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 <linux/uaccess.h>
52
53#include <net/ip_vs.h>
54
55
56static DEFINE_MUTEX(__ip_vs_mutex);
57
58
59
60#ifdef CONFIG_IP_VS_DEBUG
61static int sysctl_ip_vs_debug_level = 0;
62
63int ip_vs_get_debug_level(void)
64{
65 return sysctl_ip_vs_debug_level;
66}
67#endif
68
69
70
71static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup);
72
73
74#ifdef CONFIG_IP_VS_IPV6
75
76static bool __ip_vs_addr_is_local_v6(struct net *net,
77 const struct in6_addr *addr)
78{
79 struct flowi6 fl6 = {
80 .daddr = *addr,
81 };
82 struct dst_entry *dst = ip6_route_output(net, NULL, &fl6);
83 bool is_local;
84
85 is_local = !dst->error && dst->dev && (dst->dev->flags & IFF_LOOPBACK);
86
87 dst_release(dst);
88 return is_local;
89}
90#endif
91
92#ifdef CONFIG_SYSCTL
93
94
95
96
97static void update_defense_level(struct netns_ipvs *ipvs)
98{
99 struct sysinfo i;
100 static int old_secure_tcp = 0;
101 int availmem;
102 int nomem;
103 int to_change = -1;
104
105
106 si_meminfo(&i);
107 availmem = i.freeram + i.bufferram;
108
109
110
111
112
113 nomem = (availmem < ipvs->sysctl_amemthresh);
114
115 local_bh_disable();
116
117
118 spin_lock(&ipvs->dropentry_lock);
119 switch (ipvs->sysctl_drop_entry) {
120 case 0:
121 atomic_set(&ipvs->dropentry, 0);
122 break;
123 case 1:
124 if (nomem) {
125 atomic_set(&ipvs->dropentry, 1);
126 ipvs->sysctl_drop_entry = 2;
127 } else {
128 atomic_set(&ipvs->dropentry, 0);
129 }
130 break;
131 case 2:
132 if (nomem) {
133 atomic_set(&ipvs->dropentry, 1);
134 } else {
135 atomic_set(&ipvs->dropentry, 0);
136 ipvs->sysctl_drop_entry = 1;
137 };
138 break;
139 case 3:
140 atomic_set(&ipvs->dropentry, 1);
141 break;
142 }
143 spin_unlock(&ipvs->dropentry_lock);
144
145
146 spin_lock(&ipvs->droppacket_lock);
147 switch (ipvs->sysctl_drop_packet) {
148 case 0:
149 ipvs->drop_rate = 0;
150 break;
151 case 1:
152 if (nomem) {
153 ipvs->drop_rate = ipvs->drop_counter
154 = ipvs->sysctl_amemthresh /
155 (ipvs->sysctl_amemthresh-availmem);
156 ipvs->sysctl_drop_packet = 2;
157 } else {
158 ipvs->drop_rate = 0;
159 }
160 break;
161 case 2:
162 if (nomem) {
163 ipvs->drop_rate = ipvs->drop_counter
164 = ipvs->sysctl_amemthresh /
165 (ipvs->sysctl_amemthresh-availmem);
166 } else {
167 ipvs->drop_rate = 0;
168 ipvs->sysctl_drop_packet = 1;
169 }
170 break;
171 case 3:
172 ipvs->drop_rate = ipvs->sysctl_am_droprate;
173 break;
174 }
175 spin_unlock(&ipvs->droppacket_lock);
176
177
178 spin_lock(&ipvs->securetcp_lock);
179 switch (ipvs->sysctl_secure_tcp) {
180 case 0:
181 if (old_secure_tcp >= 2)
182 to_change = 0;
183 break;
184 case 1:
185 if (nomem) {
186 if (old_secure_tcp < 2)
187 to_change = 1;
188 ipvs->sysctl_secure_tcp = 2;
189 } else {
190 if (old_secure_tcp >= 2)
191 to_change = 0;
192 }
193 break;
194 case 2:
195 if (nomem) {
196 if (old_secure_tcp < 2)
197 to_change = 1;
198 } else {
199 if (old_secure_tcp >= 2)
200 to_change = 0;
201 ipvs->sysctl_secure_tcp = 1;
202 }
203 break;
204 case 3:
205 if (old_secure_tcp < 2)
206 to_change = 1;
207 break;
208 }
209 old_secure_tcp = ipvs->sysctl_secure_tcp;
210 if (to_change >= 0)
211 ip_vs_protocol_timeout_change(ipvs,
212 ipvs->sysctl_secure_tcp > 1);
213 spin_unlock(&ipvs->securetcp_lock);
214
215 local_bh_enable();
216}
217
218
219
220
221
222#define DEFENSE_TIMER_PERIOD 1*HZ
223
224static void defense_work_handler(struct work_struct *work)
225{
226 struct netns_ipvs *ipvs =
227 container_of(work, struct netns_ipvs, defense_work.work);
228
229 update_defense_level(ipvs);
230 if (atomic_read(&ipvs->dropentry))
231 ip_vs_random_dropentry(ipvs);
232 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
233}
234#endif
235
236int
237ip_vs_use_count_inc(void)
238{
239 return try_module_get(THIS_MODULE);
240}
241
242void
243ip_vs_use_count_dec(void)
244{
245 module_put(THIS_MODULE);
246}
247
248
249
250
251
252#define IP_VS_SVC_TAB_BITS 8
253#define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
254#define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)
255
256
257static struct hlist_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE];
258
259static struct hlist_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE];
260
261
262
263
264
265static inline unsigned int
266ip_vs_svc_hashkey(struct netns_ipvs *ipvs, int af, unsigned int proto,
267 const union nf_inet_addr *addr, __be16 port)
268{
269 register unsigned int porth = ntohs(port);
270 __be32 addr_fold = addr->ip;
271 __u32 ahash;
272
273#ifdef CONFIG_IP_VS_IPV6
274 if (af == AF_INET6)
275 addr_fold = addr->ip6[0]^addr->ip6[1]^
276 addr->ip6[2]^addr->ip6[3];
277#endif
278 ahash = ntohl(addr_fold);
279 ahash ^= ((size_t) ipvs >> 8);
280
281 return (proto ^ ahash ^ (porth >> IP_VS_SVC_TAB_BITS) ^ porth) &
282 IP_VS_SVC_TAB_MASK;
283}
284
285
286
287
288static inline unsigned int ip_vs_svc_fwm_hashkey(struct netns_ipvs *ipvs, __u32 fwmark)
289{
290 return (((size_t)ipvs>>8) ^ fwmark) & IP_VS_SVC_TAB_MASK;
291}
292
293
294
295
296
297
298static int ip_vs_svc_hash(struct ip_vs_service *svc)
299{
300 unsigned int hash;
301
302 if (svc->flags & IP_VS_SVC_F_HASHED) {
303 pr_err("%s(): request for already hashed, called from %pS\n",
304 __func__, __builtin_return_address(0));
305 return 0;
306 }
307
308 if (svc->fwmark == 0) {
309
310
311
312 hash = ip_vs_svc_hashkey(svc->ipvs, svc->af, svc->protocol,
313 &svc->addr, svc->port);
314 hlist_add_head_rcu(&svc->s_list, &ip_vs_svc_table[hash]);
315 } else {
316
317
318
319 hash = ip_vs_svc_fwm_hashkey(svc->ipvs, svc->fwmark);
320 hlist_add_head_rcu(&svc->f_list, &ip_vs_svc_fwm_table[hash]);
321 }
322
323 svc->flags |= IP_VS_SVC_F_HASHED;
324
325 atomic_inc(&svc->refcnt);
326 return 1;
327}
328
329
330
331
332
333
334static int ip_vs_svc_unhash(struct ip_vs_service *svc)
335{
336 if (!(svc->flags & IP_VS_SVC_F_HASHED)) {
337 pr_err("%s(): request for unhash flagged, called from %pS\n",
338 __func__, __builtin_return_address(0));
339 return 0;
340 }
341
342 if (svc->fwmark == 0) {
343
344 hlist_del_rcu(&svc->s_list);
345 } else {
346
347 hlist_del_rcu(&svc->f_list);
348 }
349
350 svc->flags &= ~IP_VS_SVC_F_HASHED;
351 atomic_dec(&svc->refcnt);
352 return 1;
353}
354
355
356
357
358
359static inline struct ip_vs_service *
360__ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u16 protocol,
361 const union nf_inet_addr *vaddr, __be16 vport)
362{
363 unsigned int hash;
364 struct ip_vs_service *svc;
365
366
367 hash = ip_vs_svc_hashkey(ipvs, af, protocol, vaddr, vport);
368
369 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[hash], s_list) {
370 if ((svc->af == af)
371 && ip_vs_addr_equal(af, &svc->addr, vaddr)
372 && (svc->port == vport)
373 && (svc->protocol == protocol)
374 && (svc->ipvs == ipvs)) {
375
376 return svc;
377 }
378 }
379
380 return NULL;
381}
382
383
384
385
386
387static inline struct ip_vs_service *
388__ip_vs_svc_fwm_find(struct netns_ipvs *ipvs, int af, __u32 fwmark)
389{
390 unsigned int hash;
391 struct ip_vs_service *svc;
392
393
394 hash = ip_vs_svc_fwm_hashkey(ipvs, fwmark);
395
396 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[hash], f_list) {
397 if (svc->fwmark == fwmark && svc->af == af
398 && (svc->ipvs == ipvs)) {
399
400 return svc;
401 }
402 }
403
404 return NULL;
405}
406
407
408struct ip_vs_service *
409ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
410 const union nf_inet_addr *vaddr, __be16 vport)
411{
412 struct ip_vs_service *svc;
413
414
415
416
417 if (fwmark) {
418 svc = __ip_vs_svc_fwm_find(ipvs, af, fwmark);
419 if (svc)
420 goto out;
421 }
422
423
424
425
426
427 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, vport);
428
429 if (!svc && protocol == IPPROTO_TCP &&
430 atomic_read(&ipvs->ftpsvc_counter) &&
431 (vport == FTPDATA || ntohs(vport) >= inet_prot_sock(ipvs->net))) {
432
433
434
435
436 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, FTPPORT);
437 }
438
439 if (svc == NULL
440 && atomic_read(&ipvs->nullsvc_counter)) {
441
442
443
444 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, 0);
445 }
446
447 out:
448 IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
449 fwmark, ip_vs_proto_name(protocol),
450 IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
451 svc ? "hit" : "not hit");
452
453 return svc;
454}
455
456
457static inline void
458__ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
459{
460 atomic_inc(&svc->refcnt);
461 rcu_assign_pointer(dest->svc, svc);
462}
463
464static void ip_vs_service_free(struct ip_vs_service *svc)
465{
466 free_percpu(svc->stats.cpustats);
467 kfree(svc);
468}
469
470static void ip_vs_service_rcu_free(struct rcu_head *head)
471{
472 struct ip_vs_service *svc;
473
474 svc = container_of(head, struct ip_vs_service, rcu_head);
475 ip_vs_service_free(svc);
476}
477
478static void __ip_vs_svc_put(struct ip_vs_service *svc, bool do_delay)
479{
480 if (atomic_dec_and_test(&svc->refcnt)) {
481 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n",
482 svc->fwmark,
483 IP_VS_DBG_ADDR(svc->af, &svc->addr),
484 ntohs(svc->port));
485 if (do_delay)
486 call_rcu(&svc->rcu_head, ip_vs_service_rcu_free);
487 else
488 ip_vs_service_free(svc);
489 }
490}
491
492
493
494
495
496static inline unsigned int ip_vs_rs_hashkey(int af,
497 const union nf_inet_addr *addr,
498 __be16 port)
499{
500 register unsigned int porth = ntohs(port);
501 __be32 addr_fold = addr->ip;
502
503#ifdef CONFIG_IP_VS_IPV6
504 if (af == AF_INET6)
505 addr_fold = addr->ip6[0]^addr->ip6[1]^
506 addr->ip6[2]^addr->ip6[3];
507#endif
508
509 return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
510 & IP_VS_RTAB_MASK;
511}
512
513
514static void ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
515{
516 unsigned int hash;
517
518 if (dest->in_rs_table)
519 return;
520
521
522
523
524
525 hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
526
527 hlist_add_head_rcu(&dest->d_list, &ipvs->rs_table[hash]);
528 dest->in_rs_table = 1;
529}
530
531
532static void ip_vs_rs_unhash(struct ip_vs_dest *dest)
533{
534
535
536
537 if (dest->in_rs_table) {
538 hlist_del_rcu(&dest->d_list);
539 dest->in_rs_table = 0;
540 }
541}
542
543
544bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
545 const union nf_inet_addr *daddr, __be16 dport)
546{
547 unsigned int hash;
548 struct ip_vs_dest *dest;
549
550
551 hash = ip_vs_rs_hashkey(af, daddr, dport);
552
553 hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
554 if (dest->port == dport &&
555 dest->af == af &&
556 ip_vs_addr_equal(af, &dest->addr, daddr) &&
557 (dest->protocol == protocol || dest->vfwmark)) {
558
559 return true;
560 }
561 }
562
563 return false;
564}
565
566
567
568
569
570
571
572struct ip_vs_dest *ip_vs_find_real_service(struct netns_ipvs *ipvs, int af,
573 __u16 protocol,
574 const union nf_inet_addr *daddr,
575 __be16 dport)
576{
577 unsigned int hash;
578 struct ip_vs_dest *dest;
579
580
581 hash = ip_vs_rs_hashkey(af, daddr, dport);
582
583 hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
584 if (dest->port == dport &&
585 dest->af == af &&
586 ip_vs_addr_equal(af, &dest->addr, daddr) &&
587 (dest->protocol == protocol || dest->vfwmark)) {
588
589 return dest;
590 }
591 }
592
593 return NULL;
594}
595
596
597
598
599static struct ip_vs_dest *
600ip_vs_lookup_dest(struct ip_vs_service *svc, int dest_af,
601 const union nf_inet_addr *daddr, __be16 dport)
602{
603 struct ip_vs_dest *dest;
604
605
606
607
608 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
609 if ((dest->af == dest_af) &&
610 ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
611 (dest->port == dport)) {
612
613 return dest;
614 }
615 }
616
617 return NULL;
618}
619
620
621
622
623
624
625
626
627
628struct ip_vs_dest *ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
629 const union nf_inet_addr *daddr,
630 __be16 dport,
631 const union nf_inet_addr *vaddr,
632 __be16 vport, __u16 protocol, __u32 fwmark,
633 __u32 flags)
634{
635 struct ip_vs_dest *dest;
636 struct ip_vs_service *svc;
637 __be16 port = dport;
638
639 svc = ip_vs_service_find(ipvs, svc_af, fwmark, protocol, vaddr, vport);
640 if (!svc)
641 return NULL;
642 if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
643 port = 0;
644 dest = ip_vs_lookup_dest(svc, dest_af, daddr, port);
645 if (!dest)
646 dest = ip_vs_lookup_dest(svc, dest_af, daddr, port ^ dport);
647 return dest;
648}
649
650void ip_vs_dest_dst_rcu_free(struct rcu_head *head)
651{
652 struct ip_vs_dest_dst *dest_dst = container_of(head,
653 struct ip_vs_dest_dst,
654 rcu_head);
655
656 dst_release(dest_dst->dst_cache);
657 kfree(dest_dst);
658}
659
660
661static void __ip_vs_dst_cache_reset(struct ip_vs_dest *dest)
662{
663 struct ip_vs_dest_dst *old;
664
665 old = rcu_dereference_protected(dest->dest_dst, 1);
666 if (old) {
667 RCU_INIT_POINTER(dest->dest_dst, NULL);
668 call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
669 }
670}
671
672
673
674
675
676
677
678
679
680
681
682static struct ip_vs_dest *
683ip_vs_trash_get_dest(struct ip_vs_service *svc, int dest_af,
684 const union nf_inet_addr *daddr, __be16 dport)
