projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
[deliverable/linux.git] / net / netfilter / ipvs / ip_vs_ctl.c
1 /*
2 * IPVS An implementation of the IP virtual server support for the
3 * LINUX operating system. IPVS is now implemented as a module
4 * over the NetFilter framework. IPVS can be used to build a
5 * high-performance and highly available server based on a
6 * cluster of servers.
7 *
8 * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
9 * Peter Kese <peter.kese@ijs.si>
10 * Julian Anastasov <ja@ssi.bg>
11 *
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
16 *
17 * Changes:
18 *
19 */
20
21 #define KMSG_COMPONENT "IPVS"
22 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
23
24 #include <linux/module.h>
25 #include <linux/init.h>
26 #include <linux/types.h>
27 #include <linux/capability.h>
28 #include <linux/fs.h>
29 #include <linux/sysctl.h>
30 #include <linux/proc_fs.h>
31 #include <linux/workqueue.h>
32 #include <linux/swap.h>
33 #include <linux/seq_file.h>
34 #include <linux/slab.h>
35
36 #include <linux/netfilter.h>
37 #include <linux/netfilter_ipv4.h>
38 #include <linux/mutex.h>
39
40 #include <net/net_namespace.h>
41 #include <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 /* semaphore for IPVS sockopts. And, [gs]etsockopt may sleep. */
56 static DEFINE_MUTEX(__ip_vs_mutex);
57
58 /* lock for service table */
59 static DEFINE_RWLOCK(__ip_vs_svc_lock);
60
61 /* sysctl variables */
62
63 #ifdef CONFIG_IP_VS_DEBUG
64 static int sysctl_ip_vs_debug_level = 0;
65
66 int ip_vs_get_debug_level(void)
67 {
68 return sysctl_ip_vs_debug_level;
69 }
70 #endif
71
72
73 /* Protos */
74 static void __ip_vs_del_service(struct ip_vs_service *svc);
75
76
77 #ifdef CONFIG_IP_VS_IPV6
78 /* Taken from rt6_fill_node() in net/ipv6/route.c, is there a better way? */
79 static int __ip_vs_addr_is_local_v6(struct net *net,
80 const struct in6_addr *addr)
81 {
82 struct rt6_info *rt;
83 struct flowi6 fl6 = {
84 .daddr = *addr,
85 };
86
87 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
88 if (rt && rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
89 return 1;
90
91 return 0;
92 }
93 #endif
94
95 #ifdef CONFIG_SYSCTL
96 /*
97 * update_defense_level is called from keventd and from sysctl,
98 * so it needs to protect itself from softirqs
99 */
100 static void update_defense_level(struct netns_ipvs *ipvs)
101 {
102 struct sysinfo i;
103 static int old_secure_tcp = 0;
104 int availmem;
105 int nomem;
106 int to_change = -1;
107
108 /* we only count free and buffered memory (in pages) */
109 si_meminfo(&i);
110 availmem = i.freeram + i.bufferram;
111 /* however in linux 2.5 the i.bufferram is total page cache size,
112 we need adjust it */
113 /* si_swapinfo(&i); */
114 /* availmem = availmem - (i.totalswap - i.freeswap); */
115
116 nomem = (availmem < ipvs->sysctl_amemthresh);
117
118 local_bh_disable();
119
120 /* drop_entry */
121 spin_lock(&ipvs->dropentry_lock);
122 switch (ipvs->sysctl_drop_entry) {
123 case 0:
124 atomic_set(&ipvs->dropentry, 0);
125 break;
126 case 1:
127 if (nomem) {
128 atomic_set(&ipvs->dropentry, 1);
129 ipvs->sysctl_drop_entry = 2;
130 } else {
131 atomic_set(&ipvs->dropentry, 0);
132 }
133 break;
134 case 2:
135 if (nomem) {
136 atomic_set(&ipvs->dropentry, 1);
137 } else {
138 atomic_set(&ipvs->dropentry, 0);
139 ipvs->sysctl_drop_entry = 1;
140 };
141 break;
142 case 3:
143 atomic_set(&ipvs->dropentry, 1);
144 break;
145 }
146 spin_unlock(&ipvs->dropentry_lock);
147
148 /* drop_packet */
149 spin_lock(&ipvs->droppacket_lock);
150 switch (ipvs->sysctl_drop_packet) {
151 case 0:
152 ipvs->drop_rate = 0;
153 break;
154 case 1:
155 if (nomem) {
156 ipvs->drop_rate = ipvs->drop_counter
157 = ipvs->sysctl_amemthresh /
158 (ipvs->sysctl_amemthresh-availmem);
159 ipvs->sysctl_drop_packet = 2;
160 } else {
161 ipvs->drop_rate = 0;
162 }
163 break;
164 case 2:
165 if (nomem) {
166 ipvs->drop_rate = ipvs->drop_counter
167 = ipvs->sysctl_amemthresh /
168 (ipvs->sysctl_amemthresh-availmem);
169 } else {
170 ipvs->drop_rate = 0;
171 ipvs->sysctl_drop_packet = 1;
172 }
173 break;
174 case 3:
175 ipvs->drop_rate = ipvs->sysctl_am_droprate;
176 break;
177 }
178 spin_unlock(&ipvs->droppacket_lock);
179
180 /* secure_tcp */
181 spin_lock(&ipvs->securetcp_lock);
182 switch (ipvs->sysctl_secure_tcp) {
183 case 0:
184 if (old_secure_tcp >= 2)
185 to_change = 0;
186 break;
187 case 1:
188 if (nomem) {
189 if (old_secure_tcp < 2)
190 to_change = 1;
191 ipvs->sysctl_secure_tcp = 2;
192 } else {
193 if (old_secure_tcp >= 2)
194 to_change = 0;
195 }
196 break;
197 case 2:
198 if (nomem) {
199 if (old_secure_tcp < 2)
200 to_change = 1;
201 } else {
202 if (old_secure_tcp >= 2)
203 to_change = 0;
204 ipvs->sysctl_secure_tcp = 1;
205 }
206 break;
207 case 3:
208 if (old_secure_tcp < 2)
209 to_change = 1;
210 break;
211 }
212 old_secure_tcp = ipvs->sysctl_secure_tcp;
213 if (to_change >= 0)
214 ip_vs_protocol_timeout_change(ipvs,
215 ipvs->sysctl_secure_tcp > 1);
216 spin_unlock(&ipvs->securetcp_lock);
217
218 local_bh_enable();
219 }
220
221
222 /*
223 * Timer for checking the defense
224 */
225 #define DEFENSE_TIMER_PERIOD 1*HZ
226
227 static void defense_work_handler(struct work_struct *work)
228 {
229 struct netns_ipvs *ipvs =
230 container_of(work, struct netns_ipvs, defense_work.work);
231
232 update_defense_level(ipvs);
233 if (atomic_read(&ipvs->dropentry))
234 ip_vs_random_dropentry(ipvs->net);
235 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
236 }
237 #endif
238
239 int
240 ip_vs_use_count_inc(void)
241 {
242 return try_module_get(THIS_MODULE);
243 }
244
245 void
246 ip_vs_use_count_dec(void)
247 {
248 module_put(THIS_MODULE);
249 }
250
251
252 /*
253 * Hash table: for virtual service lookups
254 */
255 #define IP_VS_SVC_TAB_BITS 8
256 #define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
257 #define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)
258
259 /* the service table hashed by <protocol, addr, port> */
260 static struct list_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE];
261 /* the service table hashed by fwmark */
262 static struct list_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE];
263
264
265 /*
266 * Returns hash value for virtual service
267 */
268 static inline unsigned int
269 ip_vs_svc_hashkey(struct net *net, int af, unsigned int proto,
270 const union nf_inet_addr *addr, __be16 port)
271 {
272 register unsigned int porth = ntohs(port);
273 __be32 addr_fold = addr->ip;
274
275 #ifdef CONFIG_IP_VS_IPV6
276 if (af == AF_INET6)
277 addr_fold = addr->ip6[0]^addr->ip6[1]^
278 addr->ip6[2]^addr->ip6[3];
279 #endif
280 addr_fold ^= ((size_t)net>>8);
281
282 return (proto^ntohl(addr_fold)^(porth>>IP_VS_SVC_TAB_BITS)^porth)
283 & IP_VS_SVC_TAB_MASK;
284 }
285
286 /*
287 * Returns hash value of fwmark for virtual service lookup
288 */
289 static inline unsigned int ip_vs_svc_fwm_hashkey(struct net *net, __u32 fwmark)
290 {
291 return (((size_t)net>>8) ^ fwmark) & IP_VS_SVC_TAB_MASK;
292 }
293
294 /*
295 * Hashes a service in the ip_vs_svc_table by <netns,proto,addr,port>
296 * or in the ip_vs_svc_fwm_table by fwmark.
297 * Should be called with locked tables.
298 */
299 static int ip_vs_svc_hash(struct ip_vs_service *svc)
300 {
301 unsigned int hash;
302
303 if (svc->flags & IP_VS_SVC_F_HASHED) {
304 pr_err("%s(): request for already hashed, called from %pF\n",
305 __func__, __builtin_return_address(0));
306 return 0;
307 }
308
309 if (svc->fwmark == 0) {
310 /*
311 * Hash it by <netns,protocol,addr,port> in ip_vs_svc_table
312 */
313 hash = ip_vs_svc_hashkey(svc->net, svc->af, svc->protocol,
314 &svc->addr, svc->port);
315 list_add(&svc->s_list, &ip_vs_svc_table[hash]);
316 } else {
317 /*
318 * Hash it by fwmark in svc_fwm_table
319 */
320 hash = ip_vs_svc_fwm_hashkey(svc->net, svc->fwmark);
321 list_add(&svc->f_list, &ip_vs_svc_fwm_table[hash]);
322 }
323
324 svc->flags |= IP_VS_SVC_F_HASHED;
325 /* increase its refcnt because it is referenced by the svc table */
326 atomic_inc(&svc->refcnt);
327 return 1;
328 }
329
330
331 /*
332 * Unhashes a service from svc_table / svc_fwm_table.
333 * Should be called with locked tables.
334 */
335 static int ip_vs_svc_unhash(struct ip_vs_service *svc)
336 {
337 if (!(svc->flags & IP_VS_SVC_F_HASHED)) {
338 pr_err("%s(): request for unhash flagged, called from %pF\n",
339 __func__, __builtin_return_address(0));
340 return 0;
341 }
342
343 if (svc->fwmark == 0) {
344 /* Remove it from the svc_table table */
345 list_del(&svc->s_list);
346 } else {
347 /* Remove it from the svc_fwm_table table */
348 list_del(&svc->f_list);
349 }
350
351 svc->flags &= ~IP_VS_SVC_F_HASHED;
352 atomic_dec(&svc->refcnt);
353 return 1;
354 }
355
356
357 /*
358 * Get service by {netns, proto,addr,port} in the service table.
359 */
360 static inline struct ip_vs_service *
361 __ip_vs_service_find(struct net *net, int af, __u16 protocol,
362 const union nf_inet_addr *vaddr, __be16 vport)
363 {
364 unsigned int hash;
365 struct ip_vs_service *svc;
366
367 /* Check for "full" addressed entries */
368 hash = ip_vs_svc_hashkey(net, af, protocol, vaddr, vport);
369
370 list_for_each_entry(svc, &ip_vs_svc_table[hash], s_list){
371 if ((svc->af == af)
372 && ip_vs_addr_equal(af, &svc->addr, vaddr)
373 && (svc->port == vport)
374 && (svc->protocol == protocol)
375 && net_eq(svc->net, net)) {
376 /* HIT */
377 return svc;
378 }
379 }
380
381 return NULL;
382 }
383
384
385 /*
386 * Get service by {fwmark} in the service table.
387 */
388 static inline struct ip_vs_service *
389 __ip_vs_svc_fwm_find(struct net *net, int af, __u32 fwmark)
390 {
391 unsigned int hash;
392 struct ip_vs_service *svc;
393
394 /* Check for fwmark addressed entries */
395 hash = ip_vs_svc_fwm_hashkey(net, fwmark);
396
397 list_for_each_entry(svc, &ip_vs_svc_fwm_table[hash], f_list) {
398 if (svc->fwmark == fwmark && svc->af == af
399 && net_eq(svc->net, net)) {
400 /* HIT */
401 return svc;
402 }
403 }
404
405 return NULL;
406 }
407
408 struct ip_vs_service *
409 ip_vs_service_get(struct net *net, int af, __u32 fwmark, __u16 protocol,
410 const union nf_inet_addr *vaddr, __be16 vport)
411 {
412 struct ip_vs_service *svc;
413 struct netns_ipvs *ipvs = net_ipvs(net);
414
415 read_lock(&__ip_vs_svc_lock);
416
417 /*
418 * Check the table hashed by fwmark first
419 */
420 if (fwmark) {
421 svc = __ip_vs_svc_fwm_find(net, af, fwmark);
422 if (svc)
423 goto out;
424 }
425
426 /*
427 * Check the table hashed by <protocol,addr,port>
428 * for "full" addressed entries
429 */
430 svc = __ip_vs_service_find(net, af, protocol, vaddr, vport);
431
432 if (svc == NULL
433 && protocol == IPPROTO_TCP
434 && atomic_read(&ipvs->ftpsvc_counter)
435 && (vport == FTPDATA || ntohs(vport) >= PROT_SOCK)) {
436 /*
437 * Check if ftp service entry exists, the packet
438 * might belong to FTP data connections.
439 */
440 svc = __ip_vs_service_find(net, af, protocol, vaddr, FTPPORT);
441 }
442
443 if (svc == NULL
444 && atomic_read(&ipvs->nullsvc_counter)) {
445 /*
446 * Check if the catch-all port (port zero) exists
447 */
448 svc = __ip_vs_service_find(net, af, protocol, vaddr, 0);
449 }
450
451 out:
452 if (svc)
453 atomic_inc(&svc->usecnt);
454 read_unlock(&__ip_vs_svc_lock);
455
456 IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
457 fwmark, ip_vs_proto_name(protocol),
458 IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
459 svc ? "hit" : "not hit");
460
461 return svc;
462 }
463
464
465 static inline void
466 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
467 {
468 atomic_inc(&svc->refcnt);
469 dest->svc = svc;
470 }
471
472 static void
473 __ip_vs_unbind_svc(struct ip_vs_dest *dest)
474 {
475 struct ip_vs_service *svc = dest->svc;
476
477 dest->svc = NULL;
478 if (atomic_dec_and_test(&svc->refcnt)) {
479 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
480 svc->fwmark,
481 IP_VS_DBG_ADDR(svc->af, &svc->addr),
482 ntohs(svc->port), atomic_read(&svc->usecnt));
483 free_percpu(svc->stats.cpustats);
484 kfree(svc);
485 }
486 }
487
488
489 /*
490 * Returns hash value for real service
491 */
492 static 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 /*
510 * Hashes ip_vs_dest in rs_table by <proto,addr,port>.
