1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21#define KMSG_COMPONENT "IPVS"
22#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
23
24#include <linux/module.h>
25#include <linux/init.h>
26#include <linux/types.h>
27#include <linux/capability.h>
28#include <linux/fs.h>
29#include <linux/sysctl.h>
30#include <linux/proc_fs.h>
31#include <linux/workqueue.h>
32#include <linux/swap.h>
33#include <linux/seq_file.h>
34#include <linux/slab.h>
35
36#include <linux/netfilter.h>
37#include <linux/netfilter_ipv4.h>
38#include <linux/mutex.h>
39
40#include <net/net_namespace.h>
41#include <linux/nsproxy.h>
42#include <net/ip.h>
43#ifdef CONFIG_IP_VS_IPV6
44#include <net/ipv6.h>
45#include <net/ip6_route.h>
46#endif
47#include <net/route.h>
48#include <net/sock.h>
49#include <net/genetlink.h>
50
51#include <asm/uaccess.h>
52
53#include <net/ip_vs.h>
54
55
56static DEFINE_MUTEX(__ip_vs_mutex);
57
58
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->net);
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 net *net, 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) net >> 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 net *net, __u32 fwmark)
289{
290 return (((size_t)net>>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 %pF\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->net, 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->net, 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 %pF\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 net *net, 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(net, 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 && net_eq(svc->net, net)) {
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 net *net, int af, __u32 fwmark)
389{
390 unsigned int hash;
391 struct ip_vs_service *svc;
392
393
394 hash = ip_vs_svc_fwm_hashkey(net, 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 && net_eq(svc->net, net)) {
399
400 return svc;
401 }
402 }
403
404 return NULL;
405}
406
407
408struct ip_vs_service *
409ip_vs_service_find(struct net *net, int af, __u32 fwmark, __u16 protocol,
410 const union nf_inet_addr *vaddr, __be16 vport)
411{
412 struct ip_vs_service *svc;
413 struct netns_ipvs *ipvs = net_ipvs(net);
414
415
416
417
418 if (fwmark) {
419 svc = __ip_vs_svc_fwm_find(net, af, fwmark);
420 if (svc)
421 goto out;
422 }
423
424
425
426
427
428 svc = __ip_vs_service_find(net, af, protocol, vaddr, vport);
429
430 if (svc == NULL
431 && protocol == IPPROTO_TCP
432 && atomic_read(&ipvs->ftpsvc_counter)
433 && (vport == FTPDATA || ntohs(vport) >= PROT_SOCK)) {
434
435
436
437
438 svc = __ip_vs_service_find(net, af, protocol, vaddr, FTPPORT);
439 }
440
441 if (svc == NULL
442 && atomic_read(&ipvs->nullsvc_counter)) {
443
444
445
446 svc = __ip_vs_service_find(net, af, protocol, vaddr, 0);
447 }
448
449 out:
450 IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
451 fwmark, ip_vs_proto_name(protocol),
452 IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
453 svc ? "hit" : "not hit");
454
455 return svc;
456}
457
458
459static inline void
460__ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
461{
462 atomic_inc(&svc->refcnt);
463 dest->svc = svc;
464}
465
466static void ip_vs_service_free(struct ip_vs_service *svc)
467{
468 if (svc->stats.cpustats)
469 free_percpu(svc->stats.cpustats);
470 kfree(svc);
471}
472
473static void
474__ip_vs_unbind_svc(struct ip_vs_dest *dest)
475{
476 struct ip_vs_service *svc = dest->svc;
477
478 dest->svc = NULL;
479 if (atomic_dec_and_test(&svc->refcnt)) {
480 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n",
481 svc->fwmark,
482 IP_VS_DBG_ADDR(svc->af, &svc->addr),
483 ntohs(svc->port));
484 ip_vs_service_free(svc);
485 }
486}
487
488
489
490
491
492static inline unsigned int ip_vs_rs_hashkey(int af,
493 const union nf_inet_addr *addr,
494 __be16 port)
495{
496 register unsigned int porth = ntohs(port);
497 __be32 addr_fold = addr->ip;
498
499#ifdef CONFIG_IP_VS_IPV6
500 if (af == AF_INET6)
501 addr_fold = addr->ip6[0]^addr->ip6[1]^
502 addr->ip6[2]^addr->ip6[3];
503#endif
504
505 return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
506 & IP_VS_RTAB_MASK;
507}
508
509
510static void ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
511{
512 unsigned int hash;
513
514 if (dest->in_rs_table)
515 return;
516
517
518
519
520
521 hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
522
523 hlist_add_head_rcu(&dest->d_list, &ipvs->rs_table[hash]);
524 dest->in_rs_table = 1;
525}
526
527
528static void ip_vs_rs_unhash(struct ip_vs_dest *dest)
529{
530
531
532
533 if (dest->in_rs_table) {
534 hlist_del_rcu(&dest->d_list);
535 dest->in_rs_table = 0;
536 }
537}
538
539
540bool ip_vs_has_real_service(struct net *net, int af, __u16 protocol,
541 const union nf_inet_addr *daddr, __be16 dport)
542{
543 struct netns_ipvs *ipvs = net_ipvs(net);
544 unsigned int hash;
545 struct ip_vs_dest *dest;
546
547
548 hash = ip_vs_rs_hashkey(af, daddr, dport);
549
550 rcu_read_lock();
551 hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
552 if (dest->port == dport &&
553 dest->af == af &&
554 ip_vs_addr_equal(af, &dest->addr, daddr) &&
555 (dest->protocol == protocol || dest->vfwmark)) {
556
557 rcu_read_unlock();
558 return true;
559 }
560 }
561 rcu_read_unlock();
562
563 return false;
564}
565
566
567
568
569static struct ip_vs_dest *
570ip_vs_lookup_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
571 __be16 dport)
572{
573 struct ip_vs_dest *dest;
574
575
576
577
578 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
579 if ((dest->af == svc->af)
580 && ip_vs_addr_equal(svc->af, &dest->addr, daddr)
581 && (dest->port == dport)) {
582
583 return dest;
584 }
585 }
586
587 return NULL;
588}
589
590
591
592
593
594
595
596
597
598struct ip_vs_dest *ip_vs_find_dest(struct net *net, int af,
599 const union nf_inet_addr *daddr,
600 __be16 dport,
601 const union nf_inet_addr *vaddr,
602 __be16 vport, __u16 protocol, __u32 fwmark,
603 __u32 flags)
604{
605 struct ip_vs_dest *dest;
606 struct ip_vs_service *svc;
607 __be16 port = dport;
608
609 svc = ip_vs_service_find(net, af, fwmark, protocol, vaddr, vport);
610 if (!svc)
611 return NULL;
612 if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
613 port = 0;
614 dest = ip_vs_lookup_dest(svc, daddr, port);
615 if (!dest)
616 dest = ip_vs_lookup_dest(svc, daddr, port ^ dport);
617 return dest;
618}
619
620void ip_vs_dest_dst_rcu_free(struct rcu_head *head)
621{
622 struct ip_vs_dest_dst *dest_dst = container_of(head,
623 struct ip_vs_dest_dst,
624 rcu_head);
625
626 dst_release(dest_dst->dst_cache);
627 kfree(dest_dst);
628}
629
630
631static void __ip_vs_dst_cache_reset(struct ip_vs_dest *dest)
632{
633 struct ip_vs_dest_dst *old;
634
635 old = rcu_dereference_protected(dest->dest_dst, 1);
636 if (old) {
637 RCU_INIT_POINTER(dest->dest_dst, NULL);
638 call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
639 }
640}
641
642
643
644
645
646
647
648
649
650
651
652static struct ip_vs_dest *
653ip_vs_trash_get_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
654 __be16 dport)
655{
656 struct ip_vs_dest *dest;
657 struct netns_ipvs *ipvs = net_ipvs(svc->net);
658
659
660
661
662 spin_lock_bh(&ipvs->dest_trash_lock);
663 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
664 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
665 "dest->refcnt=%d\n",
666 dest->vfwmark,
667 IP_VS_DBG_ADDR(svc->af, &dest->addr),
668 ntohs(dest->port),
669 atomic_read(&dest->refcnt));
670
671
672
673 if (test_bit(IP_VS_DEST_STATE_REMOVING, &dest->state))
674 continue;
675 if (dest->af == svc->af &&
676 ip_vs_addr_equal(svc->af, &dest->addr, daddr) &&
677 dest->port == dport &&
678 dest->vfwmark == svc->fwmark &&
679 dest->protocol == svc->protocol &&
680 (svc->fwmark ||
681 (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
682 dest->vport == svc->port))) {
683
684 list_del(&dest->t_list);
685 ip_vs_dest_hold(dest);
686 goto out;
687 }
688 }
689
690 dest = NULL;
691
692out:
693 spin_unlock_bh(&ipvs->dest_trash_lock);
694
695 return dest;
696}
697
698static void ip_vs_dest_free(struct ip_vs_dest *dest)
699{
700 __ip_vs_dst_cache_reset(dest);
701 __ip_vs_unbind_svc(dest);
702 free_percpu(dest->stats.cpustats);
703 kfree(dest);
704}
705
706
707
708
709
710
711
712
713
714
715static void ip_vs_trash_cleanup(struct net *net)
716{
717 struct ip_vs_dest *dest, *nxt;
718 struct netns_ipvs *ipvs = net_ipvs(net);
719
720 del_timer_sync(&ipvs->dest_trash_timer);
721
722 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, t_list) {
723 list_del(&dest->t_list);
724 ip_vs_dest_free(dest);
725 }
726}
727
728static void
729ip_vs_copy_stats(struct ip_vs_stats_user *dst, struct ip_vs_stats *src)
730{
731#define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->ustats.c - src->ustats0.c
732
733 spin_lock_bh(&src->lock);
734
735 IP_VS_SHOW_STATS_COUNTER(conns);
736 IP_VS_SHOW_STATS_COUNTER(inpkts);
737 IP_VS_SHOW_STATS_COUNTER(outpkts);
738 IP_VS_SHOW_STATS_COUNTER(inbytes);
739 IP_VS_SHOW_STATS_COUNTER(outbytes);
740
741 ip_vs_read_estimator(dst, src);
742
743 spin_unlock_bh(&src->lock);
744}
745
746static void
747ip_vs_zero_stats(struct ip_vs_stats *stats)
748{
749 spin_lock_bh(&stats->lock);
750
751
752
753#define IP_VS_ZERO_STATS_COUNTER(c) stats->ustats0.c = stats->ustats.c
754
755 IP_VS_ZERO_STATS_COUNTER(conns);
756 IP_VS_ZERO_STATS_COUNTER(inpkts);
757 IP_VS_ZERO_STATS_COUNTER(outpkts);
758 IP_VS_ZERO_STATS_COUNTER(inbytes);
759 IP_VS_ZERO_STATS_COUNTER(outbytes);
760
761 ip_vs_zero_estimator(stats);
762
763 spin_unlock_bh(&stats->lock);
764}
765
766
767
768
769static void
770__ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
771 struct ip_vs_dest_user_kern *udest, int add)
772{
773 struct netns_ipvs *ipvs = net_ipvs(svc->net);
774 struct ip_vs_scheduler *sched;
775 int conn_flags;
776
777
778 atomic_set(&dest->weight, udest->weight);
779 conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
780 conn_flags |= IP_VS_CONN_F_INACTIVE;
781
782
783 if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
784 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
785 } else {
786
787
788
789
790 ip_vs_rs_hash(ipvs, dest);
791 }
792 atomic_set(&dest->conn_flags, conn_flags);
793
794
795 if (!dest->svc) {
796 __ip_vs_bind_svc(dest, svc);
797 } else {
798 if (dest->svc != svc) {
799 __ip_vs_unbind_svc(dest);
800 ip_vs_zero_stats(&dest->stats);
801 __ip_vs_bind_svc(dest, svc);
802 }
803 }
804
805
806 dest->flags |= IP_VS_DEST_F_AVAILABLE;
807
808 if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
809 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
810 dest->u_threshold = udest->u_threshold;
811 dest->l_threshold = udest->l_threshold;
812
813 spin_lock_bh(&dest->dst_lock);
814 __ip_vs_dst_cache_reset(dest);
815 spin_unlock_bh(&dest->dst_lock);
816
817 sched = rcu_dereference_protected(svc->scheduler, 1);
818 if (add) {
819 ip_vs_start_estimator(svc->net, &dest->stats);
820 list_add_rcu(&dest->n_list, &svc->destinations);
821 svc->num_dests++;
822 if (sched->add_dest)
823 sched->add_dest(svc, dest);
824 } else {
825 if (sched->upd_dest)
826 sched->upd_dest(svc, dest);
827 }
828}
829
830
831
832
833
834static int
835ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
836 struct ip_vs_dest **dest_p)
837{
838 struct ip_vs_dest *dest;
839 unsigned int atype;
840
841 EnterFunction(2);
842
843#ifdef CONFIG_IP_VS_IPV6
844 if (svc->af == AF_INET6) {
845 atype = ipv6_addr_type(&udest->addr.in6);
