projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ipvs: convert dests to rcu
[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 */
578bc3ef 74static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup);
7a4f0761
HS		 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? */
c24584c0
ED		 79static bool __ip_vs_addr_is_local_v6(struct net *net,
80 const struct in6_addr *addr)
09571c7a 81{
4c9483b2
DM		 82 struct flowi6 fl6 = {
83 .daddr = *addr,
09571c7a 84 };
c24584c0
ED		 85 struct dst_entry *dst = ip6_route_output(net, NULL, &fl6);
86 bool is_local;
09571c7a 87
c24584c0 88 is_local = !dst->error && dst->dev && (dst->dev->flags & IFF_LOOPBACK);
09571c7a 89
c24584c0
ED		 90 dst_release(dst);
91 return is_local;
09571c7a
VB		 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;
e9836f24 274 __u32 ahash;
1da177e4 275
b18610de
JV		 276#ifdef CONFIG_IP_VS_IPV6
277 if (af == AF_INET6)
278 addr_fold = addr->ip6[0]^addr->ip6[1]^
279 addr->ip6[2]^addr->ip6[3];
280#endif
e9836f24
JA		 281 ahash = ntohl(addr_fold);
282 ahash ^= ((size_t) net >> 8);
b18610de 283
e9836f24
JA		 284 return (proto ^ ahash ^ (porth >> IP_VS_SVC_TAB_BITS) ^ porth) &
285 IP_VS_SVC_TAB_MASK;
1da177e4
LT		 286}
287
288/*
289 * Returns hash value of fwmark for virtual service lookup
290 */
95c96174 291static inline unsigned int ip_vs_svc_fwm_hashkey(struct net *net, __u32 fwmark)
1da177e4 292{
fc723250 293 return (((size_t)net>>8) ^ fwmark) & IP_VS_SVC_TAB_MASK;
1da177e4
LT		 294}
295
296/*
fc723250 297 * Hashes a service in the ip_vs_svc_table by <netns,proto,addr,port>
1da177e4
LT		 298 * or in the ip_vs_svc_fwm_table by fwmark.
299 * Should be called with locked tables.
300 */
301static int ip_vs_svc_hash(struct ip_vs_service *svc)
302{
95c96174 303 unsigned int hash;
1da177e4
LT		 304
305 if (svc->flags & IP_VS_SVC_F_HASHED) {
1e3e238e
HE		 306 pr_err("%s(): request for already hashed, called from %pF\n",
307 __func__, __builtin_return_address(0));
1da177e4
LT		 308 return 0;
309 }
310
311 if (svc->fwmark == 0) {
312 /*
fc723250 313 * Hash it by <netns,protocol,addr,port> in ip_vs_svc_table
1da177e4 314 */
fc723250
HS		 315 hash = ip_vs_svc_hashkey(svc->net, svc->af, svc->protocol,
316 &svc->addr, svc->port);
1da177e4
LT		 317 list_add(&svc->s_list, &ip_vs_svc_table[hash]);
318 } else {
319 /*
fc723250 320 * Hash it by fwmark in svc_fwm_table
1da177e4 321 */
fc723250 322 hash = ip_vs_svc_fwm_hashkey(svc->net, svc->fwmark);
1da177e4
LT		 323 list_add(&svc->f_list, &ip_vs_svc_fwm_table[hash]);
324 }
325
326 svc->flags |= IP_VS_SVC_F_HASHED;
327 /* increase its refcnt because it is referenced by the svc table */
328 atomic_inc(&svc->refcnt);
329 return 1;
330}
331
332
333/*
fc723250 334 * Unhashes a service from svc_table / svc_fwm_table.
1da177e4
LT		 335 * Should be called with locked tables.
336 */
337static int ip_vs_svc_unhash(struct ip_vs_service *svc)
338{
339 if (!(svc->flags & IP_VS_SVC_F_HASHED)) {
1e3e238e
HE		 340 pr_err("%s(): request for unhash flagged, called from %pF\n",
341 __func__, __builtin_return_address(0));
1da177e4
LT		 342 return 0;
343 }
344
345 if (svc->fwmark == 0) {
fc723250 346 /* Remove it from the svc_table table */
1da177e4
LT		 347 list_del(&svc->s_list);
348 } else {
fc723250 349 /* Remove it from the svc_fwm_table table */
1da177e4
LT		 350 list_del(&svc->f_list);
351 }
352
353 svc->flags &= ~IP_VS_SVC_F_HASHED;
354 atomic_dec(&svc->refcnt);
355 return 1;
356}
357
358
359/*
fc723250 360 * Get service by {netns, proto,addr,port} in the service table.
1da177e4 361 */
b18610de 362static inline struct ip_vs_service *
fc723250
HS		 363__ip_vs_service_find(struct net *net, int af, __u16 protocol,
364 const union nf_inet_addr *vaddr, __be16 vport)
1da177e4 365{
95c96174 366 unsigned int hash;
1da177e4
LT		 367 struct ip_vs_service *svc;
368
369 /* Check for "full" addressed entries */
fc723250 370 hash = ip_vs_svc_hashkey(net, af, protocol, vaddr, vport);
1da177e4
LT		 371
372 list_for_each_entry(svc, &ip_vs_svc_table[hash], s_list){
b18610de
JV		 373 if ((svc->af == af)
374 && ip_vs_addr_equal(af, &svc->addr, vaddr)
1da177e4 375 && (svc->port == vport)
fc723250
HS		 376 && (svc->protocol == protocol)
377 && net_eq(svc->net, net)) {
1da177e4 378 /* HIT */
1da177e4
LT		 379 return svc;
380 }
381 }
382
383 return NULL;
384}
385
386
387/*
388 * Get service by {fwmark} in the service table.
389 */
b18610de 390static inline struct ip_vs_service *
fc723250 391__ip_vs_svc_fwm_find(struct net *net, int af, __u32 fwmark)
1da177e4 392{
95c96174 393 unsigned int hash;
1da177e4
LT		 394 struct ip_vs_service *svc;
395
396 /* Check for fwmark addressed entries */
fc723250 397 hash = ip_vs_svc_fwm_hashkey(net, fwmark);
1da177e4
LT		 398
399 list_for_each_entry(svc, &ip_vs_svc_fwm_table[hash], f_list) {
fc723250
HS		 400 if (svc->fwmark == fwmark && svc->af == af
401 && net_eq(svc->net, net)) {
1da177e4 402 /* HIT */
1da177e4
LT		 403 return svc;
404 }
405 }
406
407 return NULL;
408}
409
410struct ip_vs_service *
fc723250 411ip_vs_service_get(struct net *net, int af, __u32 fwmark, __u16 protocol,
3c2e0505 412 const union nf_inet_addr *vaddr, __be16 vport)
1da177e4
LT		 413{
414 struct ip_vs_service *svc;
763f8d0e 415 struct netns_ipvs *ipvs = net_ipvs(net);
3c2e0505 416
1da177e4
LT		 417 read_lock(&__ip_vs_svc_lock);
418
419 /*
420 * Check the table hashed by fwmark first
421 */
097fc76a
JA		 422 if (fwmark) {
423 svc = __ip_vs_svc_fwm_find(net, af, fwmark);
424 if (svc)
425 goto out;
426 }
1da177e4
LT		 427
428 /*
429 * Check the table hashed by <protocol,addr,port>
430 * for "full" addressed entries
431 */
fc723250 432 svc = __ip_vs_service_find(net, af, protocol, vaddr, vport);
1da177e4
LT		 433
434 if (svc == NULL
435 && protocol == IPPROTO_TCP
763f8d0e 436 && atomic_read(&ipvs->ftpsvc_counter)
1da177e4
LT		 437 && (vport == FTPDATA || ntohs(vport) >= PROT_SOCK)) {
438 /*
439 * Check if ftp service entry exists, the packet
440 * might belong to FTP data connections.
