2 * IPVS An implementation of the IP virtual server support for the
3 * LINUX operating system. IPVS is now implemented as a module
4 * over the NetFilter framework. IPVS can be used to build a
5 * high-performance and highly available server based on a
8 * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
9 * Peter Kese <peter.kese@ijs.si>
10 * Julian Anastasov <ja@ssi.bg>
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
21 #define KMSG_COMPONENT "IPVS"
22 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
24 #include <linux/module.h>
25 #include <linux/init.h>
26 #include <linux/types.h>
27 #include <linux/capability.h>
29 #include <linux/sysctl.h>
30 #include <linux/proc_fs.h>
31 #include <linux/workqueue.h>
32 #include <linux/swap.h>
33 #include <linux/seq_file.h>
34 #include <linux/slab.h>
36 #include <linux/netfilter.h>
37 #include <linux/netfilter_ipv4.h>
38 #include <linux/mutex.h>
40 #include <net/net_namespace.h>
41 #include <linux/nsproxy.h>
43 #ifdef CONFIG_IP_VS_IPV6
45 #include <net/ip6_route.h>
47 #include <net/route.h>
49 #include <net/genetlink.h>
51 #include <linux/uaccess.h>
53 #include <net/ip_vs.h>
55 /* semaphore for IPVS sockopts. And, [gs]etsockopt may sleep. */
56 static DEFINE_MUTEX(__ip_vs_mutex
);
58 /* sysctl variables */
60 #ifdef CONFIG_IP_VS_DEBUG
61 static int sysctl_ip_vs_debug_level
= 0;
63 int ip_vs_get_debug_level(void)
65 return sysctl_ip_vs_debug_level
;
71 static void __ip_vs_del_service(struct ip_vs_service
*svc
, bool cleanup
);
74 #ifdef CONFIG_IP_VS_IPV6
75 /* Taken from rt6_fill_node() in net/ipv6/route.c, is there a better way? */
76 static bool __ip_vs_addr_is_local_v6(struct net
*net
,
77 const struct in6_addr
*addr
)
82 struct dst_entry
*dst
= ip6_route_output(net
, NULL
, &fl6
);
85 is_local
= !dst
->error
&& dst
->dev
&& (dst
->dev
->flags
& IFF_LOOPBACK
);
94 * update_defense_level is called from keventd and from sysctl,
95 * so it needs to protect itself from softirqs
97 static void update_defense_level(struct netns_ipvs
*ipvs
)
100 static int old_secure_tcp
= 0;
105 /* we only count free and buffered memory (in pages) */
107 availmem
= i
.freeram
+ i
.bufferram
;
108 /* however in linux 2.5 the i.bufferram is total page cache size,
110 /* si_swapinfo(&i); */
111 /* availmem = availmem - (i.totalswap - i.freeswap); */
113 nomem
= (availmem
< ipvs
->sysctl_amemthresh
);
118 spin_lock(&ipvs
->dropentry_lock
);
119 switch (ipvs
->sysctl_drop_entry
) {
121 atomic_set(&ipvs
->dropentry
, 0);
125 atomic_set(&ipvs
->dropentry
, 1);
126 ipvs
->sysctl_drop_entry
= 2;
128 atomic_set(&ipvs
->dropentry
, 0);
133 atomic_set(&ipvs
->dropentry
, 1);
135 atomic_set(&ipvs
->dropentry
, 0);
136 ipvs
->sysctl_drop_entry
= 1;
140 atomic_set(&ipvs
->dropentry
, 1);
143 spin_unlock(&ipvs
->dropentry_lock
);
146 spin_lock(&ipvs
->droppacket_lock
);
147 switch (ipvs
->sysctl_drop_packet
) {
153 ipvs
->drop_rate
= ipvs
->drop_counter
154 = ipvs
->sysctl_amemthresh
/
155 (ipvs
->sysctl_amemthresh
-availmem
);
156 ipvs
->sysctl_drop_packet
= 2;
163 ipvs
->drop_rate
= ipvs
->drop_counter
164 = ipvs
->sysctl_amemthresh
/
165 (ipvs
->sysctl_amemthresh
-availmem
);
168 ipvs
->sysctl_drop_packet
= 1;
172 ipvs
->drop_rate
= ipvs
->sysctl_am_droprate
;
175 spin_unlock(&ipvs
->droppacket_lock
);
178 spin_lock(&ipvs
->securetcp_lock
);
179 switch (ipvs
->sysctl_secure_tcp
) {
181 if (old_secure_tcp
>= 2)
186 if (old_secure_tcp
< 2)
188 ipvs
->sysctl_secure_tcp
= 2;
190 if (old_secure_tcp
>= 2)
196 if (old_secure_tcp
< 2)
199 if (old_secure_tcp
>= 2)
201 ipvs
->sysctl_secure_tcp
= 1;
205 if (old_secure_tcp
< 2)
209 old_secure_tcp
= ipvs
->sysctl_secure_tcp
;
211 ip_vs_protocol_timeout_change(ipvs
,
212 ipvs
->sysctl_secure_tcp
> 1);
213 spin_unlock(&ipvs
->securetcp_lock
);
220 * Timer for checking the defense
222 #define DEFENSE_TIMER_PERIOD 1*HZ
224 static void defense_work_handler(struct work_struct
*work
)
226 struct netns_ipvs
*ipvs
=
227 container_of(work
, struct netns_ipvs
, defense_work
.work
);
229 update_defense_level(ipvs
);
230 if (atomic_read(&ipvs
->dropentry
))
231 ip_vs_random_dropentry(ipvs
);
232 schedule_delayed_work(&ipvs
->defense_work
, DEFENSE_TIMER_PERIOD
);
237 ip_vs_use_count_inc(void)
239 return try_module_get(THIS_MODULE
);
243 ip_vs_use_count_dec(void)
245 module_put(THIS_MODULE
);
250 * Hash table: for virtual service lookups
252 #define IP_VS_SVC_TAB_BITS 8
253 #define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
254 #define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)
256 /* the service table hashed by <protocol, addr, port> */
257 static struct hlist_head ip_vs_svc_table
[IP_VS_SVC_TAB_SIZE
];
258 /* the service table hashed by fwmark */
259 static struct hlist_head ip_vs_svc_fwm_table
[IP_VS_SVC_TAB_SIZE
];
263 * Returns hash value for virtual service
265 static inline unsigned int
266 ip_vs_svc_hashkey(struct netns_ipvs
*ipvs
, int af
, unsigned int proto
,
267 const union nf_inet_addr
*addr
, __be16 port
)
269 register unsigned int porth
= ntohs(port
);
270 __be32 addr_fold
= addr
->ip
;
273 #ifdef CONFIG_IP_VS_IPV6
275 addr_fold
= addr
->ip6
[0]^addr
->ip6
[1]^
276 addr
->ip6
[2]^addr
->ip6
[3];
278 ahash
= ntohl(addr_fold
);
279 ahash
^= ((size_t) ipvs
>> 8);
281 return (proto
^ ahash
^ (porth
>> IP_VS_SVC_TAB_BITS
) ^ porth
) &
286 * Returns hash value of fwmark for virtual service lookup
288 static inline unsigned int ip_vs_svc_fwm_hashkey(struct netns_ipvs
*ipvs
, __u32 fwmark
)
290 return (((size_t)ipvs
>>8) ^ fwmark
) & IP_VS_SVC_TAB_MASK
;
294 * Hashes a service in the ip_vs_svc_table by <netns,proto,addr,port>
295 * or in the ip_vs_svc_fwm_table by fwmark.
296 * Should be called with locked tables.
298 static int ip_vs_svc_hash(struct ip_vs_service
*svc
)
302 if (svc
->flags
& IP_VS_SVC_F_HASHED
) {
303 pr_err("%s(): request for already hashed, called from %pF\n",
304 __func__
, __builtin_return_address(0));
308 if (svc
->fwmark
== 0) {
310 * Hash it by <netns,protocol,addr,port> in ip_vs_svc_table
312 hash
= ip_vs_svc_hashkey(svc
->ipvs
, svc
->af
, svc
->protocol
,
313 &svc
->addr
, svc
->port
);
314 hlist_add_head_rcu(&svc
->s_list
, &ip_vs_svc_table
[hash
]);
317 * Hash it by fwmark in svc_fwm_table
319 hash
= ip_vs_svc_fwm_hashkey(svc
->ipvs
, svc
->fwmark
);
320 hlist_add_head_rcu(&svc
->f_list
, &ip_vs_svc_fwm_table
[hash
]);
323 svc
->flags
|= IP_VS_SVC_F_HASHED
;
324 /* increase its refcnt because it is referenced by the svc table */
325 atomic_inc(&svc
->refcnt
);
331 * Unhashes a service from svc_table / svc_fwm_table.
332 * Should be called with locked tables.
334 static int ip_vs_svc_unhash(struct ip_vs_service
*svc
)
336 if (!(svc
->flags
& IP_VS_SVC_F_HASHED
)) {
337 pr_err("%s(): request for unhash flagged, called from %pF\n",
338 __func__
, __builtin_return_address(0));
342 if (svc
->fwmark
== 0) {
343 /* Remove it from the svc_table table */
344 hlist_del_rcu(&svc
->s_list
);
346 /* Remove it from the svc_fwm_table table */
347 hlist_del_rcu(&svc
->f_list
);
350 svc
->flags
&= ~IP_VS_SVC_F_HASHED
;
351 atomic_dec(&svc
->refcnt
);
357 * Get service by {netns, proto,addr,port} in the service table.
359 static inline struct ip_vs_service
*
360 __ip_vs_service_find(struct netns_ipvs
*ipvs
, int af
, __u16 protocol
,
361 const union nf_inet_addr
*vaddr
, __be16 vport
)
364 struct ip_vs_service
*svc
;
366 /* Check for "full" addressed entries */
367 hash
= ip_vs_svc_hashkey(ipvs
, af
, protocol
, vaddr
, vport
);
369 hlist_for_each_entry_rcu(svc
, &ip_vs_svc_table
[hash
], s_list
) {
371 && ip_vs_addr_equal(af
, &svc
->addr
, vaddr
)
372 && (svc
->port
== vport
)
373 && (svc
->protocol
== protocol
)
374 && (svc
->ipvs
== ipvs
)) {
385 * Get service by {fwmark} in the service table.
387 static inline struct ip_vs_service
*
388 __ip_vs_svc_fwm_find(struct netns_ipvs
*ipvs
, int af
, __u32 fwmark
)
391 struct ip_vs_service
*svc
;
393 /* Check for fwmark addressed entries */
394 hash
= ip_vs_svc_fwm_hashkey(ipvs
, fwmark
);
396 hlist_for_each_entry_rcu(svc
, &ip_vs_svc_fwm_table
[hash
], f_list
) {
397 if (svc
->fwmark
== fwmark
&& svc
->af
== af
398 && (svc
->ipvs
== ipvs
)) {
407 /* Find service, called under RCU lock */
408 struct ip_vs_service
*
409 ip_vs_service_find(struct netns_ipvs
*ipvs
, int af
, __u32 fwmark
, __u16 protocol
,
410 const union nf_inet_addr
*vaddr
, __be16 vport
)
412 struct ip_vs_service
*svc
;
415 * Check the table hashed by fwmark first
418 svc
= __ip_vs_svc_fwm_find(ipvs
, af
, fwmark
);
424 * Check the table hashed by <protocol,addr,port>
425 * for "full" addressed entries
427 svc
= __ip_vs_service_find(ipvs
, af
, protocol
, vaddr
, vport
);
429 if (!svc
&& protocol
== IPPROTO_TCP
&&
430 atomic_read(&ipvs
->ftpsvc_counter
) &&
431 (vport
== FTPDATA
|| ntohs(vport
) >= inet_prot_sock(ipvs
->net
))) {
433 * Check if ftp service entry exists, the packet
434 * might belong to FTP data connections.
436 svc
= __ip_vs_service_find(ipvs
, af
, protocol
, vaddr
, FTPPORT
);
440 && atomic_read(&ipvs
->nullsvc_counter
)) {
442 * Check if the catch-all port (port zero) exists
444 svc
= __ip_vs_service_find(ipvs
, af
, protocol
, vaddr
, 0);
448 IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
449 fwmark
, ip_vs_proto_name(protocol
),
450 IP_VS_DBG_ADDR(af
, vaddr
), ntohs(vport
),
451 svc
? "hit" : "not hit");
458 __ip_vs_bind_svc(struct ip_vs_dest
*dest
, struct ip_vs_service
*svc
)
460 atomic_inc(&svc
->refcnt
);
461 rcu_assign_pointer(dest
->svc
, svc
);
464 static void ip_vs_service_free(struct ip_vs_service
*svc
)
466 free_percpu(svc
->stats
.cpustats
);
470 static void ip_vs_service_rcu_free(struct rcu_head
*head
)
472 struct ip_vs_service
*svc
;
474 svc
= container_of(head
, struct ip_vs_service
, rcu_head
);
475 ip_vs_service_free(svc
);
478 static void __ip_vs_svc_put(struct ip_vs_service
*svc
, bool do_delay
)
480 if (atomic_dec_and_test(&svc
->refcnt
)) {
481 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n",
483 IP_VS_DBG_ADDR(svc
->af
, &svc
->addr
),
486 call_rcu(&svc
->rcu_head
, ip_vs_service_rcu_free
);
488 ip_vs_service_free(svc
);
494 * Returns hash value for real service
496 static inline unsigned int ip_vs_rs_hashkey(int af
,
497 const union nf_inet_addr
*addr
,
500 register unsigned int porth
= ntohs(port
);
501 __be32 addr_fold
= addr
->ip
;
503 #ifdef CONFIG_IP_VS_IPV6
505 addr_fold
= addr
->ip6
[0]^addr
->ip6
[1]^
506 addr
->ip6
[2]^addr
->ip6
[3];
509 return (ntohl(addr_fold
)^(porth
>>IP_VS_RTAB_BITS
)^porth
)
513 /* Hash ip_vs_dest in rs_table by <proto,addr,port>. */
514 static void ip_vs_rs_hash(struct netns_ipvs
*ipvs
, struct ip_vs_dest
*dest
)
518 if (dest
->in_rs_table
)
522 * Hash by proto,addr,port,
523 * which are the parameters of the real service.
525 hash
= ip_vs_rs_hashkey(dest
->af
, &dest
->addr
, dest
->port
);
527 hlist_add_head_rcu(&dest
->d_list
, &ipvs
->rs_table
[hash
]);
528 dest
->in_rs_table
= 1;
531 /* Unhash ip_vs_dest from rs_table. */
532 static void ip_vs_rs_unhash(struct ip_vs_dest
*dest
)
535 * Remove it from the rs_table table.
537 if (dest
->in_rs_table
) {
538 hlist_del_rcu(&dest
->d_list
);
539 dest
->in_rs_table
= 0;
543 /* Check if real service by <proto,addr,port> is present */
544 bool ip_vs_has_real_service(struct netns_ipvs
*ipvs
, int af
, __u16 protocol
,
545 const union nf_inet_addr
*daddr
, __be16 dport
)
548 struct ip_vs_dest
*dest
;
550 /* Check for "full" addressed entries */
551 hash
= ip_vs_rs_hashkey(af
, daddr
, dport
);
554 hlist_for_each_entry_rcu(dest
, &ipvs
->rs_table
[hash
], d_list
) {
555 if (dest
->port
== dport
&&
557 ip_vs_addr_equal(af
, &dest
->addr
, daddr
) &&
558 (dest
->protocol
== protocol
|| dest
->vfwmark
)) {
569 /* Find real service record by <proto,addr,port>.
570 * In case of multiple records with the same <proto,addr,port>, only
571 * the first found record is returned.
573 * To be called under RCU lock.
575 struct ip_vs_dest
*ip_vs_find_real_service(struct netns_ipvs
*ipvs
, int af
,
577 const union nf_inet_addr
*daddr
,
581 struct ip_vs_dest
*dest
;
583 /* Check for "full" addressed entries */
584 hash
= ip_vs_rs_hashkey(af
, daddr
, dport
);
586 hlist_for_each_entry_rcu(dest
, &ipvs
->rs_table
[hash
], d_list
) {
587 if (dest
->port
== dport
&&
589 ip_vs_addr_equal(af
, &dest
->addr
, daddr
) &&
590 (dest
->protocol
== protocol
|| dest
->vfwmark
)) {
599 /* Lookup destination by {addr,port} in the given service
600 * Called under RCU lock.
602 static struct ip_vs_dest
*
603 ip_vs_lookup_dest(struct ip_vs_service
*svc
, int dest_af
,
604 const union nf_inet_addr
*daddr
, __be16 dport
)
606 struct ip_vs_dest
*dest
;
609 * Find the destination for the given service
611 list_for_each_entry_rcu(dest
, &svc
->destinations
, n_list
) {
612 if ((dest
->af
== dest_af
) &&
613 ip_vs_addr_equal(dest_af
, &dest
->addr
, daddr
) &&
614 (dest
->port
== dport
)) {
624 * Find destination by {daddr,dport,vaddr,protocol}
625 * Created to be used in ip_vs_process_message() in
626 * the backup synchronization daemon. It finds the
627 * destination to be bound to the received connection
629 * Called under RCU lock, no refcnt is returned.
