2 * IPVS An implementation of the IP virtual server support for the
3 * LINUX operating system. IPVS is now implemented as a module
4 * over the NetFilter framework. IPVS can be used to build a
5 * high-performance and highly available server based on a
8 * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
9 * Peter Kese <peter.kese@ijs.si>
10 * Julian Anastasov <ja@ssi.bg>
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
21 #define KMSG_COMPONENT "IPVS"
22 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
24 #include <linux/module.h>
25 #include <linux/init.h>
26 #include <linux/types.h>
27 #include <linux/capability.h>
29 #include <linux/sysctl.h>
30 #include <linux/proc_fs.h>
31 #include <linux/workqueue.h>
32 #include <linux/swap.h>
33 #include <linux/seq_file.h>
34 #include <linux/slab.h>
36 #include <linux/netfilter.h>
37 #include <linux/netfilter_ipv4.h>
38 #include <linux/mutex.h>
40 #include <net/net_namespace.h>
41 #include <linux/nsproxy.h>
43 #ifdef CONFIG_IP_VS_IPV6
45 #include <net/ip6_route.h>
47 #include <net/route.h>
49 #include <net/genetlink.h>
51 #include <asm/uaccess.h>
53 #include <net/ip_vs.h>
55 /* semaphore for IPVS sockopts. And, [gs]etsockopt may sleep. */
56 static DEFINE_MUTEX(__ip_vs_mutex);
58 /* sysctl variables */
60 #ifdef CONFIG_IP_VS_DEBUG
61 static int sysctl_ip_vs_debug_level = 0;
63 int ip_vs_get_debug_level(void)
65 return sysctl_ip_vs_debug_level;
71 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup);
74 #ifdef CONFIG_IP_VS_IPV6
75 /* Taken from rt6_fill_node() in net/ipv6/route.c, is there a better way? */
76 static bool __ip_vs_addr_is_local_v6(struct net *net,
77 const struct in6_addr *addr)
82 struct dst_entry *dst = ip6_route_output(net, NULL, &fl6);
85 is_local = !dst->error && dst->dev && (dst->dev->flags & IFF_LOOPBACK);
94 * update_defense_level is called from keventd and from sysctl,
95 * so it needs to protect itself from softirqs
97 static void update_defense_level(struct netns_ipvs *ipvs)
100 static int old_secure_tcp = 0;
105 /* we only count free and buffered memory (in pages) */
107 availmem = i.freeram + i.bufferram;
108 /* however in linux 2.5 the i.bufferram is total page cache size,
110 /* si_swapinfo(&i); */
111 /* availmem = availmem - (i.totalswap - i.freeswap); */
113 nomem = (availmem < ipvs->sysctl_amemthresh);
118 spin_lock(&ipvs->dropentry_lock);
119 switch (ipvs->sysctl_drop_entry) {
121 atomic_set(&ipvs->dropentry, 0);
125 atomic_set(&ipvs->dropentry, 1);
126 ipvs->sysctl_drop_entry = 2;
128 atomic_set(&ipvs->dropentry, 0);
133 atomic_set(&ipvs->dropentry, 1);
135 atomic_set(&ipvs->dropentry, 0);
136 ipvs->sysctl_drop_entry = 1;
140 atomic_set(&ipvs->dropentry, 1);
143 spin_unlock(&ipvs->dropentry_lock);
146 spin_lock(&ipvs->droppacket_lock);
147 switch (ipvs->sysctl_drop_packet) {
153 ipvs->drop_rate = ipvs->drop_counter
154 = ipvs->sysctl_amemthresh /
155 (ipvs->sysctl_amemthresh-availmem);
156 ipvs->sysctl_drop_packet = 2;
163 ipvs->drop_rate = ipvs->drop_counter
164 = ipvs->sysctl_amemthresh /
165 (ipvs->sysctl_amemthresh-availmem);
168 ipvs->sysctl_drop_packet = 1;
172 ipvs->drop_rate = ipvs->sysctl_am_droprate;
175 spin_unlock(&ipvs->droppacket_lock);
178 spin_lock(&ipvs->securetcp_lock);
179 switch (ipvs->sysctl_secure_tcp) {
181 if (old_secure_tcp >= 2)
186 if (old_secure_tcp < 2)
188 ipvs->sysctl_secure_tcp = 2;
190 if (old_secure_tcp >= 2)
196 if (old_secure_tcp < 2)
199 if (old_secure_tcp >= 2)
201 ipvs->sysctl_secure_tcp = 1;
205 if (old_secure_tcp < 2)
209 old_secure_tcp = ipvs->sysctl_secure_tcp;
211 ip_vs_protocol_timeout_change(ipvs,
212 ipvs->sysctl_secure_tcp > 1);
213 spin_unlock(&ipvs->securetcp_lock);
220 * Timer for checking the defense
222 #define DEFENSE_TIMER_PERIOD 1*HZ
224 static void defense_work_handler(struct work_struct *work)
226 struct netns_ipvs *ipvs =
227 container_of(work, struct netns_ipvs, defense_work.work);
229 update_defense_level(ipvs);
230 if (atomic_read(&ipvs->dropentry))
231 ip_vs_random_dropentry(ipvs);
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);
430 && protocol == IPPROTO_TCP
431 && atomic_read(&ipvs->ftpsvc_counter)
432 && (vport == FTPDATA || ntohs(vport) >= PROT_SOCK)) {
434 * Check if ftp service entry exists, the packet
435 * might belong to FTP data connections.
437 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, FTPPORT);
441 && atomic_read(&ipvs->nullsvc_counter)) {
443 * Check if the catch-all port (port zero) exists
445 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, 0);
449 IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
450 fwmark, ip_vs_proto_name(protocol),
451 IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
452 svc ? "hit" : "not hit");
459 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
461 atomic_inc(&svc->refcnt);
462 rcu_assign_pointer(dest->svc, svc);
465 static void ip_vs_service_free(struct ip_vs_service *svc)
467 free_percpu(svc->stats.cpustats);
471 static void ip_vs_service_rcu_free(struct rcu_head *head)
473 struct ip_vs_service *svc;
475 svc = container_of(head, struct ip_vs_service, rcu_head);
476 ip_vs_service_free(svc);
479 static void __ip_vs_svc_put(struct ip_vs_service *svc, bool do_delay)
481 if (atomic_dec_and_test(&svc->refcnt)) {
482 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n",
484 IP_VS_DBG_ADDR(svc->af, &svc->addr),
487 call_rcu(&svc->rcu_head, ip_vs_service_rcu_free);
489 ip_vs_service_free(svc);
495 * Returns hash value for real service
497 static inline unsigned int ip_vs_rs_hashkey(int af,
498 const union nf_inet_addr *addr,
501 register unsigned int porth = ntohs(port);
502 __be32 addr_fold = addr->ip;
504 #ifdef CONFIG_IP_VS_IPV6
506 addr_fold = addr->ip6[0]^addr->ip6[1]^
507 addr->ip6[2]^addr->ip6[3];
510 return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
514 /* Hash ip_vs_dest in rs_table by <proto,addr,port>. */
515 static void ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
519 if (dest->in_rs_table)
523 * Hash by proto,addr,port,
524 * which are the parameters of the real service.
526 hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
528 hlist_add_head_rcu(&dest->d_list, &ipvs->rs_table[hash]);
529 dest->in_rs_table = 1;
532 /* Unhash ip_vs_dest from rs_table. */
533 static void ip_vs_rs_unhash(struct ip_vs_dest *dest)
536 * Remove it from the rs_table table.
538 if (dest->in_rs_table) {
539 hlist_del_rcu(&dest->d_list);
540 dest->in_rs_table = 0;
544 /* Check if real service by <proto,addr,port> is present */
545 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
546 const union nf_inet_addr *daddr, __be16 dport)
549 struct ip_vs_dest *dest;
551 /* Check for "full" addressed entries */
552 hash = ip_vs_rs_hashkey(af, daddr, dport);
555 hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
556 if (dest->port == dport &&
558 ip_vs_addr_equal(af, &dest->addr, daddr) &&
559 (dest->protocol == protocol || dest->vfwmark)) {
570 /* Lookup destination by {addr,port} in the given service
571 * Called under RCU lock.
573 static struct ip_vs_dest *
574 ip_vs_lookup_dest(struct ip_vs_service *svc, int dest_af,
575 const union nf_inet_addr *daddr, __be16 dport)
577 struct ip_vs_dest *dest;
580 * Find the destination for the given service
582 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
583 if ((dest->af == dest_af) &&
584 ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
585 (dest->port == dport)) {
595 * Find destination by {daddr,dport,vaddr,protocol}
596 * Created to be used in ip_vs_process_message() in
597 * the backup synchronization daemon. It finds the
598 * destination to be bound to the received connection
600 * Called under RCU lock, no refcnt is returned.
602 struct ip_vs_dest *ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
603 const union nf_inet_addr *daddr,
605 const union nf_inet_addr *vaddr,
606 __be16 vport, __u16 protocol, __u32 fwmark,
609 struct ip_vs_dest *dest;
610 struct ip_vs_service *svc;
613 svc = ip_vs_service_find(ipvs, svc_af, fwmark, protocol, vaddr, vport);
616 if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
618 dest = ip_vs_lookup_dest(svc, dest_af, daddr, port);
620 dest = ip_vs_lookup_dest(svc, dest_af, daddr, port ^ dport);
624 void ip_vs_dest_dst_rcu_free(struct rcu_head *head)
626 struct ip_vs_dest_dst *dest_dst = container_of(head,
627 struct ip_vs_dest_dst,
630 dst_release(dest_dst->dst_cache);
634 /* Release dest_dst and dst_cache for dest in user context */
635 static void __ip_vs_dst_cache_reset(struct ip_vs_dest *dest)
637 struct ip_vs_dest_dst *old;
639 old = rcu_dereference_protected(dest->dest_dst, 1);
641 RCU_INIT_POINTER(dest->dest_dst, NULL);
642 call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
647 * Lookup dest by {svc,addr,port} in the destination trash.
648 * The destination trash is used to hold the destinations that are removed
649 * from the service table but are still referenced by some conn entries.
650 * The reason to add the destination trash is when the dest is temporary
651 * down (either by administrator or by monitor program), the dest can be
652 * picked back from the trash, the remaining connections to the dest can
653 * continue, and the counting information of the dest is also useful for
656 static struct ip_vs_dest *
657 ip_vs_trash_get_dest(struct ip_vs_service *svc, int dest_af,
658 const union nf_inet_addr *daddr, __be16 dport)
660 struct ip_vs_dest *dest;
661 struct netns_ipvs *ipvs = svc->ipvs;
664 * Find the destination in trash
666 spin_lock_bh(&ipvs->dest_trash_lock);
667 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
668 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
671 IP_VS_DBG_ADDR(dest->af, &dest->addr),
673 atomic_read(&dest->refcnt));
674 if (dest->af == dest_af &&
675 ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
676 dest->port == dport &&
677 dest->vfwmark == svc->fwmark &&
678 dest->protocol == svc->protocol &&
680 (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
681 dest->vport == svc->port))) {
683 list_del(&dest->t_list);
684 ip_vs_dest_hold(dest);
692 spin_unlock_bh(&ipvs->dest_trash_lock);
697 static void ip_vs_dest_free(struct ip_vs_dest *dest)
699 struct ip_vs_service *svc = rcu_dereference_protected(dest->svc, 1);
701 __ip_vs_dst_cache_reset(dest);
702 __ip_vs_svc_put(svc, false);
703 free_percpu(dest->stats.cpustats);
704 ip_vs_dest_put_and_free(dest);
708 * Clean up all the destinations in the trash
709 * Called by the ip_vs_control_cleanup()
711 * When the ip_vs_control_clearup is activated by ipvs module exit,
712 * the service tables must have been flushed and all the connections
713 * are expired, and the refcnt of each destination in the trash must
714 * be 0, so we simply release them here.
