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)
1177 /* I'm the first user of the service */
1178 atomic_set(&svc->usecnt, 0);
1179 atomic_set(&svc->refcnt, 0);
1182 svc->protocol = u->protocol;
1183 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1184 svc->port = u->port;
1185 svc->fwmark = u->fwmark;
1186 svc->flags = u->flags;
1187 svc->timeout = u->timeout * HZ;
1188 svc->netmask = u->netmask;
1191 INIT_LIST_HEAD(&svc->destinations);
1192 rwlock_init(&svc->sched_lock);
1193 spin_lock_init(&svc->stats.lock);
1195 /* Bind the scheduler */
1196 ret = ip_vs_bind_scheduler(svc, sched);
1201 /* Bind the ct retriever */
1202 ip_vs_bind_pe(svc, pe);
1205 /* Update the virtual service counters */
1206 if (svc->port == FTPPORT)
1207 atomic_inc(&ipvs->ftpsvc_counter);
1208 else if (svc->port == 0)
1209 atomic_inc(&ipvs->nullsvc_counter);
1211 ip_vs_start_estimator(net, &svc->stats);
1213 /* Count only IPv4 services for old get/setsockopt interface */
1214 if (svc->af == AF_INET)
1215 ipvs->num_services++;
1217 /* Hash the service into the service table */
1218 write_lock_bh(&__ip_vs_svc_lock);
1219 ip_vs_svc_hash(svc);
1220 write_unlock_bh(&__ip_vs_svc_lock);
1223 /* Now there is a service - full throttle */
1230 ip_vs_unbind_scheduler(svc);
1233 ip_vs_app_inc_put(svc->inc);
1236 if (svc->stats.cpustats)
1237 free_percpu(svc->stats.cpustats);
1240 ip_vs_scheduler_put(sched);
1243 /* decrease the module use count */
1244 ip_vs_use_count_dec();
1251 * Edit a service and bind it with a new scheduler
1254 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1256 struct ip_vs_scheduler *sched, *old_sched;
1257 struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1261 * Lookup the scheduler, by 'u->sched_name'
1263 sched = ip_vs_scheduler_get(u->sched_name);
1264 if (sched == NULL) {
1265 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1270 if (u->pe_name && *u->pe_name) {
1271 pe = ip_vs_pe_getbyname(u->pe_name);
1273 pr_info("persistence engine module ip_vs_pe_%s "
1274 "not found\n", u->pe_name);
1281 #ifdef CONFIG_IP_VS_IPV6
1282 if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1288 write_lock_bh(&__ip_vs_svc_lock);
1291 * Wait until all other svc users go away.
1293 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1296 * Set the flags and timeout value
1298 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1299 svc->timeout = u->timeout * HZ;
1300 svc->netmask = u->netmask;
1302 old_sched = svc->scheduler;
1303 if (sched != old_sched) {
1305 * Unbind the old scheduler
1307 if ((ret = ip_vs_unbind_scheduler(svc))) {
1313 * Bind the new scheduler
1315 if ((ret = ip_vs_bind_scheduler(svc, sched))) {
1317 * If ip_vs_bind_scheduler fails, restore the old
1319 * The main reason of failure is out of memory.
1321 * The question is if the old scheduler can be
1322 * restored all the time. TODO: if it cannot be
1323 * restored some time, we must delete the service,
1324 * otherwise the system may crash.
1326 ip_vs_bind_scheduler(svc, old_sched);
1334 ip_vs_unbind_pe(svc);
1335 ip_vs_bind_pe(svc, pe);
1339 write_unlock_bh(&__ip_vs_svc_lock);
1341 ip_vs_scheduler_put(old_sched);
1342 ip_vs_pe_put(old_pe);
1348 * Delete a service from the service list
1349 * - The service must be unlinked, unlocked and not referenced!
1350 * - We are called under _bh lock
1352 static void __ip_vs_del_service(struct ip_vs_service *svc)
1354 struct ip_vs_dest *dest, *nxt;
1355 struct ip_vs_scheduler *old_sched;
1356 struct ip_vs_pe *old_pe;
1357 struct netns_ipvs *ipvs = net_ipvs(svc->net);
1359 pr_info("%s: enter\n", __func__);
1361 /* Count only IPv4 services for old get/setsockopt interface */
1362 if (svc->af == AF_INET)
1363 ipvs->num_services--;
1365 ip_vs_stop_estimator(svc->net, &svc->stats);
1367 /* Unbind scheduler */
1368 old_sched = svc->scheduler;
1369 ip_vs_unbind_scheduler(svc);
1370 ip_vs_scheduler_put(old_sched);
1372 /* Unbind persistence engine */
1374 ip_vs_unbind_pe(svc);
1375 ip_vs_pe_put(old_pe);
1377 /* Unbind app inc */
1379 ip_vs_app_inc_put(svc->inc);
1384 * Unlink the whole destination list
1386 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1387 __ip_vs_unlink_dest(svc, dest, 0);
1388 __ip_vs_del_dest(svc->net, dest);
1392 * Update the virtual service counters
1394 if (svc->port == FTPPORT)
1395 atomic_dec(&ipvs->ftpsvc_counter);
1396 else if (svc->port == 0)
1397 atomic_dec(&ipvs->nullsvc_counter);
1400 * Free the service if nobody refers to it
1402 if (atomic_read(&svc->refcnt) == 0) {
1403 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
1405 IP_VS_DBG_ADDR(svc->af, &svc->addr),
1406 ntohs(svc->port), atomic_read(&svc->usecnt));
1407 free_percpu(svc->stats.cpustats);
1411 /* decrease the module use count */
1412 ip_vs_use_count_dec();
1416 * Unlink a service from list and try to delete it if its refcnt reached 0
1418 static void ip_vs_unlink_service(struct ip_vs_service *svc)
1421 * Unhash it from the service table
1423 write_lock_bh(&__ip_vs_svc_lock);
1425 ip_vs_svc_unhash(svc);
1428 * Wait until all the svc users go away.
1430 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1432 __ip_vs_del_service(svc);
1434 write_unlock_bh(&__ip_vs_svc_lock);
1438 * Delete a service from the service list
1440 static int ip_vs_del_service(struct ip_vs_service *svc)
1444 ip_vs_unlink_service(svc);
1451 * Flush all the virtual services
1453 static int ip_vs_flush(struct net *net)
1456 struct ip_vs_service *svc, *nxt;
1459 * Flush the service table hashed by <netns,protocol,addr,port>
1461 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1462 list_for_each_entry_safe(svc, nxt, &ip_vs_svc_table[idx],
1464 if (net_eq(svc->net, net))
1465 ip_vs_unlink_service(svc);
1470 * Flush the service table hashed by fwmark
1472 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1473 list_for_each_entry_safe(svc, nxt,
1474 &ip_vs_svc_fwm_table[idx], f_list) {
1475 if (net_eq(svc->net, net))
1476 ip_vs_unlink_service(svc);
1484 * Delete service by {netns} in the service table.
1485 * Called by __ip_vs_cleanup()
1487 void ip_vs_service_net_cleanup(struct net *net)
1490 /* Check for "full" addressed entries */
1491 mutex_lock(&__ip_vs_mutex);
1493 mutex_unlock(&__ip_vs_mutex);
1497 * Release dst hold by dst_cache
1500 __ip_vs_dev_reset(struct ip_vs_dest *dest, struct net_device *dev)
1502 spin_lock_bh(&dest->dst_lock);
1503 if (dest->dst_cache && dest->dst_cache->dev == dev) {
1504 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1506 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1508 atomic_read(&dest->refcnt));
1509 ip_vs_dst_reset(dest);
1511 spin_unlock_bh(&dest->dst_lock);
1515 * Netdev event receiver
1516 * Currently only NETDEV_UNREGISTER is handled, i.e. if we hold a reference to
1517 * a device that is "unregister" it must be released.
