1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * IPVS An implementation of the IP virtual server support for the
4 * LINUX operating system. IPVS is now implemented as a module
5 * over the NetFilter framework. IPVS can be used to build a
6 * high-performance and highly available server based on a
9 * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
10 * Peter Kese <peter.kese@ijs.si>
11 * Julian Anastasov <ja@ssi.bg>
16 #define KMSG_COMPONENT "IPVS"
17 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
19 #include <linux/module.h>
20 #include <linux/init.h>
21 #include <linux/types.h>
22 #include <linux/capability.h>
24 #include <linux/sysctl.h>
25 #include <linux/proc_fs.h>
26 #include <linux/workqueue.h>
27 #include <linux/swap.h>
28 #include <linux/seq_file.h>
29 #include <linux/slab.h>
31 #include <linux/netfilter.h>
32 #include <linux/netfilter_ipv4.h>
33 #include <linux/mutex.h>
35 #include <net/net_namespace.h>
36 #include <linux/nsproxy.h>
38 #ifdef CONFIG_IP_VS_IPV6
40 #include <net/ip6_route.h>
41 #include <net/netfilter/ipv6/nf_defrag_ipv6.h>
43 #include <net/route.h>
45 #include <net/genetlink.h>
47 #include <linux/uaccess.h>
49 #include <net/ip_vs.h>
51 /* semaphore for IPVS sockopts. And, [gs]etsockopt may sleep. */
52 static DEFINE_MUTEX(__ip_vs_mutex);
54 /* sysctl variables */
56 #ifdef CONFIG_IP_VS_DEBUG
57 static int sysctl_ip_vs_debug_level = 0;
59 int ip_vs_get_debug_level(void)
61 return sysctl_ip_vs_debug_level;
67 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup);
70 #ifdef CONFIG_IP_VS_IPV6
71 /* Taken from rt6_fill_node() in net/ipv6/route.c, is there a better way? */
72 static bool __ip_vs_addr_is_local_v6(struct net *net,
73 const struct in6_addr *addr)
78 struct dst_entry *dst = ip6_route_output(net, NULL, &fl6);
81 is_local = !dst->error && dst->dev && (dst->dev->flags & IFF_LOOPBACK);
90 * update_defense_level is called from keventd and from sysctl,
91 * so it needs to protect itself from softirqs
93 static void update_defense_level(struct netns_ipvs *ipvs)
100 /* we only count free and buffered memory (in pages) */
102 availmem = i.freeram + i.bufferram;
103 /* however in linux 2.5 the i.bufferram is total page cache size,
105 /* si_swapinfo(&i); */
106 /* availmem = availmem - (i.totalswap - i.freeswap); */
108 nomem = (availmem < ipvs->sysctl_amemthresh);
113 spin_lock(&ipvs->dropentry_lock);
114 switch (ipvs->sysctl_drop_entry) {
116 atomic_set(&ipvs->dropentry, 0);
120 atomic_set(&ipvs->dropentry, 1);
121 ipvs->sysctl_drop_entry = 2;
123 atomic_set(&ipvs->dropentry, 0);
128 atomic_set(&ipvs->dropentry, 1);
130 atomic_set(&ipvs->dropentry, 0);
131 ipvs->sysctl_drop_entry = 1;
135 atomic_set(&ipvs->dropentry, 1);
138 spin_unlock(&ipvs->dropentry_lock);
141 spin_lock(&ipvs->droppacket_lock);
142 switch (ipvs->sysctl_drop_packet) {
148 ipvs->drop_rate = ipvs->drop_counter
149 = ipvs->sysctl_amemthresh /
150 (ipvs->sysctl_amemthresh-availmem);
151 ipvs->sysctl_drop_packet = 2;
158 ipvs->drop_rate = ipvs->drop_counter
159 = ipvs->sysctl_amemthresh /
160 (ipvs->sysctl_amemthresh-availmem);
163 ipvs->sysctl_drop_packet = 1;
167 ipvs->drop_rate = ipvs->sysctl_am_droprate;
170 spin_unlock(&ipvs->droppacket_lock);
173 spin_lock(&ipvs->securetcp_lock);
174 switch (ipvs->sysctl_secure_tcp) {
176 if (ipvs->old_secure_tcp >= 2)
181 if (ipvs->old_secure_tcp < 2)
183 ipvs->sysctl_secure_tcp = 2;
185 if (ipvs->old_secure_tcp >= 2)
191 if (ipvs->old_secure_tcp < 2)
194 if (ipvs->old_secure_tcp >= 2)
196 ipvs->sysctl_secure_tcp = 1;
200 if (ipvs->old_secure_tcp < 2)
204 ipvs->old_secure_tcp = ipvs->sysctl_secure_tcp;
206 ip_vs_protocol_timeout_change(ipvs,
207 ipvs->sysctl_secure_tcp > 1);
208 spin_unlock(&ipvs->securetcp_lock);
215 * Timer for checking the defense
217 #define DEFENSE_TIMER_PERIOD 1*HZ
219 static void defense_work_handler(struct work_struct *work)
221 struct netns_ipvs *ipvs =
222 container_of(work, struct netns_ipvs, defense_work.work);
224 update_defense_level(ipvs);
225 if (atomic_read(&ipvs->dropentry))
226 ip_vs_random_dropentry(ipvs);
227 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
232 ip_vs_use_count_inc(void)
234 return try_module_get(THIS_MODULE);
238 ip_vs_use_count_dec(void)
240 module_put(THIS_MODULE);
245 * Hash table: for virtual service lookups
247 #define IP_VS_SVC_TAB_BITS 8
248 #define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
249 #define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)
251 /* the service table hashed by <protocol, addr, port> */
252 static struct hlist_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE];
253 /* the service table hashed by fwmark */
254 static struct hlist_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE];
258 * Returns hash value for virtual service
260 static inline unsigned int
261 ip_vs_svc_hashkey(struct netns_ipvs *ipvs, int af, unsigned int proto,
262 const union nf_inet_addr *addr, __be16 port)
264 register unsigned int porth = ntohs(port);
265 __be32 addr_fold = addr->ip;
268 #ifdef CONFIG_IP_VS_IPV6
270 addr_fold = addr->ip6[0]^addr->ip6[1]^
271 addr->ip6[2]^addr->ip6[3];
273 ahash = ntohl(addr_fold);
274 ahash ^= ((size_t) ipvs >> 8);
276 return (proto ^ ahash ^ (porth >> IP_VS_SVC_TAB_BITS) ^ porth) &
281 * Returns hash value of fwmark for virtual service lookup
283 static inline unsigned int ip_vs_svc_fwm_hashkey(struct netns_ipvs *ipvs, __u32 fwmark)
285 return (((size_t)ipvs>>8) ^ fwmark) & IP_VS_SVC_TAB_MASK;
289 * Hashes a service in the ip_vs_svc_table by <netns,proto,addr,port>
290 * or in the ip_vs_svc_fwm_table by fwmark.
291 * Should be called with locked tables.
293 static int ip_vs_svc_hash(struct ip_vs_service *svc)
297 if (svc->flags & IP_VS_SVC_F_HASHED) {
298 pr_err("%s(): request for already hashed, called from %pS\n",
299 __func__, __builtin_return_address(0));
303 if (svc->fwmark == 0) {
305 * Hash it by <netns,protocol,addr,port> in ip_vs_svc_table
307 hash = ip_vs_svc_hashkey(svc->ipvs, svc->af, svc->protocol,
308 &svc->addr, svc->port);
309 hlist_add_head_rcu(&svc->s_list, &ip_vs_svc_table[hash]);
312 * Hash it by fwmark in svc_fwm_table
314 hash = ip_vs_svc_fwm_hashkey(svc->ipvs, svc->fwmark);
315 hlist_add_head_rcu(&svc->f_list, &ip_vs_svc_fwm_table[hash]);
318 svc->flags |= IP_VS_SVC_F_HASHED;
319 /* increase its refcnt because it is referenced by the svc table */
320 atomic_inc(&svc->refcnt);
326 * Unhashes a service from svc_table / svc_fwm_table.
327 * Should be called with locked tables.
329 static int ip_vs_svc_unhash(struct ip_vs_service *svc)
331 if (!(svc->flags & IP_VS_SVC_F_HASHED)) {
332 pr_err("%s(): request for unhash flagged, called from %pS\n",
333 __func__, __builtin_return_address(0));
337 if (svc->fwmark == 0) {
338 /* Remove it from the svc_table table */
339 hlist_del_rcu(&svc->s_list);
341 /* Remove it from the svc_fwm_table table */
342 hlist_del_rcu(&svc->f_list);
345 svc->flags &= ~IP_VS_SVC_F_HASHED;
346 atomic_dec(&svc->refcnt);
352 * Get service by {netns, proto,addr,port} in the service table.
354 static inline struct ip_vs_service *
355 __ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u16 protocol,
356 const union nf_inet_addr *vaddr, __be16 vport)
359 struct ip_vs_service *svc;
361 /* Check for "full" addressed entries */
362 hash = ip_vs_svc_hashkey(ipvs, af, protocol, vaddr, vport);
364 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[hash], s_list) {
366 && ip_vs_addr_equal(af, &svc->addr, vaddr)
367 && (svc->port == vport)
368 && (svc->protocol == protocol)
369 && (svc->ipvs == ipvs)) {
380 * Get service by {fwmark} in the service table.
382 static inline struct ip_vs_service *
383 __ip_vs_svc_fwm_find(struct netns_ipvs *ipvs, int af, __u32 fwmark)
386 struct ip_vs_service *svc;
388 /* Check for fwmark addressed entries */
389 hash = ip_vs_svc_fwm_hashkey(ipvs, fwmark);
391 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[hash], f_list) {
392 if (svc->fwmark == fwmark && svc->af == af
393 && (svc->ipvs == ipvs)) {
402 /* Find service, called under RCU lock */
403 struct ip_vs_service *
404 ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
405 const union nf_inet_addr *vaddr, __be16 vport)
407 struct ip_vs_service *svc;
410 * Check the table hashed by fwmark first
413 svc = __ip_vs_svc_fwm_find(ipvs, af, fwmark);
419 * Check the table hashed by <protocol,addr,port>
420 * for "full" addressed entries
422 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, vport);
424 if (!svc && protocol == IPPROTO_TCP &&
425 atomic_read(&ipvs->ftpsvc_counter) &&
426 (vport == FTPDATA || ntohs(vport) >= inet_prot_sock(ipvs->net))) {
428 * Check if ftp service entry exists, the packet
429 * might belong to FTP data connections.
431 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, FTPPORT);
435 && atomic_read(&ipvs->nullsvc_counter)) {
437 * Check if the catch-all port (port zero) exists
439 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, 0);
443 IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
444 fwmark, ip_vs_proto_name(protocol),
445 IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
446 svc ? "hit" : "not hit");
453 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
455 atomic_inc(&svc->refcnt);
456 rcu_assign_pointer(dest->svc, svc);
459 static void ip_vs_service_free(struct ip_vs_service *svc)
461 free_percpu(svc->stats.cpustats);
465 static void ip_vs_service_rcu_free(struct rcu_head *head)
467 struct ip_vs_service *svc;
469 svc = container_of(head, struct ip_vs_service, rcu_head);
470 ip_vs_service_free(svc);
473 static void __ip_vs_svc_put(struct ip_vs_service *svc, bool do_delay)
475 if (atomic_dec_and_test(&svc->refcnt)) {
476 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n",
478 IP_VS_DBG_ADDR(svc->af, &svc->addr),
481 call_rcu(&svc->rcu_head, ip_vs_service_rcu_free);
483 ip_vs_service_free(svc);
489 * Returns hash value for real service
491 static inline unsigned int ip_vs_rs_hashkey(int af,
492 const union nf_inet_addr *addr,
495 register unsigned int porth = ntohs(port);
496 __be32 addr_fold = addr->ip;
498 #ifdef CONFIG_IP_VS_IPV6
500 addr_fold = addr->ip6[0]^addr->ip6[1]^
501 addr->ip6[2]^addr->ip6[3];
504 return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
508 /* Hash ip_vs_dest in rs_table by <proto,addr,port>. */
509 static void ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
514 if (dest->in_rs_table)
517 switch (IP_VS_DFWD_METHOD(dest)) {
518 case IP_VS_CONN_F_MASQ:
521 case IP_VS_CONN_F_TUNNEL:
522 switch (dest->tun_type) {
523 case IP_VS_CONN_F_TUNNEL_TYPE_GUE:
524 port = dest->tun_port;
526 case IP_VS_CONN_F_TUNNEL_TYPE_IPIP:
527 case IP_VS_CONN_F_TUNNEL_TYPE_GRE:
539 * Hash by proto,addr,port,
540 * which are the parameters of the real service.
542 hash = ip_vs_rs_hashkey(dest->af, &dest->addr, port);
544 hlist_add_head_rcu(&dest->d_list, &ipvs->rs_table[hash]);
545 dest->in_rs_table = 1;
548 /* Unhash ip_vs_dest from rs_table. */
549 static void ip_vs_rs_unhash(struct ip_vs_dest *dest)
552 * Remove it from the rs_table table.
