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 <linux/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 %pS\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 %pS\n",
338 __func__, __builtin_return_address(0));
342 if (svc->fwmark == 0) {
343 /* Remove it from the svc_table table */
344 hlist_del_rcu(&svc->s_list);
346 /* Remove it from the svc_fwm_table table */
347 hlist_del_rcu(&svc->f_list);
350 svc->flags &= ~IP_VS_SVC_F_HASHED;
351 atomic_dec(&svc->refcnt);
357 * Get service by {netns, proto,addr,port} in the service table.
359 static inline struct ip_vs_service *
360 __ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u16 protocol,
361 const union nf_inet_addr *vaddr, __be16 vport)
364 struct ip_vs_service *svc;
366 /* Check for "full" addressed entries */
367 hash = ip_vs_svc_hashkey(ipvs, af, protocol, vaddr, vport);
369 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[hash], s_list) {
371 && ip_vs_addr_equal(af, &svc->addr, vaddr)
372 && (svc->port == vport)
373 && (svc->protocol == protocol)
374 && (svc->ipvs == ipvs)) {
385 * Get service by {fwmark} in the service table.
387 static inline struct ip_vs_service *
388 __ip_vs_svc_fwm_find(struct netns_ipvs *ipvs, int af, __u32 fwmark)
391 struct ip_vs_service *svc;
393 /* Check for fwmark addressed entries */
394 hash = ip_vs_svc_fwm_hashkey(ipvs, fwmark);
396 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[hash], f_list) {
397 if (svc->fwmark == fwmark && svc->af == af
398 && (svc->ipvs == ipvs)) {
407 /* Find service, called under RCU lock */
408 struct ip_vs_service *
409 ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
410 const union nf_inet_addr *vaddr, __be16 vport)
412 struct ip_vs_service *svc;
415 * Check the table hashed by fwmark first
418 svc = __ip_vs_svc_fwm_find(ipvs, af, fwmark);
424 * Check the table hashed by <protocol,addr,port>
425 * for "full" addressed entries
427 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, vport);
429 if (!svc && protocol == IPPROTO_TCP &&
430 atomic_read(&ipvs->ftpsvc_counter) &&
431 (vport == FTPDATA || ntohs(vport) >= inet_prot_sock(ipvs->net))) {
433 * Check if ftp service entry exists, the packet
434 * might belong to FTP data connections.
436 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, FTPPORT);
440 && atomic_read(&ipvs->nullsvc_counter)) {
442 * Check if the catch-all port (port zero) exists
444 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, 0);
448 IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
449 fwmark, ip_vs_proto_name(protocol),
450 IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
451 svc ? "hit" : "not hit");
458 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
460 atomic_inc(&svc->refcnt);
461 rcu_assign_pointer(dest->svc, svc);
464 static void ip_vs_service_free(struct ip_vs_service *svc)
466 free_percpu(svc->stats.cpustats);
470 static void ip_vs_service_rcu_free(struct rcu_head *head)
472 struct ip_vs_service *svc;
474 svc = container_of(head, struct ip_vs_service, rcu_head);
475 ip_vs_service_free(svc);
478 static void __ip_vs_svc_put(struct ip_vs_service *svc, bool do_delay)
480 if (atomic_dec_and_test(&svc->refcnt)) {
481 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n",
483 IP_VS_DBG_ADDR(svc->af, &svc->addr),
486 call_rcu(&svc->rcu_head, ip_vs_service_rcu_free);
488 ip_vs_service_free(svc);
494 * Returns hash value for real service
496 static inline unsigned int ip_vs_rs_hashkey(int af,
497 const union nf_inet_addr *addr,
500 register unsigned int porth = ntohs(port);
501 __be32 addr_fold = addr->ip;
503 #ifdef CONFIG_IP_VS_IPV6
505 addr_fold = addr->ip6[0]^addr->ip6[1]^
506 addr->ip6[2]^addr->ip6[3];
509 return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
513 /* Hash ip_vs_dest in rs_table by <proto,addr,port>. */
514 static void ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
518 if (dest->in_rs_table)
522 * Hash by proto,addr,port,
523 * which are the parameters of the real service.
525 hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
527 hlist_add_head_rcu(&dest->d_list, &ipvs->rs_table[hash]);
528 dest->in_rs_table = 1;
531 /* Unhash ip_vs_dest from rs_table. */
532 static void ip_vs_rs_unhash(struct ip_vs_dest *dest)
535 * Remove it from the rs_table table.
537 if (dest->in_rs_table) {
538 hlist_del_rcu(&dest->d_list);
539 dest->in_rs_table = 0;
543 /* Check if real service by <proto,addr,port> is present */
544 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
545 const union nf_inet_addr *daddr, __be16 dport)
548 struct ip_vs_dest *dest;
550 /* Check for "full" addressed entries */
551 hash = ip_vs_rs_hashkey(af, daddr, dport);
553 hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
554 if (dest->port == dport &&
556 ip_vs_addr_equal(af, &dest->addr, daddr) &&
557 (dest->protocol == protocol || dest->vfwmark)) {
566 /* Find real service record by <proto,addr,port>.
567 * In case of multiple records with the same <proto,addr,port>, only
568 * the first found record is returned.
570 * To be called under RCU lock.
572 struct ip_vs_dest *ip_vs_find_real_service(struct netns_ipvs *ipvs, int af,
574 const union nf_inet_addr *daddr,
578 struct ip_vs_dest *dest;
580 /* Check for "full" addressed entries */
581 hash = ip_vs_rs_hashkey(af, daddr, dport);
583 hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
584 if (dest->port == dport &&
586 ip_vs_addr_equal(af, &dest->addr, daddr) &&
587 (dest->protocol == protocol || dest->vfwmark)) {
596 /* Lookup destination by {addr,port} in the given service
597 * Called under RCU lock.
599 static struct ip_vs_dest *
600 ip_vs_lookup_dest(struct ip_vs_service *svc, int dest_af,
601 const union nf_inet_addr *daddr, __be16 dport)
603 struct ip_vs_dest *dest;
606 * Find the destination for the given service
608 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
609 if ((dest->af == dest_af) &&
610 ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
611 (dest->port == dport)) {
621 * Find destination by {daddr,dport,vaddr,protocol}
622 * Created to be used in ip_vs_process_message() in
623 * the backup synchronization daemon. It finds the
624 * destination to be bound to the received connection
626 * Called under RCU lock, no refcnt is returned.
628 struct ip_vs_dest *ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
629 const union nf_inet_addr *daddr,
631 const union nf_inet_addr *vaddr,
632 __be16 vport, __u16 protocol, __u32 fwmark,
635 struct ip_vs_dest *dest;
636 struct ip_vs_service *svc;
639 svc = ip_vs_service_find(ipvs, svc_af, fwmark, protocol, vaddr, vport);
642 if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
644 dest = ip_vs_lookup_dest(svc, dest_af, daddr, port);
646 dest = ip_vs_lookup_dest(svc, dest_af, daddr, port ^ dport);
650 void ip_vs_dest_dst_rcu_free(struct rcu_head *head)
652 struct ip_vs_dest_dst *dest_dst = container_of(head,
653 struct ip_vs_dest_dst,
656 dst_release(dest_dst->dst_cache);
660 /* Release dest_dst and dst_cache for dest in user context */
661 static void __ip_vs_dst_cache_reset(struct ip_vs_dest *dest)
663 struct ip_vs_dest_dst *old;
665 old = rcu_dereference_protected(dest->dest_dst, 1);
667 RCU_INIT_POINTER(dest->dest_dst, NULL);
668 call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
673 * Lookup dest by {svc,addr,port} in the destination trash.
674 * The destination trash is used to hold the destinations that are removed
675 * from the service table but are still referenced by some conn entries.
676 * The reason to add the destination trash is when the dest is temporary
677 * down (either by administrator or by monitor program), the dest can be
678 * picked back from the trash, the remaining connections to the dest can
679 * continue, and the counting information of the dest is also useful for
682 static struct ip_vs_dest *
683 ip_vs_trash_get_dest(struct ip_vs_service *svc, int dest_af,
684 const union nf_inet_addr *daddr, __be16 dport)
686 struct ip_vs_dest *dest;
687 struct netns_ipvs *ipvs = svc->ipvs;
690 * Find the destination in trash
692 spin_lock_bh(&ipvs->dest_trash_lock);
693 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
694 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
697 IP_VS_DBG_ADDR(dest->af, &dest->addr),
699 refcount_read(&dest->refcnt));
700 if (dest->af == dest_af &&
701 ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
702 dest->port == dport &&
703 dest->vfwmark == svc->fwmark &&
704 dest->protocol == svc->protocol &&
706 (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
707 dest->vport == svc->port))) {
709 list_del(&dest->t_list);
717 spin_unlock_bh(&ipvs->dest_trash_lock);
722 static void ip_vs_dest_free(struct ip_vs_dest *dest)
724 struct ip_vs_service *svc = rcu_dereference_protected(dest->svc, 1);
726 __ip_vs_dst_cache_reset(dest);
727 __ip_vs_svc_put(svc, false);
728 free_percpu(dest->stats.cpustats);
729 ip_vs_dest_put_and_free(dest);
733 * Clean up all the destinations in the trash
734 * Called by the ip_vs_control_cleanup()
736 * When the ip_vs_control_clearup is activated by ipvs module exit,
737 * the service tables must have been flushed and all the connections
738 * are expired, and the refcnt of each destination in the trash must
739 * be 1, so we simply release them here.
741 static void ip_vs_trash_cleanup(struct netns_ipvs *ipvs)
743 struct ip_vs_dest *dest, *nxt;
745 del_timer_sync(&ipvs->dest_trash_timer);
746 /* No need to use dest_trash_lock */
747 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, t_list) {
748 list_del(&dest->t_list);
749 ip_vs_dest_free(dest);
754 ip_vs_copy_stats(struct ip_vs_kstats *dst, struct ip_vs_stats *src)
756 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->kstats.c - src->kstats0.c
758 spin_lock_bh(&src->lock);
760 IP_VS_SHOW_STATS_COUNTER(conns);
761 IP_VS_SHOW_STATS_COUNTER(inpkts);
762 IP_VS_SHOW_STATS_COUNTER(outpkts);
763 IP_VS_SHOW_STATS_COUNTER(inbytes);
764 IP_VS_SHOW_STATS_COUNTER(outbytes);
766 ip_vs_read_estimator(dst, src);
768 spin_unlock_bh(&src->lock);
772 ip_vs_export_stats_user(struct ip_vs_stats_user *dst, struct ip_vs_kstats *src)
774 dst->conns = (u32)src->conns;
775 dst->inpkts = (u32)src->inpkts;
776 dst->outpkts = (u32)src->outpkts;
777 dst->inbytes = src->inbytes;
778 dst->outbytes = src->outbytes;
779 dst->cps = (u32)src->cps;
780 dst->inpps = (u32)src->inpps;
781 dst->outpps = (u32)src->outpps;
782 dst->inbps = (u32)src->inbps;
783 dst->outbps = (u32)src->outbps;
787 ip_vs_zero_stats(struct ip_vs_stats *stats)
789 spin_lock_bh(&stats->lock);
791 /* get current counters as zero point, rates are zeroed */
793 #define IP_VS_ZERO_STATS_COUNTER(c) stats->kstats0.c = stats->kstats.c
795 IP_VS_ZERO_STATS_COUNTER(conns);
796 IP_VS_ZERO_STATS_COUNTER(inpkts);
797 IP_VS_ZERO_STATS_COUNTER(outpkts);
798 IP_VS_ZERO_STATS_COUNTER(inbytes);
799 IP_VS_ZERO_STATS_COUNTER(outbytes);
801 ip_vs_zero_estimator(stats);
803 spin_unlock_bh(&stats->lock);
807 * Update a destination in the given service
810 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
811 struct ip_vs_dest_user_kern *udest, int add)
813 struct netns_ipvs *ipvs = svc->ipvs;
814 struct ip_vs_service *old_svc;
815 struct ip_vs_scheduler *sched;
818 /* We cannot modify an address and change the address family */
819 BUG_ON(!add && udest->af != dest->af);
821 if (add && udest->af != svc->af)
822 ipvs->mixed_address_family_dests++;
824 /* set the weight and the flags */
825 atomic_set(&dest->weight, udest->weight);
826 conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
827 conn_flags |= IP_VS_CONN_F_INACTIVE;
829 /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
830 if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
831 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
834 * Put the real service in rs_table if not present.
