Merge branch 'for-2.6.38' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq
[linux-2.6.git] / net / netfilter / ipvs / ip_vs_ctl.c
1 /*
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
6  *              cluster of servers.
7  *
8  * Authors:     Wensong Zhang <wensong@linuxvirtualserver.org>
9  *              Peter Kese <peter.kese@ijs.si>
10  *              Julian Anastasov <ja@ssi.bg>
11  *
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.
16  *
17  * Changes:
18  *
19  */
20
21 #define KMSG_COMPONENT "IPVS"
22 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
23
24 #include <linux/module.h>
25 #include <linux/init.h>
26 #include <linux/types.h>
27 #include <linux/capability.h>
28 #include <linux/fs.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>
35
36 #include <linux/netfilter.h>
37 #include <linux/netfilter_ipv4.h>
38 #include <linux/mutex.h>
39
40 #include <net/net_namespace.h>
41 #include <net/ip.h>
42 #ifdef CONFIG_IP_VS_IPV6
43 #include <net/ipv6.h>
44 #include <net/ip6_route.h>
45 #endif
46 #include <net/route.h>
47 #include <net/sock.h>
48 #include <net/genetlink.h>
49
50 #include <asm/uaccess.h>
51
52 #include <net/ip_vs.h>
53
54 /* semaphore for IPVS sockopts. And, [gs]etsockopt may sleep. */
55 static DEFINE_MUTEX(__ip_vs_mutex);
56
57 /* lock for service table */
58 static DEFINE_RWLOCK(__ip_vs_svc_lock);
59
60 /* lock for table with the real services */
61 static DEFINE_RWLOCK(__ip_vs_rs_lock);
62
63 /* lock for state and timeout tables */
64 static DEFINE_SPINLOCK(ip_vs_securetcp_lock);
65
66 /* lock for drop entry handling */
67 static DEFINE_SPINLOCK(__ip_vs_dropentry_lock);
68
69 /* lock for drop packet handling */
70 static DEFINE_SPINLOCK(__ip_vs_droppacket_lock);
71
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);
76
77 /* number of virtual services */
78 static int ip_vs_num_services = 0;
79
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;
93 #endif
94 int sysctl_ip_vs_snat_reroute = 1;
95
96
97 #ifdef CONFIG_IP_VS_DEBUG
98 static int sysctl_ip_vs_debug_level = 0;
99
100 int ip_vs_get_debug_level(void)
101 {
102         return sysctl_ip_vs_debug_level;
103 }
104 #endif
105
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)
109 {
110         struct rt6_info *rt;
111         struct flowi fl = {
112                 .oif = 0,
113                 .fl6_dst = *addr,
114                 .fl6_src = { .s6_addr32 = {0, 0, 0, 0} },
115         };
116
117         rt = (struct rt6_info *)ip6_route_output(&init_net, NULL, &fl);
118         if (rt && rt->rt6i_dev && (rt->rt6i_dev->flags & IFF_LOOPBACK))
119                         return 1;
120
121         return 0;
122 }
123 #endif
124 /*
125  *      update_defense_level is called from keventd and from sysctl,
126  *      so it needs to protect itself from softirqs
127  */
128 static void update_defense_level(void)
129 {
130         struct sysinfo i;
131         static int old_secure_tcp = 0;
132         int availmem;
133         int nomem;
134         int to_change = -1;
135
136         /* we only count free and buffered memory (in pages) */
137         si_meminfo(&i);
138         availmem = i.freeram + i.bufferram;
139         /* however in linux 2.5 the i.bufferram is total page cache size,
140            we need adjust it */
141         /* si_swapinfo(&i); */
142         /* availmem = availmem - (i.totalswap - i.freeswap); */
143
144         nomem = (availmem < sysctl_ip_vs_amemthresh);
145
146         local_bh_disable();
147
148         /* drop_entry */
149         spin_lock(&__ip_vs_dropentry_lock);
150         switch (sysctl_ip_vs_drop_entry) {
151         case 0:
152                 atomic_set(&ip_vs_dropentry, 0);
153                 break;
154         case 1:
155                 if (nomem) {
156                         atomic_set(&ip_vs_dropentry, 1);
157                         sysctl_ip_vs_drop_entry = 2;
158                 } else {
159                         atomic_set(&ip_vs_dropentry, 0);
160                 }
161                 break;
162         case 2:
163                 if (nomem) {
164                         atomic_set(&ip_vs_dropentry, 1);
165                 } else {
166                         atomic_set(&ip_vs_dropentry, 0);
167                         sysctl_ip_vs_drop_entry = 1;
168                 };
169                 break;
170         case 3:
171                 atomic_set(&ip_vs_dropentry, 1);
172                 break;
173         }
174         spin_unlock(&__ip_vs_dropentry_lock);
175
176         /* drop_packet */
177         spin_lock(&__ip_vs_droppacket_lock);
178         switch (sysctl_ip_vs_drop_packet) {
179         case 0:
180                 ip_vs_drop_rate = 0;
181                 break;
182         case 1:
183                 if (nomem) {
184                         ip_vs_drop_rate = ip_vs_drop_counter
185                                 = sysctl_ip_vs_amemthresh /
186                                 (sysctl_ip_vs_amemthresh-availmem);
187                         sysctl_ip_vs_drop_packet = 2;
188                 } else {
189                         ip_vs_drop_rate = 0;
190                 }
191                 break;
192         case 2:
193                 if (nomem) {
194                         ip_vs_drop_rate = ip_vs_drop_counter
195                                 = sysctl_ip_vs_amemthresh /
196                                 (sysctl_ip_vs_amemthresh-availmem);
197                 } else {
198                         ip_vs_drop_rate = 0;
199                         sysctl_ip_vs_drop_packet = 1;
200                 }
201                 break;
202         case 3:
203                 ip_vs_drop_rate = sysctl_ip_vs_am_droprate;
204                 break;
205         }
206         spin_unlock(&__ip_vs_droppacket_lock);
207
208         /* secure_tcp */
209         spin_lock(&ip_vs_securetcp_lock);
210         switch (sysctl_ip_vs_secure_tcp) {
211         case 0:
212                 if (old_secure_tcp >= 2)
213                         to_change = 0;
214                 break;
215         case 1:
216                 if (nomem) {
217                         if (old_secure_tcp < 2)
218                                 to_change = 1;
219                         sysctl_ip_vs_secure_tcp = 2;
220                 } else {
221                         if (old_secure_tcp >= 2)
222                                 to_change = 0;
223                 }
224                 break;
225         case 2:
226                 if (nomem) {
227                         if (old_secure_tcp < 2)
228                                 to_change = 1;
229                 } else {
230                         if (old_secure_tcp >= 2)
231                                 to_change = 0;
232                         sysctl_ip_vs_secure_tcp = 1;
233                 }
234                 break;
235         case 3:
236                 if (old_secure_tcp < 2)
237                         to_change = 1;
238                 break;
239         }
240         old_secure_tcp = sysctl_ip_vs_secure_tcp;
241         if (to_change >= 0)
242                 ip_vs_protocol_timeout_change(sysctl_ip_vs_secure_tcp>1);
243         spin_unlock(&ip_vs_securetcp_lock);
244
245         local_bh_enable();
246 }
247
248
249 /*
250  *      Timer for checking the defense
251  */
252 #define DEFENSE_TIMER_PERIOD    1*HZ
253 static void defense_work_handler(struct work_struct *work);
254 static DECLARE_DELAYED_WORK(defense_work, defense_work_handler);
255
256 static void defense_work_handler(struct work_struct *work)
257 {
258         update_defense_level();
259         if (atomic_read(&ip_vs_dropentry))
260                 ip_vs_random_dropentry();
261
262         schedule_delayed_work(&defense_work, DEFENSE_TIMER_PERIOD);
263 }
264
265 int
266 ip_vs_use_count_inc(void)
267 {
268         return try_module_get(THIS_MODULE);
269 }
270
271 void
272 ip_vs_use_count_dec(void)
273 {
274         module_put(THIS_MODULE);
275 }
276
277
278 /*
279  *      Hash table: for virtual service lookups
280  */
281 #define IP_VS_SVC_TAB_BITS 8
282 #define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
283 #define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)
284
285 /* the service table hashed by <protocol, addr, port> */
286 static struct list_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE];
287 /* the service table hashed by fwmark */
288 static struct list_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE];
289
290 /*
291  *      Hash table: for real service lookups
292  */
293 #define IP_VS_RTAB_BITS 4
294 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
295 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
296
297 static struct list_head ip_vs_rtable[IP_VS_RTAB_SIZE];
298
299 /*
300  *      Trash for destinations
301  */
302 static LIST_HEAD(ip_vs_dest_trash);
303
304 /*
305  *      FTP & NULL virtual service counters
306  */
307 static atomic_t ip_vs_ftpsvc_counter = ATOMIC_INIT(0);
308 static atomic_t ip_vs_nullsvc_counter = ATOMIC_INIT(0);
309
310
311 /*
312  *      Returns hash value for virtual service
313  */
314 static __inline__ unsigned
315 ip_vs_svc_hashkey(int af, unsigned proto, const union nf_inet_addr *addr,
316                   __be16 port)
317 {
318         register unsigned porth = ntohs(port);
319         __be32 addr_fold = addr->ip;
320
321 #ifdef CONFIG_IP_VS_IPV6
322         if (af == AF_INET6)
323                 addr_fold = addr->ip6[0]^addr->ip6[1]^
324                             addr->ip6[2]^addr->ip6[3];
325 #endif
326
327         return (proto^ntohl(addr_fold)^(porth>>IP_VS_SVC_TAB_BITS)^porth)
328                 & IP_VS_SVC_TAB_MASK;
329 }
330
331 /*
332  *      Returns hash value of fwmark for virtual service lookup
333  */
334 static __inline__ unsigned ip_vs_svc_fwm_hashkey(__u32 fwmark)
335 {
336         return fwmark & IP_VS_SVC_TAB_MASK;
337 }
338
339 /*
340  *      Hashes a service in the ip_vs_svc_table by <proto,addr,port>
341  *      or in the ip_vs_svc_fwm_table by fwmark.
342  *      Should be called with locked tables.
343  */
344 static int ip_vs_svc_hash(struct ip_vs_service *svc)
345 {
346         unsigned hash;
347
348         if (svc->flags & IP_VS_SVC_F_HASHED) {
349                 pr_err("%s(): request for already hashed, called from %pF\n",
350                        __func__, __builtin_return_address(0));
351                 return 0;
352         }
353
354         if (svc->fwmark == 0) {
355                 /*
356                  *  Hash it by <protocol,addr,port> in ip_vs_svc_table
357                  */
358                 hash = ip_vs_svc_hashkey(svc->af, svc->protocol, &svc->addr,
359                                          svc->port);
360                 list_add(&svc->s_list, &ip_vs_svc_table[hash]);
361         } else {
362                 /*
363                  *  Hash it by fwmark in ip_vs_svc_fwm_table
364                  */
365                 hash = ip_vs_svc_fwm_hashkey(svc->fwmark);
366                 list_add(&svc->f_list, &ip_vs_svc_fwm_table[hash]);
367         }
368
369         svc->flags |= IP_VS_SVC_F_HASHED;
370         /* increase its refcnt because it is referenced by the svc table */
371         atomic_inc(&svc->refcnt);
372         return 1;
373 }
374
375
376 /*
377  *      Unhashes a service from ip_vs_svc_table/ip_vs_svc_fwm_table.
378  *      Should be called with locked tables.
379  */
380 static int ip_vs_svc_unhash(struct ip_vs_service *svc)
381 {
382         if (!(svc->flags & IP_VS_SVC_F_HASHED)) {
383                 pr_err("%s(): request for unhash flagged, called from %pF\n",
384                        __func__, __builtin_return_address(0));
385                 return 0;
386         }
387
388         if (svc->fwmark == 0) {
389                 /* Remove it from the ip_vs_svc_table table */
390                 list_del(&svc->s_list);
391         } else {
392                 /* Remove it from the ip_vs_svc_fwm_table table */
393                 list_del(&svc->f_list);
394         }
395
396         svc->flags &= ~IP_VS_SVC_F_HASHED;
397         atomic_dec(&svc->refcnt);
398         return 1;
399 }
400
401
402 /*
403  *      Get service by {proto,addr,port} in the service table.
404  */
405 static inline struct ip_vs_service *
406 __ip_vs_service_find(int af, __u16 protocol, const union nf_inet_addr *vaddr,
407                     __be16 vport)
408 {
409         unsigned hash;
410         struct ip_vs_service *svc;
411
412         /* Check for "full" addressed entries */
413         hash = ip_vs_svc_hashkey(af, protocol, vaddr, vport);
414
415         list_for_each_entry(svc, &ip_vs_svc_table[hash], s_list){
416                 if ((svc->af == af)
417                     && ip_vs_addr_equal(af, &svc->addr, vaddr)
418                     && (svc->port == vport)
419                     && (svc->protocol == protocol)) {
420                         /* HIT */
421                         return svc;
422                 }
423         }
424
425         return NULL;
426 }
427
428
429 /*
430  *      Get service by {fwmark} in the service table.
431  */
432 static inline struct ip_vs_service *
433 __ip_vs_svc_fwm_find(int af, __u32 fwmark)
434 {
435         unsigned hash;
436         struct ip_vs_service *svc;
437
438         /* Check for fwmark addressed entries */
439         hash = ip_vs_svc_fwm_hashkey(fwmark);
440
441         list_for_each_entry(svc, &ip_vs_svc_fwm_table[hash], f_list) {
442                 if (svc->fwmark == fwmark && svc->af == af) {
443                         /* HIT */
444                         return svc;
445                 }
446         }
447
448         return NULL;
449 }
450
451 struct ip_vs_service *
452 ip_vs_service_get(int af, __u32 fwmark, __u16 protocol,
453                   const union nf_inet_addr *vaddr, __be16 vport)
454 {
455         struct ip_vs_service *svc;
456
457         read_lock(&__ip_vs_svc_lock);
458
459         /*
460          *      Check the table hashed by fwmark first
461          */
462         if (fwmark && (svc = __ip_vs_svc_fwm_find(af, fwmark)))
463                 goto out;
464
465         /*
466          *      Check the table hashed by <protocol,addr,port>
467          *      for "full" addressed entries
468          */
469         svc = __ip_vs_service_find(af, protocol, vaddr, vport);
470
471         if (svc == NULL
472             && protocol == IPPROTO_TCP
473             && atomic_read(&ip_vs_ftpsvc_counter)
474             && (vport == FTPDATA || ntohs(vport) >= PROT_SOCK)) {
475                 /*
476                  * Check if ftp service entry exists, the packet
477                  * might belong to FTP data connections.