554 if (dest->in_rs_table) {
555 hlist_del_rcu(&dest->d_list);
556 dest->in_rs_table = 0;
560 /* Check if real service by <proto,addr,port> is present */
561 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
562 const union nf_inet_addr *daddr, __be16 dport)
565 struct ip_vs_dest *dest;
567 /* Check for "full" addressed entries */
568 hash = ip_vs_rs_hashkey(af, daddr, dport);
570 hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
571 if (dest->port == dport &&
573 ip_vs_addr_equal(af, &dest->addr, daddr) &&
574 (dest->protocol == protocol || dest->vfwmark) &&
575 IP_VS_DFWD_METHOD(dest) == IP_VS_CONN_F_MASQ) {
584 /* Find real service record by <proto,addr,port>.
585 * In case of multiple records with the same <proto,addr,port>, only
586 * the first found record is returned.
588 * To be called under RCU lock.
590 struct ip_vs_dest *ip_vs_find_real_service(struct netns_ipvs *ipvs, int af,
592 const union nf_inet_addr *daddr,
596 struct ip_vs_dest *dest;
598 /* Check for "full" addressed entries */
599 hash = ip_vs_rs_hashkey(af, daddr, dport);
601 hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
602 if (dest->port == dport &&
604 ip_vs_addr_equal(af, &dest->addr, daddr) &&
605 (dest->protocol == protocol || dest->vfwmark) &&
606 IP_VS_DFWD_METHOD(dest) == IP_VS_CONN_F_MASQ) {
615 /* Find real service record by <af,addr,tun_port>.
616 * In case of multiple records with the same <af,addr,tun_port>, only
617 * the first found record is returned.
619 * To be called under RCU lock.
621 struct ip_vs_dest *ip_vs_find_tunnel(struct netns_ipvs *ipvs, int af,
622 const union nf_inet_addr *daddr,
625 struct ip_vs_dest *dest;
628 /* Check for "full" addressed entries */
629 hash = ip_vs_rs_hashkey(af, daddr, tun_port);
631 hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
632 if (dest->tun_port == tun_port &&
634 ip_vs_addr_equal(af, &dest->addr, daddr) &&
635 IP_VS_DFWD_METHOD(dest) == IP_VS_CONN_F_TUNNEL) {
644 /* Lookup destination by {addr,port} in the given service
645 * Called under RCU lock.
647 static struct ip_vs_dest *
648 ip_vs_lookup_dest(struct ip_vs_service *svc, int dest_af,
649 const union nf_inet_addr *daddr, __be16 dport)
651 struct ip_vs_dest *dest;
654 * Find the destination for the given service
656 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
657 if ((dest->af == dest_af) &&
658 ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
659 (dest->port == dport)) {
669 * Find destination by {daddr,dport,vaddr,protocol}
670 * Created to be used in ip_vs_process_message() in
671 * the backup synchronization daemon. It finds the
672 * destination to be bound to the received connection
674 * Called under RCU lock, no refcnt is returned.
676 struct ip_vs_dest *ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
677 const union nf_inet_addr *daddr,
679 const union nf_inet_addr *vaddr,
680 __be16 vport, __u16 protocol, __u32 fwmark,
683 struct ip_vs_dest *dest;
684 struct ip_vs_service *svc;
687 svc = ip_vs_service_find(ipvs, svc_af, fwmark, protocol, vaddr, vport);
690 if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
692 dest = ip_vs_lookup_dest(svc, dest_af, daddr, port);
694 dest = ip_vs_lookup_dest(svc, dest_af, daddr, port ^ dport);
698 void ip_vs_dest_dst_rcu_free(struct rcu_head *head)
700 struct ip_vs_dest_dst *dest_dst = container_of(head,
701 struct ip_vs_dest_dst,
704 dst_release(dest_dst->dst_cache);
708 /* Release dest_dst and dst_cache for dest in user context */
709 static void __ip_vs_dst_cache_reset(struct ip_vs_dest *dest)
711 struct ip_vs_dest_dst *old;
713 old = rcu_dereference_protected(dest->dest_dst, 1);
715 RCU_INIT_POINTER(dest->dest_dst, NULL);
716 call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
721 * Lookup dest by {svc,addr,port} in the destination trash.
722 * The destination trash is used to hold the destinations that are removed
723 * from the service table but are still referenced by some conn entries.
724 * The reason to add the destination trash is when the dest is temporary
725 * down (either by administrator or by monitor program), the dest can be
726 * picked back from the trash, the remaining connections to the dest can
727 * continue, and the counting information of the dest is also useful for
730 static struct ip_vs_dest *
731 ip_vs_trash_get_dest(struct ip_vs_service *svc, int dest_af,
732 const union nf_inet_addr *daddr, __be16 dport)
734 struct ip_vs_dest *dest;
735 struct netns_ipvs *ipvs = svc->ipvs;
738 * Find the destination in trash
740 spin_lock_bh(&ipvs->dest_trash_lock);
741 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
742 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
745 IP_VS_DBG_ADDR(dest->af, &dest->addr),
747 refcount_read(&dest->refcnt));
748 if (dest->af == dest_af &&
749 ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
750 dest->port == dport &&
751 dest->vfwmark == svc->fwmark &&
752 dest->protocol == svc->protocol &&
754 (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
755 dest->vport == svc->port))) {
757 list_del(&dest->t_list);
765 spin_unlock_bh(&ipvs->dest_trash_lock);
770 static void ip_vs_dest_free(struct ip_vs_dest *dest)
772 struct ip_vs_service *svc = rcu_dereference_protected(dest->svc, 1);
774 __ip_vs_dst_cache_reset(dest);
775 __ip_vs_svc_put(svc, false);
776 free_percpu(dest->stats.cpustats);
777 ip_vs_dest_put_and_free(dest);
781 * Clean up all the destinations in the trash
782 * Called by the ip_vs_control_cleanup()
784 * When the ip_vs_control_clearup is activated by ipvs module exit,
785 * the service tables must have been flushed and all the connections
786 * are expired, and the refcnt of each destination in the trash must
787 * be 1, so we simply release them here.
789 static void ip_vs_trash_cleanup(struct netns_ipvs *ipvs)
791 struct ip_vs_dest *dest, *nxt;
793 del_timer_sync(&ipvs->dest_trash_timer);
794 /* No need to use dest_trash_lock */
795 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, t_list) {
796 list_del(&dest->t_list);
797 ip_vs_dest_free(dest);
802 ip_vs_copy_stats(struct ip_vs_kstats *dst, struct ip_vs_stats *src)
804 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->kstats.c - src->kstats0.c
806 spin_lock_bh(&src->lock);
808 IP_VS_SHOW_STATS_COUNTER(conns);
809 IP_VS_SHOW_STATS_COUNTER(inpkts);
810 IP_VS_SHOW_STATS_COUNTER(outpkts);
811 IP_VS_SHOW_STATS_COUNTER(inbytes);
812 IP_VS_SHOW_STATS_COUNTER(outbytes);
814 ip_vs_read_estimator(dst, src);
816 spin_unlock_bh(&src->lock);
820 ip_vs_export_stats_user(struct ip_vs_stats_user *dst, struct ip_vs_kstats *src)
822 dst->conns = (u32)src->conns;
823 dst->inpkts = (u32)src->inpkts;
824 dst->outpkts = (u32)src->outpkts;
825 dst->inbytes = src->inbytes;
826 dst->outbytes = src->outbytes;
827 dst->cps = (u32)src->cps;
828 dst->inpps = (u32)src->inpps;
829 dst->outpps = (u32)src->outpps;
830 dst->inbps = (u32)src->inbps;
831 dst->outbps = (u32)src->outbps;
835 ip_vs_zero_stats(struct ip_vs_stats *stats)
837 spin_lock_bh(&stats->lock);
839 /* get current counters as zero point, rates are zeroed */
841 #define IP_VS_ZERO_STATS_COUNTER(c) stats->kstats0.c = stats->kstats.c
843 IP_VS_ZERO_STATS_COUNTER(conns);
844 IP_VS_ZERO_STATS_COUNTER(inpkts);
845 IP_VS_ZERO_STATS_COUNTER(outpkts);
846 IP_VS_ZERO_STATS_COUNTER(inbytes);
847 IP_VS_ZERO_STATS_COUNTER(outbytes);
849 ip_vs_zero_estimator(stats);
851 spin_unlock_bh(&stats->lock);
855 * Update a destination in the given service
858 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
859 struct ip_vs_dest_user_kern *udest, int add)
861 struct netns_ipvs *ipvs = svc->ipvs;
862 struct ip_vs_service *old_svc;
863 struct ip_vs_scheduler *sched;
866 /* We cannot modify an address and change the address family */
867 BUG_ON(!add && udest->af != dest->af);
869 if (add && udest->af != svc->af)
870 ipvs->mixed_address_family_dests++;
872 /* keep the last_weight with latest non-0 weight */
873 if (add || udest->weight != 0)
874 atomic_set(&dest->last_weight, udest->weight);
876 /* set the weight and the flags */
877 atomic_set(&dest->weight, udest->weight);
878 conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
879 conn_flags |= IP_VS_CONN_F_INACTIVE;
881 /* Need to rehash? */
882 if ((udest->conn_flags & IP_VS_CONN_F_FWD_MASK) !=
883 IP_VS_DFWD_METHOD(dest) ||
884 udest->tun_type != dest->tun_type ||
885 udest->tun_port != dest->tun_port)
886 ip_vs_rs_unhash(dest);
888 /* set the tunnel info */
889 dest->tun_type = udest->tun_type;
890 dest->tun_port = udest->tun_port;
891 dest->tun_flags = udest->tun_flags;
893 /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
894 if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
895 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
897 /* FTP-NAT requires conntrack for mangling */
898 if (svc->port == FTPPORT)
899 ip_vs_register_conntrack(svc);
901 atomic_set(&dest->conn_flags, conn_flags);
902 /* Put the real service in rs_table if not present. */
903 ip_vs_rs_hash(ipvs, dest);
905 /* bind the service */
906 old_svc = rcu_dereference_protected(dest->svc, 1);
908 __ip_vs_bind_svc(dest, svc);
910 if (old_svc != svc) {
911 ip_vs_zero_stats(&dest->stats);
912 __ip_vs_bind_svc(dest, svc);
913 __ip_vs_svc_put(old_svc, true);
917 /* set the dest status flags */
918 dest->flags |= IP_VS_DEST_F_AVAILABLE;
920 if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
921 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
922 dest->u_threshold = udest->u_threshold;
923 dest->l_threshold = udest->l_threshold;
925 dest->af = udest->af;
927 spin_lock_bh(&dest->dst_lock);
928 __ip_vs_dst_cache_reset(dest);
929 spin_unlock_bh(&dest->dst_lock);
932 ip_vs_start_estimator(svc->ipvs, &dest->stats);
933 list_add_rcu(&dest->n_list, &svc->destinations);
935 sched = rcu_dereference_protected(svc->scheduler, 1);
936 if (sched && sched->add_dest)
937 sched->add_dest(svc, dest);
939 sched = rcu_dereference_protected(svc->scheduler, 1);
940 if (sched && sched->upd_dest)
941 sched->upd_dest(svc, dest);
947 * Create a destination for the given service
950 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
951 struct ip_vs_dest **dest_p)
953 struct ip_vs_dest *dest;
954 unsigned int atype, i;
958 #ifdef CONFIG_IP_VS_IPV6
959 if (udest->af == AF_INET6) {
962 atype = ipv6_addr_type(&udest->addr.in6);
963 if ((!(atype & IPV6_ADDR_UNICAST) ||
964 atype & IPV6_ADDR_LINKLOCAL) &&
965 !__ip_vs_addr_is_local_v6(svc->ipvs->net, &udest->addr.in6))
968 ret = nf_defrag_ipv6_enable(svc->ipvs->net);
974 atype = inet_addr_type(svc->ipvs->net, udest->addr.ip);
975 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
979 dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
983 dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
984 if (!dest->stats.cpustats)
987 for_each_possible_cpu(i) {
988 struct ip_vs_cpu_stats *ip_vs_dest_stats;
989 ip_vs_dest_stats = per_cpu_ptr(dest->stats.cpustats, i);
990 u64_stats_init(&ip_vs_dest_stats->syncp);
993 dest->af = udest->af;
994 dest->protocol = svc->protocol;
995 dest->vaddr = svc->addr;
996 dest->vport = svc->port;
997 dest->vfwmark = svc->fwmark;
998 ip_vs_addr_copy(udest->af, &dest->addr, &udest->addr);
999 dest->port = udest->port;
1001 atomic_set(&dest->activeconns, 0);
1002 atomic_set(&dest->inactconns, 0);
1003 atomic_set(&dest->persistconns, 0);
1004 refcount_set(&dest->refcnt, 1);
1006 INIT_HLIST_NODE(&dest->d_list);
1007 spin_lock_init(&dest->dst_lock);
1008 spin_lock_init(&dest->stats.lock);
1009 __ip_vs_update_dest(svc, dest, udest, 1);
1023 * Add a destination into an existing service
1026 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1028 struct ip_vs_dest *dest;
1029 union nf_inet_addr daddr;
1030 __be16 dport = udest->port;
1035 if (udest->weight < 0) {
1036 pr_err("%s(): server weight less than zero\n", __func__);
1040 if (udest->l_threshold > udest->u_threshold) {
1041 pr_err("%s(): lower threshold is higher than upper threshold\n",
1046 if (udest->tun_type == IP_VS_CONN_F_TUNNEL_TYPE_GUE) {
1047 if (udest->tun_port == 0) {
1048 pr_err("%s(): tunnel port is zero\n", __func__);
1053 ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
1055 /* We use function that requires RCU lock */
1057 dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
1061 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
1066 * Check if the dest already exists in the trash and
1067 * is from the same service
1069 dest = ip_vs_trash_get_dest(svc, udest->af, &daddr, dport);
1072 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
1073 "dest->refcnt=%d, service %u/%s:%u\n",
1074 IP_VS_DBG_ADDR(udest->af, &daddr), ntohs(dport),
1075 refcount_read(&dest->refcnt),
1077 IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
1078 ntohs(dest->vport));
1080 __ip_vs_update_dest(svc, dest, udest, 1);
1084 * Allocate and initialize the dest structure
1086 ret = ip_vs_new_dest(svc, udest, &dest);
1095 * Edit a destination in the given service
1098 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1100 struct ip_vs_dest *dest;
1101 union nf_inet_addr daddr;
1102 __be16 dport = udest->port;
1106 if (udest->weight < 0) {
1107 pr_err("%s(): server weight less than zero\n", __func__);
1111 if (udest->l_threshold > udest->u_threshold) {
1112 pr_err("%s(): lower threshold is higher than upper threshold\n",
1117 if (udest->tun_type == IP_VS_CONN_F_TUNNEL_TYPE_GUE) {
1118 if (udest->tun_port == 0) {
1119 pr_err("%s(): tunnel port is zero\n", __func__);
1124 ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
1126 /* We use function that requires RCU lock */
1128 dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
1132 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1136 __ip_vs_update_dest(svc, dest, udest, 0);
1143 * Delete a destination (must be already unlinked from the service)
1145 static void __ip_vs_del_dest(struct netns_ipvs *ipvs, struct ip_vs_dest *dest,
1148 ip_vs_stop_estimator(ipvs, &dest->stats);
1151 * Remove it from the d-linked list with the real services.
