1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
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 <linux/nsproxy.h>
42#include <net/ip.h>
43#ifdef CONFIG_IP_VS_IPV6
44#include <net/ipv6.h>
45#include <net/ip6_route.h>
46#endif
47#include <net/route.h>
48#include <net/sock.h>
49#include <net/genetlink.h>
50
51#include <asm/uaccess.h>
52
53#include <net/ip_vs.h>
54
55
56static DEFINE_MUTEX(__ip_vs_mutex);
57
58
59static DEFINE_RWLOCK(__ip_vs_svc_lock);
60
61
62
63#ifdef CONFIG_IP_VS_DEBUG
64static int sysctl_ip_vs_debug_level = 0;
65
66int ip_vs_get_debug_level(void)
67{
68 return sysctl_ip_vs_debug_level;
69}
70#endif
71
72
73
74static void __ip_vs_del_service(struct ip_vs_service *svc);
75
76
77#ifdef CONFIG_IP_VS_IPV6
78
79static int __ip_vs_addr_is_local_v6(struct net *net,
80 const struct in6_addr *addr)
81{
82 struct rt6_info *rt;
83 struct flowi6 fl6 = {
84 .daddr = *addr,
85 };
86
87 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
88 if (rt && rt->rt6i_dev && (rt->rt6i_dev->flags & IFF_LOOPBACK))
89 return 1;
90
91 return 0;
92}
93#endif
94
95#ifdef CONFIG_SYSCTL
96
97
98
99
100static void update_defense_level(struct netns_ipvs *ipvs)
101{
102 struct sysinfo i;
103 static int old_secure_tcp = 0;
104 int availmem;
105 int nomem;
106 int to_change = -1;
107
108
109 si_meminfo(&i);
110 availmem = i.freeram + i.bufferram;
111
112
113
114
115
116 nomem = (availmem < ipvs->sysctl_amemthresh);
117
118 local_bh_disable();
119
120
121 spin_lock(&ipvs->dropentry_lock);
122 switch (ipvs->sysctl_drop_entry) {
123 case 0:
124 atomic_set(&ipvs->dropentry, 0);
125 break;
126 case 1:
127 if (nomem) {
128 atomic_set(&ipvs->dropentry, 1);
129 ipvs->sysctl_drop_entry = 2;
130 } else {
131 atomic_set(&ipvs->dropentry, 0);
132 }
133 break;
134 case 2:
135 if (nomem) {
136 atomic_set(&ipvs->dropentry, 1);
137 } else {
138 atomic_set(&ipvs->dropentry, 0);
139 ipvs->sysctl_drop_entry = 1;
140 };
141 break;
142 case 3:
143 atomic_set(&ipvs->dropentry, 1);
144 break;
145 }
146 spin_unlock(&ipvs->dropentry_lock);
147
148
149 spin_lock(&ipvs->droppacket_lock);
150 switch (ipvs->sysctl_drop_packet) {
151 case 0:
152 ipvs->drop_rate = 0;
153 break;
154 case 1:
155 if (nomem) {
156 ipvs->drop_rate = ipvs->drop_counter
157 = ipvs->sysctl_amemthresh /
158 (ipvs->sysctl_amemthresh-availmem);
159 ipvs->sysctl_drop_packet = 2;
160 } else {
161 ipvs->drop_rate = 0;
162 }
163 break;
164 case 2:
165 if (nomem) {
166 ipvs->drop_rate = ipvs->drop_counter
167 = ipvs->sysctl_amemthresh /
168 (ipvs->sysctl_amemthresh-availmem);
169 } else {
170 ipvs->drop_rate = 0;
171 ipvs->sysctl_drop_packet = 1;
172 }
173 break;
174 case 3:
175 ipvs->drop_rate = ipvs->sysctl_am_droprate;
176 break;
177 }
178 spin_unlock(&ipvs->droppacket_lock);
179
180
181 spin_lock(&ipvs->securetcp_lock);
182 switch (ipvs->sysctl_secure_tcp) {
183 case 0:
184 if (old_secure_tcp >= 2)
185 to_change = 0;
186 break;
187 case 1:
188 if (nomem) {
189 if (old_secure_tcp < 2)
190 to_change = 1;
191 ipvs->sysctl_secure_tcp = 2;
192 } else {
193 if (old_secure_tcp >= 2)
194 to_change = 0;
195 }
196 break;
197 case 2:
198 if (nomem) {
199 if (old_secure_tcp < 2)
200 to_change = 1;
201 } else {
202 if (old_secure_tcp >= 2)
203 to_change = 0;
204 ipvs->sysctl_secure_tcp = 1;
205 }
206 break;
207 case 3:
208 if (old_secure_tcp < 2)
209 to_change = 1;
210 break;
211 }
212 old_secure_tcp = ipvs->sysctl_secure_tcp;
213 if (to_change >= 0)
214 ip_vs_protocol_timeout_change(ipvs,
215 ipvs->sysctl_secure_tcp > 1);
216 spin_unlock(&ipvs->securetcp_lock);
217
218 local_bh_enable();
219}
220
221
222
223
224
225#define DEFENSE_TIMER_PERIOD 1*HZ
226
227static void defense_work_handler(struct work_struct *work)
228{
229 struct netns_ipvs *ipvs =
230 container_of(work, struct netns_ipvs, defense_work.work);
231
232 update_defense_level(ipvs);
233 if (atomic_read(&ipvs->dropentry))
234 ip_vs_random_dropentry(ipvs->net);
235 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
236}
237#endif
238
239int
240ip_vs_use_count_inc(void)
241{
242 return try_module_get(THIS_MODULE);
243}
244
245void
246ip_vs_use_count_dec(void)
247{
248 module_put(THIS_MODULE);
249}
250
251
252
253
254
255#define IP_VS_SVC_TAB_BITS 8
256#define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
257#define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)
258
259
260static struct list_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE];
261
262static struct list_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE];
263
264
265
266
267
268static inline unsigned
269ip_vs_svc_hashkey(struct net *net, int af, unsigned proto,
270 const union nf_inet_addr *addr, __be16 port)
271{
272 register unsigned porth = ntohs(port);
273 __be32 addr_fold = addr->ip;
274
275#ifdef CONFIG_IP_VS_IPV6
276 if (af == AF_INET6)
277 addr_fold = addr->ip6[0]^addr->ip6[1]^
278 addr->ip6[2]^addr->ip6[3];
279#endif
280 addr_fold ^= ((size_t)net>>8);
281
282 return (proto^ntohl(addr_fold)^(porth>>IP_VS_SVC_TAB_BITS)^porth)
283 & IP_VS_SVC_TAB_MASK;
284}
285
286
287
288
289static inline unsigned ip_vs_svc_fwm_hashkey(struct net *net, __u32 fwmark)
290{
291 return (((size_t)net>>8) ^ fwmark) & IP_VS_SVC_TAB_MASK;
292}
293
294
295
296
297
298
299static int ip_vs_svc_hash(struct ip_vs_service *svc)
300{
301 unsigned hash;
302
303 if (svc->flags & IP_VS_SVC_F_HASHED) {
304 pr_err("%s(): request for already hashed, called from %pF\n",
305 __func__, __builtin_return_address(0));
306 return 0;
307 }
308
309 if (svc->fwmark == 0) {
310
311
312
313 hash = ip_vs_svc_hashkey(svc->net, svc->af, svc->protocol,
314 &svc->addr, svc->port);
315 list_add(&svc->s_list, &ip_vs_svc_table[hash]);
316 } else {
317
318
319
320 hash = ip_vs_svc_fwm_hashkey(svc->net, svc->fwmark);
321 list_add(&svc->f_list, &ip_vs_svc_fwm_table[hash]);
322 }
323
324 svc->flags |= IP_VS_SVC_F_HASHED;
325
326 atomic_inc(&svc->refcnt);
327 return 1;
328}
329
330
331
332
333
334
335static int ip_vs_svc_unhash(struct ip_vs_service *svc)
336{
337 if (!(svc->flags & IP_VS_SVC_F_HASHED)) {
338 pr_err("%s(): request for unhash flagged, called from %pF\n",
339 __func__, __builtin_return_address(0));
340 return 0;
341 }
342
343 if (svc->fwmark == 0) {
344
345 list_del(&svc->s_list);
346 } else {
347
348 list_del(&svc->f_list);
349 }
350
351 svc->flags &= ~IP_VS_SVC_F_HASHED;
352 atomic_dec(&svc->refcnt);
353 return 1;
354}
355
356
357
358
359
360static inline struct ip_vs_service *
361__ip_vs_service_find(struct net *net, int af, __u16 protocol,
362 const union nf_inet_addr *vaddr, __be16 vport)
363{
364 unsigned hash;
365 struct ip_vs_service *svc;
366
367
368 hash = ip_vs_svc_hashkey(net, af, protocol, vaddr, vport);
369
370 list_for_each_entry(svc, &ip_vs_svc_table[hash], s_list){
371 if ((svc->af == af)
372 && ip_vs_addr_equal(af, &svc->addr, vaddr)
373 && (svc->port == vport)
374 && (svc->protocol == protocol)
375 && net_eq(svc->net, net)) {
376
377 return svc;
378 }
379 }
380
381 return NULL;
382}
383
384
385
386
387
388static inline struct ip_vs_service *
389__ip_vs_svc_fwm_find(struct net *net, int af, __u32 fwmark)
390{
391 unsigned hash;
392 struct ip_vs_service *svc;
393
394
395 hash = ip_vs_svc_fwm_hashkey(net, fwmark);
396
397 list_for_each_entry(svc, &ip_vs_svc_fwm_table[hash], f_list) {
398 if (svc->fwmark == fwmark && svc->af == af
399 && net_eq(svc->net, net)) {
400
401 return svc;
402 }
403 }
404
405 return NULL;
406}
407
408struct ip_vs_service *
409ip_vs_service_get(struct net *net, int af, __u32 fwmark, __u16 protocol,
410 const union nf_inet_addr *vaddr, __be16 vport)
411{
412 struct ip_vs_service *svc;
413 struct netns_ipvs *ipvs = net_ipvs(net);
414
415 read_lock(&__ip_vs_svc_lock);
416
417
418
419
420 if (fwmark) {
421 svc = __ip_vs_svc_fwm_find(net, af, fwmark);
422 if (svc)
423 goto out;
424 }
425
426
427
428
429
430 svc = __ip_vs_service_find(net, af, protocol, vaddr, vport);
431
432 if (svc == NULL
433 && protocol == IPPROTO_TCP
434 && atomic_read(&ipvs->ftpsvc_counter)
435 && (vport == FTPDATA || ntohs(vport) >= PROT_SOCK)) {
436
437
438
439
440 svc = __ip_vs_service_find(net, af, protocol, vaddr, FTPPORT);
441 }
442
443 if (svc == NULL
444 && atomic_read(&ipvs->nullsvc_counter)) {
445
446
447
448 svc = __ip_vs_service_find(net, af, protocol, vaddr, 0);
449 }
450
451 out:
452 if (svc)
453 atomic_inc(&svc->usecnt);
454 read_unlock(&__ip_vs_svc_lock);
455
456 IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
457 fwmark, ip_vs_proto_name(protocol),
458 IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
459 svc ? "hit" : "not hit");
460
461 return svc;
462}
463
464
465static inline void
466__ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
467{
468 atomic_inc(&svc->refcnt);
469 dest->svc = svc;
470}
471
472static void
473__ip_vs_unbind_svc(struct ip_vs_dest *dest)
474{
475 struct ip_vs_service *svc = dest->svc;
476
477 dest->svc = NULL;
478 if (atomic_dec_and_test(&svc->refcnt)) {
479 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
480 svc->fwmark,
481 IP_VS_DBG_ADDR(svc->af, &svc->addr),
482 ntohs(svc->port), atomic_read(&svc->usecnt));
483 free_percpu(svc->stats.cpustats);
484 kfree(svc);
485 }
486}
487
488
489
490
491
492static inline unsigned ip_vs_rs_hashkey(int af,
493 const union nf_inet_addr *addr,
494 __be16 port)
495{
496 register unsigned porth = ntohs(port);
497 __be32 addr_fold = addr->ip;
498
499#ifdef CONFIG_IP_VS_IPV6
500 if (af == AF_INET6)
501 addr_fold = addr->ip6[0]^addr->ip6[1]^
502 addr->ip6[2]^addr->ip6[3];
503#endif
504
505 return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
506 & IP_VS_RTAB_MASK;
507}
508
509
510
511
512
513static int ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
514{
515 unsigned hash;
516
517 if (!list_empty(&dest->d_list)) {
518 return 0;
519 }
520
521
522
523
524
525 hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
526
527 list_add(&dest->d_list, &ipvs->rs_table[hash]);
528
529 return 1;
530}
531
532
533
534
535
536static int ip_vs_rs_unhash(struct ip_vs_dest *dest)
537{
538
539
540
541 if (!list_empty(&dest->d_list)) {
542 list_del(&dest->d_list);
543 INIT_LIST_HEAD(&dest->d_list);
544 }
545
546 return 1;
547}
548
549
550
551
552struct ip_vs_dest *
553ip_vs_lookup_real_service(struct net *net, int af, __u16 protocol,
554 const union nf_inet_addr *daddr,
555 __be16 dport)
556{
557 struct netns_ipvs *ipvs = net_ipvs(net);
558 unsigned hash;
559 struct ip_vs_dest *dest;
560
561
562
563
564
565 hash = ip_vs_rs_hashkey(af, daddr, dport);
566
567 read_lock(&ipvs->rs_lock);
568 list_for_each_entry(dest, &ipvs->rs_table[hash], d_list) {
569 if ((dest->af == af)
570 && ip_vs_addr_equal(af, &dest->addr, daddr)
571 && (dest->port == dport)
572 && ((dest->protocol == protocol) ||
573 dest->vfwmark)) {
574
575 read_unlock(&ipvs->rs_lock);
576 return dest;
577 }
578 }
579 read_unlock(&ipvs->rs_lock);
580
581 return NULL;
582}
583
584
585
586
587static struct ip_vs_dest *
588ip_vs_lookup_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
589 __be16 dport)
590{
591 struct ip_vs_dest *dest;
592
593
594
595
596 list_for_each_entry(dest, &svc->destinations, n_list) {
597 if ((dest->af == svc->af)
598 && ip_vs_addr_equal(svc->af, &dest->addr, daddr)
599 && (dest->port == dport)) {
600
601 return dest;
602 }
603 }
604
605 return NULL;
606}
607
608
609
610
611
612
613
614
615
616
617
618struct ip_vs_dest *ip_vs_find_dest(struct net *net, int af,
619 const union nf_inet_addr *daddr,
620 __be16 dport,
621 const union nf_inet_addr *vaddr,
622 __be16 vport, __u16 protocol, __u32 fwmark,
623 __u32 flags)
624{
625 struct ip_vs_dest *dest;
626 struct ip_vs_service *svc;