685{
686 struct ip_vs_dest *dest;
687 struct netns_ipvs *ipvs = svc->ipvs;
688
689
690
691
692 spin_lock_bh(&ipvs->dest_trash_lock);
693 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
694 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
695 "dest->refcnt=%d\n",
696 dest->vfwmark,
697 IP_VS_DBG_ADDR(dest->af, &dest->addr),
698 ntohs(dest->port),
699 refcount_read(&dest->refcnt));
700 if (dest->af == dest_af &&
701 ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
702 dest->port == dport &&
703 dest->vfwmark == svc->fwmark &&
704 dest->protocol == svc->protocol &&
705 (svc->fwmark ||
706 (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
707 dest->vport == svc->port))) {
708
709 list_del(&dest->t_list);
710 goto out;
711 }
712 }
713
714 dest = NULL;
715
716out:
717 spin_unlock_bh(&ipvs->dest_trash_lock);
718
719 return dest;
720}
721
722static void ip_vs_dest_free(struct ip_vs_dest *dest)
723{
724 struct ip_vs_service *svc = rcu_dereference_protected(dest->svc, 1);
725
726 __ip_vs_dst_cache_reset(dest);
727 __ip_vs_svc_put(svc, false);
728 free_percpu(dest->stats.cpustats);
729 ip_vs_dest_put_and_free(dest);
730}
731
732
733
734
735
736
737
738
739
740
741static void ip_vs_trash_cleanup(struct netns_ipvs *ipvs)
742{
743 struct ip_vs_dest *dest, *nxt;
744
745 del_timer_sync(&ipvs->dest_trash_timer);
746
747 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, t_list) {
748 list_del(&dest->t_list);
749 ip_vs_dest_free(dest);
750 }
751}
752
753static void
754ip_vs_copy_stats(struct ip_vs_kstats *dst, struct ip_vs_stats *src)
755{
756#define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->kstats.c - src->kstats0.c
757
758 spin_lock_bh(&src->lock);
759
760 IP_VS_SHOW_STATS_COUNTER(conns);
761 IP_VS_SHOW_STATS_COUNTER(inpkts);
762 IP_VS_SHOW_STATS_COUNTER(outpkts);
763 IP_VS_SHOW_STATS_COUNTER(inbytes);
764 IP_VS_SHOW_STATS_COUNTER(outbytes);
765
766 ip_vs_read_estimator(dst, src);
767
768 spin_unlock_bh(&src->lock);
769}
770
771static void
772ip_vs_export_stats_user(struct ip_vs_stats_user *dst, struct ip_vs_kstats *src)
773{
774 dst->conns = (u32)src->conns;
775 dst->inpkts = (u32)src->inpkts;
776 dst->outpkts = (u32)src->outpkts;
777 dst->inbytes = src->inbytes;
778 dst->outbytes = src->outbytes;
779 dst->cps = (u32)src->cps;
780 dst->inpps = (u32)src->inpps;
781 dst->outpps = (u32)src->outpps;
782 dst->inbps = (u32)src->inbps;
783 dst->outbps = (u32)src->outbps;
784}
785
786static void
787ip_vs_zero_stats(struct ip_vs_stats *stats)
788{
789 spin_lock_bh(&stats->lock);
790
791
792
793#define IP_VS_ZERO_STATS_COUNTER(c) stats->kstats0.c = stats->kstats.c
794
795 IP_VS_ZERO_STATS_COUNTER(conns);
796 IP_VS_ZERO_STATS_COUNTER(inpkts);
797 IP_VS_ZERO_STATS_COUNTER(outpkts);
798 IP_VS_ZERO_STATS_COUNTER(inbytes);
799 IP_VS_ZERO_STATS_COUNTER(outbytes);
800
801 ip_vs_zero_estimator(stats);
802
803 spin_unlock_bh(&stats->lock);
804}
805
806
807
808
809static void
810__ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
811 struct ip_vs_dest_user_kern *udest, int add)
812{
813 struct netns_ipvs *ipvs = svc->ipvs;
814 struct ip_vs_service *old_svc;
815 struct ip_vs_scheduler *sched;
816 int conn_flags;
817
818
819 BUG_ON(!add && udest->af != dest->af);
820
821 if (add && udest->af != svc->af)
822 ipvs->mixed_address_family_dests++;
823
824
825 atomic_set(&dest->weight, udest->weight);
826 conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
827 conn_flags |= IP_VS_CONN_F_INACTIVE;
828
829
830 if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
831 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
832 } else {
833
834
835
836
837 ip_vs_rs_hash(ipvs, dest);
838 }
839 atomic_set(&dest->conn_flags, conn_flags);
840
841
842 old_svc = rcu_dereference_protected(dest->svc, 1);
843 if (!old_svc) {
844 __ip_vs_bind_svc(dest, svc);
845 } else {
846 if (old_svc != svc) {
847 ip_vs_zero_stats(&dest->stats);
848 __ip_vs_bind_svc(dest, svc);
849 __ip_vs_svc_put(old_svc, true);
850 }
851 }
852
853
854 dest->flags |= IP_VS_DEST_F_AVAILABLE;
855
856 if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
857 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
858 dest->u_threshold = udest->u_threshold;
859 dest->l_threshold = udest->l_threshold;
860
861 dest->af = udest->af;
862
863 spin_lock_bh(&dest->dst_lock);
864 __ip_vs_dst_cache_reset(dest);
865 spin_unlock_bh(&dest->dst_lock);
866
867 if (add) {
868 ip_vs_start_estimator(svc->ipvs, &dest->stats);
869 list_add_rcu(&dest->n_list, &svc->destinations);
870 svc->num_dests++;
871 sched = rcu_dereference_protected(svc->scheduler, 1);
872 if (sched && sched->add_dest)
873 sched->add_dest(svc, dest);
874 } else {
875 sched = rcu_dereference_protected(svc->scheduler, 1);
876 if (sched && sched->upd_dest)
877 sched->upd_dest(svc, dest);
878 }
879}
880
881
882
883
884
885static int
886ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
887 struct ip_vs_dest **dest_p)
888{
889 struct ip_vs_dest *dest;
890 unsigned int atype, i;
891
892 EnterFunction(2);
893
894#ifdef CONFIG_IP_VS_IPV6
895 if (udest->af == AF_INET6) {
896 atype = ipv6_addr_type(&udest->addr.in6);
897 if ((!(atype & IPV6_ADDR_UNICAST) ||
898 atype & IPV6_ADDR_LINKLOCAL) &&
899 !__ip_vs_addr_is_local_v6(svc->ipvs->net, &udest->addr.in6))
900 return -EINVAL;
901 } else
902#endif
903 {
904 atype = inet_addr_type(svc->ipvs->net, udest->addr.ip);
905 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
906 return -EINVAL;
907 }
908
909 dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
910 if (dest == NULL)
911 return -ENOMEM;
912
913 dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
914 if (!dest->stats.cpustats)
915 goto err_alloc;
916
917 for_each_possible_cpu(i) {
918 struct ip_vs_cpu_stats *ip_vs_dest_stats;
919 ip_vs_dest_stats = per_cpu_ptr(dest->stats.cpustats, i);
920 u64_stats_init(&ip_vs_dest_stats->syncp);
921 }
922
923 dest->af = udest->af;
924 dest->protocol = svc->protocol;
925 dest->vaddr = svc->addr;
926 dest->vport = svc->port;
927 dest->vfwmark = svc->fwmark;
928 ip_vs_addr_copy(udest->af, &dest->addr, &udest->addr);
929 dest->port = udest->port;
930
931 atomic_set(&dest->activeconns, 0);
932 atomic_set(&dest->inactconns, 0);
933 atomic_set(&dest->persistconns, 0);
934 refcount_set(&dest->refcnt, 1);
935
936 INIT_HLIST_NODE(&dest->d_list);
937 spin_lock_init(&dest->dst_lock);
938 spin_lock_init(&dest->stats.lock);
939 __ip_vs_update_dest(svc, dest, udest, 1);
940
941 *dest_p = dest;
942
943 LeaveFunction(2);
944 return 0;
945
946err_alloc:
947 kfree(dest);
948 return -ENOMEM;
949}
950
951
952
953
954
955static int
956ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
957{
958 struct ip_vs_dest *dest;
959 union nf_inet_addr daddr;
960 __be16 dport = udest->port;
961 int ret;
962
963 EnterFunction(2);
964
965 if (udest->weight < 0) {
966 pr_err("%s(): server weight less than zero\n", __func__);
967 return -ERANGE;
968 }
969
970 if (udest->l_threshold > udest->u_threshold) {
971 pr_err("%s(): lower threshold is higher than upper threshold\n",
972 __func__);
973 return -ERANGE;
974 }
975
976 ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
977
978
979 rcu_read_lock();
980 dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
981 rcu_read_unlock();
982
983 if (dest != NULL) {
984 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
985 return -EEXIST;
986 }
987
988
989
990
991
992 dest = ip_vs_trash_get_dest(svc, udest->af, &daddr, dport);
993
994 if (dest != NULL) {
995 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
996 "dest->refcnt=%d, service %u/%s:%u\n",
997 IP_VS_DBG_ADDR(udest->af, &daddr), ntohs(dport),
998 refcount_read(&dest->refcnt),
999 dest->vfwmark,
1000 IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
1001 ntohs(dest->vport));
1002
1003 __ip_vs_update_dest(svc, dest, udest, 1);
1004 ret = 0;
1005 } else {
1006
1007
1008
1009 ret = ip_vs_new_dest(svc, udest, &dest);
1010 }
1011 LeaveFunction(2);
1012
1013 return ret;
1014}
1015
1016
1017
1018
1019
1020static int
1021ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1022{
1023 struct ip_vs_dest *dest;
1024 union nf_inet_addr daddr;
1025 __be16 dport = udest->port;
1026
1027 EnterFunction(2);
1028
1029 if (udest->weight < 0) {
1030 pr_err("%s(): server weight less than zero\n", __func__);
1031 return -ERANGE;
1032 }
1033
1034 if (udest->l_threshold > udest->u_threshold) {
1035 pr_err("%s(): lower threshold is higher than upper threshold\n",
1036 __func__);
1037 return -ERANGE;
1038 }
1039
1040 ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
1041
1042
1043 rcu_read_lock();
1044 dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
1045 rcu_read_unlock();
1046
1047 if (dest == NULL) {
1048 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1049 return -ENOENT;
1050 }
1051
1052 __ip_vs_update_dest(svc, dest, udest, 0);
1053 LeaveFunction(2);
1054
1055 return 0;
1056}
1057
1058
1059
1060
1061static void __ip_vs_del_dest(struct netns_ipvs *ipvs, struct ip_vs_dest *dest,
1062 bool cleanup)
1063{
1064 ip_vs_stop_estimator(ipvs, &dest->stats);
1065
1066
1067
1068
1069 ip_vs_rs_unhash(dest);
1070
1071 spin_lock_bh(&ipvs->dest_trash_lock);
1072 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1073 IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port),
1074 refcount_read(&dest->refcnt));
1075 if (list_empty(&ipvs->dest_trash) && !cleanup)
1076 mod_timer(&ipvs->dest_trash_timer,
1077 jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1078
1079 list_add(&dest->t_list, &ipvs->dest_trash);
1080 dest->idle_start = 0;
1081 spin_unlock_bh(&ipvs->dest_trash_lock);
1082}
1083
1084
1085
1086
1087
1088static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1089 struct ip_vs_dest *dest,
1090 int svcupd)
1091{
1092 dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1093
1094
1095
1096
1097 list_del_rcu(&dest->n_list);
1098 svc->num_dests--;
1099
1100 if (dest->af != svc->af)
1101 svc->ipvs->mixed_address_family_dests--;
1102
1103 if (svcupd) {
1104 struct ip_vs_scheduler *sched;
1105
1106 sched = rcu_dereference_protected(svc->scheduler, 1);
1107 if (sched && sched->del_dest)
1108 sched->del_dest(svc, dest);
1109 }
1110}
1111
1112
1113
1114
1115
1116static int
1117ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1118{
1119 struct ip_vs_dest *dest;
1120 __be16 dport = udest->port;
1121
1122 EnterFunction(2);
1123
1124
1125 rcu_read_lock();
1126 dest = ip_vs_lookup_dest(svc, udest->af, &udest->addr, dport);
1127 rcu_read_unlock();
1128
1129 if (dest == NULL) {
1130 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1131 return -ENOENT;
1132 }
1133
1134
1135
1136
1137 __ip_vs_unlink_dest(svc, dest, 1);
1138
1139
1140
1141
1142 __ip_vs_del_dest(svc->ipvs, dest, false);
1143
1144 LeaveFunction(2);
1145
1146 return 0;
1147}
1148
1149static void ip_vs_dest_trash_expire(struct timer_list *t)
1150{
1151 struct netns_ipvs *ipvs = from_timer(ipvs, t, dest_trash_timer);
1152 struct ip_vs_dest *dest, *next;
1153 unsigned long now = jiffies;
1154
1155 spin_lock(&ipvs->dest_trash_lock);
1156 list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) {
1157 if (refcount_read(&dest->refcnt) > 1)
1158 continue;
1159 if (dest->idle_start) {
1160 if (time_before(now, dest->idle_start +
1161 IP_VS_DEST_TRASH_PERIOD))
1162 continue;
1163 } else {
1164 dest->idle_start = max(1UL, now);
1165 continue;
1166 }
1167 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1168 dest->vfwmark,
1169 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1170 ntohs(dest->port));
1171 list_del(&dest->t_list);
1172 ip_vs_dest_free(dest);
1173 }
1174 if (!list_empty(&ipvs->dest_trash))
1175 mod_timer(&ipvs->dest_trash_timer,
1176 jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1177 spin_unlock(&ipvs->dest_trash_lock);
1178}
1179
1180
1181
1182
1183static int
1184ip_vs_add_service(struct netns_ipvs *ipvs, struct ip_vs_service_user_kern *u,
1185 struct ip_vs_service **svc_p)
1186{
1187 int ret = 0, i;
1188 struct ip_vs_scheduler *sched = NULL;
1189 struct ip_vs_pe *pe = NULL;
1190 struct ip_vs_service *svc = NULL;
1191
1192
1193 ip_vs_use_count_inc();
1194
1195
1196 if (strcmp(u->sched_name, "none")) {
1197 sched = ip_vs_scheduler_get(u->sched_name);
1198 if (!sched) {
1199 pr_info("Scheduler module ip_vs_%s not found\n",
1200 u->sched_name);
1201 ret = -ENOENT;
1202 goto out_err;
1203 }
1204 }
1205
1206 if (u->pe_name && *u->pe_name) {
1207 pe = ip_vs_pe_getbyname(u->pe_name);
1208 if (pe == NULL) {
1209 pr_info("persistence engine module ip_vs_pe_%s "
1210 "not found\n", u->pe_name);
1211 ret = -ENOENT;
1212 goto out_err;
1213 }
1214 }
1215
1216#ifdef CONFIG_IP_VS_IPV6
1217 if (u->af == AF_INET6) {
1218 __u32 plen = (__force __u32) u->netmask;
1219
1220 if (plen < 1 || plen > 128) {
1221 ret = -EINVAL;
1222 goto out_err;
1223 }
1224 }
1225#endif
1226
1227 svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1228 if (svc == NULL) {
1229 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1230 ret = -ENOMEM;
1231 goto out_err;
1232 }
1233 svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1234 if (!svc->stats.cpustats) {
