2 * IPVS An implementation of the IP virtual server support for the
3 * LINUX operating system. IPVS is now implemented as a module
4 * over the NetFilter framework. IPVS can be used to build a
5 * high-performance and highly available server based on a
8 * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
9 * Peter Kese <peter.kese@ijs.si>
10 * Julian Anastasov <ja@ssi.bg>
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
21 #define KMSG_COMPONENT "IPVS"
22 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
24 #include <linux/module.h>
25 #include <linux/init.h>
26 #include <linux/types.h>
27 #include <linux/capability.h>
29 #include <linux/sysctl.h>
30 #include <linux/proc_fs.h>
31 #include <linux/workqueue.h>
32 #include <linux/swap.h>
33 #include <linux/seq_file.h>
34 #include <linux/slab.h>
36 #include <linux/netfilter.h>
37 #include <linux/netfilter_ipv4.h>
38 #include <linux/mutex.h>
40 #include <net/net_namespace.h>
41 #include <linux/nsproxy.h>
43 #ifdef CONFIG_IP_VS_IPV6
45 #include <net/ip6_route.h>
47 #include <net/route.h>
49 #include <net/genetlink.h>
51 #include <asm/uaccess.h>
53 #include <net/ip_vs.h>
55 /* semaphore for IPVS sockopts. And, [gs]etsockopt may sleep. */
56 static DEFINE_MUTEX(__ip_vs_mutex);
58 /* lock for service table */
59 static DEFINE_RWLOCK(__ip_vs_svc_lock);
61 /* sysctl variables */
63 #ifdef CONFIG_IP_VS_DEBUG
64 static int sysctl_ip_vs_debug_level = 0;
66 int ip_vs_get_debug_level(void)
68 return sysctl_ip_vs_debug_level;
74 static void __ip_vs_del_service(struct ip_vs_service *svc);
77 #ifdef CONFIG_IP_VS_IPV6
78 /* Taken from rt6_fill_node() in net/ipv6/route.c, is there a better way? */
79 static bool __ip_vs_addr_is_local_v6(struct net *net,
80 const struct in6_addr *addr)
85 struct dst_entry *dst = ip6_route_output(net, NULL, &fl6);
88 is_local = !dst->error && dst->dev && (dst->dev->flags & IFF_LOOPBACK);
97 * update_defense_level is called from keventd and from sysctl,
98 * so it needs to protect itself from softirqs
100 static void update_defense_level(struct netns_ipvs *ipvs)
103 static int old_secure_tcp = 0;
108 /* we only count free and buffered memory (in pages) */
110 availmem = i.freeram + i.bufferram;
111 /* however in linux 2.5 the i.bufferram is total page cache size,
113 /* si_swapinfo(&i); */
114 /* availmem = availmem - (i.totalswap - i.freeswap); */
116 nomem = (availmem < ipvs->sysctl_amemthresh);
121 spin_lock(&ipvs->dropentry_lock);
122 switch (ipvs->sysctl_drop_entry) {
124 atomic_set(&ipvs->dropentry, 0);
128 atomic_set(&ipvs->dropentry, 1);
129 ipvs->sysctl_drop_entry = 2;
131 atomic_set(&ipvs->dropentry, 0);
136 atomic_set(&ipvs->dropentry, 1);
138 atomic_set(&ipvs->dropentry, 0);
139 ipvs->sysctl_drop_entry = 1;
143 atomic_set(&ipvs->dropentry, 1);
146 spin_unlock(&ipvs->dropentry_lock);
149 spin_lock(&ipvs->droppacket_lock);
150 switch (ipvs->sysctl_drop_packet) {
156 ipvs->drop_rate = ipvs->drop_counter
157 = ipvs->sysctl_amemthresh /
158 (ipvs->sysctl_amemthresh-availmem);
159 ipvs->sysctl_drop_packet = 2;
166 ipvs->drop_rate = ipvs->drop_counter
167 = ipvs->sysctl_amemthresh /
168 (ipvs->sysctl_amemthresh-availmem);
171 ipvs->sysctl_drop_packet = 1;
175 ipvs->drop_rate = ipvs->sysctl_am_droprate;
178 spin_unlock(&ipvs->droppacket_lock);
181 spin_lock(&ipvs->securetcp_lock);
182 switch (ipvs->sysctl_secure_tcp) {
184 if (old_secure_tcp >= 2)
189 if (old_secure_tcp < 2)
191 ipvs->sysctl_secure_tcp = 2;
193 if (old_secure_tcp >= 2)
199 if (old_secure_tcp < 2)
202 if (old_secure_tcp >= 2)
204 ipvs->sysctl_secure_tcp = 1;
208 if (old_secure_tcp < 2)
212 old_secure_tcp = ipvs->sysctl_secure_tcp;
214 ip_vs_protocol_timeout_change(ipvs,
215 ipvs->sysctl_secure_tcp > 1);
216 spin_unlock(&ipvs->securetcp_lock);
223 * Timer for checking the defense
225 #define DEFENSE_TIMER_PERIOD 1*HZ
227 static void defense_work_handler(struct work_struct *work)
229 struct netns_ipvs *ipvs =
230 container_of(work, struct netns_ipvs, defense_work.work);
232 update_defense_level(ipvs);
233 if (atomic_read(&ipvs->dropentry))
234 ip_vs_random_dropentry(ipvs->net);
235 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
240 ip_vs_use_count_inc(void)
242 return try_module_get(THIS_MODULE);
246 ip_vs_use_count_dec(void)
248 module_put(THIS_MODULE);
253 * Hash table: for virtual service lookups
255 #define IP_VS_SVC_TAB_BITS 8
256 #define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
257 #define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)
259 /* the service table hashed by <protocol, addr, port> */
260 static struct list_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE];
261 /* the service table hashed by fwmark */
262 static struct list_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE];
266 * Returns hash value for virtual service
268 static inline unsigned int
269 ip_vs_svc_hashkey(struct net *net, int af, unsigned int proto,
270 const union nf_inet_addr *addr, __be16 port)
272 register unsigned int porth = ntohs(port);
273 __be32 addr_fold = addr->ip;
275 #ifdef CONFIG_IP_VS_IPV6
277 addr_fold = addr->ip6[0]^addr->ip6[1]^
278 addr->ip6[2]^addr->ip6[3];
280 addr_fold ^= ((size_t)net>>8);
282 return (proto^ntohl(addr_fold)^(porth>>IP_VS_SVC_TAB_BITS)^porth)
283 & IP_VS_SVC_TAB_MASK;
287 * Returns hash value of fwmark for virtual service lookup
289 static inline unsigned int ip_vs_svc_fwm_hashkey(struct net *net, __u32 fwmark)
291 return (((size_t)net>>8) ^ fwmark) & IP_VS_SVC_TAB_MASK;
295 * Hashes a service in the ip_vs_svc_table by <netns,proto,addr,port>
296 * or in the ip_vs_svc_fwm_table by fwmark.
297 * Should be called with locked tables.
299 static int ip_vs_svc_hash(struct ip_vs_service *svc)
303 if (svc->flags & IP_VS_SVC_F_HASHED) {
304 pr_err("%s(): request for already hashed, called from %pF\n",
305 __func__, __builtin_return_address(0));
309 if (svc->fwmark == 0) {
311 * Hash it by <netns,protocol,addr,port> in ip_vs_svc_table
313 hash = ip_vs_svc_hashkey(svc->net, svc->af, svc->protocol,
314 &svc->addr, svc->port);
315 list_add(&svc->s_list, &ip_vs_svc_table[hash]);
318 * Hash it by fwmark in svc_fwm_table
320 hash = ip_vs_svc_fwm_hashkey(svc->net, svc->fwmark);
321 list_add(&svc->f_list, &ip_vs_svc_fwm_table[hash]);
324 svc->flags |= IP_VS_SVC_F_HASHED;
325 /* increase its refcnt because it is referenced by the svc table */
326 atomic_inc(&svc->refcnt);
332 * Unhashes a service from svc_table / svc_fwm_table.
333 * Should be called with locked tables.
335 static int ip_vs_svc_unhash(struct ip_vs_service *svc)
337 if (!(svc->flags & IP_VS_SVC_F_HASHED)) {
338 pr_err("%s(): request for unhash flagged, called from %pF\n",
339 __func__, __builtin_return_address(0));
343 if (svc->fwmark == 0) {
344 /* Remove it from the svc_table table */
345 list_del(&svc->s_list);
347 /* Remove it from the svc_fwm_table table */
348 list_del(&svc->f_list);
351 svc->flags &= ~IP_VS_SVC_F_HASHED;
352 atomic_dec(&svc->refcnt);
358 * Get service by {netns, proto,addr,port} in the service table.
360 static inline struct ip_vs_service *
361 __ip_vs_service_find(struct net *net, int af, __u16 protocol,
362 const union nf_inet_addr *vaddr, __be16 vport)
365 struct ip_vs_service *svc;
367 /* Check for "full" addressed entries */
368 hash = ip_vs_svc_hashkey(net, af, protocol, vaddr, vport);
370 list_for_each_entry(svc, &ip_vs_svc_table[hash], s_list){
372 && ip_vs_addr_equal(af, &svc->addr, vaddr)
373 && (svc->port == vport)
374 && (svc->protocol == protocol)
375 && net_eq(svc->net, net)) {
386 * Get service by {fwmark} in the service table.
388 static inline struct ip_vs_service *
389 __ip_vs_svc_fwm_find(struct net *net, int af, __u32 fwmark)
392 struct ip_vs_service *svc;
394 /* Check for fwmark addressed entries */
395 hash = ip_vs_svc_fwm_hashkey(net, fwmark);
397 list_for_each_entry(svc, &ip_vs_svc_fwm_table[hash], f_list) {
398 if (svc->fwmark == fwmark && svc->af == af
399 && net_eq(svc->net, net)) {
408 struct ip_vs_service *
409 ip_vs_service_get(struct net *net, int af, __u32 fwmark, __u16 protocol,
410 const union nf_inet_addr *vaddr, __be16 vport)
412 struct ip_vs_service *svc;
413 struct netns_ipvs *ipvs = net_ipvs(net);
415 read_lock(&__ip_vs_svc_lock);
418 * Check the table hashed by fwmark first
421 svc = __ip_vs_svc_fwm_find(net, af, fwmark);
427 * Check the table hashed by <protocol,addr,port>
428 * for "full" addressed entries
430 svc = __ip_vs_service_find(net, af, protocol, vaddr, vport);
433 && protocol == IPPROTO_TCP
434 && atomic_read(&ipvs->ftpsvc_counter)
435 && (vport == FTPDATA || ntohs(vport) >= PROT_SOCK)) {
437 * Check if ftp service entry exists, the packet
438 * might belong to FTP data connections.
