2 * IPVS An implementation of the IP virtual server support for the
3 * LINUX operating system. IPVS is now implemented as a module
4 * over the NetFilter framework. IPVS can be used to build a
5 * high-performance and highly available server based on a
8 * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
9 * Peter Kese <peter.kese@ijs.si>
10 * Julian Anastasov <ja@ssi.bg>
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
21 #define KMSG_COMPONENT "IPVS"
22 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
24 #include <linux/module.h>
25 #include <linux/init.h>
26 #include <linux/types.h>
27 #include <linux/capability.h>
29 #include <linux/sysctl.h>
30 #include <linux/proc_fs.h>
31 #include <linux/workqueue.h>
32 #include <linux/swap.h>
33 #include <linux/seq_file.h>
34 #include <linux/slab.h>
36 #include <linux/netfilter.h>
37 #include <linux/netfilter_ipv4.h>
38 #include <linux/mutex.h>
40 #include <net/net_namespace.h>
41 #include <linux/nsproxy.h>
43 #ifdef CONFIG_IP_VS_IPV6
45 #include <net/ip6_route.h>
46 #include <net/netfilter/ipv6/nf_defrag_ipv6.h>
48 #include <net/route.h>
50 #include <net/genetlink.h>
52 #include <linux/uaccess.h>
54 #include <net/ip_vs.h>
56 /* semaphore for IPVS sockopts. And, [gs]etsockopt may sleep. */
57 static DEFINE_MUTEX(__ip_vs_mutex);
59 /* sysctl variables */
61 #ifdef CONFIG_IP_VS_DEBUG
62 static int sysctl_ip_vs_debug_level = 0;
64 int ip_vs_get_debug_level(void)
66 return sysctl_ip_vs_debug_level;
72 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup);
75 #ifdef CONFIG_IP_VS_IPV6
76 /* Taken from rt6_fill_node() in net/ipv6/route.c, is there a better way? */
77 static bool __ip_vs_addr_is_local_v6(struct net *net,
78 const struct in6_addr *addr)
83 struct dst_entry *dst = ip6_route_output(net, NULL, &fl6);
86 is_local = !dst->error && dst->dev && (dst->dev->flags & IFF_LOOPBACK);
95 * update_defense_level is called from keventd and from sysctl,
96 * so it needs to protect itself from softirqs
98 static void update_defense_level(struct netns_ipvs *ipvs)
105 /* we only count free and buffered memory (in pages) */
107 availmem = i.freeram + i.bufferram;
108 /* however in linux 2.5 the i.bufferram is total page cache size,
110 /* si_swapinfo(&i); */
111 /* availmem = availmem - (i.totalswap - i.freeswap); */
113 nomem = (availmem < ipvs->sysctl_amemthresh);
118 spin_lock(&ipvs->dropentry_lock);
119 switch (ipvs->sysctl_drop_entry) {
121 atomic_set(&ipvs->dropentry, 0);
125 atomic_set(&ipvs->dropentry, 1);
126 ipvs->sysctl_drop_entry = 2;
128 atomic_set(&ipvs->dropentry, 0);
133 atomic_set(&ipvs->dropentry, 1);
135 atomic_set(&ipvs->dropentry, 0);
136 ipvs->sysctl_drop_entry = 1;
140 atomic_set(&ipvs->dropentry, 1);
143 spin_unlock(&ipvs->dropentry_lock);
146 spin_lock(&ipvs->droppacket_lock);
147 switch (ipvs->sysctl_drop_packet) {
153 ipvs->drop_rate = ipvs->drop_counter
154 = ipvs->sysctl_amemthresh /
155 (ipvs->sysctl_amemthresh-availmem);
156 ipvs->sysctl_drop_packet = 2;
163 ipvs->drop_rate = ipvs->drop_counter
164 = ipvs->sysctl_amemthresh /
165 (ipvs->sysctl_amemthresh-availmem);
168 ipvs->sysctl_drop_packet = 1;
172 ipvs->drop_rate = ipvs->sysctl_am_droprate;
175 spin_unlock(&ipvs->droppacket_lock);
178 spin_lock(&ipvs->securetcp_lock);
179 switch (ipvs->sysctl_secure_tcp) {
181 if (ipvs->old_secure_tcp >= 2)
186 if (ipvs->old_secure_tcp < 2)
188 ipvs->sysctl_secure_tcp = 2;
190 if (ipvs->old_secure_tcp >= 2)
196 if (ipvs->old_secure_tcp < 2)
199 if (ipvs->old_secure_tcp >= 2)
201 ipvs->sysctl_secure_tcp = 1;
205 if (ipvs->old_secure_tcp < 2)
209 ipvs->old_secure_tcp = ipvs->sysctl_secure_tcp;
211 ip_vs_protocol_timeout_change(ipvs,
212 ipvs->sysctl_secure_tcp > 1);
213 spin_unlock(&ipvs->securetcp_lock);
220 * Timer for checking the defense
222 #define DEFENSE_TIMER_PERIOD 1*HZ
224 static void defense_work_handler(struct work_struct *work)
226 struct netns_ipvs *ipvs =
227 container_of(work, struct netns_ipvs, defense_work.work);
229 update_defense_level(ipvs);
230 if (atomic_read(&ipvs->dropentry))
231 ip_vs_random_dropentry(ipvs);
232 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
237 ip_vs_use_count_inc(void)
239 return try_module_get(THIS_MODULE);
243 ip_vs_use_count_dec(void)
245 module_put(THIS_MODULE);
250 * Hash table: for virtual service lookups
252 #define IP_VS_SVC_TAB_BITS 8
253 #define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
254 #define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)
256 /* the service table hashed by <protocol, addr, port> */
257 static struct hlist_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE];
258 /* the service table hashed by fwmark */
259 static struct hlist_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE];
263 * Returns hash value for virtual service
265 static inline unsigned int
266 ip_vs_svc_hashkey(struct netns_ipvs *ipvs, int af, unsigned int proto,
267 const union nf_inet_addr *addr, __be16 port)
269 register unsigned int porth = ntohs(port);
270 __be32 addr_fold = addr->ip;
273 #ifdef CONFIG_IP_VS_IPV6
275 addr_fold = addr->ip6[0]^addr->ip6[1]^
276 addr->ip6[2]^addr->ip6[3];
278 ahash = ntohl(addr_fold);
279 ahash ^= ((size_t) ipvs >> 8);
281 return (proto ^ ahash ^ (porth >> IP_VS_SVC_TAB_BITS) ^ porth) &
286 * Returns hash value of fwmark for virtual service lookup
288 static inline unsigned int ip_vs_svc_fwm_hashkey(struct netns_ipvs *ipvs, __u32 fwmark)
290 return (((size_t)ipvs>>8) ^ fwmark) & IP_VS_SVC_TAB_MASK;
294 * Hashes a service in the ip_vs_svc_table by <netns,proto,addr,port>
295 * or in the ip_vs_svc_fwm_table by fwmark.
296 * Should be called with locked tables.
298 static int ip_vs_svc_hash(struct ip_vs_service *svc)
302 if (svc->flags & IP_VS_SVC_F_HASHED) {
303 pr_err("%s(): request for already hashed, called from %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 /* keep the last_weight with latest non-0 weight */
825 if (add || udest->weight != 0)
826 atomic_set(&dest->last_weight, udest->weight);
828 /* set the weight and the flags */
829 atomic_set(&dest->weight, udest->weight);
830 conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
831 conn_flags |= IP_VS_CONN_F_INACTIVE;
833 /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
834 if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
835 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
838 * Put the real service in rs_table if not present.
839 * For now only for NAT!
841 ip_vs_rs_hash(ipvs, dest);
842 /* FTP-NAT requires conntrack for mangling */
843 if (svc->port == FTPPORT)
844 ip_vs_register_conntrack(svc);
846 atomic_set(&dest->conn_flags, conn_flags);
848 /* bind the service */
849 old_svc = rcu_dereference_protected(dest->svc, 1);
851 __ip_vs_bind_svc(dest, svc);
853 if (old_svc != svc) {
854 ip_vs_zero_stats(&dest->stats);
855 __ip_vs_bind_svc(dest, svc);
856 __ip_vs_svc_put(old_svc, true);
860 /* set the dest status flags */
861 dest->flags |= IP_VS_DEST_F_AVAILABLE;
863 if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
864 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
865 dest->u_threshold = udest->u_threshold;
866 dest->l_threshold = udest->l_threshold;
868 dest->af = udest->af;
870 spin_lock_bh(&dest->dst_lock);
871 __ip_vs_dst_cache_reset(dest);
872 spin_unlock_bh(&dest->dst_lock);
875 ip_vs_start_estimator(svc->ipvs, &dest->stats);
876 list_add_rcu(&dest->n_list, &svc->destinations);
878 sched = rcu_dereference_protected(svc->scheduler, 1);
879 if (sched && sched->add_dest)
880 sched->add_dest(svc, dest);
882 sched = rcu_dereference_protected(svc->scheduler, 1);
883 if (sched && sched->upd_dest)
884 sched->upd_dest(svc, dest);
890 * Create a destination for the given service
893 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
894 struct ip_vs_dest **dest_p)
896 struct ip_vs_dest *dest;
897 unsigned int atype, i;
901 #ifdef CONFIG_IP_VS_IPV6
902 if (udest->af == AF_INET6) {
905 atype = ipv6_addr_type(&udest->addr.in6);
906 if ((!(atype & IPV6_ADDR_UNICAST) ||
907 atype & IPV6_ADDR_LINKLOCAL) &&
908 !__ip_vs_addr_is_local_v6(svc->ipvs->net, &udest->addr.in6))
911 ret = nf_defrag_ipv6_enable(svc->ipvs->net);
917 atype = inet_addr_type(svc->ipvs->net, udest->addr.ip);
918 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
922 dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
926 dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
927 if (!dest->stats.cpustats)
930 for_each_possible_cpu(i) {
931 struct ip_vs_cpu_stats *ip_vs_dest_stats;
932 ip_vs_dest_stats = per_cpu_ptr(dest->stats.cpustats, i);
933 u64_stats_init(&ip_vs_dest_stats->syncp);
936 dest->af = udest->af;
937 dest->protocol = svc->protocol;
938 dest->vaddr = svc->addr;
939 dest->vport = svc->port;
940 dest->vfwmark = svc->fwmark;
941 ip_vs_addr_copy(udest->af, &dest->addr, &udest->addr);
942 dest->port = udest->port;
944 atomic_set(&dest->activeconns, 0);
945 atomic_set(&dest->inactconns, 0);
946 atomic_set(&dest->persistconns, 0);
947 refcount_set(&dest->refcnt, 1);
949 INIT_HLIST_NODE(&dest->d_list);
950 spin_lock_init(&dest->dst_lock);
951 spin_lock_init(&dest->stats.lock);
952 __ip_vs_update_dest(svc, dest, udest, 1);
966 * Add a destination into an existing service
969 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
971 struct ip_vs_dest *dest;
972 union nf_inet_addr daddr;
973 __be16 dport = udest->port;
978 if (udest->weight < 0) {
979 pr_err("%s(): server weight less than zero\n", __func__);
983 if (udest->l_threshold > udest->u_threshold) {
984 pr_err("%s(): lower threshold is higher than upper threshold\n",
989 ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
991 /* We use function that requires RCU lock */
993 dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
997 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
1002 * Check if the dest already exists in the trash and
1003 * is from the same service
1005 dest = ip_vs_trash_get_dest(svc, udest->af, &daddr, dport);
1008 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
1009 "dest->refcnt=%d, service %u/%s:%u\n",
1010 IP_VS_DBG_ADDR(udest->af, &daddr), ntohs(dport),
1011 refcount_read(&dest->refcnt),
1013 IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
1014 ntohs(dest->vport));
1016 __ip_vs_update_dest(svc, dest, udest, 1);
1020 * Allocate and initialize the dest structure
1022 ret = ip_vs_new_dest(svc, udest, &dest);
1031 * Edit a destination in the given service
1034 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1036 struct ip_vs_dest *dest;
1037 union nf_inet_addr daddr;
1038 __be16 dport = udest->port;
1042 if (udest->weight < 0) {
1043 pr_err("%s(): server weight less than zero\n", __func__);
1047 if (udest->l_threshold > udest->u_threshold) {
1048 pr_err("%s(): lower threshold is higher than upper threshold\n",
1053 ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
1055 /* We use function that requires RCU lock */
1057 dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
1061 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1065 __ip_vs_update_dest(svc, dest, udest, 0);
1072 * Delete a destination (must be already unlinked from the service)
1074 static void __ip_vs_del_dest(struct netns_ipvs *ipvs, struct ip_vs_dest *dest,
1077 ip_vs_stop_estimator(ipvs, &dest->stats);
1080 * Remove it from the d-linked list with the real services.
