projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ipvs: wakeup master thread
[deliverable/linux.git] / net / netfilter / ipvs / ip_vs_ctl.c
CommitLineData
1da177e4
LT		 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 *
1da177e4
LT		 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
9aada7ac
HE		 21#define KMSG_COMPONENT "IPVS"
22#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
23
1da177e4
LT		 24#include <linux/module.h>
25#include <linux/init.h>
26#include <linux/types.h>
4fc268d2 27#include <linux/capability.h>
1da177e4
LT		 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>
1da177e4 33#include <linux/seq_file.h>
5a0e3ad6 34#include <linux/slab.h>
1da177e4
LT		 35
36#include <linux/netfilter.h>
37#include <linux/netfilter_ipv4.h>
14cc3e2b 38#include <linux/mutex.h>
1da177e4 39
457c4cbc 40#include <net/net_namespace.h>
9330419d 41#include <linux/nsproxy.h>
1da177e4 42#include <net/ip.h>
09571c7a
VB		 43#ifdef CONFIG_IP_VS_IPV6
44#include <net/ipv6.h>
45#include <net/ip6_route.h>
46#endif
14c85021 47#include <net/route.h>
1da177e4 48#include <net/sock.h>
9a812198 49#include <net/genetlink.h>
1da177e4
LT		 50
51#include <asm/uaccess.h>
52
53#include <net/ip_vs.h>
54
55/* semaphore for IPVS sockopts. And, [gs]etsockopt may sleep. */
14cc3e2b 56static DEFINE_MUTEX(__ip_vs_mutex);
1da177e4
LT		 57
58/* lock for service table */
59static DEFINE_RWLOCK(__ip_vs_svc_lock);
60
1da177e4 61/* sysctl variables */
1da177e4
LT		 62
63#ifdef CONFIG_IP_VS_DEBUG
64static int sysctl_ip_vs_debug_level = 0;
65
66int ip_vs_get_debug_level(void)
67{
68 return sysctl_ip_vs_debug_level;
69}
70#endif
71
7a4f0761
HS		 72
73/* Protos */
74static void __ip_vs_del_service(struct ip_vs_service *svc);
75
76
09571c7a
VB		 77#ifdef CONFIG_IP_VS_IPV6
78/* Taken from rt6_fill_node() in net/ipv6/route.c, is there a better way? */
4a98480b
HS		 79static int __ip_vs_addr_is_local_v6(struct net *net,
80 const struct in6_addr *addr)
09571c7a
VB		 81{
82 struct rt6_info *rt;
4c9483b2
DM		 83 struct flowi6 fl6 = {
84 .daddr = *addr,
09571c7a
VB		 85 };
86
4c9483b2 87 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
d1918542 88 if (rt && rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
4c9483b2 89 return 1;
09571c7a
VB		 90
91 return 0;
92}
93#endif
14e40546
SH		 94
95#ifdef CONFIG_SYSCTL
1da177e4 96/*
af9debd4
JA		 97 * update_defense_level is called from keventd and from sysctl,
98 * so it needs to protect itself from softirqs
1da177e4 99 */
9330419d 100static void update_defense_level(struct netns_ipvs *ipvs)
1da177e4
LT		 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
a0840e2e 116 nomem = (availmem < ipvs->sysctl_amemthresh);
1da177e4 117
af9debd4
JA		 118 local_bh_disable();
119
1da177e4 120 /* drop_entry */
a0840e2e
HS		 121 spin_lock(&ipvs->dropentry_lock);
122 switch (ipvs->sysctl_drop_entry) {
1da177e4 123 case 0:
a0840e2e 124 atomic_set(&ipvs->dropentry, 0);
1da177e4
LT		 125 break;
126 case 1:
127 if (nomem) {
a0840e2e
HS		 128 atomic_set(&ipvs->dropentry, 1);
129 ipvs->sysctl_drop_entry = 2;
1da177e4 130 } else {
a0840e2e 131 atomic_set(&ipvs->dropentry, 0);
1da177e4
LT		 132 }
133 break;
134 case 2:
135 if (nomem) {
a0840e2e 136 atomic_set(&ipvs->dropentry, 1);
1da177e4 137 } else {
a0840e2e
HS		 138 atomic_set(&ipvs->dropentry, 0);
139 ipvs->sysctl_drop_entry = 1;
1da177e4
LT		 140 };
141 break;
142 case 3:
a0840e2e 143 atomic_set(&ipvs->dropentry, 1);
1da177e4
LT		 144 break;
145 }
a0840e2e 146 spin_unlock(&ipvs->dropentry_lock);
1da177e4
LT		 147
148 /* drop_packet */
a0840e2e
HS		 149 spin_lock(&ipvs->droppacket_lock);
150 switch (ipvs->sysctl_drop_packet) {
1da177e4 151 case 0:
a0840e2e 152 ipvs->drop_rate = 0;
1da177e4
LT		 153 break;
154 case 1:
155 if (nomem) {
a0840e2e
HS		 156 ipvs->drop_rate = ipvs->drop_counter
157 = ipvs->sysctl_amemthresh /
158 (ipvs->sysctl_amemthresh-availmem);
159 ipvs->sysctl_drop_packet = 2;
1da177e4 160 } else {
a0840e2e 161 ipvs->drop_rate = 0;
1da177e4
LT		 162 }
163 break;
164 case 2:
165 if (nomem) {
a0840e2e
HS		 166 ipvs->drop_rate = ipvs->drop_counter
167 = ipvs->sysctl_amemthresh /
168 (ipvs->sysctl_amemthresh-availmem);
1da177e4 169 } else {
a0840e2e
HS		 170 ipvs->drop_rate = 0;
171 ipvs->sysctl_drop_packet = 1;
1da177e4
LT		 172 }
173 break;
174 case 3:
a0840e2e 175 ipvs->drop_rate = ipvs->sysctl_am_droprate;
1da177e4
LT		 176 break;
177 }
a0840e2e 178 spin_unlock(&ipvs->droppacket_lock);
1da177e4
LT		 179
180 /* secure_tcp */
a0840e2e
HS		 181 spin_lock(&ipvs->securetcp_lock);
182 switch (ipvs->sysctl_secure_tcp) {
1da177e4
LT		 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;
a0840e2e 191 ipvs->sysctl_secure_tcp = 2;
1da177e4
LT		 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;
a0840e2e 204 ipvs->sysctl_secure_tcp = 1;
1da177e4
LT		 205 }
206 break;
207 case 3:
208 if (old_secure_tcp < 2)
209 to_change = 1;
210 break;
211 }
a0840e2e 212 old_secure_tcp = ipvs->sysctl_secure_tcp;
1da177e4 213 if (to_change >= 0)
9330419d 214 ip_vs_protocol_timeout_change(ipvs,
a0840e2e
HS		 215 ipvs->sysctl_secure_tcp > 1);
216 spin_unlock(&ipvs->securetcp_lock);
af9debd4
JA		 217
218 local_bh_enable();
1da177e4
LT		 219}
220
221
222/*
223 * Timer for checking the defense
224 */
225#define DEFENSE_TIMER_PERIOD 1*HZ
1da177e4 226
c4028958 227static void defense_work_handler(struct work_struct *work)
1da177e4 228{
f6340ee0
HS		 229 struct netns_ipvs *ipvs =
230 container_of(work, struct netns_ipvs, defense_work.work);
9330419d
HS		 231
232 update_defense_level(ipvs);
a0840e2e 233 if (atomic_read(&ipvs->dropentry))
f6340ee0
HS		 234 ip_vs_random_dropentry(ipvs->net);
235 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
1da177e4 236}
14e40546 237#endif
1da177e4
LT		 238
239int
240ip_vs_use_count_inc(void)
241{
242 return try_module_get(THIS_MODULE);
243}
244
245void
246ip_vs_use_count_dec(void)
247{
248 module_put(THIS_MODULE);
249}
250
251
252/*
253 * 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> */
260static struct list_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE];
261/* the service table hashed by fwmark */
262static struct list_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE];
263
1da177e4
LT		 264
265/*
266 * Returns hash value for virtual service
267 */
95c96174
ED		 268static inline unsigned int
269ip_vs_svc_hashkey(struct net *net, int af, unsigned int proto,
fc723250 270 const union nf_inet_addr *addr, __be16 port)
1da177e4 271{
95c96174 272 register unsigned int porth = ntohs(port);
b18610de 273 __be32 addr_fold = addr->ip;
1da177e4 274
b18610de
JV		 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
fc723250 280 addr_fold ^= ((size_t)net>>8);
b18610de
JV		 281
282 return (proto^ntohl(addr_fold)^(porth>>IP_VS_SVC_TAB_BITS)^porth)
1da177e4
LT		 283 & IP_VS_SVC_TAB_MASK;
284}
285
286/*
287 * Returns hash value of fwmark for virtual service lookup
288 */
95c96174 289static inline unsigned int ip_vs_svc_fwm_hashkey(struct net *net, __u32 fwmark)
1da177e4 290{
fc723250 291 return (((size_t)net>>8) ^ fwmark) & IP_VS_SVC_TAB_MASK;
1da177e4
LT		 292}
293
294/*
fc723250 295 * Hashes a service in the ip_vs_svc_table by <netns,proto,addr,port>
1da177e4
LT		 296 * or in the ip_vs_svc_fwm_table by fwmark.
297 * Should be called with locked tables.
298 */
299static int ip_vs_svc_hash(struct ip_vs_service *svc)
300{
95c96174 301 unsigned int hash;
1da177e4
LT		 302
303 if (svc->flags & IP_VS_SVC_F_HASHED) {
1e3e238e
HE		 304 pr_err("%s(): request for already hashed, called from %pF\n",
305 __func__, __builtin_return_address(0));
1da177e4
LT		 306 return 0;
307 }
308
309 if (svc->fwmark == 0) {
310 /*
fc723250 311 * Hash it by <netns,protocol,addr,port> in ip_vs_svc_table
1da177e4 312 */
fc723250
HS		 313 hash = ip_vs_svc_hashkey(svc->net, svc->af, svc->protocol,
314 &svc->addr, svc->port);
1da177e4
LT		 315 list_add(&svc->s_list, &ip_vs_svc_table[hash]);
316 } else {
317 /*
fc723250 318 * Hash it by fwmark in svc_fwm_table
1da177e4 319 */
fc723250 320 hash = ip_vs_svc_fwm_hashkey(svc->net, svc->fwmark);
1da177e4
LT		 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/*
fc723250 332 * Unhashes a service from svc_table / svc_fwm_table.
1da177e4
LT		 333 * Should be called with locked tables.
334 */
335static int ip_vs_svc_unhash(struct ip_vs_service *svc)
336{
337 if (!(svc->flags & IP_VS_SVC_F_HASHED)) {
1e3e238e
HE		 338 pr_err("%s(): request for unhash flagged, called from %pF\n",
339 __func__, __builtin_return_address(0));
1da177e4
LT		 340 return 0;
341 }
342
343 if (svc->fwmark == 0) {
fc723250 344 /* Remove it from the svc_table table */
1da177e4
LT		 345 list_del(&svc->s_list);
346 } else {
fc723250 347 /* Remove it from the svc_fwm_table table */
1da177e4
LT		 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/*
fc723250 358 * Get service by {netns, proto,addr,port} in the service table.
1da177e4 359 */
b18610de 360static inline struct ip_vs_service *
fc723250
HS		 361__ip_vs_service_find(struct net *net, int af, __u16 protocol,
362 const union nf_inet_addr *vaddr, __be16 vport)
1da177e4 363{
95c96174 364 unsigned int hash;
1da177e4
LT		 365 struct ip_vs_service *svc;
366
367 /* Check for "full" addressed entries */
fc723250 368 hash = ip_vs_svc_hashkey(net, af, protocol, vaddr, vport);
1da177e4
LT		 369
370 list_for_each_entry(svc, &ip_vs_svc_table[hash], s_list){
b18610de
JV		 371 if ((svc->af == af)
372 && ip_vs_addr_equal(af, &svc->addr, vaddr)
1da177e4 373 && (svc->port == vport)
fc723250
HS		 374 && (svc->protocol == protocol)
375 && net_eq(svc->net, net)) {
1da177e4 376 /* HIT */
1da177e4
LT		 377 return svc;
378 }
379 }
380
381 return NULL;
382}
383
384
385/*
386 * Get service by {fwmark} in the service table.
387 */
b18610de 388static inline struct ip_vs_service *
fc723250 389__ip_vs_svc_fwm_find(struct net *net, int af, __u32 fwmark)
1da177e4 390{
95c96174 391 unsigned int hash;
1da177e4
LT		 392 struct ip_vs_service *svc;
393
394 /* Check for fwmark addressed entries */
fc723250 395 hash = ip_vs_svc_fwm_hashkey(net, fwmark);
1da177e4
LT		 396
397 list_for_each_entry(svc, &ip_vs_svc_fwm_table[hash], f_list) {
fc723250
HS		 398 if (svc->fwmark == fwmark && svc->af == af
399 && net_eq(svc->net, net)) {
1da177e4 400 /* HIT */
1da177e4
LT		 401 return svc;
402 }
403 }
404
405 return NULL;
406}
407
408struct ip_vs_service *
fc723250 409ip_vs_service_get(struct net *net, int af, __u32 fwmark, __u16 protocol,
3c2e0505 410 const union nf_inet_addr *vaddr, __be16 vport)
1da177e4
LT		 411{
412 struct ip_vs_service *svc;
763f8d0e 413 struct netns_ipvs *ipvs = net_ipvs(net);
3c2e0505 414
1da177e4
LT		 415 read_lock(&__ip_vs_svc_lock);
416
417 /*
418 * Check the table hashed by fwmark first
419 */
097fc76a
JA		 420 if (fwmark) {
421 svc = __ip_vs_svc_fwm_find(net, af, fwmark);
422 if (svc)
423 goto out;
424 }
1da177e4
LT		 425
426 /*
427 * Check the table hashed by <protocol,addr,port>
428 * for "full" addressed entries
429 */
fc723250 430 svc = __ip_vs_service_find(net, af, protocol, vaddr, vport);
1da177e4
LT		 431
432 if (svc == NULL
433 && protocol == IPPROTO_TCP
763f8d0e 434 && atomic_read(&ipvs->ftpsvc_counter)
1da177e4
LT		 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 */
fc723250 440 svc = __ip_vs_service_find(net, af, protocol, vaddr, FTPPORT);
1da177e4
LT		 441 }
442
443 if (svc == NULL
763f8d0e 444 && atomic_read(&ipvs->nullsvc_counter)) {
1da177e4
LT		 445 /*
446 * Check if the catch-all port (port zero) exists
447 */
fc723250 448 svc = __ip_vs_service_find(net, af, protocol, vaddr, 0);
1da177e4
LT		 449 }
450
451 out:
26c15cfd
JA		 452 if (svc)
453 atomic_inc(&svc->usecnt);
1da177e4
LT		 454 read_unlock(&__ip_vs_svc_lock);
455
3c2e0505
JV		 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");
1da177e4
LT		 460
461 return svc;
462}
463
464
465static inline void
466__ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
467{
468 atomic_inc(&svc->refcnt);
469 dest->svc = svc;
470}
471
26c15cfd 472static void
1da177e4
LT		 473__ip_vs_unbind_svc(struct ip_vs_dest *dest)
474{
475 struct ip_vs_service *svc = dest->svc;
476
477 dest->svc = NULL;
26c15cfd
JA		 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));
b17fc996 483 free_percpu(svc->stats.cpustats);
1da177e4 484 kfree(svc);
26c15cfd 485 }
1da177e4
LT		 486}
487
488
489/*
490 * Returns hash value for real service
491 */
95c96174 492static inline unsigned int ip_vs_rs_hashkey(int af,
7937df15
JV		 493 const union nf_inet_addr *addr,
494 __be16 port)
1da177e4 495{
95c96174 496 register unsigned int porth = ntohs(port);
7937df15
JV		 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
1da177e4 504
7937df15 505 return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
1da177e4
LT		 506 & IP_VS_RTAB_MASK;
507}
508
509/*
fc723250 510 * Hashes ip_vs_dest in rs_table by <proto,addr,port>.
