IPVS: Allow null argument to ip_vs_scheduler_put()
[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                 .nl_u = {
114                         .ip6_u = {
115                                 .daddr = *addr,
116                                 .saddr = { .s6_addr32 = {0, 0, 0, 0} }, } },
117         };
118
119         rt = (struct rt6_info *)ip6_route_output(&init_net, NULL, &fl);
120         if (rt && rt->rt6i_dev && (rt->rt6i_dev->flags & IFF_LOOPBACK))
121                         return 1;
122
123         return 0;
124 }
125 #endif
126 /*
127  *      update_defense_level is called from keventd and from sysctl,
128  *      so it needs to protect itself from softirqs
129  */
130 static void update_defense_level(void)
131 {
132         struct sysinfo i;
133         static int old_secure_tcp = 0;
134         int availmem;
135         int nomem;
136         int to_change = -1;
137
138         /* we only count free and buffered memory (in pages) */
139         si_meminfo(&i);
140         availmem = i.freeram + i.bufferram;
141         /* however in linux 2.5 the i.bufferram is total page cache size,
142            we need adjust it */
143         /* si_swapinfo(&i); */
144         /* availmem = availmem - (i.totalswap - i.freeswap); */
145
146         nomem = (availmem < sysctl_ip_vs_amemthresh);
147
148         local_bh_disable();
149
150         /* drop_entry */
151         spin_lock(&__ip_vs_dropentry_lock);
152         switch (sysctl_ip_vs_drop_entry) {
153         case 0:
154                 atomic_set(&ip_vs_dropentry, 0);
155                 break;
156         case 1:
157                 if (nomem) {
158                         atomic_set(&ip_vs_dropentry, 1);
159                         sysctl_ip_vs_drop_entry = 2;
160                 } else {
161                         atomic_set(&ip_vs_dropentry, 0);
162                 }
163                 break;
164         case 2:
165                 if (nomem) {
166                         atomic_set(&ip_vs_dropentry, 1);
167                 } else {
168                         atomic_set(&ip_vs_dropentry, 0);
169                         sysctl_ip_vs_drop_entry = 1;
170                 };
171                 break;
172         case 3:
173                 atomic_set(&ip_vs_dropentry, 1);
174                 break;
175         }
176         spin_unlock(&__ip_vs_dropentry_lock);
177
178         /* drop_packet */
179         spin_lock(&__ip_vs_droppacket_lock);
180         switch (sysctl_ip_vs_drop_packet) {
181         case 0:
182                 ip_vs_drop_rate = 0;
183                 break;
184         case 1:
185                 if (nomem) {
186                         ip_vs_drop_rate = ip_vs_drop_counter
187                                 = sysctl_ip_vs_amemthresh /
188                                 (sysctl_ip_vs_amemthresh-availmem);
189                         sysctl_ip_vs_drop_packet = 2;
190                 } else {
191                         ip_vs_drop_rate = 0;
192                 }
193                 break;
194         case 2:
195                 if (nomem) {
196                         ip_vs_drop_rate = ip_vs_drop_counter
197                                 = sysctl_ip_vs_amemthresh /
198                                 (sysctl_ip_vs_amemthresh-availmem);
199                 } else {
200                         ip_vs_drop_rate = 0;
201                         sysctl_ip_vs_drop_packet = 1;
202                 }
203                 break;
204         case 3:
205                 ip_vs_drop_rate = sysctl_ip_vs_am_droprate;
206                 break;
207         }
208         spin_unlock(&__ip_vs_droppacket_lock);
209
210         /* secure_tcp */
211         spin_lock(&ip_vs_securetcp_lock);
212         switch (sysctl_ip_vs_secure_tcp) {
213         case 0:
214                 if (old_secure_tcp >= 2)
215                         to_change = 0;
216                 break;
217         case 1:
218                 if (nomem) {
219                         if (old_secure_tcp < 2)
220                                 to_change = 1;
221                         sysctl_ip_vs_secure_tcp = 2;
222                 } else {
223                         if (old_secure_tcp >= 2)
224                                 to_change = 0;
225                 }
226                 break;
227         case 2:
228                 if (nomem) {
229                         if (old_secure_tcp < 2)
230                                 to_change = 1;
231                 } else {
232                         if (old_secure_tcp >= 2)
233                                 to_change = 0;
234                         sysctl_ip_vs_secure_tcp = 1;
235                 }
236                 break;
237         case 3:
238                 if (old_secure_tcp < 2)
239                         to_change = 1;
240                 break;
241         }
242         old_secure_tcp = sysctl_ip_vs_secure_tcp;
243         if (to_change >= 0)
244                 ip_vs_protocol_timeout_change(sysctl_ip_vs_secure_tcp>1);
245         spin_unlock(&ip_vs_securetcp_lock);
246
247         local_bh_enable();
248 }
249
250
251 /*
252  *      Timer for checking the defense
253  */
254 #define DEFENSE_TIMER_PERIOD    1*HZ
255 static void defense_work_handler(struct work_struct *work);
256 static DECLARE_DELAYED_WORK(defense_work, defense_work_handler);
257
258 static void defense_work_handler(struct work_struct *work)
259 {
260         update_defense_level();
261         if (atomic_read(&ip_vs_dropentry))
262                 ip_vs_random_dropentry();
263
264         schedule_delayed_work(&defense_work, DEFENSE_TIMER_PERIOD);
265 }
266
267 int
268 ip_vs_use_count_inc(void)
269 {
270         return try_module_get(THIS_MODULE);
271 }
272
273 void
274 ip_vs_use_count_dec(void)
275 {
276         module_put(THIS_MODULE);
277 }
278
279
280 /*
281  *      Hash table: for virtual service lookups
282  */
283 #define IP_VS_SVC_TAB_BITS 8
284 #define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
285 #define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)
286
287 /* the service table hashed by <protocol, addr, port> */
288 static struct list_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE];
289 /* the service table hashed by fwmark */
290 static struct list_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE];
291
292 /*
293  *      Hash table: for real service lookups
294  */
295 #define IP_VS_RTAB_BITS 4
296 #define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
297 #define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)
298
299 static struct list_head ip_vs_rtable[IP_VS_RTAB_SIZE];
300
301 /*
302  *      Trash for destinations
303  */
304 static LIST_HEAD(ip_vs_dest_trash);
305
306 /*
307  *      FTP & NULL virtual service counters
308  */
309 static atomic_t ip_vs_ftpsvc_counter = ATOMIC_INIT(0);
310 static atomic_t ip_vs_nullsvc_counter = ATOMIC_INIT(0);
311
312
313 /*
314  *      Returns hash value for virtual service
315  */
316 static __inline__ unsigned
317 ip_vs_svc_hashkey(int af, unsigned proto, const union nf_inet_addr *addr,
318                   __be16 port)
319 {
320         register unsigned porth = ntohs(port);
321         __be32 addr_fold = addr->ip;
322
323 #ifdef CONFIG_IP_VS_IPV6
324         if (af == AF_INET6)
325                 addr_fold = addr->ip6[0]^addr->ip6[1]^
326                             addr->ip6[2]^addr->ip6[3];
327 #endif
328
329         return (proto^ntohl(addr_fold)^(porth>>IP_VS_SVC_TAB_BITS)^porth)
330                 & IP_VS_SVC_TAB_MASK;
331 }
332
333 /*
334  *      Returns hash value of fwmark for virtual service lookup
335  */
336 static __inline__ unsigned ip_vs_svc_fwm_hashkey(__u32 fwmark)
337 {
338         return fwmark & IP_VS_SVC_TAB_MASK;
339 }
340
341 /*
342  *      Hashes a service in the ip_vs_svc_table by <proto,addr,port>
343  *      or in the ip_vs_svc_fwm_table by fwmark.
344  *      Should be called with locked tables.
345  */
346 static int ip_vs_svc_hash(struct ip_vs_service *svc)
347 {
348         unsigned hash;
349
350         if (svc->flags & IP_VS_SVC_F_HASHED) {
351                 pr_err("%s(): request for already hashed, called from %pF\n",
352                        __func__, __builtin_return_address(0));
353                 return 0;
354         }
355
356         if (svc->fwmark == 0) {
357                 /*
358                  *  Hash it by <protocol,addr,port> in ip_vs_svc_table
359                  */
360                 hash = ip_vs_svc_hashkey(svc->af, svc->protocol, &svc->addr,
361                                          svc->port);
362                 list_add(&svc->s_list, &ip_vs_svc_table[hash]);
363         } else {
364                 /*
365                  *  Hash it by fwmark in ip_vs_svc_fwm_table
366                  */
367                 hash = ip_vs_svc_fwm_hashkey(svc->fwmark);
368                 list_add(&svc->f_list, &ip_vs_svc_fwm_table[hash]);
369         }
370
371         svc->flags |= IP_VS_SVC_F_HASHED;
372         /* increase its refcnt because it is referenced by the svc table */
373         atomic_inc(&svc->refcnt);
374         return 1;
375 }
376
377
378 /*
379  *      Unhashes a service from ip_vs_svc_table/ip_vs_svc_fwm_table.
380  *      Should be called with locked tables.
381  */
382 static int ip_vs_svc_unhash(struct ip_vs_service *svc)
383 {
384         if (!(svc->flags & IP_VS_SVC_F_HASHED)) {
385                 pr_err("%s(): request for unhash flagged, called from %pF\n",
386                        __func__, __builtin_return_address(0));
387                 return 0;
388         }
389
390         if (svc->fwmark == 0) {
391                 /* Remove it from the ip_vs_svc_table table */
392                 list_del(&svc->s_list);
393         } else {
394                 /* Remove it from the ip_vs_svc_fwm_table table */
395                 list_del(&svc->f_list);
396         }
397
398         svc->flags &= ~IP_VS_SVC_F_HASHED;
399         atomic_dec(&svc->refcnt);
400         return 1;
401 }
402
403
404 /*
405  *      Get service by {proto,addr,port} in the service table.
406  */
407 static inline struct ip_vs_service *
408 __ip_vs_service_find(int af, __u16 protocol, const union nf_inet_addr *vaddr,
409                     __be16 vport)
410 {
411         unsigned hash;
412         struct ip_vs_service *svc;
413
414         /* Check for "full" addressed entries */
415         hash = ip_vs_svc_hashkey(af, protocol, vaddr, vport);
416
417         list_for_each_entry(svc, &ip_vs_svc_table[hash], s_list){
418                 if ((svc->af == af)
419                     && ip_vs_addr_equal(af, &svc->addr, vaddr)
420                     && (svc->port == vport)
421                     && (svc->protocol == protocol)) {
422                         /* HIT */
423                         return svc;
424                 }
425         }
426
427         return NULL;
428 }
429
430
431 /*
432  *      Get service by {fwmark} in the service table.
433  */
434 static inline struct ip_vs_service *
435 __ip_vs_svc_fwm_find(int af, __u32 fwmark)
436 {
437         unsigned hash;
438         struct ip_vs_service *svc;
439
440         /* Check for fwmark addressed entries */
441         hash = ip_vs_svc_fwm_hashkey(fwmark);
442
443         list_for_each_entry(svc, &ip_vs_svc_fwm_table[hash], f_list) {
444                 if (svc->fwmark == fwmark && svc->af == af) {
445                         /* HIT */
446                         return svc;
447                 }
448         }
449
450         return NULL;
451 }
452
453 struct ip_vs_service *
454 ip_vs_service_get(int af, __u32 fwmark, __u16 protocol,
455                   const union nf_inet_addr *vaddr, __be16 vport)
456 {
457         struct ip_vs_service *svc;
458
459         read_lock(&__ip_vs_svc_lock);
460
461         /*
462          *      Check the table hashed by fwmark first
463          */
464         if (fwmark && (svc = __ip_vs_svc_fwm_find(af, fwmark)))
465                 goto out;
466
467         /*
468          *      Check the table hashed by <protocol,addr,port>
469          *      for "full" addressed entries
470          */
471         svc = __ip_vs_service_find(af, protocol, vaddr, vport);
472
473         if (svc == NULL
474             && protocol == IPPROTO_TCP
475             && atomic_read(&ip_vs_ftpsvc_counter)
476             && (vport == FTPDATA || ntohs(vport) >= PROT_SOCK)) {
477                 /*
478                  * Check if ftp service entry exists, the packet
479                  * might belong to FTP data connections.
480                  */
481                 svc = __ip_vs_service_find(af, protocol, vaddr, FTPPORT);
482         }
483
484         if (svc == NULL
485             && atomic_read(&ip_vs_nullsvc_counter)) {
486                 /*
487                  * Check if the catch-all port (port zero) exists
488                  */
489                 svc = __ip_vs_service_find(af, protocol, vaddr, 0);
490         }
491
492   out:
493         if (svc)
494                 atomic_inc(&svc->usecnt);
495         read_unlock(&__ip_vs_svc_lock);
496
497         IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
498                       fwmark, ip_vs_proto_name(protocol),
499                       IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
500                       svc ? "hit" : "not hit");
501
502         return svc;
503 }
504
505
506 static inline void
507 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
508 {
509         atomic_inc(&svc->refcnt);
510         dest->svc = svc;
511 }
512
513 static void
514 __ip_vs_unbind_svc(struct ip_vs_dest *dest)
515 {
516         struct ip_vs_service *svc = dest->svc;
517
518         dest->svc = NULL;
519         if (atomic_dec_and_test(&svc->refcnt)) {
520                 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
521                               svc->fwmark,
522                               IP_VS_DBG_ADDR(svc->af, &svc->addr),
523                               ntohs(svc->port), atomic_read(&svc->usecnt));
524                 kfree(svc);
525         }
526 }
527
528
529 /*
530  *      Returns hash value for real service
531  */
532 static inline unsigned ip_vs_rs_hashkey(int af,
533                                             const union nf_inet_addr *addr,
534                                             __be16 port)
535 {
536         register unsigned porth = ntohs(port);
537         __be32 addr_fold = addr->ip;
538
539 #ifdef CONFIG_IP_VS_IPV6
540         if (af == AF_INET6)
541                 addr_fold = addr->ip6[0]^addr->ip6[1]^
542                             addr->ip6[2]^addr->ip6[3];
543 #endif
544
545         return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
546                 & IP_VS_RTAB_MASK;
547 }
548
549 /*
550  *      Hashes ip_vs_dest in ip_vs_rtable by <proto,addr,port>.