1235 ret = -ENOMEM;
1236 goto out_err;
1237 }
1238
1239 for_each_possible_cpu(i) {
1240 struct ip_vs_cpu_stats *ip_vs_stats;
1241 ip_vs_stats = per_cpu_ptr(svc->stats.cpustats, i);
1242 u64_stats_init(&ip_vs_stats->syncp);
1243 }
1244
1245
1246
1247 atomic_set(&svc->refcnt, 0);
1248
1249 svc->af = u->af;
1250 svc->protocol = u->protocol;
1251 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1252 svc->port = u->port;
1253 svc->fwmark = u->fwmark;
1254 svc->flags = u->flags;
1255 svc->timeout = u->timeout * HZ;
1256 svc->netmask = u->netmask;
1257 svc->ipvs = ipvs;
1258
1259 INIT_LIST_HEAD(&svc->destinations);
1260 spin_lock_init(&svc->sched_lock);
1261 spin_lock_init(&svc->stats.lock);
1262
1263
1264 if (sched) {
1265 ret = ip_vs_bind_scheduler(svc, sched);
1266 if (ret)
1267 goto out_err;
1268 sched = NULL;
1269 }
1270
1271
1272 RCU_INIT_POINTER(svc->pe, pe);
1273 pe = NULL;
1274
1275
1276 if (svc->port == FTPPORT)
1277 atomic_inc(&ipvs->ftpsvc_counter);
1278 else if (svc->port == 0)
1279 atomic_inc(&ipvs->nullsvc_counter);
1280 if (svc->pe && svc->pe->conn_out)
1281 atomic_inc(&ipvs->conn_out_counter);
1282
1283 ip_vs_start_estimator(ipvs, &svc->stats);
1284
1285
1286 if (svc->af == AF_INET)
1287 ipvs->num_services++;
1288
1289
1290 ip_vs_svc_hash(svc);
1291
1292 *svc_p = svc;
1293
1294 ipvs->enable = 1;
1295 return 0;
1296
1297
1298 out_err:
1299 if (svc != NULL) {
1300 ip_vs_unbind_scheduler(svc, sched);
1301 ip_vs_service_free(svc);
1302 }
1303 ip_vs_scheduler_put(sched);
1304 ip_vs_pe_put(pe);
1305
1306
1307 ip_vs_use_count_dec();
1308
1309 return ret;
1310}
1311
1312
1313
1314
1315
1316static int
1317ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1318{
1319 struct ip_vs_scheduler *sched = NULL, *old_sched;
1320 struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1321 int ret = 0;
1322 bool new_pe_conn_out, old_pe_conn_out;
1323
1324
1325
1326
1327 if (strcmp(u->sched_name, "none")) {
1328 sched = ip_vs_scheduler_get(u->sched_name);
1329 if (!sched) {
1330 pr_info("Scheduler module ip_vs_%s not found\n",
1331 u->sched_name);
1332 return -ENOENT;
1333 }
1334 }
1335 old_sched = sched;
1336
1337 if (u->pe_name && *u->pe_name) {
1338 pe = ip_vs_pe_getbyname(u->pe_name);
1339 if (pe == NULL) {
1340 pr_info("persistence engine module ip_vs_pe_%s "
1341 "not found\n", u->pe_name);
1342 ret = -ENOENT;
1343 goto out;
1344 }
1345 old_pe = pe;
1346 }
1347
1348#ifdef CONFIG_IP_VS_IPV6
1349 if (u->af == AF_INET6) {
1350 __u32 plen = (__force __u32) u->netmask;
1351
1352 if (plen < 1 || plen > 128) {
1353 ret = -EINVAL;
1354 goto out;
1355 }
1356 }
1357#endif
1358
1359 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1360 if (sched != old_sched) {
1361 if (old_sched) {
1362 ip_vs_unbind_scheduler(svc, old_sched);
1363 RCU_INIT_POINTER(svc->scheduler, NULL);
1364
1365 synchronize_rcu();
1366 }
1367
1368 if (sched) {
1369 ret = ip_vs_bind_scheduler(svc, sched);
1370 if (ret) {
1371 ip_vs_scheduler_put(sched);
1372 goto out;
1373 }
1374 }
1375 }
1376
1377
1378
1379
1380 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1381 svc->timeout = u->timeout * HZ;
1382 svc->netmask = u->netmask;
1383
1384 old_pe = rcu_dereference_protected(svc->pe, 1);
1385 if (pe != old_pe) {
1386 rcu_assign_pointer(svc->pe, pe);
1387
1388 new_pe_conn_out = (pe && pe->conn_out) ? true : false;
1389 old_pe_conn_out = (old_pe && old_pe->conn_out) ? true : false;
1390 if (new_pe_conn_out && !old_pe_conn_out)
1391 atomic_inc(&svc->ipvs->conn_out_counter);
1392 if (old_pe_conn_out && !new_pe_conn_out)
1393 atomic_dec(&svc->ipvs->conn_out_counter);
1394 }
1395
1396out:
1397 ip_vs_scheduler_put(old_sched);
1398 ip_vs_pe_put(old_pe);
1399 return ret;
1400}
1401
1402
1403
1404
1405
1406
1407static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1408{
1409 struct ip_vs_dest *dest, *nxt;
1410 struct ip_vs_scheduler *old_sched;
1411 struct ip_vs_pe *old_pe;
1412 struct netns_ipvs *ipvs = svc->ipvs;
1413
1414
1415 if (svc->af == AF_INET)
1416 ipvs->num_services--;
1417
1418 ip_vs_stop_estimator(svc->ipvs, &svc->stats);
1419
1420
1421 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1422 ip_vs_unbind_scheduler(svc, old_sched);
1423 ip_vs_scheduler_put(old_sched);
1424
1425
1426 old_pe = rcu_dereference_protected(svc->pe, 1);
1427 if (old_pe && old_pe->conn_out)
1428 atomic_dec(&ipvs->conn_out_counter);
1429 ip_vs_pe_put(old_pe);
1430
1431
1432
1433
1434 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1435 __ip_vs_unlink_dest(svc, dest, 0);
1436 __ip_vs_del_dest(svc->ipvs, dest, cleanup);
1437 }
1438
1439
1440
1441
1442 if (svc->port == FTPPORT)
1443 atomic_dec(&ipvs->ftpsvc_counter);
1444 else if (svc->port == 0)
1445 atomic_dec(&ipvs->nullsvc_counter);
1446
1447
1448
1449
1450 __ip_vs_svc_put(svc, true);
1451
1452
1453 ip_vs_use_count_dec();
1454}
1455
1456
1457
1458
1459static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1460{
1461
1462 atomic_inc(&svc->refcnt);
1463
1464
1465
1466 ip_vs_svc_unhash(svc);
1467
1468 __ip_vs_del_service(svc, cleanup);
1469}
1470
1471
1472
1473
1474static int ip_vs_del_service(struct ip_vs_service *svc)
1475{
1476 if (svc == NULL)
1477 return -EEXIST;
1478 ip_vs_unlink_service(svc, false);
1479
1480 return 0;
1481}
1482
1483
1484
1485
1486
1487static int ip_vs_flush(struct netns_ipvs *ipvs, bool cleanup)
1488{
1489 int idx;
1490 struct ip_vs_service *svc;
1491 struct hlist_node *n;
1492
1493
1494
1495
1496 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1497 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1498 s_list) {
1499 if (svc->ipvs == ipvs)
1500 ip_vs_unlink_service(svc, cleanup);
1501 }
1502 }
1503
1504
1505
1506
1507 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1508 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1509 f_list) {
1510 if (svc->ipvs == ipvs)
1511 ip_vs_unlink_service(svc, cleanup);
1512 }
1513 }
1514
1515 return 0;
1516}
1517
1518
1519
1520
1521
1522void ip_vs_service_net_cleanup(struct netns_ipvs *ipvs)
1523{
1524 EnterFunction(2);
1525
1526 mutex_lock(&__ip_vs_mutex);
1527 ip_vs_flush(ipvs, true);
1528 mutex_unlock(&__ip_vs_mutex);
1529 LeaveFunction(2);
1530}
1531
1532
1533static inline void
1534ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1535{
1536 struct ip_vs_dest_dst *dest_dst;
1537
1538 spin_lock_bh(&dest->dst_lock);
1539 dest_dst = rcu_dereference_protected(dest->dest_dst, 1);
1540 if (dest_dst && dest_dst->dst_cache->dev == dev) {
1541 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1542 dev->name,
1543 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1544 ntohs(dest->port),
1545 refcount_read(&dest->refcnt));
1546 __ip_vs_dst_cache_reset(dest);
1547 }
1548 spin_unlock_bh(&dest->dst_lock);
1549
1550}
1551
1552
1553
1554static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1555 void *ptr)
1556{
1557 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1558 struct net *net = dev_net(dev);
1559 struct netns_ipvs *ipvs = net_ipvs(net);
1560 struct ip_vs_service *svc;
1561 struct ip_vs_dest *dest;
1562 unsigned int idx;
1563
1564 if (event != NETDEV_DOWN || !ipvs)
1565 return NOTIFY_DONE;
1566 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1567 EnterFunction(2);
1568 mutex_lock(&__ip_vs_mutex);
1569 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1570 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1571 if (svc->ipvs == ipvs) {
1572 list_for_each_entry(dest, &svc->destinations,
1573 n_list) {
1574 ip_vs_forget_dev(dest, dev);
1575 }
1576 }
1577 }
1578
1579 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1580 if (svc->ipvs == ipvs) {
1581 list_for_each_entry(dest, &svc->destinations,
1582 n_list) {
1583 ip_vs_forget_dev(dest, dev);
1584 }
1585 }
1586
1587 }
1588 }
1589
1590 spin_lock_bh(&ipvs->dest_trash_lock);
1591 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1592 ip_vs_forget_dev(dest, dev);
1593 }
1594 spin_unlock_bh(&ipvs->dest_trash_lock);
1595 mutex_unlock(&__ip_vs_mutex);
1596 LeaveFunction(2);
1597 return NOTIFY_DONE;
1598}
1599
1600
1601
1602
1603static int ip_vs_zero_service(struct ip_vs_service *svc)
1604{
1605 struct ip_vs_dest *dest;
1606
1607 list_for_each_entry(dest, &svc->destinations, n_list) {
1608 ip_vs_zero_stats(&dest->stats);
1609 }
1610 ip_vs_zero_stats(&svc->stats);
1611 return 0;
1612}
1613
1614static int ip_vs_zero_all(struct netns_ipvs *ipvs)
1615{
1616 int idx;
1617 struct ip_vs_service *svc;
1618
1619 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1620 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1621 if (svc->ipvs == ipvs)
1622 ip_vs_zero_service(svc);
1623 }
1624 }
1625
1626 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1627 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1628 if (svc->ipvs == ipvs)
1629 ip_vs_zero_service(svc);
1630 }
1631 }
1632
1633 ip_vs_zero_stats(&ipvs->tot_stats);
1634 return 0;
1635}
1636
1637#ifdef CONFIG_SYSCTL
1638
1639static int zero;
1640static int three = 3;
1641
1642static int
1643proc_do_defense_mode(struct ctl_table *table, int write,
1644 void __user *buffer, size_t *lenp, loff_t *ppos)
1645{
1646 struct netns_ipvs *ipvs = table->extra2;
1647 int *valp = table->data;
1648 int val = *valp;
1649 int rc;
1650
1651 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1652 if (write && (*valp != val)) {
1653 if ((*valp < 0) || (*valp > 3)) {
1654
1655 *valp = val;
1656 } else {
1657 update_defense_level(ipvs);
1658 }
1659 }
1660 return rc;
1661}
1662
1663static int
1664proc_do_sync_threshold(struct ctl_table *table, int write,
1665 void __user *buffer, size_t *lenp, loff_t *ppos)
1666{
1667 int *valp = table->data;
1668 int val[2];
1669 int rc;
1670
1671
1672 memcpy(val, valp, sizeof(val));
1673
1674 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1675 if (write && (valp[0] < 0 || valp[1] < 0 ||
1676 (valp[0] >= valp[1] && valp[1]))) {
1677
1678 memcpy(valp, val, sizeof(val));
1679 }
1680 return rc;
1681}
1682
1683static int
1684proc_do_sync_mode(struct ctl_table *table, int write,
1685 void __user *buffer, size_t *lenp, loff_t *ppos)
1686{
1687 int *valp = table->data;
1688 int val = *valp;
1689 int rc;
1690
1691 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1692 if (write && (*valp != val)) {
1693 if ((*valp < 0) || (*valp > 1)) {
1694
1695 *valp = val;
1696 }
1697 }
1698 return rc;
1699}
1700
1701static int
1702proc_do_sync_ports(struct ctl_table *table, int write,
1703 void __user *buffer, size_t *lenp, loff_t *ppos)
1704{
1705 int *valp = table->data;
1706 int val = *valp;
1707 int rc;
1708
1709 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1710 if (write && (*valp != val)) {
1711 if (*valp < 1 || !is_power_of_2(*valp)) {
1712
1713 *valp = val;
1714 }
1715 }
1716 return rc;
1717}
1718
1719
1720
1721
1722
1723
1724
1725static struct ctl_table vs_vars[] = {
1726 {
1727 .procname = "amemthresh",
1728 .maxlen = sizeof(int),
1729 .mode = 0644,
1730 .proc_handler = proc_dointvec,
1731 },
1732 {
1733 .procname = "am_droprate",
1734 .maxlen = sizeof(int),
1735 .mode = 0644,
1736 .proc_handler = proc_dointvec,
1737 },
1738 {
1739 .procname = "drop_entry",
1740 .maxlen = sizeof(int),
1741 .mode = 0644,
1742 .proc_handler = proc_do_defense_mode,
1743 },
1744 {
1745 .procname = "drop_packet",
1746 .maxlen = sizeof(int),
1747 .mode = 0644,
1748 .proc_handler = proc_do_defense_mode,
1749 },
1750#ifdef CONFIG_IP_VS_NFCT
1751 {
1752 .procname = "conntrack",
1753 .maxlen = sizeof(int),
1754 .mode = 0644,
1755 .proc_handler = &proc_dointvec,
1756 },
1757#endif
1758 {
1759 .procname = "secure_tcp",
1760 .maxlen = sizeof(int),
1761 .mode = 0644,
1762 .proc_handler = proc_do_defense_mode,
1763 },
1764 {
1765 .procname = "snat_reroute",
1766 .maxlen = sizeof(int),
1767 .mode = 0644,
1768 .proc_handler = &proc_dointvec,
1769 },
1770 {
1771 .procname = "sync_version",
1772 .maxlen = sizeof(int),
1773 .mode = 0644,
1774 .proc_handler = proc_do_sync_mode,
1775 },
1776 {
1777 .procname = "sync_ports",
1778 .maxlen = sizeof(int),
1779 .mode = 0644,
1780 .proc_handler = proc_do_sync_ports,
1781 },
1782 {
1783 .procname = "sync_persist_mode",
1784 .maxlen = sizeof(int),
1785 .mode = 0644,
1786 .proc_handler = proc_dointvec,
1787 },
1788 {
1789 .procname = "sync_qlen_max",
1790 .maxlen = sizeof(unsigned long),
1791 .mode = 0644,
1792 .proc_handler = proc_doulongvec_minmax,
1793 },
1794 {
1795 .procname = "sync_sock_size",
1796 .maxlen = sizeof(int),
1797 .mode = 0644,
1798 .proc_handler = proc_dointvec,
1799 },
1800 {
1801 .procname = "cache_bypass",
1802 .maxlen = sizeof(int),
1803 .mode = 0644,
1804 .proc_handler = proc_dointvec,
1805 },
1806 {
1807 .procname = "expire_nodest_conn",
1808 .maxlen = sizeof(int),
1809 .mode = 0644,
1810 .proc_handler = proc_dointvec,
1811 },
1812 {
1813 .procname = "sloppy_tcp",
1814 .maxlen = sizeof(int),
1815 .mode = 0644,
1816 .proc_handler = proc_dointvec,
1817 },
1818 {
1819 .procname = "sloppy_sctp",
1820 .maxlen = sizeof(int),
1821 .mode = 0644,
1822 .proc_handler = proc_dointvec,
1823 },
1824 {
1825 .procname = "expire_quiescent_template",