1da177e4
LT		 511 * should be called with locked tables.
512 */
fc723250 513static int ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
1da177e4 514{
95c96174 515 unsigned int hash;
1da177e4
LT		 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 */
7937df15
JV		 525 hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
526
fc723250 527 list_add(&dest->d_list, &ipvs->rs_table[hash]);
1da177e4
LT		 528
529 return 1;
530}
531
532/*
fc723250 533 * UNhashes ip_vs_dest from rs_table.
1da177e4
LT		 534 * should be called with locked tables.
535 */
536static int ip_vs_rs_unhash(struct ip_vs_dest *dest)
537{
538 /*
fc723250 539 * Remove it from the rs_table table.
1da177e4
LT		 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 */
552struct ip_vs_dest *
fc723250 553ip_vs_lookup_real_service(struct net *net, int af, __u16 protocol,
7937df15
JV		 554 const union nf_inet_addr *daddr,
555 __be16 dport)
1da177e4 556{
fc723250 557 struct netns_ipvs *ipvs = net_ipvs(net);
95c96174 558 unsigned int hash;
1da177e4
LT		 559 struct ip_vs_dest *dest;
560
561 /*
562 * Check for "full" addressed entries
563 * Return the first found entry
564 */
7937df15 565 hash = ip_vs_rs_hashkey(af, daddr, dport);
1da177e4 566
a0840e2e 567 read_lock(&ipvs->rs_lock);
fc723250 568 list_for_each_entry(dest, &ipvs->rs_table[hash], d_list) {
7937df15
JV		 569 if ((dest->af == af)
570 && ip_vs_addr_equal(af, &dest->addr, daddr)
1da177e4
LT		 571 && (dest->port == dport)
572 && ((dest->protocol == protocol) ||
573 dest->vfwmark)) {
574 /* HIT */
a0840e2e 575 read_unlock(&ipvs->rs_lock);
1da177e4
LT		 576 return dest;
577 }
578 }
a0840e2e 579 read_unlock(&ipvs->rs_lock);
1da177e4
LT		 580
581 return NULL;
582}
583
584/*
585 * Lookup destination by {addr,port} in the given service
586 */
587static struct ip_vs_dest *
7937df15
JV		 588ip_vs_lookup_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
589 __be16 dport)
1da177e4
LT		 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) {
7937df15
JV		 597 if ((dest->af == svc->af)
598 && ip_vs_addr_equal(svc->af, &dest->addr, daddr)
599 && (dest->port == dport)) {
1da177e4
LT		 600 /* HIT */
601 return dest;
602 }
603 }
604
605 return NULL;
606}
607
1e356f9c
RB		 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 */
fc723250
HS		 618struct ip_vs_dest *ip_vs_find_dest(struct net *net, int af,
619 const union nf_inet_addr *daddr,
7937df15
JV		 620 __be16 dport,
621 const union nf_inet_addr *vaddr,
52793dbe
JA		 622 __be16 vport, __u16 protocol, __u32 fwmark,
623 __u32 flags)
1e356f9c
RB		 624{
625 struct ip_vs_dest *dest;
626 struct ip_vs_service *svc;
52793dbe 627 __be16 port = dport;
1e356f9c 628
fc723250 629 svc = ip_vs_service_get(net, af, fwmark, protocol, vaddr, vport);
1e356f9c
RB		 630 if (!svc)
631 return NULL;
52793dbe
JA		 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);
1e356f9c
RB		 637 if (dest)
638 atomic_inc(&dest->refcnt);
639 ip_vs_service_put(svc);
640 return dest;
641}
1da177e4
LT		 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 */
653static struct ip_vs_dest *
7937df15
JV		 654ip_vs_trash_get_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
655 __be16 dport)
1da177e4
LT		 656{
657 struct ip_vs_dest *dest, *nxt;
f2431e6e 658 struct netns_ipvs *ipvs = net_ipvs(svc->net);
1da177e4
LT		 659
660 /*
661 * Find the destination in trash
662 */
f2431e6e 663 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, n_list) {
7937df15
JV		 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) &&
1da177e4
LT		 672 dest->port == dport &&
673 dest->vfwmark == svc->fwmark &&
674 dest->protocol == svc->protocol &&
675 (svc->fwmark ||
7937df15 676 (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
1da177e4
LT		 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) {
7937df15
JV		 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));
1da177e4
LT		 691 list_del(&dest->n_list);
692 ip_vs_dst_reset(dest);
693 __ip_vs_unbind_svc(dest);
b17fc996 694 free_percpu(dest->stats.cpustats);
1da177e4
LT		 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 */
f2431e6e 712static void ip_vs_trash_cleanup(struct net *net)
1da177e4
LT		 713{
714 struct ip_vs_dest *dest, *nxt;
f2431e6e 715 struct netns_ipvs *ipvs = net_ipvs(net);
1da177e4 716
f2431e6e 717 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, n_list) {
1da177e4
LT		 718 list_del(&dest->n_list);
719 ip_vs_dst_reset(dest);
720 __ip_vs_unbind_svc(dest);
b17fc996 721 free_percpu(dest->stats.cpustats);
1da177e4
LT		 722 kfree(dest);
723 }
724}
725
55a3d4e1
JA		 726static void
727ip_vs_copy_stats(struct ip_vs_stats_user *dst, struct ip_vs_stats *src)
728{
729#define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->ustats.c - src->ustats0.c
55a3d4e1
JA		 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
ea9f22cc 739 ip_vs_read_estimator(dst, src);
55a3d4e1
JA		 740
741 spin_unlock_bh(&src->lock);
742}
1da177e4
LT		 743
744static void
745ip_vs_zero_stats(struct ip_vs_stats *stats)
746{
747 spin_lock_bh(&stats->lock);
e93615d0 748
55a3d4e1
JA		 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
55a3d4e1
JA		 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
1da177e4 759 ip_vs_zero_estimator(stats);
e93615d0 760
3a14a313 761 spin_unlock_bh(&stats->lock);
1da177e4
LT		 762}
763
764/*
765 * Update a destination in the given service
766 */
767static void
26c15cfd
JA		 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)
1da177e4 770{
fc723250 771 struct netns_ipvs *ipvs = net_ipvs(svc->net);
1da177e4
LT		 772 int conn_flags;
773
774 /* set the weight and the flags */
775 atomic_set(&dest->weight, udest->weight);
3575792e
JA		 776 conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
777 conn_flags |= IP_VS_CONN_F_INACTIVE;
1da177e4 778
1da177e4 779 /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
3575792e 780 if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
1da177e4
LT		 781 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
782 } else {
783 /*
fc723250 784 * Put the real service in rs_table if not present.
1da177e4
LT		 785 * For now only for NAT!
786 */
a0840e2e 787 write_lock_bh(&ipvs->rs_lock);
fc723250 788 ip_vs_rs_hash(ipvs, dest);
a0840e2e 789 write_unlock_bh(&ipvs->rs_lock);
1da177e4
LT		 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;
26c15cfd 811
ff75f40f 812 spin_lock_bh(&dest->dst_lock);
fc604767 813 ip_vs_dst_reset(dest);
ff75f40f 814 spin_unlock_bh(&dest->dst_lock);
fc604767 815
26c15cfd 816 if (add)
6ef757f9 817 ip_vs_start_estimator(svc->net, &dest->stats);
26c15cfd
JA		 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);
1da177e4
LT		 834}
835
836
837/*
838 * Create a destination for the given service
839 */
840static int
c860c6b1 841ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
1da177e4
LT		 842 struct ip_vs_dest **dest_p)
843{
844 struct ip_vs_dest *dest;
95c96174 845 unsigned int atype;
1da177e4
LT		 846
847 EnterFunction(2);
848
09571c7a
VB		 849#ifdef CONFIG_IP_VS_IPV6
850 if (svc->af == AF_INET6) {
851 atype = ipv6_addr_type(&udest->addr.in6);
3bfb92f4
SW		 852 if ((!(atype & IPV6_ADDR_UNICAST) ||
853 atype & IPV6_ADDR_LINKLOCAL) &&
4a98480b 854 !__ip_vs_addr_is_local_v6(svc->net, &udest->addr.in6))
09571c7a
VB		 855 return -EINVAL;
856 } else
857#endif
858 {
4a98480b 859 atype = inet_addr_type(svc->net, udest->addr.ip);
09571c7a
VB		 860 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
861 return -EINVAL;
862 }
1da177e4 863
dee06e47 864 dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
0a9ee813 865 if (dest == NULL)
1da177e4 866 return -ENOMEM;
0a9ee813 867
b17fc996 868 dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
0a9ee813 869 if (!dest->stats.cpustats)
b17fc996 870 goto err_alloc;
1da177e4 871
c860c6b1 872 dest->af = svc->af;
1da177e4 873 dest->protocol = svc->protocol;
c860c6b1 874 dest->vaddr = svc->addr;
1da177e4
LT		 875 dest->vport = svc->port;
876 dest->vfwmark = svc->fwmark;
c860c6b1 877 ip_vs_addr_copy(svc->af, &dest->addr, &udest->addr);
1da177e4
LT		 878 dest->port = udest->port;
879
880 atomic_set(&dest->activeconns, 0);
881 atomic_set(&dest->inactconns, 0);
882 atomic_set(&dest->persistconns, 0);
26c15cfd 883 atomic_set(&dest->refcnt, 1);
1da177e4
LT		 884
885 INIT_LIST_HEAD(&dest->d_list);
886 spin_lock_init(&dest->dst_lock);
887 spin_lock_init(&dest->stats.lock);
26c15cfd 888 __ip_vs_update_dest(svc, dest, udest, 1);
1da177e4
LT		 889
890 *dest_p = dest;
891
892 LeaveFunction(2);
893 return 0;
b17fc996
HS		 894
895err_alloc:
896 kfree(dest);
897 return -ENOMEM;
1da177e4
LT		 898}
899
900
901/*
902 * Add a destination into an existing service
903 */
904static int
c860c6b1 905ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1da177e4
LT		 906{
907 struct ip_vs_dest *dest;
c860c6b1 908 union nf_inet_addr daddr;
014d730d 909 __be16 dport = udest->port;
1da177e4
LT		 910 int ret;
911
912 EnterFunction(2);
913
914 if (udest->weight < 0) {
1e3e238e 915 pr_err("%s(): server weight less than zero\n", __func__);
1da177e4
LT		 916 return -ERANGE;
917 }
918
919 if (udest->l_threshold > udest->u_threshold) {
1e3e238e
HE		 920 pr_err("%s(): lower threshold is higher than upper threshold\n",
921 __func__);
1da177e4
LT		 922 return -ERANGE;
923 }
924
c860c6b1
JV		 925 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
926
1da177e4
LT		 927 /*
928 * Check if the dest already exists in the list
929 */
7937df15
JV		 930 dest = ip_vs_lookup_dest(svc, &daddr, dport);
931
1da177e4 932 if (dest != NULL) {
1e3e238e 933 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
1da177e4
LT		 934 return -EEXIST;
935 }
936
937 /*
938 * Check if the dest already exists in the trash and
939 * is from the same service
940 */
7937df15
JV		 941 dest = ip_vs_trash_get_dest(svc, &daddr, dport);
942
1da177e4 943 if (dest != NULL) {
cfc78c5a
JV		 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
1da177e4
LT		 952 /*
953 * Get the destination from the trash
954 */
955 list_del(&dest->n_list);
956
26c15cfd
JA		 957 __ip_vs_update_dest(svc, dest, udest, 1);
958 ret = 0;
959 } else {
1da177e4 960 /*
26c15cfd 961 * Allocate and initialize the dest structure
1da177e4 962 */
26c15cfd 963 ret = ip_vs_new_dest(svc, udest, &dest);
1da177e4 964 }
1da177e4
LT		 965 LeaveFunction(2);
966
26c15cfd 967 return ret;
1da177e4
LT		 968}
969
970
971/*
972 * Edit a destination in the given service
973 */
974static int
c860c6b1 975ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1da177e4
LT		 976{
977 struct ip_vs_dest *dest;
c860c6b1 978 union nf_inet_addr daddr;
014d730d 979 __be16 dport = udest->port;
1da177e4
LT		 980
981 EnterFunction(2);
982
983 if (udest->weight < 0) {
1e3e238e 984 pr_err("%s(): server weight less than zero\n", __func__);
1da177e4
LT		 985 return -ERANGE;
986 }
987
988 if (udest->l_threshold > udest->u_threshold) {
1e3e238e
HE		 989 pr_err("%s(): lower threshold is higher than upper threshold\n",
990 __func__);
1da177e4
LT		 991 return -ERANGE;
992 }
993
c860c6b1
JV		 994 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
995
1da177e4
LT		 996 /*
997 * Lookup the destination list
998 */
7937df15
JV		 999 dest = ip_vs_lookup_dest(svc, &daddr, dport);
1000
1da177e4 1001 if (dest == NULL) {
1e3e238e 1002 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1da177e4
LT		 1003 return -ENOENT;
1004 }
1005
26c15cfd 1006 __ip_vs_update_dest(svc, dest, udest, 0);
1da177e4
LT		 1007 LeaveFunction(2);
1008
1009 return 0;
1010}
1011
1012
1013/*
1014 * Delete a destination (must be already unlinked from the service)
1015 */
29c2026f 1016static void __ip_vs_del_dest(struct net *net, struct ip_vs_dest *dest)
1da177e4 1017{
a0840e2e
HS		 1018 struct netns_ipvs *ipvs = net_ipvs(net);
1019
6ef757f9 1020 ip_vs_stop_estimator(net, &dest->stats);
1da177e4
LT		 1021
1022 /*
1023 * Remove it from the d-linked list with the real services.
1024 */
a0840e2e 1025 write_lock_bh(&ipvs->rs_lock);
1da177e4 1026 ip_vs_rs_unhash(dest);
a0840e2e 1027 write_unlock_bh(&ipvs->rs_lock);
1da177e4
LT		 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)) {
26c15cfd
JA		 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));
1da177e4
LT		 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);
b17fc996 1046 free_percpu(dest->stats.cpustats);
1da177e4
LT		 1047 kfree(dest);
1048 } else {
cfc78c5a
JV		 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));
f2431e6e 1054 list_add(&dest->n_list, &ipvs->dest_trash);
1da177e4
LT		 1055 atomic_inc(&dest->refcnt);
1056 }
1057}
1058
1059
1060/*
1061 * Unlink a destination from the given service
1062 */
1063static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1064 struct ip_vs_dest *dest,
1065 int svcupd)
1066{
1067 dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1068
1069 /*
1070 * Remove it from the d-linked destination list.