1082 ip_vs_rs_unhash(dest);
1084 spin_lock_bh(&ipvs->dest_trash_lock);
1085 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1086 IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port),
1087 refcount_read(&dest->refcnt));
1088 if (list_empty(&ipvs->dest_trash) && !cleanup)
1089 mod_timer(&ipvs->dest_trash_timer,
1090 jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1091 /* dest lives in trash with reference */
1092 list_add(&dest->t_list, &ipvs->dest_trash);
1093 dest->idle_start = 0;
1094 spin_unlock_bh(&ipvs->dest_trash_lock);
1099 * Unlink a destination from the given service
1101 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1102 struct ip_vs_dest *dest,
1105 dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1108 * Remove it from the d-linked destination list.
1110 list_del_rcu(&dest->n_list);
1113 if (dest->af != svc->af)
1114 svc->ipvs->mixed_address_family_dests--;
1117 struct ip_vs_scheduler *sched;
1119 sched = rcu_dereference_protected(svc->scheduler, 1);
1120 if (sched && sched->del_dest)
1121 sched->del_dest(svc, dest);
1127 * Delete a destination server in the given service
1130 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1132 struct ip_vs_dest *dest;
1133 __be16 dport = udest->port;
1137 /* We use function that requires RCU lock */
1139 dest = ip_vs_lookup_dest(svc, udest->af, &udest->addr, dport);
1143 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1148 * Unlink dest from the service
1150 __ip_vs_unlink_dest(svc, dest, 1);
1153 * Delete the destination
1155 __ip_vs_del_dest(svc->ipvs, dest, false);
1162 static void ip_vs_dest_trash_expire(struct timer_list *t)
1164 struct netns_ipvs *ipvs = from_timer(ipvs, t, dest_trash_timer);
1165 struct ip_vs_dest *dest, *next;
1166 unsigned long now = jiffies;
1168 spin_lock(&ipvs->dest_trash_lock);
1169 list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) {
1170 if (refcount_read(&dest->refcnt) > 1)
1172 if (dest->idle_start) {
1173 if (time_before(now, dest->idle_start +
1174 IP_VS_DEST_TRASH_PERIOD))
1177 dest->idle_start = max(1UL, now);
1180 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1182 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1184 list_del(&dest->t_list);
1185 ip_vs_dest_free(dest);
1187 if (!list_empty(&ipvs->dest_trash))
1188 mod_timer(&ipvs->dest_trash_timer,
1189 jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1190 spin_unlock(&ipvs->dest_trash_lock);
1194 * Add a service into the service hash table
1197 ip_vs_add_service(struct netns_ipvs *ipvs, struct ip_vs_service_user_kern *u,
1198 struct ip_vs_service **svc_p)
1201 struct ip_vs_scheduler *sched = NULL;
1202 struct ip_vs_pe *pe = NULL;
1203 struct ip_vs_service *svc = NULL;
1205 /* increase the module use count */
1206 if (!ip_vs_use_count_inc())
1207 return -ENOPROTOOPT;
1209 /* Lookup the scheduler by 'u->sched_name' */
1210 if (strcmp(u->sched_name, "none")) {
1211 sched = ip_vs_scheduler_get(u->sched_name);
1213 pr_info("Scheduler module ip_vs_%s not found\n",
1220 if (u->pe_name && *u->pe_name) {
1221 pe = ip_vs_pe_getbyname(u->pe_name);
1223 pr_info("persistence engine module ip_vs_pe_%s "
1224 "not found\n", u->pe_name);
1230 #ifdef CONFIG_IP_VS_IPV6
1231 if (u->af == AF_INET6) {
1232 __u32 plen = (__force __u32) u->netmask;
1234 if (plen < 1 || plen > 128) {
1239 ret = nf_defrag_ipv6_enable(ipvs->net);
1245 svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1247 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1251 svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1252 if (!svc->stats.cpustats) {
1257 for_each_possible_cpu(i) {
1258 struct ip_vs_cpu_stats *ip_vs_stats;
1259 ip_vs_stats = per_cpu_ptr(svc->stats.cpustats, i);
1260 u64_stats_init(&ip_vs_stats->syncp);
1264 /* I'm the first user of the service */
1265 atomic_set(&svc->refcnt, 0);
1268 svc->protocol = u->protocol;
1269 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1270 svc->port = u->port;
1271 svc->fwmark = u->fwmark;
1272 svc->flags = u->flags & ~IP_VS_SVC_F_HASHED;
1273 svc->timeout = u->timeout * HZ;
1274 svc->netmask = u->netmask;
1277 INIT_LIST_HEAD(&svc->destinations);
1278 spin_lock_init(&svc->sched_lock);
1279 spin_lock_init(&svc->stats.lock);
1281 /* Bind the scheduler */
1283 ret = ip_vs_bind_scheduler(svc, sched);
1289 /* Bind the ct retriever */
1290 RCU_INIT_POINTER(svc->pe, pe);
1293 /* Update the virtual service counters */
1294 if (svc->port == FTPPORT)
1295 atomic_inc(&ipvs->ftpsvc_counter);
1296 else if (svc->port == 0)
1297 atomic_inc(&ipvs->nullsvc_counter);
1298 if (svc->pe && svc->pe->conn_out)
1299 atomic_inc(&ipvs->conn_out_counter);
1301 ip_vs_start_estimator(ipvs, &svc->stats);
1303 /* Count only IPv4 services for old get/setsockopt interface */
1304 if (svc->af == AF_INET)
1305 ipvs->num_services++;
1307 /* Hash the service into the service table */
1308 ip_vs_svc_hash(svc);
1311 /* Now there is a service - full throttle */
1318 ip_vs_unbind_scheduler(svc, sched);
1319 ip_vs_service_free(svc);
1321 ip_vs_scheduler_put(sched);
1324 /* decrease the module use count */
1325 ip_vs_use_count_dec();
1332 * Edit a service and bind it with a new scheduler
1335 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1337 struct ip_vs_scheduler *sched = NULL, *old_sched;
1338 struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1340 bool new_pe_conn_out, old_pe_conn_out;
1343 * Lookup the scheduler, by 'u->sched_name'
1345 if (strcmp(u->sched_name, "none")) {
1346 sched = ip_vs_scheduler_get(u->sched_name);
1348 pr_info("Scheduler module ip_vs_%s not found\n",
1355 if (u->pe_name && *u->pe_name) {
1356 pe = ip_vs_pe_getbyname(u->pe_name);
1358 pr_info("persistence engine module ip_vs_pe_%s "
1359 "not found\n", u->pe_name);
1366 #ifdef CONFIG_IP_VS_IPV6
1367 if (u->af == AF_INET6) {
1368 __u32 plen = (__force __u32) u->netmask;
1370 if (plen < 1 || plen > 128) {
1377 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1378 if (sched != old_sched) {
1380 ip_vs_unbind_scheduler(svc, old_sched);
1381 RCU_INIT_POINTER(svc->scheduler, NULL);
1382 /* Wait all svc->sched_data users */
1385 /* Bind the new scheduler */
1387 ret = ip_vs_bind_scheduler(svc, sched);
1389 ip_vs_scheduler_put(sched);
1396 * Set the flags and timeout value
1398 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1399 svc->timeout = u->timeout * HZ;
1400 svc->netmask = u->netmask;
1402 old_pe = rcu_dereference_protected(svc->pe, 1);
1404 rcu_assign_pointer(svc->pe, pe);
1405 /* check for optional methods in new pe */
1406 new_pe_conn_out = (pe && pe->conn_out) ? true : false;
1407 old_pe_conn_out = (old_pe && old_pe->conn_out) ? true : false;
1408 if (new_pe_conn_out && !old_pe_conn_out)
1409 atomic_inc(&svc->ipvs->conn_out_counter);
1410 if (old_pe_conn_out && !new_pe_conn_out)
1411 atomic_dec(&svc->ipvs->conn_out_counter);
1415 ip_vs_scheduler_put(old_sched);
1416 ip_vs_pe_put(old_pe);
1421 * Delete a service from the service list
1422 * - The service must be unlinked, unlocked and not referenced!
1423 * - We are called under _bh lock
1425 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1427 struct ip_vs_dest *dest, *nxt;
1428 struct ip_vs_scheduler *old_sched;
1429 struct ip_vs_pe *old_pe;
1430 struct netns_ipvs *ipvs = svc->ipvs;
1432 /* Count only IPv4 services for old get/setsockopt interface */
1433 if (svc->af == AF_INET)
1434 ipvs->num_services--;
1436 ip_vs_stop_estimator(svc->ipvs, &svc->stats);
1438 /* Unbind scheduler */
1439 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1440 ip_vs_unbind_scheduler(svc, old_sched);
1441 ip_vs_scheduler_put(old_sched);
1443 /* Unbind persistence engine, keep svc->pe */
1444 old_pe = rcu_dereference_protected(svc->pe, 1);
1445 if (old_pe && old_pe->conn_out)
1446 atomic_dec(&ipvs->conn_out_counter);
1447 ip_vs_pe_put(old_pe);
1450 * Unlink the whole destination list
1452 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1453 __ip_vs_unlink_dest(svc, dest, 0);
1454 __ip_vs_del_dest(svc->ipvs, dest, cleanup);
1458 * Update the virtual service counters
1460 if (svc->port == FTPPORT)
1461 atomic_dec(&ipvs->ftpsvc_counter);
1462 else if (svc->port == 0)
1463 atomic_dec(&ipvs->nullsvc_counter);
1466 * Free the service if nobody refers to it
1468 __ip_vs_svc_put(svc, true);
1470 /* decrease the module use count */
1471 ip_vs_use_count_dec();
1475 * Unlink a service from list and try to delete it if its refcnt reached 0
1477 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1479 ip_vs_unregister_conntrack(svc);
1480 /* Hold svc to avoid double release from dest_trash */
1481 atomic_inc(&svc->refcnt);
1483 * Unhash it from the service table
1485 ip_vs_svc_unhash(svc);
1487 __ip_vs_del_service(svc, cleanup);
1491 * Delete a service from the service list
1493 static int ip_vs_del_service(struct ip_vs_service *svc)
1497 ip_vs_unlink_service(svc, false);
1504 * Flush all the virtual services
1506 static int ip_vs_flush(struct netns_ipvs *ipvs, bool cleanup)
1509 struct ip_vs_service *svc;
1510 struct hlist_node *n;
1513 * Flush the service table hashed by <netns,protocol,addr,port>
1515 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1516 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1518 if (svc->ipvs == ipvs)
1519 ip_vs_unlink_service(svc, cleanup);
1524 * Flush the service table hashed by fwmark
1526 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1527 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1529 if (svc->ipvs == ipvs)
1530 ip_vs_unlink_service(svc, cleanup);
1538 * Delete service by {netns} in the service table.
