2 * IPVS An implementation of the IP virtual server support for the
3 * LINUX operating system. IPVS is now implemented as a module
4 * over the NetFilter framework. IPVS can be used to build a
5 * high-performance and highly available server based on a
8 * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
9 * Peter Kese <peter.kese@ijs.si>
10 * Julian Anastasov <ja@ssi.bg>
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
21 #define KMSG_COMPONENT "IPVS"
22 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
24 #include <linux/module.h>
25 #include <linux/init.h>
26 #include <linux/types.h>
27 #include <linux/capability.h>
29 #include <linux/sysctl.h>
30 #include <linux/proc_fs.h>
31 #include <linux/workqueue.h>
32 #include <linux/swap.h>
33 #include <linux/seq_file.h>
34 #include <linux/slab.h>
36 #include <linux/netfilter.h>
37 #include <linux/netfilter_ipv4.h>
38 #include <linux/mutex.h>
40 #include <net/net_namespace.h>
41 #include <linux/nsproxy.h>
43 #ifdef CONFIG_IP_VS_IPV6
45 #include <net/ip6_route.h>
47 #include <net/route.h>
49 #include <net/genetlink.h>
51 #include <asm/uaccess.h>
53 #include <net/ip_vs.h>
55 /* semaphore for IPVS sockopts. And, [gs]etsockopt may sleep. */
56 static DEFINE_MUTEX(__ip_vs_mutex
);
58 /* 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
->net
);
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
);
430 && protocol
== IPPROTO_TCP
431 && atomic_read(&ipvs
->ftpsvc_counter
)
432 && (vport
== FTPDATA
|| ntohs(vport
) >= PROT_SOCK
)) {
434 * Check if ftp service entry exists, the packet
435 * might belong to FTP data connections.
437 svc
= __ip_vs_service_find(ipvs
, af
, protocol
, vaddr
, FTPPORT
);
441 && atomic_read(&ipvs
->nullsvc_counter
)) {
443 * Check if the catch-all port (port zero) exists
445 svc
= __ip_vs_service_find(ipvs
, af
, protocol
, vaddr
, 0);
449 IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
450 fwmark
, ip_vs_proto_name(protocol
),
451 IP_VS_DBG_ADDR(af
, vaddr
), ntohs(vport
),
452 svc
? "hit" : "not hit");
459 __ip_vs_bind_svc(struct ip_vs_dest
*dest
, struct ip_vs_service
*svc
)
461 atomic_inc(&svc
->refcnt
);
462 rcu_assign_pointer(dest
->svc
, svc
);
465 static void ip_vs_service_free(struct ip_vs_service
*svc
)
467 free_percpu(svc
->stats
.cpustats
);
471 static void ip_vs_service_rcu_free(struct rcu_head
*head
)
473 struct ip_vs_service
*svc
;
475 svc
= container_of(head
, struct ip_vs_service
, rcu_head
);
476 ip_vs_service_free(svc
);
479 static void __ip_vs_svc_put(struct ip_vs_service
*svc
, bool do_delay
)
481 if (atomic_dec_and_test(&svc
->refcnt
)) {
482 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n",
484 IP_VS_DBG_ADDR(svc
->af
, &svc
->addr
),
487 call_rcu(&svc
->rcu_head
, ip_vs_service_rcu_free
);
489 ip_vs_service_free(svc
);
495 * Returns hash value for real service
497 static inline unsigned int ip_vs_rs_hashkey(int af
,
498 const union nf_inet_addr
*addr
,
501 register unsigned int porth
= ntohs(port
);
502 __be32 addr_fold
= addr
->ip
;
504 #ifdef CONFIG_IP_VS_IPV6
506 addr_fold
= addr
->ip6
[0]^addr
->ip6
[1]^
507 addr
->ip6
[2]^addr
->ip6
[3];
510 return (ntohl(addr_fold
)^(porth
>>IP_VS_RTAB_BITS
)^porth
)
514 /* Hash ip_vs_dest in rs_table by <proto,addr,port>. */
515 static void ip_vs_rs_hash(struct netns_ipvs
*ipvs
, struct ip_vs_dest
*dest
)
519 if (dest
->in_rs_table
)
523 * Hash by proto,addr,port,
524 * which are the parameters of the real service.
526 hash
= ip_vs_rs_hashkey(dest
->af
, &dest
->addr
, dest
->port
);
528 hlist_add_head_rcu(&dest
->d_list
, &ipvs
->rs_table
[hash
]);
529 dest
->in_rs_table
= 1;
532 /* Unhash ip_vs_dest from rs_table. */
533 static void ip_vs_rs_unhash(struct ip_vs_dest
*dest
)
536 * Remove it from the rs_table table.
538 if (dest
->in_rs_table
) {
539 hlist_del_rcu(&dest
->d_list
);
540 dest
->in_rs_table
= 0;
544 /* Check if real service by <proto,addr,port> is present */
545 bool ip_vs_has_real_service(struct netns_ipvs
*ipvs
, int af
, __u16 protocol
,
546 const union nf_inet_addr
*daddr
, __be16 dport
)
549 struct ip_vs_dest
*dest
;
551 /* Check for "full" addressed entries */
552 hash
= ip_vs_rs_hashkey(af
, daddr
, dport
);
555 hlist_for_each_entry_rcu(dest
, &ipvs
->rs_table
[hash
], d_list
) {
556 if (dest
->port
== dport
&&
558 ip_vs_addr_equal(af
, &dest
->addr
, daddr
) &&
559 (dest
->protocol
== protocol
|| dest
->vfwmark
)) {
570 /* Lookup destination by {addr,port} in the given service
571 * Called under RCU lock.
573 static struct ip_vs_dest
*
574 ip_vs_lookup_dest(struct ip_vs_service
*svc
, int dest_af
,
575 const union nf_inet_addr
*daddr
, __be16 dport
)
577 struct ip_vs_dest
*dest
;
580 * Find the destination for the given service
582 list_for_each_entry_rcu(dest
, &svc
->destinations
, n_list
) {
583 if ((dest
->af
== dest_af
) &&
584 ip_vs_addr_equal(dest_af
, &dest
->addr
, daddr
) &&
585 (dest
->port
== dport
)) {
595 * Find destination by {daddr,dport,vaddr,protocol}
596 * Created to be used in ip_vs_process_message() in
597 * the backup synchronization daemon. It finds the
598 * destination to be bound to the received connection
600 * Called under RCU lock, no refcnt is returned.
602 struct ip_vs_dest
*ip_vs_find_dest(struct netns_ipvs
*ipvs
, int svc_af
, int dest_af
,
603 const union nf_inet_addr
*daddr
,
605 const union nf_inet_addr
*vaddr
,
606 __be16 vport
, __u16 protocol
, __u32 fwmark
,
609 struct ip_vs_dest
*dest
;
610 struct ip_vs_service
*svc
;
613 svc
= ip_vs_service_find(ipvs
, svc_af
, fwmark
, protocol
, vaddr
, vport
);
616 if (fwmark
&& (flags
& IP_VS_CONN_F_FWD_MASK
) != IP_VS_CONN_F_MASQ
)
618 dest
= ip_vs_lookup_dest(svc
, dest_af
, daddr
, port
);
620 dest
= ip_vs_lookup_dest(svc
, dest_af
, daddr
, port
^ dport
);
624 void ip_vs_dest_dst_rcu_free(struct rcu_head
*head
)
626 struct ip_vs_dest_dst
*dest_dst
= container_of(head
,
627 struct ip_vs_dest_dst
,
630 dst_release(dest_dst
->dst_cache
);
634 /* Release dest_dst and dst_cache for dest in user context */
635 static void __ip_vs_dst_cache_reset(struct ip_vs_dest
*dest
)
637 struct ip_vs_dest_dst
*old
;
639 old
= rcu_dereference_protected(dest
->dest_dst
, 1);
641 RCU_INIT_POINTER(dest
->dest_dst
, NULL
);
642 call_rcu(&old
->rcu_head
, ip_vs_dest_dst_rcu_free
);
647 * Lookup dest by {svc,addr,port} in the destination trash.
648 * The destination trash is used to hold the destinations that are removed
649 * from the service table but are still referenced by some conn entries.
650 * The reason to add the destination trash is when the dest is temporary
651 * down (either by administrator or by monitor program), the dest can be
652 * picked back from the trash, the remaining connections to the dest can
653 * continue, and the counting information of the dest is also useful for
656 static struct ip_vs_dest
*
657 ip_vs_trash_get_dest(struct ip_vs_service
*svc
, int dest_af
,
658 const union nf_inet_addr
*daddr
, __be16 dport
)
660 struct ip_vs_dest
*dest
;
661 struct netns_ipvs
*ipvs
= svc
->ipvs
;
664 * Find the destination in trash
666 spin_lock_bh(&ipvs
->dest_trash_lock
);
667 list_for_each_entry(dest
, &ipvs
->dest_trash
, t_list
) {
668 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
671 IP_VS_DBG_ADDR(dest
->af
, &dest
->addr
),
673 atomic_read(&dest
->refcnt
));
674 if (dest
->af
== dest_af
&&
675 ip_vs_addr_equal(dest_af
, &dest
->addr
, daddr
) &&
676 dest
->port
== dport
&&
677 dest
->vfwmark
== svc
->fwmark
&&
678 dest
->protocol
== svc
->protocol
&&
680 (ip_vs_addr_equal(svc
->af
, &dest
->vaddr
, &svc
->addr
) &&
681 dest
->vport
== svc
->port
))) {
683 list_del(&dest
->t_list
);
684 ip_vs_dest_hold(dest
);
692 spin_unlock_bh(&ipvs
->dest_trash_lock
);
697 static void ip_vs_dest_free(struct ip_vs_dest
*dest
)
699 struct ip_vs_service
*svc
= rcu_dereference_protected(dest
->svc
, 1);
701 __ip_vs_dst_cache_reset(dest
);
702 __ip_vs_svc_put(svc
, false);
703 free_percpu(dest
->stats
.cpustats
);
704 ip_vs_dest_put_and_free(dest
);
708 * Clean up all the destinations in the trash
709 * Called by the ip_vs_control_cleanup()
711 * When the ip_vs_control_clearup is activated by ipvs module exit,
712 * the service tables must have been flushed and all the connections
713 * are expired, and the refcnt of each destination in the trash must
714 * be 0, so we simply release them here.
716 static void ip_vs_trash_cleanup(struct netns_ipvs
*ipvs
)
718 struct ip_vs_dest
*dest
, *nxt
;
720 del_timer_sync(&ipvs
->dest_trash_timer
);
721 /* No need to use dest_trash_lock */
722 list_for_each_entry_safe(dest
, nxt
, &ipvs
->dest_trash
, t_list
) {
723 list_del(&dest
->t_list
);
724 ip_vs_dest_free(dest
);
729 ip_vs_copy_stats(struct ip_vs_kstats
*dst
, struct ip_vs_stats
*src
)
731 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->kstats.c - src->kstats0.c
733 spin_lock_bh(&src
->lock
);
735 IP_VS_SHOW_STATS_COUNTER(conns
);
736 IP_VS_SHOW_STATS_COUNTER(inpkts
);
737 IP_VS_SHOW_STATS_COUNTER(outpkts
);
738 IP_VS_SHOW_STATS_COUNTER(inbytes
);
739 IP_VS_SHOW_STATS_COUNTER(outbytes
);
741 ip_vs_read_estimator(dst
, src
);
743 spin_unlock_bh(&src
->lock
);
747 ip_vs_export_stats_user(struct ip_vs_stats_user
*dst
, struct ip_vs_kstats
*src
)
749 dst
->conns
= (u32
)src
->conns
;
750 dst
->inpkts
= (u32
)src
->inpkts
;
751 dst
->outpkts
= (u32
)src
->outpkts
;
752 dst
->inbytes
= src
->inbytes
;
753 dst
->outbytes
= src
->outbytes
;
754 dst
->cps
= (u32
)src
->cps
;
755 dst
->inpps
= (u32
)src
->inpps
;
756 dst
->outpps
= (u32
)src
->outpps
;
757 dst
->inbps
= (u32
)src
->inbps
;
758 dst
->outbps
= (u32
)src
->outbps
;
762 ip_vs_zero_stats(struct ip_vs_stats
*stats
)
764 spin_lock_bh(&stats
->lock
);
766 /* get current counters as zero point, rates are zeroed */
768 #define IP_VS_ZERO_STATS_COUNTER(c) stats->kstats0.c = stats->kstats.c
770 IP_VS_ZERO_STATS_COUNTER(conns
);
771 IP_VS_ZERO_STATS_COUNTER(inpkts
);
772 IP_VS_ZERO_STATS_COUNTER(outpkts
);
773 IP_VS_ZERO_STATS_COUNTER(inbytes
);
774 IP_VS_ZERO_STATS_COUNTER(outbytes
);
776 ip_vs_zero_estimator(stats
);
778 spin_unlock_bh(&stats
->lock
);
782 * Update a destination in the given service
785 __ip_vs_update_dest(struct ip_vs_service
*svc
, struct ip_vs_dest
*dest
,
786 struct ip_vs_dest_user_kern
*udest
, int add
)
788 struct netns_ipvs
*ipvs
= svc
->ipvs
;
789 struct ip_vs_service
*old_svc
;
790 struct ip_vs_scheduler
*sched
;
793 /* We cannot modify an address and change the address family */
794 BUG_ON(!add
&& udest
->af
!= dest
->af
);
796 if (add
&& udest
->af
!= svc
->af
)
797 ipvs
->mixed_address_family_dests
++;
799 /* set the weight and the flags */
800 atomic_set(&dest
->weight
, udest
->weight
);
801 conn_flags
= udest
->conn_flags
& IP_VS_CONN_F_DEST_MASK
;
802 conn_flags
|= IP_VS_CONN_F_INACTIVE
;
804 /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
805 if ((conn_flags
& IP_VS_CONN_F_FWD_MASK
) != IP_VS_CONN_F_MASQ
) {
806 conn_flags
|= IP_VS_CONN_F_NOOUTPUT
;
809 * Put the real service in rs_table if not present.
810 * For now only for NAT!