1071 */
1072 list_del(&dest->n_list);
1073 svc->num_dests--;
82dfb6f3
SW		 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);
1da177e4
LT		 1080}
1081
1082
1083/*
1084 * Delete a destination server in the given service
1085 */
1086static int
c860c6b1 1087ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1da177e4
LT		 1088{
1089 struct ip_vs_dest *dest;
014d730d 1090 __be16 dport = udest->port;
1da177e4
LT		 1091
1092 EnterFunction(2);
1093
7937df15 1094 dest = ip_vs_lookup_dest(svc, &udest->addr, dport);
c860c6b1 1095
1da177e4 1096 if (dest == NULL) {
1e3e238e 1097 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1da177e4
LT		 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 */
26c15cfd 1106 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1da177e4
LT		 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 */
a0840e2e 1118 __ip_vs_del_dest(svc->net, dest);
1da177e4
LT		 1119
1120 LeaveFunction(2);
1121
1122 return 0;
1123}
1124
1125
1126/*
1127 * Add a service into the service hash table
1128 */
1129static int
fc723250 1130ip_vs_add_service(struct net *net, struct ip_vs_service_user_kern *u,
c860c6b1 1131 struct ip_vs_service **svc_p)
1da177e4
LT		 1132{
1133 int ret = 0;
1134 struct ip_vs_scheduler *sched = NULL;
0d1e71b0 1135 struct ip_vs_pe *pe = NULL;
1da177e4 1136 struct ip_vs_service *svc = NULL;
a0840e2e 1137 struct netns_ipvs *ipvs = net_ipvs(net);
1da177e4
LT		 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) {
1e3e238e 1145 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1da177e4 1146 ret = -ENOENT;
6e08bfb8 1147 goto out_err;
1da177e4
LT		 1148 }
1149
0d1e71b0 1150 if (u->pe_name && *u->pe_name) {
e9e5eee8 1151 pe = ip_vs_pe_getbyname(u->pe_name);
0d1e71b0
SH		 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
f94fd041 1160#ifdef CONFIG_IP_VS_IPV6
48148938
JV		 1161 if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1162 ret = -EINVAL;
1163 goto out_err;
f94fd041
JV		 1164 }
1165#endif
1166
dee06e47 1167 svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1da177e4 1168 if (svc == NULL) {
1e3e238e 1169 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1da177e4
LT		 1170 ret = -ENOMEM;
1171 goto out_err;
1172 }
b17fc996 1173 svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
0a9ee813 1174 if (!svc->stats.cpustats)
b17fc996 1175 goto out_err;
1da177e4
LT		 1176
1177 /* I'm the first user of the service */
26c15cfd 1178 atomic_set(&svc->usecnt, 0);
1da177e4
LT		 1179 atomic_set(&svc->refcnt, 0);
1180
c860c6b1 1181 svc->af = u->af;
1da177e4 1182 svc->protocol = u->protocol;
c860c6b1 1183 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1da177e4
LT		 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;
fc723250 1189 svc->net = net;
1da177e4
LT		 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
0d1e71b0
SH		 1201 /* Bind the ct retriever */
1202 ip_vs_bind_pe(svc, pe);
1203 pe = NULL;
1204
1da177e4
LT		 1205 /* Update the virtual service counters */
1206 if (svc->port == FTPPORT)
763f8d0e 1207 atomic_inc(&ipvs->ftpsvc_counter);
1da177e4 1208 else if (svc->port == 0)
763f8d0e 1209 atomic_inc(&ipvs->nullsvc_counter);
1da177e4 1210
6ef757f9 1211 ip_vs_start_estimator(net, &svc->stats);
f94fd041
JV		 1212
1213 /* Count only IPv4 services for old get/setsockopt interface */
1214 if (svc->af == AF_INET)
a0840e2e 1215 ipvs->num_services++;
1da177e4
LT		 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;
7a4f0761
HS		 1223 /* Now there is a service - full throttle */
1224 ipvs->enable = 1;
1da177e4
LT		 1225 return 0;
1226
b17fc996 1227
6e08bfb8 1228 out_err:
1da177e4 1229 if (svc != NULL) {
2fabf35b 1230 ip_vs_unbind_scheduler(svc);
1da177e4
LT		 1231 if (svc->inc) {
1232 local_bh_disable();
1233 ip_vs_app_inc_put(svc->inc);
1234 local_bh_enable();
1235 }
b17fc996
HS		 1236 if (svc->stats.cpustats)
1237 free_percpu(svc->stats.cpustats);
1da177e4
LT		 1238 kfree(svc);
1239 }
1240 ip_vs_scheduler_put(sched);
0d1e71b0 1241 ip_vs_pe_put(pe);
1da177e4 1242
1da177e4
LT		 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 */
1253static int
c860c6b1 1254ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1da177e4
LT		 1255{
1256 struct ip_vs_scheduler *sched, *old_sched;
0d1e71b0 1257 struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1da177e4
LT		 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) {
1e3e238e 1265 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1da177e4
LT		 1266 return -ENOENT;
1267 }
1268 old_sched = sched;
1269
0d1e71b0 1270 if (u->pe_name && *u->pe_name) {
e9e5eee8 1271 pe = ip_vs_pe_getbyname(u->pe_name);
0d1e71b0
SH		 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
f94fd041 1281#ifdef CONFIG_IP_VS_IPV6
48148938
JV		 1282 if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1283 ret = -EINVAL;
1284 goto out;
f94fd041
JV		 1285 }
1286#endif
1287
1da177e4
LT		 1288 write_lock_bh(&__ip_vs_svc_lock);
1289
1290 /*
1291 * Wait until all other svc users go away.
1292 */
26c15cfd 1293 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1da177e4
LT		 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;
9e691ed6 1309 goto out_unlock;
1da177e4
LT		 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;
9e691ed6 1328 goto out_unlock;
1da177e4
LT		 1329 }
1330 }
1331
0d1e71b0
SH		 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
552ad65a 1338out_unlock:
1da177e4 1339 write_unlock_bh(&__ip_vs_svc_lock);
552ad65a 1340out:
6e08bfb8 1341 ip_vs_scheduler_put(old_sched);
0d1e71b0 1342 ip_vs_pe_put(old_pe);
1da177e4
LT		 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 */
1352static void __ip_vs_del_service(struct ip_vs_service *svc)
1353{
1354 struct ip_vs_dest *dest, *nxt;
1355 struct ip_vs_scheduler *old_sched;
0d1e71b0 1356 struct ip_vs_pe *old_pe;
a0840e2e 1357 struct netns_ipvs *ipvs = net_ipvs(svc->net);
0d1e71b0
SH		 1358
1359 pr_info("%s: enter\n", __func__);
1da177e4 1360
f94fd041
JV		 1361 /* Count only IPv4 services for old get/setsockopt interface */
1362 if (svc->af == AF_INET)
a0840e2e 1363 ipvs->num_services--;
f94fd041 1364
6ef757f9 1365 ip_vs_stop_estimator(svc->net, &svc->stats);
1da177e4
LT		 1366
1367 /* Unbind scheduler */
1368 old_sched = svc->scheduler;
1369 ip_vs_unbind_scheduler(svc);
6e08bfb8 1370 ip_vs_scheduler_put(old_sched);
1da177e4 1371
0d1e71b0
SH		 1372 /* Unbind persistence engine */
1373 old_pe = svc->pe;
1374 ip_vs_unbind_pe(svc);
1375 ip_vs_pe_put(old_pe);
1376
1da177e4
LT		 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);
29c2026f 1388 __ip_vs_del_dest(svc->net, dest);
1da177e4
LT		 1389 }
1390
1391 /*
1392 * Update the virtual service counters
1393 */
1394 if (svc->port == FTPPORT)
763f8d0e 1395 atomic_dec(&ipvs->ftpsvc_counter);
1da177e4 1396 else if (svc->port == 0)
763f8d0e 1397 atomic_dec(&ipvs->nullsvc_counter);
1da177e4
LT		 1398
1399 /*
1400 * Free the service if nobody refers to it
1401 */
26c15cfd
JA		 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));
b17fc996 1407 free_percpu(svc->stats.cpustats);
1da177e4 1408 kfree(svc);
26c15cfd 1409 }
1da177e4
LT		 1410
1411 /* decrease the module use count */
1412 ip_vs_use_count_dec();
1413}
1414
1415/*
26c15cfd 1416 * Unlink a service from list and try to delete it if its refcnt reached 0
1da177e4 1417 */
26c15cfd 1418static void ip_vs_unlink_service(struct ip_vs_service *svc)
1da177e4 1419{
1da177e4
LT		 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 */
26c15cfd 1430 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1da177e4
LT		 1431
1432 __ip_vs_del_service(svc);
1433
1434 write_unlock_bh(&__ip_vs_svc_lock);
26c15cfd
JA		 1435}
1436
1437/*
1438 * Delete a service from the service list
1439 */
1440static int ip_vs_del_service(struct ip_vs_service *svc)
1441{
1442 if (svc == NULL)
1443 return -EEXIST;
1444 ip_vs_unlink_service(svc);
1da177e4
LT		 1445
1446 return 0;
1447}
1448
1449
1450/*
1451 * Flush all the virtual services
1452 */
fc723250 1453static int ip_vs_flush(struct net *net)
1da177e4
LT		 1454{
1455 int idx;
1456 struct ip_vs_service *svc, *nxt;
1457
1458 /*
fc723250 1459 * Flush the service table hashed by <netns,protocol,addr,port>
1da177e4
LT		 1460 */
1461 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
fc723250
HS		 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);
1da177e4
LT		 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) {
fc723250
HS		 1475 if (net_eq(svc->net, net))
1476 ip_vs_unlink_service(svc);
1da177e4
LT		 1477 }
1478 }
1479
1480 return 0;
1481}
1482
7a4f0761
HS		 1483/*
1484 * Delete service by {netns} in the service table.
1485 * Called by __ip_vs_cleanup()
1486 */
503cf15a 1487void ip_vs_service_net_cleanup(struct net *net)
7a4f0761
HS		 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 */
1499static inline void
1500__ip_vs_dev_reset(struct ip_vs_dest *dest, struct net_device *dev)
1501{
1502 spin_lock_bh(&dest->dst_lock);
1503 if (dest->dst_cache && dest->dst_cache->dev == dev) {
1504 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1505 dev->name,
1506 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1507 ntohs(dest->port),
1508 atomic_read(&dest->refcnt));
1509 ip_vs_dst_reset(dest);
1510 }
1511 spin_unlock_bh(&dest->dst_lock);
1512
1513}
1514/*
1515 * 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 */
1519static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1520 void *ptr)
1521{
1522 struct net_device *dev = ptr;
1523 struct net *net = dev_net(dev);
1524 struct ip_vs_service *svc;
1525 struct ip_vs_dest *dest;
1526 unsigned int idx;
1527
1528 if (event != NETDEV_UNREGISTER)
1529 return NOTIFY_DONE;
1530 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1531 EnterFunction(2);
1532 mutex_lock(&__ip_vs_mutex);
1533 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1534 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1535 if (net_eq(svc->net, net)) {
1536 list_for_each_entry(dest, &svc->destinations,
1537 n_list) {
1538 __ip_vs_dev_reset(dest, dev);
1539 }
1540 }
1541 }
1542
1543 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1544 if (net_eq(svc->net, net)) {
1545 list_for_each_entry(dest, &svc->destinations,
1546 n_list) {
1547 __ip_vs_dev_reset(dest, dev);
1548 }
1549 }
1550
1551 }
1552 }
1553
1554 list_for_each_entry(dest, &net_ipvs(net)->dest_trash, n_list) {
1555 __ip_vs_dev_reset(dest, dev);
1556 }
1557 mutex_unlock(&__ip_vs_mutex);
1558 LeaveFunction(2);
1559 return NOTIFY_DONE;
1560}
1da177e4
LT		 1561
1562/*
1563 * Zero counters in a service or all services
1564 */
1565static int ip_vs_zero_service(struct ip_vs_service *svc)
1566{
1567 struct ip_vs_dest *dest;
1568
1569 write_lock_bh(&__ip_vs_svc_lock);
1570 list_for_each_entry(dest, &svc->destinations, n_list) {
1571 ip_vs_zero_stats(&dest->stats);
1572 }
1573 ip_vs_zero_stats(&svc->stats);
1574 write_unlock_bh(&__ip_vs_svc_lock);
1575 return 0;
1576}
1577
fc723250 1578static int ip_vs_zero_all(struct net *net)
1da177e4
LT		 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) {
fc723250
HS		 1585 if (net_eq(svc->net, net))
1586 ip_vs_zero_service(svc);
1da177e4
LT		 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) {
fc723250
HS		 1592 if (net_eq(svc->net, net))
1593 ip_vs_zero_service(svc);
1da177e4
LT		 1594 }
1595 }
1596
2a0751af 1597 ip_vs_zero_stats(&net_ipvs(net)->tot_stats);
1da177e4
LT		 1598 return 0;
1599}
1600
14e40546 1601#ifdef CONFIG_SYSCTL
1da177e4 1602static int
8d65af78 1603proc_do_defense_mode(ctl_table *table, int write,
1da177e4
LT		 1604 void __user *buffer, size_t *lenp, loff_t *ppos)
1605{
9330419d 1606 struct net *net = current->nsproxy->net_ns;
1da177e4
LT		 1607 int *valp = table->data;
1608 int val = *valp;
1609 int rc;
1610
8d65af78 1611 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1da177e4
LT		 1612 if (write && (*valp != val)) {
1613 if ((*valp < 0) || (*valp > 3)) {
1614 /* Restore the correct value */
1615 *valp = val;
1616 } else {
9330419d 1617 update_defense_level(net_ipvs(net));
1da177e4
LT		 1618 }
1619 }
1620 return rc;
1621}
1622
1da177e4 1623static int
8d65af78 1624proc_do_sync_threshold(ctl_table *table, int write,
1da177e4
LT		 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
8d65af78 1634 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1da177e4
LT		 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
b880c1f0
HS		 1642static int
1643proc_do_sync_mode(ctl_table *table, int write,
1644 void __user *buffer, size_t *lenp, loff_t *ppos)
1645{
1646 int *valp = table->data;
1647 int val = *valp;
1648 int rc;
1649
1650 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1651 if (write && (*valp != val)) {
1652 if ((*valp < 0) || (*valp > 1)) {
1653 /* Restore the correct value */
1654 *valp = val;
1655 } else {
f131315f
HS		 1656 struct net *net = current->nsproxy->net_ns;
1657 ip_vs_sync_switch_mode(net, val);
b880c1f0
HS		 1658 }
1659 }
1660 return rc;
1661}
1da177e4
LT		 1662
1663/*
1664 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
a0840e2e 1665 * Do not change order or insert new entries without
503cf15a 1666 * align with netns init in ip_vs_control_net_init()
1da177e4
LT		 1667 */
1668
1669static struct ctl_table vs_vars[] = {
1670 {
1da177e4 1671 .procname = "amemthresh",
1da177e4
LT		 1672 .maxlen = sizeof(int),
1673 .mode = 0644,
6d9f239a 1674 .proc_handler = proc_dointvec,
1da177e4 1675 },
1da177e4 1676 {
1da177e4 1677 .procname = "am_droprate",
1da177e4
LT		 1678 .maxlen = sizeof(int),
1679 .mode = 0644,
6d9f239a 1680 .proc_handler = proc_dointvec,
1da177e4
LT		 1681 },
1682 {
1da177e4 1683 .procname = "drop_entry",
1da177e4
LT		 1684 .maxlen = sizeof(int),
1685 .mode = 0644,
6d9f239a 1686 .proc_handler = proc_do_defense_mode,
1da177e4
LT		 1687 },
1688 {
1da177e4 1689 .procname = "drop_packet",
1da177e4
LT		 1690 .maxlen = sizeof(int),
1691 .mode = 0644,
6d9f239a 1692 .proc_handler = proc_do_defense_mode,
1da177e4 1693 },
f4bc17cd
JA		 1694#ifdef CONFIG_IP_VS_NFCT
1695 {
1696 .procname = "conntrack",
f4bc17cd
JA		 1697 .maxlen = sizeof(int),
1698 .mode = 0644,
1699 .proc_handler = &proc_dointvec,
1700 },
1701#endif
1da177e4 1702 {
1da177e4 1703 .procname = "secure_tcp",
1da177e4
LT		 1704 .maxlen = sizeof(int),
1705 .mode = 0644,
6d9f239a 1706 .proc_handler = proc_do_defense_mode,
1da177e4 1707 },
8a803040
JA		 1708 {
1709 .procname = "snat_reroute",
8a803040
JA		 1710 .maxlen = sizeof(int),
1711 .mode = 0644,
1712 .proc_handler = &proc_dointvec,
1713 },
b880c1f0
HS		 1714 {
1715 .procname = "sync_version",
b880c1f0
HS		 1716 .maxlen = sizeof(int),