631 struct ip_vs_dest
*ip_vs_find_dest(struct netns_ipvs
*ipvs
, int svc_af
, int dest_af
,
632 const union nf_inet_addr
*daddr
,
634 const union nf_inet_addr
*vaddr
,
635 __be16 vport
, __u16 protocol
, __u32 fwmark
,
638 struct ip_vs_dest
*dest
;
639 struct ip_vs_service
*svc
;
642 svc
= ip_vs_service_find(ipvs
, svc_af
, fwmark
, protocol
, vaddr
, vport
);
645 if (fwmark
&& (flags
& IP_VS_CONN_F_FWD_MASK
) != IP_VS_CONN_F_MASQ
)
647 dest
= ip_vs_lookup_dest(svc
, dest_af
, daddr
, port
);
649 dest
= ip_vs_lookup_dest(svc
, dest_af
, daddr
, port
^ dport
);
653 void ip_vs_dest_dst_rcu_free(struct rcu_head
*head
)
655 struct ip_vs_dest_dst
*dest_dst
= container_of(head
,
656 struct ip_vs_dest_dst
,
659 dst_release(dest_dst
->dst_cache
);
663 /* Release dest_dst and dst_cache for dest in user context */
664 static void __ip_vs_dst_cache_reset(struct ip_vs_dest
*dest
)
666 struct ip_vs_dest_dst
*old
;
668 old
= rcu_dereference_protected(dest
->dest_dst
, 1);
670 RCU_INIT_POINTER(dest
->dest_dst
, NULL
);
671 call_rcu(&old
->rcu_head
, ip_vs_dest_dst_rcu_free
);
676 * Lookup dest by {svc,addr,port} in the destination trash.
677 * The destination trash is used to hold the destinations that are removed
678 * from the service table but are still referenced by some conn entries.
679 * The reason to add the destination trash is when the dest is temporary
680 * down (either by administrator or by monitor program), the dest can be
681 * picked back from the trash, the remaining connections to the dest can
682 * continue, and the counting information of the dest is also useful for
685 static struct ip_vs_dest
*
686 ip_vs_trash_get_dest(struct ip_vs_service
*svc
, int dest_af
,
687 const union nf_inet_addr
*daddr
, __be16 dport
)
689 struct ip_vs_dest
*dest
;
690 struct netns_ipvs
*ipvs
= svc
->ipvs
;
693 * Find the destination in trash
695 spin_lock_bh(&ipvs
->dest_trash_lock
);
696 list_for_each_entry(dest
, &ipvs
->dest_trash
, t_list
) {
697 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
700 IP_VS_DBG_ADDR(dest
->af
, &dest
->addr
),
702 refcount_read(&dest
->refcnt
));
703 if (dest
->af
== dest_af
&&
704 ip_vs_addr_equal(dest_af
, &dest
->addr
, daddr
) &&
705 dest
->port
== dport
&&
706 dest
->vfwmark
== svc
->fwmark
&&
707 dest
->protocol
== svc
->protocol
&&
709 (ip_vs_addr_equal(svc
->af
, &dest
->vaddr
, &svc
->addr
) &&
710 dest
->vport
== svc
->port
))) {
712 list_del(&dest
->t_list
);
720 spin_unlock_bh(&ipvs
->dest_trash_lock
);
725 static void ip_vs_dest_free(struct ip_vs_dest
*dest
)
727 struct ip_vs_service
*svc
= rcu_dereference_protected(dest
->svc
, 1);
729 __ip_vs_dst_cache_reset(dest
);
730 __ip_vs_svc_put(svc
, false);
731 free_percpu(dest
->stats
.cpustats
);
732 ip_vs_dest_put_and_free(dest
);
736 * Clean up all the destinations in the trash
737 * Called by the ip_vs_control_cleanup()
739 * When the ip_vs_control_clearup is activated by ipvs module exit,
740 * the service tables must have been flushed and all the connections
741 * are expired, and the refcnt of each destination in the trash must
742 * be 1, so we simply release them here.
744 static void ip_vs_trash_cleanup(struct netns_ipvs
*ipvs
)
746 struct ip_vs_dest
*dest
, *nxt
;
748 del_timer_sync(&ipvs
->dest_trash_timer
);
749 /* No need to use dest_trash_lock */
750 list_for_each_entry_safe(dest
, nxt
, &ipvs
->dest_trash
, t_list
) {
751 list_del(&dest
->t_list
);
752 ip_vs_dest_free(dest
);
757 ip_vs_copy_stats(struct ip_vs_kstats
*dst
, struct ip_vs_stats
*src
)
759 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->kstats.c - src->kstats0.c
761 spin_lock_bh(&src
->lock
);
763 IP_VS_SHOW_STATS_COUNTER(conns
);
764 IP_VS_SHOW_STATS_COUNTER(inpkts
);
765 IP_VS_SHOW_STATS_COUNTER(outpkts
);
766 IP_VS_SHOW_STATS_COUNTER(inbytes
);
767 IP_VS_SHOW_STATS_COUNTER(outbytes
);
769 ip_vs_read_estimator(dst
, src
);
771 spin_unlock_bh(&src
->lock
);
775 ip_vs_export_stats_user(struct ip_vs_stats_user
*dst
, struct ip_vs_kstats
*src
)
777 dst
->conns
= (u32
)src
->conns
;
778 dst
->inpkts
= (u32
)src
->inpkts
;
779 dst
->outpkts
= (u32
)src
->outpkts
;
780 dst
->inbytes
= src
->inbytes
;
781 dst
->outbytes
= src
->outbytes
;
782 dst
->cps
= (u32
)src
->cps
;
783 dst
->inpps
= (u32
)src
->inpps
;
784 dst
->outpps
= (u32
)src
->outpps
;
785 dst
->inbps
= (u32
)src
->inbps
;
786 dst
->outbps
= (u32
)src
->outbps
;
790 ip_vs_zero_stats(struct ip_vs_stats
*stats
)
792 spin_lock_bh(&stats
->lock
);
794 /* get current counters as zero point, rates are zeroed */
796 #define IP_VS_ZERO_STATS_COUNTER(c) stats->kstats0.c = stats->kstats.c
798 IP_VS_ZERO_STATS_COUNTER(conns
);
799 IP_VS_ZERO_STATS_COUNTER(inpkts
);
800 IP_VS_ZERO_STATS_COUNTER(outpkts
);
801 IP_VS_ZERO_STATS_COUNTER(inbytes
);
802 IP_VS_ZERO_STATS_COUNTER(outbytes
);
804 ip_vs_zero_estimator(stats
);
806 spin_unlock_bh(&stats
->lock
);
810 * Update a destination in the given service
813 __ip_vs_update_dest(struct ip_vs_service
*svc
, struct ip_vs_dest
*dest
,
814 struct ip_vs_dest_user_kern
*udest
, int add
)
816 struct netns_ipvs
*ipvs
= svc
->ipvs
;
817 struct ip_vs_service
*old_svc
;
818 struct ip_vs_scheduler
*sched
;
821 /* We cannot modify an address and change the address family */
822 BUG_ON(!add
&& udest
->af
!= dest
->af
);
824 if (add
&& udest
->af
!= svc
->af
)
825 ipvs
->mixed_address_family_dests
++;
827 /* set the weight and the flags */
828 atomic_set(&dest
->weight
, udest
->weight
);
829 conn_flags
= udest
->conn_flags
& IP_VS_CONN_F_DEST_MASK
;
830 conn_flags
|= IP_VS_CONN_F_INACTIVE
;
832 /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
833 if ((conn_flags
& IP_VS_CONN_F_FWD_MASK
) != IP_VS_CONN_F_MASQ
) {
834 conn_flags
|= IP_VS_CONN_F_NOOUTPUT
;
837 * Put the real service in rs_table if not present.
838 * For now only for NAT!
840 ip_vs_rs_hash(ipvs
, dest
);
842 atomic_set(&dest
->conn_flags
, conn_flags
);
844 /* bind the service */
845 old_svc
= rcu_dereference_protected(dest
->svc
, 1);
847 __ip_vs_bind_svc(dest
, svc
);
849 if (old_svc
!= svc
) {
850 ip_vs_zero_stats(&dest
->stats
);
851 __ip_vs_bind_svc(dest
, svc
);
852 __ip_vs_svc_put(old_svc
, true);
856 /* set the dest status flags */
857 dest
->flags
|= IP_VS_DEST_F_AVAILABLE
;
859 if (udest
->u_threshold
== 0 || udest
->u_threshold
> dest
->u_threshold
)
860 dest
->flags
&= ~IP_VS_DEST_F_OVERLOAD
;
861 dest
->u_threshold
= udest
->u_threshold
;
862 dest
->l_threshold
= udest
->l_threshold
;
864 dest
->af
= udest
->af
;
866 spin_lock_bh(&dest
->dst_lock
);
867 __ip_vs_dst_cache_reset(dest
);
868 spin_unlock_bh(&dest
->dst_lock
);
871 ip_vs_start_estimator(svc
->ipvs
, &dest
->stats
);
872 list_add_rcu(&dest
->n_list
, &svc
->destinations
);
874 sched
= rcu_dereference_protected(svc
->scheduler
, 1);
875 if (sched
&& sched
->add_dest
)
876 sched
->add_dest(svc
, dest
);
878 sched
= rcu_dereference_protected(svc
->scheduler
, 1);
879 if (sched
&& sched
->upd_dest
)
880 sched
->upd_dest(svc
, dest
);
886 * Create a destination for the given service
889 ip_vs_new_dest(struct ip_vs_service
*svc
, struct ip_vs_dest_user_kern
*udest
,
890 struct ip_vs_dest
**dest_p
)
892 struct ip_vs_dest
*dest
;
893 unsigned int atype
, i
;
897 #ifdef CONFIG_IP_VS_IPV6
898 if (udest
->af
== AF_INET6
) {
899 atype
= ipv6_addr_type(&udest
->addr
.in6
);
900 if ((!(atype
& IPV6_ADDR_UNICAST
) ||
901 atype
& IPV6_ADDR_LINKLOCAL
) &&
902 !__ip_vs_addr_is_local_v6(svc
->ipvs
->net
, &udest
->addr
.in6
))
907 atype
= inet_addr_type(svc
->ipvs
->net
, udest
->addr
.ip
);
908 if (atype
!= RTN_LOCAL
&& atype
!= RTN_UNICAST
)
912 dest
= kzalloc(sizeof(struct ip_vs_dest
), GFP_KERNEL
);
916 dest
->stats
.cpustats
= alloc_percpu(struct ip_vs_cpu_stats
);
917 if (!dest
->stats
.cpustats
)
920 for_each_possible_cpu(i
) {
921 struct ip_vs_cpu_stats
*ip_vs_dest_stats
;
922 ip_vs_dest_stats
= per_cpu_ptr(dest
->stats
.cpustats
, i
);
923 u64_stats_init(&ip_vs_dest_stats
->syncp
);
926 dest
->af
= udest
->af
;
927 dest
->protocol
= svc
->protocol
;
928 dest
->vaddr
= svc
->addr
;
929 dest
->vport
= svc
->port
;
930 dest
->vfwmark
= svc
->fwmark
;
931 ip_vs_addr_copy(udest
->af
, &dest
->addr
, &udest
->addr
);
932 dest
->port
= udest
->port
;
934 atomic_set(&dest
->activeconns
, 0);
935 atomic_set(&dest
->inactconns
, 0);
936 atomic_set(&dest
->persistconns
, 0);
937 refcount_set(&dest
->refcnt
, 1);
939 INIT_HLIST_NODE(&dest
->d_list
);
940 spin_lock_init(&dest
->dst_lock
);
941 spin_lock_init(&dest
->stats
.lock
);
942 __ip_vs_update_dest(svc
, dest
, udest
, 1);
956 * Add a destination into an existing service
959 ip_vs_add_dest(struct ip_vs_service
*svc
, struct ip_vs_dest_user_kern
*udest
)
961 struct ip_vs_dest
*dest
;
962 union nf_inet_addr daddr
;
963 __be16 dport
= udest
->port
;
968 if (udest
->weight
< 0) {
969 pr_err("%s(): server weight less than zero\n", __func__
);
973 if (udest
->l_threshold
> udest
->u_threshold
) {
974 pr_err("%s(): lower threshold is higher than upper threshold\n",
979 ip_vs_addr_copy(udest
->af
, &daddr
, &udest
->addr
);
981 /* We use function that requires RCU lock */
983 dest
= ip_vs_lookup_dest(svc
, udest
->af
, &daddr
, dport
);
987 IP_VS_DBG(1, "%s(): dest already exists\n", __func__
);
992 * Check if the dest already exists in the trash and
993 * is from the same service
995 dest
= ip_vs_trash_get_dest(svc
, udest
->af
, &daddr
, dport
);
998 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
999 "dest->refcnt=%d, service %u/%s:%u\n",
1000 IP_VS_DBG_ADDR(udest
->af
, &daddr
), ntohs(dport
),
1001 refcount_read(&dest
->refcnt
),
1003 IP_VS_DBG_ADDR(svc
->af
, &dest
->vaddr
),
1004 ntohs(dest
->vport
));
1006 __ip_vs_update_dest(svc
, dest
, udest
, 1);
1010 * Allocate and initialize the dest structure
1012 ret
= ip_vs_new_dest(svc
, udest
, &dest
);
1021 * Edit a destination in the given service
1024 ip_vs_edit_dest(struct ip_vs_service
*svc
, struct ip_vs_dest_user_kern
*udest
)
1026 struct ip_vs_dest
*dest
;
1027 union nf_inet_addr daddr
;
1028 __be16 dport
= udest
->port
;
1032 if (udest
->weight
< 0) {
1033 pr_err("%s(): server weight less than zero\n", __func__
);
1037 if (udest
->l_threshold
> udest
->u_threshold
) {
1038 pr_err("%s(): lower threshold is higher than upper threshold\n",
1043 ip_vs_addr_copy(udest
->af
, &daddr
, &udest
->addr
);
1045 /* We use function that requires RCU lock */
1047 dest
= ip_vs_lookup_dest(svc
, udest
->af
, &daddr
, dport
);
1051 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__
);
1055 __ip_vs_update_dest(svc
, dest
, udest
, 0);
1062 * Delete a destination (must be already unlinked from the service)
1064 static void __ip_vs_del_dest(struct netns_ipvs
*ipvs
, struct ip_vs_dest
*dest
,
1067 ip_vs_stop_estimator(ipvs
, &dest
->stats
);
1070 * Remove it from the d-linked list with the real services.
1072 ip_vs_rs_unhash(dest
);
1074 spin_lock_bh(&ipvs
->dest_trash_lock
);
1075 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1076 IP_VS_DBG_ADDR(dest
->af
, &dest
->addr
), ntohs(dest
->port
),
1077 refcount_read(&dest
->refcnt
));
1078 if (list_empty(&ipvs
->dest_trash
) && !cleanup
)
1079 mod_timer(&ipvs
->dest_trash_timer
,
1080 jiffies
+ (IP_VS_DEST_TRASH_PERIOD
>> 1));
1081 /* dest lives in trash with reference */
1082 list_add(&dest
->t_list
, &ipvs
->dest_trash
);
1083 dest
->idle_start
= 0;
1084 spin_unlock_bh(&ipvs
->dest_trash_lock
);
1089 * Unlink a destination from the given service
1091 static void __ip_vs_unlink_dest(struct ip_vs_service
*svc
,
1092 struct ip_vs_dest
*dest
,
1095 dest
->flags
&= ~IP_VS_DEST_F_AVAILABLE
;
1098 * Remove it from the d-linked destination list.