1519 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1522 struct net_device *dev = ptr;
1523 struct net *net = dev_net(dev);
1524 struct netns_ipvs *ipvs = net_ipvs(net);
1525 struct ip_vs_service *svc;
1526 struct ip_vs_dest *dest;
1529 if (event != NETDEV_UNREGISTER || !ipvs)
1531 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1533 mutex_lock(&__ip_vs_mutex);
1534 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1535 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1536 if (net_eq(svc->net, net)) {
1537 list_for_each_entry(dest, &svc->destinations,
1539 __ip_vs_dev_reset(dest, dev);
1544 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1545 if (net_eq(svc->net, net)) {
1546 list_for_each_entry(dest, &svc->destinations,
1548 __ip_vs_dev_reset(dest, dev);
1555 list_for_each_entry(dest, &ipvs->dest_trash, n_list) {
1556 __ip_vs_dev_reset(dest, dev);
1558 mutex_unlock(&__ip_vs_mutex);
1564 * Zero counters in a service or all services
1566 static int ip_vs_zero_service(struct ip_vs_service *svc)
1568 struct ip_vs_dest *dest;
1570 write_lock_bh(&__ip_vs_svc_lock);
1571 list_for_each_entry(dest, &svc->destinations, n_list) {
1572 ip_vs_zero_stats(&dest->stats);
1574 ip_vs_zero_stats(&svc->stats);
1575 write_unlock_bh(&__ip_vs_svc_lock);
1579 static int ip_vs_zero_all(struct net *net)
1582 struct ip_vs_service *svc;
1584 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1585 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1586 if (net_eq(svc->net, net))
1587 ip_vs_zero_service(svc);
1591 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1592 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1593 if (net_eq(svc->net, net))
1594 ip_vs_zero_service(svc);
1598 ip_vs_zero_stats(&net_ipvs(net)->tot_stats);
1602 #ifdef CONFIG_SYSCTL
1605 static int three = 3;
1608 proc_do_defense_mode(ctl_table *table, int write,
1609 void __user *buffer, size_t *lenp, loff_t *ppos)
1611 struct net *net = current->nsproxy->net_ns;
1612 int *valp = table->data;
1616 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1617 if (write && (*valp != val)) {
1618 if ((*valp < 0) || (*valp > 3)) {
1619 /* Restore the correct value */
1622 update_defense_level(net_ipvs(net));
1629 proc_do_sync_threshold(ctl_table *table, int write,
1630 void __user *buffer, size_t *lenp, loff_t *ppos)
1632 int *valp = table->data;
1636 /* backup the value first */
1637 memcpy(val, valp, sizeof(val));
1639 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1640 if (write && (valp[0] < 0 || valp[1] < 0 ||
1641 (valp[0] >= valp[1] && valp[1]))) {
1642 /* Restore the correct value */
1643 memcpy(valp, val, sizeof(val));
1649 proc_do_sync_mode(ctl_table *table, int write,
1650 void __user *buffer, size_t *lenp, loff_t *ppos)
1652 int *valp = table->data;
1656 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1657 if (write && (*valp != val)) {
1658 if ((*valp < 0) || (*valp > 1)) {
1659 /* Restore the correct value */
1667 proc_do_sync_ports(ctl_table *table, int write,
1668 void __user *buffer, size_t *lenp, loff_t *ppos)
1670 int *valp = table->data;
1674 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1675 if (write && (*valp != val)) {
1676 if (*valp < 1 || !is_power_of_2(*valp)) {
1677 /* Restore the correct value */
1685 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1686 * Do not change order or insert new entries without
1687 * align with netns init in ip_vs_control_net_init()
1690 static struct ctl_table vs_vars[] = {
1692 .procname = "amemthresh",
1693 .maxlen = sizeof(int),
1695 .proc_handler = proc_dointvec,
1698 .procname = "am_droprate",
1699 .maxlen = sizeof(int),
1701 .proc_handler = proc_dointvec,
1704 .procname = "drop_entry",
1705 .maxlen = sizeof(int),
1707 .proc_handler = proc_do_defense_mode,
1710 .procname = "drop_packet",
1711 .maxlen = sizeof(int),
1713 .proc_handler = proc_do_defense_mode,
1715 #ifdef CONFIG_IP_VS_NFCT
1717 .procname = "conntrack",
1718 .maxlen = sizeof(int),
1720 .proc_handler = &proc_dointvec,
1724 .procname = "secure_tcp",
1725 .maxlen = sizeof(int),
1727 .proc_handler = proc_do_defense_mode,
1730 .procname = "snat_reroute",
1731 .maxlen = sizeof(int),
1733 .proc_handler = &proc_dointvec,
1736 .procname = "sync_version",
1737 .maxlen = sizeof(int),
1739 .proc_handler = &proc_do_sync_mode,
1742 .procname = "sync_ports",
1743 .maxlen = sizeof(int),
1745 .proc_handler = &proc_do_sync_ports,
1748 .procname = "sync_qlen_max",
1749 .maxlen = sizeof(int),
1751 .proc_handler = proc_dointvec,
1754 .procname = "sync_sock_size",
1755 .maxlen = sizeof(int),
1757 .proc_handler = proc_dointvec,
1760 .procname = "cache_bypass",
1761 .maxlen = sizeof(int),
1763 .proc_handler = proc_dointvec,
1766 .procname = "expire_nodest_conn",
1767 .maxlen = sizeof(int),
1769 .proc_handler = proc_dointvec,
1772 .procname = "expire_quiescent_template",
1773 .maxlen = sizeof(int),
1775 .proc_handler = proc_dointvec,
1778 .procname = "sync_threshold",
1780 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1782 .proc_handler = proc_do_sync_threshold,
1785 .procname = "sync_refresh_period",
1786 .maxlen = sizeof(int),
1788 .proc_handler = proc_dointvec_jiffies,
1791 .procname = "sync_retries",
1792 .maxlen = sizeof(int),
1794 .proc_handler = proc_dointvec_minmax,
1799 .procname = "nat_icmp_send",
1800 .maxlen = sizeof(int),
1802 .proc_handler = proc_dointvec,
1804 #ifdef CONFIG_IP_VS_DEBUG
1806 .procname = "debug_level",
1807 .data = &sysctl_ip_vs_debug_level,
1808 .maxlen = sizeof(int),
1810 .proc_handler = proc_dointvec,
1815 .procname = "timeout_established",
1816 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1817 .maxlen = sizeof(int),
1819 .proc_handler = proc_dointvec_jiffies,
1822 .procname = "timeout_synsent",
1823 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1824 .maxlen = sizeof(int),
1826 .proc_handler = proc_dointvec_jiffies,
1829 .procname = "timeout_synrecv",
1830 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1831 .maxlen = sizeof(int),
1833 .proc_handler = proc_dointvec_jiffies,
1836 .procname = "timeout_finwait",
1837 .data = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1838 .maxlen = sizeof(int),
1840 .proc_handler = proc_dointvec_jiffies,
1843 .procname = "timeout_timewait",
1844 .data = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1845 .maxlen = sizeof(int),
1847 .proc_handler = proc_dointvec_jiffies,
1850 .procname = "timeout_close",
1851 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1852 .maxlen = sizeof(int),
1854 .proc_handler = proc_dointvec_jiffies,
1857 .procname = "timeout_closewait",
1858 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1859 .maxlen = sizeof(int),
1861 .proc_handler = proc_dointvec_jiffies,
1864 .procname = "timeout_lastack",
1865 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1866 .maxlen = sizeof(int),
1868 .proc_handler = proc_dointvec_jiffies,
1871 .procname = "timeout_listen",
1872 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1873 .maxlen = sizeof(int),
1875 .proc_handler = proc_dointvec_jiffies,
1878 .procname = "timeout_synack",
1879 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1880 .maxlen = sizeof(int),
1882 .proc_handler = proc_dointvec_jiffies,
1885 .procname = "timeout_udp",
1886 .data = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1887 .maxlen = sizeof(int),
1889 .proc_handler = proc_dointvec_jiffies,
1892 .procname = "timeout_icmp",
1893 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1894 .maxlen = sizeof(int),
1896 .proc_handler = proc_dointvec_jiffies,
1904 #ifdef CONFIG_PROC_FS
1907 struct seq_net_private p; /* Do not move this, netns depends upon it*/
1908 struct list_head *table;
1913 * Write the contents of the VS rule table to a PROCfs file.