1539 * Called by __ip_vs_cleanup()
1541 void ip_vs_service_net_cleanup(struct netns_ipvs *ipvs)
1544 /* Check for "full" addressed entries */
1545 mutex_lock(&__ip_vs_mutex);
1546 ip_vs_flush(ipvs, true);
1547 mutex_unlock(&__ip_vs_mutex);
1551 /* Put all references for device (dst_cache) */
1553 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1555 struct ip_vs_dest_dst *dest_dst;
1557 spin_lock_bh(&dest->dst_lock);
1558 dest_dst = rcu_dereference_protected(dest->dest_dst, 1);
1559 if (dest_dst && dest_dst->dst_cache->dev == dev) {
1560 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1562 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1564 refcount_read(&dest->refcnt));
1565 __ip_vs_dst_cache_reset(dest);
1567 spin_unlock_bh(&dest->dst_lock);
1570 /* Netdev event receiver
1571 * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1573 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1576 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1577 struct net *net = dev_net(dev);
1578 struct netns_ipvs *ipvs = net_ipvs(net);
1579 struct ip_vs_service *svc;
1580 struct ip_vs_dest *dest;
1583 if (event != NETDEV_DOWN || !ipvs)
1585 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1587 mutex_lock(&__ip_vs_mutex);
1588 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1589 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1590 if (svc->ipvs == ipvs) {
1591 list_for_each_entry(dest, &svc->destinations,
1593 ip_vs_forget_dev(dest, dev);
1598 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1599 if (svc->ipvs == ipvs) {
1600 list_for_each_entry(dest, &svc->destinations,
1602 ip_vs_forget_dev(dest, dev);
1609 spin_lock_bh(&ipvs->dest_trash_lock);
1610 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1611 ip_vs_forget_dev(dest, dev);
1613 spin_unlock_bh(&ipvs->dest_trash_lock);
1614 mutex_unlock(&__ip_vs_mutex);
1620 * Zero counters in a service or all services
1622 static int ip_vs_zero_service(struct ip_vs_service *svc)
1624 struct ip_vs_dest *dest;
1626 list_for_each_entry(dest, &svc->destinations, n_list) {
1627 ip_vs_zero_stats(&dest->stats);
1629 ip_vs_zero_stats(&svc->stats);
1633 static int ip_vs_zero_all(struct netns_ipvs *ipvs)
1636 struct ip_vs_service *svc;
1638 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1639 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1640 if (svc->ipvs == ipvs)
1641 ip_vs_zero_service(svc);
1645 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1646 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1647 if (svc->ipvs == ipvs)
1648 ip_vs_zero_service(svc);
1652 ip_vs_zero_stats(&ipvs->tot_stats);
1656 #ifdef CONFIG_SYSCTL
1660 static int three = 3;
1663 proc_do_defense_mode(struct ctl_table *table, int write,
1664 void __user *buffer, size_t *lenp, loff_t *ppos)
1666 struct netns_ipvs *ipvs = table->extra2;
1667 int *valp = table->data;
1671 struct ctl_table tmp = {
1673 .maxlen = sizeof(int),
1674 .mode = table->mode,
1677 rc = proc_dointvec(&tmp, write, buffer, lenp, ppos);
1678 if (write && (*valp != val)) {
1679 if (val < 0 || val > 3) {
1683 update_defense_level(ipvs);
1690 proc_do_sync_threshold(struct ctl_table *table, int write,
1691 void __user *buffer, size_t *lenp, loff_t *ppos)
1693 struct netns_ipvs *ipvs = table->extra2;
1694 int *valp = table->data;
1697 struct ctl_table tmp = {
1699 .maxlen = table->maxlen,
1700 .mode = table->mode,
1703 mutex_lock(&ipvs->sync_mutex);
1704 memcpy(val, valp, sizeof(val));
1705 rc = proc_dointvec(&tmp, write, buffer, lenp, ppos);
1707 if (val[0] < 0 || val[1] < 0 ||
1708 (val[0] >= val[1] && val[1]))
1711 memcpy(valp, val, sizeof(val));
1713 mutex_unlock(&ipvs->sync_mutex);
1718 proc_do_sync_ports(struct ctl_table *table, int write,
1719 void __user *buffer, size_t *lenp, loff_t *ppos)
1721 int *valp = table->data;
1725 struct ctl_table tmp = {
1727 .maxlen = sizeof(int),
1728 .mode = table->mode,
1731 rc = proc_dointvec(&tmp, write, buffer, lenp, ppos);
1732 if (write && (*valp != val)) {
1733 if (val < 1 || !is_power_of_2(val))
1742 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1743 * Do not change order or insert new entries without
1744 * align with netns init in ip_vs_control_net_init()
1747 static struct ctl_table vs_vars[] = {
1749 .procname = "amemthresh",
1750 .maxlen = sizeof(int),
1752 .proc_handler = proc_dointvec,
1755 .procname = "am_droprate",
1756 .maxlen = sizeof(int),
1758 .proc_handler = proc_dointvec,
1761 .procname = "drop_entry",
1762 .maxlen = sizeof(int),
1764 .proc_handler = proc_do_defense_mode,
1767 .procname = "drop_packet",
1768 .maxlen = sizeof(int),
1770 .proc_handler = proc_do_defense_mode,
1772 #ifdef CONFIG_IP_VS_NFCT
1774 .procname = "conntrack",
1775 .maxlen = sizeof(int),
1777 .proc_handler = &proc_dointvec,
1781 .procname = "secure_tcp",
1782 .maxlen = sizeof(int),
1784 .proc_handler = proc_do_defense_mode,
1787 .procname = "snat_reroute",
1788 .maxlen = sizeof(int),
1790 .proc_handler = &proc_dointvec,
1793 .procname = "sync_version",
1794 .maxlen = sizeof(int),
1796 .proc_handler = proc_dointvec_minmax,
1801 .procname = "sync_ports",
1802 .maxlen = sizeof(int),
1804 .proc_handler = proc_do_sync_ports,
1807 .procname = "sync_persist_mode",
1808 .maxlen = sizeof(int),
1810 .proc_handler = proc_dointvec,
1813 .procname = "sync_qlen_max",
1814 .maxlen = sizeof(unsigned long),
1816 .proc_handler = proc_doulongvec_minmax,
1819 .procname = "sync_sock_size",
1820 .maxlen = sizeof(int),
1822 .proc_handler = proc_dointvec,
1825 .procname = "cache_bypass",
1826 .maxlen = sizeof(int),
1828 .proc_handler = proc_dointvec,
1831 .procname = "expire_nodest_conn",
1832 .maxlen = sizeof(int),
1834 .proc_handler = proc_dointvec,
1837 .procname = "sloppy_tcp",
1838 .maxlen = sizeof(int),
1840 .proc_handler = proc_dointvec,
1843 .procname = "sloppy_sctp",
1844 .maxlen = sizeof(int),
1846 .proc_handler = proc_dointvec,
1849 .procname = "expire_quiescent_template",
1850 .maxlen = sizeof(int),
1852 .proc_handler = proc_dointvec,
1855 .procname = "sync_threshold",
1857 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1859 .proc_handler = proc_do_sync_threshold,
1862 .procname = "sync_refresh_period",
1863 .maxlen = sizeof(int),
1865 .proc_handler = proc_dointvec_jiffies,
1868 .procname = "sync_retries",
1869 .maxlen = sizeof(int),
1871 .proc_handler = proc_dointvec_minmax,
1876 .procname = "nat_icmp_send",
1877 .maxlen = sizeof(int),
1879 .proc_handler = proc_dointvec,
1882 .procname = "pmtu_disc",
1883 .maxlen = sizeof(int),
1885 .proc_handler = proc_dointvec,
1888 .procname = "backup_only",
1889 .maxlen = sizeof(int),
1891 .proc_handler = proc_dointvec,
1894 .procname = "conn_reuse_mode",
1895 .maxlen = sizeof(int),
1897 .proc_handler = proc_dointvec,
1900 .procname = "schedule_icmp",
1901 .maxlen = sizeof(int),
1903 .proc_handler = proc_dointvec,
1906 .procname = "ignore_tunneled",
1907 .maxlen = sizeof(int),
1909 .proc_handler = proc_dointvec,
1911 #ifdef CONFIG_IP_VS_DEBUG
1913 .procname = "debug_level",
1914 .data = &sysctl_ip_vs_debug_level,
1915 .maxlen = sizeof(int),
1917 .proc_handler = proc_dointvec,
1925 #ifdef CONFIG_PROC_FS
1928 struct seq_net_private p; /* Do not move this, netns depends upon it*/
1929 struct hlist_head *table;
1934 * Write the contents of the VS rule table to a PROCfs file.