627 __be16 port = dport;
628
629 svc = ip_vs_service_get(net, af, fwmark, protocol, vaddr, vport);
630 if (!svc)
631 return NULL;
632 if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
633 port = 0;
634 dest = ip_vs_lookup_dest(svc, daddr, port);
635 if (!dest)
636 dest = ip_vs_lookup_dest(svc, daddr, port ^ dport);
637 if (dest)
638 atomic_inc(&dest->refcnt);
639 ip_vs_service_put(svc);
640 return dest;
641}
642
643
644
645
646
647
648
649
650
651
652
653static struct ip_vs_dest *
654ip_vs_trash_get_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
655 __be16 dport)
656{
657 struct ip_vs_dest *dest, *nxt;
658 struct netns_ipvs *ipvs = net_ipvs(svc->net);
659
660
661
662
663 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, n_list) {
664 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
665 "dest->refcnt=%d\n",
666 dest->vfwmark,
667 IP_VS_DBG_ADDR(svc->af, &dest->addr),
668 ntohs(dest->port),
669 atomic_read(&dest->refcnt));
670 if (dest->af == svc->af &&
671 ip_vs_addr_equal(svc->af, &dest->addr, daddr) &&
672 dest->port == dport &&
673 dest->vfwmark == svc->fwmark &&
674 dest->protocol == svc->protocol &&
675 (svc->fwmark ||
676 (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
677 dest->vport == svc->port))) {
678
679 return dest;
680 }
681
682
683
684
685 if (atomic_read(&dest->refcnt) == 1) {
686 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u "
687 "from trash\n",
688 dest->vfwmark,
689 IP_VS_DBG_ADDR(svc->af, &dest->addr),
690 ntohs(dest->port));
691 list_del(&dest->n_list);
692 ip_vs_dst_reset(dest);
693 __ip_vs_unbind_svc(dest);
694 free_percpu(dest->stats.cpustats);
695 kfree(dest);
696 }
697 }
698
699 return NULL;
700}
701
702
703
704
705
706
707
708
709
710
711
712static void ip_vs_trash_cleanup(struct net *net)
713{
714 struct ip_vs_dest *dest, *nxt;
715 struct netns_ipvs *ipvs = net_ipvs(net);
716
717 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, n_list) {
718 list_del(&dest->n_list);
719 ip_vs_dst_reset(dest);
720 __ip_vs_unbind_svc(dest);
721 free_percpu(dest->stats.cpustats);
722 kfree(dest);
723 }
724}
725
726static void
727ip_vs_copy_stats(struct ip_vs_stats_user *dst, struct ip_vs_stats *src)
728{
729#define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->ustats.c - src->ustats0.c
730
731 spin_lock_bh(&src->lock);
732
733 IP_VS_SHOW_STATS_COUNTER(conns);
734 IP_VS_SHOW_STATS_COUNTER(inpkts);
735 IP_VS_SHOW_STATS_COUNTER(outpkts);
736 IP_VS_SHOW_STATS_COUNTER(inbytes);
737 IP_VS_SHOW_STATS_COUNTER(outbytes);
738
739 ip_vs_read_estimator(dst, src);
740
741 spin_unlock_bh(&src->lock);
742}
743
744static void
745ip_vs_zero_stats(struct ip_vs_stats *stats)
746{
747 spin_lock_bh(&stats->lock);
748
749
750
751#define IP_VS_ZERO_STATS_COUNTER(c) stats->ustats0.c = stats->ustats.c
752
753 IP_VS_ZERO_STATS_COUNTER(conns);
754 IP_VS_ZERO_STATS_COUNTER(inpkts);
755 IP_VS_ZERO_STATS_COUNTER(outpkts);
756 IP_VS_ZERO_STATS_COUNTER(inbytes);
757 IP_VS_ZERO_STATS_COUNTER(outbytes);
758
759 ip_vs_zero_estimator(stats);
760
761 spin_unlock_bh(&stats->lock);
762}
763
764
765
766
767static void
768__ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
769 struct ip_vs_dest_user_kern *udest, int add)
770{
771 struct netns_ipvs *ipvs = net_ipvs(svc->net);
772 int conn_flags;
773
774
775 atomic_set(&dest->weight, udest->weight);
776 conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
777 conn_flags |= IP_VS_CONN_F_INACTIVE;
778
779
780 if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
781 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
782 } else {
783
784
785
786
787 write_lock_bh(&ipvs->rs_lock);
788 ip_vs_rs_hash(ipvs, dest);
789 write_unlock_bh(&ipvs->rs_lock);
790 }
791 atomic_set(&dest->conn_flags, conn_flags);
792
793
794 if (!dest->svc) {
795 __ip_vs_bind_svc(dest, svc);
796 } else {
797 if (dest->svc != svc) {
798 __ip_vs_unbind_svc(dest);
799 ip_vs_zero_stats(&dest->stats);
800 __ip_vs_bind_svc(dest, svc);
801 }
802 }
803
804
805 dest->flags |= IP_VS_DEST_F_AVAILABLE;
806
807 if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
808 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
809 dest->u_threshold = udest->u_threshold;
810 dest->l_threshold = udest->l_threshold;
811
812 spin_lock_bh(&dest->dst_lock);
813 ip_vs_dst_reset(dest);
814 spin_unlock_bh(&dest->dst_lock);
815
816 if (add)
817 ip_vs_start_estimator(svc->net, &dest->stats);
818
819 write_lock_bh(&__ip_vs_svc_lock);
820
821
822 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
823
824 if (add) {
825 list_add(&dest->n_list, &svc->destinations);
826 svc->num_dests++;
827 }
828
829
830 if (svc->scheduler->update_service)
831 svc->scheduler->update_service(svc);
832
833 write_unlock_bh(&__ip_vs_svc_lock);
834}
835
836
837
838
839
840static int
841ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
842 struct ip_vs_dest **dest_p)
843{
844 struct ip_vs_dest *dest;
845 unsigned atype;
846
847 EnterFunction(2);
848
849#ifdef CONFIG_IP_VS_IPV6
850 if (svc->af == AF_INET6) {
851 atype = ipv6_addr_type(&udest->addr.in6);
852 if ((!(atype & IPV6_ADDR_UNICAST) ||
853 atype & IPV6_ADDR_LINKLOCAL) &&
854 !__ip_vs_addr_is_local_v6(svc->net, &udest->addr.in6))
855 return -EINVAL;
856 } else
857#endif
858 {
859 atype = inet_addr_type(svc->net, udest->addr.ip);
860 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
861 return -EINVAL;
862 }
863
864 dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
865 if (dest == NULL)
866 return -ENOMEM;
867
868 dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
869 if (!dest->stats.cpustats)
870 goto err_alloc;
871
872 dest->af = svc->af;
873 dest->protocol = svc->protocol;
874 dest->vaddr = svc->addr;
875 dest->vport = svc->port;
876 dest->vfwmark = svc->fwmark;
877 ip_vs_addr_copy(svc->af, &dest->addr, &udest->addr);
878 dest->port = udest->port;
879
880 atomic_set(&dest->activeconns, 0);
881 atomic_set(&dest->inactconns, 0);
882 atomic_set(&dest->persistconns, 0);
883 atomic_set(&dest->refcnt, 1);
884
885 INIT_LIST_HEAD(&dest->d_list);
886 spin_lock_init(&dest->dst_lock);
887 spin_lock_init(&dest->stats.lock);
888 __ip_vs_update_dest(svc, dest, udest, 1);
889
890 *dest_p = dest;
891
892 LeaveFunction(2);
893 return 0;
894
895err_alloc:
896 kfree(dest);
897 return -ENOMEM;
898}
899
900
901
902
903
904static int
905ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
906{
907 struct ip_vs_dest *dest;
908 union nf_inet_addr daddr;
909 __be16 dport = udest->port;
910 int ret;
911
912 EnterFunction(2);
913
914 if (udest->weight < 0) {
915 pr_err("%s(): server weight less than zero\n", __func__);
916 return -ERANGE;
917 }
918
919 if (udest->l_threshold > udest->u_threshold) {
920 pr_err("%s(): lower threshold is higher than upper threshold\n",
921 __func__);
922 return -ERANGE;
923 }
924
925 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
926
927
928
929
930 dest = ip_vs_lookup_dest(svc, &daddr, dport);
931
932 if (dest != NULL) {
933 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
934 return -EEXIST;
935 }
936
937
938
939
940
941 dest = ip_vs_trash_get_dest(svc, &daddr, dport);
942
943 if (dest != NULL) {
944 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
945 "dest->refcnt=%d, service %u/%s:%u\n",
946 IP_VS_DBG_ADDR(svc->af, &daddr), ntohs(dport),
947 atomic_read(&dest->refcnt),
948 dest->vfwmark,
949 IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
950 ntohs(dest->vport));
951
952
953
954
955 list_del(&dest->n_list);
956
957 __ip_vs_update_dest(svc, dest, udest, 1);
958 ret = 0;
959 } else {
960
961
962
963 ret = ip_vs_new_dest(svc, udest, &dest);
964 }
965 LeaveFunction(2);
966
967 return ret;
968}
969
970
971
972
973
974static int
975ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
976{
977 struct ip_vs_dest *dest;
978 union nf_inet_addr daddr;
979 __be16 dport = udest->port;
980
981 EnterFunction(2);
982
983 if (udest->weight < 0) {
984 pr_err("%s(): server weight less than zero\n", __func__);
985 return -ERANGE;
986 }
987
988 if (udest->l_threshold > udest->u_threshold) {
989 pr_err("%s(): lower threshold is higher than upper threshold\n",
990 __func__);
991 return -ERANGE;
992 }
993
994 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
995
996
997
998
999 dest = ip_vs_lookup_dest(svc, &daddr, dport);
1000
1001 if (dest == NULL) {
1002 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1003 return -ENOENT;
1004 }
1005
1006 __ip_vs_update_dest(svc, dest, udest, 0);
1007 LeaveFunction(2);
1008
1009 return 0;
1010}
1011
1012
1013
1014
1015
1016static void __ip_vs_del_dest(struct net *net, struct ip_vs_dest *dest)
1017{
1018 struct netns_ipvs *ipvs = net_ipvs(net);
1019
1020 ip_vs_stop_estimator(net, &dest->stats);
1021
1022
1023
1024
1025 write_lock_bh(&ipvs->rs_lock);
1026 ip_vs_rs_unhash(dest);
1027 write_unlock_bh(&ipvs->rs_lock);
1028
1029
1030
1031
1032
1033
1034 if (atomic_dec_and_test(&dest->refcnt)) {
1035 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u\n",
1036 dest->vfwmark,
1037 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1038 ntohs(dest->port));
1039 ip_vs_dst_reset(dest);
1040
1041
1042
1043
1044
1045 atomic_dec(&dest->svc->refcnt);
1046 free_percpu(dest->stats.cpustats);
1047 kfree(dest);
1048 } else {
1049 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, "
1050 "dest->refcnt=%d\n",
1051 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1052 ntohs(dest->port),
1053 atomic_read(&dest->refcnt));
1054 list_add(&dest->n_list, &ipvs->dest_trash);
1055 atomic_inc(&dest->refcnt);
1056 }
1057}
1058
1059
1060
1061
1062
1063static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1064 struct ip_vs_dest *dest,
1065 int svcupd)
1066{
1067 dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1068
1069
1070
1071
1072 list_del(&dest->n_list);
1073 svc->num_dests--;
1074
1075
1076
1077
1078 if (svcupd && svc->scheduler->update_service)
1079 svc->scheduler->update_service(svc);
1080}
1081
1082
1083
1084
1085
1086static int
1087ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1088{
1089 struct ip_vs_dest *dest;
1090 __be16 dport = udest->port;
1091
1092 EnterFunction(2);
1093
1094 dest = ip_vs_lookup_dest(svc, &udest->addr, dport);
1095
1096 if (dest == NULL) {
1097 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1098 return -ENOENT;
1099 }
1100
1101 write_lock_bh(&__ip_vs_svc_lock);
1102
1103
1104
1105
1106 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1107
1108
1109
1110
1111 __ip_vs_unlink_dest(svc, dest, 1);
1112
1113 write_unlock_bh(&__ip_vs_svc_lock);
1114
1115
1116
1117
1118 __ip_vs_del_dest(svc->net, dest);
1119
1120 LeaveFunction(2);
1121
1122 return 0;
1123}
1124
1125
1126
1127
1128
1129static int
1130ip_vs_add_service(struct net *net, struct ip_vs_service_user_kern *u,
1131 struct ip_vs_service **svc_p)
1132{
1133 int ret = 0;
1134 struct ip_vs_scheduler *sched = NULL;
1135 struct ip_vs_pe *pe = NULL;
1136 struct ip_vs_service *svc = NULL;
1137 struct netns_ipvs *ipvs = net_ipvs(net);
1138
1139
1140 ip_vs_use_count_inc();
1141
1142
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 if (u->pe_name && *u->pe_name) {
1151 pe = ip_vs_pe_getbyname(u->pe_name);
1152 if (pe == NULL) {
1153 pr_info("persistence engine module ip_vs_pe_%s "
1154 "not found\n", u->pe_name);
1155 ret = -ENOENT;
1156 goto out_err;
1157 }
1158 }
1159
1160#ifdef CONFIG_IP_VS_IPV6
1161 if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1162 ret = -EINVAL;
1163 goto out_err;
1164 }
1165#endif
1166
1167 svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1168 if (svc == NULL) {
1169 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1170 ret = -ENOMEM;