1826 .maxlen = sizeof(int),
1827 .mode = 0644,
1828 .proc_handler = proc_dointvec,
1829 },
1830 {
1831 .procname = "sync_threshold",
1832 .maxlen =
1833 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1834 .mode = 0644,
1835 .proc_handler = proc_do_sync_threshold,
1836 },
1837 {
1838 .procname = "sync_refresh_period",
1839 .maxlen = sizeof(int),
1840 .mode = 0644,
1841 .proc_handler = proc_dointvec_jiffies,
1842 },
1843 {
1844 .procname = "sync_retries",
1845 .maxlen = sizeof(int),
1846 .mode = 0644,
1847 .proc_handler = proc_dointvec_minmax,
1848 .extra1 = &zero,
1849 .extra2 = &three,
1850 },
1851 {
1852 .procname = "nat_icmp_send",
1853 .maxlen = sizeof(int),
1854 .mode = 0644,
1855 .proc_handler = proc_dointvec,
1856 },
1857 {
1858 .procname = "pmtu_disc",
1859 .maxlen = sizeof(int),
1860 .mode = 0644,
1861 .proc_handler = proc_dointvec,
1862 },
1863 {
1864 .procname = "backup_only",
1865 .maxlen = sizeof(int),
1866 .mode = 0644,
1867 .proc_handler = proc_dointvec,
1868 },
1869 {
1870 .procname = "conn_reuse_mode",
1871 .maxlen = sizeof(int),
1872 .mode = 0644,
1873 .proc_handler = proc_dointvec,
1874 },
1875 {
1876 .procname = "schedule_icmp",
1877 .maxlen = sizeof(int),
1878 .mode = 0644,
1879 .proc_handler = proc_dointvec,
1880 },
1881 {
1882 .procname = "ignore_tunneled",
1883 .maxlen = sizeof(int),
1884 .mode = 0644,
1885 .proc_handler = proc_dointvec,
1886 },
1887#ifdef CONFIG_IP_VS_DEBUG
1888 {
1889 .procname = "debug_level",
1890 .data = &sysctl_ip_vs_debug_level,
1891 .maxlen = sizeof(int),
1892 .mode = 0644,
1893 .proc_handler = proc_dointvec,
1894 },
1895#endif
1896 { }
1897};
1898
1899#endif
1900
1901#ifdef CONFIG_PROC_FS
1902
1903struct ip_vs_iter {
1904 struct seq_net_private p;
1905 struct hlist_head *table;
1906 int bucket;
1907};
1908
1909
1910
1911
1912
1913static inline const char *ip_vs_fwd_name(unsigned int flags)
1914{
1915 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1916 case IP_VS_CONN_F_LOCALNODE:
1917 return "Local";
1918 case IP_VS_CONN_F_TUNNEL:
1919 return "Tunnel";
1920 case IP_VS_CONN_F_DROUTE:
1921 return "Route";
1922 default:
1923 return "Masq";
1924 }
1925}
1926
1927
1928
1929static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1930{
1931 struct net *net = seq_file_net(seq);
1932 struct netns_ipvs *ipvs = net_ipvs(net);
1933 struct ip_vs_iter *iter = seq->private;
1934 int idx;
1935 struct ip_vs_service *svc;
1936
1937
1938 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1939 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
1940 if ((svc->ipvs == ipvs) && pos-- == 0) {
1941 iter->table = ip_vs_svc_table;
1942 iter->bucket = idx;
1943 return svc;
1944 }
1945 }
1946 }
1947
1948
1949 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1950 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
1951 f_list) {
1952 if ((svc->ipvs == ipvs) && pos-- == 0) {
1953 iter->table = ip_vs_svc_fwm_table;
1954 iter->bucket = idx;
1955 return svc;
1956 }
1957 }
1958 }
1959
1960 return NULL;
1961}
1962
1963static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1964 __acquires(RCU)
1965{
1966 rcu_read_lock();
1967 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1968}
1969
1970
1971static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1972{
1973 struct hlist_node *e;
1974 struct ip_vs_iter *iter;
1975 struct ip_vs_service *svc;
1976
1977 ++*pos;
1978 if (v == SEQ_START_TOKEN)
1979 return ip_vs_info_array(seq,0);
1980
1981 svc = v;
1982 iter = seq->private;
1983
1984 if (iter->table == ip_vs_svc_table) {
1985
1986 e = rcu_dereference(hlist_next_rcu(&svc->s_list));
1987 if (e)
1988 return hlist_entry(e, struct ip_vs_service, s_list);
1989
1990 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1991 hlist_for_each_entry_rcu(svc,
1992 &ip_vs_svc_table[iter->bucket],
1993 s_list) {
1994 return svc;
1995 }
1996 }
1997
1998 iter->table = ip_vs_svc_fwm_table;
1999 iter->bucket = -1;
2000 goto scan_fwmark;
2001 }
2002
2003
2004 e = rcu_dereference(hlist_next_rcu(&svc->f_list));
2005 if (e)
2006 return hlist_entry(e, struct ip_vs_service, f_list);
2007
2008 scan_fwmark:
2009 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2010 hlist_for_each_entry_rcu(svc,
2011 &ip_vs_svc_fwm_table[iter->bucket],
2012 f_list)
2013 return svc;
2014 }
2015
2016 return NULL;
2017}
2018
2019static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
2020 __releases(RCU)
2021{
2022 rcu_read_unlock();
2023}
2024
2025
2026static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
2027{
2028 if (v == SEQ_START_TOKEN) {
2029 seq_printf(seq,
2030 "IP Virtual Server version %d.%d.%d (size=%d)\n",
2031 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2032 seq_puts(seq,
2033 "Prot LocalAddress:Port Scheduler Flags\n");
2034 seq_puts(seq,
2035 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2036 } else {
2037 struct net *net = seq_file_net(seq);
2038 struct netns_ipvs *ipvs = net_ipvs(net);
2039 const struct ip_vs_service *svc = v;
2040 const struct ip_vs_iter *iter = seq->private;
2041 const struct ip_vs_dest *dest;
2042 struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler);
2043 char *sched_name = sched ? sched->name : "none";
2044
2045 if (svc->ipvs != ipvs)
2046 return 0;
2047 if (iter->table == ip_vs_svc_table) {
2048#ifdef CONFIG_IP_VS_IPV6
2049 if (svc->af == AF_INET6)
2050 seq_printf(seq, "%s [%pI6]:%04X %s ",
2051 ip_vs_proto_name(svc->protocol),
2052 &svc->addr.in6,
2053 ntohs(svc->port),
2054 sched_name);
2055 else
2056#endif
2057 seq_printf(seq, "%s %08X:%04X %s %s ",
2058 ip_vs_proto_name(svc->protocol),
2059 ntohl(svc->addr.ip),
2060 ntohs(svc->port),
2061 sched_name,
2062 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2063 } else {
2064 seq_printf(seq, "FWM %08X %s %s",
2065 svc->fwmark, sched_name,
2066 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2067 }
2068
2069 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2070 seq_printf(seq, "persistent %d %08X\n",
2071 svc->timeout,
2072 ntohl(svc->netmask));
2073 else
2074 seq_putc(seq, '\n');
2075
2076 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
2077#ifdef CONFIG_IP_VS_IPV6
2078 if (dest->af == AF_INET6)
2079 seq_printf(seq,
2080 " -> [%pI6]:%04X"
2081 " %-7s %-6d %-10d %-10d\n",
2082 &dest->addr.in6,
2083 ntohs(dest->port),
2084 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2085 atomic_read(&dest->weight),
2086 atomic_read(&dest->activeconns),
2087 atomic_read(&dest->inactconns));
2088 else
2089#endif
2090 seq_printf(seq,
2091 " -> %08X:%04X "
2092 "%-7s %-6d %-10d %-10d\n",
2093 ntohl(dest->addr.ip),
2094 ntohs(dest->port),
2095 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2096 atomic_read(&dest->weight),
2097 atomic_read(&dest->activeconns),
2098 atomic_read(&dest->inactconns));
2099
2100 }
2101 }
2102 return 0;
2103}
2104
2105static const struct seq_operations ip_vs_info_seq_ops = {
2106 .start = ip_vs_info_seq_start,
2107 .next = ip_vs_info_seq_next,
2108 .stop = ip_vs_info_seq_stop,
2109 .show = ip_vs_info_seq_show,
2110};
2111
2112static int ip_vs_info_open(struct inode *inode, struct file *file)
2113{
2114 return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2115 sizeof(struct ip_vs_iter));
2116}
2117
2118static const struct file_operations ip_vs_info_fops = {
2119 .open = ip_vs_info_open,
2120 .read = seq_read,
2121 .llseek = seq_lseek,
2122 .release = seq_release_net,
2123};
2124
2125static int ip_vs_stats_show(struct seq_file *seq, void *v)
2126{
2127 struct net *net = seq_file_single_net(seq);
2128 struct ip_vs_kstats show;
2129
2130
2131 seq_puts(seq,
2132 " Total Incoming Outgoing Incoming Outgoing\n");
2133 seq_puts(seq,
2134 " Conns Packets Packets Bytes Bytes\n");
2135
2136 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2137 seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n\n",
2138 (unsigned long long)show.conns,
2139 (unsigned long long)show.inpkts,
2140 (unsigned long long)show.outpkts,
2141 (unsigned long long)show.inbytes,
2142 (unsigned long long)show.outbytes);
2143
2144
2145 seq_puts(seq,
2146 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2147 seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n",
2148 (unsigned long long)show.cps,
2149 (unsigned long long)show.inpps,
2150 (unsigned long long)show.outpps,
2151 (unsigned long long)show.inbps,
2152 (unsigned long long)show.outbps);
2153
2154 return 0;
2155}
2156
2157static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2158{
2159 return single_open_net(inode, file, ip_vs_stats_show);
2160}
2161
2162static const struct file_operations ip_vs_stats_fops = {
2163 .open = ip_vs_stats_seq_open,
2164 .read = seq_read,
2165 .llseek = seq_lseek,
2166 .release = single_release_net,
2167};
2168
2169static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2170{
2171 struct net *net = seq_file_single_net(seq);
2172 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2173 struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats;
2174 struct ip_vs_kstats kstats;
2175 int i;
2176
2177
2178 seq_puts(seq,
2179 " Total Incoming Outgoing Incoming Outgoing\n");
2180 seq_puts(seq,
2181 "CPU Conns Packets Packets Bytes Bytes\n");
2182
2183 for_each_possible_cpu(i) {
2184 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2185 unsigned int start;
2186 u64 conns, inpkts, outpkts, inbytes, outbytes;
2187
2188 do {
2189 start = u64_stats_fetch_begin_irq(&u->syncp);
2190 conns = u->cnt.conns;
2191 inpkts = u->cnt.inpkts;
2192 outpkts = u->cnt.outpkts;
2193 inbytes = u->cnt.inbytes;
2194 outbytes = u->cnt.outbytes;
2195 } while (u64_stats_fetch_retry_irq(&u->syncp, start));
2196
2197 seq_printf(seq, "%3X %8LX %8LX %8LX %16LX %16LX\n",
2198 i, (u64)conns, (u64)inpkts,
2199 (u64)outpkts, (u64)inbytes,
2200 (u64)outbytes);
2201 }
2202
2203 ip_vs_copy_stats(&kstats, tot_stats);
2204
2205 seq_printf(seq, " ~ %8LX %8LX %8LX %16LX %16LX\n\n",
2206 (unsigned long long)kstats.conns,
2207 (unsigned long long)kstats.inpkts,
2208 (unsigned long long)kstats.outpkts,
2209 (unsigned long long)kstats.inbytes,
2210 (unsigned long long)kstats.outbytes);
2211
2212
2213 seq_puts(seq,
2214 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2215 seq_printf(seq, " %8LX %8LX %8LX %16LX %16LX\n",
2216 kstats.cps,
2217 kstats.inpps,
2218 kstats.outpps,
2219 kstats.inbps,
2220 kstats.outbps);
2221
2222 return 0;
2223}
2224
2225static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2226{
2227 return single_open_net(inode, file, ip_vs_stats_percpu_show);
2228}
2229
2230static const struct file_operations ip_vs_stats_percpu_fops = {
2231 .open = ip_vs_stats_percpu_seq_open,
2232 .read = seq_read,
2233 .llseek = seq_lseek,
2234 .release = single_release_net,
2235};
2236#endif
2237
2238
2239
2240
2241static int ip_vs_set_timeout(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2242{
2243#if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2244 struct ip_vs_proto_data *pd;
2245#endif
2246
2247 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2248 u->tcp_timeout,
2249 u->tcp_fin_timeout,
2250 u->udp_timeout);
2251
2252#ifdef CONFIG_IP_VS_PROTO_TCP
2253 if (u->tcp_timeout) {
2254 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2255 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2256 = u->tcp_timeout * HZ;
2257 }
2258
2259 if (u->tcp_fin_timeout) {
2260 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2261 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2262 = u->tcp_fin_timeout * HZ;
2263 }
2264#endif
2265
2266#ifdef CONFIG_IP_VS_PROTO_UDP
2267 if (u->udp_timeout) {
2268 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2269 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2270 = u->udp_timeout * HZ;
2271 }
2272#endif
2273 return 0;
2274}
2275
2276#define CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2277
2278struct ip_vs_svcdest_user {
2279 struct ip_vs_service_user s;
2280 struct ip_vs_dest_user d;
2281};
2282
2283static const unsigned char set_arglen[CMDID(IP_VS_SO_SET_MAX) + 1] = {
2284 [CMDID(IP_VS_SO_SET_ADD)] = sizeof(struct ip_vs_service_user),
2285 [CMDID(IP_VS_SO_SET_EDIT)] = sizeof(struct ip_vs_service_user),
2286 [CMDID(IP_VS_SO_SET_DEL)] = sizeof(struct ip_vs_service_user),
2287 [CMDID(IP_VS_SO_SET_ADDDEST)] = sizeof(struct ip_vs_svcdest_user),
2288 [CMDID(IP_VS_SO_SET_DELDEST)] = sizeof(struct ip_vs_svcdest_user),
2289 [CMDID(IP_VS_SO_SET_EDITDEST)] = sizeof(struct ip_vs_svcdest_user),
2290 [CMDID(IP_VS_SO_SET_TIMEOUT)] = sizeof(struct ip_vs_timeout_user),
2291 [CMDID(IP_VS_SO_SET_STARTDAEMON)] = sizeof(struct ip_vs_daemon_user),
2292 [CMDID(IP_VS_SO_SET_STOPDAEMON)] = sizeof(struct ip_vs_daemon_user),
2293 [CMDID(IP_VS_SO_SET_ZERO)] = sizeof(struct ip_vs_service_user),
2294};
2295
2296union ip_vs_set_arglen {
2297 struct ip_vs_service_user field_IP_VS_SO_SET_ADD;
2298 struct ip_vs_service_user field_IP_VS_SO_SET_EDIT;
2299 struct ip_vs_service_user field_IP_VS_SO_SET_DEL;
2300 struct ip_vs_svcdest_user field_IP_VS_SO_SET_ADDDEST;
2301 struct ip_vs_svcdest_user field_IP_VS_SO_SET_DELDEST;