1153 ip_vs_rs_unhash(dest);
1155 spin_lock_bh(&ipvs->dest_trash_lock);
1156 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1157 IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port),
1158 refcount_read(&dest->refcnt));
1159 if (list_empty(&ipvs->dest_trash) && !cleanup)
1160 mod_timer(&ipvs->dest_trash_timer,
1161 jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1162 /* dest lives in trash with reference */
1163 list_add(&dest->t_list, &ipvs->dest_trash);
1164 dest->idle_start = 0;
1165 spin_unlock_bh(&ipvs->dest_trash_lock);
1170 * Unlink a destination from the given service
1172 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1173 struct ip_vs_dest *dest,
1176 dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1179 * Remove it from the d-linked destination list.
1181 list_del_rcu(&dest->n_list);
1184 if (dest->af != svc->af)
1185 svc->ipvs->mixed_address_family_dests--;
1188 struct ip_vs_scheduler *sched;
1190 sched = rcu_dereference_protected(svc->scheduler, 1);
1191 if (sched && sched->del_dest)
1192 sched->del_dest(svc, dest);
1198 * Delete a destination server in the given service
1201 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1203 struct ip_vs_dest *dest;
1204 __be16 dport = udest->port;
1208 /* We use function that requires RCU lock */
1210 dest = ip_vs_lookup_dest(svc, udest->af, &udest->addr, dport);
1214 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1219 * Unlink dest from the service
1221 __ip_vs_unlink_dest(svc, dest, 1);
1224 * Delete the destination
1226 __ip_vs_del_dest(svc->ipvs, dest, false);
1233 static void ip_vs_dest_trash_expire(struct timer_list *t)
1235 struct netns_ipvs *ipvs = from_timer(ipvs, t, dest_trash_timer);
1236 struct ip_vs_dest *dest, *next;
1237 unsigned long now = jiffies;
1239 spin_lock(&ipvs->dest_trash_lock);
1240 list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) {
1241 if (refcount_read(&dest->refcnt) > 1)
1243 if (dest->idle_start) {
1244 if (time_before(now, dest->idle_start +
1245 IP_VS_DEST_TRASH_PERIOD))
1248 dest->idle_start = max(1UL, now);
1251 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1253 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1255 list_del(&dest->t_list);
1256 ip_vs_dest_free(dest);
1258 if (!list_empty(&ipvs->dest_trash))
1259 mod_timer(&ipvs->dest_trash_timer,
1260 jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1261 spin_unlock(&ipvs->dest_trash_lock);
1265 * Add a service into the service hash table
1268 ip_vs_add_service(struct netns_ipvs *ipvs, struct ip_vs_service_user_kern *u,
1269 struct ip_vs_service **svc_p)
1272 struct ip_vs_scheduler *sched = NULL;
1273 struct ip_vs_pe *pe = NULL;
1274 struct ip_vs_service *svc = NULL;
1276 /* increase the module use count */
1277 if (!ip_vs_use_count_inc())
1278 return -ENOPROTOOPT;
1280 /* Lookup the scheduler by 'u->sched_name' */
1281 if (strcmp(u->sched_name, "none")) {
1282 sched = ip_vs_scheduler_get(u->sched_name);
1284 pr_info("Scheduler module ip_vs_%s not found\n",
1291 if (u->pe_name && *u->pe_name) {
1292 pe = ip_vs_pe_getbyname(u->pe_name);
1294 pr_info("persistence engine module ip_vs_pe_%s "
1295 "not found\n", u->pe_name);
1301 #ifdef CONFIG_IP_VS_IPV6
1302 if (u->af == AF_INET6) {
1303 __u32 plen = (__force __u32) u->netmask;
1305 if (plen < 1 || plen > 128) {
1310 ret = nf_defrag_ipv6_enable(ipvs->net);
1316 svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1318 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1322 svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1323 if (!svc->stats.cpustats) {
1328 for_each_possible_cpu(i) {
1329 struct ip_vs_cpu_stats *ip_vs_stats;
1330 ip_vs_stats = per_cpu_ptr(svc->stats.cpustats, i);
1331 u64_stats_init(&ip_vs_stats->syncp);
1335 /* I'm the first user of the service */
1336 atomic_set(&svc->refcnt, 0);
1339 svc->protocol = u->protocol;
1340 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1341 svc->port = u->port;
1342 svc->fwmark = u->fwmark;
1343 svc->flags = u->flags;
1344 svc->timeout = u->timeout * HZ;
1345 svc->netmask = u->netmask;
1348 INIT_LIST_HEAD(&svc->destinations);
1349 spin_lock_init(&svc->sched_lock);
1350 spin_lock_init(&svc->stats.lock);
1352 /* Bind the scheduler */
1354 ret = ip_vs_bind_scheduler(svc, sched);
1360 /* Bind the ct retriever */
1361 RCU_INIT_POINTER(svc->pe, pe);
1364 /* Update the virtual service counters */
1365 if (svc->port == FTPPORT)
1366 atomic_inc(&ipvs->ftpsvc_counter);
1367 else if (svc->port == 0)
1368 atomic_inc(&ipvs->nullsvc_counter);
1369 if (svc->pe && svc->pe->conn_out)
1370 atomic_inc(&ipvs->conn_out_counter);
1372 ip_vs_start_estimator(ipvs, &svc->stats);
1374 /* Count only IPv4 services for old get/setsockopt interface */
1375 if (svc->af == AF_INET)
1376 ipvs->num_services++;
1378 /* Hash the service into the service table */
1379 ip_vs_svc_hash(svc);
1382 /* Now there is a service - full throttle */
1389 ip_vs_unbind_scheduler(svc, sched);
1390 ip_vs_service_free(svc);
1392 ip_vs_scheduler_put(sched);
1395 /* decrease the module use count */
1396 ip_vs_use_count_dec();
1403 * Edit a service and bind it with a new scheduler
1406 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1408 struct ip_vs_scheduler *sched = NULL, *old_sched;
1409 struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1411 bool new_pe_conn_out, old_pe_conn_out;
1414 * Lookup the scheduler, by 'u->sched_name'
1416 if (strcmp(u->sched_name, "none")) {
1417 sched = ip_vs_scheduler_get(u->sched_name);
1419 pr_info("Scheduler module ip_vs_%s not found\n",
1426 if (u->pe_name && *u->pe_name) {
1427 pe = ip_vs_pe_getbyname(u->pe_name);
1429 pr_info("persistence engine module ip_vs_pe_%s "
1430 "not found\n", u->pe_name);
1437 #ifdef CONFIG_IP_VS_IPV6
1438 if (u->af == AF_INET6) {
1439 __u32 plen = (__force __u32) u->netmask;
1441 if (plen < 1 || plen > 128) {
1448 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1449 if (sched != old_sched) {
1451 ip_vs_unbind_scheduler(svc, old_sched);
1452 RCU_INIT_POINTER(svc->scheduler, NULL);
1453 /* Wait all svc->sched_data users */
1456 /* Bind the new scheduler */
1458 ret = ip_vs_bind_scheduler(svc, sched);
1460 ip_vs_scheduler_put(sched);
1467 * Set the flags and timeout value
1469 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1470 svc->timeout = u->timeout * HZ;
1471 svc->netmask = u->netmask;
1473 old_pe = rcu_dereference_protected(svc->pe, 1);
1475 rcu_assign_pointer(svc->pe, pe);
1476 /* check for optional methods in new pe */
1477 new_pe_conn_out = (pe && pe->conn_out) ? true : false;
1478 old_pe_conn_out = (old_pe && old_pe->conn_out) ? true : false;
1479 if (new_pe_conn_out && !old_pe_conn_out)
1480 atomic_inc(&svc->ipvs->conn_out_counter);
1481 if (old_pe_conn_out && !new_pe_conn_out)
1482 atomic_dec(&svc->ipvs->conn_out_counter);
1486 ip_vs_scheduler_put(old_sched);
1487 ip_vs_pe_put(old_pe);
1492 * Delete a service from the service list
1493 * - The service must be unlinked, unlocked and not referenced!
1494 * - We are called under _bh lock
1496 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1498 struct ip_vs_dest *dest, *nxt;
1499 struct ip_vs_scheduler *old_sched;
1500 struct ip_vs_pe *old_pe;
1501 struct netns_ipvs *ipvs = svc->ipvs;
1503 /* Count only IPv4 services for old get/setsockopt interface */
1504 if (svc->af == AF_INET)
1505 ipvs->num_services--;
1507 ip_vs_stop_estimator(svc->ipvs, &svc->stats);
1509 /* Unbind scheduler */
1510 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1511 ip_vs_unbind_scheduler(svc, old_sched);
1512 ip_vs_scheduler_put(old_sched);
1514 /* Unbind persistence engine, keep svc->pe */
1515 old_pe = rcu_dereference_protected(svc->pe, 1);
1516 if (old_pe && old_pe->conn_out)
1517 atomic_dec(&ipvs->conn_out_counter);
1518 ip_vs_pe_put(old_pe);
1521 * Unlink the whole destination list
1523 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1524 __ip_vs_unlink_dest(svc, dest, 0);
1525 __ip_vs_del_dest(svc->ipvs, dest, cleanup);
1529 * Update the virtual service counters
1531 if (svc->port == FTPPORT)
1532 atomic_dec(&ipvs->ftpsvc_counter);
1533 else if (svc->port == 0)
1534 atomic_dec(&ipvs->nullsvc_counter);
1537 * Free the service if nobody refers to it
1539 __ip_vs_svc_put(svc, true);
1541 /* decrease the module use count */
1542 ip_vs_use_count_dec();
1546 * Unlink a service from list and try to delete it if its refcnt reached 0
1548 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1550 ip_vs_unregister_conntrack(svc);
1551 /* Hold svc to avoid double release from dest_trash */
1552 atomic_inc(&svc->refcnt);
1554 * Unhash it from the service table
1556 ip_vs_svc_unhash(svc);
1558 __ip_vs_del_service(svc, cleanup);
1562 * Delete a service from the service list
1564 static int ip_vs_del_service(struct ip_vs_service *svc)
1568 ip_vs_unlink_service(svc, false);
1575 * Flush all the virtual services
1577 static int ip_vs_flush(struct netns_ipvs *ipvs, bool cleanup)
1580 struct ip_vs_service *svc;
1581 struct hlist_node *n;
1584 * Flush the service table hashed by <netns,protocol,addr,port>
1586 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1587 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1589 if (svc->ipvs == ipvs)
1590 ip_vs_unlink_service(svc, cleanup);
1595 * Flush the service table hashed by fwmark
1597 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1598 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1600 if (svc->ipvs == ipvs)
1601 ip_vs_unlink_service(svc, cleanup);
1609 * Delete service by {netns} in the service table.