1935 * (It is kept just for backward compatibility)
1937 static inline const char *ip_vs_fwd_name(unsigned int flags)
1939 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1940 case IP_VS_CONN_F_LOCALNODE:
1942 case IP_VS_CONN_F_TUNNEL:
1944 case IP_VS_CONN_F_DROUTE:
1952 /* Get the Nth entry in the two lists */
1953 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1955 struct net *net = seq_file_net(seq);
1956 struct netns_ipvs *ipvs = net_ipvs(net);
1957 struct ip_vs_iter *iter = seq->private;
1959 struct ip_vs_service *svc;
1961 /* look in hash by protocol */
1962 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1963 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
1964 if ((svc->ipvs == ipvs) && pos-- == 0) {
1965 iter->table = ip_vs_svc_table;
1972 /* keep looking in fwmark */
1973 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1974 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
1976 if ((svc->ipvs == ipvs) && pos-- == 0) {
1977 iter->table = ip_vs_svc_fwm_table;
1987 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1991 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1995 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1997 struct hlist_node *e;
1998 struct ip_vs_iter *iter;
1999 struct ip_vs_service *svc;
2002 if (v == SEQ_START_TOKEN)
2003 return ip_vs_info_array(seq,0);
2006 iter = seq->private;
2008 if (iter->table == ip_vs_svc_table) {
2009 /* next service in table hashed by protocol */
2010 e = rcu_dereference(hlist_next_rcu(&svc->s_list));
2012 return hlist_entry(e, struct ip_vs_service, s_list);
2014 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2015 hlist_for_each_entry_rcu(svc,
2016 &ip_vs_svc_table[iter->bucket],
2022 iter->table = ip_vs_svc_fwm_table;
2027 /* next service in hashed by fwmark */
2028 e = rcu_dereference(hlist_next_rcu(&svc->f_list));
2030 return hlist_entry(e, struct ip_vs_service, f_list);
2033 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2034 hlist_for_each_entry_rcu(svc,
2035 &ip_vs_svc_fwm_table[iter->bucket],
2043 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
2050 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
2052 if (v == SEQ_START_TOKEN) {
2054 "IP Virtual Server version %d.%d.%d (size=%d)\n",
2055 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2057 "Prot LocalAddress:Port Scheduler Flags\n");
2059 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2061 struct net *net = seq_file_net(seq);
2062 struct netns_ipvs *ipvs = net_ipvs(net);
2063 const struct ip_vs_service *svc = v;
2064 const struct ip_vs_iter *iter = seq->private;
2065 const struct ip_vs_dest *dest;
2066 struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler);
2067 char *sched_name = sched ? sched->name : "none";
2069 if (svc->ipvs != ipvs)
2071 if (iter->table == ip_vs_svc_table) {
2072 #ifdef CONFIG_IP_VS_IPV6
2073 if (svc->af == AF_INET6)
2074 seq_printf(seq, "%s [%pI6]:%04X %s ",
2075 ip_vs_proto_name(svc->protocol),
2081 seq_printf(seq, "%s %08X:%04X %s %s ",
2082 ip_vs_proto_name(svc->protocol),
2083 ntohl(svc->addr.ip),
2086 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2088 seq_printf(seq, "FWM %08X %s %s",
2089 svc->fwmark, sched_name,
2090 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2093 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2094 seq_printf(seq, "persistent %d %08X\n",
2096 ntohl(svc->netmask));
2098 seq_putc(seq, '\n');
2100 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
2101 #ifdef CONFIG_IP_VS_IPV6
2102 if (dest->af == AF_INET6)
2105 " %-7s %-6d %-10d %-10d\n",
2108 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2109 atomic_read(&dest->weight),
2110 atomic_read(&dest->activeconns),
2111 atomic_read(&dest->inactconns));
2116 "%-7s %-6d %-10d %-10d\n",
2117 ntohl(dest->addr.ip),
2119 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2120 atomic_read(&dest->weight),
2121 atomic_read(&dest->activeconns),
2122 atomic_read(&dest->inactconns));
2129 static const struct seq_operations ip_vs_info_seq_ops = {
2130 .start = ip_vs_info_seq_start,
2131 .next = ip_vs_info_seq_next,
2132 .stop = ip_vs_info_seq_stop,
2133 .show = ip_vs_info_seq_show,
2136 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2138 struct net *net = seq_file_single_net(seq);
2139 struct ip_vs_kstats show;
2141 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2143 " Total Incoming Outgoing Incoming Outgoing\n");
2145 " Conns Packets Packets Bytes Bytes\n");
2147 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2148 seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n\n",
2149 (unsigned long long)show.conns,
2150 (unsigned long long)show.inpkts,
2151 (unsigned long long)show.outpkts,
2152 (unsigned long long)show.inbytes,
2153 (unsigned long long)show.outbytes);
2155 /* 01234567 01234567 01234567 0123456701234567 0123456701234567*/
2157 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2158 seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n",
2159 (unsigned long long)show.cps,
2160 (unsigned long long)show.inpps,
2161 (unsigned long long)show.outpps,
2162 (unsigned long long)show.inbps,
2163 (unsigned long long)show.outbps);
2168 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2170 struct net *net = seq_file_single_net(seq);
2171 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2172 struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats;
2173 struct ip_vs_kstats kstats;
2176 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2178 " Total Incoming Outgoing Incoming Outgoing\n");
2180 "CPU Conns Packets Packets Bytes Bytes\n");
2182 for_each_possible_cpu(i) {
2183 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2185 u64 conns, inpkts, outpkts, inbytes, outbytes;
2188 start = u64_stats_fetch_begin_irq(&u->syncp);
2189 conns = u->cnt.conns;
2190 inpkts = u->cnt.inpkts;
2191 outpkts = u->cnt.outpkts;
2192 inbytes = u->cnt.inbytes;
2193 outbytes = u->cnt.outbytes;
2194 } while (u64_stats_fetch_retry_irq(&u->syncp, start));
2196 seq_printf(seq, "%3X %8LX %8LX %8LX %16LX %16LX\n",
2197 i, (u64)conns, (u64)inpkts,
2198 (u64)outpkts, (u64)inbytes,
2202 ip_vs_copy_stats(&kstats, tot_stats);
2204 seq_printf(seq, " ~ %8LX %8LX %8LX %16LX %16LX\n\n",
2205 (unsigned long long)kstats.conns,
2206 (unsigned long long)kstats.inpkts,
2207 (unsigned long long)kstats.outpkts,
2208 (unsigned long long)kstats.inbytes,
2209 (unsigned long long)kstats.outbytes);
2211 /* ... 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2213 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2214 seq_printf(seq, " %8LX %8LX %8LX %16LX %16LX\n",
2226 * Set timeout values for tcp tcpfin udp in the timeout_table.
2228 static int ip_vs_set_timeout(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2230 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2231 struct ip_vs_proto_data *pd;
2234 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2239 #ifdef CONFIG_IP_VS_PROTO_TCP
2240 if (u->tcp_timeout < 0 || u->tcp_timeout > (INT_MAX / HZ) ||
2241 u->tcp_fin_timeout < 0 || u->tcp_fin_timeout > (INT_MAX / HZ)) {
2246 #ifdef CONFIG_IP_VS_PROTO_UDP
2247 if (u->udp_timeout < 0 || u->udp_timeout > (INT_MAX / HZ))
2251 #ifdef CONFIG_IP_VS_PROTO_TCP
2252 if (u->tcp_timeout) {
2253 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2254 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2255 = u->tcp_timeout * HZ;
2258 if (u->tcp_fin_timeout) {
2259 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2260 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2261 = u->tcp_fin_timeout * HZ;
2265 #ifdef CONFIG_IP_VS_PROTO_UDP
2266 if (u->udp_timeout) {
2267 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2268 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2269 = u->udp_timeout * HZ;
2275 #define CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2277 struct ip_vs_svcdest_user {
2278 struct ip_vs_service_user s;
2279 struct ip_vs_dest_user d;
2282 static const unsigned char set_arglen[CMDID(IP_VS_SO_SET_MAX) + 1] = {
2283 [CMDID(IP_VS_SO_SET_ADD)] = sizeof(struct ip_vs_service_user),
2284 [CMDID(IP_VS_SO_SET_EDIT)] = sizeof(struct ip_vs_service_user),
2285 [CMDID(IP_VS_SO_SET_DEL)] = sizeof(struct ip_vs_service_user),
2286 [CMDID(IP_VS_SO_SET_ADDDEST)] = sizeof(struct ip_vs_svcdest_user),
2287 [CMDID(IP_VS_SO_SET_DELDEST)] = sizeof(struct ip_vs_svcdest_user),
2288 [CMDID(IP_VS_SO_SET_EDITDEST)] = sizeof(struct ip_vs_svcdest_user),
2289 [CMDID(IP_VS_SO_SET_TIMEOUT)] = sizeof(struct ip_vs_timeout_user),
2290 [CMDID(IP_VS_SO_SET_STARTDAEMON)] = sizeof(struct ip_vs_daemon_user),
2291 [CMDID(IP_VS_SO_SET_STOPDAEMON)] = sizeof(struct ip_vs_daemon_user),
2292 [CMDID(IP_VS_SO_SET_ZERO)] = sizeof(struct ip_vs_service_user),
2295 union ip_vs_set_arglen {
2296 struct ip_vs_service_user field_IP_VS_SO_SET_ADD;
2297 struct ip_vs_service_user field_IP_VS_SO_SET_EDIT;
2298 struct ip_vs_service_user field_IP_VS_SO_SET_DEL;
2299 struct ip_vs_svcdest_user field_IP_VS_SO_SET_ADDDEST;
2300 struct ip_vs_svcdest_user field_IP_VS_SO_SET_DELDEST;
2301 struct ip_vs_svcdest_user field_IP_VS_SO_SET_EDITDEST;
2302 struct ip_vs_timeout_user field_IP_VS_SO_SET_TIMEOUT;
2303 struct ip_vs_daemon_user field_IP_VS_SO_SET_STARTDAEMON;
2304 struct ip_vs_daemon_user field_IP_VS_SO_SET_STOPDAEMON;
2305 struct ip_vs_service_user field_IP_VS_SO_SET_ZERO;
2308 #define MAX_SET_ARGLEN sizeof(union ip_vs_set_arglen)
2310 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2311 struct ip_vs_service_user *usvc_compat)
2313 memset(usvc, 0, sizeof(*usvc));
2316 usvc->protocol = usvc_compat->protocol;
2317 usvc->addr.ip = usvc_compat->addr;
2318 usvc->port = usvc_compat->port;
2319 usvc->fwmark = usvc_compat->fwmark;
2321 /* Deep copy of sched_name is not needed here */
2322 usvc->sched_name = usvc_compat->sched_name;
2324 usvc->flags = usvc_compat->flags;
2325 usvc->timeout = usvc_compat->timeout;
2326 usvc->netmask = usvc_compat->netmask;
2329 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2330 struct ip_vs_dest_user *udest_compat)
2332 memset(udest, 0, sizeof(*udest));
2334 udest->addr.ip = udest_compat->addr;
2335 udest->port = udest_compat->port;
2336 udest->conn_flags = udest_compat->conn_flags;
2337 udest->weight = udest_compat->weight;
2338 udest->u_threshold = udest_compat->u_threshold;
2339 udest->l_threshold = udest_compat->l_threshold;
2340 udest->af = AF_INET;
2344 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2346 struct net *net = sock_net(sk);
2348 unsigned char arg[MAX_SET_ARGLEN];
2349 struct ip_vs_service_user *usvc_compat;
2350 struct ip_vs_service_user_kern usvc;
2351 struct ip_vs_service *svc;
2352 struct ip_vs_dest_user *udest_compat;
2353 struct ip_vs_dest_user_kern udest;
2354 struct netns_ipvs *ipvs = net_ipvs(net);