1914 * (It is kept just for backward compatibility)
1916 static inline const char *ip_vs_fwd_name(unsigned int flags)
1918 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1919 case IP_VS_CONN_F_LOCALNODE:
1921 case IP_VS_CONN_F_TUNNEL:
1923 case IP_VS_CONN_F_DROUTE:
1931 /* Get the Nth entry in the two lists */
1932 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1934 struct net *net = seq_file_net(seq);
1935 struct ip_vs_iter *iter = seq->private;
1937 struct ip_vs_service *svc;
1939 /* look in hash by protocol */
1940 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1941 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1942 if (net_eq(svc->net, net) && pos-- == 0) {
1943 iter->table = ip_vs_svc_table;
1950 /* keep looking in fwmark */
1951 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1952 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1953 if (net_eq(svc->net, net) && pos-- == 0) {
1954 iter->table = ip_vs_svc_fwm_table;
1964 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1965 __acquires(__ip_vs_svc_lock)
1968 read_lock_bh(&__ip_vs_svc_lock);
1969 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1973 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1975 struct list_head *e;
1976 struct ip_vs_iter *iter;
1977 struct ip_vs_service *svc;
1980 if (v == SEQ_START_TOKEN)
1981 return ip_vs_info_array(seq,0);
1984 iter = seq->private;
1986 if (iter->table == ip_vs_svc_table) {
1987 /* next service in table hashed by protocol */
1988 if ((e = svc->s_list.next) != &ip_vs_svc_table[iter->bucket])
1989 return list_entry(e, struct ip_vs_service, s_list);
1992 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1993 list_for_each_entry(svc,&ip_vs_svc_table[iter->bucket],
1999 iter->table = ip_vs_svc_fwm_table;
2004 /* next service in hashed by fwmark */
2005 if ((e = svc->f_list.next) != &ip_vs_svc_fwm_table[iter->bucket])
2006 return list_entry(e, struct ip_vs_service, f_list);
2009 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2010 list_for_each_entry(svc, &ip_vs_svc_fwm_table[iter->bucket],
2018 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
2019 __releases(__ip_vs_svc_lock)
2021 read_unlock_bh(&__ip_vs_svc_lock);
2025 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
2027 if (v == SEQ_START_TOKEN) {
2029 "IP Virtual Server version %d.%d.%d (size=%d)\n",
2030 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2032 "Prot LocalAddress:Port Scheduler Flags\n");
2034 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2036 const struct ip_vs_service *svc = v;
2037 const struct ip_vs_iter *iter = seq->private;
2038 const struct ip_vs_dest *dest;
2040 if (iter->table == ip_vs_svc_table) {
2041 #ifdef CONFIG_IP_VS_IPV6
2042 if (svc->af == AF_INET6)
2043 seq_printf(seq, "%s [%pI6]:%04X %s ",
2044 ip_vs_proto_name(svc->protocol),
2047 svc->scheduler->name);
2050 seq_printf(seq, "%s %08X:%04X %s %s ",
2051 ip_vs_proto_name(svc->protocol),
2052 ntohl(svc->addr.ip),
2054 svc->scheduler->name,
2055 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2057 seq_printf(seq, "FWM %08X %s %s",
2058 svc->fwmark, svc->scheduler->name,
2059 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2062 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2063 seq_printf(seq, "persistent %d %08X\n",
2065 ntohl(svc->netmask));
2067 seq_putc(seq, '\n');
2069 list_for_each_entry(dest, &svc->destinations, n_list) {
2070 #ifdef CONFIG_IP_VS_IPV6
2071 if (dest->af == AF_INET6)
2074 " %-7s %-6d %-10d %-10d\n",
2077 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2078 atomic_read(&dest->weight),
2079 atomic_read(&dest->activeconns),
2080 atomic_read(&dest->inactconns));
2085 "%-7s %-6d %-10d %-10d\n",
2086 ntohl(dest->addr.ip),
2088 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2089 atomic_read(&dest->weight),
2090 atomic_read(&dest->activeconns),
2091 atomic_read(&dest->inactconns));
2098 static const struct seq_operations ip_vs_info_seq_ops = {
2099 .start = ip_vs_info_seq_start,
2100 .next = ip_vs_info_seq_next,
2101 .stop = ip_vs_info_seq_stop,
2102 .show = ip_vs_info_seq_show,
2105 static int ip_vs_info_open(struct inode *inode, struct file *file)
2107 return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2108 sizeof(struct ip_vs_iter));
2111 static const struct file_operations ip_vs_info_fops = {
2112 .owner = THIS_MODULE,
2113 .open = ip_vs_info_open,
2115 .llseek = seq_lseek,
2116 .release = seq_release_net,
2119 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2121 struct net *net = seq_file_single_net(seq);
2122 struct ip_vs_stats_user show;
2124 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2126 " Total Incoming Outgoing Incoming Outgoing\n");
2128 " Conns Packets Packets Bytes Bytes\n");
2130 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2131 seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", show.conns,
2132 show.inpkts, show.outpkts,
2133 (unsigned long long) show.inbytes,
2134 (unsigned long long) show.outbytes);
2136 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2138 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2139 seq_printf(seq, "%8X %8X %8X %16X %16X\n",
2140 show.cps, show.inpps, show.outpps,
2141 show.inbps, show.outbps);
2146 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2148 return single_open_net(inode, file, ip_vs_stats_show);
2151 static const struct file_operations ip_vs_stats_fops = {
2152 .owner = THIS_MODULE,
2153 .open = ip_vs_stats_seq_open,
2155 .llseek = seq_lseek,
2156 .release = single_release_net,
2159 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2161 struct net *net = seq_file_single_net(seq);
2162 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2163 struct ip_vs_cpu_stats *cpustats = tot_stats->cpustats;
2164 struct ip_vs_stats_user rates;
2167 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2169 " Total Incoming Outgoing Incoming Outgoing\n");
2171 "CPU Conns Packets Packets Bytes Bytes\n");
2173 for_each_possible_cpu(i) {
2174 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2176 __u64 inbytes, outbytes;
2179 start = u64_stats_fetch_begin_bh(&u->syncp);
2180 inbytes = u->ustats.inbytes;
2181 outbytes = u->ustats.outbytes;
2182 } while (u64_stats_fetch_retry_bh(&u->syncp, start));
2184 seq_printf(seq, "%3X %8X %8X %8X %16LX %16LX\n",
2185 i, u->ustats.conns, u->ustats.inpkts,
2186 u->ustats.outpkts, (__u64)inbytes,
2190 spin_lock_bh(&tot_stats->lock);
2192 seq_printf(seq, " ~ %8X %8X %8X %16LX %16LX\n\n",
2193 tot_stats->ustats.conns, tot_stats->ustats.inpkts,
2194 tot_stats->ustats.outpkts,
2195 (unsigned long long) tot_stats->ustats.inbytes,
2196 (unsigned long long) tot_stats->ustats.outbytes);
2198 ip_vs_read_estimator(&rates, tot_stats);
2200 spin_unlock_bh(&tot_stats->lock);
2202 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2204 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2205 seq_printf(seq, " %8X %8X %8X %16X %16X\n",
2215 static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2217 return single_open_net(inode, file, ip_vs_stats_percpu_show);
2220 static const struct file_operations ip_vs_stats_percpu_fops = {
2221 .owner = THIS_MODULE,
2222 .open = ip_vs_stats_percpu_seq_open,
2224 .llseek = seq_lseek,
2225 .release = single_release_net,
2230 * Set timeout values for tcp tcpfin udp in the timeout_table.