846 if ((!(atype & IPV6_ADDR_UNICAST) ||
847 atype & IPV6_ADDR_LINKLOCAL) &&
848 !__ip_vs_addr_is_local_v6(svc->net, &udest->addr.in6))
849 return -EINVAL;
850 } else
851#endif
852 {
853 atype = inet_addr_type(svc->net, udest->addr.ip);
854 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
855 return -EINVAL;
856 }
857
858 dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
859 if (dest == NULL)
860 return -ENOMEM;
861
862 dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
863 if (!dest->stats.cpustats)
864 goto err_alloc;
865
866 dest->af = svc->af;
867 dest->protocol = svc->protocol;
868 dest->vaddr = svc->addr;
869 dest->vport = svc->port;
870 dest->vfwmark = svc->fwmark;
871 ip_vs_addr_copy(svc->af, &dest->addr, &udest->addr);
872 dest->port = udest->port;
873
874 atomic_set(&dest->activeconns, 0);
875 atomic_set(&dest->inactconns, 0);
876 atomic_set(&dest->persistconns, 0);
877 atomic_set(&dest->refcnt, 1);
878
879 INIT_HLIST_NODE(&dest->d_list);
880 spin_lock_init(&dest->dst_lock);
881 spin_lock_init(&dest->stats.lock);
882 __ip_vs_update_dest(svc, dest, udest, 1);
883
884 *dest_p = dest;
885
886 LeaveFunction(2);
887 return 0;
888
889err_alloc:
890 kfree(dest);
891 return -ENOMEM;
892}
893
894
895
896
897
898static int
899ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
900{
901 struct ip_vs_dest *dest;
902 union nf_inet_addr daddr;
903 __be16 dport = udest->port;
904 int ret;
905
906 EnterFunction(2);
907
908 if (udest->weight < 0) {
909 pr_err("%s(): server weight less than zero\n", __func__);
910 return -ERANGE;
911 }
912
913 if (udest->l_threshold > udest->u_threshold) {
914 pr_err("%s(): lower threshold is higher than upper threshold\n",
915 __func__);
916 return -ERANGE;
917 }
918
919 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
920
921
922 rcu_read_lock();
923 dest = ip_vs_lookup_dest(svc, &daddr, dport);
924 rcu_read_unlock();
925
926 if (dest != NULL) {
927 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
928 return -EEXIST;
929 }
930
931
932
933
934
935 dest = ip_vs_trash_get_dest(svc, &daddr, dport);
936
937 if (dest != NULL) {
938 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
939 "dest->refcnt=%d, service %u/%s:%u\n",
940 IP_VS_DBG_ADDR(svc->af, &daddr), ntohs(dport),
941 atomic_read(&dest->refcnt),
942 dest->vfwmark,
943 IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
944 ntohs(dest->vport));
945
946 __ip_vs_update_dest(svc, dest, udest, 1);
947 ret = 0;
948 } else {
949
950
951
952 ret = ip_vs_new_dest(svc, udest, &dest);
953 }
954 LeaveFunction(2);
955
956 return ret;
957}
958
959
960
961
962
963static int
964ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
965{
966 struct ip_vs_dest *dest;
967 union nf_inet_addr daddr;
968 __be16 dport = udest->port;
969
970 EnterFunction(2);
971
972 if (udest->weight < 0) {
973 pr_err("%s(): server weight less than zero\n", __func__);
974 return -ERANGE;
975 }
976
977 if (udest->l_threshold > udest->u_threshold) {
978 pr_err("%s(): lower threshold is higher than upper threshold\n",
979 __func__);
980 return -ERANGE;
981 }
982
983 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
984
985
986 rcu_read_lock();
987 dest = ip_vs_lookup_dest(svc, &daddr, dport);
988 rcu_read_unlock();
989
990 if (dest == NULL) {
991 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
992 return -ENOENT;
993 }
994
995 __ip_vs_update_dest(svc, dest, udest, 0);
996 LeaveFunction(2);
997
998 return 0;
999}
1000
1001static void ip_vs_dest_wait_readers(struct rcu_head *head)
1002{
1003 struct ip_vs_dest *dest = container_of(head, struct ip_vs_dest,
1004 rcu_head);
1005
1006
1007 clear_bit(IP_VS_DEST_STATE_REMOVING, &dest->state);
1008}
1009
1010
1011
1012
1013
1014static void __ip_vs_del_dest(struct net *net, struct ip_vs_dest *dest,
1015 bool cleanup)
1016{
1017 struct netns_ipvs *ipvs = net_ipvs(net);
1018
1019 ip_vs_stop_estimator(net, &dest->stats);
1020
1021
1022
1023
1024 ip_vs_rs_unhash(dest);
1025
1026 if (!cleanup) {
1027 set_bit(IP_VS_DEST_STATE_REMOVING, &dest->state);
1028 call_rcu(&dest->rcu_head, ip_vs_dest_wait_readers);
1029 }
1030
1031 spin_lock_bh(&ipvs->dest_trash_lock);
1032 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1033 IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port),
1034 atomic_read(&dest->refcnt));
1035 if (list_empty(&ipvs->dest_trash) && !cleanup)
1036 mod_timer(&ipvs->dest_trash_timer,
1037 jiffies + IP_VS_DEST_TRASH_PERIOD);
1038
1039 list_add(&dest->t_list, &ipvs->dest_trash);
1040 spin_unlock_bh(&ipvs->dest_trash_lock);
1041 ip_vs_dest_put(dest);
1042}
1043
1044
1045
1046
1047
1048static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1049 struct ip_vs_dest *dest,
1050 int svcupd)
1051{
1052 dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1053
1054
1055
1056
1057 list_del_rcu(&dest->n_list);
1058 svc->num_dests--;
1059
1060 if (svcupd) {
1061 struct ip_vs_scheduler *sched;
1062
1063 sched = rcu_dereference_protected(svc->scheduler, 1);
1064 if (sched->del_dest)
1065 sched->del_dest(svc, dest);
1066 }
1067}
1068
1069
1070
1071
1072
1073static int
1074ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1075{
1076 struct ip_vs_dest *dest;
1077 __be16 dport = udest->port;
1078
1079 EnterFunction(2);
1080
1081
1082 rcu_read_lock();
1083 dest = ip_vs_lookup_dest(svc, &udest->addr, dport);
1084 rcu_read_unlock();
1085
1086 if (dest == NULL) {
1087 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1088 return -ENOENT;
1089 }
1090
1091
1092
1093
1094 __ip_vs_unlink_dest(svc, dest, 1);
1095
1096
1097
1098
1099 __ip_vs_del_dest(svc->net, dest, false);
1100
1101 LeaveFunction(2);
1102
1103 return 0;
1104}
1105
1106static void ip_vs_dest_trash_expire(unsigned long data)
1107{
1108 struct net *net = (struct net *) data;
1109 struct netns_ipvs *ipvs = net_ipvs(net);
1110 struct ip_vs_dest *dest, *next;
1111
1112 spin_lock(&ipvs->dest_trash_lock);
1113 list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) {
1114
1115 if (test_bit(IP_VS_DEST_STATE_REMOVING, &dest->state))
1116 continue;
1117 if (atomic_read(&dest->refcnt) > 0)
1118 continue;
1119 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1120 dest->vfwmark,
1121 IP_VS_DBG_ADDR(dest->svc->af, &dest->addr),
1122 ntohs(dest->port));
1123 list_del(&dest->t_list);
1124 ip_vs_dest_free(dest);
1125 }
1126 if (!list_empty(&ipvs->dest_trash))
1127 mod_timer(&ipvs->dest_trash_timer,
1128 jiffies + IP_VS_DEST_TRASH_PERIOD);
1129 spin_unlock(&ipvs->dest_trash_lock);
1130}
1131
1132
1133
1134
1135static int
1136ip_vs_add_service(struct net *net, struct ip_vs_service_user_kern *u,
1137 struct ip_vs_service **svc_p)
1138{
1139 int ret = 0;
1140 struct ip_vs_scheduler *sched = NULL;
1141 struct ip_vs_pe *pe = NULL;
1142 struct ip_vs_service *svc = NULL;
1143 struct netns_ipvs *ipvs = net_ipvs(net);
1144
1145
1146 ip_vs_use_count_inc();
1147
1148
1149 sched = ip_vs_scheduler_get(u->sched_name);
1150 if (sched == NULL) {
1151 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1152 ret = -ENOENT;
1153 goto out_err;
1154 }
1155
1156 if (u->pe_name && *u->pe_name) {
1157 pe = ip_vs_pe_getbyname(u->pe_name);
1158 if (pe == NULL) {
1159 pr_info("persistence engine module ip_vs_pe_%s "
1160 "not found\n", u->pe_name);
1161 ret = -ENOENT;
1162 goto out_err;
1163 }
1164 }
1165
1166#ifdef CONFIG_IP_VS_IPV6
1167 if (u->af == AF_INET6) {
1168 __u32 plen = (__force __u32) u->netmask;
1169
1170 if (plen < 1 || plen > 128) {
1171 ret = -EINVAL;
1172 goto out_err;
1173 }
1174 }
1175#endif
1176
1177 svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1178 if (svc == NULL) {
1179 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1180 ret = -ENOMEM;
1181 goto out_err;
1182 }
1183 svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1184 if (!svc->stats.cpustats) {
1185 ret = -ENOMEM;
1186 goto out_err;
1187 }
1188
1189
1190 atomic_set(&svc->refcnt, 0);
1191
1192 svc->af = u->af;
1193 svc->protocol = u->protocol;
1194 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1195 svc->port = u->port;
1196 svc->fwmark = u->fwmark;
1197 svc->flags = u->flags;
1198 svc->timeout = u->timeout * HZ;
1199 svc->netmask = u->netmask;
1200 svc->net = net;
1201
1202 INIT_LIST_HEAD(&svc->destinations);
1203 spin_lock_init(&svc->sched_lock);
1204 spin_lock_init(&svc->stats.lock);
1205
1206
1207 ret = ip_vs_bind_scheduler(svc, sched);
1208 if (ret)
1209 goto out_err;
1210 sched = NULL;
1211
1212
1213 RCU_INIT_POINTER(svc->pe, pe);
1214 pe = NULL;
1215
1216
1217 if (svc->port == FTPPORT)
1218 atomic_inc(&ipvs->ftpsvc_counter);
1219 else if (svc->port == 0)
1220 atomic_inc(&ipvs->nullsvc_counter);
1221
1222 ip_vs_start_estimator(net, &svc->stats);
1223
1224
1225 if (svc->af == AF_INET)
1226 ipvs->num_services++;
1227
1228
1229 ip_vs_svc_hash(svc);
1230
1231 *svc_p = svc;
1232
1233 ipvs->enable = 1;
1234 return 0;
1235
1236
1237 out_err:
1238 if (svc != NULL) {
1239 ip_vs_unbind_scheduler(svc, sched);
1240 ip_vs_service_free(svc);
1241 }
1242 ip_vs_scheduler_put(sched);
1243 ip_vs_pe_put(pe);
1244
1245
1246 ip_vs_use_count_dec();
1247
1248 return ret;
1249}
1250
1251
1252
1253
1254
1255static int
1256ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1257{
1258 struct ip_vs_scheduler *sched, *old_sched;
1259 struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1260 int ret = 0;
1261
1262
1263
1264
1265 sched = ip_vs_scheduler_get(u->sched_name);
1266 if (sched == NULL) {
1267 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1268 return -ENOENT;
1269 }
1270 old_sched = sched;
1271
1272 if (u->pe_name && *u->pe_name) {
1273 pe = ip_vs_pe_getbyname(u->pe_name);
1274 if (pe == NULL) {
1275 pr_info("persistence engine module ip_vs_pe_%s "
1276 "not found\n", u->pe_name);
1277 ret = -ENOENT;
1278 goto out;
1279 }
1280 old_pe = pe;
1281 }
1282
1283#ifdef CONFIG_IP_VS_IPV6
1284 if (u->af == AF_INET6) {
1285 __u32 plen = (__force __u32) u->netmask;
1286
1287 if (plen < 1 || plen > 128) {
1288 ret = -EINVAL;
1289 goto out;
1290 }
1291 }
1292#endif
1293
1294 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1295 if (sched != old_sched) {
1296
1297 ret = ip_vs_bind_scheduler(svc, sched);
1298 if (ret) {
1299 old_sched = sched;
1300 goto out;
1301 }
1302
1303 ip_vs_unbind_scheduler(svc, old_sched);
1304 }
1305
1306
1307
1308
1309 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1310 svc->timeout = u->timeout * HZ;
1311 svc->netmask = u->netmask;
1312
1313 old_pe = rcu_dereference_protected(svc->pe, 1);
1314 if (pe != old_pe)
1315 rcu_assign_pointer(svc->pe, pe);
1316
1317out:
1318 ip_vs_scheduler_put(old_sched);
1319 ip_vs_pe_put(old_pe);
1320 return ret;
1321}
1322
1323static void ip_vs_service_rcu_free(struct rcu_head *head)
1324{
1325 struct ip_vs_service *svc;
1326
1327 svc = container_of(head, struct ip_vs_service, rcu_head);
1328 ip_vs_service_free(svc);
1329}
1330
1331
1332
1333
1334
1335
1336static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1337{
1338 struct ip_vs_dest *dest, *nxt;
1339 struct ip_vs_scheduler *old_sched;
1340 struct ip_vs_pe *old_pe;
1341 struct netns_ipvs *ipvs = net_ipvs(svc->net);
1342
1343 pr_info("%s: enter\n", __func__);
1344
1345
1346 if (svc->af == AF_INET)
1347 ipvs->num_services--;
1348
1349 ip_vs_stop_estimator(svc->net, &svc->stats);
1350
1351
1352 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1353 ip_vs_unbind_scheduler(svc, old_sched);
1354 ip_vs_scheduler_put(old_sched);
1355
1356
1357 old_pe = rcu_dereference_protected(svc->pe, 1);
1358 ip_vs_pe_put(old_pe);
