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
);
553 hlist_for_each_entry_rcu(dest
, &ipvs
->rs_table
[hash
], d_list
) {
554 if (dest
->port
== dport
&&
556 ip_vs_addr_equal(af
, &dest
->addr
, daddr
) &&
557 (dest
->protocol
== protocol
|| dest
->vfwmark
)) {
566 /* Find real service record by <proto,addr,port>.
567 * In case of multiple records with the same <proto,addr,port>, only
568 * the first found record is returned.
570 * To be called under RCU lock.
572 struct ip_vs_dest
*ip_vs_find_real_service(struct netns_ipvs
*ipvs
, int af
,
574 const union nf_inet_addr
*daddr
,
578 struct ip_vs_dest
*dest
;
580 /* Check for "full" addressed entries */
581 hash
= ip_vs_rs_hashkey(af
, daddr
, dport
);
583 hlist_for_each_entry_rcu(dest
, &ipvs
->rs_table
[hash
], d_list
) {
584 if (dest
->port
== dport
&&
586 ip_vs_addr_equal(af
, &dest
->addr
, daddr
) &&
587 (dest
->protocol
== protocol
|| dest
->vfwmark
)) {
596 /* Lookup destination by {addr,port} in the given service
597 * Called under RCU lock.
599 static struct ip_vs_dest
*
600 ip_vs_lookup_dest(struct ip_vs_service
*svc
, int dest_af
,
601 const union nf_inet_addr
*daddr
, __be16 dport
)
603 struct ip_vs_dest
*dest
;
606 * Find the destination for the given service
608 list_for_each_entry_rcu(dest
, &svc
->destinations
, n_list
) {
609 if ((dest
->af
== dest_af
) &&
610 ip_vs_addr_equal(dest_af
, &dest
->addr
, daddr
) &&
611 (dest
->port
== dport
)) {
621 * Find destination by {daddr,dport,vaddr,protocol}
622 * Created to be used in ip_vs_process_message() in
623 * the backup synchronization daemon. It finds the
624 * destination to be bound to the received connection
626 * Called under RCU lock, no refcnt is returned.
628 struct ip_vs_dest
*ip_vs_find_dest(struct netns_ipvs
*ipvs
, int svc_af
, int dest_af
,
629 const union nf_inet_addr
*daddr
,
631 const union nf_inet_addr
*vaddr
,
632 __be16 vport
, __u16 protocol
, __u32 fwmark
,
635 struct ip_vs_dest
*dest
;
636 struct ip_vs_service
*svc
;
639 svc
= ip_vs_service_find(ipvs
, svc_af
, fwmark
, protocol
, vaddr
, vport
);
642 if (fwmark
&& (flags
& IP_VS_CONN_F_FWD_MASK
) != IP_VS_CONN_F_MASQ
)
644 dest
= ip_vs_lookup_dest(svc
, dest_af
, daddr
, port
);
646 dest
= ip_vs_lookup_dest(svc
, dest_af
, daddr
, port
^ dport
);
650 void ip_vs_dest_dst_rcu_free(struct rcu_head
*head
)
652 struct ip_vs_dest_dst
*dest_dst
= container_of(head
,
653 struct ip_vs_dest_dst
,
656 dst_release(dest_dst
->dst_cache
);
660 /* Release dest_dst and dst_cache for dest in user context */
661 static void __ip_vs_dst_cache_reset(struct ip_vs_dest
*dest
)
663 struct ip_vs_dest_dst
*old
;
665 old
= rcu_dereference_protected(dest
->dest_dst
, 1);
667 RCU_INIT_POINTER(dest
->dest_dst
, NULL
);
668 call_rcu(&old
->rcu_head
, ip_vs_dest_dst_rcu_free
);
673 * Lookup dest by {svc,addr,port} in the destination trash.
674 * The destination trash is used to hold the destinations that are removed
675 * from the service table but are still referenced by some conn entries.
676 * The reason to add the destination trash is when the dest is temporary
677 * down (either by administrator or by monitor program), the dest can be
678 * picked back from the trash, the remaining connections to the dest can
679 * continue, and the counting information of the dest is also useful for
682 static struct ip_vs_dest
*
683 ip_vs_trash_get_dest(struct ip_vs_service
*svc
, int dest_af
,
684 const union nf_inet_addr
*daddr
, __be16 dport
)
686 struct ip_vs_dest
*dest
;
687 struct netns_ipvs
*ipvs
= svc
->ipvs
;
690 * Find the destination in trash
692 spin_lock_bh(&ipvs
->dest_trash_lock
);
693 list_for_each_entry(dest
, &ipvs
->dest_trash
, t_list
) {
694 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
697 IP_VS_DBG_ADDR(dest
->af
, &dest
->addr
),
699 refcount_read(&dest
->refcnt
));
700 if (dest
->af
== dest_af
&&
701 ip_vs_addr_equal(dest_af
, &dest
->addr
, daddr
) &&
702 dest
->port
== dport
&&
703 dest
->vfwmark
== svc
->fwmark
&&
704 dest
->protocol
== svc
->protocol
&&
706 (ip_vs_addr_equal(svc
->af
, &dest
->vaddr
, &svc
->addr
) &&
707 dest
->vport
== svc
->port
))) {
709 list_del(&dest
->t_list
);
717 spin_unlock_bh(&ipvs
->dest_trash_lock
);
722 static void ip_vs_dest_free(struct ip_vs_dest
*dest
)
724 struct ip_vs_service
*svc
= rcu_dereference_protected(dest
->svc
, 1);
726 __ip_vs_dst_cache_reset(dest
);
727 __ip_vs_svc_put(svc
, false);
728 free_percpu(dest
->stats
.cpustats
);
729 ip_vs_dest_put_and_free(dest
);
733 * Clean up all the destinations in the trash
734 * Called by the ip_vs_control_cleanup()
736 * When the ip_vs_control_clearup is activated by ipvs module exit,
737 * the service tables must have been flushed and all the connections
738 * are expired, and the refcnt of each destination in the trash must
739 * be 1, so we simply release them here.
741 static void ip_vs_trash_cleanup(struct netns_ipvs
*ipvs
)
743 struct ip_vs_dest
*dest
, *nxt
;
745 del_timer_sync(&ipvs
->dest_trash_timer
);
746 /* No need to use dest_trash_lock */
747 list_for_each_entry_safe(dest
, nxt
, &ipvs
->dest_trash
, t_list
) {
748 list_del(&dest
->t_list
);
749 ip_vs_dest_free(dest
);
754 ip_vs_copy_stats(struct ip_vs_kstats
*dst
, struct ip_vs_stats
*src
)
756 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->kstats.c - src->kstats0.c
758 spin_lock_bh(&src
->lock
);
760 IP_VS_SHOW_STATS_COUNTER(conns
);
761 IP_VS_SHOW_STATS_COUNTER(inpkts
);
762 IP_VS_SHOW_STATS_COUNTER(outpkts
);
763 IP_VS_SHOW_STATS_COUNTER(inbytes
);
764 IP_VS_SHOW_STATS_COUNTER(outbytes
);
766 ip_vs_read_estimator(dst
, src
);
768 spin_unlock_bh(&src
->lock
);
772 ip_vs_export_stats_user(struct ip_vs_stats_user
*dst
, struct ip_vs_kstats
*src
)
774 dst
->conns
= (u32
)src
->conns
;
775 dst
->inpkts
= (u32
)src
->inpkts
;
776 dst
->outpkts
= (u32
)src
->outpkts
;
777 dst
->inbytes
= src
->inbytes
;
778 dst
->outbytes
= src
->outbytes
;
779 dst
->cps
= (u32
)src
->cps
;
780 dst
->inpps
= (u32
)src
->inpps
;
781 dst
->outpps
= (u32
)src
->outpps
;
782 dst
->inbps
= (u32
)src
->inbps
;
783 dst
->outbps
= (u32
)src
->outbps
;
787 ip_vs_zero_stats(struct ip_vs_stats
*stats
)
789 spin_lock_bh(&stats
->lock
);
791 /* get current counters as zero point, rates are zeroed */
793 #define IP_VS_ZERO_STATS_COUNTER(c) stats->kstats0.c = stats->kstats.c
795 IP_VS_ZERO_STATS_COUNTER(conns
);
796 IP_VS_ZERO_STATS_COUNTER(inpkts
);
797 IP_VS_ZERO_STATS_COUNTER(outpkts
);
798 IP_VS_ZERO_STATS_COUNTER(inbytes
);
799 IP_VS_ZERO_STATS_COUNTER(outbytes
);
801 ip_vs_zero_estimator(stats
);
803 spin_unlock_bh(&stats
->lock
);
807 * Update a destination in the given service
810 __ip_vs_update_dest(struct ip_vs_service
*svc
, struct ip_vs_dest
*dest
,
811 struct ip_vs_dest_user_kern
*udest
, int add
)
813 struct netns_ipvs
*ipvs
= svc
->ipvs
;
814 struct ip_vs_service
*old_svc
;
815 struct ip_vs_scheduler
*sched
;
818 /* We cannot modify an address and change the address family */
819 BUG_ON(!add
&& udest
->af
!= dest
->af
);
821 if (add
&& udest
->af
!= svc
->af
)
822 ipvs
->mixed_address_family_dests
++;
824 /* set the weight and the flags */
825 atomic_set(&dest
->weight
, udest
->weight
);
826 conn_flags
= udest
->conn_flags
& IP_VS_CONN_F_DEST_MASK
;
827 conn_flags
|= IP_VS_CONN_F_INACTIVE
;
829 /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
830 if ((conn_flags
& IP_VS_CONN_F_FWD_MASK
) != IP_VS_CONN_F_MASQ
) {
831 conn_flags
|= IP_VS_CONN_F_NOOUTPUT
;
834 * Put the real service in rs_table if not present.
835 * For now only for NAT!
837 ip_vs_rs_hash(ipvs
, dest
);
839 atomic_set(&dest
->conn_flags
, conn_flags
);
841 /* bind the service */
842 old_svc
= rcu_dereference_protected(dest
->svc
, 1);
844 __ip_vs_bind_svc(dest
, svc
);
846 if (old_svc
!= svc
) {
847 ip_vs_zero_stats(&dest
->stats
);
848 __ip_vs_bind_svc(dest
, svc
);
849 __ip_vs_svc_put(old_svc
, true);
853 /* set the dest status flags */
854 dest
->flags
|= IP_VS_DEST_F_AVAILABLE
;
856 if (udest
->u_threshold
== 0 || udest
->u_threshold
> dest
->u_threshold
)
857 dest
->flags
&= ~IP_VS_DEST_F_OVERLOAD
;
858 dest
->u_threshold
= udest
->u_threshold
;
859 dest
->l_threshold
= udest
->l_threshold
;
861 dest
->af
= udest
->af
;
863 spin_lock_bh(&dest
->dst_lock
);
864 __ip_vs_dst_cache_reset(dest
);
865 spin_unlock_bh(&dest
->dst_lock
);
868 ip_vs_start_estimator(svc
->ipvs
, &dest
->stats
);
869 list_add_rcu(&dest
->n_list
, &svc
->destinations
);
871 sched
= rcu_dereference_protected(svc
->scheduler
, 1);
872 if (sched
&& sched
->add_dest
)
873 sched
->add_dest(svc
, dest
);
875 sched
= rcu_dereference_protected(svc
->scheduler
, 1);
876 if (sched
&& sched
->upd_dest
)
877 sched
->upd_dest(svc
, dest
);
883 * Create a destination for the given service
886 ip_vs_new_dest(struct ip_vs_service
*svc
, struct ip_vs_dest_user_kern
*udest
,
887 struct ip_vs_dest
**dest_p
)
889 struct ip_vs_dest
*dest
;
890 unsigned int atype
, i
;
894 #ifdef CONFIG_IP_VS_IPV6
895 if (udest
->af
== AF_INET6
) {
896 atype
= ipv6_addr_type(&udest
->addr
.in6
);
897 if ((!(atype
& IPV6_ADDR_UNICAST
) ||
898 atype
& IPV6_ADDR_LINKLOCAL
) &&
899 !__ip_vs_addr_is_local_v6(svc
->ipvs
->net
, &udest
->addr
.in6
))
904 atype
= inet_addr_type(svc
->ipvs
->net
, udest
->addr
.ip
);
905 if (atype
!= RTN_LOCAL
&& atype
!= RTN_UNICAST
)
909 dest
= kzalloc(sizeof(struct ip_vs_dest
), GFP_KERNEL
);
913 dest
->stats
.cpustats
= alloc_percpu(struct ip_vs_cpu_stats
);
914 if (!dest
->stats
.cpustats
)
917 for_each_possible_cpu(i
) {
918 struct ip_vs_cpu_stats
*ip_vs_dest_stats
;
919 ip_vs_dest_stats
= per_cpu_ptr(dest
->stats
.cpustats
, i
);
920 u64_stats_init(&ip_vs_dest_stats
->syncp
);
923 dest
->af
= udest
->af
;
924 dest
->protocol
= svc
->protocol
;
925 dest
->vaddr
= svc
->addr
;
926 dest
->vport
= svc
->port
;
927 dest
->vfwmark
= svc
->fwmark
;
928 ip_vs_addr_copy(udest
->af
, &dest
->addr
, &udest
->addr
);
929 dest
->port
= udest
->port
;
931 atomic_set(&dest
->activeconns
, 0);
932 atomic_set(&dest
->inactconns
, 0);
933 atomic_set(&dest
->persistconns
, 0);
934 refcount_set(&dest
->refcnt
, 1);
936 INIT_HLIST_NODE(&dest
->d_list
);
937 spin_lock_init(&dest
->dst_lock
);
938 spin_lock_init(&dest
->stats
.lock
);
939 __ip_vs_update_dest(svc
, dest
, udest
, 1);
953 * Add a destination into an existing service
956 ip_vs_add_dest(struct ip_vs_service
*svc
, struct ip_vs_dest_user_kern
*udest
)
958 struct ip_vs_dest
*dest
;
959 union nf_inet_addr daddr
;
960 __be16 dport
= udest
->port
;
965 if (udest
->weight
< 0) {
966 pr_err("%s(): server weight less than zero\n", __func__
);
970 if (udest
->l_threshold
> udest
->u_threshold
) {
971 pr_err("%s(): lower threshold is higher than upper threshold\n",
976 ip_vs_addr_copy(udest
->af
, &daddr
, &udest
->addr
);
978 /* We use function that requires RCU lock */
980 dest
= ip_vs_lookup_dest(svc
, udest
->af
, &daddr
, dport
);
984 IP_VS_DBG(1, "%s(): dest already exists\n", __func__
);
989 * Check if the dest already exists in the trash and
990 * is from the same service
992 dest
= ip_vs_trash_get_dest(svc
, udest
->af
, &daddr
, dport
);
995 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
996 "dest->refcnt=%d, service %u/%s:%u\n",
997 IP_VS_DBG_ADDR(udest
->af
, &daddr
), ntohs(dport
),
998 refcount_read(&dest
->refcnt
),
1000 IP_VS_DBG_ADDR(svc
->af
, &dest
->vaddr
),
1001 ntohs(dest
->vport
));
1003 __ip_vs_update_dest(svc
, dest
, udest
, 1);
1007 * Allocate and initialize the dest structure
1009 ret
= ip_vs_new_dest(svc
, udest
, &dest
);
1018 * Edit a destination in the given service
1021 ip_vs_edit_dest(struct ip_vs_service
*svc
, struct ip_vs_dest_user_kern
*udest
)
1023 struct ip_vs_dest
*dest
;
1024 union nf_inet_addr daddr
;
1025 __be16 dport
= udest
->port
;
1029 if (udest
->weight
< 0) {
1030 pr_err("%s(): server weight less than zero\n", __func__
);
1034 if (udest
->l_threshold
> udest
->u_threshold
) {
1035 pr_err("%s(): lower threshold is higher than upper threshold\n",
1040 ip_vs_addr_copy(udest
->af
, &daddr
, &udest
->addr
);
1042 /* We use function that requires RCU lock */
1044 dest
= ip_vs_lookup_dest(svc
, udest
->af
, &daddr
, dport
);
1048 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__
);
1052 __ip_vs_update_dest(svc
, dest
, udest
, 0);
1059 * Delete a destination (must be already unlinked from the service)
1061 static void __ip_vs_del_dest(struct netns_ipvs
*ipvs
, struct ip_vs_dest
*dest
,
1064 ip_vs_stop_estimator(ipvs
, &dest
->stats
);
1067 * Remove it from the d-linked list with the real services.