478                  */
479                 svc = __ip_vs_service_find(af, protocol, vaddr, FTPPORT);
480         }
481
482         if (svc == NULL
483             && atomic_read(&ip_vs_nullsvc_counter)) {
484                 /*
485                  * Check if the catch-all port (port zero) exists
486                  */
487                 svc = __ip_vs_service_find(af, protocol, vaddr, 0);
488         }
489
490   out:
491         if (svc)
492                 atomic_inc(&svc->usecnt);
493         read_unlock(&__ip_vs_svc_lock);
494
495         IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
496                       fwmark, ip_vs_proto_name(protocol),
497                       IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
498                       svc ? "hit" : "not hit");
499
500         return svc;
501 }
502
503
504 static inline void
505 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
506 {
507         atomic_inc(&svc->refcnt);
508         dest->svc = svc;
509 }
510
511 static void
512 __ip_vs_unbind_svc(struct ip_vs_dest *dest)
513 {
514         struct ip_vs_service *svc = dest->svc;
515
516         dest->svc = NULL;
517         if (atomic_dec_and_test(&svc->refcnt)) {
518                 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
519                               svc->fwmark,
520                               IP_VS_DBG_ADDR(svc->af, &svc->addr),
521                               ntohs(svc->port), atomic_read(&svc->usecnt));
522                 kfree(svc);
523         }
524 }
525
526
527 /*
528  *      Returns hash value for real service
529  */
530 static inline unsigned ip_vs_rs_hashkey(int af,
531                                             const union nf_inet_addr *addr,
532                                             __be16 port)
533 {
534         register unsigned porth = ntohs(port);
535         __be32 addr_fold = addr->ip;
536
537 #ifdef CONFIG_IP_VS_IPV6
538         if (af == AF_INET6)
539                 addr_fold = addr->ip6[0]^addr->ip6[1]^
540                             addr->ip6[2]^addr->ip6[3];
541 #endif
542
543         return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
544                 & IP_VS_RTAB_MASK;
545 }
546
547 /*
548  *      Hashes ip_vs_dest in ip_vs_rtable by <proto,addr,port>.
549  *      should be called with locked tables.
550  */
551 static int ip_vs_rs_hash(struct ip_vs_dest *dest)
552 {
553         unsigned hash;
554
555         if (!list_empty(&dest->d_list)) {
556                 return 0;
557         }
558
559         /*
560          *      Hash by proto,addr,port,
561          *      which are the parameters of the real service.
562          */
563         hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
564
565         list_add(&dest->d_list, &ip_vs_rtable[hash]);
566
567         return 1;
568 }
569
570 /*
571  *      UNhashes ip_vs_dest from ip_vs_rtable.
572  *      should be called with locked tables.
573  */
574 static int ip_vs_rs_unhash(struct ip_vs_dest *dest)
575 {
576         /*
577          * Remove it from the ip_vs_rtable table.
578          */
579         if (!list_empty(&dest->d_list)) {
580                 list_del(&dest->d_list);
581                 INIT_LIST_HEAD(&dest->d_list);
582         }
583
584         return 1;
585 }
586
587 /*
588  *      Lookup real service by <proto,addr,port> in the real service table.
589  */
590 struct ip_vs_dest *
591 ip_vs_lookup_real_service(int af, __u16 protocol,
592                           const union nf_inet_addr *daddr,
593                           __be16 dport)
594 {
595         unsigned hash;
596         struct ip_vs_dest *dest;
597
598         /*
599          *      Check for "full" addressed entries
600          *      Return the first found entry
601          */
602         hash = ip_vs_rs_hashkey(af, daddr, dport);
603
604         read_lock(&__ip_vs_rs_lock);
605         list_for_each_entry(dest, &ip_vs_rtable[hash], d_list) {
606                 if ((dest->af == af)
607                     && ip_vs_addr_equal(af, &dest->addr, daddr)
608                     && (dest->port == dport)
609                     && ((dest->protocol == protocol) ||
610                         dest->vfwmark)) {
611                         /* HIT */
612                         read_unlock(&__ip_vs_rs_lock);
613                         return dest;
614                 }
615         }
616         read_unlock(&__ip_vs_rs_lock);
617
618         return NULL;
619 }
620
621 /*
622  *      Lookup destination by {addr,port} in the given service
623  */
624 static struct ip_vs_dest *
625 ip_vs_lookup_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
626                   __be16 dport)
627 {
628         struct ip_vs_dest *dest;
629
630         /*
631          * Find the destination for the given service
632          */
633         list_for_each_entry(dest, &svc->destinations, n_list) {
634                 if ((dest->af == svc->af)
635                     && ip_vs_addr_equal(svc->af, &dest->addr, daddr)
636                     && (dest->port == dport)) {
637                         /* HIT */
638                         return dest;
639                 }
640         }
641
642         return NULL;
643 }
644
645 /*
646  * Find destination by {daddr,dport,vaddr,protocol}
647  * Cretaed to be used in ip_vs_process_message() in
648  * the backup synchronization daemon. It finds the
649  * destination to be bound to the received connection
650  * on the backup.
651  *
652  * ip_vs_lookup_real_service() looked promissing, but
653  * seems not working as expected.
654  */
655 struct ip_vs_dest *ip_vs_find_dest(int af, const union nf_inet_addr *daddr,
656                                    __be16 dport,
657                                    const union nf_inet_addr *vaddr,
658                                    __be16 vport, __u16 protocol)
659 {
660         struct ip_vs_dest *dest;
661         struct ip_vs_service *svc;
662
663         svc = ip_vs_service_get(af, 0, protocol, vaddr, vport);
664         if (!svc)
665                 return NULL;
666         dest = ip_vs_lookup_dest(svc, daddr, dport);
667         if (dest)
668                 atomic_inc(&dest->refcnt);
669         ip_vs_service_put(svc);
670         return dest;
671 }
672
673 /*
674  *  Lookup dest by {svc,addr,port} in the destination trash.
675  *  The destination trash is used to hold the destinations that are removed
676  *  from the service table but are still referenced by some conn entries.
677  *  The reason to add the destination trash is when the dest is temporary
678  *  down (either by administrator or by monitor program), the dest can be
679  *  picked back from the trash, the remaining connections to the dest can
680  *  continue, and the counting information of the dest is also useful for
681  *  scheduling.
682  */
683 static struct ip_vs_dest *
684 ip_vs_trash_get_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
685                      __be16 dport)
686 {
687         struct ip_vs_dest *dest, *nxt;
688
689         /*
690          * Find the destination in trash
691          */
692         list_for_each_entry_safe(dest, nxt, &ip_vs_dest_trash, n_list) {
693                 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
694                               "dest->refcnt=%d\n",
695                               dest->vfwmark,
696                               IP_VS_DBG_ADDR(svc->af, &dest->addr),
697                               ntohs(dest->port),
698                               atomic_read(&dest->refcnt));
699                 if (dest->af == svc->af &&
700                     ip_vs_addr_equal(svc->af, &dest->addr, daddr) &&
701                     dest->port == dport &&
702                     dest->vfwmark == svc->fwmark &&
703                     dest->protocol == svc->protocol &&
704                     (svc->fwmark ||
705                      (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
706                       dest->vport == svc->port))) {
707                         /* HIT */
708                         return dest;
709                 }
710
711                 /*
712                  * Try to purge the destination from trash if not referenced
713                  */
714                 if (atomic_read(&dest->refcnt) == 1) {
715                         IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u "
716                                       "from trash\n",
717                                       dest->vfwmark,
718                                       IP_VS_DBG_ADDR(svc->af, &dest->addr),
719                                       ntohs(dest->port));
720                         list_del(&dest->n_list);
721                         ip_vs_dst_reset(dest);
722                         __ip_vs_unbind_svc(dest);
723                         kfree(dest);
724                 }
725         }
726
727         return NULL;
728 }
729
730
731 /*
732  *  Clean up all the destinations in the trash
733  *  Called by the ip_vs_control_cleanup()
734  *
735  *  When the ip_vs_control_clearup is activated by ipvs module exit,
736  *  the service tables must have been flushed and all the connections
737  *  are expired, and the refcnt of each destination in the trash must
738  *  be 1, so we simply release them here.
739  */
740 static void ip_vs_trash_cleanup(void)
741 {
742         struct ip_vs_dest *dest, *nxt;
743
744         list_for_each_entry_safe(dest, nxt, &ip_vs_dest_trash, n_list) {
745                 list_del(&dest->n_list);
746                 ip_vs_dst_reset(dest);
747                 __ip_vs_unbind_svc(dest);
748                 kfree(dest);
749         }
750 }
751
752
753 static void
754 ip_vs_zero_stats(struct ip_vs_stats *stats)
755 {
756         spin_lock_bh(&stats->lock);
757
758         memset(&stats->ustats, 0, sizeof(stats->ustats));
759         ip_vs_zero_estimator(stats);
760
761         spin_unlock_bh(&stats->lock);
762 }
763
764 /*
765  *      Update a destination in the given service
766  */
767 static void
768 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
769                     struct ip_vs_dest_user_kern *udest, int add)
770 {
771         int conn_flags;
772
773         /* set the weight and the flags */
774         atomic_set(&dest->weight, udest->weight);
775         conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
776         conn_flags |= IP_VS_CONN_F_INACTIVE;
777
778         /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
779         if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
780                 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
781         } else {
782                 /*
783                  *    Put the real service in ip_vs_rtable if not present.
784                  *    For now only for NAT!
785                  */
786                 write_lock_bh(&__ip_vs_rs_lock);
787                 ip_vs_rs_hash(dest);
788                 write_unlock_bh(&__ip_vs_rs_lock);
789         }
790         atomic_set(&dest->conn_flags, conn_flags);
791
792         /* bind the service */
793         if (!dest->svc) {
794                 __ip_vs_bind_svc(dest, svc);
795         } else {
796                 if (dest->svc != svc) {
797                         __ip_vs_unbind_svc(dest);
798                         ip_vs_zero_stats(&dest->stats);
799                         __ip_vs_bind_svc(dest, svc);
800                 }
801         }
802
803         /* set the dest status flags */
804         dest->flags |= IP_VS_DEST_F_AVAILABLE;
805
806         if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
807                 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
808         dest->u_threshold = udest->u_threshold;
809         dest->l_threshold = udest->l_threshold;
810
811         spin_lock(&dest->dst_lock);
812         ip_vs_dst_reset(dest);
813         spin_unlock(&dest->dst_lock);
814
815         if (add)
816                 ip_vs_new_estimator(&dest->stats);
817
818         write_lock_bh(&__ip_vs_svc_lock);
819
820         /* Wait until all other svc users go away */
821         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
822
823         if (add) {
824                 list_add(&dest->n_list, &svc->destinations);
825                 svc->num_dests++;
826         }
827
828         /* call the update_service, because server weight may be changed */
829         if (svc->scheduler->update_service)
830                 svc->scheduler->update_service(svc);
831
832         write_unlock_bh(&__ip_vs_svc_lock);
833 }
834
835
836 /*
837  *      Create a destination for the given service
838  */
839 static int
840 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
841                struct ip_vs_dest **dest_p)
842 {
843         struct ip_vs_dest *dest;
844         unsigned atype;
845
846         EnterFunction(2);
847
848 #ifdef CONFIG_IP_VS_IPV6
849         if (svc->af == AF_INET6) {
850                 atype = ipv6_addr_type(&udest->addr.in6);
851                 if ((!(atype & IPV6_ADDR_UNICAST) ||
852                         atype & IPV6_ADDR_LINKLOCAL) &&
853                         !__ip_vs_addr_is_local_v6(&udest->addr.in6))
854                         return -EINVAL;
855         } else
856 #endif
857         {
858                 atype = inet_addr_type(&init_net, udest->addr.ip);
859                 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
860                         return -EINVAL;
861         }
862
863         dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
864         if (dest == NULL) {
865                 pr_err("%s(): no memory.\n", __func__);
866                 return -ENOMEM;
867         }
868
869         dest->af = svc->af;
870         dest->protocol = svc->protocol;
871         dest->vaddr = svc->addr;
872         dest->vport = svc->port;
873         dest->vfwmark = svc->fwmark;
874         ip_vs_addr_copy(svc->af, &dest->addr, &udest->addr);
875         dest->port = udest->port;
876
877         atomic_set(&dest->activeconns, 0);
878         atomic_set(&dest->inactconns, 0);
879         atomic_set(&dest->persistconns, 0);
880         atomic_set(&dest->refcnt, 1);
881
882         INIT_LIST_HEAD(&dest->d_list);
883         spin_lock_init(&dest->dst_lock);
884         spin_lock_init(&dest->stats.lock);
885         __ip_vs_update_dest(svc, dest, udest, 1);
886
887         *dest_p = dest;
888
889         LeaveFunction(2);
890         return 0;
891 }
892
893
894 /*
895  *      Add a destination into an existing service
896  */
897 static int
898 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
899 {
900         struct ip_vs_dest *dest;
901         union nf_inet_addr daddr;
902         __be16 dport = udest->port;
903         int ret;
904
905         EnterFunction(2);
906
907         if (udest->weight < 0) {
908                 pr_err("%s(): server weight less than zero\n", __func__);
909                 return -ERANGE;
910         }
911
912         if (udest->l_threshold > udest->u_threshold) {
913                 pr_err("%s(): lower threshold is higher than upper threshold\n",
914                         __func__);
915                 return -ERANGE;
916         }
917
918         ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
919
920         /*
921          * Check if the dest already exists in the list
922          */
923         dest = ip_vs_lookup_dest(svc, &daddr, dport);
924
925         if (dest != NULL) {
926                 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
927                 return -EEXIST;
928         }
929
930         /*
931          * Check if the dest already exists in the trash and
932          * is from the same service
933          */
934         dest = ip_vs_trash_get_dest(svc, &daddr, dport);
935
936         if (dest != NULL) {
937                 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
938                               "dest->refcnt=%d, service %u/%s:%u\n",
939                               IP_VS_DBG_ADDR(svc->af, &daddr), ntohs(dport),
940                               atomic_read(&dest->refcnt),
941                               dest->vfwmark,
942                               IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
943                               ntohs(dest->vport));
944
945                 /*
946                  * Get the destination from the trash
947                  */
948                 list_del(&dest->n_list);
949
950                 __ip_vs_update_dest(svc, dest, udest, 1);
951                 ret = 0;
952         } else {
953                 /*
954                  * Allocate and initialize the dest structure
955                  */
956                 ret = ip_vs_new_dest(svc, udest, &dest);
957         }
958         LeaveFunction(2);
959
960         return ret;
961 }
962
963
964 /*
965  *      Edit a destination in the given service
966  */
967 static int
968 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
969 {
970         struct ip_vs_dest *dest;
971         union nf_inet_addr daddr;
972         __be16 dport = udest->port;
973
974         EnterFunction(2);
975
976         if (udest->weight < 0) {
977                 pr_err("%s(): server weight less than zero\n", __func__);
978                 return -ERANGE;
979         }
980
981         if (udest->l_threshold > udest->u_threshold) {
982                 pr_err("%s(): lower threshold is higher than upper threshold\n",
983                         __func__);
984                 return -ERANGE;
985         }
986
987         ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
988
989         /*
990          *  Lookup the destination list
991          */
992         dest = ip_vs_lookup_dest(svc, &daddr, dport);
993
994         if (dest == NULL) {
995                 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
996                 return -ENOENT;
997         }
998
999         __ip_vs_update_dest(svc, dest, udest, 0);
1000         LeaveFunction(2);
1001
1002         return 0;
1003 }
1004
1005
1006 /*
1007  *      Delete a destination (must be already unlinked from the service)
1008  */
1009 static void __ip_vs_del_dest(struct ip_vs_dest *dest)
1010 {
1011         ip_vs_kill_estimator(&dest->stats);
1012
1013         /*
1014          *  Remove it from the d-linked list with the real services.