441 */
fc723250 442 svc = __ip_vs_service_find(net, af, protocol, vaddr, FTPPORT);
1da177e4
LT		 443 }
444
445 if (svc == NULL
763f8d0e 446 && atomic_read(&ipvs->nullsvc_counter)) {
1da177e4
LT		 447 /*
448 * Check if the catch-all port (port zero) exists
449 */
fc723250 450 svc = __ip_vs_service_find(net, af, protocol, vaddr, 0);
1da177e4
LT		 451 }
452
453 out:
26c15cfd
JA		 454 if (svc)
455 atomic_inc(&svc->usecnt);
1da177e4
LT		 456 read_unlock(&__ip_vs_svc_lock);
457
3c2e0505
JV		 458 IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
459 fwmark, ip_vs_proto_name(protocol),
460 IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
461 svc ? "hit" : "not hit");
1da177e4
LT		 462
463 return svc;
464}
465
466
467static inline void
468__ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
469{
470 atomic_inc(&svc->refcnt);
471 dest->svc = svc;
472}
473
26c15cfd 474static void
1da177e4
LT		 475__ip_vs_unbind_svc(struct ip_vs_dest *dest)
476{
477 struct ip_vs_service *svc = dest->svc;
478
479 dest->svc = NULL;
26c15cfd
JA		 480 if (atomic_dec_and_test(&svc->refcnt)) {
481 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
482 svc->fwmark,
483 IP_VS_DBG_ADDR(svc->af, &svc->addr),
484 ntohs(svc->port), atomic_read(&svc->usecnt));
b17fc996 485 free_percpu(svc->stats.cpustats);
1da177e4 486 kfree(svc);
26c15cfd 487 }
1da177e4
LT		 488}
489
490
491/*
492 * Returns hash value for real service
493 */
95c96174 494static inline unsigned int ip_vs_rs_hashkey(int af,
7937df15
JV		 495 const union nf_inet_addr *addr,
496 __be16 port)
1da177e4 497{
95c96174 498 register unsigned int porth = ntohs(port);
7937df15
JV		 499 __be32 addr_fold = addr->ip;
500
501#ifdef CONFIG_IP_VS_IPV6
502 if (af == AF_INET6)
503 addr_fold = addr->ip6[0]^addr->ip6[1]^
504 addr->ip6[2]^addr->ip6[3];
505#endif
1da177e4 506
7937df15 507 return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
1da177e4
LT		 508 & IP_VS_RTAB_MASK;
509}
510
276472ea
JA		 511/* Hash ip_vs_dest in rs_table by <proto,addr,port>. */
512static void ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
1da177e4 513{
95c96174 514 unsigned int hash;
1da177e4 515
276472ea
JA		 516 if (dest->in_rs_table)
517 return;
1da177e4
LT		 518
519 /*
520 * Hash by proto,addr,port,
521 * which are the parameters of the real service.
522 */
7937df15
JV		 523 hash = ip_vs_rs_hashkey(dest->af, &dest->addr, dest->port);
524
276472ea
JA		 525 hlist_add_head_rcu(&dest->d_list, &ipvs->rs_table[hash]);
526 dest->in_rs_table = 1;
1da177e4
LT		 527}
528
276472ea
JA		 529/* Unhash ip_vs_dest from rs_table. */
530static void ip_vs_rs_unhash(struct ip_vs_dest *dest)
1da177e4
LT		 531{
532 /*
fc723250 533 * Remove it from the rs_table table.
1da177e4 534 */
276472ea
JA		 535 if (dest->in_rs_table) {
536 hlist_del_rcu(&dest->d_list);
537 dest->in_rs_table = 0;
1da177e4 538 }
1da177e4
LT		 539}
540
276472ea
JA		 541/* Check if real service by <proto,addr,port> is present */
542bool ip_vs_has_real_service(struct net *net, int af, __u16 protocol,
543 const union nf_inet_addr *daddr, __be16 dport)
1da177e4 544{
fc723250 545 struct netns_ipvs *ipvs = net_ipvs(net);
95c96174 546 unsigned int hash;
1da177e4
LT		 547 struct ip_vs_dest *dest;
548
276472ea 549 /* Check for "full" addressed entries */
7937df15 550 hash = ip_vs_rs_hashkey(af, daddr, dport);
1da177e4 551
276472ea
JA		 552 rcu_read_lock();
553 hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
554 if (dest->port == dport &&
555 dest->af == af &&
556 ip_vs_addr_equal(af, &dest->addr, daddr) &&
557 (dest->protocol == protocol || dest->vfwmark)) {
1da177e4 558 /* HIT */
276472ea
JA		 559 rcu_read_unlock();
560 return true;
1da177e4
LT		 561 }
562 }
276472ea 563 rcu_read_unlock();
1da177e4 564
276472ea 565 return false;
1da177e4
LT		 566}
567
413c2d04
JA		 568/* Lookup destination by {addr,port} in the given service
569 * Called under RCU lock.
1da177e4
LT		 570 */
571static struct ip_vs_dest *
7937df15
JV		 572ip_vs_lookup_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
573 __be16 dport)
1da177e4
LT		 574{
575 struct ip_vs_dest *dest;
576
577 /*
578 * Find the destination for the given service
579 */
413c2d04 580 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
7937df15
JV		 581 if ((dest->af == svc->af)
582 && ip_vs_addr_equal(svc->af, &dest->addr, daddr)
583 && (dest->port == dport)) {
1da177e4
LT		 584 /* HIT */
585 return dest;
586 }
587 }
588
589 return NULL;
590}
591
1e356f9c
RB		 592/*
593 * Find destination by {daddr,dport,vaddr,protocol}
413c2d04 594 * Created to be used in ip_vs_process_message() in
1e356f9c
RB		 595 * the backup synchronization daemon. It finds the
596 * destination to be bound to the received connection
597 * on the backup.
413c2d04 598 * Called under RCU lock, no refcnt is returned.