812 ip_vs_rs_hash(ipvs
, dest
);
814 atomic_set(&dest
->conn_flags
, conn_flags
);
816 /* bind the service */
817 old_svc
= rcu_dereference_protected(dest
->svc
, 1);
819 __ip_vs_bind_svc(dest
, svc
);
821 if (old_svc
!= svc
) {
822 ip_vs_zero_stats(&dest
->stats
);
823 __ip_vs_bind_svc(dest
, svc
);
824 __ip_vs_svc_put(old_svc
, true);
828 /* set the dest status flags */
829 dest
->flags
|= IP_VS_DEST_F_AVAILABLE
;
831 if (udest
->u_threshold
== 0 || udest
->u_threshold
> dest
->u_threshold
)
832 dest
->flags
&= ~IP_VS_DEST_F_OVERLOAD
;
833 dest
->u_threshold
= udest
->u_threshold
;
834 dest
->l_threshold
= udest
->l_threshold
;
836 dest
->af
= udest
->af
;
838 spin_lock_bh(&dest
->dst_lock
);
839 __ip_vs_dst_cache_reset(dest
);
840 spin_unlock_bh(&dest
->dst_lock
);
843 ip_vs_start_estimator(svc
->ipvs
->net
, &dest
->stats
);
844 list_add_rcu(&dest
->n_list
, &svc
->destinations
);
846 sched
= rcu_dereference_protected(svc
->scheduler
, 1);
847 if (sched
&& sched
->add_dest
)
848 sched
->add_dest(svc
, dest
);
850 sched
= rcu_dereference_protected(svc
->scheduler
, 1);
851 if (sched
&& sched
->upd_dest
)
852 sched
->upd_dest(svc
, dest
);
858 * Create a destination for the given service
861 ip_vs_new_dest(struct ip_vs_service
*svc
, struct ip_vs_dest_user_kern
*udest
,
862 struct ip_vs_dest
**dest_p
)
864 struct ip_vs_dest
*dest
;
865 unsigned int atype
, i
;
869 #ifdef CONFIG_IP_VS_IPV6
870 if (udest
->af
== AF_INET6
) {
871 atype
= ipv6_addr_type(&udest
->addr
.in6
);
872 if ((!(atype
& IPV6_ADDR_UNICAST
) ||
873 atype
& IPV6_ADDR_LINKLOCAL
) &&
874 !__ip_vs_addr_is_local_v6(svc
->ipvs
->net
, &udest
->addr
.in6
))
879 atype
= inet_addr_type(svc
->ipvs
->net
, udest
->addr
.ip
);
880 if (atype
!= RTN_LOCAL
&& atype
!= RTN_UNICAST
)
884 dest
= kzalloc(sizeof(struct ip_vs_dest
), GFP_KERNEL
);
888 dest
->stats
.cpustats
= alloc_percpu(struct ip_vs_cpu_stats
);
889 if (!dest
->stats
.cpustats
)
892 for_each_possible_cpu(i
) {
893 struct ip_vs_cpu_stats
*ip_vs_dest_stats
;
894 ip_vs_dest_stats
= per_cpu_ptr(dest
->stats
.cpustats
, i
);
895 u64_stats_init(&ip_vs_dest_stats
->syncp
);
898 dest
->af
= udest
->af
;
899 dest
->protocol
= svc
->protocol
;
900 dest
->vaddr
= svc
->addr
;
901 dest
->vport
= svc
->port
;
902 dest
->vfwmark
= svc
->fwmark
;
903 ip_vs_addr_copy(udest
->af
, &dest
->addr
, &udest
->addr
);
904 dest
->port
= udest
->port
;
906 atomic_set(&dest
->activeconns
, 0);
907 atomic_set(&dest
->inactconns
, 0);
908 atomic_set(&dest
->persistconns
, 0);
909 atomic_set(&dest
->refcnt
, 1);
911 INIT_HLIST_NODE(&dest
->d_list
);
912 spin_lock_init(&dest
->dst_lock
);
913 spin_lock_init(&dest
->stats
.lock
);
914 __ip_vs_update_dest(svc
, dest
, udest
, 1);
928 * Add a destination into an existing service
931 ip_vs_add_dest(struct ip_vs_service
*svc
, struct ip_vs_dest_user_kern
*udest
)
933 struct ip_vs_dest
*dest
;
934 union nf_inet_addr daddr
;
935 __be16 dport
= udest
->port
;
940 if (udest
->weight
< 0) {
941 pr_err("%s(): server weight less than zero\n", __func__
);
945 if (udest
->l_threshold
> udest
->u_threshold
) {
946 pr_err("%s(): lower threshold is higher than upper threshold\n",
951 ip_vs_addr_copy(udest
->af
, &daddr
, &udest
->addr
);
953 /* We use function that requires RCU lock */
955 dest
= ip_vs_lookup_dest(svc
, udest
->af
, &daddr
, dport
);
959 IP_VS_DBG(1, "%s(): dest already exists\n", __func__
);
964 * Check if the dest already exists in the trash and
965 * is from the same service
967 dest
= ip_vs_trash_get_dest(svc
, udest
->af
, &daddr
, dport
);
970 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
971 "dest->refcnt=%d, service %u/%s:%u\n",
972 IP_VS_DBG_ADDR(udest
->af
, &daddr
), ntohs(dport
),
973 atomic_read(&dest
->refcnt
),
975 IP_VS_DBG_ADDR(svc
->af
, &dest
->vaddr
),
978 __ip_vs_update_dest(svc
, dest
, udest
, 1);
982 * Allocate and initialize the dest structure
984 ret
= ip_vs_new_dest(svc
, udest
, &dest
);
993 * Edit a destination in the given service
996 ip_vs_edit_dest(struct ip_vs_service
*svc
, struct ip_vs_dest_user_kern
*udest
)
998 struct ip_vs_dest
*dest
;
999 union nf_inet_addr daddr
;
1000 __be16 dport
= udest
->port
;
1004 if (udest
->weight
< 0) {
1005 pr_err("%s(): server weight less than zero\n", __func__
);
1009 if (udest
->l_threshold
> udest
->u_threshold
) {
1010 pr_err("%s(): lower threshold is higher than upper threshold\n",
1015 ip_vs_addr_copy(udest
->af
, &daddr
, &udest
->addr
);
1017 /* We use function that requires RCU lock */
1019 dest
= ip_vs_lookup_dest(svc
, udest
->af
, &daddr
, dport
);
1023 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__
);
1027 __ip_vs_update_dest(svc
, dest
, udest
, 0);
1034 * Delete a destination (must be already unlinked from the service)
1036 static void __ip_vs_del_dest(struct netns_ipvs
*ipvs
, struct ip_vs_dest
*dest
,
1039 ip_vs_stop_estimator(ipvs
->net
, &dest
->stats
);
1042 * Remove it from the d-linked list with the real services.
1044 ip_vs_rs_unhash(dest
);
1046 spin_lock_bh(&ipvs
->dest_trash_lock
);
1047 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1048 IP_VS_DBG_ADDR(dest
->af
, &dest
->addr
), ntohs(dest
->port
),
1049 atomic_read(&dest
->refcnt
));
1050 if (list_empty(&ipvs
->dest_trash
) && !cleanup
)
1051 mod_timer(&ipvs
->dest_trash_timer
,
1052 jiffies
+ (IP_VS_DEST_TRASH_PERIOD
>> 1));
1053 /* dest lives in trash without reference */
1054 list_add(&dest
->t_list
, &ipvs
->dest_trash
);
1055 dest
->idle_start
= 0;
1056 spin_unlock_bh(&ipvs
->dest_trash_lock
);
1057 ip_vs_dest_put(dest
);
1062 * Unlink a destination from the given service
1064 static void __ip_vs_unlink_dest(struct ip_vs_service
*svc
,
1065 struct ip_vs_dest
*dest
,
1068 dest
->flags
&= ~IP_VS_DEST_F_AVAILABLE
;
1071 * Remove it from the d-linked destination list.
1073 list_del_rcu(&dest
->n_list
);
1076 if (dest
->af
!= svc
->af
)
1077 svc
->ipvs
->mixed_address_family_dests
--;
1080 struct ip_vs_scheduler
*sched
;
1082 sched
= rcu_dereference_protected(svc
->scheduler
, 1);
1083 if (sched
&& sched
->del_dest
)
1084 sched
->del_dest(svc
, dest
);
1090 * Delete a destination server in the given service
1093 ip_vs_del_dest(struct ip_vs_service
*svc
, struct ip_vs_dest_user_kern
*udest
)
1095 struct ip_vs_dest
*dest
;
1096 __be16 dport
= udest
->port
;
1100 /* We use function that requires RCU lock */
1102 dest
= ip_vs_lookup_dest(svc
, udest
->af
, &udest
->addr
, dport
);
1106 IP_VS_DBG(1, "%s(): destination not found!\n", __func__
);
1111 * Unlink dest from the service
1113 __ip_vs_unlink_dest(svc
, dest
, 1);
1116 * Delete the destination
1118 __ip_vs_del_dest(svc
->ipvs
, dest
, false);
1125 static void ip_vs_dest_trash_expire(unsigned long data
)
1127 struct netns_ipvs
*ipvs
= (struct netns_ipvs
*)data
;
1128 struct ip_vs_dest
*dest
, *next
;
1129 unsigned long now
= jiffies
;
1131 spin_lock(&ipvs
->dest_trash_lock
);
1132 list_for_each_entry_safe(dest
, next
, &ipvs
->dest_trash
, t_list
) {
1133 if (atomic_read(&dest
->refcnt
) > 0)
1135 if (dest
->idle_start
) {
1136 if (time_before(now
, dest
->idle_start
+
1137 IP_VS_DEST_TRASH_PERIOD
))
1140 dest
->idle_start
= max(1UL, now
);
1143 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1145 IP_VS_DBG_ADDR(dest
->af
, &dest
->addr
),
1147 list_del(&dest
->t_list
);
1148 ip_vs_dest_free(dest
);
1150 if (!list_empty(&ipvs
->dest_trash
))
1151 mod_timer(&ipvs
->dest_trash_timer
,
1152 jiffies
+ (IP_VS_DEST_TRASH_PERIOD
>> 1));
1153 spin_unlock(&ipvs
->dest_trash_lock
);
1157 * Add a service into the service hash table
1160 ip_vs_add_service(struct net
*net
, struct ip_vs_service_user_kern
*u
,
1161 struct ip_vs_service
**svc_p
)
1164 struct ip_vs_scheduler
*sched
= NULL
;
1165 struct ip_vs_pe
*pe
= NULL
;
1166 struct ip_vs_service
*svc
= NULL
;
1167 struct netns_ipvs
*ipvs
= net_ipvs(net
);
1169 /* increase the module use count */
1170 ip_vs_use_count_inc();
1172 /* Lookup the scheduler by 'u->sched_name' */
1173 if (strcmp(u
->sched_name
, "none")) {
1174 sched
= ip_vs_scheduler_get(u
->sched_name
);
1176 pr_info("Scheduler module ip_vs_%s not found\n",
1183 if (u
->pe_name
&& *u
->pe_name
) {
1184 pe
= ip_vs_pe_getbyname(u
->pe_name
);
1186 pr_info("persistence engine module ip_vs_pe_%s "
1187 "not found\n", u
->pe_name
);
1193 #ifdef CONFIG_IP_VS_IPV6
1194 if (u
->af
== AF_INET6
) {
1195 __u32 plen
= (__force __u32
) u
->netmask
;
1197 if (plen
< 1 || plen
> 128) {
1204 svc
= kzalloc(sizeof(struct ip_vs_service
), GFP_KERNEL
);
1206 IP_VS_DBG(1, "%s(): no memory\n", __func__
);
1210 svc
->stats
.cpustats
= alloc_percpu(struct ip_vs_cpu_stats
);
1211 if (!svc
->stats
.cpustats
) {
1216 for_each_possible_cpu(i
) {
1217 struct ip_vs_cpu_stats
*ip_vs_stats
;
1218 ip_vs_stats
= per_cpu_ptr(svc
->stats
.cpustats
, i
);
1219 u64_stats_init(&ip_vs_stats
->syncp
);
1223 /* I'm the first user of the service */
1224 atomic_set(&svc
->refcnt
, 0);
1227 svc
->protocol
= u
->protocol
;
1228 ip_vs_addr_copy(svc
->af
, &svc
->addr
, &u
->addr
);
1229 svc
->port
= u
->port
;
1230 svc
->fwmark
= u
->fwmark
;
1231 svc
->flags
= u
->flags
;
1232 svc
->timeout
= u
->timeout
* HZ
;
1233 svc
->netmask
= u
->netmask
;
1236 INIT_LIST_HEAD(&svc
->destinations
);
1237 spin_lock_init(&svc
->sched_lock
);
1238 spin_lock_init(&svc
->stats
.lock
);
1240 /* Bind the scheduler */
1242 ret
= ip_vs_bind_scheduler(svc
, sched
);
1248 /* Bind the ct retriever */
1249 RCU_INIT_POINTER(svc
->pe
, pe
);
1252 /* Update the virtual service counters */
1253 if (svc
->port
== FTPPORT
)
1254 atomic_inc(&ipvs
->ftpsvc_counter
);
1255 else if (svc
->port
== 0)
1256 atomic_inc(&ipvs
->nullsvc_counter
);
1258 ip_vs_start_estimator(net
, &svc
->stats
);
1260 /* Count only IPv4 services for old get/setsockopt interface */
1261 if (svc
->af
== AF_INET
)
1262 ipvs
->num_services
++;
1264 /* Hash the service into the service table */
1265 ip_vs_svc_hash(svc
);
1268 /* Now there is a service - full throttle */
1275 ip_vs_unbind_scheduler(svc
, sched
);
1276 ip_vs_service_free(svc
);
1278 ip_vs_scheduler_put(sched
);
1281 /* decrease the module use count */
1282 ip_vs_use_count_dec();
1289 * Edit a service and bind it with a new scheduler
1292 ip_vs_edit_service(struct ip_vs_service
*svc
, struct ip_vs_service_user_kern
*u
)
1294 struct ip_vs_scheduler
*sched
= NULL
, *old_sched
;
1295 struct ip_vs_pe
*pe
= NULL
, *old_pe
= NULL
;
1299 * Lookup the scheduler, by 'u->sched_name'
1301 if (strcmp(u
->sched_name
, "none")) {
1302 sched
= ip_vs_scheduler_get(u
->sched_name
);
1304 pr_info("Scheduler module ip_vs_%s not found\n",
1311 if (u
->pe_name
&& *u
->pe_name
) {
1312 pe
= ip_vs_pe_getbyname(u
->pe_name
);
1314 pr_info("persistence engine module ip_vs_pe_%s "
1315 "not found\n", u
->pe_name
);
1322 #ifdef CONFIG_IP_VS_IPV6
1323 if (u
->af
== AF_INET6
) {
1324 __u32 plen
= (__force __u32
) u
->netmask
;
1326 if (plen
< 1 || plen
> 128) {
1333 old_sched
= rcu_dereference_protected(svc
->scheduler
, 1);
1334 if (sched
!= old_sched
) {
1336 ip_vs_unbind_scheduler(svc
, old_sched
);
1337 RCU_INIT_POINTER(svc
->scheduler
, NULL
);
1338 /* Wait all svc->sched_data users */
1341 /* Bind the new scheduler */
1343 ret
= ip_vs_bind_scheduler(svc
, sched
);
1345 ip_vs_scheduler_put(sched
);
1352 * Set the flags and timeout value
1354 svc
->flags
= u
->flags
| IP_VS_SVC_F_HASHED
;
1355 svc
->timeout
= u
->timeout
* HZ
;
1356 svc
->netmask
= u
->netmask
;
1358 old_pe
= rcu_dereference_protected(svc
->pe
, 1);
1360 rcu_assign_pointer(svc
->pe
, pe
);
1363 ip_vs_scheduler_put(old_sched
);
1364 ip_vs_pe_put(old_pe
);
1369 * Delete a service from the service list
1370 * - The service must be unlinked, unlocked and not referenced!
1371 * - We are called under _bh lock
1373 static void __ip_vs_del_service(struct ip_vs_service
*svc
, bool cleanup
)
1375 struct ip_vs_dest
*dest
, *nxt
;
1376 struct ip_vs_scheduler
*old_sched
;
1377 struct ip_vs_pe
*old_pe
;
1378 struct netns_ipvs
*ipvs
= svc
->ipvs
;
1380 pr_info("%s: enter\n", __func__
);
1382 /* Count only IPv4 services for old get/setsockopt interface */
1383 if (svc
->af
== AF_INET
)
1384 ipvs
->num_services
--;
1386 ip_vs_stop_estimator(svc
->ipvs
->net
, &svc
->stats
);
1388 /* Unbind scheduler */
1389 old_sched
= rcu_dereference_protected(svc
->scheduler
, 1);
1390 ip_vs_unbind_scheduler(svc
, old_sched
);
1391 ip_vs_scheduler_put(old_sched
);
1393 /* Unbind persistence engine, keep svc->pe */
1394 old_pe
= rcu_dereference_protected(svc
->pe
, 1);
1395 ip_vs_pe_put(old_pe
);
1398 * Unlink the whole destination list
1400 list_for_each_entry_safe(dest
, nxt
, &svc
->destinations
, n_list
) {
1401 __ip_vs_unlink_dest(svc
, dest
, 0);
1402 __ip_vs_del_dest(svc
->ipvs
, dest
, cleanup
);
1406 * Update the virtual service counters
1408 if (svc
->port
== FTPPORT
)
1409 atomic_dec(&ipvs
->ftpsvc_counter
);
1410 else if (svc
->port
== 0)
1411 atomic_dec(&ipvs
->nullsvc_counter
);
1414 * Free the service if nobody refers to it
1416 __ip_vs_svc_put(svc
, true);
1418 /* decrease the module use count */
1419 ip_vs_use_count_dec();
1423 * Unlink a service from list and try to delete it if its refcnt reached 0
1425 static void ip_vs_unlink_service(struct ip_vs_service
*svc
, bool cleanup
)
1427 /* Hold svc to avoid double release from dest_trash */
1428 atomic_inc(&svc
->refcnt
);
1430 * Unhash it from the service table
1432 ip_vs_svc_unhash(svc
);
1434 __ip_vs_del_service(svc
, cleanup
);
1438 * Delete a service from the service list
1440 static int ip_vs_del_service(struct ip_vs_service
*svc
)
1444 ip_vs_unlink_service(svc
, false);
1451 * Flush all the virtual services
1453 static int ip_vs_flush(struct net
*net
, bool cleanup
)
1455 struct netns_ipvs
*ipvs
= net_ipvs(net
);
1457 struct ip_vs_service
*svc
;
1458 struct hlist_node
*n
;
1461 * Flush the service table hashed by <netns,protocol,addr,port>
1463 for(idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1464 hlist_for_each_entry_safe(svc
, n
, &ip_vs_svc_table
[idx
],
1466 if (svc
->ipvs
== ipvs
)
1467 ip_vs_unlink_service(svc
, cleanup
);
1472 * Flush the service table hashed by fwmark
1474 for(idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1475 hlist_for_each_entry_safe(svc
, n
, &ip_vs_svc_fwm_table
[idx
],
1477 if (svc
->ipvs
== ipvs
)
1478 ip_vs_unlink_service(svc
, cleanup
);
1486 * Delete service by {netns} in the service table.