551  *      should be called with locked tables.
552  */
553 static int ip_vs_rs_hash(struct ip_vs_dest *dest)
554 {
555         unsigned hash;
556
557         if (!list_empty(&dest->d_list)) {
558                 return 0;
559         }
560
561         /*
562          *      Hash by proto,addr,port,
563          *      which are the parameters of the real service.
564          */
565         hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
566
567         list_add(&dest->d_list, &ip_vs_rtable[hash]);
568
569         return 1;
570 }
571
572 /*
573  *      UNhashes ip_vs_dest from ip_vs_rtable.
574  *      should be called with locked tables.
575  */
576 static int ip_vs_rs_unhash(struct ip_vs_dest *dest)
577 {
578         /*
579          * Remove it from the ip_vs_rtable table.
580          */
581         if (!list_empty(&dest->d_list)) {
582                 list_del(&dest->d_list);
583                 INIT_LIST_HEAD(&dest->d_list);
584         }
585
586         return 1;
587 }
588
589 /*
590  *      Lookup real service by <proto,addr,port> in the real service table.
591  */
592 struct ip_vs_dest *
593 ip_vs_lookup_real_service(int af, __u16 protocol,
594                           const union nf_inet_addr *daddr,
595                           __be16 dport)
596 {
597         unsigned hash;
598         struct ip_vs_dest *dest;
599
600         /*
601          *      Check for "full" addressed entries
602          *      Return the first found entry
603          */
604         hash = ip_vs_rs_hashkey(af, daddr, dport);
605
606         read_lock(&__ip_vs_rs_lock);
607         list_for_each_entry(dest, &ip_vs_rtable[hash], d_list) {
608                 if ((dest->af == af)
609                     && ip_vs_addr_equal(af, &dest->addr, daddr)
610                     && (dest->port == dport)
611                     && ((dest->protocol == protocol) ||
612                         dest->vfwmark)) {
613                         /* HIT */
614                         read_unlock(&__ip_vs_rs_lock);
615                         return dest;
616                 }
617         }
618         read_unlock(&__ip_vs_rs_lock);
619
620         return NULL;
621 }
622
623 /*
624  *      Lookup destination by {addr,port} in the given service
625  */
626 static struct ip_vs_dest *
627 ip_vs_lookup_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
628                   __be16 dport)
629 {
630         struct ip_vs_dest *dest;
631
632         /*
633          * Find the destination for the given service
634          */
635         list_for_each_entry(dest, &svc->destinations, n_list) {
636                 if ((dest->af == svc->af)
637                     && ip_vs_addr_equal(svc->af, &dest->addr, daddr)
638                     && (dest->port == dport)) {
639                         /* HIT */
640                         return dest;
641                 }
642         }
643
644         return NULL;
645 }
646
647 /*
648  * Find destination by {daddr,dport,vaddr,protocol}
649  * Cretaed to be used in ip_vs_process_message() in
650  * the backup synchronization daemon. It finds the
651  * destination to be bound to the received connection
652  * on the backup.
653  *
654  * ip_vs_lookup_real_service() looked promissing, but
655  * seems not working as expected.
656  */
657 struct ip_vs_dest *ip_vs_find_dest(int af, const union nf_inet_addr *daddr,
658                                    __be16 dport,
659                                    const union nf_inet_addr *vaddr,
660                                    __be16 vport, __u16 protocol)
661 {
662         struct ip_vs_dest *dest;
663         struct ip_vs_service *svc;
664
665         svc = ip_vs_service_get(af, 0, protocol, vaddr, vport);
666         if (!svc)
667                 return NULL;
668         dest = ip_vs_lookup_dest(svc, daddr, dport);
669         if (dest)
670                 atomic_inc(&dest->refcnt);
671         ip_vs_service_put(svc);
672         return dest;
673 }
674
675 /*
676  *  Lookup dest by {svc,addr,port} in the destination trash.
677  *  The destination trash is used to hold the destinations that are removed
678  *  from the service table but are still referenced by some conn entries.
679  *  The reason to add the destination trash is when the dest is temporary
680  *  down (either by administrator or by monitor program), the dest can be
681  *  picked back from the trash, the remaining connections to the dest can
682  *  continue, and the counting information of the dest is also useful for
683  *  scheduling.
684  */
685 static struct ip_vs_dest *
686 ip_vs_trash_get_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
687                      __be16 dport)
688 {
689         struct ip_vs_dest *dest, *nxt;
690
691         /*
692          * Find the destination in trash
693          */
694         list_for_each_entry_safe(dest, nxt, &ip_vs_dest_trash, n_list) {
695                 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
696                               "dest->refcnt=%d\n",
697                               dest->vfwmark,
698                               IP_VS_DBG_ADDR(svc->af, &dest->addr),
699                               ntohs(dest->port),
700                               atomic_read(&dest->refcnt));
701                 if (dest->af == svc->af &&
702                     ip_vs_addr_equal(svc->af, &dest->addr, daddr) &&
703                     dest->port == dport &&
704                     dest->vfwmark == svc->fwmark &&
705                     dest->protocol == svc->protocol &&
706                     (svc->fwmark ||
707                      (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
708                       dest->vport == svc->port))) {
709                         /* HIT */
710                         return dest;
711                 }
712
713                 /*
714                  * Try to purge the destination from trash if not referenced
715                  */
716                 if (atomic_read(&dest->refcnt) == 1) {
717                         IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u "
718                                       "from trash\n",
719                                       dest->vfwmark,
720                                       IP_VS_DBG_ADDR(svc->af, &dest->addr),
721                                       ntohs(dest->port));
722                         list_del(&dest->n_list);
723                         ip_vs_dst_reset(dest);
724                         __ip_vs_unbind_svc(dest);
725                         kfree(dest);
726                 }
727         }
728
729         return NULL;
730 }
731
732
733 /*
734  *  Clean up all the destinations in the trash
735  *  Called by the ip_vs_control_cleanup()
736  *
737  *  When the ip_vs_control_clearup is activated by ipvs module exit,
738  *  the service tables must have been flushed and all the connections
739  *  are expired, and the refcnt of each destination in the trash must
740  *  be 1, so we simply release them here.
741  */
742 static void ip_vs_trash_cleanup(void)
743 {
744         struct ip_vs_dest *dest, *nxt;
745
746         list_for_each_entry_safe(dest, nxt, &ip_vs_dest_trash, n_list) {
747                 list_del(&dest->n_list);
748                 ip_vs_dst_reset(dest);
749                 __ip_vs_unbind_svc(dest);
750                 kfree(dest);
751         }
752 }
753
754
755 static void
756 ip_vs_zero_stats(struct ip_vs_stats *stats)
757 {
758         spin_lock_bh(&stats->lock);
759
760         memset(&stats->ustats, 0, sizeof(stats->ustats));
761         ip_vs_zero_estimator(stats);
762
763         spin_unlock_bh(&stats->lock);
764 }
765
766 /*
767  *      Update a destination in the given service
768  */
769 static void
770 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
771                     struct ip_vs_dest_user_kern *udest, int add)
772 {
773         int conn_flags;
774
775         /* set the weight and the flags */
776         atomic_set(&dest->weight, udest->weight);
777         conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
778         conn_flags |= IP_VS_CONN_F_INACTIVE;
779
780         /* check if local node and update the flags */
781 #ifdef CONFIG_IP_VS_IPV6
782         if (svc->af == AF_INET6) {
783                 if (__ip_vs_addr_is_local_v6(&udest->addr.in6)) {
784                         conn_flags = (conn_flags & ~IP_VS_CONN_F_FWD_MASK)
785                                 | IP_VS_CONN_F_LOCALNODE;
786                 }
787         } else
788 #endif
789                 if (inet_addr_type(&init_net, udest->addr.ip) == RTN_LOCAL) {
790                         conn_flags = (conn_flags & ~IP_VS_CONN_F_FWD_MASK)
791                                 | IP_VS_CONN_F_LOCALNODE;
792                 }
793
794         /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
795         if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
796                 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
797         } else {
798                 /*
799                  *    Put the real service in ip_vs_rtable if not present.
800                  *    For now only for NAT!
801                  */
802                 write_lock_bh(&__ip_vs_rs_lock);
803                 ip_vs_rs_hash(dest);
804                 write_unlock_bh(&__ip_vs_rs_lock);
805         }
806         atomic_set(&dest->conn_flags, conn_flags);
807
808         /* bind the service */
809         if (!dest->svc) {
810                 __ip_vs_bind_svc(dest, svc);
811         } else {
812                 if (dest->svc != svc) {
813                         __ip_vs_unbind_svc(dest);
814                         ip_vs_zero_stats(&dest->stats);
815                         __ip_vs_bind_svc(dest, svc);
816                 }
817         }
818
819         /* set the dest status flags */
820         dest->flags |= IP_VS_DEST_F_AVAILABLE;
821
822         if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
823                 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
824         dest->u_threshold = udest->u_threshold;
825         dest->l_threshold = udest->l_threshold;
826
827         if (add)
828                 ip_vs_new_estimator(&dest->stats);
829
830         write_lock_bh(&__ip_vs_svc_lock);
831
832         /* Wait until all other svc users go away */
833         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
834
835         if (add) {
836                 list_add(&dest->n_list, &svc->destinations);
837                 svc->num_dests++;
838         }
839
840         /* call the update_service, because server weight may be changed */
841         if (svc->scheduler->update_service)
842                 svc->scheduler->update_service(svc);
843
844         write_unlock_bh(&__ip_vs_svc_lock);
845 }
846
847
848 /*
849  *      Create a destination for the given service
850  */
851 static int
852 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
853                struct ip_vs_dest **dest_p)
854 {
855         struct ip_vs_dest *dest;
856         unsigned atype;
857
858         EnterFunction(2);
859
860 #ifdef CONFIG_IP_VS_IPV6
861         if (svc->af == AF_INET6) {
862                 atype = ipv6_addr_type(&udest->addr.in6);
863                 if ((!(atype & IPV6_ADDR_UNICAST) ||
864                         atype & IPV6_ADDR_LINKLOCAL) &&
865                         !__ip_vs_addr_is_local_v6(&udest->addr.in6))
866                         return -EINVAL;
867         } else
868 #endif
869         {
870                 atype = inet_addr_type(&init_net, udest->addr.ip);
871                 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
872                         return -EINVAL;
873         }
874
875         dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
876         if (dest == NULL) {
877                 pr_err("%s(): no memory.\n", __func__);
878                 return -ENOMEM;
879         }
880
881         dest->af = svc->af;
882         dest->protocol = svc->protocol;
883         dest->vaddr = svc->addr;
884         dest->vport = svc->port;
885         dest->vfwmark = svc->fwmark;
886         ip_vs_addr_copy(svc->af, &dest->addr, &udest->addr);
887         dest->port = udest->port;
888
889         atomic_set(&dest->activeconns, 0);
890         atomic_set(&dest->inactconns, 0);
891         atomic_set(&dest->persistconns, 0);
892         atomic_set(&dest->refcnt, 1);
893
894         INIT_LIST_HEAD(&dest->d_list);
895         spin_lock_init(&dest->dst_lock);
896         spin_lock_init(&dest->stats.lock);
897         __ip_vs_update_dest(svc, dest, udest, 1);
898
899         *dest_p = dest;
900
901         LeaveFunction(2);
902         return 0;
903 }
904
905
906 /*
907  *      Add a destination into an existing service
908  */
909 static int
910 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
911 {
912         struct ip_vs_dest *dest;
913         union nf_inet_addr daddr;
914         __be16 dport = udest->port;
915         int ret;
916
917         EnterFunction(2);
918
919         if (udest->weight < 0) {
920                 pr_err("%s(): server weight less than zero\n", __func__);
921                 return -ERANGE;
922         }
923
924         if (udest->l_threshold > udest->u_threshold) {
925                 pr_err("%s(): lower threshold is higher than upper threshold\n",
926                         __func__);
927                 return -ERANGE;
928         }
929
930         ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
931
932         /*
933          * Check if the dest already exists in the list
934          */
935         dest = ip_vs_lookup_dest(svc, &daddr, dport);
936
937         if (dest != NULL) {
938                 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
939                 return -EEXIST;
940         }
941
942         /*
943          * Check if the dest already exists in the trash and
944          * is from the same service
945          */
946         dest = ip_vs_trash_get_dest(svc, &daddr, dport);
947
948         if (dest != NULL) {
949                 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
950                               "dest->refcnt=%d, service %u/%s:%u\n",
951                               IP_VS_DBG_ADDR(svc->af, &daddr), ntohs(dport),
952                               atomic_read(&dest->refcnt),
953                               dest->vfwmark,
954                               IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
955                               ntohs(dest->vport));
956
957                 /*
958                  * Get the destination from the trash
959                  */
960                 list_del(&dest->n_list);
961
962                 __ip_vs_update_dest(svc, dest, udest, 1);
963                 ret = 0;
964         } else {
965                 /*
966                  * Allocate and initialize the dest structure
967                  */
968                 ret = ip_vs_new_dest(svc, udest, &dest);
969         }
970         LeaveFunction(2);
971
972         return ret;
973 }
974
975
976 /*
977  *      Edit a destination in the given service
978  */
979 static int
980 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
981 {
982         struct ip_vs_dest *dest;
983         union nf_inet_addr daddr;
984         __be16 dport = udest->port;
985
986         EnterFunction(2);
987
988         if (udest->weight < 0) {
989                 pr_err("%s(): server weight less than zero\n", __func__);
990                 return -ERANGE;
991         }
992
993         if (udest->l_threshold > udest->u_threshold) {
994                 pr_err("%s(): lower threshold is higher than upper threshold\n",
995                         __func__);
996                 return -ERANGE;
997         }
998
999         ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
1000
1001         /*
1002          *  Lookup the destination list
1003          */
1004         dest = ip_vs_lookup_dest(svc, &daddr, dport);
1005
1006         if (dest == NULL) {
1007                 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1008                 return -ENOENT;
1009         }
1010
1011         __ip_vs_update_dest(svc, dest, udest, 0);
1012         LeaveFunction(2);
1013
1014         return 0;
1015 }
1016
1017
1018 /*
1019  *      Delete a destination (must be already unlinked from the service)
1020  */
1021 static void __ip_vs_del_dest(struct ip_vs_dest *dest)
1022 {
1023         ip_vs_kill_estimator(&dest->stats);
1024
1025         /*
1026          *  Remove it from the d-linked list with the real services.