1100 list_del_rcu(&dest
->n_list
);
1103 if (dest
->af
!= svc
->af
)
1104 svc
->ipvs
->mixed_address_family_dests
--;
1107 struct ip_vs_scheduler
*sched
;
1109 sched
= rcu_dereference_protected(svc
->scheduler
, 1);
1110 if (sched
&& sched
->del_dest
)
1111 sched
->del_dest(svc
, dest
);
1117 * Delete a destination server in the given service
1120 ip_vs_del_dest(struct ip_vs_service
*svc
, struct ip_vs_dest_user_kern
*udest
)
1122 struct ip_vs_dest
*dest
;
1123 __be16 dport
= udest
->port
;
1127 /* We use function that requires RCU lock */
1129 dest
= ip_vs_lookup_dest(svc
, udest
->af
, &udest
->addr
, dport
);
1133 IP_VS_DBG(1, "%s(): destination not found!\n", __func__
);
1138 * Unlink dest from the service
1140 __ip_vs_unlink_dest(svc
, dest
, 1);
1143 * Delete the destination
1145 __ip_vs_del_dest(svc
->ipvs
, dest
, false);
1152 static void ip_vs_dest_trash_expire(unsigned long data
)
1154 struct netns_ipvs
*ipvs
= (struct netns_ipvs
*)data
;
1155 struct ip_vs_dest
*dest
, *next
;
1156 unsigned long now
= jiffies
;
1158 spin_lock(&ipvs
->dest_trash_lock
);
1159 list_for_each_entry_safe(dest
, next
, &ipvs
->dest_trash
, t_list
) {
1160 if (refcount_read(&dest
->refcnt
) > 1)
1162 if (dest
->idle_start
) {
1163 if (time_before(now
, dest
->idle_start
+
1164 IP_VS_DEST_TRASH_PERIOD
))
1167 dest
->idle_start
= max(1UL, now
);
1170 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1172 IP_VS_DBG_ADDR(dest
->af
, &dest
->addr
),
1174 list_del(&dest
->t_list
);
1175 ip_vs_dest_free(dest
);
1177 if (!list_empty(&ipvs
->dest_trash
))
1178 mod_timer(&ipvs
->dest_trash_timer
,
1179 jiffies
+ (IP_VS_DEST_TRASH_PERIOD
>> 1));
1180 spin_unlock(&ipvs
->dest_trash_lock
);
1184 * Add a service into the service hash table
1187 ip_vs_add_service(struct netns_ipvs
*ipvs
, struct ip_vs_service_user_kern
*u
,
1188 struct ip_vs_service
**svc_p
)
1191 struct ip_vs_scheduler
*sched
= NULL
;
1192 struct ip_vs_pe
*pe
= NULL
;
1193 struct ip_vs_service
*svc
= NULL
;
1195 /* increase the module use count */
1196 ip_vs_use_count_inc();
1198 /* Lookup the scheduler by 'u->sched_name' */
1199 if (strcmp(u
->sched_name
, "none")) {
1200 sched
= ip_vs_scheduler_get(u
->sched_name
);
1202 pr_info("Scheduler module ip_vs_%s not found\n",
1209 if (u
->pe_name
&& *u
->pe_name
) {
1210 pe
= ip_vs_pe_getbyname(u
->pe_name
);
1212 pr_info("persistence engine module ip_vs_pe_%s "
1213 "not found\n", u
->pe_name
);
1219 #ifdef CONFIG_IP_VS_IPV6
1220 if (u
->af
== AF_INET6
) {
1221 __u32 plen
= (__force __u32
) u
->netmask
;
1223 if (plen
< 1 || plen
> 128) {
1230 svc
= kzalloc(sizeof(struct ip_vs_service
), GFP_KERNEL
);
1232 IP_VS_DBG(1, "%s(): no memory\n", __func__
);
1236 svc
->stats
.cpustats
= alloc_percpu(struct ip_vs_cpu_stats
);
1237 if (!svc
->stats
.cpustats
) {
1242 for_each_possible_cpu(i
) {
1243 struct ip_vs_cpu_stats
*ip_vs_stats
;
1244 ip_vs_stats
= per_cpu_ptr(svc
->stats
.cpustats
, i
);
1245 u64_stats_init(&ip_vs_stats
->syncp
);
1249 /* I'm the first user of the service */
1250 atomic_set(&svc
->refcnt
, 0);
1253 svc
->protocol
= u
->protocol
;
1254 ip_vs_addr_copy(svc
->af
, &svc
->addr
, &u
->addr
);
1255 svc
->port
= u
->port
;
1256 svc
->fwmark
= u
->fwmark
;
1257 svc
->flags
= u
->flags
;
1258 svc
->timeout
= u
->timeout
* HZ
;
1259 svc
->netmask
= u
->netmask
;
1262 INIT_LIST_HEAD(&svc
->destinations
);
1263 spin_lock_init(&svc
->sched_lock
);
1264 spin_lock_init(&svc
->stats
.lock
);
1266 /* Bind the scheduler */
1268 ret
= ip_vs_bind_scheduler(svc
, sched
);
1274 /* Bind the ct retriever */
1275 RCU_INIT_POINTER(svc
->pe
, pe
);
1278 /* Update the virtual service counters */
1279 if (svc
->port
== FTPPORT
)
1280 atomic_inc(&ipvs
->ftpsvc_counter
);
1281 else if (svc
->port
== 0)
1282 atomic_inc(&ipvs
->nullsvc_counter
);
1283 if (svc
->pe
&& svc
->pe
->conn_out
)
1284 atomic_inc(&ipvs
->conn_out_counter
);
1286 ip_vs_start_estimator(ipvs
, &svc
->stats
);
1288 /* Count only IPv4 services for old get/setsockopt interface */
1289 if (svc
->af
== AF_INET
)
1290 ipvs
->num_services
++;
1292 /* Hash the service into the service table */
1293 ip_vs_svc_hash(svc
);
1296 /* Now there is a service - full throttle */
1303 ip_vs_unbind_scheduler(svc
, sched
);
1304 ip_vs_service_free(svc
);
1306 ip_vs_scheduler_put(sched
);
1309 /* decrease the module use count */
1310 ip_vs_use_count_dec();
1317 * Edit a service and bind it with a new scheduler
1320 ip_vs_edit_service(struct ip_vs_service
*svc
, struct ip_vs_service_user_kern
*u
)
1322 struct ip_vs_scheduler
*sched
= NULL
, *old_sched
;
1323 struct ip_vs_pe
*pe
= NULL
, *old_pe
= NULL
;
1325 bool new_pe_conn_out
, old_pe_conn_out
;
1328 * Lookup the scheduler, by 'u->sched_name'
1330 if (strcmp(u
->sched_name
, "none")) {
1331 sched
= ip_vs_scheduler_get(u
->sched_name
);
1333 pr_info("Scheduler module ip_vs_%s not found\n",
1340 if (u
->pe_name
&& *u
->pe_name
) {
1341 pe
= ip_vs_pe_getbyname(u
->pe_name
);
1343 pr_info("persistence engine module ip_vs_pe_%s "
1344 "not found\n", u
->pe_name
);
1351 #ifdef CONFIG_IP_VS_IPV6
1352 if (u
->af
== AF_INET6
) {
1353 __u32 plen
= (__force __u32
) u
->netmask
;
1355 if (plen
< 1 || plen
> 128) {
1362 old_sched
= rcu_dereference_protected(svc
->scheduler
, 1);
1363 if (sched
!= old_sched
) {
1365 ip_vs_unbind_scheduler(svc
, old_sched
);
1366 RCU_INIT_POINTER(svc
->scheduler
, NULL
);
1367 /* Wait all svc->sched_data users */
1370 /* Bind the new scheduler */
1372 ret
= ip_vs_bind_scheduler(svc
, sched
);
1374 ip_vs_scheduler_put(sched
);
1381 * Set the flags and timeout value
1383 svc
->flags
= u
->flags
| IP_VS_SVC_F_HASHED
;
1384 svc
->timeout
= u
->timeout
* HZ
;
1385 svc
->netmask
= u
->netmask
;
1387 old_pe
= rcu_dereference_protected(svc
->pe
, 1);
1389 rcu_assign_pointer(svc
->pe
, pe
);
1390 /* check for optional methods in new pe */
1391 new_pe_conn_out
= (pe
&& pe
->conn_out
) ? true : false;
1392 old_pe_conn_out
= (old_pe
&& old_pe
->conn_out
) ? true : false;
1393 if (new_pe_conn_out
&& !old_pe_conn_out
)
1394 atomic_inc(&svc
->ipvs
->conn_out_counter
);
1395 if (old_pe_conn_out
&& !new_pe_conn_out
)
1396 atomic_dec(&svc
->ipvs
->conn_out_counter
);
1400 ip_vs_scheduler_put(old_sched
);
1401 ip_vs_pe_put(old_pe
);
1406 * Delete a service from the service list
1407 * - The service must be unlinked, unlocked and not referenced!
1408 * - We are called under _bh lock
1410 static void __ip_vs_del_service(struct ip_vs_service
*svc
, bool cleanup
)
1412 struct ip_vs_dest
*dest
, *nxt
;
1413 struct ip_vs_scheduler
*old_sched
;
1414 struct ip_vs_pe
*old_pe
;
1415 struct netns_ipvs
*ipvs
= svc
->ipvs
;
1417 /* Count only IPv4 services for old get/setsockopt interface */
1418 if (svc
->af
== AF_INET
)
1419 ipvs
->num_services
--;
1421 ip_vs_stop_estimator(svc
->ipvs
, &svc
->stats
);
1423 /* Unbind scheduler */
1424 old_sched
= rcu_dereference_protected(svc
->scheduler
, 1);
1425 ip_vs_unbind_scheduler(svc
, old_sched
);
1426 ip_vs_scheduler_put(old_sched
);
1428 /* Unbind persistence engine, keep svc->pe */
1429 old_pe
= rcu_dereference_protected(svc
->pe
, 1);
1430 if (old_pe
&& old_pe
->conn_out
)
1431 atomic_dec(&ipvs
->conn_out_counter
);
1432 ip_vs_pe_put(old_pe
);
1435 * Unlink the whole destination list
1437 list_for_each_entry_safe(dest
, nxt
, &svc
->destinations
, n_list
) {
1438 __ip_vs_unlink_dest(svc
, dest
, 0);
1439 __ip_vs_del_dest(svc
->ipvs
, dest
, cleanup
);
1443 * Update the virtual service counters
1445 if (svc
->port
== FTPPORT
)
1446 atomic_dec(&ipvs
->ftpsvc_counter
);
1447 else if (svc
->port
== 0)
1448 atomic_dec(&ipvs
->nullsvc_counter
);
1451 * Free the service if nobody refers to it
1453 __ip_vs_svc_put(svc
, true);
1455 /* decrease the module use count */
1456 ip_vs_use_count_dec();
1460 * Unlink a service from list and try to delete it if its refcnt reached 0
1462 static void ip_vs_unlink_service(struct ip_vs_service
*svc
, bool cleanup
)
1464 /* Hold svc to avoid double release from dest_trash */
1465 atomic_inc(&svc
->refcnt
);
1467 * Unhash it from the service table
1469 ip_vs_svc_unhash(svc
);
1471 __ip_vs_del_service(svc
, cleanup
);
1475 * Delete a service from the service list
1477 static int ip_vs_del_service(struct ip_vs_service
*svc
)
1481 ip_vs_unlink_service(svc
, false);
1488 * Flush all the virtual services
1490 static int ip_vs_flush(struct netns_ipvs
*ipvs
, bool cleanup
)
1493 struct ip_vs_service
*svc
;
1494 struct hlist_node
*n
;
1497 * Flush the service table hashed by <netns,protocol,addr,port>
1499 for(idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1500 hlist_for_each_entry_safe(svc
, n
, &ip_vs_svc_table
[idx
],
1502 if (svc
->ipvs
== ipvs
)
1503 ip_vs_unlink_service(svc
, cleanup
);
1508 * Flush the service table hashed by fwmark
1510 for(idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1511 hlist_for_each_entry_safe(svc
, n
, &ip_vs_svc_fwm_table
[idx
],
1513 if (svc
->ipvs
== ipvs
)
1514 ip_vs_unlink_service(svc
, cleanup
);
1522 * Delete service by {netns} in the service table.
1523 * Called by __ip_vs_cleanup()
1525 void ip_vs_service_net_cleanup(struct netns_ipvs
*ipvs
)
1528 /* Check for "full" addressed entries */
1529 mutex_lock(&__ip_vs_mutex
);
1530 ip_vs_flush(ipvs
, true);
1531 mutex_unlock(&__ip_vs_mutex
);
1535 /* Put all references for device (dst_cache) */
1537 ip_vs_forget_dev(struct ip_vs_dest
*dest
, struct net_device
*dev
)
1539 struct ip_vs_dest_dst
*dest_dst
;
1541 spin_lock_bh(&dest
->dst_lock
);
1542 dest_dst
= rcu_dereference_protected(dest
->dest_dst
, 1);
1543 if (dest_dst
&& dest_dst
->dst_cache
->dev
== dev
) {
1544 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1546 IP_VS_DBG_ADDR(dest
->af
, &dest
->addr
),
1548 refcount_read(&dest
->refcnt
));
1549 __ip_vs_dst_cache_reset(dest
);
1551 spin_unlock_bh(&dest
->dst_lock
);
1554 /* Netdev event receiver
1555 * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1557 static int ip_vs_dst_event(struct notifier_block
*this, unsigned long event
,
1560 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
1561 struct net
*net
= dev_net(dev
);
1562 struct netns_ipvs
*ipvs
= net_ipvs(net
);
1563 struct ip_vs_service
*svc
;
1564 struct ip_vs_dest
*dest
;
1567 if (event
!= NETDEV_DOWN
|| !ipvs
)
1569 IP_VS_DBG(3, "%s() dev=%s\n", __func__
, dev
->name
);
1571 mutex_lock(&__ip_vs_mutex
);
1572 for (idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1573 hlist_for_each_entry(svc
, &ip_vs_svc_table
[idx
], s_list
) {
1574 if (svc
->ipvs
== ipvs
) {
1575 list_for_each_entry(dest
, &svc
->destinations
,
1577 ip_vs_forget_dev(dest
, dev
);
1582 hlist_for_each_entry(svc
, &ip_vs_svc_fwm_table
[idx
], f_list
) {
1583 if (svc
->ipvs
== ipvs
) {
1584 list_for_each_entry(dest
, &svc
->destinations
,
1586 ip_vs_forget_dev(dest
, dev
);
1593 spin_lock_bh(&ipvs
->dest_trash_lock
);
1594 list_for_each_entry(dest
, &ipvs
->dest_trash
, t_list
) {
1595 ip_vs_forget_dev(dest
, dev
);
1597 spin_unlock_bh(&ipvs
->dest_trash_lock
);
1598 mutex_unlock(&__ip_vs_mutex
);
1604 * Zero counters in a service or all services
1606 static int ip_vs_zero_service(struct ip_vs_service
*svc
)
1608 struct ip_vs_dest
*dest
;
1610 list_for_each_entry(dest
, &svc
->destinations
, n_list
) {
1611 ip_vs_zero_stats(&dest
->stats
);
1613 ip_vs_zero_stats(&svc
->stats
);
1617 static int ip_vs_zero_all(struct netns_ipvs
*ipvs
)
1620 struct ip_vs_service
*svc
;
1622 for(idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1623 hlist_for_each_entry(svc
, &ip_vs_svc_table
[idx
], s_list
) {
1624 if (svc
->ipvs
== ipvs
)
1625 ip_vs_zero_service(svc
);
1629 for(idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1630 hlist_for_each_entry(svc
, &ip_vs_svc_fwm_table
[idx
], f_list
) {
1631 if (svc
->ipvs
== ipvs
)
1632 ip_vs_zero_service(svc
);
1636 ip_vs_zero_stats(&ipvs
->tot_stats
);
1640 #ifdef CONFIG_SYSCTL
1643 static int three
= 3;
1646 proc_do_defense_mode(struct ctl_table
*table
, int write
,
1647 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
1649 struct netns_ipvs
*ipvs
= table
->extra2
;
1650 int *valp
= table
->data
;
1654 rc
= proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
1655 if (write
&& (*valp
!= val
)) {
1656 if ((*valp
< 0) || (*valp
> 3)) {
1657 /* Restore the correct value */
1660 update_defense_level(ipvs
);
1667 proc_do_sync_threshold(struct ctl_table
*table
, int write
,
1668 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
1670 int *valp
= table
->data
;
1674 /* backup the value first */
1675 memcpy(val
, valp
, sizeof(val
));
1677 rc
= proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
1678 if (write
&& (valp
[0] < 0 || valp
[1] < 0 ||
1679 (valp
[0] >= valp
[1] && valp
[1]))) {
1680 /* Restore the correct value */
1681 memcpy(valp
, val
, sizeof(val
));
1687 proc_do_sync_mode(struct ctl_table
*table
, int write
,
1688 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
1690 int *valp
= table
->data
;
1694 rc
= proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
1695 if (write
&& (*valp
!= val
)) {
1696 if ((*valp
< 0) || (*valp
> 1)) {
1697 /* Restore the correct value */
1705 proc_do_sync_ports(struct ctl_table
*table
, int write
,
1706 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
1708 int *valp
= table
->data
;
1712 rc
= proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
1713 if (write
&& (*valp
!= val
)) {
1714 if (*valp
< 1 || !is_power_of_2(*valp
)) {
1715 /* Restore the correct value */
1723 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1724 * Do not change order or insert new entries without
1725 * align with netns init in ip_vs_control_net_init()
1728 static struct ctl_table vs_vars
[] = {
1730 .procname
= "amemthresh",
1731 .maxlen
= sizeof(int),
1733 .proc_handler
= proc_dointvec
,
1736 .procname
= "am_droprate",
1737 .maxlen
= sizeof(int),
1739 .proc_handler
= proc_dointvec
,
1742 .procname
= "drop_entry",
1743 .maxlen
= sizeof(int),
1745 .proc_handler
= proc_do_defense_mode
,
1748 .procname
= "drop_packet",
1749 .maxlen
= sizeof(int),
1751 .proc_handler
= proc_do_defense_mode
,
1753 #ifdef CONFIG_IP_VS_NFCT
1755 .procname
= "conntrack",
1756 .maxlen
= sizeof(int),
1758 .proc_handler
= &proc_dointvec
,
1762 .procname
= "secure_tcp",
1763 .maxlen
= sizeof(int),
1765 .proc_handler
= proc_do_defense_mode
,
1768 .procname
= "snat_reroute",
1769 .maxlen
= sizeof(int),
1771 .proc_handler
= &proc_dointvec
,
1774 .procname
= "sync_version",
1775 .maxlen
= sizeof(int),
1777 .proc_handler
= proc_do_sync_mode
,
1780 .procname
= "sync_ports",
1781 .maxlen
= sizeof(int),
1783 .proc_handler
= proc_do_sync_ports
,
1786 .procname
= "sync_persist_mode",
1787 .maxlen
= sizeof(int),
1789 .proc_handler
= proc_dointvec
,
1792 .procname
= "sync_qlen_max",
1793 .maxlen
= sizeof(unsigned long),
1795 .proc_handler
= proc_doulongvec_minmax
,
1798 .procname
= "sync_sock_size",
1799 .maxlen
= sizeof(int),
1801 .proc_handler
= proc_dointvec
,
1804 .procname
= "cache_bypass",
1805 .maxlen
= sizeof(int),
1807 .proc_handler
= proc_dointvec
,
1810 .procname
= "expire_nodest_conn",
1811 .maxlen
= sizeof(int),
1813 .proc_handler
= proc_dointvec
,
1816 .procname
= "sloppy_tcp",
1817 .maxlen
= sizeof(int),
1819 .proc_handler
= proc_dointvec
,
1822 .procname
= "sloppy_sctp",
1823 .maxlen
= sizeof(int),
1825 .proc_handler
= proc_dointvec
,
1828 .procname
= "expire_quiescent_template",
1829 .maxlen
= sizeof(int),
1831 .proc_handler
= proc_dointvec
,
1834 .procname
= "sync_threshold",
1836 sizeof(((struct netns_ipvs
*)0)->sysctl_sync_threshold
),
1838 .proc_handler
= proc_do_sync_threshold
,
1841 .procname
= "sync_refresh_period",
1842 .maxlen
= sizeof(int),
1844 .proc_handler
= proc_dointvec_jiffies
,
1847 .procname
= "sync_retries",
1848 .maxlen
= sizeof(int),
1850 .proc_handler
= proc_dointvec_minmax
,
1855 .procname
= "nat_icmp_send",
1856 .maxlen
= sizeof(int),
1858 .proc_handler
= proc_dointvec
,
1861 .procname
= "pmtu_disc",
1862 .maxlen
= sizeof(int),
1864 .proc_handler
= proc_dointvec
,
1867 .procname
= "backup_only",
1868 .maxlen
= sizeof(int),
1870 .proc_handler
= proc_dointvec
,
1873 .procname
= "conn_reuse_mode",
1874 .maxlen
= sizeof(int),
1876 .proc_handler
= proc_dointvec
,
1879 .procname
= "schedule_icmp",
1880 .maxlen
= sizeof(int),
1882 .proc_handler
= proc_dointvec
,
1885 .procname
= "ignore_tunneled",
1886 .maxlen
= sizeof(int),
1888 .proc_handler
= proc_dointvec
,
1890 #ifdef CONFIG_IP_VS_DEBUG
1892 .procname
= "debug_level",
1893 .data
= &sysctl_ip_vs_debug_level
,
1894 .maxlen
= sizeof(int),
1896 .proc_handler
= proc_dointvec
,
1904 #ifdef CONFIG_PROC_FS
1907 struct seq_net_private p
; /* Do not move this, netns depends upon it*/
1908 struct hlist_head
*table
;
1913 * Write the contents of the VS rule table to a PROCfs file.