1069 ip_vs_rs_unhash(dest
);
1071 spin_lock_bh(&ipvs
->dest_trash_lock
);
1072 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1073 IP_VS_DBG_ADDR(dest
->af
, &dest
->addr
), ntohs(dest
->port
),
1074 refcount_read(&dest
->refcnt
));
1075 if (list_empty(&ipvs
->dest_trash
) && !cleanup
)
1076 mod_timer(&ipvs
->dest_trash_timer
,
1077 jiffies
+ (IP_VS_DEST_TRASH_PERIOD
>> 1));
1078 /* dest lives in trash with reference */
1079 list_add(&dest
->t_list
, &ipvs
->dest_trash
);
1080 dest
->idle_start
= 0;
1081 spin_unlock_bh(&ipvs
->dest_trash_lock
);
1086 * Unlink a destination from the given service
1088 static void __ip_vs_unlink_dest(struct ip_vs_service
*svc
,
1089 struct ip_vs_dest
*dest
,
1092 dest
->flags
&= ~IP_VS_DEST_F_AVAILABLE
;
1095 * Remove it from the d-linked destination list.
1097 list_del_rcu(&dest
->n_list
);
1100 if (dest
->af
!= svc
->af
)
1101 svc
->ipvs
->mixed_address_family_dests
--;
1104 struct ip_vs_scheduler
*sched
;
1106 sched
= rcu_dereference_protected(svc
->scheduler
, 1);
1107 if (sched
&& sched
->del_dest
)
1108 sched
->del_dest(svc
, dest
);
1114 * Delete a destination server in the given service
1117 ip_vs_del_dest(struct ip_vs_service
*svc
, struct ip_vs_dest_user_kern
*udest
)
1119 struct ip_vs_dest
*dest
;
1120 __be16 dport
= udest
->port
;
1124 /* We use function that requires RCU lock */
1126 dest
= ip_vs_lookup_dest(svc
, udest
->af
, &udest
->addr
, dport
);
1130 IP_VS_DBG(1, "%s(): destination not found!\n", __func__
);
1135 * Unlink dest from the service
1137 __ip_vs_unlink_dest(svc
, dest
, 1);
1140 * Delete the destination
1142 __ip_vs_del_dest(svc
->ipvs
, dest
, false);
1149 static void ip_vs_dest_trash_expire(unsigned long data
)
1151 struct netns_ipvs
*ipvs
= (struct netns_ipvs
*)data
;
1152 struct ip_vs_dest
*dest
, *next
;
1153 unsigned long now
= jiffies
;
1155 spin_lock(&ipvs
->dest_trash_lock
);
1156 list_for_each_entry_safe(dest
, next
, &ipvs
->dest_trash
, t_list
) {
1157 if (refcount_read(&dest
->refcnt
) > 1)
1159 if (dest
->idle_start
) {
1160 if (time_before(now
, dest
->idle_start
+
1161 IP_VS_DEST_TRASH_PERIOD
))
1164 dest
->idle_start
= max(1UL, now
);
1167 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1169 IP_VS_DBG_ADDR(dest
->af
, &dest
->addr
),
1171 list_del(&dest
->t_list
);
1172 ip_vs_dest_free(dest
);
1174 if (!list_empty(&ipvs
->dest_trash
))
1175 mod_timer(&ipvs
->dest_trash_timer
,
1176 jiffies
+ (IP_VS_DEST_TRASH_PERIOD
>> 1));
1177 spin_unlock(&ipvs
->dest_trash_lock
);
1181 * Add a service into the service hash table
1184 ip_vs_add_service(struct netns_ipvs
*ipvs
, struct ip_vs_service_user_kern
*u
,
1185 struct ip_vs_service
**svc_p
)
1188 struct ip_vs_scheduler
*sched
= NULL
;
1189 struct ip_vs_pe
*pe
= NULL
;
1190 struct ip_vs_service
*svc
= NULL
;
1192 /* increase the module use count */
1193 ip_vs_use_count_inc();
1195 /* Lookup the scheduler by 'u->sched_name' */
1196 if (strcmp(u
->sched_name
, "none")) {
1197 sched
= ip_vs_scheduler_get(u
->sched_name
);
1199 pr_info("Scheduler module ip_vs_%s not found\n",
1206 if (u
->pe_name
&& *u
->pe_name
) {
1207 pe
= ip_vs_pe_getbyname(u
->pe_name
);
1209 pr_info("persistence engine module ip_vs_pe_%s "
1210 "not found\n", u
->pe_name
);
1216 #ifdef CONFIG_IP_VS_IPV6
1217 if (u
->af
== AF_INET6
) {
1218 __u32 plen
= (__force __u32
) u
->netmask
;
1220 if (plen
< 1 || plen
> 128) {
1227 svc
= kzalloc(sizeof(struct ip_vs_service
), GFP_KERNEL
);
1229 IP_VS_DBG(1, "%s(): no memory\n", __func__
);
1233 svc
->stats
.cpustats
= alloc_percpu(struct ip_vs_cpu_stats
);
1234 if (!svc
->stats
.cpustats
) {
1239 for_each_possible_cpu(i
) {
1240 struct ip_vs_cpu_stats
*ip_vs_stats
;
1241 ip_vs_stats
= per_cpu_ptr(svc
->stats
.cpustats
, i
);
1242 u64_stats_init(&ip_vs_stats
->syncp
);
1246 /* I'm the first user of the service */
1247 atomic_set(&svc
->refcnt
, 0);
1250 svc
->protocol
= u
->protocol
;
1251 ip_vs_addr_copy(svc
->af
, &svc
->addr
, &u
->addr
);
1252 svc
->port
= u
->port
;
1253 svc
->fwmark
= u
->fwmark
;
1254 svc
->flags
= u
->flags
;
1255 svc
->timeout
= u
->timeout
* HZ
;
1256 svc
->netmask
= u
->netmask
;
1259 INIT_LIST_HEAD(&svc
->destinations
);
1260 spin_lock_init(&svc
->sched_lock
);
1261 spin_lock_init(&svc
->stats
.lock
);
1263 /* Bind the scheduler */
1265 ret
= ip_vs_bind_scheduler(svc
, sched
);
1271 /* Bind the ct retriever */
1272 RCU_INIT_POINTER(svc
->pe
, pe
);
1275 /* Update the virtual service counters */
1276 if (svc
->port
== FTPPORT
)
1277 atomic_inc(&ipvs
->ftpsvc_counter
);
1278 else if (svc
->port
== 0)
1279 atomic_inc(&ipvs
->nullsvc_counter
);
1280 if (svc
->pe
&& svc
->pe
->conn_out
)
1281 atomic_inc(&ipvs
->conn_out_counter
);
1283 ip_vs_start_estimator(ipvs
, &svc
->stats
);
1285 /* Count only IPv4 services for old get/setsockopt interface */
1286 if (svc
->af
== AF_INET
)
1287 ipvs
->num_services
++;
1289 /* Hash the service into the service table */
1290 ip_vs_svc_hash(svc
);
1293 /* Now there is a service - full throttle */
1300 ip_vs_unbind_scheduler(svc
, sched
);
1301 ip_vs_service_free(svc
);
1303 ip_vs_scheduler_put(sched
);
1306 /* decrease the module use count */
1307 ip_vs_use_count_dec();
1314 * Edit a service and bind it with a new scheduler
1317 ip_vs_edit_service(struct ip_vs_service
*svc
, struct ip_vs_service_user_kern
*u
)
1319 struct ip_vs_scheduler
*sched
= NULL
, *old_sched
;
1320 struct ip_vs_pe
*pe
= NULL
, *old_pe
= NULL
;
1322 bool new_pe_conn_out
, old_pe_conn_out
;
1325 * Lookup the scheduler, by 'u->sched_name'
1327 if (strcmp(u
->sched_name
, "none")) {
1328 sched
= ip_vs_scheduler_get(u
->sched_name
);
1330 pr_info("Scheduler module ip_vs_%s not found\n",
1337 if (u
->pe_name
&& *u
->pe_name
) {
1338 pe
= ip_vs_pe_getbyname(u
->pe_name
);
1340 pr_info("persistence engine module ip_vs_pe_%s "
1341 "not found\n", u
->pe_name
);
1348 #ifdef CONFIG_IP_VS_IPV6
1349 if (u
->af
== AF_INET6
) {
1350 __u32 plen
= (__force __u32
) u
->netmask
;
1352 if (plen
< 1 || plen
> 128) {
1359 old_sched
= rcu_dereference_protected(svc
->scheduler
, 1);
1360 if (sched
!= old_sched
) {
1362 ip_vs_unbind_scheduler(svc
, old_sched
);
1363 RCU_INIT_POINTER(svc
->scheduler
, NULL
);
1364 /* Wait all svc->sched_data users */
1367 /* Bind the new scheduler */
1369 ret
= ip_vs_bind_scheduler(svc
, sched
);
1371 ip_vs_scheduler_put(sched
);
1378 * Set the flags and timeout value
1380 svc
->flags
= u
->flags
| IP_VS_SVC_F_HASHED
;
1381 svc
->timeout
= u
->timeout
* HZ
;
1382 svc
->netmask
= u
->netmask
;
1384 old_pe
= rcu_dereference_protected(svc
->pe
, 1);
1386 rcu_assign_pointer(svc
->pe
, pe
);
1387 /* check for optional methods in new pe */
1388 new_pe_conn_out
= (pe
&& pe
->conn_out
) ? true : false;
1389 old_pe_conn_out
= (old_pe
&& old_pe
->conn_out
) ? true : false;
1390 if (new_pe_conn_out
&& !old_pe_conn_out
)
1391 atomic_inc(&svc
->ipvs
->conn_out_counter
);
1392 if (old_pe_conn_out
&& !new_pe_conn_out
)
1393 atomic_dec(&svc
->ipvs
->conn_out_counter
);
1397 ip_vs_scheduler_put(old_sched
);
1398 ip_vs_pe_put(old_pe
);
1403 * Delete a service from the service list
1404 * - The service must be unlinked, unlocked and not referenced!
1405 * - We are called under _bh lock
1407 static void __ip_vs_del_service(struct ip_vs_service
*svc
, bool cleanup
)
1409 struct ip_vs_dest
*dest
, *nxt
;
1410 struct ip_vs_scheduler
*old_sched
;
1411 struct ip_vs_pe
*old_pe
;
1412 struct netns_ipvs
*ipvs
= svc
->ipvs
;
1414 /* Count only IPv4 services for old get/setsockopt interface */
1415 if (svc
->af
== AF_INET
)
1416 ipvs
->num_services
--;
1418 ip_vs_stop_estimator(svc
->ipvs
, &svc
->stats
);
1420 /* Unbind scheduler */
1421 old_sched
= rcu_dereference_protected(svc
->scheduler
, 1);
1422 ip_vs_unbind_scheduler(svc
, old_sched
);
1423 ip_vs_scheduler_put(old_sched
);
1425 /* Unbind persistence engine, keep svc->pe */
1426 old_pe
= rcu_dereference_protected(svc
->pe
, 1);
1427 if (old_pe
&& old_pe
->conn_out
)
1428 atomic_dec(&ipvs
->conn_out_counter
);
1429 ip_vs_pe_put(old_pe
);
1432 * Unlink the whole destination list
1434 list_for_each_entry_safe(dest
, nxt
, &svc
->destinations
, n_list
) {
1435 __ip_vs_unlink_dest(svc
, dest
, 0);
1436 __ip_vs_del_dest(svc
->ipvs
, dest
, cleanup
);
1440 * Update the virtual service counters
1442 if (svc
->port
== FTPPORT
)
1443 atomic_dec(&ipvs
->ftpsvc_counter
);
1444 else if (svc
->port
== 0)
1445 atomic_dec(&ipvs
->nullsvc_counter
);
1448 * Free the service if nobody refers to it
1450 __ip_vs_svc_put(svc
, true);
1452 /* decrease the module use count */
1453 ip_vs_use_count_dec();
1457 * Unlink a service from list and try to delete it if its refcnt reached 0
1459 static void ip_vs_unlink_service(struct ip_vs_service
*svc
, bool cleanup
)
1461 /* Hold svc to avoid double release from dest_trash */
1462 atomic_inc(&svc
->refcnt
);
1464 * Unhash it from the service table
1466 ip_vs_svc_unhash(svc
);
1468 __ip_vs_del_service(svc
, cleanup
);
1472 * Delete a service from the service list
1474 static int ip_vs_del_service(struct ip_vs_service
*svc
)
1478 ip_vs_unlink_service(svc
, false);
1485 * Flush all the virtual services
1487 static int ip_vs_flush(struct netns_ipvs
*ipvs
, bool cleanup
)
1490 struct ip_vs_service
*svc
;
1491 struct hlist_node
*n
;
1494 * Flush the service table hashed by <netns,protocol,addr,port>
1496 for(idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1497 hlist_for_each_entry_safe(svc
, n
, &ip_vs_svc_table
[idx
],
1499 if (svc
->ipvs
== ipvs
)
1500 ip_vs_unlink_service(svc
, cleanup
);
1505 * Flush the service table hashed by fwmark
1507 for(idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1508 hlist_for_each_entry_safe(svc
, n
, &ip_vs_svc_fwm_table
[idx
],
1510 if (svc
->ipvs
== ipvs
)
1511 ip_vs_unlink_service(svc
, cleanup
);
1519 * Delete service by {netns} in the service table.