1015          */
1016         write_lock_bh(&__ip_vs_rs_lock);
1017         ip_vs_rs_unhash(dest);
1018         write_unlock_bh(&__ip_vs_rs_lock);
1019
1020         /*
1021          *  Decrease the refcnt of the dest, and free the dest
1022          *  if nobody refers to it (refcnt=0). Otherwise, throw
1023          *  the destination into the trash.
1024          */
1025         if (atomic_dec_and_test(&dest->refcnt)) {
1026                 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u\n",
1027                               dest->vfwmark,
1028                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1029                               ntohs(dest->port));
1030                 ip_vs_dst_reset(dest);
1031                 /* simply decrease svc->refcnt here, let the caller check
1032                    and release the service if nobody refers to it.
1033                    Only user context can release destination and service,
1034                    and only one user context can update virtual service at a
1035                    time, so the operation here is OK */
1036                 atomic_dec(&dest->svc->refcnt);
1037                 kfree(dest);
1038         } else {
1039                 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, "
1040                               "dest->refcnt=%d\n",
1041                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1042                               ntohs(dest->port),
1043                               atomic_read(&dest->refcnt));
1044                 list_add(&dest->n_list, &ip_vs_dest_trash);
1045                 atomic_inc(&dest->refcnt);
1046         }
1047 }
1048
1049
1050 /*
1051  *      Unlink a destination from the given service
1052  */
1053 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1054                                 struct ip_vs_dest *dest,
1055                                 int svcupd)
1056 {
1057         dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1058
1059         /*
1060          *  Remove it from the d-linked destination list.
1061          */
1062         list_del(&dest->n_list);
1063         svc->num_dests--;
1064
1065         /*
1066          *  Call the update_service function of its scheduler
1067          */
1068         if (svcupd && svc->scheduler->update_service)
1069                         svc->scheduler->update_service(svc);
1070 }
1071
1072
1073 /*
1074  *      Delete a destination server in the given service
1075  */
1076 static int
1077 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1078 {
1079         struct ip_vs_dest *dest;
1080         __be16 dport = udest->port;
1081
1082         EnterFunction(2);
1083
1084         dest = ip_vs_lookup_dest(svc, &udest->addr, dport);
1085
1086         if (dest == NULL) {
1087                 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1088                 return -ENOENT;
1089         }
1090
1091         write_lock_bh(&__ip_vs_svc_lock);
1092
1093         /*
1094          *      Wait until all other svc users go away.
1095          */
1096         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1097
1098         /*
1099          *      Unlink dest from the service
1100          */
1101         __ip_vs_unlink_dest(svc, dest, 1);
1102
1103         write_unlock_bh(&__ip_vs_svc_lock);
1104
1105         /*
1106          *      Delete the destination
1107          */
1108         __ip_vs_del_dest(dest);
1109
1110         LeaveFunction(2);
1111
1112         return 0;
1113 }
1114
1115
1116 /*
1117  *      Add a service into the service hash table
1118  */
1119 static int
1120 ip_vs_add_service(struct ip_vs_service_user_kern *u,
1121                   struct ip_vs_service **svc_p)
1122 {
1123         int ret = 0;
1124         struct ip_vs_scheduler *sched = NULL;
1125         struct ip_vs_pe *pe = NULL;
1126         struct ip_vs_service *svc = NULL;
1127
1128         /* increase the module use count */
1129         ip_vs_use_count_inc();
1130
1131         /* Lookup the scheduler by 'u->sched_name' */
1132         sched = ip_vs_scheduler_get(u->sched_name);
1133         if (sched == NULL) {
1134                 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1135                 ret = -ENOENT;
1136                 goto out_err;
1137         }
1138
1139         if (u->pe_name && *u->pe_name) {
1140                 pe = ip_vs_pe_get(u->pe_name);
1141                 if (pe == NULL) {
1142                         pr_info("persistence engine module ip_vs_pe_%s "
1143                                 "not found\n", u->pe_name);
1144                         ret = -ENOENT;
1145                         goto out_err;
1146                 }
1147         }
1148
1149 #ifdef CONFIG_IP_VS_IPV6
1150         if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1151                 ret = -EINVAL;
1152                 goto out_err;
1153         }
1154 #endif
1155
1156         svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1157         if (svc == NULL) {
1158                 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1159                 ret = -ENOMEM;
1160                 goto out_err;
1161         }
1162
1163         /* I'm the first user of the service */
1164         atomic_set(&svc->usecnt, 0);
1165         atomic_set(&svc->refcnt, 0);
1166
1167         svc->af = u->af;
1168         svc->protocol = u->protocol;
1169         ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1170         svc->port = u->port;
1171         svc->fwmark = u->fwmark;
1172         svc->flags = u->flags;
1173         svc->timeout = u->timeout * HZ;
1174         svc->netmask = u->netmask;
1175
1176         INIT_LIST_HEAD(&svc->destinations);
1177         rwlock_init(&svc->sched_lock);
1178         spin_lock_init(&svc->stats.lock);
1179
1180         /* Bind the scheduler */
1181         ret = ip_vs_bind_scheduler(svc, sched);
1182         if (ret)
1183                 goto out_err;
1184         sched = NULL;
1185
1186         /* Bind the ct retriever */
1187         ip_vs_bind_pe(svc, pe);
1188         pe = NULL;
1189
1190         /* Update the virtual service counters */
1191         if (svc->port == FTPPORT)
1192                 atomic_inc(&ip_vs_ftpsvc_counter);
1193         else if (svc->port == 0)
1194                 atomic_inc(&ip_vs_nullsvc_counter);
1195
1196         ip_vs_new_estimator(&svc->stats);
1197
1198         /* Count only IPv4 services for old get/setsockopt interface */
1199         if (svc->af == AF_INET)
1200                 ip_vs_num_services++;
1201
1202         /* Hash the service into the service table */
1203         write_lock_bh(&__ip_vs_svc_lock);
1204         ip_vs_svc_hash(svc);
1205         write_unlock_bh(&__ip_vs_svc_lock);
1206
1207         *svc_p = svc;
1208         return 0;
1209
1210  out_err:
1211         if (svc != NULL) {
1212                 ip_vs_unbind_scheduler(svc);
1213                 if (svc->inc) {
1214                         local_bh_disable();
1215                         ip_vs_app_inc_put(svc->inc);
1216                         local_bh_enable();
1217                 }
1218                 kfree(svc);
1219         }
1220         ip_vs_scheduler_put(sched);
1221         ip_vs_pe_put(pe);
1222
1223         /* decrease the module use count */
1224         ip_vs_use_count_dec();
1225
1226         return ret;
1227 }
1228
1229
1230 /*
1231  *      Edit a service and bind it with a new scheduler
1232  */
1233 static int
1234 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1235 {
1236         struct ip_vs_scheduler *sched, *old_sched;
1237         struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1238         int ret = 0;
1239
1240         /*
1241          * Lookup the scheduler, by 'u->sched_name'
1242          */
1243         sched = ip_vs_scheduler_get(u->sched_name);
1244         if (sched == NULL) {
1245                 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1246                 return -ENOENT;
1247         }
1248         old_sched = sched;
1249
1250         if (u->pe_name && *u->pe_name) {
1251                 pe = ip_vs_pe_get(u->pe_name);
1252                 if (pe == NULL) {
1253                         pr_info("persistence engine module ip_vs_pe_%s "
1254                                 "not found\n", u->pe_name);
1255                         ret = -ENOENT;
1256                         goto out;
1257                 }
1258                 old_pe = pe;
1259         }
1260
1261 #ifdef CONFIG_IP_VS_IPV6
1262         if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1263                 ret = -EINVAL;
1264                 goto out;
1265         }
1266 #endif
1267
1268         write_lock_bh(&__ip_vs_svc_lock);
1269
1270         /*
1271          * Wait until all other svc users go away.
1272          */
1273         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1274
1275         /*
1276          * Set the flags and timeout value
1277          */
1278         svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1279         svc->timeout = u->timeout * HZ;
1280         svc->netmask = u->netmask;
1281
1282         old_sched = svc->scheduler;
1283         if (sched != old_sched) {
1284                 /*
1285                  * Unbind the old scheduler
1286                  */
1287                 if ((ret = ip_vs_unbind_scheduler(svc))) {
1288                         old_sched = sched;
1289                         goto out_unlock;
1290                 }
1291
1292                 /*
1293                  * Bind the new scheduler
1294                  */
1295                 if ((ret = ip_vs_bind_scheduler(svc, sched))) {
1296                         /*
1297                          * If ip_vs_bind_scheduler fails, restore the old
1298                          * scheduler.
1299                          * The main reason of failure is out of memory.
1300                          *
1301                          * The question is if the old scheduler can be
1302                          * restored all the time. TODO: if it cannot be
1303                          * restored some time, we must delete the service,
1304                          * otherwise the system may crash.
1305                          */
1306                         ip_vs_bind_scheduler(svc, old_sched);
1307                         old_sched = sched;
1308                         goto out_unlock;
1309                 }
1310         }
1311
1312         old_pe = svc->pe;
1313         if (pe != old_pe) {
1314                 ip_vs_unbind_pe(svc);
1315                 ip_vs_bind_pe(svc, pe);
1316         }
1317
1318   out_unlock:
1319         write_unlock_bh(&__ip_vs_svc_lock);
1320   out:
1321         ip_vs_scheduler_put(old_sched);
1322         ip_vs_pe_put(old_pe);
1323         return ret;
1324 }
1325
1326
1327 /*
1328  *      Delete a service from the service list
1329  *      - The service must be unlinked, unlocked and not referenced!
1330  *      - We are called under _bh lock
1331  */
1332 static void __ip_vs_del_service(struct ip_vs_service *svc)
1333 {
1334         struct ip_vs_dest *dest, *nxt;
1335         struct ip_vs_scheduler *old_sched;
1336         struct ip_vs_pe *old_pe;
1337
1338         pr_info("%s: enter\n", __func__);
1339
1340         /* Count only IPv4 services for old get/setsockopt interface */
1341         if (svc->af == AF_INET)
1342                 ip_vs_num_services--;
1343
1344         ip_vs_kill_estimator(&svc->stats);
1345
1346         /* Unbind scheduler */
1347         old_sched = svc->scheduler;
1348         ip_vs_unbind_scheduler(svc);
1349         ip_vs_scheduler_put(old_sched);
1350
1351         /* Unbind persistence engine */
1352         old_pe = svc->pe;
1353         ip_vs_unbind_pe(svc);
1354         ip_vs_pe_put(old_pe);
1355
1356         /* Unbind app inc */
1357         if (svc->inc) {
1358                 ip_vs_app_inc_put(svc->inc);
1359                 svc->inc = NULL;
1360         }
1361
1362         /*
1363          *    Unlink the whole destination list
1364          */
1365         list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1366                 __ip_vs_unlink_dest(svc, dest, 0);
1367                 __ip_vs_del_dest(dest);
1368         }
1369
1370         /*
1371          *    Update the virtual service counters
1372          */
1373         if (svc->port == FTPPORT)
1374                 atomic_dec(&ip_vs_ftpsvc_counter);
1375         else if (svc->port == 0)
1376                 atomic_dec(&ip_vs_nullsvc_counter);
1377
1378         /*
1379          *    Free the service if nobody refers to it
1380          */
1381         if (atomic_read(&svc->refcnt) == 0) {
1382                 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
1383                               svc->fwmark,
1384                               IP_VS_DBG_ADDR(svc->af, &svc->addr),
1385                               ntohs(svc->port), atomic_read(&svc->usecnt));
1386                 kfree(svc);
1387         }
1388
1389         /* decrease the module use count */
1390         ip_vs_use_count_dec();
1391 }
1392
1393 /*
1394  * Unlink a service from list and try to delete it if its refcnt reached 0
1395  */
1396 static void ip_vs_unlink_service(struct ip_vs_service *svc)
1397 {
1398         /*
1399          * Unhash it from the service table
1400          */
1401         write_lock_bh(&__ip_vs_svc_lock);
1402
1403         ip_vs_svc_unhash(svc);
1404
1405         /*
1406          * Wait until all the svc users go away.