716 static void ip_vs_trash_cleanup(struct netns_ipvs *ipvs)
718 struct ip_vs_dest *dest, *nxt;
720 del_timer_sync(&ipvs->dest_trash_timer);
721 /* No need to use dest_trash_lock */
722 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, t_list) {
723 list_del(&dest->t_list);
724 ip_vs_dest_free(dest);
729 ip_vs_copy_stats(struct ip_vs_kstats *dst, struct ip_vs_stats *src)
731 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->kstats.c - src->kstats0.c
733 spin_lock_bh(&src->lock);
735 IP_VS_SHOW_STATS_COUNTER(conns);
736 IP_VS_SHOW_STATS_COUNTER(inpkts);
737 IP_VS_SHOW_STATS_COUNTER(outpkts);
738 IP_VS_SHOW_STATS_COUNTER(inbytes);
739 IP_VS_SHOW_STATS_COUNTER(outbytes);
741 ip_vs_read_estimator(dst, src);
743 spin_unlock_bh(&src->lock);
747 ip_vs_export_stats_user(struct ip_vs_stats_user *dst, struct ip_vs_kstats *src)
749 dst->conns = (u32)src->conns;
750 dst->inpkts = (u32)src->inpkts;
751 dst->outpkts = (u32)src->outpkts;
752 dst->inbytes = src->inbytes;
753 dst->outbytes = src->outbytes;
754 dst->cps = (u32)src->cps;
755 dst->inpps = (u32)src->inpps;
756 dst->outpps = (u32)src->outpps;
757 dst->inbps = (u32)src->inbps;
758 dst->outbps = (u32)src->outbps;
762 ip_vs_zero_stats(struct ip_vs_stats *stats)
764 spin_lock_bh(&stats->lock);
766 /* get current counters as zero point, rates are zeroed */
768 #define IP_VS_ZERO_STATS_COUNTER(c) stats->kstats0.c = stats->kstats.c
770 IP_VS_ZERO_STATS_COUNTER(conns);
771 IP_VS_ZERO_STATS_COUNTER(inpkts);
772 IP_VS_ZERO_STATS_COUNTER(outpkts);
773 IP_VS_ZERO_STATS_COUNTER(inbytes);
774 IP_VS_ZERO_STATS_COUNTER(outbytes);
776 ip_vs_zero_estimator(stats);
778 spin_unlock_bh(&stats->lock);
782 * Update a destination in the given service
785 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
786 struct ip_vs_dest_user_kern *udest, int add)
788 struct netns_ipvs *ipvs = svc->ipvs;
789 struct ip_vs_service *old_svc;
790 struct ip_vs_scheduler *sched;
793 /* We cannot modify an address and change the address family */
794 BUG_ON(!add && udest->af != dest->af);
796 if (add && udest->af != svc->af)
797 ipvs->mixed_address_family_dests++;
799 /* set the weight and the flags */
800 atomic_set(&dest->weight, udest->weight);
801 conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
802 conn_flags |= IP_VS_CONN_F_INACTIVE;
804 /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
805 if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
806 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
809 * Put the real service in rs_table if not present.
810 * For now only for NAT!
812 ip_vs_rs_hash(ipvs, dest);
814 atomic_set(&dest->conn_flags, conn_flags);
816 /* bind the service */
817 old_svc = rcu_dereference_protected(dest->svc, 1);
819 __ip_vs_bind_svc(dest, svc);
821 if (old_svc != svc) {
822 ip_vs_zero_stats(&dest->stats);
823 __ip_vs_bind_svc(dest, svc);
824 __ip_vs_svc_put(old_svc, true);
828 /* set the dest status flags */
829 dest->flags |= IP_VS_DEST_F_AVAILABLE;
831 if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
832 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
833 dest->u_threshold = udest->u_threshold;
834 dest->l_threshold = udest->l_threshold;
836 dest->af = udest->af;
838 spin_lock_bh(&dest->dst_lock);
839 __ip_vs_dst_cache_reset(dest);
840 spin_unlock_bh(&dest->dst_lock);
843 ip_vs_start_estimator(svc->ipvs, &dest->stats);
844 list_add_rcu(&dest->n_list, &svc->destinations);
846 sched = rcu_dereference_protected(svc->scheduler, 1);
847 if (sched && sched->add_dest)
848 sched->add_dest(svc, dest);
850 sched = rcu_dereference_protected(svc->scheduler, 1);
851 if (sched && sched->upd_dest)
852 sched->upd_dest(svc, dest);
858 * Create a destination for the given service
861 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
862 struct ip_vs_dest **dest_p)
864 struct ip_vs_dest *dest;
865 unsigned int atype, i;
869 #ifdef CONFIG_IP_VS_IPV6
870 if (udest->af == AF_INET6) {
871 atype = ipv6_addr_type(&udest->addr.in6);
872 if ((!(atype & IPV6_ADDR_UNICAST) ||
873 atype & IPV6_ADDR_LINKLOCAL) &&
874 !__ip_vs_addr_is_local_v6(svc->ipvs->net, &udest->addr.in6))
879 atype = inet_addr_type(svc->ipvs->net, udest->addr.ip);
880 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
884 dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
888 dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
889 if (!dest->stats.cpustats)
892 for_each_possible_cpu(i) {
893 struct ip_vs_cpu_stats *ip_vs_dest_stats;
894 ip_vs_dest_stats = per_cpu_ptr(dest->stats.cpustats, i);
895 u64_stats_init(&ip_vs_dest_stats->syncp);
898 dest->af = udest->af;
899 dest->protocol = svc->protocol;
900 dest->vaddr = svc->addr;
901 dest->vport = svc->port;
902 dest->vfwmark = svc->fwmark;
903 ip_vs_addr_copy(udest->af, &dest->addr, &udest->addr);
904 dest->port = udest->port;
906 atomic_set(&dest->activeconns, 0);
907 atomic_set(&dest->inactconns, 0);
908 atomic_set(&dest->persistconns, 0);
909 atomic_set(&dest->refcnt, 1);
911 INIT_HLIST_NODE(&dest->d_list);
912 spin_lock_init(&dest->dst_lock);
913 spin_lock_init(&dest->stats.lock);
914 __ip_vs_update_dest(svc, dest, udest, 1);
928 * Add a destination into an existing service
931 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
933 struct ip_vs_dest *dest;
934 union nf_inet_addr daddr;
935 __be16 dport = udest->port;
940 if (udest->weight < 0) {
941 pr_err("%s(): server weight less than zero\n", __func__);
945 if (udest->l_threshold > udest->u_threshold) {
946 pr_err("%s(): lower threshold is higher than upper threshold\n",
951 ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
953 /* We use function that requires RCU lock */
955 dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
959 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
964 * Check if the dest already exists in the trash and
965 * is from the same service
967 dest = ip_vs_trash_get_dest(svc, udest->af, &daddr, dport);
970 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
971 "dest->refcnt=%d, service %u/%s:%u\n",
972 IP_VS_DBG_ADDR(udest->af, &daddr), ntohs(dport),
973 atomic_read(&dest->refcnt),
975 IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
978 __ip_vs_update_dest(svc, dest, udest, 1);
982 * Allocate and initialize the dest structure
984 ret = ip_vs_new_dest(svc, udest, &dest);
993 * Edit a destination in the given service
996 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
998 struct ip_vs_dest *dest;
999 union nf_inet_addr daddr;
1000 __be16 dport = udest->port;
1004 if (udest->weight < 0) {
1005 pr_err("%s(): server weight less than zero\n", __func__);
1009 if (udest->l_threshold > udest->u_threshold) {
1010 pr_err("%s(): lower threshold is higher than upper threshold\n",
1015 ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
1017 /* We use function that requires RCU lock */
1019 dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
1023 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1027 __ip_vs_update_dest(svc, dest, udest, 0);
1034 * Delete a destination (must be already unlinked from the service)
1036 static void __ip_vs_del_dest(struct netns_ipvs *ipvs, struct ip_vs_dest *dest,
1039 ip_vs_stop_estimator(ipvs, &dest->stats);
1042 * Remove it from the d-linked list with the real services.
1044 ip_vs_rs_unhash(dest);
1046 spin_lock_bh(&ipvs->dest_trash_lock);
1047 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1048 IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port),
1049 atomic_read(&dest->refcnt));
1050 if (list_empty(&ipvs->dest_trash) && !cleanup)
1051 mod_timer(&ipvs->dest_trash_timer,
1052 jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1053 /* dest lives in trash without reference */
1054 list_add(&dest->t_list, &ipvs->dest_trash);
1055 dest->idle_start = 0;
1056 spin_unlock_bh(&ipvs->dest_trash_lock);
1057 ip_vs_dest_put(dest);
1062 * Unlink a destination from the given service
1064 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1065 struct ip_vs_dest *dest,
1068 dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1071 * Remove it from the d-linked destination list.
1073 list_del_rcu(&dest->n_list);
1076 if (dest->af != svc->af)
1077 svc->ipvs->mixed_address_family_dests--;
1080 struct ip_vs_scheduler *sched;
1082 sched = rcu_dereference_protected(svc->scheduler, 1);
1083 if (sched && sched->del_dest)
1084 sched->del_dest(svc, dest);
1090 * Delete a destination server in the given service
1093 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1095 struct ip_vs_dest *dest;
1096 __be16 dport = udest->port;
1100 /* We use function that requires RCU lock */
1102 dest = ip_vs_lookup_dest(svc, udest->af, &udest->addr, dport);
1106 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1111 * Unlink dest from the service
1113 __ip_vs_unlink_dest(svc, dest, 1);
1116 * Delete the destination
1118 __ip_vs_del_dest(svc->ipvs, dest, false);
1125 static void ip_vs_dest_trash_expire(unsigned long data)
1127 struct netns_ipvs *ipvs = (struct netns_ipvs *)data;
1128 struct ip_vs_dest *dest, *next;
1129 unsigned long now = jiffies;
1131 spin_lock(&ipvs->dest_trash_lock);
1132 list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) {
1133 if (atomic_read(&dest->refcnt) > 0)
1135 if (dest->idle_start) {
1136 if (time_before(now, dest->idle_start +
1137 IP_VS_DEST_TRASH_PERIOD))
1140 dest->idle_start = max(1UL, now);
1143 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1145 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1147 list_del(&dest->t_list);
1148 ip_vs_dest_free(dest);
1150 if (!list_empty(&ipvs->dest_trash))
1151 mod_timer(&ipvs->dest_trash_timer,
1152 jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1153 spin_unlock(&ipvs->dest_trash_lock);
1157 * Add a service into the service hash table
1160 ip_vs_add_service(struct netns_ipvs *ipvs, struct ip_vs_service_user_kern *u,
1161 struct ip_vs_service **svc_p)
1164 struct ip_vs_scheduler *sched = NULL;
1165 struct ip_vs_pe *pe = NULL;
1166 struct ip_vs_service *svc = NULL;
1168 /* increase the module use count */
1169 ip_vs_use_count_inc();
1171 /* Lookup the scheduler by 'u->sched_name' */
1172 if (strcmp(u->sched_name, "none")) {
1173 sched = ip_vs_scheduler_get(u->sched_name);
1175 pr_info("Scheduler module ip_vs_%s not found\n",
1182 if (u->pe_name && *u->pe_name) {
1183 pe = ip_vs_pe_getbyname(u->pe_name);
1185 pr_info("persistence engine module ip_vs_pe_%s "
1186 "not found\n", u->pe_name);
1192 #ifdef CONFIG_IP_VS_IPV6
1193 if (u->af == AF_INET6) {
1194 __u32 plen = (__force __u32) u->netmask;
1196 if (plen < 1 || plen > 128) {
1203 svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1205 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1209 svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1210 if (!svc->stats.cpustats) {
1215 for_each_possible_cpu(i) {
1216 struct ip_vs_cpu_stats *ip_vs_stats;
1217 ip_vs_stats = per_cpu_ptr(svc->stats.cpustats, i);
1218 u64_stats_init(&ip_vs_stats->syncp);
1222 /* I'm the first user of the service */
1223 atomic_set(&svc->refcnt, 0);
1226 svc->protocol = u->protocol;
1227 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1228 svc->port = u->port;
1229 svc->fwmark = u->fwmark;
1230 svc->flags = u->flags;
1231 svc->timeout = u->timeout * HZ;
1232 svc->netmask = u->netmask;
1235 INIT_LIST_HEAD(&svc->destinations);
1236 spin_lock_init(&svc->sched_lock);
1237 spin_lock_init(&svc->stats.lock);
1239 /* Bind the scheduler */
1241 ret = ip_vs_bind_scheduler(svc, sched);
1247 /* Bind the ct retriever */
1248 RCU_INIT_POINTER(svc->pe, pe);
1251 /* Update the virtual service counters */
1252 if (svc->port == FTPPORT)
1253 atomic_inc(&ipvs->ftpsvc_counter);
1254 else if (svc->port == 0)
1255 atomic_inc(&ipvs->nullsvc_counter);
1257 ip_vs_start_estimator(ipvs, &svc->stats);
1259 /* Count only IPv4 services for old get/setsockopt interface */
1260 if (svc->af == AF_INET)
1261 ipvs->num_services++;
1263 /* Hash the service into the service table */
1264 ip_vs_svc_hash(svc);
1267 /* Now there is a service - full throttle */
1274 ip_vs_unbind_scheduler(svc, sched);
1275 ip_vs_service_free(svc);
1277 ip_vs_scheduler_put(sched);
1280 /* decrease the module use count */
1281 ip_vs_use_count_dec();
1288 * Edit a service and bind it with a new scheduler
1291 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1293 struct ip_vs_scheduler *sched = NULL, *old_sched;
1294 struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1298 * Lookup the scheduler, by 'u->sched_name'
1300 if (strcmp(u->sched_name, "none")) {
1301 sched = ip_vs_scheduler_get(u->sched_name);
1303 pr_info("Scheduler module ip_vs_%s not found\n",
1310 if (u->pe_name && *u->pe_name) {
1311 pe = ip_vs_pe_getbyname(u->pe_name);
1313 pr_info("persistence engine module ip_vs_pe_%s "
1314 "not found\n", u->pe_name);
1321 #ifdef CONFIG_IP_VS_IPV6
1322 if (u->af == AF_INET6) {
1323 __u32 plen = (__force __u32) u->netmask;
1325 if (plen < 1 || plen > 128) {
1332 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1333 if (sched != old_sched) {
1335 ip_vs_unbind_scheduler(svc, old_sched);
1336 RCU_INIT_POINTER(svc->scheduler, NULL);
1337 /* Wait all svc->sched_data users */
1340 /* Bind the new scheduler */
1342 ret = ip_vs_bind_scheduler(svc, sched);
1344 ip_vs_scheduler_put(sched);
1351 * Set the flags and timeout value
1353 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1354 svc->timeout = u->timeout * HZ;
1355 svc->netmask = u->netmask;
1357 old_pe = rcu_dereference_protected(svc->pe, 1);
1359 rcu_assign_pointer(svc->pe, pe);
1362 ip_vs_scheduler_put(old_sched);
1363 ip_vs_pe_put(old_pe);
1368 * Delete a service from the service list
1369 * - The service must be unlinked, unlocked and not referenced!
