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 /* Count only IPv4 services for old get/setsockopt interface */
1380 if (svc->af == AF_INET)
1381 ipvs->num_services--;
1383 ip_vs_stop_estimator(svc->ipvs, &svc->stats);
1385 /* Unbind scheduler */
1386 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1387 ip_vs_unbind_scheduler(svc, old_sched);
1388 ip_vs_scheduler_put(old_sched);
1390 /* Unbind persistence engine, keep svc->pe */
1391 old_pe = rcu_dereference_protected(svc->pe, 1);
1392 ip_vs_pe_put(old_pe);
1395 * Unlink the whole destination list
1397 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1398 __ip_vs_unlink_dest(svc, dest, 0);
1399 __ip_vs_del_dest(svc->ipvs, dest, cleanup);
1403 * Update the virtual service counters
1405 if (svc->port == FTPPORT)
1406 atomic_dec(&ipvs->ftpsvc_counter);
1407 else if (svc->port == 0)
1408 atomic_dec(&ipvs->nullsvc_counter);
1411 * Free the service if nobody refers to it
1413 __ip_vs_svc_put(svc, true);
1415 /* decrease the module use count */
1416 ip_vs_use_count_dec();
1420 * Unlink a service from list and try to delete it if its refcnt reached 0
1422 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1424 /* Hold svc to avoid double release from dest_trash */
1425 atomic_inc(&svc->refcnt);
1427 * Unhash it from the service table
1429 ip_vs_svc_unhash(svc);
1431 __ip_vs_del_service(svc, cleanup);
1435 * Delete a service from the service list
1437 static int ip_vs_del_service(struct ip_vs_service *svc)
1441 ip_vs_unlink_service(svc, false);
1448 * Flush all the virtual services
1450 static int ip_vs_flush(struct netns_ipvs *ipvs, bool cleanup)
1453 struct ip_vs_service *svc;
1454 struct hlist_node *n;
1457 * Flush the service table hashed by <netns,protocol,addr,port>
1459 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1460 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1462 if (svc->ipvs == ipvs)
1463 ip_vs_unlink_service(svc, cleanup);
1468 * Flush the service table hashed by fwmark
1470 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1471 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1473 if (svc->ipvs == ipvs)
1474 ip_vs_unlink_service(svc, cleanup);
1482 * Delete service by {netns} in the service table.
1483 * Called by __ip_vs_cleanup()
1485 void ip_vs_service_net_cleanup(struct netns_ipvs *ipvs)
1488 /* Check for "full" addressed entries */
1489 mutex_lock(&__ip_vs_mutex);
1490 ip_vs_flush(ipvs, true);
1491 mutex_unlock(&__ip_vs_mutex);
1495 /* Put all references for device (dst_cache) */
1497 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1499 struct ip_vs_dest_dst *dest_dst;
1501 spin_lock_bh(&dest->dst_lock);
1502 dest_dst = rcu_dereference_protected(dest->dest_dst, 1);
1503 if (dest_dst && dest_dst->dst_cache->dev == dev) {
1504 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1506 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1508 atomic_read(&dest->refcnt));
1509 __ip_vs_dst_cache_reset(dest);
1511 spin_unlock_bh(&dest->dst_lock);
1514 /* Netdev event receiver
1515 * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1517 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1520 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1521 struct net *net = dev_net(dev);
1522 struct netns_ipvs *ipvs = net_ipvs(net);
1523 struct ip_vs_service *svc;
1524 struct ip_vs_dest *dest;
1527 if (event != NETDEV_DOWN || !ipvs)
1529 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1531 mutex_lock(&__ip_vs_mutex);
1532 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1533 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1534 if (svc->ipvs == ipvs) {
1535 list_for_each_entry(dest, &svc->destinations,
1537 ip_vs_forget_dev(dest, dev);
1542 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1543 if (svc->ipvs == ipvs) {
1544 list_for_each_entry(dest, &svc->destinations,
1546 ip_vs_forget_dev(dest, dev);
1553 spin_lock_bh(&ipvs->dest_trash_lock);
1554 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1555 ip_vs_forget_dev(dest, dev);
1557 spin_unlock_bh(&ipvs->dest_trash_lock);
1558 mutex_unlock(&__ip_vs_mutex);
1564 * Zero counters in a service or all services
1566 static int ip_vs_zero_service(struct ip_vs_service *svc)
1568 struct ip_vs_dest *dest;
1570 list_for_each_entry(dest, &svc->destinations, n_list) {
1571 ip_vs_zero_stats(&dest->stats);
1573 ip_vs_zero_stats(&svc->stats);
1577 static int ip_vs_zero_all(struct netns_ipvs *ipvs)
1580 struct ip_vs_service *svc;
1582 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1583 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1584 if (svc->ipvs == ipvs)
1585 ip_vs_zero_service(svc);
1589 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1590 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1591 if (svc->ipvs == ipvs)
1592 ip_vs_zero_service(svc);
1596 ip_vs_zero_stats(&ipvs->tot_stats);
1600 #ifdef CONFIG_SYSCTL
1603 static int three = 3;
1606 proc_do_defense_mode(struct ctl_table *table, int write,
1607 void __user *buffer, size_t *lenp, loff_t *ppos)
1609 struct netns_ipvs *ipvs = table->extra2;
1610 int *valp = table->data;
1614 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1615 if (write && (*valp != val)) {
1616 if ((*valp < 0) || (*valp > 3)) {
1617 /* Restore the correct value */
1620 update_defense_level(ipvs);
1627 proc_do_sync_threshold(struct ctl_table *table, int write,
1628 void __user *buffer, size_t *lenp, loff_t *ppos)
1630 int *valp = table->data;
1634 /* backup the value first */
1635 memcpy(val, valp, sizeof(val));
1637 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1638 if (write && (valp[0] < 0 || valp[1] < 0 ||
1639 (valp[0] >= valp[1] && valp[1]))) {
1640 /* Restore the correct value */
1641 memcpy(valp, val, sizeof(val));
1647 proc_do_sync_mode(struct ctl_table *table, int write,
1648 void __user *buffer, size_t *lenp, loff_t *ppos)
1650 int *valp = table->data;
1654 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1655 if (write && (*valp != val)) {
1656 if ((*valp < 0) || (*valp > 1)) {
1657 /* Restore the correct value */
1665 proc_do_sync_ports(struct ctl_table *table, int write,
1666 void __user *buffer, size_t *lenp, loff_t *ppos)
1668 int *valp = table->data;
1672 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1673 if (write && (*valp != val)) {
1674 if (*valp < 1 || !is_power_of_2(*valp)) {
1675 /* Restore the correct value */
1683 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1684 * Do not change order or insert new entries without
1685 * align with netns init in ip_vs_control_net_init()
1688 static struct ctl_table vs_vars[] = {
1690 .procname = "amemthresh",
1691 .maxlen = sizeof(int),
1693 .proc_handler = proc_dointvec,
1696 .procname = "am_droprate",
1697 .maxlen = sizeof(int),
1699 .proc_handler = proc_dointvec,
1702 .procname = "drop_entry",
1703 .maxlen = sizeof(int),
1705 .proc_handler = proc_do_defense_mode,
1708 .procname = "drop_packet",
1709 .maxlen = sizeof(int),
1711 .proc_handler = proc_do_defense_mode,
1713 #ifdef CONFIG_IP_VS_NFCT
1715 .procname = "conntrack",
1716 .maxlen = sizeof(int),
1718 .proc_handler = &proc_dointvec,
1722 .procname = "secure_tcp",
1723 .maxlen = sizeof(int),
1725 .proc_handler = proc_do_defense_mode,
1728 .procname = "snat_reroute",
1729 .maxlen = sizeof(int),
1731 .proc_handler = &proc_dointvec,
1734 .procname = "sync_version",
1735 .maxlen = sizeof(int),
1737 .proc_handler = &proc_do_sync_mode,
1740 .procname = "sync_ports",
1741 .maxlen = sizeof(int),
1743 .proc_handler = &proc_do_sync_ports,
1746 .procname = "sync_persist_mode",
1747 .maxlen = sizeof(int),
1749 .proc_handler = proc_dointvec,
1752 .procname = "sync_qlen_max",
1753 .maxlen = sizeof(unsigned long),
1755 .proc_handler = proc_doulongvec_minmax,
1758 .procname = "sync_sock_size",
1759 .maxlen = sizeof(int),
1761 .proc_handler = proc_dointvec,
1764 .procname = "cache_bypass",
1765 .maxlen = sizeof(int),
1767 .proc_handler = proc_dointvec,
1770 .procname = "expire_nodest_conn",
1771 .maxlen = sizeof(int),
1773 .proc_handler = proc_dointvec,
1776 .procname = "sloppy_tcp",
1777 .maxlen = sizeof(int),
1779 .proc_handler = proc_dointvec,
1782 .procname = "sloppy_sctp",
1783 .maxlen = sizeof(int),
1785 .proc_handler = proc_dointvec,
1788 .procname = "expire_quiescent_template",
1789 .maxlen = sizeof(int),
1791 .proc_handler = proc_dointvec,
1794 .procname = "sync_threshold",
1796 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1798 .proc_handler = proc_do_sync_threshold,
1801 .procname = "sync_refresh_period",
1802 .maxlen = sizeof(int),
1804 .proc_handler = proc_dointvec_jiffies,
1807 .procname = "sync_retries",
1808 .maxlen = sizeof(int),
1810 .proc_handler = proc_dointvec_minmax,
1815 .procname = "nat_icmp_send",
1816 .maxlen = sizeof(int),
1818 .proc_handler = proc_dointvec,
1821 .procname = "pmtu_disc",
1822 .maxlen = sizeof(int),
1824 .proc_handler = proc_dointvec,
1827 .procname = "backup_only",
1828 .maxlen = sizeof(int),
1830 .proc_handler = proc_dointvec,
1833 .procname = "conn_reuse_mode",
1834 .maxlen = sizeof(int),
1836 .proc_handler = proc_dointvec,
1839 .procname = "schedule_icmp",
1840 .maxlen = sizeof(int),
1842 .proc_handler = proc_dointvec,
1845 .procname = "ignore_tunneled",
1846 .maxlen = sizeof(int),
1848 .proc_handler = proc_dointvec,
1850 #ifdef CONFIG_IP_VS_DEBUG
1852 .procname = "debug_level",
1853 .data = &sysctl_ip_vs_debug_level,
1854 .maxlen = sizeof(int),
1856 .proc_handler = proc_dointvec,
1864 #ifdef CONFIG_PROC_FS
1867 struct seq_net_private p; /* Do not move this, netns depends upon it*/
1868 struct hlist_head *table;
1873 * Write the contents of the VS rule table to a PROCfs file.