440 svc = __ip_vs_service_find(net, af, protocol, vaddr, FTPPORT);
444 && atomic_read(&ipvs->nullsvc_counter)) {
446 * Check if the catch-all port (port zero) exists
448 svc = __ip_vs_service_find(net, af, protocol, vaddr, 0);
453 atomic_inc(&svc->usecnt);
454 read_unlock(&__ip_vs_svc_lock);
456 IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
457 fwmark, ip_vs_proto_name(protocol),
458 IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
459 svc ? "hit" : "not hit");
466 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
468 atomic_inc(&svc->refcnt);
473 __ip_vs_unbind_svc(struct ip_vs_dest *dest)
475 struct ip_vs_service *svc = dest->svc;
478 if (atomic_dec_and_test(&svc->refcnt)) {
479 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
481 IP_VS_DBG_ADDR(svc->af, &svc->addr),
482 ntohs(svc->port), atomic_read(&svc->usecnt));
483 free_percpu(svc->stats.cpustats);
490 * Returns hash value for real service
492 static inline unsigned int ip_vs_rs_hashkey(int af,
493 const union nf_inet_addr *addr,
496 register unsigned int porth = ntohs(port);
497 __be32 addr_fold = addr->ip;
499 #ifdef CONFIG_IP_VS_IPV6
501 addr_fold = addr->ip6[0]^addr->ip6[1]^
502 addr->ip6[2]^addr->ip6[3];
505 return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
510 * Hashes ip_vs_dest in rs_table by <proto,addr,port>.
511 * should be called with locked tables.
513 static int ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
517 if (!list_empty(&dest->d_list)) {
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 list_add(&dest->d_list, &ipvs->rs_table[hash]);
533 * UNhashes ip_vs_dest from rs_table.
534 * should be called with locked tables.
536 static int ip_vs_rs_unhash(struct ip_vs_dest *dest)
539 * Remove it from the rs_table table.
541 if (!list_empty(&dest->d_list)) {
542 list_del(&dest->d_list);
543 INIT_LIST_HEAD(&dest->d_list);
550 * Lookup real service by <proto,addr,port> in the real service table.
553 ip_vs_lookup_real_service(struct net *net, int af, __u16 protocol,
554 const union nf_inet_addr *daddr,
557 struct netns_ipvs *ipvs = net_ipvs(net);
559 struct ip_vs_dest *dest;
562 * Check for "full" addressed entries
563 * Return the first found entry
565 hash = ip_vs_rs_hashkey(af, daddr, dport);
567 read_lock(&ipvs->rs_lock);
568 list_for_each_entry(dest, &ipvs->rs_table[hash], d_list) {
570 && ip_vs_addr_equal(af, &dest->addr, daddr)
571 && (dest->port == dport)
572 && ((dest->protocol == protocol) ||
575 read_unlock(&ipvs->rs_lock);
579 read_unlock(&ipvs->rs_lock);
585 * Lookup destination by {addr,port} in the given service
587 static struct ip_vs_dest *
588 ip_vs_lookup_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
591 struct ip_vs_dest *dest;
594 * Find the destination for the given service
596 list_for_each_entry(dest, &svc->destinations, n_list) {
597 if ((dest->af == svc->af)
598 && ip_vs_addr_equal(svc->af, &dest->addr, daddr)
599 && (dest->port == dport)) {
609 * Find destination by {daddr,dport,vaddr,protocol}
610 * Cretaed to be used in ip_vs_process_message() in
611 * the backup synchronization daemon. It finds the
612 * destination to be bound to the received connection
615 * ip_vs_lookup_real_service() looked promissing, but
616 * seems not working as expected.
618 struct ip_vs_dest *ip_vs_find_dest(struct net *net, int af,
619 const union nf_inet_addr *daddr,
621 const union nf_inet_addr *vaddr,
622 __be16 vport, __u16 protocol, __u32 fwmark,
625 struct ip_vs_dest *dest;
626 struct ip_vs_service *svc;
629 svc = ip_vs_service_get(net, af, fwmark, protocol, vaddr, vport);
632 if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
634 dest = ip_vs_lookup_dest(svc, daddr, port);
636 dest = ip_vs_lookup_dest(svc, daddr, port ^ dport);
638 atomic_inc(&dest->refcnt);
639 ip_vs_service_put(svc);
644 * Lookup dest by {svc,addr,port} in the destination trash.
645 * The destination trash is used to hold the destinations that are removed
646 * from the service table but are still referenced by some conn entries.
647 * The reason to add the destination trash is when the dest is temporary
648 * down (either by administrator or by monitor program), the dest can be
649 * picked back from the trash, the remaining connections to the dest can
650 * continue, and the counting information of the dest is also useful for
653 static struct ip_vs_dest *
654 ip_vs_trash_get_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
657 struct ip_vs_dest *dest, *nxt;
658 struct netns_ipvs *ipvs = net_ipvs(svc->net);
661 * Find the destination in trash
663 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, n_list) {
664 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
667 IP_VS_DBG_ADDR(svc->af, &dest->addr),
669 atomic_read(&dest->refcnt));
670 if (dest->af == svc->af &&
671 ip_vs_addr_equal(svc->af, &dest->addr, daddr) &&
672 dest->port == dport &&
673 dest->vfwmark == svc->fwmark &&
674 dest->protocol == svc->protocol &&
676 (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
677 dest->vport == svc->port))) {
683 * Try to purge the destination from trash if not referenced
685 if (atomic_read(&dest->refcnt) == 1) {
686 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u "
689 IP_VS_DBG_ADDR(svc->af, &dest->addr),
691 list_del(&dest->n_list);
692 ip_vs_dst_reset(dest);
693 __ip_vs_unbind_svc(dest);
694 free_percpu(dest->stats.cpustats);
704 * Clean up all the destinations in the trash
705 * Called by the ip_vs_control_cleanup()
707 * When the ip_vs_control_clearup is activated by ipvs module exit,
708 * the service tables must have been flushed and all the connections
709 * are expired, and the refcnt of each destination in the trash must
710 * be 1, so we simply release them here.
712 static void ip_vs_trash_cleanup(struct net *net)
714 struct ip_vs_dest *dest, *nxt;
715 struct netns_ipvs *ipvs = net_ipvs(net);
717 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, n_list) {
718 list_del(&dest->n_list);
719 ip_vs_dst_reset(dest);
720 __ip_vs_unbind_svc(dest);
721 free_percpu(dest->stats.cpustats);
727 ip_vs_copy_stats(struct ip_vs_stats_user *dst, struct ip_vs_stats *src)
729 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->ustats.c - src->ustats0.c
731 spin_lock_bh(&src->lock);
733 IP_VS_SHOW_STATS_COUNTER(conns);
734 IP_VS_SHOW_STATS_COUNTER(inpkts);
735 IP_VS_SHOW_STATS_COUNTER(outpkts);
736 IP_VS_SHOW_STATS_COUNTER(inbytes);
737 IP_VS_SHOW_STATS_COUNTER(outbytes);
739 ip_vs_read_estimator(dst, src);
741 spin_unlock_bh(&src->lock);
745 ip_vs_zero_stats(struct ip_vs_stats *stats)
747 spin_lock_bh(&stats->lock);
749 /* get current counters as zero point, rates are zeroed */
751 #define IP_VS_ZERO_STATS_COUNTER(c) stats->ustats0.c = stats->ustats.c
753 IP_VS_ZERO_STATS_COUNTER(conns);
754 IP_VS_ZERO_STATS_COUNTER(inpkts);
755 IP_VS_ZERO_STATS_COUNTER(outpkts);
756 IP_VS_ZERO_STATS_COUNTER(inbytes);
757 IP_VS_ZERO_STATS_COUNTER(outbytes);
759 ip_vs_zero_estimator(stats);
761 spin_unlock_bh(&stats->lock);
765 * Update a destination in the given service
768 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
769 struct ip_vs_dest_user_kern *udest, int add)
771 struct netns_ipvs *ipvs = net_ipvs(svc->net);
774 /* set the weight and the flags */
775 atomic_set(&dest->weight, udest->weight);
776 conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
777 conn_flags |= IP_VS_CONN_F_INACTIVE;
779 /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
780 if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
781 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
784 * Put the real service in rs_table if not present.
785 * For now only for NAT!
787 write_lock_bh(&ipvs->rs_lock);
788 ip_vs_rs_hash(ipvs, dest);
789 write_unlock_bh(&ipvs->rs_lock);
791 atomic_set(&dest->conn_flags, conn_flags);
793 /* bind the service */
795 __ip_vs_bind_svc(dest, svc);
797 if (dest->svc != svc) {
798 __ip_vs_unbind_svc(dest);
799 ip_vs_zero_stats(&dest->stats);
800 __ip_vs_bind_svc(dest, svc);
804 /* set the dest status flags */
805 dest->flags |= IP_VS_DEST_F_AVAILABLE;
807 if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
808 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
809 dest->u_threshold = udest->u_threshold;
810 dest->l_threshold = udest->l_threshold;
812 spin_lock_bh(&dest->dst_lock);
813 ip_vs_dst_reset(dest);
814 spin_unlock_bh(&dest->dst_lock);
817 ip_vs_start_estimator(svc->net, &dest->stats);
819 write_lock_bh(&__ip_vs_svc_lock);
821 /* Wait until all other svc users go away */
822 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
825 list_add(&dest->n_list, &svc->destinations);
829 /* call the update_service, because server weight may be changed */
830 if (svc->scheduler->update_service)
831 svc->scheduler->update_service(svc);
833 write_unlock_bh(&__ip_vs_svc_lock);
838 * Create a destination for the given service
841 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
842 struct ip_vs_dest **dest_p)
844 struct ip_vs_dest *dest;
849 #ifdef CONFIG_IP_VS_IPV6
850 if (svc->af == AF_INET6) {
851 atype = ipv6_addr_type(&udest->addr.in6);
852 if ((!(atype & IPV6_ADDR_UNICAST) ||
853 atype & IPV6_ADDR_LINKLOCAL) &&
854 !__ip_vs_addr_is_local_v6(svc->net, &udest->addr.in6))
859 atype = inet_addr_type(svc->net, udest->addr.ip);
860 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
864 dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
868 dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
869 if (!dest->stats.cpustats)
873 dest->protocol = svc->protocol;
874 dest->vaddr = svc->addr;
875 dest->vport = svc->port;
876 dest->vfwmark = svc->fwmark;
877 ip_vs_addr_copy(svc->af, &dest->addr, &udest->addr);
878 dest->port = udest->port;
880 atomic_set(&dest->activeconns, 0);
881 atomic_set(&dest->inactconns, 0);
882 atomic_set(&dest->persistconns, 0);
883 atomic_set(&dest->refcnt, 1);
885 INIT_LIST_HEAD(&dest->d_list);
886 spin_lock_init(&dest->dst_lock);
887 spin_lock_init(&dest->stats.lock);
888 __ip_vs_update_dest(svc, dest, udest, 1);
902 * Add a destination into an existing service
905 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
907 struct ip_vs_dest *dest;
908 union nf_inet_addr daddr;
909 __be16 dport = udest->port;
914 if (udest->weight < 0) {
915 pr_err("%s(): server weight less than zero\n", __func__);
919 if (udest->l_threshold > udest->u_threshold) {
920 pr_err("%s(): lower threshold is higher than upper threshold\n",
925 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
928 * Check if the dest already exists in the list
930 dest = ip_vs_lookup_dest(svc, &daddr, dport);
933 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
938 * Check if the dest already exists in the trash and
939 * is from the same service
941 dest = ip_vs_trash_get_dest(svc, &daddr, dport);
944 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
945 "dest->refcnt=%d, service %u/%s:%u\n",
946 IP_VS_DBG_ADDR(svc->af, &daddr), ntohs(dport),
947 atomic_read(&dest->refcnt),
949 IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
953 * Get the destination from the trash
955 list_del(&dest->n_list);
957 __ip_vs_update_dest(svc, dest, udest, 1);
961 * Allocate and initialize the dest structure
963 ret = ip_vs_new_dest(svc, udest, &dest);
972 * Edit a destination in the given service
975 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
977 struct ip_vs_dest *dest;
978 union nf_inet_addr daddr;
979 __be16 dport = udest->port;
983 if (udest->weight < 0) {
984 pr_err("%s(): server weight less than zero\n", __func__);
988 if (udest->l_threshold > udest->u_threshold) {
989 pr_err("%s(): lower threshold is higher than upper threshold\n",
994 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
997 * Lookup the destination list
999 dest = ip_vs_lookup_dest(svc, &daddr, dport);
1002 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1006 __ip_vs_update_dest(svc, dest, udest, 0);
1014 * Delete a destination (must be already unlinked from the service)
1016 static void __ip_vs_del_dest(struct net *net, struct ip_vs_dest *dest)
1018 struct netns_ipvs *ipvs = net_ipvs(net);
1020 ip_vs_stop_estimator(net, &dest->stats);
1023 * Remove it from the d-linked list with the real services.