1610 * Called by __ip_vs_cleanup()
1612 void ip_vs_service_net_cleanup(struct netns_ipvs *ipvs)
1615 /* Check for "full" addressed entries */
1616 mutex_lock(&__ip_vs_mutex);
1617 ip_vs_flush(ipvs, true);
1618 mutex_unlock(&__ip_vs_mutex);
1622 /* Put all references for device (dst_cache) */
1624 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1626 struct ip_vs_dest_dst *dest_dst;
1628 spin_lock_bh(&dest->dst_lock);
1629 dest_dst = rcu_dereference_protected(dest->dest_dst, 1);
1630 if (dest_dst && dest_dst->dst_cache->dev == dev) {
1631 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1633 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1635 refcount_read(&dest->refcnt));
1636 __ip_vs_dst_cache_reset(dest);
1638 spin_unlock_bh(&dest->dst_lock);
1641 /* Netdev event receiver
1642 * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1644 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1647 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1648 struct net *net = dev_net(dev);
1649 struct netns_ipvs *ipvs = net_ipvs(net);
1650 struct ip_vs_service *svc;
1651 struct ip_vs_dest *dest;
1654 if (event != NETDEV_DOWN || !ipvs)
1656 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1658 mutex_lock(&__ip_vs_mutex);
1659 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1660 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1661 if (svc->ipvs == ipvs) {
1662 list_for_each_entry(dest, &svc->destinations,
1664 ip_vs_forget_dev(dest, dev);
1669 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1670 if (svc->ipvs == ipvs) {
1671 list_for_each_entry(dest, &svc->destinations,
1673 ip_vs_forget_dev(dest, dev);
1680 spin_lock_bh(&ipvs->dest_trash_lock);
1681 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1682 ip_vs_forget_dev(dest, dev);
1684 spin_unlock_bh(&ipvs->dest_trash_lock);
1685 mutex_unlock(&__ip_vs_mutex);
1691 * Zero counters in a service or all services
1693 static int ip_vs_zero_service(struct ip_vs_service *svc)
1695 struct ip_vs_dest *dest;
1697 list_for_each_entry(dest, &svc->destinations, n_list) {
1698 ip_vs_zero_stats(&dest->stats);
1700 ip_vs_zero_stats(&svc->stats);
1704 static int ip_vs_zero_all(struct netns_ipvs *ipvs)
1707 struct ip_vs_service *svc;
1709 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1710 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1711 if (svc->ipvs == ipvs)
1712 ip_vs_zero_service(svc);
1716 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1717 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1718 if (svc->ipvs == ipvs)
1719 ip_vs_zero_service(svc);
1723 ip_vs_zero_stats(&ipvs->tot_stats);
1727 #ifdef CONFIG_SYSCTL
1729 static int three = 3;
1732 proc_do_defense_mode(struct ctl_table *table, int write,
1733 void __user *buffer, size_t *lenp, loff_t *ppos)
1735 struct netns_ipvs *ipvs = table->extra2;
1736 int *valp = table->data;
1740 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1741 if (write && (*valp != val)) {
1742 if ((*valp < 0) || (*valp > 3)) {
1743 /* Restore the correct value */
1746 update_defense_level(ipvs);
1753 proc_do_sync_threshold(struct ctl_table *table, int write,
1754 void __user *buffer, size_t *lenp, loff_t *ppos)
1756 int *valp = table->data;
1760 /* backup the value first */
1761 memcpy(val, valp, sizeof(val));
1763 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1764 if (write && (valp[0] < 0 || valp[1] < 0 ||
1765 (valp[0] >= valp[1] && valp[1]))) {
1766 /* Restore the correct value */
1767 memcpy(valp, val, sizeof(val));
1773 proc_do_sync_mode(struct ctl_table *table, int write,
1774 void __user *buffer, size_t *lenp, loff_t *ppos)
1776 int *valp = table->data;
1780 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1781 if (write && (*valp != val)) {
1782 if ((*valp < 0) || (*valp > 1)) {
1783 /* Restore the correct value */
1791 proc_do_sync_ports(struct ctl_table *table, int write,
1792 void __user *buffer, size_t *lenp, loff_t *ppos)
1794 int *valp = table->data;
1798 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1799 if (write && (*valp != val)) {
1800 if (*valp < 1 || !is_power_of_2(*valp)) {
1801 /* Restore the correct value */
1809 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1810 * Do not change order or insert new entries without
1811 * align with netns init in ip_vs_control_net_init()
1814 static struct ctl_table vs_vars[] = {
1816 .procname = "amemthresh",
1817 .maxlen = sizeof(int),
1819 .proc_handler = proc_dointvec,
1822 .procname = "am_droprate",
1823 .maxlen = sizeof(int),
1825 .proc_handler = proc_dointvec,
1828 .procname = "drop_entry",
1829 .maxlen = sizeof(int),
1831 .proc_handler = proc_do_defense_mode,
1834 .procname = "drop_packet",
1835 .maxlen = sizeof(int),
1837 .proc_handler = proc_do_defense_mode,
1839 #ifdef CONFIG_IP_VS_NFCT
1841 .procname = "conntrack",
1842 .maxlen = sizeof(int),
1844 .proc_handler = &proc_dointvec,
1848 .procname = "secure_tcp",
1849 .maxlen = sizeof(int),
1851 .proc_handler = proc_do_defense_mode,
1854 .procname = "snat_reroute",
1855 .maxlen = sizeof(int),
1857 .proc_handler = &proc_dointvec,
1860 .procname = "sync_version",
1861 .maxlen = sizeof(int),
1863 .proc_handler = proc_do_sync_mode,
1866 .procname = "sync_ports",
1867 .maxlen = sizeof(int),
1869 .proc_handler = proc_do_sync_ports,
1872 .procname = "sync_persist_mode",
1873 .maxlen = sizeof(int),
1875 .proc_handler = proc_dointvec,
1878 .procname = "sync_qlen_max",
1879 .maxlen = sizeof(unsigned long),
1881 .proc_handler = proc_doulongvec_minmax,
1884 .procname = "sync_sock_size",
1885 .maxlen = sizeof(int),
1887 .proc_handler = proc_dointvec,
1890 .procname = "cache_bypass",
1891 .maxlen = sizeof(int),
1893 .proc_handler = proc_dointvec,
1896 .procname = "expire_nodest_conn",
1897 .maxlen = sizeof(int),
1899 .proc_handler = proc_dointvec,
1902 .procname = "sloppy_tcp",
1903 .maxlen = sizeof(int),
1905 .proc_handler = proc_dointvec,
1908 .procname = "sloppy_sctp",
1909 .maxlen = sizeof(int),
1911 .proc_handler = proc_dointvec,
1914 .procname = "expire_quiescent_template",
1915 .maxlen = sizeof(int),
1917 .proc_handler = proc_dointvec,
1920 .procname = "sync_threshold",
1922 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1924 .proc_handler = proc_do_sync_threshold,
1927 .procname = "sync_refresh_period",
1928 .maxlen = sizeof(int),
1930 .proc_handler = proc_dointvec_jiffies,
1933 .procname = "sync_retries",
1934 .maxlen = sizeof(int),
1936 .proc_handler = proc_dointvec_minmax,
1937 .extra1 = SYSCTL_ZERO,
1941 .procname = "nat_icmp_send",
1942 .maxlen = sizeof(int),
1944 .proc_handler = proc_dointvec,
1947 .procname = "pmtu_disc",
1948 .maxlen = sizeof(int),
1950 .proc_handler = proc_dointvec,
1953 .procname = "backup_only",
1954 .maxlen = sizeof(int),
1956 .proc_handler = proc_dointvec,
1959 .procname = "conn_reuse_mode",
1960 .maxlen = sizeof(int),
1962 .proc_handler = proc_dointvec,
1965 .procname = "schedule_icmp",
1966 .maxlen = sizeof(int),
1968 .proc_handler = proc_dointvec,
1971 .procname = "ignore_tunneled",
1972 .maxlen = sizeof(int),
1974 .proc_handler = proc_dointvec,
1976 #ifdef CONFIG_IP_VS_DEBUG
1978 .procname = "debug_level",
1979 .data = &sysctl_ip_vs_debug_level,
1980 .maxlen = sizeof(int),
1982 .proc_handler = proc_dointvec,
1990 #ifdef CONFIG_PROC_FS
1993 struct seq_net_private p; /* Do not move this, netns depends upon it*/
1994 struct hlist_head *table;
1999 * Write the contents of the VS rule table to a PROCfs file.
2000 * (It is kept just for backward compatibility)
2002 static inline const char *ip_vs_fwd_name(unsigned int flags)
2004 switch (flags & IP_VS_CONN_F_FWD_MASK) {
2005 case IP_VS_CONN_F_LOCALNODE:
2007 case IP_VS_CONN_F_TUNNEL:
2009 case IP_VS_CONN_F_DROUTE:
2017 /* Get the Nth entry in the two lists */
2018 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
2020 struct net *net = seq_file_net(seq);
2021 struct netns_ipvs *ipvs = net_ipvs(net);
2022 struct ip_vs_iter *iter = seq->private;
2024 struct ip_vs_service *svc;
2026 /* look in hash by protocol */
2027 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2028 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
2029 if ((svc->ipvs == ipvs) && pos-- == 0) {
2030 iter->table = ip_vs_svc_table;
2037 /* keep looking in fwmark */
2038 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2039 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
2041 if ((svc->ipvs == ipvs) && pos-- == 0) {
2042 iter->table = ip_vs_svc_fwm_table;
2052 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
2056 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
2060 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2062 struct hlist_node *e;
2063 struct ip_vs_iter *iter;
2064 struct ip_vs_service *svc;
2067 if (v == SEQ_START_TOKEN)
2068 return ip_vs_info_array(seq,0);
2071 iter = seq->private;
2073 if (iter->table == ip_vs_svc_table) {
2074 /* next service in table hashed by protocol */
2075 e = rcu_dereference(hlist_next_rcu(&svc->s_list));
2077 return hlist_entry(e, struct ip_vs_service, s_list);
2079 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2080 hlist_for_each_entry_rcu(svc,
2081 &ip_vs_svc_table[iter->bucket],
2087 iter->table = ip_vs_svc_fwm_table;
2092 /* next service in hashed by fwmark */
2093 e = rcu_dereference(hlist_next_rcu(&svc->f_list));
2095 return hlist_entry(e, struct ip_vs_service, f_list);
2098 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2099 hlist_for_each_entry_rcu(svc,
2100 &ip_vs_svc_fwm_table[iter->bucket],
2108 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
2115 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
2117 if (v == SEQ_START_TOKEN) {
2119 "IP Virtual Server version %d.%d.%d (size=%d)\n",
2120 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2122 "Prot LocalAddress:Port Scheduler Flags\n");
2124 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2126 struct net *net = seq_file_net(seq);
2127 struct netns_ipvs *ipvs = net_ipvs(net);
2128 const struct ip_vs_service *svc = v;
2129 const struct ip_vs_iter *iter = seq->private;
2130 const struct ip_vs_dest *dest;
2131 struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler);
2132 char *sched_name = sched ? sched->name : "none";
2134 if (svc->ipvs != ipvs)
2136 if (iter->table == ip_vs_svc_table) {
2137 #ifdef CONFIG_IP_VS_IPV6
2138 if (svc->af == AF_INET6)
2139 seq_printf(seq, "%s [%pI6]:%04X %s ",
2140 ip_vs_proto_name(svc->protocol),
2146 seq_printf(seq, "%s %08X:%04X %s %s ",
2147 ip_vs_proto_name(svc->protocol),
2148 ntohl(svc->addr.ip),
2151 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2153 seq_printf(seq, "FWM %08X %s %s",
2154 svc->fwmark, sched_name,
2155 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2158 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2159 seq_printf(seq, "persistent %d %08X\n",
2161 ntohl(svc->netmask));
2163 seq_putc(seq, '\n');
2165 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
2166 #ifdef CONFIG_IP_VS_IPV6
2167 if (dest->af == AF_INET6)
2170 " %-7s %-6d %-10d %-10d\n",
2173 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2174 atomic_read(&dest->weight),
2175 atomic_read(&dest->activeconns),
2176 atomic_read(&dest->inactconns));
2181 "%-7s %-6d %-10d %-10d\n",
2182 ntohl(dest->addr.ip),
2184 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2185 atomic_read(&dest->weight),
2186 atomic_read(&dest->activeconns),
2187 atomic_read(&dest->inactconns));
2194 static const struct seq_operations ip_vs_info_seq_ops = {
2195 .start = ip_vs_info_seq_start,
2196 .next = ip_vs_info_seq_next,
2197 .stop = ip_vs_info_seq_stop,
2198 .show = ip_vs_info_seq_show,
2201 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2203 struct net *net = seq_file_single_net(seq);
2204 struct ip_vs_kstats show;
2206 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2208 " Total Incoming Outgoing Incoming Outgoing\n");
2210 " Conns Packets Packets Bytes Bytes\n");
2212 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2213 seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n\n",
2214 (unsigned long long)show.conns,
2215 (unsigned long long)show.inpkts,
2216 (unsigned long long)show.outpkts,
2217 (unsigned long long)show.inbytes,
2218 (unsigned long long)show.outbytes);
2220 /* 01234567 01234567 01234567 0123456701234567 0123456701234567*/
2222 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2223 seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n",
2224 (unsigned long long)show.cps,
2225 (unsigned long long)show.inpps,
2226 (unsigned long long)show.outpps,
2227 (unsigned long long)show.inbps,
2228 (unsigned long long)show.outbps);
2233 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2235 struct net *net = seq_file_single_net(seq);
2236 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2237 struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats;
2238 struct ip_vs_kstats kstats;
2241 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2243 " Total Incoming Outgoing Incoming Outgoing\n");
2245 "CPU Conns Packets Packets Bytes Bytes\n");
2247 for_each_possible_cpu(i) {
2248 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2250 u64 conns, inpkts, outpkts, inbytes, outbytes;
2253 start = u64_stats_fetch_begin_irq(&u->syncp);
2254 conns = u->cnt.conns;
2255 inpkts = u->cnt.inpkts;
2256 outpkts = u->cnt.outpkts;
2257 inbytes = u->cnt.inbytes;
2258 outbytes = u->cnt.outbytes;
2259 } while (u64_stats_fetch_retry_irq(&u->syncp, start));
2261 seq_printf(seq, "%3X %8LX %8LX %8LX %16LX %16LX\n",
2262 i, (u64)conns, (u64)inpkts,
2263 (u64)outpkts, (u64)inbytes,
2267 ip_vs_copy_stats(&kstats, tot_stats);
2269 seq_printf(seq, " ~ %8LX %8LX %8LX %16LX %16LX\n\n",
2270 (unsigned long long)kstats.conns,
2271 (unsigned long long)kstats.inpkts,
2272 (unsigned long long)kstats.outpkts,
2273 (unsigned long long)kstats.inbytes,
2274 (unsigned long long)kstats.outbytes);
2276 /* ... 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2278 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2279 seq_printf(seq, " %8LX %8LX %8LX %16LX %16LX\n",
2291 * Set timeout values for tcp tcpfin udp in the timeout_table.