1717 .mode = 0644,
1718 .proc_handler = &proc_do_sync_mode,
1719 },
1c003b15
PNA		 1720 {
1721 .procname = "sync_qlen_max",
1722 .maxlen = sizeof(int),
1723 .mode = 0644,
1724 .proc_handler = proc_dointvec,
1725 },
1726 {
1727 .procname = "sync_sock_size",
1728 .maxlen = sizeof(int),
1729 .mode = 0644,
1730 .proc_handler = proc_dointvec,
1731 },
a0840e2e
HS		 1732 {
1733 .procname = "cache_bypass",
1734 .maxlen = sizeof(int),
1735 .mode = 0644,
1736 .proc_handler = proc_dointvec,
1737 },
1738 {
1739 .procname = "expire_nodest_conn",
1740 .maxlen = sizeof(int),
1741 .mode = 0644,
1742 .proc_handler = proc_dointvec,
1743 },
1744 {
1745 .procname = "expire_quiescent_template",
1746 .maxlen = sizeof(int),
1747 .mode = 0644,
1748 .proc_handler = proc_dointvec,
1749 },
1750 {
1751 .procname = "sync_threshold",
1752 .maxlen =
1753 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1754 .mode = 0644,
1755 .proc_handler = proc_do_sync_threshold,
1756 },
1757 {
1758 .procname = "nat_icmp_send",
1759 .maxlen = sizeof(int),
1760 .mode = 0644,
1761 .proc_handler = proc_dointvec,
1762 },
1763#ifdef CONFIG_IP_VS_DEBUG
1764 {
1765 .procname = "debug_level",
1766 .data = &sysctl_ip_vs_debug_level,
1767 .maxlen = sizeof(int),
1768 .mode = 0644,
1769 .proc_handler = proc_dointvec,
1770 },
1771#endif
1da177e4
LT		 1772#if 0
1773 {
1da177e4
LT		 1774 .procname = "timeout_established",
1775 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1776 .maxlen = sizeof(int),
1777 .mode = 0644,
6d9f239a 1778 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1779 },
1780 {
1da177e4
LT		 1781 .procname = "timeout_synsent",
1782 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1783 .maxlen = sizeof(int),
1784 .mode = 0644,
6d9f239a 1785 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1786 },
1787 {
1da177e4
LT		 1788 .procname = "timeout_synrecv",
1789 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1790 .maxlen = sizeof(int),
1791 .mode = 0644,
6d9f239a 1792 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1793 },
1794 {
1da177e4
LT		 1795 .procname = "timeout_finwait",
1796 .data = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1797 .maxlen = sizeof(int),
1798 .mode = 0644,
6d9f239a 1799 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1800 },
1801 {
1da177e4
LT		 1802 .procname = "timeout_timewait",
1803 .data = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1804 .maxlen = sizeof(int),
1805 .mode = 0644,
6d9f239a 1806 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1807 },
1808 {
1da177e4
LT		 1809 .procname = "timeout_close",
1810 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1811 .maxlen = sizeof(int),
1812 .mode = 0644,
6d9f239a 1813 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1814 },
1815 {
1da177e4
LT		 1816 .procname = "timeout_closewait",
1817 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1818 .maxlen = sizeof(int),
1819 .mode = 0644,
6d9f239a 1820 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1821 },
1822 {
1da177e4
LT		 1823 .procname = "timeout_lastack",
1824 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1825 .maxlen = sizeof(int),
1826 .mode = 0644,
6d9f239a 1827 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1828 },
1829 {
1da177e4
LT		 1830 .procname = "timeout_listen",
1831 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1832 .maxlen = sizeof(int),
1833 .mode = 0644,
6d9f239a 1834 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1835 },
1836 {
1da177e4
LT		 1837 .procname = "timeout_synack",
1838 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1839 .maxlen = sizeof(int),
1840 .mode = 0644,
6d9f239a 1841 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1842 },
1843 {
1da177e4
LT		 1844 .procname = "timeout_udp",
1845 .data = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1846 .maxlen = sizeof(int),
1847 .mode = 0644,
6d9f239a 1848 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1849 },
1850 {
1da177e4
LT		 1851 .procname = "timeout_icmp",
1852 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1853 .maxlen = sizeof(int),
1854 .mode = 0644,
6d9f239a 1855 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1856 },
1857#endif
f8572d8f 1858 { }
1da177e4
LT		 1859};
1860
14e40546 1861#endif
1da177e4 1862
1da177e4
LT		 1863#ifdef CONFIG_PROC_FS
1864
1865struct ip_vs_iter {
fc723250 1866 struct seq_net_private p; /* Do not move this, netns depends upon it*/
1da177e4
LT		 1867 struct list_head *table;
1868 int bucket;
1869};
1870
1871/*
1872 * Write the contents of the VS rule table to a PROCfs file.
1873 * (It is kept just for backward compatibility)
1874 */
95c96174 1875static inline const char *ip_vs_fwd_name(unsigned int flags)
1da177e4
LT		 1876{
1877 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1878 case IP_VS_CONN_F_LOCALNODE:
1879 return "Local";
1880 case IP_VS_CONN_F_TUNNEL:
1881 return "Tunnel";
1882 case IP_VS_CONN_F_DROUTE:
1883 return "Route";
1884 default:
1885 return "Masq";
1886 }
1887}
1888
1889
1890/* Get the Nth entry in the two lists */
1891static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1892{
fc723250 1893 struct net *net = seq_file_net(seq);
1da177e4
LT		 1894 struct ip_vs_iter *iter = seq->private;
1895 int idx;
1896 struct ip_vs_service *svc;
1897
1898 /* look in hash by protocol */
1899 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1900 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
fc723250 1901 if (net_eq(svc->net, net) && pos-- == 0) {
1da177e4
LT		 1902 iter->table = ip_vs_svc_table;
1903 iter->bucket = idx;
1904 return svc;
1905 }
1906 }
1907 }
1908
1909 /* keep looking in fwmark */
1910 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1911 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
fc723250 1912 if (net_eq(svc->net, net) && pos-- == 0) {
1da177e4
LT		 1913 iter->table = ip_vs_svc_fwm_table;
1914 iter->bucket = idx;
1915 return svc;
1916 }
1917 }
1918 }
1919
1920 return NULL;
1921}
1922
1923static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
563e94f0 1924__acquires(__ip_vs_svc_lock)
1da177e4
LT		 1925{
1926
1927 read_lock_bh(&__ip_vs_svc_lock);
1928 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1929}
1930
1931
1932static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1933{
1934 struct list_head *e;
1935 struct ip_vs_iter *iter;
1936 struct ip_vs_service *svc;
1937
1938 ++*pos;
1939 if (v == SEQ_START_TOKEN)
1940 return ip_vs_info_array(seq,0);
1941
1942 svc = v;
1943 iter = seq->private;
1944
1945 if (iter->table == ip_vs_svc_table) {
1946 /* next service in table hashed by protocol */
1947 if ((e = svc->s_list.next) != &ip_vs_svc_table[iter->bucket])
1948 return list_entry(e, struct ip_vs_service, s_list);
1949
1950
1951 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1952 list_for_each_entry(svc,&ip_vs_svc_table[iter->bucket],
1953 s_list) {
1954 return svc;
1955 }
1956 }
1957
1958 iter->table = ip_vs_svc_fwm_table;
1959 iter->bucket = -1;
1960 goto scan_fwmark;
1961 }
1962
1963 /* next service in hashed by fwmark */
1964 if ((e = svc->f_list.next) != &ip_vs_svc_fwm_table[iter->bucket])
1965 return list_entry(e, struct ip_vs_service, f_list);
1966
1967 scan_fwmark:
1968 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1969 list_for_each_entry(svc, &ip_vs_svc_fwm_table[iter->bucket],
1970 f_list)
1971 return svc;
1972 }
1973
1974 return NULL;
1975}
1976
1977static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
563e94f0 1978__releases(__ip_vs_svc_lock)
1da177e4
LT		 1979{
1980 read_unlock_bh(&__ip_vs_svc_lock);
1981}
1982
1983
1984static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
1985{
1986 if (v == SEQ_START_TOKEN) {
1987 seq_printf(seq,
1988 "IP Virtual Server version %d.%d.%d (size=%d)\n",
6f7edb48 1989 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
1da177e4
LT		 1990 seq_puts(seq,
1991 "Prot LocalAddress:Port Scheduler Flags\n");
1992 seq_puts(seq,
1993 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
1994 } else {
1995 const struct ip_vs_service *svc = v;
1996 const struct ip_vs_iter *iter = seq->private;
1997 const struct ip_vs_dest *dest;
1998
667a5f18
VB		 1999 if (iter->table == ip_vs_svc_table) {
2000#ifdef CONFIG_IP_VS_IPV6
2001 if (svc->af == AF_INET6)
5b095d98 2002 seq_printf(seq, "%s [%pI6]:%04X %s ",
667a5f18 2003 ip_vs_proto_name(svc->protocol),
38ff4fa4 2004 &svc->addr.in6,
667a5f18
VB		 2005 ntohs(svc->port),
2006 svc->scheduler->name);
2007 else
2008#endif
26ec037f 2009 seq_printf(seq, "%s %08X:%04X %s %s ",
667a5f18
VB		 2010 ip_vs_proto_name(svc->protocol),
2011 ntohl(svc->addr.ip),
2012 ntohs(svc->port),
26ec037f
NC		 2013 svc->scheduler->name,
2014 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
667a5f18 2015 } else {
26ec037f
NC		 2016 seq_printf(seq, "FWM %08X %s %s",
2017 svc->fwmark, svc->scheduler->name,
2018 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
667a5f18 2019 }
1da177e4
LT		 2020
2021 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2022 seq_printf(seq, "persistent %d %08X\n",
2023 svc->timeout,
2024 ntohl(svc->netmask));
2025 else
2026 seq_putc(seq, '\n');
2027
2028 list_for_each_entry(dest, &svc->destinations, n_list) {
667a5f18
VB		 2029#ifdef CONFIG_IP_VS_IPV6
2030 if (dest->af == AF_INET6)
2031 seq_printf(seq,
5b095d98 2032 " -> [%pI6]:%04X"
667a5f18 2033 " %-7s %-6d %-10d %-10d\n",
38ff4fa4 2034 &dest->addr.in6,
667a5f18
VB		 2035 ntohs(dest->port),
2036 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2037 atomic_read(&dest->weight),
2038 atomic_read(&dest->activeconns),
2039 atomic_read(&dest->inactconns));
2040 else
2041#endif
2042 seq_printf(seq,
2043 " -> %08X:%04X "
2044 "%-7s %-6d %-10d %-10d\n",
2045 ntohl(dest->addr.ip),
2046 ntohs(dest->port),
2047 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2048 atomic_read(&dest->weight),
2049 atomic_read(&dest->activeconns),
2050 atomic_read(&dest->inactconns));
2051
1da177e4
LT		 2052 }
2053 }
2054 return 0;
2055}
2056
56b3d975 2057static const struct seq_operations ip_vs_info_seq_ops = {
1da177e4
LT		 2058 .start = ip_vs_info_seq_start,
2059 .next = ip_vs_info_seq_next,
2060 .stop = ip_vs_info_seq_stop,
2061 .show = ip_vs_info_seq_show,
2062};
2063
2064static int ip_vs_info_open(struct inode *inode, struct file *file)
2065{
fc723250 2066 return seq_open_net(inode, file, &ip_vs_info_seq_ops,
cf7732e4 2067 sizeof(struct ip_vs_iter));
1da177e4
LT		 2068}
2069
9a32144e 2070static const struct file_operations ip_vs_info_fops = {
1da177e4
LT		 2071 .owner = THIS_MODULE,
2072 .open = ip_vs_info_open,
2073 .read = seq_read,
2074 .llseek = seq_lseek,
0f08190f 2075 .release = seq_release_net,
1da177e4
LT		 2076};
2077
1da177e4
LT		 2078static int ip_vs_stats_show(struct seq_file *seq, void *v)
2079{
b17fc996 2080 struct net *net = seq_file_single_net(seq);
55a3d4e1 2081 struct ip_vs_stats_user show;
1da177e4
LT		 2082
2083/* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2084 seq_puts(seq,
2085 " Total Incoming Outgoing Incoming Outgoing\n");
2086 seq_printf(seq,
2087 " Conns Packets Packets Bytes Bytes\n");
2088
55a3d4e1
JA		 2089 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2090 seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", show.conns,
2091 show.inpkts, show.outpkts,
2092 (unsigned long long) show.inbytes,
2093 (unsigned long long) show.outbytes);
1da177e4
LT		 2094
2095/* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2096 seq_puts(seq,
2097 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
55a3d4e1
JA		 2098 seq_printf(seq, "%8X %8X %8X %16X %16X\n",
2099 show.cps, show.inpps, show.outpps,
2100 show.inbps, show.outbps);
1da177e4
LT		 2101
2102 return 0;
2103}
2104
2105static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2106{
fc723250 2107 return single_open_net(inode, file, ip_vs_stats_show);
1da177e4
LT		 2108}
2109
9a32144e 2110static const struct file_operations ip_vs_stats_fops = {
1da177e4
LT		 2111 .owner = THIS_MODULE,
2112 .open = ip_vs_stats_seq_open,
2113 .read = seq_read,
2114 .llseek = seq_lseek,
0f08190f 2115 .release = single_release_net,
1da177e4
LT		 2116};
2117
b17fc996
HS		 2118static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2119{
2120 struct net *net = seq_file_single_net(seq);
2a0751af
JA		 2121 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2122 struct ip_vs_cpu_stats *cpustats = tot_stats->cpustats;
ea9f22cc 2123 struct ip_vs_stats_user rates;
b17fc996
HS		 2124 int i;
2125
2126/* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2127 seq_puts(seq,
2128 " Total Incoming Outgoing Incoming Outgoing\n");
2129 seq_printf(seq,
2130 "CPU Conns Packets Packets Bytes Bytes\n");
2131
2132 for_each_possible_cpu(i) {
2a0751af
JA		 2133 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2134 unsigned int start;
2135 __u64 inbytes, outbytes;
2136
2137 do {
2138 start = u64_stats_fetch_begin_bh(&u->syncp);
2139 inbytes = u->ustats.inbytes;
2140 outbytes = u->ustats.outbytes;
2141 } while (u64_stats_fetch_retry_bh(&u->syncp, start));
2142
b17fc996 2143 seq_printf(seq, "%3X %8X %8X %8X %16LX %16LX\n",
2a0751af
JA		 2144 i, u->ustats.conns, u->ustats.inpkts,
2145 u->ustats.outpkts, (__u64)inbytes,
2146 (__u64)outbytes);
b17fc996
HS		 2147 }
2148
2149 spin_lock_bh(&tot_stats->lock);
ea9f22cc 2150
b17fc996
HS		 2151 seq_printf(seq, " ~ %8X %8X %8X %16LX %16LX\n\n",
2152 tot_stats->ustats.conns, tot_stats->ustats.inpkts,
2153 tot_stats->ustats.outpkts,
2154 (unsigned long long) tot_stats->ustats.inbytes,
2155 (unsigned long long) tot_stats->ustats.outbytes);
2156
ea9f22cc
JA		 2157 ip_vs_read_estimator(&rates, tot_stats);
2158
2159 spin_unlock_bh(&tot_stats->lock);
2160
b17fc996
HS		 2161/* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2162 seq_puts(seq,
2163 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2164 seq_printf(seq, " %8X %8X %8X %16X %16X\n",
ea9f22cc
JA		 2165 rates.cps,
2166 rates.inpps,
2167 rates.outpps,
2168 rates.inbps,
2169 rates.outbps);
b17fc996
HS		 2170
2171 return 0;
2172}
2173
2174static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2175{
2176 return single_open_net(inode, file, ip_vs_stats_percpu_show);
2177}
2178
2179static const struct file_operations ip_vs_stats_percpu_fops = {
2180 .owner = THIS_MODULE,
2181 .open = ip_vs_stats_percpu_seq_open,
2182 .read = seq_read,
2183 .llseek = seq_lseek,
0f08190f 2184 .release = single_release_net,
b17fc996 2185};
1da177e4
LT		 2186#endif
2187
2188/*
2189 * Set timeout values for tcp tcpfin udp in the timeout_table.