1025 write_lock_bh(&ipvs->rs_lock);
1026 ip_vs_rs_unhash(dest);
1027 write_unlock_bh(&ipvs->rs_lock);
1030 * Decrease the refcnt of the dest, and free the dest
1031 * if nobody refers to it (refcnt=0). Otherwise, throw
1032 * the destination into the trash.
1034 if (atomic_dec_and_test(&dest->refcnt)) {
1035 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u\n",
1037 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1039 ip_vs_dst_reset(dest);
1040 /* simply decrease svc->refcnt here, let the caller check
1041 and release the service if nobody refers to it.
1042 Only user context can release destination and service,
1043 and only one user context can update virtual service at a
1044 time, so the operation here is OK */
1045 atomic_dec(&dest->svc->refcnt);
1046 free_percpu(dest->stats.cpustats);
1049 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, "
1050 "dest->refcnt=%d\n",
1051 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1053 atomic_read(&dest->refcnt));
1054 list_add(&dest->n_list, &ipvs->dest_trash);
1055 atomic_inc(&dest->refcnt);
1061 * Unlink a destination from the given service
1063 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1064 struct ip_vs_dest *dest,
1067 dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1070 * Remove it from the d-linked destination list.
1072 list_del(&dest->n_list);
1076 * Call the update_service function of its scheduler
1078 if (svcupd && svc->scheduler->update_service)
1079 svc->scheduler->update_service(svc);
1084 * Delete a destination server in the given service
1087 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1089 struct ip_vs_dest *dest;
1090 __be16 dport = udest->port;
1094 dest = ip_vs_lookup_dest(svc, &udest->addr, dport);
1097 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1101 write_lock_bh(&__ip_vs_svc_lock);
1104 * Wait until all other svc users go away.
1106 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1109 * Unlink dest from the service
1111 __ip_vs_unlink_dest(svc, dest, 1);
1113 write_unlock_bh(&__ip_vs_svc_lock);
1116 * Delete the destination
1118 __ip_vs_del_dest(svc->net, dest);
1127 * Add a service into the service hash table
1130 ip_vs_add_service(struct net *net, struct ip_vs_service_user_kern *u,
1131 struct ip_vs_service **svc_p)
1134 struct ip_vs_scheduler *sched = NULL;
1135 struct ip_vs_pe *pe = NULL;
1136 struct ip_vs_service *svc = NULL;
1137 struct netns_ipvs *ipvs = net_ipvs(net);
1139 /* increase the module use count */
1140 ip_vs_use_count_inc();
1142 /* Lookup the scheduler by 'u->sched_name' */
1143 sched = ip_vs_scheduler_get(u->sched_name);
1144 if (sched == NULL) {
1145 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1150 if (u->pe_name && *u->pe_name) {
1151 pe = ip_vs_pe_getbyname(u->pe_name);
1153 pr_info("persistence engine module ip_vs_pe_%s "
1154 "not found\n", u->pe_name);
1160 #ifdef CONFIG_IP_VS_IPV6
1161 if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1167 svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1169 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1173 svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1174 if (!svc->stats.cpustats) {
1179 /* I'm the first user of the service */
1180 atomic_set(&svc->usecnt, 0);
1181 atomic_set(&svc->refcnt, 0);
1184 svc->protocol = u->protocol;
1185 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1186 svc->port = u->port;
1187 svc->fwmark = u->fwmark;
1188 svc->flags = u->flags;
1189 svc->timeout = u->timeout * HZ;
1190 svc->netmask = u->netmask;
1193 INIT_LIST_HEAD(&svc->destinations);
1194 rwlock_init(&svc->sched_lock);
1195 spin_lock_init(&svc->stats.lock);
1197 /* Bind the scheduler */
1198 ret = ip_vs_bind_scheduler(svc, sched);
1203 /* Bind the ct retriever */
1204 ip_vs_bind_pe(svc, pe);
1207 /* Update the virtual service counters */
1208 if (svc->port == FTPPORT)
1209 atomic_inc(&ipvs->ftpsvc_counter);
1210 else if (svc->port == 0)
1211 atomic_inc(&ipvs->nullsvc_counter);
1213 ip_vs_start_estimator(net, &svc->stats);
1215 /* Count only IPv4 services for old get/setsockopt interface */
1216 if (svc->af == AF_INET)
1217 ipvs->num_services++;
1219 /* Hash the service into the service table */
1220 write_lock_bh(&__ip_vs_svc_lock);
1221 ip_vs_svc_hash(svc);
1222 write_unlock_bh(&__ip_vs_svc_lock);
1225 /* Now there is a service - full throttle */
1232 ip_vs_unbind_scheduler(svc);
1235 ip_vs_app_inc_put(svc->inc);
1238 if (svc->stats.cpustats)
1239 free_percpu(svc->stats.cpustats);
1242 ip_vs_scheduler_put(sched);
1245 /* decrease the module use count */
1246 ip_vs_use_count_dec();
1253 * Edit a service and bind it with a new scheduler
1256 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1258 struct ip_vs_scheduler *sched, *old_sched;
1259 struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1263 * Lookup the scheduler, by 'u->sched_name'
1265 sched = ip_vs_scheduler_get(u->sched_name);
1266 if (sched == NULL) {
1267 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1272 if (u->pe_name && *u->pe_name) {
1273 pe = ip_vs_pe_getbyname(u->pe_name);
1275 pr_info("persistence engine module ip_vs_pe_%s "
1276 "not found\n", u->pe_name);
1283 #ifdef CONFIG_IP_VS_IPV6
1284 if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1290 write_lock_bh(&__ip_vs_svc_lock);
1293 * Wait until all other svc users go away.
1295 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1298 * Set the flags and timeout value
1300 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1301 svc->timeout = u->timeout * HZ;
1302 svc->netmask = u->netmask;
1304 old_sched = svc->scheduler;
1305 if (sched != old_sched) {
1307 * Unbind the old scheduler
1309 if ((ret = ip_vs_unbind_scheduler(svc))) {
1315 * Bind the new scheduler
1317 if ((ret = ip_vs_bind_scheduler(svc, sched))) {
1319 * If ip_vs_bind_scheduler fails, restore the old
1321 * The main reason of failure is out of memory.
1323 * The question is if the old scheduler can be
1324 * restored all the time. TODO: if it cannot be
1325 * restored some time, we must delete the service,
1326 * otherwise the system may crash.
1328 ip_vs_bind_scheduler(svc, old_sched);
1336 ip_vs_unbind_pe(svc);
1337 ip_vs_bind_pe(svc, pe);
1341 write_unlock_bh(&__ip_vs_svc_lock);
1343 ip_vs_scheduler_put(old_sched);
1344 ip_vs_pe_put(old_pe);
1350 * Delete a service from the service list
1351 * - The service must be unlinked, unlocked and not referenced!
1352 * - We are called under _bh lock
1354 static void __ip_vs_del_service(struct ip_vs_service *svc)
1356 struct ip_vs_dest *dest, *nxt;
1357 struct ip_vs_scheduler *old_sched;
1358 struct ip_vs_pe *old_pe;
1359 struct netns_ipvs *ipvs = net_ipvs(svc->net);
1361 pr_info("%s: enter\n", __func__);
1363 /* Count only IPv4 services for old get/setsockopt interface */
1364 if (svc->af == AF_INET)
1365 ipvs->num_services--;
1367 ip_vs_stop_estimator(svc->net, &svc->stats);
1369 /* Unbind scheduler */
1370 old_sched = svc->scheduler;
1371 ip_vs_unbind_scheduler(svc);
1372 ip_vs_scheduler_put(old_sched);
1374 /* Unbind persistence engine */
1376 ip_vs_unbind_pe(svc);
1377 ip_vs_pe_put(old_pe);
1379 /* Unbind app inc */
1381 ip_vs_app_inc_put(svc->inc);
1386 * Unlink the whole destination list
1388 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1389 __ip_vs_unlink_dest(svc, dest, 0);
1390 __ip_vs_del_dest(svc->net, dest);
1394 * Update the virtual service counters
1396 if (svc->port == FTPPORT)
1397 atomic_dec(&ipvs->ftpsvc_counter);
1398 else if (svc->port == 0)
1399 atomic_dec(&ipvs->nullsvc_counter);
1402 * Free the service if nobody refers to it
1404 if (atomic_read(&svc->refcnt) == 0) {
1405 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
1407 IP_VS_DBG_ADDR(svc->af, &svc->addr),
1408 ntohs(svc->port), atomic_read(&svc->usecnt));
1409 free_percpu(svc->stats.cpustats);
1413 /* decrease the module use count */
1414 ip_vs_use_count_dec();
1418 * Unlink a service from list and try to delete it if its refcnt reached 0
1420 static void ip_vs_unlink_service(struct ip_vs_service *svc)
1423 * Unhash it from the service table
1425 write_lock_bh(&__ip_vs_svc_lock);
1427 ip_vs_svc_unhash(svc);
1430 * Wait until all the svc users go away.
1432 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1434 __ip_vs_del_service(svc);
1436 write_unlock_bh(&__ip_vs_svc_lock);
1440 * Delete a service from the service list
1442 static int ip_vs_del_service(struct ip_vs_service *svc)
1446 ip_vs_unlink_service(svc);
1453 * Flush all the virtual services
1455 static int ip_vs_flush(struct net *net)
1458 struct ip_vs_service *svc, *nxt;
1461 * Flush the service table hashed by <netns,protocol,addr,port>
1463 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1464 list_for_each_entry_safe(svc, nxt, &ip_vs_svc_table[idx],
1466 if (net_eq(svc->net, net))
1467 ip_vs_unlink_service(svc);
1472 * Flush the service table hashed by fwmark
1474 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1475 list_for_each_entry_safe(svc, nxt,
1476 &ip_vs_svc_fwm_table[idx], f_list) {
1477 if (net_eq(svc->net, net))
1478 ip_vs_unlink_service(svc);
1486 * Delete service by {netns} in the service table.
1487 * Called by __ip_vs_cleanup()
1489 void ip_vs_service_net_cleanup(struct net *net)
1492 /* Check for "full" addressed entries */
1493 mutex_lock(&__ip_vs_mutex);
1495 mutex_unlock(&__ip_vs_mutex);
1499 * Release dst hold by dst_cache
1502 __ip_vs_dev_reset(struct ip_vs_dest *dest, struct net_device *dev)
1504 spin_lock_bh(&dest->dst_lock);
1505 if (dest->dst_cache && dest->dst_cache->dev == dev) {
1506 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1508 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1510 atomic_read(&dest->refcnt));
1511 ip_vs_dst_reset(dest);
1513 spin_unlock_bh(&dest->dst_lock);
1517 * Netdev event receiver
1518 * Currently only NETDEV_UNREGISTER is handled, i.e. if we hold a reference to
1519 * a device that is "unregister" it must be released.