2232 static int ip_vs_set_timeout(struct net *net, struct ip_vs_timeout_user *u)
2234 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2235 struct ip_vs_proto_data *pd;
2238 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2243 #ifdef CONFIG_IP_VS_PROTO_TCP
2244 if (u->tcp_timeout) {
2245 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2246 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2247 = u->tcp_timeout * HZ;
2250 if (u->tcp_fin_timeout) {
2251 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2252 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2253 = u->tcp_fin_timeout * HZ;
2257 #ifdef CONFIG_IP_VS_PROTO_UDP
2258 if (u->udp_timeout) {
2259 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2260 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2261 = u->udp_timeout * HZ;
2268 #define SET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2269 #define SERVICE_ARG_LEN (sizeof(struct ip_vs_service_user))
2270 #define SVCDEST_ARG_LEN (sizeof(struct ip_vs_service_user) + \
2271 sizeof(struct ip_vs_dest_user))
2272 #define TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2273 #define DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user))
2274 #define MAX_ARG_LEN SVCDEST_ARG_LEN
2276 static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
2277 [SET_CMDID(IP_VS_SO_SET_ADD)] = SERVICE_ARG_LEN,
2278 [SET_CMDID(IP_VS_SO_SET_EDIT)] = SERVICE_ARG_LEN,
2279 [SET_CMDID(IP_VS_SO_SET_DEL)] = SERVICE_ARG_LEN,
2280 [SET_CMDID(IP_VS_SO_SET_FLUSH)] = 0,
2281 [SET_CMDID(IP_VS_SO_SET_ADDDEST)] = SVCDEST_ARG_LEN,
2282 [SET_CMDID(IP_VS_SO_SET_DELDEST)] = SVCDEST_ARG_LEN,
2283 [SET_CMDID(IP_VS_SO_SET_EDITDEST)] = SVCDEST_ARG_LEN,
2284 [SET_CMDID(IP_VS_SO_SET_TIMEOUT)] = TIMEOUT_ARG_LEN,
2285 [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)] = DAEMON_ARG_LEN,
2286 [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)] = DAEMON_ARG_LEN,
2287 [SET_CMDID(IP_VS_SO_SET_ZERO)] = SERVICE_ARG_LEN,
2290 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2291 struct ip_vs_service_user *usvc_compat)
2293 memset(usvc, 0, sizeof(*usvc));
2296 usvc->protocol = usvc_compat->protocol;
2297 usvc->addr.ip = usvc_compat->addr;
2298 usvc->port = usvc_compat->port;
2299 usvc->fwmark = usvc_compat->fwmark;
2301 /* Deep copy of sched_name is not needed here */
2302 usvc->sched_name = usvc_compat->sched_name;
2304 usvc->flags = usvc_compat->flags;
2305 usvc->timeout = usvc_compat->timeout;
2306 usvc->netmask = usvc_compat->netmask;
2309 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2310 struct ip_vs_dest_user *udest_compat)
2312 memset(udest, 0, sizeof(*udest));
2314 udest->addr.ip = udest_compat->addr;
2315 udest->port = udest_compat->port;
2316 udest->conn_flags = udest_compat->conn_flags;
2317 udest->weight = udest_compat->weight;
2318 udest->u_threshold = udest_compat->u_threshold;
2319 udest->l_threshold = udest_compat->l_threshold;
2323 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2325 struct net *net = sock_net(sk);
2327 unsigned char arg[MAX_ARG_LEN];
2328 struct ip_vs_service_user *usvc_compat;
2329 struct ip_vs_service_user_kern usvc;
2330 struct ip_vs_service *svc;
2331 struct ip_vs_dest_user *udest_compat;
2332 struct ip_vs_dest_user_kern udest;
2333 struct netns_ipvs *ipvs = net_ipvs(net);
2335 if (!capable(CAP_NET_ADMIN))
2338 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2340 if (len < 0 || len > MAX_ARG_LEN)
2342 if (len != set_arglen[SET_CMDID(cmd)]) {
2343 pr_err("set_ctl: len %u != %u\n",
2344 len, set_arglen[SET_CMDID(cmd)]);
2348 if (copy_from_user(arg, user, len) != 0)
2351 /* increase the module use count */
2352 ip_vs_use_count_inc();
2354 /* Handle daemons since they have another lock */
2355 if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2356 cmd == IP_VS_SO_SET_STOPDAEMON) {
2357 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2359 if (mutex_lock_interruptible(&ipvs->sync_mutex)) {
2363 if (cmd == IP_VS_SO_SET_STARTDAEMON)
2364 ret = start_sync_thread(net, dm->state, dm->mcast_ifn,
2367 ret = stop_sync_thread(net, dm->state);
2368 mutex_unlock(&ipvs->sync_mutex);
2372 if (mutex_lock_interruptible(&__ip_vs_mutex)) {
2377 if (cmd == IP_VS_SO_SET_FLUSH) {
2378 /* Flush the virtual service */
2379 ret = ip_vs_flush(net);
2381 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2382 /* Set timeout values for (tcp tcpfin udp) */
2383 ret = ip_vs_set_timeout(net, (struct ip_vs_timeout_user *)arg);
2387 usvc_compat = (struct ip_vs_service_user *)arg;
2388 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2390 /* We only use the new structs internally, so copy userspace compat
2391 * structs to extended internal versions */
2392 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2393 ip_vs_copy_udest_compat(&udest, udest_compat);
2395 if (cmd == IP_VS_SO_SET_ZERO) {
2396 /* if no service address is set, zero counters in all */
2397 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2398 ret = ip_vs_zero_all(net);
2403 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2404 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2405 usvc.protocol != IPPROTO_SCTP) {
2406 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2407 usvc.protocol, &usvc.addr.ip,
2408 ntohs(usvc.port), usvc.sched_name);
2413 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2414 if (usvc.fwmark == 0)
2415 svc = __ip_vs_service_find(net, usvc.af, usvc.protocol,
2416 &usvc.addr, usvc.port);
2418 svc = __ip_vs_svc_fwm_find(net, usvc.af, usvc.fwmark);
2420 if (cmd != IP_VS_SO_SET_ADD
2421 && (svc == NULL || svc->protocol != usvc.protocol)) {
2427 case IP_VS_SO_SET_ADD:
2431 ret = ip_vs_add_service(net, &usvc, &svc);
2433 case IP_VS_SO_SET_EDIT:
2434 ret = ip_vs_edit_service(svc, &usvc);
2436 case IP_VS_SO_SET_DEL:
2437 ret = ip_vs_del_service(svc);
2441 case IP_VS_SO_SET_ZERO:
2442 ret = ip_vs_zero_service(svc);
2444 case IP_VS_SO_SET_ADDDEST:
2445 ret = ip_vs_add_dest(svc, &udest);
2447 case IP_VS_SO_SET_EDITDEST:
2448 ret = ip_vs_edit_dest(svc, &udest);
2450 case IP_VS_SO_SET_DELDEST:
2451 ret = ip_vs_del_dest(svc, &udest);
2458 mutex_unlock(&__ip_vs_mutex);
2460 /* decrease the module use count */
2461 ip_vs_use_count_dec();
2468 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2470 dst->protocol = src->protocol;
2471 dst->addr = src->addr.ip;
2472 dst->port = src->port;
2473 dst->fwmark = src->fwmark;
2474 strlcpy(dst->sched_name, src->scheduler->name, sizeof(dst->sched_name));
2475 dst->flags = src->flags;
2476 dst->timeout = src->timeout / HZ;
2477 dst->netmask = src->netmask;
2478 dst->num_dests = src->num_dests;
2479 ip_vs_copy_stats(&dst->stats, &src->stats);
2483 __ip_vs_get_service_entries(struct net *net,
2484 const struct ip_vs_get_services *get,
2485 struct ip_vs_get_services __user *uptr)
2488 struct ip_vs_service *svc;
2489 struct ip_vs_service_entry entry;
2492 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2493 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2494 /* Only expose IPv4 entries to old interface */
2495 if (svc->af != AF_INET || !net_eq(svc->net, net))
2498 if (count >= get->num_services)
2500 memset(&entry, 0, sizeof(entry));
2501 ip_vs_copy_service(&entry, svc);
2502 if (copy_to_user(&uptr->entrytable[count],
2503 &entry, sizeof(entry))) {
2511 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2512 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2513 /* Only expose IPv4 entries to old interface */
2514 if (svc->af != AF_INET || !net_eq(svc->net, net))
2517 if (count >= get->num_services)
2519 memset(&entry, 0, sizeof(entry));
2520 ip_vs_copy_service(&entry, svc);
2521 if (copy_to_user(&uptr->entrytable[count],
2522 &entry, sizeof(entry))) {
2534 __ip_vs_get_dest_entries(struct net *net, const struct ip_vs_get_dests *get,
2535 struct ip_vs_get_dests __user *uptr)
2537 struct ip_vs_service *svc;
2538 union nf_inet_addr addr = { .ip = get->addr };
2542 svc = __ip_vs_svc_fwm_find(net, AF_INET, get->fwmark);
2544 svc = __ip_vs_service_find(net, AF_INET, get->protocol, &addr,
2549 struct ip_vs_dest *dest;
2550 struct ip_vs_dest_entry entry;