1e356f9c 599 */
fc723250
HS		 600struct ip_vs_dest *ip_vs_find_dest(struct net *net, int af,
601 const union nf_inet_addr *daddr,
7937df15
JV		 602 __be16 dport,
603 const union nf_inet_addr *vaddr,
52793dbe
JA		 604 __be16 vport, __u16 protocol, __u32 fwmark,
605 __u32 flags)
1e356f9c
RB		 606{
607 struct ip_vs_dest *dest;
608 struct ip_vs_service *svc;
52793dbe 609 __be16 port = dport;
1e356f9c 610
fc723250 611 svc = ip_vs_service_get(net, af, fwmark, protocol, vaddr, vport);
1e356f9c
RB		 612 if (!svc)
613 return NULL;
52793dbe
JA		 614 if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
615 port = 0;
616 dest = ip_vs_lookup_dest(svc, daddr, port);
617 if (!dest)
618 dest = ip_vs_lookup_dest(svc, daddr, port ^ dport);
1e356f9c
RB		 619 ip_vs_service_put(svc);
620 return dest;
621}
1da177e4 622
026ace06
JA		 623void ip_vs_dest_dst_rcu_free(struct rcu_head *head)
624{
625 struct ip_vs_dest_dst *dest_dst = container_of(head,
626 struct ip_vs_dest_dst,
627 rcu_head);
628
629 dst_release(dest_dst->dst_cache);
630 kfree(dest_dst);
631}
632
633/* Release dest_dst and dst_cache for dest in user context */
d1deae4d
JA		 634static void __ip_vs_dst_cache_reset(struct ip_vs_dest *dest)
635{
026ace06 636 struct ip_vs_dest_dst *old;
d1deae4d 637
026ace06
JA		 638 old = rcu_dereference_protected(dest->dest_dst, 1);
639 if (old) {
640 RCU_INIT_POINTER(dest->dest_dst, NULL);
641 call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
642 }
d1deae4d
JA		 643}
644
1da177e4
LT		 645/*
646 * Lookup dest by {svc,addr,port} in the destination trash.
647 * The destination trash is used to hold the destinations that are removed
648 * from the service table but are still referenced by some conn entries.
649 * The reason to add the destination trash is when the dest is temporary
650 * down (either by administrator or by monitor program), the dest can be
651 * picked back from the trash, the remaining connections to the dest can
652 * continue, and the counting information of the dest is also useful for
653 * scheduling.
654 */
655static struct ip_vs_dest *
7937df15
JV		 656ip_vs_trash_get_dest(struct ip_vs_service *svc, const union nf_inet_addr *daddr,
657 __be16 dport)
1da177e4 658{
578bc3ef 659 struct ip_vs_dest *dest;
f2431e6e 660 struct netns_ipvs *ipvs = net_ipvs(svc->net);
1da177e4
LT		 661
662 /*
663 * Find the destination in trash
664 */
578bc3ef
JA		 665 spin_lock_bh(&ipvs->dest_trash_lock);
666 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
7937df15
JV		 667 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
668 "dest->refcnt=%d\n",
669 dest->vfwmark,
670 IP_VS_DBG_ADDR(svc->af, &dest->addr),
671 ntohs(dest->port),
672 atomic_read(&dest->refcnt));
578bc3ef
JA		 673 /* We can not reuse dest while in grace period
674 * because conns still can use dest->svc
675 */
676 if (test_bit(IP_VS_DEST_STATE_REMOVING, &dest->state))
677 continue;
7937df15
JV		 678 if (dest->af == svc->af &&
679 ip_vs_addr_equal(svc->af, &dest->addr, daddr) &&
1da177e4
LT		 680 dest->port == dport &&
681 dest->vfwmark == svc->fwmark &&
682 dest->protocol == svc->protocol &&
683 (svc->fwmark ||
7937df15 684 (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
1da177e4
LT		 685 dest->vport == svc->port))) {
686 /* HIT */
578bc3ef
JA		 687 list_del(&dest->t_list);
688 ip_vs_dest_hold(dest);
689 goto out;
1da177e4
LT		 690 }
691 }
692
578bc3ef
JA		 693 dest = NULL;
694
695out:
696 spin_unlock_bh(&ipvs->dest_trash_lock);
697
698 return dest;
1da177e4
LT		 699}
700
578bc3ef
JA		 701static void ip_vs_dest_free(struct ip_vs_dest *dest)
702{
703 __ip_vs_dst_cache_reset(dest);
704 __ip_vs_unbind_svc(dest);
705 free_percpu(dest->stats.cpustats);
706 kfree(dest);
707}
1da177e4
LT		 708
709/*
710 * Clean up all the destinations in the trash
711 * Called by the ip_vs_control_cleanup()
712 *
713 * When the ip_vs_control_clearup is activated by ipvs module exit,
714 * the service tables must have been flushed and all the connections
715 * are expired, and the refcnt of each destination in the trash must
578bc3ef 716 * be 0, so we simply release them here.
1da177e4 717 */
f2431e6e 718static void ip_vs_trash_cleanup(struct net *net)
1da177e4
LT		 719{
720 struct ip_vs_dest *dest, *nxt;
f2431e6e 721 struct netns_ipvs *ipvs = net_ipvs(net);
1da177e4 722
578bc3ef
JA		 723 del_timer_sync(&ipvs->dest_trash_timer);
724 /* No need to use dest_trash_lock */
725 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, t_list) {
726 list_del(&dest->t_list);
727 ip_vs_dest_free(dest);
1da177e4
LT		 728 }
729}
730
55a3d4e1
JA		 731static void
732ip_vs_copy_stats(struct ip_vs_stats_user *dst, struct ip_vs_stats *src)
733{
734#define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->ustats.c - src->ustats0.c
55a3d4e1
JA		 735
736 spin_lock_bh(&src->lock);
737
738 IP_VS_SHOW_STATS_COUNTER(conns);
739 IP_VS_SHOW_STATS_COUNTER(inpkts);
740 IP_VS_SHOW_STATS_COUNTER(outpkts);
741 IP_VS_SHOW_STATS_COUNTER(inbytes);
742 IP_VS_SHOW_STATS_COUNTER(outbytes);
743
ea9f22cc 744 ip_vs_read_estimator(dst, src);
55a3d4e1
JA		 745
746 spin_unlock_bh(&src->lock);
747}
1da177e4
LT		 748
749static void
750ip_vs_zero_stats(struct ip_vs_stats *stats)
751{
752 spin_lock_bh(&stats->lock);
e93615d0 753
55a3d4e1
JA		 754 /* get current counters as zero point, rates are zeroed */
755
756#define IP_VS_ZERO_STATS_COUNTER(c) stats->ustats0.c = stats->ustats.c
55a3d4e1
JA		 757
758 IP_VS_ZERO_STATS_COUNTER(conns);
759 IP_VS_ZERO_STATS_COUNTER(inpkts);
760 IP_VS_ZERO_STATS_COUNTER(outpkts);
761 IP_VS_ZERO_STATS_COUNTER(inbytes);
762 IP_VS_ZERO_STATS_COUNTER(outbytes);
763
1da177e4 764 ip_vs_zero_estimator(stats);
e93615d0 765
3a14a313 766 spin_unlock_bh(&stats->lock);
1da177e4
LT		 767}
768
769/*
770 * Update a destination in the given service
771 */
772static void
26c15cfd
JA		 773__ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
774 struct ip_vs_dest_user_kern *udest, int add)
1da177e4 775{
fc723250 776 struct netns_ipvs *ipvs = net_ipvs(svc->net);
1da177e4
LT		 777 int conn_flags;
778
779 /* set the weight and the flags */
780 atomic_set(&dest->weight, udest->weight);
3575792e
JA		 781 conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
782 conn_flags |= IP_VS_CONN_F_INACTIVE;
1da177e4 783
1da177e4 784 /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
3575792e 785 if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
1da177e4
LT		 786 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
787 } else {
788 /*
fc723250 789 * Put the real service in rs_table if not present.
1da177e4
LT		 790 * For now only for NAT!
