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>
42 #ifdef CONFIG_IP_VS_IPV6
44 #include <net/ip6_route.h>
46 #include <net/route.h>
48 #include <net/genetlink.h>
50 #include <asm/uaccess.h>
52 #include <net/ip_vs.h>
54 /* semaphore for IPVS sockopts. And, [gs]etsockopt may sleep. */
55 static DEFINE_MUTEX(__ip_vs_mutex
);
57 /* lock for service table */
58 static DEFINE_RWLOCK(__ip_vs_svc_lock
);
60 /* lock for table with the real services */
61 static DEFINE_RWLOCK(__ip_vs_rs_lock
);
63 /* lock for state and timeout tables */
64 static DEFINE_SPINLOCK(ip_vs_securetcp_lock
);
66 /* lock for drop entry handling */
67 static DEFINE_SPINLOCK(__ip_vs_dropentry_lock
);
69 /* lock for drop packet handling */
70 static DEFINE_SPINLOCK(__ip_vs_droppacket_lock
);
72 /* 1/rate drop and drop-entry variables */
73 int ip_vs_drop_rate
= 0;
74 int ip_vs_drop_counter
= 0;
75 static atomic_t ip_vs_dropentry
= ATOMIC_INIT(0);
77 /* number of virtual services */
78 static int ip_vs_num_services
= 0;
80 /* sysctl variables */
81 static int sysctl_ip_vs_drop_entry
= 0;
82 static int sysctl_ip_vs_drop_packet
= 0;
83 static int sysctl_ip_vs_secure_tcp
= 0;
84 static int sysctl_ip_vs_amemthresh
= 1024;
85 static int sysctl_ip_vs_am_droprate
= 10;
86 int sysctl_ip_vs_cache_bypass
= 0;
87 int sysctl_ip_vs_expire_nodest_conn
= 0;
88 int sysctl_ip_vs_expire_quiescent_template
= 0;
89 int sysctl_ip_vs_sync_threshold
[2] = { 3, 50 };
90 int sysctl_ip_vs_nat_icmp_send
= 0;
91 #ifdef CONFIG_IP_VS_NFCT
92 int sysctl_ip_vs_conntrack
;
94 int sysctl_ip_vs_snat_reroute
= 1;
97 #ifdef CONFIG_IP_VS_DEBUG
98 static int sysctl_ip_vs_debug_level
= 0;
100 int ip_vs_get_debug_level(void)
102 return sysctl_ip_vs_debug_level
;
106 #ifdef CONFIG_IP_VS_IPV6
107 /* Taken from rt6_fill_node() in net/ipv6/route.c, is there a better way? */
108 static int __ip_vs_addr_is_local_v6(const struct in6_addr
*addr
)
116 .saddr
= { .s6_addr32
= {0, 0, 0, 0} }, } },
119 rt
= (struct rt6_info
*)ip6_route_output(&init_net
, NULL
, &fl
);
120 if (rt
&& rt
->rt6i_dev
&& (rt
->rt6i_dev
->flags
& IFF_LOOPBACK
))
127 * update_defense_level is called from keventd and from sysctl,
128 * so it needs to protect itself from softirqs
130 static void update_defense_level(void)
133 static int old_secure_tcp
= 0;
138 /* we only count free and buffered memory (in pages) */
140 availmem
= i
.freeram
+ i
.bufferram
;
141 /* however in linux 2.5 the i.bufferram is total page cache size,
143 /* si_swapinfo(&i); */
144 /* availmem = availmem - (i.totalswap - i.freeswap); */
146 nomem
= (availmem
< sysctl_ip_vs_amemthresh
);
151 spin_lock(&__ip_vs_dropentry_lock
);
152 switch (sysctl_ip_vs_drop_entry
) {
154 atomic_set(&ip_vs_dropentry
, 0);
158 atomic_set(&ip_vs_dropentry
, 1);
159 sysctl_ip_vs_drop_entry
= 2;
161 atomic_set(&ip_vs_dropentry
, 0);
166 atomic_set(&ip_vs_dropentry
, 1);
168 atomic_set(&ip_vs_dropentry
, 0);
169 sysctl_ip_vs_drop_entry
= 1;
173 atomic_set(&ip_vs_dropentry
, 1);
176 spin_unlock(&__ip_vs_dropentry_lock
);
179 spin_lock(&__ip_vs_droppacket_lock
);
180 switch (sysctl_ip_vs_drop_packet
) {
186 ip_vs_drop_rate
= ip_vs_drop_counter
187 = sysctl_ip_vs_amemthresh
/
188 (sysctl_ip_vs_amemthresh
-availmem
);
189 sysctl_ip_vs_drop_packet
= 2;
196 ip_vs_drop_rate
= ip_vs_drop_counter
197 = sysctl_ip_vs_amemthresh
/
198 (sysctl_ip_vs_amemthresh
-availmem
);
201 sysctl_ip_vs_drop_packet
= 1;
205 ip_vs_drop_rate
= sysctl_ip_vs_am_droprate
;
208 spin_unlock(&__ip_vs_droppacket_lock
);
211 spin_lock(&ip_vs_securetcp_lock
);
212 switch (sysctl_ip_vs_secure_tcp
) {
214 if (old_secure_tcp
>= 2)
219 if (old_secure_tcp
< 2)
221 sysctl_ip_vs_secure_tcp
= 2;
223 if (old_secure_tcp
>= 2)
229 if (old_secure_tcp
< 2)
232 if (old_secure_tcp
>= 2)
234 sysctl_ip_vs_secure_tcp
= 1;
238 if (old_secure_tcp
< 2)
242 old_secure_tcp
= sysctl_ip_vs_secure_tcp
;
244 ip_vs_protocol_timeout_change(sysctl_ip_vs_secure_tcp
>1);
245 spin_unlock(&ip_vs_securetcp_lock
);
252 * Timer for checking the defense
254 #define DEFENSE_TIMER_PERIOD 1*HZ
255 static void defense_work_handler(struct work_struct
*work
);
256 static DECLARE_DELAYED_WORK(defense_work
, defense_work_handler
);
258 static void defense_work_handler(struct work_struct
*work
)
260 update_defense_level();
261 if (atomic_read(&ip_vs_dropentry
))
262 ip_vs_random_dropentry();
264 schedule_delayed_work(&defense_work
, DEFENSE_TIMER_PERIOD
);
268 ip_vs_use_count_inc(void)
270 return try_module_get(THIS_MODULE
);
274 ip_vs_use_count_dec(void)
276 module_put(THIS_MODULE
);
281 * Hash table: for virtual service lookups
283 #define IP_VS_SVC_TAB_BITS 8
284 #define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
285 #define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)
287 /* the service table hashed by <protocol, addr, port> */
288 static struct list_head ip_vs_svc_table
[IP_VS_SVC_TAB_SIZE
];
289 /* the service table hashed by fwmark */
290 static struct list_head ip_vs_svc_fwm_table
[IP_VS_SVC_TAB_SIZE
];
293 * Hash table: for real service lookups
295 #define IP_VS_RTAB_BITS 4
296 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
297 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
299 static struct list_head ip_vs_rtable
[IP_VS_RTAB_SIZE
];
302 * Trash for destinations
304 static LIST_HEAD(ip_vs_dest_trash
);
307 * FTP & NULL virtual service counters
309 static atomic_t ip_vs_ftpsvc_counter
= ATOMIC_INIT(0);
310 static atomic_t ip_vs_nullsvc_counter
= ATOMIC_INIT(0);
314 * Returns hash value for virtual service
316 static __inline__
unsigned
317 ip_vs_svc_hashkey(int af
, unsigned proto
, const union nf_inet_addr
*addr
,
320 register unsigned porth
= ntohs(port
);
321 __be32 addr_fold
= addr
->ip
;
323 #ifdef CONFIG_IP_VS_IPV6
325 addr_fold
= addr
->ip6
[0]^addr
->ip6
[1]^
326 addr
->ip6
[2]^addr
->ip6
[3];
329 return (proto
^ntohl(addr_fold
)^(porth
>>IP_VS_SVC_TAB_BITS
)^porth
)
330 & IP_VS_SVC_TAB_MASK
;
334 * Returns hash value of fwmark for virtual service lookup
336 static __inline__
unsigned ip_vs_svc_fwm_hashkey(__u32 fwmark
)
338 return fwmark
& IP_VS_SVC_TAB_MASK
;
342 * Hashes a service in the ip_vs_svc_table by <proto,addr,port>
343 * or in the ip_vs_svc_fwm_table by fwmark.
344 * Should be called with locked tables.
346 static int ip_vs_svc_hash(struct ip_vs_service
*svc
)
350 if (svc
->flags
& IP_VS_SVC_F_HASHED
) {
351 pr_err("%s(): request for already hashed, called from %pF\n",
352 __func__
, __builtin_return_address(0));
356 if (svc
->fwmark
== 0) {
358 * Hash it by <protocol,addr,port> in ip_vs_svc_table
360 hash
= ip_vs_svc_hashkey(svc
->af
, svc
->protocol
, &svc
->addr
,
362 list_add(&svc
->s_list
, &ip_vs_svc_table
[hash
]);
365 * Hash it by fwmark in ip_vs_svc_fwm_table
367 hash
= ip_vs_svc_fwm_hashkey(svc
->fwmark
);
368 list_add(&svc
->f_list
, &ip_vs_svc_fwm_table
[hash
]);
371 svc
->flags
|= IP_VS_SVC_F_HASHED
;
372 /* increase its refcnt because it is referenced by the svc table */
373 atomic_inc(&svc
->refcnt
);
379 * Unhashes a service from ip_vs_svc_table/ip_vs_svc_fwm_table.
380 * Should be called with locked tables.
382 static int ip_vs_svc_unhash(struct ip_vs_service
*svc
)
384 if (!(svc
->flags
& IP_VS_SVC_F_HASHED
)) {
385 pr_err("%s(): request for unhash flagged, called from %pF\n",
386 __func__
, __builtin_return_address(0));
390 if (svc
->fwmark
== 0) {
391 /* Remove it from the ip_vs_svc_table table */
392 list_del(&svc
->s_list
);
394 /* Remove it from the ip_vs_svc_fwm_table table */
395 list_del(&svc
->f_list
);
398 svc
->flags
&= ~IP_VS_SVC_F_HASHED
;
399 atomic_dec(&svc
->refcnt
);
405 * Get service by {proto,addr,port} in the service table.
407 static inline struct ip_vs_service
*
408 __ip_vs_service_find(int af
, __u16 protocol
, const union nf_inet_addr
*vaddr
,
412 struct ip_vs_service
*svc
;
414 /* Check for "full" addressed entries */
415 hash
= ip_vs_svc_hashkey(af
, protocol
, vaddr
, vport
);
417 list_for_each_entry(svc
, &ip_vs_svc_table
[hash
], s_list
){
419 && ip_vs_addr_equal(af
, &svc
->addr
, vaddr
)
420 && (svc
->port
== vport
)
421 && (svc
->protocol
== protocol
)) {
432 * Get service by {fwmark} in the service table.
434 static inline struct ip_vs_service
*
435 __ip_vs_svc_fwm_find(int af
, __u32 fwmark
)
438 struct ip_vs_service
*svc
;
440 /* Check for fwmark addressed entries */
441 hash
= ip_vs_svc_fwm_hashkey(fwmark
);
443 list_for_each_entry(svc
, &ip_vs_svc_fwm_table
[hash
], f_list
) {
444 if (svc
->fwmark
== fwmark
&& svc
->af
== af
) {
453 struct ip_vs_service
*
454 ip_vs_service_get(int af
, __u32 fwmark
, __u16 protocol
,
455 const union nf_inet_addr
*vaddr
, __be16 vport
)
457 struct ip_vs_service
*svc
;
459 read_lock(&__ip_vs_svc_lock
);
462 * Check the table hashed by fwmark first
464 if (fwmark
&& (svc
= __ip_vs_svc_fwm_find(af
, fwmark
)))
468 * Check the table hashed by <protocol,addr,port>
469 * for "full" addressed entries
471 svc
= __ip_vs_service_find(af
, protocol
, vaddr
, vport
);
474 && protocol
== IPPROTO_TCP
475 && atomic_read(&ip_vs_ftpsvc_counter
)
476 && (vport
== FTPDATA
|| ntohs(vport
) >= PROT_SOCK
)) {
478 * Check if ftp service entry exists, the packet
479 * might belong to FTP data connections.
481 svc
= __ip_vs_service_find(af
, protocol
, vaddr
, FTPPORT
);
485 && atomic_read(&ip_vs_nullsvc_counter
)) {
487 * Check if the catch-all port (port zero) exists
489 svc
= __ip_vs_service_find(af
, protocol
, vaddr
, 0);
494 atomic_inc(&svc
->usecnt
);
495 read_unlock(&__ip_vs_svc_lock
);
497 IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
498 fwmark
, ip_vs_proto_name(protocol
),
499 IP_VS_DBG_ADDR(af
, vaddr
), ntohs(vport
),
500 svc
? "hit" : "not hit");
507 __ip_vs_bind_svc(struct ip_vs_dest
*dest
, struct ip_vs_service
*svc
)
509 atomic_inc(&svc
->refcnt
);
514 __ip_vs_unbind_svc(struct ip_vs_dest
*dest
)
516 struct ip_vs_service
*svc
= dest
->svc
;
519 if (atomic_dec_and_test(&svc
->refcnt
)) {
520 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
522 IP_VS_DBG_ADDR(svc
->af
, &svc
->addr
),
523 ntohs(svc
->port
), atomic_read(&svc
->usecnt
));
530 * Returns hash value for real service
532 static inline unsigned ip_vs_rs_hashkey(int af
,
533 const union nf_inet_addr
*addr
,
536 register unsigned porth
= ntohs(port
);
537 __be32 addr_fold
= addr
->ip
;
539 #ifdef CONFIG_IP_VS_IPV6
541 addr_fold
= addr
->ip6
[0]^addr
->ip6
[1]^
542 addr
->ip6
[2]^addr
->ip6
[3];
545 return (ntohl(addr_fold
)^(porth
>>IP_VS_RTAB_BITS
)^porth
)
550 * Hashes ip_vs_dest in ip_vs_rtable by <proto,addr,port>.