2293 static int ip_vs_set_timeout(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2295 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2296 struct ip_vs_proto_data *pd;
2299 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2304 #ifdef CONFIG_IP_VS_PROTO_TCP
2305 if (u->tcp_timeout < 0 || u->tcp_timeout > (INT_MAX / HZ) ||
2306 u->tcp_fin_timeout < 0 || u->tcp_fin_timeout > (INT_MAX / HZ)) {
2311 #ifdef CONFIG_IP_VS_PROTO_UDP
2312 if (u->udp_timeout < 0 || u->udp_timeout > (INT_MAX / HZ))
2316 #ifdef CONFIG_IP_VS_PROTO_TCP
2317 if (u->tcp_timeout) {
2318 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2319 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2320 = u->tcp_timeout * HZ;
2323 if (u->tcp_fin_timeout) {
2324 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2325 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2326 = u->tcp_fin_timeout * HZ;
2330 #ifdef CONFIG_IP_VS_PROTO_UDP
2331 if (u->udp_timeout) {
2332 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2333 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2334 = u->udp_timeout * HZ;
2340 #define CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2342 struct ip_vs_svcdest_user {
2343 struct ip_vs_service_user s;
2344 struct ip_vs_dest_user d;
2347 static const unsigned char set_arglen[CMDID(IP_VS_SO_SET_MAX) + 1] = {
2348 [CMDID(IP_VS_SO_SET_ADD)] = sizeof(struct ip_vs_service_user),
2349 [CMDID(IP_VS_SO_SET_EDIT)] = sizeof(struct ip_vs_service_user),
2350 [CMDID(IP_VS_SO_SET_DEL)] = sizeof(struct ip_vs_service_user),
2351 [CMDID(IP_VS_SO_SET_ADDDEST)] = sizeof(struct ip_vs_svcdest_user),
2352 [CMDID(IP_VS_SO_SET_DELDEST)] = sizeof(struct ip_vs_svcdest_user),
2353 [CMDID(IP_VS_SO_SET_EDITDEST)] = sizeof(struct ip_vs_svcdest_user),
2354 [CMDID(IP_VS_SO_SET_TIMEOUT)] = sizeof(struct ip_vs_timeout_user),
2355 [CMDID(IP_VS_SO_SET_STARTDAEMON)] = sizeof(struct ip_vs_daemon_user),
2356 [CMDID(IP_VS_SO_SET_STOPDAEMON)] = sizeof(struct ip_vs_daemon_user),
2357 [CMDID(IP_VS_SO_SET_ZERO)] = sizeof(struct ip_vs_service_user),
2360 union ip_vs_set_arglen {
2361 struct ip_vs_service_user field_IP_VS_SO_SET_ADD;
2362 struct ip_vs_service_user field_IP_VS_SO_SET_EDIT;
2363 struct ip_vs_service_user field_IP_VS_SO_SET_DEL;
2364 struct ip_vs_svcdest_user field_IP_VS_SO_SET_ADDDEST;
2365 struct ip_vs_svcdest_user field_IP_VS_SO_SET_DELDEST;
2366 struct ip_vs_svcdest_user field_IP_VS_SO_SET_EDITDEST;
2367 struct ip_vs_timeout_user field_IP_VS_SO_SET_TIMEOUT;
2368 struct ip_vs_daemon_user field_IP_VS_SO_SET_STARTDAEMON;
2369 struct ip_vs_daemon_user field_IP_VS_SO_SET_STOPDAEMON;
2370 struct ip_vs_service_user field_IP_VS_SO_SET_ZERO;
2373 #define MAX_SET_ARGLEN sizeof(union ip_vs_set_arglen)
2375 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2376 struct ip_vs_service_user *usvc_compat)
2378 memset(usvc, 0, sizeof(*usvc));
2381 usvc->protocol = usvc_compat->protocol;
2382 usvc->addr.ip = usvc_compat->addr;
2383 usvc->port = usvc_compat->port;
2384 usvc->fwmark = usvc_compat->fwmark;
2386 /* Deep copy of sched_name is not needed here */
2387 usvc->sched_name = usvc_compat->sched_name;
2389 usvc->flags = usvc_compat->flags;
2390 usvc->timeout = usvc_compat->timeout;
2391 usvc->netmask = usvc_compat->netmask;
2394 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2395 struct ip_vs_dest_user *udest_compat)
2397 memset(udest, 0, sizeof(*udest));
2399 udest->addr.ip = udest_compat->addr;
2400 udest->port = udest_compat->port;
2401 udest->conn_flags = udest_compat->conn_flags;
2402 udest->weight = udest_compat->weight;
2403 udest->u_threshold = udest_compat->u_threshold;
2404 udest->l_threshold = udest_compat->l_threshold;
2405 udest->af = AF_INET;
2406 udest->tun_type = IP_VS_CONN_F_TUNNEL_TYPE_IPIP;
2410 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2412 struct net *net = sock_net(sk);
2414 unsigned char arg[MAX_SET_ARGLEN];
2415 struct ip_vs_service_user *usvc_compat;
2416 struct ip_vs_service_user_kern usvc;
2417 struct ip_vs_service *svc;
2418 struct ip_vs_dest_user *udest_compat;
2419 struct ip_vs_dest_user_kern udest;
2420 struct netns_ipvs *ipvs = net_ipvs(net);
2422 BUILD_BUG_ON(sizeof(arg) > 255);
2423 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2426 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2428 if (len != set_arglen[CMDID(cmd)]) {
2429 IP_VS_DBG(1, "set_ctl: len %u != %u\n",
2430 len, set_arglen[CMDID(cmd)]);
2434 if (copy_from_user(arg, user, len) != 0)
2437 /* Handle daemons since they have another lock */
2438 if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2439 cmd == IP_VS_SO_SET_STOPDAEMON) {
2440 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2442 if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2443 struct ipvs_sync_daemon_cfg cfg;
2445 memset(&cfg, 0, sizeof(cfg));
2447 if (strscpy(cfg.mcast_ifn, dm->mcast_ifn,
2448 sizeof(cfg.mcast_ifn)) <= 0)
2450 cfg.syncid = dm->syncid;
2451 ret = start_sync_thread(ipvs, &cfg, dm->state);
2453 ret = stop_sync_thread(ipvs, dm->state);
2458 mutex_lock(&__ip_vs_mutex);
2459 if (cmd == IP_VS_SO_SET_FLUSH) {
2460 /* Flush the virtual service */
2461 ret = ip_vs_flush(ipvs, false);
2463 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2464 /* Set timeout values for (tcp tcpfin udp) */
2465 ret = ip_vs_set_timeout(ipvs, (struct ip_vs_timeout_user *)arg);
2469 usvc_compat = (struct ip_vs_service_user *)arg;
2470 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2472 /* We only use the new structs internally, so copy userspace compat
2473 * structs to extended internal versions */
2474 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2475 ip_vs_copy_udest_compat(&udest, udest_compat);
2477 if (cmd == IP_VS_SO_SET_ZERO) {
2478 /* if no service address is set, zero counters in all */
2479 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2480 ret = ip_vs_zero_all(ipvs);
2485 if ((cmd == IP_VS_SO_SET_ADD || cmd == IP_VS_SO_SET_EDIT) &&
2486 strnlen(usvc.sched_name, IP_VS_SCHEDNAME_MAXLEN) ==
2487 IP_VS_SCHEDNAME_MAXLEN) {
2492 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2493 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2494 usvc.protocol != IPPROTO_SCTP) {
2495 pr_err("set_ctl: invalid protocol: %d %pI4:%d\n",
2496 usvc.protocol, &usvc.addr.ip,
2502 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2504 if (usvc.fwmark == 0)
2505 svc = __ip_vs_service_find(ipvs, usvc.af, usvc.protocol,
2506 &usvc.addr, usvc.port);
2508 svc = __ip_vs_svc_fwm_find(ipvs, usvc.af, usvc.fwmark);
2511 if (cmd != IP_VS_SO_SET_ADD
2512 && (svc == NULL || svc->protocol != usvc.protocol)) {
2518 case IP_VS_SO_SET_ADD:
2522 ret = ip_vs_add_service(ipvs, &usvc, &svc);
2524 case IP_VS_SO_SET_EDIT:
2525 ret = ip_vs_edit_service(svc, &usvc);
2527 case IP_VS_SO_SET_DEL:
2528 ret = ip_vs_del_service(svc);
2532 case IP_VS_SO_SET_ZERO:
2533 ret = ip_vs_zero_service(svc);
2535 case IP_VS_SO_SET_ADDDEST:
2536 ret = ip_vs_add_dest(svc, &udest);
2538 case IP_VS_SO_SET_EDITDEST:
2539 ret = ip_vs_edit_dest(svc, &udest);
2541 case IP_VS_SO_SET_DELDEST:
2542 ret = ip_vs_del_dest(svc, &udest);
2549 mutex_unlock(&__ip_vs_mutex);
2555 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2557 struct ip_vs_scheduler *sched;
2558 struct ip_vs_kstats kstats;
2561 sched = rcu_dereference_protected(src->scheduler, 1);
2562 sched_name = sched ? sched->name : "none";
2563 dst->protocol = src->protocol;
2564 dst->addr = src->addr.ip;
2565 dst->port = src->port;
2566 dst->fwmark = src->fwmark;
2567 strlcpy(dst->sched_name, sched_name, sizeof(dst->sched_name));
2568 dst->flags = src->flags;
2569 dst->timeout = src->timeout / HZ;
2570 dst->netmask = src->netmask;
2571 dst->num_dests = src->num_dests;
2572 ip_vs_copy_stats(&kstats, &src->stats);
2573 ip_vs_export_stats_user(&dst->stats, &kstats);
2577 __ip_vs_get_service_entries(struct netns_ipvs *ipvs,
2578 const struct ip_vs_get_services *get,
2579 struct ip_vs_get_services __user *uptr)
2582 struct ip_vs_service *svc;
2583 struct ip_vs_service_entry entry;
2586 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2587 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2588 /* Only expose IPv4 entries to old interface */
2589 if (svc->af != AF_INET || (svc->ipvs != ipvs))
2592 if (count >= get->num_services)
2594 memset(&entry, 0, sizeof(entry));
2595 ip_vs_copy_service(&entry, svc);
2596 if (copy_to_user(&uptr->entrytable[count],
2597 &entry, sizeof(entry))) {
2605 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2606 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2607 /* Only expose IPv4 entries to old interface */
2608 if (svc->af != AF_INET || (svc->ipvs != ipvs))
2611 if (count >= get->num_services)
2613 memset(&entry, 0, sizeof(entry));
2614 ip_vs_copy_service(&entry, svc);
2615 if (copy_to_user(&uptr->entrytable[count],
2616 &entry, sizeof(entry))) {
2628 __ip_vs_get_dest_entries(struct netns_ipvs *ipvs, const struct ip_vs_get_dests *get,
2629 struct ip_vs_get_dests __user *uptr)
2631 struct ip_vs_service *svc;
2632 union nf_inet_addr addr = { .ip = get->addr };
2637 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, get->fwmark);
2639 svc = __ip_vs_service_find(ipvs, AF_INET, get->protocol, &addr,
2645 struct ip_vs_dest *dest;
2646 struct ip_vs_dest_entry entry;
2647 struct ip_vs_kstats kstats;
2649 memset(&entry, 0, sizeof(entry));
2650 list_for_each_entry(dest, &svc->destinations, n_list) {
2651 if (count >= get->num_dests)
2654 /* Cannot expose heterogeneous members via sockopt
2657 if (dest->af != svc->af)
2660 entry.addr = dest->addr.ip;
2661 entry.port = dest->port;
2662 entry.conn_flags = atomic_read(&dest->conn_flags);
2663 entry.weight = atomic_read(&dest->weight);
2664 entry.u_threshold = dest->u_threshold;
2665 entry.l_threshold = dest->l_threshold;
2666 entry.activeconns = atomic_read(&dest->activeconns);
2667 entry.inactconns = atomic_read(&dest->inactconns);
2668 entry.persistconns = atomic_read(&dest->persistconns);
2669 ip_vs_copy_stats(&kstats, &dest->stats);
2670 ip_vs_export_stats_user(&entry.stats, &kstats);
2671 if (copy_to_user(&uptr->entrytable[count],
2672 &entry, sizeof(entry))) {
2684 __ip_vs_get_timeouts(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2686 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2687 struct ip_vs_proto_data *pd;
2690 memset(u, 0, sizeof (*u));
2692 #ifdef CONFIG_IP_VS_PROTO_TCP
2693 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2694 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2695 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2697 #ifdef CONFIG_IP_VS_PROTO_UDP
2698 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2700 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2704 static const unsigned char get_arglen[CMDID(IP_VS_SO_GET_MAX) + 1] = {
2705 [CMDID(IP_VS_SO_GET_VERSION)] = 64,
2706 [CMDID(IP_VS_SO_GET_INFO)] = sizeof(struct ip_vs_getinfo),
2707 [CMDID(IP_VS_SO_GET_SERVICES)] = sizeof(struct ip_vs_get_services),
2708 [CMDID(IP_VS_SO_GET_SERVICE)] = sizeof(struct ip_vs_service_entry),
2709 [CMDID(IP_VS_SO_GET_DESTS)] = sizeof(struct ip_vs_get_dests),
2710 [CMDID(IP_VS_SO_GET_TIMEOUT)] = sizeof(struct ip_vs_timeout_user),
2711 [CMDID(IP_VS_SO_GET_DAEMON)] = 2 * sizeof(struct ip_vs_daemon_user),
2714 union ip_vs_get_arglen {
2715 char field_IP_VS_SO_GET_VERSION[64];
2716 struct ip_vs_getinfo field_IP_VS_SO_GET_INFO;
2717 struct ip_vs_get_services field_IP_VS_SO_GET_SERVICES;
2718 struct ip_vs_service_entry field_IP_VS_SO_GET_SERVICE;
2719 struct ip_vs_get_dests field_IP_VS_SO_GET_DESTS;
2720 struct ip_vs_timeout_user field_IP_VS_SO_GET_TIMEOUT;
2721 struct ip_vs_daemon_user field_IP_VS_SO_GET_DAEMON[2];
2724 #define MAX_GET_ARGLEN sizeof(union ip_vs_get_arglen)
2727 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2729 unsigned char arg[MAX_GET_ARGLEN];
2731 unsigned int copylen;
2732 struct net *net = sock_net(sk);
2733 struct netns_ipvs *ipvs = net_ipvs(net);
2736 BUILD_BUG_ON(sizeof(arg) > 255);
2737 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2740 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2743 copylen = get_arglen[CMDID(cmd)];
2744 if (*len < (int) copylen) {
2745 IP_VS_DBG(1, "get_ctl: len %d < %u\n", *len, copylen);
2749 if (copy_from_user(arg, user, copylen) != 0)
2752 * Handle daemons first since it has its own locking
2754 if (cmd == IP_VS_SO_GET_DAEMON) {