2190 */
9330419d 2191static int ip_vs_set_timeout(struct net *net, struct ip_vs_timeout_user *u)
1da177e4 2192{
091bb34c 2193#if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
9330419d 2194 struct ip_vs_proto_data *pd;
091bb34c 2195#endif
9330419d 2196
1da177e4
LT		 2197 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2198 u->tcp_timeout,
2199 u->tcp_fin_timeout,
2200 u->udp_timeout);
2201
2202#ifdef CONFIG_IP_VS_PROTO_TCP
2203 if (u->tcp_timeout) {
9330419d
HS		 2204 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2205 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
1da177e4
LT		 2206 = u->tcp_timeout * HZ;
2207 }
2208
2209 if (u->tcp_fin_timeout) {
9330419d
HS		 2210 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2211 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
1da177e4
LT		 2212 = u->tcp_fin_timeout * HZ;
2213 }
2214#endif
2215
2216#ifdef CONFIG_IP_VS_PROTO_UDP
2217 if (u->udp_timeout) {
9330419d
HS		 2218 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2219 pd->timeout_table[IP_VS_UDP_S_NORMAL]
1da177e4
LT		 2220 = u->udp_timeout * HZ;
2221 }
2222#endif
2223 return 0;
2224}
2225
2226
2227#define SET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2228#define SERVICE_ARG_LEN (sizeof(struct ip_vs_service_user))
2229#define SVCDEST_ARG_LEN (sizeof(struct ip_vs_service_user) + \
2230 sizeof(struct ip_vs_dest_user))
2231#define TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2232#define DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user))
2233#define MAX_ARG_LEN SVCDEST_ARG_LEN
2234
9b5b5cff 2235static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
1da177e4
LT		 2236 [SET_CMDID(IP_VS_SO_SET_ADD)] = SERVICE_ARG_LEN,
2237 [SET_CMDID(IP_VS_SO_SET_EDIT)] = SERVICE_ARG_LEN,
2238 [SET_CMDID(IP_VS_SO_SET_DEL)] = SERVICE_ARG_LEN,
2239 [SET_CMDID(IP_VS_SO_SET_FLUSH)] = 0,
2240 [SET_CMDID(IP_VS_SO_SET_ADDDEST)] = SVCDEST_ARG_LEN,
2241 [SET_CMDID(IP_VS_SO_SET_DELDEST)] = SVCDEST_ARG_LEN,
2242 [SET_CMDID(IP_VS_SO_SET_EDITDEST)] = SVCDEST_ARG_LEN,
2243 [SET_CMDID(IP_VS_SO_SET_TIMEOUT)] = TIMEOUT_ARG_LEN,
2244 [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)] = DAEMON_ARG_LEN,
2245 [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)] = DAEMON_ARG_LEN,
2246 [SET_CMDID(IP_VS_SO_SET_ZERO)] = SERVICE_ARG_LEN,
2247};
2248
c860c6b1
JV		 2249static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2250 struct ip_vs_service_user *usvc_compat)
2251{
0d1e71b0
SH		 2252 memset(usvc, 0, sizeof(*usvc));
2253
c860c6b1
JV		 2254 usvc->af = AF_INET;
2255 usvc->protocol = usvc_compat->protocol;
2256 usvc->addr.ip = usvc_compat->addr;
2257 usvc->port = usvc_compat->port;
2258 usvc->fwmark = usvc_compat->fwmark;
2259
2260 /* Deep copy of sched_name is not needed here */
2261 usvc->sched_name = usvc_compat->sched_name;
2262
2263 usvc->flags = usvc_compat->flags;
2264 usvc->timeout = usvc_compat->timeout;
2265 usvc->netmask = usvc_compat->netmask;
2266}
2267
2268static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2269 struct ip_vs_dest_user *udest_compat)
2270{
0d1e71b0
SH		 2271 memset(udest, 0, sizeof(*udest));
2272
c860c6b1
JV		 2273 udest->addr.ip = udest_compat->addr;
2274 udest->port = udest_compat->port;
2275 udest->conn_flags = udest_compat->conn_flags;
2276 udest->weight = udest_compat->weight;
2277 udest->u_threshold = udest_compat->u_threshold;
2278 udest->l_threshold = udest_compat->l_threshold;
2279}
2280
1da177e4
LT		 2281static int
2282do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2283{
fc723250 2284 struct net *net = sock_net(sk);
1da177e4
LT		 2285 int ret;
2286 unsigned char arg[MAX_ARG_LEN];
c860c6b1
JV		 2287 struct ip_vs_service_user *usvc_compat;
2288 struct ip_vs_service_user_kern usvc;
1da177e4 2289 struct ip_vs_service *svc;
c860c6b1
JV		 2290 struct ip_vs_dest_user *udest_compat;
2291 struct ip_vs_dest_user_kern udest;
ae1d48b2 2292 struct netns_ipvs *ipvs = net_ipvs(net);
1da177e4
LT		 2293
2294 if (!capable(CAP_NET_ADMIN))
2295 return -EPERM;
2296
04bcef2a
AV		 2297 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2298 return -EINVAL;
2299 if (len < 0 || len > MAX_ARG_LEN)
2300 return -EINVAL;
1da177e4 2301 if (len != set_arglen[SET_CMDID(cmd)]) {
1e3e238e
HE		 2302 pr_err("set_ctl: len %u != %u\n",
2303 len, set_arglen[SET_CMDID(cmd)]);
1da177e4
LT		 2304 return -EINVAL;
2305 }
2306
2307 if (copy_from_user(arg, user, len) != 0)
2308 return -EFAULT;
2309
2310 /* increase the module use count */
2311 ip_vs_use_count_inc();
2312
ae1d48b2
HS		 2313 /* Handle daemons since they have another lock */
2314 if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2315 cmd == IP_VS_SO_SET_STOPDAEMON) {
2316 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2317
2318 if (mutex_lock_interruptible(&ipvs->sync_mutex)) {
2319 ret = -ERESTARTSYS;
2320 goto out_dec;
2321 }
2322 if (cmd == IP_VS_SO_SET_STARTDAEMON)
2323 ret = start_sync_thread(net, dm->state, dm->mcast_ifn,
2324 dm->syncid);
2325 else
2326 ret = stop_sync_thread(net, dm->state);
2327 mutex_unlock(&ipvs->sync_mutex);
2328 goto out_dec;
2329 }
2330
14cc3e2b 2331 if (mutex_lock_interruptible(&__ip_vs_mutex)) {
1da177e4
LT		 2332 ret = -ERESTARTSYS;
2333 goto out_dec;
2334 }
2335
2336 if (cmd == IP_VS_SO_SET_FLUSH) {
2337 /* Flush the virtual service */
fc723250 2338 ret = ip_vs_flush(net);
1da177e4
LT		 2339 goto out_unlock;
2340 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2341 /* Set timeout values for (tcp tcpfin udp) */
9330419d 2342 ret = ip_vs_set_timeout(net, (struct ip_vs_timeout_user *)arg);
1da177e4 2343 goto out_unlock;
1da177e4
LT		 2344 }
2345
c860c6b1
JV		 2346 usvc_compat = (struct ip_vs_service_user *)arg;
2347 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2348
2349 /* We only use the new structs internally, so copy userspace compat
2350 * structs to extended internal versions */
2351 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2352 ip_vs_copy_udest_compat(&udest, udest_compat);
1da177e4
LT		 2353
2354 if (cmd == IP_VS_SO_SET_ZERO) {
2355 /* if no service address is set, zero counters in all */
c860c6b1 2356 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
fc723250 2357 ret = ip_vs_zero_all(net);
1da177e4
LT		 2358 goto out_unlock;
2359 }
2360 }
2361
2906f66a
VMR		 2362 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2363 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2364 usvc.protocol != IPPROTO_SCTP) {
1e3e238e
HE		 2365 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2366 usvc.protocol, &usvc.addr.ip,
2367 ntohs(usvc.port), usvc.sched_name);
1da177e4
LT		 2368 ret = -EFAULT;
2369 goto out_unlock;
2370 }
2371
2372 /* Lookup the exact service by <protocol, addr, port> or fwmark */
c860c6b1 2373 if (usvc.fwmark == 0)
fc723250 2374 svc = __ip_vs_service_find(net, usvc.af, usvc.protocol,
26c15cfd 2375 &usvc.addr, usvc.port);
1da177e4 2376 else
fc723250 2377 svc = __ip_vs_svc_fwm_find(net, usvc.af, usvc.fwmark);
1da177e4
LT		 2378
2379 if (cmd != IP_VS_SO_SET_ADD
c860c6b1 2380 && (svc == NULL || svc->protocol != usvc.protocol)) {
1da177e4 2381 ret = -ESRCH;
26c15cfd 2382 goto out_unlock;
1da177e4
LT		 2383 }
2384
2385 switch (cmd) {
2386 case IP_VS_SO_SET_ADD:
2387 if (svc != NULL)
2388 ret = -EEXIST;
2389 else
fc723250 2390 ret = ip_vs_add_service(net, &usvc, &svc);
1da177e4
LT		 2391 break;
2392 case IP_VS_SO_SET_EDIT:
c860c6b1 2393 ret = ip_vs_edit_service(svc, &usvc);
1da177e4
LT		 2394 break;
2395 case IP_VS_SO_SET_DEL:
2396 ret = ip_vs_del_service(svc);
2397 if (!ret)
2398 goto out_unlock;
2399 break;
2400 case IP_VS_SO_SET_ZERO:
2401 ret = ip_vs_zero_service(svc);
2402 break;
2403 case IP_VS_SO_SET_ADDDEST:
c860c6b1 2404 ret = ip_vs_add_dest(svc, &udest);
1da177e4
LT		 2405 break;
2406 case IP_VS_SO_SET_EDITDEST:
c860c6b1 2407 ret = ip_vs_edit_dest(svc, &udest);
1da177e4
LT		 2408 break;
2409 case IP_VS_SO_SET_DELDEST:
c860c6b1 2410 ret = ip_vs_del_dest(svc, &udest);
1da177e4
LT		 2411 break;
2412 default:
2413 ret = -EINVAL;
2414 }
2415
1da177e4 2416 out_unlock:
14cc3e2b 2417 mutex_unlock(&__ip_vs_mutex);
1da177e4
LT		 2418 out_dec:
2419 /* decrease the module use count */
2420 ip_vs_use_count_dec();
2421
2422 return ret;
2423}
2424
2425
1da177e4
LT		 2426static void
2427ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2428{
2429 dst->protocol = src->protocol;
e7ade46a 2430 dst->addr = src->addr.ip;
1da177e4
LT		 2431 dst->port = src->port;
2432 dst->fwmark = src->fwmark;
4da62fc7 2433 strlcpy(dst->sched_name, src->scheduler->name, sizeof(dst->sched_name));
1da177e4
LT		 2434 dst->flags = src->flags;
2435 dst->timeout = src->timeout / HZ;
2436 dst->netmask = src->netmask;
2437 dst->num_dests = src->num_dests;
2438 ip_vs_copy_stats(&dst->stats, &src->stats);
2439}
2440
2441static inline int
fc723250
HS		 2442__ip_vs_get_service_entries(struct net *net,
2443 const struct ip_vs_get_services *get,
1da177e4
LT		 2444 struct ip_vs_get_services __user *uptr)
2445{
2446 int idx, count=0;
2447 struct ip_vs_service *svc;
2448 struct ip_vs_service_entry entry;
2449 int ret = 0;
2450
2451 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2452 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
f94fd041 2453 /* Only expose IPv4 entries to old interface */
fc723250 2454 if (svc->af != AF_INET || !net_eq(svc->net, net))
f94fd041
JV		 2455 continue;
2456
1da177e4
LT		 2457 if (count >= get->num_services)
2458 goto out;
4da62fc7 2459 memset(&entry, 0, sizeof(entry));
1da177e4
LT		 2460 ip_vs_copy_service(&entry, svc);
2461 if (copy_to_user(&uptr->entrytable[count],
2462 &entry, sizeof(entry))) {
2463 ret = -EFAULT;
2464 goto out;
2465 }
2466 count++;
2467 }
2468 }
2469
2470 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2471 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
f94fd041 2472 /* Only expose IPv4 entries to old interface */
fc723250 2473 if (svc->af != AF_INET || !net_eq(svc->net, net))
f94fd041
JV		 2474 continue;
2475
1da177e4
LT		 2476 if (count >= get->num_services)
2477 goto out;
4da62fc7 2478 memset(&entry, 0, sizeof(entry));
1da177e4
LT		 2479 ip_vs_copy_service(&entry, svc);
2480 if (copy_to_user(&uptr->entrytable[count],
2481 &entry, sizeof(entry))) {
2482 ret = -EFAULT;
2483 goto out;
2484 }
2485 count++;
2486 }
2487 }
552ad65a 2488out:
1da177e4
LT		 2489 return ret;
2490}
2491
2492static inline int
fc723250 2493__ip_vs_get_dest_entries(struct net *net, const struct ip_vs_get_dests *get,
1da177e4
LT		 2494 struct ip_vs_get_dests __user *uptr)
2495{
2496 struct ip_vs_service *svc;
b18610de 2497 union nf_inet_addr addr = { .ip = get->addr };
1da177e4
LT		 2498 int ret = 0;
2499
2500 if (get->fwmark)
fc723250 2501 svc = __ip_vs_svc_fwm_find(net, AF_INET, get->fwmark);
1da177e4 2502 else
fc723250 2503 svc = __ip_vs_service_find(net, AF_INET, get->protocol, &addr,
26c15cfd 2504 get->port);
b18610de 2505
1da177e4
LT		 2506 if (svc) {
2507 int count = 0;
2508 struct ip_vs_dest *dest;
2509 struct ip_vs_dest_entry entry;
2510
2511 list_for_each_entry(dest, &svc->destinations, n_list) {
2512 if (count >= get->num_dests)
2513 break;
2514
e7ade46a 2515 entry.addr = dest->addr.ip;
1da177e4
LT		 2516 entry.port = dest->port;
2517 entry.conn_flags = atomic_read(&dest->conn_flags);
2518 entry.weight = atomic_read(&dest->weight);
2519 entry.u_threshold = dest->u_threshold;
2520 entry.l_threshold = dest->l_threshold;
2521 entry.activeconns = atomic_read(&dest->activeconns);
2522 entry.inactconns = atomic_read(&dest->inactconns);
2523 entry.persistconns = atomic_read(&dest->persistconns);
2524 ip_vs_copy_stats(&entry.stats, &dest->stats);
2525 if (copy_to_user(&uptr->entrytable[count],
2526 &entry, sizeof(entry))) {
2527 ret = -EFAULT;
2528 break;
2529 }
2530 count++;
2531 }
1da177e4
LT		 2532 } else
2533 ret = -ESRCH;
2534 return ret;
2535}
2536
2537static inline void
9330419d 2538__ip_vs_get_timeouts(struct net *net, struct ip_vs_timeout_user *u)
1da177e4 2539{
091bb34c 2540#if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
9330419d 2541 struct ip_vs_proto_data *pd;
091bb34c 2542#endif
9330419d 2543
1da177e4 2544#ifdef CONFIG_IP_VS_PROTO_TCP
9330419d
HS		 2545 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2546 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2547 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
1da177e4
LT		 2548#endif
2549#ifdef CONFIG_IP_VS_PROTO_UDP
9330419d 2550 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
1da177e4 2551 u->udp_timeout =
9330419d 2552 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
1da177e4
LT		 2553#endif
2554}
2555
2556
2557#define GET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2558#define GET_INFO_ARG_LEN (sizeof(struct ip_vs_getinfo))
2559#define GET_SERVICES_ARG_LEN (sizeof(struct ip_vs_get_services))
2560#define GET_SERVICE_ARG_LEN (sizeof(struct ip_vs_service_entry))
2561#define GET_DESTS_ARG_LEN (sizeof(struct ip_vs_get_dests))
2562#define GET_TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2563#define GET_DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user) * 2)
2564
9b5b5cff 2565static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
1da177e4
LT		 2566 [GET_CMDID(IP_VS_SO_GET_VERSION)] = 64,
2567 [GET_CMDID(IP_VS_SO_GET_INFO)] = GET_INFO_ARG_LEN,
2568 [GET_CMDID(IP_VS_SO_GET_SERVICES)] = GET_SERVICES_ARG_LEN,