2356 BUILD_BUG_ON(sizeof(arg) > 255);
2357 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2360 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2362 if (len != set_arglen[CMDID(cmd)]) {
2363 IP_VS_DBG(1, "set_ctl: len %u != %u\n",
2364 len, set_arglen[CMDID(cmd)]);
2368 if (copy_from_user(arg, user, len) != 0)
2371 /* Handle daemons since they have another lock */
2372 if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2373 cmd == IP_VS_SO_SET_STOPDAEMON) {
2374 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2376 if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2377 struct ipvs_sync_daemon_cfg cfg;
2379 memset(&cfg, 0, sizeof(cfg));
2381 if (strscpy(cfg.mcast_ifn, dm->mcast_ifn,
2382 sizeof(cfg.mcast_ifn)) <= 0)
2384 cfg.syncid = dm->syncid;
2385 ret = start_sync_thread(ipvs, &cfg, dm->state);
2387 ret = stop_sync_thread(ipvs, dm->state);
2392 mutex_lock(&__ip_vs_mutex);
2393 if (cmd == IP_VS_SO_SET_FLUSH) {
2394 /* Flush the virtual service */
2395 ret = ip_vs_flush(ipvs, false);
2397 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2398 /* Set timeout values for (tcp tcpfin udp) */
2399 ret = ip_vs_set_timeout(ipvs, (struct ip_vs_timeout_user *)arg);
2402 /* No more commands with len == 0 below */
2407 usvc_compat = (struct ip_vs_service_user *)arg;
2408 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2410 /* We only use the new structs internally, so copy userspace compat
2411 * structs to extended internal versions */
2412 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2413 ip_vs_copy_udest_compat(&udest, udest_compat);
2415 if (cmd == IP_VS_SO_SET_ZERO) {
2416 /* if no service address is set, zero counters in all */
2417 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2418 ret = ip_vs_zero_all(ipvs);
2423 if ((cmd == IP_VS_SO_SET_ADD || cmd == IP_VS_SO_SET_EDIT) &&
2424 strnlen(usvc.sched_name, IP_VS_SCHEDNAME_MAXLEN) ==
2425 IP_VS_SCHEDNAME_MAXLEN) {
2430 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2431 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2432 usvc.protocol != IPPROTO_SCTP) {
2433 pr_err("set_ctl: invalid protocol: %d %pI4:%d\n",
2434 usvc.protocol, &usvc.addr.ip,
2440 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2442 if (usvc.fwmark == 0)
2443 svc = __ip_vs_service_find(ipvs, usvc.af, usvc.protocol,
2444 &usvc.addr, usvc.port);
2446 svc = __ip_vs_svc_fwm_find(ipvs, usvc.af, usvc.fwmark);
2449 if (cmd != IP_VS_SO_SET_ADD
2450 && (svc == NULL || svc->protocol != usvc.protocol)) {
2456 case IP_VS_SO_SET_ADD:
2460 ret = ip_vs_add_service(ipvs, &usvc, &svc);
2462 case IP_VS_SO_SET_EDIT:
2463 ret = ip_vs_edit_service(svc, &usvc);
2465 case IP_VS_SO_SET_DEL:
2466 ret = ip_vs_del_service(svc);
2470 case IP_VS_SO_SET_ZERO:
2471 ret = ip_vs_zero_service(svc);
2473 case IP_VS_SO_SET_ADDDEST:
2474 ret = ip_vs_add_dest(svc, &udest);
2476 case IP_VS_SO_SET_EDITDEST:
2477 ret = ip_vs_edit_dest(svc, &udest);
2479 case IP_VS_SO_SET_DELDEST:
2480 ret = ip_vs_del_dest(svc, &udest);
2484 mutex_unlock(&__ip_vs_mutex);
2490 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2492 struct ip_vs_scheduler *sched;
2493 struct ip_vs_kstats kstats;
2496 sched = rcu_dereference_protected(src->scheduler, 1);
2497 sched_name = sched ? sched->name : "none";
2498 dst->protocol = src->protocol;
2499 dst->addr = src->addr.ip;
2500 dst->port = src->port;
2501 dst->fwmark = src->fwmark;
2502 strlcpy(dst->sched_name, sched_name, sizeof(dst->sched_name));
2503 dst->flags = src->flags;
2504 dst->timeout = src->timeout / HZ;
2505 dst->netmask = src->netmask;
2506 dst->num_dests = src->num_dests;
2507 ip_vs_copy_stats(&kstats, &src->stats);
2508 ip_vs_export_stats_user(&dst->stats, &kstats);
2512 __ip_vs_get_service_entries(struct netns_ipvs *ipvs,
2513 const struct ip_vs_get_services *get,
2514 struct ip_vs_get_services __user *uptr)
2517 struct ip_vs_service *svc;
2518 struct ip_vs_service_entry entry;
2521 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2522 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2523 /* Only expose IPv4 entries to old interface */
2524 if (svc->af != AF_INET || (svc->ipvs != ipvs))
2527 if (count >= get->num_services)
2529 memset(&entry, 0, sizeof(entry));
2530 ip_vs_copy_service(&entry, svc);
2531 if (copy_to_user(&uptr->entrytable[count],
2532 &entry, sizeof(entry))) {
2540 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2541 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2542 /* Only expose IPv4 entries to old interface */
2543 if (svc->af != AF_INET || (svc->ipvs != ipvs))
2546 if (count >= get->num_services)
2548 memset(&entry, 0, sizeof(entry));
2549 ip_vs_copy_service(&entry, svc);
2550 if (copy_to_user(&uptr->entrytable[count],
2551 &entry, sizeof(entry))) {
2563 __ip_vs_get_dest_entries(struct netns_ipvs *ipvs, const struct ip_vs_get_dests *get,
2564 struct ip_vs_get_dests __user *uptr)
2566 struct ip_vs_service *svc;
2567 union nf_inet_addr addr = { .ip = get->addr };
2572 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, get->fwmark);
2574 svc = __ip_vs_service_find(ipvs, AF_INET, get->protocol, &addr,
2580 struct ip_vs_dest *dest;
2581 struct ip_vs_dest_entry entry;
2582 struct ip_vs_kstats kstats;
2584 memset(&entry, 0, sizeof(entry));
2585 list_for_each_entry(dest, &svc->destinations, n_list) {
2586 if (count >= get->num_dests)
2589 /* Cannot expose heterogeneous members via sockopt
2592 if (dest->af != svc->af)
2595 entry.addr = dest->addr.ip;
2596 entry.port = dest->port;
2597 entry.conn_flags = atomic_read(&dest->conn_flags);
2598 entry.weight = atomic_read(&dest->weight);
2599 entry.u_threshold = dest->u_threshold;
2600 entry.l_threshold = dest->l_threshold;
2601 entry.activeconns = atomic_read(&dest->activeconns);
2602 entry.inactconns = atomic_read(&dest->inactconns);
2603 entry.persistconns = atomic_read(&dest->persistconns);
2604 ip_vs_copy_stats(&kstats, &dest->stats);
2605 ip_vs_export_stats_user(&entry.stats, &kstats);
2606 if (copy_to_user(&uptr->entrytable[count],
2607 &entry, sizeof(entry))) {
2619 __ip_vs_get_timeouts(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2621 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2622 struct ip_vs_proto_data *pd;
2625 memset(u, 0, sizeof (*u));
2627 #ifdef CONFIG_IP_VS_PROTO_TCP
2628 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2629 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2630 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2632 #ifdef CONFIG_IP_VS_PROTO_UDP
2633 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2635 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2639 static const unsigned char get_arglen[CMDID(IP_VS_SO_GET_MAX) + 1] = {
2640 [CMDID(IP_VS_SO_GET_VERSION)] = 64,
2641 [CMDID(IP_VS_SO_GET_INFO)] = sizeof(struct ip_vs_getinfo),
2642 [CMDID(IP_VS_SO_GET_SERVICES)] = sizeof(struct ip_vs_get_services),
2643 [CMDID(IP_VS_SO_GET_SERVICE)] = sizeof(struct ip_vs_service_entry),
2644 [CMDID(IP_VS_SO_GET_DESTS)] = sizeof(struct ip_vs_get_dests),
2645 [CMDID(IP_VS_SO_GET_TIMEOUT)] = sizeof(struct ip_vs_timeout_user),
2646 [CMDID(IP_VS_SO_GET_DAEMON)] = 2 * sizeof(struct ip_vs_daemon_user),
2649 union ip_vs_get_arglen {
2650 char field_IP_VS_SO_GET_VERSION[64];
2651 struct ip_vs_getinfo field_IP_VS_SO_GET_INFO;
2652 struct ip_vs_get_services field_IP_VS_SO_GET_SERVICES;
2653 struct ip_vs_service_entry field_IP_VS_SO_GET_SERVICE;
2654 struct ip_vs_get_dests field_IP_VS_SO_GET_DESTS;
2655 struct ip_vs_timeout_user field_IP_VS_SO_GET_TIMEOUT;
2656 struct ip_vs_daemon_user field_IP_VS_SO_GET_DAEMON[2];
2659 #define MAX_GET_ARGLEN sizeof(union ip_vs_get_arglen)
2662 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2664 unsigned char arg[MAX_GET_ARGLEN];
2666 unsigned int copylen;
2667 struct net *net = sock_net(sk);
2668 struct netns_ipvs *ipvs = net_ipvs(net);
2671 BUILD_BUG_ON(sizeof(arg) > 255);
2672 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2675 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2678 copylen = get_arglen[CMDID(cmd)];
2679 if (*len < (int) copylen) {
2680 IP_VS_DBG(1, "get_ctl: len %d < %u\n", *len, copylen);
2684 if (copy_from_user(arg, user, copylen) != 0)
2687 * Handle daemons first since it has its own locking
2689 if (cmd == IP_VS_SO_GET_DAEMON) {
2690 struct ip_vs_daemon_user d[2];
2692 memset(&d, 0, sizeof(d));
2693 mutex_lock(&ipvs->sync_mutex);
2694 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2695 d[0].state = IP_VS_STATE_MASTER;
2696 strlcpy(d[0].mcast_ifn, ipvs->mcfg.mcast_ifn,
2697 sizeof(d[0].mcast_ifn));
2698 d[0].syncid = ipvs->mcfg.syncid;
2700 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2701 d[1].state = IP_VS_STATE_BACKUP;
2702 strlcpy(d[1].mcast_ifn, ipvs->bcfg.mcast_ifn,
2703 sizeof(d[1].mcast_ifn));
2704 d[1].syncid = ipvs->bcfg.syncid;
2706 if (copy_to_user(user, &d, sizeof(d)) != 0)
2708 mutex_unlock(&ipvs->sync_mutex);
2712 mutex_lock(&__ip_vs_mutex);
2714 case IP_VS_SO_GET_VERSION:
2718 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2719 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2720 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2724 *len = strlen(buf)+1;
2728 case IP_VS_SO_GET_INFO:
2730 struct ip_vs_getinfo info;
2731 info.version = IP_VS_VERSION_CODE;
2732 info.size = ip_vs_conn_tab_size;
2733 info.num_services = ipvs->num_services;
2734 if (copy_to_user(user, &info, sizeof(info)) != 0)
2739 case IP_VS_SO_GET_SERVICES:
2741 struct ip_vs_get_services *get;
2744 get = (struct ip_vs_get_services *)arg;
2745 size = sizeof(*get) +
2746 sizeof(struct ip_vs_service_entry) * get->num_services;
2748 pr_err("length: %u != %u\n", *len, size);
2752 ret = __ip_vs_get_service_entries(ipvs, get, user);
2756 case IP_VS_SO_GET_SERVICE:
2758 struct ip_vs_service_entry *entry;
2759 struct ip_vs_service *svc;
2760 union nf_inet_addr addr;
2762 entry = (struct ip_vs_service_entry *)arg;
2763 addr.ip = entry->addr;
2766 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, entry->fwmark);
2768 svc = __ip_vs_service_find(ipvs, AF_INET,
2769 entry->protocol, &addr,
2773 ip_vs_copy_service(entry, svc);
2774 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2781 case IP_VS_SO_GET_DESTS:
2783 struct ip_vs_get_dests *get;
2786 get = (struct ip_vs_get_dests *)arg;
2787 size = sizeof(*get) +
2788 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2790 pr_err("length: %u != %u\n", *len, size);
2794 ret = __ip_vs_get_dest_entries(ipvs, get, user);
2798 case IP_VS_SO_GET_TIMEOUT:
2800 struct ip_vs_timeout_user t;
2802 __ip_vs_get_timeouts(ipvs, &t);
2803 if (copy_to_user(user, &t, sizeof(t)) != 0)
2813 mutex_unlock(&__ip_vs_mutex);
2818 static struct nf_sockopt_ops ip_vs_sockopts = {
2820 .set_optmin = IP_VS_BASE_CTL,