1359
1360
1361
1362
1363 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1364 __ip_vs_unlink_dest(svc, dest, 0);
1365 __ip_vs_del_dest(svc->net, dest, cleanup);
1366 }
1367
1368
1369
1370
1371 if (svc->port == FTPPORT)
1372 atomic_dec(&ipvs->ftpsvc_counter);
1373 else if (svc->port == 0)
1374 atomic_dec(&ipvs->nullsvc_counter);
1375
1376
1377
1378
1379 if (atomic_dec_and_test(&svc->refcnt)) {
1380 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n",
1381 svc->fwmark,
1382 IP_VS_DBG_ADDR(svc->af, &svc->addr),
1383 ntohs(svc->port));
1384 call_rcu(&svc->rcu_head, ip_vs_service_rcu_free);
1385 }
1386
1387
1388 ip_vs_use_count_dec();
1389}
1390
1391
1392
1393
1394static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1395{
1396
1397 atomic_inc(&svc->refcnt);
1398
1399
1400
1401 ip_vs_svc_unhash(svc);
1402
1403 __ip_vs_del_service(svc, cleanup);
1404}
1405
1406
1407
1408
1409static int ip_vs_del_service(struct ip_vs_service *svc)
1410{
1411 if (svc == NULL)
1412 return -EEXIST;
1413 ip_vs_unlink_service(svc, false);
1414
1415 return 0;
1416}
1417
1418
1419
1420
1421
1422static int ip_vs_flush(struct net *net, bool cleanup)
1423{
1424 int idx;
1425 struct ip_vs_service *svc;
1426 struct hlist_node *n;
1427
1428
1429
1430
1431 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1432 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1433 s_list) {
1434 if (net_eq(svc->net, net))
1435 ip_vs_unlink_service(svc, cleanup);
1436 }
1437 }
1438
1439
1440
1441
1442 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1443 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1444 f_list) {
1445 if (net_eq(svc->net, net))
1446 ip_vs_unlink_service(svc, cleanup);
1447 }
1448 }
1449
1450 return 0;
1451}
1452
1453
1454
1455
1456
1457void ip_vs_service_net_cleanup(struct net *net)
1458{
1459 EnterFunction(2);
1460
1461 mutex_lock(&__ip_vs_mutex);
1462 ip_vs_flush(net, true);
1463 mutex_unlock(&__ip_vs_mutex);
1464 LeaveFunction(2);
1465}
1466
1467
1468static inline void
1469ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1470{
1471 struct ip_vs_dest_dst *dest_dst;
1472
1473 spin_lock_bh(&dest->dst_lock);
1474 dest_dst = rcu_dereference_protected(dest->dest_dst, 1);
1475 if (dest_dst && dest_dst->dst_cache->dev == dev) {
1476 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1477 dev->name,
1478 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1479 ntohs(dest->port),
1480 atomic_read(&dest->refcnt));
1481 __ip_vs_dst_cache_reset(dest);
1482 }
1483 spin_unlock_bh(&dest->dst_lock);
1484
1485}
1486
1487
1488
1489static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1490 void *ptr)
1491{
1492 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1493 struct net *net = dev_net(dev);
1494 struct netns_ipvs *ipvs = net_ipvs(net);
1495 struct ip_vs_service *svc;
1496 struct ip_vs_dest *dest;
1497 unsigned int idx;
1498
1499 if (event != NETDEV_DOWN || !ipvs)
1500 return NOTIFY_DONE;
1501 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1502 EnterFunction(2);
1503 mutex_lock(&__ip_vs_mutex);
1504 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1505 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1506 if (net_eq(svc->net, net)) {
1507 list_for_each_entry(dest, &svc->destinations,
1508 n_list) {
1509 ip_vs_forget_dev(dest, dev);
1510 }
1511 }
1512 }
1513
1514 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1515 if (net_eq(svc->net, net)) {
1516 list_for_each_entry(dest, &svc->destinations,
1517 n_list) {
1518 ip_vs_forget_dev(dest, dev);
1519 }
1520 }
1521
1522 }
1523 }
1524
1525 spin_lock_bh(&ipvs->dest_trash_lock);
1526 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1527 ip_vs_forget_dev(dest, dev);
1528 }
1529 spin_unlock_bh(&ipvs->dest_trash_lock);
1530 mutex_unlock(&__ip_vs_mutex);
1531 LeaveFunction(2);
1532 return NOTIFY_DONE;
1533}
1534
1535
1536
1537
1538static int ip_vs_zero_service(struct ip_vs_service *svc)
1539{
1540 struct ip_vs_dest *dest;
1541
1542 list_for_each_entry(dest, &svc->destinations, n_list) {
1543 ip_vs_zero_stats(&dest->stats);
1544 }
1545 ip_vs_zero_stats(&svc->stats);
1546 return 0;
1547}
1548
1549static int ip_vs_zero_all(struct net *net)
1550{
1551 int idx;
1552 struct ip_vs_service *svc;
1553
1554 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1555 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1556 if (net_eq(svc->net, net))
1557 ip_vs_zero_service(svc);
1558 }
1559 }
1560
1561 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1562 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1563 if (net_eq(svc->net, net))
1564 ip_vs_zero_service(svc);
1565 }
1566 }
1567
1568 ip_vs_zero_stats(&net_ipvs(net)->tot_stats);
1569 return 0;
1570}
1571
1572#ifdef CONFIG_SYSCTL
1573
1574static int zero;
1575static int three = 3;
1576
1577static int
1578proc_do_defense_mode(struct ctl_table *table, int write,
1579 void __user *buffer, size_t *lenp, loff_t *ppos)
1580{
1581 struct net *net = current->nsproxy->net_ns;
1582 int *valp = table->data;
1583 int val = *valp;
1584 int rc;
1585
1586 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1587 if (write && (*valp != val)) {
1588 if ((*valp < 0) || (*valp > 3)) {
1589
1590 *valp = val;
1591 } else {
1592 update_defense_level(net_ipvs(net));
1593 }
1594 }
1595 return rc;
1596}
1597
1598static int
1599proc_do_sync_threshold(struct ctl_table *table, int write,
1600 void __user *buffer, size_t *lenp, loff_t *ppos)
1601{
1602 int *valp = table->data;
1603 int val[2];
1604 int rc;
1605
1606
1607 memcpy(val, valp, sizeof(val));
1608
1609 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1610 if (write && (valp[0] < 0 || valp[1] < 0 ||
1611 (valp[0] >= valp[1] && valp[1]))) {
1612
1613 memcpy(valp, val, sizeof(val));
1614 }
1615 return rc;
1616}
1617
1618static int
1619proc_do_sync_mode(struct ctl_table *table, int write,
1620 void __user *buffer, size_t *lenp, loff_t *ppos)
1621{
1622 int *valp = table->data;
1623 int val = *valp;
1624 int rc;
1625
1626 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1627 if (write && (*valp != val)) {
1628 if ((*valp < 0) || (*valp > 1)) {
1629
1630 *valp = val;
1631 }
1632 }
1633 return rc;
1634}
1635
1636static int
1637proc_do_sync_ports(struct ctl_table *table, int write,
1638 void __user *buffer, size_t *lenp, loff_t *ppos)
1639{
1640 int *valp = table->data;
1641 int val = *valp;
1642 int rc;
1643
1644 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1645 if (write && (*valp != val)) {
1646 if (*valp < 1 || !is_power_of_2(*valp)) {
1647
1648 *valp = val;
1649 }
1650 }
1651 return rc;
1652}
1653
1654
1655
1656
1657
1658
1659
1660static struct ctl_table vs_vars[] = {
1661 {
1662 .procname = "amemthresh",
1663 .maxlen = sizeof(int),
1664 .mode = 0644,
1665 .proc_handler = proc_dointvec,
1666 },
1667 {
1668 .procname = "am_droprate",
1669 .maxlen = sizeof(int),
1670 .mode = 0644,
1671 .proc_handler = proc_dointvec,
1672 },
1673 {
1674 .procname = "drop_entry",
1675 .maxlen = sizeof(int),
1676 .mode = 0644,
1677 .proc_handler = proc_do_defense_mode,
1678 },
1679 {
1680 .procname = "drop_packet",
1681 .maxlen = sizeof(int),
1682 .mode = 0644,
1683 .proc_handler = proc_do_defense_mode,
1684 },
1685#ifdef CONFIG_IP_VS_NFCT
1686 {
1687 .procname = "conntrack",
1688 .maxlen = sizeof(int),
1689 .mode = 0644,
1690 .proc_handler = &proc_dointvec,
1691 },
1692#endif
1693 {
1694 .procname = "secure_tcp",
1695 .maxlen = sizeof(int),
1696 .mode = 0644,
1697 .proc_handler = proc_do_defense_mode,
1698 },
1699 {
1700 .procname = "snat_reroute",
1701 .maxlen = sizeof(int),
1702 .mode = 0644,
1703 .proc_handler = &proc_dointvec,
1704 },
1705 {
1706 .procname = "sync_version",
1707 .maxlen = sizeof(int),
1708 .mode = 0644,
1709 .proc_handler = &proc_do_sync_mode,
1710 },
1711 {
1712 .procname = "sync_ports",
1713 .maxlen = sizeof(int),
1714 .mode = 0644,
1715 .proc_handler = &proc_do_sync_ports,
1716 },
1717 {
1718 .procname = "sync_persist_mode",
1719 .maxlen = sizeof(int),
1720 .mode = 0644,
1721 .proc_handler = proc_dointvec,
1722 },
1723 {
1724 .procname = "sync_qlen_max",
1725 .maxlen = sizeof(unsigned long),
1726 .mode = 0644,
1727 .proc_handler = proc_doulongvec_minmax,
1728 },
1729 {
1730 .procname = "sync_sock_size",
1731 .maxlen = sizeof(int),
1732 .mode = 0644,
1733 .proc_handler = proc_dointvec,
1734 },
1735 {
1736 .procname = "cache_bypass",
1737 .maxlen = sizeof(int),
1738 .mode = 0644,
1739 .proc_handler = proc_dointvec,
1740 },
1741 {
1742 .procname = "expire_nodest_conn",
1743 .maxlen = sizeof(int),
1744 .mode = 0644,
1745 .proc_handler = proc_dointvec,
1746 },
1747 {
1748 .procname = "sloppy_tcp",
1749 .maxlen = sizeof(int),
1750 .mode = 0644,
1751 .proc_handler = proc_dointvec,
1752 },
1753 {
1754 .procname = "sloppy_sctp",
1755 .maxlen = sizeof(int),
1756 .mode = 0644,
1757 .proc_handler = proc_dointvec,
1758 },
1759 {
1760 .procname = "expire_quiescent_template",
1761 .maxlen = sizeof(int),
1762 .mode = 0644,
1763 .proc_handler = proc_dointvec,
1764 },
1765 {
1766 .procname = "sync_threshold",
1767 .maxlen =
1768 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1769 .mode = 0644,
1770 .proc_handler = proc_do_sync_threshold,
1771 },
1772 {
1773 .procname = "sync_refresh_period",
1774 .maxlen = sizeof(int),
1775 .mode = 0644,
1776 .proc_handler = proc_dointvec_jiffies,
1777 },
1778 {
1779 .procname = "sync_retries",
1780 .maxlen = sizeof(int),
1781 .mode = 0644,
1782 .proc_handler = proc_dointvec_minmax,
1783 .extra1 = &zero,
1784 .extra2 = &three,
1785 },
1786 {
1787 .procname = "nat_icmp_send",
1788 .maxlen = sizeof(int),
1789 .mode = 0644,
1790 .proc_handler = proc_dointvec,
1791 },
1792 {
1793 .procname = "pmtu_disc",
1794 .maxlen = sizeof(int),
1795 .mode = 0644,
1796 .proc_handler = proc_dointvec,
1797 },
1798 {
1799 .procname = "backup_only",
1800 .maxlen = sizeof(int),
1801 .mode = 0644,
1802 .proc_handler = proc_dointvec,
1803 },
1804#ifdef CONFIG_IP_VS_DEBUG
1805 {
1806 .procname = "debug_level",
1807 .data = &sysctl_ip_vs_debug_level,
1808 .maxlen = sizeof(int),
1809 .mode = 0644,
1810 .proc_handler = proc_dointvec,
1811 },
1812#endif
1813#if 0
1814 {
1815 .procname = "timeout_established",
1816 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1817 .maxlen = sizeof(int),
1818 .mode = 0644,
1819 .proc_handler = proc_dointvec_jiffies,
1820 },
1821 {
1822 .procname = "timeout_synsent",
1823 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1824 .maxlen = sizeof(int),
1825 .mode = 0644,
1826 .proc_handler = proc_dointvec_jiffies,
1827 },
1828 {
1829 .procname = "timeout_synrecv",
1830 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1831 .maxlen = sizeof(int),
1832 .mode = 0644,
1833 .proc_handler = proc_dointvec_jiffies,
1834 },
1835 {
1836 .procname = "timeout_finwait",
1837 .data = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1838 .maxlen = sizeof(int),
1839 .mode = 0644,
1840 .proc_handler = proc_dointvec_jiffies,
1841 },
1842 {
1843 .procname = "timeout_timewait",
1844 .data = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1845 .maxlen = sizeof(int),
1846 .mode = 0644,
1847 .proc_handler = proc_dointvec_jiffies,
1848 },
1849 {
1850 .procname = "timeout_close",
1851 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1852 .maxlen = sizeof(int),
1853 .mode = 0644,
1854 .proc_handler = proc_dointvec_jiffies,
1855 },
1856 {
1857 .procname = "timeout_closewait",
1858 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1859 .maxlen = sizeof(int),
1860 .mode = 0644,