1874 * (It is kept just for backward compatibility)
1876 static inline const char *ip_vs_fwd_name(unsigned int flags)
1878 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1879 case IP_VS_CONN_F_LOCALNODE:
1881 case IP_VS_CONN_F_TUNNEL:
1883 case IP_VS_CONN_F_DROUTE:
1891 /* Get the Nth entry in the two lists */
1892 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1894 struct net *net = seq_file_net(seq);
1895 struct netns_ipvs *ipvs = net_ipvs(net);
1896 struct ip_vs_iter *iter = seq->private;
1898 struct ip_vs_service *svc;
1900 /* look in hash by protocol */
1901 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1902 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
1903 if ((svc->ipvs == ipvs) && pos-- == 0) {
1904 iter->table = ip_vs_svc_table;
1911 /* keep looking in fwmark */
1912 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1913 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
1915 if ((svc->ipvs == ipvs) && pos-- == 0) {
1916 iter->table = ip_vs_svc_fwm_table;
1926 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1930 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1934 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1936 struct hlist_node *e;
1937 struct ip_vs_iter *iter;
1938 struct ip_vs_service *svc;
1941 if (v == SEQ_START_TOKEN)
1942 return ip_vs_info_array(seq,0);
1945 iter = seq->private;
1947 if (iter->table == ip_vs_svc_table) {
1948 /* next service in table hashed by protocol */
1949 e = rcu_dereference(hlist_next_rcu(&svc->s_list));
1951 return hlist_entry(e, struct ip_vs_service, s_list);
1953 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1954 hlist_for_each_entry_rcu(svc,
1955 &ip_vs_svc_table[iter->bucket],
1961 iter->table = ip_vs_svc_fwm_table;
1966 /* next service in hashed by fwmark */
1967 e = rcu_dereference(hlist_next_rcu(&svc->f_list));
1969 return hlist_entry(e, struct ip_vs_service, f_list);
1972 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1973 hlist_for_each_entry_rcu(svc,
1974 &ip_vs_svc_fwm_table[iter->bucket],
1982 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
1989 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
1991 if (v == SEQ_START_TOKEN) {
1993 "IP Virtual Server version %d.%d.%d (size=%d)\n",
1994 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
1996 "Prot LocalAddress:Port Scheduler Flags\n");
1998 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2000 const struct ip_vs_service *svc = v;
2001 const struct ip_vs_iter *iter = seq->private;
2002 const struct ip_vs_dest *dest;
2003 struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler);
2004 char *sched_name = sched ? sched->name : "none";
2006 if (iter->table == ip_vs_svc_table) {
2007 #ifdef CONFIG_IP_VS_IPV6
2008 if (svc->af == AF_INET6)
2009 seq_printf(seq, "%s [%pI6]:%04X %s ",
2010 ip_vs_proto_name(svc->protocol),
2016 seq_printf(seq, "%s %08X:%04X %s %s ",
2017 ip_vs_proto_name(svc->protocol),
2018 ntohl(svc->addr.ip),
2021 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2023 seq_printf(seq, "FWM %08X %s %s",
2024 svc->fwmark, sched_name,
2025 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2028 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2029 seq_printf(seq, "persistent %d %08X\n",
2031 ntohl(svc->netmask));
2033 seq_putc(seq, '\n');
2035 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
2036 #ifdef CONFIG_IP_VS_IPV6
2037 if (dest->af == AF_INET6)
2040 " %-7s %-6d %-10d %-10d\n",
2043 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2044 atomic_read(&dest->weight),
2045 atomic_read(&dest->activeconns),
2046 atomic_read(&dest->inactconns));
2051 "%-7s %-6d %-10d %-10d\n",
2052 ntohl(dest->addr.ip),
2054 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2055 atomic_read(&dest->weight),
2056 atomic_read(&dest->activeconns),
2057 atomic_read(&dest->inactconns));
2064 static const struct seq_operations ip_vs_info_seq_ops = {
2065 .start = ip_vs_info_seq_start,
2066 .next = ip_vs_info_seq_next,
2067 .stop = ip_vs_info_seq_stop,
2068 .show = ip_vs_info_seq_show,
2071 static int ip_vs_info_open(struct inode *inode, struct file *file)
2073 return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2074 sizeof(struct ip_vs_iter));
2077 static const struct file_operations ip_vs_info_fops = {
2078 .owner = THIS_MODULE,
2079 .open = ip_vs_info_open,
2081 .llseek = seq_lseek,
2082 .release = seq_release_net,
2085 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2087 struct net *net = seq_file_single_net(seq);
2088 struct ip_vs_kstats show;
2090 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2092 " Total Incoming Outgoing Incoming Outgoing\n");
2094 " Conns Packets Packets Bytes Bytes\n");
2096 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2097 seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n\n",
2098 (unsigned long long)show.conns,
2099 (unsigned long long)show.inpkts,
2100 (unsigned long long)show.outpkts,
2101 (unsigned long long)show.inbytes,
2102 (unsigned long long)show.outbytes);
2104 /* 01234567 01234567 01234567 0123456701234567 0123456701234567*/
2106 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2107 seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n",
2108 (unsigned long long)show.cps,
2109 (unsigned long long)show.inpps,
2110 (unsigned long long)show.outpps,
2111 (unsigned long long)show.inbps,
2112 (unsigned long long)show.outbps);
2117 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2119 return single_open_net(inode, file, ip_vs_stats_show);
2122 static const struct file_operations ip_vs_stats_fops = {
2123 .owner = THIS_MODULE,
2124 .open = ip_vs_stats_seq_open,
2126 .llseek = seq_lseek,
2127 .release = single_release_net,
2130 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2132 struct net *net = seq_file_single_net(seq);
2133 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2134 struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats;
2135 struct ip_vs_kstats kstats;
2138 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2140 " Total Incoming Outgoing Incoming Outgoing\n");
2142 "CPU Conns Packets Packets Bytes Bytes\n");
2144 for_each_possible_cpu(i) {
2145 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2147 u64 conns, inpkts, outpkts, inbytes, outbytes;
2150 start = u64_stats_fetch_begin_irq(&u->syncp);
2151 conns = u->cnt.conns;
2152 inpkts = u->cnt.inpkts;
2153 outpkts = u->cnt.outpkts;
2154 inbytes = u->cnt.inbytes;
2155 outbytes = u->cnt.outbytes;
2156 } while (u64_stats_fetch_retry_irq(&u->syncp, start));
2158 seq_printf(seq, "%3X %8LX %8LX %8LX %16LX %16LX\n",
2159 i, (u64)conns, (u64)inpkts,
2160 (u64)outpkts, (u64)inbytes,
2164 ip_vs_copy_stats(&kstats, tot_stats);
2166 seq_printf(seq, " ~ %8LX %8LX %8LX %16LX %16LX\n\n",
2167 (unsigned long long)kstats.conns,
2168 (unsigned long long)kstats.inpkts,
2169 (unsigned long long)kstats.outpkts,
2170 (unsigned long long)kstats.inbytes,
2171 (unsigned long long)kstats.outbytes);
2173 /* ... 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2175 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2176 seq_printf(seq, " %8LX %8LX %8LX %16LX %16LX\n",
2186 static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2188 return single_open_net(inode, file, ip_vs_stats_percpu_show);
2191 static const struct file_operations ip_vs_stats_percpu_fops = {
2192 .owner = THIS_MODULE,
2193 .open = ip_vs_stats_percpu_seq_open,
2195 .llseek = seq_lseek,
2196 .release = single_release_net,
2201 * Set timeout values for tcp tcpfin udp in the timeout_table.
