projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ipvs: reduce sync rate with time thresholds
[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
749c42b6
JA		 1602
1603static int zero;
1604static int three = 3;
1605
1da177e4 1606static int
8d65af78 1607proc_do_defense_mode(ctl_table *table, int write,
1da177e4
LT		 1608 void __user *buffer, size_t *lenp, loff_t *ppos)
1609{
9330419d 1610 struct net *net = current->nsproxy->net_ns;
1da177e4
LT		 1611 int *valp = table->data;
1612 int val = *valp;
1613 int rc;
1614
8d65af78 1615 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1da177e4
LT		 1616 if (write && (*valp != val)) {
1617 if ((*valp < 0) || (*valp > 3)) {
1618 /* Restore the correct value */
1619 *valp = val;
1620 } else {
9330419d 1621 update_defense_level(net_ipvs(net));
1da177e4
LT		 1622 }
1623 }
1624 return rc;
1625}
1626
1da177e4 1627static int
8d65af78 1628proc_do_sync_threshold(ctl_table *table, int write,
1da177e4
LT		 1629 void __user *buffer, size_t *lenp, loff_t *ppos)
1630{
1631 int *valp = table->data;
1632 int val[2];
1633 int rc;
1634
1635 /* backup the value first */
1636 memcpy(val, valp, sizeof(val));
1637
8d65af78 1638 rc = proc_dointvec(table, write, buffer, lenp, ppos);
749c42b6
JA		 1639 if (write && (valp[0] < 0 || valp[1] < 0 ||
1640 (valp[0] >= valp[1] && valp[1]))) {
1da177e4
LT		 1641 /* Restore the correct value */
1642 memcpy(valp, val, sizeof(val));
1643 }
1644 return rc;
1645}
1646
b880c1f0
HS		 1647static int
1648proc_do_sync_mode(ctl_table *table, int write,
1649 void __user *buffer, size_t *lenp, loff_t *ppos)
1650{
1651 int *valp = table->data;
1652 int val = *valp;
1653 int rc;
1654
1655 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1656 if (write && (*valp != val)) {
1657 if ((*valp < 0) || (*valp > 1)) {
1658 /* Restore the correct value */
1659 *valp = val;
1660 } else {
f131315f
HS		 1661 struct net *net = current->nsproxy->net_ns;
1662 ip_vs_sync_switch_mode(net, val);
b880c1f0
HS		 1663 }
1664 }
1665 return rc;
1666}
1da177e4
LT		 1667
1668/*
1669 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
a0840e2e 1670 * Do not change order or insert new entries without
503cf15a 1671 * align with netns init in ip_vs_control_net_init()
1da177e4
LT		 1672 */
1673
1674static struct ctl_table vs_vars[] = {
1675 {
1da177e4 1676 .procname = "amemthresh",
1da177e4
LT		 1677 .maxlen = sizeof(int),
1678 .mode = 0644,
6d9f239a 1679 .proc_handler = proc_dointvec,
1da177e4 1680 },
1da177e4 1681 {
1da177e4 1682 .procname = "am_droprate",
1da177e4
LT		 1683 .maxlen = sizeof(int),
1684 .mode = 0644,
6d9f239a 1685 .proc_handler = proc_dointvec,
1da177e4
LT		 1686 },
1687 {
1da177e4 1688 .procname = "drop_entry",
1da177e4
LT		 1689 .maxlen = sizeof(int),
1690 .mode = 0644,
6d9f239a 1691 .proc_handler = proc_do_defense_mode,
1da177e4
LT		 1692 },
1693 {
1da177e4 1694 .procname = "drop_packet",
1da177e4
LT		 1695 .maxlen = sizeof(int),
1696 .mode = 0644,
6d9f239a 1697 .proc_handler = proc_do_defense_mode,
1da177e4 1698 },
f4bc17cd
JA		 1699#ifdef CONFIG_IP_VS_NFCT
1700 {
1701 .procname = "conntrack",
f4bc17cd
JA		 1702 .maxlen = sizeof(int),
1703 .mode = 0644,
1704 .proc_handler = &proc_dointvec,
1705 },
1706#endif
1da177e4 1707 {
1da177e4 1708 .procname = "secure_tcp",
1da177e4
LT		 1709 .maxlen = sizeof(int),
1710 .mode = 0644,
6d9f239a 1711 .proc_handler = proc_do_defense_mode,
1da177e4 1712 },
8a803040
JA		 1713 {
1714 .procname = "snat_reroute",
8a803040
JA		 1715 .maxlen = sizeof(int),
1716 .mode = 0644,
1717 .proc_handler = &proc_dointvec,
1718 },
b880c1f0
HS		 1719 {
1720 .procname = "sync_version",
b880c1f0
HS		 1721 .maxlen = sizeof(int),
1722 .mode = 0644,
1723 .proc_handler = &proc_do_sync_mode,
1724 },
1c003b15
PNA		 1725 {
1726 .procname = "sync_qlen_max",
1727 .maxlen = sizeof(int),
1728 .mode = 0644,
1729 .proc_handler = proc_dointvec,
1730 },
1731 {
1732 .procname = "sync_sock_size",
1733 .maxlen = sizeof(int),
1734 .mode = 0644,
1735 .proc_handler = proc_dointvec,
1736 },
a0840e2e
HS		 1737 {
1738 .procname = "cache_bypass",
1739 .maxlen = sizeof(int),
1740 .mode = 0644,
1741 .proc_handler = proc_dointvec,
1742 },
1743 {
1744 .procname = "expire_nodest_conn",
1745 .maxlen = sizeof(int),
1746 .mode = 0644,
1747 .proc_handler = proc_dointvec,
1748 },
1749 {
1750 .procname = "expire_quiescent_template",
1751 .maxlen = sizeof(int),
1752 .mode = 0644,
1753 .proc_handler = proc_dointvec,
1754 },
1755 {
1756 .procname = "sync_threshold",
1757 .maxlen =
1758 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1759 .mode = 0644,
1760 .proc_handler = proc_do_sync_threshold,
1761 },
749c42b6
JA		 1762 {
1763 .procname = "sync_refresh_period",
1764 .maxlen = sizeof(int),
1765 .mode = 0644,
1766 .proc_handler = proc_dointvec_jiffies,
1767 },
1768 {
1769 .procname = "sync_retries",
1770 .maxlen = sizeof(int),
1771 .mode = 0644,
1772 .proc_handler = proc_dointvec_minmax,
1773 .extra1 = &zero,
1774 .extra2 = &three,
1775 },
a0840e2e
HS		 1776 {
1777 .procname = "nat_icmp_send",
1778 .maxlen = sizeof(int),
1779 .mode = 0644,
1780 .proc_handler = proc_dointvec,
1781 },
1782#ifdef CONFIG_IP_VS_DEBUG
1783 {
1784 .procname = "debug_level",
1785 .data = &sysctl_ip_vs_debug_level,
1786 .maxlen = sizeof(int),
1787 .mode = 0644,
1788 .proc_handler = proc_dointvec,
1789 },
1790#endif
1da177e4
LT		 1791#if 0
1792 {
1da177e4
LT		 1793 .procname = "timeout_established",
1794 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1795 .maxlen = sizeof(int),
1796 .mode = 0644,
6d9f239a 1797 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1798 },
1799 {
1da177e4
LT		 1800 .procname = "timeout_synsent",
1801 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1802 .maxlen = sizeof(int),
1803 .mode = 0644,
6d9f239a 1804 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1805 },
1806 {
1da177e4
LT		 1807 .procname = "timeout_synrecv",
1808 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1809 .maxlen = sizeof(int),
1810 .mode = 0644,
6d9f239a 1811 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1812 },
1813 {
1da177e4
LT		 1814 .procname = "timeout_finwait",
1815 .data = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1816 .maxlen = sizeof(int),
1817 .mode = 0644,
6d9f239a 1818 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1819 },
1820 {
1da177e4
LT		 1821 .procname = "timeout_timewait",
1822 .data = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1823 .maxlen = sizeof(int),
1824 .mode = 0644,
6d9f239a 1825 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1826 },
1827 {
1da177e4
LT		 1828 .procname = "timeout_close",
1829 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1830 .maxlen = sizeof(int),
1831 .mode = 0644,
6d9f239a 1832 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1833 },
1834 {
1da177e4
LT		 1835 .procname = "timeout_closewait",
1836 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1837 .maxlen = sizeof(int),
1838 .mode = 0644,
6d9f239a 1839 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1840 },
1841 {
1da177e4
LT		 1842 .procname = "timeout_lastack",
1843 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1844 .maxlen = sizeof(int),
1845 .mode = 0644,
6d9f239a 1846 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1847 },
1848 {
1da177e4
LT		 1849 .procname = "timeout_listen",
1850 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1851 .maxlen = sizeof(int),
1852 .mode = 0644,
6d9f239a 1853 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1854 },
1855 {
1da177e4
LT		 1856 .procname = "timeout_synack",
1857 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1858 .maxlen = sizeof(int),
1859 .mode = 0644,
6d9f239a 1860 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1861 },
1862 {
1da177e4
LT		 1863 .procname = "timeout_udp",
1864 .data = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1865 .maxlen = sizeof(int),
1866 .mode = 0644,
6d9f239a 1867 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1868 },
1869 {
1da177e4
LT		 1870 .procname = "timeout_icmp",
1871 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1872 .maxlen = sizeof(int),
1873 .mode = 0644,
6d9f239a 1874 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1875 },
1876#endif
f8572d8f 1877 { }
1da177e4
LT		 1878};
1879
14e40546 1880#endif
1da177e4 1881
1da177e4
LT		 1882#ifdef CONFIG_PROC_FS
1883
1884struct ip_vs_iter {
fc723250 1885 struct seq_net_private p; /* Do not move this, netns depends upon it*/
1da177e4
LT		 1886 struct list_head *table;
1887 int bucket;
1888};
1889
1890/*
1891 * Write the contents of the VS rule table to a PROCfs file.