2302 struct ip_vs_svcdest_user field_IP_VS_SO_SET_EDITDEST;
2303 struct ip_vs_timeout_user field_IP_VS_SO_SET_TIMEOUT;
2304 struct ip_vs_daemon_user field_IP_VS_SO_SET_STARTDAEMON;
2305 struct ip_vs_daemon_user field_IP_VS_SO_SET_STOPDAEMON;
2306 struct ip_vs_service_user field_IP_VS_SO_SET_ZERO;
2307};
2308
2309#define MAX_SET_ARGLEN sizeof(union ip_vs_set_arglen)
2310
2311static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2312 struct ip_vs_service_user *usvc_compat)
2313{
2314 memset(usvc, 0, sizeof(*usvc));
2315
2316 usvc->af = AF_INET;
2317 usvc->protocol = usvc_compat->protocol;
2318 usvc->addr.ip = usvc_compat->addr;
2319 usvc->port = usvc_compat->port;
2320 usvc->fwmark = usvc_compat->fwmark;
2321
2322
2323 usvc->sched_name = usvc_compat->sched_name;
2324
2325 usvc->flags = usvc_compat->flags;
2326 usvc->timeout = usvc_compat->timeout;
2327 usvc->netmask = usvc_compat->netmask;
2328}
2329
2330static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2331 struct ip_vs_dest_user *udest_compat)
2332{
2333 memset(udest, 0, sizeof(*udest));
2334
2335 udest->addr.ip = udest_compat->addr;
2336 udest->port = udest_compat->port;
2337 udest->conn_flags = udest_compat->conn_flags;
2338 udest->weight = udest_compat->weight;
2339 udest->u_threshold = udest_compat->u_threshold;
2340 udest->l_threshold = udest_compat->l_threshold;
2341 udest->af = AF_INET;
2342}
2343
2344static int
2345do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2346{
2347 struct net *net = sock_net(sk);
2348 int ret;
2349 unsigned char arg[MAX_SET_ARGLEN];
2350 struct ip_vs_service_user *usvc_compat;
2351 struct ip_vs_service_user_kern usvc;
2352 struct ip_vs_service *svc;
2353 struct ip_vs_dest_user *udest_compat;
2354 struct ip_vs_dest_user_kern udest;
2355 struct netns_ipvs *ipvs = net_ipvs(net);
2356
2357 BUILD_BUG_ON(sizeof(arg) > 255);
2358 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2359 return -EPERM;
2360
2361 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2362 return -EINVAL;
2363 if (len != set_arglen[CMDID(cmd)]) {
2364 IP_VS_DBG(1, "set_ctl: len %u != %u\n",
2365 len, set_arglen[CMDID(cmd)]);
2366 return -EINVAL;
2367 }
2368
2369 if (copy_from_user(arg, user, len) != 0)
2370 return -EFAULT;
2371
2372
2373 ip_vs_use_count_inc();
2374
2375
2376 if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2377 cmd == IP_VS_SO_SET_STOPDAEMON) {
2378 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2379
2380 if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2381 struct ipvs_sync_daemon_cfg cfg;
2382
2383 memset(&cfg, 0, sizeof(cfg));
2384 strlcpy(cfg.mcast_ifn, dm->mcast_ifn,
2385 sizeof(cfg.mcast_ifn));
2386 cfg.syncid = dm->syncid;
2387 rtnl_lock();
2388 mutex_lock(&ipvs->sync_mutex);
2389 ret = start_sync_thread(ipvs, &cfg, dm->state);
2390 mutex_unlock(&ipvs->sync_mutex);
2391 rtnl_unlock();
2392 } else {
2393 mutex_lock(&ipvs->sync_mutex);
2394 ret = stop_sync_thread(ipvs, dm->state);
2395 mutex_unlock(&ipvs->sync_mutex);
2396 }
2397 goto out_dec;
2398 }
2399
2400 mutex_lock(&__ip_vs_mutex);
2401 if (cmd == IP_VS_SO_SET_FLUSH) {
2402
2403 ret = ip_vs_flush(ipvs, false);
2404 goto out_unlock;
2405 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2406
2407 ret = ip_vs_set_timeout(ipvs, (struct ip_vs_timeout_user *)arg);
2408 goto out_unlock;
2409 }
2410
2411 usvc_compat = (struct ip_vs_service_user *)arg;
2412 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2413
2414
2415
2416 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2417 ip_vs_copy_udest_compat(&udest, udest_compat);
2418
2419 if (cmd == IP_VS_SO_SET_ZERO) {
2420
2421 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2422 ret = ip_vs_zero_all(ipvs);
2423 goto out_unlock;
2424 }
2425 }
2426
2427
2428 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2429 usvc.protocol != IPPROTO_SCTP) {
2430 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2431 usvc.protocol, &usvc.addr.ip,
2432 ntohs(usvc.port), usvc.sched_name);
2433 ret = -EFAULT;
2434 goto out_unlock;
2435 }
2436
2437
2438 rcu_read_lock();
2439 if (usvc.fwmark == 0)
2440 svc = __ip_vs_service_find(ipvs, usvc.af, usvc.protocol,
2441 &usvc.addr, usvc.port);
2442 else
2443 svc = __ip_vs_svc_fwm_find(ipvs, usvc.af, usvc.fwmark);
2444 rcu_read_unlock();
2445
2446 if (cmd != IP_VS_SO_SET_ADD
2447 && (svc == NULL || svc->protocol != usvc.protocol)) {
2448 ret = -ESRCH;
2449 goto out_unlock;
2450 }
2451
2452 switch (cmd) {
2453 case IP_VS_SO_SET_ADD:
2454 if (svc != NULL)
2455 ret = -EEXIST;
2456 else
2457 ret = ip_vs_add_service(ipvs, &usvc, &svc);
2458 break;
2459 case IP_VS_SO_SET_EDIT:
2460 ret = ip_vs_edit_service(svc, &usvc);
2461 break;
2462 case IP_VS_SO_SET_DEL:
2463 ret = ip_vs_del_service(svc);
2464 if (!ret)
2465 goto out_unlock;
2466 break;
2467 case IP_VS_SO_SET_ZERO:
2468 ret = ip_vs_zero_service(svc);
2469 break;
2470 case IP_VS_SO_SET_ADDDEST:
2471 ret = ip_vs_add_dest(svc, &udest);
2472 break;
2473 case IP_VS_SO_SET_EDITDEST:
2474 ret = ip_vs_edit_dest(svc, &udest);
2475 break;
2476 case IP_VS_SO_SET_DELDEST:
2477 ret = ip_vs_del_dest(svc, &udest);
2478 break;
2479 default:
2480 ret = -EINVAL;
2481 }
2482
2483 out_unlock:
2484 mutex_unlock(&__ip_vs_mutex);
2485 out_dec:
2486
2487 ip_vs_use_count_dec();
2488
2489 return ret;
2490}
2491
2492
2493static void
2494ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2495{
2496 struct ip_vs_scheduler *sched;
2497 struct ip_vs_kstats kstats;
2498 char *sched_name;
2499
2500 sched = rcu_dereference_protected(src->scheduler, 1);
2501 sched_name = sched ? sched->name : "none";
2502 dst->protocol = src->protocol;
2503 dst->addr = src->addr.ip;
2504 dst->port = src->port;
2505 dst->fwmark = src->fwmark;
2506 strlcpy(dst->sched_name, sched_name, sizeof(dst->sched_name));
2507 dst->flags = src->flags;
2508 dst->timeout = src->timeout / HZ;
2509 dst->netmask = src->netmask;
2510 dst->num_dests = src->num_dests;
2511 ip_vs_copy_stats(&kstats, &src->stats);
2512 ip_vs_export_stats_user(&dst->stats, &kstats);
2513}
2514
2515static inline int
2516__ip_vs_get_service_entries(struct netns_ipvs *ipvs,
2517 const struct ip_vs_get_services *get,
2518 struct ip_vs_get_services __user *uptr)
2519{
2520 int idx, count=0;
2521 struct ip_vs_service *svc;
2522 struct ip_vs_service_entry entry;
2523 int ret = 0;
2524
2525 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2526 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2527
2528 if (svc->af != AF_INET || (svc->ipvs != ipvs))
2529 continue;
2530
2531 if (count >= get->num_services)
2532 goto out;
2533 memset(&entry, 0, sizeof(entry));
2534 ip_vs_copy_service(&entry, svc);
2535 if (copy_to_user(&uptr->entrytable[count],
2536 &entry, sizeof(entry))) {
2537 ret = -EFAULT;
2538 goto out;
2539 }
2540 count++;
2541 }
2542 }
2543
2544 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2545 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2546
2547 if (svc->af != AF_INET || (svc->ipvs != ipvs))
2548 continue;
2549
2550 if (count >= get->num_services)
2551 goto out;
2552 memset(&entry, 0, sizeof(entry));
2553 ip_vs_copy_service(&entry, svc);
2554 if (copy_to_user(&uptr->entrytable[count],
2555 &entry, sizeof(entry))) {
2556 ret = -EFAULT;
2557 goto out;
2558 }
2559 count++;
2560 }
2561 }
2562out:
2563 return ret;
2564}
2565
2566static inline int
2567__ip_vs_get_dest_entries(struct netns_ipvs *ipvs, const struct ip_vs_get_dests *get,
2568 struct ip_vs_get_dests __user *uptr)
2569{
2570 struct ip_vs_service *svc;
2571 union nf_inet_addr addr = { .ip = get->addr };
2572 int ret = 0;
2573
2574 rcu_read_lock();
2575 if (get->fwmark)
2576 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, get->fwmark);
2577 else
2578 svc = __ip_vs_service_find(ipvs, AF_INET, get->protocol, &addr,
2579 get->port);
2580 rcu_read_unlock();
2581
2582 if (svc) {
2583 int count = 0;
2584 struct ip_vs_dest *dest;
2585 struct ip_vs_dest_entry entry;
2586 struct ip_vs_kstats kstats;
2587
2588 memset(&entry, 0, sizeof(entry));
2589 list_for_each_entry(dest, &svc->destinations, n_list) {
2590 if (count >= get->num_dests)
2591 break;
2592
2593
2594
2595
2596 if (dest->af != svc->af)
2597 continue;
2598
2599 entry.addr = dest->addr.ip;
2600 entry.port = dest->port;
2601 entry.conn_flags = atomic_read(&dest->conn_flags);
2602 entry.weight = atomic_read(&dest->weight);
2603 entry.u_threshold = dest->u_threshold;
2604 entry.l_threshold = dest->l_threshold;
2605 entry.activeconns = atomic_read(&dest->activeconns);
2606 entry.inactconns = atomic_read(&dest->inactconns);
2607 entry.persistconns = atomic_read(&dest->persistconns);
2608 ip_vs_copy_stats(&kstats, &dest->stats);
2609 ip_vs_export_stats_user(&entry.stats, &kstats);
2610 if (copy_to_user(&uptr->entrytable[count],
2611 &entry, sizeof(entry))) {
2612 ret = -EFAULT;
2613 break;
2614 }
2615 count++;
2616 }
2617 } else
2618 ret = -ESRCH;
2619 return ret;
2620}
2621
2622static inline void
2623__ip_vs_get_timeouts(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2624{
2625#if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2626 struct ip_vs_proto_data *pd;
2627#endif
2628
2629 memset(u, 0, sizeof (*u));
2630
2631#ifdef CONFIG_IP_VS_PROTO_TCP
2632 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2633 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2634 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2635#endif
2636#ifdef CONFIG_IP_VS_PROTO_UDP
2637 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2638 u->udp_timeout =
2639 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2640#endif
2641}
2642
2643static const unsigned char get_arglen[CMDID(IP_VS_SO_GET_MAX) + 1] = {
2644 [CMDID(IP_VS_SO_GET_VERSION)] = 64,
2645 [CMDID(IP_VS_SO_GET_INFO)] = sizeof(struct ip_vs_getinfo),
2646 [CMDID(IP_VS_SO_GET_SERVICES)] = sizeof(struct ip_vs_get_services),
2647 [CMDID(IP_VS_SO_GET_SERVICE)] = sizeof(struct ip_vs_service_entry),
2648 [CMDID(IP_VS_SO_GET_DESTS)] = sizeof(struct ip_vs_get_dests),
2649 [CMDID(IP_VS_SO_GET_TIMEOUT)] = sizeof(struct ip_vs_timeout_user),
2650 [CMDID(IP_VS_SO_GET_DAEMON)] = 2 * sizeof(struct ip_vs_daemon_user),
2651};
2652
2653union ip_vs_get_arglen {
2654 char field_IP_VS_SO_GET_VERSION[64];
2655 struct ip_vs_getinfo field_IP_VS_SO_GET_INFO;
2656 struct ip_vs_get_services field_IP_VS_SO_GET_SERVICES;
2657 struct ip_vs_service_entry field_IP_VS_SO_GET_SERVICE;
2658 struct ip_vs_get_dests field_IP_VS_SO_GET_DESTS;
2659 struct ip_vs_timeout_user field_IP_VS_SO_GET_TIMEOUT;
2660 struct ip_vs_daemon_user field_IP_VS_SO_GET_DAEMON[2];
2661};
2662
2663#define MAX_GET_ARGLEN sizeof(union ip_vs_get_arglen)
2664
2665static int
2666do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2667{
2668 unsigned char arg[MAX_GET_ARGLEN];
2669 int ret = 0;
2670 unsigned int copylen;
2671 struct net *net = sock_net(sk);
2672 struct netns_ipvs *ipvs = net_ipvs(net);
2673
2674 BUG_ON(!net);
2675 BUILD_BUG_ON(sizeof(arg) > 255);
2676 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2677 return -EPERM;
2678
2679 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2680 return -EINVAL;
2681
2682 copylen = get_arglen[CMDID(cmd)];
2683 if (*len < (int) copylen) {
2684 IP_VS_DBG(1, "get_ctl: len %d < %u\n", *len, copylen);
2685 return -EINVAL;
2686 }
2687
2688 if (copy_from_user(arg, user, copylen) != 0)
2689 return -EFAULT;
2690
2691
2692
2693 if (cmd == IP_VS_SO_GET_DAEMON) {
2694 struct ip_vs_daemon_user d[2];
2695
2696 memset(&d, 0, sizeof(d));
2697 mutex_lock(&ipvs->sync_mutex);
2698 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2699 d[0].state = IP_VS_STATE_MASTER;
2700 strlcpy(d[0].mcast_ifn, ipvs->mcfg.mcast_ifn,
2701 sizeof(d[0].mcast_ifn));
2702 d[0].syncid = ipvs->mcfg.syncid;
2703 }
2704 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2705 d[1].state = IP_VS_STATE_BACKUP;
2706 strlcpy(d[1].mcast_ifn, ipvs->bcfg.mcast_ifn,
2707 sizeof(d[1].mcast_ifn));
2708 d[1].syncid = ipvs->bcfg.syncid;
2709 }
2710 if (copy_to_user(user, &d, sizeof(d)) != 0)
2711 ret = -EFAULT;
2712 mutex_unlock(&ipvs->sync_mutex);
2713 return ret;
2714 }
2715
2716 mutex_lock(&__ip_vs_mutex);
2717 switch (cmd) {
2718 case IP_VS_SO_GET_VERSION:
2719 {
2720 char buf[64];
2721
2722 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2723 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2724 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2725 ret = -EFAULT;
2726 goto out;
2727 }
2728 *len = strlen(buf)+1;
2729 }
2730 break;
2731
2732 case IP_VS_SO_GET_INFO:
2733 {
2734 struct ip_vs_getinfo info;
2735 info.version = IP_VS_VERSION_CODE;
2736 info.size = ip_vs_conn_tab_size;
2737 info.num_services = ipvs->num_services;
2738 if (copy_to_user(user, &info, sizeof(info)) != 0)
2739 ret = -EFAULT;
2740 }
2741 break;
2742
2743 case IP_VS_SO_GET_SERVICES:
2744 {
2745 struct ip_vs_get_services *get;
2746 int size;
2747
2748 get = (struct ip_vs_get_services *)arg;
2749 size = sizeof(*get) +
2750 sizeof(struct ip_vs_service_entry) * get->num_services;
2751 if (*len != size) {
2752 pr_err("length: %u != %u\n", *len, size);
2753 ret = -EINVAL;
2754 goto out;
2755 }