1861 .proc_handler = proc_dointvec_jiffies,
1862 },
1863 {
1864 .procname = "timeout_lastack",
1865 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1866 .maxlen = sizeof(int),
1867 .mode = 0644,
1868 .proc_handler = proc_dointvec_jiffies,
1869 },
1870 {
1871 .procname = "timeout_listen",
1872 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1873 .maxlen = sizeof(int),
1874 .mode = 0644,
1875 .proc_handler = proc_dointvec_jiffies,
1876 },
1877 {
1878 .procname = "timeout_synack",
1879 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1880 .maxlen = sizeof(int),
1881 .mode = 0644,
1882 .proc_handler = proc_dointvec_jiffies,
1883 },
1884 {
1885 .procname = "timeout_udp",
1886 .data = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1887 .maxlen = sizeof(int),
1888 .mode = 0644,
1889 .proc_handler = proc_dointvec_jiffies,
1890 },
1891 {
1892 .procname = "timeout_icmp",
1893 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1894 .maxlen = sizeof(int),
1895 .mode = 0644,
1896 .proc_handler = proc_dointvec_jiffies,
1897 },
1898#endif
1899 { }
1900};
1901
1902#endif
1903
1904#ifdef CONFIG_PROC_FS
1905
1906struct ip_vs_iter {
1907 struct seq_net_private p;
1908 struct hlist_head *table;
1909 int bucket;
1910};
1911
1912
1913
1914
1915
1916static inline const char *ip_vs_fwd_name(unsigned int flags)
1917{
1918 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1919 case IP_VS_CONN_F_LOCALNODE:
1920 return "Local";
1921 case IP_VS_CONN_F_TUNNEL:
1922 return "Tunnel";
1923 case IP_VS_CONN_F_DROUTE:
1924 return "Route";
1925 default:
1926 return "Masq";
1927 }
1928}
1929
1930
1931
1932static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1933{
1934 struct net *net = seq_file_net(seq);
1935 struct ip_vs_iter *iter = seq->private;
1936 int idx;
1937 struct ip_vs_service *svc;
1938
1939
1940 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1941 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
1942 if (net_eq(svc->net, net) && pos-- == 0) {
1943 iter->table = ip_vs_svc_table;
1944 iter->bucket = idx;
1945 return svc;
1946 }
1947 }
1948 }
1949
1950
1951 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1952 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
1953 f_list) {
1954 if (net_eq(svc->net, net) && pos-- == 0) {
1955 iter->table = ip_vs_svc_fwm_table;
1956 iter->bucket = idx;
1957 return svc;
1958 }
1959 }
1960 }
1961
1962 return NULL;
1963}
1964
1965static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1966 __acquires(RCU)
1967{
1968 rcu_read_lock();
1969 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1970}
1971
1972
1973static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1974{
1975 struct hlist_node *e;
1976 struct ip_vs_iter *iter;
1977 struct ip_vs_service *svc;
1978
1979 ++*pos;
1980 if (v == SEQ_START_TOKEN)
1981 return ip_vs_info_array(seq,0);
1982
1983 svc = v;
1984 iter = seq->private;
1985
1986 if (iter->table == ip_vs_svc_table) {
1987
1988 e = rcu_dereference(hlist_next_rcu(&svc->s_list));
1989 if (e)
1990 return hlist_entry(e, struct ip_vs_service, s_list);
1991
1992 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1993 hlist_for_each_entry_rcu(svc,
1994 &ip_vs_svc_table[iter->bucket],
1995 s_list) {
1996 return svc;
1997 }
1998 }
1999
2000 iter->table = ip_vs_svc_fwm_table;
2001 iter->bucket = -1;
2002 goto scan_fwmark;
2003 }
2004
2005
2006 e = rcu_dereference(hlist_next_rcu(&svc->f_list));
2007 if (e)
2008 return hlist_entry(e, struct ip_vs_service, f_list);
2009
2010 scan_fwmark:
2011 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2012 hlist_for_each_entry_rcu(svc,
2013 &ip_vs_svc_fwm_table[iter->bucket],
2014 f_list)
2015 return svc;
2016 }
2017
2018 return NULL;
2019}
2020
2021static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
2022 __releases(RCU)
2023{
2024 rcu_read_unlock();
2025}
2026
2027
2028static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
2029{
2030 if (v == SEQ_START_TOKEN) {
2031 seq_printf(seq,
2032 "IP Virtual Server version %d.%d.%d (size=%d)\n",
2033 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2034 seq_puts(seq,
2035 "Prot LocalAddress:Port Scheduler Flags\n");
2036 seq_puts(seq,
2037 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2038 } else {
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
2044 if (iter->table == ip_vs_svc_table) {
2045#ifdef CONFIG_IP_VS_IPV6
2046 if (svc->af == AF_INET6)
2047 seq_printf(seq, "%s [%pI6]:%04X %s ",
2048 ip_vs_proto_name(svc->protocol),
2049 &svc->addr.in6,
2050 ntohs(svc->port),
2051 sched->name);
2052 else
2053#endif
2054 seq_printf(seq, "%s %08X:%04X %s %s ",
2055 ip_vs_proto_name(svc->protocol),
2056 ntohl(svc->addr.ip),
2057 ntohs(svc->port),
2058 sched->name,
2059 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2060 } else {
2061 seq_printf(seq, "FWM %08X %s %s",
2062 svc->fwmark, sched->name,
2063 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2064 }
2065
2066 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2067 seq_printf(seq, "persistent %d %08X\n",
2068 svc->timeout,
2069 ntohl(svc->netmask));
2070 else
2071 seq_putc(seq, '\n');
2072
2073 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
2074#ifdef CONFIG_IP_VS_IPV6
2075 if (dest->af == AF_INET6)
2076 seq_printf(seq,
2077 " -> [%pI6]:%04X"
2078 " %-7s %-6d %-10d %-10d\n",
2079 &dest->addr.in6,
2080 ntohs(dest->port),
2081 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2082 atomic_read(&dest->weight),
2083 atomic_read(&dest->activeconns),
2084 atomic_read(&dest->inactconns));
2085 else
2086#endif
2087 seq_printf(seq,
2088 " -> %08X:%04X "
2089 "%-7s %-6d %-10d %-10d\n",
2090 ntohl(dest->addr.ip),
2091 ntohs(dest->port),
2092 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2093 atomic_read(&dest->weight),
2094 atomic_read(&dest->activeconns),
2095 atomic_read(&dest->inactconns));
2096
2097 }
2098 }
2099 return 0;
2100}
2101
2102static const struct seq_operations ip_vs_info_seq_ops = {
2103 .start = ip_vs_info_seq_start,
2104 .next = ip_vs_info_seq_next,
2105 .stop = ip_vs_info_seq_stop,
2106 .show = ip_vs_info_seq_show,
2107};
2108
2109static int ip_vs_info_open(struct inode *inode, struct file *file)
2110{
2111 return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2112 sizeof(struct ip_vs_iter));
2113}
2114
2115static const struct file_operations ip_vs_info_fops = {
2116 .owner = THIS_MODULE,
2117 .open = ip_vs_info_open,
2118 .read = seq_read,
2119 .llseek = seq_lseek,
2120 .release = seq_release_net,
2121};
2122
2123static int ip_vs_stats_show(struct seq_file *seq, void *v)
2124{
2125 struct net *net = seq_file_single_net(seq);
2126 struct ip_vs_stats_user show;
2127
2128
2129 seq_puts(seq,
2130 " Total Incoming Outgoing Incoming Outgoing\n");
2131 seq_printf(seq,
2132 " Conns Packets Packets Bytes Bytes\n");
2133
2134 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2135 seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", show.conns,
2136 show.inpkts, show.outpkts,
2137 (unsigned long long) show.inbytes,
2138 (unsigned long long) show.outbytes);
2139
2140
2141 seq_puts(seq,
2142 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2143 seq_printf(seq, "%8X %8X %8X %16X %16X\n",
2144 show.cps, show.inpps, show.outpps,
2145 show.inbps, show.outbps);
2146
2147 return 0;
2148}
2149
2150static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2151{
2152 return single_open_net(inode, file, ip_vs_stats_show);
2153}
2154
2155static const struct file_operations ip_vs_stats_fops = {
2156 .owner = THIS_MODULE,
2157 .open = ip_vs_stats_seq_open,
2158 .read = seq_read,
2159 .llseek = seq_lseek,
2160 .release = single_release_net,
2161};
2162
2163static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2164{
2165 struct net *net = seq_file_single_net(seq);
2166 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2167 struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats;
2168 struct ip_vs_stats_user rates;
2169 int i;
2170
2171
2172 seq_puts(seq,
2173 " Total Incoming Outgoing Incoming Outgoing\n");
2174 seq_printf(seq,
2175 "CPU Conns Packets Packets Bytes Bytes\n");
2176
2177 for_each_possible_cpu(i) {
2178 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2179 unsigned int start;
2180 __u64 inbytes, outbytes;
2181
2182 do {
2183 start = u64_stats_fetch_begin_bh(&u->syncp);
2184 inbytes = u->ustats.inbytes;
2185 outbytes = u->ustats.outbytes;
2186 } while (u64_stats_fetch_retry_bh(&u->syncp, start));
2187
2188 seq_printf(seq, "%3X %8X %8X %8X %16LX %16LX\n",
2189 i, u->ustats.conns, u->ustats.inpkts,
2190 u->ustats.outpkts, (__u64)inbytes,
2191 (__u64)outbytes);
2192 }
2193
2194 spin_lock_bh(&tot_stats->lock);
2195
2196 seq_printf(seq, " ~ %8X %8X %8X %16LX %16LX\n\n",
2197 tot_stats->ustats.conns, tot_stats->ustats.inpkts,
2198 tot_stats->ustats.outpkts,
2199 (unsigned long long) tot_stats->ustats.inbytes,
2200 (unsigned long long) tot_stats->ustats.outbytes);
2201
2202 ip_vs_read_estimator(&rates, tot_stats);
2203
2204 spin_unlock_bh(&tot_stats->lock);
2205
2206
2207 seq_puts(seq,
2208 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2209 seq_printf(seq, " %8X %8X %8X %16X %16X\n",
2210 rates.cps,
2211 rates.inpps,
2212 rates.outpps,
2213 rates.inbps,
2214 rates.outbps);
2215
2216 return 0;
2217}
2218
2219static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2220{
2221 return single_open_net(inode, file, ip_vs_stats_percpu_show);
2222}
2223
2224static const struct file_operations ip_vs_stats_percpu_fops = {
2225 .owner = THIS_MODULE,
2226 .open = ip_vs_stats_percpu_seq_open,
2227 .read = seq_read,
2228 .llseek = seq_lseek,
2229 .release = single_release_net,
2230};
2231#endif
2232
2233
2234
2235
2236static int ip_vs_set_timeout(struct net *net, struct ip_vs_timeout_user *u)
2237{
2238#if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2239 struct ip_vs_proto_data *pd;
2240#endif
2241
2242 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2243 u->tcp_timeout,
2244 u->tcp_fin_timeout,
2245 u->udp_timeout);
2246
2247#ifdef CONFIG_IP_VS_PROTO_TCP
2248 if (u->tcp_timeout) {
2249 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2250 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2251 = u->tcp_timeout * HZ;
2252 }
2253
2254 if (u->tcp_fin_timeout) {
2255 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2256 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2257 = u->tcp_fin_timeout * HZ;
2258 }
2259#endif
2260
2261#ifdef CONFIG_IP_VS_PROTO_UDP
2262 if (u->udp_timeout) {
2263 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2264 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2265 = u->udp_timeout * HZ;
2266 }
2267#endif
2268 return 0;
2269}
2270
2271
2272#define SET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2273#define SERVICE_ARG_LEN (sizeof(struct ip_vs_service_user))
2274#define SVCDEST_ARG_LEN (sizeof(struct ip_vs_service_user) + \
2275 sizeof(struct ip_vs_dest_user))
2276#define TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2277#define DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user))
2278#define MAX_ARG_LEN SVCDEST_ARG_LEN
2279
2280static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
2281 [SET_CMDID(IP_VS_SO_SET_ADD)] = SERVICE_ARG_LEN,
2282 [SET_CMDID(IP_VS_SO_SET_EDIT)] = SERVICE_ARG_LEN,
2283 [SET_CMDID(IP_VS_SO_SET_DEL)] = SERVICE_ARG_LEN,
2284 [SET_CMDID(IP_VS_SO_SET_FLUSH)] = 0,
2285 [SET_CMDID(IP_VS_SO_SET_ADDDEST)] = SVCDEST_ARG_LEN,
2286 [SET_CMDID(IP_VS_SO_SET_DELDEST)] = SVCDEST_ARG_LEN,