2821 .set_optmax = IP_VS_SO_SET_MAX+1,
2822 .set = do_ip_vs_set_ctl,
2823 .get_optmin = IP_VS_BASE_CTL,
2824 .get_optmax = IP_VS_SO_GET_MAX+1,
2825 .get = do_ip_vs_get_ctl,
2826 .owner = THIS_MODULE,
2830 * Generic Netlink interface
2833 /* IPVS genetlink family */
2834 static struct genl_family ip_vs_genl_family;
2836 /* Policy used for first-level command attributes */
2837 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2838 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2839 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2840 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2841 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2842 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2843 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2846 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2847 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2848 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2849 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2850 .len = IP_VS_IFNAME_MAXLEN - 1 },
2851 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2852 [IPVS_DAEMON_ATTR_SYNC_MAXLEN] = { .type = NLA_U16 },
2853 [IPVS_DAEMON_ATTR_MCAST_GROUP] = { .type = NLA_U32 },
2854 [IPVS_DAEMON_ATTR_MCAST_GROUP6] = { .len = sizeof(struct in6_addr) },
2855 [IPVS_DAEMON_ATTR_MCAST_PORT] = { .type = NLA_U16 },
2856 [IPVS_DAEMON_ATTR_MCAST_TTL] = { .type = NLA_U8 },
2859 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2860 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2861 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2862 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2863 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2864 .len = sizeof(union nf_inet_addr) },
2865 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2866 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2867 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2868 .len = IP_VS_SCHEDNAME_MAXLEN - 1 },
2869 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2870 .len = IP_VS_PENAME_MAXLEN },
2871 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2872 .len = sizeof(struct ip_vs_flags) },
2873 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2874 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2875 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2878 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2879 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2880 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2881 .len = sizeof(union nf_inet_addr) },
2882 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2883 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2884 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2885 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2886 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2887 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2888 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2889 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2890 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2891 [IPVS_DEST_ATTR_ADDR_FAMILY] = { .type = NLA_U16 },
2894 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2895 struct ip_vs_kstats *kstats)
2897 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2902 if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, (u32)kstats->conns) ||
2903 nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, (u32)kstats->inpkts) ||
2904 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, (u32)kstats->outpkts) ||
2905 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes,
2906 IPVS_STATS_ATTR_PAD) ||
2907 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes,
2908 IPVS_STATS_ATTR_PAD) ||
2909 nla_put_u32(skb, IPVS_STATS_ATTR_CPS, (u32)kstats->cps) ||
2910 nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, (u32)kstats->inpps) ||
2911 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, (u32)kstats->outpps) ||
2912 nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, (u32)kstats->inbps) ||
2913 nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, (u32)kstats->outbps))
2914 goto nla_put_failure;
2915 nla_nest_end(skb, nl_stats);
2920 nla_nest_cancel(skb, nl_stats);
2924 static int ip_vs_genl_fill_stats64(struct sk_buff *skb, int container_type,
2925 struct ip_vs_kstats *kstats)
2927 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2932 if (nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CONNS, kstats->conns,
2933 IPVS_STATS_ATTR_PAD) ||
2934 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPKTS, kstats->inpkts,
2935 IPVS_STATS_ATTR_PAD) ||
2936 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPKTS, kstats->outpkts,
2937 IPVS_STATS_ATTR_PAD) ||
2938 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes,
2939 IPVS_STATS_ATTR_PAD) ||
2940 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes,
2941 IPVS_STATS_ATTR_PAD) ||
2942 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CPS, kstats->cps,
2943 IPVS_STATS_ATTR_PAD) ||
2944 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPPS, kstats->inpps,
2945 IPVS_STATS_ATTR_PAD) ||
2946 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPPS, kstats->outpps,
2947 IPVS_STATS_ATTR_PAD) ||
2948 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBPS, kstats->inbps,
2949 IPVS_STATS_ATTR_PAD) ||
2950 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBPS, kstats->outbps,
2951 IPVS_STATS_ATTR_PAD))
2952 goto nla_put_failure;
2953 nla_nest_end(skb, nl_stats);
2958 nla_nest_cancel(skb, nl_stats);
2962 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2963 struct ip_vs_service *svc)
2965 struct ip_vs_scheduler *sched;
2966 struct ip_vs_pe *pe;
2967 struct nlattr *nl_service;
2968 struct ip_vs_flags flags = { .flags = svc->flags,
2970 struct ip_vs_kstats kstats;
2973 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2977 if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2978 goto nla_put_failure;
2980 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2981 goto nla_put_failure;
2983 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2984 nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2985 nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2986 goto nla_put_failure;
2989 sched = rcu_dereference_protected(svc->scheduler, 1);
2990 sched_name = sched ? sched->name : "none";
2991 pe = rcu_dereference_protected(svc->pe, 1);
2992 if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched_name) ||
2993 (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
2994 nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2995 nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2996 nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
2997 goto nla_put_failure;
2998 ip_vs_copy_stats(&kstats, &svc->stats);
2999 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &kstats))
3000 goto nla_put_failure;
3001 if (ip_vs_genl_fill_stats64(skb, IPVS_SVC_ATTR_STATS64, &kstats))
3002 goto nla_put_failure;
3004 nla_nest_end(skb, nl_service);
3009 nla_nest_cancel(skb, nl_service);
3013 static int ip_vs_genl_dump_service(struct sk_buff *skb,
3014 struct ip_vs_service *svc,
3015 struct netlink_callback *cb)
3019 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3020 &ip_vs_genl_family, NLM_F_MULTI,
3021 IPVS_CMD_NEW_SERVICE);
3025 if (ip_vs_genl_fill_service(skb, svc) < 0)
3026 goto nla_put_failure;
3028 genlmsg_end(skb, hdr);
3032 genlmsg_cancel(skb, hdr);
3036 static int ip_vs_genl_dump_services(struct sk_buff *skb,
3037 struct netlink_callback *cb)
3040 int start = cb->args[0];
3041 struct ip_vs_service *svc;
3042 struct net *net = sock_net(skb->sk);
3043 struct netns_ipvs *ipvs = net_ipvs(net);
3045 mutex_lock(&__ip_vs_mutex);
3046 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3047 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
3048 if (++idx <= start || (svc->ipvs != ipvs))
3050 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3052 goto nla_put_failure;
3057 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3058 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
3059 if (++idx <= start || (svc->ipvs != ipvs))
3061 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3063 goto nla_put_failure;
3069 mutex_unlock(&__ip_vs_mutex);
3075 static bool ip_vs_is_af_valid(int af)
3079 #ifdef CONFIG_IP_VS_IPV6
3080 if (af == AF_INET6 && ipv6_mod_enabled())
3086 static int ip_vs_genl_parse_service(struct netns_ipvs *ipvs,
3087 struct ip_vs_service_user_kern *usvc,
3088 struct nlattr *nla, int full_entry,
3089 struct ip_vs_service **ret_svc)
3091 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
3092 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
3093 struct ip_vs_service *svc;
3095 /* Parse mandatory identifying service fields first */
3097 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla,
3098 ip_vs_svc_policy, NULL))
3101 nla_af = attrs[IPVS_SVC_ATTR_AF];
3102 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
3103 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
3104 nla_port = attrs[IPVS_SVC_ATTR_PORT];
3105 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
3107 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3110 memset(usvc, 0, sizeof(*usvc));
3112 usvc->af = nla_get_u16(nla_af);
3113 if (!ip_vs_is_af_valid(usvc->af))
3114 return -EAFNOSUPPORT;
3117 usvc->protocol = IPPROTO_TCP;
3118 usvc->fwmark = nla_get_u32(nla_fwmark);
3120 usvc->protocol = nla_get_u16(nla_protocol);
3121 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3122 usvc->port = nla_get_be16(nla_port);
3128 svc = __ip_vs_svc_fwm_find(ipvs, usvc->af, usvc->fwmark);
3130 svc = __ip_vs_service_find(ipvs, usvc->af, usvc->protocol,
3131 &usvc->addr, usvc->port);
3135 /* If a full entry was requested, check for the additional fields */
3137 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3139 struct ip_vs_flags flags;
3141 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3142 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3143 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3144 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3145 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3147 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3150 nla_memcpy(&flags, nla_flags, sizeof(flags));
3152 /* prefill flags from service if it already exists */
3154 usvc->flags = svc->flags;
3156 /* set new flags from userland */
3157 usvc->flags = (usvc->flags & ~flags.mask) |
3158 (flags.flags & flags.mask);
3159 usvc->sched_name = nla_data(nla_sched);
3160 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3161 usvc->timeout = nla_get_u32(nla_timeout);
3162 usvc->netmask = nla_get_be32(nla_netmask);