1914 * (It is kept just for backward compatibility)
1916 static inline const char *ip_vs_fwd_name(unsigned int flags
)
1918 switch (flags
& IP_VS_CONN_F_FWD_MASK
) {
1919 case IP_VS_CONN_F_LOCALNODE
:
1921 case IP_VS_CONN_F_TUNNEL
:
1923 case IP_VS_CONN_F_DROUTE
:
1931 /* Get the Nth entry in the two lists */
1932 static struct ip_vs_service
*ip_vs_info_array(struct seq_file
*seq
, loff_t pos
)
1934 struct net
*net
= seq_file_net(seq
);
1935 struct netns_ipvs
*ipvs
= net_ipvs(net
);
1936 struct ip_vs_iter
*iter
= seq
->private;
1938 struct ip_vs_service
*svc
;
1940 /* look in hash by protocol */
1941 for (idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1942 hlist_for_each_entry_rcu(svc
, &ip_vs_svc_table
[idx
], s_list
) {
1943 if ((svc
->ipvs
== ipvs
) && pos
-- == 0) {
1944 iter
->table
= ip_vs_svc_table
;
1951 /* keep looking in fwmark */
1952 for (idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1953 hlist_for_each_entry_rcu(svc
, &ip_vs_svc_fwm_table
[idx
],
1955 if ((svc
->ipvs
== ipvs
) && pos
-- == 0) {
1956 iter
->table
= ip_vs_svc_fwm_table
;
1966 static void *ip_vs_info_seq_start(struct seq_file
*seq
, loff_t
*pos
)
1970 return *pos
? ip_vs_info_array(seq
, *pos
- 1) : SEQ_START_TOKEN
;
1974 static void *ip_vs_info_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
1976 struct hlist_node
*e
;
1977 struct ip_vs_iter
*iter
;
1978 struct ip_vs_service
*svc
;
1981 if (v
== SEQ_START_TOKEN
)
1982 return ip_vs_info_array(seq
,0);
1985 iter
= seq
->private;
1987 if (iter
->table
== ip_vs_svc_table
) {
1988 /* next service in table hashed by protocol */
1989 e
= rcu_dereference(hlist_next_rcu(&svc
->s_list
));
1991 return hlist_entry(e
, struct ip_vs_service
, s_list
);
1993 while (++iter
->bucket
< IP_VS_SVC_TAB_SIZE
) {
1994 hlist_for_each_entry_rcu(svc
,
1995 &ip_vs_svc_table
[iter
->bucket
],
2001 iter
->table
= ip_vs_svc_fwm_table
;
2006 /* next service in hashed by fwmark */
2007 e
= rcu_dereference(hlist_next_rcu(&svc
->f_list
));
2009 return hlist_entry(e
, struct ip_vs_service
, f_list
);
2012 while (++iter
->bucket
< IP_VS_SVC_TAB_SIZE
) {
2013 hlist_for_each_entry_rcu(svc
,
2014 &ip_vs_svc_fwm_table
[iter
->bucket
],
2022 static void ip_vs_info_seq_stop(struct seq_file
*seq
, void *v
)
2029 static int ip_vs_info_seq_show(struct seq_file
*seq
, void *v
)
2031 if (v
== SEQ_START_TOKEN
) {
2033 "IP Virtual Server version %d.%d.%d (size=%d)\n",
2034 NVERSION(IP_VS_VERSION_CODE
), ip_vs_conn_tab_size
);
2036 "Prot LocalAddress:Port Scheduler Flags\n");
2038 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2040 const struct ip_vs_service
*svc
= v
;
2041 const struct ip_vs_iter
*iter
= seq
->private;
2042 const struct ip_vs_dest
*dest
;
2043 struct ip_vs_scheduler
*sched
= rcu_dereference(svc
->scheduler
);
2044 char *sched_name
= sched
? sched
->name
: "none";
2046 if (iter
->table
== ip_vs_svc_table
) {
2047 #ifdef CONFIG_IP_VS_IPV6
2048 if (svc
->af
== AF_INET6
)
2049 seq_printf(seq
, "%s [%pI6]:%04X %s ",
2050 ip_vs_proto_name(svc
->protocol
),
2056 seq_printf(seq
, "%s %08X:%04X %s %s ",
2057 ip_vs_proto_name(svc
->protocol
),
2058 ntohl(svc
->addr
.ip
),
2061 (svc
->flags
& IP_VS_SVC_F_ONEPACKET
)?"ops ":"");
2063 seq_printf(seq
, "FWM %08X %s %s",
2064 svc
->fwmark
, sched_name
,
2065 (svc
->flags
& IP_VS_SVC_F_ONEPACKET
)?"ops ":"");
2068 if (svc
->flags
& IP_VS_SVC_F_PERSISTENT
)
2069 seq_printf(seq
, "persistent %d %08X\n",
2071 ntohl(svc
->netmask
));
2073 seq_putc(seq
, '\n');
2075 list_for_each_entry_rcu(dest
, &svc
->destinations
, n_list
) {
2076 #ifdef CONFIG_IP_VS_IPV6
2077 if (dest
->af
== AF_INET6
)
2080 " %-7s %-6d %-10d %-10d\n",
2083 ip_vs_fwd_name(atomic_read(&dest
->conn_flags
)),
2084 atomic_read(&dest
->weight
),
2085 atomic_read(&dest
->activeconns
),
2086 atomic_read(&dest
->inactconns
));
2091 "%-7s %-6d %-10d %-10d\n",
2092 ntohl(dest
->addr
.ip
),
2094 ip_vs_fwd_name(atomic_read(&dest
->conn_flags
)),
2095 atomic_read(&dest
->weight
),
2096 atomic_read(&dest
->activeconns
),
2097 atomic_read(&dest
->inactconns
));
2104 static const struct seq_operations ip_vs_info_seq_ops
= {
2105 .start
= ip_vs_info_seq_start
,
2106 .next
= ip_vs_info_seq_next
,
2107 .stop
= ip_vs_info_seq_stop
,
2108 .show
= ip_vs_info_seq_show
,
2111 static int ip_vs_info_open(struct inode
*inode
, struct file
*file
)
2113 return seq_open_net(inode
, file
, &ip_vs_info_seq_ops
,
2114 sizeof(struct ip_vs_iter
));
2117 static const struct file_operations ip_vs_info_fops
= {
2118 .owner
= THIS_MODULE
,
2119 .open
= ip_vs_info_open
,
2121 .llseek
= seq_lseek
,
2122 .release
= seq_release_net
,
2125 static int ip_vs_stats_show(struct seq_file
*seq
, void *v
)
2127 struct net
*net
= seq_file_single_net(seq
);
2128 struct ip_vs_kstats show
;
2130 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2132 " Total Incoming Outgoing Incoming Outgoing\n");
2134 " Conns Packets Packets Bytes Bytes\n");
2136 ip_vs_copy_stats(&show
, &net_ipvs(net
)->tot_stats
);
2137 seq_printf(seq
, "%8LX %8LX %8LX %16LX %16LX\n\n",
2138 (unsigned long long)show
.conns
,
2139 (unsigned long long)show
.inpkts
,
2140 (unsigned long long)show
.outpkts
,
2141 (unsigned long long)show
.inbytes
,
2142 (unsigned long long)show
.outbytes
);
2144 /* 01234567 01234567 01234567 0123456701234567 0123456701234567*/
2146 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2147 seq_printf(seq
, "%8LX %8LX %8LX %16LX %16LX\n",
2148 (unsigned long long)show
.cps
,
2149 (unsigned long long)show
.inpps
,
2150 (unsigned long long)show
.outpps
,
2151 (unsigned long long)show
.inbps
,
2152 (unsigned long long)show
.outbps
);
2157 static int ip_vs_stats_seq_open(struct inode
*inode
, struct file
*file
)
2159 return single_open_net(inode
, file
, ip_vs_stats_show
);
2162 static const struct file_operations ip_vs_stats_fops
= {
2163 .owner
= THIS_MODULE
,
2164 .open
= ip_vs_stats_seq_open
,
2166 .llseek
= seq_lseek
,
2167 .release
= single_release_net
,
2170 static int ip_vs_stats_percpu_show(struct seq_file
*seq
, void *v
)
2172 struct net
*net
= seq_file_single_net(seq
);
2173 struct ip_vs_stats
*tot_stats
= &net_ipvs(net
)->tot_stats
;
2174 struct ip_vs_cpu_stats __percpu
*cpustats
= tot_stats
->cpustats
;
2175 struct ip_vs_kstats kstats
;
2178 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2180 " Total Incoming Outgoing Incoming Outgoing\n");
2182 "CPU Conns Packets Packets Bytes Bytes\n");
2184 for_each_possible_cpu(i
) {
2185 struct ip_vs_cpu_stats
*u
= per_cpu_ptr(cpustats
, i
);
2187 u64 conns
, inpkts
, outpkts
, inbytes
, outbytes
;
2190 start
= u64_stats_fetch_begin_irq(&u
->syncp
);
2191 conns
= u
->cnt
.conns
;
2192 inpkts
= u
->cnt
.inpkts
;
2193 outpkts
= u
->cnt
.outpkts
;
2194 inbytes
= u
->cnt
.inbytes
;
2195 outbytes
= u
->cnt
.outbytes
;
2196 } while (u64_stats_fetch_retry_irq(&u
->syncp
, start
));
2198 seq_printf(seq
, "%3X %8LX %8LX %8LX %16LX %16LX\n",
2199 i
, (u64
)conns
, (u64
)inpkts
,
2200 (u64
)outpkts
, (u64
)inbytes
,
2204 ip_vs_copy_stats(&kstats
, tot_stats
);
2206 seq_printf(seq
, " ~ %8LX %8LX %8LX %16LX %16LX\n\n",
2207 (unsigned long long)kstats
.conns
,
2208 (unsigned long long)kstats
.inpkts
,
2209 (unsigned long long)kstats
.outpkts
,
2210 (unsigned long long)kstats
.inbytes
,
2211 (unsigned long long)kstats
.outbytes
);
2213 /* ... 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2215 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2216 seq_printf(seq
, " %8LX %8LX %8LX %16LX %16LX\n",
2226 static int ip_vs_stats_percpu_seq_open(struct inode
*inode
, struct file
*file
)
2228 return single_open_net(inode
, file
, ip_vs_stats_percpu_show
);
2231 static const struct file_operations ip_vs_stats_percpu_fops
= {
2232 .owner
= THIS_MODULE
,
2233 .open
= ip_vs_stats_percpu_seq_open
,
2235 .llseek
= seq_lseek
,
2236 .release
= single_release_net
,
2241 * Set timeout values for tcp tcpfin udp in the timeout_table.
2243 static int ip_vs_set_timeout(struct netns_ipvs
*ipvs
, struct ip_vs_timeout_user
*u
)
2245 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2246 struct ip_vs_proto_data
*pd
;
2249 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2254 #ifdef CONFIG_IP_VS_PROTO_TCP
2255 if (u
->tcp_timeout
) {
2256 pd
= ip_vs_proto_data_get(ipvs
, IPPROTO_TCP
);
2257 pd
->timeout_table
[IP_VS_TCP_S_ESTABLISHED
]
2258 = u
->tcp_timeout
* HZ
;
2261 if (u
->tcp_fin_timeout
) {
2262 pd
= ip_vs_proto_data_get(ipvs
, IPPROTO_TCP
);
2263 pd
->timeout_table
[IP_VS_TCP_S_FIN_WAIT
]
2264 = u
->tcp_fin_timeout
* HZ
;
2268 #ifdef CONFIG_IP_VS_PROTO_UDP
2269 if (u
->udp_timeout
) {
2270 pd
= ip_vs_proto_data_get(ipvs
, IPPROTO_UDP
);
2271 pd
->timeout_table
[IP_VS_UDP_S_NORMAL
]
2272 = u
->udp_timeout
* HZ
;
2278 #define CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2280 struct ip_vs_svcdest_user
{
2281 struct ip_vs_service_user s
;
2282 struct ip_vs_dest_user d
;
2285 static const unsigned char set_arglen
[CMDID(IP_VS_SO_SET_MAX
) + 1] = {
2286 [CMDID(IP_VS_SO_SET_ADD
)] = sizeof(struct ip_vs_service_user
),
2287 [CMDID(IP_VS_SO_SET_EDIT
)] = sizeof(struct ip_vs_service_user
),
2288 [CMDID(IP_VS_SO_SET_DEL
)] = sizeof(struct ip_vs_service_user
),
2289 [CMDID(IP_VS_SO_SET_ADDDEST
)] = sizeof(struct ip_vs_svcdest_user
),
2290 [CMDID(IP_VS_SO_SET_DELDEST
)] = sizeof(struct ip_vs_svcdest_user
),
2291 [CMDID(IP_VS_SO_SET_EDITDEST
)] = sizeof(struct ip_vs_svcdest_user
),
2292 [CMDID(IP_VS_SO_SET_TIMEOUT
)] = sizeof(struct ip_vs_timeout_user
),
2293 [CMDID(IP_VS_SO_SET_STARTDAEMON
)] = sizeof(struct ip_vs_daemon_user
),
2294 [CMDID(IP_VS_SO_SET_STOPDAEMON
)] = sizeof(struct ip_vs_daemon_user
),
2295 [CMDID(IP_VS_SO_SET_ZERO
)] = sizeof(struct ip_vs_service_user
),
2298 union ip_vs_set_arglen
{
2299 struct ip_vs_service_user field_IP_VS_SO_SET_ADD
;
2300 struct ip_vs_service_user field_IP_VS_SO_SET_EDIT
;
2301 struct ip_vs_service_user field_IP_VS_SO_SET_DEL
;
2302 struct ip_vs_svcdest_user field_IP_VS_SO_SET_ADDDEST
;
2303 struct ip_vs_svcdest_user field_IP_VS_SO_SET_DELDEST
;
2304 struct ip_vs_svcdest_user field_IP_VS_SO_SET_EDITDEST
;
2305 struct ip_vs_timeout_user field_IP_VS_SO_SET_TIMEOUT
;
2306 struct ip_vs_daemon_user field_IP_VS_SO_SET_STARTDAEMON
;
2307 struct ip_vs_daemon_user field_IP_VS_SO_SET_STOPDAEMON
;
2308 struct ip_vs_service_user field_IP_VS_SO_SET_ZERO
;
2311 #define MAX_SET_ARGLEN sizeof(union ip_vs_set_arglen)
2313 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern
*usvc
,
2314 struct ip_vs_service_user
*usvc_compat
)
2316 memset(usvc
, 0, sizeof(*usvc
));
2319 usvc
->protocol
= usvc_compat
->protocol
;
2320 usvc
->addr
.ip
= usvc_compat
->addr
;
2321 usvc
->port
= usvc_compat
->port
;
2322 usvc
->fwmark
= usvc_compat
->fwmark
;
2324 /* Deep copy of sched_name is not needed here */
2325 usvc
->sched_name
= usvc_compat
->sched_name
;
2327 usvc
->flags
= usvc_compat
->flags
;
2328 usvc
->timeout
= usvc_compat
->timeout
;
2329 usvc
->netmask
= usvc_compat
->netmask
;
2332 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern
*udest
,
2333 struct ip_vs_dest_user
*udest_compat
)
2335 memset(udest
, 0, sizeof(*udest
));
2337 udest
->addr
.ip
= udest_compat
->addr
;
2338 udest
->port
= udest_compat
->port
;
2339 udest
->conn_flags
= udest_compat
->conn_flags
;
2340 udest
->weight
= udest_compat
->weight
;
2341 udest
->u_threshold
= udest_compat
->u_threshold
;
2342 udest
->l_threshold
= udest_compat
->l_threshold
;
2343 udest
->af
= AF_INET
;
2347 do_ip_vs_set_ctl(struct sock
*sk
, int cmd
, void __user
*user
, unsigned int len
)
2349 struct net
*net
= sock_net(sk
);