1171 goto out_err;
1172 }
1173 svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1174 if (!svc->stats.cpustats)
1175 goto out_err;
1176
1177
1178 atomic_set(&svc->usecnt, 0);
1179 atomic_set(&svc->refcnt, 0);
1180
1181 svc->af = u->af;
1182 svc->protocol = u->protocol;
1183 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1184 svc->port = u->port;
1185 svc->fwmark = u->fwmark;
1186 svc->flags = u->flags;
1187 svc->timeout = u->timeout * HZ;
1188 svc->netmask = u->netmask;
1189 svc->net = net;
1190
1191 INIT_LIST_HEAD(&svc->destinations);
1192 rwlock_init(&svc->sched_lock);
1193 spin_lock_init(&svc->stats.lock);
1194
1195
1196 ret = ip_vs_bind_scheduler(svc, sched);
1197 if (ret)
1198 goto out_err;
1199 sched = NULL;
1200
1201
1202 ip_vs_bind_pe(svc, pe);
1203 pe = NULL;
1204
1205
1206 if (svc->port == FTPPORT)
1207 atomic_inc(&ipvs->ftpsvc_counter);
1208 else if (svc->port == 0)
1209 atomic_inc(&ipvs->nullsvc_counter);
1210
1211 ip_vs_start_estimator(net, &svc->stats);
1212
1213
1214 if (svc->af == AF_INET)
1215 ipvs->num_services++;
1216
1217
1218 write_lock_bh(&__ip_vs_svc_lock);
1219 ip_vs_svc_hash(svc);
1220 write_unlock_bh(&__ip_vs_svc_lock);
1221
1222 *svc_p = svc;
1223
1224 ipvs->enable = 1;
1225 return 0;
1226
1227
1228 out_err:
1229 if (svc != NULL) {
1230 ip_vs_unbind_scheduler(svc);
1231 if (svc->inc) {
1232 local_bh_disable();
1233 ip_vs_app_inc_put(svc->inc);
1234 local_bh_enable();
1235 }
1236 if (svc->stats.cpustats)
1237 free_percpu(svc->stats.cpustats);
1238 kfree(svc);
1239 }
1240 ip_vs_scheduler_put(sched);
1241 ip_vs_pe_put(pe);
1242
1243
1244 ip_vs_use_count_dec();
1245
1246 return ret;
1247}
1248
1249
1250
1251
1252
1253static int
1254ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1255{
1256 struct ip_vs_scheduler *sched, *old_sched;
1257 struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1258 int ret = 0;
1259
1260
1261
1262
1263 sched = ip_vs_scheduler_get(u->sched_name);
1264 if (sched == NULL) {
1265 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1266 return -ENOENT;
1267 }
1268 old_sched = sched;
1269
1270 if (u->pe_name && *u->pe_name) {
1271 pe = ip_vs_pe_getbyname(u->pe_name);
1272 if (pe == NULL) {
1273 pr_info("persistence engine module ip_vs_pe_%s "
1274 "not found\n", u->pe_name);
1275 ret = -ENOENT;
1276 goto out;
1277 }
1278 old_pe = pe;
1279 }
1280
1281#ifdef CONFIG_IP_VS_IPV6
1282 if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1283 ret = -EINVAL;
1284 goto out;
1285 }
1286#endif
1287
1288 write_lock_bh(&__ip_vs_svc_lock);
1289
1290
1291
1292
1293 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1294
1295
1296
1297
1298 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1299 svc->timeout = u->timeout * HZ;
1300 svc->netmask = u->netmask;
1301
1302 old_sched = svc->scheduler;
1303 if (sched != old_sched) {
1304
1305
1306
1307 if ((ret = ip_vs_unbind_scheduler(svc))) {
1308 old_sched = sched;
1309 goto out_unlock;
1310 }
1311
1312
1313
1314
1315 if ((ret = ip_vs_bind_scheduler(svc, sched))) {
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326 ip_vs_bind_scheduler(svc, old_sched);
1327 old_sched = sched;
1328 goto out_unlock;
1329 }
1330 }
1331
1332 old_pe = svc->pe;
1333 if (pe != old_pe) {
1334 ip_vs_unbind_pe(svc);
1335 ip_vs_bind_pe(svc, pe);
1336 }
1337
1338out_unlock:
1339 write_unlock_bh(&__ip_vs_svc_lock);
1340out:
1341 ip_vs_scheduler_put(old_sched);
1342 ip_vs_pe_put(old_pe);
1343 return ret;
1344}
1345
1346
1347
1348
1349
1350
1351
1352static void __ip_vs_del_service(struct ip_vs_service *svc)
1353{
1354 struct ip_vs_dest *dest, *nxt;
1355 struct ip_vs_scheduler *old_sched;
1356 struct ip_vs_pe *old_pe;
1357 struct netns_ipvs *ipvs = net_ipvs(svc->net);
1358
1359 pr_info("%s: enter\n", __func__);
1360
1361
1362 if (svc->af == AF_INET)
1363 ipvs->num_services--;
1364
1365 ip_vs_stop_estimator(svc->net, &svc->stats);
1366
1367
1368 old_sched = svc->scheduler;
1369 ip_vs_unbind_scheduler(svc);
1370 ip_vs_scheduler_put(old_sched);
1371
1372
1373 old_pe = svc->pe;
1374 ip_vs_unbind_pe(svc);
1375 ip_vs_pe_put(old_pe);
1376
1377
1378 if (svc->inc) {
1379 ip_vs_app_inc_put(svc->inc);
1380 svc->inc = NULL;
1381 }
1382
1383
1384
1385
1386 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1387 __ip_vs_unlink_dest(svc, dest, 0);
1388 __ip_vs_del_dest(svc->net, dest);
1389 }
1390
1391
1392
1393
1394 if (svc->port == FTPPORT)
1395 atomic_dec(&ipvs->ftpsvc_counter);
1396 else if (svc->port == 0)
1397 atomic_dec(&ipvs->nullsvc_counter);
1398
1399
1400
1401
1402 if (atomic_read(&svc->refcnt) == 0) {
1403 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
1404 svc->fwmark,
1405 IP_VS_DBG_ADDR(svc->af, &svc->addr),
1406 ntohs(svc->port), atomic_read(&svc->usecnt));
1407 free_percpu(svc->stats.cpustats);
1408 kfree(svc);
1409 }
1410
1411
1412 ip_vs_use_count_dec();
1413}
1414
1415
1416
1417
1418static void ip_vs_unlink_service(struct ip_vs_service *svc)
1419{
1420
1421
1422
1423 write_lock_bh(&__ip_vs_svc_lock);
1424
1425 ip_vs_svc_unhash(svc);
1426
1427
1428
1429
1430 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1431
1432 __ip_vs_del_service(svc);
1433
1434 write_unlock_bh(&__ip_vs_svc_lock);
1435}
1436
1437
1438
1439
1440static int ip_vs_del_service(struct ip_vs_service *svc)
1441{
1442 if (svc == NULL)
1443 return -EEXIST;
1444 ip_vs_unlink_service(svc);
1445
1446 return 0;
1447}
1448
1449
1450
1451
1452
1453static int ip_vs_flush(struct net *net)
1454{
1455 int idx;
1456 struct ip_vs_service *svc, *nxt;
1457
1458
1459
1460
1461 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1462 list_for_each_entry_safe(svc, nxt, &ip_vs_svc_table[idx],
1463 s_list) {
1464 if (net_eq(svc->net, net))
1465 ip_vs_unlink_service(svc);
1466 }
1467 }
1468
1469
1470
1471
1472 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1473 list_for_each_entry_safe(svc, nxt,
1474 &ip_vs_svc_fwm_table[idx], f_list) {
1475 if (net_eq(svc->net, net))
1476 ip_vs_unlink_service(svc);
1477 }
1478 }
1479
1480 return 0;
1481}
1482
1483
1484
1485
1486
1487void ip_vs_service_net_cleanup(struct net *net)
1488{
1489 EnterFunction(2);
1490
1491 mutex_lock(&__ip_vs_mutex);
1492 ip_vs_flush(net);
1493 mutex_unlock(&__ip_vs_mutex);
1494 LeaveFunction(2);
1495}
1496
1497
1498
1499static inline void
1500__ip_vs_dev_reset(struct ip_vs_dest *dest, struct net_device *dev)
1501{
1502 spin_lock_bh(&dest->dst_lock);
1503 if (dest->dst_cache && dest->dst_cache->dev == dev) {
1504 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1505 dev->name,
1506 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1507 ntohs(dest->port),
1508 atomic_read(&dest->refcnt));
1509 ip_vs_dst_reset(dest);
1510 }
1511 spin_unlock_bh(&dest->dst_lock);
1512
1513}
1514
1515
1516
1517
1518
1519static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1520 void *ptr)
1521{
1522 struct net_device *dev = ptr;
1523 struct net *net = dev_net(dev);
1524 struct ip_vs_service *svc;
1525 struct ip_vs_dest *dest;
1526 unsigned int idx;
1527
1528 if (event != NETDEV_UNREGISTER)
1529 return NOTIFY_DONE;
1530 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1531 EnterFunction(2);
1532 mutex_lock(&__ip_vs_mutex);
1533 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1534 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1535 if (net_eq(svc->net, net)) {
1536 list_for_each_entry(dest, &svc->destinations,
1537 n_list) {
1538 __ip_vs_dev_reset(dest, dev);
1539 }
1540 }
1541 }
1542
1543 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1544 if (net_eq(svc->net, net)) {
1545 list_for_each_entry(dest, &svc->destinations,
1546 n_list) {
1547 __ip_vs_dev_reset(dest, dev);
1548 }
1549 }
1550
1551 }
1552 }
1553
1554 list_for_each_entry(dest, &net_ipvs(net)->dest_trash, n_list) {
1555 __ip_vs_dev_reset(dest, dev);
1556 }
1557 mutex_unlock(&__ip_vs_mutex);
1558 LeaveFunction(2);
1559 return NOTIFY_DONE;
1560}
1561
1562
1563
1564
1565static int ip_vs_zero_service(struct ip_vs_service *svc)
1566{
1567 struct ip_vs_dest *dest;
1568
1569 write_lock_bh(&__ip_vs_svc_lock);
1570 list_for_each_entry(dest, &svc->destinations, n_list) {
1571 ip_vs_zero_stats(&dest->stats);
1572 }
1573 ip_vs_zero_stats(&svc->stats);
1574 write_unlock_bh(&__ip_vs_svc_lock);
1575 return 0;
1576}
1577
1578static int ip_vs_zero_all(struct net *net)
1579{
1580 int idx;
1581 struct ip_vs_service *svc;
1582
1583 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1584 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1585 if (net_eq(svc->net, net))
1586 ip_vs_zero_service(svc);
1587 }
1588 }
1589
1590 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1591 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1592 if (net_eq(svc->net, net))
1593 ip_vs_zero_service(svc);
1594 }
1595 }
1596
1597 ip_vs_zero_stats(&net_ipvs(net)->tot_stats);
1598 return 0;
1599}
1600
1601#ifdef CONFIG_SYSCTL
1602static int
1603proc_do_defense_mode(ctl_table *table, int write,
1604 void __user *buffer, size_t *lenp, loff_t *ppos)
1605{
1606 struct net *net = current->nsproxy->net_ns;
1607 int *valp = table->data;
1608 int val = *valp;
1609 int rc;
1610
1611 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1612 if (write && (*valp != val)) {
1613 if ((*valp < 0) || (*valp > 3)) {
1614
1615 *valp = val;
1616 } else {
1617 update_defense_level(net_ipvs(net));
1618 }
1619 }
1620 return rc;
1621}
1622
1623static int
1624proc_do_sync_threshold(ctl_table *table, int write,
1625 void __user *buffer, size_t *lenp, loff_t *ppos)
1626{
1627 int *valp = table->data;
1628 int val[2];
1629 int rc;
1630
1631
1632 memcpy(val, valp, sizeof(val));
1633
1634 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1635 if (write && (valp[0] < 0 || valp[1] < 0 || valp[0] >= valp[1])) {
1636
1637 memcpy(valp, val, sizeof(val));
1638 }
1639 return rc;
1640}
1641
1642static int
1643proc_do_sync_mode(ctl_table *table, int write,
1644 void __user *buffer, size_t *lenp, loff_t *ppos)
1645{
1646 int *valp = table->data;
1647 int val = *valp;
1648 int rc;
1649
1650 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1651 if (write && (*valp != val)) {
1652 if ((*valp < 0) || (*valp > 1)) {
1653
1654 *valp = val;
1655 } else {
1656 struct net *net = current->nsproxy->net_ns;
1657 ip_vs_sync_switch_mode(net, val);
1658 }
1659 }
1660 return rc;
1661}
1662
1663
1664
1665
1666
1667
1668
1669static struct ctl_table vs_vars[] = {
1670 {
1671 .procname = "amemthresh",
1672 .maxlen = sizeof(int),
1673 .mode = 0644,
1674 .proc_handler = proc_dointvec,
1675 },
1676 {
1677 .procname = "am_droprate",
1678 .maxlen = sizeof(int),
1679 .mode = 0644,
1680 .proc_handler = proc_dointvec,
1681 },
1682 {
1683 .procname = "drop_entry",
1684 .maxlen = sizeof(int),
1685 .mode = 0644,
1686 .proc_handler = proc_do_defense_mode,
1687 },
1688 {
1689 .procname = "drop_packet",
1690 .maxlen = sizeof(int),
1691 .mode = 0644,
1692 .proc_handler = proc_do_defense_mode,
1693 },
1694#ifdef CONFIG_IP_VS_NFCT
1695 {
1696 .procname = "conntrack",
1697 .maxlen = sizeof(int),
1698 .mode = 0644,
1699 .proc_handler = &proc_dointvec,
1700 },
1701#endif
1702 {
1703 .procname = "secure_tcp",
1704 .maxlen = sizeof(int),
1705 .mode = 0644,
1706 .proc_handler = proc_do_defense_mode,
1707 },
1708 {
1709 .procname = "snat_reroute",
1710 .maxlen = sizeof(int),
1711 .mode = 0644,
1712 .proc_handler = &proc_dointvec,
1713 },
1714 {
1715 .procname = "sync_version",
1716 .maxlen = sizeof(int),
1717 .mode = 0644,
1718 .proc_handler = &proc_do_sync_mode,
1719 },
1720 {
1721 .procname = "cache_bypass",
1722 .maxlen = sizeof(int),
1723 .mode = 0644,
1724 .proc_handler = proc_dointvec,