835 * For now only for NAT!
837 ip_vs_rs_hash(ipvs, dest);
839 atomic_set(&dest->conn_flags, conn_flags);
841 /* bind the service */
842 old_svc = rcu_dereference_protected(dest->svc, 1);
844 __ip_vs_bind_svc(dest, svc);
846 if (old_svc != svc) {
847 ip_vs_zero_stats(&dest->stats);
848 __ip_vs_bind_svc(dest, svc);
849 __ip_vs_svc_put(old_svc, true);
853 /* set the dest status flags */
854 dest->flags |= IP_VS_DEST_F_AVAILABLE;
856 if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
857 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
858 dest->u_threshold = udest->u_threshold;
859 dest->l_threshold = udest->l_threshold;
861 dest->af = udest->af;
863 spin_lock_bh(&dest->dst_lock);
864 __ip_vs_dst_cache_reset(dest);
865 spin_unlock_bh(&dest->dst_lock);
868 ip_vs_start_estimator(svc->ipvs, &dest->stats);
869 list_add_rcu(&dest->n_list, &svc->destinations);
871 sched = rcu_dereference_protected(svc->scheduler, 1);
872 if (sched && sched->add_dest)
873 sched->add_dest(svc, dest);
875 sched = rcu_dereference_protected(svc->scheduler, 1);
876 if (sched && sched->upd_dest)
877 sched->upd_dest(svc, dest);
883 * Create a destination for the given service
886 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
887 struct ip_vs_dest **dest_p)
889 struct ip_vs_dest *dest;
890 unsigned int atype, i;
894 #ifdef CONFIG_IP_VS_IPV6
895 if (udest->af == AF_INET6) {
896 atype = ipv6_addr_type(&udest->addr.in6);
897 if ((!(atype & IPV6_ADDR_UNICAST) ||
898 atype & IPV6_ADDR_LINKLOCAL) &&
899 !__ip_vs_addr_is_local_v6(svc->ipvs->net, &udest->addr.in6))
904 atype = inet_addr_type(svc->ipvs->net, udest->addr.ip);
905 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
909 dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
913 dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
914 if (!dest->stats.cpustats)
917 for_each_possible_cpu(i) {
918 struct ip_vs_cpu_stats *ip_vs_dest_stats;
919 ip_vs_dest_stats = per_cpu_ptr(dest->stats.cpustats, i);
920 u64_stats_init(&ip_vs_dest_stats->syncp);
923 dest->af = udest->af;
924 dest->protocol = svc->protocol;
925 dest->vaddr = svc->addr;
926 dest->vport = svc->port;
927 dest->vfwmark = svc->fwmark;
928 ip_vs_addr_copy(udest->af, &dest->addr, &udest->addr);
929 dest->port = udest->port;
931 atomic_set(&dest->activeconns, 0);
932 atomic_set(&dest->inactconns, 0);
933 atomic_set(&dest->persistconns, 0);
934 refcount_set(&dest->refcnt, 1);
936 INIT_HLIST_NODE(&dest->d_list);
937 spin_lock_init(&dest->dst_lock);
938 spin_lock_init(&dest->stats.lock);
939 __ip_vs_update_dest(svc, dest, udest, 1);
953 * Add a destination into an existing service
956 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
958 struct ip_vs_dest *dest;
959 union nf_inet_addr daddr;
960 __be16 dport = udest->port;
965 if (udest->weight < 0) {
966 pr_err("%s(): server weight less than zero\n", __func__);
970 if (udest->l_threshold > udest->u_threshold) {
971 pr_err("%s(): lower threshold is higher than upper threshold\n",
976 ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
978 /* We use function that requires RCU lock */
980 dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
984 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
989 * Check if the dest already exists in the trash and
990 * is from the same service
992 dest = ip_vs_trash_get_dest(svc, udest->af, &daddr, dport);
995 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
996 "dest->refcnt=%d, service %u/%s:%u\n",
997 IP_VS_DBG_ADDR(udest->af, &daddr), ntohs(dport),
998 refcount_read(&dest->refcnt),
1000 IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
1001 ntohs(dest->vport));
1003 __ip_vs_update_dest(svc, dest, udest, 1);
1007 * Allocate and initialize the dest structure
1009 ret = ip_vs_new_dest(svc, udest, &dest);
1018 * Edit a destination in the given service
1021 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1023 struct ip_vs_dest *dest;
1024 union nf_inet_addr daddr;
1025 __be16 dport = udest->port;
1029 if (udest->weight < 0) {
1030 pr_err("%s(): server weight less than zero\n", __func__);
1034 if (udest->l_threshold > udest->u_threshold) {
1035 pr_err("%s(): lower threshold is higher than upper threshold\n",
1040 ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
1042 /* We use function that requires RCU lock */
1044 dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
1048 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1052 __ip_vs_update_dest(svc, dest, udest, 0);
1059 * Delete a destination (must be already unlinked from the service)
1061 static void __ip_vs_del_dest(struct netns_ipvs *ipvs, struct ip_vs_dest *dest,
1064 ip_vs_stop_estimator(ipvs, &dest->stats);
1067 * Remove it from the d-linked list with the real services.
1069 ip_vs_rs_unhash(dest);
1071 spin_lock_bh(&ipvs->dest_trash_lock);
1072 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1073 IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port),
1074 refcount_read(&dest->refcnt));
1075 if (list_empty(&ipvs->dest_trash) && !cleanup)
1076 mod_timer(&ipvs->dest_trash_timer,
1077 jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1078 /* dest lives in trash with reference */
1079 list_add(&dest->t_list, &ipvs->dest_trash);
1080 dest->idle_start = 0;
1081 spin_unlock_bh(&ipvs->dest_trash_lock);
1086 * Unlink a destination from the given service
1088 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1089 struct ip_vs_dest *dest,
1092 dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1095 * Remove it from the d-linked destination list.
1097 list_del_rcu(&dest->n_list);
1100 if (dest->af != svc->af)
1101 svc->ipvs->mixed_address_family_dests--;
1104 struct ip_vs_scheduler *sched;
1106 sched = rcu_dereference_protected(svc->scheduler, 1);
1107 if (sched && sched->del_dest)
1108 sched->del_dest(svc, dest);
1114 * Delete a destination server in the given service
1117 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1119 struct ip_vs_dest *dest;
1120 __be16 dport = udest->port;
1124 /* We use function that requires RCU lock */
1126 dest = ip_vs_lookup_dest(svc, udest->af, &udest->addr, dport);
1130 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1135 * Unlink dest from the service
1137 __ip_vs_unlink_dest(svc, dest, 1);
1140 * Delete the destination
1142 __ip_vs_del_dest(svc->ipvs, dest, false);
1149 static void ip_vs_dest_trash_expire(struct timer_list *t)
1151 struct netns_ipvs *ipvs = from_timer(ipvs, t, dest_trash_timer);
1152 struct ip_vs_dest *dest, *next;
1153 unsigned long now = jiffies;
1155 spin_lock(&ipvs->dest_trash_lock);
1156 list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) {
1157 if (refcount_read(&dest->refcnt) > 1)
1159 if (dest->idle_start) {
1160 if (time_before(now, dest->idle_start +
1161 IP_VS_DEST_TRASH_PERIOD))
1164 dest->idle_start = max(1UL, now);
1167 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1169 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1171 list_del(&dest->t_list);
1172 ip_vs_dest_free(dest);
1174 if (!list_empty(&ipvs->dest_trash))
1175 mod_timer(&ipvs->dest_trash_timer,
1176 jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1177 spin_unlock(&ipvs->dest_trash_lock);
1181 * Add a service into the service hash table
1184 ip_vs_add_service(struct netns_ipvs *ipvs, struct ip_vs_service_user_kern *u,
1185 struct ip_vs_service **svc_p)
1188 struct ip_vs_scheduler *sched = NULL;
1189 struct ip_vs_pe *pe = NULL;
1190 struct ip_vs_service *svc = NULL;
1192 /* increase the module use count */
1193 ip_vs_use_count_inc();
1195 /* Lookup the scheduler by 'u->sched_name' */
1196 if (strcmp(u->sched_name, "none")) {
1197 sched = ip_vs_scheduler_get(u->sched_name);
1199 pr_info("Scheduler module ip_vs_%s not found\n",
1206 if (u->pe_name && *u->pe_name) {
1207 pe = ip_vs_pe_getbyname(u->pe_name);
1209 pr_info("persistence engine module ip_vs_pe_%s "
1210 "not found\n", u->pe_name);
1216 #ifdef CONFIG_IP_VS_IPV6
1217 if (u->af == AF_INET6) {
1218 __u32 plen = (__force __u32) u->netmask;
1220 if (plen < 1 || plen > 128) {
1227 svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1229 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1233 svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1234 if (!svc->stats.cpustats) {
1239 for_each_possible_cpu(i) {
1240 struct ip_vs_cpu_stats *ip_vs_stats;
1241 ip_vs_stats = per_cpu_ptr(svc->stats.cpustats, i);
1242 u64_stats_init(&ip_vs_stats->syncp);
1246 /* I'm the first user of the service */
1247 atomic_set(&svc->refcnt, 0);
1250 svc->protocol = u->protocol;
1251 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1252 svc->port = u->port;
1253 svc->fwmark = u->fwmark;
1254 svc->flags = u->flags;
1255 svc->timeout = u->timeout * HZ;
1256 svc->netmask = u->netmask;
1259 INIT_LIST_HEAD(&svc->destinations);
1260 spin_lock_init(&svc->sched_lock);
1261 spin_lock_init(&svc->stats.lock);
1263 /* Bind the scheduler */
1265 ret = ip_vs_bind_scheduler(svc, sched);
1271 /* Bind the ct retriever */
1272 RCU_INIT_POINTER(svc->pe, pe);
1275 /* Update the virtual service counters */
1276 if (svc->port == FTPPORT)
1277 atomic_inc(&ipvs->ftpsvc_counter);
1278 else if (svc->port == 0)
1279 atomic_inc(&ipvs->nullsvc_counter);
1280 if (svc->pe && svc->pe->conn_out)
1281 atomic_inc(&ipvs->conn_out_counter);
1283 ip_vs_start_estimator(ipvs, &svc->stats);
1285 /* Count only IPv4 services for old get/setsockopt interface */
1286 if (svc->af == AF_INET)
1287 ipvs->num_services++;
1289 /* Hash the service into the service table */
1290 ip_vs_svc_hash(svc);
1293 /* Now there is a service - full throttle */
1300 ip_vs_unbind_scheduler(svc, sched);
1301 ip_vs_service_free(svc);
1303 ip_vs_scheduler_put(sched);
1306 /* decrease the module use count */
1307 ip_vs_use_count_dec();
1314 * Edit a service and bind it with a new scheduler
1317 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1319 struct ip_vs_scheduler *sched = NULL, *old_sched;
1320 struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1322 bool new_pe_conn_out, old_pe_conn_out;
1325 * Lookup the scheduler, by 'u->sched_name'
1327 if (strcmp(u->sched_name, "none")) {
1328 sched = ip_vs_scheduler_get(u->sched_name);
1330 pr_info("Scheduler module ip_vs_%s not found\n",
1337 if (u->pe_name && *u->pe_name) {
1338 pe = ip_vs_pe_getbyname(u->pe_name);
1340 pr_info("persistence engine module ip_vs_pe_%s "
1341 "not found\n", u->pe_name);
1348 #ifdef CONFIG_IP_VS_IPV6
1349 if (u->af == AF_INET6) {
1350 __u32 plen = (__force __u32) u->netmask;
1352 if (plen < 1 || plen > 128) {
1359 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1360 if (sched != old_sched) {
1362 ip_vs_unbind_scheduler(svc, old_sched);
1363 RCU_INIT_POINTER(svc->scheduler, NULL);
1364 /* Wait all svc->sched_data users */
1367 /* Bind the new scheduler */
1369 ret = ip_vs_bind_scheduler(svc, sched);
1371 ip_vs_scheduler_put(sched);
1378 * Set the flags and timeout value
1380 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1381 svc->timeout = u->timeout * HZ;
1382 svc->netmask = u->netmask;
1384 old_pe = rcu_dereference_protected(svc->pe, 1);
1386 rcu_assign_pointer(svc->pe, pe);
1387 /* check for optional methods in new pe */
1388 new_pe_conn_out = (pe && pe->conn_out) ? true : false;
1389 old_pe_conn_out = (old_pe && old_pe->conn_out) ? true : false;
1390 if (new_pe_conn_out && !old_pe_conn_out)
1391 atomic_inc(&svc->ipvs->conn_out_counter);
1392 if (old_pe_conn_out && !new_pe_conn_out)
1393 atomic_dec(&svc->ipvs->conn_out_counter);
1397 ip_vs_scheduler_put(old_sched);
1398 ip_vs_pe_put(old_pe);
1403 * Delete a service from the service list
1404 * - The service must be unlinked, unlocked and not referenced!