2552 list_for_each_entry(dest, &svc->destinations, n_list) {
2553 if (count >= get->num_dests)
2556 entry.addr = dest->addr.ip;
2557 entry.port = dest->port;
2558 entry.conn_flags = atomic_read(&dest->conn_flags);
2559 entry.weight = atomic_read(&dest->weight);
2560 entry.u_threshold = dest->u_threshold;
2561 entry.l_threshold = dest->l_threshold;
2562 entry.activeconns = atomic_read(&dest->activeconns);
2563 entry.inactconns = atomic_read(&dest->inactconns);
2564 entry.persistconns = atomic_read(&dest->persistconns);
2565 ip_vs_copy_stats(&entry.stats, &dest->stats);
2566 if (copy_to_user(&uptr->entrytable[count],
2567 &entry, sizeof(entry))) {
2579 __ip_vs_get_timeouts(struct net *net, struct ip_vs_timeout_user *u)
2581 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2582 struct ip_vs_proto_data *pd;
2585 #ifdef CONFIG_IP_VS_PROTO_TCP
2586 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2587 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2588 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2590 #ifdef CONFIG_IP_VS_PROTO_UDP
2591 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2593 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2598 #define GET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2599 #define GET_INFO_ARG_LEN (sizeof(struct ip_vs_getinfo))
2600 #define GET_SERVICES_ARG_LEN (sizeof(struct ip_vs_get_services))
2601 #define GET_SERVICE_ARG_LEN (sizeof(struct ip_vs_service_entry))
2602 #define GET_DESTS_ARG_LEN (sizeof(struct ip_vs_get_dests))
2603 #define GET_TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2604 #define GET_DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user) * 2)
2606 static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
2607 [GET_CMDID(IP_VS_SO_GET_VERSION)] = 64,
2608 [GET_CMDID(IP_VS_SO_GET_INFO)] = GET_INFO_ARG_LEN,
2609 [GET_CMDID(IP_VS_SO_GET_SERVICES)] = GET_SERVICES_ARG_LEN,
2610 [GET_CMDID(IP_VS_SO_GET_SERVICE)] = GET_SERVICE_ARG_LEN,
2611 [GET_CMDID(IP_VS_SO_GET_DESTS)] = GET_DESTS_ARG_LEN,
2612 [GET_CMDID(IP_VS_SO_GET_TIMEOUT)] = GET_TIMEOUT_ARG_LEN,
2613 [GET_CMDID(IP_VS_SO_GET_DAEMON)] = GET_DAEMON_ARG_LEN,
2617 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2619 unsigned char arg[128];
2621 unsigned int copylen;
2622 struct net *net = sock_net(sk);
2623 struct netns_ipvs *ipvs = net_ipvs(net);
2626 if (!capable(CAP_NET_ADMIN))
2629 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2632 if (*len < get_arglen[GET_CMDID(cmd)]) {
2633 pr_err("get_ctl: len %u < %u\n",
2634 *len, get_arglen[GET_CMDID(cmd)]);
2638 copylen = get_arglen[GET_CMDID(cmd)];
2642 if (copy_from_user(arg, user, copylen) != 0)
2645 * Handle daemons first since it has its own locking
2647 if (cmd == IP_VS_SO_GET_DAEMON) {
2648 struct ip_vs_daemon_user d[2];
2650 memset(&d, 0, sizeof(d));
2651 if (mutex_lock_interruptible(&ipvs->sync_mutex))
2652 return -ERESTARTSYS;
2654 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2655 d[0].state = IP_VS_STATE_MASTER;
2656 strlcpy(d[0].mcast_ifn, ipvs->master_mcast_ifn,
2657 sizeof(d[0].mcast_ifn));
2658 d[0].syncid = ipvs->master_syncid;
2660 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2661 d[1].state = IP_VS_STATE_BACKUP;
2662 strlcpy(d[1].mcast_ifn, ipvs->backup_mcast_ifn,
2663 sizeof(d[1].mcast_ifn));
2664 d[1].syncid = ipvs->backup_syncid;
2666 if (copy_to_user(user, &d, sizeof(d)) != 0)
2668 mutex_unlock(&ipvs->sync_mutex);
2672 if (mutex_lock_interruptible(&__ip_vs_mutex))
2673 return -ERESTARTSYS;
2676 case IP_VS_SO_GET_VERSION:
2680 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2681 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2682 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2686 *len = strlen(buf)+1;
2690 case IP_VS_SO_GET_INFO:
2692 struct ip_vs_getinfo info;
2693 info.version = IP_VS_VERSION_CODE;
2694 info.size = ip_vs_conn_tab_size;
2695 info.num_services = ipvs->num_services;
2696 if (copy_to_user(user, &info, sizeof(info)) != 0)
2701 case IP_VS_SO_GET_SERVICES:
2703 struct ip_vs_get_services *get;
2706 get = (struct ip_vs_get_services *)arg;
2707 size = sizeof(*get) +
2708 sizeof(struct ip_vs_service_entry) * get->num_services;
2710 pr_err("length: %u != %u\n", *len, size);
2714 ret = __ip_vs_get_service_entries(net, get, user);
2718 case IP_VS_SO_GET_SERVICE:
2720 struct ip_vs_service_entry *entry;
2721 struct ip_vs_service *svc;
2722 union nf_inet_addr addr;
2724 entry = (struct ip_vs_service_entry *)arg;
2725 addr.ip = entry->addr;
2727 svc = __ip_vs_svc_fwm_find(net, AF_INET, entry->fwmark);
2729 svc = __ip_vs_service_find(net, AF_INET,
2730 entry->protocol, &addr,
2733 ip_vs_copy_service(entry, svc);
2734 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2741 case IP_VS_SO_GET_DESTS:
2743 struct ip_vs_get_dests *get;
2746 get = (struct ip_vs_get_dests *)arg;
2747 size = sizeof(*get) +
2748 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2750 pr_err("length: %u != %u\n", *len, size);
2754 ret = __ip_vs_get_dest_entries(net, get, user);
2758 case IP_VS_SO_GET_TIMEOUT:
2760 struct ip_vs_timeout_user t;
2762 memset(&t, 0, sizeof(t));
2763 __ip_vs_get_timeouts(net, &t);
2764 if (copy_to_user(user, &t, sizeof(t)) != 0)
2774 mutex_unlock(&__ip_vs_mutex);
2779 static struct nf_sockopt_ops ip_vs_sockopts = {
2781 .set_optmin = IP_VS_BASE_CTL,
2782 .set_optmax = IP_VS_SO_SET_MAX+1,
2783 .set = do_ip_vs_set_ctl,
2784 .get_optmin = IP_VS_BASE_CTL,
2785 .get_optmax = IP_VS_SO_GET_MAX+1,
2786 .get = do_ip_vs_get_ctl,
2787 .owner = THIS_MODULE,
2791 * Generic Netlink interface
2794 /* IPVS genetlink family */
2795 static struct genl_family ip_vs_genl_family = {
2796 .id = GENL_ID_GENERATE,
2798 .name = IPVS_GENL_NAME,
2799 .version = IPVS_GENL_VERSION,
2800 .maxattr = IPVS_CMD_MAX,
2801 .netnsok = true, /* Make ipvsadm to work on netns */
2804 /* Policy used for first-level command attributes */
2805 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2806 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2807 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2808 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2809 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2810 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2811 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2814 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2815 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2816 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2817 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2818 .len = IP_VS_IFNAME_MAXLEN },
2819 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2822 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2823 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2824 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2825 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2826 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2827 .len = sizeof(union nf_inet_addr) },
2828 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2829 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2830 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2831 .len = IP_VS_SCHEDNAME_MAXLEN },
2832 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2833 .len = IP_VS_PENAME_MAXLEN },
2834 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2835 .len = sizeof(struct ip_vs_flags) },
2836 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2837 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2838 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2841 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2842 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2843 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2844 .len = sizeof(union nf_inet_addr) },
2845 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2846 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2847 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2848 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2849 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2850 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2851 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2852 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2853 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2856 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2857 struct ip_vs_stats *stats)