1892 * (It is kept just for backward compatibility)
1893 */
95c96174 1894static inline const char *ip_vs_fwd_name(unsigned int flags)
1da177e4
LT		 1895{
1896 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1897 case IP_VS_CONN_F_LOCALNODE:
1898 return "Local";
1899 case IP_VS_CONN_F_TUNNEL:
1900 return "Tunnel";
1901 case IP_VS_CONN_F_DROUTE:
1902 return "Route";
1903 default:
1904 return "Masq";
1905 }
1906}
1907
1908
1909/* Get the Nth entry in the two lists */
1910static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1911{
fc723250 1912 struct net *net = seq_file_net(seq);
1da177e4
LT		 1913 struct ip_vs_iter *iter = seq->private;
1914 int idx;
1915 struct ip_vs_service *svc;
1916
1917 /* look in hash by protocol */
1918 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1919 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
fc723250 1920 if (net_eq(svc->net, net) && pos-- == 0) {
1da177e4
LT		 1921 iter->table = ip_vs_svc_table;
1922 iter->bucket = idx;
1923 return svc;
1924 }
1925 }
1926 }
1927
1928 /* keep looking in fwmark */
1929 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1930 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
fc723250 1931 if (net_eq(svc->net, net) && pos-- == 0) {
1da177e4
LT		 1932 iter->table = ip_vs_svc_fwm_table;
1933 iter->bucket = idx;
1934 return svc;
1935 }
1936 }
1937 }
1938
1939 return NULL;
1940}
1941
1942static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
563e94f0 1943__acquires(__ip_vs_svc_lock)
1da177e4
LT		 1944{
1945
1946 read_lock_bh(&__ip_vs_svc_lock);
1947 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
1948}
1949
1950
1951static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1952{
1953 struct list_head *e;
1954 struct ip_vs_iter *iter;
1955 struct ip_vs_service *svc;
1956
1957 ++*pos;
1958 if (v == SEQ_START_TOKEN)
1959 return ip_vs_info_array(seq,0);
1960
1961 svc = v;
1962 iter = seq->private;
1963
1964 if (iter->table == ip_vs_svc_table) {
1965 /* next service in table hashed by protocol */
1966 if ((e = svc->s_list.next) != &ip_vs_svc_table[iter->bucket])
1967 return list_entry(e, struct ip_vs_service, s_list);
1968
1969
1970 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1971 list_for_each_entry(svc,&ip_vs_svc_table[iter->bucket],
1972 s_list) {
1973 return svc;
1974 }
1975 }
1976
1977 iter->table = ip_vs_svc_fwm_table;
1978 iter->bucket = -1;
1979 goto scan_fwmark;
1980 }
1981
1982 /* next service in hashed by fwmark */
1983 if ((e = svc->f_list.next) != &ip_vs_svc_fwm_table[iter->bucket])
1984 return list_entry(e, struct ip_vs_service, f_list);
1985
1986 scan_fwmark:
1987 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
1988 list_for_each_entry(svc, &ip_vs_svc_fwm_table[iter->bucket],
1989 f_list)
1990 return svc;
1991 }
1992
1993 return NULL;
1994}
1995
1996static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
563e94f0 1997__releases(__ip_vs_svc_lock)
1da177e4
LT		 1998{
1999 read_unlock_bh(&__ip_vs_svc_lock);
2000}
2001
2002
2003static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
2004{
2005 if (v == SEQ_START_TOKEN) {
2006 seq_printf(seq,
2007 "IP Virtual Server version %d.%d.%d (size=%d)\n",
6f7edb48 2008 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
1da177e4
LT		 2009 seq_puts(seq,
2010 "Prot LocalAddress:Port Scheduler Flags\n");
2011 seq_puts(seq,
2012 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2013 } else {
2014 const struct ip_vs_service *svc = v;
2015 const struct ip_vs_iter *iter = seq->private;
2016 const struct ip_vs_dest *dest;
2017
667a5f18
VB		 2018 if (iter->table == ip_vs_svc_table) {
2019#ifdef CONFIG_IP_VS_IPV6
2020 if (svc->af == AF_INET6)
5b095d98 2021 seq_printf(seq, "%s [%pI6]:%04X %s ",
667a5f18 2022 ip_vs_proto_name(svc->protocol),
38ff4fa4 2023 &svc->addr.in6,
667a5f18
VB		 2024 ntohs(svc->port),
2025 svc->scheduler->name);
2026 else
2027#endif
26ec037f 2028 seq_printf(seq, "%s %08X:%04X %s %s ",
667a5f18
VB		 2029 ip_vs_proto_name(svc->protocol),
2030 ntohl(svc->addr.ip),
2031 ntohs(svc->port),
26ec037f
NC		 2032 svc->scheduler->name,
2033 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
667a5f18 2034 } else {
26ec037f
NC		 2035 seq_printf(seq, "FWM %08X %s %s",
2036 svc->fwmark, svc->scheduler->name,
2037 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
667a5f18 2038 }
1da177e4
LT		 2039
2040 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2041 seq_printf(seq, "persistent %d %08X\n",
2042 svc->timeout,
2043 ntohl(svc->netmask));
2044 else
2045 seq_putc(seq, '\n');
2046
2047 list_for_each_entry(dest, &svc->destinations, n_list) {
667a5f18
VB		 2048#ifdef CONFIG_IP_VS_IPV6
2049 if (dest->af == AF_INET6)
2050 seq_printf(seq,
5b095d98 2051 " -> [%pI6]:%04X"
667a5f18 2052 " %-7s %-6d %-10d %-10d\n",
38ff4fa4 2053 &dest->addr.in6,
667a5f18
VB		 2054 ntohs(dest->port),
2055 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2056 atomic_read(&dest->weight),
2057 atomic_read(&dest->activeconns),
2058 atomic_read(&dest->inactconns));
2059 else
2060#endif
2061 seq_printf(seq,
2062 " -> %08X:%04X "
2063 "%-7s %-6d %-10d %-10d\n",
2064 ntohl(dest->addr.ip),
2065 ntohs(dest->port),
2066 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2067 atomic_read(&dest->weight),
2068 atomic_read(&dest->activeconns),
2069 atomic_read(&dest->inactconns));
2070
1da177e4
LT		 2071 }
2072 }
2073 return 0;
2074}
2075
56b3d975 2076static const struct seq_operations ip_vs_info_seq_ops = {
1da177e4
LT		 2077 .start = ip_vs_info_seq_start,
2078 .next = ip_vs_info_seq_next,
2079 .stop = ip_vs_info_seq_stop,
2080 .show = ip_vs_info_seq_show,
2081};
2082
2083static int ip_vs_info_open(struct inode *inode, struct file *file)
2084{
fc723250 2085 return seq_open_net(inode, file, &ip_vs_info_seq_ops,
cf7732e4 2086 sizeof(struct ip_vs_iter));
1da177e4
LT		 2087}
2088
9a32144e 2089static const struct file_operations ip_vs_info_fops = {
1da177e4
LT		 2090 .owner = THIS_MODULE,
2091 .open = ip_vs_info_open,
2092 .read = seq_read,
2093 .llseek = seq_lseek,
0f08190f 2094 .release = seq_release_net,
1da177e4
LT		 2095};
2096
1da177e4
LT		 2097static int ip_vs_stats_show(struct seq_file *seq, void *v)
2098{
b17fc996 2099 struct net *net = seq_file_single_net(seq);
55a3d4e1 2100 struct ip_vs_stats_user show;
1da177e4
LT		 2101
2102/* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2103 seq_puts(seq,
2104 " Total Incoming Outgoing Incoming Outgoing\n");
2105 seq_printf(seq,
2106 " Conns Packets Packets Bytes Bytes\n");
2107
55a3d4e1
JA		 2108 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2109 seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", show.conns,
2110 show.inpkts, show.outpkts,
2111 (unsigned long long) show.inbytes,
2112 (unsigned long long) show.outbytes);
1da177e4
LT		 2113
2114/* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2115 seq_puts(seq,
2116 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
55a3d4e1
JA		 2117 seq_printf(seq, "%8X %8X %8X %16X %16X\n",
2118 show.cps, show.inpps, show.outpps,
2119 show.inbps, show.outbps);
1da177e4
LT		 2120
2121 return 0;
2122}
2123
2124static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2125{
fc723250 2126 return single_open_net(inode, file, ip_vs_stats_show);
1da177e4
LT		 2127}
2128
9a32144e 2129static const struct file_operations ip_vs_stats_fops = {
1da177e4
LT		 2130 .owner = THIS_MODULE,
2131 .open = ip_vs_stats_seq_open,
2132 .read = seq_read,
2133 .llseek = seq_lseek,
0f08190f 2134 .release = single_release_net,
1da177e4
LT		 2135};
2136
b17fc996
HS		 2137static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2138{
2139 struct net *net = seq_file_single_net(seq);
2a0751af
JA		 2140 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2141 struct ip_vs_cpu_stats *cpustats = tot_stats->cpustats;
ea9f22cc 2142 struct ip_vs_stats_user rates;
b17fc996
HS		 2143 int i;
2144
2145/* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2146 seq_puts(seq,
2147 " Total Incoming Outgoing Incoming Outgoing\n");
2148 seq_printf(seq,
2149 "CPU Conns Packets Packets Bytes Bytes\n");
2150
2151 for_each_possible_cpu(i) {
2a0751af
JA		 2152 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2153 unsigned int start;
2154 __u64 inbytes, outbytes;
2155
2156 do {
2157 start = u64_stats_fetch_begin_bh(&u->syncp);
2158 inbytes = u->ustats.inbytes;
2159 outbytes = u->ustats.outbytes;
2160 } while (u64_stats_fetch_retry_bh(&u->syncp, start));
2161
b17fc996 2162 seq_printf(seq, "%3X %8X %8X %8X %16LX %16LX\n",
2a0751af
JA		 2163 i, u->ustats.conns, u->ustats.inpkts,
2164 u->ustats.outpkts, (__u64)inbytes,
2165 (__u64)outbytes);
b17fc996
HS		 2166 }
2167
2168 spin_lock_bh(&tot_stats->lock);
ea9f22cc 2169
b17fc996
HS		 2170 seq_printf(seq, " ~ %8X %8X %8X %16LX %16LX\n\n",
2171 tot_stats->ustats.conns, tot_stats->ustats.inpkts,
2172 tot_stats->ustats.outpkts,
2173 (unsigned long long) tot_stats->ustats.inbytes,
2174 (unsigned long long) tot_stats->ustats.outbytes);
2175
ea9f22cc
JA		 2176 ip_vs_read_estimator(&rates, tot_stats);
2177
2178 spin_unlock_bh(&tot_stats->lock);
2179
b17fc996
HS		 2180/* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2181 seq_puts(seq,
2182 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2183 seq_printf(seq, " %8X %8X %8X %16X %16X\n",
ea9f22cc
JA		 2184 rates.cps,
2185 rates.inpps,
2186 rates.outpps,
2187 rates.inbps,
2188 rates.outbps);
b17fc996
HS		 2189
2190 return 0;
2191}
2192
2193static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2194{
2195 return single_open_net(inode, file, ip_vs_stats_percpu_show);
2196}
2197
2198static const struct file_operations ip_vs_stats_percpu_fops = {
2199 .owner = THIS_MODULE,
2200 .open = ip_vs_stats_percpu_seq_open,
2201 .read = seq_read,
2202 .llseek = seq_lseek,
0f08190f 2203 .release = single_release_net,
b17fc996 2204};
1da177e4
LT		 2205#endif
2206
2207/*
2208 * Set timeout values for tcp tcpfin udp in the timeout_table.