1725 },
1726 {
1727 .procname = "expire_nodest_conn",
1728 .maxlen = sizeof(int),
1729 .mode = 0644,
1730 .proc_handler = proc_dointvec,
1731 },
1732 {
1733 .procname = "expire_quiescent_template",
1734 .maxlen = sizeof(int),
1735 .mode = 0644,
1736 .proc_handler = proc_dointvec,
1737 },
1738 {
1739 .procname = "sync_threshold",
1740 .maxlen =
1741 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1742 .mode = 0644,
1743 .proc_handler = proc_do_sync_threshold,
1744 },
1745 {
1746 .procname = "nat_icmp_send",
1747 .maxlen = sizeof(int),
1748 .mode = 0644,
1749 .proc_handler = proc_dointvec,
1750 },
1751#ifdef CONFIG_IP_VS_DEBUG
1752 {
1753 .procname = "debug_level",
1754 .data = &sysctl_ip_vs_debug_level,
1755 .maxlen = sizeof(int),
1756 .mode = 0644,
1757 .proc_handler = proc_dointvec,
1758 },
1759#endif
1760#if 0
1761 {
1762 .procname = "timeout_established",
1763 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1764 .maxlen = sizeof(int),
1765 .mode = 0644,
1766 .proc_handler = proc_dointvec_jiffies,
1767 },
1768 {
1769 .procname = "timeout_synsent",
1770 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1771 .maxlen = sizeof(int),
1772 .mode = 0644,
1773 .proc_handler = proc_dointvec_jiffies,
1774 },
1775 {
1776 .procname = "timeout_synrecv",
1777 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1778 .maxlen = sizeof(int),
1779 .mode = 0644,
1780 .proc_handler = proc_dointvec_jiffies,
1781 },
1782 {
1783 .procname = "timeout_finwait",
1784 .data = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1785 .maxlen = sizeof(int),
1786 .mode = 0644,
1787 .proc_handler = proc_dointvec_jiffies,
1788 },
1789 {
1790 .procname = "timeout_timewait",
1791 .data = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1792 .maxlen = sizeof(int),
1793 .mode = 0644,
1794 .proc_handler = proc_dointvec_jiffies,
1795 },
1796 {
1797 .procname = "timeout_close",
1798 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1799 .maxlen = sizeof(int),
1800 .mode = 0644,
1801 .proc_handler = proc_dointvec_jiffies,
1802 },
1803 {
1804 .procname = "timeout_closewait",
1805 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1806 .maxlen = sizeof(int),
1807 .mode = 0644,
1808 .proc_handler = proc_dointvec_jiffies,
1809 },
1810 {
1811 .procname = "timeout_lastack",
1812 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1813 .maxlen = sizeof(int),
1814 .mode = 0644,
1815 .proc_handler = proc_dointvec_jiffies,
1816 },
1817 {
1818 .procname = "timeout_listen",
1819 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1820 .maxlen = sizeof(int),
1821 .mode = 0644,
1822 .proc_handler = proc_dointvec_jiffies,
1823 },
1824 {
1825 .procname = "timeout_synack",
1826 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1827 .maxlen = sizeof(int),
1828 .mode = 0644,
1829 .proc_handler = proc_dointvec_jiffies,
1830 },
1831 {
1832 .procname = "timeout_udp",
1833 .data = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1834 .maxlen = sizeof(int),
1835 .mode = 0644,
1836 .proc_handler = proc_dointvec_jiffies,
1837 },
1838 {
1839 .procname = "timeout_icmp",
1840 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1841 .maxlen = sizeof(int),
1842 .mode = 0644,
1843 .proc_handler = proc_dointvec_jiffies,
1844 },
1845#endif
1846 { }
1847};
1848
1849const struct ctl_path net_vs_ctl_path[] = {
1850 { .procname = "net", },
1851 { .procname = "ipv4", },
1852 { .procname = "vs", },
1853 { }
1854};
1855EXPORT_SYMBOL_GPL(net_vs_ctl_path);
1856#endif
1857
1858#ifdef CONFIG_PROC_FS
1859
1860struct ip_vs_iter {
1861 struct seq_net_private p;
1862 struct list_head *table;
1863 int bucket;
1864};
1865
1866
1867
1868
1869
1870static inline const char *ip_vs_fwd_name(unsigned flags)
1871{
1872 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1873 case IP_VS_CONN_F_LOCALNODE:
1874 return "Local";
1875 case IP_VS_CONN_F_TUNNEL:
1876 return "Tunnel";
1877 case IP_VS_CONN_F_DROUTE:
1878 return "Route";
1879 default:
1880 return "Masq";
1881 }
1882}
1883
1884
1885
1886static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1887{
1888 struct net *net = seq_file_net(seq);
1889 struct ip_vs_iter *iter = seq->private;
1890 int idx;
1891 struct ip_vs_service *svc;
1892
1893
1894 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1895 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1896 if (net_eq(svc->net, net) && pos-- == 0) {
1897 iter->table = ip_vs_svc_table;
1898 iter->bucket = idx;
1899 return svc;
1900 }
1901 }
1902 }
1903
1904
1905 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1906 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1907 if (net_eq(svc->net, net) && pos-- == 0) {
1908 iter->table = ip_vs_svc_fwm_table;
1909 iter->bucket = idx;
1910 return svc;
1911 }
1912 }
1913 }
1914
1915 return NULL;
1916}
1917
1918static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1919__acquires(__ip_vs_svc_lock)
1920{
1921
1922 read_lock_bh(&__ip_vs_svc_lock);
1923 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1924}
1925
1926
1927static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1928{
1929 struct list_head *e;
1930 struct ip_vs_iter *iter;
1931 struct ip_vs_service *svc;
1932
1933 ++*pos;
1934 if (v == SEQ_START_TOKEN)
1935 return ip_vs_info_array(seq,0);
1936
1937 svc = v;
1938 iter = seq->private;
1939
1940 if (iter->table == ip_vs_svc_table) {
1941
1942 if ((e = svc->s_list.next) != &ip_vs_svc_table[iter->bucket])
1943 return list_entry(e, struct ip_vs_service, s_list);
1944
1945
1946 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1947 list_for_each_entry(svc,&ip_vs_svc_table[iter->bucket],
1948 s_list) {
1949 return svc;
1950 }
1951 }
1952
1953 iter->table = ip_vs_svc_fwm_table;
1954 iter->bucket = -1;
1955 goto scan_fwmark;
1956 }
1957
1958
1959 if ((e = svc->f_list.next) != &ip_vs_svc_fwm_table[iter->bucket])
1960 return list_entry(e, struct ip_vs_service, f_list);
1961
1962 scan_fwmark:
1963 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1964 list_for_each_entry(svc, &ip_vs_svc_fwm_table[iter->bucket],
1965 f_list)
1966 return svc;
1967 }
1968
1969 return NULL;
1970}
1971
1972static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
1973__releases(__ip_vs_svc_lock)
1974{
1975 read_unlock_bh(&__ip_vs_svc_lock);
1976}
1977
1978
1979static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
1980{
1981 if (v == SEQ_START_TOKEN) {
1982 seq_printf(seq,
1983 "IP Virtual Server version %d.%d.%d (size=%d)\n",
1984 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
1985 seq_puts(seq,
1986 "Prot LocalAddress:Port Scheduler Flags\n");
1987 seq_puts(seq,
1988 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
1989 } else {
1990 const struct ip_vs_service *svc = v;
1991 const struct ip_vs_iter *iter = seq->private;
1992 const struct ip_vs_dest *dest;
1993
1994 if (iter->table == ip_vs_svc_table) {
1995#ifdef CONFIG_IP_VS_IPV6
1996 if (svc->af == AF_INET6)
1997 seq_printf(seq, "%s [%pI6]:%04X %s ",
1998 ip_vs_proto_name(svc->protocol),
1999 &svc->addr.in6,
2000 ntohs(svc->port),
2001 svc->scheduler->name);
2002 else
2003#endif
2004 seq_printf(seq, "%s %08X:%04X %s %s ",
2005 ip_vs_proto_name(svc->protocol),
2006 ntohl(svc->addr.ip),
2007 ntohs(svc->port),
2008 svc->scheduler->name,
2009 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2010 } else {
2011 seq_printf(seq, "FWM %08X %s %s",
2012 svc->fwmark, svc->scheduler->name,
2013 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2014 }
2015
2016 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2017 seq_printf(seq, "persistent %d %08X\n",
2018 svc->timeout,
2019 ntohl(svc->netmask));
2020 else
2021 seq_putc(seq, '\n');
2022
2023 list_for_each_entry(dest, &svc->destinations, n_list) {
2024#ifdef CONFIG_IP_VS_IPV6
2025 if (dest->af == AF_INET6)
2026 seq_printf(seq,
2027 " -> [%pI6]:%04X"
2028 " %-7s %-6d %-10d %-10d\n",
2029 &dest->addr.in6,
2030 ntohs(dest->port),
2031 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2032 atomic_read(&dest->weight),
2033 atomic_read(&dest->activeconns),
2034 atomic_read(&dest->inactconns));
2035 else
2036#endif
2037 seq_printf(seq,
2038 " -> %08X:%04X "
2039 "%-7s %-6d %-10d %-10d\n",
2040 ntohl(dest->addr.ip),
2041 ntohs(dest->port),
2042 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2043 atomic_read(&dest->weight),
2044 atomic_read(&dest->activeconns),
2045 atomic_read(&dest->inactconns));
2046
2047 }
2048 }
2049 return 0;
2050}
2051
2052static const struct seq_operations ip_vs_info_seq_ops = {
2053 .start = ip_vs_info_seq_start,
2054 .next = ip_vs_info_seq_next,
2055 .stop = ip_vs_info_seq_stop,
2056 .show = ip_vs_info_seq_show,
2057};
2058
2059static int ip_vs_info_open(struct inode *inode, struct file *file)
2060{
2061 return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2062 sizeof(struct ip_vs_iter));
2063}
2064
2065static const struct file_operations ip_vs_info_fops = {
2066 .owner = THIS_MODULE,
2067 .open = ip_vs_info_open,
2068 .read = seq_read,
2069 .llseek = seq_lseek,
2070 .release = seq_release_net,
2071};
2072
2073static int ip_vs_stats_show(struct seq_file *seq, void *v)
2074{
2075 struct net *net = seq_file_single_net(seq);
2076 struct ip_vs_stats_user show;
2077
2078
2079 seq_puts(seq,
2080 " Total Incoming Outgoing Incoming Outgoing\n");
2081 seq_printf(seq,
2082 " Conns Packets Packets Bytes Bytes\n");
2083
2084 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2085 seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", show.conns,
2086 show.inpkts, show.outpkts,
2087 (unsigned long long) show.inbytes,
2088 (unsigned long long) show.outbytes);
2089
2090
2091 seq_puts(seq,
2092 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2093 seq_printf(seq, "%8X %8X %8X %16X %16X\n",
2094 show.cps, show.inpps, show.outpps,
2095 show.inbps, show.outbps);
2096
2097 return 0;
2098}
2099
2100static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2101{
2102 return single_open_net(inode, file, ip_vs_stats_show);
2103}
2104
2105static const struct file_operations ip_vs_stats_fops = {
2106 .owner = THIS_MODULE,
2107 .open = ip_vs_stats_seq_open,
2108 .read = seq_read,
2109 .llseek = seq_lseek,
2110 .release = single_release_net,
2111};
2112
2113static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2114{
2115 struct net *net = seq_file_single_net(seq);
2116 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2117 struct ip_vs_cpu_stats *cpustats = tot_stats->cpustats;
2118 struct ip_vs_stats_user rates;
2119 int i;
2120
2121
2122 seq_puts(seq,
2123 " Total Incoming Outgoing Incoming Outgoing\n");
2124 seq_printf(seq,
2125 "CPU Conns Packets Packets Bytes Bytes\n");
2126
2127 for_each_possible_cpu(i) {
2128 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2129 unsigned int start;
2130 __u64 inbytes, outbytes;
2131
2132 do {
2133 start = u64_stats_fetch_begin_bh(&u->syncp);
2134 inbytes = u->ustats.inbytes;
2135 outbytes = u->ustats.outbytes;
2136 } while (u64_stats_fetch_retry_bh(&u->syncp, start));
2137
2138 seq_printf(seq, "%3X %8X %8X %8X %16LX %16LX\n",
2139 i, u->ustats.conns, u->ustats.inpkts,
2140 u->ustats.outpkts, (__u64)inbytes,
2141 (__u64)outbytes);
2142 }
2143
2144 spin_lock_bh(&tot_stats->lock);
2145
2146 seq_printf(seq, " ~ %8X %8X %8X %16LX %16LX\n\n",
2147 tot_stats->ustats.conns, tot_stats->ustats.inpkts,
2148 tot_stats->ustats.outpkts,
2149 (unsigned long long) tot_stats->ustats.inbytes,
2150 (unsigned long long) tot_stats->ustats.outbytes);
2151
2152 ip_vs_read_estimator(&rates, tot_stats);
2153
2154 spin_unlock_bh(&tot_stats->lock);
2155
2156
2157 seq_puts(seq,