551 * should be called with locked tables.
553 static int ip_vs_rs_hash(struct ip_vs_dest
*dest
)
557 if (!list_empty(&dest
->d_list
)) {
562 * Hash by proto,addr,port,
563 * which are the parameters of the real service.
565 hash
= ip_vs_rs_hashkey(dest
->af
, &dest
->addr
, dest
->port
);
567 list_add(&dest
->d_list
, &ip_vs_rtable
[hash
]);
573 * UNhashes ip_vs_dest from ip_vs_rtable.
574 * should be called with locked tables.
576 static int ip_vs_rs_unhash(struct ip_vs_dest
*dest
)
579 * Remove it from the ip_vs_rtable table.
581 if (!list_empty(&dest
->d_list
)) {
582 list_del(&dest
->d_list
);
583 INIT_LIST_HEAD(&dest
->d_list
);
590 * Lookup real service by <proto,addr,port> in the real service table.
593 ip_vs_lookup_real_service(int af
, __u16 protocol
,
594 const union nf_inet_addr
*daddr
,
598 struct ip_vs_dest
*dest
;
601 * Check for "full" addressed entries
602 * Return the first found entry
604 hash
= ip_vs_rs_hashkey(af
, daddr
, dport
);
606 read_lock(&__ip_vs_rs_lock
);
607 list_for_each_entry(dest
, &ip_vs_rtable
[hash
], d_list
) {
609 && ip_vs_addr_equal(af
, &dest
->addr
, daddr
)
610 && (dest
->port
== dport
)
611 && ((dest
->protocol
== protocol
) ||
614 read_unlock(&__ip_vs_rs_lock
);
618 read_unlock(&__ip_vs_rs_lock
);
624 * Lookup destination by {addr,port} in the given service
626 static struct ip_vs_dest
*
627 ip_vs_lookup_dest(struct ip_vs_service
*svc
, const union nf_inet_addr
*daddr
,
630 struct ip_vs_dest
*dest
;
633 * Find the destination for the given service
635 list_for_each_entry(dest
, &svc
->destinations
, n_list
) {
636 if ((dest
->af
== svc
->af
)
637 && ip_vs_addr_equal(svc
->af
, &dest
->addr
, daddr
)
638 && (dest
->port
== dport
)) {
648 * Find destination by {daddr,dport,vaddr,protocol}
649 * Cretaed to be used in ip_vs_process_message() in
650 * the backup synchronization daemon. It finds the
651 * destination to be bound to the received connection
654 * ip_vs_lookup_real_service() looked promissing, but
655 * seems not working as expected.
657 struct ip_vs_dest
*ip_vs_find_dest(int af
, const union nf_inet_addr
*daddr
,
659 const union nf_inet_addr
*vaddr
,
660 __be16 vport
, __u16 protocol
)
662 struct ip_vs_dest
*dest
;
663 struct ip_vs_service
*svc
;
665 svc
= ip_vs_service_get(af
, 0, protocol
, vaddr
, vport
);
668 dest
= ip_vs_lookup_dest(svc
, daddr
, dport
);
670 atomic_inc(&dest
->refcnt
);
671 ip_vs_service_put(svc
);
676 * Lookup dest by {svc,addr,port} in the destination trash.
677 * The destination trash is used to hold the destinations that are removed
678 * from the service table but are still referenced by some conn entries.
679 * The reason to add the destination trash is when the dest is temporary
680 * down (either by administrator or by monitor program), the dest can be
681 * picked back from the trash, the remaining connections to the dest can
682 * continue, and the counting information of the dest is also useful for
685 static struct ip_vs_dest
*
686 ip_vs_trash_get_dest(struct ip_vs_service
*svc
, const union nf_inet_addr
*daddr
,
689 struct ip_vs_dest
*dest
, *nxt
;
692 * Find the destination in trash
694 list_for_each_entry_safe(dest
, nxt
, &ip_vs_dest_trash
, n_list
) {
695 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
698 IP_VS_DBG_ADDR(svc
->af
, &dest
->addr
),
700 atomic_read(&dest
->refcnt
));
701 if (dest
->af
== svc
->af
&&
702 ip_vs_addr_equal(svc
->af
, &dest
->addr
, daddr
) &&
703 dest
->port
== dport
&&
704 dest
->vfwmark
== svc
->fwmark
&&
705 dest
->protocol
== svc
->protocol
&&
707 (ip_vs_addr_equal(svc
->af
, &dest
->vaddr
, &svc
->addr
) &&
708 dest
->vport
== svc
->port
))) {
714 * Try to purge the destination from trash if not referenced
716 if (atomic_read(&dest
->refcnt
) == 1) {
717 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u "
720 IP_VS_DBG_ADDR(svc
->af
, &dest
->addr
),
722 list_del(&dest
->n_list
);
723 ip_vs_dst_reset(dest
);
724 __ip_vs_unbind_svc(dest
);
734 * Clean up all the destinations in the trash
735 * Called by the ip_vs_control_cleanup()
737 * When the ip_vs_control_clearup is activated by ipvs module exit,
738 * the service tables must have been flushed and all the connections
739 * are expired, and the refcnt of each destination in the trash must
740 * be 1, so we simply release them here.
742 static void ip_vs_trash_cleanup(void)
744 struct ip_vs_dest
*dest
, *nxt
;
746 list_for_each_entry_safe(dest
, nxt
, &ip_vs_dest_trash
, n_list
) {
747 list_del(&dest
->n_list
);
748 ip_vs_dst_reset(dest
);
749 __ip_vs_unbind_svc(dest
);
756 ip_vs_zero_stats(struct ip_vs_stats
*stats
)
758 spin_lock_bh(&stats
->lock
);
760 memset(&stats
->ustats
, 0, sizeof(stats
->ustats
));
761 ip_vs_zero_estimator(stats
);
763 spin_unlock_bh(&stats
->lock
);
767 * Update a destination in the given service
770 __ip_vs_update_dest(struct ip_vs_service
*svc
, struct ip_vs_dest
*dest
,
771 struct ip_vs_dest_user_kern
*udest
, int add
)
775 /* set the weight and the flags */
776 atomic_set(&dest
->weight
, udest
->weight
);
777 conn_flags
= udest
->conn_flags
& IP_VS_CONN_F_DEST_MASK
;
778 conn_flags
|= IP_VS_CONN_F_INACTIVE
;
780 /* check if local node and update the flags */
781 #ifdef CONFIG_IP_VS_IPV6
782 if (svc
->af
== AF_INET6
) {
783 if (__ip_vs_addr_is_local_v6(&udest
->addr
.in6
)) {
784 conn_flags
= (conn_flags
& ~IP_VS_CONN_F_FWD_MASK
)
785 | IP_VS_CONN_F_LOCALNODE
;
789 if (inet_addr_type(&init_net
, udest
->addr
.ip
) == RTN_LOCAL
) {
790 conn_flags
= (conn_flags
& ~IP_VS_CONN_F_FWD_MASK
)
791 | IP_VS_CONN_F_LOCALNODE
;
794 /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
795 if ((conn_flags
& IP_VS_CONN_F_FWD_MASK
) != IP_VS_CONN_F_MASQ
) {
796 conn_flags
|= IP_VS_CONN_F_NOOUTPUT
;
799 * Put the real service in ip_vs_rtable if not present.
800 * For now only for NAT!
802 write_lock_bh(&__ip_vs_rs_lock
);
804 write_unlock_bh(&__ip_vs_rs_lock
);
806 atomic_set(&dest
->conn_flags
, conn_flags
);
808 /* bind the service */
810 __ip_vs_bind_svc(dest
, svc
);
812 if (dest
->svc
!= svc
) {
813 __ip_vs_unbind_svc(dest
);
814 ip_vs_zero_stats(&dest
->stats
);
815 __ip_vs_bind_svc(dest
, svc
);
819 /* set the dest status flags */
820 dest
->flags
|= IP_VS_DEST_F_AVAILABLE
;
822 if (udest
->u_threshold
== 0 || udest
->u_threshold
> dest
->u_threshold
)
823 dest
->flags
&= ~IP_VS_DEST_F_OVERLOAD
;
824 dest
->u_threshold
= udest
->u_threshold
;
825 dest
->l_threshold
= udest
->l_threshold
;
828 ip_vs_new_estimator(&dest
->stats
);
830 write_lock_bh(&__ip_vs_svc_lock
);
832 /* Wait until all other svc users go away */
833 IP_VS_WAIT_WHILE(atomic_read(&svc
->usecnt
) > 0);
836 list_add(&dest
->n_list
, &svc
->destinations
);
840 /* call the update_service, because server weight may be changed */
841 if (svc
->scheduler
->update_service
)
842 svc
->scheduler
->update_service(svc
);
844 write_unlock_bh(&__ip_vs_svc_lock
);
849 * Create a destination for the given service
852 ip_vs_new_dest(struct ip_vs_service
*svc
, struct ip_vs_dest_user_kern
*udest
,
853 struct ip_vs_dest
**dest_p
)
855 struct ip_vs_dest
*dest
;
860 #ifdef CONFIG_IP_VS_IPV6
861 if (svc
->af
== AF_INET6
) {
862 atype
= ipv6_addr_type(&udest
->addr
.in6
);
863 if ((!(atype
& IPV6_ADDR_UNICAST
) ||
864 atype
& IPV6_ADDR_LINKLOCAL
) &&
865 !__ip_vs_addr_is_local_v6(&udest
->addr
.in6
))
870 atype
= inet_addr_type(&init_net
, udest
->addr
.ip
);
871 if (atype
!= RTN_LOCAL
&& atype
!= RTN_UNICAST
)
875 dest
= kzalloc(sizeof(struct ip_vs_dest
), GFP_KERNEL
);
877 pr_err("%s(): no memory.\n", __func__
);
882 dest
->protocol
= svc
->protocol
;
883 dest
->vaddr
= svc
->addr
;
884 dest
->vport
= svc
->port
;
885 dest
->vfwmark
= svc
->fwmark
;
886 ip_vs_addr_copy(svc
->af
, &dest
->addr
, &udest
->addr
);
887 dest
->port
= udest
->port
;
889 atomic_set(&dest
->activeconns
, 0);
890 atomic_set(&dest
->inactconns
, 0);
891 atomic_set(&dest
->persistconns
, 0);
892 atomic_set(&dest
->refcnt
, 1);
894 INIT_LIST_HEAD(&dest
->d_list
);
895 spin_lock_init(&dest
->dst_lock
);
896 spin_lock_init(&dest
->stats
.lock
);
897 __ip_vs_update_dest(svc
, dest
, udest
, 1);
907 * Add a destination into an existing service
910 ip_vs_add_dest(struct ip_vs_service
*svc
, struct ip_vs_dest_user_kern
*udest
)
912 struct ip_vs_dest
*dest
;
913 union nf_inet_addr daddr
;
914 __be16 dport
= udest
->port
;
919 if (udest
->weight
< 0) {
920 pr_err("%s(): server weight less than zero\n", __func__
);
924 if (udest
->l_threshold
> udest
->u_threshold
) {
925 pr_err("%s(): lower threshold is higher than upper threshold\n",
930 ip_vs_addr_copy(svc
->af
, &daddr
, &udest
->addr
);
933 * Check if the dest already exists in the list
935 dest
= ip_vs_lookup_dest(svc
, &daddr
, dport
);
938 IP_VS_DBG(1, "%s(): dest already exists\n", __func__
);
943 * Check if the dest already exists in the trash and
944 * is from the same service
946 dest
= ip_vs_trash_get_dest(svc
, &daddr
, dport
);
949 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
950 "dest->refcnt=%d, service %u/%s:%u\n",
951 IP_VS_DBG_ADDR(svc
->af
, &daddr
), ntohs(dport
),
952 atomic_read(&dest
->refcnt
),
954 IP_VS_DBG_ADDR(svc
->af
, &dest
->vaddr
),
958 * Get the destination from the trash
960 list_del(&dest
->n_list
);
962 __ip_vs_update_dest(svc
, dest
, udest
, 1);
966 * Allocate and initialize the dest structure
968 ret
= ip_vs_new_dest(svc
, udest
, &dest
);
977 * Edit a destination in the given service
980 ip_vs_edit_dest(struct ip_vs_service
*svc
, struct ip_vs_dest_user_kern
*udest
)
982 struct ip_vs_dest
*dest
;
983 union nf_inet_addr daddr
;
984 __be16 dport
= udest
->port
;
988 if (udest
->weight
< 0) {
989 pr_err("%s(): server weight less than zero\n", __func__
);
993 if (udest
->l_threshold
> udest
->u_threshold
) {
994 pr_err("%s(): lower threshold is higher than upper threshold\n",
999 ip_vs_addr_copy(svc
->af
, &daddr
, &udest
->addr
);
1002 * Lookup the destination list
1004 dest
= ip_vs_lookup_dest(svc
, &daddr
, dport
);
1007 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__
);
1011 __ip_vs_update_dest(svc
, dest
, udest
, 0);
1019 * Delete a destination (must be already unlinked from the service)
1021 static void __ip_vs_del_dest(struct ip_vs_dest
*dest
)
1023 ip_vs_kill_estimator(&dest
->stats
);
1026 * Remove it from the d-linked list with the real services.