1521 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1524 struct net_device *dev = ptr;
1525 struct net *net = dev_net(dev);
1526 struct netns_ipvs *ipvs = net_ipvs(net);
1527 struct ip_vs_service *svc;
1528 struct ip_vs_dest *dest;
1531 if (event != NETDEV_UNREGISTER || !ipvs)
1533 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1535 mutex_lock(&__ip_vs_mutex);
1536 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1537 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1538 if (net_eq(svc->net, net)) {
1539 list_for_each_entry(dest, &svc->destinations,
1541 __ip_vs_dev_reset(dest, dev);
1546 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1547 if (net_eq(svc->net, net)) {
1548 list_for_each_entry(dest, &svc->destinations,
1550 __ip_vs_dev_reset(dest, dev);
1557 list_for_each_entry(dest, &ipvs->dest_trash, n_list) {
1558 __ip_vs_dev_reset(dest, dev);
1560 mutex_unlock(&__ip_vs_mutex);
1566 * Zero counters in a service or all services
1568 static int ip_vs_zero_service(struct ip_vs_service *svc)
1570 struct ip_vs_dest *dest;
1572 write_lock_bh(&__ip_vs_svc_lock);
1573 list_for_each_entry(dest, &svc->destinations, n_list) {
1574 ip_vs_zero_stats(&dest->stats);
1576 ip_vs_zero_stats(&svc->stats);
1577 write_unlock_bh(&__ip_vs_svc_lock);
1581 static int ip_vs_zero_all(struct net *net)
1584 struct ip_vs_service *svc;
1586 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1587 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1588 if (net_eq(svc->net, net))
1589 ip_vs_zero_service(svc);
1593 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1594 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1595 if (net_eq(svc->net, net))
1596 ip_vs_zero_service(svc);
1600 ip_vs_zero_stats(&net_ipvs(net)->tot_stats);
1604 #ifdef CONFIG_SYSCTL
1607 static int three = 3;
1610 proc_do_defense_mode(ctl_table *table, int write,
1611 void __user *buffer, size_t *lenp, loff_t *ppos)
1613 struct net *net = current->nsproxy->net_ns;
1614 int *valp = table->data;
1618 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1619 if (write && (*valp != val)) {
1620 if ((*valp < 0) || (*valp > 3)) {
1621 /* Restore the correct value */
1624 update_defense_level(net_ipvs(net));
1631 proc_do_sync_threshold(ctl_table *table, int write,
1632 void __user *buffer, size_t *lenp, loff_t *ppos)
1634 int *valp = table->data;
1638 /* backup the value first */
1639 memcpy(val, valp, sizeof(val));
1641 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1642 if (write && (valp[0] < 0 || valp[1] < 0 ||
1643 (valp[0] >= valp[1] && valp[1]))) {
1644 /* Restore the correct value */
1645 memcpy(valp, val, sizeof(val));
1651 proc_do_sync_mode(ctl_table *table, int write,
1652 void __user *buffer, size_t *lenp, loff_t *ppos)
1654 int *valp = table->data;
1658 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1659 if (write && (*valp != val)) {
1660 if ((*valp < 0) || (*valp > 1)) {
1661 /* Restore the correct value */
1669 proc_do_sync_ports(ctl_table *table, int write,
1670 void __user *buffer, size_t *lenp, loff_t *ppos)
1672 int *valp = table->data;
1676 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1677 if (write && (*valp != val)) {
1678 if (*valp < 1 || !is_power_of_2(*valp)) {
1679 /* Restore the correct value */
1687 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1688 * Do not change order or insert new entries without
1689 * align with netns init in ip_vs_control_net_init()
1692 static struct ctl_table vs_vars[] = {
1694 .procname = "amemthresh",
1695 .maxlen = sizeof(int),
1697 .proc_handler = proc_dointvec,
1700 .procname = "am_droprate",
1701 .maxlen = sizeof(int),
1703 .proc_handler = proc_dointvec,
1706 .procname = "drop_entry",
1707 .maxlen = sizeof(int),
1709 .proc_handler = proc_do_defense_mode,
1712 .procname = "drop_packet",
1713 .maxlen = sizeof(int),
1715 .proc_handler = proc_do_defense_mode,
1717 #ifdef CONFIG_IP_VS_NFCT
1719 .procname = "conntrack",
1720 .maxlen = sizeof(int),
1722 .proc_handler = &proc_dointvec,
1726 .procname = "secure_tcp",
1727 .maxlen = sizeof(int),
1729 .proc_handler = proc_do_defense_mode,
1732 .procname = "snat_reroute",
1733 .maxlen = sizeof(int),
1735 .proc_handler = &proc_dointvec,
1738 .procname = "sync_version",
1739 .maxlen = sizeof(int),
1741 .proc_handler = &proc_do_sync_mode,
1744 .procname = "sync_ports",
1745 .maxlen = sizeof(int),
1747 .proc_handler = &proc_do_sync_ports,
1750 .procname = "sync_qlen_max",
1751 .maxlen = sizeof(int),
1753 .proc_handler = proc_dointvec,
1756 .procname = "sync_sock_size",
1757 .maxlen = sizeof(int),
1759 .proc_handler = proc_dointvec,
1762 .procname = "cache_bypass",
1763 .maxlen = sizeof(int),
1765 .proc_handler = proc_dointvec,
1768 .procname = "expire_nodest_conn",
1769 .maxlen = sizeof(int),
1771 .proc_handler = proc_dointvec,
1774 .procname = "expire_quiescent_template",
1775 .maxlen = sizeof(int),
1777 .proc_handler = proc_dointvec,
1780 .procname = "sync_threshold",
1782 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1784 .proc_handler = proc_do_sync_threshold,
1787 .procname = "sync_refresh_period",
1788 .maxlen = sizeof(int),
1790 .proc_handler = proc_dointvec_jiffies,
1793 .procname = "sync_retries",
1794 .maxlen = sizeof(int),
1796 .proc_handler = proc_dointvec_minmax,
1801 .procname = "nat_icmp_send",
1802 .maxlen = sizeof(int),
1804 .proc_handler = proc_dointvec,
1806 #ifdef CONFIG_IP_VS_DEBUG
1808 .procname = "debug_level",
1809 .data = &sysctl_ip_vs_debug_level,
1810 .maxlen = sizeof(int),
1812 .proc_handler = proc_dointvec,
1817 .procname = "timeout_established",
1818 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1819 .maxlen = sizeof(int),
1821 .proc_handler = proc_dointvec_jiffies,
1824 .procname = "timeout_synsent",
1825 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1826 .maxlen = sizeof(int),
1828 .proc_handler = proc_dointvec_jiffies,
1831 .procname = "timeout_synrecv",
1832 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1833 .maxlen = sizeof(int),
1835 .proc_handler = proc_dointvec_jiffies,
1838 .procname = "timeout_finwait",
1839 .data = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1840 .maxlen = sizeof(int),
1842 .proc_handler = proc_dointvec_jiffies,
1845 .procname = "timeout_timewait",
1846 .data = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1847 .maxlen = sizeof(int),
1849 .proc_handler = proc_dointvec_jiffies,
1852 .procname = "timeout_close",
1853 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1854 .maxlen = sizeof(int),
1856 .proc_handler = proc_dointvec_jiffies,
1859 .procname = "timeout_closewait",
1860 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1861 .maxlen = sizeof(int),
1863 .proc_handler = proc_dointvec_jiffies,
1866 .procname = "timeout_lastack",
1867 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1868 .maxlen = sizeof(int),
1870 .proc_handler = proc_dointvec_jiffies,
1873 .procname = "timeout_listen",
1874 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1875 .maxlen = sizeof(int),
1877 .proc_handler = proc_dointvec_jiffies,
1880 .procname = "timeout_synack",
1881 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1882 .maxlen = sizeof(int),
1884 .proc_handler = proc_dointvec_jiffies,
1887 .procname = "timeout_udp",
1888 .data = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1889 .maxlen = sizeof(int),
1891 .proc_handler = proc_dointvec_jiffies,
1894 .procname = "timeout_icmp",
1895 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1896 .maxlen = sizeof(int),
1898 .proc_handler = proc_dointvec_jiffies,
1906 #ifdef CONFIG_PROC_FS
1909 struct seq_net_private p; /* Do not move this, netns depends upon it*/
1910 struct list_head *table;
1915 * Write the contents of the VS rule table to a PROCfs file.