1520 * Called by __ip_vs_cleanup()
1522 void ip_vs_service_net_cleanup(struct netns_ipvs
*ipvs
)
1525 /* Check for "full" addressed entries */
1526 mutex_lock(&__ip_vs_mutex
);
1527 ip_vs_flush(ipvs
, true);
1528 mutex_unlock(&__ip_vs_mutex
);
1532 /* Put all references for device (dst_cache) */
1534 ip_vs_forget_dev(struct ip_vs_dest
*dest
, struct net_device
*dev
)
1536 struct ip_vs_dest_dst
*dest_dst
;
1538 spin_lock_bh(&dest
->dst_lock
);
1539 dest_dst
= rcu_dereference_protected(dest
->dest_dst
, 1);
1540 if (dest_dst
&& dest_dst
->dst_cache
->dev
== dev
) {
1541 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1543 IP_VS_DBG_ADDR(dest
->af
, &dest
->addr
),
1545 refcount_read(&dest
->refcnt
));
1546 __ip_vs_dst_cache_reset(dest
);
1548 spin_unlock_bh(&dest
->dst_lock
);
1551 /* Netdev event receiver
1552 * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1554 static int ip_vs_dst_event(struct notifier_block
*this, unsigned long event
,
1557 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
1558 struct net
*net
= dev_net(dev
);
1559 struct netns_ipvs
*ipvs
= net_ipvs(net
);
1560 struct ip_vs_service
*svc
;
1561 struct ip_vs_dest
*dest
;
1564 if (event
!= NETDEV_DOWN
|| !ipvs
)
1566 IP_VS_DBG(3, "%s() dev=%s\n", __func__
, dev
->name
);
1568 mutex_lock(&__ip_vs_mutex
);
1569 for (idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1570 hlist_for_each_entry(svc
, &ip_vs_svc_table
[idx
], s_list
) {
1571 if (svc
->ipvs
== ipvs
) {
1572 list_for_each_entry(dest
, &svc
->destinations
,
1574 ip_vs_forget_dev(dest
, dev
);
1579 hlist_for_each_entry(svc
, &ip_vs_svc_fwm_table
[idx
], f_list
) {
1580 if (svc
->ipvs
== ipvs
) {
1581 list_for_each_entry(dest
, &svc
->destinations
,
1583 ip_vs_forget_dev(dest
, dev
);
1590 spin_lock_bh(&ipvs
->dest_trash_lock
);
1591 list_for_each_entry(dest
, &ipvs
->dest_trash
, t_list
) {
1592 ip_vs_forget_dev(dest
, dev
);
1594 spin_unlock_bh(&ipvs
->dest_trash_lock
);
1595 mutex_unlock(&__ip_vs_mutex
);
1601 * Zero counters in a service or all services
1603 static int ip_vs_zero_service(struct ip_vs_service
*svc
)
1605 struct ip_vs_dest
*dest
;
1607 list_for_each_entry(dest
, &svc
->destinations
, n_list
) {
1608 ip_vs_zero_stats(&dest
->stats
);
1610 ip_vs_zero_stats(&svc
->stats
);
1614 static int ip_vs_zero_all(struct netns_ipvs
*ipvs
)
1617 struct ip_vs_service
*svc
;
1619 for(idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1620 hlist_for_each_entry(svc
, &ip_vs_svc_table
[idx
], s_list
) {
1621 if (svc
->ipvs
== ipvs
)
1622 ip_vs_zero_service(svc
);
1626 for(idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1627 hlist_for_each_entry(svc
, &ip_vs_svc_fwm_table
[idx
], f_list
) {
1628 if (svc
->ipvs
== ipvs
)
1629 ip_vs_zero_service(svc
);
1633 ip_vs_zero_stats(&ipvs
->tot_stats
);
1637 #ifdef CONFIG_SYSCTL
1640 static int three
= 3;
1643 proc_do_defense_mode(struct ctl_table
*table
, int write
,
1644 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
1646 struct netns_ipvs
*ipvs
= table
->extra2
;
1647 int *valp
= table
->data
;
1651 rc
= proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
1652 if (write
&& (*valp
!= val
)) {
1653 if ((*valp
< 0) || (*valp
> 3)) {
1654 /* Restore the correct value */
1657 update_defense_level(ipvs
);
1664 proc_do_sync_threshold(struct ctl_table
*table
, int write
,
1665 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
1667 int *valp
= table
->data
;
1671 /* backup the value first */
1672 memcpy(val
, valp
, sizeof(val
));
1674 rc
= proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
1675 if (write
&& (valp
[0] < 0 || valp
[1] < 0 ||
1676 (valp
[0] >= valp
[1] && valp
[1]))) {
1677 /* Restore the correct value */
1678 memcpy(valp
, val
, sizeof(val
));
1684 proc_do_sync_mode(struct ctl_table
*table
, int write
,
1685 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
1687 int *valp
= table
->data
;
1691 rc
= proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
1692 if (write
&& (*valp
!= val
)) {
1693 if ((*valp
< 0) || (*valp
> 1)) {
1694 /* Restore the correct value */
1702 proc_do_sync_ports(struct ctl_table
*table
, int write
,
1703 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
1705 int *valp
= table
->data
;
1709 rc
= proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
1710 if (write
&& (*valp
!= val
)) {
1711 if (*valp
< 1 || !is_power_of_2(*valp
)) {
1712 /* Restore the correct value */
1720 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1721 * Do not change order or insert new entries without
1722 * align with netns init in ip_vs_control_net_init()
1725 static struct ctl_table vs_vars
[] = {
1727 .procname
= "amemthresh",
1728 .maxlen
= sizeof(int),
1730 .proc_handler
= proc_dointvec
,
1733 .procname
= "am_droprate",
1734 .maxlen
= sizeof(int),
1736 .proc_handler
= proc_dointvec
,
1739 .procname
= "drop_entry",
1740 .maxlen
= sizeof(int),
1742 .proc_handler
= proc_do_defense_mode
,
1745 .procname
= "drop_packet",
1746 .maxlen
= sizeof(int),
1748 .proc_handler
= proc_do_defense_mode
,
1750 #ifdef CONFIG_IP_VS_NFCT
1752 .procname
= "conntrack",
1753 .maxlen
= sizeof(int),
1755 .proc_handler
= &proc_dointvec
,
1759 .procname
= "secure_tcp",
1760 .maxlen
= sizeof(int),
1762 .proc_handler
= proc_do_defense_mode
,
1765 .procname
= "snat_reroute",
1766 .maxlen
= sizeof(int),
1768 .proc_handler
= &proc_dointvec
,
1771 .procname
= "sync_version",
1772 .maxlen
= sizeof(int),
1774 .proc_handler
= proc_do_sync_mode
,
1777 .procname
= "sync_ports",
1778 .maxlen
= sizeof(int),
1780 .proc_handler
= proc_do_sync_ports
,
1783 .procname
= "sync_persist_mode",
1784 .maxlen
= sizeof(int),
1786 .proc_handler
= proc_dointvec
,
1789 .procname
= "sync_qlen_max",
1790 .maxlen
= sizeof(unsigned long),
1792 .proc_handler
= proc_doulongvec_minmax
,
1795 .procname
= "sync_sock_size",
1796 .maxlen
= sizeof(int),
1798 .proc_handler
= proc_dointvec
,
1801 .procname
= "cache_bypass",
1802 .maxlen
= sizeof(int),
1804 .proc_handler
= proc_dointvec
,
1807 .procname
= "expire_nodest_conn",
1808 .maxlen
= sizeof(int),
1810 .proc_handler
= proc_dointvec
,
1813 .procname
= "sloppy_tcp",
1814 .maxlen
= sizeof(int),
1816 .proc_handler
= proc_dointvec
,
1819 .procname
= "sloppy_sctp",
1820 .maxlen
= sizeof(int),
1822 .proc_handler
= proc_dointvec
,
1825 .procname
= "expire_quiescent_template",
1826 .maxlen
= sizeof(int),
1828 .proc_handler
= proc_dointvec
,
1831 .procname
= "sync_threshold",
1833 sizeof(((struct netns_ipvs
*)0)->sysctl_sync_threshold
),
1835 .proc_handler
= proc_do_sync_threshold
,
1838 .procname
= "sync_refresh_period",
1839 .maxlen
= sizeof(int),
1841 .proc_handler
= proc_dointvec_jiffies
,
1844 .procname
= "sync_retries",
1845 .maxlen
= sizeof(int),
1847 .proc_handler
= proc_dointvec_minmax
,
1852 .procname
= "nat_icmp_send",
1853 .maxlen
= sizeof(int),
1855 .proc_handler
= proc_dointvec
,
1858 .procname
= "pmtu_disc",
1859 .maxlen
= sizeof(int),
1861 .proc_handler
= proc_dointvec
,
1864 .procname
= "backup_only",
1865 .maxlen
= sizeof(int),
1867 .proc_handler
= proc_dointvec
,
1870 .procname
= "conn_reuse_mode",
1871 .maxlen
= sizeof(int),
1873 .proc_handler
= proc_dointvec
,
1876 .procname
= "schedule_icmp",
1877 .maxlen
= sizeof(int),
1879 .proc_handler
= proc_dointvec
,
1882 .procname
= "ignore_tunneled",
1883 .maxlen
= sizeof(int),
1885 .proc_handler
= proc_dointvec
,
1887 #ifdef CONFIG_IP_VS_DEBUG
1889 .procname
= "debug_level",
1890 .data
= &sysctl_ip_vs_debug_level
,
1891 .maxlen
= sizeof(int),
1893 .proc_handler
= proc_dointvec
,
1901 #ifdef CONFIG_PROC_FS
1904 struct seq_net_private p
; /* Do not move this, netns depends upon it*/
1905 struct hlist_head
*table
;
1910 * Write the contents of the VS rule table to a PROCfs file.
1911 * (It is kept just for backward compatibility)
1913 static inline const char *ip_vs_fwd_name(unsigned int flags
)
1915 switch (flags
& IP_VS_CONN_F_FWD_MASK
) {
1916 case IP_VS_CONN_F_LOCALNODE
:
1918 case IP_VS_CONN_F_TUNNEL
:
1920 case IP_VS_CONN_F_DROUTE
:
1928 /* Get the Nth entry in the two lists */
1929 static struct ip_vs_service
*ip_vs_info_array(struct seq_file
*seq
, loff_t pos
)
1931 struct net
*net
= seq_file_net(seq
);
1932 struct netns_ipvs
*ipvs
= net_ipvs(net
);
1933 struct ip_vs_iter
*iter
= seq
->private;
1935 struct ip_vs_service
*svc
;
1937 /* look in hash by protocol */
1938 for (idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1939 hlist_for_each_entry_rcu(svc
, &ip_vs_svc_table
[idx
], s_list
) {
1940 if ((svc
->ipvs
== ipvs
) && pos
-- == 0) {
1941 iter
->table
= ip_vs_svc_table
;
1948 /* keep looking in fwmark */
1949 for (idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1950 hlist_for_each_entry_rcu(svc
, &ip_vs_svc_fwm_table
[idx
],
1952 if ((svc
->ipvs
== ipvs
) && pos
-- == 0) {
1953 iter
->table
= ip_vs_svc_fwm_table
;
1963 static void *ip_vs_info_seq_start(struct seq_file
*seq
, loff_t
*pos
)
1967 return *pos
? ip_vs_info_array(seq
, *pos
- 1) : SEQ_START_TOKEN
;
1971 static void *ip_vs_info_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
1973 struct hlist_node
*e
;
1974 struct ip_vs_iter
*iter
;
1975 struct ip_vs_service
*svc
;
1978 if (v
== SEQ_START_TOKEN
)
1979 return ip_vs_info_array(seq
,0);
1982 iter
= seq
->private;
1984 if (iter
->table
== ip_vs_svc_table
) {
1985 /* next service in table hashed by protocol */
1986 e
= rcu_dereference(hlist_next_rcu(&svc
->s_list
));
1988 return hlist_entry(e
, struct ip_vs_service
, s_list
);
1990 while (++iter
->bucket
< IP_VS_SVC_TAB_SIZE
) {
1991 hlist_for_each_entry_rcu(svc
,
1992 &ip_vs_svc_table
[iter
->bucket
],
1998 iter
->table
= ip_vs_svc_fwm_table
;
2003 /* next service in hashed by fwmark */
2004 e
= rcu_dereference(hlist_next_rcu(&svc
->f_list
));
2006 return hlist_entry(e
, struct ip_vs_service
, f_list
);
2009 while (++iter
->bucket
< IP_VS_SVC_TAB_SIZE
) {
2010 hlist_for_each_entry_rcu(svc
,
2011 &ip_vs_svc_fwm_table
[iter
->bucket
],
2019 static void ip_vs_info_seq_stop(struct seq_file
*seq
, void *v
)
2026 static int ip_vs_info_seq_show(struct seq_file
*seq
, void *v
)
2028 if (v
== SEQ_START_TOKEN
) {
2030 "IP Virtual Server version %d.%d.%d (size=%d)\n",
2031 NVERSION(IP_VS_VERSION_CODE
), ip_vs_conn_tab_size
);
2033 "Prot LocalAddress:Port Scheduler Flags\n");
2035 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2037 struct net
*net
= seq_file_net(seq
);
2038 struct netns_ipvs
*ipvs
= net_ipvs(net
);
2039 const struct ip_vs_service
*svc
= v
;
2040 const struct ip_vs_iter
*iter
= seq
->private;
2041 const struct ip_vs_dest
*dest
;
2042 struct ip_vs_scheduler
*sched
= rcu_dereference(svc
->scheduler
);
2043 char *sched_name
= sched
? sched
->name
: "none";
2045 if (svc
->ipvs
!= ipvs
)
2047 if (iter
->table
== ip_vs_svc_table
) {
2048 #ifdef CONFIG_IP_VS_IPV6
2049 if (svc
->af
== AF_INET6
)
2050 seq_printf(seq
, "%s [%pI6]:%04X %s ",
2051 ip_vs_proto_name(svc
->protocol
),
2057 seq_printf(seq
, "%s %08X:%04X %s %s ",
2058 ip_vs_proto_name(svc
->protocol
),
2059 ntohl(svc
->addr
.ip
),
2062 (svc
->flags
& IP_VS_SVC_F_ONEPACKET
)?"ops ":"");
2064 seq_printf(seq
, "FWM %08X %s %s",
2065 svc
->fwmark
, sched_name
,
2066 (svc
->flags
& IP_VS_SVC_F_ONEPACKET
)?"ops ":"");
2069 if (svc
->flags
& IP_VS_SVC_F_PERSISTENT
)
2070 seq_printf(seq
, "persistent %d %08X\n",
2072 ntohl(svc
->netmask
));
2074 seq_putc(seq
, '\n');
2076 list_for_each_entry_rcu(dest
, &svc
->destinations
, n_list
) {
2077 #ifdef CONFIG_IP_VS_IPV6
2078 if (dest
->af
== AF_INET6
)
2081 " %-7s %-6d %-10d %-10d\n",
2084 ip_vs_fwd_name(atomic_read(&dest
->conn_flags
)),
2085 atomic_read(&dest
->weight
),
2086 atomic_read(&dest
->activeconns
),
2087 atomic_read(&dest
->inactconns
));
2092 "%-7s %-6d %-10d %-10d\n",
2093 ntohl(dest
->addr
.ip
),
2095 ip_vs_fwd_name(atomic_read(&dest
->conn_flags
)),
2096 atomic_read(&dest
->weight
),
2097 atomic_read(&dest
->activeconns
),
2098 atomic_read(&dest
->inactconns
));
2105 static const struct seq_operations ip_vs_info_seq_ops
= {
2106 .start
= ip_vs_info_seq_start
,
2107 .next
= ip_vs_info_seq_next
,
2108 .stop
= ip_vs_info_seq_stop
,
2109 .show
= ip_vs_info_seq_show
,
2112 static int ip_vs_info_open(struct inode
*inode
, struct file
*file
)
2114 return seq_open_net(inode
, file
, &ip_vs_info_seq_ops
,
2115 sizeof(struct ip_vs_iter
));
2118 static const struct file_operations ip_vs_info_fops
= {
2119 .owner
= THIS_MODULE
,
2120 .open
= ip_vs_info_open
,
2122 .llseek
= seq_lseek
,
2123 .release
= seq_release_net
,
2126 static int ip_vs_stats_show(struct seq_file
*seq
, void *v
)
2128 struct net
*net
= seq_file_single_net(seq
);
2129 struct ip_vs_kstats show
;
2131 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2133 " Total Incoming Outgoing Incoming Outgoing\n");
2135 " Conns Packets Packets Bytes Bytes\n");
2137 ip_vs_copy_stats(&show
, &net_ipvs(net
)->tot_stats
);
2138 seq_printf(seq
, "%8LX %8LX %8LX %16LX %16LX\n\n",
2139 (unsigned long long)show
.conns
,
2140 (unsigned long long)show
.inpkts
,
2141 (unsigned long long)show
.outpkts
,
2142 (unsigned long long)show
.inbytes
,
2143 (unsigned long long)show
.outbytes
);
2145 /* 01234567 01234567 01234567 0123456701234567 0123456701234567*/
2147 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2148 seq_printf(seq
, "%8LX %8LX %8LX %16LX %16LX\n",
2149 (unsigned long long)show
.cps
,
2150 (unsigned long long)show
.inpps
,
2151 (unsigned long long)show
.outpps
,
2152 (unsigned long long)show
.inbps
,
2153 (unsigned long long)show
.outbps
);
2158 static int ip_vs_stats_seq_open(struct inode
*inode
, struct file
*file
)
2160 return single_open_net(inode
, file
, ip_vs_stats_show
);
2163 static const struct file_operations ip_vs_stats_fops
= {
2164 .owner
= THIS_MODULE
,
2165 .open
= ip_vs_stats_seq_open
,
2167 .llseek
= seq_lseek
,
2168 .release
= single_release_net
,
2171 static int ip_vs_stats_percpu_show(struct seq_file
*seq
, void *v
)
2173 struct net
*net
= seq_file_single_net(seq
);
2174 struct ip_vs_stats
*tot_stats
= &net_ipvs(net
)->tot_stats
;
2175 struct ip_vs_cpu_stats __percpu
*cpustats
= tot_stats
->cpustats
;
2176 struct ip_vs_kstats kstats
;
2179 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2181 " Total Incoming Outgoing Incoming Outgoing\n");
2183 "CPU Conns Packets Packets Bytes Bytes\n");
2185 for_each_possible_cpu(i
) {
2186 struct ip_vs_cpu_stats
*u
= per_cpu_ptr(cpustats
, i
);
2188 u64 conns
, inpkts
, outpkts
, inbytes
, outbytes
;
2191 start
= u64_stats_fetch_begin_irq(&u
->syncp
);
2192 conns
= u
->cnt
.conns
;
2193 inpkts
= u
->cnt
.inpkts
;
2194 outpkts
= u
->cnt
.outpkts
;
2195 inbytes
= u
->cnt
.inbytes
;
2196 outbytes
= u
->cnt
.outbytes
;
2197 } while (u64_stats_fetch_retry_irq(&u
->syncp
, start
));
2199 seq_printf(seq
, "%3X %8LX %8LX %8LX %16LX %16LX\n",
2200 i
, (u64
)conns
, (u64
)inpkts
,
2201 (u64
)outpkts
, (u64
)inbytes
,
2205 ip_vs_copy_stats(&kstats
, tot_stats
);
2207 seq_printf(seq
, " ~ %8LX %8LX %8LX %16LX %16LX\n\n",
2208 (unsigned long long)kstats
.conns
,
2209 (unsigned long long)kstats
.inpkts
,
2210 (unsigned long long)kstats
.outpkts
,
2211 (unsigned long long)kstats
.inbytes
,
2212 (unsigned long long)kstats
.outbytes
);
2214 /* ... 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2216 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2217 seq_printf(seq
, " %8LX %8LX %8LX %16LX %16LX\n",
2227 static int ip_vs_stats_percpu_seq_open(struct inode
*inode
, struct file
*file
)
2229 return single_open_net(inode
, file
, ip_vs_stats_percpu_show
);
2232 static const struct file_operations ip_vs_stats_percpu_fops
= {
2233 .owner
= THIS_MODULE
,
2234 .open
= ip_vs_stats_percpu_seq_open
,
2236 .llseek
= seq_lseek
,
2237 .release
= single_release_net
,
2242 * Set timeout values for tcp tcpfin udp in the timeout_table.