2209 */
9330419d 2210static int ip_vs_set_timeout(struct net *net, struct ip_vs_timeout_user *u)
1da177e4 2211{
091bb34c 2212#if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
9330419d 2213 struct ip_vs_proto_data *pd;
091bb34c 2214#endif
9330419d 2215
1da177e4
LT		 2216 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2217 u->tcp_timeout,
2218 u->tcp_fin_timeout,
2219 u->udp_timeout);
2220
2221#ifdef CONFIG_IP_VS_PROTO_TCP
2222 if (u->tcp_timeout) {
9330419d
HS		 2223 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2224 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
1da177e4
LT		 2225 = u->tcp_timeout * HZ;
2226 }
2227
2228 if (u->tcp_fin_timeout) {
9330419d
HS		 2229 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2230 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
1da177e4
LT		 2231 = u->tcp_fin_timeout * HZ;
2232 }
2233#endif
2234
2235#ifdef CONFIG_IP_VS_PROTO_UDP
2236 if (u->udp_timeout) {
9330419d
HS		 2237 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2238 pd->timeout_table[IP_VS_UDP_S_NORMAL]
1da177e4
LT		 2239 = u->udp_timeout * HZ;
2240 }
2241#endif
2242 return 0;
2243}
2244
2245
2246#define SET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2247#define SERVICE_ARG_LEN (sizeof(struct ip_vs_service_user))
2248#define SVCDEST_ARG_LEN (sizeof(struct ip_vs_service_user) + \
2249 sizeof(struct ip_vs_dest_user))
2250#define TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2251#define DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user))
2252#define MAX_ARG_LEN SVCDEST_ARG_LEN
2253
9b5b5cff 2254static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
1da177e4
LT		 2255 [SET_CMDID(IP_VS_SO_SET_ADD)] = SERVICE_ARG_LEN,
2256 [SET_CMDID(IP_VS_SO_SET_EDIT)] = SERVICE_ARG_LEN,
2257 [SET_CMDID(IP_VS_SO_SET_DEL)] = SERVICE_ARG_LEN,
2258 [SET_CMDID(IP_VS_SO_SET_FLUSH)] = 0,
2259 [SET_CMDID(IP_VS_SO_SET_ADDDEST)] = SVCDEST_ARG_LEN,
2260 [SET_CMDID(IP_VS_SO_SET_DELDEST)] = SVCDEST_ARG_LEN,
2261 [SET_CMDID(IP_VS_SO_SET_EDITDEST)] = SVCDEST_ARG_LEN,
2262 [SET_CMDID(IP_VS_SO_SET_TIMEOUT)] = TIMEOUT_ARG_LEN,
2263 [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)] = DAEMON_ARG_LEN,
2264 [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)] = DAEMON_ARG_LEN,
2265 [SET_CMDID(IP_VS_SO_SET_ZERO)] = SERVICE_ARG_LEN,
2266};
2267
c860c6b1
JV		 2268static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2269 struct ip_vs_service_user *usvc_compat)
2270{
0d1e71b0
SH		 2271 memset(usvc, 0, sizeof(*usvc));
2272
c860c6b1
JV		 2273 usvc->af = AF_INET;
2274 usvc->protocol = usvc_compat->protocol;
2275 usvc->addr.ip = usvc_compat->addr;
2276 usvc->port = usvc_compat->port;
2277 usvc->fwmark = usvc_compat->fwmark;
2278
2279 /* Deep copy of sched_name is not needed here */
2280 usvc->sched_name = usvc_compat->sched_name;
2281
2282 usvc->flags = usvc_compat->flags;
2283 usvc->timeout = usvc_compat->timeout;
2284 usvc->netmask = usvc_compat->netmask;
2285}
2286
2287static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2288 struct ip_vs_dest_user *udest_compat)
2289{
0d1e71b0
SH		 2290 memset(udest, 0, sizeof(*udest));
2291
c860c6b1
JV		 2292 udest->addr.ip = udest_compat->addr;
2293 udest->port = udest_compat->port;
2294 udest->conn_flags = udest_compat->conn_flags;
2295 udest->weight = udest_compat->weight;
2296 udest->u_threshold = udest_compat->u_threshold;
2297 udest->l_threshold = udest_compat->l_threshold;
2298}
2299
1da177e4
LT		 2300static int
2301do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2302{
fc723250 2303 struct net *net = sock_net(sk);
1da177e4
LT		 2304 int ret;
2305 unsigned char arg[MAX_ARG_LEN];
c860c6b1
JV		 2306 struct ip_vs_service_user *usvc_compat;
2307 struct ip_vs_service_user_kern usvc;
1da177e4 2308 struct ip_vs_service *svc;
c860c6b1
JV		 2309 struct ip_vs_dest_user *udest_compat;
2310 struct ip_vs_dest_user_kern udest;
ae1d48b2 2311 struct netns_ipvs *ipvs = net_ipvs(net);
1da177e4
LT		 2312
2313 if (!capable(CAP_NET_ADMIN))
2314 return -EPERM;
2315
04bcef2a
AV		 2316 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2317 return -EINVAL;
2318 if (len < 0 || len > MAX_ARG_LEN)
2319 return -EINVAL;
1da177e4 2320 if (len != set_arglen[SET_CMDID(cmd)]) {
1e3e238e
HE		 2321 pr_err("set_ctl: len %u != %u\n",
2322 len, set_arglen[SET_CMDID(cmd)]);
1da177e4
LT		 2323 return -EINVAL;
2324 }
2325
2326 if (copy_from_user(arg, user, len) != 0)
2327 return -EFAULT;
2328
2329 /* increase the module use count */
2330 ip_vs_use_count_inc();
2331
ae1d48b2
HS		 2332 /* Handle daemons since they have another lock */
2333 if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2334 cmd == IP_VS_SO_SET_STOPDAEMON) {
2335 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2336
2337 if (mutex_lock_interruptible(&ipvs->sync_mutex)) {
2338 ret = -ERESTARTSYS;
2339 goto out_dec;
2340 }
2341 if (cmd == IP_VS_SO_SET_STARTDAEMON)
2342 ret = start_sync_thread(net, dm->state, dm->mcast_ifn,
2343 dm->syncid);
2344 else
2345 ret = stop_sync_thread(net, dm->state);
2346 mutex_unlock(&ipvs->sync_mutex);
2347 goto out_dec;
2348 }
2349
14cc3e2b 2350 if (mutex_lock_interruptible(&__ip_vs_mutex)) {
1da177e4
LT		 2351 ret = -ERESTARTSYS;
2352 goto out_dec;
2353 }
2354
2355 if (cmd == IP_VS_SO_SET_FLUSH) {
2356 /* Flush the virtual service */
fc723250 2357 ret = ip_vs_flush(net);
1da177e4
LT		 2358 goto out_unlock;
2359 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2360 /* Set timeout values for (tcp tcpfin udp) */
9330419d 2361 ret = ip_vs_set_timeout(net, (struct ip_vs_timeout_user *)arg);
1da177e4 2362 goto out_unlock;
1da177e4
LT		 2363 }
2364
c860c6b1
JV		 2365 usvc_compat = (struct ip_vs_service_user *)arg;
2366 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2367
2368 /* We only use the new structs internally, so copy userspace compat
2369 * structs to extended internal versions */
2370 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2371 ip_vs_copy_udest_compat(&udest, udest_compat);
1da177e4
LT		 2372
2373 if (cmd == IP_VS_SO_SET_ZERO) {
2374 /* if no service address is set, zero counters in all */
c860c6b1 2375 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
fc723250 2376 ret = ip_vs_zero_all(net);
1da177e4
LT		 2377 goto out_unlock;
2378 }
2379 }
2380
2906f66a
VMR		 2381 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2382 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2383 usvc.protocol != IPPROTO_SCTP) {
1e3e238e
HE		 2384 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2385 usvc.protocol, &usvc.addr.ip,
2386 ntohs(usvc.port), usvc.sched_name);
1da177e4
LT		 2387 ret = -EFAULT;
2388 goto out_unlock;
2389 }
2390
2391 /* Lookup the exact service by <protocol, addr, port> or fwmark */
c860c6b1 2392 if (usvc.fwmark == 0)
fc723250 2393 svc = __ip_vs_service_find(net, usvc.af, usvc.protocol,
26c15cfd 2394 &usvc.addr, usvc.port);
1da177e4 2395 else
fc723250 2396 svc = __ip_vs_svc_fwm_find(net, usvc.af, usvc.fwmark);
1da177e4
LT		 2397
2398 if (cmd != IP_VS_SO_SET_ADD
c860c6b1 2399 && (svc == NULL || svc->protocol != usvc.protocol)) {
1da177e4 2400 ret = -ESRCH;
26c15cfd 2401 goto out_unlock;
1da177e4
LT		 2402 }
2403
2404 switch (cmd) {
2405 case IP_VS_SO_SET_ADD:
2406 if (svc != NULL)
2407 ret = -EEXIST;
2408 else
fc723250 2409 ret = ip_vs_add_service(net, &usvc, &svc);
1da177e4
LT		 2410 break;
2411 case IP_VS_SO_SET_EDIT:
c860c6b1 2412 ret = ip_vs_edit_service(svc, &usvc);
1da177e4
LT		 2413 break;
2414 case IP_VS_SO_SET_DEL:
2415 ret = ip_vs_del_service(svc);
2416 if (!ret)
2417 goto out_unlock;
2418 break;
2419 case IP_VS_SO_SET_ZERO:
2420 ret = ip_vs_zero_service(svc);
2421 break;
2422 case IP_VS_SO_SET_ADDDEST:
c860c6b1 2423 ret = ip_vs_add_dest(svc, &udest);
1da177e4
LT		 2424 break;
2425 case IP_VS_SO_SET_EDITDEST:
c860c6b1 2426 ret = ip_vs_edit_dest(svc, &udest);
1da177e4
LT		 2427 break;
2428 case IP_VS_SO_SET_DELDEST:
c860c6b1 2429 ret = ip_vs_del_dest(svc, &udest);
1da177e4
LT		 2430 break;
2431 default:
2432 ret = -EINVAL;
2433 }
2434
1da177e4 2435 out_unlock:
14cc3e2b 2436 mutex_unlock(&__ip_vs_mutex);
1da177e4
LT		 2437 out_dec:
2438 /* decrease the module use count */
2439 ip_vs_use_count_dec();
2440
2441 return ret;
2442}
2443
2444
1da177e4
LT		 2445static void
2446ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2447{
2448 dst->protocol = src->protocol;
e7ade46a 2449 dst->addr = src->addr.ip;
1da177e4
LT		 2450 dst->port = src->port;
2451 dst->fwmark = src->fwmark;
4da62fc7 2452 strlcpy(dst->sched_name, src->scheduler->name, sizeof(dst->sched_name));
1da177e4
LT		 2453 dst->flags = src->flags;
2454 dst->timeout = src->timeout / HZ;
2455 dst->netmask = src->netmask;
2456 dst->num_dests = src->num_dests;
2457 ip_vs_copy_stats(&dst->stats, &src->stats);
2458}
2459
2460static inline int
fc723250
HS		 2461__ip_vs_get_service_entries(struct net *net,
2462 const struct ip_vs_get_services *get,
1da177e4
LT		 2463 struct ip_vs_get_services __user *uptr)
2464{
2465 int idx, count=0;
2466 struct ip_vs_service *svc;
2467 struct ip_vs_service_entry entry;
2468 int ret = 0;
2469
2470 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2471 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_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 }
2488