2287 [SET_CMDID(IP_VS_SO_SET_EDITDEST)] = SVCDEST_ARG_LEN,
2288 [SET_CMDID(IP_VS_SO_SET_TIMEOUT)] = TIMEOUT_ARG_LEN,
2289 [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)] = DAEMON_ARG_LEN,
2290 [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)] = DAEMON_ARG_LEN,
2291 [SET_CMDID(IP_VS_SO_SET_ZERO)] = SERVICE_ARG_LEN,
2292};
2293
2294static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2295 struct ip_vs_service_user *usvc_compat)
2296{
2297 memset(usvc, 0, sizeof(*usvc));
2298
2299 usvc->af = AF_INET;
2300 usvc->protocol = usvc_compat->protocol;
2301 usvc->addr.ip = usvc_compat->addr;
2302 usvc->port = usvc_compat->port;
2303 usvc->fwmark = usvc_compat->fwmark;
2304
2305
2306 usvc->sched_name = usvc_compat->sched_name;
2307
2308 usvc->flags = usvc_compat->flags;
2309 usvc->timeout = usvc_compat->timeout;
2310 usvc->netmask = usvc_compat->netmask;
2311}
2312
2313static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2314 struct ip_vs_dest_user *udest_compat)
2315{
2316 memset(udest, 0, sizeof(*udest));
2317
2318 udest->addr.ip = udest_compat->addr;
2319 udest->port = udest_compat->port;
2320 udest->conn_flags = udest_compat->conn_flags;
2321 udest->weight = udest_compat->weight;
2322 udest->u_threshold = udest_compat->u_threshold;
2323 udest->l_threshold = udest_compat->l_threshold;
2324}
2325
2326static int
2327do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2328{
2329 struct net *net = sock_net(sk);
2330 int ret;
2331 unsigned char arg[MAX_ARG_LEN];
2332 struct ip_vs_service_user *usvc_compat;
2333 struct ip_vs_service_user_kern usvc;
2334 struct ip_vs_service *svc;
2335 struct ip_vs_dest_user *udest_compat;
2336 struct ip_vs_dest_user_kern udest;
2337 struct netns_ipvs *ipvs = net_ipvs(net);
2338
2339 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2340 return -EPERM;
2341
2342 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2343 return -EINVAL;
2344 if (len < 0 || len > MAX_ARG_LEN)
2345 return -EINVAL;
2346 if (len != set_arglen[SET_CMDID(cmd)]) {
2347 pr_err("set_ctl: len %u != %u\n",
2348 len, set_arglen[SET_CMDID(cmd)]);
2349 return -EINVAL;
2350 }
2351
2352 if (copy_from_user(arg, user, len) != 0)
2353 return -EFAULT;
2354
2355
2356 ip_vs_use_count_inc();
2357
2358
2359 if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2360 cmd == IP_VS_SO_SET_STOPDAEMON) {
2361 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2362
2363 if (mutex_lock_interruptible(&ipvs->sync_mutex)) {
2364 ret = -ERESTARTSYS;
2365 goto out_dec;
2366 }
2367 if (cmd == IP_VS_SO_SET_STARTDAEMON)
2368 ret = start_sync_thread(net, dm->state, dm->mcast_ifn,
2369 dm->syncid);
2370 else
2371 ret = stop_sync_thread(net, dm->state);
2372 mutex_unlock(&ipvs->sync_mutex);
2373 goto out_dec;
2374 }
2375
2376 if (mutex_lock_interruptible(&__ip_vs_mutex)) {
2377 ret = -ERESTARTSYS;
2378 goto out_dec;
2379 }
2380
2381 if (cmd == IP_VS_SO_SET_FLUSH) {
2382
2383 ret = ip_vs_flush(net, false);
2384 goto out_unlock;
2385 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2386
2387 ret = ip_vs_set_timeout(net, (struct ip_vs_timeout_user *)arg);
2388 goto out_unlock;
2389 }
2390
2391 usvc_compat = (struct ip_vs_service_user *)arg;
2392 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2393
2394
2395
2396 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2397 ip_vs_copy_udest_compat(&udest, udest_compat);
2398
2399 if (cmd == IP_VS_SO_SET_ZERO) {
2400
2401 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2402 ret = ip_vs_zero_all(net);
2403 goto out_unlock;
2404 }
2405 }
2406
2407
2408 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2409 usvc.protocol != IPPROTO_SCTP) {
2410 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2411 usvc.protocol, &usvc.addr.ip,
2412 ntohs(usvc.port), usvc.sched_name);
2413 ret = -EFAULT;
2414 goto out_unlock;
2415 }
2416
2417
2418 rcu_read_lock();
2419 if (usvc.fwmark == 0)
2420 svc = __ip_vs_service_find(net, usvc.af, usvc.protocol,
2421 &usvc.addr, usvc.port);
2422 else
2423 svc = __ip_vs_svc_fwm_find(net, usvc.af, usvc.fwmark);
2424 rcu_read_unlock();
2425
2426 if (cmd != IP_VS_SO_SET_ADD
2427 && (svc == NULL || svc->protocol != usvc.protocol)) {
2428 ret = -ESRCH;
2429 goto out_unlock;
2430 }
2431
2432 switch (cmd) {
2433 case IP_VS_SO_SET_ADD:
2434 if (svc != NULL)
2435 ret = -EEXIST;
2436 else
2437 ret = ip_vs_add_service(net, &usvc, &svc);
2438 break;
2439 case IP_VS_SO_SET_EDIT:
2440 ret = ip_vs_edit_service(svc, &usvc);
2441 break;
2442 case IP_VS_SO_SET_DEL:
2443 ret = ip_vs_del_service(svc);
2444 if (!ret)
2445 goto out_unlock;
2446 break;
2447 case IP_VS_SO_SET_ZERO:
2448 ret = ip_vs_zero_service(svc);
2449 break;
2450 case IP_VS_SO_SET_ADDDEST:
2451 ret = ip_vs_add_dest(svc, &udest);
2452 break;
2453 case IP_VS_SO_SET_EDITDEST:
2454 ret = ip_vs_edit_dest(svc, &udest);
2455 break;
2456 case IP_VS_SO_SET_DELDEST:
2457 ret = ip_vs_del_dest(svc, &udest);
2458 break;
2459 default:
2460 ret = -EINVAL;
2461 }
2462
2463 out_unlock:
2464 mutex_unlock(&__ip_vs_mutex);
2465 out_dec:
2466
2467 ip_vs_use_count_dec();
2468
2469 return ret;
2470}
2471
2472
2473static void
2474ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2475{
2476 struct ip_vs_scheduler *sched;
2477
2478 sched = rcu_dereference_protected(src->scheduler, 1);
2479 dst->protocol = src->protocol;
2480 dst->addr = src->addr.ip;
2481 dst->port = src->port;
2482 dst->fwmark = src->fwmark;
2483 strlcpy(dst->sched_name, sched->name, sizeof(dst->sched_name));
2484 dst->flags = src->flags;
2485 dst->timeout = src->timeout / HZ;
2486 dst->netmask = src->netmask;
2487 dst->num_dests = src->num_dests;
2488 ip_vs_copy_stats(&dst->stats, &src->stats);
2489}
2490
2491static inline int
2492__ip_vs_get_service_entries(struct net *net,
2493 const struct ip_vs_get_services *get,
2494 struct ip_vs_get_services __user *uptr)
2495{
2496 int idx, count=0;
2497 struct ip_vs_service *svc;
2498 struct ip_vs_service_entry entry;
2499 int ret = 0;
2500
2501 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2502 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2503
2504 if (svc->af != AF_INET || !net_eq(svc->net, net))
2505 continue;
2506
2507 if (count >= get->num_services)
2508 goto out;
2509 memset(&entry, 0, sizeof(entry));
2510 ip_vs_copy_service(&entry, svc);
2511 if (copy_to_user(&uptr->entrytable[count],
2512 &entry, sizeof(entry))) {
2513 ret = -EFAULT;
2514 goto out;
2515 }
2516 count++;
2517 }
2518 }
2519
2520 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2521 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2522
2523 if (svc->af != AF_INET || !net_eq(svc->net, net))
2524 continue;
2525
2526 if (count >= get->num_services)
2527 goto out;
2528 memset(&entry, 0, sizeof(entry));
2529 ip_vs_copy_service(&entry, svc);
2530 if (copy_to_user(&uptr->entrytable[count],
2531 &entry, sizeof(entry))) {
2532 ret = -EFAULT;
2533 goto out;
2534 }
2535 count++;
2536 }
2537 }
2538out:
2539 return ret;
2540}
2541
2542static inline int
2543__ip_vs_get_dest_entries(struct net *net, const struct ip_vs_get_dests *get,
2544 struct ip_vs_get_dests __user *uptr)
2545{
2546 struct ip_vs_service *svc;
2547 union nf_inet_addr addr = { .ip = get->addr };
2548 int ret = 0;
2549
2550 rcu_read_lock();
2551 if (get->fwmark)
2552 svc = __ip_vs_svc_fwm_find(net, AF_INET, get->fwmark);
2553 else
2554 svc = __ip_vs_service_find(net, AF_INET, get->protocol, &addr,
2555 get->port);
2556 rcu_read_unlock();
2557
2558 if (svc) {
2559 int count = 0;
2560 struct ip_vs_dest *dest;
2561 struct ip_vs_dest_entry entry;
2562
2563 memset(&entry, 0, sizeof(entry));
2564 list_for_each_entry(dest, &svc->destinations, n_list) {
2565 if (count >= get->num_dests)
2566 break;
2567
2568 entry.addr = dest->addr.ip;
2569 entry.port = dest->port;
2570 entry.conn_flags = atomic_read(&dest->conn_flags);
2571 entry.weight = atomic_read(&dest->weight);
2572 entry.u_threshold = dest->u_threshold;
2573 entry.l_threshold = dest->l_threshold;
2574 entry.activeconns = atomic_read(&dest->activeconns);
2575 entry.inactconns = atomic_read(&dest->inactconns);
2576 entry.persistconns = atomic_read(&dest->persistconns);
2577 ip_vs_copy_stats(&entry.stats, &dest->stats);
2578 if (copy_to_user(&uptr->entrytable[count],
2579 &entry, sizeof(entry))) {
2580 ret = -EFAULT;
2581 break;
2582 }
2583 count++;
2584 }
2585 } else
2586 ret = -ESRCH;
2587 return ret;
2588}
2589
2590static inline void
2591__ip_vs_get_timeouts(struct net *net, struct ip_vs_timeout_user *u)
2592{
2593#if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2594 struct ip_vs_proto_data *pd;
2595#endif
2596
2597 memset(u, 0, sizeof (*u));
2598
2599#ifdef CONFIG_IP_VS_PROTO_TCP
2600 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2601 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2602 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2603#endif
2604#ifdef CONFIG_IP_VS_PROTO_UDP
2605 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2606 u->udp_timeout =
2607 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2608#endif
2609}
2610
2611
2612#define GET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2613#define GET_INFO_ARG_LEN (sizeof(struct ip_vs_getinfo))
2614#define GET_SERVICES_ARG_LEN (sizeof(struct ip_vs_get_services))
2615#define GET_SERVICE_ARG_LEN (sizeof(struct ip_vs_service_entry))
2616#define GET_DESTS_ARG_LEN (sizeof(struct ip_vs_get_dests))
2617#define GET_TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2618#define GET_DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user) * 2)
2619
2620static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
2621 [GET_CMDID(IP_VS_SO_GET_VERSION)] = 64,
2622 [GET_CMDID(IP_VS_SO_GET_INFO)] = GET_INFO_ARG_LEN,
2623 [GET_CMDID(IP_VS_SO_GET_SERVICES)] = GET_SERVICES_ARG_LEN,
2624 [GET_CMDID(IP_VS_SO_GET_SERVICE)] = GET_SERVICE_ARG_LEN,
2625 [GET_CMDID(IP_VS_SO_GET_DESTS)] = GET_DESTS_ARG_LEN,
2626 [GET_CMDID(IP_VS_SO_GET_TIMEOUT)] = GET_TIMEOUT_ARG_LEN,
2627 [GET_CMDID(IP_VS_SO_GET_DAEMON)] = GET_DAEMON_ARG_LEN,
2628};
2629
2630static int
2631do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2632{
2633 unsigned char arg[128];
2634 int ret = 0;
2635 unsigned int copylen;
2636 struct net *net = sock_net(sk);
2637 struct netns_ipvs *ipvs = net_ipvs(net);
2638
2639 BUG_ON(!net);
2640 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2641 return -EPERM;
2642
2643 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2644 return -EINVAL;
2645
2646 if (*len < get_arglen[GET_CMDID(cmd)]) {
2647 pr_err("get_ctl: len %u < %u\n",
2648 *len, get_arglen[GET_CMDID(cmd)]);
2649 return -EINVAL;
2650 }
2651
2652 copylen = get_arglen[GET_CMDID(cmd)];
2653 if (copylen > 128)
2654 return -EINVAL;
2655
2656 if (copy_from_user(arg, user, copylen) != 0)
2657 return -EFAULT;
2658
2659
2660
2661 if (cmd == IP_VS_SO_GET_DAEMON) {
2662 struct ip_vs_daemon_user d[2];
2663
2664 memset(&d, 0, sizeof(d));
2665 if (mutex_lock_interruptible(&ipvs->sync_mutex))
2666 return -ERESTARTSYS;
2667
2668 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2669 d[0].state = IP_VS_STATE_MASTER;
2670 strlcpy(d[0].mcast_ifn, ipvs->master_mcast_ifn,
2671 sizeof(d[0].mcast_ifn));
2672 d[0].syncid = ipvs->master_syncid;
2673 }
2674 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2675 d[1].state = IP_VS_STATE_BACKUP;
2676 strlcpy(d[1].mcast_ifn, ipvs->backup_mcast_ifn,
2677 sizeof(d[1].mcast_ifn));
2678 d[1].syncid = ipvs->backup_syncid;
2679 }
2680 if (copy_to_user(user, &d, sizeof(d)) != 0)
2681 ret = -EFAULT;