791 */
fc723250 792 ip_vs_rs_hash(ipvs, dest);
1da177e4
LT		 793 }
794 atomic_set(&dest->conn_flags, conn_flags);
795
796 /* bind the service */
797 if (!dest->svc) {
798 __ip_vs_bind_svc(dest, svc);
799 } else {
800 if (dest->svc != svc) {
801 __ip_vs_unbind_svc(dest);
802 ip_vs_zero_stats(&dest->stats);
803 __ip_vs_bind_svc(dest, svc);
804 }
805 }
806
807 /* set the dest status flags */
808 dest->flags |= IP_VS_DEST_F_AVAILABLE;
809
810 if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
811 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
812 dest->u_threshold = udest->u_threshold;
813 dest->l_threshold = udest->l_threshold;
26c15cfd 814
ff75f40f 815 spin_lock_bh(&dest->dst_lock);
d1deae4d 816 __ip_vs_dst_cache_reset(dest);
ff75f40f 817 spin_unlock_bh(&dest->dst_lock);
fc604767 818
26c15cfd 819 if (add)
6ef757f9 820 ip_vs_start_estimator(svc->net, &dest->stats);
26c15cfd
JA		 821
822 write_lock_bh(&__ip_vs_svc_lock);
823
824 /* Wait until all other svc users go away */
825 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
826
827 if (add) {
413c2d04 828 list_add_rcu(&dest->n_list, &svc->destinations);
26c15cfd 829 svc->num_dests++;
6b6df466
JA		 830 if (svc->scheduler->add_dest)
831 svc->scheduler->add_dest(svc, dest);
832 } else {
833 if (svc->scheduler->upd_dest)
834 svc->scheduler->upd_dest(svc, dest);
26c15cfd
JA		 835 }
836
837 /* call the update_service, because server weight may be changed */
838 if (svc->scheduler->update_service)
839 svc->scheduler->update_service(svc);
840
841 write_unlock_bh(&__ip_vs_svc_lock);
1da177e4
LT		 842}
843
844
845/*
846 * Create a destination for the given service
847 */
848static int
c860c6b1 849ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
1da177e4
LT		 850 struct ip_vs_dest **dest_p)
851{
852 struct ip_vs_dest *dest;
95c96174 853 unsigned int atype;
1da177e4
LT		 854
855 EnterFunction(2);
856
09571c7a
VB		 857#ifdef CONFIG_IP_VS_IPV6
858 if (svc->af == AF_INET6) {
859 atype = ipv6_addr_type(&udest->addr.in6);
3bfb92f4
SW		 860 if ((!(atype & IPV6_ADDR_UNICAST) ||
861 atype & IPV6_ADDR_LINKLOCAL) &&
4a98480b 862 !__ip_vs_addr_is_local_v6(svc->net, &udest->addr.in6))
09571c7a
VB		 863 return -EINVAL;
864 } else
865#endif
866 {
4a98480b 867 atype = inet_addr_type(svc->net, udest->addr.ip);
09571c7a
VB		 868 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
869 return -EINVAL;
870 }
1da177e4 871
dee06e47 872 dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
0a9ee813 873 if (dest == NULL)
1da177e4 874 return -ENOMEM;
0a9ee813 875
b17fc996 876 dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
0a9ee813 877 if (!dest->stats.cpustats)
b17fc996 878 goto err_alloc;
1da177e4 879
c860c6b1 880 dest->af = svc->af;
1da177e4 881 dest->protocol = svc->protocol;
c860c6b1 882 dest->vaddr = svc->addr;
1da177e4
LT		 883 dest->vport = svc->port;
884 dest->vfwmark = svc->fwmark;
c860c6b1 885 ip_vs_addr_copy(svc->af, &dest->addr, &udest->addr);
1da177e4
LT		 886 dest->port = udest->port;
887
888 atomic_set(&dest->activeconns, 0);
889 atomic_set(&dest->inactconns, 0);
890 atomic_set(&dest->persistconns, 0);
26c15cfd 891 atomic_set(&dest->refcnt, 1);
1da177e4 892
276472ea 893 INIT_HLIST_NODE(&dest->d_list);
1da177e4
LT		 894 spin_lock_init(&dest->dst_lock);
895 spin_lock_init(&dest->stats.lock);
26c15cfd 896 __ip_vs_update_dest(svc, dest, udest, 1);
1da177e4
LT		 897
898 *dest_p = dest;
899
900 LeaveFunction(2);
901 return 0;
b17fc996
HS		 902
903err_alloc:
904 kfree(dest);
905 return -ENOMEM;
1da177e4
LT		 906}
907
908
909/*
910 * Add a destination into an existing service
911 */
912static int
c860c6b1 913ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1da177e4
LT		 914{
915 struct ip_vs_dest *dest;
c860c6b1 916 union nf_inet_addr daddr;
014d730d 917 __be16 dport = udest->port;
1da177e4
LT		 918 int ret;
919
920 EnterFunction(2);
921
922 if (udest->weight < 0) {
1e3e238e 923 pr_err("%s(): server weight less than zero\n", __func__);
1da177e4
LT		 924 return -ERANGE;
925 }
926
927 if (udest->l_threshold > udest->u_threshold) {
1e3e238e
HE		 928 pr_err("%s(): lower threshold is higher than upper threshold\n",
929 __func__);
1da177e4
LT		 930 return -ERANGE;
931 }
932
c860c6b1
JV		 933 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
934
413c2d04
JA		 935 /* We use function that requires RCU lock */
936 rcu_read_lock();
7937df15 937 dest = ip_vs_lookup_dest(svc, &daddr, dport);
413c2d04 938 rcu_read_unlock();
7937df15 939
1da177e4 940 if (dest != NULL) {
1e3e238e 941 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
1da177e4
LT		 942 return -EEXIST;
943 }
944
945 /*
946 * Check if the dest already exists in the trash and
947 * is from the same service
948 */
7937df15
JV		 949 dest = ip_vs_trash_get_dest(svc, &daddr, dport);
950
1da177e4 951 if (dest != NULL) {
cfc78c5a
JV		 952 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
953 "dest->refcnt=%d, service %u/%s:%u\n",
954 IP_VS_DBG_ADDR(svc->af, &daddr), ntohs(dport),
955 atomic_read(&dest->refcnt),
956 dest->vfwmark,
957 IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
958 ntohs(dest->vport));
959
26c15cfd
JA		 960 __ip_vs_update_dest(svc, dest, udest, 1);
961 ret = 0;
962 } else {
1da177e4 963 /*
26c15cfd 964 * Allocate and initialize the dest structure
1da177e4 965 */
26c15cfd 966 ret = ip_vs_new_dest(svc, udest, &dest);
1da177e4 967 }
1da177e4
LT		 968 LeaveFunction(2);
969
26c15cfd 970 return ret;
1da177e4
LT		 971}
972
973
974/*
975 * Edit a destination in the given service
976 */
977static int
c860c6b1 978ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1da177e4
LT		 979{
980 struct ip_vs_dest *dest;
c860c6b1 981 union nf_inet_addr daddr;
014d730d 982 __be16 dport = udest->port;
1da177e4
LT		 983
984 EnterFunction(2);
985
986 if (udest->weight < 0) {
1e3e238e 987 pr_err("%s(): server weight less than zero\n", __func__);
1da177e4
LT		 988 return -ERANGE;
989 }
990
991 if (udest->l_threshold > udest->u_threshold) {
1e3e238e
HE		 992 pr_err("%s(): lower threshold is higher than upper threshold\n",
993 __func__);
1da177e4
LT		 994 return -ERANGE;
995 }
996
c860c6b1
JV		 997 ip_vs_addr_copy(svc->af, &daddr, &udest->addr);
998
413c2d04
JA		 999 /* We use function that requires RCU lock */
1000 rcu_read_lock();
7937df15 1001 dest = ip_vs_lookup_dest(svc, &daddr, dport);
413c2d04 1002 rcu_read_unlock();
7937df15 1003
1da177e4 1004 if (dest == NULL) {
1e3e238e 1005 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1da177e4
LT		 1006 return -ENOENT;
1007 }
1008
26c15cfd 1009 __ip_vs_update_dest(svc, dest, udest, 0);
1da177e4
LT		 1010 LeaveFunction(2);
1011
1012 return 0;
1013}
1014
578bc3ef
JA		 1015static void ip_vs_dest_wait_readers(struct rcu_head *head)
1016{
1017 struct ip_vs_dest *dest = container_of(head, struct ip_vs_dest,
1018 rcu_head);
1019
1020 /* End of grace period after unlinking */
1021 clear_bit(IP_VS_DEST_STATE_REMOVING, &dest->state);
1022}
1023
1da177e4
LT		 1024
1025/*
1026 * Delete a destination (must be already unlinked from the service)
1027 */
578bc3ef
JA		 1028static void __ip_vs_del_dest(struct net *net, struct ip_vs_dest *dest,
1029 bool cleanup)
1da177e4 1030{
a0840e2e
HS		 1031 struct netns_ipvs *ipvs = net_ipvs(net);
1032
6ef757f9 1033 ip_vs_stop_estimator(net, &dest->stats);
1da177e4
LT		 1034
1035 /*
1036 * Remove it from the d-linked list with the real services.