2158 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2159 seq_printf(seq, " %8X %8X %8X %16X %16X\n",
2160 rates.cps,
2161 rates.inpps,
2162 rates.outpps,
2163 rates.inbps,
2164 rates.outbps);
2165
2166 return 0;
2167}
2168
2169static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2170{
2171 return single_open_net(inode, file, ip_vs_stats_percpu_show);
2172}
2173
2174static const struct file_operations ip_vs_stats_percpu_fops = {
2175 .owner = THIS_MODULE,
2176 .open = ip_vs_stats_percpu_seq_open,
2177 .read = seq_read,
2178 .llseek = seq_lseek,
2179 .release = single_release_net,
2180};
2181#endif
2182
2183
2184
2185
2186static int ip_vs_set_timeout(struct net *net, struct ip_vs_timeout_user *u)
2187{
2188#if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2189 struct ip_vs_proto_data *pd;
2190#endif
2191
2192 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2193 u->tcp_timeout,
2194 u->tcp_fin_timeout,
2195 u->udp_timeout);
2196
2197#ifdef CONFIG_IP_VS_PROTO_TCP
2198 if (u->tcp_timeout) {
2199 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2200 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2201 = u->tcp_timeout * HZ;
2202 }
2203
2204 if (u->tcp_fin_timeout) {
2205 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2206 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2207 = u->tcp_fin_timeout * HZ;
2208 }
2209#endif
2210
2211#ifdef CONFIG_IP_VS_PROTO_UDP
2212 if (u->udp_timeout) {
2213 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2214 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2215 = u->udp_timeout * HZ;
2216 }
2217#endif
2218 return 0;
2219}
2220
2221
2222#define SET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2223#define SERVICE_ARG_LEN (sizeof(struct ip_vs_service_user))
2224#define SVCDEST_ARG_LEN (sizeof(struct ip_vs_service_user) + \
2225 sizeof(struct ip_vs_dest_user))
2226#define TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2227#define DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user))
2228#define MAX_ARG_LEN SVCDEST_ARG_LEN
2229
2230static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
2231 [SET_CMDID(IP_VS_SO_SET_ADD)] = SERVICE_ARG_LEN,
2232 [SET_CMDID(IP_VS_SO_SET_EDIT)] = SERVICE_ARG_LEN,
2233 [SET_CMDID(IP_VS_SO_SET_DEL)] = SERVICE_ARG_LEN,
2234 [SET_CMDID(IP_VS_SO_SET_FLUSH)] = 0,
2235 [SET_CMDID(IP_VS_SO_SET_ADDDEST)] = SVCDEST_ARG_LEN,
2236 [SET_CMDID(IP_VS_SO_SET_DELDEST)] = SVCDEST_ARG_LEN,
2237 [SET_CMDID(IP_VS_SO_SET_EDITDEST)] = SVCDEST_ARG_LEN,
2238 [SET_CMDID(IP_VS_SO_SET_TIMEOUT)] = TIMEOUT_ARG_LEN,
2239 [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)] = DAEMON_ARG_LEN,
2240 [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)] = DAEMON_ARG_LEN,
2241 [SET_CMDID(IP_VS_SO_SET_ZERO)] = SERVICE_ARG_LEN,
2242};
2243
2244static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2245 struct ip_vs_service_user *usvc_compat)
2246{
2247 memset(usvc, 0, sizeof(*usvc));
2248
2249 usvc->af = AF_INET;
2250 usvc->protocol = usvc_compat->protocol;
2251 usvc->addr.ip = usvc_compat->addr;
2252 usvc->port = usvc_compat->port;
2253 usvc->fwmark = usvc_compat->fwmark;
2254
2255
2256 usvc->sched_name = usvc_compat->sched_name;
2257
2258 usvc->flags = usvc_compat->flags;
2259 usvc->timeout = usvc_compat->timeout;
2260 usvc->netmask = usvc_compat->netmask;
2261}
2262
2263static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2264 struct ip_vs_dest_user *udest_compat)
2265{
2266 memset(udest, 0, sizeof(*udest));
2267
2268 udest->addr.ip = udest_compat->addr;
2269 udest->port = udest_compat->port;
2270 udest->conn_flags = udest_compat->conn_flags;
2271 udest->weight = udest_compat->weight;
2272 udest->u_threshold = udest_compat->u_threshold;
2273 udest->l_threshold = udest_compat->l_threshold;
2274}
2275
2276static int
2277do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2278{
2279 struct net *net = sock_net(sk);
2280 int ret;
2281 unsigned char arg[MAX_ARG_LEN];
2282 struct ip_vs_service_user *usvc_compat;
2283 struct ip_vs_service_user_kern usvc;
2284 struct ip_vs_service *svc;
2285 struct ip_vs_dest_user *udest_compat;
2286 struct ip_vs_dest_user_kern udest;
2287 struct netns_ipvs *ipvs = net_ipvs(net);
2288
2289 if (!capable(CAP_NET_ADMIN))
2290 return -EPERM;
2291
2292 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2293 return -EINVAL;
2294 if (len < 0 || len > MAX_ARG_LEN)
2295 return -EINVAL;
2296 if (len != set_arglen[SET_CMDID(cmd)]) {
2297 pr_err("set_ctl: len %u != %u\n",
2298 len, set_arglen[SET_CMDID(cmd)]);
2299 return -EINVAL;
2300 }
2301
2302 if (copy_from_user(arg, user, len) != 0)
2303 return -EFAULT;
2304
2305
2306 ip_vs_use_count_inc();
2307
2308
2309 if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2310 cmd == IP_VS_SO_SET_STOPDAEMON) {
2311 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2312
2313 if (mutex_lock_interruptible(&ipvs->sync_mutex)) {
2314 ret = -ERESTARTSYS;
2315 goto out_dec;
2316 }
2317 if (cmd == IP_VS_SO_SET_STARTDAEMON)
2318 ret = start_sync_thread(net, dm->state, dm->mcast_ifn,
2319 dm->syncid);
2320 else
2321 ret = stop_sync_thread(net, dm->state);
2322 mutex_unlock(&ipvs->sync_mutex);
2323 goto out_dec;
2324 }
2325
2326 if (mutex_lock_interruptible(&__ip_vs_mutex)) {
2327 ret = -ERESTARTSYS;
2328 goto out_dec;
2329 }
2330
2331 if (cmd == IP_VS_SO_SET_FLUSH) {
2332
2333 ret = ip_vs_flush(net);
2334 goto out_unlock;
2335 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2336
2337 ret = ip_vs_set_timeout(net, (struct ip_vs_timeout_user *)arg);
2338 goto out_unlock;
2339 }
2340
2341 usvc_compat = (struct ip_vs_service_user *)arg;
2342 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2343
2344
2345
2346 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2347 ip_vs_copy_udest_compat(&udest, udest_compat);
2348
2349 if (cmd == IP_VS_SO_SET_ZERO) {
2350
2351 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2352 ret = ip_vs_zero_all(net);
2353 goto out_unlock;
2354 }
2355 }
2356
2357
2358 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2359 usvc.protocol != IPPROTO_SCTP) {
2360 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2361 usvc.protocol, &usvc.addr.ip,
2362 ntohs(usvc.port), usvc.sched_name);
2363 ret = -EFAULT;
2364 goto out_unlock;
2365 }
2366
2367
2368 if (usvc.fwmark == 0)
2369 svc = __ip_vs_service_find(net, usvc.af, usvc.protocol,
2370 &usvc.addr, usvc.port);
2371 else
2372 svc = __ip_vs_svc_fwm_find(net, usvc.af, usvc.fwmark);
2373
2374 if (cmd != IP_VS_SO_SET_ADD
2375 && (svc == NULL || svc->protocol != usvc.protocol)) {
2376 ret = -ESRCH;
2377 goto out_unlock;
2378 }
2379
2380 switch (cmd) {
2381 case IP_VS_SO_SET_ADD:
2382 if (svc != NULL)
2383 ret = -EEXIST;
2384 else
2385 ret = ip_vs_add_service(net, &usvc, &svc);
2386 break;
2387 case IP_VS_SO_SET_EDIT:
2388 ret = ip_vs_edit_service(svc, &usvc);
2389 break;
2390 case IP_VS_SO_SET_DEL:
2391 ret = ip_vs_del_service(svc);
2392 if (!ret)
2393 goto out_unlock;
2394 break;
2395 case IP_VS_SO_SET_ZERO:
2396 ret = ip_vs_zero_service(svc);
2397 break;
2398 case IP_VS_SO_SET_ADDDEST:
2399 ret = ip_vs_add_dest(svc, &udest);
2400 break;
2401 case IP_VS_SO_SET_EDITDEST:
2402 ret = ip_vs_edit_dest(svc, &udest);
2403 break;
2404 case IP_VS_SO_SET_DELDEST:
2405 ret = ip_vs_del_dest(svc, &udest);
2406 break;
2407 default:
2408 ret = -EINVAL;
2409 }
2410
2411 out_unlock:
2412 mutex_unlock(&__ip_vs_mutex);
2413 out_dec:
2414
2415 ip_vs_use_count_dec();
2416
2417 return ret;
2418}
2419
2420
2421static void
2422ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2423{
2424 dst->protocol = src->protocol;
2425 dst->addr = src->addr.ip;
2426 dst->port = src->port;
2427 dst->fwmark = src->fwmark;
2428 strlcpy(dst->sched_name, src->scheduler->name, sizeof(dst->sched_name));
2429 dst->flags = src->flags;
2430 dst->timeout = src->timeout / HZ;
2431 dst->netmask = src->netmask;
2432 dst->num_dests = src->num_dests;
2433 ip_vs_copy_stats(&dst->stats, &src->stats);
2434}
2435
2436static inline int
2437__ip_vs_get_service_entries(struct net *net,
2438 const struct ip_vs_get_services *get,
2439 struct ip_vs_get_services __user *uptr)
2440{
2441 int idx, count=0;
2442 struct ip_vs_service *svc;
2443 struct ip_vs_service_entry entry;
2444 int ret = 0;
2445
2446 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2447 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2448
2449 if (svc->af != AF_INET || !net_eq(svc->net, net))
2450 continue;
2451
2452 if (count >= get->num_services)
2453 goto out;
2454 memset(&entry, 0, sizeof(entry));
2455 ip_vs_copy_service(&entry, svc);
2456 if (copy_to_user(&uptr->entrytable[count],
2457 &entry, sizeof(entry))) {
2458 ret = -EFAULT;
2459 goto out;
2460 }
2461 count++;
2462 }
2463 }
2464
2465 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2466 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2467
2468 if (svc->af != AF_INET || !net_eq(svc->net, net))
2469 continue;
2470
2471 if (count >= get->num_services)
2472 goto out;
2473 memset(&entry, 0, sizeof(entry));
2474 ip_vs_copy_service(&entry, svc);
2475 if (copy_to_user(&uptr->entrytable[count],
2476 &entry, sizeof(entry))) {
2477 ret = -EFAULT;
2478 goto out;
2479 }
2480 count++;
2481 }
2482 }
2483out:
2484 return ret;
2485}
2486
2487static inline int
2488__ip_vs_get_dest_entries(struct net *net, const struct ip_vs_get_dests *get,
2489 struct ip_vs_get_dests __user *uptr)
2490{
2491 struct ip_vs_service *svc;
2492 union nf_inet_addr addr = { .ip = get->addr };
2493 int ret = 0;
2494
2495 if (get->fwmark)
2496 svc = __ip_vs_svc_fwm_find(net, AF_INET, get->fwmark);
2497 else
2498 svc = __ip_vs_service_find(net, AF_INET, get->protocol, &addr,
2499 get->port);
2500
2501 if (svc) {
2502 int count = 0;
2503 struct ip_vs_dest *dest;
2504 struct ip_vs_dest_entry entry;
2505
2506 list_for_each_entry(dest, &svc->destinations, n_list) {
2507 if (count >= get->num_dests)
2508 break;
2509
2510 entry.addr = dest->addr.ip;
2511 entry.port = dest->port;
2512 entry.conn_flags = atomic_read(&dest->conn_flags);
2513 entry.weight = atomic_read(&dest->weight);
2514 entry.u_threshold = dest->u_threshold;
2515 entry.l_threshold = dest->l_threshold;
2516 entry.activeconns = atomic_read(&dest->activeconns);
2517 entry.inactconns = atomic_read(&dest->inactconns);
2518 entry.persistconns = atomic_read(&dest->persistconns);
2519 ip_vs_copy_stats(&entry.stats, &dest->stats);
2520 if (copy_to_user(&uptr->entrytable[count],
2521 &entry, sizeof(entry))) {
2522 ret = -EFAULT;
2523 break;
2524 }
2525 count++;
2526 }
2527 } else
2528 ret = -ESRCH;
2529 return ret;
2530}
2531
2532static inline void
2533__ip_vs_get_timeouts(struct net *net, struct ip_vs_timeout_user *u)
2534{
2535#if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2536 struct ip_vs_proto_data *pd;
2537#endif
2538
2539#ifdef CONFIG_IP_VS_PROTO_TCP
2540 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2541 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2542 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2543#endif
2544#ifdef CONFIG_IP_VS_PROTO_UDP
2545 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2546 u->udp_timeout =
2547 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2548#endif
2549}
2550
2551
2552#define GET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2553#define GET_INFO_ARG_LEN (sizeof(struct ip_vs_getinfo))
2554#define GET_SERVICES_ARG_LEN (sizeof(struct ip_vs_get_services))
2555#define GET_SERVICE_ARG_LEN (sizeof(struct ip_vs_service_entry))
2556#define GET_DESTS_ARG_LEN (sizeof(struct ip_vs_get_dests))
2557#define GET_TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2558#define GET_DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user) * 2)