2203 static int ip_vs_set_timeout(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2205 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2206 struct ip_vs_proto_data *pd;
2209 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2214 #ifdef CONFIG_IP_VS_PROTO_TCP
2215 if (u->tcp_timeout) {
2216 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2217 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2218 = u->tcp_timeout * HZ;
2221 if (u->tcp_fin_timeout) {
2222 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2223 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2224 = u->tcp_fin_timeout * HZ;
2228 #ifdef CONFIG_IP_VS_PROTO_UDP
2229 if (u->udp_timeout) {
2230 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2231 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2232 = u->udp_timeout * HZ;
2238 #define CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2240 struct ip_vs_svcdest_user {
2241 struct ip_vs_service_user s;
2242 struct ip_vs_dest_user d;
2245 static const unsigned char set_arglen[CMDID(IP_VS_SO_SET_MAX) + 1] = {
2246 [CMDID(IP_VS_SO_SET_ADD)] = sizeof(struct ip_vs_service_user),
2247 [CMDID(IP_VS_SO_SET_EDIT)] = sizeof(struct ip_vs_service_user),
2248 [CMDID(IP_VS_SO_SET_DEL)] = sizeof(struct ip_vs_service_user),
2249 [CMDID(IP_VS_SO_SET_ADDDEST)] = sizeof(struct ip_vs_svcdest_user),
2250 [CMDID(IP_VS_SO_SET_DELDEST)] = sizeof(struct ip_vs_svcdest_user),
2251 [CMDID(IP_VS_SO_SET_EDITDEST)] = sizeof(struct ip_vs_svcdest_user),
2252 [CMDID(IP_VS_SO_SET_TIMEOUT)] = sizeof(struct ip_vs_timeout_user),
2253 [CMDID(IP_VS_SO_SET_STARTDAEMON)] = sizeof(struct ip_vs_daemon_user),
2254 [CMDID(IP_VS_SO_SET_STOPDAEMON)] = sizeof(struct ip_vs_daemon_user),
2255 [CMDID(IP_VS_SO_SET_ZERO)] = sizeof(struct ip_vs_service_user),
2258 union ip_vs_set_arglen {
2259 struct ip_vs_service_user field_IP_VS_SO_SET_ADD;
2260 struct ip_vs_service_user field_IP_VS_SO_SET_EDIT;
2261 struct ip_vs_service_user field_IP_VS_SO_SET_DEL;
2262 struct ip_vs_svcdest_user field_IP_VS_SO_SET_ADDDEST;
2263 struct ip_vs_svcdest_user field_IP_VS_SO_SET_DELDEST;
2264 struct ip_vs_svcdest_user field_IP_VS_SO_SET_EDITDEST;
2265 struct ip_vs_timeout_user field_IP_VS_SO_SET_TIMEOUT;
2266 struct ip_vs_daemon_user field_IP_VS_SO_SET_STARTDAEMON;
2267 struct ip_vs_daemon_user field_IP_VS_SO_SET_STOPDAEMON;
2268 struct ip_vs_service_user field_IP_VS_SO_SET_ZERO;
2271 #define MAX_SET_ARGLEN sizeof(union ip_vs_set_arglen)
2273 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2274 struct ip_vs_service_user *usvc_compat)
2276 memset(usvc, 0, sizeof(*usvc));
2279 usvc->protocol = usvc_compat->protocol;
2280 usvc->addr.ip = usvc_compat->addr;
2281 usvc->port = usvc_compat->port;
2282 usvc->fwmark = usvc_compat->fwmark;
2284 /* Deep copy of sched_name is not needed here */
2285 usvc->sched_name = usvc_compat->sched_name;
2287 usvc->flags = usvc_compat->flags;
2288 usvc->timeout = usvc_compat->timeout;
2289 usvc->netmask = usvc_compat->netmask;
2292 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2293 struct ip_vs_dest_user *udest_compat)
2295 memset(udest, 0, sizeof(*udest));
2297 udest->addr.ip = udest_compat->addr;
2298 udest->port = udest_compat->port;
2299 udest->conn_flags = udest_compat->conn_flags;
2300 udest->weight = udest_compat->weight;
2301 udest->u_threshold = udest_compat->u_threshold;
2302 udest->l_threshold = udest_compat->l_threshold;
2303 udest->af = AF_INET;
2307 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2309 struct net *net = sock_net(sk);
2311 unsigned char arg[MAX_SET_ARGLEN];
2312 struct ip_vs_service_user *usvc_compat;
2313 struct ip_vs_service_user_kern usvc;
2314 struct ip_vs_service *svc;
2315 struct ip_vs_dest_user *udest_compat;
2316 struct ip_vs_dest_user_kern udest;
2317 struct netns_ipvs *ipvs = net_ipvs(net);
2319 BUILD_BUG_ON(sizeof(arg) > 255);
2320 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2323 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2325 if (len != set_arglen[CMDID(cmd)]) {
2326 IP_VS_DBG(1, "set_ctl: len %u != %u\n",
2327 len, set_arglen[CMDID(cmd)]);
2331 if (copy_from_user(arg, user, len) != 0)
2334 /* increase the module use count */
2335 ip_vs_use_count_inc();
2337 /* Handle daemons since they have another lock */
2338 if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2339 cmd == IP_VS_SO_SET_STOPDAEMON) {
2340 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2342 if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2343 struct ipvs_sync_daemon_cfg cfg;
2345 memset(&cfg, 0, sizeof(cfg));
2346 strlcpy(cfg.mcast_ifn, dm->mcast_ifn,
2347 sizeof(cfg.mcast_ifn));
2348 cfg.syncid = dm->syncid;
2350 mutex_lock(&ipvs->sync_mutex);
2351 ret = start_sync_thread(ipvs, &cfg, dm->state);
2352 mutex_unlock(&ipvs->sync_mutex);
2355 mutex_lock(&ipvs->sync_mutex);
2356 ret = stop_sync_thread(ipvs, dm->state);
2357 mutex_unlock(&ipvs->sync_mutex);
2362 mutex_lock(&__ip_vs_mutex);
2363 if (cmd == IP_VS_SO_SET_FLUSH) {
2364 /* Flush the virtual service */
2365 ret = ip_vs_flush(ipvs, false);
2367 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2368 /* Set timeout values for (tcp tcpfin udp) */
2369 ret = ip_vs_set_timeout(ipvs, (struct ip_vs_timeout_user *)arg);
2373 usvc_compat = (struct ip_vs_service_user *)arg;
2374 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2376 /* We only use the new structs internally, so copy userspace compat
2377 * structs to extended internal versions */
2378 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2379 ip_vs_copy_udest_compat(&udest, udest_compat);
2381 if (cmd == IP_VS_SO_SET_ZERO) {
2382 /* if no service address is set, zero counters in all */
2383 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2384 ret = ip_vs_zero_all(ipvs);
2389 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2390 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2391 usvc.protocol != IPPROTO_SCTP) {
2392 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2393 usvc.protocol, &usvc.addr.ip,
2394 ntohs(usvc.port), usvc.sched_name);
2399 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2401 if (usvc.fwmark == 0)
2402 svc = __ip_vs_service_find(ipvs, usvc.af, usvc.protocol,
2403 &usvc.addr, usvc.port);
2405 svc = __ip_vs_svc_fwm_find(ipvs, usvc.af, usvc.fwmark);
2408 if (cmd != IP_VS_SO_SET_ADD
2409 && (svc == NULL || svc->protocol != usvc.protocol)) {
2415 case IP_VS_SO_SET_ADD:
2419 ret = ip_vs_add_service(ipvs, &usvc, &svc);
2421 case IP_VS_SO_SET_EDIT:
2422 ret = ip_vs_edit_service(svc, &usvc);
2424 case IP_VS_SO_SET_DEL:
2425 ret = ip_vs_del_service(svc);
2429 case IP_VS_SO_SET_ZERO:
2430 ret = ip_vs_zero_service(svc);
2432 case IP_VS_SO_SET_ADDDEST:
2433 ret = ip_vs_add_dest(svc, &udest);
2435 case IP_VS_SO_SET_EDITDEST:
2436 ret = ip_vs_edit_dest(svc, &udest);
2438 case IP_VS_SO_SET_DELDEST:
2439 ret = ip_vs_del_dest(svc, &udest);
2446 mutex_unlock(&__ip_vs_mutex);
2448 /* decrease the module use count */
2449 ip_vs_use_count_dec();
2456 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2458 struct ip_vs_scheduler *sched;
2459 struct ip_vs_kstats kstats;
2462 sched = rcu_dereference_protected(src->scheduler, 1);
2463 sched_name = sched ? sched->name : "none";
2464 dst->protocol = src->protocol;
2465 dst->addr = src->addr.ip;
2466 dst->port = src->port;
2467 dst->fwmark = src->fwmark;
2468 strlcpy(dst->sched_name, sched_name, sizeof(dst->sched_name));
2469 dst->flags = src->flags;
2470 dst->timeout = src->timeout / HZ;
2471 dst->netmask = src->netmask;
2472 dst->num_dests = src->num_dests;
2473 ip_vs_copy_stats(&kstats, &src->stats);
2474 ip_vs_export_stats_user(&dst->stats, &kstats);
2478 __ip_vs_get_service_entries(struct netns_ipvs *ipvs,
2479 const struct ip_vs_get_services *get,
2480 struct ip_vs_get_services __user *uptr)
2483 struct ip_vs_service *svc;
2484 struct ip_vs_service_entry entry;
2487 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2488 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2489 /* Only expose IPv4 entries to old interface */
2490 if (svc->af != AF_INET || (svc->ipvs != ipvs))
2493 if (count >= get->num_services)
2495 memset(&entry, 0, sizeof(entry));
2496 ip_vs_copy_service(&entry, svc);