511 * should be called with locked tables.
512 */
513 static int ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
514 {
515 unsigned int hash;
516
517 if (!list_empty(&dest->d_list)) {
518 return 0;
519 }
520
521 /*
522 * Hash by proto,addr,port,
523 * which are the parameters of the real service.
524 */
525 hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
526
527 list_add(&dest->d_list, &ipvs->rs_table[hash]);
528
529 return 1;
530 }
531
532 /*
533 * UNhashes ip_vs_dest from rs_table.
534 * should be called with locked tables.
535 */
536 static int ip_vs_rs_unhash(struct ip_vs_dest *dest)
537 {
538 /*
539 * Remove it from the rs_table table.
540 */
541 if (!list_empty(&dest->d_list)) {
542 list_del(&dest->d_list);
543 INIT_LIST_HEAD(&dest->d_list);
544 }
545
546 return 1;
547 }
548
549 /*
550 * Lookup real service by <proto,addr,port> in the real service table.
551 */
552 struct ip_vs_dest *
553 ip_vs_lookup_real_service(struct net *net, int af, __u16 protocol,
554 const union nf_inet_addr *daddr,
555 __be16 dport)
556 {
557 struct netns_ipvs *ipvs = net_ipvs(net);
558 unsigned int hash;
559 struct ip_vs_dest *dest;
560
561 /*
562 * Check for "full" addressed entries
563 * Return the first found entry
564 */
565 hash = ip_vs_rs_hashkey(af, daddr, dport);
566
567 read_lock(&ipvs->rs_lock);
568 list_for_each_entry(dest, &ipvs->rs_table[hash], d_list) {
569 if ((dest->af == af)
570 && ip_vs_addr_equal(af, &dest->addr, daddr)
571 && (dest->port == dport)
572 && ((dest->protocol == protocol) ||
573 dest->vfwmark)) {
574 /* HIT */
575 read_unlock(&ipvs->rs_lock);
576 return dest;
577 }
578 }
579 read_unlock(&ipvs->rs_lock);
580
581 return NULL;
582 }
583
584 /*
585 * Lookup destination by {addr,port} in the given service
586 */
587 static struct ip_vs_dest *
588 ip_vs_lookup_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
589 __be16 dport)
590 {
591 struct ip_vs_dest *dest;
592
593 /*
594 * Find the destination for the given service
595 */
596 list_for_each_entry(dest, &svc->destinations, n_list) {
597 if ((dest->af == svc->af)
598 && ip_vs_addr_equal(svc->af, &dest->addr, daddr)
599 && (dest->port == dport)) {
600 /* HIT */
601 return dest;
602 }
603 }
604
605 return NULL;
606 }
607
608 /*
609 * Find destination by {daddr,dport,vaddr,protocol}
610 * Cretaed to be used in ip_vs_process_message() in
611 * the backup synchronization daemon. It finds the
612 * destination to be bound to the received connection
613 * on the backup.
614 *
615 * ip_vs_lookup_real_service() looked promissing, but
616 * seems not working as expected.
617 */
618 struct ip_vs_dest *ip_vs_find_dest(struct net *net, int af,
619 const union nf_inet_addr *daddr,
620 __be16 dport,
621 const union nf_inet_addr *vaddr,
622 __be16 vport, __u16 protocol, __u32 fwmark,
623 __u32 flags)
624 {
625 struct ip_vs_dest *dest;
626 struct ip_vs_service *svc;
627 __be16 port = dport;
628
629 svc = ip_vs_service_get(net, af, fwmark, protocol, vaddr, vport);
630 if (!svc)
631 return NULL;
632 if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
633 port = 0;
634 dest = ip_vs_lookup_dest(svc, daddr, port);
635 if (!dest)
636 dest = ip_vs_lookup_dest(svc, daddr, port ^ dport);
637 if (dest)
638 atomic_inc(&dest->refcnt);
639 ip_vs_service_put(svc);
640 return dest;
641 }
642
643 /*
644 * Lookup dest by {svc,addr,port} in the destination trash.
645 * The destination trash is used to hold the destinations that are removed
646 * from the service table but are still referenced by some conn entries.
647 * The reason to add the destination trash is when the dest is temporary
648 * down (either by administrator or by monitor program), the dest can be
649 * picked back from the trash, the remaining connections to the dest can
650 * continue, and the counting information of the dest is also useful for
651 * scheduling.
652 */
653 static struct ip_vs_dest *
654 ip_vs_trash_get_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
655 __be16 dport)
656 {
657 struct ip_vs_dest *dest, *nxt;
658 struct netns_ipvs *ipvs = net_ipvs(svc->net);
659
660 /*
661 * Find the destination in trash
662 */
663 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, n_list) {
664 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
665 "dest->refcnt=%d\n",
666 dest->vfwmark,
667 IP_VS_DBG_ADDR(svc->af, &dest->addr),
668 ntohs(dest->port),
669 atomic_read(&dest->refcnt));
670 if (dest->af == svc->af &&
671 ip_vs_addr_equal(svc->af, &dest->addr, daddr) &&
672 dest->port == dport &&
673 dest->vfwmark == svc->fwmark &&
674 dest->protocol == svc->protocol &&
675 (svc->fwmark ||
676 (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
677 dest->vport == svc->port))) {
678 /* HIT */
679 return dest;
680 }
681
682 /*
683 * Try to purge the destination from trash if not referenced
684 */
685 if (atomic_read(&dest->refcnt) == 1) {
686 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u "
687 "from trash\n",
688 dest->vfwmark,
689 IP_VS_DBG_ADDR(svc->af, &dest->addr),
690 ntohs(dest->port));
691 list_del(&dest->n_list);
692 ip_vs_dst_reset(dest);
693 __ip_vs_unbind_svc(dest);
694 free_percpu(dest->stats.cpustats);
695 kfree(dest);
696 }
697 }
698
699 return NULL;
700 }
701
702
703 /*
704 * Clean up all the destinations in the trash
705 * Called by the ip_vs_control_cleanup()
706 *
707 * When the ip_vs_control_clearup is activated by ipvs module exit,
708 * the service tables must have been flushed and all the connections
709 * are expired, and the refcnt of each destination in the trash must
710 * be 1, so we simply release them here.
711 */
712 static void ip_vs_trash_cleanup(struct net *net)
713 {
714 struct ip_vs_dest *dest, *nxt;
715 struct netns_ipvs *ipvs = net_ipvs(net);
716
717 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, n_list) {
718 list_del(&dest->n_list);
719 ip_vs_dst_reset(dest);
720 __ip_vs_unbind_svc(dest);
721 free_percpu(dest->stats.cpustats);
722 kfree(dest);
723 }
724 }
725
726 static void
727 ip_vs_copy_stats(struct ip_vs_stats_user *dst, struct ip_vs_stats *src)
728 {
729 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->ustats.c - src->ustats0.c
730
731 spin_lock_bh(&src->lock);
732
733 IP_VS_SHOW_STATS_COUNTER(conns);
734 IP_VS_SHOW_STATS_COUNTER(inpkts);
735 IP_VS_SHOW_STATS_COUNTER(outpkts);
736 IP_VS_SHOW_STATS_COUNTER(inbytes);
737 IP_VS_SHOW_STATS_COUNTER(outbytes);
738
739 ip_vs_read_estimator(dst, src);
740
741 spin_unlock_bh(&src->lock);
742 }
743
744 static void
745 ip_vs_zero_stats(struct ip_vs_stats *stats)
746 {
747 spin_lock_bh(&stats->lock);
748
749 /* get current counters as zero point, rates are zeroed */
750
751 #define IP_VS_ZERO_STATS_COUNTER(c) stats->ustats0.c = stats->ustats.c
752
753 IP_VS_ZERO_STATS_COUNTER(conns);
754 IP_VS_ZERO_STATS_COUNTER(inpkts);
755 IP_VS_ZERO_STATS_COUNTER(outpkts);
756 IP_VS_ZERO_STATS_COUNTER(inbytes);
757 IP_VS_ZERO_STATS_COUNTER(outbytes);
758
759 ip_vs_zero_estimator(stats);
760
761 spin_unlock_bh(&stats->lock);
762 }
763
764 /*
765 * Update a destination in the given service
766 */
767 static void
768 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
769 struct ip_vs_dest_user_kern *udest, int add)
770 {
771 struct netns_ipvs *ipvs = net_ipvs(svc->net);
772 int conn_flags;
773
774 /* set the weight and the flags */
775 atomic_set(&dest->weight, udest->weight);
776 conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
777 conn_flags |= IP_VS_CONN_F_INACTIVE;
778
779 /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
780 if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
781 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
782 } else {
783 /*
784 * Put the real service in rs_table if not present.
785 * For now only for NAT!
786 */
787 write_lock_bh(&ipvs->rs_lock);
788 ip_vs_rs_hash(ipvs, dest);
789 write_unlock_bh(&ipvs->rs_lock);
790 }
791 atomic_set(&dest->conn_flags, conn_flags);
792
793 /* bind the service */
794 if (!dest->svc) {
795 __ip_vs_bind_svc(dest, svc);
796 } else {
797 if (dest->svc != svc) {
798 __ip_vs_unbind_svc(dest);
799 ip_vs_zero_stats(&dest->stats);
800 __ip_vs_bind_svc(dest, svc);
801 }
802 }
803
804 /* set the dest status flags */
805 dest->flags |= IP_VS_DEST_F_AVAILABLE;
806
807 if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
808 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
809 dest->u_threshold = udest->u_threshold;
810 dest->l_threshold = udest->l_threshold;
811
812 spin_lock_bh(&dest->dst_lock);
813 ip_vs_dst_reset(dest);
814 spin_unlock_bh(&dest->dst_lock);
815
816 if (add)
817 ip_vs_start_estimator(svc->net, &dest->stats);
818
819 write_lock_bh(&__ip_vs_svc_lock);
820
821 /* Wait until all other svc users go away */
822 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
823
824 if (add) {
825 list_add(&dest->n_list, &svc->destinations);
826 svc->num_dests++;
827 }
828
829 /* call the update_service, because server weight may be changed */
830 if (svc->scheduler->update_service)
831 svc->scheduler->update_service(svc);
832
833 write_unlock_bh(&__ip_vs_svc_lock);
834 }
835
836
837 /*
838 * Create a destination for the given service
839 */
840 static int
841 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
842 struct ip_vs_dest **dest_p)
843 {
844 struct ip_vs_dest *dest;
845 unsigned int atype;
846
847 EnterFunction(2);
848
849 #ifdef CONFIG_IP_VS_IPV6
850 if (svc->af == AF_INET6) {
851 atype = ipv6_addr_type(&udest->addr.in6);
852 if ((!(atype & IPV6_ADDR_UNICAST) ||
853 atype & IPV6_ADDR_LINKLOCAL) &&
854 !__ip_vs_addr_is_local_v6(svc->net, &udest->addr.in6))
855 return -EINVAL;
856 } else
857 #endif
858 {
859 atype = inet_addr_type(svc->net, udest->addr.ip);
860 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
861 return -EINVAL;
862 }
863
864 dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
865 if (dest == NULL)
866 return -ENOMEM;
867
868 dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
869 if (!dest->stats.cpustats)
870 goto err_alloc;
871
872 dest->af = svc->af;
873 dest->protocol = svc->protocol;
874 dest->vaddr = svc->addr;
875 dest->vport = svc->port;
876 dest->vfwmark = svc->fwmark;
877 ip_vs_addr_copy(svc->af, &dest->addr, &udest->addr);
878 dest->port = udest->port;
879
880 atomic_set(&dest->activeconns, 0);
881 atomic_set(&dest->inactconns, 0);
882 atomic_set(&dest->persistconns, 0);
883 atomic_set(&dest->refcnt, 1);
884
885 INIT_LIST_HEAD(&dest->d_list);
886 spin_lock_init(&dest->dst_lock);
887 spin_lock_init(&dest->stats.lock);
888 __ip_vs_update_dest(svc, dest, udest, 1);
889
890 *dest_p = dest;
891
892 LeaveFunction(2);
893 return 0;
894
895 err_alloc:
896 kfree(dest);
897 return -ENOMEM;
898 }
899
900
901 /*
902 * Add a destination into an existing service
903 */
904 static int
905 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
906 {
907 struct ip_vs_dest *dest;
908 union nf_inet_addr daddr;
909 __be16 dport = udest->port;
910 int ret;
911
912 EnterFunction(2);
913
914 if (udest->weight < 0) {
915 pr_err("%s(): server weight less than zero\n", __func__);
916 return -ERANGE;
917 }
918
919 if (udest->l_threshold > udest->u_threshold) {
920 pr_err("%s(): lower threshold is higher than upper threshold\n",
921 __func__);
922 return -ERANGE;
923 }
924
925 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
926
927 /*
928 * Check if the dest already exists in the list
929 */
930 dest = ip_vs_lookup_dest(svc, &daddr, dport);
931
932 if (dest != NULL) {
933 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
934 return -EEXIST;
935 }
936
937 /*
938 * Check if the dest already exists in the trash and
939 * is from the same service
940 */
941 dest = ip_vs_trash_get_dest(svc, &daddr, dport);
942
943 if (dest != NULL) {
944 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
945 "dest->refcnt=%d, service %u/%s:%u\n",
946 IP_VS_DBG_ADDR(svc->af, &daddr), ntohs(dport),
947 atomic_read(&dest->refcnt),
948 dest->vfwmark,
949 IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
950 ntohs(dest->vport));
951
952 /*
953 * Get the destination from the trash
954 */
955 list_del(&dest->n_list);
956
957 __ip_vs_update_dest(svc, dest, udest, 1);
958 ret = 0;
959 } else {
960 /*
961 * Allocate and initialize the dest structure
962 */
963 ret = ip_vs_new_dest(svc, udest, &dest);
964 }
965 LeaveFunction(2);
966
967 return ret;
968 }
969
970
971 /*
972 * Edit a destination in the given service
973 */
974 static int
975 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
976 {
977 struct ip_vs_dest *dest;
978 union nf_inet_addr daddr;
979 __be16 dport = udest->port;
980
981 EnterFunction(2);
982
983 if (udest->weight < 0) {
984 pr_err("%s(): server weight less than zero\n", __func__);
985 return -ERANGE;
986 }
987
988 if (udest->l_threshold > udest->u_threshold) {
989 pr_err("%s(): lower threshold is higher than upper threshold\n",
990 __func__);
991 return -ERANGE;
992 }
993
994 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
995
996 /*
997 * Lookup the destination list
998 */
999 dest = ip_vs_lookup_dest(svc, &daddr, dport);
1000
1001 if (dest == NULL) {
1002 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1003 return -ENOENT;
1004 }
1005
1006 __ip_vs_update_dest(svc, dest, udest, 0);
1007 LeaveFunction(2);
1008
1009 return 0;
1010 }
1011
1012
1013 /*
1014 * Delete a destination (must be already unlinked from the service)
1015 */
1016 static void __ip_vs_del_dest(struct net *net, struct ip_vs_dest *dest)
1017 {
1018 struct netns_ipvs *ipvs = net_ipvs(net);
1019
1020 ip_vs_stop_estimator(net, &dest->stats);
1021
1022 /*
1023 * Remove it from the d-linked list with the real services.