1916 * (It is kept just for backward compatibility)
1918 static inline const char *ip_vs_fwd_name(unsigned int flags)
1920 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1921 case IP_VS_CONN_F_LOCALNODE:
1923 case IP_VS_CONN_F_TUNNEL:
1925 case IP_VS_CONN_F_DROUTE:
1933 /* Get the Nth entry in the two lists */
1934 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1936 struct net *net = seq_file_net(seq);
1937 struct ip_vs_iter *iter = seq->private;
1939 struct ip_vs_service *svc;
1941 /* look in hash by protocol */
1942 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1943 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1944 if (net_eq(svc->net, net) && pos-- == 0) {
1945 iter->table = ip_vs_svc_table;
1952 /* keep looking in fwmark */
1953 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1954 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1955 if (net_eq(svc->net, net) && pos-- == 0) {
1956 iter->table = ip_vs_svc_fwm_table;
1966 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1967 __acquires(__ip_vs_svc_lock)
1970 read_lock_bh(&__ip_vs_svc_lock);
1971 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1975 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1977 struct list_head *e;
1978 struct ip_vs_iter *iter;
1979 struct ip_vs_service *svc;
1982 if (v == SEQ_START_TOKEN)
1983 return ip_vs_info_array(seq,0);
1986 iter = seq->private;
1988 if (iter->table == ip_vs_svc_table) {
1989 /* next service in table hashed by protocol */
1990 if ((e = svc->s_list.next) != &ip_vs_svc_table[iter->bucket])
1991 return list_entry(e, struct ip_vs_service, s_list);
1994 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1995 list_for_each_entry(svc,&ip_vs_svc_table[iter->bucket],
2001 iter->table = ip_vs_svc_fwm_table;
2006 /* next service in hashed by fwmark */
2007 if ((e = svc->f_list.next) != &ip_vs_svc_fwm_table[iter->bucket])
2008 return list_entry(e, struct ip_vs_service, f_list);
2011 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2012 list_for_each_entry(svc, &ip_vs_svc_fwm_table[iter->bucket],
2020 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
2021 __releases(__ip_vs_svc_lock)
2023 read_unlock_bh(&__ip_vs_svc_lock);
2027 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
2029 if (v == SEQ_START_TOKEN) {
2031 "IP Virtual Server version %d.%d.%d (size=%d)\n",
2032 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2034 "Prot LocalAddress:Port Scheduler Flags\n");
2036 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2038 const struct ip_vs_service *svc = v;
2039 const struct ip_vs_iter *iter = seq->private;
2040 const struct ip_vs_dest *dest;
2042 if (iter->table == ip_vs_svc_table) {
2043 #ifdef CONFIG_IP_VS_IPV6
2044 if (svc->af == AF_INET6)
2045 seq_printf(seq, "%s [%pI6]:%04X %s ",
2046 ip_vs_proto_name(svc->protocol),
2049 svc->scheduler->name);
2052 seq_printf(seq, "%s %08X:%04X %s %s ",
2053 ip_vs_proto_name(svc->protocol),
2054 ntohl(svc->addr.ip),
2056 svc->scheduler->name,
2057 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2059 seq_printf(seq, "FWM %08X %s %s",
2060 svc->fwmark, svc->scheduler->name,
2061 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2064 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2065 seq_printf(seq, "persistent %d %08X\n",
2067 ntohl(svc->netmask));
2069 seq_putc(seq, '\n');
2071 list_for_each_entry(dest, &svc->destinations, n_list) {
2072 #ifdef CONFIG_IP_VS_IPV6
2073 if (dest->af == AF_INET6)
2076 " %-7s %-6d %-10d %-10d\n",
2079 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2080 atomic_read(&dest->weight),
2081 atomic_read(&dest->activeconns),
2082 atomic_read(&dest->inactconns));
2087 "%-7s %-6d %-10d %-10d\n",
2088 ntohl(dest->addr.ip),
2090 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2091 atomic_read(&dest->weight),
2092 atomic_read(&dest->activeconns),
2093 atomic_read(&dest->inactconns));
2100 static const struct seq_operations ip_vs_info_seq_ops = {
2101 .start = ip_vs_info_seq_start,
2102 .next = ip_vs_info_seq_next,
2103 .stop = ip_vs_info_seq_stop,
2104 .show = ip_vs_info_seq_show,
2107 static int ip_vs_info_open(struct inode *inode, struct file *file)
2109 return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2110 sizeof(struct ip_vs_iter));
2113 static const struct file_operations ip_vs_info_fops = {
2114 .owner = THIS_MODULE,
2115 .open = ip_vs_info_open,
2117 .llseek = seq_lseek,
2118 .release = seq_release_net,
2121 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2123 struct net *net = seq_file_single_net(seq);
2124 struct ip_vs_stats_user show;
2126 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2128 " Total Incoming Outgoing Incoming Outgoing\n");
2130 " Conns Packets Packets Bytes Bytes\n");
2132 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2133 seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", show.conns,
2134 show.inpkts, show.outpkts,
2135 (unsigned long long) show.inbytes,
2136 (unsigned long long) show.outbytes);
2138 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2140 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2141 seq_printf(seq, "%8X %8X %8X %16X %16X\n",
2142 show.cps, show.inpps, show.outpps,
2143 show.inbps, show.outbps);
2148 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2150 return single_open_net(inode, file, ip_vs_stats_show);
2153 static const struct file_operations ip_vs_stats_fops = {
2154 .owner = THIS_MODULE,
2155 .open = ip_vs_stats_seq_open,
2157 .llseek = seq_lseek,
2158 .release = single_release_net,
2161 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2163 struct net *net = seq_file_single_net(seq);
2164 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2165 struct ip_vs_cpu_stats *cpustats = tot_stats->cpustats;
2166 struct ip_vs_stats_user rates;
2169 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2171 " Total Incoming Outgoing Incoming Outgoing\n");
2173 "CPU Conns Packets Packets Bytes Bytes\n");
2175 for_each_possible_cpu(i) {
2176 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2178 __u64 inbytes, outbytes;
2181 start = u64_stats_fetch_begin_bh(&u->syncp);
2182 inbytes = u->ustats.inbytes;
2183 outbytes = u->ustats.outbytes;
2184 } while (u64_stats_fetch_retry_bh(&u->syncp, start));
2186 seq_printf(seq, "%3X %8X %8X %8X %16LX %16LX\n",
2187 i, u->ustats.conns, u->ustats.inpkts,
2188 u->ustats.outpkts, (__u64)inbytes,
2192 spin_lock_bh(&tot_stats->lock);
2194 seq_printf(seq, " ~ %8X %8X %8X %16LX %16LX\n\n",
2195 tot_stats->ustats.conns, tot_stats->ustats.inpkts,
2196 tot_stats->ustats.outpkts,
2197 (unsigned long long) tot_stats->ustats.inbytes,
2198 (unsigned long long) tot_stats->ustats.outbytes);
2200 ip_vs_read_estimator(&rates, tot_stats);
2202 spin_unlock_bh(&tot_stats->lock);
2204 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2206 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2207 seq_printf(seq, " %8X %8X %8X %16X %16X\n",
2217 static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2219 return single_open_net(inode, file, ip_vs_stats_percpu_show);
2222 static const struct file_operations ip_vs_stats_percpu_fops = {
2223 .owner = THIS_MODULE,
2224 .open = ip_vs_stats_percpu_seq_open,
2226 .llseek = seq_lseek,
2227 .release = single_release_net,
2232 * Set timeout values for tcp tcpfin udp in the timeout_table.
2234 static int ip_vs_set_timeout(struct net *net, struct ip_vs_timeout_user *u)
2236 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2237 struct ip_vs_proto_data *pd;
2240 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2245 #ifdef CONFIG_IP_VS_PROTO_TCP
2246 if (u->tcp_timeout) {
2247 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2248 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2249 = u->tcp_timeout * HZ;
2252 if (u->tcp_fin_timeout) {
2253 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2254 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2255 = u->tcp_fin_timeout * HZ;
2259 #ifdef CONFIG_IP_VS_PROTO_UDP
2260 if (u->udp_timeout) {
2261 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2262 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2263 = u->udp_timeout * HZ;
2270 #define SET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2271 #define SERVICE_ARG_LEN (sizeof(struct ip_vs_service_user))
2272 #define SVCDEST_ARG_LEN (sizeof(struct ip_vs_service_user) + \
2273 sizeof(struct ip_vs_dest_user))
2274 #define TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2275 #define DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user))
2276 #define MAX_ARG_LEN SVCDEST_ARG_LEN
2278 static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
2279 [SET_CMDID(IP_VS_SO_SET_ADD)] = SERVICE_ARG_LEN,
2280 [SET_CMDID(IP_VS_SO_SET_EDIT)] = SERVICE_ARG_LEN,
2281 [SET_CMDID(IP_VS_SO_SET_DEL)] = SERVICE_ARG_LEN,
2282 [SET_CMDID(IP_VS_SO_SET_FLUSH)] = 0,
2283 [SET_CMDID(IP_VS_SO_SET_ADDDEST)] = SVCDEST_ARG_LEN,
2284 [SET_CMDID(IP_VS_SO_SET_DELDEST)] = SVCDEST_ARG_LEN,
2285 [SET_CMDID(IP_VS_SO_SET_EDITDEST)] = SVCDEST_ARG_LEN,
2286 [SET_CMDID(IP_VS_SO_SET_TIMEOUT)] = TIMEOUT_ARG_LEN,
2287 [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)] = DAEMON_ARG_LEN,
2288 [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)] = DAEMON_ARG_LEN,
2289 [SET_CMDID(IP_VS_SO_SET_ZERO)] = SERVICE_ARG_LEN,
2292 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2293 struct ip_vs_service_user *usvc_compat)
2295 memset(usvc, 0, sizeof(*usvc));
2298 usvc->protocol = usvc_compat->protocol;
2299 usvc->addr.ip = usvc_compat->addr;
2300 usvc->port = usvc_compat->port;
2301 usvc->fwmark = usvc_compat->fwmark;
2303 /* Deep copy of sched_name is not needed here */
2304 usvc->sched_name = usvc_compat->sched_name;
2306 usvc->flags = usvc_compat->flags;
2307 usvc->timeout = usvc_compat->timeout;
2308 usvc->netmask = usvc_compat->netmask;
2311 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2312 struct ip_vs_dest_user *udest_compat)
2314 memset(udest, 0, sizeof(*udest));
2316 udest->addr.ip = udest_compat->addr;
2317 udest->port = udest_compat->port;
2318 udest->conn_flags = udest_compat->conn_flags;
2319 udest->weight = udest_compat->weight;
2320 udest->u_threshold = udest_compat->u_threshold;
2321 udest->l_threshold = udest_compat->l_threshold;
2325 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2327 struct net *net = sock_net(sk);
2329 unsigned char arg[MAX_ARG_LEN];
2330 struct ip_vs_service_user *usvc_compat;
2331 struct ip_vs_service_user_kern usvc;
2332 struct ip_vs_service *svc;
2333 struct ip_vs_dest_user *udest_compat;
2334 struct ip_vs_dest_user_kern udest;
2335 struct netns_ipvs *ipvs = net_ipvs(net);
2337 if (!capable(CAP_NET_ADMIN))
2340 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2342 if (len < 0 || len > MAX_ARG_LEN)
2344 if (len != set_arglen[SET_CMDID(cmd)]) {
2345 pr_err("set_ctl: len %u != %u\n",
2346 len, set_arglen[SET_CMDID(cmd)]);
2350 if (copy_from_user(arg, user, len) != 0)
2353 /* increase the module use count */
2354 ip_vs_use_count_inc();
2356 /* Handle daemons since they have another lock */
2357 if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2358 cmd == IP_VS_SO_SET_STOPDAEMON) {
2359 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2361 if (mutex_lock_interruptible(&ipvs->sync_mutex)) {
2365 if (cmd == IP_VS_SO_SET_STARTDAEMON)
2366 ret = start_sync_thread(net, dm->state, dm->mcast_ifn,
2369 ret = stop_sync_thread(net, dm->state);
2370 mutex_unlock(&ipvs->sync_mutex);
2374 if (mutex_lock_interruptible(&__ip_vs_mutex)) {
2379 if (cmd == IP_VS_SO_SET_FLUSH) {
2380 /* Flush the virtual service */
2381 ret = ip_vs_flush(net);
2383 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2384 /* Set timeout values for (tcp tcpfin udp) */
2385 ret = ip_vs_set_timeout(net, (struct ip_vs_timeout_user *)arg);
2389 usvc_compat = (struct ip_vs_service_user *)arg;
2390 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2392 /* We only use the new structs internally, so copy userspace compat