1487 * Called by __ip_vs_cleanup()
1489 void ip_vs_service_net_cleanup(struct net
*net
)
1492 /* Check for "full" addressed entries */
1493 mutex_lock(&__ip_vs_mutex
);
1494 ip_vs_flush(net
, true);
1495 mutex_unlock(&__ip_vs_mutex
);
1499 /* Put all references for device (dst_cache) */
1501 ip_vs_forget_dev(struct ip_vs_dest
*dest
, struct net_device
*dev
)
1503 struct ip_vs_dest_dst
*dest_dst
;
1505 spin_lock_bh(&dest
->dst_lock
);
1506 dest_dst
= rcu_dereference_protected(dest
->dest_dst
, 1);
1507 if (dest_dst
&& dest_dst
->dst_cache
->dev
== dev
) {
1508 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1510 IP_VS_DBG_ADDR(dest
->af
, &dest
->addr
),
1512 atomic_read(&dest
->refcnt
));
1513 __ip_vs_dst_cache_reset(dest
);
1515 spin_unlock_bh(&dest
->dst_lock
);
1518 /* Netdev event receiver
1519 * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1521 static int ip_vs_dst_event(struct notifier_block
*this, unsigned long event
,
1524 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
1525 struct net
*net
= dev_net(dev
);
1526 struct netns_ipvs
*ipvs
= net_ipvs(net
);
1527 struct ip_vs_service
*svc
;
1528 struct ip_vs_dest
*dest
;
1531 if (event
!= NETDEV_DOWN
|| !ipvs
)
1533 IP_VS_DBG(3, "%s() dev=%s\n", __func__
, dev
->name
);
1535 mutex_lock(&__ip_vs_mutex
);
1536 for (idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1537 hlist_for_each_entry(svc
, &ip_vs_svc_table
[idx
], s_list
) {
1538 if (svc
->ipvs
== ipvs
) {
1539 list_for_each_entry(dest
, &svc
->destinations
,
1541 ip_vs_forget_dev(dest
, dev
);
1546 hlist_for_each_entry(svc
, &ip_vs_svc_fwm_table
[idx
], f_list
) {
1547 if (svc
->ipvs
== ipvs
) {
1548 list_for_each_entry(dest
, &svc
->destinations
,
1550 ip_vs_forget_dev(dest
, dev
);
1557 spin_lock_bh(&ipvs
->dest_trash_lock
);
1558 list_for_each_entry(dest
, &ipvs
->dest_trash
, t_list
) {
1559 ip_vs_forget_dev(dest
, dev
);
1561 spin_unlock_bh(&ipvs
->dest_trash_lock
);
1562 mutex_unlock(&__ip_vs_mutex
);
1568 * Zero counters in a service or all services
1570 static int ip_vs_zero_service(struct ip_vs_service
*svc
)
1572 struct ip_vs_dest
*dest
;
1574 list_for_each_entry(dest
, &svc
->destinations
, n_list
) {
1575 ip_vs_zero_stats(&dest
->stats
);
1577 ip_vs_zero_stats(&svc
->stats
);
1581 static int ip_vs_zero_all(struct net
*net
)
1583 struct netns_ipvs
*ipvs
= net_ipvs(net
);
1585 struct ip_vs_service
*svc
;
1587 for(idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1588 hlist_for_each_entry(svc
, &ip_vs_svc_table
[idx
], s_list
) {
1589 if (svc
->ipvs
== ipvs
)
1590 ip_vs_zero_service(svc
);
1594 for(idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1595 hlist_for_each_entry(svc
, &ip_vs_svc_fwm_table
[idx
], f_list
) {
1596 if (svc
->ipvs
== ipvs
)
1597 ip_vs_zero_service(svc
);
1601 ip_vs_zero_stats(&ipvs
->tot_stats
);
1605 #ifdef CONFIG_SYSCTL
1608 static int three
= 3;
1611 proc_do_defense_mode(struct ctl_table
*table
, int write
,
1612 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
1614 struct netns_ipvs
*ipvs
= table
->extra2
;
1615 int *valp
= table
->data
;
1619 rc
= proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
1620 if (write
&& (*valp
!= val
)) {
1621 if ((*valp
< 0) || (*valp
> 3)) {
1622 /* Restore the correct value */
1625 update_defense_level(ipvs
);
1632 proc_do_sync_threshold(struct ctl_table
*table
, int write
,
1633 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
1635 int *valp
= table
->data
;
1639 /* backup the value first */
1640 memcpy(val
, valp
, sizeof(val
));
1642 rc
= proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
1643 if (write
&& (valp
[0] < 0 || valp
[1] < 0 ||
1644 (valp
[0] >= valp
[1] && valp
[1]))) {
1645 /* Restore the correct value */
1646 memcpy(valp
, val
, sizeof(val
));
1652 proc_do_sync_mode(struct ctl_table
*table
, int write
,
1653 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
1655 int *valp
= table
->data
;
1659 rc
= proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
1660 if (write
&& (*valp
!= val
)) {
1661 if ((*valp
< 0) || (*valp
> 1)) {
1662 /* Restore the correct value */
1670 proc_do_sync_ports(struct ctl_table
*table
, int write
,
1671 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
1673 int *valp
= table
->data
;
1677 rc
= proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
1678 if (write
&& (*valp
!= val
)) {
1679 if (*valp
< 1 || !is_power_of_2(*valp
)) {
1680 /* Restore the correct value */
1688 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1689 * Do not change order or insert new entries without
1690 * align with netns init in ip_vs_control_net_init()
1693 static struct ctl_table vs_vars
[] = {
1695 .procname
= "amemthresh",
1696 .maxlen
= sizeof(int),
1698 .proc_handler
= proc_dointvec
,
1701 .procname
= "am_droprate",
1702 .maxlen
= sizeof(int),
1704 .proc_handler
= proc_dointvec
,
1707 .procname
= "drop_entry",
1708 .maxlen
= sizeof(int),
1710 .proc_handler
= proc_do_defense_mode
,
1713 .procname
= "drop_packet",
1714 .maxlen
= sizeof(int),
1716 .proc_handler
= proc_do_defense_mode
,
1718 #ifdef CONFIG_IP_VS_NFCT
1720 .procname
= "conntrack",
1721 .maxlen
= sizeof(int),
1723 .proc_handler
= &proc_dointvec
,
1727 .procname
= "secure_tcp",
1728 .maxlen
= sizeof(int),
1730 .proc_handler
= proc_do_defense_mode
,
1733 .procname
= "snat_reroute",
1734 .maxlen
= sizeof(int),
1736 .proc_handler
= &proc_dointvec
,
1739 .procname
= "sync_version",
1740 .maxlen
= sizeof(int),
1742 .proc_handler
= &proc_do_sync_mode
,
1745 .procname
= "sync_ports",
1746 .maxlen
= sizeof(int),
1748 .proc_handler
= &proc_do_sync_ports
,
1751 .procname
= "sync_persist_mode",
1752 .maxlen
= sizeof(int),
1754 .proc_handler
= proc_dointvec
,
1757 .procname
= "sync_qlen_max",
1758 .maxlen
= sizeof(unsigned long),
1760 .proc_handler
= proc_doulongvec_minmax
,
1763 .procname
= "sync_sock_size",
1764 .maxlen
= sizeof(int),
1766 .proc_handler
= proc_dointvec
,
1769 .procname
= "cache_bypass",
1770 .maxlen
= sizeof(int),
1772 .proc_handler
= proc_dointvec
,
1775 .procname
= "expire_nodest_conn",
1776 .maxlen
= sizeof(int),
1778 .proc_handler
= proc_dointvec
,
1781 .procname
= "sloppy_tcp",
1782 .maxlen
= sizeof(int),
1784 .proc_handler
= proc_dointvec
,
1787 .procname
= "sloppy_sctp",
1788 .maxlen
= sizeof(int),
1790 .proc_handler
= proc_dointvec
,
1793 .procname
= "expire_quiescent_template",
1794 .maxlen
= sizeof(int),
1796 .proc_handler
= proc_dointvec
,
1799 .procname
= "sync_threshold",
1801 sizeof(((struct netns_ipvs
*)0)->sysctl_sync_threshold
),
1803 .proc_handler
= proc_do_sync_threshold
,
1806 .procname
= "sync_refresh_period",
1807 .maxlen
= sizeof(int),
1809 .proc_handler
= proc_dointvec_jiffies
,
1812 .procname
= "sync_retries",
1813 .maxlen
= sizeof(int),
1815 .proc_handler
= proc_dointvec_minmax
,
1820 .procname
= "nat_icmp_send",
1821 .maxlen
= sizeof(int),
1823 .proc_handler
= proc_dointvec
,
1826 .procname
= "pmtu_disc",
1827 .maxlen
= sizeof(int),
1829 .proc_handler
= proc_dointvec
,
1832 .procname
= "backup_only",
1833 .maxlen
= sizeof(int),
1835 .proc_handler
= proc_dointvec
,
1838 .procname
= "conn_reuse_mode",
1839 .maxlen
= sizeof(int),
1841 .proc_handler
= proc_dointvec
,
1844 .procname
= "schedule_icmp",
1845 .maxlen
= sizeof(int),
1847 .proc_handler
= proc_dointvec
,
1850 .procname
= "ignore_tunneled",
1851 .maxlen
= sizeof(int),
1853 .proc_handler
= proc_dointvec
,
1855 #ifdef CONFIG_IP_VS_DEBUG
1857 .procname
= "debug_level",
1858 .data
= &sysctl_ip_vs_debug_level
,
1859 .maxlen
= sizeof(int),
1861 .proc_handler
= proc_dointvec
,
1869 #ifdef CONFIG_PROC_FS
1872 struct seq_net_private p
; /* Do not move this, netns depends upon it*/
1873 struct hlist_head
*table
;
1878 * Write the contents of the VS rule table to a PROCfs file.
1879 * (It is kept just for backward compatibility)
1881 static inline const char *ip_vs_fwd_name(unsigned int flags
)
1883 switch (flags
& IP_VS_CONN_F_FWD_MASK
) {
1884 case IP_VS_CONN_F_LOCALNODE
:
1886 case IP_VS_CONN_F_TUNNEL
:
1888 case IP_VS_CONN_F_DROUTE
:
1896 /* Get the Nth entry in the two lists */
1897 static struct ip_vs_service
*ip_vs_info_array(struct seq_file
*seq
, loff_t pos
)
1899 struct net
*net
= seq_file_net(seq
);
1900 struct netns_ipvs
*ipvs
= net_ipvs(net
);
1901 struct ip_vs_iter
*iter
= seq
->private;
1903 struct ip_vs_service
*svc
;
1905 /* look in hash by protocol */
1906 for (idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1907 hlist_for_each_entry_rcu(svc
, &ip_vs_svc_table
[idx
], s_list
) {
1908 if ((svc
->ipvs
== ipvs
) && pos
-- == 0) {
1909 iter
->table
= ip_vs_svc_table
;
1916 /* keep looking in fwmark */
1917 for (idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1918 hlist_for_each_entry_rcu(svc
, &ip_vs_svc_fwm_table
[idx
],
1920 if ((svc
->ipvs
== ipvs
) && pos
-- == 0) {
1921 iter
->table
= ip_vs_svc_fwm_table
;
1931 static void *ip_vs_info_seq_start(struct seq_file
*seq
, loff_t
*pos
)
1935 return *pos
? ip_vs_info_array(seq
, *pos
- 1) : SEQ_START_TOKEN
;
1939 static void *ip_vs_info_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
1941 struct hlist_node
*e
;
1942 struct ip_vs_iter
*iter
;
1943 struct ip_vs_service
*svc
;
1946 if (v
== SEQ_START_TOKEN
)
1947 return ip_vs_info_array(seq
,0);
1950 iter
= seq
->private;
1952 if (iter
->table
== ip_vs_svc_table
) {
1953 /* next service in table hashed by protocol */
1954 e
= rcu_dereference(hlist_next_rcu(&svc
->s_list
));
1956 return hlist_entry(e
, struct ip_vs_service
, s_list
);
1958 while (++iter
->bucket
< IP_VS_SVC_TAB_SIZE
) {
1959 hlist_for_each_entry_rcu(svc
,
1960 &ip_vs_svc_table
[iter
->bucket
],
1966 iter
->table
= ip_vs_svc_fwm_table
;
1971 /* next service in hashed by fwmark */
1972 e
= rcu_dereference(hlist_next_rcu(&svc
->f_list
));
1974 return hlist_entry(e
, struct ip_vs_service
, f_list
);
1977 while (++iter
->bucket
< IP_VS_SVC_TAB_SIZE
) {
1978 hlist_for_each_entry_rcu(svc
,
1979 &ip_vs_svc_fwm_table
[iter
->bucket
],
1987 static void ip_vs_info_seq_stop(struct seq_file
*seq
, void *v
)
1994 static int ip_vs_info_seq_show(struct seq_file
*seq
, void *v
)
1996 if (v
== SEQ_START_TOKEN
) {
1998 "IP Virtual Server version %d.%d.%d (size=%d)\n",
1999 NVERSION(IP_VS_VERSION_CODE
), ip_vs_conn_tab_size
);
2001 "Prot LocalAddress:Port Scheduler Flags\n");
2003 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2005 const struct ip_vs_service
*svc
= v
;
2006 const struct ip_vs_iter
*iter
= seq
->private;
2007 const struct ip_vs_dest
*dest
;
2008 struct ip_vs_scheduler
*sched
= rcu_dereference(svc
->scheduler
);
2009 char *sched_name
= sched
? sched
->name
: "none";
2011 if (iter
->table
== ip_vs_svc_table
) {
2012 #ifdef CONFIG_IP_VS_IPV6
2013 if (svc
->af
== AF_INET6
)
2014 seq_printf(seq
, "%s [%pI6]:%04X %s ",
2015 ip_vs_proto_name(svc