2756 ret = __ip_vs_get_service_entries(ipvs, get, user);
2757 }
2758 break;
2759
2760 case IP_VS_SO_GET_SERVICE:
2761 {
2762 struct ip_vs_service_entry *entry;
2763 struct ip_vs_service *svc;
2764 union nf_inet_addr addr;
2765
2766 entry = (struct ip_vs_service_entry *)arg;
2767 addr.ip = entry->addr;
2768 rcu_read_lock();
2769 if (entry->fwmark)
2770 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, entry->fwmark);
2771 else
2772 svc = __ip_vs_service_find(ipvs, AF_INET,
2773 entry->protocol, &addr,
2774 entry->port);
2775 rcu_read_unlock();
2776 if (svc) {
2777 ip_vs_copy_service(entry, svc);
2778 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2779 ret = -EFAULT;
2780 } else
2781 ret = -ESRCH;
2782 }
2783 break;
2784
2785 case IP_VS_SO_GET_DESTS:
2786 {
2787 struct ip_vs_get_dests *get;
2788 int size;
2789
2790 get = (struct ip_vs_get_dests *)arg;
2791 size = sizeof(*get) +
2792 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2793 if (*len != size) {
2794 pr_err("length: %u != %u\n", *len, size);
2795 ret = -EINVAL;
2796 goto out;
2797 }
2798 ret = __ip_vs_get_dest_entries(ipvs, get, user);
2799 }
2800 break;
2801
2802 case IP_VS_SO_GET_TIMEOUT:
2803 {
2804 struct ip_vs_timeout_user t;
2805
2806 __ip_vs_get_timeouts(ipvs, &t);
2807 if (copy_to_user(user, &t, sizeof(t)) != 0)
2808 ret = -EFAULT;
2809 }
2810 break;
2811
2812 default:
2813 ret = -EINVAL;
2814 }
2815
2816out:
2817 mutex_unlock(&__ip_vs_mutex);
2818 return ret;
2819}
2820
2821
2822static struct nf_sockopt_ops ip_vs_sockopts = {
2823 .pf = PF_INET,
2824 .set_optmin = IP_VS_BASE_CTL,
2825 .set_optmax = IP_VS_SO_SET_MAX+1,
2826 .set = do_ip_vs_set_ctl,
2827 .get_optmin = IP_VS_BASE_CTL,
2828 .get_optmax = IP_VS_SO_GET_MAX+1,
2829 .get = do_ip_vs_get_ctl,
2830 .owner = THIS_MODULE,
2831};
2832
2833
2834
2835
2836
2837
2838static struct genl_family ip_vs_genl_family;
2839
2840
2841static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2842 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2843 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2844 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2845 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2846 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2847 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2848};
2849
2850
2851static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2852 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2853 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2854 .len = IP_VS_IFNAME_MAXLEN },
2855 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2856 [IPVS_DAEMON_ATTR_SYNC_MAXLEN] = { .type = NLA_U16 },
2857 [IPVS_DAEMON_ATTR_MCAST_GROUP] = { .type = NLA_U32 },
2858 [IPVS_DAEMON_ATTR_MCAST_GROUP6] = { .len = sizeof(struct in6_addr) },
2859 [IPVS_DAEMON_ATTR_MCAST_PORT] = { .type = NLA_U16 },
2860 [IPVS_DAEMON_ATTR_MCAST_TTL] = { .type = NLA_U8 },
2861};
2862
2863
2864static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2865 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2866 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2867 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2868 .len = sizeof(union nf_inet_addr) },
2869 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2870 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2871 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2872 .len = IP_VS_SCHEDNAME_MAXLEN },
2873 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2874 .len = IP_VS_PENAME_MAXLEN },
2875 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2876 .len = sizeof(struct ip_vs_flags) },
2877 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2878 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2879 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2880};
2881
2882
2883static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2884 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2885 .len = sizeof(union nf_inet_addr) },
2886 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2887 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2888 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2889 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2890 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2891 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2892 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2893 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2894 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2895 [IPVS_DEST_ATTR_ADDR_FAMILY] = { .type = NLA_U16 },
2896};
2897
2898static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2899 struct ip_vs_kstats *kstats)
2900{
2901 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2902
2903 if (!nl_stats)
2904 return -EMSGSIZE;
2905
2906 if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, (u32)kstats->conns) ||
2907 nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, (u32)kstats->inpkts) ||
2908 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, (u32)kstats->outpkts) ||
2909 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes,
2910 IPVS_STATS_ATTR_PAD) ||
2911 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes,
2912 IPVS_STATS_ATTR_PAD) ||
2913 nla_put_u32(skb, IPVS_STATS_ATTR_CPS, (u32)kstats->cps) ||
2914 nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, (u32)kstats->inpps) ||
2915 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, (u32)kstats->outpps) ||
2916 nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, (u32)kstats->inbps) ||
2917 nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, (u32)kstats->outbps))
2918 goto nla_put_failure;
2919 nla_nest_end(skb, nl_stats);
2920
2921 return 0;
2922
2923nla_put_failure:
2924 nla_nest_cancel(skb, nl_stats);
2925 return -EMSGSIZE;
2926}
2927
2928static int ip_vs_genl_fill_stats64(struct sk_buff *skb, int container_type,
2929 struct ip_vs_kstats *kstats)
2930{
2931 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2932
2933 if (!nl_stats)
2934 return -EMSGSIZE;
2935
2936 if (nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CONNS, kstats->conns,
2937 IPVS_STATS_ATTR_PAD) ||
2938 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPKTS, kstats->inpkts,
2939 IPVS_STATS_ATTR_PAD) ||
2940 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPKTS, kstats->outpkts,
2941 IPVS_STATS_ATTR_PAD) ||
2942 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes,
2943 IPVS_STATS_ATTR_PAD) ||
2944 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes,
2945 IPVS_STATS_ATTR_PAD) ||
2946 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CPS, kstats->cps,
2947 IPVS_STATS_ATTR_PAD) ||
2948 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPPS, kstats->inpps,
2949 IPVS_STATS_ATTR_PAD) ||
2950 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPPS, kstats->outpps,
2951 IPVS_STATS_ATTR_PAD) ||
2952 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBPS, kstats->inbps,
2953 IPVS_STATS_ATTR_PAD) ||
2954 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBPS, kstats->outbps,
2955 IPVS_STATS_ATTR_PAD))
2956 goto nla_put_failure;
2957 nla_nest_end(skb, nl_stats);
2958
2959 return 0;
2960
2961nla_put_failure:
2962 nla_nest_cancel(skb, nl_stats);
2963 return -EMSGSIZE;
2964}
2965
2966static int ip_vs_genl_fill_service(struct sk_buff *skb,
2967 struct ip_vs_service *svc)
2968{
2969 struct ip_vs_scheduler *sched;
2970 struct ip_vs_pe *pe;
2971 struct nlattr *nl_service;
2972 struct ip_vs_flags flags = { .flags = svc->flags,
2973 .mask = ~0 };
2974 struct ip_vs_kstats kstats;
2975 char *sched_name;
2976
2977 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2978 if (!nl_service)
2979 return -EMSGSIZE;
2980
2981 if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2982 goto nla_put_failure;
2983 if (svc->fwmark) {
2984 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2985 goto nla_put_failure;
2986 } else {
2987 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2988 nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2989 nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2990 goto nla_put_failure;
2991 }
2992
2993 sched = rcu_dereference_protected(svc->scheduler, 1);
2994 sched_name = sched ? sched->name : "none";
2995 pe = rcu_dereference_protected(svc->pe, 1);
2996 if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched_name) ||
2997 (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
2998 nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2999 nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
3000 nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
3001 goto nla_put_failure;
3002 ip_vs_copy_stats(&kstats, &svc->stats);
3003 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &kstats))
3004 goto nla_put_failure;
3005 if (ip_vs_genl_fill_stats64(skb, IPVS_SVC_ATTR_STATS64, &kstats))
3006 goto nla_put_failure;
3007
3008 nla_nest_end(skb, nl_service);
3009
3010 return 0;
3011
3012nla_put_failure:
3013 nla_nest_cancel(skb, nl_service);
3014 return -EMSGSIZE;
3015}
3016
3017static int ip_vs_genl_dump_service(struct sk_buff *skb,
3018 struct ip_vs_service *svc,
3019 struct netlink_callback *cb)
3020{
3021 void *hdr;
3022
3023 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3024 &ip_vs_genl_family, NLM_F_MULTI,
3025 IPVS_CMD_NEW_SERVICE);
3026 if (!hdr)
3027 return -EMSGSIZE;
3028
3029 if (ip_vs_genl_fill_service(skb, svc) < 0)
3030 goto nla_put_failure;
3031
3032 genlmsg_end(skb, hdr);
3033 return 0;
3034
3035nla_put_failure:
3036 genlmsg_cancel(skb, hdr);
3037 return -EMSGSIZE;
3038}
3039
3040static int ip_vs_genl_dump_services(struct sk_buff *skb,
3041 struct netlink_callback *cb)
3042{
3043 int idx = 0, i;
3044 int start = cb->args[0];
3045 struct ip_vs_service *svc;
3046 struct net *net = sock_net(skb->sk);
3047 struct netns_ipvs *ipvs = net_ipvs(net);
3048
3049 mutex_lock(&__ip_vs_mutex);
3050 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3051 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
3052 if (++idx <= start || (svc->ipvs != ipvs))
3053 continue;
3054 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3055 idx--;
3056 goto nla_put_failure;
3057 }
3058 }
3059 }
3060
3061 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3062 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
3063 if (++idx <= start || (svc->ipvs != ipvs))
3064 continue;
3065 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3066 idx--;
3067 goto nla_put_failure;
3068 }
3069 }
3070 }
3071
3072nla_put_failure:
3073 mutex_unlock(&__ip_vs_mutex);
3074 cb->args[0] = idx;
3075
3076 return skb->len;
3077}
3078
3079static bool ip_vs_is_af_valid(int af)
3080{
3081 if (af == AF_INET)
3082 return true;
3083#ifdef CONFIG_IP_VS_IPV6
3084 if (af == AF_INET6 && ipv6_mod_enabled())
3085 return true;
3086#endif
3087 return false;
3088}
3089
3090static int ip_vs_genl_parse_service(struct netns_ipvs *ipvs,
3091 struct ip_vs_service_user_kern *usvc,
3092 struct nlattr *nla, int full_entry,
3093 struct ip_vs_service **ret_svc)
3094{
3095 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
3096 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
3097 struct ip_vs_service *svc;
3098
3099
3100 if (nla == NULL ||
3101 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla,
3102 ip_vs_svc_policy, NULL))
3103 return -EINVAL;
3104
3105 nla_af = attrs[IPVS_SVC_ATTR_AF];
3106 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
3107 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
3108 nla_port = attrs[IPVS_SVC_ATTR_PORT];
3109 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
3110
3111 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3112 return -EINVAL;
3113
3114 memset(usvc, 0, sizeof(*usvc));
3115
3116 usvc->af = nla_get_u16(nla_af);
3117 if (!ip_vs_is_af_valid(usvc->af))
3118 return -EAFNOSUPPORT;
3119
3120 if (nla_fwmark) {
3121 usvc->protocol = IPPROTO_TCP;
3122 usvc->fwmark = nla_get_u32(nla_fwmark);
3123 } else {
3124 usvc->protocol = nla_get_u16(nla_protocol);
3125 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3126 usvc->port = nla_get_be16(nla_port);
3127 usvc->fwmark = 0;
3128 }
3129
3130 rcu_read_lock();
3131 if (usvc->fwmark)
3132 svc = __ip_vs_svc_fwm_find(ipvs, usvc->af, usvc->fwmark);
3133 else
3134 svc = __ip_vs_service_find(ipvs, usvc->af, usvc->protocol,
3135 &usvc->addr, usvc->port);
3136 rcu_read_unlock();
3137 *ret_svc = svc;
3138
3139
3140 if (full_entry) {
3141 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3142 *nla_netmask;
3143 struct ip_vs_flags flags;
3144
3145 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3146 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3147 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3148 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3149 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3150
3151 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3152 return -EINVAL;
3153
3154 nla_memcpy(&flags, nla_flags, sizeof(flags));
3155
3156
3157 if (svc)
3158 usvc->flags = svc->flags;
3159
3160
3161 usvc->flags = (usvc->flags & ~flags.mask) |
3162 (flags.flags & flags.mask);
3163 usvc->sched_name = nla_data(nla_sched);