3168 static struct ip_vs_service *ip_vs_genl_find_service(struct netns_ipvs *ipvs,
3171 struct ip_vs_service_user_kern usvc;
3172 struct ip_vs_service *svc;
3175 ret = ip_vs_genl_parse_service(ipvs, &usvc, nla, 0, &svc);
3176 return ret ? ERR_PTR(ret) : svc;
3179 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3181 struct nlattr *nl_dest;
3182 struct ip_vs_kstats kstats;
3184 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3188 if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3189 nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3190 nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3191 (atomic_read(&dest->conn_flags) &
3192 IP_VS_CONN_F_FWD_MASK)) ||
3193 nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3194 atomic_read(&dest->weight)) ||
3195 nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3196 nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3197 nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3198 atomic_read(&dest->activeconns)) ||
3199 nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3200 atomic_read(&dest->inactconns)) ||
3201 nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3202 atomic_read(&dest->persistconns)) ||
3203 nla_put_u16(skb, IPVS_DEST_ATTR_ADDR_FAMILY, dest->af))
3204 goto nla_put_failure;
3205 ip_vs_copy_stats(&kstats, &dest->stats);
3206 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &kstats))
3207 goto nla_put_failure;
3208 if (ip_vs_genl_fill_stats64(skb, IPVS_DEST_ATTR_STATS64, &kstats))
3209 goto nla_put_failure;
3211 nla_nest_end(skb, nl_dest);
3216 nla_nest_cancel(skb, nl_dest);
3220 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3221 struct netlink_callback *cb)
3225 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3226 &ip_vs_genl_family, NLM_F_MULTI,
3231 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3232 goto nla_put_failure;
3234 genlmsg_end(skb, hdr);
3238 genlmsg_cancel(skb, hdr);
3242 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3243 struct netlink_callback *cb)
3246 int start = cb->args[0];
3247 struct ip_vs_service *svc;
3248 struct ip_vs_dest *dest;
3249 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3250 struct net *net = sock_net(skb->sk);
3251 struct netns_ipvs *ipvs = net_ipvs(net);
3253 mutex_lock(&__ip_vs_mutex);
3255 /* Try to find the service for which to dump destinations */
3256 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs, IPVS_CMD_ATTR_MAX,
3257 ip_vs_cmd_policy, NULL))
3261 svc = ip_vs_genl_find_service(ipvs, attrs[IPVS_CMD_ATTR_SERVICE]);
3262 if (IS_ERR_OR_NULL(svc))
3265 /* Dump the destinations */
3266 list_for_each_entry(dest, &svc->destinations, n_list) {
3269 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3271 goto nla_put_failure;
3279 mutex_unlock(&__ip_vs_mutex);
3284 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3285 struct nlattr *nla, int full_entry)
3287 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3288 struct nlattr *nla_addr, *nla_port;
3289 struct nlattr *nla_addr_family;
3291 /* Parse mandatory identifying destination fields first */
3293 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla,
3294 ip_vs_dest_policy, NULL))
3297 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3298 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3299 nla_addr_family = attrs[IPVS_DEST_ATTR_ADDR_FAMILY];
3301 if (!(nla_addr && nla_port))
3304 memset(udest, 0, sizeof(*udest));
3306 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3307 udest->port = nla_get_be16(nla_port);
3309 if (nla_addr_family)
3310 udest->af = nla_get_u16(nla_addr_family);
3314 /* If a full entry was requested, check for the additional fields */
3316 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3319 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3320 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3321 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3322 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3324 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3327 udest->conn_flags = nla_get_u32(nla_fwd)
3328 & IP_VS_CONN_F_FWD_MASK;
3329 udest->weight = nla_get_u32(nla_weight);
3330 udest->u_threshold = nla_get_u32(nla_u_thresh);
3331 udest->l_threshold = nla_get_u32(nla_l_thresh);
3337 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3338 struct ipvs_sync_daemon_cfg *c)
3340 struct nlattr *nl_daemon;
3342 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3346 if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3347 nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, c->mcast_ifn) ||
3348 nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, c->syncid) ||
3349 nla_put_u16(skb, IPVS_DAEMON_ATTR_SYNC_MAXLEN, c->sync_maxlen) ||
3350 nla_put_u16(skb, IPVS_DAEMON_ATTR_MCAST_PORT, c->mcast_port) ||
3351 nla_put_u8(skb, IPVS_DAEMON_ATTR_MCAST_TTL, c->mcast_ttl))
3352 goto nla_put_failure;
3353 #ifdef CONFIG_IP_VS_IPV6
3354 if (c->mcast_af == AF_INET6) {
3355 if (nla_put_in6_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP6,
3356 &c->mcast_group.in6))
3357 goto nla_put_failure;
3360 if (c->mcast_af == AF_INET &&
3361 nla_put_in_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP,
3363 goto nla_put_failure;
3364 nla_nest_end(skb, nl_daemon);
3369 nla_nest_cancel(skb, nl_daemon);
3373 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3374 struct ipvs_sync_daemon_cfg *c,
3375 struct netlink_callback *cb)
3378 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3379 &ip_vs_genl_family, NLM_F_MULTI,
3380 IPVS_CMD_NEW_DAEMON);
3384 if (ip_vs_genl_fill_daemon(skb, state, c))
3385 goto nla_put_failure;
3387 genlmsg_end(skb, hdr);
3391 genlmsg_cancel(skb, hdr);
3395 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3396 struct netlink_callback *cb)
3398 struct net *net = sock_net(skb->sk);
3399 struct netns_ipvs *ipvs = net_ipvs(net);
3401 mutex_lock(&ipvs->sync_mutex);
3402 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3403 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3404 &ipvs->mcfg, cb) < 0)
3405 goto nla_put_failure;
3410 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3411 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3412 &ipvs->bcfg, cb) < 0)
3413 goto nla_put_failure;
3419 mutex_unlock(&ipvs->sync_mutex);
3424 static int ip_vs_genl_new_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3426 struct ipvs_sync_daemon_cfg c;
3430 memset(&c, 0, sizeof(c));
3431 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3432 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3433 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3435 strlcpy(c.mcast_ifn, nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3436 sizeof(c.mcast_ifn));
3437 c.syncid = nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]);
3439 a = attrs[IPVS_DAEMON_ATTR_SYNC_MAXLEN];
3441 c.sync_maxlen = nla_get_u16(a);
3443 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP];
3445 c.mcast_af = AF_INET;
3446 c.mcast_group.ip = nla_get_in_addr(a);
3447 if (!ipv4_is_multicast(c.mcast_group.ip))
3450 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP6];
3452 #ifdef CONFIG_IP_VS_IPV6
3455 c.mcast_af = AF_INET6;
3456 c.mcast_group.in6 = nla_get_in6_addr(a);
3457 addr_type = ipv6_addr_type(&c.mcast_group.in6);
3458 if (!(addr_type & IPV6_ADDR_MULTICAST))
3461 return -EAFNOSUPPORT;
3466 a = attrs[IPVS_DAEMON_ATTR_MCAST_PORT];
3468 c.mcast_port = nla_get_u16(a);
3470 a = attrs[IPVS_DAEMON_ATTR_MCAST_TTL];
3472 c.mcast_ttl = nla_get_u8(a);
3474 /* The synchronization protocol is incompatible with mixed family
3477 if (ipvs->mixed_address_family_dests > 0)
3480 ret = start_sync_thread(ipvs, &c,
3481 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3485 static int ip_vs_genl_del_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3489 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3492 ret = stop_sync_thread(ipvs,
3493 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3497 static int ip_vs_genl_set_config(struct netns_ipvs *ipvs, struct nlattr **attrs)
3499 struct ip_vs_timeout_user t;
3501 __ip_vs_get_timeouts(ipvs, &t);
3503 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3504 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3506 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3508 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3510 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3511 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3513 return ip_vs_set_timeout(ipvs, &t);
3516 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3518 int ret = -EINVAL, cmd;
3519 struct net *net = sock_net(skb->sk);
3520 struct netns_ipvs *ipvs = net_ipvs(net);
3522 cmd = info->genlhdr->cmd;
3524 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3525 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3527 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3528 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3529 info->attrs[IPVS_CMD_ATTR_DAEMON],
3530 ip_vs_daemon_policy, info->extack))
3533 if (cmd == IPVS_CMD_NEW_DAEMON)
3534 ret = ip_vs_genl_new_daemon(ipvs, daemon_attrs);
3536 ret = ip_vs_genl_del_daemon(ipvs, daemon_attrs);
3543 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3545 struct ip_vs_service *svc = NULL;
3546 struct ip_vs_service_user_kern usvc;
3547 struct ip_vs_dest_user_kern udest;
3549 int need_full_svc = 0, need_full_dest = 0;
3550 struct net *net = sock_net(skb->sk);
3551 struct netns_ipvs *ipvs = net_ipvs(net);
3553 cmd = info->genlhdr->cmd;
3555 mutex_lock(&__ip_vs_mutex);
3557 if (cmd == IPVS_CMD_FLUSH) {
3558 ret = ip_vs_flush(ipvs, false);
3560 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3561 ret = ip_vs_genl_set_config(ipvs, info->attrs);
3563 } else if (cmd == IPVS_CMD_ZERO &&
3564 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3565 ret = ip_vs_zero_all(ipvs);
3569 /* All following commands require a service argument, so check if we
3570 * received a valid one. We need a full service specification when
3571 * adding / editing a service. Only identifying members otherwise. */
3572 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3575 ret = ip_vs_genl_parse_service(ipvs, &usvc,
3576 info->attrs[IPVS_CMD_ATTR_SERVICE],
3577 need_full_svc, &svc);
3581 /* Unless we're adding a new service, the service must already exist */
3582 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3587 /* Destination commands require a valid destination argument. For
3588 * adding / editing a destination, we need a full destination
3590 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3591 cmd == IPVS_CMD_DEL_DEST) {
3592 if (cmd != IPVS_CMD_DEL_DEST)
3595 ret = ip_vs_genl_parse_dest(&udest,
3596 info->attrs[IPVS_CMD_ATTR_DEST],
3601 /* Old protocols did not allow the user to specify address