->protocol
),
2021 seq_printf(seq
, "%s %08X:%04X %s %s ",
2022 ip_vs_proto_name(svc
->protocol
),
2023 ntohl(svc
->addr
.ip
),
2026 (svc
->flags
& IP_VS_SVC_F_ONEPACKET
)?"ops ":"");
2028 seq_printf(seq
, "FWM %08X %s %s",
2029 svc
->fwmark
, sched_name
,
2030 (svc
->flags
& IP_VS_SVC_F_ONEPACKET
)?"ops ":"");
2033 if (svc
->flags
& IP_VS_SVC_F_PERSISTENT
)
2034 seq_printf(seq
, "persistent %d %08X\n",
2036 ntohl(svc
->netmask
));
2038 seq_putc(seq
, '\n');
2040 list_for_each_entry_rcu(dest
, &svc
->destinations
, n_list
) {
2041 #ifdef CONFIG_IP_VS_IPV6
2042 if (dest
->af
== AF_INET6
)
2045 " %-7s %-6d %-10d %-10d\n",
2048 ip_vs_fwd_name(atomic_read(&dest
->conn_flags
)),
2049 atomic_read(&dest
->weight
),
2050 atomic_read(&dest
->activeconns
),
2051 atomic_read(&dest
->inactconns
));
2056 "%-7s %-6d %-10d %-10d\n",
2057 ntohl(dest
->addr
.ip
),
2059 ip_vs_fwd_name(atomic_read(&dest
->conn_flags
)),
2060 atomic_read(&dest
->weight
),
2061 atomic_read(&dest
->activeconns
),
2062 atomic_read(&dest
->inactconns
));
2069 static const struct seq_operations ip_vs_info_seq_ops
= {
2070 .start
= ip_vs_info_seq_start
,
2071 .next
= ip_vs_info_seq_next
,
2072 .stop
= ip_vs_info_seq_stop
,
2073 .show
= ip_vs_info_seq_show
,
2076 static int ip_vs_info_open(struct inode
*inode
, struct file
*file
)
2078 return seq_open_net(inode
, file
, &ip_vs_info_seq_ops
,
2079 sizeof(struct ip_vs_iter
));
2082 static const struct file_operations ip_vs_info_fops
= {
2083 .owner
= THIS_MODULE
,
2084 .open
= ip_vs_info_open
,
2086 .llseek
= seq_lseek
,
2087 .release
= seq_release_net
,
2090 static int ip_vs_stats_show(struct seq_file
*seq
, void *v
)
2092 struct net
*net
= seq_file_single_net(seq
);
2093 struct ip_vs_kstats show
;
2095 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2097 " Total Incoming Outgoing Incoming Outgoing\n");
2099 " Conns Packets Packets Bytes Bytes\n");
2101 ip_vs_copy_stats(&show
, &net_ipvs(net
)->tot_stats
);
2102 seq_printf(seq
, "%8LX %8LX %8LX %16LX %16LX\n\n",
2103 (unsigned long long)show
.conns
,
2104 (unsigned long long)show
.inpkts
,
2105 (unsigned long long)show
.outpkts
,
2106 (unsigned long long)show
.inbytes
,
2107 (unsigned long long)show
.outbytes
);
2109 /* 01234567 01234567 01234567 0123456701234567 0123456701234567*/
2111 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2112 seq_printf(seq
, "%8LX %8LX %8LX %16LX %16LX\n",
2113 (unsigned long long)show
.cps
,
2114 (unsigned long long)show
.inpps
,
2115 (unsigned long long)show
.outpps
,
2116 (unsigned long long)show
.inbps
,
2117 (unsigned long long)show
.outbps
);
2122 static int ip_vs_stats_seq_open(struct inode
*inode
, struct file
*file
)
2124 return single_open_net(inode
, file
, ip_vs_stats_show
);
2127 static const struct file_operations ip_vs_stats_fops
= {
2128 .owner
= THIS_MODULE
,
2129 .open
= ip_vs_stats_seq_open
,
2131 .llseek
= seq_lseek
,
2132 .release
= single_release_net
,
2135 static int ip_vs_stats_percpu_show(struct seq_file
*seq
, void *v
)
2137 struct net
*net
= seq_file_single_net(seq
);
2138 struct ip_vs_stats
*tot_stats
= &net_ipvs(net
)->tot_stats
;
2139 struct ip_vs_cpu_stats __percpu
*cpustats
= tot_stats
->cpustats
;
2140 struct ip_vs_kstats kstats
;
2143 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2145 " Total Incoming Outgoing Incoming Outgoing\n");
2147 "CPU Conns Packets Packets Bytes Bytes\n");
2149 for_each_possible_cpu(i
) {
2150 struct ip_vs_cpu_stats
*u
= per_cpu_ptr(cpustats
, i
);
2152 u64 conns
, inpkts
, outpkts
, inbytes
, outbytes
;
2155 start
= u64_stats_fetch_begin_irq(&u
->syncp
);
2156 conns
= u
->cnt
.conns
;
2157 inpkts
= u
->cnt
.inpkts
;
2158 outpkts
= u
->cnt
.outpkts
;
2159 inbytes
= u
->cnt
.inbytes
;
2160 outbytes
= u
->cnt
.outbytes
;
2161 } while (u64_stats_fetch_retry_irq(&u
->syncp
, start
));
2163 seq_printf(seq
, "%3X %8LX %8LX %8LX %16LX %16LX\n",
2164 i
, (u64
)conns
, (u64
)inpkts
,
2165 (u64
)outpkts
, (u64
)inbytes
,
2169 ip_vs_copy_stats(&kstats
, tot_stats
);
2171 seq_printf(seq
, " ~ %8LX %8LX %8LX %16LX %16LX\n\n",
2172 (unsigned long long)kstats
.conns
,
2173 (unsigned long long)kstats
.inpkts
,
2174 (unsigned long long)kstats
.outpkts
,
2175 (unsigned long long)kstats
.inbytes
,
2176 (unsigned long long)kstats
.outbytes
);
2178 /* ... 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2180 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2181 seq_printf(seq
, " %8LX %8LX %8LX %16LX %16LX\n",
2191 static int ip_vs_stats_percpu_seq_open(struct inode
*inode
, struct file
*file
)
2193 return single_open_net(inode
, file
, ip_vs_stats_percpu_show
);
2196 static const struct file_operations ip_vs_stats_percpu_fops
= {
2197 .owner
= THIS_MODULE
,
2198 .open
= ip_vs_stats_percpu_seq_open
,
2200 .llseek
= seq_lseek
,
2201 .release
= single_release_net
,
2206 * Set timeout values for tcp tcpfin udp in the timeout_table.
2208 static int ip_vs_set_timeout(struct net
*net
, struct ip_vs_timeout_user
*u
)
2210 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2211 struct ip_vs_proto_data
*pd
;
2214 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2219 #ifdef CONFIG_IP_VS_PROTO_TCP
2220 if (u
->tcp_timeout
) {
2221 pd
= ip_vs_proto_data_get(net
, IPPROTO_TCP
);
2222 pd
->timeout_table
[IP_VS_TCP_S_ESTABLISHED
]
2223 = u
->tcp_timeout
* HZ
;
2226 if (u
->tcp_fin_timeout
) {
2227 pd
= ip_vs_proto_data_get(net
, IPPROTO_TCP
);
2228 pd
->timeout_table
[IP_VS_TCP_S_FIN_WAIT
]
2229 = u
->tcp_fin_timeout
* HZ
;
2233 #ifdef CONFIG_IP_VS_PROTO_UDP
2234 if (u
->udp_timeout
) {
2235 pd
= ip_vs_proto_data_get(net
, IPPROTO_UDP
);
2236 pd
->timeout_table
[IP_VS_UDP_S_NORMAL
]
2237 = u
->udp_timeout
* HZ
;
2243 #define CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2245 struct ip_vs_svcdest_user
{
2246 struct ip_vs_service_user s
;
2247 struct ip_vs_dest_user d
;
2250 static const unsigned char set_arglen
[CMDID(IP_VS_SO_SET_MAX
) + 1] = {
2251 [CMDID(IP_VS_SO_SET_ADD
)] = sizeof(struct ip_vs_service_user
),
2252 [CMDID(IP_VS_SO_SET_EDIT
)] = sizeof(struct ip_vs_service_user
),
2253 [CMDID(IP_VS_SO_SET_DEL
)] = sizeof(struct ip_vs_service_user
),
2254 [CMDID(IP_VS_SO_SET_ADDDEST
)] = sizeof(struct ip_vs_svcdest_user
),
2255 [CMDID(IP_VS_SO_SET_DELDEST
)] = sizeof(struct ip_vs_svcdest_user
),
2256 [CMDID(IP_VS_SO_SET_EDITDEST
)] = sizeof(struct ip_vs_svcdest_user
),
2257 [CMDID(IP_VS_SO_SET_TIMEOUT
)] = sizeof(struct ip_vs_timeout_user
),
2258 [CMDID(IP_VS_SO_SET_STARTDAEMON
)] = sizeof(struct ip_vs_daemon_user
),
2259 [CMDID(IP_VS_SO_SET_STOPDAEMON
)] = sizeof(struct ip_vs_daemon_user
),
2260 [CMDID(IP_VS_SO_SET_ZERO
)] = sizeof(struct ip_vs_service_user
),
2263 union ip_vs_set_arglen
{
2264 struct ip_vs_service_user field_IP_VS_SO_SET_ADD
;
2265 struct ip_vs_service_user field_IP_VS_SO_SET_EDIT
;
2266 struct ip_vs_service_user field_IP_VS_SO_SET_DEL
;
2267 struct ip_vs_svcdest_user field_IP_VS_SO_SET_ADDDEST
;
2268 struct ip_vs_svcdest_user field_IP_VS_SO_SET_DELDEST
;
2269 struct ip_vs_svcdest_user field_IP_VS_SO_SET_EDITDEST
;
2270 struct ip_vs_timeout_user field_IP_VS_SO_SET_TIMEOUT
;
2271 struct ip_vs_daemon_user field_IP_VS_SO_SET_STARTDAEMON
;
2272 struct ip_vs_daemon_user field_IP_VS_SO_SET_STOPDAEMON
;
2273 struct ip_vs_service_user field_IP_VS_SO_SET_ZERO
;
2276 #define MAX_SET_ARGLEN sizeof(union ip_vs_set_arglen)
2278 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern
*usvc
,
2279 struct ip_vs_service_user
*usvc_compat
)
2281 memset(usvc
, 0, sizeof(*usvc
));
2284 usvc
->protocol
= usvc_compat
->protocol
;
2285 usvc
->addr
.ip
= usvc_compat
->addr
;
2286 usvc
->port
= usvc_compat
->port
;
2287 usvc
->fwmark
= usvc_compat
->fwmark
;
2289 /* Deep copy of sched_name is not needed here */
2290 usvc
->sched_name
= usvc_compat
->sched_name
;
2292 usvc
->flags
= usvc_compat
->flags
;
2293 usvc
->timeout
= usvc_compat
->timeout
;
2294 usvc
->netmask
= usvc_compat
->netmask
;
2297 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern
*udest
,
2298 struct ip_vs_dest_user
*udest_compat
)
2300 memset(udest
, 0, sizeof(*udest
));
2302 udest
->addr
.ip
= udest_compat
->addr
;
2303 udest
->port
= udest_compat
->port
;
2304 udest
->conn_flags
= udest_compat
->conn_flags
;
2305 udest
->weight
= udest_compat
->weight
;
2306 udest
->u_threshold
= udest_compat
->u_threshold
;
2307 udest
->l_threshold
= udest_compat
->l_threshold
;
2308 udest
->af
= AF_INET
;
2312 do_ip_vs_set_ctl(struct sock
*sk
, int cmd
, void __user
*user
, unsigned int len
)
2314 struct net
*net
= sock_net(sk
);
2316 unsigned char arg
[MAX_SET_ARGLEN
];
2317 struct ip_vs_service_user
*usvc_compat
;
2318 struct ip_vs_service_user_kern usvc
;
2319 struct ip_vs_service
*svc
;
2320 struct ip_vs_dest_user
*udest_compat
;
2321 struct ip_vs_dest_user_kern udest
;
2322 struct netns_ipvs
*ipvs
= net_ipvs(net
);
2324 BUILD_BUG_ON(sizeof(arg
) > 255);
2325 if (!ns_capable(sock_net(sk
)->user_ns
, CAP_NET_ADMIN
))
2328 if (cmd
< IP_VS_BASE_CTL
|| cmd
> IP_VS_SO_SET_MAX
)
2330 if (len
!= set_arglen
[CMDID(cmd
)]) {
2331 IP_VS_DBG(1, "set_ctl: len %u != %u\n",
2332 len
, set_arglen
[CMDID(cmd
)]);
2336 if (copy_from_user(arg
, user
, len
) != 0)
2339 /* increase the module use count */
2340 ip_vs_use_count_inc();
2342 /* Handle daemons since they have another lock */
2343 if (cmd
== IP_VS_SO_SET_STARTDAEMON
||
2344 cmd
== IP_VS_SO_SET_STOPDAEMON
) {
2345 struct ip_vs_daemon_user
*dm
= (struct ip_vs_daemon_user
*)arg
;
2347 if (cmd
== IP_VS_SO_SET_STARTDAEMON
) {
2348 struct ipvs_sync_daemon_cfg cfg
;
2350 memset(&cfg
, 0, sizeof(cfg
));
2351 strlcpy(cfg
.mcast_ifn
, dm
->mcast_ifn
,
2352 sizeof(cfg
.mcast_ifn
));
2353 cfg
.syncid
= dm
->syncid
;
2355 mutex_lock(&ipvs
->sync_mutex
);
2356 ret
= start_sync_thread(net
, &cfg
, dm
->state
);
2357 mutex_unlock(&ipvs
->sync_mutex
);
2360 mutex_lock(&ipvs
->sync_mutex
);
2361 ret
= stop_sync_thread(net
, dm
->state
);
2362 mutex_unlock(&ipvs
->sync_mutex
);
2367 mutex_lock(&__ip_vs_mutex
);
2368 if (cmd
== IP_VS_SO_SET_FLUSH
) {
2369 /* Flush the virtual service */
2370 ret
= ip_vs_flush(net
, false);
2372 } else if (cmd
== IP_VS_SO_SET_TIMEOUT
) {
2373 /* Set timeout values for (tcp tcpfin udp) */
2374 ret
= ip_vs_set_timeout(net
, (struct ip_vs_timeout_user
*)arg
);
2378 usvc_compat
= (struct ip_vs_service_user
*)arg
;
2379 udest_compat
= (struct ip_vs_dest_user
*)(usvc_compat
+ 1);
2381 /* We only use the new structs internally, so copy userspace compat
2382 * structs to extended internal versions */
2383 ip_vs_copy_usvc_compat(&usvc
, usvc_compat
);
2384 ip_vs_copy_udest_compat(&udest
, udest_compat
);
2386 if (cmd
== IP_VS_SO_SET_ZERO
) {
2387 /* if no service address is set, zero counters in all */
2388 if (!usvc
.fwmark
&& !usvc
.addr
.ip
&& !usvc
.port
) {
2389 ret
= ip_vs_zero_all(net
);