3164 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3165 usvc->timeout = nla_get_u32(nla_timeout);
3166 usvc->netmask = nla_get_be32(nla_netmask);
3167 }
3168
3169 return 0;
3170}
3171
3172static struct ip_vs_service *ip_vs_genl_find_service(struct netns_ipvs *ipvs,
3173 struct nlattr *nla)
3174{
3175 struct ip_vs_service_user_kern usvc;
3176 struct ip_vs_service *svc;
3177 int ret;
3178
3179 ret = ip_vs_genl_parse_service(ipvs, &usvc, nla, 0, &svc);
3180 return ret ? ERR_PTR(ret) : svc;
3181}
3182
3183static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3184{
3185 struct nlattr *nl_dest;
3186 struct ip_vs_kstats kstats;
3187
3188 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3189 if (!nl_dest)
3190 return -EMSGSIZE;
3191
3192 if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3193 nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3194 nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3195 (atomic_read(&dest->conn_flags) &
3196 IP_VS_CONN_F_FWD_MASK)) ||
3197 nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3198 atomic_read(&dest->weight)) ||
3199 nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3200 nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3201 nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3202 atomic_read(&dest->activeconns)) ||
3203 nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3204 atomic_read(&dest->inactconns)) ||
3205 nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3206 atomic_read(&dest->persistconns)) ||
3207 nla_put_u16(skb, IPVS_DEST_ATTR_ADDR_FAMILY, dest->af))
3208 goto nla_put_failure;
3209 ip_vs_copy_stats(&kstats, &dest->stats);
3210 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &kstats))
3211 goto nla_put_failure;
3212 if (ip_vs_genl_fill_stats64(skb, IPVS_DEST_ATTR_STATS64, &kstats))
3213 goto nla_put_failure;
3214
3215 nla_nest_end(skb, nl_dest);
3216
3217 return 0;
3218
3219nla_put_failure:
3220 nla_nest_cancel(skb, nl_dest);
3221 return -EMSGSIZE;
3222}
3223
3224static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3225 struct netlink_callback *cb)
3226{
3227 void *hdr;
3228
3229 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3230 &ip_vs_genl_family, NLM_F_MULTI,
3231 IPVS_CMD_NEW_DEST);
3232 if (!hdr)
3233 return -EMSGSIZE;
3234
3235 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3236 goto nla_put_failure;
3237
3238 genlmsg_end(skb, hdr);
3239 return 0;
3240
3241nla_put_failure:
3242 genlmsg_cancel(skb, hdr);
3243 return -EMSGSIZE;
3244}
3245
3246static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3247 struct netlink_callback *cb)
3248{
3249 int idx = 0;
3250 int start = cb->args[0];
3251 struct ip_vs_service *svc;
3252 struct ip_vs_dest *dest;
3253 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3254 struct net *net = sock_net(skb->sk);
3255 struct netns_ipvs *ipvs = net_ipvs(net);
3256
3257 mutex_lock(&__ip_vs_mutex);
3258
3259
3260 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs, IPVS_CMD_ATTR_MAX,
3261 ip_vs_cmd_policy, NULL))
3262 goto out_err;
3263
3264
3265 svc = ip_vs_genl_find_service(ipvs, attrs[IPVS_CMD_ATTR_SERVICE]);
3266 if (IS_ERR_OR_NULL(svc))
3267 goto out_err;
3268
3269
3270 list_for_each_entry(dest, &svc->destinations, n_list) {
3271 if (++idx <= start)
3272 continue;
3273 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3274 idx--;
3275 goto nla_put_failure;
3276 }
3277 }
3278
3279nla_put_failure:
3280 cb->args[0] = idx;
3281
3282out_err:
3283 mutex_unlock(&__ip_vs_mutex);
3284
3285 return skb->len;
3286}
3287
3288static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3289 struct nlattr *nla, int full_entry)
3290{
3291 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3292 struct nlattr *nla_addr, *nla_port;
3293 struct nlattr *nla_addr_family;
3294
3295
3296 if (nla == NULL ||
3297 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla,
3298 ip_vs_dest_policy, NULL))
3299 return -EINVAL;
3300
3301 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3302 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3303 nla_addr_family = attrs[IPVS_DEST_ATTR_ADDR_FAMILY];
3304
3305 if (!(nla_addr && nla_port))
3306 return -EINVAL;
3307
3308 memset(udest, 0, sizeof(*udest));
3309
3310 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3311 udest->port = nla_get_be16(nla_port);
3312
3313 if (nla_addr_family)
3314 udest->af = nla_get_u16(nla_addr_family);
3315 else
3316 udest->af = 0;
3317
3318
3319 if (full_entry) {
3320 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3321 *nla_l_thresh;
3322
3323 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3324 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3325 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3326 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3327
3328 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3329 return -EINVAL;
3330
3331 udest->conn_flags = nla_get_u32(nla_fwd)
3332 & IP_VS_CONN_F_FWD_MASK;
3333 udest->weight = nla_get_u32(nla_weight);
3334 udest->u_threshold = nla_get_u32(nla_u_thresh);
3335 udest->l_threshold = nla_get_u32(nla_l_thresh);
3336 }
3337
3338 return 0;
3339}
3340
3341static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3342 struct ipvs_sync_daemon_cfg *c)
3343{
3344 struct nlattr *nl_daemon;
3345
3346 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3347 if (!nl_daemon)
3348 return -EMSGSIZE;
3349
3350 if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3351 nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, c->mcast_ifn) ||
3352 nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, c->syncid) ||
3353 nla_put_u16(skb, IPVS_DAEMON_ATTR_SYNC_MAXLEN, c->sync_maxlen) ||
3354 nla_put_u16(skb, IPVS_DAEMON_ATTR_MCAST_PORT, c->mcast_port) ||
3355 nla_put_u8(skb, IPVS_DAEMON_ATTR_MCAST_TTL, c->mcast_ttl))
3356 goto nla_put_failure;
3357#ifdef CONFIG_IP_VS_IPV6
3358 if (c->mcast_af == AF_INET6) {
3359 if (nla_put_in6_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP6,
3360 &c->mcast_group.in6))
3361 goto nla_put_failure;
3362 } else
3363#endif
3364 if (c->mcast_af == AF_INET &&
3365 nla_put_in_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP,
3366 c->mcast_group.ip))
3367 goto nla_put_failure;
3368 nla_nest_end(skb, nl_daemon);
3369
3370 return 0;
3371
3372nla_put_failure:
3373 nla_nest_cancel(skb, nl_daemon);
3374 return -EMSGSIZE;
3375}
3376
3377static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3378 struct ipvs_sync_daemon_cfg *c,
3379 struct netlink_callback *cb)
3380{
3381 void *hdr;
3382 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3383 &ip_vs_genl_family, NLM_F_MULTI,
3384 IPVS_CMD_NEW_DAEMON);
3385 if (!hdr)
3386 return -EMSGSIZE;
3387
3388 if (ip_vs_genl_fill_daemon(skb, state, c))
3389 goto nla_put_failure;
3390
3391 genlmsg_end(skb, hdr);
3392 return 0;
3393
3394nla_put_failure:
3395 genlmsg_cancel(skb, hdr);
3396 return -EMSGSIZE;
3397}
3398
3399static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3400 struct netlink_callback *cb)
3401{
3402 struct net *net = sock_net(skb->sk);
3403 struct netns_ipvs *ipvs = net_ipvs(net);
3404
3405 mutex_lock(&ipvs->sync_mutex);
3406 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3407 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3408 &ipvs->mcfg, cb) < 0)
3409 goto nla_put_failure;
3410
3411 cb->args[0] = 1;
3412 }
3413
3414 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3415 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3416 &ipvs->bcfg, cb) < 0)
3417 goto nla_put_failure;
3418
3419 cb->args[1] = 1;
3420 }
3421
3422nla_put_failure:
3423 mutex_unlock(&ipvs->sync_mutex);
3424
3425 return skb->len;
3426}
3427
3428static int ip_vs_genl_new_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3429{
3430 struct ipvs_sync_daemon_cfg c;
3431 struct nlattr *a;
3432 int ret;
3433
3434 memset(&c, 0, sizeof(c));
3435 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3436 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3437 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3438 return -EINVAL;
3439 strlcpy(c.mcast_ifn, nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3440 sizeof(c.mcast_ifn));
3441 c.syncid = nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]);
3442
3443 a = attrs[IPVS_DAEMON_ATTR_SYNC_MAXLEN];
3444 if (a)
3445 c.sync_maxlen = nla_get_u16(a);
3446
3447 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP];
3448 if (a) {
3449 c.mcast_af = AF_INET;
3450 c.mcast_group.ip = nla_get_in_addr(a);
3451 if (!ipv4_is_multicast(c.mcast_group.ip))
3452 return -EINVAL;
3453 } else {
3454 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP6];
3455 if (a) {
3456#ifdef CONFIG_IP_VS_IPV6
3457 int addr_type;
3458
3459 c.mcast_af = AF_INET6;
3460 c.mcast_group.in6 = nla_get_in6_addr(a);
3461 addr_type = ipv6_addr_type(&c.mcast_group.in6);
3462 if (!(addr_type & IPV6_ADDR_MULTICAST))
3463 return -EINVAL;
3464#else
3465 return -EAFNOSUPPORT;
3466#endif
3467 }
3468 }
3469
3470 a = attrs[IPVS_DAEMON_ATTR_MCAST_PORT];
3471 if (a)
3472 c.mcast_port = nla_get_u16(a);
3473
3474 a = attrs[IPVS_DAEMON_ATTR_MCAST_TTL];
3475 if (a)
3476 c.mcast_ttl = nla_get_u8(a);
3477
3478
3479
3480
3481 if (ipvs->mixed_address_family_dests > 0)
3482 return -EINVAL;
3483
3484 rtnl_lock();
3485 mutex_lock(&ipvs->sync_mutex);
3486 ret = start_sync_thread(ipvs, &c,
3487 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3488 mutex_unlock(&ipvs->sync_mutex);
3489 rtnl_unlock();
3490 return ret;
3491}
3492
3493static int ip_vs_genl_del_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3494{
3495 int ret;
3496
3497 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3498 return -EINVAL;
3499
3500 mutex_lock(&ipvs->sync_mutex);
3501 ret = stop_sync_thread(ipvs,
3502 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3503 mutex_unlock(&ipvs->sync_mutex);
3504 return ret;
3505}
3506
3507static int ip_vs_genl_set_config(struct netns_ipvs *ipvs, struct nlattr **attrs)
3508{
3509 struct ip_vs_timeout_user t;
3510
3511 __ip_vs_get_timeouts(ipvs, &t);
3512
3513 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3514 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3515
3516 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3517 t.tcp_fin_timeout =
3518 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3519
3520 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3521 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3522
3523 return ip_vs_set_timeout(ipvs, &t);
3524}
3525
3526static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3527{
3528 int ret = -EINVAL, cmd;
3529 struct net *net = sock_net(skb->sk);
3530 struct netns_ipvs *ipvs = net_ipvs(net);
3531
3532 cmd = info->genlhdr->cmd;
3533
3534 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3535 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3536
3537 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3538 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3539 info->attrs[IPVS_CMD_ATTR_DAEMON],
3540 ip_vs_daemon_policy, info->extack))
3541 goto out;
3542
3543 if (cmd == IPVS_CMD_NEW_DAEMON)
3544 ret = ip_vs_genl_new_daemon(ipvs, daemon_attrs);
3545 else
3546 ret = ip_vs_genl_del_daemon(ipvs, daemon_attrs);
3547 }
3548
3549out:
3550 return ret;
3551}
3552
3553static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3554{
3555 struct ip_vs_service *svc = NULL;
3556 struct ip_vs_service_user_kern usvc;
3557 struct ip_vs_dest_user_kern udest;
3558 int ret = 0, cmd;
3559 int need_full_svc = 0, need_full_dest = 0;
3560 struct net *net = sock_net(skb->sk);
3561 struct netns_ipvs *ipvs = net_ipvs(net);
3562
3563 cmd = info->genlhdr->cmd;
3564
3565 mutex_lock(&__ip_vs_mutex);
3566
3567 if (cmd == IPVS_CMD_FLUSH) {
3568 ret = ip_vs_flush(ipvs, false);
3569 goto out;
3570 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3571 ret = ip_vs_genl_set_config(ipvs, info->attrs);
3572 goto out;
3573 } else if (cmd == IPVS_CMD_ZERO &&
3574 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3575 ret = ip_vs_zero_all(ipvs);
3576 goto out;
3577 }
3578
3579
3580
3581
3582 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3583 need_full_svc = 1;
3584
3585 ret = ip_vs_genl_parse_service(ipvs, &usvc,
3586 info->attrs[IPVS_CMD_ATTR_SERVICE],
3587 need_full_svc, &svc);
3588 if (ret)
3589 goto out;
3590
3591
3592 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3593 ret = -ESRCH;
3594 goto out;
3595 }
3596
3597
3598
3599
3600 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3601 cmd == IPVS_CMD_DEL_DEST) {
3602 if (cmd != IPVS_CMD_DEL_DEST)
3603 need_full_dest = 1;
3604
3605 ret = ip_vs_genl_parse_dest(&udest,
3606 info->attrs[IPVS_CMD_ATTR_DEST],
3607 need_full_dest);
3608 if (ret)
3609 goto out;
3610
3611
3612
3613
3614
3615
3616
3617 if (udest.af == 0)
3618 udest.af = svc->af;
3619
3620 if (!ip_vs_is_af_valid(udest.af)) {
3621 ret = -EAFNOSUPPORT;
3622 goto out;
3623 }
3624
3625 if (udest.af != svc->af && cmd != IPVS_CMD_DEL_DEST) {
3626
3627
3628
3629 if (ipvs->sync_state) {
3630 ret = -EINVAL;
3631 goto out;
3632 }
3633
3634