2489 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2490 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
f94fd041 2491 /* Only expose IPv4 entries to old interface */
fc723250 2492 if (svc->af != AF_INET || !net_eq(svc->net, net))
f94fd041
JV		 2493 continue;
2494
1da177e4
LT		 2495 if (count >= get->num_services)
2496 goto out;
4da62fc7 2497 memset(&entry, 0, sizeof(entry));
1da177e4
LT		 2498 ip_vs_copy_service(&entry, svc);
2499 if (copy_to_user(&uptr->entrytable[count],
2500 &entry, sizeof(entry))) {
2501 ret = -EFAULT;
2502 goto out;
2503 }
2504 count++;
2505 }
2506 }
552ad65a 2507out:
1da177e4
LT		 2508 return ret;
2509}
2510
2511static inline int
fc723250 2512__ip_vs_get_dest_entries(struct net *net, const struct ip_vs_get_dests *get,
1da177e4
LT		 2513 struct ip_vs_get_dests __user *uptr)
2514{
2515 struct ip_vs_service *svc;
b18610de 2516 union nf_inet_addr addr = { .ip = get->addr };
1da177e4
LT		 2517 int ret = 0;
2518
2519 if (get->fwmark)
fc723250 2520 svc = __ip_vs_svc_fwm_find(net, AF_INET, get->fwmark);
1da177e4 2521 else
fc723250 2522 svc = __ip_vs_service_find(net, AF_INET, get->protocol, &addr,
26c15cfd 2523 get->port);
b18610de 2524
1da177e4
LT		 2525 if (svc) {
2526 int count = 0;
2527 struct ip_vs_dest *dest;
2528 struct ip_vs_dest_entry entry;
2529
2530 list_for_each_entry(dest, &svc->destinations, n_list) {
2531 if (count >= get->num_dests)
2532 break;
2533
e7ade46a 2534 entry.addr = dest->addr.ip;
1da177e4
LT		 2535 entry.port = dest->port;
2536 entry.conn_flags = atomic_read(&dest->conn_flags);
2537 entry.weight = atomic_read(&dest->weight);
2538 entry.u_threshold = dest->u_threshold;
2539 entry.l_threshold = dest->l_threshold;
2540 entry.activeconns = atomic_read(&dest->activeconns);
2541 entry.inactconns = atomic_read(&dest->inactconns);
2542 entry.persistconns = atomic_read(&dest->persistconns);
2543 ip_vs_copy_stats(&entry.stats, &dest->stats);
2544 if (copy_to_user(&uptr->entrytable[count],
2545 &entry, sizeof(entry))) {
2546 ret = -EFAULT;
2547 break;
2548 }
2549 count++;
2550 }
1da177e4
LT		 2551 } else
2552 ret = -ESRCH;
2553 return ret;
2554}
2555
2556static inline void
9330419d 2557__ip_vs_get_timeouts(struct net *net, struct ip_vs_timeout_user *u)
1da177e4 2558{
091bb34c 2559#if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
9330419d 2560 struct ip_vs_proto_data *pd;
091bb34c 2561#endif
9330419d 2562
1da177e4 2563#ifdef CONFIG_IP_VS_PROTO_TCP
9330419d
HS		 2564 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2565 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2566 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
1da177e4
LT		 2567#endif
2568#ifdef CONFIG_IP_VS_PROTO_UDP
9330419d 2569 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
1da177e4 2570 u->udp_timeout =
9330419d 2571 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
1da177e4
LT		 2572#endif
2573}
2574
2575
2576#define GET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2577#define GET_INFO_ARG_LEN (sizeof(struct ip_vs_getinfo))
2578#define GET_SERVICES_ARG_LEN (sizeof(struct ip_vs_get_services))
2579#define GET_SERVICE_ARG_LEN (sizeof(struct ip_vs_service_entry))
2580#define GET_DESTS_ARG_LEN (sizeof(struct ip_vs_get_dests))
2581#define GET_TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2582#define GET_DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user) * 2)
2583
9b5b5cff 2584static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
1da177e4
LT		 2585 [GET_CMDID(IP_VS_SO_GET_VERSION)] = 64,
2586 [GET_CMDID(IP_VS_SO_GET_INFO)] = GET_INFO_ARG_LEN,
2587 [GET_CMDID(IP_VS_SO_GET_SERVICES)] = GET_SERVICES_ARG_LEN,
2588 [GET_CMDID(IP_VS_SO_GET_SERVICE)] = GET_SERVICE_ARG_LEN,
2589 [GET_CMDID(IP_VS_SO_GET_DESTS)] = GET_DESTS_ARG_LEN,
2590 [GET_CMDID(IP_VS_SO_GET_TIMEOUT)] = GET_TIMEOUT_ARG_LEN,
2591 [GET_CMDID(IP_VS_SO_GET_DAEMON)] = GET_DAEMON_ARG_LEN,
2592};
2593
2594static int
2595do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2596{
2597 unsigned char arg[128];
2598 int ret = 0;
04bcef2a 2599 unsigned int copylen;
fc723250 2600 struct net *net = sock_net(sk);
f131315f 2601 struct netns_ipvs *ipvs = net_ipvs(net);
1da177e4 2602
fc723250 2603 BUG_ON(!net);
1da177e4
LT		 2604 if (!capable(CAP_NET_ADMIN))
2605 return -EPERM;
2606
04bcef2a
AV		 2607 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2608 return -EINVAL;
2609
1da177e4 2610 if (*len < get_arglen[GET_CMDID(cmd)]) {
1e3e238e
HE		 2611 pr_err("get_ctl: len %u < %u\n",
2612 *len, get_arglen[GET_CMDID(cmd)]);
1da177e4
LT		 2613 return -EINVAL;
2614 }
2615
04bcef2a
AV		 2616 copylen = get_arglen[GET_CMDID(cmd)];
2617 if (copylen > 128)
2618 return -EINVAL;
2619
2620 if (copy_from_user(arg, user, copylen) != 0)
1da177e4 2621 return -EFAULT;
ae1d48b2
HS		 2622 /*
2623 * Handle daemons first since it has its own locking
2624 */
2625 if (cmd == IP_VS_SO_GET_DAEMON) {
2626 struct ip_vs_daemon_user d[2];
2627
2628 memset(&d, 0, sizeof(d));
2629 if (mutex_lock_interruptible(&ipvs->sync_mutex))
2630 return -ERESTARTSYS;
2631
2632 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2633 d[0].state = IP_VS_STATE_MASTER;
2634 strlcpy(d[0].mcast_ifn, ipvs->master_mcast_ifn,
2635 sizeof(d[0].mcast_ifn));
2636 d[0].syncid = ipvs->master_syncid;
2637 }
2638 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2639 d[1].state = IP_VS_STATE_BACKUP;
2640 strlcpy(d[1].mcast_ifn, ipvs->backup_mcast_ifn,
2641 sizeof(d[1].mcast_ifn));
2642 d[1].syncid = ipvs->backup_syncid;
2643 }
2644 if (copy_to_user(user, &d, sizeof(d)) != 0)
2645 ret = -EFAULT;
2646 mutex_unlock(&ipvs->sync_mutex);
2647 return ret;
2648 }
1da177e4 2649
14cc3e2b 2650 if (mutex_lock_interruptible(&__ip_vs_mutex))
1da177e4
LT		 2651 return -ERESTARTSYS;
2652
2653 switch (cmd) {
2654 case IP_VS_SO_GET_VERSION:
2655 {
2656 char buf[64];
2657
2658 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
6f7edb48 2659 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
1da177e4
LT		 2660 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2661 ret = -EFAULT;
2662 goto out;
2663 }
2664 *len = strlen(buf)+1;
2665 }
2666 break;
2667
2668 case IP_VS_SO_GET_INFO:
2669 {
2670 struct ip_vs_getinfo info;
2671 info.version = IP_VS_VERSION_CODE;
6f7edb48 2672 info.size = ip_vs_conn_tab_size;
a0840e2e 2673 info.num_services = ipvs->num_services;
1da177e4
LT		 2674 if (copy_to_user(user, &info, sizeof(info)) != 0)
2675 ret = -EFAULT;
2676 }
2677 break;
2678
2679 case IP_VS_SO_GET_SERVICES:
2680 {
2681 struct ip_vs_get_services *get;
2682 int size;
2683
2684 get = (struct ip_vs_get_services *)arg;
2685 size = sizeof(*get) +
2686 sizeof(struct ip_vs_service_entry) * get->num_services;
2687 if (*len != size) {
1e3e238e 2688 pr_err("length: %u != %u\n", *len, size);
1da177e4
LT		 2689 ret = -EINVAL;
2690 goto out;
2691 }
fc723250 2692 ret = __ip_vs_get_service_entries(net, get, user);
1da177e4
LT		 2693 }
2694 break;
2695
2696 case IP_VS_SO_GET_SERVICE:
2697 {
2698 struct ip_vs_service_entry *entry;
2699 struct ip_vs_service *svc;
b18610de 2700 union nf_inet_addr addr;
1da177e4
LT		 2701
2702 entry = (struct ip_vs_service_entry *)arg;
b18610de 2703 addr.ip = entry->addr;
1da177e4 2704 if (entry->fwmark)
fc723250 2705 svc = __ip_vs_svc_fwm_find(net, AF_INET, entry->fwmark);
1da177e4 2706 else
fc723250
HS		 2707 svc = __ip_vs_service_find(net, AF_INET,
2708 entry->protocol, &addr,
2709 entry->port);
1da177e4
LT		 2710 if (svc) {
2711 ip_vs_copy_service(entry, svc);
2712 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2713 ret = -EFAULT;
1da177e4
LT		 2714 } else
2715 ret = -ESRCH;
2716 }
2717 break;
2718
2719 case IP_VS_SO_GET_DESTS:
2720 {
2721 struct ip_vs_get_dests *get;
2722 int size;
2723
2724 get = (struct ip_vs_get_dests *)arg;
2725 size = sizeof(*get) +
2726 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2727 if (*len != size) {
1e3e238e 2728 pr_err("length: %u != %u\n", *len, size);
1da177e4
LT		 2729 ret = -EINVAL;
2730 goto out;
2731 }
fc723250 2732 ret = __ip_vs_get_dest_entries(net, get, user);
1da177e4
LT		 2733 }
2734 break;
2735
2736 case IP_VS_SO_GET_TIMEOUT:
2737 {
2738 struct ip_vs_timeout_user t;
2739
9330419d 2740 __ip_vs_get_timeouts(net, &t);
1da177e4
LT		 2741 if (copy_to_user(user, &t, sizeof(t)) != 0)
2742 ret = -EFAULT;
2743 }
2744 break;
2745
1da177e4
LT		 2746 default:
2747 ret = -EINVAL;
2748 }
2749
552ad65a 2750out:
14cc3e2b 2751 mutex_unlock(&__ip_vs_mutex);
1da177e4
LT		 2752 return ret;
2753}
2754
2755
2756static struct nf_sockopt_ops ip_vs_sockopts = {
2757 .pf = PF_INET,
2758 .set_optmin = IP_VS_BASE_CTL,
2759 .set_optmax = IP_VS_SO_SET_MAX+1,
2760 .set = do_ip_vs_set_ctl,
2761 .get_optmin = IP_VS_BASE_CTL,
2762 .get_optmax = IP_VS_SO_GET_MAX+1,
2763 .get = do_ip_vs_get_ctl,
16fcec35 2764 .owner = THIS_MODULE,
1da177e4
LT		 2765};
2766
9a812198
JV		 2767/*
2768 * Generic Netlink interface
2769 */
2770
2771/* IPVS genetlink family */
2772static struct genl_family ip_vs_genl_family = {
2773 .id = GENL_ID_GENERATE,
2774 .hdrsize = 0,
2775 .name = IPVS_GENL_NAME,
2776 .version = IPVS_GENL_VERSION,
2777 .maxattr = IPVS_CMD_MAX,
c6d2d445 2778 .netnsok = true, /* Make ipvsadm to work on netns */
9a812198
JV		 2779};
2780
2781/* Policy used for first-level command attributes */