1037 */
1da177e4 1038 ip_vs_rs_unhash(dest);
1da177e4 1039
578bc3ef
JA		 1040 if (!cleanup) {
1041 set_bit(IP_VS_DEST_STATE_REMOVING, &dest->state);
1042 call_rcu(&dest->rcu_head, ip_vs_dest_wait_readers);
1da177e4 1043 }
578bc3ef
JA		 1044
1045 spin_lock_bh(&ipvs->dest_trash_lock);
1046 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1047 IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port),
1048 atomic_read(&dest->refcnt));
1049 if (list_empty(&ipvs->dest_trash) && !cleanup)
1050 mod_timer(&ipvs->dest_trash_timer,
1051 jiffies + IP_VS_DEST_TRASH_PERIOD);
1052 /* dest lives in trash without reference */
1053 list_add(&dest->t_list, &ipvs->dest_trash);
1054 spin_unlock_bh(&ipvs->dest_trash_lock);
1055 ip_vs_dest_put(dest);
1da177e4
LT		 1056}
1057
1058
1059/*
1060 * Unlink a destination from the given service
1061 */
1062static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1063 struct ip_vs_dest *dest,
1064 int svcupd)
1065{
1066 dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1067
1068 /*
1069 * Remove it from the d-linked destination list.
1070 */
413c2d04 1071 list_del_rcu(&dest->n_list);
1da177e4 1072 svc->num_dests--;
82dfb6f3 1073
6b6df466
JA		 1074 if (svcupd && svc->scheduler->del_dest)
1075 svc->scheduler->del_dest(svc, dest);
1076
82dfb6f3
SW		 1077 /*
1078 * Call the update_service function of its scheduler
1079 */
1080 if (svcupd && svc->scheduler->update_service)
1081 svc->scheduler->update_service(svc);
1da177e4
LT		 1082}
1083
1084
1085/*
1086 * Delete a destination server in the given service
1087 */
1088static int
c860c6b1 1089ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1da177e4
LT		 1090{
1091 struct ip_vs_dest *dest;
014d730d 1092 __be16 dport = udest->port;
1da177e4
LT		 1093
1094 EnterFunction(2);
1095
413c2d04
JA		 1096 /* We use function that requires RCU lock */
1097 rcu_read_lock();
7937df15 1098 dest = ip_vs_lookup_dest(svc, &udest->addr, dport);
413c2d04 1099 rcu_read_unlock();
c860c6b1 1100
1da177e4 1101 if (dest == NULL) {
1e3e238e 1102 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1da177e4
LT		 1103 return -ENOENT;
1104 }
1105
1106 write_lock_bh(&__ip_vs_svc_lock);
1107
1108 /*
1109 * Wait until all other svc users go away.
1110 */
26c15cfd 1111 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1da177e4
LT		 1112
1113 /*
1114 * Unlink dest from the service
1115 */
1116 __ip_vs_unlink_dest(svc, dest, 1);
1117
1118 write_unlock_bh(&__ip_vs_svc_lock);
1119
1120 /*
1121 * Delete the destination
1122 */
578bc3ef 1123 __ip_vs_del_dest(svc->net, dest, false);
1da177e4
LT		 1124
1125 LeaveFunction(2);
1126
1127 return 0;
1128}
1129
578bc3ef
JA		 1130static void ip_vs_dest_trash_expire(unsigned long data)
1131{
1132 struct net *net = (struct net *) data;
1133 struct netns_ipvs *ipvs = net_ipvs(net);
1134 struct ip_vs_dest *dest, *next;
1135
1136 spin_lock(&ipvs->dest_trash_lock);
1137 list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) {
1138 /* Skip if dest is in grace period */
1139 if (test_bit(IP_VS_DEST_STATE_REMOVING, &dest->state))
1140 continue;
1141 if (atomic_read(&dest->refcnt) > 0)
1142 continue;
1143 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1144 dest->vfwmark,
1145 IP_VS_DBG_ADDR(dest->svc->af, &dest->addr),
1146 ntohs(dest->port));
1147 list_del(&dest->t_list);
1148 ip_vs_dest_free(dest);
1149 }
1150 if (!list_empty(&ipvs->dest_trash))
1151 mod_timer(&ipvs->dest_trash_timer,
1152 jiffies + IP_VS_DEST_TRASH_PERIOD);
1153 spin_unlock(&ipvs->dest_trash_lock);
1154}
1da177e4
LT		 1155
1156/*
1157 * Add a service into the service hash table
1158 */
1159static int
fc723250 1160ip_vs_add_service(struct net *net, struct ip_vs_service_user_kern *u,
c860c6b1 1161 struct ip_vs_service **svc_p)
1da177e4
LT		 1162{
1163 int ret = 0;
1164 struct ip_vs_scheduler *sched = NULL;
0d1e71b0 1165 struct ip_vs_pe *pe = NULL;
1da177e4 1166 struct ip_vs_service *svc = NULL;
a0840e2e 1167 struct netns_ipvs *ipvs = net_ipvs(net);
1da177e4
LT		 1168
1169 /* increase the module use count */
1170 ip_vs_use_count_inc();
1171
1172 /* Lookup the scheduler by 'u->sched_name' */
1173 sched = ip_vs_scheduler_get(u->sched_name);
1174 if (sched == NULL) {
1e3e238e 1175 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1da177e4 1176 ret = -ENOENT;
6e08bfb8 1177 goto out_err;
1da177e4
LT		 1178 }
1179
0d1e71b0 1180 if (u->pe_name && *u->pe_name) {
e9e5eee8 1181 pe = ip_vs_pe_getbyname(u->pe_name);
0d1e71b0
SH		 1182 if (pe == NULL) {
1183 pr_info("persistence engine module ip_vs_pe_%s "
1184 "not found\n", u->pe_name);
1185 ret = -ENOENT;
1186 goto out_err;
1187 }
1188 }
1189
f94fd041 1190#ifdef CONFIG_IP_VS_IPV6
48148938
JV		 1191 if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1192 ret = -EINVAL;
1193 goto out_err;
f94fd041
JV		 1194 }
1195#endif
1196
dee06e47 1197 svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1da177e4 1198 if (svc == NULL) {
1e3e238e 1199 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1da177e4
LT		 1200 ret = -ENOMEM;
1201 goto out_err;
1202 }
b17fc996 1203 svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