2682 mutex_unlock(&ipvs->sync_mutex);
2683 return ret;
2684 }
2685
2686 if (mutex_lock_interruptible(&__ip_vs_mutex))
2687 return -ERESTARTSYS;
2688
2689 switch (cmd) {
2690 case IP_VS_SO_GET_VERSION:
2691 {
2692 char buf[64];
2693
2694 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2695 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2696 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2697 ret = -EFAULT;
2698 goto out;
2699 }
2700 *len = strlen(buf)+1;
2701 }
2702 break;
2703
2704 case IP_VS_SO_GET_INFO:
2705 {
2706 struct ip_vs_getinfo info;
2707 info.version = IP_VS_VERSION_CODE;
2708 info.size = ip_vs_conn_tab_size;
2709 info.num_services = ipvs->num_services;
2710 if (copy_to_user(user, &info, sizeof(info)) != 0)
2711 ret = -EFAULT;
2712 }
2713 break;
2714
2715 case IP_VS_SO_GET_SERVICES:
2716 {
2717 struct ip_vs_get_services *get;
2718 int size;
2719
2720 get = (struct ip_vs_get_services *)arg;
2721 size = sizeof(*get) +
2722 sizeof(struct ip_vs_service_entry) * get->num_services;
2723 if (*len != size) {
2724 pr_err("length: %u != %u\n", *len, size);
2725 ret = -EINVAL;
2726 goto out;
2727 }
2728 ret = __ip_vs_get_service_entries(net, get, user);
2729 }
2730 break;
2731
2732 case IP_VS_SO_GET_SERVICE:
2733 {
2734 struct ip_vs_service_entry *entry;
2735 struct ip_vs_service *svc;
2736 union nf_inet_addr addr;
2737
2738 entry = (struct ip_vs_service_entry *)arg;
2739 addr.ip = entry->addr;
2740 rcu_read_lock();
2741 if (entry->fwmark)
2742 svc = __ip_vs_svc_fwm_find(net, AF_INET, entry->fwmark);
2743 else
2744 svc = __ip_vs_service_find(net, AF_INET,
2745 entry->protocol, &addr,
2746 entry->port);
2747 rcu_read_unlock();
2748 if (svc) {
2749 ip_vs_copy_service(entry, svc);
2750 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2751 ret = -EFAULT;
2752 } else
2753 ret = -ESRCH;
2754 }
2755 break;
2756
2757 case IP_VS_SO_GET_DESTS:
2758 {
2759 struct ip_vs_get_dests *get;
2760 int size;
2761
2762 get = (struct ip_vs_get_dests *)arg;
2763 size = sizeof(*get) +
2764 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2765 if (*len != size) {
2766 pr_err("length: %u != %u\n", *len, size);
2767 ret = -EINVAL;
2768 goto out;
2769 }
2770 ret = __ip_vs_get_dest_entries(net, get, user);
2771 }
2772 break;
2773
2774 case IP_VS_SO_GET_TIMEOUT:
2775 {
2776 struct ip_vs_timeout_user t;
2777
2778 __ip_vs_get_timeouts(net, &t);
2779 if (copy_to_user(user, &t, sizeof(t)) != 0)
2780 ret = -EFAULT;
2781 }
2782 break;
2783
2784 default:
2785 ret = -EINVAL;
2786 }
2787
2788out:
2789 mutex_unlock(&__ip_vs_mutex);
2790 return ret;
2791}
2792
2793
2794static struct nf_sockopt_ops ip_vs_sockopts = {
2795 .pf = PF_INET,
2796 .set_optmin = IP_VS_BASE_CTL,
2797 .set_optmax = IP_VS_SO_SET_MAX+1,
2798 .set = do_ip_vs_set_ctl,
2799 .get_optmin = IP_VS_BASE_CTL,
2800 .get_optmax = IP_VS_SO_GET_MAX+1,
2801 .get = do_ip_vs_get_ctl,
2802 .owner = THIS_MODULE,
2803};
2804
2805
2806
2807
2808
2809
2810static struct genl_family ip_vs_genl_family = {
2811 .id = GENL_ID_GENERATE,
2812 .hdrsize = 0,
2813 .name = IPVS_GENL_NAME,
2814 .version = IPVS_GENL_VERSION,
2815 .maxattr = IPVS_CMD_MAX,
2816 .netnsok = true,
2817};
2818
2819
2820static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2821 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2822 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2823 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2824 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2825 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2826 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2827};
2828
2829
2830static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2831 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2832 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2833 .len = IP_VS_IFNAME_MAXLEN },
2834 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2835};
2836
2837
2838static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2839 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2840 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2841 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2842 .len = sizeof(union nf_inet_addr) },
2843 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2844 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2845 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2846 .len = IP_VS_SCHEDNAME_MAXLEN },
2847 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2848 .len = IP_VS_PENAME_MAXLEN },
2849 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2850 .len = sizeof(struct ip_vs_flags) },
2851 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2852 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2853 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2854};
2855
2856
2857static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2858 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2859 .len = sizeof(union nf_inet_addr) },
2860 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2861 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2862 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2863 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2864 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2865 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2866 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2867 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2868 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2869};
2870
2871static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2872 struct ip_vs_stats *stats)
2873{
2874 struct ip_vs_stats_user ustats;
2875 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2876 if (!nl_stats)
2877 return -EMSGSIZE;
2878
2879 ip_vs_copy_stats(&ustats, stats);
2880
2881 if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, ustats.conns) ||
2882 nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, ustats.inpkts) ||
2883 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, ustats.outpkts) ||
2884 nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, ustats.inbytes) ||
2885 nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, ustats.outbytes) ||
2886 nla_put_u32(skb, IPVS_STATS_ATTR_CPS, ustats.cps) ||
2887 nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, ustats.inpps) ||
2888 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, ustats.outpps) ||
2889 nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, ustats.inbps) ||
2890 nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, ustats.outbps))
2891 goto nla_put_failure;
2892 nla_nest_end(skb, nl_stats);
2893
2894 return 0;
2895
2896nla_put_failure:
2897 nla_nest_cancel(skb, nl_stats);
2898 return -EMSGSIZE;
2899}
2900
2901static int ip_vs_genl_fill_service(struct sk_buff *skb,
2902 struct ip_vs_service *svc)
2903{
2904 struct ip_vs_scheduler *sched;
2905 struct ip_vs_pe *pe;
2906 struct nlattr *nl_service;
2907 struct ip_vs_flags flags = { .flags = svc->flags,
2908 .mask = ~0 };
2909
2910 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2911 if (!nl_service)
2912 return -EMSGSIZE;
2913
2914 if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2915 goto nla_put_failure;
2916 if (svc->fwmark) {
2917 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2918 goto nla_put_failure;
2919 } else {
2920 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2921 nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2922 nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2923 goto nla_put_failure;
2924 }
2925
2926 sched = rcu_dereference_protected(svc->scheduler, 1);
2927 pe = rcu_dereference_protected(svc->pe, 1);
2928 if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched->name) ||
2929 (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
2930 nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2931 nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2932 nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
2933 goto nla_put_failure;
2934 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2935 goto nla_put_failure;
2936
2937 nla_nest_end(skb, nl_service);
2938
2939 return 0;
2940
2941nla_put_failure:
2942 nla_nest_cancel(skb, nl_service);
2943 return -EMSGSIZE;
2944}
2945
2946static int ip_vs_genl_dump_service(struct sk_buff *skb,
2947 struct ip_vs_service *svc,
2948 struct netlink_callback *cb)
2949{
2950 void *hdr;
2951
2952 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
2953 &ip_vs_genl_family, NLM_F_MULTI,
2954 IPVS_CMD_NEW_SERVICE);
2955 if (!hdr)
2956 return -EMSGSIZE;
2957
2958 if (ip_vs_genl_fill_service(skb, svc) < 0)
2959 goto nla_put_failure;
2960
2961 return genlmsg_end(skb, hdr);
2962
2963nla_put_failure:
2964 genlmsg_cancel(skb, hdr);
2965 return -EMSGSIZE;
2966}
2967
2968static int ip_vs_genl_dump_services(struct sk_buff *skb,
2969 struct netlink_callback *cb)
2970{
2971 int idx = 0, i;
2972 int start = cb->args[0];
2973 struct ip_vs_service *svc;
2974 struct net *net = skb_sknet(skb);
2975
2976 mutex_lock(&__ip_vs_mutex);
2977 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2978 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2979 if (++idx <= start || !net_eq(svc->net, net))
2980 continue;
2981 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2982 idx--;
2983 goto nla_put_failure;
2984 }
2985 }
2986 }
2987
2988 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2989 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
2990 if (++idx <= start || !net_eq(svc->net, net))
2991 continue;
2992 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2993 idx--;
2994 goto nla_put_failure;
2995 }
2996 }
2997 }
2998
2999nla_put_failure:
3000 mutex_unlock(&__ip_vs_mutex);
3001 cb->args[0] = idx;
3002
3003 return skb->len;
3004}
3005
3006static int ip_vs_genl_parse_service(struct net *net,
3007 struct ip_vs_service_user_kern *usvc,
3008 struct nlattr *nla, int full_entry,
3009 struct ip_vs_service **ret_svc)
3010{
3011 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
3012 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
3013 struct ip_vs_service *svc;
3014
3015
3016 if (nla == NULL ||
3017 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
3018 return -EINVAL;
3019
3020 nla_af = attrs[IPVS_SVC_ATTR_AF];
3021 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
3022 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
3023 nla_port = attrs[IPVS_SVC_ATTR_PORT];
3024 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
3025
3026 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3027 return -EINVAL;
3028
3029 memset(usvc, 0, sizeof(*usvc));
3030
3031 usvc->af = nla_get_u16(nla_af);
3032#ifdef CONFIG_IP_VS_IPV6
3033 if (usvc->af != AF_INET && usvc->af != AF_INET6)
3034#else
3035 if (usvc->af != AF_INET)
3036#endif
3037 return -EAFNOSUPPORT;
3038
3039 if (nla_fwmark) {
3040 usvc->protocol = IPPROTO_TCP;
3041 usvc->fwmark = nla_get_u32(nla_fwmark);
3042 } else {
3043 usvc->protocol = nla_get_u16(nla_protocol);
3044 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3045 usvc->port = nla_get_be16(nla_port);
3046 usvc->fwmark = 0;
3047 }
3048
3049 rcu_read_lock();
3050 if (usvc->fwmark)
3051 svc = __ip_vs_svc_fwm_find(net, usvc->af, usvc->fwmark);
3052 else
3053 svc = __ip_vs_service_find(net, usvc->af, usvc->protocol,
3054 &usvc->addr, usvc->port);
3055 rcu_read_unlock();
3056 *ret_svc = svc;
3057
3058
3059 if (full_entry) {
3060 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3061 *nla_netmask;
3062 struct ip_vs_flags flags;
3063
