEN] = { .type = NLA_U16 },
2885 [IPVS_DAEMON_ATTR_MCAST_GROUP] = { .type = NLA_U32 },
2886 [IPVS_DAEMON_ATTR_MCAST_GROUP6] = { .len = sizeof(struct in6_addr) },
2887 [IPVS_DAEMON_ATTR_MCAST_PORT] = { .type = NLA_U16 },
2888 [IPVS_DAEMON_ATTR_MCAST_TTL] = { .type = NLA_U8 },
2891 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2892 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2893 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2894 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2895 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2896 .len = sizeof(union nf_inet_addr) },
2897 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2898 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2899 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2900 .len = IP_VS_SCHEDNAME_MAXLEN - 1 },
2901 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2902 .len = IP_VS_PENAME_MAXLEN },
2903 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2904 .len = sizeof(struct ip_vs_flags) },
2905 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2906 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2907 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2910 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2911 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2912 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2913 .len = sizeof(union nf_inet_addr) },
2914 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2915 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2916 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2917 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2918 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2919 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2920 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2921 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2922 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2923 [IPVS_DEST_ATTR_ADDR_FAMILY] = { .type = NLA_U16 },
2926 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2927 struct ip_vs_kstats *kstats)
2929 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2934 if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, (u32)kstats->conns) ||
2935 nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, (u32)kstats->inpkts) ||
2936 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, (u32)kstats->outpkts) ||
2937 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes,
2938 IPVS_STATS_ATTR_PAD) ||
2939 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes,
2940 IPVS_STATS_ATTR_PAD) ||
2941 nla_put_u32(skb, IPVS_STATS_ATTR_CPS, (u32)kstats->cps) ||
2942 nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, (u32)kstats->inpps) ||
2943 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, (u32)kstats->outpps) ||
2944 nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, (u32)kstats->inbps) ||
2945 nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, (u32)kstats->outbps))
2946 goto nla_put_failure;
2947 nla_nest_end(skb, nl_stats);
2952 nla_nest_cancel(skb, nl_stats);
2956 static int ip_vs_genl_fill_stats64(struct sk_buff *skb, int container_type,
2957 struct ip_vs_kstats *kstats)
2959 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2964 if (nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CONNS, kstats->conns,
2965 IPVS_STATS_ATTR_PAD) ||
2966 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPKTS, kstats->inpkts,
2967 IPVS_STATS_ATTR_PAD) ||
2968 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPKTS, kstats->outpkts,
2969 IPVS_STATS_ATTR_PAD) ||
2970 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes,
2971 IPVS_STATS_ATTR_PAD) ||
2972 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes,
2973 IPVS_STATS_ATTR_PAD) ||
2974 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CPS, kstats->cps,
2975 IPVS_STATS_ATTR_PAD) ||
2976 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPPS, kstats->inpps,
2977 IPVS_STATS_ATTR_PAD) ||
2978 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPPS, kstats->outpps,