0a54e939
JL		 1204 if (!svc->stats.cpustats) {
1205 ret = -ENOMEM;
b17fc996 1206 goto out_err;
0a54e939 1207 }
1da177e4
LT		 1208
1209 /* I'm the first user of the service */
26c15cfd 1210 atomic_set(&svc->usecnt, 0);
1da177e4
LT		 1211 atomic_set(&svc->refcnt, 0);
1212
c860c6b1 1213 svc->af = u->af;
1da177e4 1214 svc->protocol = u->protocol;
c860c6b1 1215 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1da177e4
LT		 1216 svc->port = u->port;
1217 svc->fwmark = u->fwmark;
1218 svc->flags = u->flags;
1219 svc->timeout = u->timeout * HZ;
1220 svc->netmask = u->netmask;
fc723250 1221 svc->net = net;
1da177e4
LT		 1222
1223 INIT_LIST_HEAD(&svc->destinations);
ba3a3ce1 1224 spin_lock_init(&svc->sched_lock);
1da177e4
LT		 1225 spin_lock_init(&svc->stats.lock);
1226
1227 /* Bind the scheduler */
1228 ret = ip_vs_bind_scheduler(svc, sched);
1229 if (ret)
1230 goto out_err;
1231 sched = NULL;
1232
0d1e71b0
SH		 1233 /* Bind the ct retriever */
1234 ip_vs_bind_pe(svc, pe);
1235 pe = NULL;
1236
1da177e4
LT		 1237 /* Update the virtual service counters */
1238 if (svc->port == FTPPORT)
763f8d0e 1239 atomic_inc(&ipvs->ftpsvc_counter);
1da177e4 1240 else if (svc->port == 0)
763f8d0e 1241 atomic_inc(&ipvs->nullsvc_counter);
1da177e4 1242
6ef757f9 1243 ip_vs_start_estimator(net, &svc->stats);
f94fd041
JV		 1244
1245 /* Count only IPv4 services for old get/setsockopt interface */
1246 if (svc->af == AF_INET)
a0840e2e 1247 ipvs->num_services++;
1da177e4
LT		 1248
1249 /* Hash the service into the service table */
1250 write_lock_bh(&__ip_vs_svc_lock);
1251 ip_vs_svc_hash(svc);
1252 write_unlock_bh(&__ip_vs_svc_lock);
1253
1254 *svc_p = svc;
7a4f0761
HS		 1255 /* Now there is a service - full throttle */
1256 ipvs->enable = 1;
1da177e4
LT		 1257 return 0;
1258
b17fc996 1259
6e08bfb8 1260 out_err:
1da177e4 1261 if (svc != NULL) {
2fabf35b 1262 ip_vs_unbind_scheduler(svc);
1da177e4
LT		 1263 if (svc->inc) {
1264 local_bh_disable();
1265 ip_vs_app_inc_put(svc->inc);
1266 local_bh_enable();
1267 }
b17fc996
HS		 1268 if (svc->stats.cpustats)
1269 free_percpu(svc->stats.cpustats);
1da177e4
LT		 1270 kfree(svc);
1271 }
1272 ip_vs_scheduler_put(sched);
0d1e71b0 1273 ip_vs_pe_put(pe);
1da177e4 1274
1da177e4
LT		 1275 /* decrease the module use count */
1276 ip_vs_use_count_dec();
1277
1278 return ret;
1279}
1280
1281
1282/*
1283 * Edit a service and bind it with a new scheduler
1284 */
1285static int
c860c6b1 1286ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1da177e4
LT		 1287{
1288 struct ip_vs_scheduler *sched, *old_sched;
0d1e71b0 1289 struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1da177e4
LT		 1290 int ret = 0;
1291
1292 /*
1293 * Lookup the scheduler, by 'u->sched_name'
1294 */
1295 sched = ip_vs_scheduler_get(u->sched_name);
1296 if (sched == NULL) {
1e3e238e 1297 pr_info("Scheduler module ip_vs_%s not found\n", u->sched_name);
1da177e4
LT		 1298 return -ENOENT;
1299 }
1300 old_sched = sched;
1301
0d1e71b0 1302 if (u->pe_name && *u->pe_name) {
e9e5eee8 1303 pe = ip_vs_pe_getbyname(u->pe_name);
0d1e71b0
SH		 1304 if (pe == NULL) {
1305 pr_info("persistence engine module ip_vs_pe_%s "
1306 "not found\n", u->pe_name);
1307 ret = -ENOENT;
1308 goto out;
1309 }
1310 old_pe = pe;
1311 }
1312
f94fd041 1313#ifdef CONFIG_IP_VS_IPV6
48148938
JV		 1314 if (u->af == AF_INET6 && (u->netmask < 1 || u->netmask > 128)) {
1315 ret = -EINVAL;
1316 goto out;
f94fd041
JV		 1317 }
1318#endif
1319
1da177e4
LT		 1320 write_lock_bh(&__ip_vs_svc_lock);
1321
1322 /*
1323 * Wait until all other svc users go away.
1324 */
26c15cfd 1325 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1da177e4
LT		 1326
1327 /*
1328 * Set the flags and timeout value
1329 */
1330 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1331 svc->timeout = u->timeout * HZ;
1332 svc->netmask = u->netmask;
1333
1334 old_sched = svc->scheduler;
1335 if (sched != old_sched) {
1336 /*
1337 * Unbind the old scheduler
1338 */
ed3ffc4e 1339 ip_vs_unbind_scheduler(svc);
1da177e4
LT		 1340
1341 /*
1342 * Bind the new scheduler
1343 */
1344 if ((ret = ip_vs_bind_scheduler(svc, sched))) {
1345 /*
1346 * If ip_vs_bind_scheduler fails, restore the old
1347 * scheduler.
1348 * The main reason of failure is out of memory.
1349 *
1350 * The question is if the old scheduler can be
1351 * restored all the time. TODO: if it cannot be
1352 * restored some time, we must delete the service,
1353 * otherwise the system may crash.
1354 */
1355 ip_vs_bind_scheduler(svc, old_sched);
1356 old_sched = sched;
9e691ed6 1357 goto out_unlock;
1da177e4
LT		 1358 }
1359 }
1360
0d1e71b0
SH		 1361 old_pe = svc->pe;
1362 if (pe != old_pe) {
1363 ip_vs_unbind_pe(svc);
1364 ip_vs_bind_pe(svc, pe);
1365 }
1366
552ad65a 1367out_unlock:
1da177e4 1368 write_unlock_bh(&__ip_vs_svc_lock);
552ad65a 1369out:
6e08bfb8 1370 ip_vs_scheduler_put(old_sched);
0d1e71b0 1371 ip_vs_pe_put(old_pe);
1da177e4
LT		 1372 return ret;
1373}
1374
1375
1376/*
1377 * Delete a service from the service list
1378 * - The service must be unlinked, unlocked and not referenced!