2393 * structs to extended internal versions */
2394 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2395 ip_vs_copy_udest_compat(&udest, udest_compat);
2397 if (cmd == IP_VS_SO_SET_ZERO) {
2398 /* if no service address is set, zero counters in all */
2399 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2400 ret = ip_vs_zero_all(net);
2405 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2406 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2407 usvc.protocol != IPPROTO_SCTP) {
2408 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2409 usvc.protocol, &usvc.addr.ip,
2410 ntohs(usvc.port), usvc.sched_name);
2415 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2416 if (usvc.fwmark == 0)
2417 svc = __ip_vs_service_find(net, usvc.af, usvc.protocol,
2418 &usvc.addr, usvc.port);
2420 svc = __ip_vs_svc_fwm_find(net, usvc.af, usvc.fwmark);
2422 if (cmd != IP_VS_SO_SET_ADD
2423 && (svc == NULL || svc->protocol != usvc.protocol)) {
2429 case IP_VS_SO_SET_ADD:
2433 ret = ip_vs_add_service(net, &usvc, &svc);
2435 case IP_VS_SO_SET_EDIT:
2436 ret = ip_vs_edit_service(svc, &usvc);
2438 case IP_VS_SO_SET_DEL:
2439 ret = ip_vs_del_service(svc);
2443 case IP_VS_SO_SET_ZERO:
2444 ret = ip_vs_zero_service(svc);
2446 case IP_VS_SO_SET_ADDDEST:
2447 ret = ip_vs_add_dest(svc, &udest);
2449 case IP_VS_SO_SET_EDITDEST:
2450 ret = ip_vs_edit_dest(svc, &udest);
2452 case IP_VS_SO_SET_DELDEST:
2453 ret = ip_vs_del_dest(svc, &udest);
2460 mutex_unlock(&__ip_vs_mutex);
2462 /* decrease the module use count */
2463 ip_vs_use_count_dec();
2470 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2472 dst->protocol = src->protocol;
2473 dst->addr = src->addr.ip;
2474 dst->port = src->port;
2475 dst->fwmark = src->fwmark;
2476 strlcpy(dst->sched_name, src->scheduler->name, sizeof(dst->sched_name));
2477 dst->flags = src->flags;
2478 dst->timeout = src->timeout / HZ;
2479 dst->netmask = src->netmask;
2480 dst->num_dests = src->num_dests;
2481 ip_vs_copy_stats(&dst->stats, &src->stats);
2485 __ip_vs_get_service_entries(struct net *net,
2486 const struct ip_vs_get_services *get,
2487 struct ip_vs_get_services __user *uptr)
2490 struct ip_vs_service *svc;
2491 struct ip_vs_service_entry entry;
2494 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2495 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2496 /* Only expose IPv4 entries to old interface */
2497 if (svc->af != AF_INET || !net_eq(svc->net, net))
2500 if (count >= get->num_services)
2502 memset(&entry, 0, sizeof(entry));
2503 ip_vs_copy_service(&entry, svc);
2504 if (copy_to_user(&uptr->entrytable[count],
2505 &entry, sizeof(entry))) {
2513 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2514 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2515 /* Only expose IPv4 entries to old interface */
2516 if (svc->af != AF_INET || !net_eq(svc->net, net))
2519 if (count >= get->num_services)
2521 memset(&entry, 0, sizeof(entry));
2522 ip_vs_copy_service(&entry, svc);
2523 if (copy_to_user(&uptr->entrytable[count],
2524 &entry, sizeof(entry))) {
2536 __ip_vs_get_dest_entries(struct net *net, const struct ip_vs_get_dests *get,
2537 struct ip_vs_get_dests __user *uptr)
2539 struct ip_vs_service *svc;
2540 union nf_inet_addr addr = { .ip = get->addr };
2544 svc = __ip_vs_svc_fwm_find(net, AF_INET, get->fwmark);
2546 svc = __ip_vs_service_find(net, AF_INET, get->protocol, &addr,
2551 struct ip_vs_dest *dest;
2552 struct ip_vs_dest_entry entry;
2554 list_for_each_entry(dest, &svc->destinations, n_list) {
2555 if (count >= get->num_dests)
2558 entry.addr = dest->addr.ip;
2559 entry.port = dest->port;
2560 entry.conn_flags = atomic_read(&dest->conn_flags);
2561 entry.weight = atomic_read(&dest->weight);
2562 entry.u_threshold = dest->u_threshold;
2563 entry.l_threshold = dest->l_threshold;
2564 entry.activeconns = atomic_read(&dest->activeconns);
2565 entry.inactconns = atomic_read(&dest->inactconns);
2566 entry.persistconns = atomic_read(&dest->persistconns);
2567 ip_vs_copy_stats(&entry.stats, &dest->stats);
2568 if (copy_to_user(&uptr->entrytable[count],
2569 &entry, sizeof(entry))) {
2581 __ip_vs_get_timeouts(struct net *net, struct ip_vs_timeout_user *u)
2583 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2584 struct ip_vs_proto_data *pd;
2587 #ifdef CONFIG_IP_VS_PROTO_TCP
2588 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2589 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2590 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2592 #ifdef CONFIG_IP_VS_PROTO_UDP
2593 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2595 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2600 #define GET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2601 #define GET_INFO_ARG_LEN (sizeof(struct ip_vs_getinfo))
2602 #define GET_SERVICES_ARG_LEN (sizeof(struct ip_vs_get_services))
2603 #define GET_SERVICE_ARG_LEN (sizeof(struct ip_vs_service_entry))
2604 #define GET_DESTS_ARG_LEN (sizeof(struct ip_vs_get_dests))
2605 #define GET_TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2606 #define GET_DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user) * 2)
2608 static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
2609 [GET_CMDID(IP_VS_SO_GET_VERSION)] = 64,
2610 [GET_CMDID(IP_VS_SO_GET_INFO)] = GET_INFO_ARG_LEN,
2611 [GET_CMDID(IP_VS_SO_GET_SERVICES)] = GET_SERVICES_ARG_LEN,
2612 [GET_CMDID(IP_VS_SO_GET_SERVICE)] = GET_SERVICE_ARG_LEN,
2613 [GET_CMDID(IP_VS_SO_GET_DESTS)] = GET_DESTS_ARG_LEN,
2614 [GET_CMDID(IP_VS_SO_GET_TIMEOUT)] = GET_TIMEOUT_ARG_LEN,
2615 [GET_CMDID(IP_VS_SO_GET_DAEMON)] = GET_DAEMON_ARG_LEN,
2619 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2621 unsigned char arg[128];
2623 unsigned int copylen;
2624 struct net *net = sock_net(sk);
2625 struct netns_ipvs *ipvs = net_ipvs(net);
2628 if (!capable(CAP_NET_ADMIN))
2631 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2634 if (*len < get_arglen[GET_CMDID(cmd)]) {
2635 pr_err("get_ctl: len %u < %u\n",
2636 *len, get_arglen[GET_CMDID(cmd)]);
2640 copylen = get_arglen[GET_CMDID(cmd)];
2644 if (copy_from_user(arg, user, copylen) != 0)
2647 * Handle daemons first since it has its own locking
2649 if (cmd == IP_VS_SO_GET_DAEMON) {
2650 struct ip_vs_daemon_user d[2];
2652 memset(&d, 0, sizeof(d));
2653 if (mutex_lock_interruptible(&ipvs->sync_mutex))
2654 return -ERESTARTSYS;
2656 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2657 d[0].state = IP_VS_STATE_MASTER;
2658 strlcpy(d[0].mcast_ifn, ipvs->master_mcast_ifn,
2659 sizeof(d[0].mcast_ifn));
2660 d[0].syncid = ipvs->master_syncid;
2662 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2663 d[1].state = IP_VS_STATE_BACKUP;
2664 strlcpy(d[1].mcast_ifn, ipvs->backup_mcast_ifn,
2665 sizeof(d[1].mcast_ifn));
2666 d[1].syncid = ipvs->backup_syncid;
2668 if (copy_to_user(user, &d, sizeof(d)) != 0)
2670 mutex_unlock(&ipvs->sync_mutex);
2674 if (mutex_lock_interruptible(&__ip_vs_mutex))
2675 return -ERESTARTSYS;
2678 case IP_VS_SO_GET_VERSION:
2682 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2683 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2684 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2688 *len = strlen(buf)+1;
2692 case IP_VS_SO_GET_INFO:
2694 struct ip_vs_getinfo info;
2695 info.version = IP_VS_VERSION_CODE;
2696 info.size = ip_vs_conn_tab_size;
2697 info.num_services = ipvs->num_services;
2698 if (copy_to_user(user, &info, sizeof(info)) != 0)
2703 case IP_VS_SO_GET_SERVICES:
2705 struct ip_vs_get_services *get;
2708 get = (struct ip_vs_get_services *)arg;
2709 size = sizeof(*get) +
2710 sizeof(struct ip_vs_service_entry) * get->num_services;
2712 pr_err("length: %u != %u\n", *len, size);
2716 ret = __ip_vs_get_service_entries(net, get, user);
2720 case IP_VS_SO_GET_SERVICE:
2722 struct ip_vs_service_entry *entry;
2723 struct ip_vs_service *svc;
2724 union nf_inet_addr addr;
2726 entry = (struct ip_vs_service_entry *)arg;
2727 addr.ip = entry->addr;
2729 svc = __ip_vs_svc_fwm_find(net, AF_INET, entry->fwmark);
2731 svc = __ip_vs_service_find(net, AF_INET,
2732 entry->protocol, &addr,
2735 ip_vs_copy_service(entry, svc);
2736 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2743 case IP_VS_SO_GET_DESTS:
2745 struct ip_vs_get_dests *get;
2748 get = (struct ip_vs_get_dests *)arg;
2749 size = sizeof(*get) +
2750 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2752 pr_err("length: %u != %u\n", *len, size);
2756 ret = __ip_vs_get_dest_entries(net, get, user);
2760 case IP_VS_SO_GET_TIMEOUT:
2762 struct ip_vs_timeout_user t;
2764 memset(&t, 0, sizeof(t));
2765 __ip_vs_get_timeouts(net, &t);
2766 if (copy_to_user(user, &t, sizeof(t)) != 0)
2776 mutex_unlock(&__ip_vs_mutex);
2781 static struct nf_sockopt_ops ip_vs_sockopts = {
2783 .set_optmin = IP_VS_BASE_CTL,
2784 .set_optmax = IP_VS_SO_SET_MAX+1,
2785 .set = do_ip_vs_set_ctl,
2786 .get_optmin = IP_VS_BASE_CTL,
2787 .get_optmax = IP_VS_SO_GET_MAX+1,
2788 .get = do_ip_vs_get_ctl,
2789 .owner = THIS_MODULE,
2793 * Generic Netlink interface
2796 /* IPVS genetlink family */
2797 static struct genl_family ip_vs_genl_family = {
2798 .id = GENL_ID_GENERATE,
2800 .name = IPVS_GENL_NAME,
2801 .version = IPVS_GENL_VERSION,
2802 .maxattr = IPVS_CMD_MAX,
2803 .netnsok = true, /* Make ipvsadm to work on netns */
2806 /* Policy used for first-level command attributes */
2807 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2808 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2809 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2810 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2811 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2812 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2813 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2816 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2817 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2818 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2819 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2820 .len = IP_VS_IFNAME_MAXLEN },
2821 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2824 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2825 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2826 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2827 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2828 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2829 .len = sizeof(union nf_inet_addr) },
2830 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2831 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2832 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2833 .len = IP_VS_SCHEDNAME_MAXLEN },
2834 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2835 .len = IP_VS_PENAME_MAXLEN },
2836 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2837 .len = sizeof(struct ip_vs_flags) },
2838 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2839 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2840 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2843 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2844 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2845 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2846 .len = sizeof(union nf_inet_addr) },
2847 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2848 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2849 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2850 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2851 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2852 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2853 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2854 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2855 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2858 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2859 struct ip_vs_stats *stats)
2861 struct ip_vs_stats_user ustats;
2862 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2866 ip_vs_copy_stats(&ustats, stats);