1370 * - We are called under _bh lock
1372 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1374 struct ip_vs_dest *dest, *nxt;
1375 struct ip_vs_scheduler *old_sched;
1376 struct ip_vs_pe *old_pe;
1377 struct netns_ipvs *ipvs = svc->ipvs;
1379 pr_info("%s: enter\n", __func__);
1381 /* Count only IPv4 services for old get/setsockopt interface */
1382 if (svc->af == AF_INET)
1383 ipvs->num_services--;
1385 ip_vs_stop_estimator(svc->ipvs, &svc->stats);
1387 /* Unbind scheduler */
1388 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1389 ip_vs_unbind_scheduler(svc, old_sched);
1390 ip_vs_scheduler_put(old_sched);
1392 /* Unbind persistence engine, keep svc->pe */
1393 old_pe = rcu_dereference_protected(svc->pe, 1);
1394 ip_vs_pe_put(old_pe);
1397 * Unlink the whole destination list
1399 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1400 __ip_vs_unlink_dest(svc, dest, 0);
1401 __ip_vs_del_dest(svc->ipvs, dest, cleanup);
1405 * Update the virtual service counters
1407 if (svc->port == FTPPORT)
1408 atomic_dec(&ipvs->ftpsvc_counter);
1409 else if (svc->port == 0)
1410 atomic_dec(&ipvs->nullsvc_counter);
1413 * Free the service if nobody refers to it
1415 __ip_vs_svc_put(svc, true);
1417 /* decrease the module use count */
1418 ip_vs_use_count_dec();
1422 * Unlink a service from list and try to delete it if its refcnt reached 0
1424 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1426 /* Hold svc to avoid double release from dest_trash */
1427 atomic_inc(&svc->refcnt);
1429 * Unhash it from the service table
1431 ip_vs_svc_unhash(svc);
1433 __ip_vs_del_service(svc, cleanup);
1437 * Delete a service from the service list
1439 static int ip_vs_del_service(struct ip_vs_service *svc)
1443 ip_vs_unlink_service(svc, false);
1450 * Flush all the virtual services
1452 static int ip_vs_flush(struct netns_ipvs *ipvs, bool cleanup)
1455 struct ip_vs_service *svc;
1456 struct hlist_node *n;
1459 * Flush the service table hashed by <netns,protocol,addr,port>
1461 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1462 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1464 if (svc->ipvs == ipvs)
1465 ip_vs_unlink_service(svc, cleanup);
1470 * Flush the service table hashed by fwmark
1472 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1473 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1475 if (svc->ipvs == ipvs)
1476 ip_vs_unlink_service(svc, cleanup);
1484 * Delete service by {netns} in the service table.
1485 * Called by __ip_vs_cleanup()
1487 void ip_vs_service_net_cleanup(struct netns_ipvs *ipvs)
1490 /* Check for "full" addressed entries */
1491 mutex_lock(&__ip_vs_mutex);
1492 ip_vs_flush(ipvs, true);
1493 mutex_unlock(&__ip_vs_mutex);
1497 /* Put all references for device (dst_cache) */
1499 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1501 struct ip_vs_dest_dst *dest_dst;
1503 spin_lock_bh(&dest->dst_lock);
1504 dest_dst = rcu_dereference_protected(dest->dest_dst, 1);
1505 if (dest_dst && dest_dst->dst_cache->dev == dev) {
1506 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1508 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1510 atomic_read(&dest->refcnt));
1511 __ip_vs_dst_cache_reset(dest);
1513 spin_unlock_bh(&dest->dst_lock);
1516 /* Netdev event receiver
1517 * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1519 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1522 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1523 struct net *net = dev_net(dev);
1524 struct netns_ipvs *ipvs = net_ipvs(net);
1525 struct ip_vs_service *svc;
1526 struct ip_vs_dest *dest;
1529 if (event != NETDEV_DOWN || !ipvs)
1531 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1533 mutex_lock(&__ip_vs_mutex);
1534 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1535 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1536 if (svc->ipvs == ipvs) {
1537 list_for_each_entry(dest, &svc->destinations,
1539 ip_vs_forget_dev(dest, dev);
1544 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1545 if (svc->ipvs == ipvs) {
1546 list_for_each_entry(dest, &svc->destinations,
1548 ip_vs_forget_dev(dest, dev);
1555 spin_lock_bh(&ipvs->dest_trash_lock);
1556 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1557 ip_vs_forget_dev(dest, dev);
1559 spin_unlock_bh(&ipvs->dest_trash_lock);
1560 mutex_unlock(&__ip_vs_mutex);
1566 * Zero counters in a service or all services
1568 static int ip_vs_zero_service(struct ip_vs_service *svc)
1570 struct ip_vs_dest *dest;
1572 list_for_each_entry(dest, &svc->destinations, n_list) {
1573 ip_vs_zero_stats(&dest->stats);
1575 ip_vs_zero_stats(&svc->stats);
1579 static int ip_vs_zero_all(struct netns_ipvs *ipvs)
1582 struct ip_vs_service *svc;
1584 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1585 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1586 if (svc->ipvs == ipvs)
1587 ip_vs_zero_service(svc);
1591 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1592 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1593 if (svc->ipvs == ipvs)
1594 ip_vs_zero_service(svc);
1598 ip_vs_zero_stats(&ipvs->tot_stats);
1602 #ifdef CONFIG_SYSCTL
1605 static int three = 3;
1608 proc_do_defense_mode(struct ctl_table *table, int write,
1609 void __user *buffer, size_t *lenp, loff_t *ppos)
1611 struct netns_ipvs *ipvs = table->extra2;
1612 int *valp = table->data;
1616 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1617 if (write && (*valp != val)) {
1618 if ((*valp < 0) || (*valp > 3)) {
1619 /* Restore the correct value */
1622 update_defense_level(ipvs);
1629 proc_do_sync_threshold(struct ctl_table *table, int write,
1630 void __user *buffer, size_t *lenp, loff_t *ppos)
1632 int *valp = table->data;
1636 /* backup the value first */
1637 memcpy(val, valp, sizeof(val));
1639 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1640 if (write && (valp[0] < 0 || valp[1] < 0 ||
1641 (valp[0] >= valp[1] && valp[1]))) {
1642 /* Restore the correct value */
1643 memcpy(valp, val, sizeof(val));
1649 proc_do_sync_mode(struct ctl_table *table, int write,
1650 void __user *buffer, size_t *lenp, loff_t *ppos)
1652 int *valp = table->data;
1656 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1657 if (write && (*valp != val)) {
1658 if ((*valp < 0) || (*valp > 1)) {
1659 /* Restore the correct value */
1667 proc_do_sync_ports(struct ctl_table *table, int write,
1668 void __user *buffer, size_t *lenp, loff_t *ppos)
1670 int *valp = table->data;
1674 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1675 if (write && (*valp != val)) {
1676 if (*valp < 1 || !is_power_of_2(*valp)) {
1677 /* Restore the correct value */
1685 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1686 * Do not change order or insert new entries without
1687 * align with netns init in ip_vs_control_net_init()
1690 static struct ctl_table vs_vars[] = {
1692 .procname = "amemthresh",
1693 .maxlen = sizeof(int),
1695 .proc_handler = proc_dointvec,
1698 .procname = "am_droprate",
1699 .maxlen = sizeof(int),
1701 .proc_handler = proc_dointvec,
1704 .procname = "drop_entry",
1705 .maxlen = sizeof(int),
1707 .proc_handler = proc_do_defense_mode,
1710 .procname = "drop_packet",
1711 .maxlen = sizeof(int),
1713 .proc_handler = proc_do_defense_mode,
1715 #ifdef CONFIG_IP_VS_NFCT
1717 .procname = "conntrack",
1718 .maxlen = sizeof(int),
1720 .proc_handler = &proc_dointvec,
1724 .procname = "secure_tcp",
1725 .maxlen = sizeof(int),
1727 .proc_handler = proc_do_defense_mode,
1730 .procname = "snat_reroute",
1731 .maxlen = sizeof(int),
1733 .proc_handler = &proc_dointvec,
1736 .procname = "sync_version",
1737 .maxlen = sizeof(int),
1739 .proc_handler = &proc_do_sync_mode,
1742 .procname = "sync_ports",
1743 .maxlen = sizeof(int),
1745 .proc_handler = &proc_do_sync_ports,
1748 .procname = "sync_persist_mode",
1749 .maxlen = sizeof(int),
1751 .proc_handler = proc_dointvec,
1754 .procname = "sync_qlen_max",
1755 .maxlen = sizeof(unsigned long),
1757 .proc_handler = proc_doulongvec_minmax,
1760 .procname = "sync_sock_size",
1761 .maxlen = sizeof(int),
1763 .proc_handler = proc_dointvec,
1766 .procname = "cache_bypass",
1767 .maxlen = sizeof(int),
1769 .proc_handler = proc_dointvec,
1772 .procname = "expire_nodest_conn",
1773 .maxlen = sizeof(int),
1775 .proc_handler = proc_dointvec,
1778 .procname = "sloppy_tcp",
1779 .maxlen = sizeof(int),
1781 .proc_handler = proc_dointvec,
1784 .procname = "sloppy_sctp",
1785 .maxlen = sizeof(int),
1787 .proc_handler = proc_dointvec,
1790 .procname = "expire_quiescent_template",
1791 .maxlen = sizeof(int),
1793 .proc_handler = proc_dointvec,
1796 .procname = "sync_threshold",
1798 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1800 .proc_handler = proc_do_sync_threshold,
1803 .procname = "sync_refresh_period",
1804 .maxlen = sizeof(int),
1806 .proc_handler = proc_dointvec_jiffies,
1809 .procname = "sync_retries",
1810 .maxlen = sizeof(int),
1812 .proc_handler = proc_dointvec_minmax,
1817 .procname = "nat_icmp_send",
1818 .maxlen = sizeof(int),
1820 .proc_handler = proc_dointvec,
1823 .procname = "pmtu_disc",
1824 .maxlen = sizeof(int),
1826 .proc_handler = proc_dointvec,
1829 .procname = "backup_only",
1830 .maxlen = sizeof(int),
1832 .proc_handler = proc_dointvec,
1835 .procname = "conn_reuse_mode",
1836 .maxlen = sizeof(int),
1838 .proc_handler = proc_dointvec,
1841 .procname = "schedule_icmp",
1842 .maxlen = sizeof(int),
1844 .proc_handler = proc_dointvec,
1847 .procname = "ignore_tunneled",
1848 .maxlen = sizeof(int),
1850 .proc_handler = proc_dointvec,
1852 #ifdef CONFIG_IP_VS_DEBUG
1854 .procname = "debug_level",
1855 .data = &sysctl_ip_vs_debug_level,
1856 .maxlen = sizeof(int),
1858 .proc_handler = proc_dointvec,
1866 #ifdef CONFIG_PROC_FS
1869 struct seq_net_private p; /* Do not move this, netns depends upon it*/
1870 struct hlist_head *table;
1875 * Write the contents of the VS rule table to a PROCfs file.