1024 */
1025 write_lock_bh(&ipvs->rs_lock);
1026 ip_vs_rs_unhash(dest);
1027 write_unlock_bh(&ipvs->rs_lock);
1028
1029 /*
1030 * Decrease the refcnt of the dest, and free the dest
1031 * if nobody refers to it (refcnt=0). Otherwise, throw
1032 * the destination into the trash.
1033 */
1034 if (atomic_dec_and_test(&dest->refcnt)) {
1035 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u\n",
1036 dest->vfwmark,
1037 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1038 ntohs(dest->port));
1039 ip_vs_dst_reset(dest);
1040 /* simply decrease svc->refcnt here, let the caller check
1041 and release the service if nobody refers to it.
1042 Only user context can release destination and service,
1043 and only one user context can update virtual service at a
1044 time, so the operation here is OK */
1045 atomic_dec(&dest->svc->refcnt);
1046 free_percpu(dest->stats.cpustats);
1047 kfree(dest);
1048 } else {
1049 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, "
1050 "dest->refcnt=%d\n",
1051 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1052 ntohs(dest->port),
1053 atomic_read(&dest->refcnt));
1054 list_add(&dest->n_list, &ipvs->dest_trash);
1055 atomic_inc(&dest->refcnt);
1056 }
1057 }
1058
1059
1060 /*
1061 * Unlink a destination from the given service
1062 */
1063 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1064 struct ip_vs_dest *dest,
1065 int svcupd)
1066 {
1067 dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1068
1069 /*
1070 * Remove it from the d-linked destination list.
1071 */
1072 list_del(&dest->n_list);
1073 svc->num_dests--;
1074
1075 /*
1076 * Call the update_service function of its scheduler
1077 */
1078 if (svcupd && svc->scheduler->update_service)
1079 svc->scheduler->update_service(svc);
1080 }
1081
1082
1083 /*
1084 * Delete a destination server in the given service
1085 */
1086 static int
1087 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1088 {
1089 struct ip_vs_dest *dest;
1090 __be16 dport = udest->port;
1091
1092 EnterFunction(2);
1093
1094 dest = ip_vs_lookup_dest(svc, &udest->addr, dport);
1095
1096 if (dest == NULL) {
1097 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1098 return -ENOENT;
1099 }
1100
1101 write_lock_bh(&__ip_vs_svc_lock);
1102
1103 /*
1104 * Wait until all other svc users go away.
1105 */
1106 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1107
1108 /*
1109 * Unlink dest from the service
1110 */
1111 __ip_vs_unlink_dest(svc, dest, 1);
1112
1113 write_unlock_bh(&__ip_vs_svc_lock);
1114
1115 /*
1116 * Delete the destination
1117 */
1118 __ip_vs_del_dest(svc->net, dest);
1119
1120 LeaveFunction(2);
1121
1122 return 0;
1123 }
1124
1125
1126 /*
1127 * Add a service into the service hash table
1128 */
1129 static int
1130 ip_vs_add_service(struct net *net, struct ip_vs_service_user_kern *u,
1131 struct ip_vs_service **svc_p)
1132 {
1133 int ret = 0;
1134 struct ip_vs_scheduler *sched = NULL;
1135 struct ip_vs_pe *pe = NULL;
1136 struct ip_vs_service *svc = NULL;
1137 struct netns_ipvs *ipvs = net_ipvs(net);
1138
1139 /* increase the module use count */
1140 ip_vs_use_count_inc();
1141
1142 /* Lookup the scheduler by 'u->sched_name' */
1143 sched = ip_vs_scheduler_get(u->sched_name);
1144 if (sched == NULL) {
1145 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1146 ret = -ENOENT;
1147 goto out_err;
1148 }
1149
1150 if (u->pe_name && *u->pe_name) {
1151 pe = ip_vs_pe_getbyname(u->pe_name);
1152 if (pe == NULL) {
1153 pr_info("persistence engine module ip_vs_pe_%s "
1154 "not found\n", u->pe_name);
1155 ret = -ENOENT;
1156 goto out_err;
1157 }
1158 }
1159
1160 #ifdef CONFIG_IP_VS_IPV6
1161 if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1162 ret = -EINVAL;
1163 goto out_err;
1164 }
1165 #endif
1166
1167 svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1168 if (svc == NULL) {
1169 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1170 ret = -ENOMEM;
1171 goto out_err;
1172 }
1173 svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1174 if (!svc->stats.cpustats)
1175 goto out_err;
1176
1177 /* I'm the first user of the service */
1178 atomic_set(&svc->usecnt, 0);
1179 atomic_set(&svc->refcnt, 0);
1180
1181 svc->af = u->af;
1182 svc->protocol = u->protocol;
1183 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1184 svc->port = u->port;
1185 svc->fwmark = u->fwmark;
1186 svc->flags = u->flags;
1187 svc->timeout = u->timeout * HZ;
1188 svc->netmask = u->netmask;
1189 svc->net = net;
1190
1191 INIT_LIST_HEAD(&svc->destinations);
1192 rwlock_init(&svc->sched_lock);
1193 spin_lock_init(&svc->stats.lock);
1194
1195 /* Bind the scheduler */
1196 ret = ip_vs_bind_scheduler(svc, sched);
1197 if (ret)
1198 goto out_err;
1199 sched = NULL;
1200
1201 /* Bind the ct retriever */
1202 ip_vs_bind_pe(svc, pe);
1203 pe = NULL;
1204
1205 /* Update the virtual service counters */
1206 if (svc->port == FTPPORT)
1207 atomic_inc(&ipvs->ftpsvc_counter);
1208 else if (svc->port == 0)
1209 atomic_inc(&ipvs->nullsvc_counter);
1210
1211 ip_vs_start_estimator(net, &svc->stats);
1212
1213 /* Count only IPv4 services for old get/setsockopt interface */
1214 if (svc->af == AF_INET)
1215 ipvs->num_services++;
1216
1217 /* Hash the service into the service table */
1218 write_lock_bh(&__ip_vs_svc_lock);
1219 ip_vs_svc_hash(svc);
1220 write_unlock_bh(&__ip_vs_svc_lock);
1221
1222 *svc_p = svc;
1223 /* Now there is a service - full throttle */
1224 ipvs->enable = 1;
1225 return 0;
1226
1227
1228 out_err:
1229 if (svc != NULL) {
1230 ip_vs_unbind_scheduler(svc);
1231 if (svc->inc) {
1232 local_bh_disable();
1233 ip_vs_app_inc_put(svc->inc);
1234 local_bh_enable();
1235 }
1236 if (svc->stats.cpustats)
1237 free_percpu(svc->stats.cpustats);
1238 kfree(svc);
1239 }
1240 ip_vs_scheduler_put(sched);
1241 ip_vs_pe_put(pe);
1242
1243 /* decrease the module use count */
1244 ip_vs_use_count_dec();
1245
1246 return ret;
1247 }
1248
1249
1250 /*
1251 * Edit a service and bind it with a new scheduler
1252 */
1253 static int
1254 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1255 {
1256 struct ip_vs_scheduler *sched, *old_sched;
1257 struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1258 int ret = 0;
1259
1260 /*
1261 * Lookup the scheduler, by 'u->sched_name'
1262 */
1263 sched = ip_vs_scheduler_get(u->sched_name);
1264 if (sched == NULL) {
1265 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1266 return -ENOENT;
1267 }
1268 old_sched = sched;
1269
1270 if (u->pe_name && *u->pe_name) {
1271 pe = ip_vs_pe_getbyname(u->pe_name);
1272 if (pe == NULL) {
1273 pr_info("persistence engine module ip_vs_pe_%s "
1274 "not found\n", u->pe_name);
1275 ret = -ENOENT;
1276 goto out;
1277 }
1278 old_pe = pe;
1279 }
1280
1281 #ifdef CONFIG_IP_VS_IPV6
1282 if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1283 ret = -EINVAL;
1284 goto out;
1285 }
1286 #endif
1287
1288 write_lock_bh(&__ip_vs_svc_lock);
1289
1290 /*
1291 * Wait until all other svc users go away.
1292 */
1293 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1294
1295 /*
1296 * Set the flags and timeout value
1297 */
1298 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1299 svc->timeout = u->timeout * HZ;
1300 svc->netmask = u->netmask;
1301
1302 old_sched = svc->scheduler;
1303 if (sched != old_sched) {
1304 /*
1305 * Unbind the old scheduler
1306 */
1307 if ((ret = ip_vs_unbind_scheduler(svc))) {
1308 old_sched = sched;
1309 goto out_unlock;
1310 }
1311
1312 /*
1313 * Bind the new scheduler
1314 */
1315 if ((ret = ip_vs_bind_scheduler(svc, sched))) {
1316 /*
1317 * If ip_vs_bind_scheduler fails, restore the old
1318 * scheduler.
1319 * The main reason of failure is out of memory.
1320 *
1321 * The question is if the old scheduler can be
1322 * restored all the time. TODO: if it cannot be
1323 * restored some time, we must delete the service,
1324 * otherwise the system may crash.
1325 */
1326 ip_vs_bind_scheduler(svc, old_sched);
1327 old_sched = sched;
1328 goto out_unlock;
1329 }
1330 }
1331
1332 old_pe = svc->pe;
1333 if (pe != old_pe) {
1334 ip_vs_unbind_pe(svc);
1335 ip_vs_bind_pe(svc, pe);
1336 }
1337
1338 out_unlock:
1339 write_unlock_bh(&__ip_vs_svc_lock);
1340 out:
1341 ip_vs_scheduler_put(old_sched);
1342 ip_vs_pe_put(old_pe);
1343 return ret;
1344 }
1345
1346
1347 /*
1348 * Delete a service from the service list
1349 * - The service must be unlinked, unlocked and not referenced!
1350 * - We are called under _bh lock
1351 */
1352 static void __ip_vs_del_service(struct ip_vs_service *svc)
1353 {
1354 struct ip_vs_dest *dest, *nxt;
1355 struct ip_vs_scheduler *old_sched;
1356 struct ip_vs_pe *old_pe;
1357 struct netns_ipvs *ipvs = net_ipvs(svc->net);
1358
1359 pr_info("%s: enter\n", __func__);
1360
1361 /* Count only IPv4 services for old get/setsockopt interface */
1362 if (svc->af == AF_INET)
1363 ipvs->num_services--;
1364
1365 ip_vs_stop_estimator(svc->net, &svc->stats);
1366
1367 /* Unbind scheduler */
1368 old_sched = svc->scheduler;
1369 ip_vs_unbind_scheduler(svc);
1370 ip_vs_scheduler_put(old_sched);
1371
1372 /* Unbind persistence engine */
1373 old_pe = svc->pe;
1374 ip_vs_unbind_pe(svc);
1375 ip_vs_pe_put(old_pe);
1376
1377 /* Unbind app inc */
1378 if (svc->inc) {
1379 ip_vs_app_inc_put(svc->inc);
1380 svc->inc = NULL;
1381 }
1382
1383 /*
1384 * Unlink the whole destination list
1385 */
1386 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1387 __ip_vs_unlink_dest(svc, dest, 0);
1388 __ip_vs_del_dest(svc->net, dest);
1389 }
1390
1391 /*
1392 * Update the virtual service counters
1393 */
1394 if (svc->port == FTPPORT)
1395 atomic_dec(&ipvs->ftpsvc_counter);
1396 else if (svc->port == 0)
1397 atomic_dec(&ipvs->nullsvc_counter);
1398
1399 /*
1400 * Free the service if nobody refers to it
1401 */
1402 if (atomic_read(&svc->refcnt) == 0) {
1403 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
1404 svc->fwmark,
1405 IP_VS_DBG_ADDR(svc->af, &svc->addr),
1406 ntohs(svc->port), atomic_read(&svc->usecnt));
1407 free_percpu(svc->stats.cpustats);
1408 kfree(svc);
1409 }
1410
1411 /* decrease the module use count */
1412 ip_vs_use_count_dec();
1413 }
1414
1415 /*
1416 * Unlink a service from list and try to delete it if its refcnt reached 0
1417 */
1418 static void ip_vs_unlink_service(struct ip_vs_service *svc)
1419 {
1420 /*
1421 * Unhash it from the service table
1422 */
1423 write_lock_bh(&__ip_vs_svc_lock);
1424
1425 ip_vs_svc_unhash(svc);
1426
1427 /*
1428 * Wait until all the svc users go away.
1429 */
1430 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1431
1432 __ip_vs_del_service(svc);
1433
1434 write_unlock_bh(&__ip_vs_svc_lock);
1435 }
1436
1437 /*
1438 * Delete a service from the service list
1439 */
1440 static int ip_vs_del_service(struct ip_vs_service *svc)
1441 {
1442 if (svc == NULL)
1443 return -EEXIST;
1444 ip_vs_unlink_service(svc);
1445
1446 return 0;
1447 }
1448
1449
1450 /*
1451 * Flush all the virtual services
1452 */
1453 static int ip_vs_flush(struct net *net)
1454 {
1455 int idx;
1456 struct ip_vs_service *svc, *nxt;
1457
1458 /*
1459 * Flush the service table hashed by <netns,protocol,addr,port>
1460 */
1461 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1462 list_for_each_entry_safe(svc, nxt, &ip_vs_svc_table[idx],
1463 s_list) {
1464 if (net_eq(svc->net, net))
1465 ip_vs_unlink_service(svc);
1466 }
1467 }
1468
1469 /*
1470 * Flush the service table hashed by fwmark
1471 */
1472 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1473 list_for_each_entry_safe(svc, nxt,
1474 &ip_vs_svc_fwm_table[idx], f_list) {
1475 if (net_eq(svc->net, net))
1476 ip_vs_unlink_service(svc);
1477 }
1478 }
1479
1480 return 0;
1481 }
1482
1483 /*
1484 * Delete service by {netns} in the service table.