1027          */
1028         write_lock_bh(&__ip_vs_rs_lock);
1029         ip_vs_rs_unhash(dest);
1030         write_unlock_bh(&__ip_vs_rs_lock);
1031
1032         /*
1033          *  Decrease the refcnt of the dest, and free the dest
1034          *  if nobody refers to it (refcnt=0). Otherwise, throw
1035          *  the destination into the trash.
1036          */
1037         if (atomic_dec_and_test(&dest->refcnt)) {
1038                 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u\n",
1039                               dest->vfwmark,
1040                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1041                               ntohs(dest->port));
1042                 ip_vs_dst_reset(dest);
1043                 /* simply decrease svc->refcnt here, let the caller check
1044                    and release the service if nobody refers to it.
1045                    Only user context can release destination and service,
1046                    and only one user context can update virtual service at a
1047                    time, so the operation here is OK */
1048                 atomic_dec(&dest->svc->refcnt);
1049                 kfree(dest);
1050         } else {
1051                 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, "
1052                               "dest->refcnt=%d\n",
1053                               IP_VS_DBG_ADDR(dest->af, &dest->addr),
1054                               ntohs(dest->port),
1055                               atomic_read(&dest->refcnt));
1056                 list_add(&dest->n_list, &ip_vs_dest_trash);
1057                 atomic_inc(&dest->refcnt);
1058         }
1059 }
1060
1061
1062 /*
1063  *      Unlink a destination from the given service
1064  */
1065 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1066                                 struct ip_vs_dest *dest,
1067                                 int svcupd)
1068 {
1069         dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1070
1071         /*
1072          *  Remove it from the d-linked destination list.
1073          */
1074         list_del(&dest->n_list);
1075         svc->num_dests--;
1076
1077         /*
1078          *  Call the update_service function of its scheduler
1079          */
1080         if (svcupd && svc->scheduler->update_service)
1081                         svc->scheduler->update_service(svc);
1082 }
1083
1084
1085 /*
1086  *      Delete a destination server in the given service
1087  */
1088 static int
1089 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1090 {
1091         struct ip_vs_dest *dest;
1092         __be16 dport = udest->port;
1093
1094         EnterFunction(2);
1095
1096         dest = ip_vs_lookup_dest(svc, &udest->addr, dport);
1097
1098         if (dest == NULL) {
1099                 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1100                 return -ENOENT;
1101         }
1102
1103         write_lock_bh(&__ip_vs_svc_lock);
1104
1105         /*
1106          *      Wait until all other svc users go away.
1107          */
1108         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1109
1110         /*
1111          *      Unlink dest from the service
1112          */
1113         __ip_vs_unlink_dest(svc, dest, 1);
1114
1115         write_unlock_bh(&__ip_vs_svc_lock);
1116
1117         /*
1118          *      Delete the destination
1119          */
1120         __ip_vs_del_dest(dest);
1121
1122         LeaveFunction(2);
1123
1124         return 0;
1125 }
1126
1127
1128 /*
1129  *      Add a service into the service hash table
1130  */
1131 static int
1132 ip_vs_add_service(struct ip_vs_service_user_kern *u,
1133                   struct ip_vs_service **svc_p)
1134 {
1135         int ret = 0;
1136         struct ip_vs_scheduler *sched = NULL;
1137         struct ip_vs_service *svc = NULL;
1138
1139         /* increase the module use count */
1140         ip_vs_use_count_inc();
1141
1142         /* Lookup the scheduler by 'u->sched_name' */
1143         sched = ip_vs_scheduler_get(u->sched_name);
1144         if (sched == NULL) {
1145                 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1146                 ret = -ENOENT;
1147                 goto out_err;
1148         }
1149
1150 #ifdef CONFIG_IP_VS_IPV6
1151         if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1152                 ret = -EINVAL;
1153                 goto out_err;
1154         }
1155 #endif
1156
1157         svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1158         if (svc == NULL) {
1159                 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1160                 ret = -ENOMEM;
1161                 goto out_err;
1162         }
1163
1164         /* I'm the first user of the service */
1165         atomic_set(&svc->usecnt, 0);
1166         atomic_set(&svc->refcnt, 0);
1167
1168         svc->af = u->af;
1169         svc->protocol = u->protocol;
1170         ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1171         svc->port = u->port;
1172         svc->fwmark = u->fwmark;
1173         svc->flags = u->flags;
1174         svc->timeout = u->timeout * HZ;
1175         svc->netmask = u->netmask;
1176
1177         INIT_LIST_HEAD(&svc->destinations);
1178         rwlock_init(&svc->sched_lock);
1179         spin_lock_init(&svc->stats.lock);
1180
1181         /* Bind the scheduler */
1182         ret = ip_vs_bind_scheduler(svc, sched);
1183         if (ret)
1184                 goto out_err;
1185         sched = NULL;
1186
1187         /* Update the virtual service counters */
1188         if (svc->port == FTPPORT)
1189                 atomic_inc(&ip_vs_ftpsvc_counter);
1190         else if (svc->port == 0)
1191                 atomic_inc(&ip_vs_nullsvc_counter);
1192
1193         ip_vs_new_estimator(&svc->stats);
1194
1195         /* Count only IPv4 services for old get/setsockopt interface */
1196         if (svc->af == AF_INET)
1197                 ip_vs_num_services++;
1198
1199         /* Hash the service into the service table */
1200         write_lock_bh(&__ip_vs_svc_lock);
1201         ip_vs_svc_hash(svc);
1202         write_unlock_bh(&__ip_vs_svc_lock);
1203
1204         *svc_p = svc;
1205         return 0;
1206
1207  out_err:
1208         if (svc != NULL) {
1209                 if (svc->scheduler)
1210                         ip_vs_unbind_scheduler(svc);
1211                 if (svc->inc) {
1212                         local_bh_disable();
1213                         ip_vs_app_inc_put(svc->inc);
1214                         local_bh_enable();
1215                 }
1216                 kfree(svc);
1217         }
1218         ip_vs_scheduler_put(sched);
1219
1220         /* decrease the module use count */
1221         ip_vs_use_count_dec();
1222
1223         return ret;
1224 }
1225
1226
1227 /*
1228  *      Edit a service and bind it with a new scheduler
1229  */
1230 static int
1231 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1232 {
1233         struct ip_vs_scheduler *sched, *old_sched;
1234         int ret = 0;
1235
1236         /*
1237          * Lookup the scheduler, by 'u->sched_name'
1238          */
1239         sched = ip_vs_scheduler_get(u->sched_name);
1240         if (sched == NULL) {
1241                 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1242                 return -ENOENT;
1243         }
1244         old_sched = sched;
1245
1246 #ifdef CONFIG_IP_VS_IPV6
1247         if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1248                 ret = -EINVAL;
1249                 goto out;
1250         }
1251 #endif
1252
1253         write_lock_bh(&__ip_vs_svc_lock);
1254
1255         /*
1256          * Wait until all other svc users go away.
1257          */
1258         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1259
1260         /*
1261          * Set the flags and timeout value
1262          */
1263         svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1264         svc->timeout = u->timeout * HZ;
1265         svc->netmask = u->netmask;
1266
1267         old_sched = svc->scheduler;
1268         if (sched != old_sched) {
1269                 /*
1270                  * Unbind the old scheduler
1271                  */
1272                 if ((ret = ip_vs_unbind_scheduler(svc))) {
1273                         old_sched = sched;
1274                         goto out_unlock;
1275                 }
1276
1277                 /*
1278                  * Bind the new scheduler
1279                  */
1280                 if ((ret = ip_vs_bind_scheduler(svc, sched))) {
1281                         /*
1282                          * If ip_vs_bind_scheduler fails, restore the old
1283                          * scheduler.
1284                          * The main reason of failure is out of memory.
1285                          *
1286                          * The question is if the old scheduler can be
1287                          * restored all the time. TODO: if it cannot be
1288                          * restored some time, we must delete the service,
1289                          * otherwise the system may crash.
1290                          */
1291                         ip_vs_bind_scheduler(svc, old_sched);
1292                         old_sched = sched;
1293                         goto out_unlock;
1294                 }
1295         }
1296
1297   out_unlock:
1298         write_unlock_bh(&__ip_vs_svc_lock);
1299 #ifdef CONFIG_IP_VS_IPV6
1300   out:
1301 #endif
1302         ip_vs_scheduler_put(old_sched);
1303         return ret;
1304 }
1305
1306
1307 /*
1308  *      Delete a service from the service list
1309  *      - The service must be unlinked, unlocked and not referenced!
1310  *      - We are called under _bh lock
1311  */
1312 static void __ip_vs_del_service(struct ip_vs_service *svc)
1313 {
1314         struct ip_vs_dest *dest, *nxt;
1315         struct ip_vs_scheduler *old_sched;
1316
1317         /* Count only IPv4 services for old get/setsockopt interface */
1318         if (svc->af == AF_INET)
1319                 ip_vs_num_services--;
1320
1321         ip_vs_kill_estimator(&svc->stats);
1322
1323         /* Unbind scheduler */
1324         old_sched = svc->scheduler;
1325         ip_vs_unbind_scheduler(svc);
1326         ip_vs_scheduler_put(old_sched);
1327
1328         /* Unbind app inc */
1329         if (svc->inc) {
1330                 ip_vs_app_inc_put(svc->inc);
1331                 svc->inc = NULL;
1332         }
1333
1334         /*
1335          *    Unlink the whole destination list
1336          */
1337         list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1338                 __ip_vs_unlink_dest(svc, dest, 0);
1339                 __ip_vs_del_dest(dest);
1340         }
1341
1342         /*
1343          *    Update the virtual service counters
1344          */
1345         if (svc->port == FTPPORT)
1346                 atomic_dec(&ip_vs_ftpsvc_counter);
1347         else if (svc->port == 0)
1348                 atomic_dec(&ip_vs_nullsvc_counter);
1349
1350         /*
1351          *    Free the service if nobody refers to it
1352          */
1353         if (atomic_read(&svc->refcnt) == 0) {
1354                 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
1355                               svc->fwmark,
1356                               IP_VS_DBG_ADDR(svc->af, &svc->addr),
1357                               ntohs(svc->port), atomic_read(&svc->usecnt));
1358                 kfree(svc);
1359         }
1360
1361         /* decrease the module use count */
1362         ip_vs_use_count_dec();
1363 }
1364
1365 /*
1366  * Unlink a service from list and try to delete it if its refcnt reached 0
1367  */
1368 static void ip_vs_unlink_service(struct ip_vs_service *svc)
1369 {
1370         /*
1371          * Unhash it from the service table
1372          */
1373         write_lock_bh(&__ip_vs_svc_lock);
1374
1375         ip_vs_svc_unhash(svc);
1376
1377         /*
1378          * Wait until all the svc users go away.