3635 switch (udest.conn_flags) {
3636 case IP_VS_CONN_F_TUNNEL:
3637
3638 break;
3639 default:
3640 ret = -EINVAL;
3641 goto out;
3642 }
3643 }
3644 }
3645
3646 switch (cmd) {
3647 case IPVS_CMD_NEW_SERVICE:
3648 if (svc == NULL)
3649 ret = ip_vs_add_service(ipvs, &usvc, &svc);
3650 else
3651 ret = -EEXIST;
3652 break;
3653 case IPVS_CMD_SET_SERVICE:
3654 ret = ip_vs_edit_service(svc, &usvc);
3655 break;
3656 case IPVS_CMD_DEL_SERVICE:
3657 ret = ip_vs_del_service(svc);
3658
3659 break;
3660 case IPVS_CMD_NEW_DEST:
3661 ret = ip_vs_add_dest(svc, &udest);
3662 break;
3663 case IPVS_CMD_SET_DEST:
3664 ret = ip_vs_edit_dest(svc, &udest);
3665 break;
3666 case IPVS_CMD_DEL_DEST:
3667 ret = ip_vs_del_dest(svc, &udest);
3668 break;
3669 case IPVS_CMD_ZERO:
3670 ret = ip_vs_zero_service(svc);
3671 break;
3672 default:
3673 ret = -EINVAL;
3674 }
3675
3676out:
3677 mutex_unlock(&__ip_vs_mutex);
3678
3679 return ret;
3680}
3681
3682static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3683{
3684 struct sk_buff *msg;
3685 void *reply;
3686 int ret, cmd, reply_cmd;
3687 struct net *net = sock_net(skb->sk);
3688 struct netns_ipvs *ipvs = net_ipvs(net);
3689
3690 cmd = info->genlhdr->cmd;
3691
3692 if (cmd == IPVS_CMD_GET_SERVICE)
3693 reply_cmd = IPVS_CMD_NEW_SERVICE;
3694 else if (cmd == IPVS_CMD_GET_INFO)
3695 reply_cmd = IPVS_CMD_SET_INFO;
3696 else if (cmd == IPVS_CMD_GET_CONFIG)
3697 reply_cmd = IPVS_CMD_SET_CONFIG;
3698 else {
3699 pr_err("unknown Generic Netlink command\n");
3700 return -EINVAL;
3701 }
3702
3703 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3704 if (!msg)
3705 return -ENOMEM;
3706
3707 mutex_lock(&__ip_vs_mutex);
3708
3709 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3710 if (reply == NULL)
3711 goto nla_put_failure;
3712
3713 switch (cmd) {
3714 case IPVS_CMD_GET_SERVICE:
3715 {
3716 struct ip_vs_service *svc;
3717
3718 svc = ip_vs_genl_find_service(ipvs,
3719 info->attrs[IPVS_CMD_ATTR_SERVICE]);
3720 if (IS_ERR(svc)) {
3721 ret = PTR_ERR(svc);
3722 goto out_err;
3723 } else if (svc) {
3724 ret = ip_vs_genl_fill_service(msg, svc);
3725 if (ret)
3726 goto nla_put_failure;
3727 } else {
3728 ret = -ESRCH;
3729 goto out_err;
3730 }
3731
3732 break;
3733 }
3734
3735 case IPVS_CMD_GET_CONFIG:
3736 {
3737 struct ip_vs_timeout_user t;
3738
3739 __ip_vs_get_timeouts(ipvs, &t);
3740#ifdef CONFIG_IP_VS_PROTO_TCP
3741 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3742 t.tcp_timeout) ||
3743 nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3744 t.tcp_fin_timeout))
3745 goto nla_put_failure;
3746#endif
3747#ifdef CONFIG_IP_VS_PROTO_UDP
3748 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3749 goto nla_put_failure;
3750#endif
3751
3752 break;
3753 }
3754
3755 case IPVS_CMD_GET_INFO:
3756 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3757 IP_VS_VERSION_CODE) ||
3758 nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3759 ip_vs_conn_tab_size))
3760 goto nla_put_failure;
3761 break;
3762 }
3763
3764 genlmsg_end(msg, reply);
3765 ret = genlmsg_reply(msg, info);
3766 goto out;
3767
3768nla_put_failure:
3769 pr_err("not enough space in Netlink message\n");
3770 ret = -EMSGSIZE;
3771
3772out_err:
3773 nlmsg_free(msg);
3774out:
3775 mutex_unlock(&__ip_vs_mutex);
3776
3777 return ret;
3778}
3779
3780
3781static const struct genl_ops ip_vs_genl_ops[] = {
3782 {
3783 .cmd = IPVS_CMD_NEW_SERVICE,
3784 .flags = GENL_ADMIN_PERM,
3785 .policy = ip_vs_cmd_policy,
3786 .doit = ip_vs_genl_set_cmd,
3787 },
3788 {
3789 .cmd = IPVS_CMD_SET_SERVICE,
3790 .flags = GENL_ADMIN_PERM,
3791 .policy = ip_vs_cmd_policy,
3792 .doit = ip_vs_genl_set_cmd,
3793 },
3794 {
3795 .cmd = IPVS_CMD_DEL_SERVICE,
3796 .flags = GENL_ADMIN_PERM,
3797 .policy = ip_vs_cmd_policy,
3798 .doit = ip_vs_genl_set_cmd,
3799 },
3800 {
3801 .cmd = IPVS_CMD_GET_SERVICE,
3802 .flags = GENL_ADMIN_PERM,
3803 .doit = ip_vs_genl_get_cmd,
3804 .dumpit = ip_vs_genl_dump_services,
3805 .policy = ip_vs_cmd_policy,
3806 },
3807 {
3808 .cmd = IPVS_CMD_NEW_DEST,
3809 .flags = GENL_ADMIN_PERM,
3810 .policy = ip_vs_cmd_policy,
3811 .doit = ip_vs_genl_set_cmd,
3812 },
3813 {
3814 .cmd = IPVS_CMD_SET_DEST,
3815 .flags = GENL_ADMIN_PERM,
3816 .policy = ip_vs_cmd_policy,
3817 .doit = ip_vs_genl_set_cmd,
3818 },
3819 {
3820 .cmd = IPVS_CMD_DEL_DEST,
3821 .flags = GENL_ADMIN_PERM,
3822 .policy = ip_vs_cmd_policy,
3823 .doit = ip_vs_genl_set_cmd,
3824 },
3825 {
3826 .cmd = IPVS_CMD_GET_DEST,
3827 .flags = GENL_ADMIN_PERM,
3828 .policy = ip_vs_cmd_policy,
3829 .dumpit = ip_vs_genl_dump_dests,
3830 },
3831 {
3832 .cmd = IPVS_CMD_NEW_DAEMON,
3833 .flags = GENL_ADMIN_PERM,
3834 .policy = ip_vs_cmd_policy,
3835 .doit = ip_vs_genl_set_daemon,
3836 },
3837 {
3838 .cmd = IPVS_CMD_DEL_DAEMON,
3839 .flags = GENL_ADMIN_PERM,
3840 .policy = ip_vs_cmd_policy,
3841 .doit = ip_vs_genl_set_daemon,
3842 },
3843 {
3844 .cmd = IPVS_CMD_GET_DAEMON,
3845 .flags = GENL_ADMIN_PERM,
3846 .dumpit = ip_vs_genl_dump_daemons,
3847 },
3848 {
3849 .cmd = IPVS_CMD_SET_CONFIG,
3850 .flags = GENL_ADMIN_PERM,
3851 .policy = ip_vs_cmd_policy,
3852 .doit = ip_vs_genl_set_cmd,
3853 },
3854 {
3855 .cmd = IPVS_CMD_GET_CONFIG,
3856 .flags = GENL_ADMIN_PERM,
3857 .doit = ip_vs_genl_get_cmd,
3858 },
3859 {
3860 .cmd = IPVS_CMD_GET_INFO,
3861 .flags = GENL_ADMIN_PERM,
3862 .doit = ip_vs_genl_get_cmd,
3863 },
3864 {
3865 .cmd = IPVS_CMD_ZERO,
3866 .flags = GENL_ADMIN_PERM,
3867 .policy = ip_vs_cmd_policy,
3868 .doit = ip_vs_genl_set_cmd,
3869 },
3870 {
3871 .cmd = IPVS_CMD_FLUSH,
3872 .flags = GENL_ADMIN_PERM,
3873 .doit = ip_vs_genl_set_cmd,
3874 },
3875};
3876
3877static struct genl_family ip_vs_genl_family __ro_after_init = {
3878 .hdrsize = 0,
3879 .name = IPVS_GENL_NAME,
3880 .version = IPVS_GENL_VERSION,
3881 .maxattr = IPVS_CMD_ATTR_MAX,
3882 .netnsok = true,
3883 .module = THIS_MODULE,
3884 .ops = ip_vs_genl_ops,
3885 .n_ops = ARRAY_SIZE(ip_vs_genl_ops),
3886};
3887
3888static int __init ip_vs_genl_register(void)
3889{
3890 return genl_register_family(&ip_vs_genl_family);
3891}
3892
3893static void ip_vs_genl_unregister(void)
3894{
3895 genl_unregister_family(&ip_vs_genl_family);
3896}
3897
3898
3899
3900
3901
3902
3903#ifdef CONFIG_SYSCTL
3904static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs)
3905{
3906 struct net *net = ipvs->net;
3907 int idx;
3908 struct ctl_table *tbl;
3909
3910 atomic_set(&ipvs->dropentry, 0);
3911 spin_lock_init(&ipvs->dropentry_lock);
3912 spin_lock_init(&ipvs->droppacket_lock);
3913 spin_lock_init(&ipvs->securetcp_lock);
3914
3915 if (!net_eq(net, &init_net)) {
3916 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3917 if (tbl == NULL)
3918 return -ENOMEM;
3919
3920
3921 if (net->user_ns != &init_user_ns)
3922 tbl[0].procname = NULL;
3923 } else
3924 tbl = vs_vars;
3925
3926 for (idx = 0; idx < ARRAY_SIZE(vs_vars); idx++) {
3927 if (tbl[idx].proc_handler == proc_do_defense_mode)
3928 tbl[idx].extra2 = ipvs;
3929 }
3930 idx = 0;
3931 ipvs->sysctl_amemthresh = 1024;
3932 tbl[idx++].data = &ipvs->sysctl_amemthresh;
3933 ipvs->sysctl_am_droprate = 10;
3934 tbl[idx++].data = &ipvs->sysctl_am_droprate;
3935 tbl[idx++].data = &ipvs->sysctl_drop_entry;
3936 tbl[idx++].data = &ipvs->sysctl_drop_packet;
3937#ifdef CONFIG_IP_VS_NFCT
3938 tbl[idx++].data = &ipvs->sysctl_conntrack;
3939#endif
3940 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3941 ipvs->sysctl_snat_reroute = 1;
3942 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3943 ipvs->sysctl_sync_ver = 1;
3944 tbl[idx++].data = &ipvs->sysctl_sync_ver;
3945 ipvs->sysctl_sync_ports = 1;
3946 tbl[idx++].data = &ipvs->sysctl_sync_ports;
3947 tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
3948 ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3949 tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3950 ipvs->sysctl_sync_sock_size = 0;
3951 tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3952 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3953 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3954 tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
3955 tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
3956 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3957 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3958 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3959 tbl[idx].data = &ipvs->sysctl_sync_threshold;
3960 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3961 ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3962 tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3963 ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3964 tbl[idx++].data = &ipvs->sysctl_sync_retries;
3965 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3966 ipvs->sysctl_pmtu_disc = 1;
3967 tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
3968 tbl[idx++].data = &ipvs->sysctl_backup_only;
3969 ipvs->sysctl_conn_reuse_mode = 1;
3970 tbl[idx++].data = &ipvs->sysctl_conn_reuse_mode;
3971 tbl[idx++].data = &ipvs->sysctl_schedule_icmp;
3972 tbl[idx++].data = &ipvs->sysctl_ignore_tunneled;
3973
3974 ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3975 if (ipvs->sysctl_hdr == NULL) {
3976 if (!net_eq(net, &init_net))
3977 kfree(tbl);
3978 return -ENOMEM;
3979 }
3980 ip_vs_start_estimator(ipvs, &ipvs->tot_stats);
3981 ipvs->sysctl_tbl = tbl;
3982
3983 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3984 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3985
3986 return 0;
3987}
3988
3989static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs)
3990{
3991 struct net *net = ipvs->net;
3992
3993 cancel_delayed_work_sync(&ipvs->defense_work);
3994 cancel_work_sync(&ipvs->defense_work.work);
3995 unregister_net_sysctl_table(ipvs->sysctl_hdr);
3996 ip_vs_stop_estimator(ipvs, &ipvs->tot_stats);
3997
3998 if (!net_eq(net, &init_net))
3999 kfree(ipvs->sysctl_tbl);
4000}
4001
4002#else
4003
4004static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs) { return 0; }
4005static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs) { }
4006
4007#endif
4008
4009static struct notifier_block ip_vs_dst_notifier = {
4010 .notifier_call = ip_vs_dst_event,
4011};
4012
4013int __net_init ip_vs_control_net_init(struct netns_ipvs *ipvs)
4014{
4015 int i, idx;
4016
4017
4018 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
4019 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
4020
4021 INIT_LIST_HEAD(&ipvs->dest_trash);
4022 spin_lock_init(&ipvs->dest_trash_lock);
4023 timer_setup(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire, 0);
4024 atomic_set(&ipvs->ftpsvc_counter, 0);
4025 atomic_set(&ipvs->nullsvc_counter, 0);
4026 atomic_set(&ipvs->conn_out_counter, 0);
4027
4028
4029 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
4030 if (!ipvs->tot_stats.cpustats)
4031 return -ENOMEM;
4032
4033 for_each_possible_cpu(i) {
4034 struct ip_vs_cpu_stats *ipvs_tot_stats;
4035 ipvs_tot_stats = per_cpu_ptr(ipvs->tot_stats.cpustats, i);
4036 u64_stats_init(&ipvs_tot_stats->syncp);
4037 }
4038
4039 spin_lock_init(&ipvs->tot_stats.lock);
4040
4041 proc_create("ip_vs", 0, ipvs->net->proc_net, &ip_vs_info_fops);
4042 proc_create("ip_vs_stats", 0, ipvs->net->proc_net, &ip_vs_stats_fops);
4043 proc_create("ip_vs_stats_percpu", 0, ipvs->net->proc_net,
4044 &ip_vs_stats_percpu_fops);
4045
4046 if (ip_vs_control_net_init_sysctl(ipvs))
4047 goto err;
4048
4049 return 0;
4050
4051err:
4052 free_percpu(ipvs->tot_stats.cpustats);
4053 return -ENOMEM;
4054}
4055
4056void __net_exit ip_vs_control_net_cleanup(struct netns_ipvs *ipvs)
4057{
4058 ip_vs_trash_cleanup(ipvs);
4059 ip_vs_control_net_cleanup_sysctl(ipvs);
4060 remove_proc_entry("ip_vs_stats_percpu", ipvs->net->proc_net);
4061 remove_proc_entry("ip_vs_stats", ipvs->net->proc_net);
4062 remove_proc_entry("ip_vs", ipvs->net->proc_net);
4063 free_percpu(ipvs->tot_stats.cpustats);
4064}
4065
4066int __init ip_vs_register_nl_ioctl(void)
4067{
4068 int ret;
4069
4070 ret = nf_register_sockopt(&ip_vs_sockopts);
4071 if (ret) {
4072 pr_err("cannot register sockopt.\n");
4073 goto err_sock;
4074 }
4075
4076 ret = ip_vs_genl_register();
4077 if (ret) {
4078 pr_err("cannot register Generic Netlink interface.\n");
4079 goto err_genl;
4080 }
4081 return 0;
4082
4083err_genl:
4084 nf_unregister_sockopt(&ip_vs_sockopts);
4085err_sock:
4086 return ret;
4087}
4088
4089void ip_vs_unregister_nl_ioctl(void)
4090{
4091 ip_vs_genl_unregister();
4092 nf_unregister_sockopt(&ip_vs_sockopts);
4093}
4094
4095int __init ip_vs_control_init(void)
4096{
4097 int idx;
4098 int ret;
4099
4100 EnterFunction(2);
4101
4102
4103 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
4104 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
4105 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
4106 }
4107
4108 smp_wmb();
4109
4110 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
4111 if (ret < 0)
4112 return ret;
4113
4114 LeaveFunction(2);
4115 return 0;
4116}
4117
4118
4119void ip_vs_control_cleanup(void)
4120{
4121 EnterFunction(2);
4122 unregister_netdevice_notifier(&ip_vs_dst_notifier);
4123 LeaveFunction(2);
4124}
4125