1876 * (It is kept just for backward compatibility)
1878 static inline const char *ip_vs_fwd_name(unsigned int flags)
1880 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1881 case IP_VS_CONN_F_LOCALNODE:
1883 case IP_VS_CONN_F_TUNNEL:
1885 case IP_VS_CONN_F_DROUTE:
1893 /* Get the Nth entry in the two lists */
1894 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1896 struct net *net = seq_file_net(seq);
1897 struct netns_ipvs *ipvs = net_ipvs(net);
1898 struct ip_vs_iter *iter = seq->private;
1900 struct ip_vs_service *svc;
1902 /* look in hash by protocol */
1903 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1904 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
1905 if ((svc->ipvs == ipvs) && pos-- == 0) {
1906 iter->table = ip_vs_svc_table;
1913 /* keep looking in fwmark */
1914 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1915 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
1917 if ((svc->ipvs == ipvs) && pos-- == 0) {
1918 iter->table = ip_vs_svc_fwm_table;
1928 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1932 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1936 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1938 struct hlist_node *e;
1939 struct ip_vs_iter *iter;
1940 struct ip_vs_service *svc;
1943 if (v == SEQ_START_TOKEN)
1944 return ip_vs_info_array(seq,0);
1947 iter = seq->private;
1949 if (iter->table == ip_vs_svc_table) {
1950 /* next service in table hashed by protocol */
1951 e = rcu_dereference(hlist_next_rcu(&svc->s_list));
1953 return hlist_entry(e, struct ip_vs_service, s_list);
1955 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1956 hlist_for_each_entry_rcu(svc,
1957 &ip_vs_svc_table[iter->bucket],
1963 iter->table = ip_vs_svc_fwm_table;
1968 /* next service in hashed by fwmark */
1969 e = rcu_dereference(hlist_next_rcu(&svc->f_list));
1971 return hlist_entry(e, struct ip_vs_service, f_list);
1974 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1975 hlist_for_each_entry_rcu(svc,
1976 &ip_vs_svc_fwm_table[iter->bucket],
1984 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
1991 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
1993 if (v == SEQ_START_TOKEN) {
1995 "IP Virtual Server version %d.%d.%d (size=%d)\n",
1996 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
1998 "Prot LocalAddress:Port Scheduler Flags\n");
2000 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2002 const struct ip_vs_service *svc = v;
2003 const struct ip_vs_iter *iter = seq->private;
2004 const struct ip_vs_dest *dest;
2005 struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler);
2006 char *sched_name = sched ? sched->name : "none";
2008 if (iter->table == ip_vs_svc_table) {
2009 #ifdef CONFIG_IP_VS_IPV6
2010 if (svc->af == AF_INET6)
2011 seq_printf(seq, "%s [%pI6]:%04X %s ",
2012 ip_vs_proto_name(svc->protocol),
2018 seq_printf(seq, "%s %08X:%04X %s %s ",
2019 ip_vs_proto_name(svc->protocol),
2020 ntohl(svc->addr.ip),
2023 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2025 seq_printf(seq, "FWM %08X %s %s",
2026 svc->fwmark, sched_name,
2027 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2030 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2031 seq_printf(seq, "persistent %d %08X\n",
2033 ntohl(svc->netmask));
2035 seq_putc(seq, '\n');
2037 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
2038 #ifdef CONFIG_IP_VS_IPV6
2039 if (dest->af == AF_INET6)
2042 " %-7s %-6d %-10d %-10d\n",
2045 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2046 atomic_read(&dest->weight),
2047 atomic_read(&dest->activeconns),
2048 atomic_read(&dest->inactconns));
2053 "%-7s %-6d %-10d %-10d\n",
2054 ntohl(dest->addr.ip),
2056 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2057 atomic_read(&dest->weight),
2058 atomic_read(&dest->activeconns),
2059 atomic_read(&dest->inactconns));
2066 static const struct seq_operations ip_vs_info_seq_ops = {
2067 .start = ip_vs_info_seq_start,
2068 .next = ip_vs_info_seq_next,
2069 .stop = ip_vs_info_seq_stop,
2070 .show = ip_vs_info_seq_show,
2073 static int ip_vs_info_open(struct inode *inode, struct file *file)
2075 return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2076 sizeof(struct ip_vs_iter));
2079 static const struct file_operations ip_vs_info_fops = {
2080 .owner = THIS_MODULE,
2081 .open = ip_vs_info_open,
2083 .llseek = seq_lseek,
2084 .release = seq_release_net,
2087 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2089 struct net *net = seq_file_single_net(seq);
2090 struct ip_vs_kstats show;
2092 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2094 " Total Incoming Outgoing Incoming Outgoing\n");
2096 " Conns Packets Packets Bytes Bytes\n");
2098 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2099 seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n\n",
2100 (unsigned long long)show.conns,
2101 (unsigned long long)show.inpkts,
2102 (unsigned long long)show.outpkts,
2103 (unsigned long long)show.inbytes,
2104 (unsigned long long)show.outbytes);
2106 /* 01234567 01234567 01234567 0123456701234567 0123456701234567*/
2108 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2109 seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n",
2110 (unsigned long long)show.cps,
2111 (unsigned long long)show.inpps,
2112 (unsigned long long)show.outpps,
2113 (unsigned long long)show.inbps,
2114 (unsigned long long)show.outbps);
2119 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2121 return single_open_net(inode, file, ip_vs_stats_show);
2124 static const struct file_operations ip_vs_stats_fops = {
2125 .owner = THIS_MODULE,
2126 .open = ip_vs_stats_seq_open,
2128 .llseek = seq_lseek,
2129 .release = single_release_net,
2132 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2134 struct net *net = seq_file_single_net(seq);
2135 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2136 struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats;
2137 struct ip_vs_kstats kstats;
2140 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2142 " Total Incoming Outgoing Incoming Outgoing\n");
2144 "CPU Conns Packets Packets Bytes Bytes\n");
2146 for_each_possible_cpu(i) {
2147 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2149 u64 conns, inpkts, outpkts, inbytes, outbytes;
2152 start = u64_stats_fetch_begin_irq(&u->syncp);
2153 conns = u->cnt.conns;
2154 inpkts = u->cnt.inpkts;
2155 outpkts = u->cnt.outpkts;
2156 inbytes = u->cnt.inbytes;
2157 outbytes = u->cnt.outbytes;
2158 } while (u64_stats_fetch_retry_irq(&u->syncp, start));
2160 seq_printf(seq, "%3X %8LX %8LX %8LX %16LX %16LX\n",
2161 i, (u64)conns, (u64)inpkts,
2162 (u64)outpkts, (u64)inbytes,
2166 ip_vs_copy_stats(&kstats, tot_stats);
2168 seq_printf(seq, " ~ %8LX %8LX %8LX %16LX %16LX\n\n",
2169 (unsigned long long)kstats.conns,
2170 (unsigned long long)kstats.inpkts,
2171 (unsigned long long)kstats.outpkts,
2172 (unsigned long long)kstats.inbytes,
2173 (unsigned long long)kstats.outbytes);
2175 /* ... 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2177 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2178 seq_printf(seq, " %8LX %8LX %8LX %16LX %16LX\n",
2188 static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2190 return single_open_net(inode, file, ip_vs_stats_percpu_show);
2193 static const struct file_operations ip_vs_stats_percpu_fops = {
2194 .owner = THIS_MODULE,
2195 .open = ip_vs_stats_percpu_seq_open,
2197 .llseek = seq_lseek,
2198 .release = single_release_net,
2203 * Set timeout values for tcp tcpfin udp in the timeout_table.