2979 IPVS_STATS_ATTR_PAD) ||
2980 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBPS, kstats->inbps,
2981 IPVS_STATS_ATTR_PAD) ||
2982 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBPS, kstats->outbps,
2983 IPVS_STATS_ATTR_PAD))
2984 goto nla_put_failure;
2985 nla_nest_end(skb, nl_stats);
2990 nla_nest_cancel(skb, nl_stats);
2994 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2995 struct ip_vs_service *svc)
2997 struct ip_vs_scheduler *sched;
2998 struct ip_vs_pe *pe;
2999 struct nlattr *nl_service;
3000 struct ip_vs_flags flags = { .flags = svc->flags,
3002 struct ip_vs_kstats kstats;
3005 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
3009 if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
3010 goto nla_put_failure;
3012 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
3013 goto nla_put_failure;
3015 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
3016 nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
3017 nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
3018 goto nla_put_failure;
3021 sched = rcu_dereference_protected(svc->scheduler, 1);
3022 sched_name = sched ? sched->name : "none";
3023 pe = rcu_dereference_protected(svc->pe, 1);
3024 if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched_name) ||
3025 (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
3026 nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
3027 nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
3028 nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
3029 goto nla_put_failure;
3030 ip_vs_copy_stats(&kstats, &svc->stats);
3031 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &kstats))
3032 goto nla_put_failure;
3033 if (ip_vs_genl_fill_stats64(skb, IPVS_SVC_ATTR_STATS64, &kstats))
3034 goto nla_put_failure;
3036 nla_nest_end(skb, nl_service);
3041 nla_nest_cancel(skb, nl_service);
3045 static int ip_vs_genl_dump_service(struct sk_buff *skb,
3046 struct ip_vs_service *svc,
3047 struct netlink_callback *cb)
3051 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3052 &ip_vs_genl_family, NLM_F_MULTI,
3053 IPVS_CMD_NEW_SERVICE);
3057 if (ip_vs_genl_fill_service(skb, svc) < 0)
3058 goto nla_put_failure;
3060 genlmsg_end(skb, hdr);
3064 genlmsg_cancel(skb, hdr);
3068 static int ip_vs_genl_dump_services(struct sk_buff *skb,
3069 struct netlink_callback *cb)
3072 int start = cb->args[0];
3073 struct ip_vs_service *svc;
3074 struct net *net = sock_net(skb->sk);
3075 struct netns_ipvs *ipvs = net_ipvs(net);
3077 mutex_lock(&__ip_vs_mutex);
3078 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3079 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
3080 if (++idx <= start || (svc->ipvs != ipvs))
3082 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3084 goto nla_put_failure;
3089 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3090 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
3091 if (++idx <= start || (svc->ipvs != ipvs))
3093 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3095 goto nla_put_failure;
3101 mutex_unlock(&__ip_vs_mutex);
3107 static bool ip_vs_is_af_valid(int af)
3111 #ifdef CONFIG_IP_VS_IPV6
3112 if (af == AF_INET6 && ipv6_mod_enabled())
3118 static int ip_vs_genl_parse_service(struct netns_ipvs *ipvs,
3119 struct ip_vs_service_user_kern *usvc,
3120 struct nlattr *nla, int full_entry,
3121 struct ip_vs_service **ret_svc)
3123 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
3124 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
3125 struct ip_vs_service *svc;
3127 /* Parse mandatory identifying service fields first */
3129 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla,
3130 ip_vs_svc_policy, NULL))
3133 nla_af = attrs[IPVS_SVC_ATTR_AF];
3134 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
3135 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
3136 nla_port = attrs[IPVS_SVC_ATTR_PORT];
3137 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
3139 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3142 memset(usvc, 0, sizeof(*usvc));