1028 write_lock_bh(&__ip_vs_rs_lock
);
1029 ip_vs_rs_unhash(dest
);
1030 write_unlock_bh(&__ip_vs_rs_lock
);
1033 * Decrease the refcnt of the dest, and free the dest
1034 * if nobody refers to it (refcnt=0). Otherwise, throw
1035 * the destination into the trash.
1037 if (atomic_dec_and_test(&dest
->refcnt
)) {
1038 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u\n",
1040 IP_VS_DBG_ADDR(dest
->af
, &dest
->addr
),
1042 ip_vs_dst_reset(dest
);
1043 /* simply decrease svc->refcnt here, let the caller check
1044 and release the service if nobody refers to it.
1045 Only user context can release destination and service,
1046 and only one user context can update virtual service at a
1047 time, so the operation here is OK */
1048 atomic_dec(&dest
->svc
->refcnt
);
1051 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, "
1052 "dest->refcnt=%d\n",
1053 IP_VS_DBG_ADDR(dest
->af
, &dest
->addr
),
1055 atomic_read(&dest
->refcnt
));
1056 list_add(&dest
->n_list
, &ip_vs_dest_trash
);
1057 atomic_inc(&dest
->refcnt
);
1063 * Unlink a destination from the given service
1065 static void __ip_vs_unlink_dest(struct ip_vs_service
*svc
,
1066 struct ip_vs_dest
*dest
,
1069 dest
->flags
&= ~IP_VS_DEST_F_AVAILABLE
;
1072 * Remove it from the d-linked destination list.
1074 list_del(&dest
->n_list
);
1078 * Call the update_service function of its scheduler
1080 if (svcupd
&& svc
->scheduler
->update_service
)
1081 svc
->scheduler
->update_service(svc
);
1086 * Delete a destination server in the given service
1089 ip_vs_del_dest(struct ip_vs_service
*svc
, struct ip_vs_dest_user_kern
*udest
)
1091 struct ip_vs_dest
*dest
;
1092 __be16 dport
= udest
->port
;
1096 dest
= ip_vs_lookup_dest(svc
, &udest
->addr
, dport
);
1099 IP_VS_DBG(1, "%s(): destination not found!\n", __func__
);
1103 write_lock_bh(&__ip_vs_svc_lock
);
1106 * Wait until all other svc users go away.
1108 IP_VS_WAIT_WHILE(atomic_read(&svc
->usecnt
) > 0);
1111 * Unlink dest from the service
1113 __ip_vs_unlink_dest(svc
, dest
, 1);
1115 write_unlock_bh(&__ip_vs_svc_lock
);
1118 * Delete the destination
1120 __ip_vs_del_dest(dest
);
1129 * Add a service into the service hash table
1132 ip_vs_add_service(struct ip_vs_service_user_kern
*u
,
1133 struct ip_vs_service
**svc_p
)
1136 struct ip_vs_scheduler
*sched
= NULL
;
1137 struct ip_vs_service
*svc
= NULL
;
1139 /* increase the module use count */
1140 ip_vs_use_count_inc();
1142 /* Lookup the scheduler by 'u->sched_name' */
1143 sched
= ip_vs_scheduler_get(u
->sched_name
);
1144 if (sched
== NULL
) {
1145 pr_info("Scheduler module ip_vs_%s not found\n", u
->sched_name
);
1150 #ifdef CONFIG_IP_VS_IPV6
1151 if (u
->af
== AF_INET6
&& (u
->netmask
< 1 || u
->netmask
> 128)) {
1157 svc
= kzalloc(sizeof(struct ip_vs_service
), GFP_KERNEL
);
1159 IP_VS_DBG(1, "%s(): no memory\n", __func__
);
1164 /* I'm the first user of the service */
1165 atomic_set(&svc
->usecnt
, 0);
1166 atomic_set(&svc
->refcnt
, 0);
1169 svc
->protocol
= u
->protocol
;
1170 ip_vs_addr_copy(svc
->af
, &svc
->addr
, &u
->addr
);
1171 svc
->port
= u
->port
;
1172 svc
->fwmark
= u
->fwmark
;
1173 svc
->flags
= u
->flags
;
1174 svc
->timeout
= u
->timeout
* HZ
;
1175 svc
->netmask
= u
->netmask
;
1177 INIT_LIST_HEAD(&svc
->destinations
);
1178 rwlock_init(&svc
->sched_lock
);
1179 spin_lock_init(&svc
->stats
.lock
);
1181 /* Bind the scheduler */
1182 ret
= ip_vs_bind_scheduler(svc
, sched
);
1187 /* Update the virtual service counters */
1188 if (svc
->port
== FTPPORT
)
1189 atomic_inc(&ip_vs_ftpsvc_counter
);
1190 else if (svc
->port
== 0)
1191 atomic_inc(&ip_vs_nullsvc_counter
);
1193 ip_vs_new_estimator(&svc
->stats
);
1195 /* Count only IPv4 services for old get/setsockopt interface */
1196 if (svc
->af
== AF_INET
)
1197 ip_vs_num_services
++;
1199 /* Hash the service into the service table */
1200 write_lock_bh(&__ip_vs_svc_lock
);
1201 ip_vs_svc_hash(svc
);
1202 write_unlock_bh(&__ip_vs_svc_lock
);
1210 ip_vs_unbind_scheduler(svc
);
1213 ip_vs_app_inc_put(svc
->inc
);
1218 ip_vs_scheduler_put(sched
);
1221 /* decrease the module use count */
1222 ip_vs_use_count_dec();
1229 * Edit a service and bind it with a new scheduler
1232 ip_vs_edit_service(struct ip_vs_service
*svc
, struct ip_vs_service_user_kern
*u
)
1234 struct ip_vs_scheduler
*sched
, *old_sched
;
1238 * Lookup the scheduler, by 'u->sched_name'
1240 sched
= ip_vs_scheduler_get(u
->sched_name
);
1241 if (sched
== NULL
) {
1242 pr_info("Scheduler module ip_vs_%s not found\n", u
->sched_name
);
1247 #ifdef CONFIG_IP_VS_IPV6
1248 if (u
->af
== AF_INET6
&& (u
->netmask
< 1 || u
->netmask
> 128)) {
1254 write_lock_bh(&__ip_vs_svc_lock
);
1257 * Wait until all other svc users go away.
1259 IP_VS_WAIT_WHILE(atomic_read(&svc
->usecnt
) > 0);
1262 * Set the flags and timeout value
1264 svc
->flags
= u
->flags
| IP_VS_SVC_F_HASHED
;
1265 svc
->timeout
= u
->timeout
* HZ
;
1266 svc
->netmask
= u
->netmask
;
1268 old_sched
= svc
->scheduler
;
1269 if (sched
!= old_sched
) {
1271 * Unbind the old scheduler
1273 if ((ret
= ip_vs_unbind_scheduler(svc
))) {
1279 * Bind the new scheduler
1281 if ((ret
= ip_vs_bind_scheduler(svc
, sched
))) {
1283 * If ip_vs_bind_scheduler fails, restore the old
1285 * The main reason of failure is out of memory.
1287 * The question is if the old scheduler can be
1288 * restored all the time. TODO: if it cannot be
1289 * restored some time, we must delete the service,
1290 * otherwise the system may crash.
1292 ip_vs_bind_scheduler(svc
, old_sched
);
1299 write_unlock_bh(&__ip_vs_svc_lock
);
1300 #ifdef CONFIG_IP_VS_IPV6
1305 ip_vs_scheduler_put(old_sched
);
1312 * Delete a service from the service list
1313 * - The service must be unlinked, unlocked and not referenced!
1314 * - We are called under _bh lock
1316 static void __ip_vs_del_service(struct ip_vs_service
*svc
)
1318 struct ip_vs_dest
*dest
, *nxt
;
1319 struct ip_vs_scheduler
*old_sched
;
1321 /* Count only IPv4 services for old get/setsockopt interface */
1322 if (svc
->af
== AF_INET
)
1323 ip_vs_num_services
--;
1325 ip_vs_kill_estimator(&svc
->stats
);
1327 /* Unbind scheduler */
1328 old_sched
= svc
->scheduler
;
1329 ip_vs_unbind_scheduler(svc
);
1331 ip_vs_scheduler_put(old_sched
);
1333 /* Unbind app inc */
1335 ip_vs_app_inc_put(svc
->inc
);
1340 * Unlink the whole destination list
1342 list_for_each_entry_safe(dest
, nxt
, &svc
->destinations
, n_list
) {
1343 __ip_vs_unlink_dest(svc
, dest
, 0);
1344 __ip_vs_del_dest(dest
);
1348 * Update the virtual service counters
1350 if (svc
->port
== FTPPORT
)
1351 atomic_dec(&ip_vs_ftpsvc_counter
);
1352 else if (svc
->port
== 0)
1353 atomic_dec(&ip_vs_nullsvc_counter
);
1356 * Free the service if nobody refers to it
1358 if (atomic_read(&svc
->refcnt
) == 0) {
1359 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
1361 IP_VS_DBG_ADDR(svc
->af
, &svc
->addr
),
1362 ntohs(svc
->port
), atomic_read(&svc
->usecnt
));
1366 /* decrease the module use count */
1367 ip_vs_use_count_dec();
1371 * Unlink a service from list and try to delete it if its refcnt reached 0
1373 static void ip_vs_unlink_service(struct ip_vs_service
*svc
)
1376 * Unhash it from the service table
1378 write_lock_bh(&__ip_vs_svc_lock
);
1380 ip_vs_svc_unhash(svc
);
1383 * Wait until all the svc users go away.
1385 IP_VS_WAIT_WHILE(atomic_read(&svc
->usecnt
) > 0);
1387 __ip_vs_del_service(svc
);
1389 write_unlock_bh(&__ip_vs_svc_lock
);
1393 * Delete a service from the service list
1395 static int ip_vs_del_service(struct ip_vs_service
*svc
)
1399 ip_vs_unlink_service(svc
);
1406 * Flush all the virtual services
1408 static int ip_vs_flush(void)
1411 struct ip_vs_service
*svc
, *nxt
;
1414 * Flush the service table hashed by <protocol,addr,port>
1416 for(idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1417 list_for_each_entry_safe(svc
, nxt
, &ip_vs_svc_table
[idx
], s_list
) {
1418 ip_vs_unlink_service(svc
);
1423 * Flush the service table hashed by fwmark
1425 for(idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1426 list_for_each_entry_safe(svc
, nxt
,
1427 &ip_vs_svc_fwm_table
[idx
], f_list
) {
1428 ip_vs_unlink_service(svc
);
1437 * Zero counters in a service or all services
1439 static int ip_vs_zero_service(struct ip_vs_service
*svc
)
1441 struct ip_vs_dest
*dest
;
1443 write_lock_bh(&__ip_vs_svc_lock
);
1444 list_for_each_entry(dest
, &svc
->destinations
, n_list
) {
1445 ip_vs_zero_stats(&dest
->stats
);
1447 ip_vs_zero_stats(&svc
->stats
);
1448 write_unlock_bh(&__ip_vs_svc_lock
);
1452 static int ip_vs_zero_all(void)
1455 struct ip_vs_service
*svc
;
1457 for(idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1458 list_for_each_entry(svc
, &ip_vs_svc_table
[idx
], s_list
) {
1459 ip_vs_zero_service(svc
);
1463 for(idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1464 list_for_each_entry(svc
, &ip_vs_svc_fwm_table
[idx
], f_list
) {
1465 ip_vs_zero_service(svc
);
1469 ip_vs_zero_stats(&ip_vs_stats
);
1475 proc_do_defense_mode(ctl_table
*table
, int write
,
1476 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
1478 int *valp
= table
->data
;
1482 rc
= proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
1483 if (write
&& (*valp
!= val
)) {
1484 if ((*valp
< 0) || (*valp
> 3)) {
1485 /* Restore the correct value */
1488 update_defense_level();
1496 proc_do_sync_threshold(ctl_table
*table
, int write
,
1497 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
1499 int *valp
= table
->data
;
1503 /* backup the value first */
1504 memcpy(val
, valp
, sizeof(val
));
1506 rc
= proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
1507 if (write
&& (valp
[0] < 0 || valp
[1] < 0 || valp
[0] >= valp
[1])) {
1508 /* Restore the correct value */
1509 memcpy(valp
, val
, sizeof(val
));
1516 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1519 static struct ctl_table vs_vars
[] = {
1521 .procname
= "amemthresh",
1522 .data
= &sysctl_ip_vs_amemthresh
,
1523 .maxlen
= sizeof(int),
1525 .proc_handler
= proc_dointvec
,
1527 #ifdef CONFIG_IP_VS_DEBUG
1529 .procname
= "debug_level",
1530 .data
= &sysctl_ip_vs_debug_level
,