1379          */
1380         IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1381
1382         __ip_vs_del_service(svc);
1383
1384         write_unlock_bh(&__ip_vs_svc_lock);
1385 }
1386
1387 /*
1388  *      Delete a service from the service list
1389  */
1390 static int ip_vs_del_service(struct ip_vs_service *svc)
1391 {
1392         if (svc == NULL)
1393                 return -EEXIST;
1394         ip_vs_unlink_service(svc);
1395
1396         return 0;
1397 }
1398
1399
1400 /*
1401  *      Flush all the virtual services
1402  */
1403 static int ip_vs_flush(void)
1404 {
1405         int idx;
1406         struct ip_vs_service *svc, *nxt;
1407
1408         /*
1409          * Flush the service table hashed by <protocol,addr,port>
1410          */
1411         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1412                 list_for_each_entry_safe(svc, nxt, &ip_vs_svc_table[idx], s_list) {
1413                         ip_vs_unlink_service(svc);
1414                 }
1415         }
1416
1417         /*
1418          * Flush the service table hashed by fwmark
1419          */
1420         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1421                 list_for_each_entry_safe(svc, nxt,
1422                                          &ip_vs_svc_fwm_table[idx], f_list) {
1423                         ip_vs_unlink_service(svc);
1424                 }
1425         }
1426
1427         return 0;
1428 }
1429
1430
1431 /*
1432  *      Zero counters in a service or all services
1433  */
1434 static int ip_vs_zero_service(struct ip_vs_service *svc)
1435 {
1436         struct ip_vs_dest *dest;
1437
1438         write_lock_bh(&__ip_vs_svc_lock);
1439         list_for_each_entry(dest, &svc->destinations, n_list) {
1440                 ip_vs_zero_stats(&dest->stats);
1441         }
1442         ip_vs_zero_stats(&svc->stats);
1443         write_unlock_bh(&__ip_vs_svc_lock);
1444         return 0;
1445 }
1446
1447 static int ip_vs_zero_all(void)
1448 {
1449         int idx;
1450         struct ip_vs_service *svc;
1451
1452         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1453                 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1454                         ip_vs_zero_service(svc);
1455                 }
1456         }
1457
1458         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1459                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1460                         ip_vs_zero_service(svc);
1461                 }
1462         }
1463
1464         ip_vs_zero_stats(&ip_vs_stats);
1465         return 0;
1466 }
1467
1468
1469 static int
1470 proc_do_defense_mode(ctl_table *table, int write,
1471                      void __user *buffer, size_t *lenp, loff_t *ppos)
1472 {
1473         int *valp = table->data;
1474         int val = *valp;
1475         int rc;
1476
1477         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1478         if (write && (*valp != val)) {
1479                 if ((*valp < 0) || (*valp > 3)) {
1480                         /* Restore the correct value */
1481                         *valp = val;
1482                 } else {
1483                         update_defense_level();
1484                 }
1485         }
1486         return rc;
1487 }
1488
1489
1490 static int
1491 proc_do_sync_threshold(ctl_table *table, int write,
1492                        void __user *buffer, size_t *lenp, loff_t *ppos)
1493 {
1494         int *valp = table->data;
1495         int val[2];
1496         int rc;
1497
1498         /* backup the value first */
1499         memcpy(val, valp, sizeof(val));
1500
1501         rc = proc_dointvec(table, write, buffer, lenp, ppos);
1502         if (write && (valp[0] < 0 || valp[1] < 0 || valp[0] >= valp[1])) {
1503                 /* Restore the correct value */
1504                 memcpy(valp, val, sizeof(val));
1505         }
1506         return rc;
1507 }
1508
1509
1510 /*
1511  *      IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1512  */
1513
1514 static struct ctl_table vs_vars[] = {
1515         {
1516                 .procname       = "amemthresh",
1517                 .data           = &sysctl_ip_vs_amemthresh,
1518                 .maxlen         = sizeof(int),
1519                 .mode           = 0644,
1520                 .proc_handler   = proc_dointvec,
1521         },
1522 #ifdef CONFIG_IP_VS_DEBUG
1523         {
1524                 .procname       = "debug_level",
1525                 .data           = &sysctl_ip_vs_debug_level,
1526                 .maxlen         = sizeof(int),
1527                 .mode           = 0644,
1528                 .proc_handler   = proc_dointvec,
1529         },
1530 #endif
1531         {
1532                 .procname       = "am_droprate",
1533                 .data           = &sysctl_ip_vs_am_droprate,
1534                 .maxlen         = sizeof(int),
1535                 .mode           = 0644,
1536                 .proc_handler   = proc_dointvec,
1537         },
1538         {
1539                 .procname       = "drop_entry",
1540                 .data           = &sysctl_ip_vs_drop_entry,
1541                 .maxlen         = sizeof(int),
1542                 .mode           = 0644,
1543                 .proc_handler   = proc_do_defense_mode,
1544         },
1545         {
1546                 .procname       = "drop_packet",
1547                 .data           = &sysctl_ip_vs_drop_packet,
1548                 .maxlen         = sizeof(int),
1549                 .mode           = 0644,
1550                 .proc_handler   = proc_do_defense_mode,
1551         },
1552 #ifdef CONFIG_IP_VS_NFCT
1553         {
1554                 .procname       = "conntrack",
1555                 .data           = &sysctl_ip_vs_conntrack,
1556                 .maxlen         = sizeof(int),
1557                 .mode           = 0644,
1558                 .proc_handler   = &proc_dointvec,
1559         },
1560 #endif
1561         {
1562                 .procname       = "secure_tcp",
1563                 .data           = &sysctl_ip_vs_secure_tcp,
1564                 .maxlen         = sizeof(int),
1565                 .mode           = 0644,
1566                 .proc_handler   = proc_do_defense_mode,
1567         },
1568         {
1569                 .procname       = "snat_reroute",
1570                 .data           = &sysctl_ip_vs_snat_reroute,
1571                 .maxlen         = sizeof(int),
1572                 .mode           = 0644,
1573                 .proc_handler   = &proc_dointvec,
1574         },
1575 #if 0
1576         {
1577                 .procname       = "timeout_established",
1578                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1579                 .maxlen         = sizeof(int),
1580                 .mode           = 0644,
1581                 .proc_handler   = proc_dointvec_jiffies,
1582         },
1583         {
1584                 .procname       = "timeout_synsent",
1585                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1586                 .maxlen         = sizeof(int),
1587                 .mode           = 0644,
1588                 .proc_handler   = proc_dointvec_jiffies,
1589         },
1590         {
1591                 .procname       = "timeout_synrecv",
1592                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1593                 .maxlen         = sizeof(int),
1594                 .mode           = 0644,
1595                 .proc_handler   = proc_dointvec_jiffies,
1596         },
1597         {
1598                 .procname       = "timeout_finwait",
1599                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1600                 .maxlen         = sizeof(int),
1601                 .mode           = 0644,
1602                 .proc_handler   = proc_dointvec_jiffies,
1603         },
1604         {
1605                 .procname       = "timeout_timewait",
1606                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1607                 .maxlen         = sizeof(int),
1608                 .mode           = 0644,
1609                 .proc_handler   = proc_dointvec_jiffies,
1610         },
1611         {
1612                 .procname       = "timeout_close",
1613                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1614                 .maxlen         = sizeof(int),
1615                 .mode           = 0644,
1616                 .proc_handler   = proc_dointvec_jiffies,
1617         },
1618         {
1619                 .procname       = "timeout_closewait",
1620                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1621                 .maxlen         = sizeof(int),
1622                 .mode           = 0644,
1623                 .proc_handler   = proc_dointvec_jiffies,
1624         },
1625         {
1626                 .procname       = "timeout_lastack",
1627                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1628                 .maxlen         = sizeof(int),
1629                 .mode           = 0644,
1630                 .proc_handler   = proc_dointvec_jiffies,
1631         },
1632         {
1633                 .procname       = "timeout_listen",
1634                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1635                 .maxlen         = sizeof(int),
1636                 .mode           = 0644,
1637                 .proc_handler   = proc_dointvec_jiffies,
1638         },
1639         {
1640                 .procname       = "timeout_synack",
1641                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1642                 .maxlen         = sizeof(int),
1643                 .mode           = 0644,
1644                 .proc_handler   = proc_dointvec_jiffies,
1645         },
1646         {
1647                 .procname       = "timeout_udp",
1648                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1649                 .maxlen         = sizeof(int),
1650                 .mode           = 0644,
1651                 .proc_handler   = proc_dointvec_jiffies,
1652         },
1653         {
1654                 .procname       = "timeout_icmp",
1655                 .data   = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1656                 .maxlen         = sizeof(int),
1657                 .mode           = 0644,
1658                 .proc_handler   = proc_dointvec_jiffies,
1659         },
1660 #endif
1661         {
1662                 .procname       = "cache_bypass",
1663                 .data           = &sysctl_ip_vs_cache_bypass,
1664                 .maxlen         = sizeof(int),
1665                 .mode           = 0644,
1666                 .proc_handler   = proc_dointvec,
1667         },
1668         {
1669                 .procname       = "expire_nodest_conn",
1670                 .data           = &sysctl_ip_vs_expire_nodest_conn,
1671                 .maxlen         = sizeof(int),
1672                 .mode           = 0644,
1673                 .proc_handler   = proc_dointvec,
1674         },
1675         {
1676                 .procname       = "expire_quiescent_template",
1677                 .data           = &sysctl_ip_vs_expire_quiescent_template,
1678                 .maxlen         = sizeof(int),
1679                 .mode           = 0644,
1680                 .proc_handler   = proc_dointvec,
1681         },
1682         {
1683                 .procname       = "sync_threshold",
1684                 .data           = &sysctl_ip_vs_sync_threshold,
1685                 .maxlen         = sizeof(sysctl_ip_vs_sync_threshold),
1686                 .mode           = 0644,
1687                 .proc_handler   = proc_do_sync_threshold,
1688         },
1689         {
1690                 .procname       = "nat_icmp_send",
1691                 .data           = &sysctl_ip_vs_nat_icmp_send,
1692                 .maxlen         = sizeof(int),
1693                 .mode           = 0644,
1694                 .proc_handler   = proc_dointvec,
1695         },
1696         { }
1697 };
1698
1699 const struct ctl_path net_vs_ctl_path[] = {
1700         { .procname = "net", },
1701         { .procname = "ipv4", },
1702         { .procname = "vs", },
1703         { }
1704 };
1705 EXPORT_SYMBOL_GPL(net_vs_ctl_path);
1706
1707 static struct ctl_table_header * sysctl_header;
1708
1709 #ifdef CONFIG_PROC_FS
1710
1711 struct ip_vs_iter {
1712         struct list_head *table;
1713         int bucket;
1714 };
1715
1716 /*
1717  *      Write the contents of the VS rule table to a PROCfs file.
1718  *      (It is kept just for backward compatibility)
1719  */
1720 static inline const char *ip_vs_fwd_name(unsigned flags)
1721 {
1722         switch (flags & IP_VS_CONN_F_FWD_MASK) {
1723         case IP_VS_CONN_F_LOCALNODE:
1724                 return "Local";
1725         case IP_VS_CONN_F_TUNNEL:
1726                 return "Tunnel";
1727         case IP_VS_CONN_F_DROUTE:
1728                 return "Route";
1729         default:
1730                 return "Masq";
1731         }
1732 }
1733
1734
1735 /* Get the Nth entry in the two lists */
1736 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1737 {
1738         struct ip_vs_iter *iter = seq->private;
1739         int idx;
1740         struct ip_vs_service *svc;
1741
1742         /* look in hash by protocol */
1743         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1744                 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1745                         if (pos-- == 0){
1746                                 iter->table = ip_vs_svc_table;
1747                                 iter->bucket = idx;
1748                                 return svc;
1749                         }
1750                 }
1751         }
1752
1753         /* keep looking in fwmark */
1754         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1755                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1756                         if (pos-- == 0) {
1757                                 iter->table = ip_vs_svc_fwm_table;
1758                                 iter->bucket = idx;
1759                                 return svc;
1760                         }
1761                 }
1762         }
1763
1764         return NULL;
1765 }
1766
1767 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1768 __acquires(__ip_vs_svc_lock)
1769 {
1770
1771         read_lock_bh(&__ip_vs_svc_lock);
1772         return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1773 }
1774
1775
1776 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1777 {
1778         struct list_head *e;
1779         struct ip_vs_iter *iter;
1780         struct ip_vs_service *svc;
1781
1782         ++*pos;
1783         if (v == SEQ_START_TOKEN)
1784                 return ip_vs_info_array(seq,0);
1785
1786         svc = v;
1787         iter = seq->private;
1788
1789         if (iter->table == ip_vs_svc_table) {
1790                 /* next service in table hashed by protocol */
1791                 if ((e = svc->s_list.next) != &ip_vs_svc_table[iter->bucket])
1792                         return list_entry(e, struct ip_vs_service, s_list);
1793
1794
1795                 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1796                         list_for_each_entry(svc,&ip_vs_svc_table[iter->bucket],
1797                                             s_list) {
1798                                 return svc;
1799                         }
1800                 }
1801
1802                 iter->table = ip_vs_svc_fwm_table;
1803                 iter->bucket = -1;
1804                 goto scan_fwmark;
1805         }
1806
1807         /* next service in hashed by fwmark */
1808         if ((e = svc->f_list.next) != &ip_vs_svc_fwm_table[iter->bucket])
1809                 return list_entry(e, struct ip_vs_service, f_list);
1810
1811  scan_fwmark:
1812         while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1813                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[iter->bucket],
1814                                     f_list)
1815                         return svc;
1816         }
1817
1818         return NULL;
1819 }
1820
1821 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
1822 __releases(__ip_vs_svc_lock)
1823 {
1824         read_unlock_bh(&__ip_vs_svc_lock);
1825 }
1826
1827
1828 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
1829 {
1830         if (v == SEQ_START_TOKEN) {
1831                 seq_printf(seq,
1832                         "IP Virtual Server version %d.%d.%d (size=%d)\n",
1833                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
1834                 seq_puts(seq,
1835                          "Prot LocalAddress:Port Scheduler Flags\n");
1836                 seq_puts(seq,
1837                          "  -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
1838         } else {
1839                 const struct ip_vs_service *svc = v;
1840                 const struct ip_vs_iter *iter = seq->private;
1841                 const struct ip_vs_dest *dest;
1842
1843                 if (iter->table == ip_vs_svc_table) {
1844 #ifdef CONFIG_IP_VS_IPV6
1845                         if (svc->af == AF_INET6)
1846                                 seq_printf(seq, "%s  [%pI6]:%04X %s ",
1847                                            ip_vs_proto_name(svc->protocol),
1848                                            &svc->addr.in6,
1849                                            ntohs(svc->port),
1850                                            svc->scheduler->name);
1851                         else
1852 #endif
1853                                 seq_printf(seq, "%s  %08X:%04X %s %s ",
1854                                            ip_vs_proto_name(svc->protocol),
1855                                            ntohl(svc->addr.ip),
1856                                            ntohs(svc->port),
1857                                            svc->scheduler->name,
1858                                            (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
1859                 } else {
1860                         seq_printf(seq, "FWM  %08X %s %s",
1861                                    svc->fwmark, svc->scheduler->name,
1862                                    (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
1863                 }
1864
1865                 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
1866                         seq_printf(seq, "persistent %d %08X\n",
1867                                 svc->timeout,
1868                                 ntohl(svc->netmask));
1869                 else
1870                         seq_putc(seq, '\n');
1871
1872                 list_for_each_entry(dest, &svc->destinations, n_list) {
1873 #ifdef CONFIG_IP_VS_IPV6
1874                         if (dest->af == AF_INET6)
1875                                 seq_printf(seq,
1876                                            "  -> [%pI6]:%04X"
1877                                            "      %-7s %-6d %-10d %-10d\n",
1878                                            &dest->addr.in6,
1879                                            ntohs(dest->port),
1880                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
1881                                            atomic_read(&dest->weight),
1882                                            atomic_read(&dest->activeconns),
1883                                            atomic_read(&dest->inactconns));
1884                         else
1885 #endif
1886                                 seq_printf(seq,
1887                                            "  -> %08X:%04X      "
1888                                            "%-7s %-6d %-10d %-10d\n",
1889                                            ntohl(dest->addr.ip),
1890                                            ntohs(dest->port),
1891                                            ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
1892                                            atomic_read(&dest->weight),
1893                                            atomic_read(&dest->activeconns),
1894                                            atomic_read(&dest->inactconns));
1895
1896                 }
1897         }
1898         return 0;
1899 }
1900
1901 static const struct seq_operations ip_vs_info_seq_ops = {
1902         .start = ip_vs_info_seq_start,
1903         .next  = ip_vs_info_seq_next,
1904         .stop  = ip_vs_info_seq_stop,
1905         .show  = ip_vs_info_seq_show,
1906 };
1907
1908 static int ip_vs_info_open(struct inode *inode, struct file *file)
1909 {
1910         return seq_open_private(file, &ip_vs_info_seq_ops,
1911                         sizeof(struct ip_vs_iter));
1912 }
1913
1914 static const struct file_operations ip_vs_info_fops = {
1915         .owner   = THIS_MODULE,
1916         .open    = ip_vs_info_open,
1917         .read    = seq_read,
1918         .llseek  = seq_lseek,
1919         .release = seq_release_private,
1920 };
1921
1922 #endif
1923
1924 struct ip_vs_stats ip_vs_stats = {
1925         .lock = __SPIN_LOCK_UNLOCKED(ip_vs_stats.lock),
1926 };
1927
1928 #ifdef CONFIG_PROC_FS
1929 static int ip_vs_stats_show(struct seq_file *seq, void *v)
1930 {
1931
1932 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
1933         seq_puts(seq,
1934                  "   Total Incoming Outgoing         Incoming         Outgoing\n");
1935         seq_printf(seq,
1936                    "   Conns  Packets  Packets            Bytes            Bytes\n");
1937
1938         spin_lock_bh(&ip_vs_stats.lock);
1939         seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", ip_vs_stats.ustats.conns,
1940                    ip_vs_stats.ustats.inpkts, ip_vs_stats.ustats.outpkts,
1941                    (unsigned long long) ip_vs_stats.ustats.inbytes,
1942                    (unsigned long long) ip_vs_stats.ustats.outbytes);
1943
1944 /*                 01234567 01234567 01234567 0123456701234567 0123456701234567 */
1945         seq_puts(seq,
1946                    " Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
1947         seq_printf(seq,"%8X %8X %8X %16X %16X\n",
1948                         ip_vs_stats.ustats.cps,
1949                         ip_vs_stats.ustats.inpps,
1950                         ip_vs_stats.ustats.outpps,
1951                         ip_vs_stats.ustats.inbps,
1952                         ip_vs_stats.ustats.outbps);
1953         spin_unlock_bh(&ip_vs_stats.lock);
1954
1955         return 0;
1956 }
1957
1958 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
1959 {
1960         return single_open(file, ip_vs_stats_show, NULL);
1961 }
1962
1963 static const struct file_operations ip_vs_stats_fops = {
1964         .owner = THIS_MODULE,
1965         .open = ip_vs_stats_seq_open,
1966         .read = seq_read,
1967         .llseek = seq_lseek,
1968         .release = single_release,
1969 };
1970
1971 #endif
1972
1973 /*
1974  *      Set timeout values for tcp tcpfin udp in the timeout_table.