1405 * - We are called under _bh lock
1407 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1409 struct ip_vs_dest *dest, *nxt;
1410 struct ip_vs_scheduler *old_sched;
1411 struct ip_vs_pe *old_pe;
1412 struct netns_ipvs *ipvs = svc->ipvs;
1414 /* Count only IPv4 services for old get/setsockopt interface */
1415 if (svc->af == AF_INET)
1416 ipvs->num_services--;
1418 ip_vs_stop_estimator(svc->ipvs, &svc->stats);
1420 /* Unbind scheduler */
1421 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1422 ip_vs_unbind_scheduler(svc, old_sched);
1423 ip_vs_scheduler_put(old_sched);
1425 /* Unbind persistence engine, keep svc->pe */
1426 old_pe = rcu_dereference_protected(svc->pe, 1);
1427 if (old_pe && old_pe->conn_out)
1428 atomic_dec(&ipvs->conn_out_counter);
1429 ip_vs_pe_put(old_pe);
1432 * Unlink the whole destination list
1434 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1435 __ip_vs_unlink_dest(svc, dest, 0);
1436 __ip_vs_del_dest(svc->ipvs, dest, cleanup);
1440 * Update the virtual service counters
1442 if (svc->port == FTPPORT)
1443 atomic_dec(&ipvs->ftpsvc_counter);
1444 else if (svc->port == 0)
1445 atomic_dec(&ipvs->nullsvc_counter);
1448 * Free the service if nobody refers to it
1450 __ip_vs_svc_put(svc, true);
1452 /* decrease the module use count */
1453 ip_vs_use_count_dec();
1457 * Unlink a service from list and try to delete it if its refcnt reached 0
1459 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1461 /* Hold svc to avoid double release from dest_trash */
1462 atomic_inc(&svc->refcnt);
1464 * Unhash it from the service table
1466 ip_vs_svc_unhash(svc);
1468 __ip_vs_del_service(svc, cleanup);
1472 * Delete a service from the service list
1474 static int ip_vs_del_service(struct ip_vs_service *svc)
1478 ip_vs_unlink_service(svc, false);
1485 * Flush all the virtual services
1487 static int ip_vs_flush(struct netns_ipvs *ipvs, bool cleanup)
1490 struct ip_vs_service *svc;
1491 struct hlist_node *n;
1494 * Flush the service table hashed by <netns,protocol,addr,port>
1496 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1497 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1499 if (svc->ipvs == ipvs)
1500 ip_vs_unlink_service(svc, cleanup);
1505 * Flush the service table hashed by fwmark
1507 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1508 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1510 if (svc->ipvs == ipvs)
1511 ip_vs_unlink_service(svc, cleanup);
1519 * Delete service by {netns} in the service table.
1520 * Called by __ip_vs_cleanup()
1522 void ip_vs_service_net_cleanup(struct netns_ipvs *ipvs)
1525 /* Check for "full" addressed entries */
1526 mutex_lock(&__ip_vs_mutex);
1527 ip_vs_flush(ipvs, true);
1528 mutex_unlock(&__ip_vs_mutex);
1532 /* Put all references for device (dst_cache) */
1534 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1536 struct ip_vs_dest_dst *dest_dst;
1538 spin_lock_bh(&dest->dst_lock);
1539 dest_dst = rcu_dereference_protected(dest->dest_dst, 1);
1540 if (dest_dst && dest_dst->dst_cache->dev == dev) {
1541 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1543 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1545 refcount_read(&dest->refcnt));
1546 __ip_vs_dst_cache_reset(dest);
1548 spin_unlock_bh(&dest->dst_lock);
1551 /* Netdev event receiver
1552 * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1554 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1557 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1558 struct net *net = dev_net(dev);
1559 struct netns_ipvs *ipvs = net_ipvs(net);
1560 struct ip_vs_service *svc;
1561 struct ip_vs_dest *dest;
1564 if (event != NETDEV_DOWN || !ipvs)
1566 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1568 mutex_lock(&__ip_vs_mutex);
1569 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1570 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1571 if (svc->ipvs == ipvs) {
1572 list_for_each_entry(dest, &svc->destinations,
1574 ip_vs_forget_dev(dest, dev);
1579 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1580 if (svc->ipvs == ipvs) {
1581 list_for_each_entry(dest, &svc->destinations,
1583 ip_vs_forget_dev(dest, dev);
1590 spin_lock_bh(&ipvs->dest_trash_lock);
1591 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1592 ip_vs_forget_dev(dest, dev);
1594 spin_unlock_bh(&ipvs->dest_trash_lock);
1595 mutex_unlock(&__ip_vs_mutex);
1601 * Zero counters in a service or all services
1603 static int ip_vs_zero_service(struct ip_vs_service *svc)
1605 struct ip_vs_dest *dest;
1607 list_for_each_entry(dest, &svc->destinations, n_list) {
1608 ip_vs_zero_stats(&dest->stats);
1610 ip_vs_zero_stats(&svc->stats);
1614 static int ip_vs_zero_all(struct netns_ipvs *ipvs)
1617 struct ip_vs_service *svc;
1619 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1620 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1621 if (svc->ipvs == ipvs)
1622 ip_vs_zero_service(svc);
1626 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1627 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1628 if (svc->ipvs == ipvs)
1629 ip_vs_zero_service(svc);
1633 ip_vs_zero_stats(&ipvs->tot_stats);
1637 #ifdef CONFIG_SYSCTL
1640 static int three = 3;
1643 proc_do_defense_mode(struct ctl_table *table, int write,
1644 void __user *buffer, size_t *lenp, loff_t *ppos)
1646 struct netns_ipvs *ipvs = table->extra2;
1647 int *valp = table->data;
1651 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1652 if (write && (*valp != val)) {
1653 if ((*valp < 0) || (*valp > 3)) {
1654 /* Restore the correct value */
1657 update_defense_level(ipvs);
1664 proc_do_sync_threshold(struct ctl_table *table, int write,
1665 void __user *buffer, size_t *lenp, loff_t *ppos)
1667 int *valp = table->data;
1671 /* backup the value first */
1672 memcpy(val, valp, sizeof(val));
1674 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1675 if (write && (valp[0] < 0 || valp[1] < 0 ||
1676 (valp[0] >= valp[1] && valp[1]))) {
1677 /* Restore the correct value */
1678 memcpy(valp, val, sizeof(val));
1684 proc_do_sync_mode(struct ctl_table *table, int write,
1685 void __user *buffer, size_t *lenp, loff_t *ppos)
1687 int *valp = table->data;
1691 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1692 if (write && (*valp != val)) {
1693 if ((*valp < 0) || (*valp > 1)) {
1694 /* Restore the correct value */
1702 proc_do_sync_ports(struct ctl_table *table, int write,
1703 void __user *buffer, size_t *lenp, loff_t *ppos)
1705 int *valp = table->data;
1709 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1710 if (write && (*valp != val)) {
1711 if (*valp < 1 || !is_power_of_2(*valp)) {
1712 /* Restore the correct value */
1720 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1721 * Do not change order or insert new entries without
1722 * align with netns init in ip_vs_control_net_init()
1725 static struct ctl_table vs_vars[] = {
1727 .procname = "amemthresh",
1728 .maxlen = sizeof(int),
1730 .proc_handler = proc_dointvec,
1733 .procname = "am_droprate",
1734 .maxlen = sizeof(int),
1736 .proc_handler = proc_dointvec,
1739 .procname = "drop_entry",
1740 .maxlen = sizeof(int),
1742 .proc_handler = proc_do_defense_mode,
1745 .procname = "drop_packet",
1746 .maxlen = sizeof(int),
1748 .proc_handler = proc_do_defense_mode,
1750 #ifdef CONFIG_IP_VS_NFCT
1752 .procname = "conntrack",
1753 .maxlen = sizeof(int),
1755 .proc_handler = &proc_dointvec,
1759 .procname = "secure_tcp",
1760 .maxlen = sizeof(int),
1762 .proc_handler = proc_do_defense_mode,
1765 .procname = "snat_reroute",
1766 .maxlen = sizeof(int),
1768 .proc_handler = &proc_dointvec,
1771 .procname = "sync_version",
1772 .maxlen = sizeof(int),
1774 .proc_handler = proc_do_sync_mode,
1777 .procname = "sync_ports",
1778 .maxlen = sizeof(int),
1780 .proc_handler = proc_do_sync_ports,
1783 .procname = "sync_persist_mode",
1784 .maxlen = sizeof(int),
1786 .proc_handler = proc_dointvec,
1789 .procname = "sync_qlen_max",
1790 .maxlen = sizeof(unsigned long),
1792 .proc_handler = proc_doulongvec_minmax,
1795 .procname = "sync_sock_size",
1796 .maxlen = sizeof(int),
1798 .proc_handler = proc_dointvec,
1801 .procname = "cache_bypass",
1802 .maxlen = sizeof(int),
1804 .proc_handler = proc_dointvec,
1807 .procname = "expire_nodest_conn",
1808 .maxlen = sizeof(int),
1810 .proc_handler = proc_dointvec,
1813 .procname = "sloppy_tcp",
1814 .maxlen = sizeof(int),
1816 .proc_handler = proc_dointvec,
1819 .procname = "sloppy_sctp",
1820 .maxlen = sizeof(int),
1822 .proc_handler = proc_dointvec,
1825 .procname = "expire_quiescent_template",
1826 .maxlen = sizeof(int),
1828 .proc_handler = proc_dointvec,
1831 .procname = "sync_threshold",
1833 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1835 .proc_handler = proc_do_sync_threshold,
1838 .procname = "sync_refresh_period",
1839 .maxlen = sizeof(int),
1841 .proc_handler = proc_dointvec_jiffies,
1844 .procname = "sync_retries",
1845 .maxlen = sizeof(int),
1847 .proc_handler = proc_dointvec_minmax,
1852 .procname = "nat_icmp_send",
1853 .maxlen = sizeof(int),
1855 .proc_handler = proc_dointvec,
1858 .procname = "pmtu_disc",
1859 .maxlen = sizeof(int),
1861 .proc_handler = proc_dointvec,
1864 .procname = "backup_only",
1865 .maxlen = sizeof(int),
1867 .proc_handler = proc_dointvec,
1870 .procname = "conn_reuse_mode",
1871 .maxlen = sizeof(int),
1873 .proc_handler = proc_dointvec,
1876 .procname = "schedule_icmp",
1877 .maxlen = sizeof(int),
1879 .proc_handler = proc_dointvec,
1882 .procname = "ignore_tunneled",
1883 .maxlen = sizeof(int),
1885 .proc_handler = proc_dointvec,
1887 #ifdef CONFIG_IP_VS_DEBUG
1889 .procname = "debug_level",
1890 .data = &sysctl_ip_vs_debug_level,
1891 .maxlen = sizeof(int),
1893 .proc_handler = proc_dointvec,
1901 #ifdef CONFIG_PROC_FS
1904 struct seq_net_private p; /* Do not move this, netns depends upon it*/
1905 struct hlist_head *table;
1910 * Write the contents of the VS rule table to a PROCfs file.