1531 .maxlen
= sizeof(int),
1533 .proc_handler
= proc_dointvec
,
1537 .procname
= "am_droprate",
1538 .data
= &sysctl_ip_vs_am_droprate
,
1539 .maxlen
= sizeof(int),
1541 .proc_handler
= proc_dointvec
,
1544 .procname
= "drop_entry",
1545 .data
= &sysctl_ip_vs_drop_entry
,
1546 .maxlen
= sizeof(int),
1548 .proc_handler
= proc_do_defense_mode
,
1551 .procname
= "drop_packet",
1552 .data
= &sysctl_ip_vs_drop_packet
,
1553 .maxlen
= sizeof(int),
1555 .proc_handler
= proc_do_defense_mode
,
1557 #ifdef CONFIG_IP_VS_NFCT
1559 .procname
= "conntrack",
1560 .data
= &sysctl_ip_vs_conntrack
,
1561 .maxlen
= sizeof(int),
1563 .proc_handler
= &proc_dointvec
,
1567 .procname
= "secure_tcp",
1568 .data
= &sysctl_ip_vs_secure_tcp
,
1569 .maxlen
= sizeof(int),
1571 .proc_handler
= proc_do_defense_mode
,
1574 .procname
= "snat_reroute",
1575 .data
= &sysctl_ip_vs_snat_reroute
,
1576 .maxlen
= sizeof(int),
1578 .proc_handler
= &proc_dointvec
,
1582 .procname
= "timeout_established",
1583 .data
= &vs_timeout_table_dos
.timeout
[IP_VS_S_ESTABLISHED
],
1584 .maxlen
= sizeof(int),
1586 .proc_handler
= proc_dointvec_jiffies
,
1589 .procname
= "timeout_synsent",
1590 .data
= &vs_timeout_table_dos
.timeout
[IP_VS_S_SYN_SENT
],
1591 .maxlen
= sizeof(int),
1593 .proc_handler
= proc_dointvec_jiffies
,
1596 .procname
= "timeout_synrecv",
1597 .data
= &vs_timeout_table_dos
.timeout
[IP_VS_S_SYN_RECV
],
1598 .maxlen
= sizeof(int),
1600 .proc_handler
= proc_dointvec_jiffies
,
1603 .procname
= "timeout_finwait",
1604 .data
= &vs_timeout_table_dos
.timeout
[IP_VS_S_FIN_WAIT
],
1605 .maxlen
= sizeof(int),
1607 .proc_handler
= proc_dointvec_jiffies
,
1610 .procname
= "timeout_timewait",
1611 .data
= &vs_timeout_table_dos
.timeout
[IP_VS_S_TIME_WAIT
],
1612 .maxlen
= sizeof(int),
1614 .proc_handler
= proc_dointvec_jiffies
,
1617 .procname
= "timeout_close",
1618 .data
= &vs_timeout_table_dos
.timeout
[IP_VS_S_CLOSE
],
1619 .maxlen
= sizeof(int),
1621 .proc_handler
= proc_dointvec_jiffies
,
1624 .procname
= "timeout_closewait",
1625 .data
= &vs_timeout_table_dos
.timeout
[IP_VS_S_CLOSE_WAIT
],
1626 .maxlen
= sizeof(int),
1628 .proc_handler
= proc_dointvec_jiffies
,
1631 .procname
= "timeout_lastack",
1632 .data
= &vs_timeout_table_dos
.timeout
[IP_VS_S_LAST_ACK
],
1633 .maxlen
= sizeof(int),
1635 .proc_handler
= proc_dointvec_jiffies
,
1638 .procname
= "timeout_listen",
1639 .data
= &vs_timeout_table_dos
.timeout
[IP_VS_S_LISTEN
],
1640 .maxlen
= sizeof(int),
1642 .proc_handler
= proc_dointvec_jiffies
,
1645 .procname
= "timeout_synack",
1646 .data
= &vs_timeout_table_dos
.timeout
[IP_VS_S_SYNACK
],
1647 .maxlen
= sizeof(int),
1649 .proc_handler
= proc_dointvec_jiffies
,
1652 .procname
= "timeout_udp",
1653 .data
= &vs_timeout_table_dos
.timeout
[IP_VS_S_UDP
],
1654 .maxlen
= sizeof(int),
1656 .proc_handler
= proc_dointvec_jiffies
,
1659 .procname
= "timeout_icmp",
1660 .data
= &vs_timeout_table_dos
.timeout
[IP_VS_S_ICMP
],
1661 .maxlen
= sizeof(int),
1663 .proc_handler
= proc_dointvec_jiffies
,
1667 .procname
= "cache_bypass",
1668 .data
= &sysctl_ip_vs_cache_bypass
,
1669 .maxlen
= sizeof(int),
1671 .proc_handler
= proc_dointvec
,
1674 .procname
= "expire_nodest_conn",
1675 .data
= &sysctl_ip_vs_expire_nodest_conn
,
1676 .maxlen
= sizeof(int),
1678 .proc_handler
= proc_dointvec
,
1681 .procname
= "expire_quiescent_template",
1682 .data
= &sysctl_ip_vs_expire_quiescent_template
,
1683 .maxlen
= sizeof(int),
1685 .proc_handler
= proc_dointvec
,
1688 .procname
= "sync_threshold",
1689 .data
= &sysctl_ip_vs_sync_threshold
,
1690 .maxlen
= sizeof(sysctl_ip_vs_sync_threshold
),
1692 .proc_handler
= proc_do_sync_threshold
,
1695 .procname
= "nat_icmp_send",
1696 .data
= &sysctl_ip_vs_nat_icmp_send
,
1697 .maxlen
= sizeof(int),
1699 .proc_handler
= proc_dointvec
,
1704 const struct ctl_path net_vs_ctl_path
[] = {
1705 { .procname
= "net", },
1706 { .procname
= "ipv4", },
1707 { .procname
= "vs", },
1710 EXPORT_SYMBOL_GPL(net_vs_ctl_path
);
1712 static struct ctl_table_header
* sysctl_header
;
1714 #ifdef CONFIG_PROC_FS
1717 struct list_head
*table
;
1722 * Write the contents of the VS rule table to a PROCfs file.
1723 * (It is kept just for backward compatibility)
1725 static inline const char *ip_vs_fwd_name(unsigned flags
)
1727 switch (flags
& IP_VS_CONN_F_FWD_MASK
) {
1728 case IP_VS_CONN_F_LOCALNODE
:
1730 case IP_VS_CONN_F_TUNNEL
:
1732 case IP_VS_CONN_F_DROUTE
:
1740 /* Get the Nth entry in the two lists */
1741 static struct ip_vs_service
*ip_vs_info_array(struct seq_file
*seq
, loff_t pos
)
1743 struct ip_vs_iter
*iter
= seq
->private;
1745 struct ip_vs_service
*svc
;
1747 /* look in hash by protocol */
1748 for (idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1749 list_for_each_entry(svc
, &ip_vs_svc_table
[idx
], s_list
) {
1751 iter
->table
= ip_vs_svc_table
;
1758 /* keep looking in fwmark */
1759 for (idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
1760 list_for_each_entry(svc
, &ip_vs_svc_fwm_table
[idx
], f_list
) {
1762 iter
->table
= ip_vs_svc_fwm_table
;
1772 static void *ip_vs_info_seq_start(struct seq_file
*seq
, loff_t
*pos
)
1773 __acquires(__ip_vs_svc_lock
)
1776 read_lock_bh(&__ip_vs_svc_lock
);
1777 return *pos
? ip_vs_info_array(seq
, *pos
- 1) : SEQ_START_TOKEN
;
1781 static void *ip_vs_info_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
1783 struct list_head
*e
;
1784 struct ip_vs_iter
*iter
;
1785 struct ip_vs_service
*svc
;
1788 if (v
== SEQ_START_TOKEN
)
1789 return ip_vs_info_array(seq
,0);
1792 iter
= seq
->private;
1794 if (iter
->table
== ip_vs_svc_table
) {
1795 /* next service in table hashed by protocol */
1796 if ((e
= svc
->s_list
.next
) != &ip_vs_svc_table
[iter
->bucket
])
1797 return list_entry(e
, struct ip_vs_service
, s_list
);
1800 while (++iter
->bucket
< IP_VS_SVC_TAB_SIZE
) {
1801 list_for_each_entry(svc
,&ip_vs_svc_table
[iter
->bucket
],
1807 iter
->table
= ip_vs_svc_fwm_table
;
1812 /* next service in hashed by fwmark */
1813 if ((e
= svc
->f_list
.next
) != &ip_vs_svc_fwm_table
[iter
->bucket
])
1814 return list_entry(e
, struct ip_vs_service
, f_list
);
1817 while (++iter
->bucket
< IP_VS_SVC_TAB_SIZE
) {
1818 list_for_each_entry(svc
, &ip_vs_svc_fwm_table
[iter
->bucket
],
1826 static void ip_vs_info_seq_stop(struct seq_file
*seq
, void *v
)
1827 __releases(__ip_vs_svc_lock
)
1829 read_unlock_bh(&__ip_vs_svc_lock
);
1833 static int ip_vs_info_seq_show(struct seq_file
*seq
, void *v
)
1835 if (v
== SEQ_START_TOKEN
) {
1837 "IP Virtual Server version %d.%d.%d (size=%d)\n",
1838 NVERSION(IP_VS_VERSION_CODE
), ip_vs_conn_tab_size
);
1840 "Prot LocalAddress:Port Scheduler Flags\n");
1842 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
1844 const struct ip_vs_service
*svc
= v
;
1845 const struct ip_vs_iter
*iter
= seq
->private;
1846 const struct ip_vs_dest
*dest
;
1848 if (iter
->table
== ip_vs_svc_table
) {
1849 #ifdef CONFIG_IP_VS_IPV6
1850 if (svc
->af
== AF_INET6
)
1851 seq_printf(seq
, "%s [%pI6]:%04X %s ",
1852 ip_vs_proto_name(svc
->protocol
),
1855 svc
->scheduler
->name
);
1858 seq_printf(seq
, "%s %08X:%04X %s %s ",
1859 ip_vs_proto_name(svc
->protocol
),
1860 ntohl(svc
->addr
.ip
),
1862 svc
->scheduler
->name
,
1863 (svc
->flags
& IP_VS_SVC_F_ONEPACKET
)?"ops ":"");
1865 seq_printf(seq
, "FWM %08X %s %s",
1866 svc
->fwmark
, svc
->scheduler
->name
,
1867 (svc
->flags
& IP_VS_SVC_F_ONEPACKET
)?"ops ":"");
1870 if (svc
->flags
& IP_VS_SVC_F_PERSISTENT
)
1871 seq_printf(seq
, "persistent %d %08X\n",
1873 ntohl(svc
->netmask
));
1875 seq_putc(seq
, '\n');
1877 list_for_each_entry(dest
, &svc
->destinations
, n_list
) {
1878 #ifdef CONFIG_IP_VS_IPV6
1879 if (dest
->af
== AF_INET6
)
1882 " %-7s %-6d %-10d %-10d\n",
1885 ip_vs_fwd_name(atomic_read(&dest
->conn_flags
)),
1886 atomic_read(&dest
->weight
),
1887 atomic_read(&dest
->activeconns
),
1888 atomic_read(&dest
->inactconns
));
1893 "%-7s %-6d %-10d %-10d\n",
1894 ntohl(dest
->addr
.ip
),
1896 ip_vs_fwd_name(atomic_read(&dest
->conn_flags
)),
1897 atomic_read(&dest
->weight
),
1898 atomic_read(&dest
->activeconns
),
1899 atomic_read(&dest
->inactconns
));
1906 static const struct seq_operations ip_vs_info_seq_ops
= {
1907 .start
= ip_vs_info_seq_start
,
1908 .next
= ip_vs_info_seq_next
,
1909 .stop
= ip_vs_info_seq_stop
,
1910 .show
= ip_vs_info_seq_show
,
1913 static int ip_vs_info_open(struct inode
*inode
, struct file
*file
)
1915 return seq_open_private(file
, &ip_vs_info_seq_ops
,
1916 sizeof(struct ip_vs_iter
));
1919 static const struct file_operations ip_vs_info_fops
= {
1920 .owner
= THIS_MODULE
,
1921 .open
= ip_vs_info_open
,
1923 .llseek
= seq_lseek
,
1924 .release
= seq_release_private
,
1929 struct ip_vs_stats ip_vs_stats
= {
1930 .lock
= __SPIN_LOCK_UNLOCKED(ip_vs_stats
.lock
),
1933 #ifdef CONFIG_PROC_FS
1934 static int ip_vs_stats_show(struct seq_file
*seq
, void *v
)
1937 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
1939 " Total Incoming Outgoing Incoming Outgoing\n");
1941 " Conns Packets Packets Bytes Bytes\n");
1943 spin_lock_bh(&ip_vs_stats
.lock
);
1944 seq_printf(seq
, "%8X %8X %8X %16LX %16LX\n\n", ip_vs_stats
.ustats
.conns
,
1945 ip_vs_stats
.ustats
.inpkts
, ip_vs_stats
.ustats
.outpkts
,
1946 (unsigned long long) ip_vs_stats
.ustats
.inbytes
,
1947 (unsigned long long) ip_vs_stats
.ustats
.outbytes
);
1949 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
1951 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
1952 seq_printf(seq
,"%8X %8X %8X %16X %16X\n",
1953 ip_vs_stats
.ustats
.cps
,
1954 ip_vs_stats
.ustats
.inpps
,
1955 ip_vs_stats
.ustats
.outpps
,
1956 ip_vs_stats
.ustats
.inbps
,
1957 ip_vs_stats
.ustats
.outbps
);
1958 spin_unlock_bh(&ip_vs_stats
.lock
);
1963 static int ip_vs_stats_seq_open(struct inode
*inode
, struct file
*file
)
1965 return single_open(file
, ip_vs_stats_show
, NULL
);
1968 static const struct file_operations ip_vs_stats_fops
= {
1969 .owner
= THIS_MODULE
,
1970 .open
= ip_vs_stats_seq_open
,
1972 .llseek
= seq_lseek
,
1973 .release
= single_release
,
1979 * Set timeout values for tcp tcpfin udp in the timeout_table.