2394 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2395 if (usvc
.protocol
!= IPPROTO_TCP
&& usvc
.protocol
!= IPPROTO_UDP
&&
2396 usvc
.protocol
!= IPPROTO_SCTP
) {
2397 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2398 usvc
.protocol
, &usvc
.addr
.ip
,
2399 ntohs(usvc
.port
), usvc
.sched_name
);
2404 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2406 if (usvc
.fwmark
== 0)
2407 svc
= __ip_vs_service_find(ipvs
, usvc
.af
, usvc
.protocol
,
2408 &usvc
.addr
, usvc
.port
);
2410 svc
= __ip_vs_svc_fwm_find(ipvs
, usvc
.af
, usvc
.fwmark
);
2413 if (cmd
!= IP_VS_SO_SET_ADD
2414 && (svc
== NULL
|| svc
->protocol
!= usvc
.protocol
)) {
2420 case IP_VS_SO_SET_ADD
:
2424 ret
= ip_vs_add_service(net
, &usvc
, &svc
);
2426 case IP_VS_SO_SET_EDIT
:
2427 ret
= ip_vs_edit_service(svc
, &usvc
);
2429 case IP_VS_SO_SET_DEL
:
2430 ret
= ip_vs_del_service(svc
);
2434 case IP_VS_SO_SET_ZERO
:
2435 ret
= ip_vs_zero_service(svc
);
2437 case IP_VS_SO_SET_ADDDEST
:
2438 ret
= ip_vs_add_dest(svc
, &udest
);
2440 case IP_VS_SO_SET_EDITDEST
:
2441 ret
= ip_vs_edit_dest(svc
, &udest
);
2443 case IP_VS_SO_SET_DELDEST
:
2444 ret
= ip_vs_del_dest(svc
, &udest
);
2451 mutex_unlock(&__ip_vs_mutex
);
2453 /* decrease the module use count */
2454 ip_vs_use_count_dec();
2461 ip_vs_copy_service(struct ip_vs_service_entry
*dst
, struct ip_vs_service
*src
)
2463 struct ip_vs_scheduler
*sched
;
2464 struct ip_vs_kstats kstats
;
2467 sched
= rcu_dereference_protected(src
->scheduler
, 1);
2468 sched_name
= sched
? sched
->name
: "none";
2469 dst
->protocol
= src
->protocol
;
2470 dst
->addr
= src
->addr
.ip
;
2471 dst
->port
= src
->port
;
2472 dst
->fwmark
= src
->fwmark
;
2473 strlcpy(dst
->sched_name
, sched_name
, sizeof(dst
->sched_name
));
2474 dst
->flags
= src
->flags
;
2475 dst
->timeout
= src
->timeout
/ HZ
;
2476 dst
->netmask
= src
->netmask
;
2477 dst
->num_dests
= src
->num_dests
;
2478 ip_vs_copy_stats(&kstats
, &src
->stats
);
2479 ip_vs_export_stats_user(&dst
->stats
, &kstats
);
2483 __ip_vs_get_service_entries(struct net
*net
,
2484 const struct ip_vs_get_services
*get
,
2485 struct ip_vs_get_services __user
*uptr
)
2487 struct netns_ipvs
*ipvs
= net_ipvs(net
);
2489 struct ip_vs_service
*svc
;
2490 struct ip_vs_service_entry entry
;
2493 for (idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
2494 hlist_for_each_entry(svc
, &ip_vs_svc_table
[idx
], s_list
) {
2495 /* Only expose IPv4 entries to old interface */
2496 if (svc
->af
!= AF_INET
|| (svc
->ipvs
!= ipvs
))
2499 if (count
>= get
->num_services
)
2501 memset(&entry
, 0, sizeof(entry
));
2502 ip_vs_copy_service(&entry
, svc
);
2503 if (copy_to_user(&uptr
->entrytable
[count
],
2504 &entry
, sizeof(entry
))) {
2512 for (idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
2513 hlist_for_each_entry(svc
, &ip_vs_svc_fwm_table
[idx
], f_list
) {
2514 /* Only expose IPv4 entries to old interface */
2515 if (svc
->af
!= AF_INET
|| (svc
->ipvs
!= ipvs
))
2518 if (count
>= get
->num_services
)
2520 memset(&entry
, 0, sizeof(entry
));
2521 ip_vs_copy_service(&entry
, svc
);
2522 if (copy_to_user(&uptr
->entrytable
[count
],
2523 &entry
, sizeof(entry
))) {
2535 __ip_vs_get_dest_entries(struct net
*net
, const struct ip_vs_get_dests
*get
,
2536 struct ip_vs_get_dests __user
*uptr
)
2538 struct netns_ipvs
*ipvs
= net_ipvs(net
);
2539 struct ip_vs_service
*svc
;
2540 union nf_inet_addr addr
= { .ip
= get
->addr
};
2545 svc
= __ip_vs_svc_fwm_find(ipvs
, AF_INET
, get
->fwmark
);
2547 svc
= __ip_vs_service_find(ipvs
, AF_INET
, get
->protocol
, &addr
,
2553 struct ip_vs_dest
*dest
;
2554 struct ip_vs_dest_entry entry
;
2555 struct ip_vs_kstats kstats
;
2557 memset(&entry
, 0, sizeof(entry
));
2558 list_for_each_entry(dest
, &svc
->destinations
, n_list
) {
2559 if (count
>= get
->num_dests
)
2562 /* Cannot expose heterogeneous members via sockopt
2565 if (dest
->af
!= svc
->af
)
2568 entry
.addr
= dest
->addr
.ip
;
2569 entry
.port
= dest
->port
;
2570 entry
.conn_flags
= atomic_read(&dest
->conn_flags
);
2571 entry
.weight
= atomic_read(&dest
->weight
);
2572 entry
.u_threshold
= dest
->u_threshold
;
2573 entry
.l_threshold
= dest
->l_threshold
;
2574 entry
.activeconns
= atomic_read(&dest
->activeconns
);
2575 entry
.inactconns
= atomic_read(&dest
->inactconns
);
2576 entry
.persistconns
= atomic_read(&dest
->persistconns
);
2577 ip_vs_copy_stats(&kstats
, &dest
->stats
);
2578 ip_vs_export_stats_user(&entry
.stats
, &kstats
);
2579 if (copy_to_user(&uptr
->entrytable
[count
],
2580 &entry
, sizeof(entry
))) {
2592 __ip_vs_get_timeouts(struct net
*net
, struct ip_vs_timeout_user
*u
)
2594 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2595 struct ip_vs_proto_data
*pd
;
2598 memset(u
, 0, sizeof (*u
));
2600 #ifdef CONFIG_IP_VS_PROTO_TCP
2601 pd
= ip_vs_proto_data_get(net
, IPPROTO_TCP
);
2602 u
->tcp_timeout
= pd
->timeout_table
[IP_VS_TCP_S_ESTABLISHED
] / HZ
;
2603 u
->tcp_fin_timeout
= pd
->timeout_table
[IP_VS_TCP_S_FIN_WAIT
] / HZ
;
2605 #ifdef CONFIG_IP_VS_PROTO_UDP
2606 pd
= ip_vs_proto_data_get(net
, IPPROTO_UDP
);
2608 pd
->timeout_table
[IP_VS_UDP_S_NORMAL
] / HZ
;
2612 static const unsigned char get_arglen
[CMDID(IP_VS_SO_GET_MAX
) + 1] = {
2613 [CMDID(IP_VS_SO_GET_VERSION
)] = 64,
2614 [CMDID(IP_VS_SO_GET_INFO
)] = sizeof(struct ip_vs_getinfo
),
2615 [CMDID(IP_VS_SO_GET_SERVICES
)] = sizeof(struct ip_vs_get_services
),
2616 [CMDID(IP_VS_SO_GET_SERVICE
)] = sizeof(struct ip_vs_service_entry
),
2617 [CMDID(IP_VS_SO_GET_DESTS
)] = sizeof(struct ip_vs_get_dests
),
2618 [CMDID(IP_VS_SO_GET_TIMEOUT
)] = sizeof(struct ip_vs_timeout_user
),
2619 [CMDID(IP_VS_SO_GET_DAEMON
)] = 2 * sizeof(struct ip_vs_daemon_user
),
2622 union ip_vs_get_arglen
{
2623 char field_IP_VS_SO_GET_VERSION
[64];
2624 struct ip_vs_getinfo field_IP_VS_SO_GET_INFO
;
2625 struct ip_vs_get_services field_IP_VS_SO_GET_SERVICES
;
2626 struct ip_vs_service_entry field_IP_VS_SO_GET_SERVICE
;
2627 struct ip_vs_get_dests field_IP_VS_SO_GET_DESTS
;
2628 struct ip_vs_timeout_user field_IP_VS_SO_GET_TIMEOUT
;
2629 struct ip_vs_daemon_user field_IP_VS_SO_GET_DAEMON
[2];
2632 #define MAX_GET_ARGLEN sizeof(union ip_vs_get_arglen)
2635 do_ip_vs_get_ctl(struct sock
*sk
, int cmd
, void __user
*user
, int *len
)
2637 unsigned char arg
[MAX_GET_ARGLEN
];
2639 unsigned int copylen
;
2640 struct net
*net
= sock_net(sk
);
2641 struct netns_ipvs
*ipvs
= net_ipvs(net
);
2644 BUILD_BUG_ON(sizeof(arg
) > 255);
2645 if (!ns_capable(sock_net(sk
)->user_ns
, CAP_NET_ADMIN
))
2648 if (cmd
< IP_VS_BASE_CTL
|| cmd
> IP_VS_SO_GET_MAX
)
2651 copylen
= get_arglen
[CMDID(cmd
)];
2652 if (*len
< (int) copylen
) {
2653 IP_VS_DBG(1, "get_ctl: len %d < %u\n", *len
, copylen
);
2657 if (copy_from_user(arg
, user
, copylen
) != 0)
2660 * Handle daemons first since it has its own locking
2662 if (cmd
== IP_VS_SO_GET_DAEMON
) {
2663 struct ip_vs_daemon_user d
[2];
2665 memset(&d
, 0, sizeof(d
));
2666 mutex_lock(&ipvs
->sync_mutex
);
2667 if (ipvs
->sync_state
& IP_VS_STATE_MASTER
) {
2668 d
[0].state
= IP_VS_STATE_MASTER
;
2669 strlcpy(d
[0].mcast_ifn
, ipvs
->mcfg
.mcast_ifn
,
2670 sizeof(d
[0].mcast_ifn
));
2671 d
[0].syncid
= ipvs
->mcfg
.syncid
;
2673 if (ipvs
->sync_state
& IP_VS_STATE_BACKUP
) {
2674 d
[1].state
= IP_VS_STATE_BACKUP
;
2675 strlcpy(d
[1].mcast_ifn
, ipvs
->bcfg
.mcast_ifn
,
2676 sizeof(d
[1].mcast_ifn
));
2677 d
[1].syncid
= ipvs
->bcfg
.syncid
;
2679 if (copy_to_user(user
, &d
, sizeof(d
)) != 0)
2681 mutex_unlock(&ipvs
->sync_mutex
);
2685 mutex_lock(&__ip_vs_mutex
);
2687 case IP_VS_SO_GET_VERSION
:
2691 sprintf(buf
, "IP Virtual Server version %d.%d.%d (size=%d)",
2692 NVERSION(IP_VS_VERSION_CODE
), ip_vs_conn_tab_size
);
2693 if (copy_to_user(user
, buf
, strlen(buf
)+1) != 0) {
2697 *len
= strlen(buf
)+1;
2701 case IP_VS_SO_GET_INFO
:
2703 struct ip_vs_getinfo info
;
2704 info
.version
= IP_VS_VERSION_CODE
;
2705 info
.size
= ip_vs_conn_tab_size
;
2706 info
.num_services
= ipvs
->num_services
;
2707 if (copy_to_user(user
, &info
, sizeof(info
)) != 0)
2712 case IP_VS_SO_GET_SERVICES
:
2714 struct ip_vs_get_services
*get
;
2717 get
= (struct ip_vs_get_services
*)arg
;
2718 size
= sizeof(*get
) +
2719 sizeof(struct ip_vs_service_entry
) * get
->num_services
;
2721 pr_err("length: %u != %u\n", *len
, size
);
2725 ret
= __ip_vs_get_service_entries(net
, get
, user
);
2729 case IP_VS_SO_GET_SERVICE
:
2731 struct ip_vs_service_entry
*entry
;
2732 struct ip_vs_service
*svc
;
2733 union nf_inet_addr addr
;
2735 entry
= (struct ip_vs_service_entry
*)arg
;
2736 addr
.ip
= entry
->addr
;
2739 svc
= __ip_vs_svc_fwm_find(ipvs
, AF_INET
, entry
->fwmark
);
2741 svc
= __ip_vs_service_find(ipvs
, AF_INET
,
2742 entry
->protocol
, &addr
,
2746 ip_vs_copy_service(entry
, svc
);
2747 if (copy_to_user(user
, entry
, sizeof(*entry
)) != 0)
2754 case IP_VS_SO_GET_DESTS
:
2756 struct ip_vs_get_dests
*get
;
2759 get
= (struct ip_vs_get_dests
*)arg
;
2760 size
= sizeof(*get
) +
2761 sizeof(struct ip_vs_dest_entry
) * get
->num_dests
;
2763 pr_err("length: %u != %u\n", *len
, size
);
2767 ret
= __ip_vs_get_dest_entries(net
, get
, user
);
2771 case IP_VS_SO_GET_TIMEOUT
:
2773 struct ip_vs_timeout_user t
;
2775 __ip_vs_get_timeouts(net
, &t
);
2776 if (copy_to_user(user
, &t
, sizeof(t
)) != 0)
2786 mutex_unlock(&__ip_vs_mutex
);
2791 static struct nf_sockopt_ops ip_vs_sockopts
= {
2793 .set_optmin
= IP_VS_BASE_CTL
,
2794 .set_optmax
= IP_VS_SO_SET_MAX
+1,
2795 .set
= do_ip_vs_set_ctl
,
2796 .get_optmin
= IP_VS_BASE_CTL
,
2797 .get_optmax
= IP_VS_SO_GET_MAX
+1,
2798 .get
= do_ip_vs_get_ctl
,
2799 .owner
= THIS_MODULE
,
2803 * Generic Netlink interface
2806 /* IPVS genetlink family */
2807 static struct genl_family ip_vs_genl_family
= {
2808 .id
= GENL_ID_GENERATE
,
2810 .name
= IPVS_GENL_NAME
,
2811 .version
= IPVS_GENL_VERSION
,
2812 .maxattr
= IPVS_CMD_MAX
,
2813 .netnsok
= true, /* Make ipvsadm to work on netns */
2816 /* Policy used for first-level command attributes */
2817 static const struct nla_policy ip_vs_cmd_policy
[IPVS_CMD_ATTR_MAX
+ 1] = {
2818 [IPVS_CMD_ATTR_SERVICE
] = { .type
= NLA_NESTED
},
2819 [IPVS_CMD_ATTR_DEST
] = { .type
= NLA_NESTED
},
2820 [IPVS_CMD_ATTR_DAEMON
] = { .type
= NLA_NESTED
},
2821 [IPVS_CMD_ATTR_TIMEOUT_TCP
] = { .type
= NLA_U32
},
2822 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN
] = { .type
= NLA_U32
},
2823 [IPVS_CMD_ATTR_TIMEOUT_UDP
] = { .type
= NLA_U32
},
2826 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2827 static const struct nla_policy ip_vs_daemon_policy
[IPVS_DAEMON_ATTR_MAX
+ 1] = {
2828 [IPVS_DAEMON_ATTR_STATE
] = { .type
= NLA_U32
},
2829 [IPVS_DAEMON_ATTR_MCAST_IFN
] = { .type
= NLA_NUL_STRING
,
2830 .len
= IP_VS_IFNAME_MAXLEN
},
2831 [IPVS_DAEMON_ATTR_SYNC_ID
] = { .type