2244 static int ip_vs_set_timeout(struct netns_ipvs
*ipvs
, struct ip_vs_timeout_user
*u
)
2246 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2247 struct ip_vs_proto_data
*pd
;
2250 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2255 #ifdef CONFIG_IP_VS_PROTO_TCP
2256 if (u
->tcp_timeout
< 0 || u
->tcp_timeout
> (INT_MAX
/ HZ
) ||
2257 u
->tcp_fin_timeout
< 0 || u
->tcp_fin_timeout
> (INT_MAX
/ HZ
)) {
2262 #ifdef CONFIG_IP_VS_PROTO_UDP
2263 if (u
->udp_timeout
< 0 || u
->udp_timeout
> (INT_MAX
/ HZ
))
2267 #ifdef CONFIG_IP_VS_PROTO_TCP
2268 if (u
->tcp_timeout
) {
2269 pd
= ip_vs_proto_data_get(ipvs
, IPPROTO_TCP
);
2270 pd
->timeout_table
[IP_VS_TCP_S_ESTABLISHED
]
2271 = u
->tcp_timeout
* HZ
;
2274 if (u
->tcp_fin_timeout
) {
2275 pd
= ip_vs_proto_data_get(ipvs
, IPPROTO_TCP
);
2276 pd
->timeout_table
[IP_VS_TCP_S_FIN_WAIT
]
2277 = u
->tcp_fin_timeout
* HZ
;
2281 #ifdef CONFIG_IP_VS_PROTO_UDP
2282 if (u
->udp_timeout
) {
2283 pd
= ip_vs_proto_data_get(ipvs
, IPPROTO_UDP
);
2284 pd
->timeout_table
[IP_VS_UDP_S_NORMAL
]
2285 = u
->udp_timeout
* HZ
;
2291 #define CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2293 struct ip_vs_svcdest_user
{
2294 struct ip_vs_service_user s
;
2295 struct ip_vs_dest_user d
;
2298 static const unsigned char set_arglen
[CMDID(IP_VS_SO_SET_MAX
) + 1] = {
2299 [CMDID(IP_VS_SO_SET_ADD
)] = sizeof(struct ip_vs_service_user
),
2300 [CMDID(IP_VS_SO_SET_EDIT
)] = sizeof(struct ip_vs_service_user
),
2301 [CMDID(IP_VS_SO_SET_DEL
)] = sizeof(struct ip_vs_service_user
),
2302 [CMDID(IP_VS_SO_SET_ADDDEST
)] = sizeof(struct ip_vs_svcdest_user
),
2303 [CMDID(IP_VS_SO_SET_DELDEST
)] = sizeof(struct ip_vs_svcdest_user
),
2304 [CMDID(IP_VS_SO_SET_EDITDEST
)] = sizeof(struct ip_vs_svcdest_user
),
2305 [CMDID(IP_VS_SO_SET_TIMEOUT
)] = sizeof(struct ip_vs_timeout_user
),
2306 [CMDID(IP_VS_SO_SET_STARTDAEMON
)] = sizeof(struct ip_vs_daemon_user
),
2307 [CMDID(IP_VS_SO_SET_STOPDAEMON
)] = sizeof(struct ip_vs_daemon_user
),
2308 [CMDID(IP_VS_SO_SET_ZERO
)] = sizeof(struct ip_vs_service_user
),
2311 union ip_vs_set_arglen
{
2312 struct ip_vs_service_user field_IP_VS_SO_SET_ADD
;
2313 struct ip_vs_service_user field_IP_VS_SO_SET_EDIT
;
2314 struct ip_vs_service_user field_IP_VS_SO_SET_DEL
;
2315 struct ip_vs_svcdest_user field_IP_VS_SO_SET_ADDDEST
;
2316 struct ip_vs_svcdest_user field_IP_VS_SO_SET_DELDEST
;
2317 struct ip_vs_svcdest_user field_IP_VS_SO_SET_EDITDEST
;
2318 struct ip_vs_timeout_user field_IP_VS_SO_SET_TIMEOUT
;
2319 struct ip_vs_daemon_user field_IP_VS_SO_SET_STARTDAEMON
;
2320 struct ip_vs_daemon_user field_IP_VS_SO_SET_STOPDAEMON
;
2321 struct ip_vs_service_user field_IP_VS_SO_SET_ZERO
;
2324 #define MAX_SET_ARGLEN sizeof(union ip_vs_set_arglen)
2326 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern
*usvc
,
2327 struct ip_vs_service_user
*usvc_compat
)
2329 memset(usvc
, 0, sizeof(*usvc
));
2332 usvc
->protocol
= usvc_compat
->protocol
;
2333 usvc
->addr
.ip
= usvc_compat
->addr
;
2334 usvc
->port
= usvc_compat
->port
;
2335 usvc
->fwmark
= usvc_compat
->fwmark
;
2337 /* Deep copy of sched_name is not needed here */
2338 usvc
->sched_name
= usvc_compat
->sched_name
;
2340 usvc
->flags
= usvc_compat
->flags
;
2341 usvc
->timeout
= usvc_compat
->timeout
;
2342 usvc
->netmask
= usvc_compat
->netmask
;
2345 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern
*udest
,
2346 struct ip_vs_dest_user
*udest_compat
)
2348 memset(udest
, 0, sizeof(*udest
));
2350 udest
->addr
.ip
= udest_compat
->addr
;
2351 udest
->port
= udest_compat
->port
;
2352 udest
->conn_flags
= udest_compat
->conn_flags
;
2353 udest
->weight
= udest_compat
->weight
;
2354 udest
->u_threshold
= udest_compat
->u_threshold
;
2355 udest
->l_threshold
= udest_compat
->l_threshold
;
2356 udest
->af
= AF_INET
;
2360 do_ip_vs_set_ctl(struct sock
*sk
, int cmd
, void __user
*user
, unsigned int len
)
2362 struct net
*net
= sock_net(sk
);
2364 unsigned char arg
[MAX_SET_ARGLEN
];
2365 struct ip_vs_service_user
*usvc_compat
;
2366 struct ip_vs_service_user_kern usvc
;
2367 struct ip_vs_service
*svc
;
2368 struct ip_vs_dest_user
*udest_compat
;
2369 struct ip_vs_dest_user_kern udest
;
2370 struct netns_ipvs
*ipvs
= net_ipvs(net
);
2372 BUILD_BUG_ON(sizeof(arg
) > 255);
2373 if (!ns_capable(sock_net(sk
)->user_ns
, CAP_NET_ADMIN
))
2376 if (cmd
< IP_VS_BASE_CTL
|| cmd
> IP_VS_SO_SET_MAX
)
2378 if (len
!= set_arglen
[CMDID(cmd
)]) {
2379 IP_VS_DBG(1, "set_ctl: len %u != %u\n",
2380 len
, set_arglen
[CMDID(cmd
)]);
2384 if (copy_from_user(arg
, user
, len
) != 0)
2387 /* increase the module use count */
2388 ip_vs_use_count_inc();
2390 /* Handle daemons since they have another lock */
2391 if (cmd
== IP_VS_SO_SET_STARTDAEMON
||
2392 cmd
== IP_VS_SO_SET_STOPDAEMON
) {
2393 struct ip_vs_daemon_user
*dm
= (struct ip_vs_daemon_user
*)arg
;
2395 if (cmd
== IP_VS_SO_SET_STARTDAEMON
) {
2396 struct ipvs_sync_daemon_cfg cfg
;
2398 memset(&cfg
, 0, sizeof(cfg
));
2400 if (strscpy(cfg
.mcast_ifn
, dm
->mcast_ifn
,
2401 sizeof(cfg
.mcast_ifn
)) <= 0)
2403 cfg
.syncid
= dm
->syncid
;
2404 ret
= start_sync_thread(ipvs
, &cfg
, dm
->state
);
2406 mutex_lock(&ipvs
->sync_mutex
);
2407 ret
= stop_sync_thread(ipvs
, dm
->state
);
2408 mutex_unlock(&ipvs
->sync_mutex
);
2413 mutex_lock(&__ip_vs_mutex
);
2414 if (cmd
== IP_VS_SO_SET_FLUSH
) {
2415 /* Flush the virtual service */
2416 ret
= ip_vs_flush(ipvs
, false);
2418 } else if (cmd
== IP_VS_SO_SET_TIMEOUT
) {
2419 /* Set timeout values for (tcp tcpfin udp) */
2420 ret
= ip_vs_set_timeout(ipvs
, (struct ip_vs_timeout_user
*)arg
);
2424 usvc_compat
= (struct ip_vs_service_user
*)arg
;
2425 udest_compat
= (struct ip_vs_dest_user
*)(usvc_compat
+ 1);
2427 /* We only use the new structs internally, so copy userspace compat
2428 * structs to extended internal versions */
2429 ip_vs_copy_usvc_compat(&usvc
, usvc_compat
);
2430 ip_vs_copy_udest_compat(&udest
, udest_compat
);
2432 if (cmd
== IP_VS_SO_SET_ZERO
) {
2433 /* if no service address is set, zero counters in all */
2434 if (!usvc
.fwmark
&& !usvc
.addr
.ip
&& !usvc
.port
) {
2435 ret
= ip_vs_zero_all(ipvs
);
2440 if ((cmd
== IP_VS_SO_SET_ADD
|| cmd
== IP_VS_SO_SET_EDIT
) &&
2441 strnlen(usvc
.sched_name
, IP_VS_SCHEDNAME_MAXLEN
) ==
2442 IP_VS_SCHEDNAME_MAXLEN
) {
2447 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2448 if (usvc
.protocol
!= IPPROTO_TCP
&& usvc
.protocol
!= IPPROTO_UDP
&&
2449 usvc
.protocol
!= IPPROTO_SCTP
) {
2450 pr_err("set_ctl: invalid protocol: %d %pI4:%d\n",
2451 usvc
.protocol
, &usvc
.addr
.ip
,
2457 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2459 if (usvc
.fwmark
== 0)
2460 svc
= __ip_vs_service_find(ipvs
, usvc
.af
, usvc
.protocol
,
2461 &usvc
.addr
, usvc
.port
);
2463 svc
= __ip_vs_svc_fwm_find(ipvs
, usvc
.af
, usvc
.fwmark
);
2466 if (cmd
!= IP_VS_SO_SET_ADD
2467 && (svc
== NULL
|| svc
->protocol
!= usvc
.protocol
)) {
2473 case IP_VS_SO_SET_ADD
:
2477 ret
= ip_vs_add_service(ipvs
, &usvc
, &svc
);
2479 case IP_VS_SO_SET_EDIT
:
2480 ret
= ip_vs_edit_service(svc
, &usvc
);
2482 case IP_VS_SO_SET_DEL
:
2483 ret
= ip_vs_del_service(svc
);
2487 case IP_VS_SO_SET_ZERO
:
2488 ret
= ip_vs_zero_service(svc
);
2490 case IP_VS_SO_SET_ADDDEST
:
2491 ret
= ip_vs_add_dest(svc
, &udest
);
2493 case IP_VS_SO_SET_EDITDEST
:
2494 ret
= ip_vs_edit_dest(svc
, &udest
);
2496 case IP_VS_SO_SET_DELDEST
:
2497 ret
= ip_vs_del_dest(svc
, &udest
);
2504 mutex_unlock(&__ip_vs_mutex
);
2506 /* decrease the module use count */
2507 ip_vs_use_count_dec();
2514 ip_vs_copy_service(struct ip_vs_service_entry
*dst
, struct ip_vs_service
*src
)
2516 struct ip_vs_scheduler
*sched
;
2517 struct ip_vs_kstats kstats
;
2520 sched
= rcu_dereference_protected(src
->scheduler
, 1);
2521 sched_name
= sched
? sched
->name
: "none";
2522 dst
->protocol
= src
->protocol
;
2523 dst
->addr
= src
->addr
.ip
;
2524 dst
->port
= src
->port
;
2525 dst
->fwmark
= src
->fwmark
;
2526 strlcpy(dst
->sched_name
, sched_name
, sizeof(dst
->sched_name
));
2527 dst
->flags
= src
->flags
;
2528 dst
->timeout
= src
->timeout
/ HZ
;
2529 dst
->netmask
= src
->netmask
;
2530 dst
->num_dests
= src
->num_dests
;
2531 ip_vs_copy_stats(&kstats
, &src
->stats
);
2532 ip_vs_export_stats_user(&dst
->stats
, &kstats
);
2536 __ip_vs_get_service_entries(struct netns_ipvs
*ipvs
,
2537 const struct ip_vs_get_services
*get
,
2538 struct ip_vs_get_services __user
*uptr
)
2541 struct ip_vs_service
*svc
;
2542 struct ip_vs_service_entry entry
;
2545 for (idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
2546 hlist_for_each_entry(svc
, &ip_vs_svc_table
[idx
], s_list
) {
2547 /* Only expose IPv4 entries to old interface */
2548 if (svc
->af
!= AF_INET
|| (svc
->ipvs
!= ipvs
))
2551 if (count
>= get
->num_services
)
2553 memset(&entry
, 0, sizeof(entry
));
2554 ip_vs_copy_service(&entry
, svc
);
2555 if (copy_to_user(&uptr
->entrytable
[count
],
2556 &entry
, sizeof(entry
))) {
2564 for (idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
2565 hlist_for_each_entry(svc
, &ip_vs_svc_fwm_table
[idx
], f_list
) {
2566 /* Only expose IPv4 entries to old interface */
2567 if (svc
->af
!= AF_INET
|| (svc
->ipvs
!= ipvs
))
2570 if (count
>= get
->num_services
)
2572 memset(&entry
, 0, sizeof(entry
));
2573 ip_vs_copy_service(&entry
, svc
);
2574 if (copy_to_user(&uptr
->entrytable
[count
],
2575 &entry
, sizeof(entry
))) {
2587 __ip_vs_get_dest_entries(struct netns_ipvs
*ipvs
, const struct ip_vs_get_dests
*get
,
2588 struct ip_vs_get_dests __user
*uptr
)
2590 struct ip_vs_service
*svc
;
2591 union nf_inet_addr addr
= { .ip
= get
->addr
};
2596 svc
= __ip_vs_svc_fwm_find(ipvs
, AF_INET
, get
->fwmark
);
2598 svc
= __ip_vs_service_find(ipvs
, AF_INET
, get
->protocol
, &addr
,
2604 struct ip_vs_dest
*dest
;
2605 struct ip_vs_dest_entry entry
;
2606 struct ip_vs_kstats kstats
;
2608 memset(&entry
, 0, sizeof(entry
));
2609 list_for_each_entry(dest
, &svc
->destinations
, n_list
) {
2610 if (count
>= get
->num_dests
)
2613 /* Cannot expose heterogeneous members via sockopt