1981 static int ip_vs_set_timeout(struct ip_vs_timeout_user
*u
)
1983 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
1988 #ifdef CONFIG_IP_VS_PROTO_TCP
1989 if (u
->tcp_timeout
) {
1990 ip_vs_protocol_tcp
.timeout_table
[IP_VS_TCP_S_ESTABLISHED
]
1991 = u
->tcp_timeout
* HZ
;
1994 if (u
->tcp_fin_timeout
) {
1995 ip_vs_protocol_tcp
.timeout_table
[IP_VS_TCP_S_FIN_WAIT
]
1996 = u
->tcp_fin_timeout
* HZ
;
2000 #ifdef CONFIG_IP_VS_PROTO_UDP
2001 if (u
->udp_timeout
) {
2002 ip_vs_protocol_udp
.timeout_table
[IP_VS_UDP_S_NORMAL
]
2003 = u
->udp_timeout
* HZ
;
2010 #define SET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2011 #define SERVICE_ARG_LEN (sizeof(struct ip_vs_service_user))
2012 #define SVCDEST_ARG_LEN (sizeof(struct ip_vs_service_user) + \
2013 sizeof(struct ip_vs_dest_user))
2014 #define TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2015 #define DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user))
2016 #define MAX_ARG_LEN SVCDEST_ARG_LEN
2018 static const unsigned char set_arglen
[SET_CMDID(IP_VS_SO_SET_MAX
)+1] = {
2019 [SET_CMDID(IP_VS_SO_SET_ADD
)] = SERVICE_ARG_LEN
,
2020 [SET_CMDID(IP_VS_SO_SET_EDIT
)] = SERVICE_ARG_LEN
,
2021 [SET_CMDID(IP_VS_SO_SET_DEL
)] = SERVICE_ARG_LEN
,
2022 [SET_CMDID(IP_VS_SO_SET_FLUSH
)] = 0,
2023 [SET_CMDID(IP_VS_SO_SET_ADDDEST
)] = SVCDEST_ARG_LEN
,
2024 [SET_CMDID(IP_VS_SO_SET_DELDEST
)] = SVCDEST_ARG_LEN
,
2025 [SET_CMDID(IP_VS_SO_SET_EDITDEST
)] = SVCDEST_ARG_LEN
,
2026 [SET_CMDID(IP_VS_SO_SET_TIMEOUT
)] = TIMEOUT_ARG_LEN
,
2027 [SET_CMDID(IP_VS_SO_SET_STARTDAEMON
)] = DAEMON_ARG_LEN
,
2028 [SET_CMDID(IP_VS_SO_SET_STOPDAEMON
)] = DAEMON_ARG_LEN
,
2029 [SET_CMDID(IP_VS_SO_SET_ZERO
)] = SERVICE_ARG_LEN
,
2032 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern
*usvc
,
2033 struct ip_vs_service_user
*usvc_compat
)
2036 usvc
->protocol
= usvc_compat
->protocol
;
2037 usvc
->addr
.ip
= usvc_compat
->addr
;
2038 usvc
->port
= usvc_compat
->port
;
2039 usvc
->fwmark
= usvc_compat
->fwmark
;
2041 /* Deep copy of sched_name is not needed here */
2042 usvc
->sched_name
= usvc_compat
->sched_name
;
2044 usvc
->flags
= usvc_compat
->flags
;
2045 usvc
->timeout
= usvc_compat
->timeout
;
2046 usvc
->netmask
= usvc_compat
->netmask
;
2049 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern
*udest
,
2050 struct ip_vs_dest_user
*udest_compat
)
2052 udest
->addr
.ip
= udest_compat
->addr
;
2053 udest
->port
= udest_compat
->port
;
2054 udest
->conn_flags
= udest_compat
->conn_flags
;
2055 udest
->weight
= udest_compat
->weight
;
2056 udest
->u_threshold
= udest_compat
->u_threshold
;
2057 udest
->l_threshold
= udest_compat
->l_threshold
;
2061 do_ip_vs_set_ctl(struct sock
*sk
, int cmd
, void __user
*user
, unsigned int len
)
2064 unsigned char arg
[MAX_ARG_LEN
];
2065 struct ip_vs_service_user
*usvc_compat
;
2066 struct ip_vs_service_user_kern usvc
;
2067 struct ip_vs_service
*svc
;
2068 struct ip_vs_dest_user
*udest_compat
;
2069 struct ip_vs_dest_user_kern udest
;
2071 if (!capable(CAP_NET_ADMIN
))
2074 if (cmd
< IP_VS_BASE_CTL
|| cmd
> IP_VS_SO_SET_MAX
)
2076 if (len
< 0 || len
> MAX_ARG_LEN
)
2078 if (len
!= set_arglen
[SET_CMDID(cmd
)]) {
2079 pr_err("set_ctl: len %u != %u\n",
2080 len
, set_arglen
[SET_CMDID(cmd
)]);
2084 if (copy_from_user(arg
, user
, len
) != 0)
2087 /* increase the module use count */
2088 ip_vs_use_count_inc();
2090 if (mutex_lock_interruptible(&__ip_vs_mutex
)) {
2095 if (cmd
== IP_VS_SO_SET_FLUSH
) {
2096 /* Flush the virtual service */
2097 ret
= ip_vs_flush();
2099 } else if (cmd
== IP_VS_SO_SET_TIMEOUT
) {
2100 /* Set timeout values for (tcp tcpfin udp) */
2101 ret
= ip_vs_set_timeout((struct ip_vs_timeout_user
*)arg
);
2103 } else if (cmd
== IP_VS_SO_SET_STARTDAEMON
) {
2104 struct ip_vs_daemon_user
*dm
= (struct ip_vs_daemon_user
*)arg
;
2105 ret
= start_sync_thread(dm
->state
, dm
->mcast_ifn
, dm
->syncid
);
2107 } else if (cmd
== IP_VS_SO_SET_STOPDAEMON
) {
2108 struct ip_vs_daemon_user
*dm
= (struct ip_vs_daemon_user
*)arg
;
2109 ret
= stop_sync_thread(dm
->state
);
2113 usvc_compat
= (struct ip_vs_service_user
*)arg
;
2114 udest_compat
= (struct ip_vs_dest_user
*)(usvc_compat
+ 1);
2116 /* We only use the new structs internally, so copy userspace compat
2117 * structs to extended internal versions */
2118 ip_vs_copy_usvc_compat(&usvc
, usvc_compat
);
2119 ip_vs_copy_udest_compat(&udest
, udest_compat
);
2121 if (cmd
== IP_VS_SO_SET_ZERO
) {
2122 /* if no service address is set, zero counters in all */
2123 if (!usvc
.fwmark
&& !usvc
.addr
.ip
&& !usvc
.port
) {
2124 ret
= ip_vs_zero_all();
2129 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2130 if (usvc
.protocol
!= IPPROTO_TCP
&& usvc
.protocol
!= IPPROTO_UDP
&&
2131 usvc
.protocol
!= IPPROTO_SCTP
) {
2132 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2133 usvc
.protocol
, &usvc
.addr
.ip
,
2134 ntohs(usvc
.port
), usvc
.sched_name
);
2139 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2140 if (usvc
.fwmark
== 0)
2141 svc
= __ip_vs_service_find(usvc
.af
, usvc
.protocol
,
2142 &usvc
.addr
, usvc
.port
);
2144 svc
= __ip_vs_svc_fwm_find(usvc
.af
, usvc
.fwmark
);
2146 if (cmd
!= IP_VS_SO_SET_ADD
2147 && (svc
== NULL
|| svc
->protocol
!= usvc
.protocol
)) {
2153 case IP_VS_SO_SET_ADD
:
2157 ret
= ip_vs_add_service(&usvc
, &svc
);
2159 case IP_VS_SO_SET_EDIT
:
2160 ret
= ip_vs_edit_service(svc
, &usvc
);
2162 case IP_VS_SO_SET_DEL
:
2163 ret
= ip_vs_del_service(svc
);
2167 case IP_VS_SO_SET_ZERO
:
2168 ret
= ip_vs_zero_service(svc
);
2170 case IP_VS_SO_SET_ADDDEST
:
2171 ret
= ip_vs_add_dest(svc
, &udest
);
2173 case IP_VS_SO_SET_EDITDEST
:
2174 ret
= ip_vs_edit_dest(svc
, &udest
);
2176 case IP_VS_SO_SET_DELDEST
:
2177 ret
= ip_vs_del_dest(svc
, &udest
);
2184 mutex_unlock(&__ip_vs_mutex
);
2186 /* decrease the module use count */
2187 ip_vs_use_count_dec();
2194 ip_vs_copy_stats(struct ip_vs_stats_user
*dst
, struct ip_vs_stats
*src
)
2196 spin_lock_bh(&src
->lock
);
2197 memcpy(dst
, &src
->ustats
, sizeof(*dst
));
2198 spin_unlock_bh(&src
->lock
);
2202 ip_vs_copy_service(struct ip_vs_service_entry
*dst
, struct ip_vs_service
*src
)
2204 dst
->protocol
= src
->protocol
;
2205 dst
->addr
= src
->addr
.ip
;
2206 dst
->port
= src
->port
;
2207 dst
->fwmark
= src
->fwmark
;
2208 strlcpy(dst
->sched_name
, src
->scheduler
->name
, sizeof(dst
->sched_name
));
2209 dst
->flags
= src
->flags
;
2210 dst
->timeout
= src
->timeout
/ HZ
;
2211 dst
->netmask
= src
->netmask
;
2212 dst
->num_dests
= src
->num_dests
;
2213 ip_vs_copy_stats(&dst
->stats
, &src
->stats
);
2217 __ip_vs_get_service_entries(const struct ip_vs_get_services
*get
,
2218 struct ip_vs_get_services __user
*uptr
)
2221 struct ip_vs_service
*svc
;
2222 struct ip_vs_service_entry entry
;
2225 for (idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
2226 list_for_each_entry(svc
, &ip_vs_svc_table
[idx
], s_list
) {
2227 /* Only expose IPv4 entries to old interface */
2228 if (svc
->af
!= AF_INET
)
2231 if (count
>= get
->num_services
)
2233 memset(&entry
, 0, sizeof(entry
));
2234 ip_vs_copy_service(&entry
, svc
);
2235 if (copy_to_user(&uptr
->entrytable
[count
],
2236 &entry
, sizeof(entry
))) {
2244 for (idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
2245 list_for_each_entry(svc
, &ip_vs_svc_fwm_table
[idx
], f_list
) {
2246 /* Only expose IPv4 entries to old interface */
2247 if (svc
->af
!= AF_INET
)
2250 if (count
>= get
->num_services
)
2252 memset(&entry
, 0, sizeof(entry
));
2253 ip_vs_copy_service(&entry
, svc
);
2254 if (copy_to_user(&uptr
->entrytable
[count
],
2255 &entry
, sizeof(entry
))) {
2267 __ip_vs_get_dest_entries(const struct ip_vs_get_dests
*get
,
2268 struct ip_vs_get_dests __user
*uptr
)
2270 struct ip_vs_service
*svc
;
2271 union nf_inet_addr addr
= { .ip
= get
->addr
};
2275 svc
= __ip_vs_svc_fwm_find(AF_INET
, get
->fwmark
);
2277 svc
= __ip_vs_service_find(AF_INET
, get
->protocol
, &addr
,
2282 struct ip_vs_dest
*dest
;
2283 struct ip_vs_dest_entry entry
;
2285 list_for_each_entry(dest
, &svc
->destinations
, n_list
) {
2286 if (count
>= get
->num_dests
)
2289 entry
.addr
= dest
->addr
.ip
;
2290 entry
.port
= dest
->port
;
2291 entry
.conn_flags
= atomic_read(&dest
->conn_flags
);
2292 entry
.weight
= atomic_read(&dest
->weight
);
2293 entry
.u_threshold
= dest
->u_threshold
;
2294 entry
.l_threshold
= dest
->l_threshold
;
2295 entry
.activeconns
= atomic_read(&dest
->activeconns
);
2296 entry
.inactconns
= atomic_read(&dest
->inactconns
);
2297 entry
.persistconns
= atomic_read(&dest
->persistconns
);
2298 ip_vs_copy_stats(&entry
.stats
, &dest
->stats
);
2299 if (copy_to_user(&uptr
->entrytable
[count
],
2300 &entry
, sizeof(entry
))) {
2312 __ip_vs_get_timeouts(struct ip_vs_timeout_user
*u
)
2314 #ifdef CONFIG_IP_VS_PROTO_TCP
2316 ip_vs_protocol_tcp
.timeout_table
[IP_VS_TCP_S_ESTABLISHED
] / HZ
;
2317 u
->tcp_fin_timeout
=
2318 ip_vs_protocol_tcp
.timeout_table
[IP_VS_TCP_S_FIN_WAIT