1407          */
1408         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1409
1410         __ip_vs_del_service(svc);
1411
1412         write_unlock_bh(&__ip_vs_svc_lock);
1413 }
1414
1415 /*
1416  *      Delete a service from the service list
1417  */
1418 static int ip_vs_del_service(struct ip_vs_service *svc)
1419 {
1420         if (svc == NULL)
1421                 return -EEXIST;
1422         ip_vs_unlink_service(svc);
1423
1424         return 0;
1425 }
1426
1427
1428 /*
1429  *      Flush all the virtual services
1430  */
1431 static int ip_vs_flush(void)
1432 {
1433         int idx;
1434         struct ip_vs_service *svc, *nxt;
1435
1436         /*
1437          * Flush the service table hashed by <protocol,addr,port>
1438          */
1439         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1440                 list_for_each_entry_safe(svc, nxt, &ip_vs_svc_table[idx], s_list) {
1441                         ip_vs_unlink_service(svc);
1442                 }
1443         }
1444
1445         /*
1446          * Flush the service table hashed by fwmark
1447          */
1448         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1449                 list_for_each_entry_safe(svc, nxt,
1450                                          &ip_vs_svc_fwm_table[idx], f_list) {
1451                         ip_vs_unlink_service(svc);
1452                 }
1453         }
1454
1455         return 0;
1456 }
1457
1458
1459 /*
1460  *      Zero counters in a service or all services
1461  */
1462 static int ip_vs_zero_service(struct ip_vs_service *svc)
1463 {
1464         struct ip_vs_dest *dest;
1465
1466         write_lock_bh(&__ip_vs_svc_lock);
1467         list_for_each_entry(dest, &svc->destinations, n_list) {
1468                 ip_vs_zero_stats(&dest->stats);
1469         }
1470         ip_vs_zero_stats(&svc->stats);
1471         write_unlock_bh(&__ip_vs_svc_lock);
1472         return 0;
1473 }
1474
1475 static int ip_vs_zero_all(void)
1476 {
1477         int idx;
1478         struct ip_vs_service *svc;
1479
1480         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1481                 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1482                         ip_vs_zero_service(svc);
1483                 }
1484         }
1485
1486         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1487                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1488                         ip_vs_zero_service(svc);
1489                 }
1490         }
1491
1492         ip_vs_zero_stats(&ip_vs_stats);
1493         return 0;
1494 }
1495
1496
1497 static int
1498 proc_do_defense_mode(ctl_table *table, int write,
1499                      void __user *buffer, size_t *lenp, loff_t *ppos)
1500 {
1501         int *valp = table->data;
1502         int val = *valp;
1503         int rc;
1504
1505         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1506         if (write && (*valp != val)) {
1507                 if ((*valp < 0) || (*valp > 3)) {
1508                         /* Restore the correct value */
1509                         *valp = val;
1510                 } else {
1511                         update_defense_level();
1512                 }
1513         }
1514         return rc;
1515 }
1516
1517
1518 static int
1519 proc_do_sync_threshold(ctl_table *table, int write,
1520                        void __user *buffer, size_t *lenp, loff_t *ppos)
1521 {
1522         int *valp = table->data;
1523         int val[2];
1524         int rc;
1525
1526         /* backup the value first */
1527         memcpy(val, valp, sizeof(val));
1528
1529         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1530         if (write && (valp[0] < 0 || valp[1] < 0 || valp[0] >= valp[1])) {
1531                 /* Restore the correct value */
1532                 memcpy(valp, val, sizeof(val));
1533         }
1534         return rc;
1535 }
1536
1537
1538 /*
1539  *      IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1540  */
1541
1542 static struct ctl_table vs_vars[] = {
1543         {
1544                 .procname       = "amemthresh",
1545                 .data           = &sysctl_ip_vs_amemthresh,
1546                 .maxlen         = sizeof(int),
1547                 .mode           = 0644,
1548                 .proc_handler   = proc_dointvec,
1549         },
1550 #ifdef CONFIG_IP_VS_DEBUG
1551         {
1552                 .procname       = "debug_level",
1553                 .data           = &sysctl_ip_vs_debug_level,
1554                 .maxlen         = sizeof(int),
1555                 .mode           = 0644,
1556                 .proc_handler   = proc_dointvec,
1557         },
1558 #endif
1559         {
1560                 .procname       = "am_droprate",
1561                 .data           = &sysctl_ip_vs_am_droprate,
1562                 .maxlen         = sizeof(int),
1563                 .mode           = 0644,
1564                 .proc_handler   = proc_dointvec,
1565         },
1566         {
1567                 .procname       = "drop_entry",
1568                 .data           = &sysctl_ip_vs_drop_entry,
1569                 .maxlen         = sizeof(int),
1570                 .mode           = 0644,
1571                 .proc_handler   = proc_do_defense_mode,
1572         },
1573         {
1574                 .procname       = "drop_packet",
1575                 .data           = &sysctl_ip_vs_drop_packet,
1576                 .maxlen         = sizeof(int),
1577                 .mode           = 0644,
1578                 .proc_handler   = proc_do_defense_mode,
1579         },
1580 #ifdef CONFIG_IP_VS_NFCT
1581         {
1582                 .procname       = "conntrack",
1583                 .data           = &sysctl_ip_vs_conntrack,
1584                 .maxlen         = sizeof(int),
1585                 .mode           = 0644,
1586                 .proc_handler   = &proc_dointvec,
1587         },
1588 #endif
1589         {
1590                 .procname       = "secure_tcp",
1591                 .data           = &sysctl_ip_vs_secure_tcp,
1592                 .maxlen         = sizeof(int),
1593                 .mode           = 0644,
1594                 .proc_handler   = proc_do_defense_mode,
1595         },
1596         {
1597                 .procname       = "snat_reroute",
1598                 .data           = &sysctl_ip_vs_snat_reroute,
1599                 .maxlen         = sizeof(int),
1600                 .mode           = 0644,
1601                 .proc_handler   = &proc_dointvec,
1602         },
1603 #if 0
1604         {
1605                 .procname       = "timeout_established",
1606                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1607                 .maxlen         = sizeof(int),
1608                 .mode           = 0644,
1609                 .proc_handler   = proc_dointvec_jiffies,
1610         },
1611         {
1612                 .procname       = "timeout_synsent",
1613                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1614                 .maxlen         = sizeof(int),
1615                 .mode           = 0644,
1616                 .proc_handler   = proc_dointvec_jiffies,
1617         },
1618         {
1619                 .procname       = "timeout_synrecv",
1620                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1621                 .maxlen         = sizeof(int),
1622                 .mode           = 0644,
1623                 .proc_handler   = proc_dointvec_jiffies,
1624         },
1625         {
1626                 .procname       = "timeout_finwait",
1627                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1628                 .maxlen         = sizeof(int),
1629                 .mode           = 0644,
1630                 .proc_handler   = proc_dointvec_jiffies,
1631         },
1632         {
1633                 .procname       = "timeout_timewait",
1634                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1635                 .maxlen         = sizeof(int),
1636                 .mode           = 0644,
1637                 .proc_handler   = proc_dointvec_jiffies,
1638         },
1639         {
1640                 .procname       = "timeout_close",
1641                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1642                 .maxlen         = sizeof(int),
1643                 .mode           = 0644,
1644                 .proc_handler   = proc_dointvec_jiffies,
1645         },
1646         {
1647                 .procname       = "timeout_closewait",
1648                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1649                 .maxlen         = sizeof(int),
1650                 .mode           = 0644,
1651                 .proc_handler   = proc_dointvec_jiffies,
1652         },
1653         {
1654                 .procname       = "timeout_lastack",
1655                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1656                 .maxlen         = sizeof(int),
1657                 .mode           = 0644,
1658                 .proc_handler   = proc_dointvec_jiffies,
1659         },
1660         {
1661                 .procname       = "timeout_listen",
1662                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1663                 .maxlen         = sizeof(int),
1664                 .mode           = 0644,
1665                 .proc_handler   = proc_dointvec_jiffies,
1666         },
1667         {
1668                 .procname       = "timeout_synack",
1669                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1670                 .maxlen         = sizeof(int),
1671                 .mode           = 0644,
1672                 .proc_handler   = proc_dointvec_jiffies,
1673         },
1674         {
1675                 .procname       = "timeout_udp",
1676                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1677                 .maxlen         = sizeof(int),
1678                 .mode           = 0644,
1679                 .proc_handler   = proc_dointvec_jiffies,
1680         },
1681         {
1682                 .procname       = "timeout_icmp",
1683                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1684                 .maxlen         = sizeof(int),
1685                 .mode           = 0644,
1686                 .proc_handler   = proc_dointvec_jiffies,
1687         },
1688 #endif
1689         {
1690                 .procname       = "cache_bypass",
1691                 .data           = &sysctl_ip_vs_cache_bypass,
1692                 .maxlen         = sizeof(int),
1693                 .mode           = 0644,
1694                 .proc_handler   = proc_dointvec,
1695         },
1696         {
1697                 .procname       = "expire_nodest_conn",
1698                 .data           = &sysctl_ip_vs_expire_nodest_conn,
1699                 .maxlen         = sizeof(int),
1700                 .mode           = 0644,
1701                 .proc_handler   = proc_dointvec,
1702         },
1703         {
1704                 .procname       = "expire_quiescent_template",
1705                 .data           = &sysctl_ip_vs_expire_quiescent_template,
1706                 .maxlen         = sizeof(int),
1707                 .mode           = 0644,
1708                 .proc_handler   = proc_dointvec,
1709         },
1710         {
1711                 .procname       = "sync_threshold",
1712                 .data           = &sysctl_ip_vs_sync_threshold,
1713                 .maxlen         = sizeof(sysctl_ip_vs_sync_threshold),
1714                 .mode           = 0644,
1715                 .proc_handler   = proc_do_sync_threshold,
1716         },
1717         {
1718                 .procname       = "nat_icmp_send",
1719                 .data           = &sysctl_ip_vs_nat_icmp_send,
1720                 .maxlen         = sizeof(int),
1721                 .mode           = 0644,
1722                 .proc_handler   = proc_dointvec,
1723         },
1724         { }
1725 };
1726
1727 const struct ctl_path net_vs_ctl_path[] = {
1728         { .procname = "net", },
1729         { .procname = "ipv4", },
1730         { .procname = "vs", },
1731         { }
1732 };
1733 EXPORT_SYMBOL_GPL(net_vs_ctl_path);
1734
1735 static struct ctl_table_header * sysctl_header;
1736
1737 #ifdef CONFIG_PROC_FS
1738
1739 struct ip_vs_iter {
1740         struct list_head *table;
1741         int bucket;
1742 };
1743
1744 /*
1745  *      Write the contents of the VS rule table to a PROCfs file.
1746  *      (It is kept just for backward compatibility)
1747  */
1748 static inline const char *ip_vs_fwd_name(unsigned flags)
1749 {
1750         switch (flags & IP_VS_CONN_F_FWD_MASK) {
1751         case IP_VS_CONN_F_LOCALNODE:
1752                 return "Local";
1753         case IP_VS_CONN_F_TUNNEL:
1754                 return "Tunnel";
1755         case IP_VS_CONN_F_DROUTE:
1756                 return "Route";
1757         default:
1758                 return "Masq";
1759         }
1760 }
1761
1762
1763 /* Get the Nth entry in the two lists */
1764 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1765 {
1766         struct ip_vs_iter *iter = seq->private;
1767         int idx;
1768         struct ip_vs_service *svc;
1769
1770         /* look in hash by protocol */
1771         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1772                 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1773                         if (pos-- == 0){
1774                                 iter->table = ip_vs_svc_table;
1775                                 iter->bucket = idx;
1776                                 return svc;
1777                         }
1778                 }
1779         }
1780
1781         /* keep looking in fwmark */
1782         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1783                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1784                         if (pos-- == 0) {
1785                                 iter->table = ip_vs_svc_fwm_table;
1786                                 iter->bucket = idx;
1787                                 return svc;
1788                         }
1789                 }
1790         }
1791
1792         return NULL;
1793 }
1794
1795 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1796 __acquires(__ip_vs_svc_lock)
1797 {
1798
1799         read_lock_bh(&__ip_vs_svc_lock);
1800         return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1801 }
1802
1803
1804 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1805 {
1806         struct list_head *e;
1807         struct ip_vs_iter *iter;
1808         struct ip_vs_service *svc;
1809
1810         ++*pos;
1811         if (v == SEQ_START_TOKEN)
1812                 return ip_vs_info_array(seq,0);
1813
1814         svc = v;
1815         iter = seq->private;
1816
1817         if (iter->table == ip_vs_svc_table) {
1818                 /* next service in table hashed by protocol */
1819                 if ((e = svc->s_list.next) != &ip_vs_svc_table[iter->bucket])
1820                         return list_entry(e, struct ip_vs_service, s_list);
1821
1822
1823                 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1824                         list_for_each_entry(svc,&ip_vs_svc_table[iter->bucket],
1825                                             s_list) {
1826                                 return svc;
1827                         }
1828                 }
1829
1830                 iter->table = ip_vs_svc_fwm_table;
1831                 iter->bucket = -1;
1832                 goto scan_fwmark;
1833         }
1834
1835         /* next service in hashed by fwmark */
1836         if ((e = svc->f_list.next) != &ip_vs_svc_fwm_table[iter->bucket])
1837                 return list_entry(e, struct ip_vs_service, f_list);
1838
1839  scan_fwmark:
1840         while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1841                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[iter->bucket],
1842                                     f_list)
1843                         return svc;
1844         }
1845
1846         return NULL;
1847 }
1848
1849 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
1850 __releases(__ip_vs_svc_lock)
1851 {
1852         read_unlock_bh(&__ip_vs_svc_lock);
1853 }
1854
1855
1856 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
1857 {
1858         if (v == SEQ_START_TOKEN) {
1859                 seq_printf(seq,
1860                         "IP Virtual Server version %d.%d.%d (size=%d)\n",
1861                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
1862                 seq_puts(seq,
1863                          "Prot LocalAddress:Port Scheduler Flags\n");
1864                 seq_puts(seq,
1865                          "  -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
1866         } else {
1867                 const struct ip_vs_service *svc = v;
1868                 const struct ip_vs_iter *iter = seq->private;
1869                 const struct ip_vs_dest *dest;
1870
1871                 if (iter->table == ip_vs_svc_table) {
1872 #ifdef CONFIG_IP_VS_IPV6
1873                         if (svc->af == AF_INET6)
1874                                 seq_printf(seq, "%s  [%pI6]:%04X %s ",
1875                                            ip_vs_proto_name(svc->protocol),
1876                                            &svc->addr.in6,
1877                                            ntohs(svc->port),
1878                                            svc->scheduler->name);
1879                         else
1880 #endif
1881                                 seq_printf(seq, "%s  %08X:%04X %s %s ",
1882                                            ip_vs_proto_name(svc->protocol),
1883                                            ntohl(svc->addr.ip),
1884                                            ntohs(svc->port),
1885                                            svc->scheduler->name,
1886                                            (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
1887                 } else {
1888                         seq_printf(seq, "FWM  %08X %s %s",
1889                                    svc->fwmark, svc->scheduler->name,
1890                                    (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
1891                 }
1892
1893                 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
1894                         seq_printf(seq, "persistent %d %08X\n",
1895                                 svc->timeout,
1896                                 ntohl(svc->netmask));
1897                 else
1898                         seq_putc(seq, '\n');
1899
1900                 list_for_each_entry(dest, &svc->destinations, n_list) {
1901 #ifdef CONFIG_IP_VS_IPV6
1902                         if (dest->af == AF_INET6)
1903                                 seq_printf(seq,
1904                                            "  -> [%pI6]:%04X"
1905                                            "      %-7s %-6d %-10d %-10d\n",
1906                                            &dest->addr.in6,
1907                                            ntohs(dest->port),
1908                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
1909                                            atomic_read(&dest->weight),
1910                                            atomic_read(&dest->activeconns),
1911                                            atomic_read(&dest->inactconns));
1912                         else
1913 #endif
1914                                 seq_printf(seq,
1915                                            "  -> %08X:%04X      "
1916                                            "%-7s %-6d %-10d %-10d\n",
1917                                            ntohl(dest->addr.ip),
1918                                            ntohs(dest->port),
1919                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
1920                                            atomic_read(&dest->weight),
1921                                            atomic_read(&dest->activeconns),
1922                                            atomic_read(&dest->inactconns));
1923
1924                 }
1925         }
1926         return 0;
1927 }
1928
1929 static const struct seq_operations ip_vs_info_seq_ops = {
1930         .start = ip_vs_info_seq_start,
1931         .next  = ip_vs_info_seq_next,
1932         .stop  = ip_vs_info_seq_stop,
1933         .show  = ip_vs_info_seq_show,
1934 };
1935
1936 static int ip_vs_info_open(struct inode *inode, struct file *file)
1937 {
1938         return seq_open_private(file, &ip_vs_info_seq_ops,
1939                         sizeof(struct ip_vs_iter));
1940 }
1941
1942 static const struct file_operations ip_vs_info_fops = {
1943         .owner   = THIS_MODULE,
1944         .open    = ip_vs_info_open,
1945         .read    = seq_read,
1946         .llseek  = seq_lseek,
1947         .release = seq_release_private,
1948 };
1949
1950 #endif
1951
1952 struct ip_vs_stats ip_vs_stats = {
1953         .lock = __SPIN_LOCK_UNLOCKED(ip_vs_stats.lock),
1954 };
1955
1956 #ifdef CONFIG_PROC_FS
1957 static int ip_vs_stats_show(struct seq_file *seq, void *v)
1958 {
1959
1960 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
1961         seq_puts(seq,
1962                  "   Total Incoming Outgoing         Incoming         Outgoing\n");
1963         seq_printf(seq,
1964                    "   Conns  Packets  Packets            Bytes            Bytes\n");
1965
1966         spin_lock_bh(&ip_vs_stats.lock);
1967         seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", ip_vs_stats.ustats.conns,
1968                    ip_vs_stats.ustats.inpkts, ip_vs_stats.ustats.outpkts,
1969                    (unsigned long long) ip_vs_stats.ustats.inbytes,
1970                    (unsigned long long) ip_vs_stats.ustats.outbytes);
1971
1972 /*                 01234567 01234567 01234567 0123456701234567 0123456701234567 */
1973         seq_puts(seq,
1974                    " Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
1975         seq_printf(seq,"%8X %8X %8X %16X %16X\n",
1976                         ip_vs_stats.ustats.cps,
1977                         ip_vs_stats.ustats.inpps,
1978                         ip_vs_stats.ustats.outpps,
1979                         ip_vs_stats.ustats.inbps,
1980                         ip_vs_stats.ustats.outbps);
1981         spin_unlock_bh(&ip_vs_stats.lock);
1982
1983         return 0;
1984 }
1985
1986 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
1987 {
1988         return single_open(file, ip_vs_stats_show, NULL);
1989 }
1990
1991 static const struct file_operations ip_vs_stats_fops = {
1992         .owner = THIS_MODULE,
1993         .open = ip_vs_stats_seq_open,
1994         .read = seq_read,
1995         .llseek = seq_lseek,
1996         .release = single_release,
1997 };
1998
1999 #endif
2000
2001 /*
2002  *      Set timeout values for tcp tcpfin udp in the timeout_table.