2497 if (copy_to_user(&uptr->entrytable[count],
2498 &entry, sizeof(entry))) {
2506 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2507 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2508 /* Only expose IPv4 entries to old interface */
2509 if (svc->af != AF_INET || (svc->ipvs != ipvs))
2512 if (count >= get->num_services)
2514 memset(&entry, 0, sizeof(entry));
2515 ip_vs_copy_service(&entry, svc);
2516 if (copy_to_user(&uptr->entrytable[count],
2517 &entry, sizeof(entry))) {
2529 __ip_vs_get_dest_entries(struct netns_ipvs *ipvs, const struct ip_vs_get_dests *get,
2530 struct ip_vs_get_dests __user *uptr)
2532 struct ip_vs_service *svc;
2533 union nf_inet_addr addr = { .ip = get->addr };
2538 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, get->fwmark);
2540 svc = __ip_vs_service_find(ipvs, AF_INET, get->protocol, &addr,
2546 struct ip_vs_dest *dest;
2547 struct ip_vs_dest_entry entry;
2548 struct ip_vs_kstats kstats;
2550 memset(&entry, 0, sizeof(entry));
2551 list_for_each_entry(dest, &svc->destinations, n_list) {
2552 if (count >= get->num_dests)
2555 /* Cannot expose heterogeneous members via sockopt
2558 if (dest->af != svc->af)
2561 entry.addr = dest->addr.ip;
2562 entry.port = dest->port;
2563 entry.conn_flags = atomic_read(&dest->conn_flags);
2564 entry.weight = atomic_read(&dest->weight);
2565 entry.u_threshold = dest->u_threshold;
2566 entry.l_threshold = dest->l_threshold;
2567 entry.activeconns = atomic_read(&dest->activeconns);
2568 entry.inactconns = atomic_read(&dest->inactconns);
2569 entry.persistconns = atomic_read(&dest->persistconns);
2570 ip_vs_copy_stats(&kstats, &dest->stats);
2571 ip_vs_export_stats_user(&entry.stats, &kstats);
2572 if (copy_to_user(&uptr->entrytable[count],
2573 &entry, sizeof(entry))) {
2585 __ip_vs_get_timeouts(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2587 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2588 struct ip_vs_proto_data *pd;
2591 memset(u, 0, sizeof (*u));
2593 #ifdef CONFIG_IP_VS_PROTO_TCP
2594 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2595 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2596 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2598 #ifdef CONFIG_IP_VS_PROTO_UDP
2599 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2601 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2605 static const unsigned char get_arglen[CMDID(IP_VS_SO_GET_MAX) + 1] = {
2606 [CMDID(IP_VS_SO_GET_VERSION)] = 64,
2607 [CMDID(IP_VS_SO_GET_INFO)] = sizeof(struct ip_vs_getinfo),
2608 [CMDID(IP_VS_SO_GET_SERVICES)] = sizeof(struct ip_vs_get_services),
2609 [CMDID(IP_VS_SO_GET_SERVICE)] = sizeof(struct ip_vs_service_entry),
2610 [CMDID(IP_VS_SO_GET_DESTS)] = sizeof(struct ip_vs_get_dests),
2611 [CMDID(IP_VS_SO_GET_TIMEOUT)] = sizeof(struct ip_vs_timeout_user),
2612 [CMDID(IP_VS_SO_GET_DAEMON)] = 2 * sizeof(struct ip_vs_daemon_user),
2615 union ip_vs_get_arglen {
2616 char field_IP_VS_SO_GET_VERSION[64];
2617 struct ip_vs_getinfo field_IP_VS_SO_GET_INFO;
2618 struct ip_vs_get_services field_IP_VS_SO_GET_SERVICES;
2619 struct ip_vs_service_entry field_IP_VS_SO_GET_SERVICE;
2620 struct ip_vs_get_dests field_IP_VS_SO_GET_DESTS;
2621 struct ip_vs_timeout_user field_IP_VS_SO_GET_TIMEOUT;
2622 struct ip_vs_daemon_user field_IP_VS_SO_GET_DAEMON[2];
2625 #define MAX_GET_ARGLEN sizeof(union ip_vs_get_arglen)
2628 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2630 unsigned char arg[MAX_GET_ARGLEN];
2632 unsigned int copylen;
2633 struct net *net = sock_net(sk);
2634 struct netns_ipvs *ipvs = net_ipvs(net);
2637 BUILD_BUG_ON(sizeof(arg) > 255);
2638 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2641 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2644 copylen = get_arglen[CMDID(cmd)];
2645 if (*len < (int) copylen) {
2646 IP_VS_DBG(1, "get_ctl: len %d < %u\n", *len, copylen);
2650 if (copy_from_user(arg, user, copylen) != 0)
2653 * Handle daemons first since it has its own locking
2655 if (cmd == IP_VS_SO_GET_DAEMON) {
2656 struct ip_vs_daemon_user d[2];
2658 memset(&d, 0, sizeof(d));
2659 mutex_lock(&ipvs->sync_mutex);
2660 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2661 d[0].state = IP_VS_STATE_MASTER;
2662 strlcpy(d[0].mcast_ifn, ipvs->mcfg.mcast_ifn,
2663 sizeof(d[0].mcast_ifn));
2664 d[0].syncid = ipvs->mcfg.syncid;
2666 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2667 d[1].state = IP_VS_STATE_BACKUP;
2668 strlcpy(d[1].mcast_ifn, ipvs->bcfg.mcast_ifn,
2669 sizeof(d[1].mcast_ifn));
2670 d[1].syncid = ipvs->bcfg.syncid;
2672 if (copy_to_user(user, &d, sizeof(d)) != 0)
2674 mutex_unlock(&ipvs->sync_mutex);
2678 mutex_lock(&__ip_vs_mutex);
2680 case IP_VS_SO_GET_VERSION:
2684 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2685 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2686 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2690 *len = strlen(buf)+1;
2694 case IP_VS_SO_GET_INFO:
2696 struct ip_vs_getinfo info;
2697 info.version = IP_VS_VERSION_CODE;
2698 info.size = ip_vs_conn_tab_size;
2699 info.num_services = ipvs->num_services;
2700 if (copy_to_user(user, &info, sizeof(info)) != 0)
2705 case IP_VS_SO_GET_SERVICES:
2707 struct ip_vs_get_services *get;
2710 get = (struct ip_vs_get_services *)arg;
2711 size = sizeof(*get) +
2712 sizeof(struct ip_vs_service_entry) * get->num_services;
2714 pr_err("length: %u != %u\n", *len, size);
2718 ret = __ip_vs_get_service_entries(ipvs, get, user);
2722 case IP_VS_SO_GET_SERVICE:
2724 struct ip_vs_service_entry *entry;
2725 struct ip_vs_service *svc;
2726 union nf_inet_addr addr;
2728 entry = (struct ip_vs_service_entry *)arg;
2729 addr.ip = entry->addr;
2732 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, entry->fwmark);
2734 svc = __ip_vs_service_find(ipvs, AF_INET,
2735 entry->protocol, &addr,
2739 ip_vs_copy_service(entry, svc);
2740 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2747 case IP_VS_SO_GET_DESTS:
2749 struct ip_vs_get_dests *get;
2752 get = (struct ip_vs_get_dests *)arg;
2753 size = sizeof(*get) +
2754 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2756 pr_err("length: %u != %u\n", *len, size);
2760 ret = __ip_vs_get_dest_entries(ipvs, get, user);
2764 case IP_VS_SO_GET_TIMEOUT:
2766 struct ip_vs_timeout_user t;
2768 __ip_vs_get_timeouts(ipvs, &t);
2769 if (copy_to_user(user, &t, sizeof(t)) != 0)
2779 mutex_unlock(&__ip_vs_mutex);
2784 static struct nf_sockopt_ops ip_vs_sockopts = {
2786 .set_optmin = IP_VS_BASE_CTL,
2787 .set_optmax = IP_VS_SO_SET_MAX+1,
2788 .set = do_ip_vs_set_ctl,
2789 .get_optmin = IP_VS_BASE_CTL,
2790 .get_optmax = IP_VS_SO_GET_MAX+1,
2791 .get = do_ip_vs_get_ctl,
2792 .owner = THIS_MODULE,
2796 * Generic Netlink interface
2799 /* IPVS genetlink family */
2800 static struct genl_family ip_vs_genl_family = {
2801 .id = GENL_ID_GENERATE,
2803 .name = IPVS_GENL_NAME,
2804 .version = IPVS_GENL_VERSION,
2805 .maxattr = IPVS_CMD_MAX,
2806 .netnsok = true, /* Make ipvsadm to work on netns */
2809 /* Policy used for first-level command attributes */
2810 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2811 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2812 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2813 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2814 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2815 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2816 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2819 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2820 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2821 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2822 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2823 .len = IP_VS_IFNAME_MAXLEN },
2824 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2825 [IPVS_DAEMON_ATTR_SYNC_MAXLEN] = { .type = NLA_U16 },
2826 [IPVS_DAEMON_ATTR_MCAST_GROUP] = { .type = NLA_U32 },
2827 [IPVS_DAEMON_ATTR_MCAST_GROUP6] = { .len = sizeof(struct in6_addr) },
2828 [IPVS_DAEMON_ATTR_MCAST_PORT] = { .type = NLA_U16 },
2829 [IPVS_DAEMON_ATTR_MCAST_TTL] = { .type = NLA_U8 },