3064 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3065 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3066 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3067 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3068 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3069
3070 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3071 return -EINVAL;
3072
3073 nla_memcpy(&flags, nla_flags, sizeof(flags));
3074
3075
3076 if (svc)
3077 usvc->flags = svc->flags;
3078
3079
3080 usvc->flags = (usvc->flags & ~flags.mask) |
3081 (flags.flags & flags.mask);
3082 usvc->sched_name = nla_data(nla_sched);
3083 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3084 usvc->timeout = nla_get_u32(nla_timeout);
3085 usvc->netmask = nla_get_be32(nla_netmask);
3086 }
3087
3088 return 0;
3089}
3090
3091static struct ip_vs_service *ip_vs_genl_find_service(struct net *net,
3092 struct nlattr *nla)
3093{
3094 struct ip_vs_service_user_kern usvc;
3095 struct ip_vs_service *svc;
3096 int ret;
3097
3098 ret = ip_vs_genl_parse_service(net, &usvc, nla, 0, &svc);
3099 return ret ? ERR_PTR(ret) : svc;
3100}
3101
3102static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3103{
3104 struct nlattr *nl_dest;
3105
3106 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3107 if (!nl_dest)
3108 return -EMSGSIZE;
3109
3110 if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3111 nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3112 nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3113 (atomic_read(&dest->conn_flags) &
3114 IP_VS_CONN_F_FWD_MASK)) ||
3115 nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3116 atomic_read(&dest->weight)) ||
3117 nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3118 nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3119 nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3120 atomic_read(&dest->activeconns)) ||
3121 nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3122 atomic_read(&dest->inactconns)) ||
3123 nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3124 atomic_read(&dest->persistconns)))
3125 goto nla_put_failure;
3126 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
3127 goto nla_put_failure;
3128
3129 nla_nest_end(skb, nl_dest);
3130
3131 return 0;
3132
3133nla_put_failure:
3134 nla_nest_cancel(skb, nl_dest);
3135 return -EMSGSIZE;
3136}
3137
3138static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3139 struct netlink_callback *cb)
3140{
3141 void *hdr;
3142
3143 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3144 &ip_vs_genl_family, NLM_F_MULTI,
3145 IPVS_CMD_NEW_DEST);
3146 if (!hdr)
3147 return -EMSGSIZE;
3148
3149 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3150 goto nla_put_failure;
3151
3152 return genlmsg_end(skb, hdr);
3153
3154nla_put_failure:
3155 genlmsg_cancel(skb, hdr);
3156 return -EMSGSIZE;
3157}
3158
3159static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3160 struct netlink_callback *cb)
3161{
3162 int idx = 0;
3163 int start = cb->args[0];
3164 struct ip_vs_service *svc;
3165 struct ip_vs_dest *dest;
3166 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3167 struct net *net = skb_sknet(skb);
3168
3169 mutex_lock(&__ip_vs_mutex);
3170
3171
3172 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3173 IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3174 goto out_err;
3175
3176
3177 svc = ip_vs_genl_find_service(net, attrs[IPVS_CMD_ATTR_SERVICE]);
3178 if (IS_ERR(svc) || svc == NULL)
3179 goto out_err;
3180
3181
3182 list_for_each_entry(dest, &svc->destinations, n_list) {
3183 if (++idx <= start)
3184 continue;
3185 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3186 idx--;
3187 goto nla_put_failure;
3188 }
3189 }
3190
3191nla_put_failure:
3192 cb->args[0] = idx;
3193
3194out_err:
3195 mutex_unlock(&__ip_vs_mutex);
3196
3197 return skb->len;
3198}
3199
3200static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3201 struct nlattr *nla, int full_entry)
3202{
3203 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3204 struct nlattr *nla_addr, *nla_port;
3205
3206
3207 if (nla == NULL ||
3208 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3209 return -EINVAL;
3210
3211 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3212 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3213
3214 if (!(nla_addr && nla_port))
3215 return -EINVAL;
3216
3217 memset(udest, 0, sizeof(*udest));
3218
3219 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3220 udest->port = nla_get_be16(nla_port);
3221
3222
3223 if (full_entry) {
3224 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3225 *nla_l_thresh;
3226
3227 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3228 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3229 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3230 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3231
3232 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3233 return -EINVAL;
3234
3235 udest->conn_flags = nla_get_u32(nla_fwd)
3236 & IP_VS_CONN_F_FWD_MASK;
3237 udest->weight = nla_get_u32(nla_weight);
3238 udest->u_threshold = nla_get_u32(nla_u_thresh);
3239 udest->l_threshold = nla_get_u32(nla_l_thresh);
3240 }
3241
3242 return 0;
3243}
3244
3245static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3246 const char *mcast_ifn, __u32 syncid)
3247{
3248 struct nlattr *nl_daemon;
3249
3250 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3251 if (!nl_daemon)
3252 return -EMSGSIZE;
3253
3254 if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3255 nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn) ||
3256 nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid))
3257 goto nla_put_failure;
3258 nla_nest_end(skb, nl_daemon);
3259
3260 return 0;
3261
3262nla_put_failure:
3263 nla_nest_cancel(skb, nl_daemon);
3264 return -EMSGSIZE;
3265}
3266
3267static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3268 const char *mcast_ifn, __u32 syncid,
3269 struct netlink_callback *cb)
3270{
3271 void *hdr;
3272 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3273 &ip_vs_genl_family, NLM_F_MULTI,
3274 IPVS_CMD_NEW_DAEMON);
3275 if (!hdr)
3276 return -EMSGSIZE;
3277
3278 if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
3279 goto nla_put_failure;
3280
3281 return genlmsg_end(skb, hdr);
3282
3283nla_put_failure:
3284 genlmsg_cancel(skb, hdr);
3285 return -EMSGSIZE;
3286}
3287
3288static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3289 struct netlink_callback *cb)
3290{
3291 struct net *net = skb_sknet(skb);
3292 struct netns_ipvs *ipvs = net_ipvs(net);
3293
3294 mutex_lock(&ipvs->sync_mutex);
3295 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3296 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3297 ipvs->master_mcast_ifn,
3298 ipvs->master_syncid, cb) < 0)
3299 goto nla_put_failure;
3300
3301 cb->args[0] = 1;
3302 }
3303
3304 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3305 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3306 ipvs->backup_mcast_ifn,
3307 ipvs->backup_syncid, cb) < 0)
3308 goto nla_put_failure;
3309
3310 cb->args[1] = 1;
3311 }
3312
3313nla_put_failure:
3314 mutex_unlock(&ipvs->sync_mutex);
3315
3316 return skb->len;
3317}
3318
3319static int ip_vs_genl_new_daemon(struct net *net, struct nlattr **attrs)
3320{
3321 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3322 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3323 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3324 return -EINVAL;
3325
3326 return start_sync_thread(net,
3327 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3328 nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3329 nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3330}
3331
3332static int ip_vs_genl_del_daemon(struct net *net, struct nlattr **attrs)
3333{
3334 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3335 return -EINVAL;
3336
3337 return stop_sync_thread(net,
3338 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3339}
3340
3341static int ip_vs_genl_set_config(struct net *net, struct nlattr **attrs)
3342{
3343 struct ip_vs_timeout_user t;
3344
3345 __ip_vs_get_timeouts(net, &t);
3346
3347 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3348 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3349
3350 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3351 t.tcp_fin_timeout =
3352 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3353
3354 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3355 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3356
3357 return ip_vs_set_timeout(net, &t);
3358}
3359
3360static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3361{
3362 int ret = 0, cmd;
3363 struct net *net;
3364 struct netns_ipvs *ipvs;
3365
3366 net = skb_sknet(skb);
3367 ipvs = net_ipvs(net);
3368 cmd = info->genlhdr->cmd;
3369
3370 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3371 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3372
3373 mutex_lock(&ipvs->sync_mutex);
3374 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3375 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3376 info->attrs[IPVS_CMD_ATTR_DAEMON],
3377 ip_vs_daemon_policy)) {
3378 ret = -EINVAL;
3379 goto out;
3380 }
3381
3382 if (cmd == IPVS_CMD_NEW_DAEMON)
3383 ret = ip_vs_genl_new_daemon(net, daemon_attrs);
3384 else
3385 ret = ip_vs_genl_del_daemon(net, daemon_attrs);
3386out:
3387 mutex_unlock(&ipvs->sync_mutex);
3388 }
3389 return ret;
3390}
3391
3392static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3393{
3394 struct ip_vs_service *svc = NULL;
3395 struct ip_vs_service_user_kern usvc;
3396 struct ip_vs_dest_user_kern udest;
3397 int ret = 0, cmd;
3398 int need_full_svc = 0, need_full_dest = 0;
3399 struct net *net;
3400
3401 net = skb_sknet(skb);
3402 cmd = info->genlhdr->cmd;
3403
3404 mutex_lock(&__ip_vs_mutex);
3405
3406 if (cmd == IPVS_CMD_FLUSH) {
3407 ret = ip_vs_flush(net, false);
3408 goto out;
3409 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3410 ret = ip_vs_genl_set_config(net, info->attrs);
3411 goto out;
3412 } else if (cmd == IPVS_CMD_ZERO &&
3413 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3414 ret = ip_vs_zero_all(net);
3415 goto out;
3416 }
3417
3418
3419
3420
3421 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3422 need_full_svc = 1;
3423
3424 ret = ip_vs_genl_parse_service(net, &usvc,
3425 info->attrs[IPVS_CMD_ATTR_SERVICE],
3426 need_full_svc, &svc);
3427 if (ret)
3428 goto out;
3429
3430
3431 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3432 ret = -ESRCH;
3433 goto out;
3434 }
3435
3436
3437
3438
3439 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3440 cmd == IPVS_CMD_DEL_DEST) {
3441 if (cmd != IPVS_CMD_DEL_DEST)
3442 need_full_dest = 1;
3443
3444 ret = ip_vs_genl_parse_dest(&udest,
3445 info->attrs[IPVS_CMD_ATTR_DEST],
3446 need_full_dest);
3447 if (ret)
3448 goto out;
3449 }
3450
3451 switch (cmd) {
3452 case IPVS_CMD_NEW_SERVICE:
3453 if (svc == NULL)
3454 ret = ip_vs_add_service(net, &usvc, &svc);
3455 else
3456 ret = -EEXIST;
3457 break;
3458 case IPVS_CMD_SET_SERVICE:
3459 ret = ip_vs_edit_service(svc, &usvc);
3460 break;
3461 case IPVS_CMD_DEL_SERVICE:
3462 ret = ip_vs_del_service(svc);
3463
3464 break;
3465 case IPVS_CMD_NEW_DEST:
3466 ret = ip_vs_add_dest(svc, &udest);
3467 break;
3468 case IPVS_CMD_SET_DEST:
3469 ret = ip_vs_edit_dest(svc, &udest);
3470 break;
3471 case IPVS_CMD_DEL_DEST:
3472 ret = ip_vs_del_dest(svc, &udest);
3473 break;
3474 case IPVS_CMD_ZERO:
3475 ret = ip_vs_zero_service(svc);