2003  */
2004 static int ip_vs_set_timeout(struct ip_vs_timeout_user *u)
2005 {
2006         IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2007                   u->tcp_timeout,
2008                   u->tcp_fin_timeout,
2009                   u->udp_timeout);
2010
2011 #ifdef CONFIG_IP_VS_PROTO_TCP
2012         if (u->tcp_timeout) {
2013                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_ESTABLISHED]
2014                         = u->tcp_timeout * HZ;
2015         }
2016
2017         if (u->tcp_fin_timeout) {
2018                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_FIN_WAIT]
2019                         = u->tcp_fin_timeout * HZ;
2020         }
2021 #endif
2022
2023 #ifdef CONFIG_IP_VS_PROTO_UDP
2024         if (u->udp_timeout) {
2025                 ip_vs_protocol_udp.timeout_table[IP_VS_UDP_S_NORMAL]
2026                         = u->udp_timeout * HZ;
2027         }
2028 #endif
2029         return 0;
2030 }
2031
2032
2033 #define SET_CMDID(cmd)          (cmd - IP_VS_BASE_CTL)
2034 #define SERVICE_ARG_LEN         (sizeof(struct ip_vs_service_user))
2035 #define SVCDEST_ARG_LEN         (sizeof(struct ip_vs_service_user) +    \
2036                                  sizeof(struct ip_vs_dest_user))
2037 #define TIMEOUT_ARG_LEN         (sizeof(struct ip_vs_timeout_user))
2038 #define DAEMON_ARG_LEN          (sizeof(struct ip_vs_daemon_user))
2039 #define MAX_ARG_LEN             SVCDEST_ARG_LEN
2040
2041 static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
2042         [SET_CMDID(IP_VS_SO_SET_ADD)]           = SERVICE_ARG_LEN,
2043         [SET_CMDID(IP_VS_SO_SET_EDIT)]          = SERVICE_ARG_LEN,
2044         [SET_CMDID(IP_VS_SO_SET_DEL)]           = SERVICE_ARG_LEN,
2045         [SET_CMDID(IP_VS_SO_SET_FLUSH)]         = 0,
2046         [SET_CMDID(IP_VS_SO_SET_ADDDEST)]       = SVCDEST_ARG_LEN,
2047         [SET_CMDID(IP_VS_SO_SET_DELDEST)]       = SVCDEST_ARG_LEN,
2048         [SET_CMDID(IP_VS_SO_SET_EDITDEST)]      = SVCDEST_ARG_LEN,
2049         [SET_CMDID(IP_VS_SO_SET_TIMEOUT)]       = TIMEOUT_ARG_LEN,
2050         [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)]   = DAEMON_ARG_LEN,
2051         [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)]    = DAEMON_ARG_LEN,
2052         [SET_CMDID(IP_VS_SO_SET_ZERO)]          = SERVICE_ARG_LEN,
2053 };
2054
2055 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2056                                   struct ip_vs_service_user *usvc_compat)
2057 {
2058         memset(usvc, 0, sizeof(*usvc));
2059
2060         usvc->af                = AF_INET;
2061         usvc->protocol          = usvc_compat->protocol;
2062         usvc->addr.ip           = usvc_compat->addr;
2063         usvc->port              = usvc_compat->port;
2064         usvc->fwmark            = usvc_compat->fwmark;
2065
2066         /* Deep copy of sched_name is not needed here */
2067         usvc->sched_name        = usvc_compat->sched_name;
2068
2069         usvc->flags             = usvc_compat->flags;
2070         usvc->timeout           = usvc_compat->timeout;
2071         usvc->netmask           = usvc_compat->netmask;
2072 }
2073
2074 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2075                                    struct ip_vs_dest_user *udest_compat)
2076 {
2077         memset(udest, 0, sizeof(*udest));
2078
2079         udest->addr.ip          = udest_compat->addr;
2080         udest->port             = udest_compat->port;
2081         udest->conn_flags       = udest_compat->conn_flags;
2082         udest->weight           = udest_compat->weight;
2083         udest->u_threshold      = udest_compat->u_threshold;
2084         udest->l_threshold      = udest_compat->l_threshold;
2085 }
2086
2087 static int
2088 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2089 {
2090         int ret;
2091         unsigned char arg[MAX_ARG_LEN];
2092         struct ip_vs_service_user *usvc_compat;
2093         struct ip_vs_service_user_kern usvc;
2094         struct ip_vs_service *svc;
2095         struct ip_vs_dest_user *udest_compat;
2096         struct ip_vs_dest_user_kern udest;
2097
2098         if (!capable(CAP_NET_ADMIN))
2099                 return -EPERM;
2100
2101         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2102                 return -EINVAL;
2103         if (len < 0 || len >  MAX_ARG_LEN)
2104                 return -EINVAL;
2105         if (len != set_arglen[SET_CMDID(cmd)]) {
2106                 pr_err("set_ctl: len %u != %u\n",
2107                        len, set_arglen[SET_CMDID(cmd)]);
2108                 return -EINVAL;
2109         }
2110
2111         if (copy_from_user(arg, user, len) != 0)
2112                 return -EFAULT;
2113
2114         /* increase the module use count */
2115         ip_vs_use_count_inc();
2116
2117         if (mutex_lock_interruptible(&__ip_vs_mutex)) {
2118                 ret = -ERESTARTSYS;
2119                 goto out_dec;
2120         }
2121
2122         if (cmd == IP_VS_SO_SET_FLUSH) {
2123                 /* Flush the virtual service */
2124                 ret = ip_vs_flush();
2125                 goto out_unlock;
2126         } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2127                 /* Set timeout values for (tcp tcpfin udp) */
2128                 ret = ip_vs_set_timeout((struct ip_vs_timeout_user *)arg);
2129                 goto out_unlock;
2130         } else if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2131                 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2132                 ret = start_sync_thread(dm->state, dm->mcast_ifn, dm->syncid);
2133                 goto out_unlock;
2134         } else if (cmd == IP_VS_SO_SET_STOPDAEMON) {
2135                 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2136                 ret = stop_sync_thread(dm->state);
2137                 goto out_unlock;
2138         }
2139
2140         usvc_compat = (struct ip_vs_service_user *)arg;
2141         udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2142
2143         /* We only use the new structs internally, so copy userspace compat
2144          * structs to extended internal versions */
2145         ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2146         ip_vs_copy_udest_compat(&udest, udest_compat);
2147
2148         if (cmd == IP_VS_SO_SET_ZERO) {
2149                 /* if no service address is set, zero counters in all */
2150                 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2151                         ret = ip_vs_zero_all();
2152                         goto out_unlock;
2153                 }
2154         }
2155
2156         /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2157         if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2158             usvc.protocol != IPPROTO_SCTP) {
2159                 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2160                        usvc.protocol, &usvc.addr.ip,
2161                        ntohs(usvc.port), usvc.sched_name);
2162                 ret = -EFAULT;
2163                 goto out_unlock;
2164         }
2165
2166         /* Lookup the exact service by <protocol, addr, port> or fwmark */
2167         if (usvc.fwmark == 0)
2168                 svc = __ip_vs_service_find(usvc.af, usvc.protocol,
2169                                            &usvc.addr, usvc.port);
2170         else
2171                 svc = __ip_vs_svc_fwm_find(usvc.af, usvc.fwmark);
2172
2173         if (cmd != IP_VS_SO_SET_ADD
2174             && (svc == NULL || svc->protocol != usvc.protocol)) {
2175                 ret = -ESRCH;
2176                 goto out_unlock;
2177         }
2178
2179         switch (cmd) {
2180         case IP_VS_SO_SET_ADD:
2181                 if (svc != NULL)
2182                         ret = -EEXIST;
2183                 else
2184                         ret = ip_vs_add_service(&usvc, &svc);
2185                 break;
2186         case IP_VS_SO_SET_EDIT:
2187                 ret = ip_vs_edit_service(svc, &usvc);
2188                 break;
2189         case IP_VS_SO_SET_DEL:
2190                 ret = ip_vs_del_service(svc);
2191                 if (!ret)
2192                         goto out_unlock;
2193                 break;
2194         case IP_VS_SO_SET_ZERO:
2195                 ret = ip_vs_zero_service(svc);
2196                 break;
2197         case IP_VS_SO_SET_ADDDEST:
2198                 ret = ip_vs_add_dest(svc, &udest);
2199                 break;
2200         case IP_VS_SO_SET_EDITDEST:
2201                 ret = ip_vs_edit_dest(svc, &udest);
2202                 break;
2203         case IP_VS_SO_SET_DELDEST:
2204                 ret = ip_vs_del_dest(svc, &udest);
2205                 break;
2206         default:
2207                 ret = -EINVAL;
2208         }
2209
2210   out_unlock:
2211         mutex_unlock(&__ip_vs_mutex);
2212   out_dec:
2213         /* decrease the module use count */
2214         ip_vs_use_count_dec();
2215
2216         return ret;
2217 }
2218
2219
2220 static void
2221 ip_vs_copy_stats(struct ip_vs_stats_user *dst, struct ip_vs_stats *src)
2222 {
2223         spin_lock_bh(&src->lock);
2224         memcpy(dst, &src->ustats, sizeof(*dst));
2225         spin_unlock_bh(&src->lock);
2226 }
2227
2228 static void
2229 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2230 {
2231         dst->protocol = src->protocol;
2232         dst->addr = src->addr.ip;
2233         dst->port = src->port;
2234         dst->fwmark = src->fwmark;
2235         strlcpy(dst->sched_name, src->scheduler->name, sizeof(dst->sched_name));
2236         dst->flags = src->flags;
2237         dst->timeout = src->timeout / HZ;
2238         dst->netmask = src->netmask;
2239         dst->num_dests = src->num_dests;
2240         ip_vs_copy_stats(&dst->stats, &src->stats);
2241 }
2242
2243 static inline int
2244 __ip_vs_get_service_entries(const struct ip_vs_get_services *get,
2245                             struct ip_vs_get_services __user *uptr)
2246 {
2247         int idx, count=0;
2248         struct ip_vs_service *svc;
2249         struct ip_vs_service_entry entry;
2250         int ret = 0;
2251
2252         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2253                 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2254                         /* Only expose IPv4 entries to old interface */
2255                         if (svc->af != AF_INET)
2256                                 continue;
2257
2258                         if (count >= get->num_services)
2259                                 goto out;
2260                         memset(&entry, 0, sizeof(entry));
2261                         ip_vs_copy_service(&entry, svc);
2262                         if (copy_to_user(&uptr->entrytable[count],
2263                                          &entry, sizeof(entry))) {
2264                                 ret = -EFAULT;
2265                                 goto out;
2266                         }
2267                         count++;
2268                 }
2269         }
2270
2271         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2272                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2273                         /* Only expose IPv4 entries to old interface */
2274                         if (svc->af != AF_INET)
2275                                 continue;
2276
2277                         if (count >= get->num_services)
2278                                 goto out;
2279                         memset(&entry, 0, sizeof(entry));
2280                         ip_vs_copy_service(&entry, svc);
2281                         if (copy_to_user(&uptr->entrytable[count],
2282                                          &entry, sizeof(entry))) {
2283                                 ret = -EFAULT;
2284                                 goto out;
2285                         }
2286                         count++;
2287                 }
2288         }
2289   out:
2290         return ret;
2291 }
2292
2293 static inline int
2294 __ip_vs_get_dest_entries(const struct ip_vs_get_dests *get,
2295                          struct ip_vs_get_dests __user *uptr)
2296 {
2297         struct ip_vs_service *svc;
2298         union nf_inet_addr addr = { .ip = get->addr };
2299         int ret = 0;
2300
2301         if (get->fwmark)
2302                 svc = __ip_vs_svc_fwm_find(AF_INET, get->fwmark);
2303         else
2304                 svc = __ip_vs_service_find(AF_INET, get->protocol, &addr,
2305                                            get->port);
2306
2307         if (svc) {
2308                 int count = 0;
2309                 struct ip_vs_dest *dest;
2310                 struct ip_vs_dest_entry entry;
2311
2312                 list_for_each_entry(dest, &svc->destinations, n_list) {
2313                         if (count >= get->num_dests)
2314                                 break;
2315
2316                         entry.addr = dest->addr.ip;
2317                         entry.port = dest->port;
2318                         entry.conn_flags = atomic_read(&dest->conn_flags);
2319                         entry.weight = atomic_read(&dest->weight);
2320                         entry.u_threshold = dest->u_threshold;
2321                         entry.l_threshold = dest->l_threshold;
2322                         entry.activeconns = atomic_read(&dest->activeconns);
2323                         entry.inactconns = atomic_read(&dest->inactconns);
2324                         entry.persistconns = atomic_read(&dest->persistconns);
2325                         ip_vs_copy_stats(&entry.stats, &dest->stats);
2326                         if (copy_to_user(&uptr->entrytable[count],
2327                                          &entry, sizeof(entry))) {
2328                                 ret = -EFAULT;
2329                                 break;
2330                         }
2331                         count++;
2332                 }
2333         } else
2334                 ret = -ESRCH;
2335         return ret;
2336 }
2337
2338 static inline void
2339 __ip_vs_get_timeouts(struct ip_vs_timeout_user *u)
2340 {
2341 #ifdef CONFIG_IP_VS_PROTO_TCP
2342         u->tcp_timeout =
2343                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2344         u->tcp_fin_timeout =
2345                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2346 #endif