2782static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2783 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2784 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2785 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2786 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2787 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2788 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2789};
2790
2791/* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2792static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2793 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2794 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2795 .len = IP_VS_IFNAME_MAXLEN },
2796 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2797};
2798
2799/* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2800static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2801 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2802 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2803 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2804 .len = sizeof(union nf_inet_addr) },
2805 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2806 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2807 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2808 .len = IP_VS_SCHEDNAME_MAXLEN },
0d1e71b0
SH		 2809 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2810 .len = IP_VS_PENAME_MAXLEN },
9a812198
JV		 2811 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2812 .len = sizeof(struct ip_vs_flags) },
2813 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2814 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2815 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2816};
2817
2818/* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2819static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2820 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2821 .len = sizeof(union nf_inet_addr) },
2822 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2823 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2824 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2825 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2826 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2827 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2828 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2829 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2830 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2831};
2832
2833static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2834 struct ip_vs_stats *stats)
2835{
55a3d4e1 2836 struct ip_vs_stats_user ustats;
9a812198
JV		 2837 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2838 if (!nl_stats)
2839 return -EMSGSIZE;
2840
55a3d4e1 2841 ip_vs_copy_stats(&ustats, stats);
9a812198 2842
969e8e25
DM		 2843 if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, ustats.conns) ||
2844 nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, ustats.inpkts) ||
2845 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, ustats.outpkts) ||
2846 nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, ustats.inbytes) ||
2847 nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, ustats.outbytes) ||
2848 nla_put_u32(skb, IPVS_STATS_ATTR_CPS, ustats.cps) ||
2849 nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, ustats.inpps) ||
2850 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, ustats.outpps) ||
2851 nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, ustats.inbps) ||
2852 nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, ustats.outbps))
2853 goto nla_put_failure;
9a812198
JV		 2854 nla_nest_end(skb, nl_stats);
2855
2856 return 0;
2857
2858nla_put_failure:
9a812198
JV		 2859 nla_nest_cancel(skb, nl_stats);
2860 return -EMSGSIZE;
2861}
2862
2863static int ip_vs_genl_fill_service(struct sk_buff *skb,
2864 struct ip_vs_service *svc)
2865{
2866 struct nlattr *nl_service;
2867 struct ip_vs_flags flags = { .flags = svc->flags,
2868 .mask = ~0 };
2869
2870 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2871 if (!nl_service)
2872 return -EMSGSIZE;
2873
969e8e25
DM		 2874 if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2875 goto nla_put_failure;
9a812198 2876 if (svc->fwmark) {
969e8e25
DM		 2877 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2878 goto nla_put_failure;
9a812198 2879 } else {
969e8e25
DM		 2880 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2881 nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2882 nla_put_u16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2883 goto nla_put_failure;
9a812198
JV		 2884 }
2885
969e8e25
DM		 2886 if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, svc->scheduler->name) ||
2887 (svc->pe &&
2888 nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, svc->pe->name)) ||
2889 nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2890 nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2891 nla_put_u32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
2892 goto nla_put_failure;
9a812198
JV		 2893 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2894 goto nla_put_failure;
2895
2896 nla_nest_end(skb, nl_service);
2897
2898 return 0;
2899
2900nla_put_failure:
2901 nla_nest_cancel(skb, nl_service);
2902 return -EMSGSIZE;
2903}
2904
2905static int ip_vs_genl_dump_service(struct sk_buff *skb,
2906 struct ip_vs_service *svc,
2907 struct netlink_callback *cb)
2908{
2909 void *hdr;
2910
2911 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
2912 &ip_vs_genl_family, NLM_F_MULTI,
2913 IPVS_CMD_NEW_SERVICE);
2914 if (!hdr)
2915 return -EMSGSIZE;
2916
2917 if (ip_vs_genl_fill_service(skb, svc) < 0)
2918 goto nla_put_failure;
2919
2920 return genlmsg_end(skb, hdr);
2921
2922nla_put_failure:
2923 genlmsg_cancel(skb, hdr);
2924 return -EMSGSIZE;
2925}
2926
2927static int ip_vs_genl_dump_services(struct sk_buff *skb,
2928 struct netlink_callback *cb)
2929{
2930 int idx = 0, i;
2931 int start = cb->args[0];
2932 struct ip_vs_service *svc;
fc723250 2933 struct net *net = skb_sknet(skb);
9a812198
JV		 2934
2935 mutex_lock(&__ip_vs_mutex);
2936 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2937 list_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
fc723250 2938 if (++idx <= start || !net_eq(svc->net, net))
9a812198
JV		 2939 continue;
2940 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2941 idx--;
2942 goto nla_put_failure;
2943 }
2944 }
2945 }
2946
2947 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
2948 list_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
fc723250 2949 if (++idx <= start || !net_eq(svc->net, net))
9a812198
JV		 2950 continue;
2951 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
2952 idx--;
2953 goto nla_put_failure;
2954 }
2955 }
2956 }
2957
2958nla_put_failure:
2959 mutex_unlock(&__ip_vs_mutex);
2960 cb->args[0] = idx;
2961
2962 return skb->len;
2963}
2964
fc723250
HS		 2965static int ip_vs_genl_parse_service(struct net *net,
2966 struct ip_vs_service_user_kern *usvc,
26c15cfd
JA		 2967 struct nlattr *nla, int full_entry,
2968 struct ip_vs_service **ret_svc)
9a812198
JV		 2969{
2970 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
2971 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
26c15cfd 2972 struct ip_vs_service *svc;
9a812198
JV		 2973
2974 /* Parse mandatory identifying service fields first */
2975 if (nla == NULL ||
2976 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
2977 return -EINVAL;
2978
2979 nla_af = attrs[IPVS_SVC_ATTR_AF];
2980 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
2981 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
2982 nla_port = attrs[IPVS_SVC_ATTR_PORT];
2983 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
2984
2985 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
2986 return -EINVAL;
2987
258c8893
SH		 2988 memset(usvc, 0, sizeof(*usvc));
2989
c860c6b1 2990 usvc->af = nla_get_u16(nla_af);
f94fd041
JV		 2991#ifdef CONFIG_IP_VS_IPV6
2992 if (usvc->af != AF_INET && usvc->af != AF_INET6)
2993#else
2994 if (usvc->af != AF_INET)
2995#endif
9a812198
JV		 2996 return -EAFNOSUPPORT;
2997
2998 if (nla_fwmark) {
2999 usvc->protocol = IPPROTO_TCP;
3000 usvc->fwmark = nla_get_u32(nla_fwmark);
3001 } else {
3002 usvc->protocol = nla_get_u16(nla_protocol);
3003 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3004 usvc->port = nla_get_u16(nla_port);
3005 usvc->fwmark = 0;
3006 }
3007
26c15cfd 3008 if (usvc->fwmark)
fc723250 3009 svc = __ip_vs_svc_fwm_find(net, usvc->af, usvc->fwmark);
26c15cfd 3010 else
fc723250 3011 svc = __ip_vs_service_find(net, usvc->af, usvc->protocol,
26c15cfd
JA		 3012 &usvc->addr, usvc->port);
3013 *ret_svc = svc;
3014
9a812198
JV		 3015 /* If a full entry was requested, check for the additional fields */
3016 if (full_entry) {
0d1e71b0 3017 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
9a812198
JV		 3018 *nla_netmask;
3019 struct ip_vs_flags flags;
9a812198
JV		 3020
3021 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
0d1e71b0 3022 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
9a812198
JV		 3023 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3024 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3025 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3026