2868 if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, ustats.conns) ||
2869 nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, ustats.inpkts) ||
2870 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, ustats.outpkts) ||
2871 nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, ustats.inbytes) ||
2872 nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, ustats.outbytes) ||
2873 nla_put_u32(skb, IPVS_STATS_ATTR_CPS, ustats.cps) ||
2874 nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, ustats.inpps) ||
2875 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, ustats.outpps) ||
2876 nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, ustats.inbps) ||
2877 nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, ustats.outbps))
2878 goto nla_put_failure;
2879 nla_nest_end(skb, nl_stats);
2884 nla_nest_cancel(skb, nl_stats);
2888 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2889 struct ip_vs_service *svc)
2891 struct nlattr *nl_service;
2892 struct ip_vs_flags flags = { .flags = svc->flags,
2895 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2899 if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2900 goto nla_put_failure;
2902 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2903 goto nla_put_failure;
2905 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2906 nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2907 nla_put_u16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2908 goto nla_put_failure;
2911 if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, svc->scheduler->name) ||
2913 nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, svc->pe->name)) ||
2914 nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2915 nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2916 nla_put_u32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
2917 goto nla_put_failure;
2918 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2919 goto nla_put_failure;
2921 nla_nest_end(skb, nl_service);
2926 nla_nest_cancel(skb, nl_service);
2930 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2931 struct ip_vs_service *svc,
2932 struct netlink_callback *cb)
2936 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2937 &ip_vs_genl_family, NLM_F_MULTI,
2938 IPVS_CMD_NEW_SERVICE);
2942 if (ip_vs_genl_fill_service(skb, svc) < 0)
2943 goto nla_put_failure;
2945 return genlmsg_end(skb, hdr);
2948 genlmsg_cancel(skb, hdr);
2952 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2953 struct netlink_callback *cb)
2956 int start = cb->args[0];
2957 struct ip_vs_service *svc;
2958 struct net *net = skb_sknet(skb);
2960 mutex_lock(&__ip_vs_mutex);
2961 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2962 list_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2963 if (++idx <= start || !net_eq(svc->net, net))
2965 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2967 goto nla_put_failure;
2972 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2973 list_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
2974 if (++idx <= start || !net_eq(svc->net, net))
2976 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2978 goto nla_put_failure;
2984 mutex_unlock(&__ip_vs_mutex);
2990 static int ip_vs_genl_parse_service(struct net *net,
2991 struct ip_vs_service_user_kern *usvc,
2992 struct nlattr *nla, int full_entry,
2993 struct ip_vs_service **ret_svc)
2995 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
2996 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
2997 struct ip_vs_service *svc;
2999 /* Parse mandatory identifying service fields first */
3001 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
3004 nla_af = attrs[IPVS_SVC_ATTR_AF];
3005 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
3006 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
3007 nla_port = attrs[IPVS_SVC_ATTR_PORT];
3008 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
3010 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3013 memset(usvc, 0, sizeof(*usvc));
3015 usvc->af = nla_get_u16(nla_af);
3016 #ifdef CONFIG_IP_VS_IPV6
3017 if (usvc->af != AF_INET && usvc->af != AF_INET6)
3019 if (usvc->af != AF_INET)
3021 return -EAFNOSUPPORT;
3024 usvc->protocol = IPPROTO_TCP;
3025 usvc->fwmark = nla_get_u32(nla_fwmark);
3027 usvc->protocol = nla_get_u16(nla_protocol);
3028 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3029 usvc->port = nla_get_u16(nla_port);
3034 svc = __ip_vs_svc_fwm_find(net, usvc->af, usvc->fwmark);
3036 svc = __ip_vs_service_find(net, usvc->af, usvc->protocol,
3037 &usvc->addr, usvc->port);
3040 /* If a full entry was requested, check for the additional fields */
3042 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3044 struct ip_vs_flags flags;
3046 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3047 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3048 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3049 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3050 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3052 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3055 nla_memcpy(&flags, nla_flags, sizeof(flags));
3057 /* prefill flags from service if it already exists */
3059 usvc->flags = svc->flags;
3061 /* set new flags from userland */
3062 usvc->flags = (usvc->flags & ~flags.mask) |
3063 (flags.flags & flags.mask);
3064 usvc->sched_name = nla_data(nla_sched);
3065 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3066 usvc->timeout = nla_get_u32(nla_timeout);
3067 usvc->netmask = nla_get_u32(nla_netmask);
3073 static struct ip_vs_service *ip_vs_genl_find_service(struct net *net,
3076 struct ip_vs_service_user_kern usvc;
3077 struct ip_vs_service *svc;
3080 ret = ip_vs_genl_parse_service(net, &usvc, nla, 0, &svc);
3081 return ret ? ERR_PTR(ret) : svc;
3084 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3086 struct nlattr *nl_dest;
3088 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3092 if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3093 nla_put_u16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3094 nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3095 (atomic_read(&dest->conn_flags) &
3096 IP_VS_CONN_F_FWD_MASK)) ||
3097 nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3098 atomic_read(&dest->weight)) ||
3099 nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3100 nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3101 nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3102 atomic_read(&dest->activeconns)) ||
3103 nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3104 atomic_read(&dest->inactconns)) ||
3105 nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3106 atomic_read(&dest->persistconns)))
3107 goto nla_put_failure;
3108 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
3109 goto nla_put_failure;
3111 nla_nest_end(skb, nl_dest);
3116 nla_nest_cancel(skb, nl_dest);
3120 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3121 struct netlink_callback *cb)
3125 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
3126 &ip_vs_genl_family, NLM_F_MULTI,
3131 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3132 goto nla_put_failure;
3134 return genlmsg_end(skb, hdr);
3137 genlmsg_cancel(skb, hdr);
3141 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3142 struct netlink_callback *cb)
3145 int start = cb->args[0];
3146 struct ip_vs_service *svc;
3147 struct ip_vs_dest *dest;
3148 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3149 struct net *net = skb_sknet(skb);
3151 mutex_lock(&__ip_vs_mutex);
3153 /* Try to find the service for which to dump destinations */
3154 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3155 IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3159 svc = ip_vs_genl_find_service(net, attrs[IPVS_CMD_ATTR_SERVICE]);
3160 if (IS_ERR(svc) || svc == NULL)
3163 /* Dump the destinations */
3164 list_for_each_entry(dest, &svc->destinations, n_list) {
3167 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3169 goto nla_put_failure;
3177 mutex_unlock(&__ip_vs_mutex);
3182 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3183 struct nlattr *nla, int full_entry)
3185 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3186 struct nlattr *nla_addr, *nla_port;
3188 /* Parse mandatory identifying destination fields first */
3190 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3193 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3194 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3196 if (!(nla_addr && nla_port))
3199 memset(udest, 0, sizeof(*udest));
3201 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3202 udest->port = nla_get_u16(nla_port);
3204 /* If a full entry was requested, check for the additional fields */
3206 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3209 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3210 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3211 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3212 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3214 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3217 udest->conn_flags = nla_get_u32(nla_fwd)
3218 & IP_VS_CONN_F_FWD_MASK;
3219 udest->weight = nla_get_u32(nla_weight);
3220 udest->u_threshold = nla_get_u32(nla_u_thresh);
3221 udest->l_threshold = nla_get_u32(nla_l_thresh);
3227 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __be32 state,
3228 const char *mcast_ifn, __be32 syncid)
3230 struct nlattr *nl_daemon;
3232 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3236 if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3237 nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn) ||
3238 nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid))
3239 goto nla_put_failure;
3240 nla_nest_end(skb, nl_daemon);
3245 nla_nest_cancel(skb, nl_daemon);
3249 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __be32 state,
3250 const char *mcast_ifn, __be32 syncid,
3251 struct netlink_callback *cb)
3254 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
3255 &ip_vs_genl_family, NLM_F_MULTI,
3256 IPVS_CMD_NEW_DAEMON);
3260 if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
3261 goto nla_put_failure;
3263 return genlmsg_end(skb, hdr);
3266 genlmsg_cancel(skb, hdr);
3270 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3271 struct netlink_callback *cb)
3273 struct net *net = skb_sknet(skb);
3274 struct netns_ipvs *ipvs = net_ipvs(net);
3276 mutex_lock(&ipvs->sync_mutex);
3277 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3278 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3279 ipvs->master_mcast_ifn,
3280 ipvs->master_syncid, cb) < 0)
3281 goto nla_put_failure;
3286 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3287 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3288 ipvs->backup_mcast_ifn,
3289 ipvs->backup_syncid, cb) < 0)
3290 goto nla_put_failure;
3296 mutex_unlock(&ipvs->sync_mutex);
3301 static int ip_vs_genl_new_daemon(struct net *net, struct nlattr **attrs)
3303 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3304 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3305 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3308 return start_sync_thread(net,
3309 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3310 nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3311 nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3314 static int ip_vs_genl_del_daemon(struct net *net, struct nlattr **attrs)
3316 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3319 return stop_sync_thread(net,
3320 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3323 static int ip_vs_genl_set_config(struct net *net, struct nlattr **attrs)
3325 struct ip_vs_timeout_user t;
3327 __ip_vs_get_timeouts(net, &t);
3329 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3330 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3332 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3334 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3336 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3337 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3339 return ip_vs_set_timeout(net, &t);