3144 usvc->af = nla_get_u16(nla_af);
3145 if (!ip_vs_is_af_valid(usvc->af))
3146 return -EAFNOSUPPORT;
3149 usvc->protocol = IPPROTO_TCP;
3150 usvc->fwmark = nla_get_u32(nla_fwmark);
3152 usvc->protocol = nla_get_u16(nla_protocol);
3153 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3154 usvc->port = nla_get_be16(nla_port);
3160 svc = __ip_vs_svc_fwm_find(ipvs, usvc->af, usvc->fwmark);
3162 svc = __ip_vs_service_find(ipvs, usvc->af, usvc->protocol,
3163 &usvc->addr, usvc->port);
3167 /* If a full entry was requested, check for the additional fields */
3169 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3171 struct ip_vs_flags flags;
3173 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3174 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3175 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3176 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3177 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3179 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3182 nla_memcpy(&flags, nla_flags, sizeof(flags));
3184 /* prefill flags from service if it already exists */
3186 usvc->flags = svc->flags;
3188 /* set new flags from userland */
3189 usvc->flags = (usvc->flags & ~flags.mask) |
3190 (flags.flags & flags.mask);
3191 usvc->sched_name = nla_data(nla_sched);
3192 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3193 usvc->timeout = nla_get_u32(nla_timeout);
3194 usvc->netmask = nla_get_be32(nla_netmask);
3200 static struct ip_vs_service *ip_vs_genl_find_service(struct netns_ipvs *ipvs,
3203 struct ip_vs_service_user_kern usvc;
3204 struct ip_vs_service *svc;
3207 ret = ip_vs_genl_parse_service(ipvs, &usvc, nla, 0, &svc);
3208 return ret ? ERR_PTR(ret) : svc;
3211 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3213 struct nlattr *nl_dest;
3214 struct ip_vs_kstats kstats;
3216 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3220 if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3221 nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3222 nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3223 (atomic_read(&dest->conn_flags) &
3224 IP_VS_CONN_F_FWD_MASK)) ||
3225 nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3226 atomic_read(&dest->weight)) ||
3227 nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3228 nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3229 nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3230 atomic_read(&dest->activeconns)) ||
3231 nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3232 atomic_read(&dest->inactconns)) ||
3233 nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3234 atomic_read(&dest->persistconns)) ||
3235 nla_put_u16(skb, IPVS_DEST_ATTR_ADDR_FAMILY, dest->af))
3236 goto nla_put_failure;
3237 ip_vs_copy_stats(&kstats, &dest->stats);
3238 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &kstats))
3239 goto nla_put_failure;
3240 if (ip_vs_genl_fill_stats64(skb, IPVS_DEST_ATTR_STATS64, &kstats))
3241 goto nla_put_failure;
3243 nla_nest_end(skb, nl_dest);
3248 nla_nest_cancel(skb, nl_dest);
3252 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3253 struct netlink_callback *cb)
3257 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3258 &ip_vs_genl_family, NLM_F_MULTI,
3263 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3264 goto nla_put_failure;
3266 genlmsg_end(skb, hdr);
3270 genlmsg_cancel(skb, hdr);
3274 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3275 struct netlink_callback *cb)
3278 int start = cb->args[0];
3279 struct ip_vs_service *svc;
3280 struct ip_vs_dest *dest;
3281 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3282 struct net *net = sock_net(skb->sk);
3283 struct netns_ipvs *ipvs = net_ipvs(net);
3285 mutex_lock(&__ip_vs_mutex);
3287 /* Try to find the service for which to dump destinations */
3288 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs, IPVS_CMD_ATTR_MAX,
3289 ip_vs_cmd_policy, NULL))
3293 svc = ip_vs_genl_find_service(ipvs, attrs[IPVS_CMD_ATTR_SERVICE]);