3027 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3028 return -EINVAL;
3029
3030 nla_memcpy(&flags, nla_flags, sizeof(flags));
3031
3032 /* prefill flags from service if it already exists */
26c15cfd 3033 if (svc)
9a812198 3034 usvc->flags = svc->flags;
9a812198
JV		 3035
3036 /* set new flags from userland */
3037 usvc->flags = (usvc->flags & ~flags.mask) |
3038 (flags.flags & flags.mask);
c860c6b1 3039 usvc->sched_name = nla_data(nla_sched);
0d1e71b0 3040 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
9a812198
JV		 3041 usvc->timeout = nla_get_u32(nla_timeout);
3042 usvc->netmask = nla_get_u32(nla_netmask);
3043 }
3044
3045 return 0;
3046}
3047
fc723250
HS		 3048static struct ip_vs_service *ip_vs_genl_find_service(struct net *net,
3049 struct nlattr *nla)
9a812198 3050{
c860c6b1 3051 struct ip_vs_service_user_kern usvc;
26c15cfd 3052 struct ip_vs_service *svc;
9a812198
JV		 3053 int ret;
3054
fc723250 3055 ret = ip_vs_genl_parse_service(net, &usvc, nla, 0, &svc);
26c15cfd 3056 return ret ? ERR_PTR(ret) : svc;
9a812198
JV		 3057}
3058
3059static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3060{
3061 struct nlattr *nl_dest;
3062
3063 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3064 if (!nl_dest)
3065 return -EMSGSIZE;
3066
969e8e25
DM		 3067 if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3068 nla_put_u16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3069 nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3070 (atomic_read(&dest->conn_flags) &
3071 IP_VS_CONN_F_FWD_MASK)) ||
3072 nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3073 atomic_read(&dest->weight)) ||
3074 nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3075 nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3076 nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3077 atomic_read(&dest->activeconns)) ||
3078 nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3079 atomic_read(&dest->inactconns)) ||
3080 nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3081 atomic_read(&dest->persistconns)))
3082 goto nla_put_failure;
9a812198
JV		 3083 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
3084 goto nla_put_failure;
3085
3086 nla_nest_end(skb, nl_dest);
3087
3088 return 0;
3089
3090nla_put_failure:
3091 nla_nest_cancel(skb, nl_dest);
3092 return -EMSGSIZE;
3093}
3094
3095static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3096 struct netlink_callback *cb)
3097{
3098 void *hdr;
3099
3100 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
3101 &ip_vs_genl_family, NLM_F_MULTI,
3102 IPVS_CMD_NEW_DEST);
3103 if (!hdr)
3104 return -EMSGSIZE;
3105
3106 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3107 goto nla_put_failure;
3108
3109 return genlmsg_end(skb, hdr);
3110
3111nla_put_failure:
3112 genlmsg_cancel(skb, hdr);
3113 return -EMSGSIZE;
3114}
3115
3116static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3117 struct netlink_callback *cb)
3118{
3119 int idx = 0;
3120 int start = cb->args[0];
3121 struct ip_vs_service *svc;
3122 struct ip_vs_dest *dest;
3123 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
a0840e2e 3124 struct net *net = skb_sknet(skb);
9a812198
JV		 3125
3126 mutex_lock(&__ip_vs_mutex);
3127
3128 /* Try to find the service for which to dump destinations */
3129 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3130 IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3131 goto out_err;
3132
a0840e2e 3133
fc723250 3134 svc = ip_vs_genl_find_service(net, attrs[IPVS_CMD_ATTR_SERVICE]);
9a812198
JV		 3135 if (IS_ERR(svc) || svc == NULL)
3136 goto out_err;
3137
3138 /* Dump the destinations */
3139 list_for_each_entry(dest, &svc->destinations, n_list) {
3140 if (++idx <= start)
3141 continue;
3142 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3143 idx--;
3144 goto nla_put_failure;
3145 }
3146 }
3147
3148nla_put_failure:
3149 cb->args[0] = idx;
9a812198
JV		 3150
3151out_err:
3152 mutex_unlock(&__ip_vs_mutex);
3153
3154 return skb->len;
3155}
3156
c860c6b1 3157static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
9a812198
JV		 3158 struct nlattr *nla, int full_entry)
3159{
3160 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3161 struct nlattr *nla_addr, *nla_port;
3162
3163 /* Parse mandatory identifying destination fields first */
3164 if (nla == NULL ||
3165 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3166 return -EINVAL;
3167
3168 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3169 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3170
3171 if (!(nla_addr && nla_port))
3172 return -EINVAL;
3173
258c8893
SH		 3174 memset(udest, 0, sizeof(*udest));
3175
9a812198
JV		 3176 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3177 udest->port = nla_get_u16(nla_port);
3178
3179 /* If a full entry was requested, check for the additional fields */
3180 if (full_entry) {
3181 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3182 *nla_l_thresh;
3183
3184 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3185 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3186 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3187 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3188
3189 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3190 return -EINVAL;
3191
3192 udest->conn_flags = nla_get_u32(nla_fwd)
3193 & IP_VS_CONN_F_FWD_MASK;
3194 udest->weight = nla_get_u32(nla_weight);
3195 udest->u_threshold = nla_get_u32(nla_u_thresh);
3196 udest->l_threshold = nla_get_u32(nla_l_thresh);
3197 }
3198
3199 return 0;
3200}
3201
3202static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __be32 state,
3203 const char *mcast_ifn, __be32 syncid)
3204{
3205 struct nlattr *nl_daemon;
3206
3207 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3208 if (!nl_daemon)
3209 return -EMSGSIZE;
3210
969e8e25
DM		 3211 if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3212 nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn) ||
3213 nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid))
3214 goto nla_put_failure;
9a812198
JV		 3215 nla_nest_end(skb, nl_daemon);
3216
3217 return 0;
3218
3219nla_put_failure:
3220 nla_nest_cancel(skb, nl_daemon);
3221 return -EMSGSIZE;
3222}
3223
3224static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __be32 state,
3225 const char *mcast_ifn, __be32 syncid,
3226 struct netlink_callback *cb)
3227{
3228 void *hdr;
3229 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
3230 &ip_vs_genl_family, NLM_F_MULTI,
3231 IPVS_CMD_NEW_DAEMON);
3232 if (!hdr)
3233 return -EMSGSIZE;
3234
3235 if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
3236 goto nla_put_failure;
3237
3238 return genlmsg_end(skb, hdr);
3239
3240nla_put_failure:
3241 genlmsg_cancel(skb, hdr);
3242 return -EMSGSIZE;
3243}
3244
3245static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3246 struct netlink_callback *cb)
3247{
a09d1977 3248 struct net *net = skb_sknet(skb);
f131315f
HS		 3249 struct netns_ipvs *ipvs = net_ipvs(net);
3250
ae1d48b2 3251 mutex_lock(&ipvs->sync_mutex);
f131315f 3252 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
9a812198 3253 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
f131315f
HS		 3254 ipvs->master_mcast_ifn,
3255 ipvs->master_syncid, cb) < 0)
9a812198
JV		 3256 goto nla_put_failure;
3257
3258 cb->args[0] = 1;
3259 }
3260
f131315f 3261 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
9a812198 3262 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
f131315f
HS		 3263 ipvs->backup_mcast_ifn,
3264 ipvs->backup_syncid, cb) < 0)
9a812198
JV		 3265 goto nla_put_failure;
3266
3267 cb->args[1] = 1;
3268 }
3269
3270nla_put_failure:
ae1d48b2 3271 mutex_unlock(&ipvs->sync_mutex);
9a812198
JV		 3272
3273 return skb->len;
3274}
3275
f131315f 3276static int ip_vs_genl_new_daemon(struct net *net, struct nlattr **attrs)
9a812198
JV		 3277{
3278 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3279 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3280 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3281 return -EINVAL;
3282
f131315f
HS		 3283 return start_sync_thread(net,
3284 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
9a812198
JV		 3285 nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3286 nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3287}
3288
f131315f 3289static int ip_vs_genl_del_daemon(struct net *net, struct nlattr **attrs)
9a812198
JV		 3290{
3291 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3292 return -EINVAL;
3293
f131315f
HS		 3294 return stop_sync_thread(net,
3295 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
9a812198
JV		 3296}
3297
9330419d 3298static int ip_vs_genl_set_config(struct net *net, struct nlattr **attrs)
9a812198
JV		 3299{
3300 struct ip_vs_timeout_user t;
3301
9330419d 3302 __ip_vs_get_timeouts(net, &t);
9a812198
JV		 3303
3304 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3305 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3306
3307 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3308 t.tcp_fin_timeout =
3309 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3310
3311 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3312 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3313
9330419d 3314 return ip_vs_set_timeout(net, &t);
9a812198
JV		 3315}