2205 static int ip_vs_set_timeout(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2207 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2208 struct ip_vs_proto_data *pd;
2211 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2216 #ifdef CONFIG_IP_VS_PROTO_TCP
2217 if (u->tcp_timeout) {
2218 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2219 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2220 = u->tcp_timeout * HZ;
2223 if (u->tcp_fin_timeout) {
2224 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2225 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2226 = u->tcp_fin_timeout * HZ;
2230 #ifdef CONFIG_IP_VS_PROTO_UDP
2231 if (u->udp_timeout) {
2232 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2233 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2234 = u->udp_timeout * HZ;
2240 #define CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2242 struct ip_vs_svcdest_user {
2243 struct ip_vs_service_user s;
2244 struct ip_vs_dest_user d;
2247 static const unsigned char set_arglen[CMDID(IP_VS_SO_SET_MAX) + 1] = {
2248 [CMDID(IP_VS_SO_SET_ADD)] = sizeof(struct ip_vs_service_user),
2249 [CMDID(IP_VS_SO_SET_EDIT)] = sizeof(struct ip_vs_service_user),
2250 [CMDID(IP_VS_SO_SET_DEL)] = sizeof(struct ip_vs_service_user),
2251 [CMDID(IP_VS_SO_SET_ADDDEST)] = sizeof(struct ip_vs_svcdest_user),
2252 [CMDID(IP_VS_SO_SET_DELDEST)] = sizeof(struct ip_vs_svcdest_user),
2253 [CMDID(IP_VS_SO_SET_EDITDEST)] = sizeof(struct ip_vs_svcdest_user),
2254 [CMDID(IP_VS_SO_SET_TIMEOUT)] = sizeof(struct ip_vs_timeout_user),
2255 [CMDID(IP_VS_SO_SET_STARTDAEMON)] = sizeof(struct ip_vs_daemon_user),
2256 [CMDID(IP_VS_SO_SET_STOPDAEMON)] = sizeof(struct ip_vs_daemon_user),
2257 [CMDID(IP_VS_SO_SET_ZERO)] = sizeof(struct ip_vs_service_user),
2260 union ip_vs_set_arglen {
2261 struct ip_vs_service_user field_IP_VS_SO_SET_ADD;
2262 struct ip_vs_service_user field_IP_VS_SO_SET_EDIT;
2263 struct ip_vs_service_user field_IP_VS_SO_SET_DEL;
2264 struct ip_vs_svcdest_user field_IP_VS_SO_SET_ADDDEST;
2265 struct ip_vs_svcdest_user field_IP_VS_SO_SET_DELDEST;
2266 struct ip_vs_svcdest_user field_IP_VS_SO_SET_EDITDEST;
2267 struct ip_vs_timeout_user field_IP_VS_SO_SET_TIMEOUT;
2268 struct ip_vs_daemon_user field_IP_VS_SO_SET_STARTDAEMON;
2269 struct ip_vs_daemon_user field_IP_VS_SO_SET_STOPDAEMON;
2270 struct ip_vs_service_user field_IP_VS_SO_SET_ZERO;
2273 #define MAX_SET_ARGLEN sizeof(union ip_vs_set_arglen)
2275 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2276 struct ip_vs_service_user *usvc_compat)
2278 memset(usvc, 0, sizeof(*usvc));
2281 usvc->protocol = usvc_compat->protocol;
2282 usvc->addr.ip = usvc_compat->addr;
2283 usvc->port = usvc_compat->port;
2284 usvc->fwmark = usvc_compat->fwmark;
2286 /* Deep copy of sched_name is not needed here */
2287 usvc->sched_name = usvc_compat->sched_name;
2289 usvc->flags = usvc_compat->flags;
2290 usvc->timeout = usvc_compat->timeout;
2291 usvc->netmask = usvc_compat->netmask;
2294 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2295 struct ip_vs_dest_user *udest_compat)
2297 memset(udest, 0, sizeof(*udest));
2299 udest->addr.ip = udest_compat->addr;
2300 udest->port = udest_compat->port;
2301 udest->conn_flags = udest_compat->conn_flags;
2302 udest->weight = udest_compat->weight;
2303 udest->u_threshold = udest_compat->u_threshold;
2304 udest->l_threshold = udest_compat->l_threshold;
2305 udest->af = AF_INET;
2309 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2311 struct net *net = sock_net(sk);
2313 unsigned char arg[MAX_SET_ARGLEN];
2314 struct ip_vs_service_user *usvc_compat;
2315 struct ip_vs_service_user_kern usvc;
2316 struct ip_vs_service *svc;
2317 struct ip_vs_dest_user *udest_compat;
2318 struct ip_vs_dest_user_kern udest;
2319 struct netns_ipvs *ipvs = net_ipvs(net);
2321 BUILD_BUG_ON(sizeof(arg) > 255);
2322 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2325 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2327 if (len != set_arglen[CMDID(cmd)]) {
2328 IP_VS_DBG(1, "set_ctl: len %u != %u\n",
2329 len, set_arglen[CMDID(cmd)]);
2333 if (copy_from_user(arg, user, len) != 0)
2336 /* increase the module use count */
2337 ip_vs_use_count_inc();
2339 /* Handle daemons since they have another lock */
2340 if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2341 cmd == IP_VS_SO_SET_STOPDAEMON) {
2342 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2344 if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2345 struct ipvs_sync_daemon_cfg cfg;
2347 memset(&cfg, 0, sizeof(cfg));
2348 strlcpy(cfg.mcast_ifn, dm->mcast_ifn,
2349 sizeof(cfg.mcast_ifn));
2350 cfg.syncid = dm->syncid;
2352 mutex_lock(&ipvs->sync_mutex);
2353 ret = start_sync_thread(ipvs, &cfg, dm->state);
2354 mutex_unlock(&ipvs->sync_mutex);
2357 mutex_lock(&ipvs->sync_mutex);
2358 ret = stop_sync_thread(ipvs, dm->state);
2359 mutex_unlock(&ipvs->sync_mutex);
2364 mutex_lock(&__ip_vs_mutex);
2365 if (cmd == IP_VS_SO_SET_FLUSH) {
2366 /* Flush the virtual service */
2367 ret = ip_vs_flush(ipvs, false);
2369 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2370 /* Set timeout values for (tcp tcpfin udp) */
2371 ret = ip_vs_set_timeout(ipvs, (struct ip_vs_timeout_user *)arg);
2375 usvc_compat = (struct ip_vs_service_user *)arg;
2376 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2378 /* We only use the new structs internally, so copy userspace compat
2379 * structs to extended internal versions */
2380 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2381 ip_vs_copy_udest_compat(&udest, udest_compat);
2383 if (cmd == IP_VS_SO_SET_ZERO) {
2384 /* if no service address is set, zero counters in all */
2385 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2386 ret = ip_vs_zero_all(ipvs);
2391 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2392 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2393 usvc.protocol != IPPROTO_SCTP) {
2394 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2395 usvc.protocol, &usvc.addr.ip,
2396 ntohs(usvc.port), usvc.sched_name);
2401 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2403 if (usvc.fwmark == 0)
2404 svc = __ip_vs_service_find(ipvs, usvc.af, usvc.protocol,
2405 &usvc.addr, usvc.port);
2407 svc = __ip_vs_svc_fwm_find(ipvs, usvc.af, usvc.fwmark);
2410 if (cmd != IP_VS_SO_SET_ADD
2411 && (svc == NULL || svc->protocol != usvc.protocol)) {
2417 case IP_VS_SO_SET_ADD:
2421 ret = ip_vs_add_service(ipvs, &usvc, &svc);
2423 case IP_VS_SO_SET_EDIT:
2424 ret = ip_vs_edit_service(svc, &usvc);
2426 case IP_VS_SO_SET_DEL:
2427 ret = ip_vs_del_service(svc);
2431 case IP_VS_SO_SET_ZERO:
2432 ret = ip_vs_zero_service(svc);
2434 case IP_VS_SO_SET_ADDDEST:
2435 ret = ip_vs_add_dest(svc, &udest);
2437 case IP_VS_SO_SET_EDITDEST:
2438 ret = ip_vs_edit_dest(svc, &udest);
2440 case IP_VS_SO_SET_DELDEST:
2441 ret = ip_vs_del_dest(svc, &udest);
2448 mutex_unlock(&__ip_vs_mutex);
2450 /* decrease the module use count */
2451 ip_vs_use_count_dec();
2458 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2460 struct ip_vs_scheduler *sched;
2461 struct ip_vs_kstats kstats;
2464 sched = rcu_dereference_protected(src->scheduler, 1);
2465 sched_name = sched ? sched->name : "none";
2466 dst->protocol = src->protocol;
2467 dst->addr = src->addr.ip;
2468 dst->port = src->port;
2469 dst->fwmark = src->fwmark;
2470 strlcpy(dst->sched_name, sched_name, sizeof(dst->sched_name));
2471 dst->flags = src->flags;
2472 dst->timeout = src->timeout / HZ;
2473 dst->netmask = src->netmask;
2474 dst->num_dests = src->num_dests;
2475 ip_vs_copy_stats(&kstats, &src->stats);
2476 ip_vs_export_stats_user(&dst->stats, &kstats);
2480 __ip_vs_get_service_entries(struct netns_ipvs *ipvs,
2481 const struct ip_vs_get_services *get,
2482 struct ip_vs_get_services __user *uptr)
2485 struct ip_vs_service *svc;
2486 struct ip_vs_service_entry entry;
2489 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2490 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2491 /* Only expose IPv4 entries to old interface */
2492 if (svc->af != AF_INET || (svc->ipvs != ipvs))
2495 if (count >= get->num_services)
2497 memset(&entry, 0, sizeof(entry));
2498 ip_vs_copy_service(&entry, svc);
2499 if (copy_to_user(&uptr->entrytable[count],
2500 &entry, sizeof(entry))) {
2508 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2509 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2510 /* Only expose IPv4 entries to old interface */
2511 if (svc->af != AF_INET || (svc->ipvs != ipvs))
2514 if (count >= get->num_services)
2516 memset(&entry, 0, sizeof(entry));
2517 ip_vs_copy_service(&entry, svc);
2518 if (copy_to_user(&uptr->entrytable[count],
2519 &entry, sizeof(entry))) {
2531 __ip_vs_get_dest_entries(struct netns_ipvs *ipvs, const struct ip_vs_get_dests *get,
2532 struct ip_vs_get_dests __user *uptr)
2534 struct ip_vs_service *svc;
2535 union nf_inet_addr addr = { .ip = get->addr };
2540 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, get->fwmark);
2542 svc = __ip_vs_service_find(ipvs, AF_INET, get->protocol, &addr,
2548 struct ip_vs_dest *dest;
2549 struct ip_vs_dest_entry entry;
2550 struct ip_vs_kstats kstats;
2552 memset(&entry, 0, sizeof(entry));
2553 list_for_each_entry(dest, &svc->destinations, n_list) {
2554 if (count >= get->num_dests)
2557 /* Cannot expose heterogeneous members via sockopt
2560 if (dest->af != svc->af)
2563 entry.addr = dest->addr.ip;
2564 entry.port = dest->port;
2565 entry.conn_flags = atomic_read(&dest->conn_flags);
2566 entry.weight = atomic_read(&dest->weight);
2567 entry.u_threshold = dest->u_threshold;
2568 entry.l_threshold = dest->l_threshold;
2569 entry.activeconns = atomic_read(&dest->activeconns);
2570 entry.inactconns = atomic_read(&dest->inactconns);
2571 entry.persistconns = atomic_read(&dest->persistconns);
2572 ip_vs_copy_stats(&kstats, &dest->stats);
2573 ip_vs_export_stats_user(&entry.stats, &kstats);
2574 if (copy_to_user(&uptr->entrytable[count],
2575 &entry, sizeof(entry))) {
2587 __ip_vs_get_timeouts(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2589 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2590 struct ip_vs_proto_data *pd;
2593 memset(u, 0, sizeof (*u));
2595 #ifdef CONFIG_IP_VS_PROTO_TCP
2596 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2597 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2598 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2600 #ifdef CONFIG_IP_VS_PROTO_UDP
2601 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2603 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2607 static const unsigned char get_arglen[CMDID(IP_VS_SO_GET_MAX) + 1] = {
2608 [CMDID(IP_VS_SO_GET_VERSION)] = 64,
2609 [CMDID(IP_VS_SO_GET_INFO)] = sizeof(struct ip_vs_getinfo),
2610 [CMDID(IP_VS_SO_GET_SERVICES)] = sizeof(struct ip_vs_get_services),
2611 [CMDID(IP_VS_SO_GET_SERVICE)] = sizeof(struct ip_vs_service_entry),
2612 [CMDID(IP_VS_SO_GET_DESTS)] = sizeof(struct ip_vs_get_dests),
2613 [CMDID(IP_VS_SO_GET_TIMEOUT)] = sizeof(struct ip_vs_timeout_user),
2614 [CMDID(IP_VS_SO_GET_DAEMON)] = 2 * sizeof(struct ip_vs_daemon_user),
2617 union ip_vs_get_arglen {
2618 char field_IP_VS_SO_GET_VERSION[64];
2619 struct ip_vs_getinfo field_IP_VS_SO_GET_INFO;
2620 struct ip_vs_get_services field_IP_VS_SO_GET_SERVICES;
2621 struct ip_vs_service_entry field_IP_VS_SO_GET_SERVICE;
2622 struct ip_vs_get_dests field_IP_VS_SO_GET_DESTS;
2623 struct ip_vs_timeout_user field_IP_VS_SO_GET_TIMEOUT;
2624 struct ip_vs_daemon_user field_IP_VS_SO_GET_DAEMON[2];
2627 #define MAX_GET_ARGLEN sizeof(union ip_vs_get_arglen)
2630 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2632 unsigned char arg[MAX_GET_ARGLEN];
2634 unsigned int copylen;
2635 struct net *net = sock_net(sk);
2636 struct netns_ipvs *ipvs = net_ipvs(net);
2639 BUILD_BUG_ON(sizeof(arg) > 255);