3294 if (IS_ERR_OR_NULL(svc))
3297 /* Dump the destinations */
3298 list_for_each_entry(dest, &svc->destinations, n_list) {
3301 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3303 goto nla_put_failure;
3311 mutex_unlock(&__ip_vs_mutex);
3316 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3317 struct nlattr *nla, int full_entry)
3319 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3320 struct nlattr *nla_addr, *nla_port;
3321 struct nlattr *nla_addr_family;
3323 /* Parse mandatory identifying destination fields first */
3325 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla,
3326 ip_vs_dest_policy, NULL))
3329 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3330 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3331 nla_addr_family = attrs[IPVS_DEST_ATTR_ADDR_FAMILY];
3333 if (!(nla_addr && nla_port))
3336 memset(udest, 0, sizeof(*udest));
3338 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3339 udest->port = nla_get_be16(nla_port);
3341 if (nla_addr_family)
3342 udest->af = nla_get_u16(nla_addr_family);
3346 /* If a full entry was requested, check for the additional fields */
3348 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3351 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3352 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3353 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3354 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3356 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3359 udest->conn_flags = nla_get_u32(nla_fwd)
3360 & IP_VS_CONN_F_FWD_MASK;
3361 udest->weight = nla_get_u32(nla_weight);
3362 udest->u_threshold = nla_get_u32(nla_u_thresh);
3363 udest->l_threshold = nla_get_u32(nla_l_thresh);
3369 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3370 struct ipvs_sync_daemon_cfg *c)
3372 struct nlattr *nl_daemon;
3374 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3378 if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3379 nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, c->mcast_ifn) ||
3380 nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, c->syncid) ||
3381 nla_put_u16(skb, IPVS_DAEMON_ATTR_SYNC_MAXLEN, c->sync_maxlen) ||
3382 nla_put_u16(skb, IPVS_DAEMON_ATTR_MCAST_PORT, c->mcast_port) ||
3383 nla_put_u8(skb, IPVS_DAEMON_ATTR_MCAST_TTL, c->mcast_ttl))
3384 goto nla_put_failure;
3385 #ifdef CONFIG_IP_VS_IPV6
3386 if (c->mcast_af == AF_INET6) {
3387 if (nla_put_in6_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP6,
3388 &c->mcast_group.in6))
3389 goto nla_put_failure;
3392 if (c->mcast_af == AF_INET &&
3393 nla_put_in_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP,
3395 goto nla_put_failure;
3396 nla_nest_end(skb, nl_daemon);
3401 nla_nest_cancel(skb, nl_daemon);
3405 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3406 struct ipvs_sync_daemon_cfg *c,
3407 struct netlink_callback *cb)
3410 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3411 &ip_vs_genl_family, NLM_F_MULTI,
3412 IPVS_CMD_NEW_DAEMON);
3416 if (ip_vs_genl_fill_daemon(skb, state, c))
3417 goto nla_put_failure;
3419 genlmsg_end(skb, hdr);
3423 genlmsg_cancel(skb, hdr);
3427 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3428 struct netlink_callback *cb)
3430 struct net *net = sock_net(skb->sk);
3431 struct netns_ipvs *ipvs = net_ipvs(net);
3433 mutex_lock(&ipvs->sync_mutex);
3434 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3435 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3436 &ipvs->mcfg, cb) < 0)
3437 goto nla_put_failure;
3442 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3443 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3444 &ipvs->bcfg, cb) < 0)
3445 goto nla_put_failure;
3451 mutex_unlock(&ipvs->sync_mutex);
3456 static int ip_vs_genl_new_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3458 struct ipvs_sync_daemon_cfg c;
3462 memset(&c, 0, sizeof(c));
3463 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3464 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3465 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3467 strlcpy(c.mcast_ifn, nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3468 sizeof(c.mcast_ifn));