2832 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2833 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2834 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2835 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2836 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2837 .len = sizeof(union nf_inet_addr) },
2838 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2839 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2840 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2841 .len = IP_VS_SCHEDNAME_MAXLEN },
2842 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2843 .len = IP_VS_PENAME_MAXLEN },
2844 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2845 .len = sizeof(struct ip_vs_flags) },
2846 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2847 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2848 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2851 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2852 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2853 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2854 .len = sizeof(union nf_inet_addr) },
2855 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2856 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2857 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2858 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2859 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2860 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2861 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2862 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2863 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2864 [IPVS_DEST_ATTR_ADDR_FAMILY] = { .type = NLA_U16 },
2867 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2868 struct ip_vs_kstats *kstats)
2870 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2875 if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, (u32)kstats->conns) ||
2876 nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, (u32)kstats->inpkts) ||
2877 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, (u32)kstats->outpkts) ||
2878 nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes) ||
2879 nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes) ||
2880 nla_put_u32(skb, IPVS_STATS_ATTR_CPS, (u32)kstats->cps) ||
2881 nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, (u32)kstats->inpps) ||
2882 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, (u32)kstats->outpps) ||
2883 nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, (u32)kstats->inbps) ||
2884 nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, (u32)kstats->outbps))
2885 goto nla_put_failure;
2886 nla_nest_end(skb, nl_stats);
2891 nla_nest_cancel(skb, nl_stats);
2895 static int ip_vs_genl_fill_stats64(struct sk_buff *skb, int container_type,
2896 struct ip_vs_kstats *kstats)
2898 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2903 if (nla_put_u64(skb, IPVS_STATS_ATTR_CONNS, kstats->conns) ||
2904 nla_put_u64(skb, IPVS_STATS_ATTR_INPKTS, kstats->inpkts) ||
2905 nla_put_u64(skb, IPVS_STATS_ATTR_OUTPKTS, kstats->outpkts) ||
2906 nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes) ||
2907 nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes) ||
2908 nla_put_u64(skb, IPVS_STATS_ATTR_CPS, kstats->cps) ||
2909 nla_put_u64(skb, IPVS_STATS_ATTR_INPPS, kstats->inpps) ||
2910 nla_put_u64(skb, IPVS_STATS_ATTR_OUTPPS, kstats->outpps) ||
2911 nla_put_u64(skb, IPVS_STATS_ATTR_INBPS, kstats->inbps) ||
2912 nla_put_u64(skb, IPVS_STATS_ATTR_OUTBPS, kstats->outbps))
2913 goto nla_put_failure;
2914 nla_nest_end(skb, nl_stats);
2919 nla_nest_cancel(skb, nl_stats);
2923 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2924 struct ip_vs_service *svc)
2926 struct ip_vs_scheduler *sched;
2927 struct ip_vs_pe *pe;
2928 struct nlattr *nl_service;
2929 struct ip_vs_flags flags = { .flags = svc->flags,
2931 struct ip_vs_kstats kstats;
2934 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2938 if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2939 goto nla_put_failure;
2941 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2942 goto nla_put_failure;
2944 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2945 nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2946 nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2947 goto nla_put_failure;
2950 sched = rcu_dereference_protected(svc->scheduler, 1);
2951 sched_name = sched ? sched->name : "none";
2952 pe = rcu_dereference_protected(svc->pe, 1);
2953 if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched_name) ||
2954 (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
2955 nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2956 nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2957 nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
2958 goto nla_put_failure;
2959 ip_vs_copy_stats(&kstats, &svc->stats);
2960 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &kstats))
2961 goto nla_put_failure;
2962 if (ip_vs_genl_fill_stats64(skb, IPVS_SVC_ATTR_STATS64, &kstats))
2963 goto nla_put_failure;
2965 nla_nest_end(skb, nl_service);
2970 nla_nest_cancel(skb, nl_service);
2974 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2975 struct ip_vs_service *svc,
2976 struct netlink_callback *cb)
2980 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2981 &ip_vs_genl_family, NLM_F_MULTI,
2982 IPVS_CMD_NEW_SERVICE);
2986 if (ip_vs_genl_fill_service(skb, svc) < 0)
2987 goto nla_put_failure;
2989 genlmsg_end(skb, hdr);
2993 genlmsg_cancel(skb, hdr);
2997 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2998 struct netlink_callback *cb)
3001 int start = cb->args[0];
3002 struct ip_vs_service *svc;
3003 struct net *net = sock_net(skb->sk);
3004 struct netns_ipvs *ipvs = net_ipvs(net);
3006 mutex_lock(&__ip_vs_mutex);
3007 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3008 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
3009 if (++idx <= start || (svc->ipvs != ipvs))
3011 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3013 goto nla_put_failure;
3018 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3019 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
3020 if (++idx <= start || (svc->ipvs != ipvs))
3022 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3024 goto nla_put_failure;
3030 mutex_unlock(&__ip_vs_mutex);
3036 static int ip_vs_genl_parse_service(struct netns_ipvs *ipvs,
3037 struct ip_vs_service_user_kern *usvc,
3038 struct nlattr *nla, int full_entry,
3039 struct ip_vs_service **ret_svc)
3041 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
3042 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
3043 struct ip_vs_service *svc;
3045 /* Parse mandatory identifying service fields first */
3047 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
3050 nla_af = attrs[IPVS_SVC_ATTR_AF];
3051 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
3052 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
3053 nla_port = attrs[IPVS_SVC_ATTR_PORT];
3054 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
3056 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3059 memset(usvc, 0, sizeof(*usvc));
3061 usvc->af = nla_get_u16(nla_af);
3062 #ifdef CONFIG_IP_VS_IPV6
3063 if (usvc->af != AF_INET && usvc->af != AF_INET6)
3065 if (usvc->af != AF_INET)
3067 return -EAFNOSUPPORT;
3070 usvc->protocol = IPPROTO_TCP;
3071 usvc->fwmark = nla_get_u32(nla_fwmark);
3073 usvc->protocol = nla_get_u16(nla_protocol);
3074 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3075 usvc->port = nla_get_be16(nla_port);
3081 svc = __ip_vs_svc_fwm_find(ipvs, usvc->af, usvc->fwmark);
3083 svc = __ip_vs_service_find(ipvs, usvc->af, usvc->protocol,
3084 &usvc->addr, usvc->port);
3088 /* If a full entry was requested, check for the additional fields */
3090 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3092 struct ip_vs_flags flags;
3094 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3095 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3096 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3097 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3098 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3100 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3103 nla_memcpy(&flags, nla_flags, sizeof(flags));