2859 struct ip_vs_stats_user ustats;
2860 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2864 ip_vs_copy_stats(&ustats, stats);
2866 if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, ustats.conns) ||
2867 nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, ustats.inpkts) ||
2868 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, ustats.outpkts) ||
2869 nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, ustats.inbytes) ||
2870 nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, ustats.outbytes) ||
2871 nla_put_u32(skb, IPVS_STATS_ATTR_CPS, ustats.cps) ||
2872 nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, ustats.inpps) ||
2873 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, ustats.outpps) ||
2874 nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, ustats.inbps) ||
2875 nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, ustats.outbps))
2876 goto nla_put_failure;
2877 nla_nest_end(skb, nl_stats);
2882 nla_nest_cancel(skb, nl_stats);
2886 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2887 struct ip_vs_service *svc)
2889 struct nlattr *nl_service;
2890 struct ip_vs_flags flags = { .flags = svc->flags,
2893 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2897 if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2898 goto nla_put_failure;
2900 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2901 goto nla_put_failure;
2903 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2904 nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2905 nla_put_u16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2906 goto nla_put_failure;
2909 if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, svc->scheduler->name) ||
2911 nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, svc->pe->name)) ||
2912 nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2913 nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2914 nla_put_u32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
2915 goto nla_put_failure;
2916 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2917 goto nla_put_failure;
2919 nla_nest_end(skb, nl_service);
2924 nla_nest_cancel(skb, nl_service);
2928 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2929 struct ip_vs_service *svc,
2930 struct netlink_callback *cb)
2934 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2935 &ip_vs_genl_family, NLM_F_MULTI,
2936 IPVS_CMD_NEW_SERVICE);
2940 if (ip_vs_genl_fill_service(skb, svc) < 0)
2941 goto nla_put_failure;
2943 return genlmsg_end(skb, hdr);
2946 genlmsg_cancel(skb, hdr);
2950 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2951 struct netlink_callback *cb)
2954 int start = cb->args[0];
2955 struct ip_vs_service *svc;
2956 struct net *net = skb_sknet(skb);
2958 mutex_lock(&__ip_vs_mutex);
2959 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2960 list_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2961 if (++idx <= start || !net_eq(svc->net, net))
2963 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2965 goto nla_put_failure;
2970 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2971 list_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
2972 if (++idx <= start || !net_eq(svc->net, net))
2974 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2976 goto nla_put_failure;
2982 mutex_unlock(&__ip_vs_mutex);
2988 static int ip_vs_genl_parse_service(struct net *net,
2989 struct ip_vs_service_user_kern *usvc,
2990 struct nlattr *nla, int full_entry,
2991 struct ip_vs_service **ret_svc)
2993 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
2994 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
2995 struct ip_vs_service *svc;
2997 /* Parse mandatory identifying service fields first */
2999 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
3002 nla_af = attrs[IPVS_SVC_ATTR_AF];
3003 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
3004 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
3005 nla_port = attrs[IPVS_SVC_ATTR_PORT];
3006 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
3008 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3011 memset(usvc, 0, sizeof(*usvc));
3013 usvc->af = nla_get_u16(nla_af);
3014 #ifdef CONFIG_IP_VS_IPV6
3015 if (usvc->af != AF_INET && usvc->af != AF_INET6)
3017 if (usvc->af != AF_INET)
3019 return -EAFNOSUPPORT;
3022 usvc->protocol = IPPROTO_TCP;
3023 usvc->fwmark = nla_get_u32(nla_fwmark);
3025 usvc->protocol = nla_get_u16(nla_protocol);
3026 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3027 usvc->port = nla_get_u16(nla_port);
3032 svc = __ip_vs_svc_fwm_find(net, usvc->af, usvc->fwmark);
3034 svc = __ip_vs_service_find(net, usvc->af, usvc->protocol,
3035 &usvc->addr, usvc->port);
3038 /* If a full entry was requested, check for the additional fields */
3040 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3042 struct ip_vs_flags flags;
3044 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3045 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3046 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3047 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3048 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3050 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3053 nla_memcpy(&flags, nla_flags, sizeof(flags));
3055 /* prefill flags from service if it already exists */
3057 usvc->flags = svc->flags;
3059 /* set new flags from userland */
3060 usvc->flags = (usvc->flags & ~flags.mask) |
3061 (flags.flags & flags.mask);
3062 usvc->sched_name = nla_data(nla_sched);
3063 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3064 usvc->timeout = nla_get_u32(nla_timeout);
3065 usvc->netmask = nla_get_u32(nla_netmask);
3071 static struct ip_vs_service *ip_vs_genl_find_service(struct net *net,
3074 struct ip_vs_service_user_kern usvc;
3075 struct ip_vs_service *svc;
3078 ret = ip_vs_genl_parse_service(net, &usvc, nla, 0, &svc);
3079 return ret ? ERR_PTR(ret) : svc;
3082 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3084 struct nlattr *nl_dest;
3086 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3090 if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3091 nla_put_u16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3092 nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3093 (atomic_read(&dest->conn_flags) &
3094 IP_VS_CONN_F_FWD_MASK)) ||
3095 nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3096 atomic_read(&dest->weight)) ||
3097 nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3098 nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3099 nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3100 atomic_read(&dest->activeconns)) ||
3101 nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3102 atomic_read(&dest->inactconns)) ||
3103 nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3104 atomic_read(&dest->persistconns)))
3105 goto nla_put_failure;
3106 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
3107 goto nla_put_failure;
3109 nla_nest_end(skb, nl_dest);
3114 nla_nest_cancel(skb, nl_dest);
3118 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3119 struct netlink_callback *cb)
3123 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
3124 &ip_vs_genl_family, NLM_F_MULTI,
3129 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3130 goto nla_put_failure;
3132 return genlmsg_end(skb, hdr);
3135 genlmsg_cancel(skb, hdr);
3139 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3140 struct netlink_callback *cb)
3143 int start = cb->args[0];
3144 struct ip_vs_service *svc;
3145 struct ip_vs_dest *dest;
3146 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3147 struct net *net = skb_sknet(skb);
3149 mutex_lock(&__ip_vs_mutex);
3151 /* Try to find the service for which to dump destinations */
3152 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3153 IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3157 svc = ip_vs_genl_find_service(net, attrs[IPVS_CMD_ATTR_SERVICE]);