2755 struct ip_vs_daemon_user d[2];
2757 memset(&d, 0, sizeof(d));
2758 mutex_lock(&ipvs->sync_mutex);
2759 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2760 d[0].state = IP_VS_STATE_MASTER;
2761 strlcpy(d[0].mcast_ifn, ipvs->mcfg.mcast_ifn,
2762 sizeof(d[0].mcast_ifn));
2763 d[0].syncid = ipvs->mcfg.syncid;
2765 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2766 d[1].state = IP_VS_STATE_BACKUP;
2767 strlcpy(d[1].mcast_ifn, ipvs->bcfg.mcast_ifn,
2768 sizeof(d[1].mcast_ifn));
2769 d[1].syncid = ipvs->bcfg.syncid;
2771 if (copy_to_user(user, &d, sizeof(d)) != 0)
2773 mutex_unlock(&ipvs->sync_mutex);
2777 mutex_lock(&__ip_vs_mutex);
2779 case IP_VS_SO_GET_VERSION:
2783 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2784 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2785 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2789 *len = strlen(buf)+1;
2793 case IP_VS_SO_GET_INFO:
2795 struct ip_vs_getinfo info;
2796 info.version = IP_VS_VERSION_CODE;
2797 info.size = ip_vs_conn_tab_size;
2798 info.num_services = ipvs->num_services;
2799 if (copy_to_user(user, &info, sizeof(info)) != 0)
2804 case IP_VS_SO_GET_SERVICES:
2806 struct ip_vs_get_services *get;
2809 get = (struct ip_vs_get_services *)arg;
2810 size = struct_size(get, entrytable, get->num_services);
2812 pr_err("length: %u != %u\n", *len, size);
2816 ret = __ip_vs_get_service_entries(ipvs, get, user);
2820 case IP_VS_SO_GET_SERVICE:
2822 struct ip_vs_service_entry *entry;
2823 struct ip_vs_service *svc;
2824 union nf_inet_addr addr;
2826 entry = (struct ip_vs_service_entry *)arg;
2827 addr.ip = entry->addr;
2830 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, entry->fwmark);
2832 svc = __ip_vs_service_find(ipvs, AF_INET,
2833 entry->protocol, &addr,
2837 ip_vs_copy_service(entry, svc);
2838 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2845 case IP_VS_SO_GET_DESTS:
2847 struct ip_vs_get_dests *get;
2850 get = (struct ip_vs_get_dests *)arg;
2851 size = struct_size(get, entrytable, get->num_dests);
2853 pr_err("length: %u != %u\n", *len, size);
2857 ret = __ip_vs_get_dest_entries(ipvs, get, user);
2861 case IP_VS_SO_GET_TIMEOUT:
2863 struct ip_vs_timeout_user t;
2865 __ip_vs_get_timeouts(ipvs, &t);
2866 if (copy_to_user(user, &t, sizeof(t)) != 0)
2876 mutex_unlock(&__ip_vs_mutex);
2881 static struct nf_sockopt_ops ip_vs_sockopts = {
2883 .set_optmin = IP_VS_BASE_CTL,
2884 .set_optmax = IP_VS_SO_SET_MAX+1,
2885 .set = do_ip_vs_set_ctl,
2886 .get_optmin = IP_VS_BASE_CTL,
2887 .get_optmax = IP_VS_SO_GET_MAX+1,
2888 .get = do_ip_vs_get_ctl,
2889 .owner = THIS_MODULE,
2893 * Generic Netlink interface
2896 /* IPVS genetlink family */
2897 static struct genl_family ip_vs_genl_family;
2899 /* Policy used for first-level command attributes */
2900 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2901 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2902 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2903 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2904 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2905 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2906 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2909 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2910 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2911 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2912 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2913 .len = IP_VS_IFNAME_MAXLEN - 1 },
2914 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2915 [IPVS_DAEMON_ATTR_SYNC_MAXLEN] = { .type = NLA_U16 },
2916 [IPVS_DAEMON_ATTR_MCAST_GROUP] = { .type = NLA_U32 },
2917 [IPVS_DAEMON_ATTR_MCAST_GROUP6] = { .len = sizeof(struct in6_addr) },
2918 [IPVS_DAEMON_ATTR_MCAST_PORT] = { .type = NLA_U16 },
2919 [IPVS_DAEMON_ATTR_MCAST_TTL] = { .type = NLA_U8 },
2922 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2923 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2924 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2925 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2926 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2927 .len = sizeof(union nf_inet_addr) },
2928 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2929 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2930 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2931 .len = IP_VS_SCHEDNAME_MAXLEN - 1 },
2932 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2933 .len = IP_VS_PENAME_MAXLEN },
2934 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2935 .len = sizeof(struct ip_vs_flags) },
2936 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2937 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2938 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2941 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2942 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2943 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2944 .len = sizeof(union nf_inet_addr) },
2945 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2946 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2947 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2948 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2949 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2950 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2951 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2952 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2953 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2954 [IPVS_DEST_ATTR_ADDR_FAMILY] = { .type = NLA_U16 },
2955 [IPVS_DEST_ATTR_TUN_TYPE] = { .type = NLA_U8 },
2956 [IPVS_DEST_ATTR_TUN_PORT] = { .type = NLA_U16 },
2957 [IPVS_DEST_ATTR_TUN_FLAGS] = { .type = NLA_U16 },
2960 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2961 struct ip_vs_kstats *kstats)
2963 struct nlattr *nl_stats = nla_nest_start_noflag(skb, container_type);
2968 if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, (u32)kstats->conns) ||
2969 nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, (u32)kstats->inpkts) ||
2970 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, (u32)kstats->outpkts) ||
2971 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes,
2972 IPVS_STATS_ATTR_PAD) ||
2973 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes,
2974 IPVS_STATS_ATTR_PAD) ||
2975 nla_put_u32(skb, IPVS_STATS_ATTR_CPS, (u32)kstats->cps) ||
2976 nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, (u32)kstats->inpps) ||
2977 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, (u32)kstats->outpps) ||
2978 nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, (u32)kstats->inbps) ||
2979 nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, (u32)kstats->outbps))
2980 goto nla_put_failure;
2981 nla_nest_end(skb, nl_stats);
2986 nla_nest_cancel(skb, nl_stats);
2990 static int ip_vs_genl_fill_stats64(struct sk_buff *skb, int container_type,
2991 struct ip_vs_kstats *kstats)
2993 struct nlattr *nl_stats = nla_nest_start_noflag(skb, container_type);
2998 if (nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CONNS, kstats->conns,
2999 IPVS_STATS_ATTR_PAD) ||
3000 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPKTS, kstats->inpkts,
3001 IPVS_STATS_ATTR_PAD) ||
3002 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPKTS, kstats->outpkts,
3003 IPVS_STATS_ATTR_PAD) ||
3004 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes,
3005 IPVS_STATS_ATTR_PAD) ||
3006 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes,
3007 IPVS_STATS_ATTR_PAD) ||
3008 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CPS, kstats->cps,
3009 IPVS_STATS_ATTR_PAD) ||
3010 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPPS, kstats->inpps,
3011 IPVS_STATS_ATTR_PAD) ||
3012 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPPS, kstats->outpps,
3013 IPVS_STATS_ATTR_PAD) ||
3014 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBPS, kstats->inbps,
3015 IPVS_STATS_ATTR_PAD) ||
3016 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBPS, kstats->outbps,
3017 IPVS_STATS_ATTR_PAD))
3018 goto nla_put_failure;
3019 nla_nest_end(skb, nl_stats);
3024 nla_nest_cancel(skb, nl_stats);
3028 static int ip_vs_genl_fill_service(struct sk_buff *skb,
3029 struct ip_vs_service *svc)
3031 struct ip_vs_scheduler *sched;
3032 struct ip_vs_pe *pe;
3033 struct nlattr *nl_service;
3034 struct ip_vs_flags flags = { .flags = svc->flags,
3036 struct ip_vs_kstats kstats;
3039 nl_service = nla_nest_start_noflag(skb, IPVS_CMD_ATTR_SERVICE);
3043 if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
3044 goto nla_put_failure;
3046 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
3047 goto nla_put_failure;
3049 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
3050 nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
3051 nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
3052 goto nla_put_failure;
3055 sched = rcu_dereference_protected(svc->scheduler, 1);
3056 sched_name = sched ? sched->name : "none";
3057 pe = rcu_dereference_protected(svc->pe, 1);
3058 if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched_name) ||
3059 (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
3060 nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
3061 nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
3062 nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
3063 goto nla_put_failure;
3064 ip_vs_copy_stats(&kstats, &svc->stats);
3065 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &kstats))
3066 goto nla_put_failure;
3067 if (ip_vs_genl_fill_stats64(skb, IPVS_SVC_ATTR_STATS64, &kstats))
3068 goto nla_put_failure;
3070 nla_nest_end(skb, nl_service);
3075 nla_nest_cancel(skb, nl_service);
3079 static int ip_vs_genl_dump_service(struct sk_buff *skb,
3080 struct ip_vs_service *svc,
3081 struct netlink_callback *cb)
3085 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3086 &ip_vs_genl_family, NLM_F_MULTI,
3087 IPVS_CMD_NEW_SERVICE);
3091 if (ip_vs_genl_fill_service(skb, svc) < 0)
3092 goto nla_put_failure;
3094 genlmsg_end(skb, hdr);
3098 genlmsg_cancel(skb, hdr);
3102 static int ip_vs_genl_dump_services(struct sk_buff *skb,
3103 struct netlink_callback *cb)
3106 int start = cb->args[0];
3107 struct ip_vs_service *svc;
3108 struct net *net = sock_net(skb->sk);
3109 struct netns_ipvs *ipvs = net_ipvs(net);
3111 mutex_lock(&__ip_vs_mutex);
3112 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3113 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
3114 if (++idx <= start || (svc->ipvs != ipvs))
3116 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3118 goto nla_put_failure;
3123 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3124 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
3125 if (++idx <= start || (svc->ipvs != ipvs))
3127 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3129 goto nla_put_failure;
3135 mutex_unlock(&__ip_vs_mutex);
3141 static bool ip_vs_is_af_valid(int af)
3145 #ifdef CONFIG_IP_VS_IPV6
3146 if (af == AF_INET6 && ipv6_mod_enabled())
3152 static int ip_vs_genl_parse_service(struct netns_ipvs *ipvs,
3153 struct ip_vs_service_user_kern *usvc,
3154 struct nlattr *nla, bool full_entry,
3155 struct ip_vs_service **ret_svc)
3157 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
3158 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
3159 struct ip_vs_service *svc;
3161 /* Parse mandatory identifying service fields first */
3163 nla_parse_nested_deprecated(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy, NULL))
3166 nla_af = attrs[IPVS_SVC_ATTR_AF];
3167 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
3168 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
3169 nla_port = attrs[IPVS_SVC_ATTR_PORT];
3170 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
3172 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3175 memset(usvc, 0, sizeof(*usvc));
3177 usvc->af = nla_get_u16(nla_af);
3178 if (!ip_vs_is_af_valid(usvc->af))
3179 return -EAFNOSUPPORT;
3182 usvc->protocol = IPPROTO_TCP;
3183 usvc->fwmark = nla_get_u32(nla_fwmark);
3185 usvc->protocol = nla_get_u16(nla_protocol);
3186 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3187 usvc->port = nla_get_be16(nla_port);
3193 svc = __ip_vs_svc_fwm_find(ipvs, usvc->af, usvc->fwmark);
3195 svc = __ip_vs_service_find(ipvs, usvc->af, usvc->protocol,
3196 &usvc->addr, usvc->port);