1911 * (It is kept just for backward compatibility)
1913 static inline const char *ip_vs_fwd_name(unsigned int flags)
1915 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1916 case IP_VS_CONN_F_LOCALNODE:
1918 case IP_VS_CONN_F_TUNNEL:
1920 case IP_VS_CONN_F_DROUTE:
1928 /* Get the Nth entry in the two lists */
1929 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1931 struct net *net = seq_file_net(seq);
1932 struct netns_ipvs *ipvs = net_ipvs(net);
1933 struct ip_vs_iter *iter = seq->private;
1935 struct ip_vs_service *svc;
1937 /* look in hash by protocol */
1938 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1939 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
1940 if ((svc->ipvs == ipvs) && pos-- == 0) {
1941 iter->table = ip_vs_svc_table;
1948 /* keep looking in fwmark */
1949 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1950 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
1952 if ((svc->ipvs == ipvs) && pos-- == 0) {
1953 iter->table = ip_vs_svc_fwm_table;
1963 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1967 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1971 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1973 struct hlist_node *e;
1974 struct ip_vs_iter *iter;
1975 struct ip_vs_service *svc;
1978 if (v == SEQ_START_TOKEN)
1979 return ip_vs_info_array(seq,0);
1982 iter = seq->private;
1984 if (iter->table == ip_vs_svc_table) {
1985 /* next service in table hashed by protocol */
1986 e = rcu_dereference(hlist_next_rcu(&svc->s_list));
1988 return hlist_entry(e, struct ip_vs_service, s_list);
1990 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1991 hlist_for_each_entry_rcu(svc,
1992 &ip_vs_svc_table[iter->bucket],
1998 iter->table = ip_vs_svc_fwm_table;
2003 /* next service in hashed by fwmark */
2004 e = rcu_dereference(hlist_next_rcu(&svc->f_list));
2006 return hlist_entry(e, struct ip_vs_service, f_list);
2009 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2010 hlist_for_each_entry_rcu(svc,
2011 &ip_vs_svc_fwm_table[iter->bucket],
2019 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
2026 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
2028 if (v == SEQ_START_TOKEN) {
2030 "IP Virtual Server version %d.%d.%d (size=%d)\n",
2031 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2033 "Prot LocalAddress:Port Scheduler Flags\n");
2035 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2037 struct net *net = seq_file_net(seq);
2038 struct netns_ipvs *ipvs = net_ipvs(net);
2039 const struct ip_vs_service *svc = v;
2040 const struct ip_vs_iter *iter = seq->private;
2041 const struct ip_vs_dest *dest;
2042 struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler);
2043 char *sched_name = sched ? sched->name : "none";
2045 if (svc->ipvs != ipvs)
2047 if (iter->table == ip_vs_svc_table) {
2048 #ifdef CONFIG_IP_VS_IPV6
2049 if (svc->af == AF_INET6)
2050 seq_printf(seq, "%s [%pI6]:%04X %s ",
2051 ip_vs_proto_name(svc->protocol),
2057 seq_printf(seq, "%s %08X:%04X %s %s ",
2058 ip_vs_proto_name(svc->protocol),
2059 ntohl(svc->addr.ip),
2062 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2064 seq_printf(seq, "FWM %08X %s %s",
2065 svc->fwmark, sched_name,
2066 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2069 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2070 seq_printf(seq, "persistent %d %08X\n",
2072 ntohl(svc->netmask));
2074 seq_putc(seq, '\n');
2076 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
2077 #ifdef CONFIG_IP_VS_IPV6
2078 if (dest->af == AF_INET6)
2081 " %-7s %-6d %-10d %-10d\n",
2084 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2085 atomic_read(&dest->weight),
2086 atomic_read(&dest->activeconns),
2087 atomic_read(&dest->inactconns));
2092 "%-7s %-6d %-10d %-10d\n",
2093 ntohl(dest->addr.ip),
2095 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2096 atomic_read(&dest->weight),
2097 atomic_read(&dest->activeconns),
2098 atomic_read(&dest->inactconns));
2105 static const struct seq_operations ip_vs_info_seq_ops = {
2106 .start = ip_vs_info_seq_start,
2107 .next = ip_vs_info_seq_next,
2108 .stop = ip_vs_info_seq_stop,
2109 .show = ip_vs_info_seq_show,
2112 static int ip_vs_info_open(struct inode *inode, struct file *file)
2114 return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2115 sizeof(struct ip_vs_iter));
2118 static const struct file_operations ip_vs_info_fops = {
2119 .open = ip_vs_info_open,
2121 .llseek = seq_lseek,
2122 .release = seq_release_net,
2125 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2127 struct net *net = seq_file_single_net(seq);
2128 struct ip_vs_kstats show;
2130 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2132 " Total Incoming Outgoing Incoming Outgoing\n");
2134 " Conns Packets Packets Bytes Bytes\n");
2136 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2137 seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n\n",
2138 (unsigned long long)show.conns,
2139 (unsigned long long)show.inpkts,
2140 (unsigned long long)show.outpkts,
2141 (unsigned long long)show.inbytes,
2142 (unsigned long long)show.outbytes);
2144 /* 01234567 01234567 01234567 0123456701234567 0123456701234567*/
2146 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2147 seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n",
2148 (unsigned long long)show.cps,
2149 (unsigned long long)show.inpps,
2150 (unsigned long long)show.outpps,
2151 (unsigned long long)show.inbps,
2152 (unsigned long long)show.outbps);
2157 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2159 return single_open_net(inode, file, ip_vs_stats_show);
2162 static const struct file_operations ip_vs_stats_fops = {
2163 .open = ip_vs_stats_seq_open,
2165 .llseek = seq_lseek,
2166 .release = single_release_net,
2169 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2171 struct net *net = seq_file_single_net(seq);
2172 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2173 struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats;
2174 struct ip_vs_kstats kstats;
2177 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2179 " Total Incoming Outgoing Incoming Outgoing\n");
2181 "CPU Conns Packets Packets Bytes Bytes\n");
2183 for_each_possible_cpu(i) {
2184 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2186 u64 conns, inpkts, outpkts, inbytes, outbytes;
2189 start = u64_stats_fetch_begin_irq(&u->syncp);
2190 conns = u->cnt.conns;
2191 inpkts = u->cnt.inpkts;
2192 outpkts = u->cnt.outpkts;
2193 inbytes = u->cnt.inbytes;
2194 outbytes = u->cnt.outbytes;
2195 } while (u64_stats_fetch_retry_irq(&u->syncp, start));
2197 seq_printf(seq, "%3X %8LX %8LX %8LX %16LX %16LX\n",
2198 i, (u64)conns, (u64)inpkts,
2199 (u64)outpkts, (u64)inbytes,
2203 ip_vs_copy_stats(&kstats, tot_stats);
2205 seq_printf(seq, " ~ %8LX %8LX %8LX %16LX %16LX\n\n",
2206 (unsigned long long)kstats.conns,
2207 (unsigned long long)kstats.inpkts,
2208 (unsigned long long)kstats.outpkts,
2209 (unsigned long long)kstats.inbytes,
2210 (unsigned long long)kstats.outbytes);
2212 /* ... 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2214 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2215 seq_printf(seq, " %8LX %8LX %8LX %16LX %16LX\n",
2225 static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2227 return single_open_net(inode, file, ip_vs_stats_percpu_show);
2230 static const struct file_operations ip_vs_stats_percpu_fops = {
2231 .open = ip_vs_stats_percpu_seq_open,
2233 .llseek = seq_lseek,
2234 .release = single_release_net,
2239 * Set timeout values for tcp tcpfin udp in the timeout_table.