1485 * Called by __ip_vs_cleanup()
1486 */
1487 void ip_vs_service_net_cleanup(struct net *net)
1488 {
1489 EnterFunction(2);
1490 /* Check for "full" addressed entries */
1491 mutex_lock(&__ip_vs_mutex);
1492 ip_vs_flush(net);
1493 mutex_unlock(&__ip_vs_mutex);
1494 LeaveFunction(2);
1495 }
1496 /*
1497 * Release dst hold by dst_cache
1498 */
1499 static inline void
1500 __ip_vs_dev_reset(struct ip_vs_dest *dest, struct net_device *dev)
1501 {
1502 spin_lock_bh(&dest->dst_lock);
1503 if (dest->dst_cache && dest->dst_cache->dev == dev) {
1504 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1505 dev->name,
1506 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1507 ntohs(dest->port),
1508 atomic_read(&dest->refcnt));
1509 ip_vs_dst_reset(dest);
1510 }
1511 spin_unlock_bh(&dest->dst_lock);
1512
1513 }
1514 /*
1515 * Netdev event receiver
1516 * Currently only NETDEV_UNREGISTER is handled, i.e. if we hold a reference to
1517 * a device that is "unregister" it must be released.
1518 */
1519 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1520 void *ptr)
1521 {
1522 struct net_device *dev = ptr;
1523 struct net *net = dev_net(dev);
1524 struct ip_vs_service *svc;
1525 struct ip_vs_dest *dest;
1526 unsigned int idx;
1527
1528 if (event != NETDEV_UNREGISTER)
1529 return NOTIFY_DONE;
1530 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1531 EnterFunction(2);
1532 mutex_lock(&__ip_vs_mutex);
1533 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1534 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1535 if (net_eq(svc->net, net)) {
1536 list_for_each_entry(dest, &svc->destinations,
1537 n_list) {
1538 __ip_vs_dev_reset(dest, dev);
1539 }
1540 }
1541 }
1542
1543 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1544 if (net_eq(svc->net, net)) {
1545 list_for_each_entry(dest, &svc->destinations,
1546 n_list) {
1547 __ip_vs_dev_reset(dest, dev);
1548 }
1549 }
1550
1551 }
1552 }
1553
1554 list_for_each_entry(dest, &net_ipvs(net)->dest_trash, n_list) {
1555 __ip_vs_dev_reset(dest, dev);
1556 }
1557 mutex_unlock(&__ip_vs_mutex);
1558 LeaveFunction(2);
1559 return NOTIFY_DONE;
1560 }
1561
1562 /*
1563 * Zero counters in a service or all services
1564 */
1565 static int ip_vs_zero_service(struct ip_vs_service *svc)
1566 {
1567 struct ip_vs_dest *dest;
1568
1569 write_lock_bh(&__ip_vs_svc_lock);
1570 list_for_each_entry(dest, &svc->destinations, n_list) {
1571 ip_vs_zero_stats(&dest->stats);
1572 }
1573 ip_vs_zero_stats(&svc->stats);
1574 write_unlock_bh(&__ip_vs_svc_lock);
1575 return 0;
1576 }
1577
1578 static int ip_vs_zero_all(struct net *net)
1579 {
1580 int idx;
1581 struct ip_vs_service *svc;
1582
1583 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1584 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1585 if (net_eq(svc->net, net))
1586 ip_vs_zero_service(svc);
1587 }
1588 }
1589
1590 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1591 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1592 if (net_eq(svc->net, net))
1593 ip_vs_zero_service(svc);
1594 }
1595 }
1596
1597 ip_vs_zero_stats(&net_ipvs(net)->tot_stats);
1598 return 0;
1599 }
1600
1601 #ifdef CONFIG_SYSCTL
1602 static int
1603 proc_do_defense_mode(ctl_table *table, int write,
1604 void __user *buffer, size_t *lenp, loff_t *ppos)
1605 {
1606 struct net *net = current->nsproxy->net_ns;
1607 int *valp = table->data;
1608 int val = *valp;
1609 int rc;
1610
1611 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1612 if (write && (*valp != val)) {
1613 if ((*valp < 0) || (*valp > 3)) {
1614 /* Restore the correct value */
1615 *valp = val;
1616 } else {
1617 update_defense_level(net_ipvs(net));
1618 }
1619 }
1620 return rc;
1621 }
1622
1623 static int
1624 proc_do_sync_threshold(ctl_table *table, int write,
1625 void __user *buffer, size_t *lenp, loff_t *ppos)
1626 {
1627 int *valp = table->data;
1628 int val[2];
1629 int rc;
1630
1631 /* backup the value first */
1632 memcpy(val, valp, sizeof(val));
1633
1634 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1635 if (write && (valp[0] < 0 || valp[1] < 0 || valp[0] >= valp[1])) {
1636 /* Restore the correct value */
1637 memcpy(valp, val, sizeof(val));
1638 }
1639 return rc;
1640 }
1641
1642 static int
1643 proc_do_sync_mode(ctl_table *table, int write,
1644 void __user *buffer, size_t *lenp, loff_t *ppos)
1645 {
1646 int *valp = table->data;
1647 int val = *valp;
1648 int rc;
1649
1650 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1651 if (write && (*valp != val)) {
1652 if ((*valp < 0) || (*valp > 1)) {
1653 /* Restore the correct value */
1654 *valp = val;
1655 } else {
1656 struct net *net = current->nsproxy->net_ns;
1657 ip_vs_sync_switch_mode(net, val);
1658 }
1659 }
1660 return rc;
1661 }
1662
1663 /*
1664 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1665 * Do not change order or insert new entries without
1666 * align with netns init in ip_vs_control_net_init()
1667 */
1668
1669 static struct ctl_table vs_vars[] = {
1670 {
1671 .procname = "amemthresh",
1672 .maxlen = sizeof(int),
1673 .mode = 0644,
1674 .proc_handler = proc_dointvec,
1675 },
1676 {
1677 .procname = "am_droprate",
1678 .maxlen = sizeof(int),
1679 .mode = 0644,
1680 .proc_handler = proc_dointvec,
1681 },
1682 {
1683 .procname = "drop_entry",
1684 .maxlen = sizeof(int),
1685 .mode = 0644,
1686 .proc_handler = proc_do_defense_mode,
1687 },
1688 {
1689 .procname = "drop_packet",
1690 .maxlen = sizeof(int),
1691 .mode = 0644,
1692 .proc_handler = proc_do_defense_mode,
1693 },
1694 #ifdef CONFIG_IP_VS_NFCT
1695 {
1696 .procname = "conntrack",
1697 .maxlen = sizeof(int),
1698 .mode = 0644,
1699 .proc_handler = &proc_dointvec,
1700 },
1701 #endif
1702 {
1703 .procname = "secure_tcp",
1704 .maxlen = sizeof(int),
1705 .mode = 0644,
1706 .proc_handler = proc_do_defense_mode,
1707 },
1708 {
1709 .procname = "snat_reroute",
1710 .maxlen = sizeof(int),
1711 .mode = 0644,
1712 .proc_handler = &proc_dointvec,
1713 },
1714 {
1715 .procname = "sync_version",
1716 .maxlen = sizeof(int),
1717 .mode = 0644,
1718 .proc_handler = &proc_do_sync_mode,
1719 },
1720 {
1721 .procname = "cache_bypass",
1722 .maxlen = sizeof(int),
1723 .mode = 0644,
1724 .proc_handler = proc_dointvec,
1725 },
1726 {
1727 .procname = "expire_nodest_conn",
1728 .maxlen = sizeof(int),
1729 .mode = 0644,
1730 .proc_handler = proc_dointvec,
1731 },
1732 {
1733 .procname = "expire_quiescent_template",
1734 .maxlen = sizeof(int),
1735 .mode = 0644,
1736 .proc_handler = proc_dointvec,
1737 },
1738 {
1739 .procname = "sync_threshold",
1740 .maxlen =
1741 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1742 .mode = 0644,
1743 .proc_handler = proc_do_sync_threshold,
1744 },
1745 {
1746 .procname = "nat_icmp_send",
1747 .maxlen = sizeof(int),
1748 .mode = 0644,
1749 .proc_handler = proc_dointvec,
1750 },
1751 #ifdef CONFIG_IP_VS_DEBUG
1752 {
1753 .procname = "debug_level",
1754 .data = &sysctl_ip_vs_debug_level,
1755 .maxlen = sizeof(int),
1756 .mode = 0644,
1757 .proc_handler = proc_dointvec,
1758 },
1759 #endif
1760 #if 0
1761 {
1762 .procname = "timeout_established",
1763 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1764 .maxlen = sizeof(int),
1765 .mode = 0644,
1766 .proc_handler = proc_dointvec_jiffies,
1767 },
1768 {
1769 .procname = "timeout_synsent",
1770 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1771 .maxlen = sizeof(int),
1772 .mode = 0644,
1773 .proc_handler = proc_dointvec_jiffies,
1774 },
1775 {
1776 .procname = "timeout_synrecv",
1777 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1778 .maxlen = sizeof(int),
1779 .mode = 0644,
1780 .proc_handler = proc_dointvec_jiffies,
1781 },
1782 {
1783 .procname = "timeout_finwait",
1784 .data = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1785 .maxlen = sizeof(int),
1786 .mode = 0644,
1787 .proc_handler = proc_dointvec_jiffies,
1788 },
1789 {
1790 .procname = "timeout_timewait",
1791 .data = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1792 .maxlen = sizeof(int),
1793 .mode = 0644,
1794 .proc_handler = proc_dointvec_jiffies,
1795 },
1796 {
1797 .procname = "timeout_close",
1798 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1799 .maxlen = sizeof(int),
1800 .mode = 0644,
1801 .proc_handler = proc_dointvec_jiffies,
1802 },
1803 {
1804 .procname = "timeout_closewait",
1805 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1806 .maxlen = sizeof(int),
1807 .mode = 0644,
1808 .proc_handler = proc_dointvec_jiffies,
1809 },
1810 {
1811 .procname = "timeout_lastack",
1812 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1813 .maxlen = sizeof(int),
1814 .mode = 0644,
1815 .proc_handler = proc_dointvec_jiffies,
1816 },
1817 {
1818 .procname = "timeout_listen",
1819 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1820 .maxlen = sizeof(int),
1821 .mode = 0644,
1822 .proc_handler = proc_dointvec_jiffies,
1823 },
1824 {
1825 .procname = "timeout_synack",
1826 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1827 .maxlen = sizeof(int),
1828 .mode = 0644,
1829 .proc_handler = proc_dointvec_jiffies,
1830 },
1831 {
1832 .procname = "timeout_udp",
1833 .data = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1834 .maxlen = sizeof(int),
1835 .mode = 0644,
1836 .proc_handler = proc_dointvec_jiffies,
1837 },
1838 {
1839 .procname = "timeout_icmp",
1840 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1841 .maxlen = sizeof(int),
1842 .mode = 0644,
1843 .proc_handler = proc_dointvec_jiffies,
1844 },
1845 #endif
1846 { }
1847 };
1848
1849 #endif
1850
1851 #ifdef CONFIG_PROC_FS
1852
1853 struct ip_vs_iter {
1854 struct seq_net_private p; /* Do not move this, netns depends upon it*/
1855 struct list_head *table;
1856 int bucket;
1857 };
1858
1859 /*
1860 * Write the contents of the VS rule table to a PROCfs file.