3469 c.syncid = nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]);
3471 a = attrs[IPVS_DAEMON_ATTR_SYNC_MAXLEN];
3473 c.sync_maxlen = nla_get_u16(a);
3475 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP];
3477 c.mcast_af = AF_INET;
3478 c.mcast_group.ip = nla_get_in_addr(a);
3479 if (!ipv4_is_multicast(c.mcast_group.ip))
3482 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP6];
3484 #ifdef CONFIG_IP_VS_IPV6
3487 c.mcast_af = AF_INET6;
3488 c.mcast_group.in6 = nla_get_in6_addr(a);
3489 addr_type = ipv6_addr_type(&c.mcast_group.in6);
3490 if (!(addr_type & IPV6_ADDR_MULTICAST))
3493 return -EAFNOSUPPORT;
3498 a = attrs[IPVS_DAEMON_ATTR_MCAST_PORT];
3500 c.mcast_port = nla_get_u16(a);
3502 a = attrs[IPVS_DAEMON_ATTR_MCAST_TTL];
3504 c.mcast_ttl = nla_get_u8(a);
3506 /* The synchronization protocol is incompatible with mixed family
3509 if (ipvs->mixed_address_family_dests > 0)
3512 ret = start_sync_thread(ipvs, &c,
3513 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3517 static int ip_vs_genl_del_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3521 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3524 ret = stop_sync_thread(ipvs,
3525 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3529 static int ip_vs_genl_set_config(struct netns_ipvs *ipvs, struct nlattr **attrs)
3531 struct ip_vs_timeout_user t;
3533 __ip_vs_get_timeouts(ipvs, &t);
3535 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3536 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3538 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3540 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3542 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3543 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3545 return ip_vs_set_timeout(ipvs, &t);
3548 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3550 int ret = -EINVAL, cmd;
3551 struct net *net = sock_net(skb->sk);
3552 struct netns_ipvs *ipvs = net_ipvs(net);
3554 cmd = info->genlhdr->cmd;
3556 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3557 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3559 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3560 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3561 info->attrs[IPVS_CMD_ATTR_DAEMON],
3562 ip_vs_daemon_policy, info->extack))
3565 if (cmd == IPVS_CMD_NEW_DAEMON)
3566 ret = ip_vs_genl_new_daemon(ipvs, daemon_attrs);
3568 ret = ip_vs_genl_del_daemon(ipvs, daemon_attrs);
3575 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3577 struct ip_vs_service *svc = NULL;
3578 struct ip_vs_service_user_kern usvc;
3579 struct ip_vs_dest_user_kern udest;
3581 int need_full_svc = 0, need_full_dest = 0;
3582 struct net *net = sock_net(skb->sk);
3583 struct netns_ipvs *ipvs = net_ipvs(net);
3585 cmd = info->genlhdr->cmd;
3587 mutex_lock(&__ip_vs_mutex);
3589 if (cmd == IPVS_CMD_FLUSH) {
3590 ret = ip_vs_flush(ipvs, false);
3592 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3593 ret = ip_vs_genl_set_config(ipvs, info->attrs);
3595 } else if (cmd == IPVS_CMD_ZERO &&
3596 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3597 ret = ip_vs_zero_all(ipvs);
3601 /* All following commands require a service argument, so check if we
3602 * received a valid one. We need a full service specification when
3603 * adding / editing a service. Only identifying members otherwise. */
3604 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3607 ret = ip_vs_genl_parse_service(ipvs, &usvc,
3608 info->attrs[IPVS_CMD_ATTR_SERVICE],
3609 need_full_svc, &svc);
3613 /* Unless we're adding a new service, the service must already exist */
3614 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3619 /* Destination commands require a valid destination argument. For
3620 * adding / editing a destination, we need a full destination
3622 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3623 cmd == IPVS_CMD_DEL_DEST) {
3624 if (cmd != IPVS_CMD_DEL_DEST)