= NLA_U32
},
2832 [IPVS_DAEMON_ATTR_SYNC_MAXLEN
] = { .type
= NLA_U16
},
2833 [IPVS_DAEMON_ATTR_MCAST_GROUP
] = { .type
= NLA_U32
},
2834 [IPVS_DAEMON_ATTR_MCAST_GROUP6
] = { .len
= sizeof(struct in6_addr
) },
2835 [IPVS_DAEMON_ATTR_MCAST_PORT
] = { .type
= NLA_U16
},
2836 [IPVS_DAEMON_ATTR_MCAST_TTL
] = { .type
= NLA_U8
},
2839 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2840 static const struct nla_policy ip_vs_svc_policy
[IPVS_SVC_ATTR_MAX
+ 1] = {
2841 [IPVS_SVC_ATTR_AF
] = { .type
= NLA_U16
},
2842 [IPVS_SVC_ATTR_PROTOCOL
] = { .type
= NLA_U16
},
2843 [IPVS_SVC_ATTR_ADDR
] = { .type
= NLA_BINARY
,
2844 .len
= sizeof(union nf_inet_addr
) },
2845 [IPVS_SVC_ATTR_PORT
] = { .type
= NLA_U16
},
2846 [IPVS_SVC_ATTR_FWMARK
] = { .type
= NLA_U32
},
2847 [IPVS_SVC_ATTR_SCHED_NAME
] = { .type
= NLA_NUL_STRING
,
2848 .len
= IP_VS_SCHEDNAME_MAXLEN
},
2849 [IPVS_SVC_ATTR_PE_NAME
] = { .type
= NLA_NUL_STRING
,
2850 .len
= IP_VS_PENAME_MAXLEN
},
2851 [IPVS_SVC_ATTR_FLAGS
] = { .type
= NLA_BINARY
,
2852 .len
= sizeof(struct ip_vs_flags
) },
2853 [IPVS_SVC_ATTR_TIMEOUT
] = { .type
= NLA_U32
},
2854 [IPVS_SVC_ATTR_NETMASK
] = { .type
= NLA_U32
},
2855 [IPVS_SVC_ATTR_STATS
] = { .type
= NLA_NESTED
},
2858 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2859 static const struct nla_policy ip_vs_dest_policy
[IPVS_DEST_ATTR_MAX
+ 1] = {
2860 [IPVS_DEST_ATTR_ADDR
] = { .type
= NLA_BINARY
,
2861 .len
= sizeof(union nf_inet_addr
) },
2862 [IPVS_DEST_ATTR_PORT
] = { .type
= NLA_U16
},
2863 [IPVS_DEST_ATTR_FWD_METHOD
] = { .type
= NLA_U32
},
2864 [IPVS_DEST_ATTR_WEIGHT
] = { .type
= NLA_U32
},
2865 [IPVS_DEST_ATTR_U_THRESH
] = { .type
= NLA_U32
},
2866 [IPVS_DEST_ATTR_L_THRESH
] = { .type
= NLA_U32
},
2867 [IPVS_DEST_ATTR_ACTIVE_CONNS
] = { .type
= NLA_U32
},
2868 [IPVS_DEST_ATTR_INACT_CONNS
] = { .type
= NLA_U32
},
2869 [IPVS_DEST_ATTR_PERSIST_CONNS
] = { .type
= NLA_U32
},
2870 [IPVS_DEST_ATTR_STATS
] = { .type
= NLA_NESTED
},
2871 [IPVS_DEST_ATTR_ADDR_FAMILY
] = { .type
= NLA_U16
},
2874 static int ip_vs_genl_fill_stats(struct sk_buff
*skb
, int container_type
,
2875 struct ip_vs_kstats
*kstats
)
2877 struct nlattr
*nl_stats
= nla_nest_start(skb
, container_type
);
2882 if (nla_put_u32(skb
, IPVS_STATS_ATTR_CONNS
, (u32
)kstats
->conns
) ||
2883 nla_put_u32(skb
, IPVS_STATS_ATTR_INPKTS
, (u32
)kstats
->inpkts
) ||
2884 nla_put_u32(skb
, IPVS_STATS_ATTR_OUTPKTS
, (u32
)kstats
->outpkts
) ||
2885 nla_put_u64(skb
, IPVS_STATS_ATTR_INBYTES
, kstats
->inbytes
) ||
2886 nla_put_u64(skb
, IPVS_STATS_ATTR_OUTBYTES
, kstats
->outbytes
) ||
2887 nla_put_u32(skb
, IPVS_STATS_ATTR_CPS
, (u32
)kstats
->cps
) ||
2888 nla_put_u32(skb
, IPVS_STATS_ATTR_INPPS
, (u32
)kstats
->inpps
) ||
2889 nla_put_u32(skb
, IPVS_STATS_ATTR_OUTPPS
, (u32
)kstats
->outpps
) ||
2890 nla_put_u32(skb
, IPVS_STATS_ATTR_INBPS
, (u32
)kstats
->inbps
) ||
2891 nla_put_u32(skb
, IPVS_STATS_ATTR_OUTBPS
, (u32
)kstats
->outbps
))
2892 goto nla_put_failure
;
2893 nla_nest_end(skb
, nl_stats
);
2898 nla_nest_cancel(skb
, nl_stats
);
2902 static int ip_vs_genl_fill_stats64(struct sk_buff
*skb
, int container_type
,
2903 struct ip_vs_kstats
*kstats
)
2905 struct nlattr
*nl_stats
= nla_nest_start(skb
, container_type
);
2910 if (nla_put_u64(skb
, IPVS_STATS_ATTR_CONNS
, kstats
->conns
) ||
2911 nla_put_u64(skb
, IPVS_STATS_ATTR_INPKTS
, kstats
->inpkts
) ||
2912 nla_put_u64(skb
, IPVS_STATS_ATTR_OUTPKTS
, kstats
->outpkts
) ||
2913 nla_put_u64(skb
, IPVS_STATS_ATTR_INBYTES
, kstats
->inbytes
) ||
2914 nla_put_u64(skb
, IPVS_STATS_ATTR_OUTBYTES
, kstats
->outbytes
) ||
2915 nla_put_u64(skb
, IPVS_STATS_ATTR_CPS
, kstats
->cps
) ||
2916 nla_put_u64(skb
, IPVS_STATS_ATTR_INPPS
, kstats
->inpps
) ||
2917 nla_put_u64(skb
, IPVS_STATS_ATTR_OUTPPS
, kstats
->outpps
) ||
2918 nla_put_u64(skb
, IPVS_STATS_ATTR_INBPS
, kstats
->inbps
) ||
2919 nla_put_u64(skb
, IPVS_STATS_ATTR_OUTBPS
, kstats
->outbps
))
2920 goto nla_put_failure
;
2921 nla_nest_end(skb
, nl_stats
);
2926 nla_nest_cancel(skb
, nl_stats
);
2930 static int ip_vs_genl_fill_service(struct sk_buff
*skb
,
2931 struct ip_vs_service
*svc
)
2933 struct ip_vs_scheduler
*sched
;
2934 struct ip_vs_pe
*pe
;
2935 struct nlattr
*nl_service
;
2936 struct ip_vs_flags flags
= { .flags
= svc
->flags
,
2938 struct ip_vs_kstats kstats
;
2941 nl_service
= nla_nest_start(skb
, IPVS_CMD_ATTR_SERVICE
);
2945 if (nla_put_u16(skb
, IPVS_SVC_ATTR_AF
, svc
->af
))
2946 goto nla_put_failure
;
2948 if (nla_put_u32(skb
, IPVS_SVC_ATTR_FWMARK
, svc
->fwmark
))
2949 goto nla_put_failure
;
2951 if (nla_put_u16(skb
, IPVS_SVC_ATTR_PROTOCOL
, svc
->protocol
) ||
2952 nla_put(skb
, IPVS_SVC_ATTR_ADDR
, sizeof(svc
->addr
), &svc
->addr
) ||
2953 nla_put_be16(skb
, IPVS_SVC_ATTR_PORT
, svc
->port
))
2954 goto nla_put_failure
;
2957 sched
= rcu_dereference_protected(svc
->scheduler
, 1);
2958 sched_name
= sched
? sched
->name
: "none";
2959 pe
= rcu_dereference_protected(svc
->pe
, 1);
2960 if (nla_put_string(skb
, IPVS_SVC_ATTR_SCHED_NAME
, sched_name
) ||
2961 (pe
&& nla_put_string(skb
, IPVS_SVC_ATTR_PE_NAME
, pe
->name
)) ||
2962 nla_put(skb
, IPVS_SVC_ATTR_FLAGS
, sizeof(flags
), &flags
) ||
2963 nla_put_u32(skb
, IPVS_SVC_ATTR_TIMEOUT
, svc
->timeout
/ HZ
) ||
2964 nla_put_be32(skb
, IPVS_SVC_ATTR_NETMASK
, svc
->netmask
))
2965 goto nla_put_failure
;
2966 ip_vs_copy_stats(&kstats
, &svc
->stats
);
2967 if (ip_vs_genl_fill_stats(skb
, IPVS_SVC_ATTR_STATS
, &kstats
))
2968 goto nla_put_failure
;
2969 if (ip_vs_genl_fill_stats64(skb
, IPVS_SVC_ATTR_STATS64
, &kstats
))
2970 goto nla_put_failure
;
2972 nla_nest_end(skb
, nl_service
);
2977 nla_nest_cancel(skb
, nl_service
);
2981 static int ip_vs_genl_dump_service(struct sk_buff
*skb
,
2982 struct ip_vs_service
*svc
,
2983 struct netlink_callback
*cb
)
2987 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
2988 &ip_vs_genl_family
, NLM_F_MULTI
,
2989 IPVS_CMD_NEW_SERVICE
);
2993 if (ip_vs_genl_fill_service(skb
, svc
) < 0)
2994 goto nla_put_failure
;
2996 genlmsg_end(skb
, hdr
);
3000 genlmsg_cancel(skb
, hdr
);
3004 static int ip_vs_genl_dump_services(struct sk_buff
*skb
,
3005 struct netlink_callback
*cb
)
3008 int start
= cb
->args
[0];
3009 struct ip_vs_service
*svc
;
3010 struct netns_ipvs
*ipvs
= net_ipvs(skb_sknet(skb
));
3012 mutex_lock(&__ip_vs_mutex
);
3013 for (i
= 0; i
< IP_VS_SVC_TAB_SIZE
; i
++) {
3014 hlist_for_each_entry(svc
, &ip_vs_svc_table
[i
], s_list
) {
3015 if (++idx
<= start
|| (svc
->ipvs
!= ipvs
))
3017 if (ip_vs_genl_dump_service(skb
, svc
, cb
) < 0) {
3019 goto nla_put_failure
;
3024 for (i
= 0; i
< IP_VS_SVC_TAB_SIZE
; i
++) {
3025 hlist_for_each_entry(svc
, &ip_vs_svc_fwm_table
[i
], f_list
) {
3026 if (++idx
<= start
|| (svc
->ipvs
!= ipvs
))
3028 if (ip_vs_genl_dump_service(skb
, svc
, cb
) < 0) {
3030 goto nla_put_failure
;
3036 mutex_unlock(&__ip_vs_mutex
);
3042 static int ip_vs_genl_parse_service(struct net
*net
,
3043 struct ip_vs_service_user_kern
*usvc
,
3044 struct nlattr
*nla
, int full_entry
,
3045 struct ip_vs_service
**ret_svc
)
3047 struct netns_ipvs
*ipvs
= net_ipvs(net
);
3048 struct nlattr
*attrs
[IPVS_SVC_ATTR_MAX
+ 1];
3049 struct nlattr
*nla_af
, *nla_port
, *nla_fwmark
, *nla_protocol
, *nla_addr
;
3050 struct ip_vs_service
*svc
;
3052 /* Parse mandatory identifying service fields first */
3054 nla_parse_nested(attrs
, IPVS_SVC_ATTR_MAX
, nla
, ip_vs_svc_policy
))
3057 nla_af
= attrs
[IPVS_SVC_ATTR_AF
];
3058 nla_protocol
= attrs
[IPVS_SVC_ATTR_PROTOCOL
];
3059 nla_addr
= attrs
[IPVS_SVC_ATTR_ADDR
];
3060 nla_port
= attrs
[IPVS_SVC_ATTR_PORT
];
3061 nla_fwmark
= attrs
[IPVS_SVC_ATTR_FWMARK
];
3063 if (!(nla_af
&& (nla_fwmark
|| (nla_port
&& nla_protocol
&& nla_addr
))))
3066 memset(usvc
, 0, sizeof(*usvc
));
3068 usvc
->af
= nla_get_u16(nla_af
);
3069 #ifdef CONFIG_IP_VS_IPV6
3070 if (usvc
->af
!= AF_INET
&& usvc
->af
!= AF_INET6
)
3072 if (usvc
->af
!= AF_INET
)
3074 return -EAFNOSUPPORT
;
3077 usvc
->protocol
= IPPROTO_TCP
;
3078 usvc
->fwmark
= nla_get_u32(nla_fwmark
);
3080 usvc
->protocol
= nla_get_u16(nla_protocol
);
3081 nla_memcpy(&usvc
->addr
, nla_addr
, sizeof(usvc
->addr
));
3082 usvc
->port
= nla_get_be16(nla_port
);
3088 svc
= __ip_vs_svc_fwm_find(ipvs
, usvc
->af
, usvc
->fwmark
);
3090 svc
= __ip_vs_service_find(ipvs
, usvc
->af
, usvc
->protocol
,
3091 &usvc
->addr
, usvc
->port
);
3095 /* If a full entry was requested, check for the additional fields */
3097 struct nlattr
*nla_sched
, *nla_flags
, *nla_pe
, *nla_timeout
,
3099 struct ip_vs_flags flags
;
3101 nla_sched
= attrs
[IPVS_SVC_ATTR_SCHED_NAME
];
3102 nla_pe
= attrs
[IPVS_SVC_ATTR_PE_NAME
];
3103 nla_flags
= attrs
[IPVS_SVC_ATTR_FLAGS
];
3104 nla_timeout
= attrs
[IPVS_SVC_ATTR_TIMEOUT
];
3105 nla_netmask
= attrs
[IPVS_SVC_ATTR_NETMASK
];
3107 if (!(nla_sched
&& nla_flags
&& nla_timeout
&& nla_netmask
))
3110 nla_memcpy(&flags
, nla_flags
, sizeof(flags
));
3112 /* prefill flags from service if it already exists */
3114 usvc
->flags
= svc
->flags
;
3116 /* set new flags from userland */
3117 usvc
->flags
= (usvc
->flags
& ~flags
.mask
) |
3118 (flags
.flags
& flags
.mask
);
3119 usvc
->sched_name
= nla_data(nla_sched
);
3120 usvc
->pe_name
= nla_pe
? nla_data(nla_pe
) : NULL
;
3121 usvc
->timeout
= nla_get_u32(nla_timeout
);
3122 usvc
->netmask
= nla_get_be32(nla_netmask
);
3128 static struct ip_vs_service
*ip_vs_genl_find_service(struct net
*net
,
3131 struct ip_vs_service_user_kern usvc
;
3132 struct ip_vs_service
*svc
;
3135 ret
= ip_vs_genl_parse_service(net
, &usvc
, nla
, 0, &svc
);
3136 return ret
? ERR_PTR(ret
) : svc
;
3139 static int ip_vs_genl_fill_dest(struct sk_buff
*skb
, struct ip_vs_dest
*dest
)
3141 struct nlattr
*nl_dest
;
3142 struct ip_vs_kstats kstats
;
3144 nl_dest
= nla_nest_start(skb
, IPVS_CMD_ATTR_DEST
);
3148 if (nla_put(skb
, IPVS_DEST_ATTR_ADDR
, sizeof(dest
->addr
), &dest
->addr
) ||
3149 nla_put_be16(skb
, IPVS_DEST_ATTR_PORT
, dest
->port
) ||
3150 nla_put_u32(skb
, IPVS_DEST_ATTR_FWD_METHOD
,
3151 (atomic_read(&dest
->conn_flags
) &
3152 IP_VS_CONN_F_FWD_MASK
)) ||
3153 nla_put_u32(skb
, IPVS_DEST_ATTR_WEIGHT
,
3154 atomic_read(&dest
->weight
)) ||
3155 nla_put_u32(skb
, IPVS_DEST_ATTR_U_THRESH
, dest
->u_threshold
) ||
3156 nla_put_u32(skb
, IPVS_DEST_ATTR_L_THRESH
, dest
->l_threshold
) ||
3157 nla_put_u32(skb
, IPVS_DEST_ATTR_ACTIVE_CONNS
,
3158 atomic_read(&dest
->activeconns
)) ||
3159 nla_put_u32(skb
, IPVS_DEST_ATTR_INACT_CONNS
,
3160 atomic_read(&dest
->inactconns
)) ||
3161 nla_put_u32(skb
, IPVS_DEST_ATTR_PERSIST_CONNS
,
3162 atomic_read(&dest
->persistconns
)) ||
3163 nla_put_u16(skb
, IPVS_DEST_ATTR_ADDR_FAMILY
, dest
->af
))
3164 goto nla_put_failure
;
3165 ip_vs_copy_stats(&kstats
, &dest
->stats
);
3166 if (ip_vs_genl_fill_stats(skb
, IPVS_DEST_ATTR_STATS