2347 #ifdef CONFIG_IP_VS_PROTO_UDP
2348         u->udp_timeout =
2349                 ip_vs_protocol_udp.timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2350 #endif
2351 }
2352
2353
2354 #define GET_CMDID(cmd)          (cmd - IP_VS_BASE_CTL)
2355 #define GET_INFO_ARG_LEN        (sizeof(struct ip_vs_getinfo))
2356 #define GET_SERVICES_ARG_LEN    (sizeof(struct ip_vs_get_services))
2357 #define GET_SERVICE_ARG_LEN     (sizeof(struct ip_vs_service_entry))
2358 #define GET_DESTS_ARG_LEN       (sizeof(struct ip_vs_get_dests))
2359 #define GET_TIMEOUT_ARG_LEN     (sizeof(struct ip_vs_timeout_user))
2360 #define GET_DAEMON_ARG_LEN      (sizeof(struct ip_vs_daemon_user) * 2)
2361
2362 static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
2363         [GET_CMDID(IP_VS_SO_GET_VERSION)]       = 64,
2364         [GET_CMDID(IP_VS_SO_GET_INFO)]          = GET_INFO_ARG_LEN,
2365         [GET_CMDID(IP_VS_SO_GET_SERVICES)]      = GET_SERVICES_ARG_LEN,
2366         [GET_CMDID(IP_VS_SO_GET_SERVICE)]       = GET_SERVICE_ARG_LEN,
2367         [GET_CMDID(IP_VS_SO_GET_DESTS)]         = GET_DESTS_ARG_LEN,
2368         [GET_CMDID(IP_VS_SO_GET_TIMEOUT)]       = GET_TIMEOUT_ARG_LEN,
2369         [GET_CMDID(IP_VS_SO_GET_DAEMON)]        = GET_DAEMON_ARG_LEN,
2370 };
2371
2372 static int
2373 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2374 {
2375         unsigned char arg[128];
2376         int ret = 0;
2377         unsigned int copylen;
2378
2379         if (!capable(CAP_NET_ADMIN))
2380                 return -EPERM;
2381
2382         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2383                 return -EINVAL;
2384
2385         if (*len < get_arglen[GET_CMDID(cmd)]) {
2386                 pr_err("get_ctl: len %u < %u\n",
2387                        *len, get_arglen[GET_CMDID(cmd)]);
2388                 return -EINVAL;
2389         }
2390
2391         copylen = get_arglen[GET_CMDID(cmd)];
2392         if (copylen > 128)
2393                 return -EINVAL;
2394
2395         if (copy_from_user(arg, user, copylen) != 0)
2396                 return -EFAULT;
2397
2398         if (mutex_lock_interruptible(&__ip_vs_mutex))
2399                 return -ERESTARTSYS;
2400
2401         switch (cmd) {
2402         case IP_VS_SO_GET_VERSION:
2403         {
2404                 char buf[64];
2405
2406                 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2407                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2408                 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2409                         ret = -EFAULT;
2410                         goto out;
2411                 }
2412                 *len = strlen(buf)+1;
2413         }
2414         break;
2415
2416         case IP_VS_SO_GET_INFO:
2417         {
2418                 struct ip_vs_getinfo info;
2419                 info.version = IP_VS_VERSION_CODE;
2420                 info.size = ip_vs_conn_tab_size;
2421                 info.num_services = ip_vs_num_services;
2422                 if (copy_to_user(user, &info, sizeof(info)) != 0)
2423                         ret = -EFAULT;
2424         }
2425         break;
2426
2427         case IP_VS_SO_GET_SERVICES:
2428         {
2429                 struct ip_vs_get_services *get;
2430                 int size;
2431
2432                 get = (struct ip_vs_get_services *)arg;
2433                 size = sizeof(*get) +
2434                         sizeof(struct ip_vs_service_entry) * get->num_services;
2435                 if (*len != size) {
2436                         pr_err("length: %u != %u\n", *len, size);
2437                         ret = -EINVAL;
2438                         goto out;
2439                 }
2440                 ret = __ip_vs_get_service_entries(get, user);
2441         }
2442         break;
2443
2444         case IP_VS_SO_GET_SERVICE:
2445         {
2446                 struct ip_vs_service_entry *entry;
2447                 struct ip_vs_service *svc;
2448                 union nf_inet_addr addr;
2449
2450                 entry = (struct ip_vs_service_entry *)arg;
2451                 addr.ip = entry->addr;
2452                 if (entry->fwmark)
2453                         svc = __ip_vs_svc_fwm_find(AF_INET, entry->fwmark);
2454                 else
2455                         svc = __ip_vs_service_find(AF_INET, entry->protocol,
2456                                                    &addr, entry->port);
2457                 if (svc) {
2458                         ip_vs_copy_service(entry, svc);
2459                         if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2460                                 ret = -EFAULT;
2461                 } else
2462                         ret = -ESRCH;
2463         }
2464         break;
2465
2466         case IP_VS_SO_GET_DESTS:
2467         {
2468                 struct ip_vs_get_dests *get;
2469                 int size;
2470
2471                 get = (struct ip_vs_get_dests *)arg;
2472                 size = sizeof(*get) +
2473                         sizeof(struct ip_vs_dest_entry) * get->num_dests;
2474                 if (*len != size) {
2475                         pr_err("length: %u != %u\n", *len, size);
2476                         ret = -EINVAL;
2477                         goto out;
2478                 }
2479                 ret = __ip_vs_get_dest_entries(get, user);
2480         }
2481         break;
2482
2483         case IP_VS_SO_GET_TIMEOUT:
2484         {
2485                 struct ip_vs_timeout_user t;
2486
2487                 __ip_vs_get_timeouts(&t);
2488                 if (copy_to_user(user, &t, sizeof(t)) != 0)
2489                         ret = -EFAULT;
2490         }
2491         break;
2492
2493         case IP_VS_SO_GET_DAEMON:
2494         {
2495                 struct ip_vs_daemon_user d[2];
2496
2497                 memset(&d, 0, sizeof(d));
2498                 if (ip_vs_sync_state & IP_VS_STATE_MASTER) {
2499                         d[0].state = IP_VS_STATE_MASTER;
2500                         strlcpy(d[0].mcast_ifn, ip_vs_master_mcast_ifn, sizeof(d[0].mcast_ifn));
2501                         d[0].syncid = ip_vs_master_syncid;
2502                 }
2503                 if (ip_vs_sync_state & IP_VS_STATE_BACKUP) {
2504                         d[1].state = IP_VS_STATE_BACKUP;
2505                         strlcpy(d[1].mcast_ifn, ip_vs_backup_mcast_ifn, sizeof(d[1].mcast_ifn));
2506                         d[1].syncid = ip_vs_backup_syncid;
2507                 }
2508                 if (copy_to_user(user, &d, sizeof(d)) != 0)
2509                         ret = -EFAULT;
2510         }
2511         break;
2512
2513         default:
2514                 ret = -EINVAL;
2515         }
2516
2517   out:
2518         mutex_unlock(&__ip_vs_mutex);
2519         return ret;
2520 }
2521
2522
2523 static struct nf_sockopt_ops ip_vs_sockopts = {
2524         .pf             = PF_INET,
2525         .set_optmin     = IP_VS_BASE_CTL,
2526         .set_optmax     = IP_VS_SO_SET_MAX+1,
2527         .set            = do_ip_vs_set_ctl,
2528         .get_optmin     = IP_VS_BASE_CTL,
2529         .get_optmax     = IP_VS_SO_GET_MAX+1,
2530         .get            = do_ip_vs_get_ctl,
2531         .owner          = THIS_MODULE,
2532 };
2533
2534 /*
2535  * Generic Netlink interface
2536  */
2537
2538 /* IPVS genetlink family */
2539 static struct genl_family ip_vs_genl_family = {
2540         .id             = GENL_ID_GENERATE,
2541         .hdrsize        = 0,
2542         .name           = IPVS_GENL_NAME,
2543         .version        = IPVS_GENL_VERSION,
2544         .maxattr        = IPVS_CMD_MAX,
2545 };
2546
2547 /* Policy used for first-level command attributes */
2548 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2549         [IPVS_CMD_ATTR_SERVICE]         = { .type = NLA_NESTED },
2550         [IPVS_CMD_ATTR_DEST]            = { .type = NLA_NESTED },
2551         [IPVS_CMD_ATTR_DAEMON]          = { .type = NLA_NESTED },
2552         [IPVS_CMD_ATTR_TIMEOUT_TCP]     = { .type = NLA_U32 },
2553         [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2554         [IPVS_CMD_ATTR_TIMEOUT_UDP]     = { .type = NLA_U32 },
2555 };
2556
2557 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2558 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2559         [IPVS_DAEMON_ATTR_STATE]        = { .type = NLA_U32 },
2560         [IPVS_DAEMON_ATTR_MCAST_IFN]    = { .type = NLA_NUL_STRING,
2561                                             .len = IP_VS_IFNAME_MAXLEN },
2562         [IPVS_DAEMON_ATTR_SYNC_ID]      = { .type = NLA_U32 },
2563 };
2564
2565 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2566 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2567         [IPVS_SVC_ATTR_AF]              = { .type = NLA_U16 },
2568         [IPVS_SVC_ATTR_PROTOCOL]        = { .type = NLA_U16 },
2569         [IPVS_SVC_ATTR_ADDR]            = { .type = NLA_BINARY,
2570                                             .len = sizeof(union nf_inet_addr) },
2571         [IPVS_SVC_ATTR_PORT]            = { .type = NLA_U16 },
2572         [IPVS_SVC_ATTR_FWMARK]          = { .type = NLA_U32 },
2573         [IPVS_SVC_ATTR_SCHED_NAME]      = { .type = NLA_NUL_STRING,
2574                                             .len = IP_VS_SCHEDNAME_MAXLEN },
2575         [IPVS_SVC_ATTR_PE_NAME]         = { .type = NLA_NUL_STRING,
2576                                             .len = IP_VS_PENAME_MAXLEN },
2577         [IPVS_SVC_ATTR_FLAGS]           = { .type = NLA_BINARY,
2578                                             .len = sizeof(struct ip_vs_flags) },
2579         [IPVS_SVC_ATTR_TIMEOUT]         = { .type = NLA_U32 },
2580         [IPVS_SVC_ATTR_NETMASK]         = { .type = NLA_U32 },
2581         [IPVS_SVC_ATTR_STATS]           = { .type = NLA_NESTED },
2582 };
2583
2584 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2585 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2586         [IPVS_DEST_ATTR_ADDR]           = { .type = NLA_BINARY,
2587                                             .len = sizeof(union nf_inet_addr) },
2588         [IPVS_DEST_ATTR_PORT]           = { .type = NLA_U16 },
2589         [IPVS_DEST_ATTR_FWD_METHOD]     = { .type = NLA_U32 },
2590         [IPVS_DEST_ATTR_WEIGHT]         = { .type = NLA_U32 },
2591         [IPVS_DEST_ATTR_U_THRESH]       = { .type = NLA_U32 },
2592         [IPVS_DEST_ATTR_L_THRESH]       = { .type = NLA_U32 },
2593         [IPVS_DEST_ATTR_ACTIVE_CONNS]   = { .type = NLA_U32 },
2594         [IPVS_DEST_ATTR_INACT_CONNS]    = { .type = NLA_U32 },
2595         [IPVS_DEST_ATTR_PERSIST_CONNS]  = { .type = NLA_U32 },
2596         [IPVS_DEST_ATTR_STATS]          = { .type = NLA_NESTED },
2597 };
2598
2599 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2600                                  struct ip_vs_stats *stats)
2601 {
2602         struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2603         if (!nl_stats)
2604                 return -EMSGSIZE;
2605
2606         spin_lock_bh(&stats->lock);
2607
2608         NLA_PUT_U32(skb, IPVS_STATS_ATTR_CONNS, stats->ustats.conns);
2609         NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPKTS, stats->ustats.inpkts);
2610         NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPKTS, stats->ustats.outpkts);
2611         NLA_PUT_U64(skb, IPVS_STATS_ATTR_INBYTES, stats->ustats.inbytes);
2612         NLA_PUT_U64(skb, IPVS_STATS_ATTR_OUTBYTES, stats->ustats.outbytes);
2613         NLA_PUT_U32(skb, IPVS_STATS_ATTR_CPS, stats->ustats.cps);
2614         NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPPS, stats->ustats.inpps);
2615         NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPPS, stats->ustats.outpps);
2616         NLA_PUT_U32(skb, IPVS_STATS_ATTR_INBPS, stats->ustats.inbps);
2617         NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTBPS, stats->ustats.outbps);
2618
2619         spin_unlock_bh(&stats->lock);
2620
2621         nla_nest_end(skb, nl_stats);
2622
2623         return 0;
2624
2625 nla_put_failure:
2626         spin_unlock_bh(&stats->lock);
2627         nla_nest_cancel(skb, nl_stats);
2628         return -EMSGSIZE;
2629 }
2630
2631 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2632                                    struct ip_vs_service *svc)
2633 {
2634         struct nlattr *nl_service;
2635         struct ip_vs_flags flags = { .flags = svc->flags,
2636                                      .mask = ~0 };
2637
2638         nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2639         if (!nl_service)
2640                 return -EMSGSIZE;
2641
2642         NLA_PUT_U16(skb, IPVS_SVC_ATTR_AF, svc->af);
2643
2644         if (svc->fwmark) {
2645                 NLA_PUT_U32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark);
2646         } else {
2647                 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol);
2648                 NLA_PUT(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr);
2649                 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PORT, svc->port);
2650         }
2651
2652         NLA_PUT_STRING(skb, IPVS_SVC_ATTR_SCHED_NAME, svc->scheduler->name);
2653         if (svc->pe)
2654                 NLA_PUT_STRING(skb, IPVS_SVC_ATTR_PE_NAME, svc->pe->name);
2655         NLA_PUT(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags);
2656         NLA_PUT_U32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ);
2657         NLA_PUT_U32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask);
2658
2659         if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2660                 goto nla_put_failure;
2661
2662         nla_nest_end(skb, nl_service);
2663
2664         return 0;
2665
2666 nla_put_failure:
2667         nla_nest_cancel(skb, nl_service);
2668         return -EMSGSIZE;
2669 }
2670
2671 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2672                                    struct ip_vs_service *svc,
2673                                    struct netlink_callback *cb)
2674 {
2675         void *hdr;
2676
2677         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2678                           &ip_vs_genl_family, NLM_F_MULTI,
2679                           IPVS_CMD_NEW_SERVICE);
2680         if (!hdr)
2681                 return -EMSGSIZE;
2682
2683         if (ip_vs_genl_fill_service(skb, svc) < 0)