2241 static int ip_vs_set_timeout(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2243 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2244 struct ip_vs_proto_data *pd;
2247 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2252 #ifdef CONFIG_IP_VS_PROTO_TCP
2253 if (u->tcp_timeout) {
2254 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2255 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2256 = u->tcp_timeout * HZ;
2259 if (u->tcp_fin_timeout) {
2260 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2261 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2262 = u->tcp_fin_timeout * HZ;
2266 #ifdef CONFIG_IP_VS_PROTO_UDP
2267 if (u->udp_timeout) {
2268 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2269 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2270 = u->udp_timeout * HZ;
2276 #define CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2278 struct ip_vs_svcdest_user {
2279 struct ip_vs_service_user s;
2280 struct ip_vs_dest_user d;
2283 static const unsigned char set_arglen[CMDID(IP_VS_SO_SET_MAX) + 1] = {
2284 [CMDID(IP_VS_SO_SET_ADD)] = sizeof(struct ip_vs_service_user),
2285 [CMDID(IP_VS_SO_SET_EDIT)] = sizeof(struct ip_vs_service_user),
2286 [CMDID(IP_VS_SO_SET_DEL)] = sizeof(struct ip_vs_service_user),
2287 [CMDID(IP_VS_SO_SET_ADDDEST)] = sizeof(struct ip_vs_svcdest_user),
2288 [CMDID(IP_VS_SO_SET_DELDEST)] = sizeof(struct ip_vs_svcdest_user),
2289 [CMDID(IP_VS_SO_SET_EDITDEST)] = sizeof(struct ip_vs_svcdest_user),
2290 [CMDID(IP_VS_SO_SET_TIMEOUT)] = sizeof(struct ip_vs_timeout_user),
2291 [CMDID(IP_VS_SO_SET_STARTDAEMON)] = sizeof(struct ip_vs_daemon_user),
2292 [CMDID(IP_VS_SO_SET_STOPDAEMON)] = sizeof(struct ip_vs_daemon_user),
2293 [CMDID(IP_VS_SO_SET_ZERO)] = sizeof(struct ip_vs_service_user),
2296 union ip_vs_set_arglen {
2297 struct ip_vs_service_user field_IP_VS_SO_SET_ADD;
2298 struct ip_vs_service_user field_IP_VS_SO_SET_EDIT;
2299 struct ip_vs_service_user field_IP_VS_SO_SET_DEL;
2300 struct ip_vs_svcdest_user field_IP_VS_SO_SET_ADDDEST;
2301 struct ip_vs_svcdest_user field_IP_VS_SO_SET_DELDEST;
2302 struct ip_vs_svcdest_user field_IP_VS_SO_SET_EDITDEST;
2303 struct ip_vs_timeout_user field_IP_VS_SO_SET_TIMEOUT;
2304 struct ip_vs_daemon_user field_IP_VS_SO_SET_STARTDAEMON;
2305 struct ip_vs_daemon_user field_IP_VS_SO_SET_STOPDAEMON;
2306 struct ip_vs_service_user field_IP_VS_SO_SET_ZERO;
2309 #define MAX_SET_ARGLEN sizeof(union ip_vs_set_arglen)
2311 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2312 struct ip_vs_service_user *usvc_compat)
2314 memset(usvc, 0, sizeof(*usvc));
2317 usvc->protocol = usvc_compat->protocol;
2318 usvc->addr.ip = usvc_compat->addr;
2319 usvc->port = usvc_compat->port;
2320 usvc->fwmark = usvc_compat->fwmark;
2322 /* Deep copy of sched_name is not needed here */
2323 usvc->sched_name = usvc_compat->sched_name;
2325 usvc->flags = usvc_compat->flags;
2326 usvc->timeout = usvc_compat->timeout;
2327 usvc->netmask = usvc_compat->netmask;
2330 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2331 struct ip_vs_dest_user *udest_compat)
2333 memset(udest, 0, sizeof(*udest));
2335 udest->addr.ip = udest_compat->addr;
2336 udest->port = udest_compat->port;
2337 udest->conn_flags = udest_compat->conn_flags;
2338 udest->weight = udest_compat->weight;
2339 udest->u_threshold = udest_compat->u_threshold;
2340 udest->l_threshold = udest_compat->l_threshold;
2341 udest->af = AF_INET;
2345 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2347 struct net *net = sock_net(sk);
2349 unsigned char arg[MAX_SET_ARGLEN];
2350 struct ip_vs_service_user *usvc_compat;
2351 struct ip_vs_service_user_kern usvc;
2352 struct ip_vs_service *svc;
2353 struct ip_vs_dest_user *udest_compat;
2354 struct ip_vs_dest_user_kern udest;
2355 struct netns_ipvs *ipvs = net_ipvs(net);
2357 BUILD_BUG_ON(sizeof(arg) > 255);
2358 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2361 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2363 if (len != set_arglen[CMDID(cmd)]) {
2364 IP_VS_DBG(1, "set_ctl: len %u != %u\n",
2365 len, set_arglen[CMDID(cmd)]);
2369 if (copy_from_user(arg, user, len) != 0)
2372 /* increase the module use count */
2373 ip_vs_use_count_inc();
2375 /* Handle daemons since they have another lock */
2376 if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2377 cmd == IP_VS_SO_SET_STOPDAEMON) {
2378 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2380 if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2381 struct ipvs_sync_daemon_cfg cfg;
2383 memset(&cfg, 0, sizeof(cfg));
2385 if (strscpy(cfg.mcast_ifn, dm->mcast_ifn,
2386 sizeof(cfg.mcast_ifn)) <= 0)
2388 cfg.syncid = dm->syncid;
2389 ret = start_sync_thread(ipvs, &cfg, dm->state);
2391 mutex_lock(&ipvs->sync_mutex);
2392 ret = stop_sync_thread(ipvs, dm->state);
2393 mutex_unlock(&ipvs->sync_mutex);
2398 mutex_lock(&__ip_vs_mutex);
2399 if (cmd == IP_VS_SO_SET_FLUSH) {
2400 /* Flush the virtual service */
2401 ret = ip_vs_flush(ipvs, false);
2403 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2404 /* Set timeout values for (tcp tcpfin udp) */
2405 ret = ip_vs_set_timeout(ipvs, (struct ip_vs_timeout_user *)arg);
2409 usvc_compat = (struct ip_vs_service_user *)arg;
2410 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2412 /* We only use the new structs internally, so copy userspace compat
2413 * structs to extended internal versions */
2414 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2415 ip_vs_copy_udest_compat(&udest, udest_compat);
2417 if (cmd == IP_VS_SO_SET_ZERO) {
2418 /* if no service address is set, zero counters in all */
2419 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2420 ret = ip_vs_zero_all(ipvs);
2425 if ((cmd == IP_VS_SO_SET_ADD || cmd == IP_VS_SO_SET_EDIT) &&
2426 strnlen(usvc.sched_name, IP_VS_SCHEDNAME_MAXLEN) ==
2427 IP_VS_SCHEDNAME_MAXLEN) {
2432 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2433 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2434 usvc.protocol != IPPROTO_SCTP) {
2435 pr_err("set_ctl: invalid protocol: %d %pI4:%d\n",
2436 usvc.protocol, &usvc.addr.ip,
2442 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2444 if (usvc.fwmark == 0)
2445 svc = __ip_vs_service_find(ipvs, usvc.af, usvc.protocol,
2446 &usvc.addr, usvc.port);
2448 svc = __ip_vs_svc_fwm_find(ipvs, usvc.af, usvc.fwmark);
2451 if (cmd != IP_VS_SO_SET_ADD
2452 && (svc == NULL || svc->protocol != usvc.protocol)) {
2458 case IP_VS_SO_SET_ADD:
2462 ret = ip_vs_add_service(ipvs, &usvc, &svc);
2464 case IP_VS_SO_SET_EDIT:
2465 ret = ip_vs_edit_service(svc, &usvc);
2467 case IP_VS_SO_SET_DEL:
2468 ret = ip_vs_del_service(svc);
2472 case IP_VS_SO_SET_ZERO:
2473 ret = ip_vs_zero_service(svc);
2475 case IP_VS_SO_SET_ADDDEST:
2476 ret = ip_vs_add_dest(svc, &udest);
2478 case IP_VS_SO_SET_EDITDEST:
2479 ret = ip_vs_edit_dest(svc, &udest);
2481 case IP_VS_SO_SET_DELDEST:
2482 ret = ip_vs_del_dest(svc, &udest);
2489 mutex_unlock(&__ip_vs_mutex);
2491 /* decrease the module use count */
2492 ip_vs_use_count_dec();
2499 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2501 struct ip_vs_scheduler *sched;
2502 struct ip_vs_kstats kstats;
2505 sched = rcu_dereference_protected(src->scheduler, 1);
2506 sched_name = sched ? sched->name : "none";
2507 dst->protocol = src->protocol;
2508 dst->addr = src->addr.ip;
2509 dst->port = src->port;
2510 dst->fwmark = src->fwmark;
2511 strlcpy(dst->sched_name, sched_name, sizeof(dst->sched_name));
2512 dst->flags = src->flags;
2513 dst->timeout = src->timeout / HZ;
2514 dst->netmask = src->netmask;
2515 dst->num_dests = src->num_dests;
2516 ip_vs_copy_stats(&kstats, &src->stats);
2517 ip_vs_export_stats_user(&dst->stats, &kstats);
2521 __ip_vs_get_service_entries(struct netns_ipvs *ipvs,
2522 const struct ip_vs_get_services *get,
2523 struct ip_vs_get_services __user *uptr)
2526 struct ip_vs_service *svc;
2527 struct ip_vs_service_entry entry;
2530 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2531 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2532 /* Only expose IPv4 entries to old interface */
2533 if (svc->af != AF_INET || (svc->ipvs != ipvs))
2536 if (count >= get->num_services)
2538 memset(&entry, 0, sizeof(entry));
2539 ip_vs_copy_service(&entry, svc);
2540 if (copy_to_user(&uptr->entrytable[count],
2541 &entry, sizeof(entry))) {
2549 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2550 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2551 /* Only expose IPv4 entries to old interface */
2552 if (svc->af != AF_INET || (svc->ipvs != ipvs))
2555 if (count >= get->num_services)
2557 memset(&entry, 0, sizeof(entry));
2558 ip_vs_copy_service(&entry, svc);
2559 if (copy_to_user(&uptr->entrytable[count],
2560 &entry, sizeof(entry))) {
2572 __ip_vs_get_dest_entries(struct netns_ipvs *ipvs, const struct ip_vs_get_dests *get,
2573 struct ip_vs_get_dests __user *uptr)
2575 struct ip_vs_service *svc;
2576 union nf_inet_addr addr = { .ip = get->addr };
2581 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, get->fwmark);
2583 svc = __ip_vs_service_find(ipvs, AF_INET, get->protocol, &addr,
2589 struct ip_vs_dest *dest;
2590 struct ip_vs_dest_entry entry;
2591 struct ip_vs_kstats kstats;
2593 memset(&entry, 0, sizeof(entry));
2594 list_for_each_entry(dest, &svc->destinations, n_list) {
2595 if (count >= get->num_dests)
2598 /* Cannot expose heterogeneous members via sockopt
2601 if (dest->af != svc->af)
2604 entry.addr = dest->addr.ip;
2605 entry.port = dest->port;
2606 entry.conn_flags = atomic_read(&dest->conn_flags);
2607 entry.weight = atomic_read(&dest->weight);
2608 entry.u_threshold = dest->u_threshold;
2609 entry.l_threshold = dest->l_threshold;
2610 entry.activeconns = atomic_read(&dest->activeconns);
2611 entry.inactconns = atomic_read(&dest->inactconns);
2612 entry.persistconns = atomic_read(&dest->persistconns);
2613 ip_vs_copy_stats(&kstats, &dest->stats);
2614 ip_vs_export_stats_user(&entry.stats, &kstats);
2615 if (copy_to_user(&uptr->entrytable[count],
2616 &entry, sizeof(entry))) {
2628 __ip_vs_get_timeouts(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2630 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2631 struct ip_vs_proto_data *pd;
2634 memset(u, 0, sizeof (*u));
2636 #ifdef CONFIG_IP_VS_PROTO_TCP
2637 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2638 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2639 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2641 #ifdef CONFIG_IP_VS_PROTO_UDP
2642 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2644 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2648 static const unsigned char get_arglen[CMDID(IP_VS_SO_GET_MAX) + 1] = {
2649 [CMDID(IP_VS_SO_GET_VERSION)] = 64,
2650 [CMDID(IP_VS_SO_GET_INFO)] = sizeof(struct ip_vs_getinfo),
2651 [CMDID(IP_VS_SO_GET_SERVICES)] = sizeof(struct ip_vs_get_services),
2652 [CMDID(IP_VS_SO_GET_SERVICE)] = sizeof(struct ip_vs_service_entry),
2653 [CMDID(IP_VS_SO_GET_DESTS)] = sizeof(struct ip_vs_get_dests),
2654 [CMDID(IP_VS_SO_GET_TIMEOUT)] = sizeof(struct ip_vs_timeout_user),
2655 [CMDID(IP_VS_SO_GET_DAEMON)] = 2 * sizeof(struct ip_vs_daemon_user),
2658 union ip_vs_get_arglen {
2659 char field_IP_VS_SO_GET_VERSION[64];
2660 struct ip_vs_getinfo field_IP_VS_SO_GET_INFO;
2661 struct ip_vs_get_services field_IP_VS_SO_GET_SERVICES;
2662 struct ip_vs_service_entry field_IP_VS_SO_GET_SERVICE;
2663 struct ip_vs_get_dests field_IP_VS_SO_GET_DESTS;
2664 struct ip_vs_timeout_user field_IP_VS_SO_GET_TIMEOUT;
2665 struct ip_vs_daemon_user field_IP_VS_SO_GET_DAEMON[2];
2668 #define MAX_GET_ARGLEN sizeof(union ip_vs_get_arglen)
2671 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2673 unsigned char arg[MAX_GET_ARGLEN];
2675 unsigned int copylen;
2676 struct net *net = sock_net(sk);
2677 struct netns_ipvs *ipvs = net_ipvs(net);
2680 BUILD_BUG_ON(sizeof(arg) > 255);
2681 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2684 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2687 copylen = get_arglen[CMDID(cmd)];