] / HZ
;
2320 #ifdef CONFIG_IP_VS_PROTO_UDP
2322 ip_vs_protocol_udp
.timeout_table
[IP_VS_UDP_S_NORMAL
] / HZ
;
2327 #define GET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2328 #define GET_INFO_ARG_LEN (sizeof(struct ip_vs_getinfo))
2329 #define GET_SERVICES_ARG_LEN (sizeof(struct ip_vs_get_services))
2330 #define GET_SERVICE_ARG_LEN (sizeof(struct ip_vs_service_entry))
2331 #define GET_DESTS_ARG_LEN (sizeof(struct ip_vs_get_dests))
2332 #define GET_TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2333 #define GET_DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user) * 2)
2335 static const unsigned char get_arglen
[GET_CMDID(IP_VS_SO_GET_MAX
)+1] = {
2336 [GET_CMDID(IP_VS_SO_GET_VERSION
)] = 64,
2337 [GET_CMDID(IP_VS_SO_GET_INFO
)] = GET_INFO_ARG_LEN
,
2338 [GET_CMDID(IP_VS_SO_GET_SERVICES
)] = GET_SERVICES_ARG_LEN
,
2339 [GET_CMDID(IP_VS_SO_GET_SERVICE
)] = GET_SERVICE_ARG_LEN
,
2340 [GET_CMDID(IP_VS_SO_GET_DESTS
)] = GET_DESTS_ARG_LEN
,
2341 [GET_CMDID(IP_VS_SO_GET_TIMEOUT
)] = GET_TIMEOUT_ARG_LEN
,
2342 [GET_CMDID(IP_VS_SO_GET_DAEMON
)] = GET_DAEMON_ARG_LEN
,
2346 do_ip_vs_get_ctl(struct sock
*sk
, int cmd
, void __user
*user
, int *len
)
2348 unsigned char arg
[128];
2350 unsigned int copylen
;
2352 if (!capable(CAP_NET_ADMIN
))
2355 if (cmd
< IP_VS_BASE_CTL
|| cmd
> IP_VS_SO_GET_MAX
)
2358 if (*len
< get_arglen
[GET_CMDID(cmd
)]) {
2359 pr_err("get_ctl: len %u < %u\n",
2360 *len
, get_arglen
[GET_CMDID(cmd
)]);
2364 copylen
= get_arglen
[GET_CMDID(cmd
)];
2368 if (copy_from_user(arg
, user
, copylen
) != 0)
2371 if (mutex_lock_interruptible(&__ip_vs_mutex
))
2372 return -ERESTARTSYS
;
2375 case IP_VS_SO_GET_VERSION
:
2379 sprintf(buf
, "IP Virtual Server version %d.%d.%d (size=%d)",
2380 NVERSION(IP_VS_VERSION_CODE
), ip_vs_conn_tab_size
);
2381 if (copy_to_user(user
, buf
, strlen(buf
)+1) != 0) {
2385 *len
= strlen(buf
)+1;
2389 case IP_VS_SO_GET_INFO
:
2391 struct ip_vs_getinfo info
;
2392 info
.version
= IP_VS_VERSION_CODE
;
2393 info
.size
= ip_vs_conn_tab_size
;
2394 info
.num_services
= ip_vs_num_services
;
2395 if (copy_to_user(user
, &info
, sizeof(info
)) != 0)
2400 case IP_VS_SO_GET_SERVICES
:
2402 struct ip_vs_get_services
*get
;
2405 get
= (struct ip_vs_get_services
*)arg
;
2406 size
= sizeof(*get
) +
2407 sizeof(struct ip_vs_service_entry
) * get
->num_services
;
2409 pr_err("length: %u != %u\n", *len
, size
);
2413 ret
= __ip_vs_get_service_entries(get
, user
);
2417 case IP_VS_SO_GET_SERVICE
:
2419 struct ip_vs_service_entry
*entry
;
2420 struct ip_vs_service
*svc
;
2421 union nf_inet_addr addr
;
2423 entry
= (struct ip_vs_service_entry
*)arg
;
2424 addr
.ip
= entry
->addr
;
2426 svc
= __ip_vs_svc_fwm_find(AF_INET
, entry
->fwmark
);
2428 svc
= __ip_vs_service_find(AF_INET
, entry
->protocol
,
2429 &addr
, entry
->port
);
2431 ip_vs_copy_service(entry
, svc
);
2432 if (copy_to_user(user
, entry
, sizeof(*entry
)) != 0)
2439 case IP_VS_SO_GET_DESTS
:
2441 struct ip_vs_get_dests
*get
;
2444 get
= (struct ip_vs_get_dests
*)arg
;
2445 size
= sizeof(*get
) +
2446 sizeof(struct ip_vs_dest_entry
) * get
->num_dests
;
2448 pr_err("length: %u != %u\n", *len
, size
);
2452 ret
= __ip_vs_get_dest_entries(get
, user
);
2456 case IP_VS_SO_GET_TIMEOUT
:
2458 struct ip_vs_timeout_user t
;
2460 __ip_vs_get_timeouts(&t
);
2461 if (copy_to_user(user
, &t
, sizeof(t
)) != 0)
2466 case IP_VS_SO_GET_DAEMON
:
2468 struct ip_vs_daemon_user d
[2];
2470 memset(&d
, 0, sizeof(d
));
2471 if (ip_vs_sync_state
& IP_VS_STATE_MASTER
) {
2472 d
[0].state
= IP_VS_STATE_MASTER
;
2473 strlcpy(d
[0].mcast_ifn
, ip_vs_master_mcast_ifn
, sizeof(d
[0].mcast_ifn
));
2474 d
[0].syncid
= ip_vs_master_syncid
;
2476 if (ip_vs_sync_state
& IP_VS_STATE_BACKUP
) {
2477 d
[1].state
= IP_VS_STATE_BACKUP
;
2478 strlcpy(d
[1].mcast_ifn
, ip_vs_backup_mcast_ifn
, sizeof(d
[1].mcast_ifn
));
2479 d
[1].syncid
= ip_vs_backup_syncid
;
2481 if (copy_to_user(user
, &d
, sizeof(d
)) != 0)
2491 mutex_unlock(&__ip_vs_mutex
);
2496 static struct nf_sockopt_ops ip_vs_sockopts
= {
2498 .set_optmin
= IP_VS_BASE_CTL
,
2499 .set_optmax
= IP_VS_SO_SET_MAX
+1,
2500 .set
= do_ip_vs_set_ctl
,
2501 .get_optmin
= IP_VS_BASE_CTL
,
2502 .get_optmax
= IP_VS_SO_GET_MAX
+1,
2503 .get
= do_ip_vs_get_ctl
,
2504 .owner
= THIS_MODULE
,
2508 * Generic Netlink interface
2511 /* IPVS genetlink family */
2512 static struct genl_family ip_vs_genl_family
= {
2513 .id
= GENL_ID_GENERATE
,
2515 .name
= IPVS_GENL_NAME
,
2516 .version
= IPVS_GENL_VERSION
,
2517 .maxattr
= IPVS_CMD_MAX
,
2520 /* Policy used for first-level command attributes */
2521 static const struct nla_policy ip_vs_cmd_policy
[IPVS_CMD_ATTR_MAX
+ 1] = {
2522 [IPVS_CMD_ATTR_SERVICE
] = { .type
= NLA_NESTED
},
2523 [IPVS_CMD_ATTR_DEST
] = { .type
= NLA_NESTED
},
2524 [IPVS_CMD_ATTR_DAEMON
] = { .type
= NLA_NESTED
},
2525 [IPVS_CMD_ATTR_TIMEOUT_TCP
] = { .type
= NLA_U32
},
2526 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN
] = { .type
= NLA_U32
},
2527 [IPVS_CMD_ATTR_TIMEOUT_UDP
] = { .type
= NLA_U32
},
2530 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2531 static const struct nla_policy ip_vs_daemon_policy
[IPVS_DAEMON_ATTR_MAX
+ 1] = {
2532 [IPVS_DAEMON_ATTR_STATE
] = { .type
= NLA_U32
},
2533 [IPVS_DAEMON_ATTR_MCAST_IFN
] = { .type
= NLA_NUL_STRING
,
2534 .len
= IP_VS_IFNAME_MAXLEN
},
2535 [IPVS_DAEMON_ATTR_SYNC_ID
] = { .type
= NLA_U32
},
2538 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2539 static const struct nla_policy ip_vs_svc_policy
[IPVS_SVC_ATTR_MAX
+ 1] = {
2540 [IPVS_SVC_ATTR_AF
] = { .type
= NLA_U16
},
2541 [IPVS_SVC_ATTR_PROTOCOL
] = { .type
= NLA_U16
},
2542 [IPVS_SVC_ATTR_ADDR
] = { .type
= NLA_BINARY
,
2543 .len
= sizeof(union nf_inet_addr
) },
2544 [IPVS_SVC_ATTR_PORT
] = { .type
= NLA_U16
},
2545 [IPVS_SVC_ATTR_FWMARK
] = { .type
= NLA_U32
},
2546 [IPVS_SVC_ATTR_SCHED_NAME
] = { .type
= NLA_NUL_STRING
,
2547 .len
= IP_VS_SCHEDNAME_MAXLEN
},
2548 [IPVS_SVC_ATTR_FLAGS
] = { .type
= NLA_BINARY
,
2549 .len
= sizeof(struct ip_vs_flags
) },
2550 [IPVS_SVC_ATTR_TIMEOUT
] = { .type
= NLA_U32
},
2551 [IPVS_SVC_ATTR_NETMASK
] = { .type
= NLA_U32
},
2552 [IPVS_SVC_ATTR_STATS
] = { .type
= NLA_NESTED
},
2555 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2556 static const struct nla_policy ip_vs_dest_policy
[IPVS_DEST_ATTR_MAX
+ 1] = {
2557 [IPVS_DEST_ATTR_ADDR
] = { .type
= NLA_BINARY
,
2558 .len
= sizeof(union nf_inet_addr
) },
2559 [IPVS_DEST_ATTR_PORT
] = { .type
= NLA_U16
},
2560 [IPVS_DEST_ATTR_FWD_METHOD
] = { .type
= NLA_U32
},
2561 [IPVS_DEST_ATTR_WEIGHT
] = { .type
= NLA_U32
},
2562 [IPVS_DEST_ATTR_U_THRESH
] = { .type
= NLA_U32
},
2563 [IPVS_DEST_ATTR_L_THRESH
] = { .type
= NLA_U32
},
2564 [IPVS_DEST_ATTR_ACTIVE_CONNS
] = { .type
= NLA_U32
},
2565 [IPVS_DEST_ATTR_INACT_CONNS
] = { .type
= NLA_U32
},
2566 [IPVS_DEST_ATTR_PERSIST_CONNS
] = { .type
= NLA_U32
},
2567 [IPVS_DEST_ATTR_STATS
] = { .type
= NLA_NESTED
},
2570 static int ip_vs_genl_fill_stats(struct sk_buff
*skb
, int container_type
,
2571 struct ip_vs_stats
*stats
)
2573 struct nlattr
*nl_stats
= nla_nest_start(skb
, container_type
);
2577 spin_lock_bh(&stats
->lock
);
2579 NLA_PUT_U32(skb
, IPVS_STATS_ATTR_CONNS
, stats
->ustats
.conns
);
2580 NLA_PUT_U32(skb
, IPVS_STATS_ATTR_INPKTS
, stats
->ustats
.inpkts
);
2581 NLA_PUT_U32(skb
, IPVS_STATS_ATTR_OUTPKTS
, stats
->ustats
.outpkts
);
2582 NLA_PUT_U64(skb
, IPVS_STATS_ATTR_INBYTES
, stats
->ustats
.inbytes
);
2583 NLA_PUT_U64(skb
, IPVS_STATS_ATTR_OUTBYTES
, stats
->ustats
.outbytes
);
2584 NLA_PUT_U32(skb
, IPVS_STATS_ATTR_CPS
, stats
->ustats
.cps
);
2585 NLA_PUT_U32(skb
, IPVS_STATS_ATTR_INPPS
, stats
->ustats
.inpps
);
2586 NLA_PUT_U32(skb
, IPVS_STATS_ATTR_OUTPPS
, stats
->ustats
.outpps
);
2587 NLA_PUT_U32(skb
, IPVS_STATS_ATTR_INBPS
, stats
->ustats
.inbps
);
2588 NLA_PUT_U32(skb
, IPVS_STATS_ATTR_OUTBPS
, stats
->ustats
.outbps
);
2590 spin_unlock_bh(&stats
->lock
);
2592 nla_nest_end(skb
, nl_stats
);
2597 spin_unlock_bh(&stats
->lock
);
2598 nla_nest_cancel(skb
, nl_stats
);
2602 static int ip_vs_genl_fill_service(struct sk_buff
*skb
,
2603 struct ip_vs_service
*svc
)
2605 struct nlattr
*nl_service
;
2606 struct ip_vs_flags flags
= { .flags
= svc
->flags
,
2609 nl_service
= nla_nest_start(skb
, IPVS_CMD_ATTR_SERVICE
);
2613 NLA_PUT_U16(skb
, IPVS_SVC_ATTR_AF
, svc
->af
);
2616 NLA_PUT_U32(skb
, IPVS_SVC_ATTR_FWMARK
, svc
->fwmark
);
2618 NLA_PUT_U16(skb
, IPVS_SVC_ATTR_PROTOCOL
, svc
->protocol
);
2619 NLA_PUT(skb
, IPVS_SVC_ATTR_ADDR
, sizeof(svc
->addr
), &svc
->addr
);
2620 NLA_PUT_U16(skb
, IPVS_SVC_ATTR_PORT
, svc
->port
);
2623 NLA_PUT_STRING(skb
, IPVS_SVC_ATTR_SCHED_NAME
, svc
->scheduler
->name
);
2624 NLA_PUT(skb
, IPVS_SVC_ATTR_FLAGS
, sizeof(flags
), &flags
);
2625 NLA_PUT_U32(skb
, IPVS_SVC_ATTR_TIMEOUT
, svc
->timeout
/ HZ
);
2626 NLA_PUT_U32(skb
, IPVS_SVC_ATTR_NETMASK
, svc
->netmask
);
2628 if (ip_vs_genl_fill_stats(skb
, IPVS_SVC_ATTR_STATS
, &svc