2351 unsigned char arg
[MAX_SET_ARGLEN
];
2352 struct ip_vs_service_user
*usvc_compat
;
2353 struct ip_vs_service_user_kern usvc
;
2354 struct ip_vs_service
*svc
;
2355 struct ip_vs_dest_user
*udest_compat
;
2356 struct ip_vs_dest_user_kern udest
;
2357 struct netns_ipvs
*ipvs
= net_ipvs(net
);
2359 BUILD_BUG_ON(sizeof(arg
) > 255);
2360 if (!ns_capable(sock_net(sk
)->user_ns
, CAP_NET_ADMIN
))
2363 if (cmd
< IP_VS_BASE_CTL
|| cmd
> IP_VS_SO_SET_MAX
)
2365 if (len
!= set_arglen
[CMDID(cmd
)]) {
2366 IP_VS_DBG(1, "set_ctl: len %u != %u\n",
2367 len
, set_arglen
[CMDID(cmd
)]);
2371 if (copy_from_user(arg
, user
, len
) != 0)
2374 /* increase the module use count */
2375 ip_vs_use_count_inc();
2377 /* Handle daemons since they have another lock */
2378 if (cmd
== IP_VS_SO_SET_STARTDAEMON
||
2379 cmd
== IP_VS_SO_SET_STOPDAEMON
) {
2380 struct ip_vs_daemon_user
*dm
= (struct ip_vs_daemon_user
*)arg
;
2382 if (cmd
== IP_VS_SO_SET_STARTDAEMON
) {
2383 struct ipvs_sync_daemon_cfg cfg
;
2385 memset(&cfg
, 0, sizeof(cfg
));
2386 strlcpy(cfg
.mcast_ifn
, dm
->mcast_ifn
,
2387 sizeof(cfg
.mcast_ifn
));
2388 cfg
.syncid
= dm
->syncid
;
2390 mutex_lock(&ipvs
->sync_mutex
);
2391 ret
= start_sync_thread(ipvs
, &cfg
, dm
->state
);
2392 mutex_unlock(&ipvs
->sync_mutex
);
2395 mutex_lock(&ipvs
->sync_mutex
);
2396 ret
= stop_sync_thread(ipvs
, dm
->state
);
2397 mutex_unlock(&ipvs
->sync_mutex
);
2402 mutex_lock(&__ip_vs_mutex
);
2403 if (cmd
== IP_VS_SO_SET_FLUSH
) {
2404 /* Flush the virtual service */
2405 ret
= ip_vs_flush(ipvs
, false);
2407 } else if (cmd
== IP_VS_SO_SET_TIMEOUT
) {
2408 /* Set timeout values for (tcp tcpfin udp) */
2409 ret
= ip_vs_set_timeout(ipvs
, (struct ip_vs_timeout_user
*)arg
);
2413 usvc_compat
= (struct ip_vs_service_user
*)arg
;
2414 udest_compat
= (struct ip_vs_dest_user
*)(usvc_compat
+ 1);
2416 /* We only use the new structs internally, so copy userspace compat
2417 * structs to extended internal versions */
2418 ip_vs_copy_usvc_compat(&usvc
, usvc_compat
);
2419 ip_vs_copy_udest_compat(&udest
, udest_compat
);
2421 if (cmd
== IP_VS_SO_SET_ZERO
) {
2422 /* if no service address is set, zero counters in all */
2423 if (!usvc
.fwmark
&& !usvc
.addr
.ip
&& !usvc
.port
) {
2424 ret
= ip_vs_zero_all(ipvs
);
2429 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2430 if (usvc
.protocol
!= IPPROTO_TCP
&& usvc
.protocol
!= IPPROTO_UDP
&&
2431 usvc
.protocol
!= IPPROTO_SCTP
) {
2432 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2433 usvc
.protocol
, &usvc
.addr
.ip
,
2434 ntohs(usvc
.port
), usvc
.sched_name
);
2439 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2441 if (usvc
.fwmark
== 0)
2442 svc
= __ip_vs_service_find(ipvs
, usvc
.af
, usvc
.protocol
,
2443 &usvc
.addr
, usvc
.port
);
2445 svc
= __ip_vs_svc_fwm_find(ipvs
, usvc
.af
, usvc
.fwmark
);
2448 if (cmd
!= IP_VS_SO_SET_ADD
2449 && (svc
== NULL
|| svc
->protocol
!= usvc
.protocol
)) {
2455 case IP_VS_SO_SET_ADD
:
2459 ret
= ip_vs_add_service(ipvs
, &usvc
, &svc
);
2461 case IP_VS_SO_SET_EDIT
:
2462 ret
= ip_vs_edit_service(svc
, &usvc
);
2464 case IP_VS_SO_SET_DEL
:
2465 ret
= ip_vs_del_service(svc
);
2469 case IP_VS_SO_SET_ZERO
:
2470 ret
= ip_vs_zero_service(svc
);
2472 case IP_VS_SO_SET_ADDDEST
:
2473 ret
= ip_vs_add_dest(svc
, &udest
);
2475 case IP_VS_SO_SET_EDITDEST
:
2476 ret
= ip_vs_edit_dest(svc
, &udest
);
2478 case IP_VS_SO_SET_DELDEST
:
2479 ret
= ip_vs_del_dest(svc
, &udest
);
2486 mutex_unlock(&__ip_vs_mutex
);
2488 /* decrease the module use count */
2489 ip_vs_use_count_dec();
2496 ip_vs_copy_service(struct ip_vs_service_entry
*dst
, struct ip_vs_service
*src
)
2498 struct ip_vs_scheduler
*sched
;
2499 struct ip_vs_kstats kstats
;
2502 sched
= rcu_dereference_protected(src
->scheduler
, 1);
2503 sched_name
= sched
? sched
->name
: "none";
2504 dst
->protocol
= src
->protocol
;
2505 dst
->addr
= src
->addr
.ip
;
2506 dst
->port
= src
->port
;
2507 dst
->fwmark
= src
->fwmark
;
2508 strlcpy(dst
->sched_name
, sched_name
, sizeof(dst
->sched_name
));
2509 dst
->flags
= src
->flags
;
2510 dst
->timeout
= src
->timeout
/ HZ
;
2511 dst
->netmask
= src
->netmask
;
2512 dst
->num_dests
= src
->num_dests
;
2513 ip_vs_copy_stats(&kstats
, &src
->stats
);
2514 ip_vs_export_stats_user(&dst
->stats
, &kstats
);
2518 __ip_vs_get_service_entries(struct netns_ipvs
*ipvs
,
2519 const struct ip_vs_get_services
*get
,
2520 struct ip_vs_get_services __user
*uptr
)
2523 struct ip_vs_service
*svc
;
2524 struct ip_vs_service_entry entry
;
2527 for (idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
2528 hlist_for_each_entry(svc
, &ip_vs_svc_table
[idx
], s_list
) {
2529 /* Only expose IPv4 entries to old interface */
2530 if (svc
->af
!= AF_INET
|| (svc
->ipvs
!= ipvs
))
2533 if (count
>= get
->num_services
)
2535 memset(&entry
, 0, sizeof(entry
));
2536 ip_vs_copy_service(&entry
, svc
);
2537 if (copy_to_user(&uptr
->entrytable
[count
],
2538 &entry
, sizeof(entry
))) {
2546 for (idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
2547 hlist_for_each_entry(svc
, &ip_vs_svc_fwm_table
[idx
], f_list
) {
2548 /* Only expose IPv4 entries to old interface */
2549 if (svc
->af
!= AF_INET
|| (svc
->ipvs
!= ipvs
))
2552 if (count
>= get
->num_services
)
2554 memset(&entry
, 0, sizeof(entry
));
2555 ip_vs_copy_service(&entry
, svc
);
2556 if (copy_to_user(&uptr
->entrytable
[count
],
2557 &entry
, sizeof(entry
))) {
2569 __ip_vs_get_dest_entries(struct netns_ipvs
*ipvs
, const struct ip_vs_get_dests
*get
,
2570 struct ip_vs_get_dests __user
*uptr
)
2572 struct ip_vs_service
*svc
;
2573 union nf_inet_addr addr
= { .ip
= get
->addr
};
2578 svc
= __ip_vs_svc_fwm_find(ipvs
, AF_INET
, get
->fwmark
);
2580 svc
= __ip_vs_service_find(ipvs
, AF_INET
, get
->protocol
, &addr
,
2586 struct ip_vs_dest
*dest
;
2587 struct ip_vs_dest_entry entry
;
2588 struct ip_vs_kstats kstats
;
2590 memset(&entry
, 0, sizeof(entry
));
2591 list_for_each_entry(dest
, &svc
->destinations
, n_list
) {
2592 if (count
>= get
->num_dests
)
2595 /* Cannot expose heterogeneous members via sockopt
2598 if (dest
->af
!= svc
->af
)
2601 entry
.addr
= dest
->addr
.ip
;
2602 entry
.port
= dest
->port
;
2603 entry
.conn_flags
= atomic_read(&dest
->conn_flags
);
2604 entry
.weight
= atomic_read(&dest
->weight
);
2605 entry
.u_threshold
= dest
->u_threshold
;
2606 entry
.l_threshold
= dest
->l_threshold
;
2607 entry
.activeconns
= atomic_read(&dest
->activeconns
);
2608 entry
.inactconns
= atomic_read(&dest
->inactconns
);
2609 entry
.persistconns
= atomic_read(&dest
->persistconns
);
2610 ip_vs_copy_stats(&kstats
, &dest
->stats
);
2611 ip_vs_export_stats_user(&entry
.stats
, &kstats
);
2612 if (copy_to_user(&uptr
->entrytable
[count
],
2613 &entry
, sizeof(entry
))) {
2625 __ip_vs_get_timeouts(struct netns_ipvs
*ipvs
, struct ip_vs_timeout_user
*u
)
2627 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2628 struct ip_vs_proto_data
*pd
;
2631 memset(u
, 0, sizeof (*u
));
2633 #ifdef CONFIG_IP_VS_PROTO_TCP
2634 pd
= ip_vs_proto_data_get(ipvs
, IPPROTO_TCP
);
2635 u
->tcp_timeout
= pd
->timeout_table
[IP_VS_TCP_S_ESTABLISHED
] / HZ
;
2636 u
->tcp_fin_timeout
= pd
->timeout_table
[IP_VS_TCP_S_FIN_WAIT
] / HZ
;
2638 #ifdef CONFIG_IP_VS_PROTO_UDP
2639 pd
= ip_vs_proto_data_get(ipvs
, IPPROTO_UDP
);
2641 pd
->timeout_table
[IP_VS_UDP_S_NORMAL
] / HZ
;
2645 static const unsigned char get_arglen
[CMDID(IP_VS_SO_GET_MAX
) + 1] = {
2646 [CMDID(IP_VS_SO_GET_VERSION
)] = 64,
2647 [CMDID(IP_VS_SO_GET_INFO
)] = sizeof(struct ip_vs_getinfo
),
2648 [CMDID(IP_VS_SO_GET_SERVICES
)] = sizeof(struct ip_vs_get_services
),
2649 [CMDID(IP_VS_SO_GET_SERVICE
)] = sizeof(struct ip_vs_service_entry
),
2650 [CMDID(IP_VS_SO_GET_DESTS
)] = sizeof(struct ip_vs_get_dests
),
2651 [CMDID(IP_VS_SO_GET_TIMEOUT
)] = sizeof(struct ip_vs_timeout_user
),
2652 [CMDID(IP_VS_SO_GET_DAEMON
)] = 2 * sizeof(struct ip_vs_daemon_user
),
2655 union ip_vs_get_arglen
{
2656 char field_IP_VS_SO_GET_VERSION
[64];
2657 struct ip_vs_getinfo field_IP_VS_SO_GET_INFO
;
2658 struct ip_vs_get_services field_IP_VS_SO_GET_SERVICES
;
2659 struct ip_vs_service_entry field_IP_VS_SO_GET_SERVICE
;
2660 struct ip_vs_get_dests field_IP_VS_SO_GET_DESTS
;
2661 struct ip_vs_timeout_user field_IP_VS_SO_GET_TIMEOUT
;
2662 struct ip_vs_daemon_user field_IP_VS_SO_GET_DAEMON
[2];
2665 #define MAX_GET_ARGLEN sizeof(union ip_vs_get_arglen)
2668 do_ip_vs_get_ctl(struct sock
*sk
, int cmd
, void __user
*user
, int *len
)
2670 unsigned char arg
[MAX_GET_ARGLEN
];
2672 unsigned int copylen
;
2673 struct net
*net
= sock_net(sk
);
2674 struct netns_ipvs
*ipvs
= net_ipvs(net
);
2677 BUILD_BUG_ON(sizeof(arg
) > 255);
2678 if (!ns_capable(sock_net(sk
)->user_ns
, CAP_NET_ADMIN
))
2681 if (cmd
< IP_VS_BASE_CTL
|| cmd
> IP_VS_SO_GET_MAX
)
2684 copylen
= get_arglen
[CMDID(cmd
)];
2685 if (*len
< (int) copylen
) {
2686 IP_VS_DBG(1, "get_ctl: len %d < %u\n", *len
, copylen
);
2690 if (copy_from_user(arg
, user
, copylen
) != 0)
2693 * Handle daemons first since it has its own locking
2695 if (cmd
== IP_VS_SO_GET_DAEMON
) {
2696 struct ip_vs_daemon_user d
[2];
2698 memset(&d
, 0, sizeof(d
));
2699 mutex_lock(&ipvs
->sync_mutex
);
2700 if (ipvs
->sync_state
& IP_VS_STATE_MASTER
) {
2701 d
[0].state
= IP_VS_STATE_MASTER
;
2702 strlcpy(d
[0].mcast_ifn
, ipvs
->mcfg
.mcast_ifn
,
2703 sizeof(d
[0].mcast_ifn
));
2704 d
[0].syncid
= ipvs
->mcfg
.syncid
;
2706 if (ipvs
->sync_state
& IP_VS_STATE_BACKUP
) {
2707 d
[1].state
= IP_VS_STATE_BACKUP
;
2708 strlcpy(d
[1].mcast_ifn
, ipvs
->bcfg
.mcast_ifn
,
2709 sizeof(d
[1].mcast_ifn
));
2710 d
[1].syncid
= ipvs
->bcfg
.syncid
;
2712 if (copy_to_user(user
, &d
, sizeof(d
)) != 0)
2714 mutex_unlock(&ipvs
->sync_mutex
);
2718 mutex_lock(&__ip_vs_mutex
);
2720 case IP_VS_SO_GET_VERSION
:
2724 sprintf(buf
, "IP Virtual Server version %d.%d.%d (size=%d)",
2725 NVERSION(IP_VS_VERSION_CODE
), ip_vs_conn_tab_size
);
2726 if (copy_to_user(user
, buf
, strlen(buf
)+1) != 0) {
2730 *len
= strlen(buf
)+1;
2734 case IP_VS_SO_GET_INFO
:
2736 struct ip_vs_getinfo info
;
2737 info
.version
= IP_VS_VERSION_CODE
;
2738 info
.size
= ip_vs_conn_tab_size
;
2739 info
.num_services
= ipvs
->num_services
;
2740 if (copy_to_user(user
, &info
, sizeof(info
)) != 0)
2745 case IP_VS_SO_GET_SERVICES
:
2747 struct ip_vs_get_services
*get
;
2750 get
= (struct ip_vs_get_services
*)arg
;
2751 size
= sizeof(*get
) +
2752 sizeof(struct ip_vs_service_entry
) * get
->num_services
;
2754 pr_err("length: %u != %u\n", *len
, size
);
2758 ret
= __ip_vs_get_service_entries(ipvs
, get
, user
);
2762 case IP_VS_SO_GET_SERVICE
:
2764 struct ip_vs_service_entry
*entry
;
2765 struct ip_vs_service
*svc
;
2766 union nf_inet_addr addr
;
2768 entry
= (struct ip_vs_service_entry
*)arg
;
2769 addr
.ip
= entry
->addr
;
2772 svc
= __ip_vs_svc_fwm_find(ipvs
, AF_INET
, entry
->fwmark
);
2774 svc
= __ip_vs_service_find(ipvs
, AF_INET
,
2775 entry
->protocol
, &addr
,
2779 ip_vs_copy_service(entry
, svc
);
2780 if (copy_to_user(user
, entry
, sizeof(*entry
)) != 0)
2787 case IP_VS_SO_GET_DESTS
:
2789 struct ip_vs_get_dests
*get
;
2792 get
= (struct ip_vs_get_dests
*)arg
;
2793 size
= sizeof(*get
) +
2794 sizeof(struct ip_vs_dest_entry
) * get
->num_dests
;
2796 pr_err("length: %u != %u\n", *len
, size
);
2800 ret
= __ip_vs_get_dest_entries(ipvs
, get
, user
);
2804 case IP_VS_SO_GET_TIMEOUT
:
2806 struct ip_vs_timeout_user t
;
2808 __ip_vs_get_timeouts(ipvs
, &t
);
2809 if (copy_to_user(user
, &t
, sizeof(t
)) != 0)
2819 mutex_unlock(&__ip_vs_mutex
);
2824 static struct nf_sockopt_ops ip_vs_sockopts