2688 if (*len < (int) copylen) {
2689 IP_VS_DBG(1, "get_ctl: len %d < %u\n", *len, copylen);
2693 if (copy_from_user(arg, user, copylen) != 0)
2696 * Handle daemons first since it has its own locking
2698 if (cmd == IP_VS_SO_GET_DAEMON) {
2699 struct ip_vs_daemon_user d[2];
2701 memset(&d, 0, sizeof(d));
2702 mutex_lock(&ipvs->sync_mutex);
2703 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2704 d[0].state = IP_VS_STATE_MASTER;
2705 strlcpy(d[0].mcast_ifn, ipvs->mcfg.mcast_ifn,
2706 sizeof(d[0].mcast_ifn));
2707 d[0].syncid = ipvs->mcfg.syncid;
2709 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2710 d[1].state = IP_VS_STATE_BACKUP;
2711 strlcpy(d[1].mcast_ifn, ipvs->bcfg.mcast_ifn,
2712 sizeof(d[1].mcast_ifn));
2713 d[1].syncid = ipvs->bcfg.syncid;
2715 if (copy_to_user(user, &d, sizeof(d)) != 0)
2717 mutex_unlock(&ipvs->sync_mutex);
2721 mutex_lock(&__ip_vs_mutex);
2723 case IP_VS_SO_GET_VERSION:
2727 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2728 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2729 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2733 *len = strlen(buf)+1;
2737 case IP_VS_SO_GET_INFO:
2739 struct ip_vs_getinfo info;
2740 info.version = IP_VS_VERSION_CODE;
2741 info.size = ip_vs_conn_tab_size;
2742 info.num_services = ipvs->num_services;
2743 if (copy_to_user(user, &info, sizeof(info)) != 0)
2748 case IP_VS_SO_GET_SERVICES:
2750 struct ip_vs_get_services *get;
2753 get = (struct ip_vs_get_services *)arg;
2754 size = sizeof(*get) +
2755 sizeof(struct ip_vs_service_entry) * get->num_services;
2757 pr_err("length: %u != %u\n", *len, size);
2761 ret = __ip_vs_get_service_entries(ipvs, get, user);
2765 case IP_VS_SO_GET_SERVICE:
2767 struct ip_vs_service_entry *entry;
2768 struct ip_vs_service *svc;
2769 union nf_inet_addr addr;
2771 entry = (struct ip_vs_service_entry *)arg;
2772 addr.ip = entry->addr;
2775 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, entry->fwmark);
2777 svc = __ip_vs_service_find(ipvs, AF_INET,
2778 entry->protocol, &addr,
2782 ip_vs_copy_service(entry, svc);
2783 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2790 case IP_VS_SO_GET_DESTS:
2792 struct ip_vs_get_dests *get;
2795 get = (struct ip_vs_get_dests *)arg;
2796 size = sizeof(*get) +
2797 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2799 pr_err("length: %u != %u\n", *len, size);
2803 ret = __ip_vs_get_dest_entries(ipvs, get, user);
2807 case IP_VS_SO_GET_TIMEOUT:
2809 struct ip_vs_timeout_user t;
2811 __ip_vs_get_timeouts(ipvs, &t);
2812 if (copy_to_user(user, &t, sizeof(t)) != 0)
2822 mutex_unlock(&__ip_vs_mutex);
2827 static struct nf_sockopt_ops ip_vs_sockopts = {
2829 .set_optmin = IP_VS_BASE_CTL,
2830 .set_optmax = IP_VS_SO_SET_MAX+1,
2831 .set = do_ip_vs_set_ctl,
2832 .get_optmin = IP_VS_BASE_CTL,
2833 .get_optmax = IP_VS_SO_GET_MAX+1,
2834 .get = do_ip_vs_get_ctl,
2835 .owner = THIS_MODULE,
2839 * Generic Netlink interface
2842 /* IPVS genetlink family */
2843 static struct genl_family ip_vs_genl_family;
2845 /* Policy used for first-level command attributes */
2846 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2847 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2848 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2849 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2850 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2851 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2852 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2855 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2856 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2857 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2858 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2859 .len = IP_VS_IFNAME_MAXLEN - 1 },
2860 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2861 [IPVS_DAEMON_ATTR_SYNC_MAXLEN] = { .type = NLA_U16 },
2862 [IPVS_DAEMON_ATTR_MCAST_GROUP] = { .type = NLA_U32 },
2863 [IPVS_DAEMON_ATTR_MCAST_GROUP6] = { .len = sizeof(struct in6_addr) },
2864 [IPVS_DAEMON_ATTR_MCAST_PORT] = { .type = NLA_U16 },
2865 [IPVS_DAEMON_ATTR_MCAST_TTL] = { .type = NLA_U8 },
2868 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2869 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2870 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2871 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2872 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2873 .len = sizeof(union nf_inet_addr) },
2874 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2875 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2876 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2877 .len = IP_VS_SCHEDNAME_MAXLEN - 1 },
2878 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2879 .len = IP_VS_PENAME_MAXLEN },
2880 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2881 .len = sizeof(struct ip_vs_flags) },
2882 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2883 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2884 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2887 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2888 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2889 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2890 .len = sizeof(union nf_inet_addr) },
2891 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2892 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2893 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2894 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2895 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2896 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2897 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2898 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2899 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2900 [IPVS_DEST_ATTR_ADDR_FAMILY] = { .type = NLA_U16 },
2903 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2904 struct ip_vs_kstats *kstats)
2906 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2911 if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, (u32)kstats->conns) ||
2912 nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, (u32)kstats->inpkts) ||
2913 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, (u32)kstats->outpkts) ||
2914 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes,
2915 IPVS_STATS_ATTR_PAD) ||
2916 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes,
2917 IPVS_STATS_ATTR_PAD) ||
2918 nla_put_u32(skb, IPVS_STATS_ATTR_CPS, (u32)kstats->cps) ||
2919 nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, (u32)kstats->inpps) ||
2920 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, (u32)kstats->outpps) ||
2921 nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, (u32)kstats->inbps) ||
2922 nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, (u32)kstats->outbps))
2923 goto nla_put_failure;
2924 nla_nest_end(skb, nl_stats);
2929 nla_nest_cancel(skb, nl_stats);
2933 static int ip_vs_genl_fill_stats64(struct sk_buff *skb, int container_type,
2934 struct ip_vs_kstats *kstats)
2936 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2941 if (nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CONNS, kstats->conns,
2942 IPVS_STATS_ATTR_PAD) ||
2943 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPKTS, kstats->inpkts,
2944 IPVS_STATS_ATTR_PAD) ||
2945 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPKTS, kstats->outpkts,
2946 IPVS_STATS_ATTR_PAD) ||
2947 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes,
2948 IPVS_STATS_ATTR_PAD) ||
2949 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes,
2950 IPVS_STATS_ATTR_PAD) ||
2951 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CPS, kstats->cps,
2952 IPVS_STATS_ATTR_PAD) ||
2953 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPPS, kstats->inpps,
2954 IPVS_STATS_ATTR_PAD) ||
2955 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPPS, kstats->outpps,
2956 IPVS_STATS_ATTR_PAD) ||
2957 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBPS, kstats->inbps,
2958 IPVS_STATS_ATTR_PAD) ||
2959 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBPS, kstats->outbps,
2960 IPVS_STATS_ATTR_PAD))
2961 goto nla_put_failure;
2962 nla_nest_end(skb, nl_stats);
2967 nla_nest_cancel(skb, nl_stats);
2971 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2972 struct ip_vs_service *svc)
2974 struct ip_vs_scheduler *sched;
2975 struct ip_vs_pe *pe;
2976 struct nlattr *nl_service;
2977 struct ip_vs_flags flags = { .flags = svc->flags,
2979 struct ip_vs_kstats kstats;
2982 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2986 if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2987 goto nla_put_failure;
2989 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2990 goto nla_put_failure;
2992 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2993 nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2994 nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2995 goto nla_put_failure;
2998 sched = rcu_dereference_protected(svc->scheduler, 1);
2999 sched_name = sched ? sched->name : "none";
3000 pe = rcu_dereference_protected(svc->pe, 1);
3001 if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched_name) ||
3002 (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
3003 nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
3004 nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
3005 nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
3006 goto nla_put_failure;
3007 ip_vs_copy_stats(&kstats, &svc->stats);
3008 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &kstats))
3009 goto nla_put_failure;
3010 if (ip_vs_genl_fill_stats64(skb, IPVS_SVC_ATTR_STATS64, &kstats))
3011 goto nla_put_failure;
3013 nla_nest_end(skb, nl_service);
3018 nla_nest_cancel(skb, nl_service);
3022 static int ip_vs_genl_dump_service(struct sk_buff *skb,
3023 struct ip_vs_service *svc,
3024 struct netlink_callback *cb)
3028 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3029 &ip_vs_genl_family, NLM_F_MULTI,
3030 IPVS_CMD_NEW_SERVICE);
3034 if (ip_vs_genl_fill_service(skb, svc) < 0)
3035 goto nla_put_failure;
3037 genlmsg_end(skb, hdr);
3041 genlmsg_cancel(skb, hdr);
3045 static int ip_vs_genl_dump_services(struct sk_buff *skb,
3046 struct netlink_callback *cb)
3049 int start = cb->args[0];
3050 struct ip_vs_service *svc;
3051 struct net *net = sock_net(skb->sk);
3052 struct netns_ipvs *ipvs = net_ipvs(net);
3054 mutex_lock(&__ip_vs_mutex);
3055 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3056 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
3057 if (++idx <= start || (svc->ipvs != ipvs))
3059 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3061 goto nla_put_failure;
3066 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3067 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
3068 if (++idx <= start || (svc->ipvs != ipvs))
3070 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3072 goto nla_put_failure;
3078 mutex_unlock(&__ip_vs_mutex);
3084 static bool ip_vs_is_af_valid(int af)
3088 #ifdef CONFIG_IP_VS_IPV6
3089 if (af == AF_INET6 && ipv6_mod_enabled())
3095 static int ip_vs_genl_parse_service(struct netns_ipvs *ipvs,
3096 struct ip_vs_service_user_kern *usvc,
3097 struct nlattr *nla, int full_entry,
3098 struct ip_vs_service **ret_svc)
3100 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
3101 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
3102 struct ip_vs_service *svc;
3104 /* Parse mandatory identifying service fields first */
3106 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla,
3107 ip_vs_svc_policy, NULL))
3110 nla_af = attrs[IPVS_SVC_ATTR_AF];
3111 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
3112 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
3113 nla_port = attrs[IPVS_SVC_ATTR_PORT];
3114 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
3116 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3119 memset(usvc, 0, sizeof(*usvc));
3121 usvc->af = nla_get_u16(nla_af);