3627 ret = ip_vs_genl_parse_dest(&udest,
3628 info->attrs[IPVS_CMD_ATTR_DEST],
3633 /* Old protocols did not allow the user to specify address
3634 * family, so we set it to zero instead. We also didn't
3635 * allow heterogeneous pools in the old code, so it's safe
3636 * to assume that this will have the same address family as
3642 if (!ip_vs_is_af_valid(udest.af)) {
3643 ret = -EAFNOSUPPORT;
3647 if (udest.af != svc->af && cmd != IPVS_CMD_DEL_DEST) {
3648 /* The synchronization protocol is incompatible
3649 * with mixed family services
3651 if (ipvs->sync_state) {
3656 /* Which connection types do we support? */
3657 switch (udest.conn_flags) {
3658 case IP_VS_CONN_F_TUNNEL:
3659 /* We are able to forward this */
3669 case IPVS_CMD_NEW_SERVICE:
3671 ret = ip_vs_add_service(ipvs, &usvc, &svc);
3675 case IPVS_CMD_SET_SERVICE:
3676 ret = ip_vs_edit_service(svc, &usvc);
3678 case IPVS_CMD_DEL_SERVICE:
3679 ret = ip_vs_del_service(svc);
3680 /* do not use svc, it can be freed */
3682 case IPVS_CMD_NEW_DEST:
3683 ret = ip_vs_add_dest(svc, &udest);
3685 case IPVS_CMD_SET_DEST:
3686 ret = ip_vs_edit_dest(svc, &udest);
3688 case IPVS_CMD_DEL_DEST:
3689 ret = ip_vs_del_dest(svc, &udest);
3692 ret = ip_vs_zero_service(svc);
3699 mutex_unlock(&__ip_vs_mutex);
3704 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3706 struct sk_buff *msg;
3708 int ret, cmd, reply_cmd;
3709 struct net *net = sock_net(skb->sk);
3710 struct netns_ipvs *ipvs = net_ipvs(net);
3712 cmd = info->genlhdr->cmd;
3714 if (cmd == IPVS_CMD_GET_SERVICE)
3715 reply_cmd = IPVS_CMD_NEW_SERVICE;
3716 else if (cmd == IPVS_CMD_GET_INFO)
3717 reply_cmd = IPVS_CMD_SET_INFO;
3718 else if (cmd == IPVS_CMD_GET_CONFIG)
3719 reply_cmd = IPVS_CMD_SET_CONFIG;
3721 pr_err("unknown Generic Netlink command\n");
3725 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3729 mutex_lock(&__ip_vs_mutex);
3731 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3733 goto nla_put_failure;
3736 case IPVS_CMD_GET_SERVICE:
3738 struct ip_vs_service *svc;
3740 svc = ip_vs_genl_find_service(ipvs,
3741 info->attrs[IPVS_CMD_ATTR_SERVICE]);
3746 ret = ip_vs_genl_fill_service(msg, svc);
3748 goto nla_put_failure;
3757 case IPVS_CMD_GET_CONFIG:
3759 struct ip_vs_timeout_user t;
3761 __ip_vs_get_timeouts(ipvs, &t);
3762 #ifdef CONFIG_IP_VS_PROTO_TCP
3763 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3765 nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3767 goto nla_put_failure;
3769 #ifdef CONFIG_IP_VS_PROTO_UDP
3770 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3771 goto nla_put_failure;
3777 case IPVS_CMD_GET_INFO:
3778 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3779 IP_VS_VERSION_CODE) ||
3780 nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3781 ip_vs_conn_tab_size))
3782 goto nla_put_failure;
3786 genlmsg_end(msg, reply);
3787 ret = genlmsg_reply(msg, info);
3791 pr_err("not enough space in Netlink message\n");
3797 mutex_unlock(&__ip_vs_mutex);
3803 static const struct genl_ops ip_vs_genl_ops[] = {
3805 .cmd = IPVS_CMD_NEW_SERVICE,
3806 .flags = GENL_ADMIN_PERM,
3807 .policy = ip_vs_cmd_policy,
3808 .doit = ip_vs_genl_set_cmd,
3811 .cmd = IPVS_CMD_SET_SERVICE,
3812 .flags = GENL_ADMIN_PERM,
3813 .policy = ip_vs_cmd_policy,
3814 .doit = ip_vs_genl_set_cmd,
3817 .cmd = IPVS_CMD_DEL_SERVICE,
3818 .flags = GENL_ADMIN_PERM,
3819 .policy = ip_vs_cmd_policy,
3820 .doit = ip_vs_genl_set_cmd,
3823 .cmd = IPVS_CMD_GET_SERVICE,
3824 .flags = GENL_ADMIN_PERM,
3825 .doit = ip_vs_genl_get_cmd,
3826 .dumpit = ip_vs_genl_dump_services,
3827 .policy = ip_vs_cmd_policy,
3830 .cmd = IPVS_CMD_NEW_DEST,
3831 .flags = GENL_ADMIN_PERM,
3832 .policy = ip_vs_cmd_policy,
3833 .doit = ip_vs_genl_set_cmd,
3836 .cmd = IPVS_CMD_SET_DEST,
3837 .flags = GENL_ADMIN_PERM,
3838 .policy = ip_vs_cmd_policy,
3839 .doit = ip_vs_genl_set_cmd,
3842 .cmd = IPVS_CMD_DEL_DEST,
3843 .flags = GENL_ADMIN_PERM,
3844 .policy = ip_vs_cmd_policy,
3845 .doit = ip_vs_genl_set_cmd,
3848 .cmd = IPVS_CMD_GET_DEST,