2569 [GET_CMDID(IP_VS_SO_GET_SERVICE)] = GET_SERVICE_ARG_LEN,
2570 [GET_CMDID(IP_VS_SO_GET_DESTS)] = GET_DESTS_ARG_LEN,
2571 [GET_CMDID(IP_VS_SO_GET_TIMEOUT)] = GET_TIMEOUT_ARG_LEN,
2572 [GET_CMDID(IP_VS_SO_GET_DAEMON)] = GET_DAEMON_ARG_LEN,
2573};
2574
2575static int
2576do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2577{
2578 unsigned char arg[128];
2579 int ret = 0;
04bcef2a 2580 unsigned int copylen;
fc723250 2581 struct net *net = sock_net(sk);
f131315f 2582 struct netns_ipvs *ipvs = net_ipvs(net);
1da177e4 2583
fc723250 2584 BUG_ON(!net);
1da177e4
LT		 2585 if (!capable(CAP_NET_ADMIN))
2586 return -EPERM;
2587
04bcef2a
AV		 2588 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2589 return -EINVAL;
2590
1da177e4 2591 if (*len < get_arglen[GET_CMDID(cmd)]) {
1e3e238e
HE		 2592 pr_err("get_ctl: len %u < %u\n",
2593 *len, get_arglen[GET_CMDID(cmd)]);
1da177e4
LT		 2594 return -EINVAL;
2595 }
2596
04bcef2a
AV		 2597 copylen = get_arglen[GET_CMDID(cmd)];
2598 if (copylen > 128)
2599 return -EINVAL;
2600
2601 if (copy_from_user(arg, user, copylen) != 0)
1da177e4 2602 return -EFAULT;
ae1d48b2
HS		 2603 /*
2604 * Handle daemons first since it has its own locking
2605 */
2606 if (cmd == IP_VS_SO_GET_DAEMON) {
2607 struct ip_vs_daemon_user d[2];
2608
2609 memset(&d, 0, sizeof(d));
2610 if (mutex_lock_interruptible(&ipvs->sync_mutex))
2611 return -ERESTARTSYS;
2612
2613 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2614 d[0].state = IP_VS_STATE_MASTER;
2615 strlcpy(d[0].mcast_ifn, ipvs->master_mcast_ifn,
2616 sizeof(d[0].mcast_ifn));
2617 d[0].syncid = ipvs->master_syncid;
2618 }
2619 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2620 d[1].state = IP_VS_STATE_BACKUP;
2621 strlcpy(d[1].mcast_ifn, ipvs->backup_mcast_ifn,
2622 sizeof(d[1].mcast_ifn));
2623 d[1].syncid = ipvs->backup_syncid;
2624 }
2625 if (copy_to_user(user, &d, sizeof(d)) != 0)
2626 ret = -EFAULT;
2627 mutex_unlock(&ipvs->sync_mutex);
2628 return ret;
2629 }
1da177e4 2630
14cc3e2b 2631 if (mutex_lock_interruptible(&__ip_vs_mutex))
1da177e4
LT		 2632 return -ERESTARTSYS;
2633
2634 switch (cmd) {
2635 case IP_VS_SO_GET_VERSION:
2636 {
2637 char buf[64];
2638
2639 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
6f7edb48 2640 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
1da177e4
LT		 2641 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2642 ret = -EFAULT;
2643 goto out;
2644 }
2645 *len = strlen(buf)+1;
2646 }
2647 break;
2648
2649 case IP_VS_SO_GET_INFO:
2650 {
2651 struct ip_vs_getinfo info;
2652 info.version = IP_VS_VERSION_CODE;
6f7edb48 2653 info.size = ip_vs_conn_tab_size;
a0840e2e 2654 info.num_services = ipvs->num_services;
1da177e4
LT		 2655 if (copy_to_user(user, &info, sizeof(info)) != 0)
2656 ret = -EFAULT;
2657 }
2658 break;
2659
2660 case IP_VS_SO_GET_SERVICES:
2661 {
2662 struct ip_vs_get_services *get;
2663 int size;
2664
2665 get = (struct ip_vs_get_services *)arg;
2666 size = sizeof(*get) +
2667 sizeof(struct ip_vs_service_entry) * get->num_services;
2668 if (*len != size) {
1e3e238e 2669 pr_err("length: %u != %u\n", *len, size);
1da177e4
LT		 2670 ret = -EINVAL;
2671 goto out;
2672 }
fc723250 2673 ret = __ip_vs_get_service_entries(net, get, user);
1da177e4
LT		 2674 }
2675 break;
2676
2677 case IP_VS_SO_GET_SERVICE:
2678 {
2679 struct ip_vs_service_entry *entry;
2680 struct ip_vs_service *svc;
b18610de 2681 union nf_inet_addr addr;
1da177e4
LT		 2682
2683 entry = (struct ip_vs_service_entry *)arg;
b18610de 2684 addr.ip = entry->addr;
1da177e4 2685 if (entry->fwmark)
fc723250 2686 svc = __ip_vs_svc_fwm_find(net, AF_INET, entry->fwmark);
1da177e4 2687 else
fc723250
HS		 2688 svc = __ip_vs_service_find(net, AF_INET,
2689 entry->protocol, &addr,
2690 entry->port);
1da177e4
LT		 2691 if (svc) {
2692 ip_vs_copy_service(entry, svc);
2693 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2694 ret = -EFAULT;
1da177e4
LT		 2695 } else
2696 ret = -ESRCH;
2697 }
2698 break;
2699
2700 case IP_VS_SO_GET_DESTS:
2701 {
2702 struct ip_vs_get_dests *get;
2703 int size;
2704
2705 get = (struct ip_vs_get_dests *)arg;
2706 size = sizeof(*get) +
2707 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2708 if (*len != size) {
1e3e238e 2709 pr_err("length: %u != %u\n", *len, size);
1da177e4
LT		 2710 ret = -EINVAL;
2711 goto out;
2712 }
fc723250 2713 ret = __ip_vs_get_dest_entries(net, get, user);
1da177e4
LT		 2714 }
2715 break;
2716
2717 case IP_VS_SO_GET_TIMEOUT:
2718 {
2719 struct ip_vs_timeout_user t;
2720
9330419d 2721 __ip_vs_get_timeouts(net, &t);
1da177e4
LT		 2722 if (copy_to_user(user, &t, sizeof(t)) != 0)
2723 ret = -EFAULT;
2724 }
2725 break;
2726
1da177e4
LT		 2727 default:
2728 ret = -EINVAL;
2729 }
2730
552ad65a 2731out:
14cc3e2b 2732 mutex_unlock(&__ip_vs_mutex);
1da177e4
LT		 2733 return ret;
2734}
2735
2736
2737static struct nf_sockopt_ops ip_vs_sockopts = {
2738 .pf = PF_INET,
2739 .set_optmin = IP_VS_BASE_CTL,
2740 .set_optmax = IP_VS_SO_SET_MAX+1,
2741 .set = do_ip_vs_set_ctl,
2742 .get_optmin = IP_VS_BASE_CTL,
2743 .get_optmax = IP_VS_SO_GET_MAX+1,
2744 .get = do_ip_vs_get_ctl,
16fcec35 2745 .owner = THIS_MODULE,
1da177e4
LT		 2746};
2747
9a812198
JV		 2748/*
2749 * Generic Netlink interface
2750 */
2751
2752/* IPVS genetlink family */
2753static struct genl_family ip_vs_genl_family = {
2754 .id = GENL_ID_GENERATE,
2755 .hdrsize = 0,
2756 .name = IPVS_GENL_NAME,
2757 .version = IPVS_GENL_VERSION,
2758 .maxattr = IPVS_CMD_MAX,
c6d2d445 2759 .netnsok = true, /* Make ipvsadm to work on netns */
9a812198
JV		 2760};
2761
2762/* Policy used for first-level command attributes */
2763static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2764 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2765 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2766 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2767 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2768 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2769 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2770};
2771
2772/* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2773static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2774 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2775 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2776 .len = IP_VS_IFNAME_MAXLEN },
2777 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2778};
2779
2780/* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2781static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2782 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2783 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2784 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2785 .len = sizeof(union nf_inet_addr) },
2786 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2787 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2788 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2789 .len = IP_VS_SCHEDNAME_MAXLEN },
0d1e71b0
SH		 2790 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2791 .len = IP_VS_PENAME_MAXLEN },
9a812198
JV		 2792 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2793 .len = sizeof(struct ip_vs_flags) },
2794 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2795 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2796 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2797};
2798
2799/* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2800static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2801 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2802 .len = sizeof(union nf_inet_addr) },
2803 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2804 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2805 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2806 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2807 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2808 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2809 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2810 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2811 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2812};
2813
2814static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2815 struct ip_vs_stats *stats)
2816{
55a3d4e1 2817 struct ip_vs_stats_user ustats;
9a812198
JV		 2818 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2819 if (!nl_stats)
2820 return -EMSGSIZE;
2821
55a3d4e1 2822 ip_vs_copy_stats(&ustats, stats);
9a812198 2823
969e8e25
DM		 2824 if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, ustats.conns) ||
2825 nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, ustats.inpkts) ||
2826 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, ustats.outpkts) ||
2827 nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, ustats.inbytes) ||
2828 nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, ustats.outbytes) ||
2829 nla_put_u32(skb, IPVS_STATS_ATTR_CPS, ustats.cps) ||
2830 nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, ustats.inpps) ||
2831 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, ustats.outpps) ||
2832 nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, ustats.inbps) ||
2833 nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, ustats.outbps))
2834 goto nla_put_failure;
9a812198
JV		 2835 nla_nest_end(skb, nl_stats);
2836
2837 return 0;
2838
2839nla_put_failure:
9a812198
JV		 2840 nla_nest_cancel(skb, nl_stats);
2841 return -EMSGSIZE;
2842}
2843
2844static int ip_vs_genl_fill_service(struct sk_buff *skb,
2845 struct ip_vs_service *svc)
2846{
2847 struct nlattr *nl_service;
2848 struct ip_vs_flags flags = { .flags = svc->flags,
2849 .mask = ~0 };
2850
2851 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2852 if (!nl_service)
2853 return -EMSGSIZE;
2854
969e8e25
DM		 2855 if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2856 goto nla_put_failure;
9a812198 2857 if (svc->fwmark) {
969e8e25
DM		 2858 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2859 goto nla_put_failure;
9a812198 2860 } else {
969e8e25
DM		 2861 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2862 nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2863 nla_put_u16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2864 goto nla_put_failure;
9a812198
JV		 2865 }
2866
969e8e25
DM		 2867 if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, svc->scheduler->name) ||
2868 (svc->pe &&
2869 nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, svc->pe->name)) ||
2870 nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2871 nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2872 nla_put_u32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
2873 goto nla_put_failure;
9a812198
JV		 2874 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2875 goto nla_put_failure;
2876
2877 nla_nest_end(skb, nl_service);
2878
2879 return 0;
2880
2881nla_put_failure:
2882 nla_nest_cancel(skb, nl_service);
2883 return -EMSGSIZE;
2884}
2885
2886static int ip_vs_genl_dump_service(struct sk_buff *skb,
2887 struct ip_vs_service *svc,
2888 struct netlink_callback *cb)
2889{
2890 void *hdr;
2891
2892 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2893 &ip_vs_genl_family, NLM_F_MULTI,
2894 IPVS_CMD_NEW_SERVICE);
2895 if (!hdr)
2896 return -EMSGSIZE;
2897
2898 if (ip_vs_genl_fill_service(skb, svc) < 0)
2899 goto nla_put_failure;
2900
2901 return genlmsg_end(skb, hdr);
2902
2903nla_put_failure:
2904 genlmsg_cancel(skb, hdr);
2905 return -EMSGSIZE;
2906}
2907
2908static int ip_vs_genl_dump_services(struct sk_buff *skb,
2909 struct netlink_callback *cb)
2910{
2911 int idx = 0, i;
2912 int start = cb->args[0];
2913 struct ip_vs_service *svc;
fc723250 2914 struct net *net = skb_sknet(skb);
9a812198
JV		 2915
2916 mutex_lock(&__ip_vs_mutex);
2917 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2918 list_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
fc723250 2919 if (++idx <= start || !net_eq(svc->net, net))
9a812198
JV		 2920 continue;
2921 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2922 idx--;
2923 goto nla_put_failure;
2924 }
2925 }
2926 }
2927
2928 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2929 list_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
fc723250 2930 if (++idx <= start || !net_eq(svc->net, net))
9a812198
JV		 2931 continue;
2932 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2933 idx--;
2934 goto nla_put_failure;
2935 }
2936 }
2937 }
2938
2939nla_put_failure:
2940 mutex_unlock(&__ip_vs_mutex);
2941 cb->args[0] = idx;
2942
2943 return skb->len;
2944}
2945
fc723250
HS		 2946static int ip_vs_genl_parse_service(struct net *net,
2947 struct ip_vs_service_user_kern *usvc,
26c15cfd
JA		 2948 struct nlattr *nla, int full_entry,
2949 struct ip_vs_service **ret_svc)
9a812198
JV		 2950{
2951 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
2952 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
26c15cfd 2953 struct ip_vs_service *svc;
9a812198
JV		 2954
2955 /* Parse mandatory identifying service fields first */
2956 if (nla == NULL ||
2957 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
2958 return -EINVAL;
2959
2960 nla_af = attrs[IPVS_SVC_ATTR_AF];
2961 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
2962 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
2963 nla_port = attrs[IPVS_SVC_ATTR_PORT];