2616 if (dest
->af
!= svc
->af
)
2619 entry
.addr
= dest
->addr
.ip
;
2620 entry
.port
= dest
->port
;
2621 entry
.conn_flags
= atomic_read(&dest
->conn_flags
);
2622 entry
.weight
= atomic_read(&dest
->weight
);
2623 entry
.u_threshold
= dest
->u_threshold
;
2624 entry
.l_threshold
= dest
->l_threshold
;
2625 entry
.activeconns
= atomic_read(&dest
->activeconns
);
2626 entry
.inactconns
= atomic_read(&dest
->inactconns
);
2627 entry
.persistconns
= atomic_read(&dest
->persistconns
);
2628 ip_vs_copy_stats(&kstats
, &dest
->stats
);
2629 ip_vs_export_stats_user(&entry
.stats
, &kstats
);
2630 if (copy_to_user(&uptr
->entrytable
[count
],
2631 &entry
, sizeof(entry
))) {
2643 __ip_vs_get_timeouts(struct netns_ipvs
*ipvs
, struct ip_vs_timeout_user
*u
)
2645 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2646 struct ip_vs_proto_data
*pd
;
2649 memset(u
, 0, sizeof (*u
));
2651 #ifdef CONFIG_IP_VS_PROTO_TCP
2652 pd
= ip_vs_proto_data_get(ipvs
, IPPROTO_TCP
);
2653 u
->tcp_timeout
= pd
->timeout_table
[IP_VS_TCP_S_ESTABLISHED
] / HZ
;
2654 u
->tcp_fin_timeout
= pd
->timeout_table
[IP_VS_TCP_S_FIN_WAIT
] / HZ
;
2656 #ifdef CONFIG_IP_VS_PROTO_UDP
2657 pd
= ip_vs_proto_data_get(ipvs
, IPPROTO_UDP
);
2659 pd
->timeout_table
[IP_VS_UDP_S_NORMAL
] / HZ
;
2663 static const unsigned char get_arglen
[CMDID(IP_VS_SO_GET_MAX
) + 1] = {
2664 [CMDID(IP_VS_SO_GET_VERSION
)] = 64,
2665 [CMDID(IP_VS_SO_GET_INFO
)] = sizeof(struct ip_vs_getinfo
),
2666 [CMDID(IP_VS_SO_GET_SERVICES
)] = sizeof(struct ip_vs_get_services
),
2667 [CMDID(IP_VS_SO_GET_SERVICE
)] = sizeof(struct ip_vs_service_entry
),
2668 [CMDID(IP_VS_SO_GET_DESTS
)] = sizeof(struct ip_vs_get_dests
),
2669 [CMDID(IP_VS_SO_GET_TIMEOUT
)] = sizeof(struct ip_vs_timeout_user
),
2670 [CMDID(IP_VS_SO_GET_DAEMON
)] = 2 * sizeof(struct ip_vs_daemon_user
),
2673 union ip_vs_get_arglen
{
2674 char field_IP_VS_SO_GET_VERSION
[64];
2675 struct ip_vs_getinfo field_IP_VS_SO_GET_INFO
;
2676 struct ip_vs_get_services field_IP_VS_SO_GET_SERVICES
;
2677 struct ip_vs_service_entry field_IP_VS_SO_GET_SERVICE
;
2678 struct ip_vs_get_dests field_IP_VS_SO_GET_DESTS
;
2679 struct ip_vs_timeout_user field_IP_VS_SO_GET_TIMEOUT
;
2680 struct ip_vs_daemon_user field_IP_VS_SO_GET_DAEMON
[2];
2683 #define MAX_GET_ARGLEN sizeof(union ip_vs_get_arglen)
2686 do_ip_vs_get_ctl(struct sock
*sk
, int cmd
, void __user
*user
, int *len
)
2688 unsigned char arg
[MAX_GET_ARGLEN
];
2690 unsigned int copylen
;
2691 struct net
*net
= sock_net(sk
);
2692 struct netns_ipvs
*ipvs
= net_ipvs(net
);
2695 BUILD_BUG_ON(sizeof(arg
) > 255);
2696 if (!ns_capable(sock_net(sk
)->user_ns
, CAP_NET_ADMIN
))
2699 if (cmd
< IP_VS_BASE_CTL
|| cmd
> IP_VS_SO_GET_MAX
)
2702 copylen
= get_arglen
[CMDID(cmd
)];
2703 if (*len
< (int) copylen
) {
2704 IP_VS_DBG(1, "get_ctl: len %d < %u\n", *len
, copylen
);
2708 if (copy_from_user(arg
, user
, copylen
) != 0)
2711 * Handle daemons first since it has its own locking
2713 if (cmd
== IP_VS_SO_GET_DAEMON
) {
2714 struct ip_vs_daemon_user d
[2];
2716 memset(&d
, 0, sizeof(d
));
2717 mutex_lock(&ipvs
->sync_mutex
);
2718 if (ipvs
->sync_state
& IP_VS_STATE_MASTER
) {
2719 d
[0].state
= IP_VS_STATE_MASTER
;
2720 strlcpy(d
[0].mcast_ifn
, ipvs
->mcfg
.mcast_ifn
,
2721 sizeof(d
[0].mcast_ifn
));
2722 d
[0].syncid
= ipvs
->mcfg
.syncid
;
2724 if (ipvs
->sync_state
& IP_VS_STATE_BACKUP
) {
2725 d
[1].state
= IP_VS_STATE_BACKUP
;
2726 strlcpy(d
[1].mcast_ifn
, ipvs
->bcfg
.mcast_ifn
,
2727 sizeof(d
[1].mcast_ifn
));
2728 d
[1].syncid
= ipvs
->bcfg
.syncid
;
2730 if (copy_to_user(user
, &d
, sizeof(d
)) != 0)
2732 mutex_unlock(&ipvs
->sync_mutex
);
2736 mutex_lock(&__ip_vs_mutex
);
2738 case IP_VS_SO_GET_VERSION
:
2742 sprintf(buf
, "IP Virtual Server version %d.%d.%d (size=%d)",
2743 NVERSION(IP_VS_VERSION_CODE
), ip_vs_conn_tab_size
);
2744 if (copy_to_user(user
, buf
, strlen(buf
)+1) != 0) {
2748 *len
= strlen(buf
)+1;
2752 case IP_VS_SO_GET_INFO
:
2754 struct ip_vs_getinfo info
;
2755 info
.version
= IP_VS_VERSION_CODE
;
2756 info
.size
= ip_vs_conn_tab_size
;
2757 info
.num_services
= ipvs
->num_services
;
2758 if (copy_to_user(user
, &info
, sizeof(info
)) != 0)
2763 case IP_VS_SO_GET_SERVICES
:
2765 struct ip_vs_get_services
*get
;
2768 get
= (struct ip_vs_get_services
*)arg
;
2769 size
= sizeof(*get
) +
2770 sizeof(struct ip_vs_service_entry
) * get
->num_services
;
2772 pr_err("length: %u != %u\n", *len
, size
);
2776 ret
= __ip_vs_get_service_entries(ipvs
, get
, user
);
2780 case IP_VS_SO_GET_SERVICE
:
2782 struct ip_vs_service_entry
*entry
;
2783 struct ip_vs_service
*svc
;
2784 union nf_inet_addr addr
;
2786 entry
= (struct ip_vs_service_entry
*)arg
;
2787 addr
.ip
= entry
->addr
;
2790 svc
= __ip_vs_svc_fwm_find(ipvs
, AF_INET
, entry
->fwmark
);
2792 svc
= __ip_vs_service_find(ipvs
, AF_INET
,
2793 entry
->protocol
, &addr
,
2797 ip_vs_copy_service(entry
, svc
);
2798 if (copy_to_user(user
, entry
, sizeof(*entry
)) != 0)
2805 case IP_VS_SO_GET_DESTS
:
2807 struct ip_vs_get_dests
*get
;
2810 get
= (struct ip_vs_get_dests
*)arg
;
2811 size
= sizeof(*get
) +
2812 sizeof(struct ip_vs_dest_entry
) * get
->num_dests
;
2814 pr_err("length: %u != %u\n", *len
, size
);
2818 ret
= __ip_vs_get_dest_entries(ipvs
, get
, user
);
2822 case IP_VS_SO_GET_TIMEOUT
:
2824 struct ip_vs_timeout_user t
;
2826 __ip_vs_get_timeouts(ipvs
, &t
);
2827 if (copy_to_user(user
, &t
, sizeof(t
)) != 0)
2837 mutex_unlock(&__ip_vs_mutex
);
2842 static struct nf_sockopt_ops ip_vs_sockopts
= {
2844 .set_optmin
= IP_VS_BASE_CTL
,
2845 .set_optmax
= IP_VS_SO_SET_MAX
+1,
2846 .set
= do_ip_vs_set_ctl
,
2847 .get_optmin
= IP_VS_BASE_CTL
,
2848 .get_optmax
= IP_VS_SO_GET_MAX
+1,
2849 .get
= do_ip_vs_get_ctl
,
2850 .owner
= THIS_MODULE
,
2854 * Generic Netlink interface
2857 /* IPVS genetlink family */
2858 static struct genl_family ip_vs_genl_family
;
2860 /* Policy used for first-level command attributes */
2861 static const struct nla_policy ip_vs_cmd_policy
[IPVS_CMD_ATTR_MAX
+ 1] = {
2862 [IPVS_CMD_ATTR_SERVICE
] = { .type
= NLA_NESTED
},
2863 [IPVS_CMD_ATTR_DEST
] = { .type
= NLA_NESTED
},
2864 [IPVS_CMD_ATTR_DAEMON
] = { .type
= NLA_NESTED
},
2865 [IPVS_CMD_ATTR_TIMEOUT_TCP
] = { .type
= NLA_U32
},
2866 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN
] = { .type
= NLA_U32
},
2867 [IPVS_CMD_ATTR_TIMEOUT_UDP
] = { .type
= NLA_U32
},
2870 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2871 static const struct nla_policy ip_vs_daemon_policy
[IPVS_DAEMON_ATTR_MAX
+ 1] = {
2872 [IPVS_DAEMON_ATTR_STATE
] = { .type
= NLA_U32
},
2873 [IPVS_DAEMON_ATTR_MCAST_IFN
] = { .type
= NLA_NUL_STRING
,
2874 .len
= IP_VS_IFNAME_MAXLEN
- 1 },
2875 [IPVS_DAEMON_ATTR_SYNC_ID
] = { .type
= NLA_U32
},
2876 [IPVS_DAEMON_ATTR_SYNC_MAXLEN
] = { .type
= NLA_U16
},
2877 [IPVS_DAEMON_ATTR_MCAST_GROUP
] = { .type
= NLA_U32
},
2878 [IPVS_DAEMON_ATTR_MCAST_GROUP6
] = { .len
= sizeof(struct in6_addr
) },
2879 [IPVS_DAEMON_ATTR_MCAST_PORT
] = { .type
= NLA_U16
},
2880 [IPVS_DAEMON_ATTR_MCAST_TTL
] = { .type
= NLA_U8
},
2883 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2884 static const struct nla_policy ip_vs_svc_policy
[IPVS_SVC_ATTR_MAX
+ 1] = {
2885 [IPVS_SVC_ATTR_AF
] = { .type
= NLA_U16
},
2886 [IPVS_SVC_ATTR_PROTOCOL
] = { .type
= NLA_U16
},
2887 [IPVS_SVC_ATTR_ADDR
] = { .type
= NLA_BINARY
,
2888 .len
= sizeof(union nf_inet_addr
) },
2889 [IPVS_SVC_ATTR_PORT
] = { .type
= NLA_U16
},
2890 [IPVS_SVC_ATTR_FWMARK
] = { .type
= NLA_U32
},
2891 [IPVS_SVC_ATTR_SCHED_NAME
] = { .type
= NLA_NUL_STRING
,
2892 .len
= IP_VS_SCHEDNAME_MAXLEN
- 1 },
2893 [IPVS_SVC_ATTR_PE_NAME
] = { .type
= NLA_NUL_STRING
,
2894 .len
= IP_VS_PENAME_MAXLEN
},
2895 [IPVS_SVC_ATTR_FLAGS
] = { .type
= NLA_BINARY
,
2896 .len
= sizeof(struct ip_vs_flags
) },
2897 [IPVS_SVC_ATTR_TIMEOUT
] = { .type
= NLA_U32
},
2898 [IPVS_SVC_ATTR_NETMASK
] = { .type
= NLA_U32
},
2899 [IPVS_SVC_ATTR_STATS
] = { .type
= NLA_NESTED
},
2902 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2903 static const struct nla_policy ip_vs_dest_policy
[IPVS_DEST_ATTR_MAX
+ 1] = {
2904 [IPVS_DEST_ATTR_ADDR
] = { .type
= NLA_BINARY
,
2905 .len
= sizeof(union nf_inet_addr
) },
2906 [IPVS_DEST_ATTR_PORT
] = { .type
= NLA_U16
},
2907 [IPVS_DEST_ATTR_FWD_METHOD
] = { .type
= NLA_U32
},
2908 [IPVS_DEST_ATTR_WEIGHT
] = { .type
= NLA_U32
},
2909 [IPVS_DEST_ATTR_U_THRESH
] = { .type
= NLA_U32
},
2910 [IPVS_DEST_ATTR_L_THRESH
] = { .type
= NLA_U32
},
2911 [IPVS_DEST_ATTR_ACTIVE_CONNS
] = { .type
= NLA_U32
},
2912 [IPVS_DEST_ATTR_INACT_CONNS
] = { .type
= NLA_U32
},
2913 [IPVS_DEST_ATTR_PERSIST_CONNS
] = { .type
= NLA_U32
},
2914 [IPVS_DEST_ATTR_STATS
] = { .type
= NLA_NESTED
},
2915 [IPVS_DEST_ATTR_ADDR_FAMILY
] = { .type
= NLA_U16
},
2918 static int ip_vs_genl_fill_stats(struct sk_buff
*skb
, int container_type
,
2919 struct ip_vs_kstats
*kstats
)
2921 struct nlattr
*nl_stats
= nla_nest_start(skb
, container_type
);
2926 if (nla_put_u32(skb
, IPVS_STATS_ATTR_CONNS
, (u32
)kstats
->conns
) ||
2927 nla_put_u32(skb
, IPVS_STATS_ATTR_INPKTS
, (u32
)kstats
->inpkts
) ||
2928 nla_put_u32(skb
, IPVS_STATS_ATTR_OUTPKTS
, (u32
)kstats
->outpkts
) ||
2929 nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_INBYTES
, kstats
->inbytes
,
2930 IPVS_STATS_ATTR_PAD
) ||
2931 nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_OUTBYTES
, kstats
->outbytes
,
2932 IPVS_STATS_ATTR_PAD
) ||
2933 nla_put_u32(skb
, IPVS_STATS_ATTR_CPS
, (u32
)kstats
->cps
) ||
2934 nla_put_u32(skb
, IPVS_STATS_ATTR_INPPS
, (u32
)kstats
->inpps
) ||
2935 nla_put_u32(skb
, IPVS_STATS_ATTR_OUTPPS
, (u32
)kstats
->outpps
) ||
2936 nla_put_u32(skb
, IPVS_STATS_ATTR_INBPS
, (u32
)kstats
->inbps
) ||
2937 nla_put_u32(skb
, IPVS_STATS_ATTR_OUTBPS
, (u32
)kstats
->outbps
))
2938 goto nla_put_failure
;
2939 nla_nest_end(skb
, nl_stats
);
2944 nla_nest_cancel(skb
, nl_stats
);
2948 static int ip_vs_genl_fill_stats64(struct sk_buff
*skb
, int container_type
,
2949 struct ip_vs_kstats
*kstats
)
2951 struct nlattr
*nl_stats
= nla_nest_start(skb