3105 /* prefill flags from service if it already exists */
3107 usvc->flags = svc->flags;
3109 /* set new flags from userland */
3110 usvc->flags = (usvc->flags & ~flags.mask) |
3111 (flags.flags & flags.mask);
3112 usvc->sched_name = nla_data(nla_sched);
3113 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3114 usvc->timeout = nla_get_u32(nla_timeout);
3115 usvc->netmask = nla_get_be32(nla_netmask);
3121 static struct ip_vs_service *ip_vs_genl_find_service(struct netns_ipvs *ipvs,
3124 struct ip_vs_service_user_kern usvc;
3125 struct ip_vs_service *svc;
3128 ret = ip_vs_genl_parse_service(ipvs, &usvc, nla, 0, &svc);
3129 return ret ? ERR_PTR(ret) : svc;
3132 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3134 struct nlattr *nl_dest;
3135 struct ip_vs_kstats kstats;
3137 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3141 if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3142 nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3143 nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3144 (atomic_read(&dest->conn_flags) &
3145 IP_VS_CONN_F_FWD_MASK)) ||
3146 nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3147 atomic_read(&dest->weight)) ||
3148 nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3149 nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3150 nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3151 atomic_read(&dest->activeconns)) ||
3152 nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3153 atomic_read(&dest->inactconns)) ||
3154 nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3155 atomic_read(&dest->persistconns)) ||
3156 nla_put_u16(skb, IPVS_DEST_ATTR_ADDR_FAMILY, dest->af))
3157 goto nla_put_failure;
3158 ip_vs_copy_stats(&kstats, &dest->stats);
3159 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &kstats))
3160 goto nla_put_failure;
3161 if (ip_vs_genl_fill_stats64(skb, IPVS_DEST_ATTR_STATS64, &kstats))
3162 goto nla_put_failure;
3164 nla_nest_end(skb, nl_dest);
3169 nla_nest_cancel(skb, nl_dest);
3173 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3174 struct netlink_callback *cb)
3178 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3179 &ip_vs_genl_family, NLM_F_MULTI,
3184 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3185 goto nla_put_failure;
3187 genlmsg_end(skb, hdr);
3191 genlmsg_cancel(skb, hdr);
3195 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3196 struct netlink_callback *cb)
3199 int start = cb->args[0];
3200 struct ip_vs_service *svc;
3201 struct ip_vs_dest *dest;
3202 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3203 struct net *net = sock_net(skb->sk);
3204 struct netns_ipvs *ipvs = net_ipvs(net);
3206 mutex_lock(&__ip_vs_mutex);
3208 /* Try to find the service for which to dump destinations */
3209 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3210 IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3214 svc = ip_vs_genl_find_service(ipvs, attrs[IPVS_CMD_ATTR_SERVICE]);
3215 if (IS_ERR(svc) || svc == NULL)
3218 /* Dump the destinations */
3219 list_for_each_entry(dest, &svc->destinations, n_list) {
3222 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3224 goto nla_put_failure;
3232 mutex_unlock(&__ip_vs_mutex);
3237 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3238 struct nlattr *nla, int full_entry)
3240 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3241 struct nlattr *nla_addr, *nla_port;
3242 struct nlattr *nla_addr_family;
3244 /* Parse mandatory identifying destination fields first */
3246 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3249 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3250 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3251 nla_addr_family = attrs[IPVS_DEST_ATTR_ADDR_FAMILY];
3253 if (!(nla_addr && nla_port))
3256 memset(udest, 0, sizeof(*udest));
3258 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3259 udest->port = nla_get_be16(nla_port);
3261 if (nla_addr_family)
3262 udest->af = nla_get_u16(nla_addr_family);
3266 /* If a full entry was requested, check for the additional fields */
3268 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3271 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3272 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3273 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3274 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3276 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3279 udest->conn_flags = nla_get_u32(nla_fwd)
3280 & IP_VS_CONN_F_FWD_MASK;
3281 udest->weight = nla_get_u32(nla_weight);
3282 udest->u_threshold = nla_get_u32(nla_u_thresh);
3283 udest->l_threshold = nla_get_u32(nla_l_thresh);
3289 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3290 struct ipvs_sync_daemon_cfg *c)
3292 struct nlattr *nl_daemon;
3294 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3298 if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3299 nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, c->mcast_ifn) ||
3300 nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, c->syncid) ||
3301 nla_put_u16(skb, IPVS_DAEMON_ATTR_SYNC_MAXLEN, c->sync_maxlen) ||
3302 nla_put_u16(skb, IPVS_DAEMON_ATTR_MCAST_PORT, c->mcast_port) ||
3303 nla_put_u8(skb, IPVS_DAEMON_ATTR_MCAST_TTL, c->mcast_ttl))
3304 goto nla_put_failure;
3305 #ifdef CONFIG_IP_VS_IPV6
3306 if (c->mcast_af == AF_INET6) {
3307 if (nla_put_in6_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP6,
3308 &c->mcast_group.in6))
3309 goto nla_put_failure;
3312 if (c->mcast_af == AF_INET &&
3313 nla_put_in_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP,
3315 goto nla_put_failure;
3316 nla_nest_end(skb, nl_daemon);
3321 nla_nest_cancel(skb, nl_daemon);
3325 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3326 struct ipvs_sync_daemon_cfg *c,
3327 struct netlink_callback *cb)
3330 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3331 &ip_vs_genl_family, NLM_F_MULTI,
3332 IPVS_CMD_NEW_DAEMON);
3336 if (ip_vs_genl_fill_daemon(skb, state, c))
3337 goto nla_put_failure;
3339 genlmsg_end(skb, hdr);
3343 genlmsg_cancel(skb, hdr);
3347 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3348 struct netlink_callback *cb)
3350 struct net *net = sock_net(skb->sk);
3351 struct netns_ipvs *ipvs = net_ipvs(net);
3353 mutex_lock(&ipvs->sync_mutex);
3354 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3355 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3356 &ipvs->mcfg, cb) < 0)
3357 goto nla_put_failure;
3362 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3363 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3364 &ipvs->bcfg, cb) < 0)
3365 goto nla_put_failure;
3371 mutex_unlock(&ipvs->sync_mutex);
3376 static int ip_vs_genl_new_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3378 struct ipvs_sync_daemon_cfg c;
3382 memset(&c, 0, sizeof(c));
3383 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3384 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3385 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3387 strlcpy(c.mcast_ifn, nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3388 sizeof(c.mcast_ifn));
3389 c.syncid = nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]);
3391 a = attrs[IPVS_DAEMON_ATTR_SYNC_MAXLEN];
3393 c.sync_maxlen = nla_get_u16(a);
3395 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP];
3397 c.mcast_af = AF_INET;
3398 c.mcast_group.ip = nla_get_in_addr(a);
3399 if (!ipv4_is_multicast(c.mcast_group.ip))
3402 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP6];
3404 #ifdef CONFIG_IP_VS_IPV6
3407 c.mcast_af = AF_INET6;
3408 c.mcast_group.in6 = nla_get_in6_addr(a);
3409 addr_type = ipv6_addr_type(&c.mcast_group.in6);
3410 if (!(addr_type & IPV6_ADDR_MULTICAST))
3413 return -EAFNOSUPPORT;
3418 a = attrs[IPVS_DAEMON_ATTR_MCAST_PORT];
3420 c.mcast_port = nla_get_u16(a);
3422 a = attrs[IPVS_DAEMON_ATTR_MCAST_TTL];
3424 c.mcast_ttl = nla_get_u8(a);
3426 /* The synchronization protocol is incompatible with mixed family
3429 if (ipvs->mixed_address_family_dests > 0)
3433 mutex_lock(&ipvs->sync_mutex);
3434 ret = start_sync_thread(ipvs, &c,
3435 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3436 mutex_unlock(&ipvs->sync_mutex);