3476 break;
3477 default:
3478 ret = -EINVAL;
3479 }
3480
3481out:
3482 mutex_unlock(&__ip_vs_mutex);
3483
3484 return ret;
3485}
3486
3487static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3488{
3489 struct sk_buff *msg;
3490 void *reply;
3491 int ret, cmd, reply_cmd;
3492 struct net *net;
3493
3494 net = skb_sknet(skb);
3495 cmd = info->genlhdr->cmd;
3496
3497 if (cmd == IPVS_CMD_GET_SERVICE)
3498 reply_cmd = IPVS_CMD_NEW_SERVICE;
3499 else if (cmd == IPVS_CMD_GET_INFO)
3500 reply_cmd = IPVS_CMD_SET_INFO;
3501 else if (cmd == IPVS_CMD_GET_CONFIG)
3502 reply_cmd = IPVS_CMD_SET_CONFIG;
3503 else {
3504 pr_err("unknown Generic Netlink command\n");
3505 return -EINVAL;
3506 }
3507
3508 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3509 if (!msg)
3510 return -ENOMEM;
3511
3512 mutex_lock(&__ip_vs_mutex);
3513
3514 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3515 if (reply == NULL)
3516 goto nla_put_failure;
3517
3518 switch (cmd) {
3519 case IPVS_CMD_GET_SERVICE:
3520 {
3521 struct ip_vs_service *svc;
3522
3523 svc = ip_vs_genl_find_service(net,
3524 info->attrs[IPVS_CMD_ATTR_SERVICE]);
3525 if (IS_ERR(svc)) {
3526 ret = PTR_ERR(svc);
3527 goto out_err;
3528 } else if (svc) {
3529 ret = ip_vs_genl_fill_service(msg, svc);
3530 if (ret)
3531 goto nla_put_failure;
3532 } else {
3533 ret = -ESRCH;
3534 goto out_err;
3535 }
3536
3537 break;
3538 }
3539
3540 case IPVS_CMD_GET_CONFIG:
3541 {
3542 struct ip_vs_timeout_user t;
3543
3544 __ip_vs_get_timeouts(net, &t);
3545#ifdef CONFIG_IP_VS_PROTO_TCP
3546 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3547 t.tcp_timeout) ||
3548 nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3549 t.tcp_fin_timeout))
3550 goto nla_put_failure;
3551#endif
3552#ifdef CONFIG_IP_VS_PROTO_UDP
3553 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3554 goto nla_put_failure;
3555#endif
3556
3557 break;
3558 }
3559
3560 case IPVS_CMD_GET_INFO:
3561 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3562 IP_VS_VERSION_CODE) ||
3563 nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3564 ip_vs_conn_tab_size))
3565 goto nla_put_failure;
3566 break;
3567 }
3568
3569 genlmsg_end(msg, reply);
3570 ret = genlmsg_reply(msg, info);
3571 goto out;
3572
3573nla_put_failure:
3574 pr_err("not enough space in Netlink message\n");
3575 ret = -EMSGSIZE;
3576
3577out_err:
3578 nlmsg_free(msg);
3579out:
3580 mutex_unlock(&__ip_vs_mutex);
3581
3582 return ret;
3583}
3584
3585
3586static struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3587 {
3588 .cmd = IPVS_CMD_NEW_SERVICE,
3589 .flags = GENL_ADMIN_PERM,
3590 .policy = ip_vs_cmd_policy,
3591 .doit = ip_vs_genl_set_cmd,
3592 },
3593 {
3594 .cmd = IPVS_CMD_SET_SERVICE,
3595 .flags = GENL_ADMIN_PERM,
3596 .policy = ip_vs_cmd_policy,
3597 .doit = ip_vs_genl_set_cmd,
3598 },
3599 {
3600 .cmd = IPVS_CMD_DEL_SERVICE,
3601 .flags = GENL_ADMIN_PERM,
3602 .policy = ip_vs_cmd_policy,
3603 .doit = ip_vs_genl_set_cmd,
3604 },
3605 {
3606 .cmd = IPVS_CMD_GET_SERVICE,
3607 .flags = GENL_ADMIN_PERM,
3608 .doit = ip_vs_genl_get_cmd,
3609 .dumpit = ip_vs_genl_dump_services,
3610 .policy = ip_vs_cmd_policy,
3611 },
3612 {
3613 .cmd = IPVS_CMD_NEW_DEST,
3614 .flags = GENL_ADMIN_PERM,
3615 .policy = ip_vs_cmd_policy,
3616 .doit = ip_vs_genl_set_cmd,
3617 },
3618 {
3619 .cmd = IPVS_CMD_SET_DEST,
3620 .flags = GENL_ADMIN_PERM,
3621 .policy = ip_vs_cmd_policy,
3622 .doit = ip_vs_genl_set_cmd,
3623 },
3624 {
3625 .cmd = IPVS_CMD_DEL_DEST,
3626 .flags = GENL_ADMIN_PERM,
3627 .policy = ip_vs_cmd_policy,
3628 .doit = ip_vs_genl_set_cmd,
3629 },
3630 {
3631 .cmd = IPVS_CMD_GET_DEST,
3632 .flags = GENL_ADMIN_PERM,
3633 .policy = ip_vs_cmd_policy,
3634 .dumpit = ip_vs_genl_dump_dests,
3635 },
3636 {
3637 .cmd = IPVS_CMD_NEW_DAEMON,
3638 .flags = GENL_ADMIN_PERM,
3639 .policy = ip_vs_cmd_policy,
3640 .doit = ip_vs_genl_set_daemon,
3641 },
3642 {
3643 .cmd = IPVS_CMD_DEL_DAEMON,
3644 .flags = GENL_ADMIN_PERM,
3645 .policy = ip_vs_cmd_policy,
3646 .doit = ip_vs_genl_set_daemon,
3647 },
3648 {
3649 .cmd = IPVS_CMD_GET_DAEMON,
3650 .flags = GENL_ADMIN_PERM,
3651 .dumpit = ip_vs_genl_dump_daemons,
3652 },
3653 {
3654 .cmd = IPVS_CMD_SET_CONFIG,
3655 .flags = GENL_ADMIN_PERM,
3656 .policy = ip_vs_cmd_policy,
3657 .doit = ip_vs_genl_set_cmd,
3658 },
3659 {
3660 .cmd = IPVS_CMD_GET_CONFIG,
3661 .flags = GENL_ADMIN_PERM,
3662 .doit = ip_vs_genl_get_cmd,
3663 },
3664 {
3665 .cmd = IPVS_CMD_GET_INFO,
3666 .flags = GENL_ADMIN_PERM,
3667 .doit = ip_vs_genl_get_cmd,
3668 },
3669 {
3670 .cmd = IPVS_CMD_ZERO,
3671 .flags = GENL_ADMIN_PERM,
3672 .policy = ip_vs_cmd_policy,
3673 .doit = ip_vs_genl_set_cmd,
3674 },
3675 {
3676 .cmd = IPVS_CMD_FLUSH,
3677 .flags = GENL_ADMIN_PERM,
3678 .doit = ip_vs_genl_set_cmd,
3679 },
3680};
3681
3682static int __init ip_vs_genl_register(void)
3683{
3684 return genl_register_family_with_ops(&ip_vs_genl_family,
3685 ip_vs_genl_ops, ARRAY_SIZE(ip_vs_genl_ops));
3686}
3687
3688static void ip_vs_genl_unregister(void)
3689{
3690 genl_unregister_family(&ip_vs_genl_family);
3691}
3692
3693
3694
3695
3696
3697
3698#ifdef CONFIG_SYSCTL
3699static int __net_init ip_vs_control_net_init_sysctl(struct net *net)
3700{
3701 int idx;
3702 struct netns_ipvs *ipvs = net_ipvs(net);
3703 struct ctl_table *tbl;
3704
3705 atomic_set(&ipvs->dropentry, 0);
3706 spin_lock_init(&ipvs->dropentry_lock);
3707 spin_lock_init(&ipvs->droppacket_lock);
3708 spin_lock_init(&ipvs->securetcp_lock);
3709
3710 if (!net_eq(net, &init_net)) {
3711 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3712 if (tbl == NULL)
3713 return -ENOMEM;
3714
3715
3716 if (net->user_ns != &init_user_ns)
3717 tbl[0].procname = NULL;
3718 } else
3719 tbl = vs_vars;
3720
3721 idx = 0;
3722 ipvs->sysctl_amemthresh = 1024;
3723 tbl[idx++].data = &ipvs->sysctl_amemthresh;
3724 ipvs->sysctl_am_droprate = 10;
3725 tbl[idx++].data = &ipvs->sysctl_am_droprate;
3726 tbl[idx++].data = &ipvs->sysctl_drop_entry;
3727 tbl[idx++].data = &ipvs->sysctl_drop_packet;
3728#ifdef CONFIG_IP_VS_NFCT
3729 tbl[idx++].data = &ipvs->sysctl_conntrack;
3730#endif
3731 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3732 ipvs->sysctl_snat_reroute = 1;
3733 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3734 ipvs->sysctl_sync_ver = 1;
3735 tbl[idx++].data = &ipvs->sysctl_sync_ver;
3736 ipvs->sysctl_sync_ports = 1;
3737 tbl[idx++].data = &ipvs->sysctl_sync_ports;
3738 tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
3739 ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3740 tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3741 ipvs->sysctl_sync_sock_size = 0;
3742 tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
3743 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3744 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3745 tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
3746 tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
3747 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3748 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3749 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3750 tbl[idx].data = &ipvs->sysctl_sync_threshold;
3751 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3752 ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3753 tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3754 ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3755 tbl[idx++].data = &ipvs->sysctl_sync_retries;
3756 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3757 ipvs->sysctl_pmtu_disc = 1;
3758 tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
3759 tbl[idx++].data = &ipvs->sysctl_backup_only;
3760
3761
3762 ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3763 if (ipvs->sysctl_hdr == NULL) {
3764 if (!net_eq(net, &init_net))
3765 kfree(tbl);
3766 return -ENOMEM;
3767 }
3768 ip_vs_start_estimator(net, &ipvs->tot_stats);
3769 ipvs->sysctl_tbl = tbl;
3770
3771 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3772 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3773
3774 return 0;
3775}
3776
3777static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net)
3778{
3779 struct netns_ipvs *ipvs = net_ipvs(net);
3780
3781 cancel_delayed_work_sync(&ipvs->defense_work);
3782 cancel_work_sync(&ipvs->defense_work.work);
3783 unregister_net_sysctl_table(ipvs->sysctl_hdr);
3784}
3785
3786#else
3787
3788static int __net_init ip_vs_control_net_init_sysctl(struct net *net) { return 0; }
3789static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net) { }
3790
3791#endif
3792
3793static struct notifier_block ip_vs_dst_notifier = {
3794 .notifier_call = ip_vs_dst_event,
3795};
3796
3797int __net_init ip_vs_control_net_init(struct net *net)
3798{
3799 int idx;
3800 struct netns_ipvs *ipvs = net_ipvs(net);
3801
3802
3803 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
3804 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
3805
3806 INIT_LIST_HEAD(&ipvs->dest_trash);
3807 spin_lock_init(&ipvs->dest_trash_lock);
3808 setup_timer(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire,
3809 (unsigned long) net);
3810 atomic_set(&ipvs->ftpsvc_counter, 0);
3811 atomic_set(&ipvs->nullsvc_counter, 0);
3812
3813
3814 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
3815 if (!ipvs->tot_stats.cpustats)
3816 return -ENOMEM;
3817
3818 spin_lock_init(&ipvs->tot_stats.lock);
3819
3820 proc_create("ip_vs", 0, net->proc_net, &ip_vs_info_fops);
3821 proc_create("ip_vs_stats", 0, net->proc_net, &ip_vs_stats_fops);
3822 proc_create("ip_vs_stats_percpu", 0, net->proc_net,
3823 &ip_vs_stats_percpu_fops);
3824
3825 if (ip_vs_control_net_init_sysctl(net))
3826 goto err;
3827
3828 return 0;
3829
3830err:
3831 free_percpu(ipvs->tot_stats.cpustats);
3832 return -ENOMEM;
3833}
3834
3835void __net_exit ip_vs_control_net_cleanup(struct net *net)
3836{
3837 struct netns_ipvs *ipvs = net_ipvs(net);
3838
3839
3840
3841
3842 rcu_barrier();
3843 ip_vs_trash_cleanup(net);
3844 ip_vs_stop_estimator(net, &ipvs->tot_stats);
3845 ip_vs_control_net_cleanup_sysctl(net);
3846 remove_proc_entry("ip_vs_stats_percpu", net->proc_net);
3847 remove_proc_entry("ip_vs_stats", net->proc_net);
3848 remove_proc_entry("ip_vs", net->proc_net);
3849 free_percpu(ipvs->tot_stats.cpustats);
3850}
3851
3852int __init ip_vs_register_nl_ioctl(void)
3853{
3854 int ret;
3855
3856 ret = nf_register_sockopt(&ip_vs_sockopts);
3857 if (ret) {
3858 pr_err("cannot register sockopt.\n");
3859 goto err_sock;
3860 }
3861
3862 ret = ip_vs_genl_register();
3863 if (ret) {
3864 pr_err("cannot register Generic Netlink interface.\n");
3865 goto err_genl;
3866 }
3867 return 0;
3868
3869err_genl:
3870 nf_unregister_sockopt(&ip_vs_sockopts);
3871err_sock:
3872 return ret;
3873}
3874
3875void ip_vs_unregister_nl_ioctl(void)
3876{
3877 ip_vs_genl_unregister();
3878 nf_unregister_sockopt(&ip_vs_sockopts);
3879}
3880
3881int __init ip_vs_control_init(void)
3882{
3883 int idx;
3884 int ret;
3885
3886 EnterFunction(2);
3887
3888
3889 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
3890 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
3891 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3892 }
3893
3894 smp_wmb();
3895
3896 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
3897 if (ret < 0)
3898 return ret;
3899
3900 LeaveFunction(2);
3901 return 0;
3902}
3903
3904
3905void ip_vs_control_cleanup(void)
3906{
3907 EnterFunction(2);
3908 unregister_netdevice_notifier(&ip_vs_dst_notifier);
3909 LeaveFunction(2);
3910}
3911