3122 if (!ip_vs_is_af_valid(usvc->af))
3123 return -EAFNOSUPPORT;
3126 usvc->protocol = IPPROTO_TCP;
3127 usvc->fwmark = nla_get_u32(nla_fwmark);
3129 usvc->protocol = nla_get_u16(nla_protocol);
3130 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3131 usvc->port = nla_get_be16(nla_port);
3137 svc = __ip_vs_svc_fwm_find(ipvs, usvc->af, usvc->fwmark);
3139 svc = __ip_vs_service_find(ipvs, usvc->af, usvc->protocol,
3140 &usvc->addr, usvc->port);
3144 /* If a full entry was requested, check for the additional fields */
3146 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3148 struct ip_vs_flags flags;
3150 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3151 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3152 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3153 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3154 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3156 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3159 nla_memcpy(&flags, nla_flags, sizeof(flags));
3161 /* prefill flags from service if it already exists */
3163 usvc->flags = svc->flags;
3165 /* set new flags from userland */
3166 usvc->flags = (usvc->flags & ~flags.mask) |
3167 (flags.flags & flags.mask);
3168 usvc->sched_name = nla_data(nla_sched);
3169 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3170 usvc->timeout = nla_get_u32(nla_timeout);
3171 usvc->netmask = nla_get_be32(nla_netmask);
3177 static struct ip_vs_service *ip_vs_genl_find_service(struct netns_ipvs *ipvs,
3180 struct ip_vs_service_user_kern usvc;
3181 struct ip_vs_service *svc;
3184 ret = ip_vs_genl_parse_service(ipvs, &usvc, nla, 0, &svc);
3185 return ret ? ERR_PTR(ret) : svc;
3188 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3190 struct nlattr *nl_dest;
3191 struct ip_vs_kstats kstats;
3193 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3197 if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3198 nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3199 nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3200 (atomic_read(&dest->conn_flags) &
3201 IP_VS_CONN_F_FWD_MASK)) ||
3202 nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3203 atomic_read(&dest->weight)) ||
3204 nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3205 nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3206 nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3207 atomic_read(&dest->activeconns)) ||
3208 nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3209 atomic_read(&dest->inactconns)) ||
3210 nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3211 atomic_read(&dest->persistconns)) ||
3212 nla_put_u16(skb, IPVS_DEST_ATTR_ADDR_FAMILY, dest->af))
3213 goto nla_put_failure;
3214 ip_vs_copy_stats(&kstats, &dest->stats);
3215 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &kstats))
3216 goto nla_put_failure;
3217 if (ip_vs_genl_fill_stats64(skb, IPVS_DEST_ATTR_STATS64, &kstats))
3218 goto nla_put_failure;
3220 nla_nest_end(skb, nl_dest);
3225 nla_nest_cancel(skb, nl_dest);
3229 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3230 struct netlink_callback *cb)
3234 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3235 &ip_vs_genl_family, NLM_F_MULTI,
3240 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3241 goto nla_put_failure;
3243 genlmsg_end(skb, hdr);
3247 genlmsg_cancel(skb, hdr);
3251 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3252 struct netlink_callback *cb)
3255 int start = cb->args[0];
3256 struct ip_vs_service *svc;
3257 struct ip_vs_dest *dest;
3258 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3259 struct net *net = sock_net(skb->sk);
3260 struct netns_ipvs *ipvs = net_ipvs(net);
3262 mutex_lock(&__ip_vs_mutex);
3264 /* Try to find the service for which to dump destinations */
3265 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs, IPVS_CMD_ATTR_MAX,
3266 ip_vs_cmd_policy, NULL))
3270 svc = ip_vs_genl_find_service(ipvs, attrs[IPVS_CMD_ATTR_SERVICE]);
3271 if (IS_ERR_OR_NULL(svc))
3274 /* Dump the destinations */
3275 list_for_each_entry(dest, &svc->destinations, n_list) {
3278 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3280 goto nla_put_failure;
3288 mutex_unlock(&__ip_vs_mutex);
3293 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3294 struct nlattr *nla, int full_entry)
3296 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3297 struct nlattr *nla_addr, *nla_port;
3298 struct nlattr *nla_addr_family;
3300 /* Parse mandatory identifying destination fields first */
3302 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla,
3303 ip_vs_dest_policy, NULL))
3306 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3307 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3308 nla_addr_family = attrs[IPVS_DEST_ATTR_ADDR_FAMILY];
3310 if (!(nla_addr && nla_port))
3313 memset(udest, 0, sizeof(*udest));
3315 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3316 udest->port = nla_get_be16(nla_port);
3318 if (nla_addr_family)
3319 udest->af = nla_get_u16(nla_addr_family);
3323 /* If a full entry was requested, check for the additional fields */
3325 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3328 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3329 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3330 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3331 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3333 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3336 udest->conn_flags = nla_get_u32(nla_fwd)
3337 & IP_VS_CONN_F_FWD_MASK;
3338 udest->weight = nla_get_u32(nla_weight);
3339 udest->u_threshold = nla_get_u32(nla_u_thresh);
3340 udest->l_threshold = nla_get_u32(nla_l_thresh);
3346 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3347 struct ipvs_sync_daemon_cfg *c)
3349 struct nlattr *nl_daemon;
3351 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3355 if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3356 nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, c->mcast_ifn) ||
3357 nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, c->syncid) ||
3358 nla_put_u16(skb, IPVS_DAEMON_ATTR_SYNC_MAXLEN, c->sync_maxlen) ||
3359 nla_put_u16(skb, IPVS_DAEMON_ATTR_MCAST_PORT, c->mcast_port) ||
3360 nla_put_u8(skb, IPVS_DAEMON_ATTR_MCAST_TTL, c->mcast_ttl))
3361 goto nla_put_failure;
3362 #ifdef CONFIG_IP_VS_IPV6
3363 if (c->mcast_af == AF_INET6) {
3364 if (nla_put_in6_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP6,
3365 &c->mcast_group.in6))
3366 goto nla_put_failure;
3369 if (c->mcast_af == AF_INET &&
3370 nla_put_in_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP,
3372 goto nla_put_failure;
3373 nla_nest_end(skb, nl_daemon);
3378 nla_nest_cancel(skb, nl_daemon);
3382 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3383 struct ipvs_sync_daemon_cfg *c,
3384 struct netlink_callback *cb)
3387 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3388 &ip_vs_genl_family, NLM_F_MULTI,
3389 IPVS_CMD_NEW_DAEMON);
3393 if (ip_vs_genl_fill_daemon(skb, state, c))
3394 goto nla_put_failure;
3396 genlmsg_end(skb, hdr);
3400 genlmsg_cancel(skb, hdr);
3404 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3405 struct netlink_callback *cb)
3407 struct net *net = sock_net(skb->sk);
3408 struct netns_ipvs *ipvs = net_ipvs(net);
3410 mutex_lock(&ipvs->sync_mutex);
3411 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3412 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3413 &ipvs->mcfg, cb) < 0)
3414 goto nla_put_failure;
3419 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3420 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3421 &ipvs->bcfg, cb) < 0)
3422 goto nla_put_failure;
3428 mutex_unlock(&ipvs->sync_mutex);
3433 static int ip_vs_genl_new_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3435 struct ipvs_sync_daemon_cfg c;
3439 memset(&c, 0, sizeof(c));
3440 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3441 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3442 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3444 strlcpy(c.mcast_ifn, nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3445 sizeof(c.mcast_ifn));
3446 c.syncid = nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]);
3448 a = attrs[IPVS_DAEMON_ATTR_SYNC_MAXLEN];
3450 c.sync_maxlen = nla_get_u16(a);
3452 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP];
3454 c.mcast_af = AF_INET;
3455 c.mcast_group.ip = nla_get_in_addr(a);
3456 if (!ipv4_is_multicast(c.mcast_group.ip))
3459 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP6];
3461 #ifdef CONFIG_IP_VS_IPV6
3464 c.mcast_af = AF_INET6;
3465 c.mcast_group.in6 = nla_get_in6_addr(a);
3466 addr_type = ipv6_addr_type(&c.mcast_group.in6);
3467 if (!(addr_type & IPV6_ADDR_MULTICAST))
3470 return -EAFNOSUPPORT;
3475 a = attrs[IPVS_DAEMON_ATTR_MCAST_PORT];
3477 c.mcast_port = nla_get_u16(a);
3479 a = attrs[IPVS_DAEMON_ATTR_MCAST_TTL];
3481 c.mcast_ttl = nla_get_u8(a);
3483 /* The synchronization protocol is incompatible with mixed family
3486 if (ipvs->mixed_address_family_dests > 0)
3489 ret = start_sync_thread(ipvs, &c,
3490 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3494 static int ip_vs_genl_del_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3498 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3501 mutex_lock(&ipvs->sync_mutex);
3502 ret = stop_sync_thread(ipvs,
3503 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3504 mutex_unlock(&ipvs->sync_mutex);
3508 static int ip_vs_genl_set_config(struct netns_ipvs *ipvs, struct nlattr **attrs)
3510 struct ip_vs_timeout_user t;
3512 __ip_vs_get_timeouts(ipvs, &t);
3514 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3515 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3517 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3519 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3521 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3522 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3524 return ip_vs_set_timeout(ipvs, &t);
3527 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3529 int ret = -EINVAL, cmd;
3530 struct net *net = sock_net(skb->sk);
3531 struct netns_ipvs *ipvs = net_ipvs(net);
3533 cmd = info->genlhdr->cmd;
3535 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3536 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3538 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3539 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3540 info->attrs[IPVS_CMD_ATTR_DAEMON],
3541 ip_vs_daemon_policy, info->extack))
3544 if (cmd == IPVS_CMD_NEW_DAEMON)
3545 ret = ip_vs_genl_new_daemon(ipvs, daemon_attrs);
3547 ret = ip_vs_genl_del_daemon(ipvs, daemon_attrs);
3554 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3556 struct ip_vs_service *svc = NULL;
3557 struct ip_vs_service_user_kern usvc;
3558 struct ip_vs_dest_user_kern udest;
3560 int need_full_svc = 0, need_full_dest = 0;
3561 struct net *net = sock_net(skb->sk);
3562 struct netns_ipvs *ipvs = net_ipvs(net);
3564 cmd = info->genlhdr->cmd;
3566 mutex_lock(&__ip_vs_mutex);
3568 if (cmd == IPVS_CMD_FLUSH) {
3569 ret = ip_vs_flush(ipvs, false);
3571 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3572 ret = ip_vs_genl_set_config(ipvs, info->attrs);
3574 } else if (cmd == IPVS_CMD_ZERO &&
3575 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3576 ret = ip_vs_zero_all(ipvs);
3580 /* All following commands require a service argument, so check if we
3581 * received a valid one. We need a full service specification when
3582 * adding / editing a service. Only identifying members otherwise. */
3583 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3586 ret = ip_vs_genl_parse_service(ipvs, &usvc,
3587 info->attrs[IPVS_CMD_ATTR_SERVICE],
3588 need_full_svc, &svc);
3592 /* Unless we're adding a new service, the service must already exist */
3593 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3598 /* Destination commands require a valid destination argument. For