1861 * (It is kept just for backward compatibility)
1862 */
1863 static inline const char *ip_vs_fwd_name(unsigned int flags)
1864 {
1865 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1866 case IP_VS_CONN_F_LOCALNODE:
1867 return "Local";
1868 case IP_VS_CONN_F_TUNNEL:
1869 return "Tunnel";
1870 case IP_VS_CONN_F_DROUTE:
1871 return "Route";
1872 default:
1873 return "Masq";
1874 }
1875 }
1876
1877
1878 /* Get the Nth entry in the two lists */
1879 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1880 {
1881 struct net *net = seq_file_net(seq);
1882 struct ip_vs_iter *iter = seq->private;
1883 int idx;
1884 struct ip_vs_service *svc;
1885
1886 /* look in hash by protocol */
1887 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1888 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1889 if (net_eq(svc->net, net) && pos-- == 0) {
1890 iter->table = ip_vs_svc_table;
1891 iter->bucket = idx;
1892 return svc;
1893 }
1894 }
1895 }
1896
1897 /* keep looking in fwmark */
1898 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1899 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1900 if (net_eq(svc->net, net) && pos-- == 0) {
1901 iter->table = ip_vs_svc_fwm_table;
1902 iter->bucket = idx;
1903 return svc;
1904 }
1905 }
1906 }
1907
1908 return NULL;
1909 }
1910
1911 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
1912 __acquires(__ip_vs_svc_lock)
1913 {
1914
1915 read_lock_bh(&__ip_vs_svc_lock);
1916 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1917 }
1918
1919
1920 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1921 {
1922 struct list_head *e;
1923 struct ip_vs_iter *iter;
1924 struct ip_vs_service *svc;
1925
1926 ++*pos;
1927 if (v == SEQ_START_TOKEN)
1928 return ip_vs_info_array(seq,0);
1929
1930 svc = v;
1931 iter = seq->private;
1932
1933 if (iter->table == ip_vs_svc_table) {
1934 /* next service in table hashed by protocol */
1935 if ((e = svc->s_list.next) != &ip_vs_svc_table[iter->bucket])
1936 return list_entry(e, struct ip_vs_service, s_list);
1937
1938
1939 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1940 list_for_each_entry(svc,&ip_vs_svc_table[iter->bucket],
1941 s_list) {
1942 return svc;
1943 }
1944 }
1945
1946 iter->table = ip_vs_svc_fwm_table;
1947 iter->bucket = -1;
1948 goto scan_fwmark;
1949 }
1950
1951 /* next service in hashed by fwmark */
1952 if ((e = svc->f_list.next) != &ip_vs_svc_fwm_table[iter->bucket])
1953 return list_entry(e, struct ip_vs_service, f_list);
1954
1955 scan_fwmark:
1956 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1957 list_for_each_entry(svc, &ip_vs_svc_fwm_table[iter->bucket],
1958 f_list)
1959 return svc;
1960 }
1961
1962 return NULL;
1963 }
1964
1965 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
1966 __releases(__ip_vs_svc_lock)
1967 {
1968 read_unlock_bh(&__ip_vs_svc_lock);
1969 }
1970
1971
1972 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
1973 {
1974 if (v == SEQ_START_TOKEN) {
1975 seq_printf(seq,
1976 "IP Virtual Server version %d.%d.%d (size=%d)\n",
1977 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
1978 seq_puts(seq,
1979 "Prot LocalAddress:Port Scheduler Flags\n");
1980 seq_puts(seq,
1981 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
1982 } else {
1983 const struct ip_vs_service *svc = v;
1984 const struct ip_vs_iter *iter = seq->private;
1985 const struct ip_vs_dest *dest;
1986
1987 if (iter->table == ip_vs_svc_table) {
1988 #ifdef CONFIG_IP_VS_IPV6
1989 if (svc->af == AF_INET6)
1990 seq_printf(seq, "%s [%pI6]:%04X %s ",
1991 ip_vs_proto_name(svc->protocol),
1992 &svc->addr.in6,
1993 ntohs(svc->port),
1994 svc->scheduler->name);
1995 else
1996 #endif
1997 seq_printf(seq, "%s %08X:%04X %s %s ",
1998 ip_vs_proto_name(svc->protocol),
1999 ntohl(svc->addr.ip),
2000 ntohs(svc->port),
2001 svc->scheduler->name,
2002 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2003 } else {
2004 seq_printf(seq, "FWM %08X %s %s",
2005 svc->fwmark, svc->scheduler->name,
2006 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2007 }
2008
2009 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2010 seq_printf(seq, "persistent %d %08X\n",
2011 svc->timeout,
2012 ntohl(svc->netmask));
2013 else
2014 seq_putc(seq, '\n');
2015
2016 list_for_each_entry(dest, &svc->destinations, n_list) {
2017 #ifdef CONFIG_IP_VS_IPV6
2018 if (dest->af == AF_INET6)
2019 seq_printf(seq,
2020 " -> [%pI6]:%04X"
2021 " %-7s %-6d %-10d %-10d\n",
2022 &dest->addr.in6,
2023 ntohs(dest->port),
2024 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2025 atomic_read(&dest->weight),
2026 atomic_read(&dest->activeconns),
2027 atomic_read(&dest->inactconns));
2028 else
2029 #endif
2030 seq_printf(seq,
2031 " -> %08X:%04X "
2032 "%-7s %-6d %-10d %-10d\n",
2033 ntohl(dest->addr.ip),
2034 ntohs(dest->port),
2035 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2036 atomic_read(&dest->weight),
2037 atomic_read(&dest->activeconns),
2038 atomic_read(&dest->inactconns));
2039
2040 }
2041 }
2042 return 0;
2043 }
2044
2045 static const struct seq_operations ip_vs_info_seq_ops = {
2046 .start = ip_vs_info_seq_start,
2047 .next = ip_vs_info_seq_next,
2048 .stop = ip_vs_info_seq_stop,
2049 .show = ip_vs_info_seq_show,
2050 };
2051
2052 static int ip_vs_info_open(struct inode *inode, struct file *file)
2053 {
2054 return seq_open_net(inode, file, &ip_vs_info_seq_ops,
2055 sizeof(struct ip_vs_iter));
2056 }
2057
2058 static const struct file_operations ip_vs_info_fops = {
2059 .owner = THIS_MODULE,
2060 .open = ip_vs_info_open,
2061 .read = seq_read,
2062 .llseek = seq_lseek,
2063 .release = seq_release_net,
2064 };
2065
2066 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2067 {
2068 struct net *net = seq_file_single_net(seq);
2069 struct ip_vs_stats_user show;
2070
2071 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2072 seq_puts(seq,
2073 " Total Incoming Outgoing Incoming Outgoing\n");
2074 seq_printf(seq,
2075 " Conns Packets Packets Bytes Bytes\n");
2076
2077 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2078 seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", show.conns,
2079 show.inpkts, show.outpkts,
2080 (unsigned long long) show.inbytes,
2081 (unsigned long long) show.outbytes);
2082
2083 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2084 seq_puts(seq,
2085 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2086 seq_printf(seq, "%8X %8X %8X %16X %16X\n",
2087 show.cps, show.inpps, show.outpps,
2088 show.inbps, show.outbps);
2089
2090 return 0;
2091 }
2092
2093 static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2094 {
2095 return single_open_net(inode, file, ip_vs_stats_show);
2096 }
2097
2098 static const struct file_operations ip_vs_stats_fops = {
2099 .owner = THIS_MODULE,
2100 .open = ip_vs_stats_seq_open,
2101 .read = seq_read,
2102 .llseek = seq_lseek,
2103 .release = single_release_net,
2104 };
2105
2106 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2107 {
2108 struct net *net = seq_file_single_net(seq);
2109 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2110 struct ip_vs_cpu_stats *cpustats = tot_stats->cpustats;
2111 struct ip_vs_stats_user rates;
2112 int i;
2113
2114 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2115 seq_puts(seq,
2116 " Total Incoming Outgoing Incoming Outgoing\n");
2117 seq_printf(seq,
2118 "CPU Conns Packets Packets Bytes Bytes\n");
2119
2120 for_each_possible_cpu(i) {
2121 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2122 unsigned int start;
2123 __u64 inbytes, outbytes;
2124
2125 do {
2126 start = u64_stats_fetch_begin_bh(&u->syncp);
2127 inbytes = u->ustats.inbytes;
2128 outbytes = u->ustats.outbytes;
2129 } while (u64_stats_fetch_retry_bh(&u->syncp, start));
2130
2131 seq_printf(seq, "%3X %8X %8X %8X %16LX %16LX\n",
2132 i, u->ustats.conns, u->ustats.inpkts,
2133 u->ustats.outpkts, (__u64)inbytes,
2134 (__u64)outbytes);
2135 }
2136
2137 spin_lock_bh(&tot_stats->lock);
2138
2139 seq_printf(seq, " ~ %8X %8X %8X %16LX %16LX\n\n",
2140 tot_stats->ustats.conns, tot_stats->ustats.inpkts,
2141 tot_stats->ustats.outpkts,
2142 (unsigned long long) tot_stats->ustats.inbytes,
2143 (unsigned long long) tot_stats->ustats.outbytes);
2144
2145 ip_vs_read_estimator(&rates, tot_stats);
2146
2147 spin_unlock_bh(&tot_stats->lock);
2148
2149 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2150 seq_puts(seq,
2151 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2152 seq_printf(seq, " %8X %8X %8X %16X %16X\n",
2153 rates.cps,
2154 rates.inpps,
2155 rates.outpps,
2156 rates.inbps,
2157 rates.outbps);
2158
2159 return 0;
2160 }
2161
2162 static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2163 {
2164 return single_open_net(inode, file, ip_vs_stats_percpu_show);
2165 }
2166
2167 static const struct file_operations ip_vs_stats_percpu_fops = {
2168 .owner = THIS_MODULE,
2169 .open = ip_vs_stats_percpu_seq_open,
2170 .read = seq_read,
2171 .llseek = seq_lseek,
2172 .release = single_release_net,
2173 };
2174 #endif
2175
2176 /*
2177 * Set timeout values for tcp tcpfin udp in the timeout_table.
2178 */
2179 static int ip_vs_set_timeout(struct net *net, struct ip_vs_timeout_user *u)
2180 {
2181 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2182 struct ip_vs_proto_data *pd;
2183 #endif
2184
2185 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2186 u->tcp_timeout,
2187 u->tcp_fin_timeout,
2188 u->udp_timeout);
2189
2190 #ifdef CONFIG_IP_VS_PROTO_TCP
2191 if (u->tcp_timeout) {
2192 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2193 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2194 = u->tcp_timeout * HZ;
2195 }
2196
2197 if (u->tcp_fin_timeout) {
2198 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2199 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2200 = u->tcp_fin_timeout * HZ;
2201 }
2202 #endif
2203
2204 #ifdef CONFIG_IP_VS_PROTO_UDP
2205 if (u->udp_timeout) {
2206 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2207 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2208 = u->udp_timeout * HZ;
2209 }
2210 #endif
2211 return 0;
2212 }
2213
2214
2215 #define SET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2216 #define SERVICE_ARG_LEN (sizeof(struct ip_vs_service_user))
2217 #define SVCDEST_ARG_LEN (sizeof(struct ip_vs_service_user) + \
2218 sizeof(struct ip_vs_dest_user))
2219 #define TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2220 #define DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user))
2221 #define MAX_ARG_LEN SVCDEST_ARG_LEN
2222
2223 static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
2224 [SET_CMDID(IP_VS_SO_SET_ADD)] = SERVICE_ARG_LEN,
2225 [SET_CMDID(IP_VS_SO_SET_EDIT)] = SERVICE_ARG_LEN,
2226 [SET_CMDID(IP_VS_SO_SET_DEL)] = SERVICE_ARG_LEN,
2227 [SET_CMDID(IP_VS_SO_SET_FLUSH)] = 0,
2228 [SET_CMDID(IP_VS_SO_SET_ADDDEST)] = SVCDEST_ARG_LEN,
2229 [SET_CMDID(IP_VS_SO_SET_DELDEST)] = SVCDEST_ARG_LEN,
2230 [SET_CMDID(IP_VS_SO_SET_EDITDEST)] = SVCDEST_ARG_LEN,
2231 [SET_CMDID(IP_VS_SO_SET_TIMEOUT)] = TIMEOUT_ARG_LEN,
2232 [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)] = DAEMON_ARG_LEN,
2233 [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)] = DAEMON_ARG_LEN,
2234 [SET_CMDID(IP_VS_SO_SET_ZERO)] = SERVICE_ARG_LEN,
2235 };
2236
2237 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2238 struct ip_vs_service_user *usvc_compat)
2239 {
2240 memset(usvc, 0, sizeof(*usvc));
2241
2242 usvc->af = AF_INET;
2243 usvc->protocol = usvc_compat->protocol;
2244 usvc->addr.ip = usvc_compat->addr;
2245 usvc->port = usvc_compat->port;
2246 usvc->fwmark = usvc_compat->fwmark;
2247
2248 /* Deep copy of sched_name is not needed here */
2249 usvc->sched_name = usvc_compat->sched_name;
2250
2251 usvc->flags = usvc_compat->flags;
2252 usvc->timeout = usvc_compat->timeout;
2253 usvc->netmask = usvc_compat->netmask;
2254 }
2255
2256 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2257 struct ip_vs_dest_user *udest_compat)
2258 {
2259 memset(udest, 0, sizeof(*udest));
2260
2261 udest->addr.ip = udest_compat->addr;
2262 udest->port = udest_compat->port;
2263 udest->conn_flags = udest_compat->conn_flags;
2264 udest->weight = udest_compat->weight;
2265 udest->u_threshold = udest_compat->u_threshold;
2266 udest->l_threshold = udest_compat->l_threshold;
2267 }
2268
2269 static int
2270 do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2271 {
2272 struct net *net = sock_net(sk);
2273 int ret;
2274 unsigned char arg[MAX_ARG_LEN];
2275 struct ip_vs_service_user *usvc_compat;
2276 struct ip_vs_service_user_kern usvc;
2277 struct ip_vs_service *svc;
2278 struct ip_vs_dest_user *udest_compat;
2279 struct ip_vs_dest_user_kern udest;
2280 struct netns_ipvs *ipvs = net_ipvs(net);
2281
2282 if (!capable(CAP_NET_ADMIN))
2283 return -EPERM;
2284
2285 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2286 return -EINVAL;
2287 if (len < 0 || len > MAX_ARG_LEN)
2288 return -EINVAL;
2289 if (len != set_arglen[SET_CMDID(cmd)]) {
2290 pr_err("set_ctl: len %u != %u\n",
2291 len, set_arglen[SET_CMDID(cmd)]);
2292 return -EINVAL;
2293 }
2294
2295 if (copy_from_user(arg, user, len) != 0)
2296 return -EFAULT;
2297
2298 /* increase the module use count */
2299 ip_vs_use_count_inc();
2300
2301 /* Handle daemons since they have another lock */
2302 if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2303 cmd == IP_VS_SO_SET_STOPDAEMON) {
2304 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2305
2306 if (mutex_lock_interruptible(&ipvs->sync_mutex)) {
2307 ret = -ERESTARTSYS;
2308 goto out_dec;
2309 }
2310 if (cmd == IP_VS_SO_SET_STARTDAEMON)
2311 ret = start_sync_thread(net, dm->state, dm->mcast_ifn,
2312 dm->syncid);
2313 else
2314 ret = stop_sync_thread(net, dm->state);
2315 mutex_unlock(&ipvs->sync_mutex);
2316 goto out_dec;
2317 }
2318
2319 if (mutex_lock_interruptible(&__ip_vs_mutex)) {
2320 ret = -ERESTARTSYS;
2321 goto out_dec;
2322 }
2323
2324 if (cmd == IP_VS_SO_SET_FLUSH) {
2325 /* Flush the virtual service */
2326 ret = ip_vs_flush(net);
2327 goto out_unlock;
2328 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2329 /* Set timeout values for (tcp tcpfin udp) */
2330 ret = ip_vs_set_timeout(net, (struct ip_vs_timeout_user *)arg);
2331 goto out_unlock;
2332 }
2333
2334 usvc_compat = (struct ip_vs_service_user *)arg;
2335 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2336
2337 /* We only use the new structs internally, so copy userspace compat