2964 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
2965
2966 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
2967 return -EINVAL;
2968
258c8893
SH		 2969 memset(usvc, 0, sizeof(*usvc));
2970
c860c6b1 2971 usvc->af = nla_get_u16(nla_af);
f94fd041
JV		 2972#ifdef CONFIG_IP_VS_IPV6
2973 if (usvc->af != AF_INET && usvc->af != AF_INET6)
2974#else
2975 if (usvc->af != AF_INET)
2976#endif
9a812198
JV		 2977 return -EAFNOSUPPORT;
2978
2979 if (nla_fwmark) {
2980 usvc->protocol = IPPROTO_TCP;
2981 usvc->fwmark = nla_get_u32(nla_fwmark);
2982 } else {
2983 usvc->protocol = nla_get_u16(nla_protocol);
2984 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
2985 usvc->port = nla_get_u16(nla_port);
2986 usvc->fwmark = 0;
2987 }
2988
26c15cfd 2989 if (usvc->fwmark)
fc723250 2990 svc = __ip_vs_svc_fwm_find(net, usvc->af, usvc->fwmark);
26c15cfd 2991 else
fc723250 2992 svc = __ip_vs_service_find(net, usvc->af, usvc->protocol,
26c15cfd
JA		 2993 &usvc->addr, usvc->port);
2994 *ret_svc = svc;
2995
9a812198
JV		 2996 /* If a full entry was requested, check for the additional fields */
2997 if (full_entry) {
0d1e71b0 2998 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
9a812198
JV		 2999 *nla_netmask;
3000 struct ip_vs_flags flags;
9a812198
JV		 3001
3002 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
0d1e71b0 3003 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
9a812198
JV		 3004 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3005 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3006 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3007
3008 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3009 return -EINVAL;
3010
3011 nla_memcpy(&flags, nla_flags, sizeof(flags));
3012
3013 /* prefill flags from service if it already exists */
26c15cfd 3014 if (svc)
9a812198 3015 usvc->flags = svc->flags;
9a812198
JV		 3016
3017 /* set new flags from userland */
3018 usvc->flags = (usvc->flags & ~flags.mask) |
3019 (flags.flags & flags.mask);
c860c6b1 3020 usvc->sched_name = nla_data(nla_sched);
0d1e71b0 3021 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
9a812198
JV		 3022 usvc->timeout = nla_get_u32(nla_timeout);
3023 usvc->netmask = nla_get_u32(nla_netmask);
3024 }
3025
3026 return 0;
3027}
3028
fc723250
HS		 3029static struct ip_vs_service *ip_vs_genl_find_service(struct net *net,
3030 struct nlattr *nla)
9a812198 3031{
c860c6b1 3032 struct ip_vs_service_user_kern usvc;
26c15cfd 3033 struct ip_vs_service *svc;
9a812198
JV		 3034 int ret;
3035
fc723250 3036 ret = ip_vs_genl_parse_service(net, &usvc, nla, 0, &svc);
26c15cfd 3037 return ret ? ERR_PTR(ret) : svc;
9a812198
JV		 3038}
3039
3040static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3041{
3042 struct nlattr *nl_dest;
3043
3044 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3045 if (!nl_dest)
3046 return -EMSGSIZE;
3047
969e8e25
DM		 3048 if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3049 nla_put_u16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3050 nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3051 (atomic_read(&dest->conn_flags) &
3052 IP_VS_CONN_F_FWD_MASK)) ||
3053 nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3054 atomic_read(&dest->weight)) ||
3055 nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3056 nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3057 nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3058 atomic_read(&dest->activeconns)) ||
3059 nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3060 atomic_read(&dest->inactconns)) ||
3061 nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3062 atomic_read(&dest->persistconns)))
3063 goto nla_put_failure;
9a812198
JV		 3064 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
3065 goto nla_put_failure;
3066
3067 nla_nest_end(skb, nl_dest);
3068
3069 return 0;
3070
3071nla_put_failure:
3072 nla_nest_cancel(skb, nl_dest);
3073 return -EMSGSIZE;
3074}
3075
3076static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3077 struct netlink_callback *cb)
3078{
3079 void *hdr;
3080
3081 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
3082 &ip_vs_genl_family, NLM_F_MULTI,
3083 IPVS_CMD_NEW_DEST);
3084 if (!hdr)
3085 return -EMSGSIZE;
3086
3087 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3088 goto nla_put_failure;
3089
3090 return genlmsg_end(skb, hdr);
3091
3092nla_put_failure:
3093 genlmsg_cancel(skb, hdr);
3094 return -EMSGSIZE;
3095}
3096
3097static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3098 struct netlink_callback *cb)
3099{
3100 int idx = 0;
3101 int start = cb->args[0];
3102 struct ip_vs_service *svc;
3103 struct ip_vs_dest *dest;
3104 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
a0840e2e 3105 struct net *net = skb_sknet(skb);
9a812198
JV		 3106
3107 mutex_lock(&__ip_vs_mutex);
3108
3109 /* Try to find the service for which to dump destinations */
3110 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3111 IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3112 goto out_err;
3113
a0840e2e 3114
fc723250 3115 svc = ip_vs_genl_find_service(net, attrs[IPVS_CMD_ATTR_SERVICE]);
9a812198
JV		 3116 if (IS_ERR(svc) || svc == NULL)
3117 goto out_err;
3118
3119 /* Dump the destinations */
3120 list_for_each_entry(dest, &svc->destinations, n_list) {
3121 if (++idx <= start)
3122 continue;
3123 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3124 idx--;
3125 goto nla_put_failure;
3126 }
3127 }
3128
3129nla_put_failure:
3130 cb->args[0] = idx;
9a812198
JV		 3131
3132out_err:
3133 mutex_unlock(&__ip_vs_mutex);
3134
3135 return skb->len;
3136}
3137
c860c6b1 3138static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
9a812198
JV		 3139 struct nlattr *nla, int full_entry)
3140{
3141 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3142 struct nlattr *nla_addr, *nla_port;
3143
3144 /* Parse mandatory identifying destination fields first */
3145 if (nla == NULL ||
3146 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3147 return -EINVAL;
3148
3149 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3150 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3151
3152 if (!(nla_addr && nla_port))
3153 return -EINVAL;
3154
258c8893
SH		 3155 memset(udest, 0, sizeof(*udest));
3156
9a812198
JV		 3157 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3158 udest->port = nla_get_u16(nla_port);
3159
3160 /* If a full entry was requested, check for the additional fields */
3161 if (full_entry) {
3162 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3163 *nla_l_thresh;
3164
3165 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3166 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3167 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3168 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3169
3170 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3171 return -EINVAL;
3172
3173 udest->conn_flags = nla_get_u32(nla_fwd)
3174 & IP_VS_CONN_F_FWD_MASK;
3175 udest->weight = nla_get_u32(nla_weight);
3176 udest->u_threshold = nla_get_u32(nla_u_thresh);
3177 udest->l_threshold = nla_get_u32(nla_l_thresh);
3178 }
3179
3180 return 0;
3181}
3182
3183static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __be32 state,
3184 const char *mcast_ifn, __be32 syncid)
3185{
3186 struct nlattr *nl_daemon;
3187
3188 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3189 if (!nl_daemon)
3190 return -EMSGSIZE;
3191
969e8e25
DM		 3192 if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3193 nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn) ||
3194 nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid))
3195 goto nla_put_failure;
9a812198
JV		 3196 nla_nest_end(skb, nl_daemon);
3197
3198 return 0;
3199
3200nla_put_failure:
3201 nla_nest_cancel(skb, nl_daemon);
3202 return -EMSGSIZE;
3203}
3204
3205static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __be32 state,
3206 const char *mcast_ifn, __be32 syncid,
3207 struct netlink_callback *cb)
3208{
3209 void *hdr;
3210 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
3211 &ip_vs_genl_family, NLM_F_MULTI,
3212 IPVS_CMD_NEW_DAEMON);
3213 if (!hdr)
3214 return -EMSGSIZE;
3215
3216 if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
3217 goto nla_put_failure;
3218
3219 return genlmsg_end(skb, hdr);
3220
3221nla_put_failure:
3222 genlmsg_cancel(skb, hdr);
3223 return -EMSGSIZE;
3224}
3225
3226static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3227 struct netlink_callback *cb)
3228{
a09d1977 3229 struct net *net = skb_sknet(skb);
f131315f
HS		 3230 struct netns_ipvs *ipvs = net_ipvs(net);
3231
ae1d48b2 3232 mutex_lock(&ipvs->sync_mutex);
f131315f 3233 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
9a812198 3234 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
f131315f
HS		 3235 ipvs->master_mcast_ifn,
3236 ipvs->master_syncid, cb) < 0)
9a812198
JV		 3237 goto nla_put_failure;
3238
3239 cb->args[0] = 1;
3240 }
3241
f131315f 3242 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
9a812198 3243 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
f131315f
HS		 3244 ipvs->backup_mcast_ifn,
3245 ipvs->backup_syncid, cb) < 0)
9a812198
JV		 3246 goto nla_put_failure;
3247
3248 cb->args[1] = 1;
3249 }
3250
3251nla_put_failure:
ae1d48b2 3252 mutex_unlock(&ipvs->sync_mutex);
9a812198
JV		 3253
3254 return skb->len;
3255}
3256
f131315f 3257static int ip_vs_genl_new_daemon(struct net *net, struct nlattr **attrs)
9a812198
JV		 3258{
3259 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3260 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3261 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3262 return -EINVAL;
3263
f131315f
HS		 3264 return start_sync_thread(net,
3265 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
9a812198
JV		 3266 nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3267 nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3268}
3269
f131315f 3270static int ip_vs_genl_del_daemon(struct net *net, struct nlattr **attrs)
9a812198
JV		 3271{
3272 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3273 return -EINVAL;
3274
f131315f
HS		 3275 return stop_sync_thread(net,
3276 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
9a812198
JV		 3277}
3278
9330419d 3279static int ip_vs_genl_set_config(struct net *net, struct nlattr **attrs)
9a812198
JV		 3280{
3281 struct ip_vs_timeout_user t;
3282
9330419d 3283 __ip_vs_get_timeouts(net, &t);
9a812198
JV		 3284
3285 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3286 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3287
3288 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3289 t.tcp_fin_timeout =
3290 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3291
3292 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3293 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3294
9330419d 3295 return ip_vs_set_timeout(net, &t);
9a812198
JV		 3296}
3297
ae1d48b2 3298static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
9a812198 3299{
9a812198 3300 int ret = 0, cmd;
fc723250 3301 struct net *net;
a0840e2e 3302 struct netns_ipvs *ipvs;
9a812198 3303
fc723250 3304 net = skb_sknet(skb);
a0840e2e 3305 ipvs = net_ipvs(net);
9a812198
JV		 3306 cmd = info->genlhdr->cmd;
3307
ae1d48b2 3308 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
9a812198
JV		 3309 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3310
ae1d48b2 3311 mutex_lock(&ipvs->sync_mutex);
9a812198
JV		 3312 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3313 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3314 info->attrs[IPVS_CMD_ATTR_DAEMON],
3315 ip_vs_daemon_policy)) {
3316 ret = -EINVAL;
3317 goto out;
3318 }
3319
3320 if (cmd == IPVS_CMD_NEW_DAEMON)
f131315f 3321 ret = ip_vs_genl_new_daemon(net, daemon_attrs);
9a812198 3322 else
f131315f 3323 ret = ip_vs_genl_del_daemon(net, daemon_attrs);
ae1d48b2
HS		 3324out:
3325 mutex_unlock(&ipvs->sync_mutex);
3326 }
3327 return ret;
3328}
3329
3330static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3331{
3332 struct ip_vs_service *svc = NULL;
3333 struct ip_vs_service_user_kern usvc;
3334 struct ip_vs_dest_user_kern udest;
3335 int ret = 0, cmd;
3336 int need_full_svc = 0, need_full_dest = 0;
3337 struct net *net;
ae1d48b2
HS		 3338
3339 net = skb_sknet(skb);
ae1d48b2
HS		 3340 cmd = info->genlhdr->cmd;
3341
3342 mutex_lock(&__ip_vs_mutex);
3343
3344 if (cmd == IPVS_CMD_FLUSH) {
3345 ret = ip_vs_flush(net);
3346 goto out;
3347 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3348 ret = ip_vs_genl_set_config(net, info->attrs);
9a812198
JV		 3349 goto out;
3350 } else if (cmd == IPVS_CMD_ZERO &&
3351 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
fc723250 3352 ret = ip_vs_zero_all(net);
9a812198
JV		 3353 goto out;
3354 }
3355
3356 /* All following commands require a service argument, so check if we
3357 * received a valid one. We need a full service specification when
3358 * adding / editing a service. Only identifying members otherwise. */
3359 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3360 need_full_svc = 1;
3361
fc723250 3362 ret = ip_vs_genl_parse_service(net, &usvc,
9a812198 3363 info->attrs[IPVS_CMD_ATTR_SERVICE],
26c15cfd 3364 need_full_svc, &svc);
9a812198
JV		 3365 if (ret)
3366 goto out;
3367
9a812198
JV		 3368 /* Unless we're adding a new service, the service must already exist */