3602 * family, so we set it to zero instead. We also didn't
3603 * allow heterogeneous pools in the old code, so it's safe
3604 * to assume that this will have the same address family as
3610 if (!ip_vs_is_af_valid(udest.af)) {
3611 ret = -EAFNOSUPPORT;
3615 if (udest.af != svc->af && cmd != IPVS_CMD_DEL_DEST) {
3616 /* The synchronization protocol is incompatible
3617 * with mixed family services
3619 if (ipvs->sync_state) {
3624 /* Which connection types do we support? */
3625 switch (udest.conn_flags) {
3626 case IP_VS_CONN_F_TUNNEL:
3627 /* We are able to forward this */
3637 case IPVS_CMD_NEW_SERVICE:
3639 ret = ip_vs_add_service(ipvs, &usvc, &svc);
3643 case IPVS_CMD_SET_SERVICE:
3644 ret = ip_vs_edit_service(svc, &usvc);
3646 case IPVS_CMD_DEL_SERVICE:
3647 ret = ip_vs_del_service(svc);
3648 /* do not use svc, it can be freed */
3650 case IPVS_CMD_NEW_DEST:
3651 ret = ip_vs_add_dest(svc, &udest);
3653 case IPVS_CMD_SET_DEST:
3654 ret = ip_vs_edit_dest(svc, &udest);
3656 case IPVS_CMD_DEL_DEST:
3657 ret = ip_vs_del_dest(svc, &udest);
3660 ret = ip_vs_zero_service(svc);
3667 mutex_unlock(&__ip_vs_mutex);
3672 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3674 struct sk_buff *msg;
3676 int ret, cmd, reply_cmd;
3677 struct net *net = sock_net(skb->sk);
3678 struct netns_ipvs *ipvs = net_ipvs(net);
3680 cmd = info->genlhdr->cmd;
3682 if (cmd == IPVS_CMD_GET_SERVICE)
3683 reply_cmd = IPVS_CMD_NEW_SERVICE;
3684 else if (cmd == IPVS_CMD_GET_INFO)
3685 reply_cmd = IPVS_CMD_SET_INFO;
3686 else if (cmd == IPVS_CMD_GET_CONFIG)
3687 reply_cmd = IPVS_CMD_SET_CONFIG;
3689 pr_err("unknown Generic Netlink command\n");
3693 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3697 mutex_lock(&__ip_vs_mutex);
3699 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3701 goto nla_put_failure;
3704 case IPVS_CMD_GET_SERVICE:
3706 struct ip_vs_service *svc;
3708 svc = ip_vs_genl_find_service(ipvs,
3709 info->attrs[IPVS_CMD_ATTR_SERVICE]);
3714 ret = ip_vs_genl_fill_service(msg, svc);
3716 goto nla_put_failure;
3725 case IPVS_CMD_GET_CONFIG:
3727 struct ip_vs_timeout_user t;
3729 __ip_vs_get_timeouts(ipvs, &t);
3730 #ifdef CONFIG_IP_VS_PROTO_TCP
3731 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3733 nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3735 goto nla_put_failure;
3737 #ifdef CONFIG_IP_VS_PROTO_UDP
3738 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3739 goto nla_put_failure;
3745 case IPVS_CMD_GET_INFO:
3746 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3747 IP_VS_VERSION_CODE) ||
3748 nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3749 ip_vs_conn_tab_size))
3750 goto nla_put_failure;
3754 genlmsg_end(msg, reply);
3755 ret = genlmsg_reply(msg, info);
3759 pr_err("not enough space in Netlink message\n");
3765 mutex_unlock(&__ip_vs_mutex);
3771 static const struct genl_ops ip_vs_genl_ops[] = {
3773 .cmd = IPVS_CMD_NEW_SERVICE,
3774 .flags = GENL_ADMIN_PERM,
3775 .policy = ip_vs_cmd_policy,
3776 .doit = ip_vs_genl_set_cmd,
3779 .cmd = IPVS_CMD_SET_SERVICE,
3780 .flags = GENL_ADMIN_PERM,
3781 .policy = ip_vs_cmd_policy,
3782 .doit = ip_vs_genl_set_cmd,
3785 .cmd = IPVS_CMD_DEL_SERVICE,
3786 .flags = GENL_ADMIN_PERM,
3787 .policy = ip_vs_cmd_policy,
3788 .doit = ip_vs_genl_set_cmd,
3791 .cmd = IPVS_CMD_GET_SERVICE,
3792 .flags = GENL_ADMIN_PERM,
3793 .doit = ip_vs_genl_get_cmd,
3794 .dumpit = ip_vs_genl_dump_services,
3795 .policy = ip_vs_cmd_policy,
3798 .cmd = IPVS_CMD_NEW_DEST,
3799 .flags = GENL_ADMIN_PERM,
3800 .policy = ip_vs_cmd_policy,
3801 .doit = ip_vs_genl_set_cmd,
3804 .cmd = IPVS_CMD_SET_DEST,
3805 .flags = GENL_ADMIN_PERM,
3806 .policy = ip_vs_cmd_policy,
3807 .doit = ip_vs_genl_set_cmd,
3810 .cmd = IPVS_CMD_DEL_DEST,
3811 .flags = GENL_ADMIN_PERM,
3812 .policy = ip_vs_cmd_policy,
3813 .doit = ip_vs_genl_set_cmd,
3816 .cmd = IPVS_CMD_GET_DEST,
3817 .flags = GENL_ADMIN_PERM,
3818 .policy = ip_vs_cmd_policy,
3819 .dumpit = ip_vs_genl_dump_dests,
3822 .cmd = IPVS_CMD_NEW_DAEMON,
3823 .flags = GENL_ADMIN_PERM,
3824 .policy = ip_vs_cmd_policy,
3825 .doit = ip_vs_genl_set_daemon,
3828 .cmd = IPVS_CMD_DEL_DAEMON,
3829 .flags = GENL_ADMIN_PERM,
3830 .policy = ip_vs_cmd_policy,
3831 .doit = ip_vs_genl_set_daemon,
3834 .cmd = IPVS_CMD_GET_DAEMON,
3835 .flags = GENL_ADMIN_PERM,
3836 .dumpit = ip_vs_genl_dump_daemons,
3839 .cmd = IPVS_CMD_SET_CONFIG,
3840 .flags = GENL_ADMIN_PERM,
3841 .policy = ip_vs_cmd_policy,
3842 .doit = ip_vs_genl_set_cmd,
3845 .cmd = IPVS_CMD_GET_CONFIG,
3846 .flags = GENL_ADMIN_PERM,
3847 .doit = ip_vs_genl_get_cmd,
3850 .cmd = IPVS_CMD_GET_INFO,
3851 .flags = GENL_ADMIN_PERM,
3852 .doit = ip_vs_genl_get_cmd,
3855 .cmd = IPVS_CMD_ZERO,
3856 .flags = GENL_ADMIN_PERM,
3857 .policy = ip_vs_cmd_policy,
3858 .doit = ip_vs_genl_set_cmd,
3861 .cmd = IPVS_CMD_FLUSH,
3862 .flags = GENL_ADMIN_PERM,
3863 .doit = ip_vs_genl_set_cmd,
3867 static struct genl_family ip_vs_genl_family __ro_after_init = {
3869 .name = IPVS_GENL_NAME,
3870 .version = IPVS_GENL_VERSION,
3871 .maxattr = IPVS_CMD_ATTR_MAX,
3872 .netnsok = true, /* Make ipvsadm to work on netns */
3873 .module = THIS_MODULE,
3874 .ops = ip_vs_genl_ops,
3875 .n_ops = ARRAY_SIZE(ip_vs_genl_ops),
3878 static int __init ip_vs_genl_register(void)
3880 return genl_register_family(&ip_vs_genl_family);
3883 static void ip_vs_genl_unregister(void)
3885 genl_unregister_family(&ip_vs_genl_family);
3888 /* End of Generic Netlink interface definitions */
3891 * per netns intit/exit func.
3893 #ifdef CONFIG_SYSCTL
3894 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs)
3896 struct net *net = ipvs->net;
3898 struct ctl_table *tbl;
3900 atomic_set(&ipvs->dropentry, 0);
3901 spin_lock_init(&ipvs->dropentry_lock);
3902 spin_lock_init(&ipvs->droppacket_lock);
3903 spin_lock_init(&ipvs->securetcp_lock);
3905 if (!net_eq(net, &init_net)) {
3906 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3910 /* Don't export sysctls to unprivileged users */
3911 if (net->user_ns != &init_user_ns)
3912 tbl[0].procname = NULL;
3915 /* Initialize sysctl defaults */
3916 for (idx = 0; idx < ARRAY_SIZE(vs_vars); idx++) {
3917 if (tbl[idx].proc_handler == proc_do_defense_mode)
3918 tbl[idx].extra2 = ipvs;
3921 ipvs->sysctl_amemthresh = 1024;
3922 tbl[idx++].data = &ipvs->sysctl_amemthresh;
3923 ipvs->sysctl_am_droprate = 10;
3924 tbl[idx++].data = &ipvs->sysctl_am_droprate;
3925 tbl[idx++].data = &ipvs->sysctl_drop_entry;
3926 tbl[idx++].data = &ipvs->sysctl_drop_packet;
3927 #ifdef CONFIG_IP_VS_NFCT
3928 tbl[idx++].data = &ipvs->sysctl_conntrack;
3930 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3931 ipvs->sysctl_snat_reroute = 1;
3932 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3933 ipvs->sysctl_sync_ver = 1;
3934 tbl[idx++].data = &ipvs->sysctl_sync_ver;
3935 ipvs->sysctl_sync_ports = 1;
3936 tbl[idx++].data = &ipvs->sysctl_sync_ports;
3937 tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
3938 ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3939 tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3940 ipvs->sysctl_sync_sock_size = 0;
3941 tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3942 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3943 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3944 tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
3945 tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
3946 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3947 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3948 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3949 tbl[idx].data = &ipvs->sysctl_sync_threshold;
3950 tbl[idx].extra2 = ipvs;
3951 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3952 ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3953 tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3954 ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3955 tbl[idx++].data = &ipvs->sysctl_sync_retries;
3956 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3957 ipvs->sysctl_pmtu_disc = 1;
3958 tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
3959 tbl[idx++].data = &ipvs->sysctl_backup_only;
3960 ipvs->sysctl_conn_reuse_mode = 1;
3961 tbl[idx++].data = &ipvs->sysctl_conn_reuse_mode;
3962 tbl[idx++].data = &ipvs->sysctl_schedule_icmp;
3963 tbl[idx++].data = &ipvs->sysctl_ignore_tunneled;
3964 #ifdef CONFIG_IP_VS_DEBUG
3965 /* Global sysctls must be ro in non-init netns */
3966 if (!net_eq(net, &init_net))
3967 tbl[idx++].mode = 0444;
3970 ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3971 if (ipvs->sysctl_hdr == NULL) {
3972 if (!net_eq(net, &init_net))
3976 ip_vs_start_estimator(ipvs, &ipvs->tot_stats);
3977 ipvs->sysctl_tbl = tbl;
3978 /* Schedule defense work */
3979 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3980 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3985 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs)
3987 struct net *net = ipvs->net;
3989 cancel_delayed_work_sync(&ipvs->defense_work);
3990 cancel_work_sync(&ipvs->defense_work.work);
3991 unregister_net_sysctl_table(ipvs->sysctl_hdr);
3992 ip_vs_stop_estimator(ipvs, &ipvs->tot_stats);
3994 if (!net_eq(net, &init_net))
3995 kfree(ipvs->sysctl_tbl);
4000 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs) { return 0; }
4001 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs) { }
4005 static struct notifier_block ip_vs_dst_notifier = {
4006 .notifier_call = ip_vs_dst_event,
4007 #ifdef CONFIG_IP_VS_IPV6
4008 .priority = ADDRCONF_NOTIFY_PRIORITY + 5,
4012 int __net_init ip_vs_control_net_init(struct netns_ipvs *ipvs)
4016 /* Initialize rs_table */
4017 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
4018 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
4020 INIT_LIST_HEAD(&ipvs->dest_trash);
4021 spin_lock_init(&ipvs->dest_trash_lock);
4022 timer_setup(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire, 0);
4023 atomic_set(&ipvs->ftpsvc_counter, 0);
4024 atomic_set(&ipvs->nullsvc_counter, 0);
4025 atomic_set(&ipvs->conn_out_counter, 0);
4028 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
4029 if (!ipvs->tot_stats.cpustats)
4032 for_each_possible_cpu(i) {
4033 struct ip_vs_cpu_stats *ipvs_tot_stats;
4034 ipvs_tot_stats = per_cpu_ptr(ipvs->tot_stats.cpustats, i);
4035 u64_stats_init(&ipvs_tot_stats->syncp);
4038 spin_lock_init(&ipvs->tot_stats.lock);
4040 proc_create_net("ip_vs", 0, ipvs->net->proc_net, &ip_vs_info_seq_ops,
4041 sizeof(struct ip_vs_iter));
4042 proc_create_net_single("ip_vs_stats", 0, ipvs->net->proc_net,
4043 ip_vs_stats_show, NULL);
4044 proc_create_net_single("ip_vs_stats_percpu", 0, ipvs->net->proc_net,
4045 ip_vs_stats_percpu_show, NULL);
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);