, container_type
);
2956 if (nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_CONNS
, kstats
->conns
,
2957 IPVS_STATS_ATTR_PAD
) ||
2958 nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_INPKTS
, kstats
->inpkts
,
2959 IPVS_STATS_ATTR_PAD
) ||
2960 nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_OUTPKTS
, kstats
->outpkts
,
2961 IPVS_STATS_ATTR_PAD
) ||
2962 nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_INBYTES
, kstats
->inbytes
,
2963 IPVS_STATS_ATTR_PAD
) ||
2964 nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_OUTBYTES
, kstats
->outbytes
,
2965 IPVS_STATS_ATTR_PAD
) ||
2966 nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_CPS
, kstats
->cps
,
2967 IPVS_STATS_ATTR_PAD
) ||
2968 nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_INPPS
, kstats
->inpps
,
2969 IPVS_STATS_ATTR_PAD
) ||
2970 nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_OUTPPS
, kstats
->outpps
,
2971 IPVS_STATS_ATTR_PAD
) ||
2972 nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_INBPS
, kstats
->inbps
,
2973 IPVS_STATS_ATTR_PAD
) ||
2974 nla_put_u64_64bit(skb
, IPVS_STATS_ATTR_OUTBPS
, kstats
->outbps
,
2975 IPVS_STATS_ATTR_PAD
))
2976 goto nla_put_failure
;
2977 nla_nest_end(skb
, nl_stats
);
2982 nla_nest_cancel(skb
, nl_stats
);
2986 static int ip_vs_genl_fill_service(struct sk_buff
*skb
,
2987 struct ip_vs_service
*svc
)
2989 struct ip_vs_scheduler
*sched
;
2990 struct ip_vs_pe
*pe
;
2991 struct nlattr
*nl_service
;
2992 struct ip_vs_flags flags
= { .flags
= svc
->flags
,
2994 struct ip_vs_kstats kstats
;
2997 nl_service
= nla_nest_start(skb
, IPVS_CMD_ATTR_SERVICE
);
3001 if (nla_put_u16(skb
, IPVS_SVC_ATTR_AF
, svc
->af
))
3002 goto nla_put_failure
;
3004 if (nla_put_u32(skb
, IPVS_SVC_ATTR_FWMARK
, svc
->fwmark
))
3005 goto nla_put_failure
;
3007 if (nla_put_u16(skb
, IPVS_SVC_ATTR_PROTOCOL
, svc
->protocol
) ||
3008 nla_put(skb
, IPVS_SVC_ATTR_ADDR
, sizeof(svc
->addr
), &svc
->addr
) ||
3009 nla_put_be16(skb
, IPVS_SVC_ATTR_PORT
, svc
->port
))
3010 goto nla_put_failure
;
3013 sched
= rcu_dereference_protected(svc
->scheduler
, 1);
3014 sched_name
= sched
? sched
->name
: "none";
3015 pe
= rcu_dereference_protected(svc
->pe
, 1);
3016 if (nla_put_string(skb
, IPVS_SVC_ATTR_SCHED_NAME
, sched_name
) ||
3017 (pe
&& nla_put_string(skb
, IPVS_SVC_ATTR_PE_NAME
, pe
->name
)) ||
3018 nla_put(skb
, IPVS_SVC_ATTR_FLAGS
, sizeof(flags
), &flags
) ||
3019 nla_put_u32(skb
, IPVS_SVC_ATTR_TIMEOUT
, svc
->timeout
/ HZ
) ||
3020 nla_put_be32(skb
, IPVS_SVC_ATTR_NETMASK
, svc
->netmask
))
3021 goto nla_put_failure
;
3022 ip_vs_copy_stats(&kstats
, &svc
->stats
);
3023 if (ip_vs_genl_fill_stats(skb
, IPVS_SVC_ATTR_STATS
, &kstats
))
3024 goto nla_put_failure
;
3025 if (ip_vs_genl_fill_stats64(skb
, IPVS_SVC_ATTR_STATS64
, &kstats
))
3026 goto nla_put_failure
;
3028 nla_nest_end(skb
, nl_service
);
3033 nla_nest_cancel(skb
, nl_service
);
3037 static int ip_vs_genl_dump_service(struct sk_buff
*skb
,
3038 struct ip_vs_service
*svc
,
3039 struct netlink_callback
*cb
)
3043 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
3044 &ip_vs_genl_family
, NLM_F_MULTI
,
3045 IPVS_CMD_NEW_SERVICE
);
3049 if (ip_vs_genl_fill_service(skb
, svc
) < 0)
3050 goto nla_put_failure
;
3052 genlmsg_end(skb
, hdr
);
3056 genlmsg_cancel(skb
, hdr
);
3060 static int ip_vs_genl_dump_services(struct sk_buff
*skb
,
3061 struct netlink_callback
*cb
)
3064 int start
= cb
->args
[0];
3065 struct ip_vs_service
*svc
;
3066 struct net
*net
= sock_net(skb
->sk
);
3067 struct netns_ipvs
*ipvs
= net_ipvs(net
);
3069 mutex_lock(&__ip_vs_mutex
);
3070 for (i
= 0; i
< IP_VS_SVC_TAB_SIZE
; i
++) {
3071 hlist_for_each_entry(svc
, &ip_vs_svc_table
[i
], s_list
) {
3072 if (++idx
<= start
|| (svc
->ipvs
!= ipvs
))
3074 if (ip_vs_genl_dump_service(skb
, svc
, cb
) < 0) {
3076 goto nla_put_failure
;
3081 for (i
= 0; i
< IP_VS_SVC_TAB_SIZE
; i
++) {
3082 hlist_for_each_entry(svc
, &ip_vs_svc_fwm_table
[i
], f_list
) {
3083 if (++idx
<= start
|| (svc
->ipvs
!= ipvs
))
3085 if (ip_vs_genl_dump_service(skb
, svc
, cb
) < 0) {
3087 goto nla_put_failure
;
3093 mutex_unlock(&__ip_vs_mutex
);
3099 static bool ip_vs_is_af_valid(int af
)
3103 #ifdef CONFIG_IP_VS_IPV6
3104 if (af
== AF_INET6
&& ipv6_mod_enabled())
3110 static int ip_vs_genl_parse_service(struct netns_ipvs
*ipvs
,
3111 struct ip_vs_service_user_kern
*usvc
,
3112 struct nlattr
*nla
, int full_entry
,
3113 struct ip_vs_service
**ret_svc
)
3115 struct nlattr
*attrs
[IPVS_SVC_ATTR_MAX
+ 1];
3116 struct nlattr
*nla_af
, *nla_port
, *nla_fwmark
, *nla_protocol
, *nla_addr
;
3117 struct ip_vs_service
*svc
;
3119 /* Parse mandatory identifying service fields first */
3121 nla_parse_nested(attrs
, IPVS_SVC_ATTR_MAX
, nla
,
3122 ip_vs_svc_policy
, NULL
))
3125 nla_af
= attrs
[IPVS_SVC_ATTR_AF
];
3126 nla_protocol
= attrs
[IPVS_SVC_ATTR_PROTOCOL
];
3127 nla_addr
= attrs
[IPVS_SVC_ATTR_ADDR
];
3128 nla_port
= attrs
[IPVS_SVC_ATTR_PORT
];
3129 nla_fwmark
= attrs
[IPVS_SVC_ATTR_FWMARK
];
3131 if (!(nla_af
&& (nla_fwmark
|| (nla_port
&& nla_protocol
&& nla_addr
))))
3134 memset(usvc
, 0, sizeof(*usvc
));
3136 usvc
->af
= nla_get_u16(nla_af
);
3137 if (!ip_vs_is_af_valid(usvc
->af
))
3138 return -EAFNOSUPPORT
;
3141 usvc
->protocol
= IPPROTO_TCP
;
3142 usvc
->fwmark
= nla_get_u32(nla_fwmark
);
3144 usvc
->protocol
= nla_get_u16(nla_protocol
);
3145 nla_memcpy(&usvc
->addr
, nla_addr
, sizeof(usvc
->addr
));
3146 usvc
->port
= nla_get_be16(nla_port
);
3152 svc
= __ip_vs_svc_fwm_find(ipvs
, usvc
->af
, usvc
->fwmark
);
3154 svc
= __ip_vs_service_find(ipvs
, usvc
->af
, usvc
->protocol
,
3155 &usvc
->addr
, usvc
->port
);
3159 /* If a full entry was requested, check for the additional fields */
3161 struct nlattr
*nla_sched
, *nla_flags
, *nla_pe
, *nla_timeout
,
3163 struct ip_vs_flags flags
;
3165 nla_sched
= attrs
[IPVS_SVC_ATTR_SCHED_NAME
];
3166 nla_pe
= attrs
[IPVS_SVC_ATTR_PE_NAME
];
3167 nla_flags
= attrs
[IPVS_SVC_ATTR_FLAGS
];
3168 nla_timeout
= attrs
[IPVS_SVC_ATTR_TIMEOUT
];
3169 nla_netmask
= attrs
[IPVS_SVC_ATTR_NETMASK
];
3171 if (!(nla_sched
&& nla_flags
&& nla_timeout
&& nla_netmask
))
3174 nla_memcpy(&flags
, nla_flags
, sizeof(flags
));
3176 /* prefill flags from service if it already exists */
3178 usvc
->flags
= svc
->flags
;
3180 /* set new flags from userland */
3181 usvc
->flags
= (usvc
->flags
& ~flags
.mask
) |
3182 (flags
.flags
& flags
.mask
);
3183 usvc
->sched_name
= nla_data(nla_sched
);
3184 usvc
->pe_name
= nla_pe
? nla_data(nla_pe
) : NULL
;
3185 usvc
->timeout
= nla_get_u32(nla_timeout
);
3186 usvc
->netmask
= nla_get_be32(nla_netmask
);
3192 static struct ip_vs_service
*ip_vs_genl_find_service(struct netns_ipvs
*ipvs
,
3195 struct ip_vs_service_user_kern usvc
;
3196 struct ip_vs_service
*svc
;
3199 ret
= ip_vs_genl_parse_service(ipvs
, &usvc
, nla
, 0, &svc
);
3200 return ret
? ERR_PTR(ret
) : svc
;
3203 static int ip_vs_genl_fill_dest(struct sk_buff
*skb
, struct ip_vs_dest
*dest
)
3205 struct nlattr
*nl_dest
;
3206 struct ip_vs_kstats kstats
;
3208 nl_dest
= nla_nest_start(skb
, IPVS_CMD_ATTR_DEST
);
3212 if (nla_put(skb
, IPVS_DEST_ATTR_ADDR
, sizeof(dest
->addr
), &dest
->addr
) ||
3213 nla_put_be16(skb
, IPVS_DEST_ATTR_PORT
, dest
->port
) ||
3214 nla_put_u32(skb
, IPVS_DEST_ATTR_FWD_METHOD
,
3215 (atomic_read(&dest
->conn_flags
) &
3216 IP_VS_CONN_F_FWD_MASK
)) ||
3217 nla_put_u32(skb
, IPVS_DEST_ATTR_WEIGHT
,
3218 atomic_read(&dest
->weight
)) ||
3219 nla_put_u32(skb
, IPVS_DEST_ATTR_U_THRESH
, dest
->u_threshold
) ||
3220 nla_put_u32(skb
, IPVS_DEST_ATTR_L_THRESH
, dest
->l_threshold
) ||
3221 nla_put_u32(skb
, IPVS_DEST_ATTR_ACTIVE_CONNS
,
3222 atomic_read(&dest
->activeconns
)) ||
3223 nla_put_u32(skb
, IPVS_DEST_ATTR_INACT_CONNS
,
3224 atomic_read(&dest
->inactconns
)) ||
3225 nla_put_u32(skb
, IPVS_DEST_ATTR_PERSIST_CONNS
,
3226 atomic_read(&dest
->persistconns
)) ||
3227 nla_put_u16(skb
, IPVS_DEST_ATTR_ADDR_FAMILY
, dest
->af
))
3228 goto nla_put_failure
;
3229 ip_vs_copy_stats(&kstats
, &dest
->stats
);
3230 if (ip_vs_genl_fill_stats(skb
, IPVS_DEST_ATTR_STATS
, &kstats
))
3231 goto nla_put_failure
;
3232 if (ip_vs_genl_fill_stats64(skb
, IPVS_DEST_ATTR_STATS64
, &kstats
))
3233 goto nla_put_failure
;
3235 nla_nest_end(skb
, nl_dest
);
3240 nla_nest_cancel(skb
, nl_dest
);
3244 static int ip_vs_genl_dump_dest(struct sk_buff
*skb
, struct ip_vs_dest
*dest
,
3245 struct netlink_callback
*cb
)
3249 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
3250 &ip_vs_genl_family
, NLM_F_MULTI
,
3255 if (ip_vs_genl_fill_dest(skb
, dest
) < 0)
3256 goto nla_put_failure
;
3258 genlmsg_end(skb
, hdr
);
3262 genlmsg_cancel(skb
, hdr
);
3266 static int ip_vs_genl_dump_dests(struct sk_buff
*skb
,
3267 struct netlink_callback
*cb
)
3270 int start
= cb
->args
[0];
3271 struct ip_vs_service
*svc
;
3272 struct ip_vs_dest
*dest
;
3273 struct nlattr
*attrs
[IPVS_CMD_ATTR_MAX
+ 1];
3274 struct net
*net
= sock_net(skb
->sk
);
3275 struct netns_ipvs
*ipvs
= net_ipvs(net
);
3277 mutex_lock(&__ip_vs_mutex
);
3279 /* Try to find the service for which to dump destinations */
3280 if (nlmsg_parse(cb
->nlh
, GENL_HDRLEN
, attrs
, IPVS_CMD_ATTR_MAX
,
3281 ip_vs_cmd_policy
, NULL
))
3285 svc
= ip_vs_genl_find_service(ipvs
, attrs
[IPVS_CMD_ATTR_SERVICE
]);
3286 if (IS_ERR_OR_NULL(svc
))
3289 /* Dump the destinations */
3290 list_for_each_entry(dest
, &svc
->destinations
, n_list
) {
3293 if (ip_vs_genl_dump_dest(skb
, dest
, cb
) < 0) {
3295 goto nla_put_failure
;
3303 mutex_unlock(&__ip_vs_mutex
);
3308 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern
*udest
,
3309 struct nlattr
*nla
, int full_entry
)
3311 struct nlattr
*attrs
[IPVS_DEST_ATTR_MAX
+ 1];
3312 struct nlattr
*nla_addr
, *nla_port
;
3313 struct nlattr
*nla_addr_family
;
3315 /* Parse mandatory identifying destination fields first */
3317 nla_parse_nested(attrs
, IPVS_DEST_ATTR_MAX
, nla
,
3318 ip_vs_dest_policy
, NULL
))
3321 nla_addr
= attrs
[IPVS_DEST_ATTR_ADDR
];
3322 nla_port
= attrs
[IPVS_DEST_ATTR_PORT
];
3323 nla_addr_family