1975  */
1976 static int ip_vs_set_timeout(struct ip_vs_timeout_user *u)
1977 {
1978         IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
1979                   u->tcp_timeout,
1980                   u->tcp_fin_timeout,
1981                   u->udp_timeout);
1982
1983 #ifdef CONFIG_IP_VS_PROTO_TCP
1984         if (u->tcp_timeout) {
1985                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_ESTABLISHED]
1986                         = u->tcp_timeout * HZ;
1987         }
1988
1989         if (u->tcp_fin_timeout) {
1990                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_FIN_WAIT]
1991                         = u->tcp_fin_timeout * HZ;
1992         }
1993 #endif
1994
1995 #ifdef CONFIG_IP_VS_PROTO_UDP
1996         if (u->udp_timeout) {
1997                 ip_vs_protocol_udp.timeout_table[IP_VS_UDP_S_NORMAL]
1998                         = u->udp_timeout * HZ;
1999         }
2000 #endif
2001         return 0;
2002 }
2003
2004
2005 #define SET_CMDID(cmd)          (cmd - IP_VS_BASE_CTL)
2006 #define SERVICE_ARG_LEN         (sizeof(struct ip_vs_service_user))
2007 #define SVCDEST_ARG_LEN         (sizeof(struct ip_vs_service_user) +    \
2008                                  sizeof(struct ip_vs_dest_user))
2009 #define TIMEOUT_ARG_LEN         (sizeof(struct ip_vs_timeout_user))
2010 #define DAEMON_ARG_LEN          (sizeof(struct ip_vs_daemon_user))
2011 #define MAX_ARG_LEN             SVCDEST_ARG_LEN
2012
2013 static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
2014         [SET_CMDID(IP_VS_SO_SET_ADD)]           = SERVICE_ARG_LEN,
2015         [SET_CMDID(IP_VS_SO_SET_EDIT)]          = SERVICE_ARG_LEN,
2016         [SET_CMDID(IP_VS_SO_SET_DEL)]           = SERVICE_ARG_LEN,
2017         [SET_CMDID(IP_VS_SO_SET_FLUSH)]         = 0,
2018         [SET_CMDID(IP_VS_SO_SET_ADDDEST)]       = SVCDEST_ARG_LEN,
2019         [SET_CMDID(IP_VS_SO_SET_DELDEST)]       = SVCDEST_ARG_LEN,
2020         [SET_CMDID(IP_VS_SO_SET_EDITDEST)]      = SVCDEST_ARG_LEN,
2021         [SET_CMDID(IP_VS_SO_SET_TIMEOUT)]       = TIMEOUT_ARG_LEN,
2022         [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)]   = DAEMON_ARG_LEN,
2023         [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)]    = DAEMON_ARG_LEN,
2024         [SET_CMDID(IP_VS_SO_SET_ZERO)]          = SERVICE_ARG_LEN,
2025 };
2026
2027 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2028                                   struct ip_vs_service_user *usvc_compat)
2029 {
2030         usvc->af                = AF_INET;
2031         usvc->protocol          = usvc_compat->protocol;
2032         usvc->addr.ip           = usvc_compat->addr;
2033         usvc->port              = usvc_compat->port;
2034         usvc->fwmark            = usvc_compat->fwmark;
2035
2036         /* Deep copy of sched_name is not needed here */
2037         usvc->sched_name        = usvc_compat->sched_name;
2038
2039         usvc->flags             = usvc_compat->flags;
2040         usvc->timeout           = usvc_compat->timeout;
2041         usvc->netmask           = usvc_compat->netmask;
2042 }
2043
2044 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2045                                    struct ip_vs_dest_user *udest_compat)
2046 {
2047         udest->addr.ip          = udest_compat->addr;
2048         udest->port             = udest_compat->port;
2049         udest->conn_flags       = udest_compat->conn_flags;
2050         udest->weight           = udest_compat->weight;
2051         udest->u_threshold      = udest_compat->u_threshold;
2052         udest->l_threshold      = udest_compat->l_threshold;
2053 }
2054
2055 static int
2056 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2057 {
2058         int ret;
2059         unsigned char arg[MAX_ARG_LEN];
2060         struct ip_vs_service_user *usvc_compat;
2061         struct ip_vs_service_user_kern usvc;
2062         struct ip_vs_service *svc;
2063         struct ip_vs_dest_user *udest_compat;
2064         struct ip_vs_dest_user_kern udest;
2065
2066         if (!capable(CAP_NET_ADMIN))
2067                 return -EPERM;
2068
2069         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2070                 return -EINVAL;
2071         if (len < 0 || len >  MAX_ARG_LEN)
2072                 return -EINVAL;
2073         if (len != set_arglen[SET_CMDID(cmd)]) {
2074                 pr_err("set_ctl: len %u != %u\n",
2075                        len, set_arglen[SET_CMDID(cmd)]);
2076                 return -EINVAL;
2077         }
2078
2079         if (copy_from_user(arg, user, len) != 0)
2080                 return -EFAULT;
2081
2082         /* increase the module use count */
2083         ip_vs_use_count_inc();
2084
2085         if (mutex_lock_interruptible(&__ip_vs_mutex)) {
2086                 ret = -ERESTARTSYS;
2087                 goto out_dec;
2088         }
2089
2090         if (cmd == IP_VS_SO_SET_FLUSH) {
2091                 /* Flush the virtual service */
2092                 ret = ip_vs_flush();
2093                 goto out_unlock;
2094         } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2095                 /* Set timeout values for (tcp tcpfin udp) */
2096                 ret = ip_vs_set_timeout((struct ip_vs_timeout_user *)arg);
2097                 goto out_unlock;
2098         } else if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2099                 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2100                 ret = start_sync_thread(dm->state, dm->mcast_ifn, dm->syncid);
2101                 goto out_unlock;
2102         } else if (cmd == IP_VS_SO_SET_STOPDAEMON) {
2103                 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2104                 ret = stop_sync_thread(dm->state);
2105                 goto out_unlock;
2106         }
2107
2108         usvc_compat = (struct ip_vs_service_user *)arg;
2109         udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2110
2111         /* We only use the new structs internally, so copy userspace compat
2112          * structs to extended internal versions */
2113         ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2114         ip_vs_copy_udest_compat(&udest, udest_compat);
2115
2116         if (cmd == IP_VS_SO_SET_ZERO) {
2117                 /* if no service address is set, zero counters in all */
2118                 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2119                         ret = ip_vs_zero_all();
2120                         goto out_unlock;
2121                 }
2122         }
2123
2124         /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2125         if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2126             usvc.protocol != IPPROTO_SCTP) {
2127                 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2128                        usvc.protocol, &usvc.addr.ip,
2129                        ntohs(usvc.port), usvc.sched_name);
2130                 ret = -EFAULT;
2131                 goto out_unlock;
2132         }
2133
2134         /* Lookup the exact service by <protocol, addr, port> or fwmark */
2135         if (usvc.fwmark == 0)
2136                 svc = __ip_vs_service_find(usvc.af, usvc.protocol,
2137                                            &usvc.addr, usvc.port);
2138         else
2139                 svc = __ip_vs_svc_fwm_find(usvc.af, usvc.fwmark);
2140
2141         if (cmd != IP_VS_SO_SET_ADD
2142             && (svc == NULL || svc->protocol != usvc.protocol)) {
2143                 ret = -ESRCH;
2144                 goto out_unlock;
2145         }
2146
2147         switch (cmd) {
2148         case IP_VS_SO_SET_ADD:
2149                 if (svc != NULL)
2150                         ret = -EEXIST;
2151                 else
2152                         ret = ip_vs_add_service(&usvc, &svc);
2153                 break;
2154         case IP_VS_SO_SET_EDIT:
2155                 ret = ip_vs_edit_service(svc, &usvc);
2156                 break;
2157         case IP_VS_SO_SET_DEL:
2158                 ret = ip_vs_del_service(svc);
2159                 if (!ret)
2160                         goto out_unlock;
2161                 break;
2162         case IP_VS_SO_SET_ZERO:
2163                 ret = ip_vs_zero_service(svc);
2164                 break;
2165         case IP_VS_SO_SET_ADDDEST:
2166                 ret = ip_vs_add_dest(svc, &udest);
2167                 break;
2168         case IP_VS_SO_SET_EDITDEST:
2169                 ret = ip_vs_edit_dest(svc, &udest);
2170                 break;
2171         case IP_VS_SO_SET_DELDEST:
2172                 ret = ip_vs_del_dest(svc, &udest);
2173                 break;
2174         default:
2175                 ret = -EINVAL;
2176         }
2177
2178   out_unlock:
2179         mutex_unlock(&__ip_vs_mutex);
2180   out_dec:
2181         /* decrease the module use count */
2182         ip_vs_use_count_dec();
2183
2184         return ret;
2185 }
2186
2187
2188 static void
2189 ip_vs_copy_stats(struct ip_vs_stats_user *dst, struct ip_vs_stats *src)
2190 {
2191         spin_lock_bh(&src->lock);
2192         memcpy(dst, &src->ustats, sizeof(*dst));
2193         spin_unlock_bh(&src->lock);
2194 }
2195
2196 static void
2197 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2198 {
2199         dst->protocol = src->protocol;
2200         dst->addr = src->addr.ip;
2201         dst->port = src->port;
2202         dst->fwmark = src->fwmark;
2203         strlcpy(dst->sched_name, src->scheduler->name, sizeof(dst->sched_name));
2204         dst->flags = src->flags;
2205         dst->timeout = src->timeout / HZ;
2206         dst->netmask = src->netmask;
2207         dst->num_dests = src->num_dests;
2208         ip_vs_copy_stats(&dst->stats, &src->stats);
2209 }
2210
2211 static inline int
2212 __ip_vs_get_service_entries(const struct ip_vs_get_services *get,
2213                             struct ip_vs_get_services __user *uptr)
2214 {
2215         int idx, count=0;
2216         struct ip_vs_service *svc;
2217         struct ip_vs_service_entry entry;
2218         int ret = 0;
2219
2220         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2221                 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2222                         /* Only expose IPv4 entries to old interface */
2223                         if (svc->af != AF_INET)
2224                                 continue;
2225
2226                         if (count >= get->num_services)
2227                                 goto out;
2228                         memset(&entry, 0, sizeof(entry));
2229                         ip_vs_copy_service(&entry, svc);
2230                         if (copy_to_user(&uptr->entrytable[count],
2231                                          &entry, sizeof(entry))) {
2232                                 ret = -EFAULT;
2233                                 goto out;
2234                         }
2235                         count++;
2236                 }
2237         }
2238
2239         for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2240                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2241                         /* Only expose IPv4 entries to old interface */
2242                         if (svc->af != AF_INET)
2243                                 continue;
2244
2245                         if (count >= get->num_services)
2246                                 goto out;
2247                         memset(&entry, 0, sizeof(entry));
2248                         ip_vs_copy_service(&entry, svc);
2249                         if (copy_to_user(&uptr->entrytable[count],
2250                                          &entry, sizeof(entry))) {
2251                                 ret = -EFAULT;
2252                                 goto out;
2253                         }
2254                         count++;
2255                 }
2256         }
2257   out:
2258         return ret;
2259 }
2260
2261 static inline int
2262 __ip_vs_get_dest_entries(const struct ip_vs_get_dests *get,
2263                          struct ip_vs_get_dests __user *uptr)
2264 {
2265         struct ip_vs_service *svc;
2266         union nf_inet_addr addr = { .ip = get->addr };
2267         int ret = 0;
2268
2269         if (get->fwmark)
2270                 svc = __ip_vs_svc_fwm_find(AF_INET, get->fwmark);
2271         else
2272                 svc = __ip_vs_service_find(AF_INET, get->protocol, &addr,
2273                                            get->port);
2274
2275         if (svc) {
2276                 int count = 0;
2277                 struct ip_vs_dest *dest;
2278                 struct ip_vs_dest_entry entry;
2279
2280                 list_for_each_entry(dest, &svc->destinations, n_list) {
2281                         if (count >= get->num_dests)
2282                                 break;
2283
2284                         entry.addr = dest->addr.ip;
2285                         entry.port = dest->port;
2286                         entry.conn_flags = atomic_read(&dest->conn_flags);
2287                         entry.weight = atomic_read(&dest->weight);
2288                         entry.u_threshold = dest->u_threshold;
2289                         entry.l_threshold = dest->l_threshold;
2290                         entry.activeconns = atomic_read(&dest->activeconns);
2291                         entry.inactconns = atomic_read(&dest->inactconns);
2292                         entry.persistconns = atomic_read(&dest->persistconns);
2293                         ip_vs_copy_stats(&entry.stats, &dest->stats);
2294                         if (copy_to_user(&uptr->entrytable[count],
2295                                          &entry, sizeof(entry))) {
2296                                 ret = -EFAULT;
2297                                 break;
2298                         }
2299                         count++;
2300                 }
2301         } else
2302                 ret = -ESRCH;
2303         return ret;
2304 }
2305
2306 static inline void
2307 __ip_vs_get_timeouts(struct ip_vs_timeout_user *u)
2308 {
2309 #ifdef CONFIG_IP_VS_PROTO_TCP
2310         u->tcp_timeout =
2311                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2312         u->tcp_fin_timeout =