2338 * structs to extended internal versions */
2339 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2340 ip_vs_copy_udest_compat(&udest, udest_compat);
2341
2342 if (cmd == IP_VS_SO_SET_ZERO) {
2343 /* if no service address is set, zero counters in all */
2344 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2345 ret = ip_vs_zero_all(net);
2346 goto out_unlock;
2347 }
2348 }
2349
2350 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2351 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2352 usvc.protocol != IPPROTO_SCTP) {
2353 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2354 usvc.protocol, &usvc.addr.ip,
2355 ntohs(usvc.port), usvc.sched_name);
2356 ret = -EFAULT;
2357 goto out_unlock;
2358 }
2359
2360 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2361 if (usvc.fwmark == 0)
2362 svc = __ip_vs_service_find(net, usvc.af, usvc.protocol,
2363 &usvc.addr, usvc.port);
2364 else
2365 svc = __ip_vs_svc_fwm_find(net, usvc.af, usvc.fwmark);
2366
2367 if (cmd != IP_VS_SO_SET_ADD
2368 && (svc == NULL || svc->protocol != usvc.protocol)) {
2369 ret = -ESRCH;
2370 goto out_unlock;
2371 }
2372
2373 switch (cmd) {
2374 case IP_VS_SO_SET_ADD:
2375 if (svc != NULL)
2376 ret = -EEXIST;
2377 else
2378 ret = ip_vs_add_service(net, &usvc, &svc);
2379 break;
2380 case IP_VS_SO_SET_EDIT:
2381 ret = ip_vs_edit_service(svc, &usvc);
2382 break;
2383 case IP_VS_SO_SET_DEL:
2384 ret = ip_vs_del_service(svc);
2385 if (!ret)
2386 goto out_unlock;
2387 break;
2388 case IP_VS_SO_SET_ZERO:
2389 ret = ip_vs_zero_service(svc);
2390 break;
2391 case IP_VS_SO_SET_ADDDEST:
2392 ret = ip_vs_add_dest(svc, &udest);
2393 break;
2394 case IP_VS_SO_SET_EDITDEST:
2395 ret = ip_vs_edit_dest(svc, &udest);
2396 break;
2397 case IP_VS_SO_SET_DELDEST:
2398 ret = ip_vs_del_dest(svc, &udest);
2399 break;
2400 default:
2401 ret = -EINVAL;
2402 }
2403
2404 out_unlock:
2405 mutex_unlock(&__ip_vs_mutex);
2406 out_dec:
2407 /* decrease the module use count */
2408 ip_vs_use_count_dec();
2409
2410 return ret;
2411 }
2412
2413
2414 static void
2415 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2416 {
2417 dst->protocol = src->protocol;
2418 dst->addr = src->addr.ip;
2419 dst->port = src->port;
2420 dst->fwmark = src->fwmark;
2421 strlcpy(dst->sched_name, src->scheduler->name, sizeof(dst->sched_name));
2422 dst->flags = src->flags;
2423 dst->timeout = src->timeout / HZ;
2424 dst->netmask = src->netmask;
2425 dst->num_dests = src->num_dests;
2426 ip_vs_copy_stats(&dst->stats, &src->stats);
2427 }
2428
2429 static inline int
2430 __ip_vs_get_service_entries(struct net *net,
2431 const struct ip_vs_get_services *get,
2432 struct ip_vs_get_services __user *uptr)
2433 {
2434 int idx, count=0;
2435 struct ip_vs_service *svc;
2436 struct ip_vs_service_entry entry;
2437 int ret = 0;
2438
2439 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2440 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2441 /* Only expose IPv4 entries to old interface */
2442 if (svc->af != AF_INET || !net_eq(svc->net, net))
2443 continue;
2444
2445 if (count >= get->num_services)
2446 goto out;
2447 memset(&entry, 0, sizeof(entry));
2448 ip_vs_copy_service(&entry, svc);
2449 if (copy_to_user(&uptr->entrytable[count],
2450 &entry, sizeof(entry))) {
2451 ret = -EFAULT;
2452 goto out;
2453 }
2454 count++;
2455 }
2456 }
2457
2458 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2459 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2460 /* Only expose IPv4 entries to old interface */
2461 if (svc->af != AF_INET || !net_eq(svc->net, net))
2462 continue;
2463
2464 if (count >= get->num_services)
2465 goto out;
2466 memset(&entry, 0, sizeof(entry));
2467 ip_vs_copy_service(&entry, svc);
2468 if (copy_to_user(&uptr->entrytable[count],
2469 &entry, sizeof(entry))) {
2470 ret = -EFAULT;
2471 goto out;
2472 }
2473 count++;
2474 }
2475 }
2476 out:
2477 return ret;
2478 }
2479
2480 static inline int
2481 __ip_vs_get_dest_entries(struct net *net, const struct ip_vs_get_dests *get,
2482 struct ip_vs_get_dests __user *uptr)
2483 {
2484 struct ip_vs_service *svc;
2485 union nf_inet_addr addr = { .ip = get->addr };
2486 int ret = 0;
2487
2488 if (get->fwmark)
2489 svc = __ip_vs_svc_fwm_find(net, AF_INET, get->fwmark);
2490 else
2491 svc = __ip_vs_service_find(net, AF_INET, get->protocol, &addr,
2492 get->port);
2493
2494 if (svc) {
2495 int count = 0;
2496 struct ip_vs_dest *dest;
2497 struct ip_vs_dest_entry entry;
2498
2499 list_for_each_entry(dest, &svc->destinations, n_list) {
2500 if (count >= get->num_dests)
2501 break;
2502
2503 entry.addr = dest->addr.ip;
2504 entry.port = dest->port;
2505 entry.conn_flags = atomic_read(&dest->conn_flags);
2506 entry.weight = atomic_read(&dest->weight);
2507 entry.u_threshold = dest->u_threshold;
2508 entry.l_threshold = dest->l_threshold;
2509 entry.activeconns = atomic_read(&dest->activeconns);
2510 entry.inactconns = atomic_read(&dest->inactconns);
2511 entry.persistconns = atomic_read(&dest->persistconns);
2512 ip_vs_copy_stats(&entry.stats, &dest->stats);
2513 if (copy_to_user(&uptr->entrytable[count],
2514 &entry, sizeof(entry))) {
2515 ret = -EFAULT;
2516 break;
2517 }
2518 count++;
2519 }
2520 } else
2521 ret = -ESRCH;
2522 return ret;
2523 }
2524
2525 static inline void
2526 __ip_vs_get_timeouts(struct net *net, struct ip_vs_timeout_user *u)
2527 {
2528 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2529 struct ip_vs_proto_data *pd;
2530 #endif
2531
2532 #ifdef CONFIG_IP_VS_PROTO_TCP
2533 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2534 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2535 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2536 #endif
2537 #ifdef CONFIG_IP_VS_PROTO_UDP
2538 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2539 u->udp_timeout =
2540 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2541 #endif
2542 }
2543
2544
2545 #define GET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2546 #define GET_INFO_ARG_LEN (sizeof(struct ip_vs_getinfo))
2547 #define GET_SERVICES_ARG_LEN (sizeof(struct ip_vs_get_services))
2548 #define GET_SERVICE_ARG_LEN (sizeof(struct ip_vs_service_entry))
2549 #define GET_DESTS_ARG_LEN (sizeof(struct ip_vs_get_dests))
2550 #define GET_TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2551 #define GET_DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user) * 2)
2552
2553 static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
2554 [GET_CMDID(IP_VS_SO_GET_VERSION)] = 64,
2555 [GET_CMDID(IP_VS_SO_GET_INFO)] = GET_INFO_ARG_LEN,
2556 [GET_CMDID(IP_VS_SO_GET_SERVICES)] = GET_SERVICES_ARG_LEN,
2557 [GET_CMDID(IP_VS_SO_GET_SERVICE)] = GET_SERVICE_ARG_LEN,
2558 [GET_CMDID(IP_VS_SO_GET_DESTS)] = GET_DESTS_ARG_LEN,
2559 [GET_CMDID(IP_VS_SO_GET_TIMEOUT)] = GET_TIMEOUT_ARG_LEN,
2560 [GET_CMDID(IP_VS_SO_GET_DAEMON)] = GET_DAEMON_ARG_LEN,
2561 };
2562
2563 static int
2564 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2565 {
2566 unsigned char arg[128];
2567 int ret = 0;
2568 unsigned int copylen;
2569 struct net *net = sock_net(sk);
2570 struct netns_ipvs *ipvs = net_ipvs(net);
2571
2572 BUG_ON(!net);
2573 if (!capable(CAP_NET_ADMIN))
2574 return -EPERM;
2575
2576 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2577 return -EINVAL;
2578
2579 if (*len < get_arglen[GET_CMDID(cmd)]) {
2580 pr_err("get_ctl: len %u < %u\n",
2581 *len, get_arglen[GET_CMDID(cmd)]);
2582 return -EINVAL;
2583 }
2584
2585 copylen = get_arglen[GET_CMDID(cmd)];
2586 if (copylen > 128)
2587 return -EINVAL;
2588
2589 if (copy_from_user(arg, user, copylen) != 0)
2590 return -EFAULT;
2591 /*
2592 * Handle daemons first since it has its own locking
2593 */
2594 if (cmd == IP_VS_SO_GET_DAEMON) {
2595 struct ip_vs_daemon_user d[2];
2596
2597 memset(&d, 0, sizeof(d));
2598 if (mutex_lock_interruptible(&ipvs->sync_mutex))
2599 return -ERESTARTSYS;
2600
2601 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2602 d[0].state = IP_VS_STATE_MASTER;
2603 strlcpy(d[0].mcast_ifn, ipvs->master_mcast_ifn,
2604 sizeof(d[0].mcast_ifn));
2605 d[0].syncid = ipvs->master_syncid;
2606 }
2607 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2608 d[1].state = IP_VS_STATE_BACKUP;
2609 strlcpy(d[1].mcast_ifn, ipvs->backup_mcast_ifn,
2610 sizeof(d[1].mcast_ifn));
2611 d[1].syncid = ipvs->backup_syncid;
2612 }
2613 if (copy_to_user(user, &d, sizeof(d)) != 0)
2614 ret = -EFAULT;
2615 mutex_unlock(&ipvs->sync_mutex);
2616 return ret;
2617 }
2618
2619 if (mutex_lock_interruptible(&__ip_vs_mutex))
2620 return -ERESTARTSYS;
2621
2622 switch (cmd) {
2623 case IP_VS_SO_GET_VERSION:
2624 {
2625 char buf[64];
2626
2627 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2628 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2629 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2630 ret = -EFAULT;
2631 goto out;
2632 }
2633 *len = strlen(buf)+1;
2634 }
2635 break;
2636
2637 case IP_VS_SO_GET_INFO:
2638 {
2639 struct ip_vs_getinfo info;
2640 info.version = IP_VS_VERSION_CODE;
2641 info.size = ip_vs_conn_tab_size;
2642 info.num_services = ipvs->num_services;
2643 if (copy_to_user(user, &info, sizeof(info)) != 0)
2644 ret = -EFAULT;
2645 }
2646 break;
2647
2648 case IP_VS_SO_GET_SERVICES:
2649 {
2650 struct ip_vs_get_services *get;
2651 int size;
2652
2653 get = (struct ip_vs_get_services *)arg;
2654 size = sizeof(*get) +
2655 sizeof(struct ip_vs_service_entry) * get->num_services;
2656 if (*len != size) {
2657 pr_err("length: %u != %u\n", *len, size);
2658 ret = -EINVAL;
2659 goto out;
2660 }
2661 ret = __ip_vs_get_service_entries(net, get, user);
2662 }
2663 break;
2664
2665 case IP_VS_SO_GET_SERVICE:
2666 {
2667 struct ip_vs_service_entry *entry;
2668 struct ip_vs_service *svc;
2669 union nf_inet_addr addr;
2670
2671 entry = (struct ip_vs_service_entry *)arg;
2672 addr.ip = entry->addr;
2673 if (entry->fwmark)
2674 svc = __ip_vs_svc_fwm_find(net, AF_INET, entry->fwmark);
2675 else
2676 svc = __ip_vs_service_find(net, AF_INET,
2677 entry->protocol, &addr,
2678 entry->port);
2679 if (svc) {
2680 ip_vs_copy_service(entry, svc);
2681 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2682 ret = -EFAULT;
2683 } else
2684 ret = -ESRCH;
2685 }
2686 break;
2687
2688 case IP_VS_SO_GET_DESTS:
2689 {
2690 struct ip_vs_get_dests *get;
2691 int size;
2692
2693 get = (struct ip_vs_get_dests *)arg;
2694 size = sizeof(*get) +
2695 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2696 if (*len != size) {
2697 pr_err("length: %u != %u\n", *len, size);
2698 ret = -EINVAL;
2699 goto out;
2700 }
2701 ret = __ip_vs_get_dest_entries(net, get, user);
2702 }
2703 break;
2704
2705 case IP_VS_SO_GET_TIMEOUT:
2706 {
2707 struct ip_vs_timeout_user t;
2708
2709 __ip_vs_get_timeouts(net, &t);
2710 if (copy_to_user(user, &t, sizeof(t)) != 0)
2711 ret = -EFAULT;
2712 }
2713 break;
2714
2715 default:
2716 ret = -EINVAL;
2717 }
2718
2719 out:
2720 mutex_unlock(&__ip_vs_mutex);
2721 return ret;
2722 }
2723
2724
2725 static struct nf_sockopt_ops ip_vs_sockopts = {
2726 .pf = PF_INET,
2727 .set_optmin = IP_VS_BASE_CTL,
2728 .set_optmax = IP_VS_SO_SET_MAX+1,
2729 .set = do_ip_vs_set_ctl,
2730 .get_optmin = IP_VS_BASE_CTL,
2731 .get_optmax = IP_VS_SO_GET_MAX+1,
2732 .get = do_ip_vs_get_ctl,
2733 .owner = THIS_MODULE,
2734 };
2735
2736 /*
2737 * Generic Netlink interface
2738 */
2739
2740 /* IPVS genetlink family */
2741 static struct genl_family ip_vs_genl_family = {
2742 .id = GENL_ID_GENERATE,
2743 .hdrsize = 0,
2744 .name = IPVS_GENL_NAME,
2745 .version = IPVS_GENL_VERSION,
2746 .maxattr = IPVS_CMD_MAX,
2747 .netnsok = true, /* Make ipvsadm to work on netns */
2748 };
2749
2750 /* Policy used for first-level command attributes */
2751 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2752 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2753 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2754 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2755 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2756 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2757 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2758 };
2759
2760 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2761 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2762 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2763 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2764 .len = IP_VS_IFNAME_MAXLEN },
2765 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2766 };
2767
2768 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2769 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2770 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2771 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2772 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2773 .len = sizeof(union nf_inet_addr) },
2774 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2775 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2776 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2777 .len = IP_VS_SCHEDNAME_MAXLEN },
2778 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2779 .len = IP_VS_PENAME_MAXLEN },
2780 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2781 .len = sizeof(struct ip_vs_flags) },
2782 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2783 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2784 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2785 };
2786
2787 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2788 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2789 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2790 .len = sizeof(union nf_inet_addr) },
2791 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2792 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2793 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2794 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2795 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2796 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2797 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2798 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2799 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2800 };
2801
2802 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2803 struct ip_vs_stats *stats)
2804 {
2805 struct ip_vs_stats_user ustats;
2806 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2807 if (!nl_stats)
2808 return -EMSGSIZE;
2809
2810 ip_vs_copy_stats(&ustats, stats);
2811