3158 if (IS_ERR(svc) || svc == NULL)
3161 /* Dump the destinations */
3162 list_for_each_entry(dest, &svc->destinations, n_list) {
3165 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3167 goto nla_put_failure;
3175 mutex_unlock(&__ip_vs_mutex);
3180 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3181 struct nlattr *nla, int full_entry)
3183 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3184 struct nlattr *nla_addr, *nla_port;
3186 /* Parse mandatory identifying destination fields first */
3188 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3191 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3192 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3194 if (!(nla_addr && nla_port))
3197 memset(udest, 0, sizeof(*udest));
3199 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3200 udest->port = nla_get_u16(nla_port);
3202 /* If a full entry was requested, check for the additional fields */
3204 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3207 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3208 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3209 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3210 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3212 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3215 udest->conn_flags = nla_get_u32(nla_fwd)
3216 & IP_VS_CONN_F_FWD_MASK;
3217 udest->weight = nla_get_u32(nla_weight);
3218 udest->u_threshold = nla_get_u32(nla_u_thresh);
3219 udest->l_threshold = nla_get_u32(nla_l_thresh);
3225 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __be32 state,
3226 const char *mcast_ifn, __be32 syncid)
3228 struct nlattr *nl_daemon;
3230 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3234 if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3235 nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn) ||
3236 nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid))
3237 goto nla_put_failure;
3238 nla_nest_end(skb, nl_daemon);
3243 nla_nest_cancel(skb, nl_daemon);
3247 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __be32 state,
3248 const char *mcast_ifn, __be32 syncid,
3249 struct netlink_callback *cb)
3252 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
3253 &ip_vs_genl_family, NLM_F_MULTI,
3254 IPVS_CMD_NEW_DAEMON);
3258 if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
3259 goto nla_put_failure;
3261 return genlmsg_end(skb, hdr);
3264 genlmsg_cancel(skb, hdr);
3268 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3269 struct netlink_callback *cb)
3271 struct net *net = skb_sknet(skb);
3272 struct netns_ipvs *ipvs = net_ipvs(net);
3274 mutex_lock(&ipvs->sync_mutex);
3275 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3276 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3277 ipvs->master_mcast_ifn,
3278 ipvs->master_syncid, cb) < 0)
3279 goto nla_put_failure;
3284 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3285 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3286 ipvs->backup_mcast_ifn,
3287 ipvs->backup_syncid, cb) < 0)
3288 goto nla_put_failure;
3294 mutex_unlock(&ipvs->sync_mutex);
3299 static int ip_vs_genl_new_daemon(struct net *net, struct nlattr **attrs)
3301 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3302 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3303 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3306 return start_sync_thread(net,
3307 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3308 nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3309 nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3312 static int ip_vs_genl_del_daemon(struct net *net, struct nlattr **attrs)
3314 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3317 return stop_sync_thread(net,
3318 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3321 static int ip_vs_genl_set_config(struct net *net, struct nlattr **attrs)
3323 struct ip_vs_timeout_user t;
3325 __ip_vs_get_timeouts(net, &t);
3327 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3328 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3330 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3332 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3334 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3335 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3337 return ip_vs_set_timeout(net, &t);
3340 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3344 struct netns_ipvs *ipvs;
3346 net = skb_sknet(skb);
3347 ipvs = net_ipvs(net);
3348 cmd = info->genlhdr->cmd;
3350 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3351 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3353 mutex_lock(&ipvs->sync_mutex);
3354 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3355 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3356 info->attrs[IPVS_CMD_ATTR_DAEMON],
3357 ip_vs_daemon_policy)) {
3362 if (cmd == IPVS_CMD_NEW_DAEMON)
3363 ret = ip_vs_genl_new_daemon(net, daemon_attrs);
3365 ret = ip_vs_genl_del_daemon(net, daemon_attrs);
3367 mutex_unlock(&ipvs->sync_mutex);
3372 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3374 struct ip_vs_service *svc = NULL;
3375 struct ip_vs_service_user_kern usvc;
3376 struct ip_vs_dest_user_kern udest;
3378 int need_full_svc = 0, need_full_dest = 0;
3381 net = skb_sknet(skb);
3382 cmd = info->genlhdr->cmd;
3384 mutex_lock(&__ip_vs_mutex);
3386 if (cmd == IPVS_CMD_FLUSH) {
3387 ret = ip_vs_flush(net);
3389 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3390 ret = ip_vs_genl_set_config(net, info->attrs);
3392 } else if (cmd == IPVS_CMD_ZERO &&
3393 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3394 ret = ip_vs_zero_all(net);
3398 /* All following commands require a service argument, so check if we
3399 * received a valid one. We need a full service specification when
3400 * adding / editing a service. Only identifying members otherwise. */
3401 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3404 ret = ip_vs_genl_parse_service(net, &usvc,
3405 info->attrs[IPVS_CMD_ATTR_SERVICE],
3406 need_full_svc, &svc);
3410 /* Unless we're adding a new service, the service must already exist */
3411 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3416 /* Destination commands require a valid destination argument. For
3417 * adding / editing a destination, we need a full destination
3419 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3420 cmd == IPVS_CMD_DEL_DEST) {
3421 if (cmd != IPVS_CMD_DEL_DEST)
3424 ret = ip_vs_genl_parse_dest(&udest,
3425 info->attrs[IPVS_CMD_ATTR_DEST],
3432 case IPVS_CMD_NEW_SERVICE:
3434 ret = ip_vs_add_service(net, &usvc, &svc);
3438 case IPVS_CMD_SET_SERVICE:
3439 ret = ip_vs_edit_service(svc, &usvc);
3441 case IPVS_CMD_DEL_SERVICE:
3442 ret = ip_vs_del_service(svc);
3443 /* do not use svc, it can be freed */
3445 case IPVS_CMD_NEW_DEST:
3446 ret = ip_vs_add_dest(svc, &udest);
3448 case IPVS_CMD_SET_DEST:
3449 ret = ip_vs_edit_dest(svc, &udest);
3451 case IPVS_CMD_DEL_DEST:
3452 ret = ip_vs_del_dest(svc, &udest);
3455 ret = ip_vs_zero_service(svc);
3462 mutex_unlock(&__ip_vs_mutex);
3467 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3469 struct sk_buff *msg;
3471 int ret, cmd, reply_cmd;
3474 net = skb_sknet(skb);
3475 cmd = info->genlhdr->cmd;
3477 if (cmd == IPVS_CMD_GET_SERVICE)
3478 reply_cmd = IPVS_CMD_NEW_SERVICE;
3479 else if (cmd == IPVS_CMD_GET_INFO)
3480 reply_cmd = IPVS_CMD_SET_INFO;
3481 else if (cmd == IPVS_CMD_GET_CONFIG)
3482 reply_cmd = IPVS_CMD_SET_CONFIG;
3484 pr_err("unknown Generic Netlink command\n");
3488 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3492 mutex_lock(&__ip_vs_mutex);
3494 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3496 goto nla_put_failure;
3499 case IPVS_CMD_GET_SERVICE:
3501 struct ip_vs_service *svc;
3503 svc = ip_vs_genl_find_service(net,
3504 info->attrs[IPVS_CMD_ATTR_SERVICE]);
3509 ret = ip_vs_genl_fill_service(msg, svc);
3511 goto nla_put_failure;
3520 case IPVS_CMD_GET_CONFIG:
3522 struct ip_vs_timeout_user t;
3524 __ip_vs_get_timeouts(net, &t);
3525 #ifdef CONFIG_IP_VS_PROTO_TCP