2313                 ip_vs_protocol_tcp.timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2314 #endif
2315 #ifdef CONFIG_IP_VS_PROTO_UDP
2316         u->udp_timeout =
2317                 ip_vs_protocol_udp.timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2318 #endif
2319 }
2320
2321
2322 #define GET_CMDID(cmd)          (cmd - IP_VS_BASE_CTL)
2323 #define GET_INFO_ARG_LEN        (sizeof(struct ip_vs_getinfo))
2324 #define GET_SERVICES_ARG_LEN    (sizeof(struct ip_vs_get_services))
2325 #define GET_SERVICE_ARG_LEN     (sizeof(struct ip_vs_service_entry))
2326 #define GET_DESTS_ARG_LEN       (sizeof(struct ip_vs_get_dests))
2327 #define GET_TIMEOUT_ARG_LEN     (sizeof(struct ip_vs_timeout_user))
2328 #define GET_DAEMON_ARG_LEN      (sizeof(struct ip_vs_daemon_user) * 2)
2329
2330 static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
2331         [GET_CMDID(IP_VS_SO_GET_VERSION)]       = 64,
2332         [GET_CMDID(IP_VS_SO_GET_INFO)]          = GET_INFO_ARG_LEN,
2333         [GET_CMDID(IP_VS_SO_GET_SERVICES)]      = GET_SERVICES_ARG_LEN,
2334         [GET_CMDID(IP_VS_SO_GET_SERVICE)]       = GET_SERVICE_ARG_LEN,
2335         [GET_CMDID(IP_VS_SO_GET_DESTS)]         = GET_DESTS_ARG_LEN,
2336         [GET_CMDID(IP_VS_SO_GET_TIMEOUT)]       = GET_TIMEOUT_ARG_LEN,
2337         [GET_CMDID(IP_VS_SO_GET_DAEMON)]        = GET_DAEMON_ARG_LEN,
2338 };
2339
2340 static int
2341 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2342 {
2343         unsigned char arg[128];
2344         int ret = 0;
2345         unsigned int copylen;
2346
2347         if (!capable(CAP_NET_ADMIN))
2348                 return -EPERM;
2349
2350         if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2351                 return -EINVAL;
2352
2353         if (*len < get_arglen[GET_CMDID(cmd)]) {
2354                 pr_err("get_ctl: len %u < %u\n",
2355                        *len, get_arglen[GET_CMDID(cmd)]);
2356                 return -EINVAL;
2357         }
2358
2359         copylen = get_arglen[GET_CMDID(cmd)];
2360         if (copylen > 128)
2361                 return -EINVAL;
2362
2363         if (copy_from_user(arg, user, copylen) != 0)
2364                 return -EFAULT;
2365
2366         if (mutex_lock_interruptible(&__ip_vs_mutex))
2367                 return -ERESTARTSYS;
2368
2369         switch (cmd) {
2370         case IP_VS_SO_GET_VERSION:
2371         {
2372                 char buf[64];
2373
2374                 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2375                         NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2376                 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2377                         ret = -EFAULT;
2378                         goto out;
2379                 }
2380                 *len = strlen(buf)+1;
2381         }
2382         break;
2383
2384         case IP_VS_SO_GET_INFO:
2385         {
2386                 struct ip_vs_getinfo info;
2387                 info.version = IP_VS_VERSION_CODE;
2388                 info.size = ip_vs_conn_tab_size;
2389                 info.num_services = ip_vs_num_services;
2390                 if (copy_to_user(user, &info, sizeof(info)) != 0)
2391                         ret = -EFAULT;
2392         }
2393         break;
2394
2395         case IP_VS_SO_GET_SERVICES:
2396         {
2397                 struct ip_vs_get_services *get;
2398                 int size;
2399
2400                 get = (struct ip_vs_get_services *)arg;
2401                 size = sizeof(*get) +
2402                         sizeof(struct ip_vs_service_entry) * get->num_services;
2403                 if (*len != size) {
2404                         pr_err("length: %u != %u\n", *len, size);
2405                         ret = -EINVAL;
2406                         goto out;
2407                 }
2408                 ret = __ip_vs_get_service_entries(get, user);
2409         }
2410         break;
2411
2412         case IP_VS_SO_GET_SERVICE:
2413         {
2414                 struct ip_vs_service_entry *entry;
2415                 struct ip_vs_service *svc;
2416                 union nf_inet_addr addr;
2417
2418                 entry = (struct ip_vs_service_entry *)arg;
2419                 addr.ip = entry->addr;
2420                 if (entry->fwmark)
2421                         svc = __ip_vs_svc_fwm_find(AF_INET, entry->fwmark);
2422                 else
2423                         svc = __ip_vs_service_find(AF_INET, entry->protocol,
2424                                                    &addr, entry->port);
2425                 if (svc) {
2426                         ip_vs_copy_service(entry, svc);
2427                         if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2428                                 ret = -EFAULT;
2429                 } else
2430                         ret = -ESRCH;
2431         }
2432         break;
2433
2434         case IP_VS_SO_GET_DESTS:
2435         {
2436                 struct ip_vs_get_dests *get;
2437                 int size;
2438
2439                 get = (struct ip_vs_get_dests *)arg;
2440                 size = sizeof(*get) +
2441                         sizeof(struct ip_vs_dest_entry) * get->num_dests;
2442                 if (*len != size) {
2443                         pr_err("length: %u != %u\n", *len, size);
2444                         ret = -EINVAL;
2445                         goto out;
2446                 }
2447                 ret = __ip_vs_get_dest_entries(get, user);
2448         }
2449         break;
2450
2451         case IP_VS_SO_GET_TIMEOUT:
2452         {
2453                 struct ip_vs_timeout_user t;
2454
2455                 __ip_vs_get_timeouts(&t);
2456                 if (copy_to_user(user, &t, sizeof(t)) != 0)
2457                         ret = -EFAULT;
2458         }
2459         break;
2460
2461         case IP_VS_SO_GET_DAEMON:
2462         {
2463                 struct ip_vs_daemon_user d[2];
2464
2465                 memset(&d, 0, sizeof(d));
2466                 if (ip_vs_sync_state & IP_VS_STATE_MASTER) {
2467                         d[0].state = IP_VS_STATE_MASTER;
2468                         strlcpy(d[0].mcast_ifn, ip_vs_master_mcast_ifn, sizeof(d[0].mcast_ifn));
2469                         d[0].syncid = ip_vs_master_syncid;
2470                 }
2471                 if (ip_vs_sync_state & IP_VS_STATE_BACKUP) {
2472                         d[1].state = IP_VS_STATE_BACKUP;
2473                         strlcpy(d[1].mcast_ifn, ip_vs_backup_mcast_ifn, sizeof(d[1].mcast_ifn));
2474                         d[1].syncid = ip_vs_backup_syncid;
2475                 }
2476                 if (copy_to_user(user, &d, sizeof(d)) != 0)
2477                         ret = -EFAULT;
2478         }
2479         break;
2480
2481         default:
2482                 ret = -EINVAL;
2483         }
2484
2485   out:
2486         mutex_unlock(&__ip_vs_mutex);
2487         return ret;
2488 }
2489
2490
2491 static struct nf_sockopt_ops ip_vs_sockopts = {
2492         .pf             = PF_INET,
2493         .set_optmin     = IP_VS_BASE_CTL,
2494         .set_optmax     = IP_VS_SO_SET_MAX+1,
2495         .set            = do_ip_vs_set_ctl,
2496         .get_optmin     = IP_VS_BASE_CTL,
2497         .get_optmax     = IP_VS_SO_GET_MAX+1,
2498         .get            = do_ip_vs_get_ctl,
2499         .owner          = THIS_MODULE,
2500 };
2501
2502 /*
2503  * Generic Netlink interface
2504  */
2505
2506 /* IPVS genetlink family */
2507 static struct genl_family ip_vs_genl_family = {
2508         .id             = GENL_ID_GENERATE,
2509         .hdrsize        = 0,
2510         .name           = IPVS_GENL_NAME,
2511         .version        = IPVS_GENL_VERSION,
2512         .maxattr        = IPVS_CMD_MAX,
2513 };
2514
2515 /* Policy used for first-level command attributes */
2516 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2517         [IPVS_CMD_ATTR_SERVICE]         = { .type = NLA_NESTED },
2518         [IPVS_CMD_ATTR_DEST]            = { .type = NLA_NESTED },
2519         [IPVS_CMD_ATTR_DAEMON]          = { .type = NLA_NESTED },
2520         [IPVS_CMD_ATTR_TIMEOUT_TCP]     = { .type = NLA_U32 },
2521         [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2522         [IPVS_CMD_ATTR_TIMEOUT_UDP]     = { .type = NLA_U32 },
2523 };
2524
2525 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2526 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2527         [IPVS_DAEMON_ATTR_STATE]        = { .type = NLA_U32 },
2528         [IPVS_DAEMON_ATTR_MCAST_IFN]    = { .type = NLA_NUL_STRING,
2529                                             .len = IP_VS_IFNAME_MAXLEN },
2530         [IPVS_DAEMON_ATTR_SYNC_ID]      = { .type = NLA_U32 },
2531 };
2532
2533 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2534 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2535         [IPVS_SVC_ATTR_AF]              = { .type = NLA_U16 },
2536         [IPVS_SVC_ATTR_PROTOCOL]        = { .type = NLA_U16 },
2537         [IPVS_SVC_ATTR_ADDR]            = { .type = NLA_BINARY,
2538                                             .len = sizeof(union nf_inet_addr) },
2539         [IPVS_SVC_ATTR_PORT]            = { .type = NLA_U16 },
2540         [IPVS_SVC_ATTR_FWMARK]          = { .type = NLA_U32 },
2541         [IPVS_SVC_ATTR_SCHED_NAME]      = { .type = NLA_NUL_STRING,
2542                                             .len = IP_VS_SCHEDNAME_MAXLEN },
2543         [IPVS_SVC_ATTR_FLAGS]           = { .type = NLA_BINARY,
2544                                             .len = sizeof(struct ip_vs_flags) },
2545         [IPVS_SVC_ATTR_TIMEOUT]         = { .type = NLA_U32 },
2546         [IPVS_SVC_ATTR_NETMASK]         = { .type = NLA_U32 },
2547         [IPVS_SVC_ATTR_STATS]           = { .type = NLA_NESTED },
2548 };
2549
2550 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2551 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2552         [IPVS_DEST_ATTR_ADDR]           = { .type = NLA_BINARY,
2553                                             .len = sizeof(union nf_inet_addr) },
2554         [IPVS_DEST_ATTR_PORT]           = { .type = NLA_U16 },
2555         [IPVS_DEST_ATTR_FWD_METHOD]     = { .type = NLA_U32 },
2556         [IPVS_DEST_ATTR_WEIGHT]         = { .type = NLA_U32 },
2557         [IPVS_DEST_ATTR_U_THRESH]       = { .type = NLA_U32 },
2558         [IPVS_DEST_ATTR_L_THRESH]       = { .type = NLA_U32 },
2559         [IPVS_DEST_ATTR_ACTIVE_CONNS]   = { .type = NLA_U32 },
2560         [IPVS_DEST_ATTR_INACT_CONNS]    = { .type = NLA_U32 },
2561         [IPVS_DEST_ATTR_PERSIST_CONNS]  = { .type = NLA_U32 },
2562         [IPVS_DEST_ATTR_STATS]          = { .type = NLA_NESTED },
2563 };
2564
2565 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2566                                  struct ip_vs_stats *stats)
2567 {
2568         struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2569         if (!nl_stats)
2570                 return -EMSGSIZE;
2571
2572         spin_lock_bh(&stats->lock);
2573
2574         NLA_PUT_U32(skb, IPVS_STATS_ATTR_CONNS, stats->ustats.conns);
2575         NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPKTS, stats->ustats.inpkts);
2576         NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPKTS, stats->ustats.outpkts);
2577         NLA_PUT_U64(skb, IPVS_STATS_ATTR_INBYTES, stats->ustats.inbytes);
2578         NLA_PUT_U64(skb, IPVS_STATS_ATTR_OUTBYTES, stats->ustats.outbytes);
2579         NLA_PUT_U32(skb, IPVS_STATS_ATTR_CPS, stats->ustats.cps);
2580         NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPPS, stats->ustats.inpps);
2581         NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPPS, stats->ustats.outpps);
2582         NLA_PUT_U32(skb, IPVS_STATS_ATTR_INBPS, stats->ustats.inbps);
2583         NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTBPS, stats->ustats.outbps);
2584
2585         spin_unlock_bh(&stats->lock);
2586
2587         nla_nest_end(skb, nl_stats);
2588
2589         return 0;
2590
2591 nla_put_failure:
2592         spin_unlock_bh(&stats->lock);
2593         nla_nest_cancel(skb, nl_stats);
2594         return -EMSGSIZE;
2595 }
2596
2597 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2598                                    struct ip_vs_service *svc)
2599 {
2600         struct nlattr *nl_service;
2601         struct ip_vs_flags flags = { .flags = svc->flags,
2602                                      .mask = ~0 };
2603
2604         nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2605         if (!nl_service)
2606                 return -EMSGSIZE;
2607
2608         NLA_PUT_U16(skb, IPVS_SVC_ATTR_AF, svc->af);
2609
2610         if (svc->fwmark) {
2611                 NLA_PUT_U32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark);
2612         } else {
2613                 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol);
2614                 NLA_PUT(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr);
2615                 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PORT, svc->port);
2616         }
2617
2618         NLA_PUT_STRING(skb, IPVS_SVC_ATTR_SCHED_NAME, svc->scheduler->name);
2619         NLA_PUT(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags);
2620         NLA_PUT_U32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ);
2621         NLA_PUT_U32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask);
2622
2623         if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2624                 goto nla_put_failure;
2625
2626         nla_nest_end(skb, nl_service);
2627
2628         return 0;
2629
2630 nla_put_failure:
2631         nla_nest_cancel(skb, nl_service);
2632         return -EMSGSIZE;
2633 }
2634
2635 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2636                                    struct ip_vs_service *svc,
2637                                    struct netlink_callback *cb)
2638 {
2639         void *hdr;
2640
2641         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2642                           &ip_vs_genl_family, NLM_F_MULTI,
2643                           IPVS_CMD_NEW_SERVICE);
2644         if (!hdr)
2645                 return -EMSGSIZE;
2646
2647         if (ip_vs_genl_fill_service(skb, svc) < 0)
2648                 goto nla_put_failure;
2649
2650         return genlmsg_end(skb, hdr);