2812 if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, ustats.conns) ||
2813 nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, ustats.inpkts) ||
2814 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, ustats.outpkts) ||
2815 nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, ustats.inbytes) ||
2816 nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, ustats.outbytes) ||
2817 nla_put_u32(skb, IPVS_STATS_ATTR_CPS, ustats.cps) ||
2818 nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, ustats.inpps) ||
2819 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, ustats.outpps) ||
2820 nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, ustats.inbps) ||
2821 nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, ustats.outbps))
2822 goto nla_put_failure;
2823 nla_nest_end(skb, nl_stats);
2824
2825 return 0;
2826
2827 nla_put_failure:
2828 nla_nest_cancel(skb, nl_stats);
2829 return -EMSGSIZE;
2830 }
2831
2832 static int ip_vs_genl_fill_service(struct sk_buff *skb,
2833 struct ip_vs_service *svc)
2834 {
2835 struct nlattr *nl_service;
2836 struct ip_vs_flags flags = { .flags = svc->flags,
2837 .mask = ~0 };
2838
2839 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2840 if (!nl_service)
2841 return -EMSGSIZE;
2842
2843 if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2844 goto nla_put_failure;
2845 if (svc->fwmark) {
2846 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2847 goto nla_put_failure;
2848 } else {
2849 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2850 nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2851 nla_put_u16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2852 goto nla_put_failure;
2853 }
2854
2855 if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, svc->scheduler->name) ||
2856 (svc->pe &&
2857 nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, svc->pe->name)) ||
2858 nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2859 nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2860 nla_put_u32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
2861 goto nla_put_failure;
2862 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2863 goto nla_put_failure;
2864
2865 nla_nest_end(skb, nl_service);
2866
2867 return 0;
2868
2869 nla_put_failure:
2870 nla_nest_cancel(skb, nl_service);
2871 return -EMSGSIZE;
2872 }
2873
2874 static int ip_vs_genl_dump_service(struct sk_buff *skb,
2875 struct ip_vs_service *svc,
2876 struct netlink_callback *cb)
2877 {
2878 void *hdr;
2879
2880 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2881 &ip_vs_genl_family, NLM_F_MULTI,
2882 IPVS_CMD_NEW_SERVICE);
2883 if (!hdr)
2884 return -EMSGSIZE;
2885
2886 if (ip_vs_genl_fill_service(skb, svc) < 0)
2887 goto nla_put_failure;
2888
2889 return genlmsg_end(skb, hdr);
2890
2891 nla_put_failure:
2892 genlmsg_cancel(skb, hdr);
2893 return -EMSGSIZE;
2894 }
2895
2896 static int ip_vs_genl_dump_services(struct sk_buff *skb,
2897 struct netlink_callback *cb)
2898 {
2899 int idx = 0, i;
2900 int start = cb->args[0];
2901 struct ip_vs_service *svc;
2902 struct net *net = skb_sknet(skb);
2903
2904 mutex_lock(&__ip_vs_mutex);
2905 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2906 list_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
2907 if (++idx <= start || !net_eq(svc->net, net))
2908 continue;
2909 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2910 idx--;
2911 goto nla_put_failure;
2912 }
2913 }
2914 }
2915
2916 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2917 list_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
2918 if (++idx <= start || !net_eq(svc->net, net))
2919 continue;
2920 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2921 idx--;
2922 goto nla_put_failure;
2923 }
2924 }
2925 }
2926
2927 nla_put_failure:
2928 mutex_unlock(&__ip_vs_mutex);
2929 cb->args[0] = idx;
2930
2931 return skb->len;
2932 }
2933
2934 static int ip_vs_genl_parse_service(struct net *net,
2935 struct ip_vs_service_user_kern *usvc,
2936 struct nlattr *nla, int full_entry,
2937 struct ip_vs_service **ret_svc)
2938 {
2939 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
2940 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
2941 struct ip_vs_service *svc;
2942
2943 /* Parse mandatory identifying service fields first */
2944 if (nla == NULL ||
2945 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
2946 return -EINVAL;
2947
2948 nla_af = attrs[IPVS_SVC_ATTR_AF];
2949 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
2950 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
2951 nla_port = attrs[IPVS_SVC_ATTR_PORT];
2952 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
2953
2954 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
2955 return -EINVAL;
2956
2957 memset(usvc, 0, sizeof(*usvc));
2958
2959 usvc->af = nla_get_u16(nla_af);
2960 #ifdef CONFIG_IP_VS_IPV6
2961 if (usvc->af != AF_INET && usvc->af != AF_INET6)
2962 #else
2963 if (usvc->af != AF_INET)
2964 #endif
2965 return -EAFNOSUPPORT;
2966
2967 if (nla_fwmark) {
2968 usvc->protocol = IPPROTO_TCP;
2969 usvc->fwmark = nla_get_u32(nla_fwmark);
2970 } else {
2971 usvc->protocol = nla_get_u16(nla_protocol);
2972 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
2973 usvc->port = nla_get_u16(nla_port);
2974 usvc->fwmark = 0;
2975 }
2976
2977 if (usvc->fwmark)
2978 svc = __ip_vs_svc_fwm_find(net, usvc->af, usvc->fwmark);
2979 else
2980 svc = __ip_vs_service_find(net, usvc->af, usvc->protocol,
2981 &usvc->addr, usvc->port);
2982 *ret_svc = svc;
2983
2984 /* If a full entry was requested, check for the additional fields */
2985 if (full_entry) {
2986 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
2987 *nla_netmask;
2988 struct ip_vs_flags flags;
2989
2990 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
2991 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
2992 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
2993 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
2994 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
2995
2996 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
2997 return -EINVAL;
2998
2999 nla_memcpy(&flags, nla_flags, sizeof(flags));
3000
3001 /* prefill flags from service if it already exists */
3002 if (svc)
3003 usvc->flags = svc->flags;
3004
3005 /* set new flags from userland */
3006 usvc->flags = (usvc->flags & ~flags.mask) |
3007 (flags.flags & flags.mask);
3008 usvc->sched_name = nla_data(nla_sched);
3009 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3010 usvc->timeout = nla_get_u32(nla_timeout);
3011 usvc->netmask = nla_get_u32(nla_netmask);
3012 }
3013
3014 return 0;
3015 }
3016
3017 static struct ip_vs_service *ip_vs_genl_find_service(struct net *net,
3018 struct nlattr *nla)
3019 {
3020 struct ip_vs_service_user_kern usvc;
3021 struct ip_vs_service *svc;
3022 int ret;
3023
3024 ret = ip_vs_genl_parse_service(net, &usvc, nla, 0, &svc);
3025 return ret ? ERR_PTR(ret) : svc;
3026 }
3027
3028 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3029 {
3030 struct nlattr *nl_dest;
3031
3032 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3033 if (!nl_dest)
3034 return -EMSGSIZE;
3035
3036 if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3037 nla_put_u16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3038 nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3039 (atomic_read(&dest->conn_flags) &
3040 IP_VS_CONN_F_FWD_MASK)) ||
3041 nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3042 atomic_read(&dest->weight)) ||
3043 nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3044 nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3045 nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3046 atomic_read(&dest->activeconns)) ||
3047 nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3048 atomic_read(&dest->inactconns)) ||
3049 nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3050 atomic_read(&dest->persistconns)))
3051 goto nla_put_failure;
3052 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
3053 goto nla_put_failure;
3054
3055 nla_nest_end(skb, nl_dest);
3056
3057 return 0;
3058
3059 nla_put_failure:
3060 nla_nest_cancel(skb, nl_dest);
3061 return -EMSGSIZE;
3062 }
3063
3064 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3065 struct netlink_callback *cb)
3066 {
3067 void *hdr;
3068
3069 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
3070 &ip_vs_genl_family, NLM_F_MULTI,
3071 IPVS_CMD_NEW_DEST);
3072 if (!hdr)
3073 return -EMSGSIZE;
3074
3075 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3076 goto nla_put_failure;
3077
3078 return genlmsg_end(skb, hdr);
3079
3080 nla_put_failure:
3081 genlmsg_cancel(skb, hdr);
3082 return -EMSGSIZE;
3083 }
3084
3085 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3086 struct netlink_callback *cb)
3087 {
3088 int idx = 0;
3089 int start = cb->args[0];
3090 struct ip_vs_service *svc;
3091 struct ip_vs_dest *dest;
3092 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3093 struct net *net = skb_sknet(skb);
3094
3095 mutex_lock(&__ip_vs_mutex);
3096
3097 /* Try to find the service for which to dump destinations */
3098 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3099 IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3100 goto out_err;
3101
3102
3103 svc = ip_vs_genl_find_service(net, attrs[IPVS_CMD_ATTR_SERVICE]);
3104 if (IS_ERR(svc) || svc == NULL)
3105 goto out_err;
3106
3107 /* Dump the destinations */
3108 list_for_each_entry(dest, &svc->destinations, n_list) {
3109 if (++idx <= start)
3110 continue;
3111 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3112 idx--;
3113 goto nla_put_failure;
3114 }
3115 }
3116
3117 nla_put_failure:
3118 cb->args[0] = idx;
3119
3120 out_err:
3121 mutex_unlock(&__ip_vs_mutex);
3122
3123 return skb->len;
3124 }
3125
3126 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3127 struct nlattr *nla, int full_entry)
3128 {
3129 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3130 struct nlattr *nla_addr, *nla_port;
3131
3132 /* Parse mandatory identifying destination fields first */
3133 if (nla == NULL ||
3134 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3135 return -EINVAL;
3136
3137 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3138 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3139
3140 if (!(nla_addr && nla_port))
3141 return -EINVAL;
3142
3143 memset(udest, 0, sizeof(*udest));
3144
3145 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3146 udest->port = nla_get_u16(nla_port);
3147
3148 /* If a full entry was requested, check for the additional fields */
3149 if (full_entry) {
3150 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3151 *nla_l_thresh;
3152
3153 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3154 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3155 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3156 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3157
3158 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3159 return -EINVAL;
3160
3161 udest->conn_flags = nla_get_u32(nla_fwd)
3162 & IP_VS_CONN_F_FWD_MASK;
3163 udest->weight = nla_get_u32(nla_weight);
3164 udest->u_threshold = nla_get_u32(nla_u_thresh);
3165 udest->l_threshold = nla_get_u32(nla_l_thresh);
3166 }
3167
3168 return 0;
3169 }
3170
3171 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __be32 state,
3172 const char *mcast_ifn, __be32 syncid)
3173 {
3174 struct nlattr *nl_daemon;
3175
3176 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3177 if (!nl_daemon)
3178 return -EMSGSIZE;
3179
3180 if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3181 nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn) ||
3182 nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid))
3183 goto nla_put_failure;
3184 nla_nest_end(skb, nl_daemon);
3185
3186 return 0;
3187
3188 nla_put_failure:
3189 nla_nest_cancel(skb, nl_daemon);
3190 return -EMSGSIZE;
3191 }
3192
3193 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __be32 state,
3194 const char *mcast_ifn, __be32 syncid,
3195 struct netlink_callback *cb)
3196 {
3197 void *hdr;
3198 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
3199 &ip_vs_genl_family, NLM_F_MULTI,
3200 IPVS_CMD_NEW_DAEMON);
3201 if (!hdr)
3202 return -EMSGSIZE;
3203
3204 if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
3205 goto nla_put_failure;
3206
3207 return genlmsg_end(skb, hdr);
3208
3209 nla_put_failure:
3210 genlmsg_cancel(skb, hdr);
3211 return -EMSGSIZE;
3212 }
3213
3214 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3215 struct netlink_callback *cb)
3216 {
3217 struct net *net = skb_sknet(skb);
3218 struct netns_ipvs *ipvs = net_ipvs(net);
3219
3220 mutex_lock(&ipvs->sync_mutex);
3221 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3222 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3223 ipvs->master_mcast_ifn,
3224 ipvs->master_syncid, cb) < 0)
3225 goto nla_put_failure;
3226
3227 cb->args[0] = 1;
3228 }
3229
3230 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3231 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3232 ipvs->backup_mcast_ifn,
3233 ipvs->backup_syncid, cb) < 0)
3234 goto nla_put_failure;
3235
3236 cb->args[1] = 1;
3237 }
3238
3239 nla_put_failure:
3240 mutex_unlock(&ipvs->sync_mutex);
3241
3242 return skb->len;
3243 }
3244
3245 static int ip_vs_genl_new_daemon(struct net *net, struct nlattr **attrs)
3246 {
3247 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3248 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3249 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3250 return -EINVAL;
3251
3252 return start_sync_thread(net,
3253 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
3254 nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3255 nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3256 }
3257
3258 static int ip_vs_genl_del_daemon(struct net *net, struct nlattr **attrs)
3259 {
3260 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3261 return -EINVAL;
3262
3263 return stop_sync_thread(net,
3264 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3265 }
3266
3267 static int ip_vs_genl_set_config(struct net *net, struct nlattr **attrs)
3268 {
3269 struct ip_vs_timeout_user t;
3270
3271 __ip_vs_get_timeouts(net, &t);
3272
3273 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3274 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3275
3276 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3277 t.tcp_fin_timeout =