3316
ae1d48b2 3317static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
9a812198 3318{
9a812198 3319 int ret = 0, cmd;
fc723250 3320 struct net *net;
a0840e2e 3321 struct netns_ipvs *ipvs;
9a812198 3322
fc723250 3323 net = skb_sknet(skb);
a0840e2e 3324 ipvs = net_ipvs(net);
9a812198
JV		 3325 cmd = info->genlhdr->cmd;
3326
ae1d48b2 3327 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
9a812198
JV		 3328 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3329
ae1d48b2 3330 mutex_lock(&ipvs->sync_mutex);
9a812198
JV		 3331 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3332 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3333 info->attrs[IPVS_CMD_ATTR_DAEMON],
3334 ip_vs_daemon_policy)) {
3335 ret = -EINVAL;
3336 goto out;
3337 }
3338
3339 if (cmd == IPVS_CMD_NEW_DAEMON)
f131315f 3340 ret = ip_vs_genl_new_daemon(net, daemon_attrs);
9a812198 3341 else
f131315f 3342 ret = ip_vs_genl_del_daemon(net, daemon_attrs);
ae1d48b2
HS		 3343out:
3344 mutex_unlock(&ipvs->sync_mutex);
3345 }
3346 return ret;
3347}
3348
3349static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3350{
3351 struct ip_vs_service *svc = NULL;
3352 struct ip_vs_service_user_kern usvc;
3353 struct ip_vs_dest_user_kern udest;
3354 int ret = 0, cmd;
3355 int need_full_svc = 0, need_full_dest = 0;
3356 struct net *net;
ae1d48b2
HS		 3357
3358 net = skb_sknet(skb);
ae1d48b2
HS		 3359 cmd = info->genlhdr->cmd;
3360
3361 mutex_lock(&__ip_vs_mutex);
3362
3363 if (cmd == IPVS_CMD_FLUSH) {
3364 ret = ip_vs_flush(net);
3365 goto out;
3366 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3367 ret = ip_vs_genl_set_config(net, info->attrs);
9a812198
JV		 3368 goto out;
3369 } else if (cmd == IPVS_CMD_ZERO &&
3370 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
fc723250 3371 ret = ip_vs_zero_all(net);
9a812198
JV		 3372 goto out;
3373 }
3374
3375 /* All following commands require a service argument, so check if we
3376 * received a valid one. We need a full service specification when
3377 * adding / editing a service. Only identifying members otherwise. */
3378 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3379 need_full_svc = 1;
3380
fc723250 3381 ret = ip_vs_genl_parse_service(net, &usvc,
9a812198 3382 info->attrs[IPVS_CMD_ATTR_SERVICE],
26c15cfd 3383 need_full_svc, &svc);
9a812198
JV		 3384 if (ret)
3385 goto out;
3386
9a812198
JV		 3387 /* Unless we're adding a new service, the service must already exist */
3388 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3389 ret = -ESRCH;
3390 goto out;
3391 }
3392
3393 /* Destination commands require a valid destination argument. For
3394 * adding / editing a destination, we need a full destination
3395 * specification. */
3396 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3397 cmd == IPVS_CMD_DEL_DEST) {
3398 if (cmd != IPVS_CMD_DEL_DEST)
3399 need_full_dest = 1;
3400
3401 ret = ip_vs_genl_parse_dest(&udest,
3402 info->attrs[IPVS_CMD_ATTR_DEST],
3403 need_full_dest);
3404 if (ret)
3405 goto out;
3406 }
3407
3408 switch (cmd) {
3409 case IPVS_CMD_NEW_SERVICE:
3410 if (svc == NULL)
fc723250 3411 ret = ip_vs_add_service(net, &usvc, &svc);
9a812198
JV		 3412 else
3413 ret = -EEXIST;
3414 break;
3415 case IPVS_CMD_SET_SERVICE:
3416 ret = ip_vs_edit_service(svc, &usvc);
3417 break;
3418 case IPVS_CMD_DEL_SERVICE:
3419 ret = ip_vs_del_service(svc);
26c15cfd 3420 /* do not use svc, it can be freed */
9a812198
JV		 3421 break;
3422 case IPVS_CMD_NEW_DEST:
3423 ret = ip_vs_add_dest(svc, &udest);
3424 break;
3425 case IPVS_CMD_SET_DEST:
3426 ret = ip_vs_edit_dest(svc, &udest);
3427 break;
3428 case IPVS_CMD_DEL_DEST:
3429 ret = ip_vs_del_dest(svc, &udest);
3430 break;
3431 case IPVS_CMD_ZERO:
3432 ret = ip_vs_zero_service(svc);
3433 break;
3434 default:
3435 ret = -EINVAL;
3436 }
3437
3438out:
9a812198
JV		 3439 mutex_unlock(&__ip_vs_mutex);
3440
3441 return ret;
3442}
3443
3444static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3445{
3446 struct sk_buff *msg;
3447 void *reply;
3448 int ret, cmd, reply_cmd;
fc723250 3449 struct net *net;
9a812198 3450
fc723250 3451 net = skb_sknet(skb);
9a812198
JV		 3452 cmd = info->genlhdr->cmd;
3453
3454 if (cmd == IPVS_CMD_GET_SERVICE)
3455 reply_cmd = IPVS_CMD_NEW_SERVICE;
3456 else if (cmd == IPVS_CMD_GET_INFO)
3457 reply_cmd = IPVS_CMD_SET_INFO;
3458 else if (cmd == IPVS_CMD_GET_CONFIG)
3459 reply_cmd = IPVS_CMD_SET_CONFIG;
3460 else {
1e3e238e 3461 pr_err("unknown Generic Netlink command\n");
9a812198
JV		 3462 return -EINVAL;
3463 }
3464
3465 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3466 if (!msg)
3467 return -ENOMEM;
3468
3469 mutex_lock(&__ip_vs_mutex);
3470
3471 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3472 if (reply == NULL)
3473 goto nla_put_failure;
3474
3475 switch (cmd) {
3476 case IPVS_CMD_GET_SERVICE:
3477 {
3478 struct ip_vs_service *svc;
3479
fc723250
HS		 3480 svc = ip_vs_genl_find_service(net,
3481 info->attrs[IPVS_CMD_ATTR_SERVICE]);
9a812198
JV		 3482 if (IS_ERR(svc)) {
3483 ret = PTR_ERR(svc);
3484 goto out_err;
3485 } else if (svc) {
3486 ret = ip_vs_genl_fill_service(msg, svc);
9a812198
JV		 3487 if (ret)
3488 goto nla_put_failure;
3489 } else {
3490 ret = -ESRCH;
3491 goto out_err;
3492 }
3493
3494 break;
3495 }
3496
3497 case IPVS_CMD_GET_CONFIG:
3498 {
3499 struct ip_vs_timeout_user t;
3500
9330419d 3501 __ip_vs_get_timeouts(net, &t);
9a812198 3502#ifdef CONFIG_IP_VS_PROTO_TCP
969e8e25
DM		 3503 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3504 t.tcp_timeout) ||
3505 nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3506 t.tcp_fin_timeout))
3507 goto nla_put_failure;
9a812198
JV		 3508#endif
3509#ifdef CONFIG_IP_VS_PROTO_UDP
969e8e25
DM		 3510 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3511 goto nla_put_failure;
9a812198
JV		 3512#endif
3513
3514 break;
3515 }
3516
3517 case IPVS_CMD_GET_INFO:
969e8e25
DM		 3518 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3519 IP_VS_VERSION_CODE) ||
3520 nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3521 ip_vs_conn_tab_size))
3522 goto nla_put_failure;
9a812198
JV		 3523 break;
3524 }
3525
3526 genlmsg_end(msg, reply);
134e6375 3527 ret = genlmsg_reply(msg, info);
9a812198
JV		 3528 goto out;
3529
3530nla_put_failure:
1e3e238e 3531 pr_err("not enough space in Netlink message\n");
9a812198
JV		 3532 ret = -EMSGSIZE;
3533
3534out_err:
3535 nlmsg_free(msg);
3536out:
3537 mutex_unlock(&__ip_vs_mutex);
3538
3539 return ret;
3540}
3541
3542
3543static struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3544 {
3545 .cmd = IPVS_CMD_NEW_SERVICE,
3546 .flags = GENL_ADMIN_PERM,
3547 .policy = ip_vs_cmd_policy,
3548 .doit = ip_vs_genl_set_cmd,
3549 },
3550 {
3551 .cmd = IPVS_CMD_SET_SERVICE,
3552 .flags = GENL_ADMIN_PERM,
3553 .policy = ip_vs_cmd_policy,
3554 .doit = ip_vs_genl_set_cmd,
3555 },
3556 {
3557 .cmd = IPVS_CMD_DEL_SERVICE,
3558 .flags = GENL_ADMIN_PERM,
3559 .policy = ip_vs_cmd_policy,
3560 .doit = ip_vs_genl_set_cmd,
3561 },
3562 {
3563 .cmd = IPVS_CMD_GET_SERVICE,
3564 .flags = GENL_ADMIN_PERM,
3565 .doit = ip_vs_genl_get_cmd,
3566 .dumpit = ip_vs_genl_dump_services,
3567 .policy = ip_vs_cmd_policy,
3568 },
3569 {
3570 .cmd = IPVS_CMD_NEW_DEST,
3571 .flags = GENL_ADMIN_PERM,
3572 .policy = ip_vs_cmd_policy,
3573 .doit = ip_vs_genl_set_cmd,
3574 },
3575 {
3576 .cmd = IPVS_CMD_SET_DEST,
3577 .flags = GENL_ADMIN_PERM,
3578 .policy = ip_vs_cmd_policy,
3579 .doit = ip_vs_genl_set_cmd,
3580 },
3581 {
3582 .cmd = IPVS_CMD_DEL_DEST,
3583 .flags = GENL_ADMIN_PERM,
3584 .policy = ip_vs_cmd_policy,
3585 .doit = ip_vs_genl_set_cmd,
3586 },
3587 {
3588 .cmd = IPVS_CMD_GET_DEST,
3589 .flags = GENL_ADMIN_PERM,
3590 .policy = ip_vs_cmd_policy,
3591 .dumpit = ip_vs_genl_dump_dests,
3592 },
3593 {
3594 .cmd = IPVS_CMD_NEW_DAEMON,
3595 .flags = GENL_ADMIN_PERM,
3596 .policy = ip_vs_cmd_policy,
ae1d48b2 3597 .doit = ip_vs_genl_set_daemon,
9a812198
JV		 3598 },
3599 {
3600 .cmd = IPVS_CMD_DEL_DAEMON,
3601 .flags = GENL_ADMIN_PERM,
3602 .policy = ip_vs_cmd_policy,
ae1d48b2 3603 .doit = ip_vs_genl_set_daemon,
9a812198
JV		 3604 },
3605 {
3606 .cmd = IPVS_CMD_GET_DAEMON,
3607 .flags = GENL_ADMIN_PERM,
3608 .dumpit = ip_vs_genl_dump_daemons,
3609 },
3610 {
3611 .cmd = IPVS_CMD_SET_CONFIG,
3612 .flags = GENL_ADMIN_PERM,
3613 .policy = ip_vs_cmd_policy,
3614 .doit = ip_vs_genl_set_cmd,
3615 },
3616 {
3617 .cmd = IPVS_CMD_GET_CONFIG,
3618 .flags = GENL_ADMIN_PERM,
3619 .doit = ip_vs_genl_get_cmd,
3620 },
3621 {
3622 .cmd = IPVS_CMD_GET_INFO,
3623 .flags = GENL_ADMIN_PERM,
3624 .doit = ip_vs_genl_get_cmd,
3625 },
3626 {
3627 .cmd = IPVS_CMD_ZERO,
3628 .flags = GENL_ADMIN_PERM,
3629 .policy = ip_vs_cmd_policy,
3630 .doit = ip_vs_genl_set_cmd,
3631 },
3632 {
3633 .cmd = IPVS_CMD_FLUSH,
3634 .flags = GENL_ADMIN_PERM,
3635 .doit = ip_vs_genl_set_cmd,
3636 },
3637};
3638
3639static int __init ip_vs_genl_register(void)
3640{
8f698d54
MM		 3641 return genl_register_family_with_ops(&ip_vs_genl_family,
3642 ip_vs_genl_ops, ARRAY_SIZE(ip_vs_genl_ops));
9a812198
JV		 3643}
3644
3645static void ip_vs_genl_unregister(void)
3646{
3647 genl_unregister_family(&ip_vs_genl_family);
3648}
3649
3650/* End of Generic Netlink interface definitions */
3651
61b1ab45
HS		 3652/*
3653 * per netns intit/exit func.