= attrs
[IPVS_DEST_ATTR_ADDR_FAMILY
];
3325 if (!(nla_addr
&& nla_port
))
3328 memset(udest
, 0, sizeof(*udest
));
3330 nla_memcpy(&udest
->addr
, nla_addr
, sizeof(udest
->addr
));
3331 udest
->port
= nla_get_be16(nla_port
);
3333 if (nla_addr_family
)
3334 udest
->af
= nla_get_u16(nla_addr_family
);
3338 /* If a full entry was requested, check for the additional fields */
3340 struct nlattr
*nla_fwd
, *nla_weight
, *nla_u_thresh
,
3343 nla_fwd
= attrs
[IPVS_DEST_ATTR_FWD_METHOD
];
3344 nla_weight
= attrs
[IPVS_DEST_ATTR_WEIGHT
];
3345 nla_u_thresh
= attrs
[IPVS_DEST_ATTR_U_THRESH
];
3346 nla_l_thresh
= attrs
[IPVS_DEST_ATTR_L_THRESH
];
3348 if (!(nla_fwd
&& nla_weight
&& nla_u_thresh
&& nla_l_thresh
))
3351 udest
->conn_flags
= nla_get_u32(nla_fwd
)
3352 & IP_VS_CONN_F_FWD_MASK
;
3353 udest
->weight
= nla_get_u32(nla_weight
);
3354 udest
->u_threshold
= nla_get_u32(nla_u_thresh
);
3355 udest
->l_threshold
= nla_get_u32(nla_l_thresh
);
3361 static int ip_vs_genl_fill_daemon(struct sk_buff
*skb
, __u32 state
,
3362 struct ipvs_sync_daemon_cfg
*c
)
3364 struct nlattr
*nl_daemon
;
3366 nl_daemon
= nla_nest_start(skb
, IPVS_CMD_ATTR_DAEMON
);
3370 if (nla_put_u32(skb
, IPVS_DAEMON_ATTR_STATE
, state
) ||
3371 nla_put_string(skb
, IPVS_DAEMON_ATTR_MCAST_IFN
, c
->mcast_ifn
) ||
3372 nla_put_u32(skb
, IPVS_DAEMON_ATTR_SYNC_ID
, c
->syncid
) ||
3373 nla_put_u16(skb
, IPVS_DAEMON_ATTR_SYNC_MAXLEN
, c
->sync_maxlen
) ||
3374 nla_put_u16(skb
, IPVS_DAEMON_ATTR_MCAST_PORT
, c
->mcast_port
) ||
3375 nla_put_u8(skb
, IPVS_DAEMON_ATTR_MCAST_TTL
, c
->mcast_ttl
))
3376 goto nla_put_failure
;
3377 #ifdef CONFIG_IP_VS_IPV6
3378 if (c
->mcast_af
== AF_INET6
) {
3379 if (nla_put_in6_addr(skb
, IPVS_DAEMON_ATTR_MCAST_GROUP6
,
3380 &c
->mcast_group
.in6
))
3381 goto nla_put_failure
;
3384 if (c
->mcast_af
== AF_INET
&&
3385 nla_put_in_addr(skb
, IPVS_DAEMON_ATTR_MCAST_GROUP
,
3387 goto nla_put_failure
;
3388 nla_nest_end(skb
, nl_daemon
);
3393 nla_nest_cancel(skb
, nl_daemon
);
3397 static int ip_vs_genl_dump_daemon(struct sk_buff
*skb
, __u32 state
,
3398 struct ipvs_sync_daemon_cfg
*c
,
3399 struct netlink_callback
*cb
)
3402 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
3403 &ip_vs_genl_family
, NLM_F_MULTI
,
3404 IPVS_CMD_NEW_DAEMON
);
3408 if (ip_vs_genl_fill_daemon(skb
, state
, c
))
3409 goto nla_put_failure
;
3411 genlmsg_end(skb
, hdr
);
3415 genlmsg_cancel(skb
, hdr
);
3419 static int ip_vs_genl_dump_daemons(struct sk_buff
*skb
,
3420 struct netlink_callback
*cb
)
3422 struct net
*net
= sock_net(skb
->sk
);
3423 struct netns_ipvs
*ipvs
= net_ipvs(net
);
3425 mutex_lock(&ipvs
->sync_mutex
);
3426 if ((ipvs
->sync_state
& IP_VS_STATE_MASTER
) && !cb
->args
[0]) {
3427 if (ip_vs_genl_dump_daemon(skb
, IP_VS_STATE_MASTER
,
3428 &ipvs
->mcfg
, cb
) < 0)
3429 goto nla_put_failure
;
3434 if ((ipvs
->sync_state
& IP_VS_STATE_BACKUP
) && !cb
->args
[1]) {
3435 if (ip_vs_genl_dump_daemon(skb
, IP_VS_STATE_BACKUP
,
3436 &ipvs
->bcfg
, cb
) < 0)
3437 goto nla_put_failure
;
3443 mutex_unlock(&ipvs
->sync_mutex
);
3448 static int ip_vs_genl_new_daemon(struct netns_ipvs
*ipvs
, struct nlattr
**attrs
)
3450 struct ipvs_sync_daemon_cfg c
;
3454 memset(&c
, 0, sizeof(c
));
3455 if (!(attrs
[IPVS_DAEMON_ATTR_STATE
] &&
3456 attrs
[IPVS_DAEMON_ATTR_MCAST_IFN
] &&
3457 attrs
[IPVS_DAEMON_ATTR_SYNC_ID
]))
3459 strlcpy(c
.mcast_ifn
, nla_data(attrs
[IPVS_DAEMON_ATTR_MCAST_IFN
]),
3460 sizeof(c
.mcast_ifn
));
3461 c
.syncid
= nla_get_u32(attrs
[IPVS_DAEMON_ATTR_SYNC_ID
]);
3463 a
= attrs
[IPVS_DAEMON_ATTR_SYNC_MAXLEN
];
3465 c
.sync_maxlen
= nla_get_u16(a
);
3467 a
= attrs
[IPVS_DAEMON_ATTR_MCAST_GROUP
];
3469 c
.mcast_af
= AF_INET
;
3470 c
.mcast_group
.ip
= nla_get_in_addr(a
);
3471 if (!ipv4_is_multicast(c
.mcast_group
.ip
))
3474 a
= attrs
[IPVS_DAEMON_ATTR_MCAST_GROUP6
];
3476 #ifdef CONFIG_IP_VS_IPV6
3479 c
.mcast_af
= AF_INET6
;
3480 c
.mcast_group
.in6
= nla_get_in6_addr(a
);
3481 addr_type
= ipv6_addr_type(&c
.mcast_group
.in6
);
3482 if (!(addr_type
& IPV6_ADDR_MULTICAST
))
3485 return -EAFNOSUPPORT
;
3490 a
= attrs
[IPVS_DAEMON_ATTR_MCAST_PORT
];
3492 c
.mcast_port
= nla_get_u16(a
);
3494 a
= attrs
[IPVS_DAEMON_ATTR_MCAST_TTL
];
3496 c
.mcast_ttl
= nla_get_u8(a
);
3498 /* The synchronization protocol is incompatible with mixed family
3501 if (ipvs
->mixed_address_family_dests
> 0)
3504 ret
= start_sync_thread(ipvs
, &c
,
3505 nla_get_u32(attrs
[IPVS_DAEMON_ATTR_STATE
]));
3509 static int ip_vs_genl_del_daemon(struct netns_ipvs
*ipvs
, struct nlattr
**attrs
)
3513 if (!attrs
[IPVS_DAEMON_ATTR_STATE
])
3516 mutex_lock(&ipvs
->sync_mutex
);
3517 ret
= stop_sync_thread(ipvs
,
3518 nla_get_u32(attrs
[IPVS_DAEMON_ATTR_STATE
]));
3519 mutex_unlock(&ipvs
->sync_mutex
);
3523 static int ip_vs_genl_set_config(struct netns_ipvs
*ipvs
, struct nlattr
**attrs
)
3525 struct ip_vs_timeout_user t
;
3527 __ip_vs_get_timeouts(ipvs
, &t
);
3529 if (attrs
[IPVS_CMD_ATTR_TIMEOUT_TCP
])
3530 t
.tcp_timeout
= nla_get_u32(attrs
[IPVS_CMD_ATTR_TIMEOUT_TCP
]);
3532 if (attrs
[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN
])
3534 nla_get_u32(attrs
[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN
]);
3536 if (attrs
[IPVS_CMD_ATTR_TIMEOUT_UDP
])
3537 t
.udp_timeout
= nla_get_u32(attrs
[IPVS_CMD_ATTR_TIMEOUT_UDP
]);
3539 return ip_vs_set_timeout(ipvs
, &t
);
3542 static int ip_vs_genl_set_daemon(struct sk_buff
*skb
, struct genl_info
*info
)
3544 int ret
= -EINVAL
, cmd
;
3545 struct net
*net
= sock_net(skb
->sk
);
3546 struct netns_ipvs
*ipvs
= net_ipvs(net
);
3548 cmd
= info
->genlhdr
->cmd
;
3550 if (cmd
== IPVS_CMD_NEW_DAEMON
|| cmd
== IPVS_CMD_DEL_DAEMON
) {
3551 struct nlattr
*daemon_attrs
[IPVS_DAEMON_ATTR_MAX
+ 1];
3553 if (!info
->attrs
[IPVS_CMD_ATTR_DAEMON
] ||
3554 nla_parse_nested(daemon_attrs
, IPVS_DAEMON_ATTR_MAX
,
3555 info
->attrs
[IPVS_CMD_ATTR_DAEMON
],
3556 ip_vs_daemon_policy
, info
->extack
))
3559 if (cmd
== IPVS_CMD_NEW_DAEMON
)
3560 ret
= ip_vs_genl_new_daemon(ipvs
, daemon_attrs
);
3562 ret
= ip_vs_genl_del_daemon(ipvs
, daemon_attrs
);
3569 static int ip_vs_genl_set_cmd(struct sk_buff
*skb
, struct genl_info
*info
)
3571 struct ip_vs_service
*svc
= NULL
;
3572 struct ip_vs_service_user_kern usvc
;
3573 struct ip_vs_dest_user_kern udest
;
3575 int need_full_svc
= 0, need_full_dest
= 0;
3576 struct net
*net
= sock_net(skb
->sk
);
3577 struct netns_ipvs
*ipvs
= net_ipvs(net
);
3579 cmd
= info
->genlhdr
->cmd
;
3581 mutex_lock(&__ip_vs_mutex
);
3583 if (cmd
== IPVS_CMD_FLUSH
) {
3584 ret
= ip_vs_flush(ipvs
, false);
3586 } else if (cmd
== IPVS_CMD_SET_CONFIG
) {
3587 ret
= ip_vs_genl_set_config(ipvs
, info
->attrs
);
3589 } else if (cmd
== IPVS_CMD_ZERO
&&
3590 !info
->attrs
[IPVS_CMD_ATTR_SERVICE
]) {
3591 ret
= ip_vs_zero_all(ipvs
);
3595 /* All following commands require a service argument, so check if we
3596 * received a valid one. We need a full service specification when
3597 * adding / editing a service. Only identifying members otherwise. */
3598 if (cmd
== IPVS_CMD_NEW_SERVICE
|| cmd
== IPVS_CMD_SET_SERVICE
)
3601 ret
= ip_vs_genl_parse_service(ipvs
, &usvc
,
3602 info
->attrs
[IPVS_CMD_ATTR_SERVICE
],
3603 need_full_svc
, &svc
);
3607 /* Unless we're adding a new service, the service must already exist */
3608 if ((cmd
!= IPVS_CMD_NEW_SERVICE
) && (svc
== NULL
)) {
3613 /* Destination commands require a valid destination argument. For
3614 * adding / editing a destination, we need a full destination
3616 if (cmd
== IPVS_CMD_NEW_DEST
|| cmd
== IPVS_CMD_SET_DEST
||
3617 cmd
== IPVS_CMD_DEL_DEST
) {
3618 if (cmd
!= IPVS_CMD_DEL_DEST
)
3621 ret
= ip_vs_genl_parse_dest(&udest
,
3622 info
->attrs
[IPVS_CMD_ATTR_DEST
],
3627 /* Old protocols did not allow the user to specify address
3628 * family, so we set it to zero instead. We also didn't
3629 * allow heterogeneous pools in the old code, so it's safe
3630 * to assume that this will have the same address family as
3636 if (!ip_vs_is_af_valid(udest
.af
)) {
3637 ret
= -EAFNOSUPPORT
;
3641 if (udest
.af
!= svc
->af
&& cmd
!= IPVS_CMD_DEL_DEST
) {
3642 /* The synchronization protocol is incompatible
3643 * with mixed family services
3645 if (ipvs
->sync_state
) {
3650 /* Which connection types do we support? */
3651 switch (udest
.conn_flags
) {
3652 case IP_VS_CONN_F_TUNNEL
:
3653 /* We are able to forward this */
3663 case IPVS_CMD_NEW_SERVICE
:
3665 ret
= ip_vs_add_service(ipvs
, &usvc
, &svc
);
3669 case IPVS_CMD_SET_SERVICE
:
3670 ret
= ip_vs_edit_service(svc
, &usvc
);
3672 case IPVS_CMD_DEL_SERVICE
:
3673 ret
= ip_vs_del_service(svc
);
3674 /* do not use svc, it can be freed */
3676 case IPVS_CMD_NEW_DEST
:
3677 ret
= ip_vs_add_dest(svc
, &udest
);
3679 case IPVS_CMD_SET_DEST
:
3680 ret
= ip_vs_edit_dest(svc
, &udest
);
3682 case IPVS_CMD_DEL_DEST
:
3683 ret
= ip_vs_del_dest(svc
, &udest
);
3686 ret
= ip_vs_zero_service(svc
);
3693 mutex_unlock(&__ip_vs_mutex
);
3698 static int ip_vs_genl_get_cmd(struct sk_buff
*skb
, struct genl_info
*info
)
3700 struct sk_buff
*msg
;
3702 int ret
, cmd
, reply_cmd
;
3703 struct net
*net
= sock_net(skb
->sk
);
3704 struct netns_ipvs
*ipvs
= net_ipvs(net
);
3706 cmd
= info
->genlhdr
->cmd
;
3708 if (cmd
== IPVS_CMD_GET_SERVICE
)
3709 reply_cmd
= IPVS_CMD_NEW_SERVICE
;
3710 else if (cmd
== IPVS_CMD_GET_INFO
)
3711 reply_cmd
= IPVS_CMD_SET_INFO
;
3712 else if (cmd
== IPVS_CMD_GET_CONFIG
)
3713 reply_cmd
= IPVS_CMD_SET_CONFIG
;
3715 pr_err("unknown Generic Netlink command\n");
3719 msg
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
3723 mutex_lock(&__ip_vs_mutex
);
3725 reply
= genlmsg_put_reply(msg
, info
, &ip_vs_genl_family
, 0, reply_cmd
);
3727 goto nla_put_failure
;
3730 case IPVS_CMD_GET_SERVICE
:
3732 struct ip_vs_service
*svc
;
3734 svc
= ip_vs_genl_find_service(ipvs
,
3735 info
->attrs
[IPVS_CMD_ATTR_SERVICE
]);
3740 ret
= ip_vs_genl_fill_service(msg
, svc
);
3742 goto nla_put_failure
;
3751 case IPVS_CMD_GET_CONFIG
:
3753 struct ip_vs_timeout_user t
;
3755 __ip_vs_get_timeouts(ipvs
, &t
);
3756 #ifdef CONFIG_IP_VS_PROTO_TCP