3526 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3528 nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3530 goto nla_put_failure;
3532 #ifdef CONFIG_IP_VS_PROTO_UDP
3533 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3534 goto nla_put_failure;
3540 case IPVS_CMD_GET_INFO:
3541 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3542 IP_VS_VERSION_CODE) ||
3543 nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3544 ip_vs_conn_tab_size))
3545 goto nla_put_failure;
3549 genlmsg_end(msg, reply);
3550 ret = genlmsg_reply(msg, info);
3554 pr_err("not enough space in Netlink message\n");
3560 mutex_unlock(&__ip_vs_mutex);
3566 static struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3568 .cmd = IPVS_CMD_NEW_SERVICE,
3569 .flags = GENL_ADMIN_PERM,
3570 .policy = ip_vs_cmd_policy,
3571 .doit = ip_vs_genl_set_cmd,
3574 .cmd = IPVS_CMD_SET_SERVICE,
3575 .flags = GENL_ADMIN_PERM,
3576 .policy = ip_vs_cmd_policy,
3577 .doit = ip_vs_genl_set_cmd,
3580 .cmd = IPVS_CMD_DEL_SERVICE,
3581 .flags = GENL_ADMIN_PERM,
3582 .policy = ip_vs_cmd_policy,
3583 .doit = ip_vs_genl_set_cmd,
3586 .cmd = IPVS_CMD_GET_SERVICE,
3587 .flags = GENL_ADMIN_PERM,
3588 .doit = ip_vs_genl_get_cmd,
3589 .dumpit = ip_vs_genl_dump_services,
3590 .policy = ip_vs_cmd_policy,
3593 .cmd = IPVS_CMD_NEW_DEST,
3594 .flags = GENL_ADMIN_PERM,
3595 .policy = ip_vs_cmd_policy,
3596 .doit = ip_vs_genl_set_cmd,
3599 .cmd = IPVS_CMD_SET_DEST,
3600 .flags = GENL_ADMIN_PERM,
3601 .policy = ip_vs_cmd_policy,
3602 .doit = ip_vs_genl_set_cmd,
3605 .cmd = IPVS_CMD_DEL_DEST,
3606 .flags = GENL_ADMIN_PERM,
3607 .policy = ip_vs_cmd_policy,
3608 .doit = ip_vs_genl_set_cmd,
3611 .cmd = IPVS_CMD_GET_DEST,
3612 .flags = GENL_ADMIN_PERM,
3613 .policy = ip_vs_cmd_policy,
3614 .dumpit = ip_vs_genl_dump_dests,
3617 .cmd = IPVS_CMD_NEW_DAEMON,
3618 .flags = GENL_ADMIN_PERM,
3619 .policy = ip_vs_cmd_policy,
3620 .doit = ip_vs_genl_set_daemon,
3623 .cmd = IPVS_CMD_DEL_DAEMON,
3624 .flags = GENL_ADMIN_PERM,
3625 .policy = ip_vs_cmd_policy,
3626 .doit = ip_vs_genl_set_daemon,
3629 .cmd = IPVS_CMD_GET_DAEMON,
3630 .flags = GENL_ADMIN_PERM,
3631 .dumpit = ip_vs_genl_dump_daemons,
3634 .cmd = IPVS_CMD_SET_CONFIG,
3635 .flags = GENL_ADMIN_PERM,
3636 .policy = ip_vs_cmd_policy,
3637 .doit = ip_vs_genl_set_cmd,
3640 .cmd = IPVS_CMD_GET_CONFIG,
3641 .flags = GENL_ADMIN_PERM,
3642 .doit = ip_vs_genl_get_cmd,
3645 .cmd = IPVS_CMD_GET_INFO,
3646 .flags = GENL_ADMIN_PERM,
3647 .doit = ip_vs_genl_get_cmd,
3650 .cmd = IPVS_CMD_ZERO,
3651 .flags = GENL_ADMIN_PERM,
3652 .policy = ip_vs_cmd_policy,
3653 .doit = ip_vs_genl_set_cmd,
3656 .cmd = IPVS_CMD_FLUSH,
3657 .flags = GENL_ADMIN_PERM,
3658 .doit = ip_vs_genl_set_cmd,
3662 static int __init ip_vs_genl_register(void)
3664 return genl_register_family_with_ops(&ip_vs_genl_family,
3665 ip_vs_genl_ops, ARRAY_SIZE(ip_vs_genl_ops));
3668 static void ip_vs_genl_unregister(void)
3670 genl_unregister_family(&ip_vs_genl_family);
3673 /* End of Generic Netlink interface definitions */
3676 * per netns intit/exit func.
3678 #ifdef CONFIG_SYSCTL
3679 int __net_init ip_vs_control_net_init_sysctl(struct net *net)
3682 struct netns_ipvs *ipvs = net_ipvs(net);
3683 struct ctl_table *tbl;
3685 atomic_set(&ipvs->dropentry, 0);
3686 spin_lock_init(&ipvs->dropentry_lock);
3687 spin_lock_init(&ipvs->droppacket_lock);
3688 spin_lock_init(&ipvs->securetcp_lock);
3690 if (!net_eq(net, &init_net)) {
3691 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3696 /* Initialize sysctl defaults */
3698 ipvs->sysctl_amemthresh = 1024;
3699 tbl[idx++].data = &ipvs->sysctl_amemthresh;
3700 ipvs->sysctl_am_droprate = 10;
3701 tbl[idx++].data = &ipvs->sysctl_am_droprate;
3702 tbl[idx++].data = &ipvs->sysctl_drop_entry;
3703 tbl[idx++].data = &ipvs->sysctl_drop_packet;
3704 #ifdef CONFIG_IP_VS_NFCT
3705 tbl[idx++].data = &ipvs->sysctl_conntrack;
3707 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3708 ipvs->sysctl_snat_reroute = 1;
3709 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3710 ipvs->sysctl_sync_ver = 1;
3711 tbl[idx++].data = &ipvs->sysctl_sync_ver;
3712 ipvs->sysctl_sync_ports = 1;
3713 tbl[idx++].data = &ipvs->sysctl_sync_ports;
3714 ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3715 tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3716 ipvs->sysctl_sync_sock_size = 0;
3717 tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3718 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3719 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3720 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3721 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3722 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3723 tbl[idx].data = &ipvs->sysctl_sync_threshold;
3724 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3725 ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3726 tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3727 ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3728 tbl[idx++].data = &ipvs->sysctl_sync_retries;
3729 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3732 ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3733 if (ipvs->sysctl_hdr == NULL) {
3734 if (!net_eq(net, &init_net))
3738 ip_vs_start_estimator(net, &ipvs->tot_stats);
3739 ipvs->sysctl_tbl = tbl;
3740 /* Schedule defense work */
3741 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3742 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3747 void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net)
3749 struct netns_ipvs *ipvs = net_ipvs(net);
3751 cancel_delayed_work_sync(&ipvs->defense_work);
3752 cancel_work_sync(&ipvs->defense_work.work);
3753 unregister_net_sysctl_table(ipvs->sysctl_hdr);
3758 int __net_init ip_vs_control_net_init_sysctl(struct net *net) { return 0; }
3759 void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net) { }
3763 static struct notifier_block ip_vs_dst_notifier = {
3764 .notifier_call = ip_vs_dst_event,
3767 int __net_init ip_vs_control_net_init(struct net *net)
3770 struct netns_ipvs *ipvs = net_ipvs(net);
3772 rwlock_init(&ipvs->rs_lock);
3774 /* Initialize rs_table */
3775 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
3776 INIT_LIST_HEAD(&ipvs->rs_table[idx]);
3778 INIT_LIST_HEAD(&ipvs->dest_trash);
3779 atomic_set(&ipvs->ftpsvc_counter, 0);
3780 atomic_set(&ipvs->nullsvc_counter, 0);
3783 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
3784 if (!ipvs->tot_stats.cpustats)
3787 spin_lock_init(&ipvs->tot_stats.lock);
3789 proc_net_fops_create(net, "ip_vs", 0, &ip_vs_info_fops);
3790 proc_net_fops_create(net, "ip_vs_stats", 0, &ip_vs_stats_fops);
3791 proc_net_fops_create(net, "ip_vs_stats_percpu", 0,
3792 &ip_vs_stats_percpu_fops);
3794 if (ip_vs_control_net_init_sysctl(net))
3800 free_percpu(ipvs->tot_stats.cpustats);
3804 void __net_exit ip_vs_control_net_cleanup(struct net *net)
3806 struct netns_ipvs *ipvs = net_ipvs(net);
3808 ip_vs_trash_cleanup(net);
3809 ip_vs_stop_estimator(net, &ipvs->tot_stats);
3810 ip_vs_control_net_cleanup_sysctl(net);
3811 proc_net_remove(net, "ip_vs_stats_percpu");
3812 proc_net_remove(net, "ip_vs_stats");
3813 proc_net_remove(net, "ip_vs");
3814 free_percpu(ipvs->tot_stats.cpustats);
3817 int __init ip_vs_register_nl_ioctl(void)
3821 ret = nf_register_sockopt(&ip_vs_sockopts);
3823 pr_err("cannot register sockopt.\n");
3827 ret = ip_vs_genl_register();
3829 pr_err("cannot register Generic Netlink interface.\n");
3835 nf_unregister_sockopt(&ip_vs_sockopts);
3840 void ip_vs_unregister_nl_ioctl(void)
3842 ip_vs_genl_unregister();
3843 nf_unregister_sockopt(&ip_vs_sockopts);
3846 int __init ip_vs_control_init(void)
3853 /* Initialize svc_table, ip_vs_svc_fwm_table, rs_table */
3854 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
3855 INIT_LIST_HEAD(&ip_vs_svc_table[idx]);
3856 INIT_LIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3859 smp_wmb(); /* Do we really need it now ? */
3861 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
3870 void ip_vs_control_cleanup(void)
3873 unregister_netdevice_notifier(&ip_vs_dst_notifier);