->stats
))
2629 goto nla_put_failure
;
2631 nla_nest_end(skb
, nl_service
);
2636 nla_nest_cancel(skb
, nl_service
);
2640 static int ip_vs_genl_dump_service(struct sk_buff
*skb
,
2641 struct ip_vs_service
*svc
,
2642 struct netlink_callback
*cb
)
2646 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).pid
, cb
->nlh
->nlmsg_seq
,
2647 &ip_vs_genl_family
, NLM_F_MULTI
,
2648 IPVS_CMD_NEW_SERVICE
);
2652 if (ip_vs_genl_fill_service(skb
, svc
) < 0)
2653 goto nla_put_failure
;
2655 return genlmsg_end(skb
, hdr
);
2658 genlmsg_cancel(skb
, hdr
);
2662 static int ip_vs_genl_dump_services(struct sk_buff
*skb
,
2663 struct netlink_callback
*cb
)
2666 int start
= cb
->args
[0];
2667 struct ip_vs_service
*svc
;
2669 mutex_lock(&__ip_vs_mutex
);
2670 for (i
= 0; i
< IP_VS_SVC_TAB_SIZE
; i
++) {
2671 list_for_each_entry(svc
, &ip_vs_svc_table
[i
], s_list
) {
2674 if (ip_vs_genl_dump_service(skb
, svc
, cb
) < 0) {
2676 goto nla_put_failure
;
2681 for (i
= 0; i
< IP_VS_SVC_TAB_SIZE
; i
++) {
2682 list_for_each_entry(svc
, &ip_vs_svc_fwm_table
[i
], f_list
) {
2685 if (ip_vs_genl_dump_service(skb
, svc
, cb
) < 0) {
2687 goto nla_put_failure
;
2693 mutex_unlock(&__ip_vs_mutex
);
2699 static int ip_vs_genl_parse_service(struct ip_vs_service_user_kern
*usvc
,
2700 struct nlattr
*nla
, int full_entry
,
2701 struct ip_vs_service
**ret_svc
)
2703 struct nlattr
*attrs
[IPVS_SVC_ATTR_MAX
+ 1];
2704 struct nlattr
*nla_af
, *nla_port
, *nla_fwmark
, *nla_protocol
, *nla_addr
;
2705 struct ip_vs_service
*svc
;
2707 /* Parse mandatory identifying service fields first */
2709 nla_parse_nested(attrs
, IPVS_SVC_ATTR_MAX
, nla
, ip_vs_svc_policy
))
2712 nla_af
= attrs
[IPVS_SVC_ATTR_AF
];
2713 nla_protocol
= attrs
[IPVS_SVC_ATTR_PROTOCOL
];
2714 nla_addr
= attrs
[IPVS_SVC_ATTR_ADDR
];
2715 nla_port
= attrs
[IPVS_SVC_ATTR_PORT
];
2716 nla_fwmark
= attrs
[IPVS_SVC_ATTR_FWMARK
];
2718 if (!(nla_af
&& (nla_fwmark
|| (nla_port
&& nla_protocol
&& nla_addr
))))
2721 memset(usvc
, 0, sizeof(*usvc
));
2723 usvc
->af
= nla_get_u16(nla_af
);
2724 #ifdef CONFIG_IP_VS_IPV6
2725 if (usvc
->af
!= AF_INET
&& usvc
->af
!= AF_INET6
)
2727 if (usvc
->af
!= AF_INET
)
2729 return -EAFNOSUPPORT
;
2732 usvc
->protocol
= IPPROTO_TCP
;
2733 usvc
->fwmark
= nla_get_u32(nla_fwmark
);
2735 usvc
->protocol
= nla_get_u16(nla_protocol
);
2736 nla_memcpy(&usvc
->addr
, nla_addr
, sizeof(usvc
->addr
));
2737 usvc
->port
= nla_get_u16(nla_port
);
2742 svc
= __ip_vs_svc_fwm_find(usvc
->af
, usvc
->fwmark
);
2744 svc
= __ip_vs_service_find(usvc
->af
, usvc
->protocol
,
2745 &usvc
->addr
, usvc
->port
);
2748 /* If a full entry was requested, check for the additional fields */
2750 struct nlattr
*nla_sched
, *nla_flags
, *nla_timeout
,
2752 struct ip_vs_flags flags
;
2754 nla_sched
= attrs
[IPVS_SVC_ATTR_SCHED_NAME
];
2755 nla_flags
= attrs
[IPVS_SVC_ATTR_FLAGS
];
2756 nla_timeout
= attrs
[IPVS_SVC_ATTR_TIMEOUT
];
2757 nla_netmask
= attrs
[IPVS_SVC_ATTR_NETMASK
];
2759 if (!(nla_sched
&& nla_flags
&& nla_timeout
&& nla_netmask
))
2762 nla_memcpy(&flags
, nla_flags
, sizeof(flags
));
2764 /* prefill flags from service if it already exists */
2766 usvc
->flags
= svc
->flags
;
2768 /* set new flags from userland */
2769 usvc
->flags
= (usvc
->flags
& ~flags
.mask
) |
2770 (flags
.flags
& flags
.mask
);
2771 usvc
->sched_name
= nla_data(nla_sched
);
2772 usvc
->timeout
= nla_get_u32(nla_timeout
);
2773 usvc
->netmask
= nla_get_u32(nla_netmask
);
2779 static struct ip_vs_service
*ip_vs_genl_find_service(struct nlattr
*nla
)
2781 struct ip_vs_service_user_kern usvc
;
2782 struct ip_vs_service
*svc
;
2785 ret
= ip_vs_genl_parse_service(&usvc
, nla
, 0, &svc
);
2786 return ret
? ERR_PTR(ret
) : svc
;
2789 static int ip_vs_genl_fill_dest(struct sk_buff
*skb
, struct ip_vs_dest
*dest
)
2791 struct nlattr
*nl_dest
;
2793 nl_dest
= nla_nest_start(skb
, IPVS_CMD_ATTR_DEST
);
2797 NLA_PUT(skb
, IPVS_DEST_ATTR_ADDR
, sizeof(dest
->addr
), &dest
->addr
);
2798 NLA_PUT_U16(skb
, IPVS_DEST_ATTR_PORT
, dest
->port
);
2800 NLA_PUT_U32(skb
, IPVS_DEST_ATTR_FWD_METHOD
,
2801 atomic_read(&dest
->conn_flags
) & IP_VS_CONN_F_FWD_MASK
);
2802 NLA_PUT_U32(skb
, IPVS_DEST_ATTR_WEIGHT
, atomic_read(&dest
->weight
));
2803 NLA_PUT_U32(skb
, IPVS_DEST_ATTR_U_THRESH
, dest
->u_threshold
);
2804 NLA_PUT_U32(skb
, IPVS_DEST_ATTR_L_THRESH
, dest
->l_threshold
);
2805 NLA_PUT_U32(skb
, IPVS_DEST_ATTR_ACTIVE_CONNS
,
2806 atomic_read(&dest
->activeconns
));
2807 NLA_PUT_U32(skb
, IPVS_DEST_ATTR_INACT_CONNS
,
2808 atomic_read(&dest
->inactconns
));
2809 NLA_PUT_U32(skb
, IPVS_DEST_ATTR_PERSIST_CONNS
,
2810 atomic_read(&dest
->persistconns
));
2812 if (ip_vs_genl_fill_stats(skb
, IPVS_DEST_ATTR_STATS
, &dest
->stats
))
2813 goto nla_put_failure
;
2815 nla_nest_end(skb
, nl_dest
);
2820 nla_nest_cancel(skb
, nl_dest
);
2824 static int ip_vs_genl_dump_dest(struct sk_buff
*skb
, struct ip_vs_dest
*dest
,
2825 struct netlink_callback
*cb
)
2829 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).pid
, cb
->nlh
->nlmsg_seq
,
2830 &ip_vs_genl_family
, NLM_F_MULTI
,
2835 if (ip_vs_genl_fill_dest(skb
, dest
) < 0)
2836 goto nla_put_failure
;
2838 return genlmsg_end(skb
, hdr
);
2841 genlmsg_cancel(skb
, hdr
);
2845 static int ip_vs_genl_dump_dests(struct sk_buff
*skb
,
2846 struct netlink_callback
*cb
)
2849 int start
= cb
->args
[0];
2850 struct ip_vs_service
*svc
;
2851 struct ip_vs_dest
*dest
;
2852 struct nlattr
*attrs
[IPVS_CMD_ATTR_MAX
+ 1];
2854 mutex_lock(&__ip_vs_mutex
);
2856 /* Try to find the service for which to dump destinations */
2857 if (nlmsg_parse(cb
->nlh
, GENL_HDRLEN
, attrs
,
2858 IPVS_CMD_ATTR_MAX
, ip_vs_cmd_policy
))
2861 svc
= ip_vs_genl_find_service(attrs
[IPVS_CMD_ATTR_SERVICE
]);
2862 if (IS_ERR(svc
) || svc
== NULL
)
2865 /* Dump the destinations */
2866 list_for_each_entry(dest
, &svc
->destinations
, n_list
) {
2869 if (ip_vs_genl_dump_dest(skb
, dest
, cb
) < 0) {
2871 goto nla_put_failure
;
2879 mutex_unlock(&__ip_vs_mutex
);
2884 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern
*udest
,
2885 struct nlattr
*nla
, int full_entry
)
2887 struct nlattr
*attrs
[IPVS_DEST_ATTR_MAX
+ 1];
2888 struct nlattr
*nla_addr
, *nla_port
;
2890 /* Parse mandatory identifying destination fields first */
2892 nla_parse_nested(attrs
, IPVS_DEST_ATTR_MAX
, nla
, ip_vs_dest_policy
))
2895 nla_addr
= attrs
[IPVS_DEST_ATTR_ADDR
];
2896 nla_port
= attrs
[IPVS_DEST_ATTR_PORT
];
2898 if (!(nla_addr
&& nla_port
))
2901 memset(udest
, 0, sizeof(*udest
));
2903 nla_memcpy(&udest
->addr
, nla_addr
, sizeof(udest
->addr
));
2904 udest
->port
= nla_get_u16(nla_port
);
2906 /* If a full entry was requested, check for the additional fields */
2908 struct nlattr
*nla_fwd
, *nla_weight
, *nla_u_thresh
,
2911 nla_fwd
= attrs
[IPVS_DEST_ATTR_FWD_METHOD
];
2912 nla_weight
= attrs
[IPVS_DEST_ATTR_WEIGHT
];
2913 nla_u_thresh
= attrs
[IPVS_DEST_ATTR_U_THRESH
];
2914 nla_l_thresh
= attrs
[IPVS_DEST_ATTR_L_THRESH
];
2916 if (!(nla_fwd
&& nla_weight
&& nla_u_thresh
&& nla_l_thresh
))
2919 udest
->conn_flags
= nla_get_u32(nla_fwd
)
2920 & IP_VS_CONN_F_FWD_MASK
;
2921 udest
->weight
= nla_get_u32(nla_weight
);
2922 udest
->u_threshold
= nla_get_u32(nla_u_thresh
);
2923 udest
->l_threshold
= nla_get_u32(nla_l_thresh
);
2929 static int ip_vs_genl_fill_daemon(struct sk_buff
*skb
, __be32 state
,
2930 const char *mcast_ifn
, __be32 syncid
)
2932 struct nlattr
*nl_daemon
;
2934 nl_daemon
= nla_nest_start(skb
, IPVS_CMD_ATTR_DAEMON
);
2938 NLA_PUT_U32(skb
, IPVS_DAEMON_ATTR_STATE
, state
);
2939 NLA_PUT_STRING(skb
, IPVS_DAEMON_ATTR_MCAST_IFN
, mcast_ifn
);
2940 NLA_PUT_U32(skb
, IPVS_DAEMON_ATTR_SYNC_ID
, syncid
);
2942 nla_nest_end(skb
, nl_daemon
);
2947 nla_nest_cancel(skb
, nl_daemon
);
2951 static int ip_vs_genl_dump_daemon(struct sk_buff
*skb
, __be32 state
,
2952 const char *mcast_ifn
, __be32 syncid
,
2953 struct netlink_callback
*cb
)
2956 hdr
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).pid
, cb
->nlh
->nlmsg_seq
,
2957 &ip_vs_genl_family
, NLM_F_MULTI
,
2958 IPVS_CMD_NEW_DAEMON
);
2962 if (ip_vs_genl_fill_daemon(skb
, state
, mcast_ifn
, syncid
))
2963 goto nla_put_failure
;
2965 return genlmsg_end(skb
, hdr
);
2968 genlmsg_cancel(skb
, hdr
);
2972 static int ip_vs_genl_dump_daemons(struct sk_buff
*skb
,
2973 struct netlink_callback
*cb
)
2975 mutex_lock(&__ip_vs_mutex
);
2976 if ((ip_vs_sync_state
& IP_VS_STATE_MASTER
) && !cb
->args
[0]) {
2977 if (ip_vs_genl_dump_daemon(skb
, IP_VS_STATE_MASTER
,
2978 ip_vs_master_mcast_ifn
,
2979 ip_vs_master_syncid
, cb
) < 0)
2980 goto nla_put_failure
;
2985 if ((ip_vs_sync_state
& IP_VS_STATE_BACKUP
) && !cb
->args
[1]) {
2986 if (ip_vs_genl_dump_daemon(skb
, IP_VS_STATE_BACKUP
,
2987 ip_vs_backup_mcast_ifn
,
2988 ip_vs_backup_syncid
, cb
) < 0)
2989 goto nla_put_failure
;
2995 mutex_unlock(&__ip_vs_mutex
);
3000 static int ip_vs_genl_new_daemon(struct nlattr
**attrs
)
3002 if (!(attrs
[IPVS_DAEMON_ATTR_STATE