2640 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2643 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2646 copylen = get_arglen[CMDID(cmd)];
2647 if (*len < (int) copylen) {
2648 IP_VS_DBG(1, "get_ctl: len %d < %u\n", *len, copylen);
2652 if (copy_from_user(arg, user, copylen) != 0)
2655 * Handle daemons first since it has its own locking
2657 if (cmd == IP_VS_SO_GET_DAEMON) {
2658 struct ip_vs_daemon_user d[2];
2660 memset(&d, 0, sizeof(d));
2661 mutex_lock(&ipvs->sync_mutex);
2662 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2663 d[0].state = IP_VS_STATE_MASTER;
2664 strlcpy(d[0].mcast_ifn, ipvs->mcfg.mcast_ifn,
2665 sizeof(d[0].mcast_ifn));
2666 d[0].syncid = ipvs->mcfg.syncid;
2668 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2669 d[1].state = IP_VS_STATE_BACKUP;
2670 strlcpy(d[1].mcast_ifn, ipvs->bcfg.mcast_ifn,
2671 sizeof(d[1].mcast_ifn));
2672 d[1].syncid = ipvs->bcfg.syncid;
2674 if (copy_to_user(user, &d, sizeof(d)) != 0)
2676 mutex_unlock(&ipvs->sync_mutex);
2680 mutex_lock(&__ip_vs_mutex);
2682 case IP_VS_SO_GET_VERSION:
2686 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2687 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2688 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2692 *len = strlen(buf)+1;
2696 case IP_VS_SO_GET_INFO:
2698 struct ip_vs_getinfo info;
2699 info.version = IP_VS_VERSION_CODE;
2700 info.size = ip_vs_conn_tab_size;
2701 info.num_services = ipvs->num_services;
2702 if (copy_to_user(user, &info, sizeof(info)) != 0)
2707 case IP_VS_SO_GET_SERVICES:
2709 struct ip_vs_get_services *get;
2712 get = (struct ip_vs_get_services *)arg;
2713 size = sizeof(*get) +
2714 sizeof(struct ip_vs_service_entry) * get->num_services;
2716 pr_err("length: %u != %u\n", *len, size);
2720 ret = __ip_vs_get_service_entries(ipvs, get, user);
2724 case IP_VS_SO_GET_SERVICE:
2726 struct ip_vs_service_entry *entry;
2727 struct ip_vs_service *svc;
2728 union nf_inet_addr addr;
2730 entry = (struct ip_vs_service_entry *)arg;
2731 addr.ip = entry->addr;
2734 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, entry->fwmark);
2736 svc = __ip_vs_service_find(ipvs, AF_INET,
2737 entry->protocol, &addr,
2741 ip_vs_copy_service(entry, svc);
2742 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2749 case IP_VS_SO_GET_DESTS:
2751 struct ip_vs_get_dests *get;
2754 get = (struct ip_vs_get_dests *)arg;
2755 size = sizeof(*get) +
2756 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2758 pr_err("length: %u != %u\n", *len, size);
2762 ret = __ip_vs_get_dest_entries(ipvs, get, user);
2766 case IP_VS_SO_GET_TIMEOUT:
2768 struct ip_vs_timeout_user t;
2770 __ip_vs_get_timeouts(ipvs, &t);
2771 if (copy_to_user(user, &t, sizeof(t)) != 0)
2781 mutex_unlock(&__ip_vs_mutex);
2786 static struct nf_sockopt_ops ip_vs_sockopts = {
2788 .set_optmin = IP_VS_BASE_CTL,
2789 .set_optmax = IP_VS_SO_SET_MAX+1,
2790 .set = do_ip_vs_set_ctl,
2791 .get_optmin = IP_VS_BASE_CTL,
2792 .get_optmax = IP_VS_SO_GET_MAX+1,
2793 .get = do_ip_vs_get_ctl,
2794 .owner = THIS_MODULE,
2798 * Generic Netlink interface
2801 /* IPVS genetlink family */
2802 static struct genl_family ip_vs_genl_family = {
2803 .id = GENL_ID_GENERATE,
2805 .name = IPVS_GENL_NAME,
2806 .version = IPVS_GENL_VERSION,
2807 .maxattr = IPVS_CMD_MAX,
2808 .netnsok = true, /* Make ipvsadm to work on netns */
2811 /* Policy used for first-level command attributes */
2812 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2813 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2814 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2815 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2816 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2817 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2818 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2821 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2822 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2823 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2824 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2825 .len = IP_VS_IFNAME_MAXLEN },
2826 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2827 [IPVS_DAEMON_ATTR_SYNC_MAXLEN] = { .type = NLA_U16 },
2828 [IPVS_DAEMON_ATTR_MCAST_GROUP] = { .type = NLA_U32 },
2829 [IPVS_DAEMON_ATTR_MCAST_GROUP6] = { .len = sizeof(struct in6_addr) },
2830 [IPVS_DAEMON_ATTR_MCAST_PORT] = { .type = NLA_U16 },
2831 [IPVS_DAEMON_ATTR_MCAST_TTL] = { .type = NLA_U8 },
2834 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2835 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2836 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2837 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2838 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2839 .len = sizeof(union nf_inet_addr) },
2840 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2841 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2842 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2843 .len = IP_VS_SCHEDNAME_MAXLEN },
2844 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2845 .len = IP_VS_PENAME_MAXLEN },
2846 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2847 .len = sizeof(struct ip_vs_flags) },
2848 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2849 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2850 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2853 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2854 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2855 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2856 .len = sizeof(union nf_inet_addr) },
2857 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2858 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2859 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2860 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2861 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2862 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2863 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2864 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2865 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2866 [IPVS_DEST_ATTR_ADDR_FAMILY] = { .type = NLA_U16 },
2869 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2870 struct ip_vs_kstats *kstats)
2872 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2877 if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, (u32)kstats->conns) ||
2878 nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, (u32)kstats->inpkts) ||
2879 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, (u32)kstats->outpkts) ||
2880 nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes) ||
2881 nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes) ||
2882 nla_put_u32(skb, IPVS_STATS_ATTR_CPS, (u32)kstats->cps) ||
2883 nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, (u32)kstats->inpps) ||
2884 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, (u32)kstats->outpps) ||
2885 nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, (u32)kstats->inbps) ||
2886 nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, (u32)kstats->outbps))
2887 goto nla_put_failure;
2888 nla_nest_end(skb, nl_stats);
2893 nla_nest_cancel(skb, nl_stats);
2897 static int ip_vs_genl_fill_stats64(struct sk_buff *skb, int container_type,
2898 struct ip_vs_kstats *kstats)
2900 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2905 if (nla_put_u64(skb, IPVS_STATS_ATTR_CONNS, kstats->conns) ||
2906 nla_put_u64(skb, IPVS_STATS_ATTR_INPKTS, kstats->inpkts) ||
2907 nla_put_u64(skb, IPVS_STATS_ATTR_OUTPKTS, kstats->outpkts) ||
2908 nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes) ||
2909 nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes) ||
2910 nla_put_u64(skb, IPVS_STATS_ATTR_CPS, kstats->cps) ||
2911 nla_put_u64(skb, IPVS_STATS_ATTR_INPPS, kstats->inpps) ||
2912 nla_put_u64(skb, IPVS_STATS_ATTR_OUTPPS, kstats->outpps) ||
2913 nla_put_u64(skb, IPVS_STATS_ATTR_INBPS, kstats->inbps) ||
2914 nla_put_u64(skb, IPVS_STATS_ATTR_OUTBPS, kstats->outbps))
2915 goto nla_put_failure;
2916 nla_nest_end(skb, nl_stats);
2921 nla_nest_cancel(skb, nl_stats);
2925 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2926 struct ip_vs_service *svc)
2928 struct ip_vs_scheduler *sched;
2929 struct ip_vs_pe *pe;
2930 struct nlattr *nl_service;
2931 struct ip_vs_flags flags = { .flags = svc->flags,
2933 struct ip_vs_kstats kstats;
2936 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2940 if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2941 goto nla_put_failure;
2943 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2944 goto nla_put_failure;
2946 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2947 nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2948 nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2949 goto nla_put_failure;
2952 sched = rcu_dereference_protected(svc->scheduler, 1);
2953 sched_name = sched ? sched->name : "none";
2954 pe = rcu_dereference_protected(svc->pe, 1);
2955 if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched_name) ||
2956 (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
2957 nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2958 nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2959 nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
2960 goto nla_put_failure;
2961 ip_vs_copy_stats(&kstats, &svc->stats);
2962 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &kstats))
2963 goto nla_put_failure;
2964 if (ip_vs_genl_fill_stats64(skb, IPVS_SVC_ATTR_STATS64, &kstats))
2965 goto nla_put_failure;
2967 nla_nest_end(skb, nl_service);
2972 nla_nest_cancel(skb, nl_service);
2976 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2977 struct ip_vs_service *svc,
2978 struct netlink_callback *cb)
2982 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2983 &ip_vs_genl_family, NLM_F_MULTI,
2984 IPVS_CMD_NEW_SERVICE);
2988 if (ip_vs_genl_fill_service(skb, svc) < 0)
2989 goto nla_put_failure;
2991 genlmsg_end(skb, hdr);
2995 genlmsg_cancel(skb, hdr);
2999 static int ip_vs_genl_dump_services(struct sk_buff *skb,
3000 struct netlink_callback *cb)
3003 int start = cb->args[0];
3004 struct ip_vs_service *svc;
3005 struct net *net = sock_net(skb->sk);
3006 struct netns_ipvs *ipvs = net_ipvs(net);
3008 mutex_lock(&__ip_vs_mutex);
3009 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3010 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
3011 if (++idx <= start || (svc->ipvs != ipvs))
3013 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3015 goto nla_put_failure;
3020 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3021 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
3022 if (++idx <= start || (svc->ipvs != ipvs))
3024 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3026 goto nla_put_failure;
3032 mutex_unlock(&__ip_vs_mutex);
3038 static int ip_vs_genl_parse_service(struct netns_ipvs *ipvs,
3039 struct ip_vs_service_user_kern *usvc,
3040 struct nlattr *nla, int full_entry,
3041 struct ip_vs_service **ret_svc)
3043 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
3044 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
3045 struct ip_vs_service *svc;
3047 /* Parse mandatory identifying service fields first */
3049 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
3052 nla_af = attrs[IPVS_SVC_ATTR_AF];
3053 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
3054 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
3055 nla_port = attrs[IPVS_SVC_ATTR_PORT];
3056 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
3058 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3061 memset(usvc, 0, sizeof(*usvc));
3063 usvc->af = nla_get_u16(nla_af);
3064 #ifdef CONFIG_IP_VS_IPV6
3065 if (usvc->af != AF_INET && usvc->af != AF_INET6)
3067 if (usvc->af != AF_INET)