2684                 goto nla_put_failure;
2685
2686         return genlmsg_end(skb, hdr);
2687
2688 nla_put_failure:
2689         genlmsg_cancel(skb, hdr);
2690         return -EMSGSIZE;
2691 }
2692
2693 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2694                                     struct netlink_callback *cb)
2695 {
2696         int idx = 0, i;
2697         int start = cb->args[0];
2698         struct ip_vs_service *svc;
2699
2700         mutex_lock(&__ip_vs_mutex);
2701         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2702                 list_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2703                         if (++idx <= start)
2704                                 continue;
2705                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2706                                 idx--;
2707                                 goto nla_put_failure;
2708                         }
2709                 }
2710         }
2711
2712         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2713                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
2714                         if (++idx <= start)
2715                                 continue;
2716                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2717                                 idx--;
2718                                 goto nla_put_failure;
2719                         }
2720                 }
2721         }
2722
2723 nla_put_failure:
2724         mutex_unlock(&__ip_vs_mutex);
2725         cb->args[0] = idx;
2726
2727         return skb->len;
2728 }
2729
2730 static int ip_vs_genl_parse_service(struct ip_vs_service_user_kern *usvc,
2731                                     struct nlattr *nla, int full_entry,
2732                                     struct ip_vs_service **ret_svc)
2733 {
2734         struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
2735         struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
2736         struct ip_vs_service *svc;
2737
2738         /* Parse mandatory identifying service fields first */
2739         if (nla == NULL ||
2740             nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
2741                 return -EINVAL;
2742
2743         nla_af          = attrs[IPVS_SVC_ATTR_AF];
2744         nla_protocol    = attrs[IPVS_SVC_ATTR_PROTOCOL];
2745         nla_addr        = attrs[IPVS_SVC_ATTR_ADDR];
2746         nla_port        = attrs[IPVS_SVC_ATTR_PORT];
2747         nla_fwmark      = attrs[IPVS_SVC_ATTR_FWMARK];
2748
2749         if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
2750                 return -EINVAL;
2751
2752         memset(usvc, 0, sizeof(*usvc));
2753
2754         usvc->af = nla_get_u16(nla_af);
2755 #ifdef CONFIG_IP_VS_IPV6
2756         if (usvc->af != AF_INET && usvc->af != AF_INET6)
2757 #else
2758         if (usvc->af != AF_INET)
2759 #endif
2760                 return -EAFNOSUPPORT;
2761
2762         if (nla_fwmark) {
2763                 usvc->protocol = IPPROTO_TCP;
2764                 usvc->fwmark = nla_get_u32(nla_fwmark);
2765         } else {
2766                 usvc->protocol = nla_get_u16(nla_protocol);
2767                 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
2768                 usvc->port = nla_get_u16(nla_port);
2769                 usvc->fwmark = 0;
2770         }
2771
2772         if (usvc->fwmark)
2773                 svc = __ip_vs_svc_fwm_find(usvc->af, usvc->fwmark);
2774         else
2775                 svc = __ip_vs_service_find(usvc->af, usvc->protocol,
2776                                            &usvc->addr, usvc->port);
2777         *ret_svc = svc;
2778
2779         /* If a full entry was requested, check for the additional fields */
2780         if (full_entry) {
2781                 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
2782                               *nla_netmask;
2783                 struct ip_vs_flags flags;
2784
2785                 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
2786                 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
2787                 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
2788                 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
2789                 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
2790
2791                 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
2792                         return -EINVAL;
2793
2794                 nla_memcpy(&flags, nla_flags, sizeof(flags));
2795
2796                 /* prefill flags from service if it already exists */
2797                 if (svc)
2798                         usvc->flags = svc->flags;
2799
2800                 /* set new flags from userland */
2801                 usvc->flags = (usvc->flags & ~flags.mask) |
2802                               (flags.flags & flags.mask);
2803                 usvc->sched_name = nla_data(nla_sched);
2804                 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
2805                 usvc->timeout = nla_get_u32(nla_timeout);
2806                 usvc->netmask = nla_get_u32(nla_netmask);
2807         }
2808
2809         return 0;
2810 }
2811
2812 static struct ip_vs_service *ip_vs_genl_find_service(struct nlattr *nla)
2813 {
2814         struct ip_vs_service_user_kern usvc;
2815         struct ip_vs_service *svc;
2816         int ret;
2817
2818         ret = ip_vs_genl_parse_service(&usvc, nla, 0, &svc);
2819         return ret ? ERR_PTR(ret) : svc;
2820 }
2821
2822 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
2823 {
2824         struct nlattr *nl_dest;
2825
2826         nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
2827         if (!nl_dest)
2828                 return -EMSGSIZE;
2829
2830         NLA_PUT(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr);
2831         NLA_PUT_U16(skb, IPVS_DEST_ATTR_PORT, dest->port);
2832
2833         NLA_PUT_U32(skb, IPVS_DEST_ATTR_FWD_METHOD,
2834                     atomic_read(&dest->conn_flags) & IP_VS_CONN_F_FWD_MASK);
2835         NLA_PUT_U32(skb, IPVS_DEST_ATTR_WEIGHT, atomic_read(&dest->weight));
2836         NLA_PUT_U32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold);
2837         NLA_PUT_U32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold);
2838         NLA_PUT_U32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
2839                     atomic_read(&dest->activeconns));
2840         NLA_PUT_U32(skb, IPVS_DEST_ATTR_INACT_CONNS,
2841                     atomic_read(&dest->inactconns));
2842         NLA_PUT_U32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
2843                     atomic_read(&dest->persistconns));
2844
2845         if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
2846                 goto nla_put_failure;
2847
2848         nla_nest_end(skb, nl_dest);
2849
2850         return 0;
2851
2852 nla_put_failure:
2853         nla_nest_cancel(skb, nl_dest);
2854         return -EMSGSIZE;
2855 }
2856
2857 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
2858                                 struct netlink_callback *cb)
2859 {
2860         void *hdr;
2861
2862         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2863                           &ip_vs_genl_family, NLM_F_MULTI,
2864                           IPVS_CMD_NEW_DEST);
2865         if (!hdr)
2866                 return -EMSGSIZE;
2867
2868         if (ip_vs_genl_fill_dest(skb, dest) < 0)
2869                 goto nla_put_failure;
2870
2871         return genlmsg_end(skb, hdr);
2872
2873 nla_put_failure:
2874         genlmsg_cancel(skb, hdr);
2875         return -EMSGSIZE;
2876 }
2877
2878 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
2879                                  struct netlink_callback *cb)
2880 {
2881         int idx = 0;
2882         int start = cb->args[0];
2883         struct ip_vs_service *svc;
2884         struct ip_vs_dest *dest;
2885         struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
2886
2887         mutex_lock(&__ip_vs_mutex);
2888
2889         /* Try to find the service for which to dump destinations */
2890         if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
2891                         IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
2892                 goto out_err;
2893
2894         svc = ip_vs_genl_find_service(attrs[IPVS_CMD_ATTR_SERVICE]);
2895         if (IS_ERR(svc) || svc == NULL)
2896                 goto out_err;
2897
2898         /* Dump the destinations */
2899         list_for_each_entry(dest, &svc->destinations, n_list) {
2900                 if (++idx <= start)
2901                         continue;
2902                 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
2903                         idx--;
2904                         goto nla_put_failure;
2905                 }
2906         }
2907
2908 nla_put_failure:
2909         cb->args[0] = idx;
2910
2911 out_err:
2912         mutex_unlock(&__ip_vs_mutex);
2913
2914         return skb->len;
2915 }
2916
2917 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
2918                                  struct nlattr *nla, int full_entry)
2919 {
2920         struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
2921         struct nlattr *nla_addr, *nla_port;
2922
2923         /* Parse mandatory identifying destination fields first */
2924         if (nla == NULL ||
2925             nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
2926                 return -EINVAL;
2927
2928         nla_addr        = attrs[IPVS_DEST_ATTR_ADDR];
2929         nla_port        = attrs[IPVS_DEST_ATTR_PORT];
2930
2931         if (!(nla_addr && nla_port))
2932                 return -EINVAL;
2933
2934         memset(udest, 0, sizeof(*udest));
2935
2936         nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
2937         udest->port = nla_get_u16(nla_port);
2938
2939         /* If a full entry was requested, check for the additional fields */
2940         if (full_entry) {
2941                 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
2942                               *nla_l_thresh;
2943
2944                 nla_fwd         = attrs[IPVS_DEST_ATTR_FWD_METHOD];
2945                 nla_weight      = attrs[IPVS_DEST_ATTR_WEIGHT];
2946                 nla_u_thresh    = attrs[IPVS_DEST_ATTR_U_THRESH];
2947                 nla_l_thresh    = attrs[IPVS_DEST_ATTR_L_THRESH];
2948
2949                 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
2950                         return -EINVAL;
2951
2952                 udest->conn_flags = nla_get_u32(nla_fwd)
2953                                     & IP_VS_CONN_F_FWD_MASK;
2954                 udest->weight = nla_get_u32(nla_weight);
2955                 udest->u_threshold = nla_get_u32(nla_u_thresh);
2956                 udest->l_threshold = nla_get_u32(nla_l_thresh);
2957         }
2958
2959         return 0;
2960 }
2961
2962 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __be32 state,
2963                                   const char *mcast_ifn, __be32 syncid)
2964 {
2965         struct nlattr *nl_daemon;
2966
2967         nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
2968         if (!nl_daemon)
2969                 return -EMSGSIZE;
2970
2971         NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_STATE, state);
2972         NLA_PUT_STRING(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn);
2973         NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid);
2974
2975         nla_nest_end(skb, nl_daemon);
2976
2977         return 0;
2978
2979 nla_put_failure:
2980         nla_nest_cancel(skb, nl_daemon);
2981         return -EMSGSIZE;
2982 }
2983
2984 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __be32 state,
2985                                   const char *mcast_ifn, __be32 syncid,
2986                                   struct netlink_callback *cb)
2987 {
2988         void *hdr;
2989         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2990                           &ip_vs_genl_family, NLM_F_MULTI,
2991                           IPVS_CMD_NEW_DAEMON);
2992         if (!hdr)
2993                 return -EMSGSIZE;
2994
2995         if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
2996                 goto nla_put_failure;
2997
2998         return genlmsg_end(skb, hdr);
2999
3000 nla_put_failure:
3001         genlmsg_cancel(skb, hdr);
3002         return -EMSGSIZE;
3003 }
3004
3005 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3006                                    struct netlink_callback *cb)
3007 {
3008         mutex_lock(&__ip_vs_mutex);
3009         if ((ip_vs_sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3010                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3011                                            ip_vs_master_mcast_ifn,
3012                                            ip_vs_master_syncid, cb) < 0)
3013                         goto nla_put_failure;
3014
3015                 cb->args[0] = 1;
3016         }
3017
3018         if ((ip_vs_sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3019                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3020                                            ip_vs_backup_mcast_ifn,
3021                                            ip_vs_backup_syncid, cb) < 0)
3022                         goto nla_put_failure;
3023
3024                 cb->args[1] = 1;
3025         }
3026
3027 nla_put_failure:
3028         mutex_unlock(&__ip_vs_mutex);
3029
3030         return skb->len;
3031 }
3032
3033 static int ip_vs_genl_new_daemon(struct nlattr **attrs)
3034 {
3035         if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3036               attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3037               attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3038                 return -EINVAL;
3039
3040         return start_sync_thread(nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3041                                  nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3042                                  nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3043 }
3044
3045 static int ip_vs_genl_del_daemon(struct nlattr **attrs)
3046 {
3047         if (!attrs[IPVS_DAEMON_ATTR_STATE])
3048                 return -EINVAL;
3049