2559
2560static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
2561 [GET_CMDID(IP_VS_SO_GET_VERSION)] = 64,
2562 [GET_CMDID(IP_VS_SO_GET_INFO)] = GET_INFO_ARG_LEN,
2563 [GET_CMDID(IP_VS_SO_GET_SERVICES)] = GET_SERVICES_ARG_LEN,
2564 [GET_CMDID(IP_VS_SO_GET_SERVICE)] = GET_SERVICE_ARG_LEN,
2565 [GET_CMDID(IP_VS_SO_GET_DESTS)] = GET_DESTS_ARG_LEN,
2566 [GET_CMDID(IP_VS_SO_GET_TIMEOUT)] = GET_TIMEOUT_ARG_LEN,
2567 [GET_CMDID(IP_VS_SO_GET_DAEMON)] = GET_DAEMON_ARG_LEN,
2568};
2569
2570static int
2571do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2572{
2573 unsigned char arg[128];
2574 int ret = 0;
2575 unsigned int copylen;
2576 struct net *net = sock_net(sk);
2577 struct netns_ipvs *ipvs = net_ipvs(net);
2578
2579 BUG_ON(!net);
2580 if (!capable(CAP_NET_ADMIN))
2581 return -EPERM;
2582
2583 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2584 return -EINVAL;
2585
2586 if (*len < get_arglen[GET_CMDID(cmd)]) {
2587 pr_err("get_ctl: len %u < %u\n",
2588 *len, get_arglen[GET_CMDID(cmd)]);
2589 return -EINVAL;
2590 }
2591
2592 copylen = get_arglen[GET_CMDID(cmd)];
2593 if (copylen > 128)
2594 return -EINVAL;
2595
2596 if (copy_from_user(arg, user, copylen) != 0)
2597 return -EFAULT;
2598
2599
2600
2601 if (cmd == IP_VS_SO_GET_DAEMON) {
2602 struct ip_vs_daemon_user d[2];
2603
2604 memset(&d, 0, sizeof(d));
2605 if (mutex_lock_interruptible(&ipvs->sync_mutex))
2606 return -ERESTARTSYS;
2607
2608 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2609 d[0].state = IP_VS_STATE_MASTER;
2610 strlcpy(d[0].mcast_ifn, ipvs->master_mcast_ifn,
2611 sizeof(d[0].mcast_ifn));
2612 d[0].syncid = ipvs->master_syncid;
2613 }
2614 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2615 d[1].state = IP_VS_STATE_BACKUP;
2616 strlcpy(d[1].mcast_ifn, ipvs->backup_mcast_ifn,
2617 sizeof(d[1].mcast_ifn));
2618 d[1].syncid = ipvs->backup_syncid;
2619 }
2620 if (copy_to_user(user, &d, sizeof(d)) != 0)
2621 ret = -EFAULT;
2622 mutex_unlock(&ipvs->sync_mutex);
2623 return ret;
2624 }
2625
2626 if (mutex_lock_interruptible(&__ip_vs_mutex))
2627 return -ERESTARTSYS;
2628
2629 switch (cmd) {
2630 case IP_VS_SO_GET_VERSION:
2631 {
2632 char buf[64];
2633
2634 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2635 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2636 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2637 ret = -EFAULT;
2638 goto out;
2639 }
2640 *len = strlen(buf)+1;
2641 }
2642 break;
2643
2644 case IP_VS_SO_GET_INFO:
2645 {
2646 struct ip_vs_getinfo info;
2647 info.version = IP_VS_VERSION_CODE;
2648 info.size = ip_vs_conn_tab_size;
2649 info.num_services = ipvs->num_services;
2650 if (copy_to_user(user, &info, sizeof(info)) != 0)
2651 ret = -EFAULT;
2652 }
2653 break;
2654
2655 case IP_VS_SO_GET_SERVICES:
2656 {
2657 struct ip_vs_get_services *get;
2658 int size;
2659
2660 get = (struct ip_vs_get_services *)arg;
2661 size = sizeof(*get) +
2662 sizeof(struct ip_vs_service_entry) * get->num_services;
2663 if (*len != size) {
2664 pr_err("length: %u != %u\n", *len, size);
2665 ret = -EINVAL;
2666 goto out;
2667 }
2668 ret = __ip_vs_get_service_entries(net, get, user);
2669 }
2670 break;
2671
2672 case IP_VS_SO_GET_SERVICE:
2673 {
2674 struct ip_vs_service_entry *entry;
2675 struct ip_vs_service *svc;
2676 union nf_inet_addr addr;
2677
2678 entry = (struct ip_vs_service_entry *)arg;
2679 addr.ip = entry->addr;
2680 if (entry->fwmark)
2681 svc = __ip_vs_svc_fwm_find(net, AF_INET, entry->fwmark);
2682 else
2683 svc = __ip_vs_service_find(net, AF_INET,
2684 entry->protocol, &addr,
2685 entry->port);
2686 if (svc) {
2687 ip_vs_copy_service(entry, svc);
2688 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2689 ret = -EFAULT;
2690 } else
2691 ret = -ESRCH;
2692 }
2693 break;
2694
2695 case IP_VS_SO_GET_DESTS:
2696 {
2697 struct ip_vs_get_dests *get;
2698 int size;
2699
2700 get = (struct ip_vs_get_dests *)arg;
2701 size = sizeof(*get) +
2702 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2703 if (*len != size) {
2704 pr_err("length: %u != %u\n", *len, size);
2705 ret = -EINVAL;
2706 goto out;
2707 }
2708 ret = __ip_vs_get_dest_entries(net, get, user);
2709 }
2710 break;
2711
2712 case IP_VS_SO_GET_TIMEOUT:
2713 {
2714 struct ip_vs_timeout_user t;
2715
2716 __ip_vs_get_timeouts(net, &t);
2717 if (copy_to_user(user, &t, sizeof(t)) != 0)
2718 ret = -EFAULT;
2719 }
2720 break;
2721
2722 default:
2723 ret = -EINVAL;
2724 }
2725
2726out:
2727 mutex_unlock(&__ip_vs_mutex);
2728 return ret;
2729}
2730
2731
2732static struct nf_sockopt_ops ip_vs_sockopts = {
2733 .pf = PF_INET,
2734 .set_optmin = IP_VS_BASE_CTL,
2735 .set_optmax = IP_VS_SO_SET_MAX+1,
2736 .set = do_ip_vs_set_ctl,
2737 .get_optmin = IP_VS_BASE_CTL,
2738 .get_optmax = IP_VS_SO_GET_MAX+1,
2739 .get = do_ip_vs_get_ctl,
2740 .owner = THIS_MODULE,
2741};
2742
2743
2744
2745
2746
2747
2748static struct genl_family ip_vs_genl_family = {
2749 .id = GENL_ID_GENERATE,
2750 .hdrsize = 0,
2751 .name = IPVS_GENL_NAME,
2752 .version = IPVS_GENL_VERSION,
2753 .maxattr = IPVS_CMD_MAX,
2754 .netnsok = true,
2755};
2756
2757
2758static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2759 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2760 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2761 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2762 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2763 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2764 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2765};
2766
2767
2768static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2769 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2770 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2771 .len = IP_VS_IFNAME_MAXLEN },
2772 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2773};
2774
2775
2776static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2777 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2778 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2779 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2780 .len = sizeof(union nf_inet_addr) },
2781 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2782 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2783 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2784 .len = IP_VS_SCHEDNAME_MAXLEN },
2785 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2786 .len = IP_VS_PENAME_MAXLEN },
2787 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2788 .len = sizeof(struct ip_vs_flags) },
2789 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2790 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2791 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2792};
2793
2794
2795static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2796 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2797 .len = sizeof(union nf_inet_addr) },
2798 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2799 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2800 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2801 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2802 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2803 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2804 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2805 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2806 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2807};
2808
2809static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2810 struct ip_vs_stats *stats)
2811{
2812 struct ip_vs_stats_user ustats;
2813 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2814 if (!nl_stats)
2815 return -EMSGSIZE;
2816
2817 ip_vs_copy_stats(&ustats, stats);
2818
2819 NLA_PUT_U32(skb, IPVS_STATS_ATTR_CONNS, ustats.conns);
2820 NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPKTS, ustats.inpkts);
2821 NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPKTS, ustats.outpkts);
2822 NLA_PUT_U64(skb, IPVS_STATS_ATTR_INBYTES, ustats.inbytes);
2823 NLA_PUT_U64(skb, IPVS_STATS_ATTR_OUTBYTES, ustats.outbytes);
2824 NLA_PUT_U32(skb, IPVS_STATS_ATTR_CPS, ustats.cps);
2825 NLA_PUT_U32(skb, IPVS_STATS_ATTR_INPPS, ustats.inpps);
2826 NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTPPS, ustats.outpps);
2827 NLA_PUT_U32(skb, IPVS_STATS_ATTR_INBPS, ustats.inbps);
2828 NLA_PUT_U32(skb, IPVS_STATS_ATTR_OUTBPS, ustats.outbps);
2829
2830 nla_nest_end(skb, nl_stats);
2831
2832 return 0;
2833
2834nla_put_failure:
2835 nla_nest_cancel(skb, nl_stats);
2836 return -EMSGSIZE;
2837}
2838
2839static int ip_vs_genl_fill_service(struct sk_buff *skb,
2840 struct ip_vs_service *svc)
2841{
2842 struct nlattr *nl_service;
2843 struct ip_vs_flags flags = { .flags = svc->flags,
2844 .mask = ~0 };
2845
2846 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2847 if (!nl_service)
2848 return -EMSGSIZE;
2849
2850 NLA_PUT_U16(skb, IPVS_SVC_ATTR_AF, svc->af);
2851
2852 if (svc->fwmark) {
2853 NLA_PUT_U32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark);
2854 } else {
2855 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol);
2856 NLA_PUT(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr);
2857 NLA_PUT_U16(skb, IPVS_SVC_ATTR_PORT, svc->port);
2858 }
2859
2860 NLA_PUT_STRING(skb, IPVS_SVC_ATTR_SCHED_NAME, svc->scheduler->name);
2861 if (svc->pe)
2862 NLA_PUT_STRING(skb, IPVS_SVC_ATTR_PE_NAME, svc->pe->name);
2863 NLA_PUT(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags);
2864 NLA_PUT_U32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ);
2865 NLA_PUT_U32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask);
2866
2867 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2868 goto nla_put_failure;
2869
2870 nla_nest_end(skb, nl_service);
2871
2872 return 0;
2873
2874nla_put_failure:
2875 nla_nest_cancel(skb, nl_service);
2876 return -EMSGSIZE;
2877}
2878
2879static int ip_vs_genl_dump_service(struct sk_buff *skb,
2880 struct ip_vs_service *svc,
2881 struct netlink_callback *cb)
2882{
2883 void *hdr;
2884
2885 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2886 &ip_vs_genl_family, NLM_F_MULTI,
2887 IPVS_CMD_NEW_SERVICE);
2888 if (!hdr)
2889 return -EMSGSIZE;
2890
2891 if (ip_vs_genl_fill_service(skb, svc) < 0)
2892 goto nla_put_failure;
2893
2894 return genlmsg_end(skb, hdr);
2895
2896nla_put_failure:
2897 genlmsg_cancel(skb, hdr);
2898 return -EMSGSIZE;
2899}
2900
2901static int ip_vs_genl_dump_services(struct sk_buff *skb,
2902 struct netlink_callback *cb)
2903{
2904 int idx = 0, i;
2905 int start = cb->args[0];
2906 struct ip_vs_service *svc;
2907 struct net *net = skb_sknet(skb);
2908
2909 mutex_lock(&__ip_vs_mutex);
2910 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2911 list_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2912 if (++idx <= start || !net_eq(svc->net, net))
2913 continue;
2914 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2915 idx--;
2916 goto nla_put_failure;
2917 }
2918 }
2919 }
2920
2921 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2922 list_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
2923 if (++idx <= start || !net_eq(svc->net, net))
2924 continue;
2925 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2926 idx--;
2927 goto nla_put_failure;
2928 }
2929 }
2930 }
2931
2932nla_put_failure:
2933 mutex_unlock(&__ip_vs_mutex);
2934 cb->args[0] = idx;
2935
2936 return skb->len;
2937}
2938
2939static int ip_vs_genl_parse_service(struct net *net,
2940 struct ip_vs_service_user_kern *usvc,
2941 struct nlattr *nla, int full_entry,
2942 struct ip_vs_service **ret_svc)
2943{
2944 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