1379 * - We are called under _bh lock
1380 */
578bc3ef 1381static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1da177e4
LT		 1382{
1383 struct ip_vs_dest *dest, *nxt;
1384 struct ip_vs_scheduler *old_sched;
0d1e71b0 1385 struct ip_vs_pe *old_pe;
a0840e2e 1386 struct netns_ipvs *ipvs = net_ipvs(svc->net);
0d1e71b0
SH		 1387
1388 pr_info("%s: enter\n", __func__);
1da177e4 1389
f94fd041
JV		 1390 /* Count only IPv4 services for old get/setsockopt interface */
1391 if (svc->af == AF_INET)
a0840e2e 1392 ipvs->num_services--;
f94fd041 1393
6ef757f9 1394 ip_vs_stop_estimator(svc->net, &svc->stats);
1da177e4
LT		 1395
1396 /* Unbind scheduler */
1397 old_sched = svc->scheduler;
1398 ip_vs_unbind_scheduler(svc);
6e08bfb8 1399 ip_vs_scheduler_put(old_sched);
1da177e4 1400
0d1e71b0
SH		 1401 /* Unbind persistence engine */
1402 old_pe = svc->pe;
1403 ip_vs_unbind_pe(svc);
1404 ip_vs_pe_put(old_pe);
1405
1da177e4
LT		 1406 /* Unbind app inc */
1407 if (svc->inc) {
1408 ip_vs_app_inc_put(svc->inc);
1409 svc->inc = NULL;
1410 }
1411
1412 /*
1413 * Unlink the whole destination list
1414 */
1415 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1416 __ip_vs_unlink_dest(svc, dest, 0);
578bc3ef 1417 __ip_vs_del_dest(svc->net, dest, cleanup);
1da177e4
LT		 1418 }
1419
1420 /*
1421 * Update the virtual service counters
1422 */
1423 if (svc->port == FTPPORT)
763f8d0e 1424 atomic_dec(&ipvs->ftpsvc_counter);
1da177e4 1425 else if (svc->port == 0)
763f8d0e 1426 atomic_dec(&ipvs->nullsvc_counter);
1da177e4
LT		 1427
1428 /*
1429 * Free the service if nobody refers to it
1430 */
26c15cfd
JA		 1431 if (atomic_read(&svc->refcnt) == 0) {
1432 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u usecnt=%d\n",
1433 svc->fwmark,
1434 IP_VS_DBG_ADDR(svc->af, &svc->addr),
1435 ntohs(svc->port), atomic_read(&svc->usecnt));
b17fc996 1436 free_percpu(svc->stats.cpustats);
1da177e4 1437 kfree(svc);
26c15cfd 1438 }
1da177e4
LT		 1439
1440 /* decrease the module use count */
1441 ip_vs_use_count_dec();
1442}
1443
1444/*
26c15cfd 1445 * Unlink a service from list and try to delete it if its refcnt reached 0
1da177e4 1446 */
578bc3ef 1447static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1da177e4 1448{
1da177e4
LT		 1449 /*
1450 * Unhash it from the service table
1451 */
1452 write_lock_bh(&__ip_vs_svc_lock);
1453
1454 ip_vs_svc_unhash(svc);
1455
1456 /*
1457 * Wait until all the svc users go away.
1458 */
26c15cfd 1459 IP_VS_WAIT_WHILE(atomic_read(&svc->usecnt) > 0);
1da177e4 1460
578bc3ef 1461 __ip_vs_del_service(svc, cleanup);
1da177e4
LT		 1462
1463 write_unlock_bh(&__ip_vs_svc_lock);
26c15cfd
JA		 1464}
1465
1466/*
1467 * Delete a service from the service list
1468 */
1469static int ip_vs_del_service(struct ip_vs_service *svc)
1470{
1471 if (svc == NULL)
1472 return -EEXIST;
578bc3ef 1473 ip_vs_unlink_service(svc, false);
1da177e4
LT		 1474
1475 return 0;
1476}
1477
1478
1479/*
1480 * Flush all the virtual services
1481 */
578bc3ef 1482static int ip_vs_flush(struct net *net, bool cleanup)
1da177e4
LT		 1483{
1484 int idx;
1485 struct ip_vs_service *svc, *nxt;
1486
1487 /*
fc723250 1488 * Flush the service table hashed by <netns,protocol,addr,port>
1da177e4
LT		 1489 */
1490 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
fc723250
HS		 1491 list_for_each_entry_safe(svc, nxt, &ip_vs_svc_table[idx],
1492 s_list) {
1493 if (net_eq(svc->net, net))
578bc3ef 1494 ip_vs_unlink_service(svc, cleanup);
1da177e4
LT		 1495 }
1496 }
1497
1498 /*
1499 * Flush the service table hashed by fwmark
1500 */
1501 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1502 list_for_each_entry_safe(svc, nxt,
1503 &ip_vs_svc_fwm_table[idx], f_list) {
fc723250 1504 if (net_eq(svc->net, net))
578bc3ef 1505 ip_vs_unlink_service(svc, cleanup);
1da177e4
LT		 1506 }
1507 }
1508
1509 return 0;
1510}
1511
7a4f0761
HS		 1512/*
1513 * Delete service by {netns} in the service table.