3050         return stop_sync_thread(nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3051 }
3052
3053 static int ip_vs_genl_set_config(struct nlattr **attrs)
3054 {
3055         struct ip_vs_timeout_user t;
3056
3057         __ip_vs_get_timeouts(&t);
3058
3059         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3060                 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3061
3062         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3063                 t.tcp_fin_timeout =
3064                         nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3065
3066         if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3067                 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3068
3069         return ip_vs_set_timeout(&t);
3070 }
3071
3072 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3073 {
3074         struct ip_vs_service *svc = NULL;
3075         struct ip_vs_service_user_kern usvc;
3076         struct ip_vs_dest_user_kern udest;
3077         int ret = 0, cmd;
3078         int need_full_svc = 0, need_full_dest = 0;
3079
3080         cmd = info->genlhdr->cmd;
3081
3082         mutex_lock(&__ip_vs_mutex);
3083
3084         if (cmd == IPVS_CMD_FLUSH) {
3085                 ret = ip_vs_flush();
3086                 goto out;
3087         } else if (cmd == IPVS_CMD_SET_CONFIG) {
3088                 ret = ip_vs_genl_set_config(info->attrs);
3089                 goto out;
3090         } else if (cmd == IPVS_CMD_NEW_DAEMON ||
3091                    cmd == IPVS_CMD_DEL_DAEMON) {
3092
3093                 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3094
3095                 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3096                     nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3097                                      info->attrs[IPVS_CMD_ATTR_DAEMON],
3098                                      ip_vs_daemon_policy)) {
3099                         ret = -EINVAL;
3100                         goto out;
3101                 }
3102
3103                 if (cmd == IPVS_CMD_NEW_DAEMON)
3104                         ret = ip_vs_genl_new_daemon(daemon_attrs);
3105                 else
3106                         ret = ip_vs_genl_del_daemon(daemon_attrs);
3107                 goto out;
3108         } else if (cmd == IPVS_CMD_ZERO &&
3109                    !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3110                 ret = ip_vs_zero_all();
3111                 goto out;
3112         }
3113
3114         /* All following commands require a service argument, so check if we
3115          * received a valid one. We need a full service specification when
3116          * adding / editing a service. Only identifying members otherwise. */
3117         if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3118                 need_full_svc = 1;
3119
3120         ret = ip_vs_genl_parse_service(&usvc,
3121                                        info->attrs[IPVS_CMD_ATTR_SERVICE],
3122                                        need_full_svc, &svc);
3123         if (ret)
3124                 goto out;
3125
3126         /* Unless we're adding a new service, the service must already exist */
3127         if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3128                 ret = -ESRCH;
3129                 goto out;
3130         }
3131
3132         /* Destination commands require a valid destination argument. For
3133          * adding / editing a destination, we need a full destination
3134          * specification. */
3135         if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3136             cmd == IPVS_CMD_DEL_DEST) {
3137                 if (cmd != IPVS_CMD_DEL_DEST)
3138                         need_full_dest = 1;
3139
3140                 ret = ip_vs_genl_parse_dest(&udest,
3141                                             info->attrs[IPVS_CMD_ATTR_DEST],
3142                                             need_full_dest);
3143                 if (ret)
3144                         goto out;
3145         }
3146
3147         switch (cmd) {
3148         case IPVS_CMD_NEW_SERVICE:
3149                 if (svc == NULL)
3150                         ret = ip_vs_add_service(&usvc, &svc);
3151                 else
3152                         ret = -EEXIST;
3153                 break;
3154         case IPVS_CMD_SET_SERVICE:
3155                 ret = ip_vs_edit_service(svc, &usvc);
3156                 break;
3157         case IPVS_CMD_DEL_SERVICE:
3158                 ret = ip_vs_del_service(svc);
3159                 /* do not use svc, it can be freed */
3160                 break;
3161         case IPVS_CMD_NEW_DEST:
3162                 ret = ip_vs_add_dest(svc, &udest);
3163                 break;
3164         case IPVS_CMD_SET_DEST:
3165                 ret = ip_vs_edit_dest(svc, &udest);
3166                 break;
3167         case IPVS_CMD_DEL_DEST:
3168                 ret = ip_vs_del_dest(svc, &udest);
3169                 break;
3170         case IPVS_CMD_ZERO:
3171                 ret = ip_vs_zero_service(svc);
3172                 break;
3173         default:
3174                 ret = -EINVAL;
3175         }
3176
3177 out:
3178         mutex_unlock(&__ip_vs_mutex);
3179
3180         return ret;
3181 }
3182
3183 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3184 {
3185         struct sk_buff *msg;
3186         void *reply;
3187         int ret, cmd, reply_cmd;
3188
3189         cmd = info->genlhdr->cmd;
3190
3191         if (cmd == IPVS_CMD_GET_SERVICE)
3192                 reply_cmd = IPVS_CMD_NEW_SERVICE;
3193         else if (cmd == IPVS_CMD_GET_INFO)
3194                 reply_cmd = IPVS_CMD_SET_INFO;
3195         else if (cmd == IPVS_CMD_GET_CONFIG)
3196                 reply_cmd = IPVS_CMD_SET_CONFIG;
3197         else {
3198                 pr_err("unknown Generic Netlink command\n");
3199                 return -EINVAL;
3200         }
3201
3202         msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3203         if (!msg)
3204                 return -ENOMEM;
3205
3206         mutex_lock(&__ip_vs_mutex);
3207
3208         reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3209         if (reply == NULL)
3210                 goto nla_put_failure;
3211
3212         switch (cmd) {
3213         case IPVS_CMD_GET_SERVICE:
3214         {
3215                 struct ip_vs_service *svc;
3216
3217                 svc = ip_vs_genl_find_service(info->attrs[IPVS_CMD_ATTR_SERVICE]);
3218                 if (IS_ERR(svc)) {
3219                         ret = PTR_ERR(svc);
3220                         goto out_err;
3221                 } else if (svc) {
3222                         ret = ip_vs_genl_fill_service(msg, svc);
3223                         if (ret)
3224                                 goto nla_put_failure;
3225                 } else {
3226                         ret = -ESRCH;
3227                         goto out_err;
3228                 }
3229
3230                 break;
3231         }
3232
3233         case IPVS_CMD_GET_CONFIG:
3234         {
3235                 struct ip_vs_timeout_user t;
3236
3237                 __ip_vs_get_timeouts(&t);
3238 #ifdef CONFIG_IP_VS_PROTO_TCP
3239                 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP, t.tcp_timeout);
3240                 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3241                             t.tcp_fin_timeout);
3242 #endif
3243 #ifdef CONFIG_IP_VS_PROTO_UDP
3244                 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout);
3245 #endif
3246
3247                 break;
3248         }
3249
3250         case IPVS_CMD_GET_INFO:
3251                 NLA_PUT_U32(msg, IPVS_INFO_ATTR_VERSION, IP_VS_VERSION_CODE);
3252                 NLA_PUT_U32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3253                             ip_vs_conn_tab_size);
3254                 break;
3255         }
3256
3257         genlmsg_end(msg, reply);
3258         ret = genlmsg_reply(msg, info);
3259         goto out;
3260
3261 nla_put_failure:
3262         pr_err("not enough space in Netlink message\n");
3263         ret = -EMSGSIZE;
3264
3265 out_err:
3266         nlmsg_free(msg);
3267 out:
3268         mutex_unlock(&__ip_vs_mutex);
3269
3270         return ret;
3271 }
3272
3273
3274 static struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3275         {
3276                 .cmd    = IPVS_CMD_NEW_SERVICE,
3277                 .flags  = GENL_ADMIN_PERM,
3278                 .policy = ip_vs_cmd_policy,
3279                 .doit   = ip_vs_genl_set_cmd,
3280         },
3281         {
3282                 .cmd    = IPVS_CMD_SET_SERVICE,
3283                 .flags  = GENL_ADMIN_PERM,
3284                 .policy = ip_vs_cmd_policy,
3285                 .doit   = ip_vs_genl_set_cmd,
3286         },
3287         {
3288                 .cmd    = IPVS_CMD_DEL_SERVICE,
3289                 .flags  = GENL_ADMIN_PERM,
3290                 .policy = ip_vs_cmd_policy,
3291                 .doit   = ip_vs_genl_set_cmd,
3292         },
3293         {
3294                 .cmd    = IPVS_CMD_GET_SERVICE,
3295                 .flags  = GENL_ADMIN_PERM,
3296                 .doit   = ip_vs_genl_get_cmd,
3297                 .dumpit = ip_vs_genl_dump_services,
3298                 .policy = ip_vs_cmd_policy,
3299         },
3300         {
3301                 .cmd    = IPVS_CMD_NEW_DEST,
3302                 .flags  = GENL_ADMIN_PERM,
3303                 .policy = ip_vs_cmd_policy,
3304                 .doit   = ip_vs_genl_set_cmd,
3305         },
3306         {
3307                 .cmd    = IPVS_CMD_SET_DEST,
3308                 .flags  = GENL_ADMIN_PERM,
3309                 .policy = ip_vs_cmd_policy,
3310                 .doit   = ip_vs_genl_set_cmd,
3311         },
3312         {
3313                 .cmd    = IPVS_CMD_DEL_DEST,
3314                 .flags  = GENL_ADMIN_PERM,
3315                 .policy = ip_vs_cmd_policy,
3316                 .doit   = ip_vs_genl_set_cmd,
3317         },
3318         {
3319                 .cmd    = IPVS_CMD_GET_DEST,
3320                 .flags  = GENL_ADMIN_PERM,
3321                 .policy = ip_vs_cmd_policy,
3322                 .dumpit = ip_vs_genl_dump_dests,
3323         },
3324         {
3325                 .cmd    = IPVS_CMD_NEW_DAEMON,
3326                 .flags  = GENL_ADMIN_PERM,
3327                 .policy = ip_vs_cmd_policy,
3328                 .doit   = ip_vs_genl_set_cmd,
3329         },
3330         {
3331                 .cmd    = IPVS_CMD_DEL_DAEMON,
3332                 .flags  = GENL_ADMIN_PERM,
3333                 .policy = ip_vs_cmd_policy,
3334                 .doit   = ip_vs_genl_set_cmd,
3335         },
3336         {
3337                 .cmd    = IPVS_CMD_GET_DAEMON,
3338                 .flags  = GENL_ADMIN_PERM,
3339                 .dumpit = ip_vs_genl_dump_daemons,
3340         },
3341         {
3342                 .cmd    = IPVS_CMD_SET_CONFIG,
3343                 .flags  = GENL_ADMIN_PERM,
3344                 .policy = ip_vs_cmd_policy,
3345                 .doit   = ip_vs_genl_set_cmd,
3346         },
3347         {
3348                 .cmd    = IPVS_CMD_GET_CONFIG,
3349                 .flags  = GENL_ADMIN_PERM,
3350                 .doit   = ip_vs_genl_get_cmd,
3351         },
3352         {
3353                 .cmd    = IPVS_CMD_GET_INFO,
3354                 .flags  = GENL_ADMIN_PERM,
3355                 .doit   = ip_vs_genl_get_cmd,
3356         },
3357         {
3358                 .cmd    = IPVS_CMD_ZERO,
3359                 .flags  = GENL_ADMIN_PERM,
3360                 .policy = ip_vs_cmd_policy,
3361                 .doit   = ip_vs_genl_set_cmd,
3362         },
3363         {
3364                 .cmd    = IPVS_CMD_FLUSH,
3365                 .flags  = GENL_ADMIN_PERM,
3366                 .doit   = ip_vs_genl_set_cmd,
3367         },
3368 };
3369
3370 static int __init ip_vs_genl_register(void)
3371 {
3372         return genl_register_family_with_ops(&ip_vs_genl_family,
3373                 ip_vs_genl_ops, ARRAY_SIZE(ip_vs_genl_ops));
3374 }
3375
3376 static void ip_vs_genl_unregister(void)
3377 {
3378         genl_unregister_family(&ip_vs_genl_family);
3379 }
3380
3381 /* End of Generic Netlink interface definitions */
3382
3383
3384 int __init ip_vs_control_init(void)
3385 {
3386         int ret;
3387         int idx;
3388
3389         EnterFunction(2);
3390
3391         /* Initialize ip_vs_svc_table, ip_vs_svc_fwm_table, ip_vs_rtable */
3392         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++)  {
3393                 INIT_LIST_HEAD(&ip_vs_svc_table[idx]);
3394                 INIT_LIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3395         }
3396         for(idx = 0; idx < IP_VS_RTAB_SIZE; idx++)  {
3397                 INIT_LIST_HEAD(&ip_vs_rtable[idx]);
3398         }
3399         smp_wmb();
3400
3401         ret = nf_register_sockopt(&ip_vs_sockopts);
3402         if (ret) {
3403                 pr_err("cannot register sockopt.\n");
3404                 return ret;
3405         }
3406
3407         ret = ip_vs_genl_register();
3408         if (ret) {
3409                 pr_err("cannot register Generic Netlink interface.\n");
3410                 nf_unregister_sockopt(&ip_vs_sockopts);
3411                 return ret;
3412         }
3413
3414         proc_net_fops_create(&init_net, "ip_vs", 0, &ip_vs_info_fops);
3415         proc_net_fops_create(&init_net, "ip_vs_stats",0, &ip_vs_stats_fops);
3416
3417         sysctl_header = register_sysctl_paths(net_vs_ctl_path, vs_vars);
3418
3419         ip_vs_new_estimator(&ip_vs_stats);
3420
3421         /* Hook the defense timer */
3422         schedule_delayed_work(&defense_work, DEFENSE_TIMER_PERIOD);
3423
3424         LeaveFunction(2);
3425         return 0;
3426 }
3427
3428
3429 void ip_vs_control_cleanup(void)
3430 {
3431         EnterFunction(2);
3432         ip_vs_trash_cleanup();
3433         cancel_delayed_work_sync(&defense_work);
3434         cancel_work_sync(&defense_work.work);
3435         ip_vs_kill_estimator(&ip_vs_stats);
3436         unregister_sysctl_table(sysctl_header);
3437         proc_net_remove(&init_net, "ip_vs_stats");
3438         proc_net_remove(&init_net, "ip_vs");
3439         ip_vs_genl_unregister();
3440         nf_unregister_sockopt(&ip_vs_sockopts);
3441         LeaveFunction(2);
3442 }