3200 /* If a full entry was requested, check for the additional fields */
3202 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3204 struct ip_vs_flags flags;
3206 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3207 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3208 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3209 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3210 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3212 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3215 nla_memcpy(&flags, nla_flags, sizeof(flags));
3217 /* prefill flags from service if it already exists */
3219 usvc->flags = svc->flags;
3221 /* set new flags from userland */
3222 usvc->flags = (usvc->flags & ~flags.mask) |
3223 (flags.flags & flags.mask);
3224 usvc->sched_name = nla_data(nla_sched);
3225 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3226 usvc->timeout = nla_get_u32(nla_timeout);
3227 usvc->netmask = nla_get_be32(nla_netmask);
3233 static struct ip_vs_service *ip_vs_genl_find_service(struct netns_ipvs *ipvs,
3236 struct ip_vs_service_user_kern usvc;
3237 struct ip_vs_service *svc;
3240 ret = ip_vs_genl_parse_service(ipvs, &usvc, nla, false, &svc);
3241 return ret ? ERR_PTR(ret) : svc;
3244 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3246 struct nlattr *nl_dest;
3247 struct ip_vs_kstats kstats;
3249 nl_dest = nla_nest_start_noflag(skb, IPVS_CMD_ATTR_DEST);
3253 if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3254 nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3255 nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3256 (atomic_read(&dest->conn_flags) &
3257 IP_VS_CONN_F_FWD_MASK)) ||
3258 nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3259 atomic_read(&dest->weight)) ||
3260 nla_put_u8(skb, IPVS_DEST_ATTR_TUN_TYPE,
3262 nla_put_be16(skb, IPVS_DEST_ATTR_TUN_PORT,
3264 nla_put_u16(skb, IPVS_DEST_ATTR_TUN_FLAGS,
3266 nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3267 nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3268 nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3269 atomic_read(&dest->activeconns)) ||
3270 nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3271 atomic_read(&dest->inactconns)) ||
3272 nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3273 atomic_read(&dest->persistconns)) ||
3274 nla_put_u16(skb, IPVS_DEST_ATTR_ADDR_FAMILY, dest->af))
3275 goto nla_put_failure;
3276 ip_vs_copy_stats(&kstats, &dest->stats);
3277 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &kstats))
3278 goto nla_put_failure;
3279 if (ip_vs_genl_fill_stats64(skb, IPVS_DEST_ATTR_STATS64, &kstats))
3280 goto nla_put_failure;
3282 nla_nest_end(skb, nl_dest);
3287 nla_nest_cancel(skb, nl_dest);
3291 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3292 struct netlink_callback *cb)
3296 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3297 &ip_vs_genl_family, NLM_F_MULTI,
3302 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3303 goto nla_put_failure;
3305 genlmsg_end(skb, hdr);
3309 genlmsg_cancel(skb, hdr);
3313 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3314 struct netlink_callback *cb)
3317 int start = cb->args[0];
3318 struct ip_vs_service *svc;
3319 struct ip_vs_dest *dest;
3320 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3321 struct net *net = sock_net(skb->sk);
3322 struct netns_ipvs *ipvs = net_ipvs(net);
3324 mutex_lock(&__ip_vs_mutex);
3326 /* Try to find the service for which to dump destinations */
3327 if (nlmsg_parse_deprecated(cb->nlh, GENL_HDRLEN, attrs, IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy, cb->extack))
3331 svc = ip_vs_genl_find_service(ipvs, attrs[IPVS_CMD_ATTR_SERVICE]);
3332 if (IS_ERR_OR_NULL(svc))
3335 /* Dump the destinations */
3336 list_for_each_entry(dest, &svc->destinations, n_list) {
3339 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3341 goto nla_put_failure;
3349 mutex_unlock(&__ip_vs_mutex);
3354 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3355 struct nlattr *nla, bool full_entry)
3357 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3358 struct nlattr *nla_addr, *nla_port;
3359 struct nlattr *nla_addr_family;
3361 /* Parse mandatory identifying destination fields first */
3363 nla_parse_nested_deprecated(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy, NULL))
3366 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3367 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3368 nla_addr_family = attrs[IPVS_DEST_ATTR_ADDR_FAMILY];
3370 if (!(nla_addr && nla_port))
3373 memset(udest, 0, sizeof(*udest));
3375 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3376 udest->port = nla_get_be16(nla_port);
3378 if (nla_addr_family)
3379 udest->af = nla_get_u16(nla_addr_family);
3383 /* If a full entry was requested, check for the additional fields */
3385 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3386 *nla_l_thresh, *nla_tun_type, *nla_tun_port,
3389 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3390 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3391 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3392 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3393 nla_tun_type = attrs[IPVS_DEST_ATTR_TUN_TYPE];
3394 nla_tun_port = attrs[IPVS_DEST_ATTR_TUN_PORT];
3395 nla_tun_flags = attrs[IPVS_DEST_ATTR_TUN_FLAGS];
3397 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3400 udest->conn_flags = nla_get_u32(nla_fwd)
3401 & IP_VS_CONN_F_FWD_MASK;
3402 udest->weight = nla_get_u32(nla_weight);
3403 udest->u_threshold = nla_get_u32(nla_u_thresh);
3404 udest->l_threshold = nla_get_u32(nla_l_thresh);
3407 udest->tun_type = nla_get_u8(nla_tun_type);
3410 udest->tun_port = nla_get_be16(nla_tun_port);
3413 udest->tun_flags = nla_get_u16(nla_tun_flags);
3419 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3420 struct ipvs_sync_daemon_cfg *c)
3422 struct nlattr *nl_daemon;
3424 nl_daemon = nla_nest_start_noflag(skb, IPVS_CMD_ATTR_DAEMON);
3428 if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3429 nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, c->mcast_ifn) ||
3430 nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, c->syncid) ||
3431 nla_put_u16(skb, IPVS_DAEMON_ATTR_SYNC_MAXLEN, c->sync_maxlen) ||
3432 nla_put_u16(skb, IPVS_DAEMON_ATTR_MCAST_PORT, c->mcast_port) ||
3433 nla_put_u8(skb, IPVS_DAEMON_ATTR_MCAST_TTL, c->mcast_ttl))
3434 goto nla_put_failure;
3435 #ifdef CONFIG_IP_VS_IPV6
3436 if (c->mcast_af == AF_INET6) {
3437 if (nla_put_in6_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP6,
3438 &c->mcast_group.in6))
3439 goto nla_put_failure;
3442 if (c->mcast_af == AF_INET &&
3443 nla_put_in_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP,
3445 goto nla_put_failure;
3446 nla_nest_end(skb, nl_daemon);
3451 nla_nest_cancel(skb, nl_daemon);
3455 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3456 struct ipvs_sync_daemon_cfg *c,
3457 struct netlink_callback *cb)
3460 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3461 &ip_vs_genl_family, NLM_F_MULTI,
3462 IPVS_CMD_NEW_DAEMON);
3466 if (ip_vs_genl_fill_daemon(skb, state, c))
3467 goto nla_put_failure;
3469 genlmsg_end(skb, hdr);
3473 genlmsg_cancel(skb, hdr);
3477 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3478 struct netlink_callback *cb)
3480 struct net *net = sock_net(skb->sk);
3481 struct netns_ipvs *ipvs = net_ipvs(net);
3483 mutex_lock(&ipvs->sync_mutex);
3484 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3485 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3486 &ipvs->mcfg, cb) < 0)
3487 goto nla_put_failure;
3492 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3493 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3494 &ipvs->bcfg, cb) < 0)
3495 goto nla_put_failure;
3501 mutex_unlock(&ipvs->sync_mutex);
3506 static int ip_vs_genl_new_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3508 struct ipvs_sync_daemon_cfg c;
3512 memset(&c, 0, sizeof(c));
3513 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3514 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3515 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3517 strlcpy(c.mcast_ifn, nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3518 sizeof(c.mcast_ifn));
3519 c.syncid = nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]);
3521 a = attrs[IPVS_DAEMON_ATTR_SYNC_MAXLEN];
3523 c.sync_maxlen = nla_get_u16(a);
3525 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP];
3527 c.mcast_af = AF_INET;
3528 c.mcast_group.ip = nla_get_in_addr(a);
3529 if (!ipv4_is_multicast(c.mcast_group.ip))
3532 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP6];
3534 #ifdef CONFIG_IP_VS_IPV6
3537 c.mcast_af = AF_INET6;
3538 c.mcast_group.in6 = nla_get_in6_addr(a);
3539 addr_type = ipv6_addr_type(&c.mcast_group.in6);
3540 if (!(addr_type & IPV6_ADDR_MULTICAST))
3543 return -EAFNOSUPPORT;
3548 a = attrs[IPVS_DAEMON_ATTR_MCAST_PORT];
3550 c.mcast_port = nla_get_u16(a);
3552 a = attrs[IPVS_DAEMON_ATTR_MCAST_TTL];
3554 c.mcast_ttl = nla_get_u8(a);
3556 /* The synchronization protocol is incompatible with mixed family
3559 if (ipvs->mixed_address_family_dests > 0)
3562 ret = start_sync_thread(ipvs, &c,
3563 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3567 static int ip_vs_genl_del_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3571 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3574 ret = stop_sync_thread(ipvs,
3575 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3579 static int ip_vs_genl_set_config(struct netns_ipvs *ipvs, struct nlattr **attrs)
3581 struct ip_vs_timeout_user t;
3583 __ip_vs_get_timeouts(ipvs, &t);
3585 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3586 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3588 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3590 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3592 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3593 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3595 return ip_vs_set_timeout(ipvs, &t);
3598 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3600 int ret = -EINVAL, cmd;
3601 struct net *net = sock_net(skb->sk);
3602 struct netns_ipvs *ipvs = net_ipvs(net);
3604 cmd = info->genlhdr->cmd;
3606 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3607 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3609 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3610 nla_parse_nested_deprecated(daemon_attrs, IPVS_DAEMON_ATTR_MAX, info->attrs[IPVS_CMD_ATTR_DAEMON], ip_vs_daemon_policy, info->extack))
3613 if (cmd == IPVS_CMD_NEW_DAEMON)
3614 ret = ip_vs_genl_new_daemon(ipvs, daemon_attrs);
3616 ret = ip_vs_genl_del_daemon(ipvs, daemon_attrs);
3623 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3625 bool need_full_svc = false, need_full_dest = false;
3626 struct ip_vs_service *svc = NULL;
3627 struct ip_vs_service_user_kern usvc;
3628 struct ip_vs_dest_user_kern udest;
3630 struct net *net = sock_net(skb->sk);
3631 struct netns_ipvs *ipvs = net_ipvs(net);
3633 cmd = info->genlhdr->cmd;
3635 mutex_lock(&__ip_vs_mutex);
3637 if (cmd == IPVS_CMD_FLUSH) {
3638 ret = ip_vs_flush(ipvs, false);
3640 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3641 ret = ip_vs_genl_set_config(ipvs, info->attrs);
3643 } else if (cmd == IPVS_CMD_ZERO &&
3644 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3645 ret = ip_vs_zero_all(ipvs);
3649 /* All following commands require a service argument, so check if we
3650 * received a valid one. We need a full service specification when
3651 * adding / editing a service. Only identifying members otherwise. */
3652 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3653 need_full_svc = true;
3655 ret = ip_vs_genl_parse_service(ipvs, &usvc,
3656 info->attrs[IPVS_CMD_ATTR_SERVICE],
3657 need_full_svc, &svc);
3661 /* Unless we're adding a new service, the service must already exist */
3662 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3667 /* Destination commands require a valid destination argument. For
3668 * adding / editing a destination, we need a full destination
3670 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3671 cmd == IPVS_CMD_DEL_DEST) {
3672 if (cmd != IPVS_CMD_DEL_DEST)
3673 need_full_dest = true;
3675 ret = ip_vs_genl_parse_dest(&udest,
3676 info->attrs[IPVS_CMD_ATTR_DEST],
3681 /* Old protocols did not allow the user to specify address
3682 * family, so we set it to zero instead. We also didn't