3369 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3370 ret = -ESRCH;
3371 goto out;
3372 }
3373
3374 /* Destination commands require a valid destination argument. For
3375 * adding / editing a destination, we need a full destination
3376 * specification. */
3377 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3378 cmd == IPVS_CMD_DEL_DEST) {
3379 if (cmd != IPVS_CMD_DEL_DEST)
3380 need_full_dest = 1;
3381
3382 ret = ip_vs_genl_parse_dest(&udest,
3383 info->attrs[IPVS_CMD_ATTR_DEST],
3384 need_full_dest);
3385 if (ret)
3386 goto out;
3387 }
3388
3389 switch (cmd) {
3390 case IPVS_CMD_NEW_SERVICE:
3391 if (svc == NULL)
fc723250 3392 ret = ip_vs_add_service(net, &usvc, &svc);
9a812198
JV		 3393 else
3394 ret = -EEXIST;
3395 break;
3396 case IPVS_CMD_SET_SERVICE:
3397 ret = ip_vs_edit_service(svc, &usvc);
3398 break;
3399 case IPVS_CMD_DEL_SERVICE:
3400 ret = ip_vs_del_service(svc);
26c15cfd 3401 /* do not use svc, it can be freed */
9a812198
JV		 3402 break;
3403 case IPVS_CMD_NEW_DEST:
3404 ret = ip_vs_add_dest(svc, &udest);
3405 break;
3406 case IPVS_CMD_SET_DEST:
3407 ret = ip_vs_edit_dest(svc, &udest);
3408 break;
3409 case IPVS_CMD_DEL_DEST:
3410 ret = ip_vs_del_dest(svc, &udest);
3411 break;
3412 case IPVS_CMD_ZERO:
3413 ret = ip_vs_zero_service(svc);
3414 break;
3415 default:
3416 ret = -EINVAL;
3417 }
3418
3419out:
9a812198
JV		 3420 mutex_unlock(&__ip_vs_mutex);
3421
3422 return ret;
3423}
3424
3425static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3426{
3427 struct sk_buff *msg;
3428 void *reply;
3429 int ret, cmd, reply_cmd;
fc723250 3430 struct net *net;
9a812198 3431
fc723250 3432 net = skb_sknet(skb);
9a812198
JV		 3433 cmd = info->genlhdr->cmd;
3434
3435 if (cmd == IPVS_CMD_GET_SERVICE)
3436 reply_cmd = IPVS_CMD_NEW_SERVICE;
3437 else if (cmd == IPVS_CMD_GET_INFO)
3438 reply_cmd = IPVS_CMD_SET_INFO;
3439 else if (cmd == IPVS_CMD_GET_CONFIG)
3440 reply_cmd = IPVS_CMD_SET_CONFIG;
3441 else {
1e3e238e 3442 pr_err("unknown Generic Netlink command\n");
9a812198
JV		 3443 return -EINVAL;
3444 }
3445
3446 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3447 if (!msg)
3448 return -ENOMEM;
3449
3450 mutex_lock(&__ip_vs_mutex);
3451
3452 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3453 if (reply == NULL)
3454 goto nla_put_failure;
3455
3456 switch (cmd) {
3457 case IPVS_CMD_GET_SERVICE:
3458 {
3459 struct ip_vs_service *svc;
3460
fc723250
HS		 3461 svc = ip_vs_genl_find_service(net,
3462 info->attrs[IPVS_CMD_ATTR_SERVICE]);
9a812198
JV		 3463 if (IS_ERR(svc)) {
3464 ret = PTR_ERR(svc);
3465 goto out_err;
3466 } else if (svc) {
3467 ret = ip_vs_genl_fill_service(msg, svc);
9a812198
JV		 3468 if (ret)
3469 goto nla_put_failure;
3470 } else {
3471 ret = -ESRCH;
3472 goto out_err;
3473 }
3474
3475 break;
3476 }
3477
3478 case IPVS_CMD_GET_CONFIG:
3479 {
3480 struct ip_vs_timeout_user t;
3481
9330419d 3482 __ip_vs_get_timeouts(net, &t);
9a812198 3483#ifdef CONFIG_IP_VS_PROTO_TCP
969e8e25
DM		 3484 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3485 t.tcp_timeout) ||
3486 nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3487 t.tcp_fin_timeout))
3488 goto nla_put_failure;
9a812198
JV		 3489#endif
3490#ifdef CONFIG_IP_VS_PROTO_UDP
969e8e25
DM		 3491 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3492 goto nla_put_failure;
9a812198
JV		 3493#endif
3494
3495 break;
3496 }
3497
3498 case IPVS_CMD_GET_INFO:
969e8e25
DM		 3499 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3500 IP_VS_VERSION_CODE) ||
3501 nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3502 ip_vs_conn_tab_size))
3503 goto nla_put_failure;
9a812198
JV		 3504 break;
3505 }
3506
3507 genlmsg_end(msg, reply);
134e6375 3508 ret = genlmsg_reply(msg, info);
9a812198
JV		 3509 goto out;
3510
3511nla_put_failure:
1e3e238e 3512 pr_err("not enough space in Netlink message\n");
9a812198
JV		 3513 ret = -EMSGSIZE;
3514
3515out_err:
3516 nlmsg_free(msg);
3517out:
3518 mutex_unlock(&__ip_vs_mutex);
3519
3520 return ret;
3521}
3522
3523
3524static struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3525 {
3526 .cmd = IPVS_CMD_NEW_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_SET_SERVICE,
3533 .flags = GENL_ADMIN_PERM,
3534 .policy = ip_vs_cmd_policy,
3535 .doit = ip_vs_genl_set_cmd,
3536 },
3537 {
3538 .cmd = IPVS_CMD_DEL_SERVICE,
3539 .flags = GENL_ADMIN_PERM,
3540 .policy = ip_vs_cmd_policy,
3541 .doit = ip_vs_genl_set_cmd,
3542 },
3543 {
3544 .cmd = IPVS_CMD_GET_SERVICE,
3545 .flags = GENL_ADMIN_PERM,
3546 .doit = ip_vs_genl_get_cmd,
3547 .dumpit = ip_vs_genl_dump_services,
3548 .policy = ip_vs_cmd_policy,
3549 },
3550 {
3551 .cmd = IPVS_CMD_NEW_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_SET_DEST,
3558 .flags = GENL_ADMIN_PERM,
3559 .policy = ip_vs_cmd_policy,
3560 .doit = ip_vs_genl_set_cmd,
3561 },
3562 {
3563 .cmd = IPVS_CMD_DEL_DEST,
3564 .flags = GENL_ADMIN_PERM,
3565 .policy = ip_vs_cmd_policy,
3566 .doit = ip_vs_genl_set_cmd,
3567 },
3568 {
3569 .cmd = IPVS_CMD_GET_DEST,
3570 .flags = GENL_ADMIN_PERM,
3571 .policy = ip_vs_cmd_policy,
3572 .dumpit = ip_vs_genl_dump_dests,
3573 },
3574 {
3575 .cmd = IPVS_CMD_NEW_DAEMON,
3576 .flags = GENL_ADMIN_PERM,
3577 .policy = ip_vs_cmd_policy,
ae1d48b2 3578 .doit = ip_vs_genl_set_daemon,
9a812198
JV		 3579 },
3580 {
3581 .cmd = IPVS_CMD_DEL_DAEMON,
3582 .flags = GENL_ADMIN_PERM,
3583 .policy = ip_vs_cmd_policy,
ae1d48b2 3584 .doit = ip_vs_genl_set_daemon,
9a812198
JV		 3585 },
3586 {
3587 .cmd = IPVS_CMD_GET_DAEMON,
3588 .flags = GENL_ADMIN_PERM,
3589 .dumpit = ip_vs_genl_dump_daemons,
3590 },
3591 {
3592 .cmd = IPVS_CMD_SET_CONFIG,
3593 .flags = GENL_ADMIN_PERM,
3594 .policy = ip_vs_cmd_policy,
3595 .doit = ip_vs_genl_set_cmd,
3596 },
3597 {
3598 .cmd = IPVS_CMD_GET_CONFIG,
3599 .flags = GENL_ADMIN_PERM,
3600 .doit = ip_vs_genl_get_cmd,
3601 },
3602 {
3603 .cmd = IPVS_CMD_GET_INFO,
3604 .flags = GENL_ADMIN_PERM,
3605 .doit = ip_vs_genl_get_cmd,
3606 },
3607 {
3608 .cmd = IPVS_CMD_ZERO,
3609 .flags = GENL_ADMIN_PERM,
3610 .policy = ip_vs_cmd_policy,
3611 .doit = ip_vs_genl_set_cmd,
3612 },
3613 {
3614 .cmd = IPVS_CMD_FLUSH,
3615 .flags = GENL_ADMIN_PERM,
3616 .doit = ip_vs_genl_set_cmd,
3617 },
3618};
3619
3620static int __init ip_vs_genl_register(void)
3621{
8f698d54
MM		 3622 return genl_register_family_with_ops(&ip_vs_genl_family,
3623 ip_vs_genl_ops, ARRAY_SIZE(ip_vs_genl_ops));
9a812198
JV		 3624}
3625
3626static void ip_vs_genl_unregister(void)
3627{
3628 genl_unregister_family(&ip_vs_genl_family);
3629}
3630
3631/* End of Generic Netlink interface definitions */
3632
61b1ab45
HS		 3633/*
3634 * per netns intit/exit func.
3635 */
14e40546 3636#ifdef CONFIG_SYSCTL
503cf15a 3637int __net_init ip_vs_control_net_init_sysctl(struct net *net)
61b1ab45 3638{
fc723250
HS		 3639 int idx;
3640 struct netns_ipvs *ipvs = net_ipvs(net);
a0840e2e 3641 struct ctl_table *tbl;
fc723250 3642
a0840e2e
HS		 3643 atomic_set(&ipvs->dropentry, 0);
3644 spin_lock_init(&ipvs->dropentry_lock);
3645 spin_lock_init(&ipvs->droppacket_lock);
3646 spin_lock_init(&ipvs->securetcp_lock);
a0840e2e
HS		 3647
3648 if (!net_eq(net, &init_net)) {
3649 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3650 if (tbl == NULL)
14e40546 3651 return -ENOMEM;
a0840e2e
HS		 3652 } else
3653 tbl = vs_vars;
3654 /* Initialize sysctl defaults */
3655 idx = 0;
3656 ipvs->sysctl_amemthresh = 1024;
3657 tbl[idx++].data = &ipvs->sysctl_amemthresh;
3658 ipvs->sysctl_am_droprate = 10;
3659 tbl[idx++].data = &ipvs->sysctl_am_droprate;
3660 tbl[idx++].data = &ipvs->sysctl_drop_entry;
3661 tbl[idx++].data = &ipvs->sysctl_drop_packet;
3662#ifdef CONFIG_IP_VS_NFCT
3663 tbl[idx++].data = &ipvs->sysctl_conntrack;
3664#endif
3665 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3666 ipvs->sysctl_snat_reroute = 1;
3667 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3668 ipvs->sysctl_sync_ver = 1;
3669 tbl[idx++].data = &ipvs->sysctl_sync_ver;
1c003b15
PNA		 3670 ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3671 tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3672 ipvs->sysctl_sync_sock_size = 0;
3673 tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
a0840e2e
HS		 3674 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3675 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3676 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
59e0350e
SH		 3677 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3678 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
a0840e2e
HS		 3679 tbl[idx].data = &ipvs->sysctl_sync_threshold;
3680 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
3681 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3682
3683
ec8f23ce 3684 ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
0443929f
SH		 3685 if (ipvs->sysctl_hdr == NULL) {
3686 if (!net_eq(net, &init_net))
3687 kfree(tbl);
14e40546 3688 return -ENOMEM;
0443929f 3689 }
6ef757f9 3690 ip_vs_start_estimator(net, &ipvs->tot_stats);
a0840e2e 3691 ipvs->sysctl_tbl = tbl;
f6340ee0
HS		 3692 /* Schedule defense work */
3693 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3694 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
61b1ab45 3695
61b1ab45 3696 return 0;
61b1ab45
HS		 3697}
3698
8f9b9a2f 3699void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net)
61b1ab45 3700{
b17fc996
HS		 3701 struct netns_ipvs *ipvs = net_ipvs(net);
3702
f2431e6e
HS		 3703 cancel_delayed_work_sync(&ipvs->defense_work);
3704 cancel_work_sync(&ipvs->defense_work.work);
a0840e2e 3705 unregister_net_sysctl_table(ipvs->sysctl_hdr);
14e40546
SH		 3706}
3707
3708#else
3709
503cf15a 3710int __net_init ip_vs_control_net_init_sysctl(struct net *net) { return 0; }
8f9b9a2f 3711void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net) { }
14e40546 3712
0443929f 3713#endif
14e40546 3714
7a4f0761
HS		 3715static struct notifier_block ip_vs_dst_notifier = {
3716 .notifier_call = ip_vs_dst_event,
3717};
3718
503cf15a 3719int __net_init ip_vs_control_net_init(struct net *net)
14e40546
SH		 3720{
3721 int idx;
3722 struct netns_ipvs *ipvs = net_ipvs(net);
3723
3458e21c 3724 rwlock_init(&ipvs->rs_lock);
14e40546
SH		 3725
3726 /* Initialize rs_table */
3727 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
3728 INIT_LIST_HEAD(&ipvs->rs_table[idx]);
3729
3730 INIT_LIST_HEAD(&ipvs->dest_trash);
3731 atomic_set(&ipvs->ftpsvc_counter, 0);
3732 atomic_set(&ipvs->nullsvc_counter, 0);
3733
3734 /* procfs stats */
3735 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
0a9ee813 3736 if (!ipvs->tot_stats.cpustats)
14e40546 3737 return -ENOMEM;
0a9ee813 3738
14e40546
SH		 3739 spin_lock_init(&ipvs->tot_stats.lock);
3740
3741 proc_net_fops_create(net, "ip_vs", 0, &ip_vs_info_fops);
3742 proc_net_fops_create(net, "ip_vs_stats", 0, &ip_vs_stats_fops);
3743 proc_net_fops_create(net, "ip_vs_stats_percpu", 0,
3744 &ip_vs_stats_percpu_fops);
3745
503cf15a 3746 if (ip_vs_control_net_init_sysctl(net))
14e40546
SH		 3747 goto err;
3748
3749 return 0;
3750
3751err:
2a0751af 3752 free_percpu(ipvs->tot_stats.cpustats);
61b1ab45
HS		 3753 return -ENOMEM;
3754}
3755
503cf15a 3756void __net_exit ip_vs_control_net_cleanup(struct net *net)
61b1ab45 3757{
b17fc996
HS		 3758 struct netns_ipvs *ipvs = net_ipvs(net);
3759
f2431e6e 3760 ip_vs_trash_cleanup(net);
6ef757f9 3761 ip_vs_stop_estimator(net, &ipvs->tot_stats);
503cf15a 3762 ip_vs_control_net_cleanup_sysctl(net);
b17fc996 3763 proc_net_remove(net, "ip_vs_stats_percpu");
61b1ab45
HS		 3764 proc_net_remove(net, "ip_vs_stats");
3765 proc_net_remove(net, "ip_vs");
2a0751af 3766 free_percpu(ipvs->tot_stats.cpustats);
61b1ab45
HS		 3767}
3768
8537de8a 3769int __init ip_vs_register_nl_ioctl(void)
1da177e4 3770{
fc723250 3771 int ret;
1da177e4 3772
1da177e4
LT		 3773 ret = nf_register_sockopt(&ip_vs_sockopts);
3774 if (ret) {
1e3e238e 3775 pr_err("cannot register sockopt.\n");
7a4f0761 3776 goto err_sock;
1da177e4
LT		 3777 }
3778
9a812198
JV		 3779 ret = ip_vs_genl_register();
3780 if (ret) {
1e3e238e 3781 pr_err("cannot register Generic Netlink interface.\n");
7a4f0761 3782 goto err_genl;
9a812198 3783 }
1da177e4 3784 return 0;
fc723250 3785
7a4f0761
HS		 3786err_genl:
3787 nf_unregister_sockopt(&ip_vs_sockopts);
3788err_sock:
fc723250 3789 return ret;
1da177e4
LT		 3790}
3791
8537de8a
HS		 3792void ip_vs_unregister_nl_ioctl(void)
3793{
3794 ip_vs_genl_unregister();
3795 nf_unregister_sockopt(&ip_vs_sockopts);
3796}
3797
3798int __init ip_vs_control_init(void)
3799{
3800 int idx;
3801 int ret;
3802
3803 EnterFunction(2);
3804
3805 /* Initialize svc_table, ip_vs_svc_fwm_table, rs_table */
3806 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
3807 INIT_LIST_HEAD(&ip_vs_svc_table[idx]);
3808 INIT_LIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3809 }
3810
3811 smp_wmb(); /* Do we really need it now ? */
3812
3813 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
3814 if (ret < 0)
3815 return ret;
3816
3817 LeaveFunction(2);
3818 return 0;
3819}
3820
1da177e4
LT		 3821
3822void ip_vs_control_cleanup(void)
3823{
3824 EnterFunction(2);
7676e345 3825 unregister_netdevice_notifier(&ip_vs_dst_notifier);
1da177e4
LT		 3826 LeaveFunction(2);
3827}
Linux kernel - Deliverables
This page took 0.980763 seconds and 5 git commands to generate.