, &kstats
))
3167 goto nla_put_failure
;
3168 if (ip_vs_genl_fill_stats64(skb
, IPVS_DEST_ATTR_STATS64
, &kstats
))
3169 goto nla_put_failure
;
3171 nla_nest_end(skb
, nl_dest
);
3176 nla_nest_cancel(skb
, nl_dest
);
3180 static int ip_vs_genl_dump_dest(struct sk_buff
*skb
, struct ip_vs_dest
*dest
,
3181 struct netlink_callback
*cb
)
3185 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
3186 &ip_vs_genl_family
, NLM_F_MULTI
,
3191 if (ip_vs_genl_fill_dest(skb
, dest
) < 0)
3192 goto nla_put_failure
;
3194 genlmsg_end(skb
, hdr
);
3198 genlmsg_cancel(skb
, hdr
);
3202 static int ip_vs_genl_dump_dests(struct sk_buff
*skb
,
3203 struct netlink_callback
*cb
)
3206 int start
= cb
->args
[0];
3207 struct ip_vs_service
*svc
;
3208 struct ip_vs_dest
*dest
;
3209 struct nlattr
*attrs
[IPVS_CMD_ATTR_MAX
+ 1];
3210 struct net
*net
= skb_sknet(skb
);
3212 mutex_lock(&__ip_vs_mutex
);
3214 /* Try to find the service for which to dump destinations */
3215 if (nlmsg_parse(cb
->nlh
, GENL_HDRLEN
, attrs
,
3216 IPVS_CMD_ATTR_MAX
, ip_vs_cmd_policy
))
3220 svc
= ip_vs_genl_find_service(net
, attrs
[IPVS_CMD_ATTR_SERVICE
]);
3221 if (IS_ERR(svc
) || svc
== NULL
)
3224 /* Dump the destinations */
3225 list_for_each_entry(dest
, &svc
->destinations
, n_list
) {
3228 if (ip_vs_genl_dump_dest(skb
, dest
, cb
) < 0) {
3230 goto nla_put_failure
;
3238 mutex_unlock(&__ip_vs_mutex
);
3243 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern
*udest
,
3244 struct nlattr
*nla
, int full_entry
)
3246 struct nlattr
*attrs
[IPVS_DEST_ATTR_MAX
+ 1];
3247 struct nlattr
*nla_addr
, *nla_port
;
3248 struct nlattr
*nla_addr_family
;
3250 /* Parse mandatory identifying destination fields first */
3252 nla_parse_nested(attrs
, IPVS_DEST_ATTR_MAX
, nla
, ip_vs_dest_policy
))
3255 nla_addr
= attrs
[IPVS_DEST_ATTR_ADDR
];
3256 nla_port
= attrs
[IPVS_DEST_ATTR_PORT
];
3257 nla_addr_family
= attrs
[IPVS_DEST_ATTR_ADDR_FAMILY
];
3259 if (!(nla_addr
&& nla_port
))
3262 memset(udest
, 0, sizeof(*udest
));
3264 nla_memcpy(&udest
->addr
, nla_addr
, sizeof(udest
->addr
));
3265 udest
->port
= nla_get_be16(nla_port
);
3267 if (nla_addr_family
)
3268 udest
->af
= nla_get_u16(nla_addr_family
);
3272 /* If a full entry was requested, check for the additional fields */
3274 struct nlattr
*nla_fwd
, *nla_weight
, *nla_u_thresh
,
3277 nla_fwd
= attrs
[IPVS_DEST_ATTR_FWD_METHOD
];
3278 nla_weight
= attrs
[IPVS_DEST_ATTR_WEIGHT
];
3279 nla_u_thresh
= attrs
[IPVS_DEST_ATTR_U_THRESH
];
3280 nla_l_thresh
= attrs
[IPVS_DEST_ATTR_L_THRESH
];
3282 if (!(nla_fwd
&& nla_weight
&& nla_u_thresh
&& nla_l_thresh
))
3285 udest
->conn_flags
= nla_get_u32(nla_fwd
)
3286 & IP_VS_CONN_F_FWD_MASK
;
3287 udest
->weight
= nla_get_u32(nla_weight
);
3288 udest
->u_threshold
= nla_get_u32(nla_u_thresh
);
3289 udest
->l_threshold
= nla_get_u32(nla_l_thresh
);
3295 static int ip_vs_genl_fill_daemon(struct sk_buff
*skb
, __u32 state
,
3296 struct ipvs_sync_daemon_cfg
*c
)
3298 struct nlattr
*nl_daemon
;
3300 nl_daemon
= nla_nest_start(skb
, IPVS_CMD_ATTR_DAEMON
);
3304 if (nla_put_u32(skb
, IPVS_DAEMON_ATTR_STATE
, state
) ||
3305 nla_put_string(skb
, IPVS_DAEMON_ATTR_MCAST_IFN
, c
->mcast_ifn
) ||
3306 nla_put_u32(skb
, IPVS_DAEMON_ATTR_SYNC_ID
, c
->syncid
) ||
3307 nla_put_u16(skb
, IPVS_DAEMON_ATTR_SYNC_MAXLEN
, c
->sync_maxlen
) ||
3308 nla_put_u16(skb
, IPVS_DAEMON_ATTR_MCAST_PORT
, c
->mcast_port
) ||
3309 nla_put_u8(skb
, IPVS_DAEMON_ATTR_MCAST_TTL
, c
->mcast_ttl
))
3310 goto nla_put_failure
;
3311 #ifdef CONFIG_IP_VS_IPV6
3312 if (c
->mcast_af
== AF_INET6
) {
3313 if (nla_put_in6_addr(skb
, IPVS_DAEMON_ATTR_MCAST_GROUP6
,
3314 &c
->mcast_group
.in6
))
3315 goto nla_put_failure
;
3318 if (c
->mcast_af
== AF_INET
&&
3319 nla_put_in_addr(skb
, IPVS_DAEMON_ATTR_MCAST_GROUP
,
3321 goto nla_put_failure
;
3322 nla_nest_end(skb
, nl_daemon
);
3327 nla_nest_cancel(skb
, nl_daemon
);
3331 static int ip_vs_genl_dump_daemon(struct sk_buff
*skb
, __u32 state
,
3332 struct ipvs_sync_daemon_cfg
*c
,
3333 struct netlink_callback
*cb
)
3336 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
, cb
->nlh
->nlmsg_seq
,
3337 &ip_vs_genl_family
, NLM_F_MULTI
,
3338 IPVS_CMD_NEW_DAEMON
);
3342 if (ip_vs_genl_fill_daemon(skb
, state
, c
))
3343 goto nla_put_failure
;
3345 genlmsg_end(skb
, hdr
);
3349 genlmsg_cancel(skb
, hdr
);
3353 static int ip_vs_genl_dump_daemons(struct sk_buff
*skb
,
3354 struct netlink_callback
*cb
)
3356 struct net
*net
= skb_sknet(skb
);
3357 struct netns_ipvs
*ipvs
= net_ipvs(net
);
3359 mutex_lock(&ipvs
->sync_mutex
);
3360 if ((ipvs
->sync_state
& IP_VS_STATE_MASTER
) && !cb
->args
[0]) {
3361 if (ip_vs_genl_dump_daemon(skb
, IP_VS_STATE_MASTER
,
3362 &ipvs
->mcfg
, cb
) < 0)
3363 goto nla_put_failure
;
3368 if ((ipvs
->sync_state
& IP_VS_STATE_BACKUP
) && !cb
->args
[1]) {
3369 if (ip_vs_genl_dump_daemon(skb
, IP_VS_STATE_BACKUP
,
3370 &ipvs
->bcfg
, cb
) < 0)
3371 goto nla_put_failure
;
3377 mutex_unlock(&ipvs
->sync_mutex
);
3382 static int ip_vs_genl_new_daemon(struct net
*net
, struct nlattr
**attrs
)
3384 struct netns_ipvs
*ipvs
= net_ipvs(net
);
3385 struct ipvs_sync_daemon_cfg c
;
3389 memset(&c
, 0, sizeof(c
));
3390 if (!(attrs
[IPVS_DAEMON_ATTR_STATE
] &&
3391 attrs
[IPVS_DAEMON_ATTR_MCAST_IFN
] &&
3392 attrs
[IPVS_DAEMON_ATTR_SYNC_ID
]))
3394 strlcpy(c
.mcast_ifn
, nla_data(attrs
[IPVS_DAEMON_ATTR_MCAST_IFN
]),
3395 sizeof(c
.mcast_ifn
));
3396 c
.syncid
= nla_get_u32(attrs
[IPVS_DAEMON_ATTR_SYNC_ID
]);
3398 a
= attrs
[IPVS_DAEMON_ATTR_SYNC_MAXLEN
];
3400 c
.sync_maxlen
= nla_get_u16(a
);
3402 a
= attrs
[IPVS_DAEMON_ATTR_MCAST_GROUP
];
3404 c
.mcast_af
= AF_INET
;
3405 c
.mcast_group
.ip
= nla_get_in_addr(a
);
3406 if (!ipv4_is_multicast(c
.mcast_group
.ip
))
3409 a
= attrs
[IPVS_DAEMON_ATTR_MCAST_GROUP6
];
3411 #ifdef CONFIG_IP_VS_IPV6
3414 c
.mcast_af
= AF_INET6
;
3415 c
.mcast_group
.in6
= nla_get_in6_addr(a
);
3416 addr_type
= ipv6_addr_type(&c
.mcast_group
.in6
);
3417 if (!(addr_type
& IPV6_ADDR_MULTICAST
))
3420 return -EAFNOSUPPORT
;
3425 a
= attrs
[IPVS_DAEMON_ATTR_MCAST_PORT
];
3427 c
.mcast_port
= nla_get_u16(a
);
3429 a
= attrs
[IPVS_DAEMON_ATTR_MCAST_TTL
];
3431 c
.mcast_ttl
= nla_get_u8(a
);
3433 /* The synchronization protocol is incompatible with mixed family
3436 if (ipvs
->mixed_address_family_dests
> 0)
3440 mutex_lock(&ipvs
->sync_mutex
);
3441 ret
= start_sync_thread(net
, &c
,
3442 nla_get_u32(attrs
[IPVS_DAEMON_ATTR_STATE
]));
3443 mutex_unlock(&ipvs
->sync_mutex
);
3448 static int ip_vs_genl_del_daemon(struct net
*net
, struct nlattr
**attrs
)
3450 struct netns_ipvs
*ipvs
= net_ipvs(net
);
3453 if (!attrs
[IPVS_DAEMON_ATTR_STATE
])
3456 mutex_lock(&ipvs
->sync_mutex
);
3457 ret
= stop_sync_thread(net
,
3458 nla_get_u32(attrs
[IPVS_DAEMON_ATTR_STATE
]));
3459 mutex_unlock(&ipvs
->sync_mutex
);
3463 static int ip_vs_genl_set_config(struct net
*net
, struct nlattr
**attrs
)
3465 struct ip_vs_timeout_user t
;
3467 __ip_vs_get_timeouts(net
, &t
);
3469 if (attrs
[IPVS_CMD_ATTR_TIMEOUT_TCP
])
3470 t
.tcp_timeout
= nla_get_u32(attrs
[IPVS_CMD_ATTR_TIMEOUT_TCP
]);
3472 if (attrs
[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN
])
3474 nla_get_u32(attrs
[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN
]);
3476 if (attrs
[IPVS_CMD_ATTR_TIMEOUT_UDP
])
3477 t
.udp_timeout
= nla_get_u32(attrs
[IPVS_CMD_ATTR_TIMEOUT_UDP
]);
3479 return ip_vs_set_timeout(net
, &t
);
3482 static int ip_vs_genl_set_daemon(struct sk_buff
*skb
, struct genl_info
*info
)
3484 int ret
= -EINVAL
, cmd
;
3486 struct netns_ipvs
*ipvs
;
3488 net
= skb_sknet(skb
);
3489 ipvs
= net_ipvs(net
);
3490 cmd
= info
->genlhdr
->cmd
;
3492 if (cmd
== IPVS_CMD_NEW_DAEMON
|| cmd
== IPVS_CMD_DEL_DAEMON
) {
3493 struct nlattr
*daemon_attrs
[IPVS_DAEMON_ATTR_MAX
+ 1];
3495 if (!info
->attrs
[IPVS_CMD_ATTR_DAEMON
] ||
3496 nla_parse_nested(daemon_attrs
, IPVS_DAEMON_ATTR_MAX
,
3497 info
->attrs
[IPVS_CMD_ATTR_DAEMON
],
3498 ip_vs_daemon_policy
))
3501 if (cmd
== IPVS_CMD_NEW_DAEMON
)
3502 ret
= ip_vs_genl_new_daemon(net
, daemon_attrs
);
3504 ret
= ip_vs_genl_del_daemon(net
, daemon_attrs
);
3511 static int ip_vs_genl_set_cmd(struct sk_buff
*skb
, struct genl_info
*info
)
3513 struct ip_vs_service
*svc
= NULL
;
3514 struct ip_vs_service_user_kern usvc
;
3515 struct ip_vs_dest_user_kern udest
;
3517 int need_full_svc
= 0, need_full_dest
= 0;
3519 struct netns_ipvs
*ipvs
;
3521 net
= skb_sknet(skb
);
3522 ipvs
= net_ipvs(net
);
3523 cmd
= info
->genlhdr
->cmd
;
3525 mutex_lock(&__ip_vs_mutex
);
3527 if (cmd
== IPVS_CMD_FLUSH
) {
3528 ret
= ip_vs_flush(net
, false);
3530 } else if (cmd
== IPVS_CMD_SET_CONFIG
) {
3531 ret
= ip_vs_genl_set_config(net
, info
->attrs
);
3533 } else if (cmd
== IPVS_CMD_ZERO
&&
3534 !info
->attrs
[IPVS_CMD_ATTR_SERVICE
]) {
3535 ret
= ip_vs_zero_all(net
);
3539 /* All following commands require a service argument, so check if we
3540 * received a valid one. We need a full service specification when
3541 * adding / editing a service. Only identifying members otherwise. */
3542 if (cmd
== IPVS_CMD_NEW_SERVICE
|| cmd
== IPVS_CMD_SET_SERVICE
)
3545 ret
= ip_vs_genl_parse_service(net
, &usvc
,
3546 info
->attrs
[IPVS_CMD_ATTR_SERVICE
],
3547 need_full_svc
, &svc
);
3551 /* Unless we're adding a new service, the service must already exist */
3552 if ((cmd
!= IPVS_CMD_NEW_SERVICE
) && (svc
== NULL
)) {
3557 /* Destination commands require a valid destination argument. For
3558 * adding / editing a destination, we need a full destination
3560 if (cmd
== IPVS_CMD_NEW_DEST
|| cmd
== IPVS_CMD_SET_DEST
||
3561 cmd
== IPVS_CMD_DEL_DEST
) {
3562 if (cmd
!= IPVS_CMD_DEL_DEST
)
3565 ret
= ip_vs_genl_parse_dest(&udest
,
3566 info
->attrs
[IPVS_CMD_ATTR_DEST
],
3571 /* Old protocols did not allow the user to specify address
3572 * family, so we set it to zero instead. We also didn't
3573 * allow heterogeneous pools in the old code, so it's safe
3574 * to assume that this will have the same address family as
3580 if (udest
.af
!= svc
->af
&& cmd
!= IPVS_CMD_DEL_DEST
) {
3581 /* The synchronization protocol is incompatible
3582 * with mixed family services
3584 if (ipvs
->sync_state
) {
3589 /* Which connection types do we support? */
3590 switch (udest
.conn_flags
) {
3591 case IP_VS_CONN_F_TUNNEL
:
3592 /* We are able to forward this */
3602 case IPVS_CMD_NEW_SERVICE
:
3604 ret
= ip_vs_add_service(net
, &usvc
, &svc
);
3608 case IPVS_CMD_SET_SERVICE
:
3609 ret
= ip_vs_edit_service(svc
, &usvc
);
3611 case IPVS_CMD_DEL_SERVICE
:
3612 ret
= ip_vs_del_service(svc
);
3613 /* do not use svc, it can be freed */
3615 case IPVS_CMD_NEW_DEST
:
3616 ret
= ip_vs_add_dest(svc
, &udest
);
3618 case IPVS_CMD_SET_DEST
:
3619 ret
= ip_vs_edit_dest(svc
, &udest
);
3621 case IPVS_CMD_DEL_DEST
:
3622 ret
= ip_vs_del_dest(svc