= {
2826 .set_optmin
= IP_VS_BASE_CTL
,
2827 .set_optmax
= IP_VS_SO_SET_MAX
+1,
2828 .set
= do_ip_vs_set_ctl
,
2829 .get_optmin
= IP_VS_BASE_CTL
,
2830 .get_optmax
= IP_VS_SO_GET_MAX
+1,
2831 .get
= do_ip_vs_get_ctl
,
2832 .owner
= THIS_MODULE
,
2836 * Generic Netlink interface
2839 /* IPVS genetlink family */
2840 static struct genl_family ip_vs_genl_family
;
2842 /* Policy used for first-level command attributes */
2843 static const struct nla_policy ip_vs_cmd_policy
[IPVS_CMD_ATTR_MAX
+ 1] = {
2844 [IPVS_CMD_ATTR_SERVICE
] = { .type
= NLA_NESTED
},
2845 [IPVS_CMD_ATTR_DEST
] = { .type
= NLA_NESTED
},
2846 [IPVS_CMD_ATTR_DAEMON
] = { .type
= NLA_NESTED
},
2847 [IPVS_CMD_ATTR_TIMEOUT_TCP
] = { .type
= NLA_U32
},
2848 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN
] = { .type
= NLA_U32
},
2849 [IPVS_CMD_ATTR_TIMEOUT_UDP
] = { .type
= NLA_U32
},
2852 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2853 static const struct nla_policy ip_vs_daemon_policy
[IPVS_DAEMON_ATTR_MAX
+ 1] = {
2854 [IPVS_DAEMON_ATTR_STATE
] = { .type
= NLA_U32
},
2855 [IPVS_DAEMON_ATTR_MCAST_IFN
] = { .type
= NLA_NUL_STRING
,
2856 .len
= IP_VS_IFNAME_MAXLEN
},
2857 [IPVS_DAEMON_ATTR_SYNC_ID
] = { .type
= NLA_U32
},
2858 [IPVS_DAEMON_ATTR_SYNC_MAXLEN
] = { .type
= NLA_U16
},
2859 [IPVS_DAEMON_ATTR_MCAST_GROUP
] = { .type
= NLA_U32
},
2860 [IPVS_DAEMON_ATTR_MCAST_GROUP6
] = { .len
= sizeof(struct in6_addr
) },
2861 [IPVS_DAEMON_ATTR_MCAST_PORT
] = { .type
= NLA_U16
},
2862 [IPVS_DAEMON_ATTR_MCAST_TTL
] = { .type
= NLA_U8
},
2865 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2866 static const struct nla_policy ip_vs_svc_policy
[IPVS_SVC_ATTR_MAX
+ 1] = {
2867 [IPVS_SVC_ATTR_AF
] = { .type
= NLA_U16
},
2868 [IPVS_SVC_ATTR_PROTOCOL
] = { .type
= NLA_U16
},
2869 [IPVS_SVC_ATTR_ADDR
] = { .type
= NLA_BINARY
,
2870 .len
= sizeof(union nf_inet_addr
) },
2871 [IPVS_SVC_ATTR_PORT
] = { .type
= NLA_U16
},
2872 [IPVS_SVC_ATTR_FWMARK
] = { .type
= NLA_U32
},
2873 [IPVS_SVC_ATTR_SCHED_NAME
] = { .type
= NLA_NUL_STRING
,
2874 .len
= IP_VS_SCHEDNAME_MAXLEN
},
2875 [IPVS_SVC_ATTR_PE_NAME
] = { .type
= NLA_NUL_STRING
,
2876 .len
= IP_VS_PENAME_MAXLEN
},
2877 [IPVS_SVC_ATTR_FLAGS
] = { .type
= NLA_BINARY
,
2878 .len
= sizeof(struct ip_vs_flags
) },
2879 [IPVS_SVC_ATTR_TIMEOUT
] = { .type
= NLA_U32
},
2880 [IPVS_SVC_ATTR_NETMASK
] = { .type
= NLA_U32
},
2881 [IPVS_SVC_ATTR_STATS
] = { .type
= NLA_NESTED
},
2884 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2885 static const struct nla_policy ip_vs_dest_policy
[IPVS_DEST_ATTR_MAX
+ 1] = {
2886 [IPVS_DEST_ATTR_ADDR
] = { .type
= NLA_BINARY
,
2887 .len
= sizeof(union nf_inet_addr
) },
2888 [IPVS_DEST_ATTR_PORT
] = { .type
= NLA_U16
},
2889 [IPVS_DEST_ATTR_FWD_METHOD
] = { .type
= NLA_U32
},
2890 [IPVS_DEST_ATTR_WEIGHT
] = { .type
= NLA_U32
},
2891 [IPVS_DEST_ATTR_U_THRESH
] = { .type
= NLA_U32
},
2892 [IPVS_DEST_ATTR_L_THRESH
] = { .type
= NLA_U32
},
2893 [IPVS_DEST_ATTR_ACTIVE_CONNS
] = { .type
= NLA_U32
},
2894 [IPVS_DEST_ATTR_INACT_CONNS
] = { .type
= NLA_U32
},
2895 [IPVS_DEST_ATTR_PERSIST_CONNS
] = { .type
= NLA_U32
},
2896 [IPVS_DEST_ATTR_STATS
] = { .type
= NLA_NESTED
},
2897 [IPVS_DEST_ATTR_ADDR_FAMILY
] = { .type
= NLA_U16
},
2900 static int ip_vs_genl_fill_stats(struct sk_buff
*skb
, int container_type
,
2901 struct ip_vs_kstats
*kstats
)
2903 struct nlattr
*nl_stats
= nla_nest_start(skb
, container_type
);
2908 if (nla_put_u32(skb
, IPVS_STATS_ATTR_CONNS
, (u32
)kstats
->conns
) ||
2909 nla_put_u32(skb
, IPVS_STATS_ATTR_INPKTS
, (u32
)kstats
->inpkts
) ||
2910 nla_put_u32(skb
, IPVS_STATS_ATTR_OUTPKTS
, (u32
)kstats
->outpkts
) ||
2911 nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_INBYTES
, kstats
->inbytes
,
2912 IPVS_STATS_ATTR_PAD
) ||
2913 nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_OUTBYTES
, kstats
->outbytes
,
2914 IPVS_STATS_ATTR_PAD
) ||
2915 nla_put_u32(skb
, IPVS_STATS_ATTR_CPS
, (u32
)kstats
->cps
) ||
2916 nla_put_u32(skb
, IPVS_STATS_ATTR_INPPS
, (u32
)kstats
->inpps
) ||
2917 nla_put_u32(skb
, IPVS_STATS_ATTR_OUTPPS
, (u32
)kstats
->outpps
) ||
2918 nla_put_u32(skb
, IPVS_STATS_ATTR_INBPS
, (u32
)kstats
->inbps
) ||
2919 nla_put_u32(skb
, IPVS_STATS_ATTR_OUTBPS
, (u32
)kstats
->outbps
))
2920 goto nla_put_failure
;
2921 nla_nest_end(skb
, nl_stats
);
2926 nla_nest_cancel(skb
, nl_stats
);
2930 static int ip_vs_genl_fill_stats64(struct sk_buff
*skb
, int container_type
,
2931 struct ip_vs_kstats
*kstats
)
2933 struct nlattr
*nl_stats
= nla_nest_start(skb
, container_type
);
2938 if (nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_CONNS
, kstats
->conns
,
2939 IPVS_STATS_ATTR_PAD
) ||
2940 nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_INPKTS
, kstats
->inpkts
,
2941 IPVS_STATS_ATTR_PAD
) ||
2942 nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_OUTPKTS
, kstats
->outpkts
,
2943 IPVS_STATS_ATTR_PAD
) ||
2944 nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_INBYTES
, kstats
->inbytes
,
2945 IPVS_STATS_ATTR_PAD
) ||
2946 nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_OUTBYTES
, kstats
->outbytes
,
2947 IPVS_STATS_ATTR_PAD
) ||
2948 nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_CPS
, kstats
->cps
,
2949 IPVS_STATS_ATTR_PAD
) ||
2950 nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_INPPS
, kstats
->inpps
,
2951 IPVS_STATS_ATTR_PAD
) ||
2952 nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_OUTPPS
, kstats
->outpps
,
2953 IPVS_STATS_ATTR_PAD
) ||
2954 nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_INBPS
, kstats
->inbps
,
2955 IPVS_STATS_ATTR_PAD
) ||
2956 nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_OUTBPS
, kstats
->outbps
,
2957 IPVS_STATS_ATTR_PAD
))
2958 goto nla_put_failure
;
2959 nla_nest_end(skb
, nl_stats
);
2964 nla_nest_cancel(skb
, nl_stats
);
2968 static int ip_vs_genl_fill_service(struct sk_buff
*skb
,
2969 struct ip_vs_service
*svc
)
2971 struct ip_vs_scheduler
*sched
;
2972 struct ip_vs_pe
*pe
;
2973 struct nlattr
*nl_service
;
2974 struct ip_vs_flags flags
= { .flags
= svc
->flags
,
2976 struct ip_vs_kstats kstats
;
2979 nl_service
= nla_nest_start(skb
, IPVS_CMD_ATTR_SERVICE
);
2983 if (nla_put_u16(skb
, IPVS_SVC_ATTR_AF
, svc
->af
))
2984 goto nla_put_failure
;
2986 if (nla_put_u32(skb
, IPVS_SVC_ATTR_FWMARK
, svc
->fwmark
))
2987 goto nla_put_failure
;
2989 if (nla_put_u16(skb
, IPVS_SVC_ATTR_PROTOCOL
, svc
->protocol
) ||
2990 nla_put(skb
, IPVS_SVC_ATTR_ADDR
, sizeof(svc
->addr
), &svc
->addr
) ||
2991 nla_put_be16(skb
, IPVS_SVC_ATTR_PORT
, svc
->port
))
2992 goto nla_put_failure
;
2995 sched
= rcu_dereference_protected(svc
->scheduler
, 1);
2996 sched_name
= sched
? sched
->name
: "none";
2997 pe
= rcu_dereference_protected(svc
->pe
, 1);
2998 if (nla_put_string(skb
, IPVS_SVC_ATTR_SCHED_NAME
, sched_name
) ||
2999 (pe
&& nla_put_string(skb
, IPVS_SVC_ATTR_PE_NAME
, pe
->name
)) ||
3000 nla_put(skb
, IPVS_SVC_ATTR_FLAGS
, sizeof(flags
), &flags
) ||
3001 nla_put_u32(skb
, IPVS_SVC_ATTR_TIMEOUT
, svc
->timeout
/ HZ
) ||
3002 nla_put_be32(skb
, IPVS_SVC_ATTR_NETMASK
, svc
->netmask
))
3003 goto nla_put_failure
;
3004 ip_vs_copy_stats(&kstats
, &svc
->stats
);
3005 if (ip_vs_genl_fill_stats(skb
, IPVS_SVC_ATTR_STATS
, &kstats
))
3006 goto nla_put_failure
;
3007 if (ip_vs_genl_fill_stats64(skb
, IPVS_SVC_ATTR_STATS64
, &kstats
))
3008 goto nla_put_failure
;
3010 nla_nest_end(skb
, nl_service
);
3015 nla_nest_cancel(skb
, nl_service
);
3019 static int ip_vs_genl_dump_service(struct sk_buff
*skb
,
3020 struct ip_vs_service
*svc
,
3021 struct netlink_callback
*cb
)
3025 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
3026 &ip_vs_genl_family
, NLM_F_MULTI
,
3027 IPVS_CMD_NEW_SERVICE
);
3031 if (ip_vs_genl_fill_service(skb
, svc
) < 0)
3032 goto nla_put_failure
;
3034 genlmsg_end(skb
, hdr
);
3038 genlmsg_cancel(skb
, hdr
);
3042 static int ip_vs_genl_dump_services(struct sk_buff
*skb
,
3043 struct netlink_callback
*cb
)
3046 int start
= cb
->args
[0];
3047 struct ip_vs_service
*svc
;
3048 struct net
*net
= sock_net(skb
->sk
);
3049 struct netns_ipvs
*ipvs
= net_ipvs(net
);
3051 mutex_lock(&__ip_vs_mutex
);
3052 for (i
= 0; i
< IP_VS_SVC_TAB_SIZE
; i
++) {
3053 hlist_for_each_entry(svc
, &ip_vs_svc_table
[i
], s_list
) {
3054 if (++idx
<= start
|| (svc
->ipvs
!= ipvs
))
3056 if (ip_vs_genl_dump_service(skb
, svc
, cb
) < 0) {
3058 goto nla_put_failure
;
3063 for (i
= 0; i
< IP_VS_SVC_TAB_SIZE
; i
++) {
3064 hlist_for_each_entry(svc
, &ip_vs_svc_fwm_table
[i
], f_list
) {
3065 if (++idx
<= start
|| (svc
->ipvs
!= ipvs
))
3067 if (ip_vs_genl_dump_service(skb
, svc
, cb
) < 0) {
3069 goto nla_put_failure
;
3075 mutex_unlock(&__ip_vs_mutex
);
3081 static bool ip_vs_is_af_valid(int af
)
3085 #ifdef CONFIG_IP_VS_IPV6
3086 if (af
== AF_INET6
&& ipv6_mod_enabled())
3092 static int ip_vs_genl_parse_service(struct netns_ipvs
*ipvs
,
3093 struct ip_vs_service_user_kern
*usvc
,
3094 struct nlattr
*nla
, int full_entry
,
3095 struct ip_vs_service
**ret_svc
)
3097 struct nlattr
*attrs
[IPVS_SVC_ATTR_MAX
+ 1];
3098 struct nlattr
*nla_af
, *nla_port
, *nla_fwmark
, *nla_protocol
, *nla_addr
;
3099 struct ip_vs_service
*svc
;
3101 /* Parse mandatory identifying service fields first */
3103 nla_parse_nested(attrs
, IPVS_SVC_ATTR_MAX
, nla
,
3104 ip_vs_svc_policy
, NULL
))
3107 nla_af
= attrs
[IPVS_SVC_ATTR_AF
];
3108 nla_protocol
= attrs
[IPVS_SVC_ATTR_PROTOCOL
];
3109 nla_addr
= attrs
[IPVS_SVC_ATTR_ADDR
];
3110 nla_port
= attrs
[IPVS_SVC_ATTR_PORT
];
3111 nla_fwmark
= attrs
[IPVS_SVC_ATTR_FWMARK
];
3113 if (!(nla_af
&& (nla_fwmark
|| (nla_port
&& nla_protocol
&& nla_addr
))))
3116 memset(usvc
, 0, sizeof(*usvc
));
3118 usvc
->af
= nla_get_u16(nla_af
);
3119 if (!ip_vs_is_af_valid(usvc
->af
))
3120 return -EAFNOSUPPORT
;
3123 usvc
->protocol
= IPPROTO_TCP
;
3124 usvc
->fwmark
= nla_get_u32(nla_fwmark
);
3126 usvc
->protocol
= nla_get_u16(nla_protocol
);
3127 nla_memcpy(&usvc
->addr
, nla_addr
, sizeof(usvc
->addr
));
3128 usvc
->port
= nla_get_be16(nla_port
);
3134 svc
= __ip_vs_svc_fwm_find(ipvs
, usvc
->af
, usvc
->fwmark
);
3136 svc
= __ip_vs_service_find(ipvs
, usvc
->af
, usvc
->protocol
,
3137 &usvc
->addr
, usvc
->port
);
3141 /* If a full entry was requested, check for the additional fields */
3143 struct nlattr
*nla_sched
, *nla_flags
, *nla_pe
, *nla_timeout
,
3145 struct ip_vs_flags flags
;
3147 nla_sched
= attrs
[IPVS_SVC_ATTR_SCHED_NAME
];
3148 nla_pe
= attrs
[IPVS_SVC_ATTR_PE_NAME
];
3149 nla_flags
= attrs
[IPVS_SVC_ATTR_FLAGS
];
3150 nla_timeout
= attrs
[IPVS_SVC_ATTR_TIMEOUT
];
3151 nla_netmask
= attrs
[IPVS_SVC_ATTR_NETMASK
];
3153 if (!(nla_sched
&& nla_flags
&& nla_timeout
&& nla_netmask
))
3156 nla_memcpy(&flags
, nla_flags
, sizeof(flags
));
3158 /* prefill flags from service if it already exists */
3160 usvc
->flags
= svc
->flags
;
3162 /* set new flags from userland */
3163 usvc
->flags
= (usvc
->flags
& ~flags
.mask
) |
3164 (flags
.flags
& flags
.mask
);
3165 usvc
->sched_name
= nla_data(nla_sched
);
3166 usvc
->pe_name
= nla_pe
? nla_data(nla_pe
) : NULL
;
3167 usvc
->timeout
= nla_get_u32(nla_timeout
);
3168 usvc
->netmask
= nla_get_be32(nla_netmask
);
3174 static struct ip_vs_service
*ip_vs_genl_find_service(struct netns_ipvs
*ipvs
,
3177 struct ip_vs_service_user_kern usvc
;
3178 struct ip_vs_service
*svc
;
3181 ret
= ip_vs_genl_parse_service(ipvs
, &usvc
, nla
, 0, &svc
);
3182 return ret
? ERR_PTR(ret
) : svc
;
3185 static int ip_vs_genl_fill_dest(struct sk_buff
*skb
, struct ip_vs_dest
*dest
)
3187 struct nlattr
*nl_dest
;