2945 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
2946 struct ip_vs_service *svc;
2947
2948
2949 if (nla == NULL ||
2950 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
2951 return -EINVAL;
2952
2953 nla_af = attrs[IPVS_SVC_ATTR_AF];
2954 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
2955 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
2956 nla_port = attrs[IPVS_SVC_ATTR_PORT];
2957 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
2958
2959 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
2960 return -EINVAL;
2961
2962 memset(usvc, 0, sizeof(*usvc));
2963
2964 usvc->af = nla_get_u16(nla_af);
2965#ifdef CONFIG_IP_VS_IPV6
2966 if (usvc->af != AF_INET && usvc->af != AF_INET6)
2967#else
2968 if (usvc->af != AF_INET)
2969#endif
2970 return -EAFNOSUPPORT;
2971
2972 if (nla_fwmark) {
2973 usvc->protocol = IPPROTO_TCP;
2974 usvc->fwmark = nla_get_u32(nla_fwmark);
2975 } else {
2976 usvc->protocol = nla_get_u16(nla_protocol);
2977 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
2978 usvc->port = nla_get_u16(nla_port);
2979 usvc->fwmark = 0;
2980 }
2981
2982 if (usvc->fwmark)
2983 svc = __ip_vs_svc_fwm_find(net, usvc->af, usvc->fwmark);
2984 else
2985 svc = __ip_vs_service_find(net, usvc->af, usvc->protocol,
2986 &usvc->addr, usvc->port);
2987 *ret_svc = svc;
2988
2989
2990 if (full_entry) {
2991 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
2992 *nla_netmask;
2993 struct ip_vs_flags flags;
2994
2995 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
2996 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
2997 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
2998 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
2999 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3000
3001 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3002 return -EINVAL;
3003
3004 nla_memcpy(&flags, nla_flags, sizeof(flags));
3005
3006
3007 if (svc)
3008 usvc->flags = svc->flags;
3009
3010
3011 usvc->flags = (usvc->flags & ~flags.mask) |
3012 (flags.flags & flags.mask);
3013 usvc->sched_name = nla_data(nla_sched);
3014 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3015 usvc->timeout = nla_get_u32(nla_timeout);
3016 usvc->netmask = nla_get_u32(nla_netmask);
3017 }
3018
3019 return 0;
3020}
3021
3022static struct ip_vs_service *ip_vs_genl_find_service(struct net *net,
3023 struct nlattr *nla)
3024{
3025 struct ip_vs_service_user_kern usvc;
3026 struct ip_vs_service *svc;
3027 int ret;
3028
3029 ret = ip_vs_genl_parse_service(net, &usvc, nla, 0, &svc);
3030 return ret ? ERR_PTR(ret) : svc;
3031}
3032
3033static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3034{
3035 struct nlattr *nl_dest;
3036
3037 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3038 if (!nl_dest)
3039 return -EMSGSIZE;
3040
3041 NLA_PUT(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr);
3042 NLA_PUT_U16(skb, IPVS_DEST_ATTR_PORT, dest->port);
3043
3044 NLA_PUT_U32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3045 atomic_read(&dest->conn_flags) & IP_VS_CONN_F_FWD_MASK);
3046 NLA_PUT_U32(skb, IPVS_DEST_ATTR_WEIGHT, atomic_read(&dest->weight));
3047 NLA_PUT_U32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold);
3048 NLA_PUT_U32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold);
3049 NLA_PUT_U32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3050 atomic_read(&dest->activeconns));
3051 NLA_PUT_U32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3052 atomic_read(&dest->inactconns));
3053 NLA_PUT_U32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3054 atomic_read(&dest->persistconns));
3055
3056 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
3057 goto nla_put_failure;
3058
3059 nla_nest_end(skb, nl_dest);
3060
3061 return 0;
3062
3063nla_put_failure:
3064 nla_nest_cancel(skb, nl_dest);
3065 return -EMSGSIZE;
3066}
3067
3068static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3069 struct netlink_callback *cb)
3070{
3071 void *hdr;
3072
3073 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
3074 &ip_vs_genl_family, NLM_F_MULTI,
3075 IPVS_CMD_NEW_DEST);
3076 if (!hdr)
3077 return -EMSGSIZE;
3078
3079 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3080 goto nla_put_failure;
3081
3082 return genlmsg_end(skb, hdr);
3083
3084nla_put_failure:
3085 genlmsg_cancel(skb, hdr);
3086 return -EMSGSIZE;
3087}
3088
3089static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3090 struct netlink_callback *cb)
3091{
3092 int idx = 0;
3093 int start = cb->args[0];
3094 struct ip_vs_service *svc;
3095 struct ip_vs_dest *dest;
3096 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3097 struct net *net = skb_sknet(skb);
3098
3099 mutex_lock(&__ip_vs_mutex);
3100
3101
3102 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3103 IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3104 goto out_err;
3105
3106
3107 svc = ip_vs_genl_find_service(net, attrs[IPVS_CMD_ATTR_SERVICE]);
3108 if (IS_ERR(svc) || svc == NULL)
3109 goto out_err;
3110
3111
3112 list_for_each_entry(dest, &svc->destinations, n_list) {
3113 if (++idx <= start)
3114 continue;
3115 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3116 idx--;
3117 goto nla_put_failure;
3118 }
3119 }
3120
3121nla_put_failure:
3122 cb->args[0] = idx;
3123
3124out_err:
3125 mutex_unlock(&__ip_vs_mutex);
3126
3127 return skb->len;
3128}
3129
3130static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3131 struct nlattr *nla, int full_entry)
3132{
3133 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3134 struct nlattr *nla_addr, *nla_port;
3135
3136
3137 if (nla == NULL ||
3138 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3139 return -EINVAL;
3140
3141 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3142 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3143
3144 if (!(nla_addr && nla_port))
3145 return -EINVAL;
3146
3147 memset(udest, 0, sizeof(*udest));
3148
3149 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3150 udest->port = nla_get_u16(nla_port);
3151
3152
3153 if (full_entry) {
3154 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3155 *nla_l_thresh;
3156
3157 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3158 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3159 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3160 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3161
3162 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3163 return -EINVAL;
3164
3165 udest->conn_flags = nla_get_u32(nla_fwd)
3166 & IP_VS_CONN_F_FWD_MASK;
3167 udest->weight = nla_get_u32(nla_weight);
3168 udest->u_threshold = nla_get_u32(nla_u_thresh);
3169 udest->l_threshold = nla_get_u32(nla_l_thresh);
3170 }
3171
3172 return 0;
3173}
3174
3175static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __be32 state,
3176 const char *mcast_ifn, __be32 syncid)
3177{
3178 struct nlattr *nl_daemon;
3179
3180 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3181 if (!nl_daemon)
3182 return -EMSGSIZE;
3183
3184 NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_STATE, state);
3185 NLA_PUT_STRING(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn);
3186 NLA_PUT_U32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid);
3187
3188 nla_nest_end(skb, nl_daemon);
3189
3190 return 0;
3191
3192nla_put_failure:
3193 nla_nest_cancel(skb, nl_daemon);
3194 return -EMSGSIZE;
3195}
3196
3197static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __be32 state,
3198 const char *mcast_ifn, __be32 syncid,
3199 struct netlink_callback *cb)
3200{
3201 void *hdr;
3202 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
3203 &ip_vs_genl_family, NLM_F_MULTI,
3204 IPVS_CMD_NEW_DAEMON);
3205 if (!hdr)
3206 return -EMSGSIZE;
3207
3208 if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
3209 goto nla_put_failure;
3210
3211 return genlmsg_end(skb, hdr);
3212
3213nla_put_failure:
3214 genlmsg_cancel(skb, hdr);
3215 return -EMSGSIZE;
3216}
3217
3218static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3219 struct netlink_callback *cb)
3220{
3221 struct net *net = skb_sknet(skb);
3222 struct netns_ipvs *ipvs = net_ipvs(net);
3223
3224 mutex_lock(&ipvs->sync_mutex);
3225 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3226 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3227 ipvs->master_mcast_ifn,
3228 ipvs->master_syncid, cb) < 0)
3229 goto nla_put_failure;
3230
3231 cb->args[0] = 1;
3232 }
3233
3234 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3235 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3236 ipvs->backup_mcast_ifn,
3237 ipvs->backup_syncid, cb) < 0)
3238 goto nla_put_failure;
3239
3240 cb->args[1] = 1;
3241 }
3242
3243nla_put_failure:
3244 mutex_unlock(&ipvs->sync_mutex);
3245
3246 return skb->len;
3247}
3248
3249static int ip_vs_genl_new_daemon(struct net *net, struct nlattr **attrs)
3250{
3251 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3252 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3253 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3254 return -EINVAL;
3255
3256 return start_sync_thread(net,
3257 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3258 nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3259 nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3260}
3261
3262static int ip_vs_genl_del_daemon(struct net *net, struct nlattr **attrs)
3263{
3264 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3265 return -EINVAL;
3266
3267 return stop_sync_thread(net,
3268 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3269}
3270
3271static int ip_vs_genl_set_config(struct net *net, struct nlattr **attrs)
3272{
3273 struct ip_vs_timeout_user t;
3274
3275 __ip_vs_get_timeouts(net, &t);
3276
3277 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3278 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3279
3280 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3281 t.tcp_fin_timeout =
3282 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3283
3284 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3285 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3286
3287 return ip_vs_set_timeout(net, &t);
3288}
3289
3290static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3291{
3292 int ret = 0, cmd;
3293 struct net *net;
3294 struct netns_ipvs *ipvs;
3295
3296 net = skb_sknet(skb);
3297 ipvs = net_ipvs(net);
3298 cmd = info->genlhdr->cmd;
3299
3300 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3301 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3302
3303 mutex_lock(&ipvs->sync_mutex);
3304 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3305 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3306 info->attrs[IPVS_CMD_ATTR_DAEMON],
3307 ip_vs_daemon_policy)) {
3308 ret = -EINVAL;
3309 goto out;
3310 }
3311
3312 if (cmd == IPVS_CMD_NEW_DAEMON)
3313 ret = ip_vs_genl_new_daemon(net, daemon_attrs);
3314 else
3315 ret = ip_vs_genl_del_daemon(net, daemon_attrs);
3316out:
3317 mutex_unlock(&ipvs->sync_mutex);
3318 }
3319 return ret;
3320}
3321
3322static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3323{
3324 struct ip_vs_service *svc = NULL;
3325 struct ip_vs_service_user_kern usvc;
3326 struct ip_vs_dest_user_kern udest;
3327 int ret = 0, cmd;
3328 int need_full_svc = 0, need_full_dest = 0;
3329 struct net *net;
3330
3331 net = skb_sknet(skb);
3332 cmd = info->genlhdr->cmd;
3333
3334 mutex_lock(&__ip_vs_mutex);
3335
3336 if (cmd == IPVS_CMD_FLUSH) {
3337 ret = ip_vs_flush(net);
3338 goto out;
3339 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3340 ret = ip_vs_genl_set_config(net, info->attrs);
3341 goto out;
3342 } else if (cmd == IPVS_CMD_ZERO &&