3599 * adding / editing a destination, we need a full destination
3601 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3602 cmd == IPVS_CMD_DEL_DEST) {
3603 if (cmd != IPVS_CMD_DEL_DEST)
3606 ret = ip_vs_genl_parse_dest(&udest,
3607 info->attrs[IPVS_CMD_ATTR_DEST],
3612 /* Old protocols did not allow the user to specify address
3613 * family, so we set it to zero instead. We also didn't
3614 * allow heterogeneous pools in the old code, so it's safe
3615 * to assume that this will have the same address family as
3621 if (!ip_vs_is_af_valid(udest.af)) {
3622 ret = -EAFNOSUPPORT;
3626 if (udest.af != svc->af && cmd != IPVS_CMD_DEL_DEST) {
3627 /* The synchronization protocol is incompatible
3628 * with mixed family services
3630 if (ipvs->sync_state) {
3635 /* Which connection types do we support? */
3636 switch (udest.conn_flags) {
3637 case IP_VS_CONN_F_TUNNEL:
3638 /* We are able to forward this */
3648 case IPVS_CMD_NEW_SERVICE:
3650 ret = ip_vs_add_service(ipvs, &usvc, &svc);
3654 case IPVS_CMD_SET_SERVICE:
3655 ret = ip_vs_edit_service(svc, &usvc);
3657 case IPVS_CMD_DEL_SERVICE:
3658 ret = ip_vs_del_service(svc);
3659 /* do not use svc, it can be freed */
3661 case IPVS_CMD_NEW_DEST:
3662 ret = ip_vs_add_dest(svc, &udest);
3664 case IPVS_CMD_SET_DEST:
3665 ret = ip_vs_edit_dest(svc, &udest);
3667 case IPVS_CMD_DEL_DEST:
3668 ret = ip_vs_del_dest(svc, &udest);
3671 ret = ip_vs_zero_service(svc);
3678 mutex_unlock(&__ip_vs_mutex);
3683 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3685 struct sk_buff *msg;
3687 int ret, cmd, reply_cmd;
3688 struct net *net = sock_net(skb->sk);
3689 struct netns_ipvs *ipvs = net_ipvs(net);
3691 cmd = info->genlhdr->cmd;
3693 if (cmd == IPVS_CMD_GET_SERVICE)
3694 reply_cmd = IPVS_CMD_NEW_SERVICE;
3695 else if (cmd == IPVS_CMD_GET_INFO)
3696 reply_cmd = IPVS_CMD_SET_INFO;
3697 else if (cmd == IPVS_CMD_GET_CONFIG)
3698 reply_cmd = IPVS_CMD_SET_CONFIG;
3700 pr_err("unknown Generic Netlink command\n");
3704 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3708 mutex_lock(&__ip_vs_mutex);
3710 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3712 goto nla_put_failure;
3715 case IPVS_CMD_GET_SERVICE:
3717 struct ip_vs_service *svc;
3719 svc = ip_vs_genl_find_service(ipvs,
3720 info->attrs[IPVS_CMD_ATTR_SERVICE]);
3725 ret = ip_vs_genl_fill_service(msg, svc);
3727 goto nla_put_failure;
3736 case IPVS_CMD_GET_CONFIG:
3738 struct ip_vs_timeout_user t;
3740 __ip_vs_get_timeouts(ipvs, &t);
3741 #ifdef CONFIG_IP_VS_PROTO_TCP
3742 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3744 nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3746 goto nla_put_failure;
3748 #ifdef CONFIG_IP_VS_PROTO_UDP
3749 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3750 goto nla_put_failure;
3756 case IPVS_CMD_GET_INFO:
3757 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3758 IP_VS_VERSION_CODE) ||
3759 nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3760 ip_vs_conn_tab_size))
3761 goto nla_put_failure;
3765 genlmsg_end(msg, reply);
3766 ret = genlmsg_reply(msg, info);
3770 pr_err("not enough space in Netlink message\n");
3776 mutex_unlock(&__ip_vs_mutex);
3782 static const struct genl_ops ip_vs_genl_ops[] = {
3784 .cmd = IPVS_CMD_NEW_SERVICE,
3785 .flags = GENL_ADMIN_PERM,
3786 .policy = ip_vs_cmd_policy,
3787 .doit = ip_vs_genl_set_cmd,
3790 .cmd = IPVS_CMD_SET_SERVICE,
3791 .flags = GENL_ADMIN_PERM,
3792 .policy = ip_vs_cmd_policy,
3793 .doit = ip_vs_genl_set_cmd,
3796 .cmd = IPVS_CMD_DEL_SERVICE,
3797 .flags = GENL_ADMIN_PERM,
3798 .policy = ip_vs_cmd_policy,
3799 .doit = ip_vs_genl_set_cmd,
3802 .cmd = IPVS_CMD_GET_SERVICE,
3803 .flags = GENL_ADMIN_PERM,
3804 .doit = ip_vs_genl_get_cmd,
3805 .dumpit = ip_vs_genl_dump_services,
3806 .policy = ip_vs_cmd_policy,
3809 .cmd = IPVS_CMD_NEW_DEST,
3810 .flags = GENL_ADMIN_PERM,
3811 .policy = ip_vs_cmd_policy,
3812 .doit = ip_vs_genl_set_cmd,
3815 .cmd = IPVS_CMD_SET_DEST,
3816 .flags = GENL_ADMIN_PERM,
3817 .policy = ip_vs_cmd_policy,
3818 .doit = ip_vs_genl_set_cmd,
3821 .cmd = IPVS_CMD_DEL_DEST,
3822 .flags = GENL_ADMIN_PERM,
3823 .policy = ip_vs_cmd_policy,
3824 .doit = ip_vs_genl_set_cmd,
3827 .cmd = IPVS_CMD_GET_DEST,
3828 .flags = GENL_ADMIN_PERM,
3829 .policy = ip_vs_cmd_policy,
3830 .dumpit = ip_vs_genl_dump_dests,
3833 .cmd = IPVS_CMD_NEW_DAEMON,
3834 .flags = GENL_ADMIN_PERM,
3835 .policy = ip_vs_cmd_policy,
3836 .doit = ip_vs_genl_set_daemon,
3839 .cmd = IPVS_CMD_DEL_DAEMON,
3840 .flags = GENL_ADMIN_PERM,
3841 .policy = ip_vs_cmd_policy,
3842 .doit = ip_vs_genl_set_daemon,
3845 .cmd = IPVS_CMD_GET_DAEMON,
3846 .flags = GENL_ADMIN_PERM,
3847 .dumpit = ip_vs_genl_dump_daemons,
3850 .cmd = IPVS_CMD_SET_CONFIG,
3851 .flags = GENL_ADMIN_PERM,
3852 .policy = ip_vs_cmd_policy,
3853 .doit = ip_vs_genl_set_cmd,
3856 .cmd = IPVS_CMD_GET_CONFIG,
3857 .flags = GENL_ADMIN_PERM,
3858 .doit = ip_vs_genl_get_cmd,
3861 .cmd = IPVS_CMD_GET_INFO,
3862 .flags = GENL_ADMIN_PERM,
3863 .doit = ip_vs_genl_get_cmd,
3866 .cmd = IPVS_CMD_ZERO,
3867 .flags = GENL_ADMIN_PERM,
3868 .policy = ip_vs_cmd_policy,
3869 .doit = ip_vs_genl_set_cmd,
3872 .cmd = IPVS_CMD_FLUSH,
3873 .flags = GENL_ADMIN_PERM,
3874 .doit = ip_vs_genl_set_cmd,
3878 static struct genl_family ip_vs_genl_family __ro_after_init = {
3880 .name = IPVS_GENL_NAME,
3881 .version = IPVS_GENL_VERSION,
3882 .maxattr = IPVS_CMD_ATTR_MAX,
3883 .netnsok = true, /* Make ipvsadm to work on netns */
3884 .module = THIS_MODULE,
3885 .ops = ip_vs_genl_ops,
3886 .n_ops = ARRAY_SIZE(ip_vs_genl_ops),
3889 static int __init ip_vs_genl_register(void)
3891 return genl_register_family(&ip_vs_genl_family);
3894 static void ip_vs_genl_unregister(void)
3896 genl_unregister_family(&ip_vs_genl_family);
3899 /* End of Generic Netlink interface definitions */
3902 * per netns intit/exit func.
3904 #ifdef CONFIG_SYSCTL
3905 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs)
3907 struct net *net = ipvs->net;
3909 struct ctl_table *tbl;
3911 atomic_set(&ipvs->dropentry, 0);
3912 spin_lock_init(&ipvs->dropentry_lock);
3913 spin_lock_init(&ipvs->droppacket_lock);
3914 spin_lock_init(&ipvs->securetcp_lock);
3916 if (!net_eq(net, &init_net)) {
3917 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3921 /* Don't export sysctls to unprivileged users */
3922 if (net->user_ns != &init_user_ns)
3923 tbl[0].procname = NULL;
3926 /* Initialize sysctl defaults */
3927 for (idx = 0; idx < ARRAY_SIZE(vs_vars); idx++) {
3928 if (tbl[idx].proc_handler == proc_do_defense_mode)
3929 tbl[idx].extra2 = ipvs;
3932 ipvs->sysctl_amemthresh = 1024;
3933 tbl[idx++].data = &ipvs->sysctl_amemthresh;
3934 ipvs->sysctl_am_droprate = 10;
3935 tbl[idx++].data = &ipvs->sysctl_am_droprate;
3936 tbl[idx++].data = &ipvs->sysctl_drop_entry;
3937 tbl[idx++].data = &ipvs->sysctl_drop_packet;
3938 #ifdef CONFIG_IP_VS_NFCT
3939 tbl[idx++].data = &ipvs->sysctl_conntrack;
3941 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3942 ipvs->sysctl_snat_reroute = 1;
3943 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3944 ipvs->sysctl_sync_ver = 1;
3945 tbl[idx++].data = &ipvs->sysctl_sync_ver;
3946 ipvs->sysctl_sync_ports = 1;
3947 tbl[idx++].data = &ipvs->sysctl_sync_ports;
3948 tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
3949 ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3950 tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3951 ipvs->sysctl_sync_sock_size = 0;
3952 tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3953 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3954 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3955 tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
3956 tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
3957 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3958 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3959 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3960 tbl[idx].data = &ipvs->sysctl_sync_threshold;
3961 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3962 ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3963 tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3964 ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3965 tbl[idx++].data = &ipvs->sysctl_sync_retries;
3966 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3967 ipvs->sysctl_pmtu_disc = 1;
3968 tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
3969 tbl[idx++].data = &ipvs->sysctl_backup_only;
3970 ipvs->sysctl_conn_reuse_mode = 1;
3971 tbl[idx++].data = &ipvs->sysctl_conn_reuse_mode;
3972 tbl[idx++].data = &ipvs->sysctl_schedule_icmp;
3973 tbl[idx++].data = &ipvs->sysctl_ignore_tunneled;
3975 ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3976 if (ipvs->sysctl_hdr == NULL) {
3977 if (!net_eq(net, &init_net))
3981 ip_vs_start_estimator(ipvs, &ipvs->tot_stats);
3982 ipvs->sysctl_tbl = tbl;
3983 /* Schedule defense work */
3984 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3985 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3990 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs)
3992 struct net *net = ipvs->net;
3994 cancel_delayed_work_sync(&ipvs->defense_work);
3995 cancel_work_sync(&ipvs->defense_work.work);
3996 unregister_net_sysctl_table(ipvs->sysctl_hdr);
3997 ip_vs_stop_estimator(ipvs, &ipvs->tot_stats);
3999 if (!net_eq(net, &init_net))
4000 kfree(ipvs->sysctl_tbl);
4005 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs) { return 0; }
4006 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs) { }
4010 static struct notifier_block ip_vs_dst_notifier = {
4011 .notifier_call = ip_vs_dst_event,
4014 int __net_init ip_vs_control_net_init(struct netns_ipvs *ipvs)
4018 /* Initialize rs_table */
4019 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
4020 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
4022 INIT_LIST_HEAD(&ipvs->dest_trash);
4023 spin_lock_init(&ipvs->dest_trash_lock);
4024 timer_setup(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire, 0);
4025 atomic_set(&ipvs->ftpsvc_counter, 0);
4026 atomic_set(&ipvs->nullsvc_counter, 0);
4027 atomic_set(&ipvs->conn_out_counter, 0);
4030 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
4031 if (!ipvs->tot_stats.cpustats)
4034 for_each_possible_cpu(i) {
4035 struct ip_vs_cpu_stats *ipvs_tot_stats;
4036 ipvs_tot_stats = per_cpu_ptr(ipvs->tot_stats.cpustats, i);
4037 u64_stats_init(&ipvs_tot_stats->syncp);
4040 spin_lock_init(&ipvs->tot_stats.lock);
4042 proc_create("ip_vs", 0, ipvs->net->proc_net, &ip_vs_info_fops);
4043 proc_create("ip_vs_stats", 0, ipvs->net->proc_net, &ip_vs_stats_fops);
4044 proc_create("ip_vs_stats_percpu", 0, ipvs->net->proc_net,
4045 &ip_vs_stats_percpu_fops);
4047 if (ip_vs_control_net_init_sysctl(ipvs))
4053 free_percpu(ipvs->tot_stats.cpustats);
4057 void __net_exit ip_vs_control_net_cleanup(struct netns_ipvs *ipvs)
4059 ip_vs_trash_cleanup(ipvs);
4060 ip_vs_control_net_cleanup_sysctl(ipvs);
4061 remove_proc_entry("ip_vs_stats_percpu", ipvs->net->proc_net);
4062 remove_proc_entry("ip_vs_stats", ipvs->net->proc_net);
4063 remove_proc_entry("ip_vs", ipvs->net->proc_net);
4064 free_percpu(ipvs->tot_stats.cpustats);
4067 int __init ip_vs_register_nl_ioctl(void)
4071 ret = nf_register_sockopt(&ip_vs_sockopts);
4073 pr_err("cannot register sockopt.\n");
4077 ret = ip_vs_genl_register();
4079 pr_err("cannot register Generic Netlink interface.\n");
4085 nf_unregister_sockopt(&ip_vs_sockopts);
4090 void ip_vs_unregister_nl_ioctl(void)
4092 ip_vs_genl_unregister();
4093 nf_unregister_sockopt(&ip_vs_sockopts);
4096 int __init ip_vs_control_init(void)
4103 /* Initialize svc_table, ip_vs_svc_fwm_table */
4104 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
4105 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
4106 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
4109 smp_wmb(); /* Do we really need it now ? */
4111 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
4120 void ip_vs_control_cleanup(void)
4123 unregister_netdevice_notifier(&ip_vs_dst_notifier);