2651
2652 nla_put_failure:
2653         genlmsg_cancel(skb, hdr);
2654         return -EMSGSIZE;
2655 }
2656
2657 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2658                                     struct netlink_callback *cb)
2659 {
2660         int idx = 0, i;
2661         int start = cb->args[0];
2662         struct ip_vs_service *svc;
2663
2664         mutex_lock(&__ip_vs_mutex);
2665         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2666                 list_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2667                         if (++idx <= start)
2668                                 continue;
2669                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2670                                 idx--;
2671                                 goto nla_put_failure;
2672                         }
2673                 }
2674         }
2675
2676         for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2677                 list_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
2678                         if (++idx <= start)
2679                                 continue;
2680                         if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2681                                 idx--;
2682                                 goto nla_put_failure;
2683                         }
2684                 }
2685         }
2686
2687 nla_put_failure:
2688         mutex_unlock(&__ip_vs_mutex);
2689         cb->args[0] = idx;
2690
2691         return skb->len;
2692 }
2693
2694 static int ip_vs_genl_parse_service(struct ip_vs_service_user_kern *usvc,
2695                                     struct nlattr *nla, int full_entry,
2696                                     struct ip_vs_service **ret_svc)
2697 {
2698         struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
2699         struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
2700         struct ip_vs_service *svc;
2701
2702         /* Parse mandatory identifying service fields first */
2703         if (nla == NULL ||
2704             nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
2705                 return -EINVAL;
2706
2707         nla_af          = attrs[IPVS_SVC_ATTR_AF];
2708         nla_protocol    = attrs[IPVS_SVC_ATTR_PROTOCOL];
2709         nla_addr        = attrs[IPVS_SVC_ATTR_ADDR];
2710         nla_port        = attrs[IPVS_SVC_ATTR_PORT];
2711         nla_fwmark      = attrs[IPVS_SVC_ATTR_FWMARK];
2712
2713         if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
2714                 return -EINVAL;
2715
2716         memset(usvc, 0, sizeof(*usvc));
2717
2718         usvc->af = nla_get_u16(nla_af);
2719 #ifdef CONFIG_IP_VS_IPV6
2720         if (usvc->af != AF_INET && usvc->af != AF_INET6)
2721 #else
2722         if (usvc->af != AF_INET)
2723 #endif
2724                 return -EAFNOSUPPORT;
2725
2726         if (nla_fwmark) {
2727                 usvc->protocol = IPPROTO_TCP;
2728                 usvc->fwmark = nla_get_u32(nla_fwmark);
2729         } else {
2730                 usvc->protocol = nla_get_u16(nla_protocol);
2731                 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
2732                 usvc->port = nla_get_u16(nla_port);
2733                 usvc->fwmark = 0;
2734         }
2735
2736         if (usvc->fwmark)
2737                 svc = __ip_vs_svc_fwm_find(usvc->af, usvc->fwmark);
2738         else
2739                 svc = __ip_vs_service_find(usvc->af, usvc->protocol,
2740                                            &usvc->addr, usvc->port);
2741         *ret_svc = svc;
2742
2743         /* If a full entry was requested, check for the additional fields */
2744         if (full_entry) {
2745                 struct nlattr *nla_sched, *nla_flags, *nla_timeout,
2746                               *nla_netmask;
2747                 struct ip_vs_flags flags;
2748
2749                 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
2750                 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
2751                 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
2752                 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
2753
2754                 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
2755                         return -EINVAL;
2756
2757                 nla_memcpy(&flags, nla_flags, sizeof(flags));
2758
2759                 /* prefill flags from service if it already exists */
2760                 if (svc)
2761                         usvc->flags = svc->flags;
2762
2763                 /* set new flags from userland */
2764                 usvc->flags = (usvc->flags & ~flags.mask) |
2765                               (flags.flags & flags.mask);
2766                 usvc->sched_name = nla_data(nla_sched);
2767                 usvc->timeout = nla_get_u32(nla_timeout);
2768                 usvc->netmask = nla_get_u32(nla_netmask);
2769         }
2770
2771         return 0;
2772 }
2773
2774 static struct ip_vs_service *ip_vs_genl_find_service(struct nlattr *nla)
2775 {
2776         struct ip_vs_service_user_kern usvc;
2777         struct ip_vs_service *svc;
2778         int ret;
2779
2780         ret = ip_vs_genl_parse_service(&usvc, nla, 0, &svc);
2781         return ret ? ERR_PTR(ret) : svc;
2782 }
2783
2784 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
2785 {
2786         struct nlattr *nl_dest;
2787
2788         nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
2789         if (!nl_dest)
2790                 return -EMSGSIZE;
2791
2792         NLA_PUT(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr);
2793         NLA_PUT_U16(skb, IPVS_DEST_ATTR_PORT, dest->port);
2794
2795         NLA_PUT_U32(skb, IPVS_DEST_ATTR_FWD_METHOD,
2796                     atomic_read(&dest->conn_flags) & IP_VS_CONN_F_FWD_MASK);
2797         NLA_PUT_U32(skb, IPVS_DEST_ATTR_WEIGHT, atomic_read(&dest->weight));
2798         NLA_PUT_U32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold);
2799         NLA_PUT_U32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold);
2800         NLA_PUT_U32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
2801                     atomic_read(&dest->activeconns));
2802         NLA_PUT_U32(skb, IPVS_DEST_ATTR_INACT_CONNS,
2803                     atomic_read(&dest->inactconns));
2804         NLA_PUT_U32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
2805                     atomic_read(&dest->persistconns));
2806
2807         if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
2808                 goto nla_put_failure;
2809
2810         nla_nest_end(skb, nl_dest);
2811
2812         return 0;
2813
2814 nla_put_failure:
2815         nla_nest_cancel(skb, nl_dest);
2816         return -EMSGSIZE;
2817 }
2818
2819 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
2820                                 struct netlink_callback *cb)
2821 {
2822         void *hdr;
2823
2824         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2825                           &ip_vs_genl_family, NLM_F_MULTI,
2826                           IPVS_CMD_NEW_DEST);
2827         if (!hdr)
2828                 return -EMSGSIZE;
2829
2830         if (ip_vs_genl_fill_dest(skb, dest) < 0)
2831                 goto nla_put_failure;
2832
2833         return genlmsg_end(skb, hdr);
2834
2835 nla_put_failure:
2836         genlmsg_cancel(skb, hdr);
2837         return -EMSGSIZE;
2838 }
2839
2840 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
2841                                  struct netlink_callback *cb)
2842 {
2843         int idx = 0;
2844         int start = cb->args[0];
2845         struct ip_vs_service *svc;
2846         struct ip_vs_dest *dest;
2847         struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
2848
2849         mutex_lock(&__ip_vs_mutex);
2850
2851         /* Try to find the service for which to dump destinations */
2852         if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
2853                         IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
2854                 goto out_err;
2855
2856         svc = ip_vs_genl_find_service(attrs[IPVS_CMD_ATTR_SERVICE]);
2857         if (IS_ERR(svc) || svc == NULL)
2858                 goto out_err;
2859
2860         /* Dump the destinations */
2861         list_for_each_entry(dest, &svc->destinations, n_list) {
2862                 if (++idx <= start)
2863                         continue;
2864                 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
2865                         idx--;
2866                         goto nla_put_failure;
2867                 }
2868         }
2869
2870 nla_put_failure:
2871         cb->args[0] = idx;
2872
2873 out_err:
2874         mutex_unlock(&__ip_vs_mutex);
2875
2876         return skb->len;
2877 }
2878
2879 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
2880                                  struct nlattr *nla, int full_entry)
2881 {
2882         struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
2883         struct nlattr *nla_addr, *nla_port;
2884
2885         /* Parse mandatory identifying destination fields first */
2886         if (nla == NULL ||
2887             nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
2888                 return -EINVAL;
2889
2890         nla_addr        = attrs[IPVS_DEST_ATTR_ADDR];
2891         nla_port        = attrs[IPVS_DEST_ATTR_PORT];
2892
2893         if (!(nla_addr && nla_port))
2894                 return -EINVAL;
2895
2896         memset(udest, 0, sizeof(*udest));
2897
2898         nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
2899         udest->port = nla_get_u16(nla_port);
2900
2901         /* If a full entry was requested, check for the additional fields */
2902         if (full_entry) {
2903                 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
2904                               *nla_l_thresh;
2905
2906                 nla_fwd         = attrs[IPVS_DEST_ATTR_FWD_METHOD];
2907                 nla_weight      = attrs[IPVS_DEST_ATTR_WEIGHT];
2908                 nla_u_thresh    = attrs[IPVS_DEST_ATTR_U_THRESH];
2909                 nla_l_thresh    = attrs[IPVS_DEST_ATTR_L_THRESH];
2910
2911                 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
2912                         return -EINVAL;
2913
2914                 udest->conn_flags = nla_get_u32(nla_fwd)
2915                                     & IP_VS_CONN_F_FWD_MASK;
2916                 udest->weight = nla_get_u32(nla_weight);
2917                 udest->u_threshold = nla_get_u32(nla_u_thresh);
2918                 udest->l_threshold = nla_get_u32(nla_l_thresh);
2919         }
2920
2921         return 0;
2922 }
2923
2924 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __be32 state,
2925                                   const char *mcast_ifn, __be32 syncid)
2926 {
2927         struct nlattr *nl_daemon;
2928
2929         nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
2930         if (!nl_daemon)
2931                 return -EMSGSIZE;
2932
2933         NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_STATE, state);
2934         NLA_PUT_STRING(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn);
2935         NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid);
2936
2937         nla_nest_end(skb, nl_daemon);
2938
2939         return 0;
2940
2941 nla_put_failure:
2942         nla_nest_cancel(skb, nl_daemon);
2943         return -EMSGSIZE;
2944 }
2945
2946 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __be32 state,
2947                                   const char *mcast_ifn, __be32 syncid,
2948                                   struct netlink_callback *cb)
2949 {
2950         void *hdr;
2951         hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2952                           &ip_vs_genl_family, NLM_F_MULTI,
2953                           IPVS_CMD_NEW_DAEMON);
2954         if (!hdr)
2955                 return -EMSGSIZE;
2956
2957         if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
2958                 goto nla_put_failure;
2959
2960         return genlmsg_end(skb, hdr);
2961
2962 nla_put_failure:
2963         genlmsg_cancel(skb, hdr);
2964         return -EMSGSIZE;
2965 }
2966
2967 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
2968                                    struct netlink_callback *cb)
2969 {
2970         mutex_lock(&__ip_vs_mutex);
2971         if ((ip_vs_sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
2972                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
2973                                            ip_vs_master_mcast_ifn,
2974                                            ip_vs_master_syncid, cb) < 0)
2975                         goto nla_put_failure;
2976
2977                 cb->args[0] = 1;
2978         }
2979
2980         if ((ip_vs_sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
2981                 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
2982                                            ip_vs_backup_mcast_ifn,
2983                                            ip_vs_backup_syncid, cb) < 0)
2984                         goto nla_put_failure;
2985
2986                 cb->args[1] = 1;
2987         }
2988
2989 nla_put_failure:
2990         mutex_unlock(&__ip_vs_mutex);
2991
2992         return skb->len;
2993 }
2994
2995 static int ip_vs_genl_new_daemon(struct nlattr **attrs)
2996 {
2997         if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
2998               attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
2999               attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3000                 return -EINVAL;
3001
3002         return start_sync_thread(nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3003                                  nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3004                                  nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3005 }
3006
3007 static int ip_vs_genl_del_daemon(struct nlattr **attrs)
3008 {
3009         if (!attrs[IPVS_DAEMON_ATTR_STATE])
3010                 return -EINVAL;
3011
3012         return stop_sync_thread(nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3013 }
3014