, &udest
);
3625 ret
= ip_vs_zero_service(svc
);
3632 mutex_unlock(&__ip_vs_mutex
);
3637 static int ip_vs_genl_get_cmd(struct sk_buff
*skb
, struct genl_info
*info
)
3639 struct sk_buff
*msg
;
3641 int ret
, cmd
, reply_cmd
;
3644 net
= skb_sknet(skb
);
3645 cmd
= info
->genlhdr
->cmd
;
3647 if (cmd
== IPVS_CMD_GET_SERVICE
)
3648 reply_cmd
= IPVS_CMD_NEW_SERVICE
;
3649 else if (cmd
== IPVS_CMD_GET_INFO
)
3650 reply_cmd
= IPVS_CMD_SET_INFO
;
3651 else if (cmd
== IPVS_CMD_GET_CONFIG
)
3652 reply_cmd
= IPVS_CMD_SET_CONFIG
;
3654 pr_err("unknown Generic Netlink command\n");
3658 msg
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
3662 mutex_lock(&__ip_vs_mutex
);
3664 reply
= genlmsg_put_reply(msg
, info
, &ip_vs_genl_family
, 0, reply_cmd
);
3666 goto nla_put_failure
;
3669 case IPVS_CMD_GET_SERVICE
:
3671 struct ip_vs_service
*svc
;
3673 svc
= ip_vs_genl_find_service(net
,
3674 info
->attrs
[IPVS_CMD_ATTR_SERVICE
]);
3679 ret
= ip_vs_genl_fill_service(msg
, svc
);
3681 goto nla_put_failure
;
3690 case IPVS_CMD_GET_CONFIG
:
3692 struct ip_vs_timeout_user t
;
3694 __ip_vs_get_timeouts(net
, &t
);
3695 #ifdef CONFIG_IP_VS_PROTO_TCP
3696 if (nla_put_u32(msg
, IPVS_CMD_ATTR_TIMEOUT_TCP
,
3698 nla_put_u32(msg
, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN
,
3700 goto nla_put_failure
;
3702 #ifdef CONFIG_IP_VS_PROTO_UDP
3703 if (nla_put_u32(msg
, IPVS_CMD_ATTR_TIMEOUT_UDP
, t
.udp_timeout
))
3704 goto nla_put_failure
;
3710 case IPVS_CMD_GET_INFO
:
3711 if (nla_put_u32(msg
, IPVS_INFO_ATTR_VERSION
,
3712 IP_VS_VERSION_CODE
) ||
3713 nla_put_u32(msg
, IPVS_INFO_ATTR_CONN_TAB_SIZE
,
3714 ip_vs_conn_tab_size
))
3715 goto nla_put_failure
;
3719 genlmsg_end(msg
, reply
);
3720 ret
= genlmsg_reply(msg
, info
);
3724 pr_err("not enough space in Netlink message\n");
3730 mutex_unlock(&__ip_vs_mutex
);
3736 static const struct genl_ops ip_vs_genl_ops
[] = {
3738 .cmd
= IPVS_CMD_NEW_SERVICE
,
3739 .flags
= GENL_ADMIN_PERM
,
3740 .policy
= ip_vs_cmd_policy
,
3741 .doit
= ip_vs_genl_set_cmd
,
3744 .cmd
= IPVS_CMD_SET_SERVICE
,
3745 .flags
= GENL_ADMIN_PERM
,
3746 .policy
= ip_vs_cmd_policy
,
3747 .doit
= ip_vs_genl_set_cmd
,
3750 .cmd
= IPVS_CMD_DEL_SERVICE
,
3751 .flags
= GENL_ADMIN_PERM
,
3752 .policy
= ip_vs_cmd_policy
,
3753 .doit
= ip_vs_genl_set_cmd
,
3756 .cmd
= IPVS_CMD_GET_SERVICE
,
3757 .flags
= GENL_ADMIN_PERM
,
3758 .doit
= ip_vs_genl_get_cmd
,
3759 .dumpit
= ip_vs_genl_dump_services
,
3760 .policy
= ip_vs_cmd_policy
,
3763 .cmd
= IPVS_CMD_NEW_DEST
,
3764 .flags
= GENL_ADMIN_PERM
,
3765 .policy
= ip_vs_cmd_policy
,
3766 .doit
= ip_vs_genl_set_cmd
,
3769 .cmd
= IPVS_CMD_SET_DEST
,
3770 .flags
= GENL_ADMIN_PERM
,
3771 .policy
= ip_vs_cmd_policy
,
3772 .doit
= ip_vs_genl_set_cmd
,
3775 .cmd
= IPVS_CMD_DEL_DEST
,
3776 .flags
= GENL_ADMIN_PERM
,
3777 .policy
= ip_vs_cmd_policy
,
3778 .doit
= ip_vs_genl_set_cmd
,
3781 .cmd
= IPVS_CMD_GET_DEST
,
3782 .flags
= GENL_ADMIN_PERM
,
3783 .policy
= ip_vs_cmd_policy
,
3784 .dumpit
= ip_vs_genl_dump_dests
,
3787 .cmd
= IPVS_CMD_NEW_DAEMON
,
3788 .flags
= GENL_ADMIN_PERM
,
3789 .policy
= ip_vs_cmd_policy
,
3790 .doit
= ip_vs_genl_set_daemon
,
3793 .cmd
= IPVS_CMD_DEL_DAEMON
,
3794 .flags
= GENL_ADMIN_PERM
,
3795 .policy
= ip_vs_cmd_policy
,
3796 .doit
= ip_vs_genl_set_daemon
,
3799 .cmd
= IPVS_CMD_GET_DAEMON
,
3800 .flags
= GENL_ADMIN_PERM
,
3801 .dumpit
= ip_vs_genl_dump_daemons
,
3804 .cmd
= IPVS_CMD_SET_CONFIG
,
3805 .flags
= GENL_ADMIN_PERM
,
3806 .policy
= ip_vs_cmd_policy
,
3807 .doit
= ip_vs_genl_set_cmd
,
3810 .cmd
= IPVS_CMD_GET_CONFIG
,
3811 .flags
= GENL_ADMIN_PERM
,
3812 .doit
= ip_vs_genl_get_cmd
,
3815 .cmd
= IPVS_CMD_GET_INFO
,
3816 .flags
= GENL_ADMIN_PERM
,
3817 .doit
= ip_vs_genl_get_cmd
,
3820 .cmd
= IPVS_CMD_ZERO
,
3821 .flags
= GENL_ADMIN_PERM
,
3822 .policy
= ip_vs_cmd_policy
,
3823 .doit
= ip_vs_genl_set_cmd
,
3826 .cmd
= IPVS_CMD_FLUSH
,
3827 .flags
= GENL_ADMIN_PERM
,
3828 .doit
= ip_vs_genl_set_cmd
,
3832 static int __init
ip_vs_genl_register(void)
3834 return genl_register_family_with_ops(&ip_vs_genl_family
,
3838 static void ip_vs_genl_unregister(void)
3840 genl_unregister_family(&ip_vs_genl_family
);
3843 /* End of Generic Netlink interface definitions */
3846 * per netns intit/exit func.
3848 #ifdef CONFIG_SYSCTL
3849 static int __net_init
ip_vs_control_net_init_sysctl(struct net
*net
)
3852 struct netns_ipvs
*ipvs
= net_ipvs(net
);
3853 struct ctl_table
*tbl
;
3855 atomic_set(&ipvs
->dropentry
, 0);
3856 spin_lock_init(&ipvs
->dropentry_lock
);
3857 spin_lock_init(&ipvs
->droppacket_lock
);
3858 spin_lock_init(&ipvs
->securetcp_lock
);
3860 if (!net_eq(net
, &init_net
)) {
3861 tbl
= kmemdup(vs_vars
, sizeof(vs_vars
), GFP_KERNEL
);
3865 /* Don't export sysctls to unprivileged users */
3866 if (net
->user_ns
!= &init_user_ns
)
3867 tbl
[0].procname
= NULL
;
3870 /* Initialize sysctl defaults */
3871 for (idx
= 0; idx
< ARRAY_SIZE(vs_vars
); idx
++) {
3872 if (tbl
[idx
].proc_handler
== proc_do_defense_mode
)
3873 tbl
[idx
].extra2
= ipvs
;
3876 ipvs
->sysctl_amemthresh
= 1024;
3877 tbl
[idx
++].data
= &ipvs
->sysctl_amemthresh
;
3878 ipvs
->sysctl_am_droprate
= 10;
3879 tbl
[idx
++].data
= &ipvs
->sysctl_am_droprate
;
3880 tbl
[idx
++].data
= &ipvs
->sysctl_drop_entry
;
3881 tbl
[idx
++].data
= &ipvs
->sysctl_drop_packet
;
3882 #ifdef CONFIG_IP_VS_NFCT
3883 tbl
[idx
++].data
= &ipvs
->sysctl_conntrack
;
3885 tbl
[idx
++].data
= &ipvs
->sysctl_secure_tcp
;
3886 ipvs
->sysctl_snat_reroute
= 1;
3887 tbl
[idx
++].data
= &ipvs
->sysctl_snat_reroute
;
3888 ipvs
->sysctl_sync_ver
= 1;
3889 tbl
[idx
++].data
= &ipvs
->sysctl_sync_ver
;
3890 ipvs
->sysctl_sync_ports
= 1;
3891 tbl
[idx
++].data
= &ipvs
->sysctl_sync_ports
;
3892 tbl
[idx
++].data
= &ipvs
->sysctl_sync_persist_mode
;
3893 ipvs
->sysctl_sync_qlen_max
= nr_free_buffer_pages() / 32;
3894 tbl
[idx
++].data
= &ipvs
->sysctl_sync_qlen_max
;
3895 ipvs
->sysctl_sync_sock_size
= 0;
3896 tbl
[idx
++].data
= &ipvs
->sysctl_sync_sock_size
;
3897 tbl
[idx
++].data
= &ipvs
->sysctl_cache_bypass
;
3898 tbl
[idx
++].data
= &ipvs
->sysctl_expire_nodest_conn
;
3899 tbl
[idx
++].data
= &ipvs
->sysctl_sloppy_tcp
;
3900 tbl
[idx
++].data
= &ipvs
->sysctl_sloppy_sctp
;
3901 tbl
[idx
++].data
= &ipvs
->sysctl_expire_quiescent_template
;
3902 ipvs
->sysctl_sync_threshold
[0] = DEFAULT_SYNC_THRESHOLD
;
3903 ipvs
->sysctl_sync_threshold
[1] = DEFAULT_SYNC_PERIOD
;
3904 tbl
[idx
].data
= &ipvs
->sysctl_sync_threshold
;
3905 tbl
[idx
++].maxlen
= sizeof(ipvs
->sysctl_sync_threshold
);
3906 ipvs
->sysctl_sync_refresh_period
= DEFAULT_SYNC_REFRESH_PERIOD
;
3907 tbl
[idx
++].data
= &ipvs
->sysctl_sync_refresh_period
;
3908 ipvs
->sysctl_sync_retries
= clamp_t(int, DEFAULT_SYNC_RETRIES
, 0, 3);
3909 tbl
[idx
++].data
= &ipvs
->sysctl_sync_retries
;
3910 tbl
[idx
++].data
= &ipvs
->sysctl_nat_icmp_send
;
3911 ipvs
->sysctl_pmtu_disc
= 1;
3912 tbl
[idx
++].data
= &ipvs
->sysctl_pmtu_disc
;
3913 tbl
[idx
++].data
= &ipvs
->sysctl_backup_only
;
3914 ipvs
->sysctl_conn_reuse_mode
= 1;
3915 tbl
[idx
++].data
= &ipvs
->sysctl_conn_reuse_mode
;
3916 tbl
[idx
++].data
= &ipvs
->sysctl_schedule_icmp
;
3917 tbl
[idx
++].data
= &ipvs
->sysctl_ignore_tunneled
;
3919 ipvs
->sysctl_hdr
= register_net_sysctl(net
, "net/ipv4/vs", tbl
);
3920 if (ipvs
->sysctl_hdr
== NULL
) {
3921 if (!net_eq(net
, &init_net
))
3925 ip_vs_start_estimator(net
, &ipvs
->tot_stats
);
3926 ipvs
->sysctl_tbl
= tbl
;
3927 /* Schedule defense work */
3928 INIT_DELAYED_WORK(&ipvs
->defense_work
, defense_work_handler
);
3929 schedule_delayed_work(&ipvs
->defense_work
, DEFENSE_TIMER_PERIOD
);
3934 static void __net_exit
ip_vs_control_net_cleanup_sysctl(struct net
*net
)
3936 struct netns_ipvs
*ipvs
= net_ipvs(net
);
3938 cancel_delayed_work_sync(&ipvs
->defense_work
);
3939 cancel_work_sync(&ipvs
->defense_work
.work
);
3940 unregister_net_sysctl_table(ipvs
->sysctl_hdr
);
3941 ip_vs_stop_estimator(net
, &ipvs
->tot_stats
);
3943 if (!net_eq(net
, &init_net
))
3944 kfree(ipvs
->sysctl_tbl
);
3949 static int __net_init
ip_vs_control_net_init_sysctl(struct net
*net
) { return 0; }
3950 static void __net_exit
ip_vs_control_net_cleanup_sysctl(struct net
*net
) { }
3954 static struct notifier_block ip_vs_dst_notifier
= {
3955 .notifier_call
= ip_vs_dst_event
,
3958 int __net_init
ip_vs_control_net_init(struct net
*net
)
3961 struct netns_ipvs
*ipvs
= net_ipvs(net
);
3963 /* Initialize rs_table */
3964 for (idx
= 0; idx
< IP_VS_RTAB_SIZE
; idx
++)
3965 INIT_HLIST_HEAD(&ipvs
->rs_table
[idx
]);
3967 INIT_LIST_HEAD(&ipvs
->dest_trash
);
3968 spin_lock_init(&ipvs
->dest_trash_lock
);
3969 setup_timer(&ipvs
->dest_trash_timer
, ip_vs_dest_trash_expire
,
3970 (unsigned long) ipvs
);
3971 atomic_set(&ipvs
->ftpsvc_counter
, 0);
3972 atomic_set(&ipvs
->nullsvc_counter
, 0);
3975 ipvs
->tot_stats
.cpustats
= alloc_percpu(struct ip_vs_cpu_stats
);
3976 if (!ipvs
->tot_stats
.cpustats
)
3979 for_each_possible_cpu(i
) {
3980 struct ip_vs_cpu_stats
*ipvs_tot_stats
;
3981 ipvs_tot_stats
= per_cpu_ptr(ipvs
->tot_stats
.cpustats
, i
);
3982 u64_stats_init(&ipvs_tot_stats
->syncp
);
3985 spin_lock_init(&ipvs
->tot_stats
.lock
);
3987 proc_create("ip_vs", 0, net
->proc_net
, &ip_vs_info_fops
);
3988 proc_create("ip_vs_stats", 0, net
->proc_net
, &ip_vs_stats_fops
);
3989 proc_create("ip_vs_stats_percpu", 0, net
->proc_net
,
3990 &ip_vs_stats_percpu_fops
);
3992 if (ip_vs_control_net_init_sysctl(net
))
3998 free_percpu(ipvs
->tot_stats
.cpustats
);
4002 void __net_exit
ip_vs_control_net_cleanup(struct net
*net
)
4004 struct netns_ipvs
*ipvs
= net_ipvs(net
);
4006 ip_vs_trash_cleanup(ipvs
);
4007 ip_vs_control_net_cleanup_sysctl(net
);
4008 remove_proc_entry("ip_vs_stats_percpu", net
->proc_net
);
4009 remove_proc_entry("ip_vs_stats", net
->proc_net
);
4010 remove_proc_entry("ip_vs", net
->proc_net
);
4011 free_percpu(ipvs
->tot_stats
.cpustats
);
4014 int __init
ip_vs_register_nl_ioctl(void)
4018 ret
= nf_register_sockopt(&ip_vs_sockopts
);
4020 pr_err("cannot register sockopt.\n");
4024 ret
= ip_vs_genl_register();
4026 pr_err("cannot register Generic Netlink interface.\n");
4032 nf_unregister_sockopt(&ip_vs_sockopts
);
4037 void ip_vs_unregister_nl_ioctl(void)
4039 ip_vs_genl_unregister();
4040 nf_unregister_sockopt(&ip_vs_sockopts
);
4043 int __init
ip_vs_control_init(void)
4050 /* Initialize svc_table, ip_vs_svc_fwm_table */
4051 for (idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
4052 INIT_HLIST_HEAD(&ip_vs_svc_table
[idx
]);
4053 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table
[idx
]);
4056 smp_wmb(); /* Do we really need it now ? */
4058 ret
= register_netdevice_notifier(&ip_vs_dst_notifier
);
4067 void ip_vs_control_cleanup(void)
4070 unregister_netdevice_notifier(&ip_vs_dst_notifier
);