3188 struct ip_vs_kstats kstats
;
3190 nl_dest
= nla_nest_start(skb
, IPVS_CMD_ATTR_DEST
);
3194 if (nla_put(skb
, IPVS_DEST_ATTR_ADDR
, sizeof(dest
->addr
), &dest
->addr
) ||
3195 nla_put_be16(skb
, IPVS_DEST_ATTR_PORT
, dest
->port
) ||
3196 nla_put_u32(skb
, IPVS_DEST_ATTR_FWD_METHOD
,
3197 (atomic_read(&dest
->conn_flags
) &
3198 IP_VS_CONN_F_FWD_MASK
)) ||
3199 nla_put_u32(skb
, IPVS_DEST_ATTR_WEIGHT
,
3200 atomic_read(&dest
->weight
)) ||
3201 nla_put_u32(skb
, IPVS_DEST_ATTR_U_THRESH
, dest
->u_threshold
) ||
3202 nla_put_u32(skb
, IPVS_DEST_ATTR_L_THRESH
, dest
->l_threshold
) ||
3203 nla_put_u32(skb
, IPVS_DEST_ATTR_ACTIVE_CONNS
,
3204 atomic_read(&dest
->activeconns
)) ||
3205 nla_put_u32(skb
, IPVS_DEST_ATTR_INACT_CONNS
,
3206 atomic_read(&dest
->inactconns
)) ||
3207 nla_put_u32(skb
, IPVS_DEST_ATTR_PERSIST_CONNS
,
3208 atomic_read(&dest
->persistconns
)) ||
3209 nla_put_u16(skb
, IPVS_DEST_ATTR_ADDR_FAMILY
, dest
->af
))
3210 goto nla_put_failure
;
3211 ip_vs_copy_stats(&kstats
, &dest
->stats
);
3212 if (ip_vs_genl_fill_stats(skb
, IPVS_DEST_ATTR_STATS
, &kstats
))
3213 goto nla_put_failure
;
3214 if (ip_vs_genl_fill_stats64(skb
, IPVS_DEST_ATTR_STATS64
, &kstats
))
3215 goto nla_put_failure
;
3217 nla_nest_end(skb
, nl_dest
);
3222 nla_nest_cancel(skb
, nl_dest
);
3226 static int ip_vs_genl_dump_dest(struct sk_buff
*skb
, struct ip_vs_dest
*dest
,
3227 struct netlink_callback
*cb
)
3231 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
3232 &ip_vs_genl_family
, NLM_F_MULTI
,
3237 if (ip_vs_genl_fill_dest(skb
, dest
) < 0)
3238 goto nla_put_failure
;
3240 genlmsg_end(skb
, hdr
);
3244 genlmsg_cancel(skb
, hdr
);
3248 static int ip_vs_genl_dump_dests(struct sk_buff
*skb
,
3249 struct netlink_callback
*cb
)
3252 int start
= cb
->args
[0];
3253 struct ip_vs_service
*svc
;
3254 struct ip_vs_dest
*dest
;
3255 struct nlattr
*attrs
[IPVS_CMD_ATTR_MAX
+ 1];
3256 struct net
*net
= sock_net(skb
->sk
);
3257 struct netns_ipvs
*ipvs
= net_ipvs(net
);
3259 mutex_lock(&__ip_vs_mutex
);
3261 /* Try to find the service for which to dump destinations */
3262 if (nlmsg_parse(cb
->nlh
, GENL_HDRLEN
, attrs
, IPVS_CMD_ATTR_MAX
,
3263 ip_vs_cmd_policy
, NULL
))
3267 svc
= ip_vs_genl_find_service(ipvs
, attrs
[IPVS_CMD_ATTR_SERVICE
]);
3268 if (IS_ERR_OR_NULL(svc
))
3271 /* Dump the destinations */
3272 list_for_each_entry(dest
, &svc
->destinations
, n_list
) {
3275 if (ip_vs_genl_dump_dest(skb
, dest
, cb
) < 0) {
3277 goto nla_put_failure
;
3285 mutex_unlock(&__ip_vs_mutex
);
3290 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern
*udest
,
3291 struct nlattr
*nla
, int full_entry
)
3293 struct nlattr
*attrs
[IPVS_DEST_ATTR_MAX
+ 1];
3294 struct nlattr
*nla_addr
, *nla_port
;
3295 struct nlattr
*nla_addr_family
;
3297 /* Parse mandatory identifying destination fields first */
3299 nla_parse_nested(attrs
, IPVS_DEST_ATTR_MAX
, nla
,
3300 ip_vs_dest_policy
, NULL
))
3303 nla_addr
= attrs
[IPVS_DEST_ATTR_ADDR
];
3304 nla_port
= attrs
[IPVS_DEST_ATTR_PORT
];
3305 nla_addr_family
= attrs
[IPVS_DEST_ATTR_ADDR_FAMILY
];
3307 if (!(nla_addr
&& nla_port
))
3310 memset(udest
, 0, sizeof(*udest
));
3312 nla_memcpy(&udest
->addr
, nla_addr
, sizeof(udest
->addr
));
3313 udest
->port
= nla_get_be16(nla_port
);
3315 if (nla_addr_family
)
3316 udest
->af
= nla_get_u16(nla_addr_family
);
3320 /* If a full entry was requested, check for the additional fields */
3322 struct nlattr
*nla_fwd
, *nla_weight
, *nla_u_thresh
,
3325 nla_fwd
= attrs
[IPVS_DEST_ATTR_FWD_METHOD
];
3326 nla_weight
= attrs
[IPVS_DEST_ATTR_WEIGHT
];
3327 nla_u_thresh
= attrs
[IPVS_DEST_ATTR_U_THRESH
];
3328 nla_l_thresh
= attrs
[IPVS_DEST_ATTR_L_THRESH
];
3330 if (!(nla_fwd
&& nla_weight
&& nla_u_thresh
&& nla_l_thresh
))
3333 udest
->conn_flags
= nla_get_u32(nla_fwd
)
3334 & IP_VS_CONN_F_FWD_MASK
;
3335 udest
->weight
= nla_get_u32(nla_weight
);
3336 udest
->u_threshold
= nla_get_u32(nla_u_thresh
);
3337 udest
->l_threshold
= nla_get_u32(nla_l_thresh
);
3343 static int ip_vs_genl_fill_daemon(struct sk_buff
*skb
, __u32 state
,
3344 struct ipvs_sync_daemon_cfg
*c
)
3346 struct nlattr
*nl_daemon
;
3348 nl_daemon
= nla_nest_start(skb
, IPVS_CMD_ATTR_DAEMON
);
3352 if (nla_put_u32(skb
, IPVS_DAEMON_ATTR_STATE
, state
) ||
3353 nla_put_string(skb
, IPVS_DAEMON_ATTR_MCAST_IFN
, c
->mcast_ifn
) ||
3354 nla_put_u32(skb
, IPVS_DAEMON_ATTR_SYNC_ID
, c
->syncid
) ||
3355 nla_put_u16(skb
, IPVS_DAEMON_ATTR_SYNC_MAXLEN
, c
->sync_maxlen
) ||
3356 nla_put_u16(skb
, IPVS_DAEMON_ATTR_MCAST_PORT
, c
->mcast_port
) ||
3357 nla_put_u8(skb
, IPVS_DAEMON_ATTR_MCAST_TTL
, c
->mcast_ttl
))
3358 goto nla_put_failure
;
3359 #ifdef CONFIG_IP_VS_IPV6
3360 if (c
->mcast_af
== AF_INET6
) {
3361 if (nla_put_in6_addr(skb
, IPVS_DAEMON_ATTR_MCAST_GROUP6
,
3362 &c
->mcast_group
.in6
))
3363 goto nla_put_failure
;
3366 if (c
->mcast_af
== AF_INET
&&
3367 nla_put_in_addr(skb
, IPVS_DAEMON_ATTR_MCAST_GROUP
,
3369 goto nla_put_failure
;
3370 nla_nest_end(skb
, nl_daemon
);
3375 nla_nest_cancel(skb
, nl_daemon
);
3379 static int ip_vs_genl_dump_daemon(struct sk_buff
*skb
, __u32 state
,
3380 struct ipvs_sync_daemon_cfg
*c
,
3381 struct netlink_callback
*cb
)
3384 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
3385 &ip_vs_genl_family
, NLM_F_MULTI
,
3386 IPVS_CMD_NEW_DAEMON
);
3390 if (ip_vs_genl_fill_daemon(skb
, state
, c
))
3391 goto nla_put_failure
;
3393 genlmsg_end(skb
, hdr
);
3397 genlmsg_cancel(skb
, hdr
);
3401 static int ip_vs_genl_dump_daemons(struct sk_buff
*skb
,
3402 struct netlink_callback
*cb
)
3404 struct net
*net
= sock_net(skb
->sk
);
3405 struct netns_ipvs
*ipvs
= net_ipvs(net
);
3407 mutex_lock(&ipvs
->sync_mutex
);
3408 if ((ipvs
->sync_state
& IP_VS_STATE_MASTER
) && !cb
->args
[0]) {
3409 if (ip_vs_genl_dump_daemon(skb
, IP_VS_STATE_MASTER
,
3410 &ipvs
->mcfg
, cb
) < 0)
3411 goto nla_put_failure
;
3416 if ((ipvs
->sync_state
& IP_VS_STATE_BACKUP
) && !cb
->args
[1]) {
3417 if (ip_vs_genl_dump_daemon(skb
, IP_VS_STATE_BACKUP
,
3418 &ipvs
->bcfg
, cb
) < 0)
3419 goto nla_put_failure
;
3425 mutex_unlock(&ipvs
->sync_mutex
);
3430 static int ip_vs_genl_new_daemon(struct netns_ipvs
*ipvs
, struct nlattr
**attrs
)
3432 struct ipvs_sync_daemon_cfg c
;
3436 memset(&c
, 0, sizeof(c
));
3437 if (!(attrs
[IPVS_DAEMON_ATTR_STATE
] &&
3438 attrs
[IPVS_DAEMON_ATTR_MCAST_IFN
] &&
3439 attrs
[IPVS_DAEMON_ATTR_SYNC_ID
]))
3441 strlcpy(c
.mcast_ifn
, nla_data(attrs
[IPVS_DAEMON_ATTR_MCAST_IFN
]),
3442 sizeof(c
.mcast_ifn
));
3443 c
.syncid
= nla_get_u32(attrs
[IPVS_DAEMON_ATTR_SYNC_ID
]);
3445 a
= attrs
[IPVS_DAEMON_ATTR_SYNC_MAXLEN
];
3447 c
.sync_maxlen
= nla_get_u16(a
);
3449 a
= attrs
[IPVS_DAEMON_ATTR_MCAST_GROUP
];
3451 c
.mcast_af
= AF_INET
;
3452 c
.mcast_group
.ip
= nla_get_in_addr(a
);
3453 if (!ipv4_is_multicast(c
.mcast_group
.ip
))
3456 a
= attrs
[IPVS_DAEMON_ATTR_MCAST_GROUP6
];
3458 #ifdef CONFIG_IP_VS_IPV6
3461 c
.mcast_af
= AF_INET6
;
3462 c
.mcast_group
.in6
= nla_get_in6_addr(a
);
3463 addr_type
= ipv6_addr_type(&c
.mcast_group
.in6
);
3464 if (!(addr_type
& IPV6_ADDR_MULTICAST
))
3467 return -EAFNOSUPPORT
;
3472 a
= attrs
[IPVS_DAEMON_ATTR_MCAST_PORT
];
3474 c
.mcast_port
= nla_get_u16(a
);
3476 a
= attrs
[IPVS_DAEMON_ATTR_MCAST_TTL
];
3478 c
.mcast_ttl
= nla_get_u8(a
);
3480 /* The synchronization protocol is incompatible with mixed family
3483 if (ipvs
->mixed_address_family_dests
> 0)
3487 mutex_lock(&ipvs
->sync_mutex
);
3488 ret
= start_sync_thread(ipvs
, &c
,
3489 nla_get_u32(attrs
[IPVS_DAEMON_ATTR_STATE
]));
3490 mutex_unlock(&ipvs
->sync_mutex
);
3495 static int ip_vs_genl_del_daemon(struct netns_ipvs
*ipvs
, struct nlattr
**attrs
)
3499 if (!attrs
[IPVS_DAEMON_ATTR_STATE
])
3502 mutex_lock(&ipvs
->sync_mutex
);
3503 ret
= stop_sync_thread(ipvs
,
3504 nla_get_u32(attrs
[IPVS_DAEMON_ATTR_STATE
]));
3505 mutex_unlock(&ipvs
->sync_mutex
);
3509 static int ip_vs_genl_set_config(struct netns_ipvs
*ipvs
, struct nlattr
**attrs
)
3511 struct ip_vs_timeout_user t
;
3513 __ip_vs_get_timeouts(ipvs
, &t
);
3515 if (attrs
[IPVS_CMD_ATTR_TIMEOUT_TCP
])
3516 t
.tcp_timeout
= nla_get_u32(attrs
[IPVS_CMD_ATTR_TIMEOUT_TCP
]);
3518 if (attrs
[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN
])
3520 nla_get_u32(attrs
[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN
]);
3522 if (attrs
[IPVS_CMD_ATTR_TIMEOUT_UDP
])
3523 t
.udp_timeout
= nla_get_u32(attrs
[IPVS_CMD_ATTR_TIMEOUT_UDP
]);
3525 return ip_vs_set_timeout(ipvs
, &t
);
3528 static int ip_vs_genl_set_daemon(struct sk_buff
*skb
, struct genl_info
*info
)
3530 int ret
= -EINVAL
, cmd
;
3531 struct net
*net
= sock_net(skb
->sk
);
3532 struct netns_ipvs
*ipvs
= net_ipvs(net
);
3534 cmd
= info
->genlhdr
->cmd
;
3536 if (cmd
== IPVS_CMD_NEW_DAEMON
|| cmd
== IPVS_CMD_DEL_DAEMON
) {
3537 struct nlattr
*daemon_attrs
[IPVS_DAEMON_ATTR_MAX
+ 1];
3539 if (!info
->attrs
[IPVS_CMD_ATTR_DAEMON
] ||
3540 nla_parse_nested(daemon_attrs
, IPVS_DAEMON_ATTR_MAX
,
3541 info
->attrs
[IPVS_CMD_ATTR_DAEMON
],
3542 ip_vs_daemon_policy
, info
->extack
))
3545 if (cmd
== IPVS_CMD_NEW_DAEMON
)
3546 ret
= ip_vs_genl_new_daemon(ipvs
, daemon_attrs
);
3548 ret
= ip_vs_genl_del_daemon(ipvs
, daemon_attrs
);
3555 static int ip_vs_genl_set_cmd(struct sk_buff
*skb
, struct genl_info
*info
)
3557 struct ip_vs_service
*svc
= NULL
;
3558 struct ip_vs_service_user_kern usvc
;
3559 struct ip_vs_dest_user_kern udest
;
3561 int need_full_svc
= 0, need_full_dest
= 0;
3562 struct net
*net
= sock_net(skb
->sk
);
3563 struct netns_ipvs
*ipvs
= net_ipvs(net
);
3565 cmd
= info
->genlhdr
->cmd
;
3567 mutex_lock(&__ip_vs_mutex
);
3569 if (cmd
== IPVS_CMD_FLUSH
) {
3570 ret
= ip_vs_flush(ipvs
, false);
3572 } else if (cmd
== IPVS_CMD_SET_CONFIG
) {
3573 ret
= ip_vs_genl_set_config(ipvs
, info
->attrs
);
3575 } else if (cmd
== IPVS_CMD_ZERO
&&
3576 !info
->attrs
[IPVS_CMD_ATTR_SERVICE
]) {
3577 ret
= ip_vs_zero_all(ipvs
);
3581 /* All following commands require a service argument, so check if we
3582 * received a valid one. We need a full service specification when
3583 * adding / editing a service. Only identifying members otherwise. */
3584 if (cmd
== IPVS_CMD_NEW_SERVICE
|| cmd
== IPVS_CMD_SET_SERVICE
)
3587 ret
= ip_vs_genl_parse_service(ipvs
, &usvc
,
3588 info
->attrs
[IPVS_CMD_ATTR_SERVICE
],
3589 need_full_svc
, &svc
);
3593 /* Unless we're adding a new service, the service must already exist */
3594 if ((cmd
!= IPVS_CMD_NEW_SERVICE
) && (svc
== NULL
)) {
3599 /* Destination commands require a valid destination argument. For
3600 * adding / editing a destination, we need a full destination
3602 if (cmd
== IPVS_CMD_NEW_DEST
|| cmd
== IPVS_CMD_SET_DEST
||
3603 cmd
== IPVS_CMD_DEL_DEST
) {
3604 if (cmd
!= IPVS_CMD_DEL_DEST
)
3607 ret
= ip_vs_genl_parse_dest(&udest
,
3608 info
->attrs
[IPVS_CMD_ATTR_DEST
],
3613 /* Old protocols did not allow the user to specify address
3614 * family, so we set it to zero instead. We also didn't
3615 * allow heterogeneous pools in the old code, so it's safe
3616 * to assume that this will have the same address family as
3622 if (!ip_vs_is_af_valid(udest
.af
)) {
3623 ret
= -EAFNOSUPPORT
;
3627 if (udest
.af
!= svc
->af
&& cmd
!= IPVS_CMD_DEL_DEST
) {
3628 /* The synchronization protocol is incompatible
3629 * with mixed family services
3631 if (ipvs
->sync_state
) {
3636 /* Which connection types do we support? */
3637 switch (udest
.conn_flags
) {
3638 case IP_VS_CONN_F_TUNNEL
:
3639 /* We are able to forward this */
3649 case IPVS_CMD_NEW_SERVICE
:
3651 ret
= ip_vs_add_service(ipvs
, &usvc
, &svc
);
3655 case IPVS_CMD_SET_SERVICE