3278 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3279
3280 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3281 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3282
3283 return ip_vs_set_timeout(net, &t);
3284 }
3285
3286 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3287 {
3288 int ret = 0, cmd;
3289 struct net *net;
3290 struct netns_ipvs *ipvs;
3291
3292 net = skb_sknet(skb);
3293 ipvs = net_ipvs(net);
3294 cmd = info->genlhdr->cmd;
3295
3296 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3297 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3298
3299 mutex_lock(&ipvs->sync_mutex);
3300 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3301 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3302 info->attrs[IPVS_CMD_ATTR_DAEMON],
3303 ip_vs_daemon_policy)) {
3304 ret = -EINVAL;
3305 goto out;
3306 }
3307
3308 if (cmd == IPVS_CMD_NEW_DAEMON)
3309 ret = ip_vs_genl_new_daemon(net, daemon_attrs);
3310 else
3311 ret = ip_vs_genl_del_daemon(net, daemon_attrs);
3312 out:
3313 mutex_unlock(&ipvs->sync_mutex);
3314 }
3315 return ret;
3316 }
3317
3318 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3319 {
3320 struct ip_vs_service *svc = NULL;
3321 struct ip_vs_service_user_kern usvc;
3322 struct ip_vs_dest_user_kern udest;
3323 int ret = 0, cmd;
3324 int need_full_svc = 0, need_full_dest = 0;
3325 struct net *net;
3326
3327 net = skb_sknet(skb);
3328 cmd = info->genlhdr->cmd;
3329
3330 mutex_lock(&__ip_vs_mutex);
3331
3332 if (cmd == IPVS_CMD_FLUSH) {
3333 ret = ip_vs_flush(net);
3334 goto out;
3335 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3336 ret = ip_vs_genl_set_config(net, info->attrs);
3337 goto out;
3338 } else if (cmd == IPVS_CMD_ZERO &&
3339 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3340 ret = ip_vs_zero_all(net);
3341 goto out;
3342 }
3343
3344 /* All following commands require a service argument, so check if we
3345 * received a valid one. We need a full service specification when
3346 * adding / editing a service. Only identifying members otherwise. */
3347 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3348 need_full_svc = 1;
3349
3350 ret = ip_vs_genl_parse_service(net, &usvc,
3351 info->attrs[IPVS_CMD_ATTR_SERVICE],
3352 need_full_svc, &svc);
3353 if (ret)
3354 goto out;
3355
3356 /* Unless we're adding a new service, the service must already exist */
3357 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3358 ret = -ESRCH;
3359 goto out;
3360 }
3361
3362 /* Destination commands require a valid destination argument. For
3363 * adding / editing a destination, we need a full destination
3364 * specification. */
3365 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3366 cmd == IPVS_CMD_DEL_DEST) {
3367 if (cmd != IPVS_CMD_DEL_DEST)
3368 need_full_dest = 1;
3369
3370 ret = ip_vs_genl_parse_dest(&udest,
3371 info->attrs[IPVS_CMD_ATTR_DEST],
3372 need_full_dest);
3373 if (ret)
3374 goto out;
3375 }
3376
3377 switch (cmd) {
3378 case IPVS_CMD_NEW_SERVICE:
3379 if (svc == NULL)
3380 ret = ip_vs_add_service(net, &usvc, &svc);
3381 else
3382 ret = -EEXIST;
3383 break;
3384 case IPVS_CMD_SET_SERVICE:
3385 ret = ip_vs_edit_service(svc, &usvc);
3386 break;
3387 case IPVS_CMD_DEL_SERVICE:
3388 ret = ip_vs_del_service(svc);
3389 /* do not use svc, it can be freed */
3390 break;
3391 case IPVS_CMD_NEW_DEST:
3392 ret = ip_vs_add_dest(svc, &udest);
3393 break;
3394 case IPVS_CMD_SET_DEST:
3395 ret = ip_vs_edit_dest(svc, &udest);
3396 break;
3397 case IPVS_CMD_DEL_DEST:
3398 ret = ip_vs_del_dest(svc, &udest);
3399 break;
3400 case IPVS_CMD_ZERO:
3401 ret = ip_vs_zero_service(svc);
3402 break;
3403 default:
3404 ret = -EINVAL;
3405 }
3406
3407 out:
3408 mutex_unlock(&__ip_vs_mutex);
3409
3410 return ret;
3411 }
3412
3413 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3414 {
3415 struct sk_buff *msg;
3416 void *reply;
3417 int ret, cmd, reply_cmd;
3418 struct net *net;
3419
3420 net = skb_sknet(skb);
3421 cmd = info->genlhdr->cmd;
3422
3423 if (cmd == IPVS_CMD_GET_SERVICE)
3424 reply_cmd = IPVS_CMD_NEW_SERVICE;
3425 else if (cmd == IPVS_CMD_GET_INFO)
3426 reply_cmd = IPVS_CMD_SET_INFO;
3427 else if (cmd == IPVS_CMD_GET_CONFIG)
3428 reply_cmd = IPVS_CMD_SET_CONFIG;
3429 else {
3430 pr_err("unknown Generic Netlink command\n");
3431 return -EINVAL;
3432 }
3433
3434 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3435 if (!msg)
3436 return -ENOMEM;
3437
3438 mutex_lock(&__ip_vs_mutex);
3439
3440 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3441 if (reply == NULL)
3442 goto nla_put_failure;
3443
3444 switch (cmd) {
3445 case IPVS_CMD_GET_SERVICE:
3446 {
3447 struct ip_vs_service *svc;
3448
3449 svc = ip_vs_genl_find_service(net,
3450 info->attrs[IPVS_CMD_ATTR_SERVICE]);
3451 if (IS_ERR(svc)) {
3452 ret = PTR_ERR(svc);
3453 goto out_err;
3454 } else if (svc) {
3455 ret = ip_vs_genl_fill_service(msg, svc);
3456 if (ret)
3457 goto nla_put_failure;
3458 } else {
3459 ret = -ESRCH;
3460 goto out_err;
3461 }
3462
3463 break;
3464 }
3465
3466 case IPVS_CMD_GET_CONFIG:
3467 {
3468 struct ip_vs_timeout_user t;
3469
3470 __ip_vs_get_timeouts(net, &t);
3471 #ifdef CONFIG_IP_VS_PROTO_TCP
3472 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3473 t.tcp_timeout) ||
3474 nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3475 t.tcp_fin_timeout))
3476 goto nla_put_failure;
3477 #endif
3478 #ifdef CONFIG_IP_VS_PROTO_UDP
3479 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3480 goto nla_put_failure;
3481 #endif
3482
3483 break;
3484 }
3485
3486 case IPVS_CMD_GET_INFO:
3487 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3488 IP_VS_VERSION_CODE) ||
3489 nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3490 ip_vs_conn_tab_size))
3491 goto nla_put_failure;
3492 break;
3493 }
3494
3495 genlmsg_end(msg, reply);
3496 ret = genlmsg_reply(msg, info);
3497 goto out;
3498
3499 nla_put_failure:
3500 pr_err("not enough space in Netlink message\n");
3501 ret = -EMSGSIZE;
3502
3503 out_err:
3504 nlmsg_free(msg);
3505 out:
3506 mutex_unlock(&__ip_vs_mutex);
3507
3508 return ret;
3509 }
3510
3511
3512 static struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3513 {
3514 .cmd = IPVS_CMD_NEW_SERVICE,
3515 .flags = GENL_ADMIN_PERM,
3516 .policy = ip_vs_cmd_policy,
3517 .doit = ip_vs_genl_set_cmd,
3518 },
3519 {
3520 .cmd = IPVS_CMD_SET_SERVICE,
3521 .flags = GENL_ADMIN_PERM,
3522 .policy = ip_vs_cmd_policy,
3523 .doit = ip_vs_genl_set_cmd,
3524 },
3525 {
3526 .cmd = IPVS_CMD_DEL_SERVICE,
3527 .flags = GENL_ADMIN_PERM,
3528 .policy = ip_vs_cmd_policy,
3529 .doit = ip_vs_genl_set_cmd,
3530 },
3531 {
3532 .cmd = IPVS_CMD_GET_SERVICE,
3533 .flags = GENL_ADMIN_PERM,
3534 .doit = ip_vs_genl_get_cmd,
3535 .dumpit = ip_vs_genl_dump_services,
3536 .policy = ip_vs_cmd_policy,
3537 },
3538 {
3539 .cmd = IPVS_CMD_NEW_DEST,
3540 .flags = GENL_ADMIN_PERM,
3541 .policy = ip_vs_cmd_policy,
3542 .doit = ip_vs_genl_set_cmd,
3543 },
3544 {
3545 .cmd = IPVS_CMD_SET_DEST,
3546 .flags = GENL_ADMIN_PERM,
3547 .policy = ip_vs_cmd_policy,
3548 .doit = ip_vs_genl_set_cmd,
3549 },
3550 {
3551 .cmd = IPVS_CMD_DEL_DEST,
3552 .flags = GENL_ADMIN_PERM,
3553 .policy = ip_vs_cmd_policy,
3554 .doit = ip_vs_genl_set_cmd,
3555 },
3556 {
3557 .cmd = IPVS_CMD_GET_DEST,
3558 .flags = GENL_ADMIN_PERM,
3559 .policy = ip_vs_cmd_policy,
3560 .dumpit = ip_vs_genl_dump_dests,
3561 },
3562 {
3563 .cmd = IPVS_CMD_NEW_DAEMON,
3564 .flags = GENL_ADMIN_PERM,
3565 .policy = ip_vs_cmd_policy,
3566 .doit = ip_vs_genl_set_daemon,
3567 },
3568 {
3569 .cmd = IPVS_CMD_DEL_DAEMON,
3570 .flags = GENL_ADMIN_PERM,
3571 .policy = ip_vs_cmd_policy,
3572 .doit = ip_vs_genl_set_daemon,
3573 },
3574 {
3575 .cmd = IPVS_CMD_GET_DAEMON,
3576 .flags = GENL_ADMIN_PERM,
3577 .dumpit = ip_vs_genl_dump_daemons,
3578 },
3579 {
3580 .cmd = IPVS_CMD_SET_CONFIG,
3581 .flags = GENL_ADMIN_PERM,
3582 .policy = ip_vs_cmd_policy,
3583 .doit = ip_vs_genl_set_cmd,
3584 },
3585 {
3586 .cmd = IPVS_CMD_GET_CONFIG,
3587 .flags = GENL_ADMIN_PERM,
3588 .doit = ip_vs_genl_get_cmd,
3589 },
3590 {
3591 .cmd = IPVS_CMD_GET_INFO,
3592 .flags = GENL_ADMIN_PERM,
3593 .doit = ip_vs_genl_get_cmd,
3594 },
3595 {
3596 .cmd = IPVS_CMD_ZERO,
3597 .flags = GENL_ADMIN_PERM,
3598 .policy = ip_vs_cmd_policy,
3599 .doit = ip_vs_genl_set_cmd,
3600 },
3601 {
3602 .cmd = IPVS_CMD_FLUSH,
3603 .flags = GENL_ADMIN_PERM,
3604 .doit = ip_vs_genl_set_cmd,
3605 },
3606 };
3607
3608 static int __init ip_vs_genl_register(void)
3609 {
3610 return genl_register_family_with_ops(&ip_vs_genl_family,
3611 ip_vs_genl_ops, ARRAY_SIZE(ip_vs_genl_ops));
3612 }
3613
3614 static void ip_vs_genl_unregister(void)
3615 {
3616 genl_unregister_family(&ip_vs_genl_family);
3617 }
3618
3619 /* End of Generic Netlink interface definitions */
3620
3621 /*
3622 * per netns intit/exit func.
3623 */
3624 #ifdef CONFIG_SYSCTL
3625 int __net_init ip_vs_control_net_init_sysctl(struct net *net)
3626 {
3627 int idx;
3628 struct netns_ipvs *ipvs = net_ipvs(net);
3629 struct ctl_table *tbl;
3630
3631 atomic_set(&ipvs->dropentry, 0);
3632 spin_lock_init(&ipvs->dropentry_lock);
3633 spin_lock_init(&ipvs->droppacket_lock);
3634 spin_lock_init(&ipvs->securetcp_lock);
3635
3636 if (!net_eq(net, &init_net)) {
3637 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3638 if (tbl == NULL)
3639 return -ENOMEM;
3640 } else
3641 tbl = vs_vars;
3642 /* Initialize sysctl defaults */
3643 idx = 0;
3644 ipvs->sysctl_amemthresh = 1024;
3645 tbl[idx++].data = &ipvs->sysctl_amemthresh;
3646 ipvs->sysctl_am_droprate = 10;
3647 tbl[idx++].data = &ipvs->sysctl_am_droprate;
3648 tbl[idx++].data = &ipvs->sysctl_drop_entry;
3649 tbl[idx++].data = &ipvs->sysctl_drop_packet;
3650 #ifdef CONFIG_IP_VS_NFCT
3651 tbl[idx++].data = &ipvs->sysctl_conntrack;
3652 #endif
3653 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3654 ipvs->sysctl_snat_reroute = 1;
3655 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3656 ipvs->sysctl_sync_ver = 1;
3657 tbl[idx++].data = &ipvs->sysctl_sync_ver;
3658 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3659 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3660 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
3661 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3662 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
3663 tbl[idx].data = &ipvs->sysctl_sync_threshold;
3664 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3665 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3666
3667
3668 ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
3669 if (ipvs->sysctl_hdr == NULL) {
3670 if (!net_eq(net, &init_net))
3671 kfree(tbl);
3672 return -ENOMEM;
3673 }
3674 ip_vs_start_estimator(net, &ipvs->tot_stats);
3675 ipvs->sysctl_tbl = tbl;
3676 /* Schedule defense work */
3677 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3678 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
3679
3680 return 0;
3681 }
3682
3683 void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net)
3684 {
3685 struct netns_ipvs *ipvs = net_ipvs(net);
3686
3687 cancel_delayed_work_sync(&ipvs->defense_work);
3688 cancel_work_sync(&ipvs->defense_work.work);
3689 unregister_net_sysctl_table(ipvs->sysctl_hdr);
3690 }
3691
3692 #else
3693
3694 int __net_init ip_vs_control_net_init_sysctl(struct net *net) { return 0; }
3695 void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net) { }
3696
3697 #endif
3698
3699 static struct notifier_block ip_vs_dst_notifier = {
3700 .notifier_call = ip_vs_dst_event,
3701 };
3702
3703 int __net_init ip_vs_control_net_init(struct net *net)
3704 {
3705 int idx;
3706 struct netns_ipvs *ipvs = net_ipvs(net);
3707
3708 rwlock_init(&ipvs->rs_lock);
3709
3710 /* Initialize rs_table */
3711 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
3712 INIT_LIST_HEAD(&ipvs->rs_table[idx]);
3713
3714 INIT_LIST_HEAD(&ipvs->dest_trash);
3715 atomic_set(&ipvs->ftpsvc_counter, 0);
3716 atomic_set(&ipvs->nullsvc_counter, 0);
3717
3718 /* procfs stats */
3719 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
3720 if (!ipvs->tot_stats.cpustats)
3721 return -ENOMEM;
3722
3723 spin_lock_init(&ipvs->tot_stats.lock);
3724
3725 proc_net_fops_create(net, "ip_vs", 0, &ip_vs_info_fops);
3726 proc_net_fops_create(net, "ip_vs_stats", 0, &ip_vs_stats_fops);
3727 proc_net_fops_create(net, "ip_vs_stats_percpu", 0,
3728 &ip_vs_stats_percpu_fops);
3729
3730 if (ip_vs_control_net_init_sysctl(net))
3731 goto err;
3732
3733 return 0;
3734
3735 err:
3736 free_percpu(ipvs->tot_stats.cpustats);
3737 return -ENOMEM;
3738 }
3739
3740 void __net_exit ip_vs_control_net_cleanup(struct net *net)
3741 {
3742 struct netns_ipvs *ipvs = net_ipvs(net);
3743
3744 ip_vs_trash_cleanup(net);
3745 ip_vs_stop_estimator(net, &ipvs->tot_stats);
3746 ip_vs_control_net_cleanup_sysctl(net);
3747 proc_net_remove(net, "ip_vs_stats_percpu");
3748 proc_net_remove(net, "ip_vs_stats");
3749 proc_net_remove(net, "ip_vs");
3750 free_percpu(ipvs->tot_stats.cpustats);
3751 }
3752
3753 int __init ip_vs_register_nl_ioctl(void)
3754 {
3755 int ret;
3756
3757 ret = nf_register_sockopt(&ip_vs_sockopts);
3758 if (ret) {
3759 pr_err("cannot register sockopt.\n");
3760 goto err_sock;
3761 }
3762
3763 ret = ip_vs_genl_register();
3764 if (ret) {
3765 pr_err("cannot register Generic Netlink interface.\n");
3766 goto err_genl;
3767 }
3768 return 0;
3769
3770 err_genl:
3771 nf_unregister_sockopt(&ip_vs_sockopts);
3772 err_sock:
3773 return ret;
3774 }
3775
3776 void ip_vs_unregister_nl_ioctl(void)
3777 {
3778 ip_vs_genl_unregister();
3779 nf_unregister_sockopt(&ip_vs_sockopts);
3780 }
3781
3782 int __init ip_vs_control_init(void)
3783 {
3784 int idx;
3785 int ret;
3786
3787 EnterFunction(2);
3788
3789 /* Initialize svc_table, ip_vs_svc_fwm_table, rs_table */
3790 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
3791 INIT_LIST_HEAD(&ip_vs_svc_table[idx]);
3792 INIT_LIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3793 }
3794
3795 smp_wmb(); /* Do we really need it now ? */
3796
3797 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
3798 if (ret < 0)
3799 return ret;
3800
3801 LeaveFunction(2);
3802 return 0;
3803 }
3804
3805
3806 void ip_vs_control_cleanup(void)
3807 {
3808 EnterFunction(2);
3809 unregister_netdevice_notifier(&ip_vs_dst_notifier);
3810 LeaveFunction(2);
3811 }
Linux kernel - Deliverables
This page took 0.108554 seconds and 5 git commands to generate.