3849 .flags = GENL_ADMIN_PERM,
3850 .policy = ip_vs_cmd_policy,
3851 .dumpit = ip_vs_genl_dump_dests,
3854 .cmd = IPVS_CMD_NEW_DAEMON,
3855 .flags = GENL_ADMIN_PERM,
3856 .policy = ip_vs_cmd_policy,
3857 .doit = ip_vs_genl_set_daemon,
3860 .cmd = IPVS_CMD_DEL_DAEMON,
3861 .flags = GENL_ADMIN_PERM,
3862 .policy = ip_vs_cmd_policy,
3863 .doit = ip_vs_genl_set_daemon,
3866 .cmd = IPVS_CMD_GET_DAEMON,
3867 .flags = GENL_ADMIN_PERM,
3868 .dumpit = ip_vs_genl_dump_daemons,
3871 .cmd = IPVS_CMD_SET_CONFIG,
3872 .flags = GENL_ADMIN_PERM,
3873 .policy = ip_vs_cmd_policy,
3874 .doit = ip_vs_genl_set_cmd,
3877 .cmd = IPVS_CMD_GET_CONFIG,
3878 .flags = GENL_ADMIN_PERM,
3879 .doit = ip_vs_genl_get_cmd,
3882 .cmd = IPVS_CMD_GET_INFO,
3883 .flags = GENL_ADMIN_PERM,
3884 .doit = ip_vs_genl_get_cmd,
3887 .cmd = IPVS_CMD_ZERO,
3888 .flags = GENL_ADMIN_PERM,
3889 .policy = ip_vs_cmd_policy,
3890 .doit = ip_vs_genl_set_cmd,
3893 .cmd = IPVS_CMD_FLUSH,
3894 .flags = GENL_ADMIN_PERM,
3895 .doit = ip_vs_genl_set_cmd,
3899 static struct genl_family ip_vs_genl_family __ro_after_init = {
3901 .name = IPVS_GENL_NAME,
3902 .version = IPVS_GENL_VERSION,
3903 .maxattr = IPVS_CMD_ATTR_MAX,
3904 .netnsok = true, /* Make ipvsadm to work on netns */
3905 .module = THIS_MODULE,
3906 .ops = ip_vs_genl_ops,
3907 .n_ops = ARRAY_SIZE(ip_vs_genl_ops),
3910 static int __init ip_vs_genl_register(void)
3912 return genl_register_family(&ip_vs_genl_family);
3915 static void ip_vs_genl_unregister(void)
3917 genl_unregister_family(&ip_vs_genl_family);
3920 /* End of Generic Netlink interface definitions */
3923 * per netns intit/exit func.
3925 #ifdef CONFIG_SYSCTL
3926 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs)
3928 struct net *net = ipvs->net;
3930 struct ctl_table *tbl;
3932 atomic_set(&ipvs->dropentry, 0);
3933 spin_lock_init(&ipvs->dropentry_lock);
3934 spin_lock_init(&ipvs->droppacket_lock);
3935 spin_lock_init(&ipvs->securetcp_lock);
3937 if (!net_eq(net, &init_net)) {
3938 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3942 /* Don't export sysctls to unprivileged users */
3943 if (net->user_ns != &init_user_ns)
3944 tbl[0].procname = NULL;
3947 /* Initialize sysctl defaults */
3948 for (idx = 0; idx < ARRAY_SIZE(vs_vars); idx++) {
3949 if (tbl[idx].proc_handler == proc_do_defense_mode)
3950 tbl[idx].extra2 = ipvs;
3953 ipvs->sysctl_amemthresh = 1024;
3954 tbl[idx++].data = &ipvs->sysctl_amemthresh;
3955 ipvs->sysctl_am_droprate = 10;
3956 tbl[idx++].data = &ipvs->sysctl_am_droprate;
3957 tbl[idx++].data = &ipvs->sysctl_drop_entry;
3958 tbl[idx++].data = &ipvs->sysctl_drop_packet;
3959 #ifdef CONFIG_IP_VS_NFCT
3960 tbl[idx++].data = &ipvs->sysctl_conntrack;
3962 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3963 ipvs->sysctl_snat_reroute = 1;
3964 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3965 ipvs->sysctl_sync_ver = 1;
3966 tbl[idx++].data = &ipvs->sysctl_sync_ver;
3967 ipvs->sysctl_sync_ports = 1;
3968 tbl[idx++].data = &ipvs->sysctl_sync_ports;
3969 tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
3970 ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3971 tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3972 ipvs->sysctl_sync_sock_size = 0;
3973 tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3974 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3975 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3976 tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
3977 tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
3978 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3979 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3980 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3981 tbl[idx].data = &ipvs->sysctl_sync_threshold;
3982 tbl[idx].extra2 = ipvs;
3983 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3984 ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3985 tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3986 ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3987 tbl[idx++].data = &ipvs->sysctl_sync_retries;