3441 static int ip_vs_genl_del_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3445 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3448 mutex_lock(&ipvs->sync_mutex);
3449 ret = stop_sync_thread(ipvs,
3450 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3451 mutex_unlock(&ipvs->sync_mutex);
3455 static int ip_vs_genl_set_config(struct netns_ipvs *ipvs, struct nlattr **attrs)
3457 struct ip_vs_timeout_user t;
3459 __ip_vs_get_timeouts(ipvs, &t);
3461 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3462 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3464 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3466 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3468 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3469 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3471 return ip_vs_set_timeout(ipvs, &t);
3474 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3476 int ret = -EINVAL, cmd;
3477 struct net *net = sock_net(skb->sk);
3478 struct netns_ipvs *ipvs = net_ipvs(net);
3480 cmd = info->genlhdr->cmd;
3482 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3483 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3485 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3486 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3487 info->attrs[IPVS_CMD_ATTR_DAEMON],
3488 ip_vs_daemon_policy))
3491 if (cmd == IPVS_CMD_NEW_DAEMON)
3492 ret = ip_vs_genl_new_daemon(ipvs, daemon_attrs);
3494 ret = ip_vs_genl_del_daemon(ipvs, daemon_attrs);
3501 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3503 struct ip_vs_service *svc = NULL;
3504 struct ip_vs_service_user_kern usvc;
3505 struct ip_vs_dest_user_kern udest;
3507 int need_full_svc = 0, need_full_dest = 0;
3508 struct net *net = sock_net(skb->sk);
3509 struct netns_ipvs *ipvs = net_ipvs(net);
3511 cmd = info->genlhdr->cmd;
3513 mutex_lock(&__ip_vs_mutex);
3515 if (cmd == IPVS_CMD_FLUSH) {
3516 ret = ip_vs_flush(ipvs, false);
3518 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3519 ret = ip_vs_genl_set_config(ipvs, info->attrs);
3521 } else if (cmd == IPVS_CMD_ZERO &&
3522 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3523 ret = ip_vs_zero_all(ipvs);
3527 /* All following commands require a service argument, so check if we
3528 * received a valid one. We need a full service specification when
3529 * adding / editing a service. Only identifying members otherwise. */
3530 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3533 ret = ip_vs_genl_parse_service(ipvs, &usvc,
3534 info->attrs[IPVS_CMD_ATTR_SERVICE],
3535 need_full_svc, &svc);
3539 /* Unless we're adding a new service, the service must already exist */
3540 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3545 /* Destination commands require a valid destination argument. For
3546 * adding / editing a destination, we need a full destination
3548 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3549 cmd == IPVS_CMD_DEL_DEST) {
3550 if (cmd != IPVS_CMD_DEL_DEST)
3553 ret = ip_vs_genl_parse_dest(&udest,
3554 info->attrs[IPVS_CMD_ATTR_DEST],
3559 /* Old protocols did not allow the user to specify address
3560 * family, so we set it to zero instead. We also didn't
3561 * allow heterogeneous pools in the old code, so it's safe
3562 * to assume that this will have the same address family as
3568 if (udest.af != svc->af && cmd != IPVS_CMD_DEL_DEST) {
3569 /* The synchronization protocol is incompatible
3570 * with mixed family services
3572 if (ipvs->sync_state) {
3577 /* Which connection types do we support? */
3578 switch (udest.conn_flags) {
3579 case IP_VS_CONN_F_TUNNEL:
3580 /* We are able to forward this */
3590 case IPVS_CMD_NEW_SERVICE:
3592 ret = ip_vs_add_service(ipvs, &usvc, &svc);
3596 case IPVS_CMD_SET_SERVICE:
3597 ret = ip_vs_edit_service(svc, &usvc);
3599 case IPVS_CMD_DEL_SERVICE:
3600 ret = ip_vs_del_service(svc);
3601 /* do not use svc, it can be freed */
3603 case IPVS_CMD_NEW_DEST:
3604 ret = ip_vs_add_dest(svc, &udest);
3606 case IPVS_CMD_SET_DEST:
3607 ret = ip_vs_edit_dest(svc, &udest);
3609 case IPVS_CMD_DEL_DEST:
3610 ret = ip_vs_del_dest(svc, &udest);
3613 ret = ip_vs_zero_service(svc);
3620 mutex_unlock(&__ip_vs_mutex);
3625 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3627 struct sk_buff *msg;
3629 int ret, cmd, reply_cmd;
3630 struct net *net = sock_net(skb->sk);
3631 struct netns_ipvs *ipvs = net_ipvs(net);
3633 cmd = info->genlhdr->cmd;
3635 if (cmd == IPVS_CMD_GET_SERVICE)
3636 reply_cmd = IPVS_CMD_NEW_SERVICE;
3637 else if (cmd == IPVS_CMD_GET_INFO)
3638 reply_cmd = IPVS_CMD_SET_INFO;
3639 else if (cmd == IPVS_CMD_GET_CONFIG)
3640 reply_cmd = IPVS_CMD_SET_CONFIG;
3642 pr_err("unknown Generic Netlink command\n");
3646 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3650 mutex_lock(&__ip_vs_mutex);
3652 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3654 goto nla_put_failure;
3657 case IPVS_CMD_GET_SERVICE:
3659 struct ip_vs_service *svc;
3661 svc = ip_vs_genl_find_service(ipvs,
3662 info->attrs[IPVS_CMD_ATTR_SERVICE]);
3667 ret = ip_vs_genl_fill_service(msg, svc);
3669 goto nla_put_failure;
3678 case IPVS_CMD_GET_CONFIG:
3680 struct ip_vs_timeout_user t;
3682 __ip_vs_get_timeouts(ipvs, &t);
3683 #ifdef CONFIG_IP_VS_PROTO_TCP
3684 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3686 nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3688 goto nla_put_failure;
3690 #ifdef CONFIG_IP_VS_PROTO_UDP
3691 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3692 goto nla_put_failure;
3698 case IPVS_CMD_GET_INFO:
3699 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3700 IP_VS_VERSION_CODE) ||
3701 nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3702 ip_vs_conn_tab_size))
3703 goto nla_put_failure;
3707 genlmsg_end(msg, reply);
3708 ret = genlmsg_reply(msg, info);
3712 pr_err("not enough space in Netlink message\n");
3718 mutex_unlock(&__ip_vs_mutex);
3724 static const struct genl_ops ip_vs_genl_ops[] = {
3726 .cmd = IPVS_CMD_NEW_SERVICE,
3727 .flags = GENL_ADMIN_PERM,
3728 .policy = ip_vs_cmd_policy,
3729 .doit = ip_vs_genl_set_cmd,
3732 .cmd = IPVS_CMD_SET_SERVICE,
3733 .flags = GENL_ADMIN_PERM,
3734 .policy = ip_vs_cmd_policy,
3735 .doit = ip_vs_genl_set_cmd,
3738 .cmd = IPVS_CMD_DEL_SERVICE,
3739 .flags = GENL_ADMIN_PERM,
3740 .policy = ip_vs_cmd_policy,
3741 .doit = ip_vs_genl_set_cmd,
3744 .cmd = IPVS_CMD_GET_SERVICE,
3745 .flags = GENL_ADMIN_PERM,
3746 .doit = ip_vs_genl_get_cmd,
3747 .dumpit = ip_vs_genl_dump_services,
3748 .policy = ip_vs_cmd_policy,
3751 .cmd = IPVS_CMD_NEW_DEST,
3752 .flags = GENL_ADMIN_PERM,
3753 .policy = ip_vs_cmd_policy,
3754 .doit = ip_vs_genl_set_cmd,
3757 .cmd = IPVS_CMD_SET_DEST,
3758 .flags = GENL_ADMIN_PERM,
3759 .policy = ip_vs_cmd_policy,
3760 .doit = ip_vs_genl_set_cmd,
3763 .cmd = IPVS_CMD_DEL_DEST,
3764 .flags = GENL_ADMIN_PERM,
3765 .policy = ip_vs_cmd_policy,
3766 .doit = ip_vs_genl_set_cmd,
3769 .cmd = IPVS_CMD_GET_DEST,
3770 .flags = GENL_ADMIN_PERM,
3771 .policy = ip_vs_cmd_policy,
3772 .dumpit = ip_vs_genl_dump_dests,
3775 .cmd = IPVS_CMD_NEW_DAEMON,
3776 .flags = GENL_ADMIN_PERM,
3777 .policy = ip_vs_cmd_policy,
3778 .doit = ip_vs_genl_set_daemon,
3781 .cmd = IPVS_CMD_DEL_DAEMON,
3782 .flags = GENL_ADMIN_PERM,
3783 .policy = ip_vs_cmd_policy,
3784 .doit = ip_vs_genl_set_daemon,
3787 .cmd = IPVS_CMD_GET_DAEMON,
3788 .flags = GENL_ADMIN_PERM,
3789 .dumpit = ip_vs_genl_dump_daemons,
3792 .cmd = IPVS_CMD_SET_CONFIG,
3793 .flags = GENL_ADMIN_PERM,
3794 .policy = ip_vs_cmd_policy,
3795 .doit = ip_vs_genl_set_cmd,
3798 .cmd = IPVS_CMD_GET_CONFIG,
3799 .flags = GENL_ADMIN_PERM,
3800 .doit = ip_vs_genl_get_cmd,
3803 .cmd = IPVS_CMD_GET_INFO,
3804 .flags = GENL_ADMIN_PERM,
3805 .doit = ip_vs_genl_get_cmd,
3808 .cmd = IPVS_CMD_ZERO,
3809 .flags = GENL_ADMIN_PERM,
3810 .policy = ip_vs_cmd_policy,
3811 .doit = ip_vs_genl_set_cmd,
3814 .cmd = IPVS_CMD_FLUSH,
3815 .flags = GENL_ADMIN_PERM,
3816 .doit = ip_vs_genl_set_cmd,
3820 static int __init ip_vs_genl_register(void)
3822 return genl_register_family_with_ops(&ip_vs_genl_family,
3826 static void ip_vs_genl_unregister(void)
3828 genl_unregister_family(&ip_vs_genl_family);
3831 /* End of Generic Netlink interface definitions */
3834 * per netns intit/exit func.