3069 return -EAFNOSUPPORT;
3072 usvc->protocol = IPPROTO_TCP;
3073 usvc->fwmark = nla_get_u32(nla_fwmark);
3075 usvc->protocol = nla_get_u16(nla_protocol);
3076 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3077 usvc->port = nla_get_be16(nla_port);
3083 svc = __ip_vs_svc_fwm_find(ipvs, usvc->af, usvc->fwmark);
3085 svc = __ip_vs_service_find(ipvs, usvc->af, usvc->protocol,
3086 &usvc->addr, usvc->port);
3090 /* If a full entry was requested, check for the additional fields */
3092 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3094 struct ip_vs_flags flags;
3096 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3097 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3098 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3099 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3100 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3102 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3105 nla_memcpy(&flags, nla_flags, sizeof(flags));
3107 /* prefill flags from service if it already exists */
3109 usvc->flags = svc->flags;
3111 /* set new flags from userland */
3112 usvc->flags = (usvc->flags & ~flags.mask) |
3113 (flags.flags & flags.mask);
3114 usvc->sched_name = nla_data(nla_sched);
3115 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3116 usvc->timeout = nla_get_u32(nla_timeout);
3117 usvc->netmask = nla_get_be32(nla_netmask);
3123 static struct ip_vs_service *ip_vs_genl_find_service(struct netns_ipvs *ipvs,
3126 struct ip_vs_service_user_kern usvc;
3127 struct ip_vs_service *svc;
3130 ret = ip_vs_genl_parse_service(ipvs, &usvc, nla, 0, &svc);
3131 return ret ? ERR_PTR(ret) : svc;
3134 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3136 struct nlattr *nl_dest;
3137 struct ip_vs_kstats kstats;
3139 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3143 if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3144 nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3145 nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3146 (atomic_read(&dest->conn_flags) &
3147 IP_VS_CONN_F_FWD_MASK)) ||
3148 nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3149 atomic_read(&dest->weight)) ||
3150 nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3151 nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3152 nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3153 atomic_read(&dest->activeconns)) ||
3154 nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3155 atomic_read(&dest->inactconns)) ||
3156 nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3157 atomic_read(&dest->persistconns)) ||
3158 nla_put_u16(skb, IPVS_DEST_ATTR_ADDR_FAMILY, dest->af))
3159 goto nla_put_failure;
3160 ip_vs_copy_stats(&kstats, &dest->stats);
3161 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &kstats))
3162 goto nla_put_failure;
3163 if (ip_vs_genl_fill_stats64(skb, IPVS_DEST_ATTR_STATS64, &kstats))
3164 goto nla_put_failure;
3166 nla_nest_end(skb, nl_dest);
3171 nla_nest_cancel(skb, nl_dest);
3175 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3176 struct netlink_callback *cb)
3180 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3181 &ip_vs_genl_family, NLM_F_MULTI,
3186 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3187 goto nla_put_failure;
3189 genlmsg_end(skb, hdr);
3193 genlmsg_cancel(skb, hdr);
3197 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3198 struct netlink_callback *cb)
3201 int start = cb->args[0];
3202 struct ip_vs_service *svc;
3203 struct ip_vs_dest *dest;
3204 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3205 struct net *net = sock_net(skb->sk);
3206 struct netns_ipvs *ipvs = net_ipvs(net);
3208 mutex_lock(&__ip_vs_mutex);
3210 /* Try to find the service for which to dump destinations */
3211 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3212 IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3216 svc = ip_vs_genl_find_service(ipvs, attrs[IPVS_CMD_ATTR_SERVICE]);
3217 if (IS_ERR(svc) || svc == NULL)
3220 /* Dump the destinations */
3221 list_for_each_entry(dest, &svc->destinations, n_list) {
3224 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3226 goto nla_put_failure;
3234 mutex_unlock(&__ip_vs_mutex);
3239 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3240 struct nlattr *nla, int full_entry)
3242 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3243 struct nlattr *nla_addr, *nla_port;
3244 struct nlattr *nla_addr_family;
3246 /* Parse mandatory identifying destination fields first */
3248 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3251 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3252 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3253 nla_addr_family = attrs[IPVS_DEST_ATTR_ADDR_FAMILY];
3255 if (!(nla_addr && nla_port))
3258 memset(udest, 0, sizeof(*udest));
3260 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3261 udest->port = nla_get_be16(nla_port);
3263 if (nla_addr_family)
3264 udest->af = nla_get_u16(nla_addr_family);
3268 /* If a full entry was requested, check for the additional fields */
3270 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3273 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3274 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3275 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3276 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3278 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3281 udest->conn_flags = nla_get_u32(nla_fwd)
3282 & IP_VS_CONN_F_FWD_MASK;
3283 udest->weight = nla_get_u32(nla_weight);
3284 udest->u_threshold = nla_get_u32(nla_u_thresh);
3285 udest->l_threshold = nla_get_u32(nla_l_thresh);
3291 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3292 struct ipvs_sync_daemon_cfg *c)
3294 struct nlattr *nl_daemon;
3296 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3300 if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3301 nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, c->mcast_ifn) ||
3302 nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, c->syncid) ||
3303 nla_put_u16(skb, IPVS_DAEMON_ATTR_SYNC_MAXLEN, c->sync_maxlen) ||
3304 nla_put_u16(skb, IPVS_DAEMON_ATTR_MCAST_PORT, c->mcast_port) ||
3305 nla_put_u8(skb, IPVS_DAEMON_ATTR_MCAST_TTL, c->mcast_ttl))
3306 goto nla_put_failure;
3307 #ifdef CONFIG_IP_VS_IPV6
3308 if (c->mcast_af == AF_INET6) {
3309 if (nla_put_in6_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP6,
3310 &c->mcast_group.in6))
3311 goto nla_put_failure;
3314 if (c->mcast_af == AF_INET &&
3315 nla_put_in_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP,
3317 goto nla_put_failure;
3318 nla_nest_end(skb, nl_daemon);
3323 nla_nest_cancel(skb, nl_daemon);
3327 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3328 struct ipvs_sync_daemon_cfg *c,
3329 struct netlink_callback *cb)
3332 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3333 &ip_vs_genl_family, NLM_F_MULTI,
3334 IPVS_CMD_NEW_DAEMON);
3338 if (ip_vs_genl_fill_daemon(skb, state, c))
3339 goto nla_put_failure;
3341 genlmsg_end(skb, hdr);
3345 genlmsg_cancel(skb, hdr);
3349 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3350 struct netlink_callback *cb)
3352 struct net *net = sock_net(skb->sk);
3353 struct netns_ipvs *ipvs = net_ipvs(net);
3355 mutex_lock(&ipvs->sync_mutex);
3356 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3357 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3358 &ipvs->mcfg, cb) < 0)
3359 goto nla_put_failure;
3364 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3365 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3366 &ipvs->bcfg, cb) < 0)
3367 goto nla_put_failure;
3373 mutex_unlock(&ipvs->sync_mutex);
3378 static int ip_vs_genl_new_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3380 struct ipvs_sync_daemon_cfg c;
3384 memset(&c, 0, sizeof(c));
3385 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3386 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3387 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3389 strlcpy(c.mcast_ifn, nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3390 sizeof(c.mcast_ifn));
3391 c.syncid = nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]);
3393 a = attrs[IPVS_DAEMON_ATTR_SYNC_MAXLEN];
3395 c.sync_maxlen = nla_get_u16(a);
3397 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP];
3399 c.mcast_af = AF_INET;
3400 c.mcast_group.ip = nla_get_in_addr(a);
3401 if (!ipv4_is_multicast(c.mcast_group.ip))
3404 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP6];
3406 #ifdef CONFIG_IP_VS_IPV6
3409 c.mcast_af = AF_INET6;
3410 c.mcast_group.in6 = nla_get_in6_addr(a);
3411 addr_type = ipv6_addr_type(&c.mcast_group.in6);
3412 if (!(addr_type & IPV6_ADDR_MULTICAST))
3415 return -EAFNOSUPPORT;
3420 a = attrs[IPVS_DAEMON_ATTR_MCAST_PORT];
3422 c.mcast_port = nla_get_u16(a);
3424 a = attrs[IPVS_DAEMON_ATTR_MCAST_TTL];
3426 c.mcast_ttl = nla_get_u8(a);
3428 /* The synchronization protocol is incompatible with mixed family
3431 if (ipvs->mixed_address_family_dests > 0)
3435 mutex_lock(&ipvs->sync_mutex);
3436 ret = start_sync_thread(ipvs, &c,
3437 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3438 mutex_unlock(&ipvs->sync_mutex);
3443 static int ip_vs_genl_del_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3447 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3450 mutex_lock(&ipvs->sync_mutex);
3451 ret = stop_sync_thread(ipvs,
3452 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3453 mutex_unlock(&ipvs->sync_mutex);
3457 static int ip_vs_genl_set_config(struct netns_ipvs *ipvs, struct nlattr **attrs)
3459 struct ip_vs_timeout_user t;
3461 __ip_vs_get_timeouts(ipvs, &t);
3463 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3464 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3466 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3468 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3470 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3471 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3473 return ip_vs_set_timeout(ipvs, &t);
3476 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3478 int ret = -EINVAL, cmd;
3479 struct net *net = sock_net(skb->sk);
3480 struct netns_ipvs *ipvs = net_ipvs(net);
3482 cmd = info->genlhdr->cmd;
3484 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3485 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3487 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3488 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3489 info->attrs[IPVS_CMD_ATTR_DAEMON],
3490 ip_vs_daemon_policy))
3493 if (cmd == IPVS_CMD_NEW_DAEMON)
3494 ret = ip_vs_genl_new_daemon(ipvs, daemon_attrs);
3496 ret = ip_vs_genl_del_daemon(ipvs, daemon_attrs);
3503 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3505 struct ip_vs_service *svc = NULL;
3506 struct ip_vs_service_user_kern usvc;
3507 struct ip_vs_dest_user_kern udest;
3509 int need_full_svc = 0, need_full_dest = 0;
3510 struct net *net = sock_net(skb->sk);
3511 struct netns_ipvs *ipvs = net_ipvs(net);
3513 cmd = info->genlhdr->cmd;
3515 mutex_lock(&__ip_vs_mutex);
3517 if (cmd == IPVS_CMD_FLUSH) {
3518 ret = ip_vs_flush(ipvs, false);
3520 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3521 ret = ip_vs_genl_set_config(ipvs, info->attrs);
3523 } else if (cmd == IPVS_CMD_ZERO &&
3524 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3525 ret = ip_vs_zero_all(ipvs);
3529 /* All following commands require a service argument, so check if we
3530 * received a valid one. We need a full service specification when
3531 * adding / editing a service. Only identifying members otherwise. */
3532 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3535 ret = ip_vs_genl_parse_service(ipvs, &usvc,
3536 info->attrs[IPVS_CMD_ATTR_SERVICE],
3537 need_full_svc, &svc);
3541 /* Unless we're adding a new service, the service must already exist */