3342 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3346 struct netns_ipvs *ipvs;
3348 net = skb_sknet(skb);
3349 ipvs = net_ipvs(net);
3350 cmd = info->genlhdr->cmd;
3352 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3353 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3355 mutex_lock(&ipvs->sync_mutex);
3356 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3357 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3358 info->attrs[IPVS_CMD_ATTR_DAEMON],
3359 ip_vs_daemon_policy)) {
3364 if (cmd == IPVS_CMD_NEW_DAEMON)
3365 ret = ip_vs_genl_new_daemon(net, daemon_attrs);
3367 ret = ip_vs_genl_del_daemon(net, daemon_attrs);
3369 mutex_unlock(&ipvs->sync_mutex);
3374 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3376 struct ip_vs_service *svc = NULL;
3377 struct ip_vs_service_user_kern usvc;
3378 struct ip_vs_dest_user_kern udest;
3380 int need_full_svc = 0, need_full_dest = 0;
3383 net = skb_sknet(skb);
3384 cmd = info->genlhdr->cmd;
3386 mutex_lock(&__ip_vs_mutex);
3388 if (cmd == IPVS_CMD_FLUSH) {
3389 ret = ip_vs_flush(net);
3391 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3392 ret = ip_vs_genl_set_config(net, info->attrs);
3394 } else if (cmd == IPVS_CMD_ZERO &&
3395 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3396 ret = ip_vs_zero_all(net);
3400 /* All following commands require a service argument, so check if we
3401 * received a valid one. We need a full service specification when
3402 * adding / editing a service. Only identifying members otherwise. */
3403 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3406 ret = ip_vs_genl_parse_service(net, &usvc,
3407 info->attrs[IPVS_CMD_ATTR_SERVICE],
3408 need_full_svc, &svc);
3412 /* Unless we're adding a new service, the service must already exist */
3413 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3418 /* Destination commands require a valid destination argument. For
3419 * adding / editing a destination, we need a full destination
3421 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3422 cmd == IPVS_CMD_DEL_DEST) {
3423 if (cmd != IPVS_CMD_DEL_DEST)
3426 ret = ip_vs_genl_parse_dest(&udest,
3427 info->attrs[IPVS_CMD_ATTR_DEST],
3434 case IPVS_CMD_NEW_SERVICE:
3436 ret = ip_vs_add_service(net, &usvc, &svc);
3440 case IPVS_CMD_SET_SERVICE:
3441 ret = ip_vs_edit_service(svc, &usvc);
3443 case IPVS_CMD_DEL_SERVICE:
3444 ret = ip_vs_del_service(svc);
3445 /* do not use svc, it can be freed */
3447 case IPVS_CMD_NEW_DEST:
3448 ret = ip_vs_add_dest(svc, &udest);
3450 case IPVS_CMD_SET_DEST:
3451 ret = ip_vs_edit_dest(svc, &udest);
3453 case IPVS_CMD_DEL_DEST:
3454 ret = ip_vs_del_dest(svc, &udest);
3457 ret = ip_vs_zero_service(svc);
3464 mutex_unlock(&__ip_vs_mutex);
3469 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3471 struct sk_buff *msg;
3473 int ret, cmd, reply_cmd;
3476 net = skb_sknet(skb);
3477 cmd = info->genlhdr->cmd;
3479 if (cmd == IPVS_CMD_GET_SERVICE)
3480 reply_cmd = IPVS_CMD_NEW_SERVICE;
3481 else if (cmd == IPVS_CMD_GET_INFO)
3482 reply_cmd = IPVS_CMD_SET_INFO;
3483 else if (cmd == IPVS_CMD_GET_CONFIG)
3484 reply_cmd = IPVS_CMD_SET_CONFIG;
3486 pr_err("unknown Generic Netlink command\n");
3490 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3494 mutex_lock(&__ip_vs_mutex);
3496 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3498 goto nla_put_failure;
3501 case IPVS_CMD_GET_SERVICE:
3503 struct ip_vs_service *svc;
3505 svc = ip_vs_genl_find_service(net,
3506 info->attrs[IPVS_CMD_ATTR_SERVICE]);
3511 ret = ip_vs_genl_fill_service(msg, svc);
3513 goto nla_put_failure;
3522 case IPVS_CMD_GET_CONFIG:
3524 struct ip_vs_timeout_user t;
3526 __ip_vs_get_timeouts(net, &t);
3527 #ifdef CONFIG_IP_VS_PROTO_TCP
3528 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3530 nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3532 goto nla_put_failure;
3534 #ifdef CONFIG_IP_VS_PROTO_UDP
3535 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3536 goto nla_put_failure;
3542 case IPVS_CMD_GET_INFO:
3543 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3544 IP_VS_VERSION_CODE) ||
3545 nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3546 ip_vs_conn_tab_size))
3547 goto nla_put_failure;
3551 genlmsg_end(msg, reply);
3552 ret = genlmsg_reply(msg, info);
3556 pr_err("not enough space in Netlink message\n");
3562 mutex_unlock(&__ip_vs_mutex);
3568 static struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3570 .cmd = IPVS_CMD_NEW_SERVICE,
3571 .flags = GENL_ADMIN_PERM,
3572 .policy = ip_vs_cmd_policy,
3573 .doit = ip_vs_genl_set_cmd,
3576 .cmd = IPVS_CMD_SET_SERVICE,
3577 .flags = GENL_ADMIN_PERM,
3578 .policy = ip_vs_cmd_policy,
3579 .doit = ip_vs_genl_set_cmd,
3582 .cmd = IPVS_CMD_DEL_SERVICE,
3583 .flags = GENL_ADMIN_PERM,
3584 .policy = ip_vs_cmd_policy,
3585 .doit = ip_vs_genl_set_cmd,
3588 .cmd = IPVS_CMD_GET_SERVICE,
3589 .flags = GENL_ADMIN_PERM,
3590 .doit = ip_vs_genl_get_cmd,
3591 .dumpit = ip_vs_genl_dump_services,
3592 .policy = ip_vs_cmd_policy,
3595 .cmd = IPVS_CMD_NEW_DEST,
3596 .flags = GENL_ADMIN_PERM,
3597 .policy = ip_vs_cmd_policy,
3598 .doit = ip_vs_genl_set_cmd,
3601 .cmd = IPVS_CMD_SET_DEST,
3602 .flags = GENL_ADMIN_PERM,
3603 .policy = ip_vs_cmd_policy,
3604 .doit = ip_vs_genl_set_cmd,
3607 .cmd = IPVS_CMD_DEL_DEST,
3608 .flags = GENL_ADMIN_PERM,
3609 .policy = ip_vs_cmd_policy,
3610 .doit = ip_vs_genl_set_cmd,
3613 .cmd = IPVS_CMD_GET_DEST,
3614 .flags = GENL_ADMIN_PERM,
3615 .policy = ip_vs_cmd_policy,
3616 .dumpit = ip_vs_genl_dump_dests,
3619 .cmd = IPVS_CMD_NEW_DAEMON,
3620 .flags = GENL_ADMIN_PERM,
3621 .policy = ip_vs_cmd_policy,
3622 .doit = ip_vs_genl_set_daemon,
3625 .cmd = IPVS_CMD_DEL_DAEMON,
3626 .flags = GENL_ADMIN_PERM,
3627 .policy = ip_vs_cmd_policy,
3628 .doit = ip_vs_genl_set_daemon,
3631 .cmd = IPVS_CMD_GET_DAEMON,
3632 .flags = GENL_ADMIN_PERM,
3633 .dumpit = ip_vs_genl_dump_daemons,
3636 .cmd = IPVS_CMD_SET_CONFIG,
3637 .flags = GENL_ADMIN_PERM,
3638 .policy = ip_vs_cmd_policy,
3639 .doit = ip_vs_genl_set_cmd,
3642 .cmd = IPVS_CMD_GET_CONFIG,
3643 .flags = GENL_ADMIN_PERM,
3644 .doit = ip_vs_genl_get_cmd,
3647 .cmd = IPVS_CMD_GET_INFO,
3648 .flags = GENL_ADMIN_PERM,
3649 .doit = ip_vs_genl_get_cmd,
3652 .cmd = IPVS_CMD_ZERO,
3653 .flags = GENL_ADMIN_PERM,
3654 .policy = ip_vs_cmd_policy,
3655 .doit = ip_vs_genl_set_cmd,
3658 .cmd = IPVS_CMD_FLUSH,
3659 .flags = GENL_ADMIN_PERM,
3660 .doit = ip_vs_genl_set_cmd,
3664 static int __init ip_vs_genl_register(void)
3666 return genl_register_family_with_ops(&ip_vs_genl_family,
3667 ip_vs_genl_ops, ARRAY_SIZE(ip_vs_genl_ops));
3670 static void ip_vs_genl_unregister(void)
3672 genl_unregister_family(&ip_vs_genl_family);
3675 /* End of Generic Netlink interface definitions */
3678 * per netns intit/exit func.
3680 #ifdef CONFIG_SYSCTL
3681 int __net_init ip_vs_control_net_init_sysctl(struct net *net)
3684 struct netns_ipvs *ipvs = net_ipvs(net);
3685 struct ctl_table *tbl;
3687 atomic_set(&ipvs->dropentry, 0);
3688 spin_lock_init(&ipvs->dropentry_lock);
3689 spin_lock_init(&ipvs->droppacket_lock);
3690 spin_lock_init(&ipvs->securetcp_lock);
3692 if (!net_eq(net, &init_net)) {
3693 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3698 /* Initialize sysctl defaults */
3700 ipvs->sysctl_amemthresh = 1024;
3701 tbl[idx++].data = &ipvs->sysctl_amemthresh;
3702 ipvs->sysctl_am_droprate = 10;
3703 tbl[idx++].data = &ipvs->sysctl_am_droprate;
3704 tbl[idx++].data = &ipvs->sysctl_drop_entry;
3705 tbl[idx++].data = &ipvs->sysctl_drop_packet;
3706 #ifdef CONFIG_IP_VS_NFCT
3707 tbl[idx++].data = &ipvs->sysctl_conntrack;
3709 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3710 ipvs->sysctl_snat_reroute = 1;
3711 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3712 ipvs->sysctl_sync_ver = 1;
3713 tbl[idx++].data = &ipvs->sysctl_sync_ver;
3714 ipvs->sysctl_sync_ports = 1;
3715 tbl[idx++].data = &ipvs->sysctl_sync_ports;
3716 ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3717 tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3718 ipvs->sysctl_sync_sock_size = 0;
3719 tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3720 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3721 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3722 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3723 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3724 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3725 tbl[idx].data = &ipvs->sysctl_sync_threshold;
3726 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3727 ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3728 tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3729 ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3730 tbl[idx++].data = &ipvs->sysctl_sync_retries;
3731 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3734 ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3735 if (ipvs->sysctl_hdr == NULL) {
3736 if (!net_eq(net, &init_net))
3740 ip_vs_start_estimator(net, &ipvs->tot_stats);
3741 ipvs->sysctl_tbl = tbl;
3742 /* Schedule defense work */
3743 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3744 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3749 void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net)
3751 struct netns_ipvs *ipvs = net_ipvs(net);
3753 cancel_delayed_work_sync(&ipvs->defense_work);
3754 cancel_work_sync(&ipvs->defense_work.work);
3755 unregister_net_sysctl_table(ipvs->sysctl_hdr);
3760 int __net_init ip_vs_control_net_init_sysctl(struct net *net) { return 0; }
3761 void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net) { }
3765 static struct notifier_block ip_vs_dst_notifier = {
3766 .notifier_call = ip_vs_dst_event,
3769 int __net_init ip_vs_control_net_init(struct net *net)
3772 struct netns_ipvs *ipvs = net_ipvs(net);
3774 rwlock_init(&ipvs->rs_lock);
3776 /* Initialize rs_table */
3777 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
3778 INIT_LIST_HEAD(&ipvs->rs_table[idx]);
3780 INIT_LIST_HEAD(&ipvs->dest_trash);
3781 atomic_set(&ipvs->ftpsvc_counter, 0);
3782 atomic_set(&ipvs->nullsvc_counter, 0);
3785 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
3786 if (!ipvs->tot_stats.cpustats)
3789 spin_lock_init(&ipvs->tot_stats.lock);
3791 proc_net_fops_create(net, "ip_vs", 0, &ip_vs_info_fops);
3792 proc_net_fops_create(net, "ip_vs_stats", 0, &ip_vs_stats_fops);
3793 proc_net_fops_create(net, "ip_vs_stats_percpu", 0,
3794 &ip_vs_stats_percpu_fops);
3796 if (ip_vs_control_net_init_sysctl(net))
3802 free_percpu(ipvs->tot_stats.cpustats);
3806 void __net_exit ip_vs_control_net_cleanup(struct net *net)
3808 struct netns_ipvs *ipvs = net_ipvs(net);
3810 ip_vs_trash_cleanup(net);
3811 ip_vs_stop_estimator(net, &ipvs->tot_stats);
3812 ip_vs_control_net_cleanup_sysctl(net);
3813 proc_net_remove(net, "ip_vs_stats_percpu");
3814 proc_net_remove(net, "ip_vs_stats");
3815 proc_net_remove(net, "ip_vs");
3816 free_percpu(ipvs->tot_stats.cpustats);
3819 int __init ip_vs_register_nl_ioctl(void)
3823 ret = nf_register_sockopt(&ip_vs_sockopts);
3825 pr_err("cannot register sockopt.\n");
3829 ret = ip_vs_genl_register();
3831 pr_err("cannot register Generic Netlink interface.\n");
3837 nf_unregister_sockopt(&ip_vs_sockopts);
3842 void ip_vs_unregister_nl_ioctl(void)
3844 ip_vs_genl_unregister();
3845 nf_unregister_sockopt(&ip_vs_sockopts);
3848 int __init ip_vs_control_init(void)
3855 /* Initialize svc_table, ip_vs_svc_fwm_table, rs_table */
3856 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
3857 INIT_LIST_HEAD(&ip_vs_svc_table[idx]);
3858 INIT_LIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3861 smp_wmb(); /* Do we really need it now ? */
3863 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
3872 void ip_vs_control_cleanup(void)
3875 unregister_netdevice_notifier(&ip_vs_dst_notifier);