3683 * allow heterogeneous pools in the old code, so it's safe
3684 * to assume that this will have the same address family as
3690 if (!ip_vs_is_af_valid(udest.af)) {
3691 ret = -EAFNOSUPPORT;
3695 if (udest.af != svc->af && cmd != IPVS_CMD_DEL_DEST) {
3696 /* The synchronization protocol is incompatible
3697 * with mixed family services
3699 if (ipvs->sync_state) {
3704 /* Which connection types do we support? */
3705 switch (udest.conn_flags) {
3706 case IP_VS_CONN_F_TUNNEL:
3707 /* We are able to forward this */
3717 case IPVS_CMD_NEW_SERVICE:
3719 ret = ip_vs_add_service(ipvs, &usvc, &svc);
3723 case IPVS_CMD_SET_SERVICE:
3724 ret = ip_vs_edit_service(svc, &usvc);
3726 case IPVS_CMD_DEL_SERVICE:
3727 ret = ip_vs_del_service(svc);
3728 /* do not use svc, it can be freed */
3730 case IPVS_CMD_NEW_DEST:
3731 ret = ip_vs_add_dest(svc, &udest);
3733 case IPVS_CMD_SET_DEST:
3734 ret = ip_vs_edit_dest(svc, &udest);
3736 case IPVS_CMD_DEL_DEST:
3737 ret = ip_vs_del_dest(svc, &udest);
3740 ret = ip_vs_zero_service(svc);
3747 mutex_unlock(&__ip_vs_mutex);
3752 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3754 struct sk_buff *msg;
3756 int ret, cmd, reply_cmd;
3757 struct net *net = sock_net(skb->sk);
3758 struct netns_ipvs *ipvs = net_ipvs(net);
3760 cmd = info->genlhdr->cmd;
3762 if (cmd == IPVS_CMD_GET_SERVICE)
3763 reply_cmd = IPVS_CMD_NEW_SERVICE;
3764 else if (cmd == IPVS_CMD_GET_INFO)
3765 reply_cmd = IPVS_CMD_SET_INFO;
3766 else if (cmd == IPVS_CMD_GET_CONFIG)
3767 reply_cmd = IPVS_CMD_SET_CONFIG;
3769 pr_err("unknown Generic Netlink command\n");
3773 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3777 mutex_lock(&__ip_vs_mutex);
3779 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3781 goto nla_put_failure;
3784 case IPVS_CMD_GET_SERVICE:
3786 struct ip_vs_service *svc;
3788 svc = ip_vs_genl_find_service(ipvs,
3789 info->attrs[IPVS_CMD_ATTR_SERVICE]);
3794 ret = ip_vs_genl_fill_service(msg, svc);
3796 goto nla_put_failure;
3805 case IPVS_CMD_GET_CONFIG:
3807 struct ip_vs_timeout_user t;
3809 __ip_vs_get_timeouts(ipvs, &t);
3810 #ifdef CONFIG_IP_VS_PROTO_TCP
3811 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3813 nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3815 goto nla_put_failure;
3817 #ifdef CONFIG_IP_VS_PROTO_UDP
3818 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3819 goto nla_put_failure;
3825 case IPVS_CMD_GET_INFO:
3826 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3827 IP_VS_VERSION_CODE) ||
3828 nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3829 ip_vs_conn_tab_size))
3830 goto nla_put_failure;
3834 genlmsg_end(msg, reply);
3835 ret = genlmsg_reply(msg, info);
3839 pr_err("not enough space in Netlink message\n");
3845 mutex_unlock(&__ip_vs_mutex);
3851 static const struct genl_ops ip_vs_genl_ops[] = {
3853 .cmd = IPVS_CMD_NEW_SERVICE,
3854 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3855 .flags = GENL_ADMIN_PERM,
3856 .doit = ip_vs_genl_set_cmd,
3859 .cmd = IPVS_CMD_SET_SERVICE,
3860 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3861 .flags = GENL_ADMIN_PERM,
3862 .doit = ip_vs_genl_set_cmd,
3865 .cmd = IPVS_CMD_DEL_SERVICE,
3866 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3867 .flags = GENL_ADMIN_PERM,
3868 .doit = ip_vs_genl_set_cmd,
3871 .cmd = IPVS_CMD_GET_SERVICE,
3872 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3873 .flags = GENL_ADMIN_PERM,
3874 .doit = ip_vs_genl_get_cmd,
3875 .dumpit = ip_vs_genl_dump_services,
3878 .cmd = IPVS_CMD_NEW_DEST,
3879 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3880 .flags = GENL_ADMIN_PERM,
3881 .doit = ip_vs_genl_set_cmd,
3884 .cmd = IPVS_CMD_SET_DEST,
3885 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3886 .flags = GENL_ADMIN_PERM,
3887 .doit = ip_vs_genl_set_cmd,
3890 .cmd = IPVS_CMD_DEL_DEST,
3891 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3892 .flags = GENL_ADMIN_PERM,
3893 .doit = ip_vs_genl_set_cmd,
3896 .cmd = IPVS_CMD_GET_DEST,
3897 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3898 .flags = GENL_ADMIN_PERM,
3899 .dumpit = ip_vs_genl_dump_dests,
3902 .cmd = IPVS_CMD_NEW_DAEMON,
3903 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3904 .flags = GENL_ADMIN_PERM,
3905 .doit = ip_vs_genl_set_daemon,
3908 .cmd = IPVS_CMD_DEL_DAEMON,
3909 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3910 .flags = GENL_ADMIN_PERM,
3911 .doit = ip_vs_genl_set_daemon,
3914 .cmd = IPVS_CMD_GET_DAEMON,
3915 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3916 .flags = GENL_ADMIN_PERM,
3917 .dumpit = ip_vs_genl_dump_daemons,
3920 .cmd = IPVS_CMD_SET_CONFIG,
3921 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3922 .flags = GENL_ADMIN_PERM,
3923 .doit = ip_vs_genl_set_cmd,
3926 .cmd = IPVS_CMD_GET_CONFIG,
3927 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3928 .flags = GENL_ADMIN_PERM,
3929 .doit = ip_vs_genl_get_cmd,
3932 .cmd = IPVS_CMD_GET_INFO,
3933 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3934 .flags = GENL_ADMIN_PERM,
3935 .doit = ip_vs_genl_get_cmd,
3938 .cmd = IPVS_CMD_ZERO,
3939 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3940 .flags = GENL_ADMIN_PERM,
3941 .doit = ip_vs_genl_set_cmd,
3944 .cmd = IPVS_CMD_FLUSH,
3945 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3946 .flags = GENL_ADMIN_PERM,
3947 .doit = ip_vs_genl_set_cmd,
3951 static struct genl_family ip_vs_genl_family __ro_after_init = {
3953 .name = IPVS_GENL_NAME,
3954 .version = IPVS_GENL_VERSION,
3955 .maxattr = IPVS_CMD_ATTR_MAX,
3956 .policy = ip_vs_cmd_policy,
3957 .netnsok = true, /* Make ipvsadm to work on netns */
3958 .module = THIS_MODULE,
3959 .ops = ip_vs_genl_ops,
3960 .n_ops = ARRAY_SIZE(ip_vs_genl_ops),
3963 static int __init ip_vs_genl_register(void)
3965 return genl_register_family(&ip_vs_genl_family);
3968 static void ip_vs_genl_unregister(void)
3970 genl_unregister_family(&ip_vs_genl_family);
3973 /* End of Generic Netlink interface definitions */
3976 * per netns intit/exit func.
3978 #ifdef CONFIG_SYSCTL
3979 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs)
3981 struct net *net = ipvs->net;
3983 struct ctl_table *tbl;
3985 atomic_set(&ipvs->dropentry, 0);
3986 spin_lock_init(&ipvs->dropentry_lock);
3987 spin_lock_init(&ipvs->droppacket_lock);
3988 spin_lock_init(&ipvs->securetcp_lock);
3990 if (!net_eq(net, &init_net)) {
3991 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3995 /* Don't export sysctls to unprivileged users */
3996 if (net->user_ns != &init_user_ns)
3997 tbl[0].procname = NULL;
4000 /* Initialize sysctl defaults */
4001 for (idx = 0; idx < ARRAY_SIZE(vs_vars); idx++) {
4002 if (tbl[idx].proc_handler == proc_do_defense_mode)
4003 tbl[idx].extra2 = ipvs;
4006 ipvs->sysctl_amemthresh = 1024;
4007 tbl[idx++].data = &ipvs->sysctl_amemthresh;
4008 ipvs->sysctl_am_droprate = 10;
4009 tbl[idx++].data = &ipvs->sysctl_am_droprate;
4010 tbl[idx++].data = &ipvs->sysctl_drop_entry;
4011 tbl[idx++].data = &ipvs->sysctl_drop_packet;
4012 #ifdef CONFIG_IP_VS_NFCT
4013 tbl[idx++].data = &ipvs->sysctl_conntrack;
4015 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
4016 ipvs->sysctl_snat_reroute = 1;
4017 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
4018 ipvs->sysctl_sync_ver = 1;
4019 tbl[idx++].data = &ipvs->sysctl_sync_ver;
4020 ipvs->sysctl_sync_ports = 1;
4021 tbl[idx++].data = &ipvs->sysctl_sync_ports;
4022 tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
4023 ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
4024 tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
4025 ipvs->sysctl_sync_sock_size = 0;
4026 tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
4027 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
4028 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
4029 tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
4030 tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
4031 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
4032 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
4033 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
4034 tbl[idx].data = &ipvs->sysctl_sync_threshold;
4035 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
4036 ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
4037 tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
4038 ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
4039 tbl[idx++].data = &ipvs->sysctl_sync_retries;
4040 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
4041 ipvs->sysctl_pmtu_disc = 1;
4042 tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
4043 tbl[idx++].data = &ipvs->sysctl_backup_only;
4044 ipvs->sysctl_conn_reuse_mode = 1;
4045 tbl[idx++].data = &ipvs->sysctl_conn_reuse_mode;
4046 tbl[idx++].data = &ipvs->sysctl_schedule_icmp;
4047 tbl[idx++].data = &ipvs->sysctl_ignore_tunneled;
4049 ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
4050 if (ipvs->sysctl_hdr == NULL) {
4051 if (!net_eq(net, &init_net))
4055 ip_vs_start_estimator(ipvs, &ipvs->tot_stats);
4056 ipvs->sysctl_tbl = tbl;
4057 /* Schedule defense work */
4058 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
4059 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
4064 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs)
4066 struct net *net = ipvs->net;
4068 cancel_delayed_work_sync(&ipvs->defense_work);
4069 cancel_work_sync(&ipvs->defense_work.work);
4070 unregister_net_sysctl_table(ipvs->sysctl_hdr);
4071 ip_vs_stop_estimator(ipvs, &ipvs->tot_stats);
4073 if (!net_eq(net, &init_net))
4074 kfree(ipvs->sysctl_tbl);
4079 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs) { return 0; }
4080 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs) { }
4084 static struct notifier_block ip_vs_dst_notifier = {
4085 .notifier_call = ip_vs_dst_event,
4086 #ifdef CONFIG_IP_VS_IPV6
4087 .priority = ADDRCONF_NOTIFY_PRIORITY + 5,
4091 int __net_init ip_vs_control_net_init(struct netns_ipvs *ipvs)
4095 /* Initialize rs_table */
4096 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
4097 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
4099 INIT_LIST_HEAD(&ipvs->dest_trash);
4100 spin_lock_init(&ipvs->dest_trash_lock);
4101 timer_setup(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire, 0);
4102 atomic_set(&ipvs->ftpsvc_counter, 0);
4103 atomic_set(&ipvs->nullsvc_counter, 0);
4104 atomic_set(&ipvs->conn_out_counter, 0);
4107 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
4108 if (!ipvs->tot_stats.cpustats)
4111 for_each_possible_cpu(i) {
4112 struct ip_vs_cpu_stats *ipvs_tot_stats;
4113 ipvs_tot_stats = per_cpu_ptr(ipvs->tot_stats.cpustats, i);
4114 u64_stats_init(&ipvs_tot_stats->syncp);
4117 spin_lock_init(&ipvs->tot_stats.lock);
4119 proc_create_net("ip_vs", 0, ipvs->net->proc_net, &ip_vs_info_seq_ops,
4120 sizeof(struct ip_vs_iter));
4121 proc_create_net_single("ip_vs_stats", 0, ipvs->net->proc_net,
4122 ip_vs_stats_show, NULL);
4123 proc_create_net_single("ip_vs_stats_percpu", 0, ipvs->net->proc_net,
4124 ip_vs_stats_percpu_show, NULL);
4126 if (ip_vs_control_net_init_sysctl(ipvs))
4132 free_percpu(ipvs->tot_stats.cpustats);
4136 void __net_exit ip_vs_control_net_cleanup(struct netns_ipvs *ipvs)
4138 ip_vs_trash_cleanup(ipvs);
4139 ip_vs_control_net_cleanup_sysctl(ipvs);
4140 remove_proc_entry("ip_vs_stats_percpu", ipvs->net->proc_net);
4141 remove_proc_entry("ip_vs_stats", ipvs->net->proc_net);
4142 remove_proc_entry("ip_vs", ipvs->net->proc_net);
4143 free_percpu(ipvs->tot_stats.cpustats);
4146 int __init ip_vs_register_nl_ioctl(void)
4150 ret = nf_register_sockopt(&ip_vs_sockopts);
4152 pr_err("cannot register sockopt.\n");
4156 ret = ip_vs_genl_register();
4158 pr_err("cannot register Generic Netlink interface.\n");
4164 nf_unregister_sockopt(&ip_vs_sockopts);
4169 void ip_vs_unregister_nl_ioctl(void)
4171 ip_vs_genl_unregister();
4172 nf_unregister_sockopt(&ip_vs_sockopts);
4175 int __init ip_vs_control_init(void)
4182 /* Initialize svc_table, ip_vs_svc_fwm_table */
4183 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
4184 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
4185 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
4188 smp_wmb(); /* Do we really need it now ? */
4190 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
4199 void ip_vs_control_cleanup(void)
4202 unregister_netdevice_notifier(&ip_vs_dst_notifier);