] &&
3003 attrs
[IPVS_DAEMON_ATTR_MCAST_IFN
] &&
3004 attrs
[IPVS_DAEMON_ATTR_SYNC_ID
]))
3007 return start_sync_thread(nla_get_u32(attrs
[IPVS_DAEMON_ATTR_STATE
]),
3008 nla_data(attrs
[IPVS_DAEMON_ATTR_MCAST_IFN
]),
3009 nla_get_u32(attrs
[IPVS_DAEMON_ATTR_SYNC_ID
]));
3012 static int ip_vs_genl_del_daemon(struct nlattr
**attrs
)
3014 if (!attrs
[IPVS_DAEMON_ATTR_STATE
])
3017 return stop_sync_thread(nla_get_u32(attrs
[IPVS_DAEMON_ATTR_STATE
]));
3020 static int ip_vs_genl_set_config(struct nlattr
**attrs
)
3022 struct ip_vs_timeout_user t
;
3024 __ip_vs_get_timeouts(&t
);
3026 if (attrs
[IPVS_CMD_ATTR_TIMEOUT_TCP
])
3027 t
.tcp_timeout
= nla_get_u32(attrs
[IPVS_CMD_ATTR_TIMEOUT_TCP
]);
3029 if (attrs
[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN
])
3031 nla_get_u32(attrs
[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN
]);
3033 if (attrs
[IPVS_CMD_ATTR_TIMEOUT_UDP
])
3034 t
.udp_timeout
= nla_get_u32(attrs
[IPVS_CMD_ATTR_TIMEOUT_UDP
]);
3036 return ip_vs_set_timeout(&t
);
3039 static int ip_vs_genl_set_cmd(struct sk_buff
*skb
, struct genl_info
*info
)
3041 struct ip_vs_service
*svc
= NULL
;
3042 struct ip_vs_service_user_kern usvc
;
3043 struct ip_vs_dest_user_kern udest
;
3045 int need_full_svc
= 0, need_full_dest
= 0;
3047 cmd
= info
->genlhdr
->cmd
;
3049 mutex_lock(&__ip_vs_mutex
);
3051 if (cmd
== IPVS_CMD_FLUSH
) {
3052 ret
= ip_vs_flush();
3054 } else if (cmd
== IPVS_CMD_SET_CONFIG
) {
3055 ret
= ip_vs_genl_set_config(info
->attrs
);
3057 } else if (cmd
== IPVS_CMD_NEW_DAEMON
||
3058 cmd
== IPVS_CMD_DEL_DAEMON
) {
3060 struct nlattr
*daemon_attrs
[IPVS_DAEMON_ATTR_MAX
+ 1];
3062 if (!info
->attrs
[IPVS_CMD_ATTR_DAEMON
] ||
3063 nla_parse_nested(daemon_attrs
, IPVS_DAEMON_ATTR_MAX
,
3064 info
->attrs
[IPVS_CMD_ATTR_DAEMON
],
3065 ip_vs_daemon_policy
)) {
3070 if (cmd
== IPVS_CMD_NEW_DAEMON
)
3071 ret
= ip_vs_genl_new_daemon(daemon_attrs
);
3073 ret
= ip_vs_genl_del_daemon(daemon_attrs
);
3075 } else if (cmd
== IPVS_CMD_ZERO
&&
3076 !info
->attrs
[IPVS_CMD_ATTR_SERVICE
]) {
3077 ret
= ip_vs_zero_all();
3081 /* All following commands require a service argument, so check if we
3082 * received a valid one. We need a full service specification when
3083 * adding / editing a service. Only identifying members otherwise. */
3084 if (cmd
== IPVS_CMD_NEW_SERVICE
|| cmd
== IPVS_CMD_SET_SERVICE
)
3087 ret
= ip_vs_genl_parse_service(&usvc
,
3088 info
->attrs
[IPVS_CMD_ATTR_SERVICE
],
3089 need_full_svc
, &svc
);
3093 /* Unless we're adding a new service, the service must already exist */
3094 if ((cmd
!= IPVS_CMD_NEW_SERVICE
) && (svc
== NULL
)) {
3099 /* Destination commands require a valid destination argument. For
3100 * adding / editing a destination, we need a full destination
3102 if (cmd
== IPVS_CMD_NEW_DEST
|| cmd
== IPVS_CMD_SET_DEST
||
3103 cmd
== IPVS_CMD_DEL_DEST
) {
3104 if (cmd
!= IPVS_CMD_DEL_DEST
)
3107 ret
= ip_vs_genl_parse_dest(&udest
,
3108 info
->attrs
[IPVS_CMD_ATTR_DEST
],
3115 case IPVS_CMD_NEW_SERVICE
:
3117 ret
= ip_vs_add_service(&usvc
, &svc
);
3121 case IPVS_CMD_SET_SERVICE
:
3122 ret
= ip_vs_edit_service(svc
, &usvc
);
3124 case IPVS_CMD_DEL_SERVICE
:
3125 ret
= ip_vs_del_service(svc
);
3126 /* do not use svc, it can be freed */
3128 case IPVS_CMD_NEW_DEST
:
3129 ret
= ip_vs_add_dest(svc
, &udest
);
3131 case IPVS_CMD_SET_DEST
:
3132 ret
= ip_vs_edit_dest(svc
, &udest
);
3134 case IPVS_CMD_DEL_DEST
:
3135 ret
= ip_vs_del_dest(svc
, &udest
);
3138 ret
= ip_vs_zero_service(svc
);
3145 mutex_unlock(&__ip_vs_mutex
);
3150 static int ip_vs_genl_get_cmd(struct sk_buff
*skb
, struct genl_info
*info
)
3152 struct sk_buff
*msg
;
3154 int ret
, cmd
, reply_cmd
;
3156 cmd
= info
->genlhdr
->cmd
;
3158 if (cmd
== IPVS_CMD_GET_SERVICE
)
3159 reply_cmd
= IPVS_CMD_NEW_SERVICE
;
3160 else if (cmd
== IPVS_CMD_GET_INFO
)
3161 reply_cmd
= IPVS_CMD_SET_INFO
;
3162 else if (cmd
== IPVS_CMD_GET_CONFIG
)
3163 reply_cmd
= IPVS_CMD_SET_CONFIG
;
3165 pr_err("unknown Generic Netlink command\n");
3169 msg
= nlmsg_new(NLMSG_DEFAULT_SIZE
, GFP_KERNEL
);
3173 mutex_lock(&__ip_vs_mutex
);
3175 reply
= genlmsg_put_reply(msg
, info
, &ip_vs_genl_family
, 0, reply_cmd
);
3177 goto nla_put_failure
;
3180 case IPVS_CMD_GET_SERVICE
:
3182 struct ip_vs_service
*svc
;
3184 svc
= ip_vs_genl_find_service(info
->attrs
[IPVS_CMD_ATTR_SERVICE
]);
3189 ret
= ip_vs_genl_fill_service(msg
, svc
);
3191 goto nla_put_failure
;
3200 case IPVS_CMD_GET_CONFIG
:
3202 struct ip_vs_timeout_user t
;
3204 __ip_vs_get_timeouts(&t
);
3205 #ifdef CONFIG_IP_VS_PROTO_TCP
3206 NLA_PUT_U32(msg
, IPVS_CMD_ATTR_TIMEOUT_TCP
, t
.tcp_timeout
);
3207 NLA_PUT_U32(msg
, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN
,
3210 #ifdef CONFIG_IP_VS_PROTO_UDP
3211 NLA_PUT_U32(msg
, IPVS_CMD_ATTR_TIMEOUT_UDP
, t
.udp_timeout
);
3217 case IPVS_CMD_GET_INFO
:
3218 NLA_PUT_U32(msg
, IPVS_INFO_ATTR_VERSION
, IP_VS_VERSION_CODE
);
3219 NLA_PUT_U32(msg
, IPVS_INFO_ATTR_CONN_TAB_SIZE
,
3220 ip_vs_conn_tab_size
);
3224 genlmsg_end(msg
, reply
);
3225 ret
= genlmsg_reply(msg
, info
);
3229 pr_err("not enough space in Netlink message\n");
3235 mutex_unlock(&__ip_vs_mutex
);
3241 static struct genl_ops ip_vs_genl_ops
[] __read_mostly
= {
3243 .cmd
= IPVS_CMD_NEW_SERVICE
,
3244 .flags
= GENL_ADMIN_PERM
,
3245 .policy
= ip_vs_cmd_policy
,
3246 .doit
= ip_vs_genl_set_cmd
,
3249 .cmd
= IPVS_CMD_SET_SERVICE
,
3250 .flags
= GENL_ADMIN_PERM
,
3251 .policy
= ip_vs_cmd_policy
,
3252 .doit
= ip_vs_genl_set_cmd
,
3255 .cmd
= IPVS_CMD_DEL_SERVICE
,
3256 .flags
= GENL_ADMIN_PERM
,
3257 .policy
= ip_vs_cmd_policy
,
3258 .doit
= ip_vs_genl_set_cmd
,
3261 .cmd
= IPVS_CMD_GET_SERVICE
,
3262 .flags
= GENL_ADMIN_PERM
,
3263 .doit
= ip_vs_genl_get_cmd
,
3264 .dumpit
= ip_vs_genl_dump_services
,
3265 .policy
= ip_vs_cmd_policy
,
3268 .cmd
= IPVS_CMD_NEW_DEST
,
3269 .flags
= GENL_ADMIN_PERM
,
3270 .policy
= ip_vs_cmd_policy
,
3271 .doit
= ip_vs_genl_set_cmd
,
3274 .cmd
= IPVS_CMD_SET_DEST
,
3275 .flags
= GENL_ADMIN_PERM
,
3276 .policy
= ip_vs_cmd_policy
,
3277 .doit
= ip_vs_genl_set_cmd
,
3280 .cmd
= IPVS_CMD_DEL_DEST
,
3281 .flags
= GENL_ADMIN_PERM
,
3282 .policy
= ip_vs_cmd_policy
,
3283 .doit
= ip_vs_genl_set_cmd
,
3286 .cmd
= IPVS_CMD_GET_DEST
,
3287 .flags
= GENL_ADMIN_PERM
,
3288 .policy
= ip_vs_cmd_policy
,
3289 .dumpit
= ip_vs_genl_dump_dests
,
3292 .cmd
= IPVS_CMD_NEW_DAEMON
,
3293 .flags
= GENL_ADMIN_PERM
,
3294 .policy
= ip_vs_cmd_policy
,
3295 .doit
= ip_vs_genl_set_cmd
,
3298 .cmd
= IPVS_CMD_DEL_DAEMON
,
3299 .flags
= GENL_ADMIN_PERM
,
3300 .policy
= ip_vs_cmd_policy
,
3301 .doit
= ip_vs_genl_set_cmd
,
3304 .cmd
= IPVS_CMD_GET_DAEMON
,
3305 .flags
= GENL_ADMIN_PERM
,
3306 .dumpit
= ip_vs_genl_dump_daemons
,
3309 .cmd
= IPVS_CMD_SET_CONFIG
,
3310 .flags
= GENL_ADMIN_PERM
,
3311 .policy
= ip_vs_cmd_policy
,
3312 .doit
= ip_vs_genl_set_cmd
,
3315 .cmd
= IPVS_CMD_GET_CONFIG
,
3316 .flags
= GENL_ADMIN_PERM
,
3317 .doit
= ip_vs_genl_get_cmd
,
3320 .cmd
= IPVS_CMD_GET_INFO
,
3321 .flags
= GENL_ADMIN_PERM
,
3322 .doit
= ip_vs_genl_get_cmd
,
3325 .cmd
= IPVS_CMD_ZERO
,
3326 .flags
= GENL_ADMIN_PERM
,
3327 .policy
= ip_vs_cmd_policy
,
3328 .doit
= ip_vs_genl_set_cmd
,
3331 .cmd
= IPVS_CMD_FLUSH
,
3332 .flags
= GENL_ADMIN_PERM
,
3333 .doit
= ip_vs_genl_set_cmd
,
3337 static int __init
ip_vs_genl_register(void)
3339 return genl_register_family_with_ops(&ip_vs_genl_family
,
3340 ip_vs_genl_ops
, ARRAY_SIZE(ip_vs_genl_ops
));
3343 static void ip_vs_genl_unregister(void)
3345 genl_unregister_family(&ip_vs_genl_family
);
3348 /* End of Generic Netlink interface definitions */
3351 int __init
ip_vs_control_init(void)
3358 ret
= nf_register_sockopt(&ip_vs_sockopts
);
3360 pr_err("cannot register sockopt.\n");
3364 ret
= ip_vs_genl_register();
3366 pr_err("cannot register Generic Netlink interface.\n");
3367 nf_unregister_sockopt(&ip_vs_sockopts
);
3371 proc_net_fops_create(&init_net
, "ip_vs", 0, &ip_vs_info_fops
);
3372 proc_net_fops_create(&init_net
, "ip_vs_stats",0, &ip_vs_stats_fops
);
3374 sysctl_header
= register_sysctl_paths(net_vs_ctl_path
, vs_vars
);
3376 /* Initialize ip_vs_svc_table, ip_vs_svc_fwm_table, ip_vs_rtable */
3377 for(idx
= 0; idx
< IP_VS_SVC_TAB_SIZE
; idx
++) {
3378 INIT_LIST_HEAD(&ip_vs_svc_table
[idx
]);
3379 INIT_LIST_HEAD(&ip_vs_svc_fwm_table
[idx
]);
3381 for(idx
= 0; idx
< IP_VS_RTAB_SIZE
; idx
++) {
3382 INIT_LIST_HEAD(&ip_vs_rtable
[idx
]);
3385 ip_vs_new_estimator(&ip_vs_stats
);
3387 /* Hook the defense timer */
3388 schedule_delayed_work(&defense_work
, DEFENSE_TIMER_PERIOD
);
3395 void ip_vs_control_cleanup(void)
3398 ip_vs_trash_cleanup();
3399 cancel_rearming_delayed_work(&defense_work
);
3400 cancel_work_sync(&defense_work
.work
);
3401 ip_vs_kill_estimator(&ip_vs_stats
);
3402 unregister_sysctl_table(sysctl_header
);
3403 proc_net_remove(&init_net
, "ip_vs_stats");
3404 proc_net_remove(&init_net
, "ip_vs");
3405 ip_vs_genl_unregister();
3406 nf_unregister_sockopt(&ip_vs_sockopts
);