3654 */
14e40546 3655#ifdef CONFIG_SYSCTL
503cf15a 3656int __net_init ip_vs_control_net_init_sysctl(struct net *net)
61b1ab45 3657{
fc723250
HS		 3658 int idx;
3659 struct netns_ipvs *ipvs = net_ipvs(net);
a0840e2e 3660 struct ctl_table *tbl;
fc723250 3661
a0840e2e
HS		 3662 atomic_set(&ipvs->dropentry, 0);
3663 spin_lock_init(&ipvs->dropentry_lock);
3664 spin_lock_init(&ipvs->droppacket_lock);
3665 spin_lock_init(&ipvs->securetcp_lock);
a0840e2e
HS		 3666
3667 if (!net_eq(net, &init_net)) {
3668 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3669 if (tbl == NULL)
14e40546 3670 return -ENOMEM;
a0840e2e
HS		 3671 } else
3672 tbl = vs_vars;
3673 /* Initialize sysctl defaults */
3674 idx = 0;
3675 ipvs->sysctl_amemthresh = 1024;
3676 tbl[idx++].data = &ipvs->sysctl_amemthresh;
3677 ipvs->sysctl_am_droprate = 10;
3678 tbl[idx++].data = &ipvs->sysctl_am_droprate;
3679 tbl[idx++].data = &ipvs->sysctl_drop_entry;
3680 tbl[idx++].data = &ipvs->sysctl_drop_packet;
3681#ifdef CONFIG_IP_VS_NFCT
3682 tbl[idx++].data = &ipvs->sysctl_conntrack;
3683#endif
3684 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3685 ipvs->sysctl_snat_reroute = 1;
3686 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3687 ipvs->sysctl_sync_ver = 1;
3688 tbl[idx++].data = &ipvs->sysctl_sync_ver;
1c003b15
PNA		 3689 ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3690 tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3691 ipvs->sysctl_sync_sock_size = 0;
3692 tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
a0840e2e
HS		 3693 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3694 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3695 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
59e0350e
SH		 3696 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3697 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
a0840e2e
HS		 3698 tbl[idx].data = &ipvs->sysctl_sync_threshold;
3699 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
749c42b6
JA		 3700 ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3701 tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3702 ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3703 tbl[idx++].data = &ipvs->sysctl_sync_retries;
a0840e2e
HS		 3704 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3705
3706
ec8f23ce 3707 ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
0443929f
SH		 3708 if (ipvs->sysctl_hdr == NULL) {
3709 if (!net_eq(net, &init_net))
3710 kfree(tbl);
14e40546 3711 return -ENOMEM;
0443929f 3712 }
6ef757f9 3713 ip_vs_start_estimator(net, &ipvs->tot_stats);
a0840e2e 3714 ipvs->sysctl_tbl = tbl;
f6340ee0
HS		 3715 /* Schedule defense work */
3716 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3717 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
61b1ab45 3718
61b1ab45 3719 return 0;
61b1ab45
HS		 3720}
3721
8f9b9a2f 3722void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net)
61b1ab45 3723{
b17fc996
HS		 3724 struct netns_ipvs *ipvs = net_ipvs(net);
3725
f2431e6e
HS		 3726 cancel_delayed_work_sync(&ipvs->defense_work);
3727 cancel_work_sync(&ipvs->defense_work.work);
a0840e2e 3728 unregister_net_sysctl_table(ipvs->sysctl_hdr);
14e40546
SH		 3729}
3730
3731#else
3732
503cf15a 3733int __net_init ip_vs_control_net_init_sysctl(struct net *net) { return 0; }
8f9b9a2f 3734void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net) { }
14e40546 3735
0443929f 3736#endif
14e40546 3737
7a4f0761
HS		 3738static struct notifier_block ip_vs_dst_notifier = {
3739 .notifier_call = ip_vs_dst_event,
3740};
3741
503cf15a 3742int __net_init ip_vs_control_net_init(struct net *net)
14e40546
SH		 3743{
3744 int idx;
3745 struct netns_ipvs *ipvs = net_ipvs(net);
3746
3458e21c 3747 rwlock_init(&ipvs->rs_lock);
14e40546
SH		 3748
3749 /* Initialize rs_table */
3750 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
3751 INIT_LIST_HEAD(&ipvs->rs_table[idx]);
3752
3753 INIT_LIST_HEAD(&ipvs->dest_trash);
3754 atomic_set(&ipvs->ftpsvc_counter, 0);
3755 atomic_set(&ipvs->nullsvc_counter, 0);
3756
3757 /* procfs stats */
3758 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
0a9ee813 3759 if (!ipvs->tot_stats.cpustats)
14e40546 3760 return -ENOMEM;
0a9ee813 3761
14e40546
SH		 3762 spin_lock_init(&ipvs->tot_stats.lock);
3763
3764 proc_net_fops_create(net, "ip_vs", 0, &ip_vs_info_fops);
3765 proc_net_fops_create(net, "ip_vs_stats", 0, &ip_vs_stats_fops);
3766 proc_net_fops_create(net, "ip_vs_stats_percpu", 0,
3767 &ip_vs_stats_percpu_fops);
3768
503cf15a 3769 if (ip_vs_control_net_init_sysctl(net))
14e40546
SH		 3770 goto err;
3771
3772 return 0;
3773
3774err:
2a0751af 3775 free_percpu(ipvs->tot_stats.cpustats);
61b1ab45
HS		 3776 return -ENOMEM;
3777}
3778
503cf15a 3779void __net_exit ip_vs_control_net_cleanup(struct net *net)
61b1ab45 3780{
b17fc996
HS		 3781 struct netns_ipvs *ipvs = net_ipvs(net);
3782
f2431e6e 3783 ip_vs_trash_cleanup(net);
6ef757f9 3784 ip_vs_stop_estimator(net, &ipvs->tot_stats);
503cf15a 3785 ip_vs_control_net_cleanup_sysctl(net);
b17fc996 3786 proc_net_remove(net, "ip_vs_stats_percpu");
61b1ab45
HS		 3787 proc_net_remove(net, "ip_vs_stats");
3788 proc_net_remove(net, "ip_vs");
2a0751af 3789 free_percpu(ipvs->tot_stats.cpustats);
61b1ab45
HS		 3790}
3791
8537de8a 3792int __init ip_vs_register_nl_ioctl(void)
1da177e4 3793{
fc723250 3794 int ret;
1da177e4 3795
1da177e4
LT		 3796 ret = nf_register_sockopt(&ip_vs_sockopts);
3797 if (ret) {
1e3e238e 3798 pr_err("cannot register sockopt.\n");
7a4f0761 3799 goto err_sock;
1da177e4
LT		 3800 }
3801
9a812198
JV		 3802 ret = ip_vs_genl_register();
3803 if (ret) {
1e3e238e 3804 pr_err("cannot register Generic Netlink interface.\n");
7a4f0761 3805 goto err_genl;
9a812198 3806 }
1da177e4 3807 return 0;
fc723250 3808
7a4f0761
HS		 3809err_genl:
3810 nf_unregister_sockopt(&ip_vs_sockopts);
3811err_sock:
fc723250 3812 return ret;
1da177e4
LT		 3813}
3814
8537de8a
HS		 3815void ip_vs_unregister_nl_ioctl(void)
3816{
3817 ip_vs_genl_unregister();
3818 nf_unregister_sockopt(&ip_vs_sockopts);
3819}
3820
3821int __init ip_vs_control_init(void)
3822{
3823 int idx;
3824 int ret;
3825
3826 EnterFunction(2);
3827
3828 /* Initialize svc_table, ip_vs_svc_fwm_table, rs_table */
3829 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
3830 INIT_LIST_HEAD(&ip_vs_svc_table[idx]);
3831 INIT_LIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3832 }
3833
3834 smp_wmb(); /* Do we really need it now ? */
3835
3836 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
3837 if (ret < 0)
3838 return ret;
3839
3840 LeaveFunction(2);
3841 return 0;
3842}
3843
1da177e4
LT		 3844
3845void ip_vs_control_cleanup(void)
3846{
3847 EnterFunction(2);
7676e345 3848 unregister_netdevice_notifier(&ip_vs_dst_notifier);
1da177e4
LT		 3849 LeaveFunction(2);
3850}
Linux kernel - Deliverables
This page took 0.981466 seconds and 5 git commands to generate.