3836 #ifdef CONFIG_SYSCTL
3837 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs)
3839 struct net *net = ipvs->net;
3841 struct ctl_table *tbl;
3843 atomic_set(&ipvs->dropentry, 0);
3844 spin_lock_init(&ipvs->dropentry_lock);
3845 spin_lock_init(&ipvs->droppacket_lock);
3846 spin_lock_init(&ipvs->securetcp_lock);
3848 if (!net_eq(net, &init_net)) {
3849 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3853 /* Don't export sysctls to unprivileged users */
3854 if (net->user_ns != &init_user_ns)
3855 tbl[0].procname = NULL;
3858 /* Initialize sysctl defaults */
3859 for (idx = 0; idx < ARRAY_SIZE(vs_vars); idx++) {
3860 if (tbl[idx].proc_handler == proc_do_defense_mode)
3861 tbl[idx].extra2 = ipvs;
3864 ipvs->sysctl_amemthresh = 1024;
3865 tbl[idx++].data = &ipvs->sysctl_amemthresh;
3866 ipvs->sysctl_am_droprate = 10;
3867 tbl[idx++].data = &ipvs->sysctl_am_droprate;
3868 tbl[idx++].data = &ipvs->sysctl_drop_entry;
3869 tbl[idx++].data = &ipvs->sysctl_drop_packet;
3870 #ifdef CONFIG_IP_VS_NFCT
3871 tbl[idx++].data = &ipvs->sysctl_conntrack;
3873 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3874 ipvs->sysctl_snat_reroute = 1;
3875 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3876 ipvs->sysctl_sync_ver = 1;
3877 tbl[idx++].data = &ipvs->sysctl_sync_ver;
3878 ipvs->sysctl_sync_ports = 1;
3879 tbl[idx++].data = &ipvs->sysctl_sync_ports;
3880 tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
3881 ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3882 tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3883 ipvs->sysctl_sync_sock_size = 0;
3884 tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3885 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3886 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3887 tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
3888 tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
3889 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3890 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3891 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3892 tbl[idx].data = &ipvs->sysctl_sync_threshold;
3893 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3894 ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3895 tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3896 ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3897 tbl[idx++].data = &ipvs->sysctl_sync_retries;
3898 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3899 ipvs->sysctl_pmtu_disc = 1;
3900 tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
3901 tbl[idx++].data = &ipvs->sysctl_backup_only;
3902 ipvs->sysctl_conn_reuse_mode = 1;
3903 tbl[idx++].data = &ipvs->sysctl_conn_reuse_mode;
3904 tbl[idx++].data = &ipvs->sysctl_schedule_icmp;
3905 tbl[idx++].data = &ipvs->sysctl_ignore_tunneled;
3907 ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3908 if (ipvs->sysctl_hdr == NULL) {
3909 if (!net_eq(net, &init_net))
3913 ip_vs_start_estimator(ipvs, &ipvs->tot_stats);
3914 ipvs->sysctl_tbl = tbl;
3915 /* Schedule defense work */
3916 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3917 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3922 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs)
3924 struct net *net = ipvs->net;
3926 cancel_delayed_work_sync(&ipvs->defense_work);
3927 cancel_work_sync(&ipvs->defense_work.work);
3928 unregister_net_sysctl_table(ipvs->sysctl_hdr);
3929 ip_vs_stop_estimator(ipvs, &ipvs->tot_stats);
3931 if (!net_eq(net, &init_net))
3932 kfree(ipvs->sysctl_tbl);
3937 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs) { return 0; }
3938 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs) { }
3942 static struct notifier_block ip_vs_dst_notifier = {
3943 .notifier_call = ip_vs_dst_event,
3946 int __net_init ip_vs_control_net_init(struct netns_ipvs *ipvs)
3950 /* Initialize rs_table */
3951 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
3952 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
3954 INIT_LIST_HEAD(&ipvs->dest_trash);
3955 spin_lock_init(&ipvs->dest_trash_lock);
3956 setup_timer(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire,
3957 (unsigned long) ipvs);
3958 atomic_set(&ipvs->ftpsvc_counter, 0);
3959 atomic_set(&ipvs->nullsvc_counter, 0);
3962 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
3963 if (!ipvs->tot_stats.cpustats)
3966 for_each_possible_cpu(i) {
3967 struct ip_vs_cpu_stats *ipvs_tot_stats;
3968 ipvs_tot_stats = per_cpu_ptr(ipvs->tot_stats.cpustats, i);
3969 u64_stats_init(&ipvs_tot_stats->syncp);
3972 spin_lock_init(&ipvs->tot_stats.lock);
3974 proc_create("ip_vs", 0, ipvs->net->proc_net, &ip_vs_info_fops);
3975 proc_create("ip_vs_stats", 0, ipvs->net->proc_net, &ip_vs_stats_fops);
3976 proc_create("ip_vs_stats_percpu", 0, ipvs->net->proc_net,
3977 &ip_vs_stats_percpu_fops);
3979 if (ip_vs_control_net_init_sysctl(ipvs))
3985 free_percpu(ipvs->tot_stats.cpustats);
3989 void __net_exit ip_vs_control_net_cleanup(struct netns_ipvs *ipvs)
3991 ip_vs_trash_cleanup(ipvs);
3992 ip_vs_control_net_cleanup_sysctl(ipvs);
3993 remove_proc_entry("ip_vs_stats_percpu", ipvs->net->proc_net);
3994 remove_proc_entry("ip_vs_stats", ipvs->net->proc_net);
3995 remove_proc_entry("ip_vs", ipvs->net->proc_net);
3996 free_percpu(ipvs->tot_stats.cpustats);
3999 int __init ip_vs_register_nl_ioctl(void)
4003 ret = nf_register_sockopt(&ip_vs_sockopts);
4005 pr_err("cannot register sockopt.\n");
4009 ret = ip_vs_genl_register();
4011 pr_err("cannot register Generic Netlink interface.\n");
4017 nf_unregister_sockopt(&ip_vs_sockopts);
4022 void ip_vs_unregister_nl_ioctl(void)
4024 ip_vs_genl_unregister();
4025 nf_unregister_sockopt(&ip_vs_sockopts);
4028 int __init ip_vs_control_init(void)
4035 /* Initialize svc_table, ip_vs_svc_fwm_table */
4036 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
4037 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
4038 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
4041 smp_wmb(); /* Do we really need it now ? */
4043 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
4052 void ip_vs_control_cleanup(void)
4055 unregister_netdevice_notifier(&ip_vs_dst_notifier);