3542 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3547 /* Destination commands require a valid destination argument. For
3548 * adding / editing a destination, we need a full destination
3550 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3551 cmd == IPVS_CMD_DEL_DEST) {
3552 if (cmd != IPVS_CMD_DEL_DEST)
3555 ret = ip_vs_genl_parse_dest(&udest,
3556 info->attrs[IPVS_CMD_ATTR_DEST],
3561 /* Old protocols did not allow the user to specify address
3562 * family, so we set it to zero instead. We also didn't
3563 * allow heterogeneous pools in the old code, so it's safe
3564 * to assume that this will have the same address family as
3570 if (udest.af != svc->af && cmd != IPVS_CMD_DEL_DEST) {
3571 /* The synchronization protocol is incompatible
3572 * with mixed family services
3574 if (ipvs->sync_state) {
3579 /* Which connection types do we support? */
3580 switch (udest.conn_flags) {
3581 case IP_VS_CONN_F_TUNNEL:
3582 /* We are able to forward this */
3592 case IPVS_CMD_NEW_SERVICE:
3594 ret = ip_vs_add_service(ipvs, &usvc, &svc);
3598 case IPVS_CMD_SET_SERVICE:
3599 ret = ip_vs_edit_service(svc, &usvc);
3601 case IPVS_CMD_DEL_SERVICE:
3602 ret = ip_vs_del_service(svc);
3603 /* do not use svc, it can be freed */
3605 case IPVS_CMD_NEW_DEST:
3606 ret = ip_vs_add_dest(svc, &udest);
3608 case IPVS_CMD_SET_DEST:
3609 ret = ip_vs_edit_dest(svc, &udest);
3611 case IPVS_CMD_DEL_DEST:
3612 ret = ip_vs_del_dest(svc, &udest);
3615 ret = ip_vs_zero_service(svc);
3622 mutex_unlock(&__ip_vs_mutex);
3627 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3629 struct sk_buff *msg;
3631 int ret, cmd, reply_cmd;
3632 struct net *net = sock_net(skb->sk);
3633 struct netns_ipvs *ipvs = net_ipvs(net);
3635 cmd = info->genlhdr->cmd;
3637 if (cmd == IPVS_CMD_GET_SERVICE)
3638 reply_cmd = IPVS_CMD_NEW_SERVICE;
3639 else if (cmd == IPVS_CMD_GET_INFO)
3640 reply_cmd = IPVS_CMD_SET_INFO;
3641 else if (cmd == IPVS_CMD_GET_CONFIG)
3642 reply_cmd = IPVS_CMD_SET_CONFIG;
3644 pr_err("unknown Generic Netlink command\n");
3648 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3652 mutex_lock(&__ip_vs_mutex);
3654 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3656 goto nla_put_failure;
3659 case IPVS_CMD_GET_SERVICE:
3661 struct ip_vs_service *svc;
3663 svc = ip_vs_genl_find_service(ipvs,
3664 info->attrs[IPVS_CMD_ATTR_SERVICE]);
3669 ret = ip_vs_genl_fill_service(msg, svc);
3671 goto nla_put_failure;
3680 case IPVS_CMD_GET_CONFIG:
3682 struct ip_vs_timeout_user t;
3684 __ip_vs_get_timeouts(ipvs, &t);
3685 #ifdef CONFIG_IP_VS_PROTO_TCP
3686 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3688 nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3690 goto nla_put_failure;
3692 #ifdef CONFIG_IP_VS_PROTO_UDP
3693 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3694 goto nla_put_failure;
3700 case IPVS_CMD_GET_INFO:
3701 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3702 IP_VS_VERSION_CODE) ||
3703 nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3704 ip_vs_conn_tab_size))
3705 goto nla_put_failure;
3709 genlmsg_end(msg, reply);
3710 ret = genlmsg_reply(msg, info);
3714 pr_err("not enough space in Netlink message\n");
3720 mutex_unlock(&__ip_vs_mutex);
3726 static const struct genl_ops ip_vs_genl_ops[] = {
3728 .cmd = IPVS_CMD_NEW_SERVICE,
3729 .flags = GENL_ADMIN_PERM,
3730 .policy = ip_vs_cmd_policy,
3731 .doit = ip_vs_genl_set_cmd,
3734 .cmd = IPVS_CMD_SET_SERVICE,
3735 .flags = GENL_ADMIN_PERM,
3736 .policy = ip_vs_cmd_policy,
3737 .doit = ip_vs_genl_set_cmd,
3740 .cmd = IPVS_CMD_DEL_SERVICE,
3741 .flags = GENL_ADMIN_PERM,
3742 .policy = ip_vs_cmd_policy,
3743 .doit = ip_vs_genl_set_cmd,
3746 .cmd = IPVS_CMD_GET_SERVICE,
3747 .flags = GENL_ADMIN_PERM,
3748 .doit = ip_vs_genl_get_cmd,
3749 .dumpit = ip_vs_genl_dump_services,
3750 .policy = ip_vs_cmd_policy,
3753 .cmd = IPVS_CMD_NEW_DEST,
3754 .flags = GENL_ADMIN_PERM,
3755 .policy = ip_vs_cmd_policy,
3756 .doit = ip_vs_genl_set_cmd,
3759 .cmd = IPVS_CMD_SET_DEST,
3760 .flags = GENL_ADMIN_PERM,
3761 .policy = ip_vs_cmd_policy,
3762 .doit = ip_vs_genl_set_cmd,
3765 .cmd = IPVS_CMD_DEL_DEST,
3766 .flags = GENL_ADMIN_PERM,
3767 .policy = ip_vs_cmd_policy,
3768 .doit = ip_vs_genl_set_cmd,
3771 .cmd = IPVS_CMD_GET_DEST,
3772 .flags = GENL_ADMIN_PERM,
3773 .policy = ip_vs_cmd_policy,
3774 .dumpit = ip_vs_genl_dump_dests,
3777 .cmd = IPVS_CMD_NEW_DAEMON,
3778 .flags = GENL_ADMIN_PERM,
3779 .policy = ip_vs_cmd_policy,
3780 .doit = ip_vs_genl_set_daemon,
3783 .cmd = IPVS_CMD_DEL_DAEMON,
3784 .flags = GENL_ADMIN_PERM,
3785 .policy = ip_vs_cmd_policy,
3786 .doit = ip_vs_genl_set_daemon,
3789 .cmd = IPVS_CMD_GET_DAEMON,
3790 .flags = GENL_ADMIN_PERM,
3791 .dumpit = ip_vs_genl_dump_daemons,
3794 .cmd = IPVS_CMD_SET_CONFIG,
3795 .flags = GENL_ADMIN_PERM,
3796 .policy = ip_vs_cmd_policy,
3797 .doit = ip_vs_genl_set_cmd,
3800 .cmd = IPVS_CMD_GET_CONFIG,
3801 .flags = GENL_ADMIN_PERM,
3802 .doit = ip_vs_genl_get_cmd,
3805 .cmd = IPVS_CMD_GET_INFO,
3806 .flags = GENL_ADMIN_PERM,
3807 .doit = ip_vs_genl_get_cmd,
3810 .cmd = IPVS_CMD_ZERO,
3811 .flags = GENL_ADMIN_PERM,
3812 .policy = ip_vs_cmd_policy,
3813 .doit = ip_vs_genl_set_cmd,
3816 .cmd = IPVS_CMD_FLUSH,
3817 .flags = GENL_ADMIN_PERM,
3818 .doit = ip_vs_genl_set_cmd,
3822 static int __init ip_vs_genl_register(void)
3824 return genl_register_family_with_ops(&ip_vs_genl_family,
3828 static void ip_vs_genl_unregister(void)
3830 genl_unregister_family(&ip_vs_genl_family);
3833 /* End of Generic Netlink interface definitions */
3836 * per netns intit/exit func.
3838 #ifdef CONFIG_SYSCTL
3839 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs)
3841 struct net *net = ipvs->net;
3843 struct ctl_table *tbl;
3845 atomic_set(&ipvs->dropentry, 0);
3846 spin_lock_init(&ipvs->dropentry_lock);
3847 spin_lock_init(&ipvs->droppacket_lock);
3848 spin_lock_init(&ipvs->securetcp_lock);
3850 if (!net_eq(net, &init_net)) {
3851 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3855 /* Don't export sysctls to unprivileged users */
3856 if (net->user_ns != &init_user_ns)
3857 tbl[0].procname = NULL;
3860 /* Initialize sysctl defaults */
3861 for (idx = 0; idx < ARRAY_SIZE(vs_vars); idx++) {
3862 if (tbl[idx].proc_handler == proc_do_defense_mode)
3863 tbl[idx].extra2 = ipvs;
3866 ipvs->sysctl_amemthresh = 1024;
3867 tbl[idx++].data = &ipvs->sysctl_amemthresh;
3868 ipvs->sysctl_am_droprate = 10;
3869 tbl[idx++].data = &ipvs->sysctl_am_droprate;
3870 tbl[idx++].data = &ipvs->sysctl_drop_entry;
3871 tbl[idx++].data = &ipvs->sysctl_drop_packet;
3872 #ifdef CONFIG_IP_VS_NFCT
3873 tbl[idx++].data = &ipvs->sysctl_conntrack;
3875 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3876 ipvs->sysctl_snat_reroute = 1;
3877 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3878 ipvs->sysctl_sync_ver = 1;
3879 tbl[idx++].data = &ipvs->sysctl_sync_ver;
3880 ipvs->sysctl_sync_ports = 1;
3881 tbl[idx++].data = &ipvs->sysctl_sync_ports;
3882 tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
3883 ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3884 tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3885 ipvs->sysctl_sync_sock_size = 0;
3886 tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3887 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3888 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3889 tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
3890 tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
3891 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3892 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3893 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3894 tbl[idx].data = &ipvs->sysctl_sync_threshold;
3895 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3896 ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3897 tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3898 ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3899 tbl[idx++].data = &ipvs->sysctl_sync_retries;
3900 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3901 ipvs->sysctl_pmtu_disc = 1;
3902 tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
3903 tbl[idx++].data = &ipvs->sysctl_backup_only;
3904 ipvs->sysctl_conn_reuse_mode = 1;
3905 tbl[idx++].data = &ipvs->sysctl_conn_reuse_mode;
3906 tbl[idx++].data = &ipvs->sysctl_schedule_icmp;
3907 tbl[idx++].data = &ipvs->sysctl_ignore_tunneled;
3909 ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3910 if (ipvs->sysctl_hdr == NULL) {
3911 if (!net_eq(net, &init_net))
3915 ip_vs_start_estimator(ipvs, &ipvs->tot_stats);
3916 ipvs->sysctl_tbl = tbl;
3917 /* Schedule defense work */
3918 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3919 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3924 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs)
3926 struct net *net = ipvs->net;
3928 cancel_delayed_work_sync(&ipvs->defense_work);
3929 cancel_work_sync(&ipvs->defense_work.work);
3930 unregister_net_sysctl_table(ipvs->sysctl_hdr);
3931 ip_vs_stop_estimator(ipvs, &ipvs->tot_stats);
3933 if (!net_eq(net, &init_net))
3934 kfree(ipvs->sysctl_tbl);
3939 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs) { return 0; }
3940 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs) { }
3944 static struct notifier_block ip_vs_dst_notifier = {
3945 .notifier_call = ip_vs_dst_event,
3948 int __net_init ip_vs_control_net_init(struct netns_ipvs *ipvs)
3950 struct net *net = ipvs->net;
3953 /* Initialize rs_table */
3954 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
3955 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
3957 INIT_LIST_HEAD(&ipvs->dest_trash);
3958 spin_lock_init(&ipvs->dest_trash_lock);
3959 setup_timer(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire,
3960 (unsigned long) ipvs);
3961 atomic_set(&ipvs->ftpsvc_counter, 0);
3962 atomic_set(&ipvs->nullsvc_counter, 0);
3965 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
3966 if (!ipvs->tot_stats.cpustats)
3969 for_each_possible_cpu(i) {
3970 struct ip_vs_cpu_stats *ipvs_tot_stats;
3971 ipvs_tot_stats = per_cpu_ptr(ipvs->tot_stats.cpustats, i);
3972 u64_stats_init(&ipvs_tot_stats->syncp);
3975 spin_lock_init(&ipvs->tot_stats.lock);
3977 proc_create("ip_vs", 0, net->proc_net, &ip_vs_info_fops);
3978 proc_create("ip_vs_stats", 0, net->proc_net, &ip_vs_stats_fops);
3979 proc_create("ip_vs_stats_percpu", 0, net->proc_net,
3980 &ip_vs_stats_percpu_fops);
3982 if (ip_vs_control_net_init_sysctl(ipvs))
3988 free_percpu(ipvs->tot_stats.cpustats);
3992 void __net_exit ip_vs_control_net_cleanup(struct netns_ipvs *ipvs)
3994 struct net *net = ipvs->net;
3996 ip_vs_trash_cleanup(ipvs);
3997 ip_vs_control_net_cleanup_sysctl(ipvs);
3998 remove_proc_entry("ip_vs_stats_percpu", net->proc_net);
3999 remove_proc_entry("ip_vs_stats", net->proc_net);
4000 remove_proc_entry("ip_vs", net->proc_net);
4001 free_percpu(ipvs->tot_stats.cpustats);
4004 int __init ip_vs_register_nl_ioctl(void)
4008 ret = nf_register_sockopt(&ip_vs_sockopts);
4010 pr_err("cannot register sockopt.\n");
4014 ret = ip_vs_genl_register();
4016 pr_err("cannot register Generic Netlink interface.\n");
4022 nf_unregister_sockopt(&ip_vs_sockopts);
4027 void ip_vs_unregister_nl_ioctl(void)
4029 ip_vs_genl_unregister();
4030 nf_unregister_sockopt(&ip_vs_sockopts);
4033 int __init ip_vs_control_init(void)
4040 /* Initialize svc_table, ip_vs_svc_fwm_table */
4041 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
4042 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
4043 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
4046 smp_wmb(); /* Do we really need it now ? */
4048 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
4057 void ip_vs_control_cleanup(void)
4060 unregister_netdevice_notifier(&ip_vs_dst_notifier);