:
3656 ret
= ip_vs_edit_service(svc
, &usvc
);
3658 case IPVS_CMD_DEL_SERVICE
:
3659 ret
= ip_vs_del_service(svc
);
3660 /* do not use svc, it can be freed */
3662 case IPVS_CMD_NEW_DEST
:
3663 ret
= ip_vs_add_dest(svc
, &udest
);
3665 case IPVS_CMD_SET_DEST
:
3666 ret
= ip_vs_edit_dest(svc
, &udest
);
3668 case IPVS_CMD_DEL_DEST
:
3669 ret
= ip_vs_del_dest(svc
, &udest
);
3672 ret
= ip_vs_zero_service(svc
);
3679 mutex_unlock(&__ip_vs_mutex
);
3684 static int ip_vs_genl_get_cmd(struct sk_buff
*skb
, struct genl_info
*info
)
3686 struct sk_buff
*msg
;
3688 int ret
, cmd
, reply_cmd
;
3689 struct net
*net
= sock_net(skb
->sk
);
3690 struct netns_ipvs
*ipvs
= net_ipvs(net
);
3692 cmd
= info
->genlhdr
->cmd
;
3694 if (cmd
== IPVS_CMD_GET_SERVICE
)
3695 reply_cmd
= IPVS_CMD_NEW_SERVICE
;
3696 else if (cmd
== IPVS_CMD_GET_INFO
)
3697 reply_cmd
= IPVS_CMD_SET_INFO
;
3698 else if (cmd
== IPVS_CMD_GET_CONFIG
)
3699 reply_cmd
= IPVS_CMD_SET_CONFIG
;
3701 pr_err("unknown Generic Netlink command\n");
3705 msg
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
3709 mutex_lock(&__ip_vs_mutex
);
3711 reply
= genlmsg_put_reply(msg
, info
, &ip_vs_genl_family
, 0, reply_cmd
);
3713 goto nla_put_failure
;
3716 case IPVS_CMD_GET_SERVICE
:
3718 struct ip_vs_service
*svc
;
3720 svc
= ip_vs_genl_find_service(ipvs
,
3721 info
->attrs
[IPVS_CMD_ATTR_SERVICE
]);
3726 ret
= ip_vs_genl_fill_service(msg
, svc
);
3728 goto nla_put_failure
;
3737 case IPVS_CMD_GET_CONFIG
:
3739 struct ip_vs_timeout_user t
;
3741 __ip_vs_get_timeouts(ipvs
, &t
);
3742 #ifdef CONFIG_IP_VS_PROTO_TCP
3743 if (nla_put_u32(msg
, IPVS_CMD_ATTR_TIMEOUT_TCP
,
3745 nla_put_u32(msg
, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN
,
3747 goto nla_put_failure
;
3749 #ifdef CONFIG_IP_VS_PROTO_UDP
3750 if (nla_put_u32(msg
, IPVS_CMD_ATTR_TIMEOUT_UDP
, t
.udp_timeout
))
3751 goto nla_put_failure
;
3757 case IPVS_CMD_GET_INFO
:
3758 if (nla_put_u32(msg
, IPVS_INFO_ATTR_VERSION
,
3759 IP_VS_VERSION_CODE
) ||
3760 nla_put_u32(msg
, IPVS_INFO_ATTR_CONN_TAB_SIZE
,
3761 ip_vs_conn_tab_size
))
3762 goto nla_put_failure
;
3766 genlmsg_end(msg
, reply
);
3767 ret
= genlmsg_reply(msg
, info
);
3771 pr_err("not enough space in Netlink message\n");
3777 mutex_unlock(&__ip_vs_mutex
);
3783 static const struct genl_ops ip_vs_genl_ops
[] = {
3785 .cmd
= IPVS_CMD_NEW_SERVICE
,
3786 .flags
= GENL_ADMIN_PERM
,
3787 .policy
= ip_vs_cmd_policy
,
3788 .doit
= ip_vs_genl_set_cmd
,
3791 .cmd
= IPVS_CMD_SET_SERVICE
,
3792 .flags
= GENL_ADMIN_PERM
,
3793 .policy
= ip_vs_cmd_policy
,
3794 .doit
= ip_vs_genl_set_cmd
,
3797 .cmd
= IPVS_CMD_DEL_SERVICE
,
3798 .flags
= GENL_ADMIN_PERM
,
3799 .policy
= ip_vs_cmd_policy
,
3800 .doit
= ip_vs_genl_set_cmd
,
3803 .cmd
= IPVS_CMD_GET_SERVICE
,
3804 .flags
= GENL_ADMIN_PERM
,
3805 .doit
= ip_vs_genl_get_cmd
,
3806 .dumpit
= ip_vs_genl_dump_services
,
3807 .policy
= ip_vs_cmd_policy
,
3810 .cmd
= IPVS_CMD_NEW_DEST
,
3811 .flags
= GENL_ADMIN_PERM
,
3812 .policy
= ip_vs_cmd_policy
,
3813 .doit
= ip_vs_genl_set_cmd
,
3816 .cmd
= IPVS_CMD_SET_DEST
,
3817 .flags
= GENL_ADMIN_PERM
,
3818 .policy
= ip_vs_cmd_policy
,
3819 .doit
= ip_vs_genl_set_cmd
,
3822 .cmd
= IPVS_CMD_DEL_DEST
,
3823 .flags
= GENL_ADMIN_PERM
,
3824 .policy
= ip_vs_cmd_policy
,
3825 .doit
= ip_vs_genl_set_cmd
,
3828 .cmd
= IPVS_CMD_GET_DEST
,
3829 .flags
= GENL_ADMIN_PERM
,
3830 .policy
= ip_vs_cmd_policy
,
3831 .dumpit
= ip_vs_genl_dump_dests
,
3834 .cmd
= IPVS_CMD_NEW_DAEMON
,
3835 .flags
= GENL_ADMIN_PERM
,
3836 .policy
= ip_vs_cmd_policy
,
3837 .doit
= ip_vs_genl_set_daemon
,
3840 .cmd
= IPVS_CMD_DEL_DAEMON
,
3841 .flags
= GENL_ADMIN_PERM
,
3842 .policy
= ip_vs_cmd_policy
,
3843 .doit
= ip_vs_genl_set_daemon
,
3846 .cmd
= IPVS_CMD_GET_DAEMON
,
3847 .flags
= GENL_ADMIN_PERM
,
3848 .dumpit
= ip_vs_genl_dump_daemons
,
3851 .cmd
= IPVS_CMD_SET_CONFIG
,
3852 .flags
= GENL_ADMIN_PERM
,
3853 .policy
= ip_vs_cmd_policy
,
3854 .doit
= ip_vs_genl_set_cmd
,
3857 .cmd
= IPVS_CMD_GET_CONFIG
,
3858 .flags
= GENL_ADMIN_PERM
,
3859 .doit
= ip_vs_genl_get_cmd
,
3862 .cmd
= IPVS_CMD_GET_INFO
,
3863 .flags
= GENL_ADMIN_PERM
,
3864 .doit
= ip_vs_genl_get_cmd
,
3867 .cmd
= IPVS_CMD_ZERO
,
3868 .flags
= GENL_ADMIN_PERM
,
3869 .policy
= ip_vs_cmd_policy
,
3870 .doit
= ip_vs_genl_set_cmd
,
3873 .cmd
= IPVS_CMD_FLUSH
,
3874 .flags
= GENL_ADMIN_PERM
,
3875 .doit
= ip_vs_genl_set_cmd
,
3879 static struct genl_family ip_vs_genl_family __ro_after_init
= {
3881 .name
= IPVS_GENL_NAME
,
3882 .version
= IPVS_GENL_VERSION
,
3883 .maxattr
= IPVS_CMD_ATTR_MAX
,
3884 .netnsok
= true, /* Make ipvsadm to work on netns */
3885 .module
= THIS_MODULE
,
3886 .ops
= ip_vs_genl_ops
,
3887 .n_ops
= ARRAY_SIZE(ip_vs_genl_ops
),
3890 static int __init
ip_vs_genl_register(void)
3892 return genl_register_family(&ip_vs_genl_family
);
3895 static void ip_vs_genl_unregister(void)
3897 genl_unregister_family(&ip_vs_genl_family
);
3900 /* End of Generic Netlink interface definitions */
3903 * per netns intit/exit func.
3905 #ifdef CONFIG_SYSCTL
3906 static int __net_init
ip_vs_control_net_init_sysctl(struct netns_ipvs
*ipvs
)
3908 struct net
*net
= ipvs
->net
;
3910 struct ctl_table
*tbl
;
3912 atomic_set(&ipvs
->dropentry
, 0);
3913 spin_lock_init(&ipvs
->dropentry_lock
);
3914 spin_lock_init(&ipvs
->droppacket_lock
);
3915 spin_lock_init(&ipvs
->securetcp_lock
);
3917 if (!net_eq(net
, &init_net
)) {
3918 tbl
= kmemdup(vs_vars
, sizeof(vs_vars
), GFP_KERNEL
);
3922 /* Don't export sysctls to unprivileged users */
3923 if (net
->user_ns
!= &init_user_ns
)
3924 tbl
[0].procname
= NULL
;
3927 /* Initialize sysctl defaults */
3928 for (idx
= 0; idx
< ARRAY_SIZE(vs_vars
); idx
++) {
3929 if (tbl
[idx
].proc_handler
== proc_do_defense_mode
)
3930 tbl
[idx
].extra2
= ipvs
;
3933 ipvs
->sysctl_amemthresh
= 1024;
3934 tbl
[idx
++].data
= &ipvs
->sysctl_amemthresh
;
3935 ipvs
->sysctl_am_droprate
= 10;
3936 tbl
[idx
++].data
= &ipvs
->sysctl_am_droprate
;
3937 tbl
[idx
++].data
= &ipvs
->sysctl_drop_entry
;
3938 tbl
[idx
++].data
= &ipvs
->sysctl_drop_packet
;
3939 #ifdef CONFIG_IP_VS_NFCT
3940 tbl
[idx
++].data
= &ipvs
->sysctl_conntrack
;
3942 tbl
[idx
++].data
= &ipvs
->sysctl_secure_tcp
;
3943 ipvs
->sysctl_snat_reroute
= 1;
3944 tbl
[idx
++].data
= &ipvs
->sysctl_snat_reroute
;
3945 ipvs
->sysctl_sync_ver
= 1;
3946 tbl
[idx
++].data
= &ipvs
->sysctl_sync_ver
;
3947 ipvs
->sysctl_sync_ports
= 1;
3948 tbl
[idx
++].data
= &ipvs
->sysctl_sync_ports
;
3949 tbl
[idx
++].data
= &ipvs
->sysctl_sync_persist_mode
;
3950 ipvs
->sysctl_sync_qlen_max
= nr_free_buffer_pages() / 32;
3951 tbl
[idx
++].data
= &ipvs
->sysctl_sync_qlen_max
;
3952 ipvs
->sysctl_sync_sock_size
= 0;
3953 tbl
[idx
++].data
= &ipvs
->sysctl_sync_sock_size
;
3954 tbl
[idx
++].data
= &ipvs
->sysctl_cache_bypass
;
3955 tbl
[idx
++].data
= &ipvs
->sysctl_expire_nodest_conn
;
3956 tbl
[idx
++].data
= &ipvs
->sysctl_sloppy_tcp
;
3957 tbl
[idx
++].data
= &ipvs
->sysctl_sloppy_sctp
;
3958 tbl
[idx
++].data
= &ipvs
->sysctl_expire_quiescent_template
;
3959 ipvs
->sysctl_sync_threshold
[0] = DEFAULT_SYNC_THRESHOLD
;
3960 ipvs
->sysctl_sync_threshold
[1] = DEFAULT_SYNC_PERIOD
;
3961 tbl
[idx
].data
= &ipvs
->sysctl_sync_threshold
;
3962 tbl
[idx
++].maxlen
= sizeof(ipvs
->sysctl_sync_threshold
);
3963 ipvs
->sysctl_sync_refresh_period
= DEFAULT_SYNC_REFRESH_PERIOD
;
3964 tbl
[idx
++].data
= &ipvs
->sysctl_sync_refresh_period
;
3965 ipvs
->sysctl_sync_retries
= clamp_t(int, DEFAULT_SYNC_RETRIES
, 0, 3);
3966 tbl
[idx
++].data
= &ipvs
->sysctl_sync_retries
;
3967 tbl
[idx
++].data
= &ipvs
->sysctl_nat_icmp_send
;
3968 ipvs
->sysctl_pmtu_disc
= 1;
3969 tbl
[idx
++].data
= &ipvs
->sysctl_pmtu_disc
;
3970 tbl
[idx
++].data
= &ipvs
->sysctl_backup_only
;
3971 ipvs
->sysctl_conn_reuse_mode
= 1;
3972 tbl
[idx
++].data
= &ipvs
->sysctl_conn_reuse_mode
;
3973 tbl
[idx
++].data
= &ipvs
->sysctl_schedule_icmp
;
3974 tbl
[idx
++].data
= &ipvs
->sysctl_ignore_tunneled
;
3976 ipvs
->sysctl_hdr
= register_net_sysctl(net
, "net/ipv4/vs", tbl
);
3977 if (ipvs
->sysctl_hdr
== NULL
) {
3978 if (!net_eq(net
, &init_net
))
3982 ip_vs_start_estimator(ipvs
, &ipvs
->tot_stats
);
3983 ipvs
->sysctl_tbl
= tbl
;
3984 /* Schedule defense work */
3985 INIT_DELAYED_WORK(&ipvs
->defense_work
, defense_work_handler
);
3986 schedule_delayed_work(&ipvs
->defense_work
, DEFENSE_TIMER_PERIOD
);
3991 static void __net_exit
ip_vs_control_net_cleanup_sysctl(struct netns_ipvs
*ipvs
)
3993 struct net
*net
= ipvs
->net
;
3995 cancel_delayed_work_sync(&ipvs
->defense_work
);
3996 cancel_work_sync(&ipvs
->defense_work
.work
);
3997 unregister_net_sysctl_table(ipvs
->sysctl_hdr
);
3998 ip_vs_stop_estimator(ipvs
, &ipvs
->tot_stats
);
4000 if (!net_eq(net
, &init_net
))
4001 kfree(ipvs
->sysctl_tbl
);
4006 static int __net_init
ip_vs_control_net_init_sysctl(struct netns_ipvs
*ipvs
) { return 0; }
4007 static void __net_exit
ip_vs_control_net_cleanup_sysctl(struct netns_ipvs
*ipvs
) { }
4011 static struct notifier_block ip_vs_dst_notifier
= {
4012 .notifier_call
= ip_vs_dst_event
,
4015 int __net_init
ip_vs_control_net_init(struct netns_ipvs
*ipvs
)
4019 /* Initialize rs_table */
4020 for (idx
= 0; idx
< IP_VS_RTAB_SIZE
; idx
++)
4021 INIT_HLIST_HEAD(&ipvs
->rs_table
[idx
]);
4023 INIT_LIST_HEAD(&ipvs
->dest_trash
);
4024 spin_lock_init(&ipvs
->dest_trash_lock
);
4025 setup_timer(&ipvs
->dest_trash_timer
, ip_vs_dest_trash_expire
,
4026 (unsigned long) ipvs
);
4027 atomic_set(&ipvs
->ftpsvc_counter
, 0);
4028 atomic_set(&ipvs
->nullsvc_counter
, 0);
4029 atomic_set(&ipvs
->conn_out_counter
, 0);
4032 ipvs
->tot_stats
.cpustats
= alloc_percpu(struct ip_vs_cpu_stats
);
4033 if (!ipvs
->tot_stats
.cpustats
)
4036 for_each_possible_cpu(i
) {
4037 struct ip_vs_cpu_stats
*ipvs_tot_stats
;
4038 ipvs_tot_stats
= per_cpu_ptr(ipvs
->tot_stats
.cpustats
, i
);
4039 u64_stats_init(&ipvs_tot_stats
->syncp
);
4042 spin_lock_init(&ipvs
->tot_stats
.lock
);
4044 proc_create("ip_vs", 0, ipvs
->net
->proc_net
, &ip_vs_info_fops
);
4045 proc_create("ip_vs_stats", 0, ipvs
->net
->proc_net
, &ip_vs_stats_fops
);
4046 proc_create("ip_vs_stats_percpu", 0, ipvs
->net
->proc_net
,
4047 &ip_vs_stats_percpu_fops
);
4049 if (ip_vs_control_net_init_sysctl(ipvs
))
4055 free_percpu(ipvs
->tot_stats
.cpustats
);
4059 void __net_exit
ip_vs_control_net_cleanup(struct netns_ipvs
*ipvs
)
4061 ip_vs_trash_cleanup(ipvs
);
4062 ip_vs_control_net_cleanup_sysctl(ipvs
);
4063 remove_proc_entry("ip_vs_stats_percpu", ipvs
->net
->proc_net
);
4064 remove_proc_entry("ip_vs_stats", ipvs
->net
->proc_net
);
4065 remove_proc_entry("ip_vs", ipvs
->net
->proc_net
);
4066 free_percpu(ipvs
->tot_stats
.cpustats
);
4069 int __init
ip_vs_register_nl_ioctl(void)
4073 ret
= nf_register_sockopt(&ip_vs_sockopts
);
4075 pr_err("cannot register sockopt.\n");
4079 ret
= ip_vs_genl_register();
4081 pr_err("cannot register Generic Netlink interface.\n");
4087 nf_unregister_sockopt(&ip_vs_sockopts
);
4092 void ip_vs_unregister_nl_ioctl(void)
4094 ip_vs_genl_unregister();
4095 nf_unregister_sockopt(&ip_vs_sockopts
);
4098 int __init
ip_vs_control_init(void)
4105 /* Initialize svc_table, ip_vs_svc_fwm_table */
4106 for (idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
4107 INIT_HLIST_HEAD(&ip_vs_svc_table
[idx
]);
4108 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table
[idx
]);
4111 smp_wmb(); /* Do we really need it now ? */
4113 ret
= register_netdevice_notifier(&ip_vs_dst_notifier
);
4122 void ip_vs_control_cleanup(void)
4125 unregister_netdevice_notifier(&ip_vs_dst_notifier
);