3988 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3989 ipvs->sysctl_pmtu_disc = 1;
3990 tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
3991 tbl[idx++].data = &ipvs->sysctl_backup_only;
3992 ipvs->sysctl_conn_reuse_mode = 1;
3993 tbl[idx++].data = &ipvs->sysctl_conn_reuse_mode;
3994 tbl[idx++].data = &ipvs->sysctl_schedule_icmp;
3995 tbl[idx++].data = &ipvs->sysctl_ignore_tunneled;
3996 #ifdef CONFIG_IP_VS_DEBUG
3997 /* Global sysctls must be ro in non-init netns */
3998 if (!net_eq(net, &init_net))
3999 tbl[idx++].mode = 0444;
4002 ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
4003 if (ipvs->sysctl_hdr == NULL) {
4004 if (!net_eq(net, &init_net))
4008 ip_vs_start_estimator(ipvs, &ipvs->tot_stats);
4009 ipvs->sysctl_tbl = tbl;
4010 /* Schedule defense work */
4011 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
4012 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
4017 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs)
4019 struct net *net = ipvs->net;
4021 cancel_delayed_work_sync(&ipvs->defense_work);
4022 cancel_work_sync(&ipvs->defense_work.work);
4023 unregister_net_sysctl_table(ipvs->sysctl_hdr);
4024 ip_vs_stop_estimator(ipvs, &ipvs->tot_stats);
4026 if (!net_eq(net, &init_net))
4027 kfree(ipvs->sysctl_tbl);
4032 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs) { return 0; }
4033 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs) { }
4037 static struct notifier_block ip_vs_dst_notifier = {
4038 .notifier_call = ip_vs_dst_event,
4039 #ifdef CONFIG_IP_VS_IPV6
4040 .priority = ADDRCONF_NOTIFY_PRIORITY + 5,
4044 int __net_init ip_vs_control_net_init(struct netns_ipvs *ipvs)
4048 /* Initialize rs_table */
4049 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
4050 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
4052 INIT_LIST_HEAD(&ipvs->dest_trash);
4053 spin_lock_init(&ipvs->dest_trash_lock);
4054 setup_timer(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire,
4055 (unsigned long) ipvs);
4056 atomic_set(&ipvs->ftpsvc_counter, 0);
4057 atomic_set(&ipvs->nullsvc_counter, 0);
4058 atomic_set(&ipvs->conn_out_counter, 0);
4061 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
4062 if (!ipvs->tot_stats.cpustats)
4065 for_each_possible_cpu(i) {
4066 struct ip_vs_cpu_stats *ipvs_tot_stats;
4067 ipvs_tot_stats = per_cpu_ptr(ipvs->tot_stats.cpustats, i);
4068 u64_stats_init(&ipvs_tot_stats->syncp);
4071 spin_lock_init(&ipvs->tot_stats.lock);
4073 proc_create("ip_vs", 0, ipvs->net->proc_net, &ip_vs_info_fops);
4074 proc_create("ip_vs_stats", 0, ipvs->net->proc_net, &ip_vs_stats_fops);
4075 proc_create("ip_vs_stats_percpu", 0, ipvs->net->proc_net,
4076 &ip_vs_stats_percpu_fops);
4078 if (ip_vs_control_net_init_sysctl(ipvs))
4084 free_percpu(ipvs->tot_stats.cpustats);
4088 void __net_exit ip_vs_control_net_cleanup(struct netns_ipvs *ipvs)
4090 ip_vs_trash_cleanup(ipvs);
4091 ip_vs_control_net_cleanup_sysctl(ipvs);
4092 remove_proc_entry("ip_vs_stats_percpu", ipvs->net->proc_net);
4093 remove_proc_entry("ip_vs_stats", ipvs->net->proc_net);
4094 remove_proc_entry("ip_vs", ipvs->net->proc_net);
4095 free_percpu(ipvs->tot_stats.cpustats);
4098 int __init ip_vs_register_nl_ioctl(void)
4102 ret = nf_register_sockopt(&ip_vs_sockopts);
4104 pr_err("cannot register sockopt.\n");
4108 ret = ip_vs_genl_register();
4110 pr_err("cannot register Generic Netlink interface.\n");
4116 nf_unregister_sockopt(&ip_vs_sockopts);
4121 void ip_vs_unregister_nl_ioctl(void)
4123 ip_vs_genl_unregister();
4124 nf_unregister_sockopt(&ip_vs_sockopts);
4127 int __init ip_vs_control_init(void)
4134 /* Initialize svc_table, ip_vs_svc_fwm_table */
4135 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
4136 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
4137 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
4140 smp_wmb(); /* Do we really need it now ? */
4142 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
4151 void ip_vs_control_cleanup(void)
4154 unregister_netdevice_notifier(&ip_vs_dst_notifier);