3015 static int ip_vs_genl_set_config(struct nlattr **attrs)
3016 {
3017         struct ip_vs_timeout_user t;
3018
3019         __ip_vs_get_timeouts(&t);
3020
3021         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3022                 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3023
3024         if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3025                 t.tcp_fin_timeout =
3026                         nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3027
3028         if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3029                 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3030
3031         return ip_vs_set_timeout(&t);
3032 }
3033
3034 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3035 {
3036         struct ip_vs_service *svc = NULL;
3037         struct ip_vs_service_user_kern usvc;
3038         struct ip_vs_dest_user_kern udest;
3039         int ret = 0, cmd;
3040         int need_full_svc = 0, need_full_dest = 0;
3041
3042         cmd = info->genlhdr->cmd;
3043
3044         mutex_lock(&__ip_vs_mutex);
3045
3046         if (cmd == IPVS_CMD_FLUSH) {
3047                 ret = ip_vs_flush();
3048                 goto out;
3049         } else if (cmd == IPVS_CMD_SET_CONFIG) {
3050                 ret = ip_vs_genl_set_config(info->attrs);
3051                 goto out;
3052         } else if (cmd == IPVS_CMD_NEW_DAEMON ||
3053                    cmd == IPVS_CMD_DEL_DAEMON) {
3054
3055                 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3056
3057                 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3058                     nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3059                                      info->attrs[IPVS_CMD_ATTR_DAEMON],
3060                                      ip_vs_daemon_policy)) {
3061                         ret = -EINVAL;
3062                         goto out;
3063                 }
3064
3065                 if (cmd == IPVS_CMD_NEW_DAEMON)
3066                         ret = ip_vs_genl_new_daemon(daemon_attrs);
3067                 else
3068                         ret = ip_vs_genl_del_daemon(daemon_attrs);
3069                 goto out;
3070         } else if (cmd == IPVS_CMD_ZERO &&
3071                    !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3072                 ret = ip_vs_zero_all();
3073                 goto out;
3074         }
3075
3076         /* All following commands require a service argument, so check if we
3077          * received a valid one. We need a full service specification when
3078          * adding / editing a service. Only identifying members otherwise. */
3079         if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3080                 need_full_svc = 1;
3081
3082         ret = ip_vs_genl_parse_service(&usvc,
3083                                        info->attrs[IPVS_CMD_ATTR_SERVICE],
3084                                        need_full_svc, &svc);
3085         if (ret)
3086                 goto out;
3087
3088         /* Unless we're adding a new service, the service must already exist */
3089         if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3090                 ret = -ESRCH;
3091                 goto out;
3092         }
3093
3094         /* Destination commands require a valid destination argument. For
3095          * adding / editing a destination, we need a full destination
3096          * specification. */
3097         if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3098             cmd == IPVS_CMD_DEL_DEST) {
3099                 if (cmd != IPVS_CMD_DEL_DEST)
3100                         need_full_dest = 1;
3101
3102                 ret = ip_vs_genl_parse_dest(&udest,
3103                                             info->attrs[IPVS_CMD_ATTR_DEST],
3104                                             need_full_dest);
3105                 if (ret)
3106                         goto out;
3107         }
3108
3109         switch (cmd) {
3110         case IPVS_CMD_NEW_SERVICE:
3111                 if (svc == NULL)
3112                         ret = ip_vs_add_service(&usvc, &svc);
3113                 else
3114                         ret = -EEXIST;
3115                 break;
3116         case IPVS_CMD_SET_SERVICE:
3117                 ret = ip_vs_edit_service(svc, &usvc);
3118                 break;
3119         case IPVS_CMD_DEL_SERVICE:
3120                 ret = ip_vs_del_service(svc);
3121                 /* do not use svc, it can be freed */
3122                 break;
3123         case IPVS_CMD_NEW_DEST:
3124                 ret = ip_vs_add_dest(svc, &udest);
3125                 break;
3126         case IPVS_CMD_SET_DEST:
3127                 ret = ip_vs_edit_dest(svc, &udest);
3128                 break;
3129         case IPVS_CMD_DEL_DEST:
3130                 ret = ip_vs_del_dest(svc, &udest);
3131                 break;
3132         case IPVS_CMD_ZERO:
3133                 ret = ip_vs_zero_service(svc);
3134                 break;
3135         default:
3136                 ret = -EINVAL;
3137         }
3138
3139 out:
3140         mutex_unlock(&__ip_vs_mutex);
3141
3142         return ret;
3143 }
3144
3145 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3146 {
3147         struct sk_buff *msg;
3148         void *reply;
3149         int ret, cmd, reply_cmd;
3150
3151         cmd = info->genlhdr->cmd;
3152
3153         if (cmd == IPVS_CMD_GET_SERVICE)
3154                 reply_cmd = IPVS_CMD_NEW_SERVICE;
3155         else if (cmd == IPVS_CMD_GET_INFO)
3156                 reply_cmd = IPVS_CMD_SET_INFO;
3157         else if (cmd == IPVS_CMD_GET_CONFIG)
3158                 reply_cmd = IPVS_CMD_SET_CONFIG;
3159         else {
3160                 pr_err("unknown Generic Netlink command\n");
3161                 return -EINVAL;
3162         }
3163
3164         msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3165         if (!msg)
3166                 return -ENOMEM;
3167
3168         mutex_lock(&__ip_vs_mutex);
3169
3170         reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3171         if (reply == NULL)
3172                 goto nla_put_failure;
3173
3174         switch (cmd) {
3175         case IPVS_CMD_GET_SERVICE:
3176         {
3177                 struct ip_vs_service *svc;
3178
3179                 svc = ip_vs_genl_find_service(info->attrs[IPVS_CMD_ATTR_SERVICE]);
3180                 if (IS_ERR(svc)) {
3181                         ret = PTR_ERR(svc);
3182                         goto out_err;
3183                 } else if (svc) {
3184                         ret = ip_vs_genl_fill_service(msg, svc);
3185                         if (ret)
3186                                 goto nla_put_failure;
3187                 } else {
3188                         ret = -ESRCH;
3189                         goto out_err;
3190                 }
3191
3192                 break;
3193         }
3194
3195         case IPVS_CMD_GET_CONFIG:
3196         {
3197                 struct ip_vs_timeout_user t;
3198
3199                 __ip_vs_get_timeouts(&t);
3200 #ifdef CONFIG_IP_VS_PROTO_TCP
3201                 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP, t.tcp_timeout);
3202                 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3203                             t.tcp_fin_timeout);
3204 #endif
3205 #ifdef CONFIG_IP_VS_PROTO_UDP
3206                 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout);
3207 #endif
3208
3209                 break;
3210         }
3211
3212         case IPVS_CMD_GET_INFO:
3213                 NLA_PUT_U32(msg, IPVS_INFO_ATTR_VERSION, IP_VS_VERSION_CODE);
3214                 NLA_PUT_U32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3215                             ip_vs_conn_tab_size);
3216                 break;
3217         }
3218
3219         genlmsg_end(msg, reply);
3220         ret = genlmsg_reply(msg, info);
3221         goto out;
3222
3223 nla_put_failure:
3224         pr_err("not enough space in Netlink message\n");
3225         ret = -EMSGSIZE;
3226
3227 out_err:
3228         nlmsg_free(msg);
3229 out:
3230         mutex_unlock(&__ip_vs_mutex);
3231
3232         return ret;
3233 }
3234
3235
3236 static struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3237         {
3238                 .cmd    = IPVS_CMD_NEW_SERVICE,
3239                 .flags  = GENL_ADMIN_PERM,
3240                 .policy = ip_vs_cmd_policy,
3241                 .doit   = ip_vs_genl_set_cmd,
3242         },
3243         {
3244                 .cmd    = IPVS_CMD_SET_SERVICE,
3245                 .flags  = GENL_ADMIN_PERM,
3246                 .policy = ip_vs_cmd_policy,
3247                 .doit   = ip_vs_genl_set_cmd,
3248         },
3249         {
3250                 .cmd    = IPVS_CMD_DEL_SERVICE,
3251                 .flags  = GENL_ADMIN_PERM,
3252                 .policy = ip_vs_cmd_policy,
3253                 .doit   = ip_vs_genl_set_cmd,
3254         },
3255         {
3256                 .cmd    = IPVS_CMD_GET_SERVICE,
3257                 .flags  = GENL_ADMIN_PERM,
3258                 .doit   = ip_vs_genl_get_cmd,
3259                 .dumpit = ip_vs_genl_dump_services,
3260                 .policy = ip_vs_cmd_policy,
3261         },
3262         {
3263                 .cmd    = IPVS_CMD_NEW_DEST,
3264                 .flags  = GENL_ADMIN_PERM,
3265                 .policy = ip_vs_cmd_policy,
3266                 .doit   = ip_vs_genl_set_cmd,
3267         },
3268         {
3269                 .cmd    = IPVS_CMD_SET_DEST,
3270                 .flags  = GENL_ADMIN_PERM,
3271                 .policy = ip_vs_cmd_policy,
3272                 .doit   = ip_vs_genl_set_cmd,
3273         },
3274         {
3275                 .cmd    = IPVS_CMD_DEL_DEST,
3276                 .flags  = GENL_ADMIN_PERM,
3277                 .policy = ip_vs_cmd_policy,
3278                 .doit   = ip_vs_genl_set_cmd,
3279         },
3280         {
3281                 .cmd    = IPVS_CMD_GET_DEST,
3282                 .flags  = GENL_ADMIN_PERM,
3283                 .policy = ip_vs_cmd_policy,
3284                 .dumpit = ip_vs_genl_dump_dests,
3285         },
3286         {
3287                 .cmd    = IPVS_CMD_NEW_DAEMON,
3288                 .flags  = GENL_ADMIN_PERM,
3289                 .policy = ip_vs_cmd_policy,
3290                 .doit   = ip_vs_genl_set_cmd,
3291         },
3292         {
3293                 .cmd    = IPVS_CMD_DEL_DAEMON,
3294                 .flags  = GENL_ADMIN_PERM,
3295                 .policy = ip_vs_cmd_policy,
3296                 .doit   = ip_vs_genl_set_cmd,
3297         },
3298         {
3299                 .cmd    = IPVS_CMD_GET_DAEMON,
3300                 .flags  = GENL_ADMIN_PERM,
3301                 .dumpit = ip_vs_genl_dump_daemons,
3302         },
3303         {
3304                 .cmd    = IPVS_CMD_SET_CONFIG,
3305                 .flags  = GENL_ADMIN_PERM,
3306                 .policy = ip_vs_cmd_policy,
3307                 .doit   = ip_vs_genl_set_cmd,
3308         },
3309         {
3310                 .cmd    = IPVS_CMD_GET_CONFIG,
3311                 .flags  = GENL_ADMIN_PERM,
3312                 .doit   = ip_vs_genl_get_cmd,
3313         },
3314         {
3315                 .cmd    = IPVS_CMD_GET_INFO,
3316                 .flags  = GENL_ADMIN_PERM,
3317                 .doit   = ip_vs_genl_get_cmd,
3318         },
3319         {
3320                 .cmd    = IPVS_CMD_ZERO,
3321                 .flags  = GENL_ADMIN_PERM,
3322                 .policy = ip_vs_cmd_policy,
3323                 .doit   = ip_vs_genl_set_cmd,
3324         },
3325         {
3326                 .cmd    = IPVS_CMD_FLUSH,
3327                 .flags  = GENL_ADMIN_PERM,
3328                 .doit   = ip_vs_genl_set_cmd,
3329         },
3330 };
3331
3332 static int __init ip_vs_genl_register(void)
3333 {
3334         return genl_register_family_with_ops(&ip_vs_genl_family,
3335                 ip_vs_genl_ops, ARRAY_SIZE(ip_vs_genl_ops));
3336 }
3337
3338 static void ip_vs_genl_unregister(void)
3339 {
3340         genl_unregister_family(&ip_vs_genl_family);
3341 }
3342
3343 /* End of Generic Netlink interface definitions */
3344
3345
3346 int __init ip_vs_control_init(void)
3347 {
3348         int ret;
3349         int idx;
3350
3351         EnterFunction(2);
3352
3353         ret = nf_register_sockopt(&ip_vs_sockopts);
3354         if (ret) {
3355                 pr_err("cannot register sockopt.\n");
3356                 return ret;
3357         }
3358
3359         ret = ip_vs_genl_register();
3360         if (ret) {
3361                 pr_err("cannot register Generic Netlink interface.\n");
3362                 nf_unregister_sockopt(&ip_vs_sockopts);
3363                 return ret;
3364         }
3365
3366         proc_net_fops_create(&init_net, "ip_vs", 0, &ip_vs_info_fops);
3367         proc_net_fops_create(&init_net, "ip_vs_stats",0, &ip_vs_stats_fops);
3368
3369         sysctl_header = register_sysctl_paths(net_vs_ctl_path, vs_vars);
3370
3371         /* Initialize ip_vs_svc_table, ip_vs_svc_fwm_table, ip_vs_rtable */
3372         for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++)  {
3373                 INIT_LIST_HEAD(&ip_vs_svc_table[idx]);
3374                 INIT_LIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3375         }
3376         for(idx = 0; idx < IP_VS_RTAB_SIZE; idx++)  {
3377                 INIT_LIST_HEAD(&ip_vs_rtable[idx]);
3378         }
3379
3380         ip_vs_new_estimator(&ip_vs_stats);
3381
3382         /* Hook the defense timer */
3383         schedule_delayed_work(&defense_work, DEFENSE_TIMER_PERIOD);
3384
3385         LeaveFunction(2);
3386         return 0;
3387 }
3388
3389
3390 void ip_vs_control_cleanup(void)
3391 {
3392         EnterFunction(2);
3393         ip_vs_trash_cleanup();
3394         cancel_rearming_delayed_work(&defense_work);
3395         cancel_work_sync(&defense_work.work);
3396         ip_vs_kill_estimator(&ip_vs_stats);
3397         unregister_sysctl_table(sysctl_header);
3398         proc_net_remove(&init_net, "ip_vs_stats");
3399         proc_net_remove(&init_net, "ip_vs");
3400         ip_vs_genl_unregister();
3401         nf_unregister_sockopt(&ip_vs_sockopts);
3402         LeaveFunction(2);
3403 }