3343 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3344 ret = ip_vs_zero_all(net);
3345 goto out;
3346 }
3347
3348
3349
3350
3351 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3352 need_full_svc = 1;
3353
3354 ret = ip_vs_genl_parse_service(net, &usvc,
3355 info->attrs[IPVS_CMD_ATTR_SERVICE],
3356 need_full_svc, &svc);
3357 if (ret)
3358 goto out;
3359
3360
3361 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3362 ret = -ESRCH;
3363 goto out;
3364 }
3365
3366
3367
3368
3369 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3370 cmd == IPVS_CMD_DEL_DEST) {
3371 if (cmd != IPVS_CMD_DEL_DEST)
3372 need_full_dest = 1;
3373
3374 ret = ip_vs_genl_parse_dest(&udest,
3375 info->attrs[IPVS_CMD_ATTR_DEST],
3376 need_full_dest);
3377 if (ret)
3378 goto out;
3379 }
3380
3381 switch (cmd) {
3382 case IPVS_CMD_NEW_SERVICE:
3383 if (svc == NULL)
3384 ret = ip_vs_add_service(net, &usvc, &svc);
3385 else
3386 ret = -EEXIST;
3387 break;
3388 case IPVS_CMD_SET_SERVICE:
3389 ret = ip_vs_edit_service(svc, &usvc);
3390 break;
3391 case IPVS_CMD_DEL_SERVICE:
3392 ret = ip_vs_del_service(svc);
3393
3394 break;
3395 case IPVS_CMD_NEW_DEST:
3396 ret = ip_vs_add_dest(svc, &udest);
3397 break;
3398 case IPVS_CMD_SET_DEST:
3399 ret = ip_vs_edit_dest(svc, &udest);
3400 break;
3401 case IPVS_CMD_DEL_DEST:
3402 ret = ip_vs_del_dest(svc, &udest);
3403 break;
3404 case IPVS_CMD_ZERO:
3405 ret = ip_vs_zero_service(svc);
3406 break;
3407 default:
3408 ret = -EINVAL;
3409 }
3410
3411out:
3412 mutex_unlock(&__ip_vs_mutex);
3413
3414 return ret;
3415}
3416
3417static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3418{
3419 struct sk_buff *msg;
3420 void *reply;
3421 int ret, cmd, reply_cmd;
3422 struct net *net;
3423
3424 net = skb_sknet(skb);
3425 cmd = info->genlhdr->cmd;
3426
3427 if (cmd == IPVS_CMD_GET_SERVICE)
3428 reply_cmd = IPVS_CMD_NEW_SERVICE;
3429 else if (cmd == IPVS_CMD_GET_INFO)
3430 reply_cmd = IPVS_CMD_SET_INFO;
3431 else if (cmd == IPVS_CMD_GET_CONFIG)
3432 reply_cmd = IPVS_CMD_SET_CONFIG;
3433 else {
3434 pr_err("unknown Generic Netlink command\n");
3435 return -EINVAL;
3436 }
3437
3438 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3439 if (!msg)
3440 return -ENOMEM;
3441
3442 mutex_lock(&__ip_vs_mutex);
3443
3444 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3445 if (reply == NULL)
3446 goto nla_put_failure;
3447
3448 switch (cmd) {
3449 case IPVS_CMD_GET_SERVICE:
3450 {
3451 struct ip_vs_service *svc;
3452
3453 svc = ip_vs_genl_find_service(net,
3454 info->attrs[IPVS_CMD_ATTR_SERVICE]);
3455 if (IS_ERR(svc)) {
3456 ret = PTR_ERR(svc);
3457 goto out_err;
3458 } else if (svc) {
3459 ret = ip_vs_genl_fill_service(msg, svc);
3460 if (ret)
3461 goto nla_put_failure;
3462 } else {
3463 ret = -ESRCH;
3464 goto out_err;
3465 }
3466
3467 break;
3468 }
3469
3470 case IPVS_CMD_GET_CONFIG:
3471 {
3472 struct ip_vs_timeout_user t;
3473
3474 __ip_vs_get_timeouts(net, &t);
3475#ifdef CONFIG_IP_VS_PROTO_TCP
3476 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP, t.tcp_timeout);
3477 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3478 t.tcp_fin_timeout);
3479#endif
3480#ifdef CONFIG_IP_VS_PROTO_UDP
3481 NLA_PUT_U32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout);
3482#endif
3483
3484 break;
3485 }
3486
3487 case IPVS_CMD_GET_INFO:
3488 NLA_PUT_U32(msg, IPVS_INFO_ATTR_VERSION, IP_VS_VERSION_CODE);
3489 NLA_PUT_U32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3490 ip_vs_conn_tab_size);
3491 break;
3492 }
3493
3494 genlmsg_end(msg, reply);
3495 ret = genlmsg_reply(msg, info);
3496 goto out;
3497
3498nla_put_failure:
3499 pr_err("not enough space in Netlink message\n");
3500 ret = -EMSGSIZE;
3501
3502out_err:
3503 nlmsg_free(msg);
3504out:
3505 mutex_unlock(&__ip_vs_mutex);
3506
3507 return ret;
3508}
3509
3510
3511static struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3512 {
3513 .cmd = IPVS_CMD_NEW_SERVICE,
3514 .flags = GENL_ADMIN_PERM,
3515 .policy = ip_vs_cmd_policy,
3516 .doit = ip_vs_genl_set_cmd,
3517 },
3518 {
3519 .cmd = IPVS_CMD_SET_SERVICE,
3520 .flags = GENL_ADMIN_PERM,
3521 .policy = ip_vs_cmd_policy,
3522 .doit = ip_vs_genl_set_cmd,
3523 },
3524 {
3525 .cmd = IPVS_CMD_DEL_SERVICE,
3526 .flags = GENL_ADMIN_PERM,
3527 .policy = ip_vs_cmd_policy,
3528 .doit = ip_vs_genl_set_cmd,
3529 },
3530 {
3531 .cmd = IPVS_CMD_GET_SERVICE,
3532 .flags = GENL_ADMIN_PERM,
3533 .doit = ip_vs_genl_get_cmd,
3534 .dumpit = ip_vs_genl_dump_services,
3535 .policy = ip_vs_cmd_policy,
3536 },
3537 {
3538 .cmd = IPVS_CMD_NEW_DEST,
3539 .flags = GENL_ADMIN_PERM,
3540 .policy = ip_vs_cmd_policy,
3541 .doit = ip_vs_genl_set_cmd,
3542 },
3543 {
3544 .cmd = IPVS_CMD_SET_DEST,
3545 .flags = GENL_ADMIN_PERM,
3546 .policy = ip_vs_cmd_policy,
3547 .doit = ip_vs_genl_set_cmd,
3548 },
3549 {
3550 .cmd = IPVS_CMD_DEL_DEST,
3551 .flags = GENL_ADMIN_PERM,
3552 .policy = ip_vs_cmd_policy,
3553 .doit = ip_vs_genl_set_cmd,
3554 },
3555 {
3556 .cmd = IPVS_CMD_GET_DEST,
3557 .flags = GENL_ADMIN_PERM,
3558 .policy = ip_vs_cmd_policy,
3559 .dumpit = ip_vs_genl_dump_dests,
3560 },
3561 {
3562 .cmd = IPVS_CMD_NEW_DAEMON,
3563 .flags = GENL_ADMIN_PERM,
3564 .policy = ip_vs_cmd_policy,
3565 .doit = ip_vs_genl_set_daemon,
3566 },
3567 {
3568 .cmd = IPVS_CMD_DEL_DAEMON,
3569 .flags = GENL_ADMIN_PERM,
3570 .policy = ip_vs_cmd_policy,
3571 .doit = ip_vs_genl_set_daemon,
3572 },
3573 {
3574 .cmd = IPVS_CMD_GET_DAEMON,
3575 .flags = GENL_ADMIN_PERM,
3576 .dumpit = ip_vs_genl_dump_daemons,
3577 },
3578 {
3579 .cmd = IPVS_CMD_SET_CONFIG,
3580 .flags = GENL_ADMIN_PERM,
3581 .policy = ip_vs_cmd_policy,
3582 .doit = ip_vs_genl_set_cmd,
3583 },
3584 {
3585 .cmd = IPVS_CMD_GET_CONFIG,
3586 .flags = GENL_ADMIN_PERM,
3587 .doit = ip_vs_genl_get_cmd,
3588 },
3589 {
3590 .cmd = IPVS_CMD_GET_INFO,
3591 .flags = GENL_ADMIN_PERM,
3592 .doit = ip_vs_genl_get_cmd,
3593 },
3594 {
3595 .cmd = IPVS_CMD_ZERO,
3596 .flags = GENL_ADMIN_PERM,
3597 .policy = ip_vs_cmd_policy,
3598 .doit = ip_vs_genl_set_cmd,
3599 },
3600 {
3601 .cmd = IPVS_CMD_FLUSH,
3602 .flags = GENL_ADMIN_PERM,
3603 .doit = ip_vs_genl_set_cmd,
3604 },
3605};
3606
3607static int __init ip_vs_genl_register(void)
3608{
3609 return genl_register_family_with_ops(&ip_vs_genl_family,
3610 ip_vs_genl_ops, ARRAY_SIZE(ip_vs_genl_ops));
3611}
3612
3613static void ip_vs_genl_unregister(void)
3614{
3615 genl_unregister_family(&ip_vs_genl_family);
3616}
3617
3618
3619
3620
3621
3622
3623#ifdef CONFIG_SYSCTL
3624int __net_init ip_vs_control_net_init_sysctl(struct net *net)
3625{
3626 int idx;
3627 struct netns_ipvs *ipvs = net_ipvs(net);
3628 struct ctl_table *tbl;
3629
3630 atomic_set(&ipvs->dropentry, 0);
3631 spin_lock_init(&ipvs->dropentry_lock);
3632 spin_lock_init(&ipvs->droppacket_lock);
3633 spin_lock_init(&ipvs->securetcp_lock);
3634
3635 if (!net_eq(net, &init_net)) {
3636 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3637 if (tbl == NULL)
3638 return -ENOMEM;
3639 } else
3640 tbl = vs_vars;
3641
3642 idx = 0;
3643 ipvs->sysctl_amemthresh = 1024;
3644 tbl[idx++].data = &ipvs->sysctl_amemthresh;
3645 ipvs->sysctl_am_droprate = 10;
3646 tbl[idx++].data = &ipvs->sysctl_am_droprate;
3647 tbl[idx++].data = &ipvs->sysctl_drop_entry;
3648 tbl[idx++].data = &ipvs->sysctl_drop_packet;
3649#ifdef CONFIG_IP_VS_NFCT
3650 tbl[idx++].data = &ipvs->sysctl_conntrack;
3651#endif
3652 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3653 ipvs->sysctl_snat_reroute = 1;
3654 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3655 ipvs->sysctl_sync_ver = 1;
3656 tbl[idx++].data = &ipvs->sysctl_sync_ver;
3657 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3658 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3659 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3660 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3661 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3662 tbl[idx].data = &ipvs->sysctl_sync_threshold;
3663 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3664 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3665
3666
3667 ipvs->sysctl_hdr = register_net_sysctl_table(net, net_vs_ctl_path,
3668 tbl);
3669 if (ipvs->sysctl_hdr == NULL) {
3670 if (!net_eq(net, &init_net))
3671 kfree(tbl);
3672 return -ENOMEM;
3673 }
3674 ip_vs_start_estimator(net, &ipvs->tot_stats);
3675 ipvs->sysctl_tbl = tbl;
3676
3677 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3678 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3679
3680 return 0;
3681}
3682
3683void __net_init ip_vs_control_net_cleanup_sysctl(struct net *net)
3684{
3685 struct netns_ipvs *ipvs = net_ipvs(net);
3686
3687 cancel_delayed_work_sync(&ipvs->defense_work);
3688 cancel_work_sync(&ipvs->defense_work.work);
3689 unregister_net_sysctl_table(ipvs->sysctl_hdr);
3690}
3691
3692#else
3693
3694int __net_init ip_vs_control_net_init_sysctl(struct net *net) { return 0; }
3695void __net_init ip_vs_control_net_cleanup_sysctl(struct net *net) { }
3696
3697#endif
3698
3699static struct notifier_block ip_vs_dst_notifier = {
3700 .notifier_call = ip_vs_dst_event,
3701};
3702
3703int __net_init ip_vs_control_net_init(struct net *net)
3704{
3705 int idx;
3706 struct netns_ipvs *ipvs = net_ipvs(net);
3707
3708 rwlock_init(&ipvs->rs_lock);
3709
3710
3711 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
3712 INIT_LIST_HEAD(&ipvs->rs_table[idx]);
3713
3714 INIT_LIST_HEAD(&ipvs->dest_trash);
3715 atomic_set(&ipvs->ftpsvc_counter, 0);
3716 atomic_set(&ipvs->nullsvc_counter, 0);
3717
3718
3719 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
3720 if (!ipvs->tot_stats.cpustats)
3721 return -ENOMEM;
3722
3723 spin_lock_init(&ipvs->tot_stats.lock);
3724
3725 proc_net_fops_create(net, "ip_vs", 0, &ip_vs_info_fops);
3726 proc_net_fops_create(net, "ip_vs_stats", 0, &ip_vs_stats_fops);
3727 proc_net_fops_create(net, "ip_vs_stats_percpu", 0,
3728 &ip_vs_stats_percpu_fops);
3729
3730 if (ip_vs_control_net_init_sysctl(net))
3731 goto err;
3732
3733 return 0;
3734
3735err:
3736 free_percpu(ipvs->tot_stats.cpustats);
3737 return -ENOMEM;
3738}
3739
3740void __net_exit ip_vs_control_net_cleanup(struct net *net)
3741{
3742 struct netns_ipvs *ipvs = net_ipvs(net);
3743
3744 ip_vs_trash_cleanup(net);
3745 ip_vs_stop_estimator(net, &ipvs->tot_stats);
3746 ip_vs_control_net_cleanup_sysctl(net);
3747 proc_net_remove(net, "ip_vs_stats_percpu");
3748 proc_net_remove(net, "ip_vs_stats");
3749 proc_net_remove(net, "ip_vs");
3750 free_percpu(ipvs->tot_stats.cpustats);
3751}
3752
3753int __init ip_vs_control_init(void)
3754{
3755 int idx;
3756 int ret;
3757
3758 EnterFunction(2);
3759
3760
3761 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
3762 INIT_LIST_HEAD(&ip_vs_svc_table[idx]);
3763 INIT_LIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3764 }
3765
3766 smp_wmb();
3767
3768 ret = nf_register_sockopt(&ip_vs_sockopts);
3769 if (ret) {
3770 pr_err("cannot register sockopt.\n");
3771 goto err_sock;
3772 }
3773
3774 ret = ip_vs_genl_register();
3775 if (ret) {
3776 pr_err("cannot register Generic Netlink interface.\n");
3777 goto err_genl;
3778 }
3779
3780 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
3781 if (ret < 0)
3782 goto err_notf;
3783
3784 LeaveFunction(2);
3785 return 0;
3786
3787err_notf:
3788 ip_vs_genl_unregister();
3789err_genl:
3790 nf_unregister_sockopt(&ip_vs_sockopts);
3791err_sock:
3792 return ret;
3793}
3794
3795
3796void ip_vs_control_cleanup(void)
3797{
3798 EnterFunction(2);
3799 unregister_netdevice_notifier(&ip_vs_dst_notifier);
3800 ip_vs_genl_unregister();
3801 nf_unregister_sockopt(&ip_vs_sockopts);
3802 LeaveFunction(2);
3803}
3804