3757 if (nla_put_u32(msg
, IPVS_CMD_ATTR_TIMEOUT_TCP
,
3759 nla_put_u32(msg
, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN
,
3761 goto nla_put_failure
;
3763 #ifdef CONFIG_IP_VS_PROTO_UDP
3764 if (nla_put_u32(msg
, IPVS_CMD_ATTR_TIMEOUT_UDP
, t
.udp_timeout
))
3765 goto nla_put_failure
;
3771 case IPVS_CMD_GET_INFO
:
3772 if (nla_put_u32(msg
, IPVS_INFO_ATTR_VERSION
,
3773 IP_VS_VERSION_CODE
) ||
3774 nla_put_u32(msg
, IPVS_INFO_ATTR_CONN_TAB_SIZE
,
3775 ip_vs_conn_tab_size
))
3776 goto nla_put_failure
;
3780 genlmsg_end(msg
, reply
);
3781 ret
= genlmsg_reply(msg
, info
);
3785 pr_err("not enough space in Netlink message\n");
3791 mutex_unlock(&__ip_vs_mutex
);
3797 static const struct genl_ops ip_vs_genl_ops
[] = {
3799 .cmd
= IPVS_CMD_NEW_SERVICE
,
3800 .flags
= GENL_ADMIN_PERM
,
3801 .policy
= ip_vs_cmd_policy
,
3802 .doit
= ip_vs_genl_set_cmd
,
3805 .cmd
= IPVS_CMD_SET_SERVICE
,
3806 .flags
= GENL_ADMIN_PERM
,
3807 .policy
= ip_vs_cmd_policy
,
3808 .doit
= ip_vs_genl_set_cmd
,
3811 .cmd
= IPVS_CMD_DEL_SERVICE
,
3812 .flags
= GENL_ADMIN_PERM
,
3813 .policy
= ip_vs_cmd_policy
,
3814 .doit
= ip_vs_genl_set_cmd
,
3817 .cmd
= IPVS_CMD_GET_SERVICE
,
3818 .flags
= GENL_ADMIN_PERM
,
3819 .doit
= ip_vs_genl_get_cmd
,
3820 .dumpit
= ip_vs_genl_dump_services
,
3821 .policy
= ip_vs_cmd_policy
,
3824 .cmd
= IPVS_CMD_NEW_DEST
,
3825 .flags
= GENL_ADMIN_PERM
,
3826 .policy
= ip_vs_cmd_policy
,
3827 .doit
= ip_vs_genl_set_cmd
,
3830 .cmd
= IPVS_CMD_SET_DEST
,
3831 .flags
= GENL_ADMIN_PERM
,
3832 .policy
= ip_vs_cmd_policy
,
3833 .doit
= ip_vs_genl_set_cmd
,
3836 .cmd
= IPVS_CMD_DEL_DEST
,
3837 .flags
= GENL_ADMIN_PERM
,
3838 .policy
= ip_vs_cmd_policy
,
3839 .doit
= ip_vs_genl_set_cmd
,
3842 .cmd
= IPVS_CMD_GET_DEST
,
3843 .flags
= GENL_ADMIN_PERM
,
3844 .policy
= ip_vs_cmd_policy
,
3845 .dumpit
= ip_vs_genl_dump_dests
,
3848 .cmd
= IPVS_CMD_NEW_DAEMON
,
3849 .flags
= GENL_ADMIN_PERM
,
3850 .policy
= ip_vs_cmd_policy
,
3851 .doit
= ip_vs_genl_set_daemon
,
3854 .cmd
= IPVS_CMD_DEL_DAEMON
,
3855 .flags
= GENL_ADMIN_PERM
,
3856 .policy
= ip_vs_cmd_policy
,
3857 .doit
= ip_vs_genl_set_daemon
,
3860 .cmd
= IPVS_CMD_GET_DAEMON
,
3861 .flags
= GENL_ADMIN_PERM
,
3862 .dumpit
= ip_vs_genl_dump_daemons
,
3865 .cmd
= IPVS_CMD_SET_CONFIG
,
3866 .flags
= GENL_ADMIN_PERM
,
3867 .policy
= ip_vs_cmd_policy
,
3868 .doit
= ip_vs_genl_set_cmd
,
3871 .cmd
= IPVS_CMD_GET_CONFIG
,
3872 .flags
= GENL_ADMIN_PERM
,
3873 .doit
= ip_vs_genl_get_cmd
,
3876 .cmd
= IPVS_CMD_GET_INFO
,
3877 .flags
= GENL_ADMIN_PERM
,
3878 .doit
= ip_vs_genl_get_cmd
,
3881 .cmd
= IPVS_CMD_ZERO
,
3882 .flags
= GENL_ADMIN_PERM
,
3883 .policy
= ip_vs_cmd_policy
,
3884 .doit
= ip_vs_genl_set_cmd
,
3887 .cmd
= IPVS_CMD_FLUSH
,
3888 .flags
= GENL_ADMIN_PERM
,
3889 .doit
= ip_vs_genl_set_cmd
,
3893 static struct genl_family ip_vs_genl_family __ro_after_init
= {
3895 .name
= IPVS_GENL_NAME
,
3896 .version
= IPVS_GENL_VERSION
,
3897 .maxattr
= IPVS_CMD_ATTR_MAX
,
3898 .netnsok
= true, /* Make ipvsadm to work on netns */
3899 .module
= THIS_MODULE
,
3900 .ops
= ip_vs_genl_ops
,
3901 .n_ops
= ARRAY_SIZE(ip_vs_genl_ops
),
3904 static int __init
ip_vs_genl_register(void)
3906 return genl_register_family(&ip_vs_genl_family
);
3909 static void ip_vs_genl_unregister(void)
3911 genl_unregister_family(&ip_vs_genl_family
);
3914 /* End of Generic Netlink interface definitions */
3917 * per netns intit/exit func.
3919 #ifdef CONFIG_SYSCTL
3920 static int __net_init
ip_vs_control_net_init_sysctl(struct netns_ipvs
*ipvs
)
3922 struct net
*net
= ipvs
->net
;
3924 struct ctl_table
*tbl
;
3926 atomic_set(&ipvs
->dropentry
, 0);
3927 spin_lock_init(&ipvs
->dropentry_lock
);
3928 spin_lock_init(&ipvs
->droppacket_lock
);
3929 spin_lock_init(&ipvs
->securetcp_lock
);
3931 if (!net_eq(net
, &init_net
)) {
3932 tbl
= kmemdup(vs_vars
, sizeof(vs_vars
), GFP_KERNEL
);
3936 /* Don't export sysctls to unprivileged users */
3937 if (net
->user_ns
!= &init_user_ns
)
3938 tbl
[0].procname
= NULL
;
3941 /* Initialize sysctl defaults */
3942 for (idx
= 0; idx
< ARRAY_SIZE(vs_vars
); idx
++) {
3943 if (tbl
[idx
].proc_handler
== proc_do_defense_mode
)
3944 tbl
[idx
].extra2
= ipvs
;
3947 ipvs
->sysctl_amemthresh
= 1024;
3948 tbl
[idx
++].data
= &ipvs
->sysctl_amemthresh
;
3949 ipvs
->sysctl_am_droprate
= 10;
3950 tbl
[idx
++].data
= &ipvs
->sysctl_am_droprate
;
3951 tbl
[idx
++].data
= &ipvs
->sysctl_drop_entry
;
3952 tbl
[idx
++].data
= &ipvs
->sysctl_drop_packet
;
3953 #ifdef CONFIG_IP_VS_NFCT
3954 tbl
[idx
++].data
= &ipvs
->sysctl_conntrack
;
3956 tbl
[idx
++].data
= &ipvs
->sysctl_secure_tcp
;
3957 ipvs
->sysctl_snat_reroute
= 1;
3958 tbl
[idx
++].data
= &ipvs
->sysctl_snat_reroute
;
3959 ipvs
->sysctl_sync_ver
= 1;
3960 tbl
[idx
++].data
= &ipvs
->sysctl_sync_ver
;
3961 ipvs
->sysctl_sync_ports
= 1;
3962 tbl
[idx
++].data
= &ipvs
->sysctl_sync_ports
;
3963 tbl
[idx
++].data
= &ipvs
->sysctl_sync_persist_mode
;
3964 ipvs
->sysctl_sync_qlen_max
= nr_free_buffer_pages() / 32;
3965 tbl
[idx
++].data
= &ipvs
->sysctl_sync_qlen_max
;
3966 ipvs
->sysctl_sync_sock_size
= 0;
3967 tbl
[idx
++].data
= &ipvs
->sysctl_sync_sock_size
;
3968 tbl
[idx
++].data
= &ipvs
->sysctl_cache_bypass
;
3969 tbl
[idx
++].data
= &ipvs
->sysctl_expire_nodest_conn
;
3970 tbl
[idx
++].data
= &ipvs
->sysctl_sloppy_tcp
;
3971 tbl
[idx
++].data
= &ipvs
->sysctl_sloppy_sctp
;
3972 tbl
[idx
++].data
= &ipvs
->sysctl_expire_quiescent_template
;
3973 ipvs
->sysctl_sync_threshold
[0] = DEFAULT_SYNC_THRESHOLD
;
3974 ipvs
->sysctl_sync_threshold
[1] = DEFAULT_SYNC_PERIOD
;
3975 tbl
[idx
].data
= &ipvs
->sysctl_sync_threshold
;
3976 tbl
[idx
++].maxlen
= sizeof(ipvs
->sysctl_sync_threshold
);
3977 ipvs
->sysctl_sync_refresh_period
= DEFAULT_SYNC_REFRESH_PERIOD
;
3978 tbl
[idx
++].data
= &ipvs
->sysctl_sync_refresh_period
;
3979 ipvs
->sysctl_sync_retries
= clamp_t(int, DEFAULT_SYNC_RETRIES
, 0, 3);
3980 tbl
[idx
++].data
= &ipvs
->sysctl_sync_retries
;
3981 tbl
[idx
++].data
= &ipvs
->sysctl_nat_icmp_send
;
3982 ipvs
->sysctl_pmtu_disc
= 1;
3983 tbl
[idx
++].data
= &ipvs
->sysctl_pmtu_disc
;
3984 tbl
[idx
++].data
= &ipvs
->sysctl_backup_only
;
3985 ipvs
->sysctl_conn_reuse_mode
= 1;
3986 tbl
[idx
++].data
= &ipvs
->sysctl_conn_reuse_mode
;
3987 tbl
[idx
++].data
= &ipvs
->sysctl_schedule_icmp
;
3988 tbl
[idx
++].data
= &ipvs
->sysctl_ignore_tunneled
;
3990 ipvs
->sysctl_hdr
= register_net_sysctl(net
, "net/ipv4/vs", tbl
);
3991 if (ipvs
->sysctl_hdr
== NULL
) {
3992 if (!net_eq(net
, &init_net
))
3996 ip_vs_start_estimator(ipvs
, &ipvs
->tot_stats
);
3997 ipvs
->sysctl_tbl
= tbl
;
3998 /* Schedule defense work */
3999 INIT_DELAYED_WORK(&ipvs
->defense_work
, defense_work_handler
);
4000 schedule_delayed_work(&ipvs
->defense_work
, DEFENSE_TIMER_PERIOD
);
4005 static void __net_exit
ip_vs_control_net_cleanup_sysctl(struct netns_ipvs
*ipvs
)
4007 struct net
*net
= ipvs
->net
;
4009 cancel_delayed_work_sync(&ipvs
->defense_work
);
4010 cancel_work_sync(&ipvs
->defense_work
.work
);
4011 unregister_net_sysctl_table(ipvs
->sysctl_hdr
);
4012 ip_vs_stop_estimator(ipvs
, &ipvs
->tot_stats
);
4014 if (!net_eq(net
, &init_net
))
4015 kfree(ipvs
->sysctl_tbl
);
4020 static int __net_init
ip_vs_control_net_init_sysctl(struct netns_ipvs
*ipvs
) { return 0; }
4021 static void __net_exit
ip_vs_control_net_cleanup_sysctl(struct netns_ipvs
*ipvs
) { }
4025 static struct notifier_block ip_vs_dst_notifier
= {
4026 .notifier_call
= ip_vs_dst_event
,
4027 #ifdef CONFIG_IP_VS_IPV6
4028 .priority
= ADDRCONF_NOTIFY_PRIORITY
+ 5,
4032 int __net_init
ip_vs_control_net_init(struct netns_ipvs
*ipvs
)
4036 /* Initialize rs_table */
4037 for (idx
= 0; idx
< IP_VS_RTAB_SIZE
; idx
++)
4038 INIT_HLIST_HEAD(&ipvs
->rs_table
[idx
]);
4040 INIT_LIST_HEAD(&ipvs
->dest_trash
);
4041 spin_lock_init(&ipvs
->dest_trash_lock
);
4042 setup_timer(&ipvs
->dest_trash_timer
, ip_vs_dest_trash_expire
,
4043 (unsigned long) ipvs
);
4044 atomic_set(&ipvs
->ftpsvc_counter
, 0);
4045 atomic_set(&ipvs
->nullsvc_counter
, 0);
4046 atomic_set(&ipvs
->conn_out_counter
, 0);
4049 ipvs
->tot_stats
.cpustats
= alloc_percpu(struct ip_vs_cpu_stats
);
4050 if (!ipvs
->tot_stats
.cpustats
)
4053 for_each_possible_cpu(i
) {
4054 struct ip_vs_cpu_stats
*ipvs_tot_stats
;
4055 ipvs_tot_stats
= per_cpu_ptr(ipvs
->tot_stats
.cpustats
, i
);
4056 u64_stats_init(&ipvs_tot_stats
->syncp
);
4059 spin_lock_init(&ipvs
->tot_stats
.lock
);
4061 proc_create("ip_vs", 0, ipvs
->net
->proc_net
, &ip_vs_info_fops
);
4062 proc_create("ip_vs_stats", 0, ipvs
->net
->proc_net
, &ip_vs_stats_fops
);
4063 proc_create("ip_vs_stats_percpu", 0, ipvs
->net
->proc_net
,
4064 &ip_vs_stats_percpu_fops
);
4066 if (ip_vs_control_net_init_sysctl(ipvs
))
4072 free_percpu(ipvs
->tot_stats
.cpustats
);
4076 void __net_exit
ip_vs_control_net_cleanup(struct netns_ipvs
*ipvs
)
4078 ip_vs_trash_cleanup(ipvs
);
4079 ip_vs_control_net_cleanup_sysctl(ipvs
);
4080 remove_proc_entry("ip_vs_stats_percpu", ipvs
->net
->proc_net
);
4081 remove_proc_entry("ip_vs_stats", ipvs
->net
->proc_net
);
4082 remove_proc_entry("ip_vs", ipvs
->net
->proc_net
);
4083 free_percpu(ipvs
->tot_stats
.cpustats
);
4086 int __init
ip_vs_register_nl_ioctl(void)
4090 ret
= nf_register_sockopt(&ip_vs_sockopts
);
4092 pr_err("cannot register sockopt.\n");
4096 ret
= ip_vs_genl_register();
4098 pr_err("cannot register Generic Netlink interface.\n");
4104 nf_unregister_sockopt(&ip_vs_sockopts
);
4109 void ip_vs_unregister_nl_ioctl(void)
4111 ip_vs_genl_unregister();
4112 nf_unregister_sockopt(&ip_vs_sockopts
);
4115 int __init
ip_vs_control_init(void)
4122 /* Initialize svc_table, ip_vs_svc_fwm_table */
4123 for (idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
4124 INIT_HLIST_HEAD(&ip_vs_svc_table
[idx
]);
4125 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table
[idx
]);
4128 smp_wmb(); /* Do we really need it now ? */
4130 ret
= register_netdevice_notifier(&ip_vs_dst_notifier
);
4139 void ip_vs_control_cleanup(void)
4142 unregister_netdevice_notifier(&ip_vs_dst_notifier
);