1514 * Called by __ip_vs_cleanup()
1515 */
503cf15a 1516void ip_vs_service_net_cleanup(struct net *net)
7a4f0761
HS		 1517{
1518 EnterFunction(2);
1519 /* Check for "full" addressed entries */
1520 mutex_lock(&__ip_vs_mutex);
578bc3ef 1521 ip_vs_flush(net, true);
7a4f0761
HS		 1522 mutex_unlock(&__ip_vs_mutex);
1523 LeaveFunction(2);
1524}
d1deae4d
JA		 1525
1526/* Put all references for device (dst_cache) */
7a4f0761 1527static inline void
d1deae4d 1528ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
7a4f0761
HS		 1529{
1530 spin_lock_bh(&dest->dst_lock);
026ace06 1531 if (dest->dest_dst && dest->dest_dst->dst_cache->dev == dev) {
7a4f0761
HS		 1532 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1533 dev->name,
1534 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1535 ntohs(dest->port),
1536 atomic_read(&dest->refcnt));
d1deae4d 1537 __ip_vs_dst_cache_reset(dest);
7a4f0761
HS		 1538 }
1539 spin_unlock_bh(&dest->dst_lock);
1540
1541}
313eae63
JA		 1542/* Netdev event receiver
1543 * Currently only NETDEV_DOWN is handled to release refs to cached dsts
7a4f0761
HS		 1544 */
1545static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1546 void *ptr)
1547{
1548 struct net_device *dev = ptr;
1549 struct net *net = dev_net(dev);
283283c4 1550 struct netns_ipvs *ipvs = net_ipvs(net);
7a4f0761
HS		 1551 struct ip_vs_service *svc;
1552 struct ip_vs_dest *dest;
1553 unsigned int idx;
1554
313eae63 1555 if (event != NETDEV_DOWN || !ipvs)
7a4f0761
HS		 1556 return NOTIFY_DONE;
1557 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1558 EnterFunction(2);
1559 mutex_lock(&__ip_vs_mutex);
1560 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1561 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1562 if (net_eq(svc->net, net)) {
1563 list_for_each_entry(dest, &svc->destinations,
1564 n_list) {
d1deae4d 1565 ip_vs_forget_dev(dest, dev);
7a4f0761
HS		 1566 }
1567 }
1568 }
1569
1570 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1571 if (net_eq(svc->net, net)) {
1572 list_for_each_entry(dest, &svc->destinations,
1573 n_list) {
d1deae4d 1574 ip_vs_forget_dev(dest, dev);
7a4f0761
HS		 1575 }
1576 }
1577
1578 }
1579 }
1580
578bc3ef
JA		 1581 spin_lock_bh(&ipvs->dest_trash_lock);
1582 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
d1deae4d 1583 ip_vs_forget_dev(dest, dev);
7a4f0761 1584 }
578bc3ef 1585 spin_unlock_bh(&ipvs->dest_trash_lock);
7a4f0761
HS		 1586 mutex_unlock(&__ip_vs_mutex);
1587 LeaveFunction(2);
1588 return NOTIFY_DONE;
1589}
1da177e4
LT		 1590
1591/*
1592 * Zero counters in a service or all services
1593 */
1594static int ip_vs_zero_service(struct ip_vs_service *svc)
1595{
1596 struct ip_vs_dest *dest;
1597
1598 write_lock_bh(&__ip_vs_svc_lock);
1599 list_for_each_entry(dest, &svc->destinations, n_list) {
1600 ip_vs_zero_stats(&dest->stats);
1601 }
1602 ip_vs_zero_stats(&svc->stats);
1603 write_unlock_bh(&__ip_vs_svc_lock);
1604 return 0;
1605}
1606
fc723250 1607static int ip_vs_zero_all(struct net *net)
1da177e4
LT		 1608{
1609 int idx;
1610 struct ip_vs_service *svc;
1611
1612 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1613 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
fc723250
HS		 1614 if (net_eq(svc->net, net))
1615 ip_vs_zero_service(svc);
1da177e4
LT		 1616 }
1617 }
1618
1619 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1620 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
fc723250
HS		 1621 if (net_eq(svc->net, net))
1622 ip_vs_zero_service(svc);
1da177e4
LT		 1623 }
1624 }
1625
2a0751af 1626 ip_vs_zero_stats(&net_ipvs(net)->tot_stats);
1da177e4
LT		 1627 return 0;
1628}
1629
14e40546 1630#ifdef CONFIG_SYSCTL
749c42b6
JA		 1631
1632static int zero;
1633static int three = 3;
1634
1da177e4 1635static int
8d65af78 1636proc_do_defense_mode(ctl_table *table, int write,
1da177e4
LT		 1637 void __user *buffer, size_t *lenp, loff_t *ppos)
1638{
9330419d 1639 struct net *net = current->nsproxy->net_ns;
1da177e4
LT		 1640 int *valp = table->data;
1641 int val = *valp;
1642 int rc;
1643
8d65af78 1644 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1da177e4
LT		 1645 if (write && (*valp != val)) {
1646 if ((*valp < 0) || (*valp > 3)) {
1647 /* Restore the correct value */
1648 *valp = val;
1649 } else {
9330419d 1650 update_defense_level(net_ipvs(net));
1da177e4
LT		 1651 }
1652 }
1653 return rc;
1654}
1655
1da177e4 1656static int
8d65af78 1657proc_do_sync_threshold(ctl_table *table, int write,
1da177e4
LT		 1658 void __user *buffer, size_t *lenp, loff_t *ppos)
1659{
1660 int *valp = table->data;
1661 int val[2];
1662 int rc;
1663
1664 /* backup the value first */
1665 memcpy(val, valp, sizeof(val));
1666
8d65af78 1667 rc = proc_dointvec(table, write, buffer, lenp, ppos);
749c42b6
JA		 1668 if (write && (valp[0] < 0 || valp[1] < 0 ||
1669 (valp[0] >= valp[1] && valp[1]))) {
1da177e4
LT		 1670 /* Restore the correct value */
1671 memcpy(valp, val, sizeof(val));
1672 }
1673 return rc;
1674}
1675
b880c1f0
HS		 1676static int
1677proc_do_sync_mode(ctl_table *table, int write,
1678 void __user *buffer, size_t *lenp, loff_t *ppos)
1679{
1680 int *valp = table->data;
1681 int val = *valp;
1682 int rc;
1683
1684 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1685 if (write && (*valp != val)) {
1686 if ((*valp < 0) || (*valp > 1)) {
1687 /* Restore the correct value */
1688 *valp = val;
f73181c8
PNA		 1689 }
1690 }
1691 return rc;
1692}
1693
1694static int
1695proc_do_sync_ports(ctl_table *table, int write,
1696 void __user *buffer, size_t *lenp, loff_t *ppos)
1697{
1698 int *valp = table->data;
1699 int val = *valp;
1700 int rc;
1701
1702 rc = proc_dointvec(table, write, buffer, lenp, ppos);
1703 if (write && (*valp != val)) {
1704 if (*valp < 1 || !is_power_of_2(*valp)) {
1705 /* Restore the correct value */
1706 *valp = val;
b880c1f0
HS		 1707 }
1708 }
1709 return rc;
1710}
1da177e4
LT		 1711
1712/*
1713 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
a0840e2e 1714 * Do not change order or insert new entries without
503cf15a 1715 * align with netns init in ip_vs_control_net_init()
1da177e4
LT		 1716 */
1717
1718static struct ctl_table vs_vars[] = {
1719 {
1da177e4 1720 .procname = "amemthresh",
1da177e4
LT		 1721 .maxlen = sizeof(int),
1722 .mode = 0644,
6d9f239a 1723 .proc_handler = proc_dointvec,
1da177e4 1724 },
1da177e4 1725 {
1da177e4 1726 .procname = "am_droprate",
1da177e4
LT		 1727 .maxlen = sizeof(int),
1728 .mode = 0644,
6d9f239a 1729 .proc_handler = proc_dointvec,
1da177e4
LT		 1730 },
1731 {
1da177e4 1732 .procname = "drop_entry",
1da177e4
LT		 1733 .maxlen = sizeof(int),
1734 .mode = 0644,
6d9f239a 1735 .proc_handler = proc_do_defense_mode,
1da177e4
LT		 1736 },
1737 {
1da177e4 1738 .procname = "drop_packet",
1da177e4
LT		 1739 .maxlen = sizeof(int),
1740 .mode = 0644,
6d9f239a 1741 .proc_handler = proc_do_defense_mode,
1da177e4 1742 },
f4bc17cd
JA		 1743#ifdef CONFIG_IP_VS_NFCT
1744 {
1745 .procname = "conntrack",
f4bc17cd
JA		 1746 .maxlen = sizeof(int),
1747 .mode = 0644,
1748 .proc_handler = &proc_dointvec,
1749 },
1750#endif
1da177e4 1751 {
1da177e4 1752 .procname = "secure_tcp",
1da177e4
LT		 1753 .maxlen = sizeof(int),
1754 .mode = 0644,
6d9f239a 1755 .proc_handler = proc_do_defense_mode,
1da177e4 1756 },
8a803040
JA		 1757 {
1758 .procname = "snat_reroute",
8a803040
JA		 1759 .maxlen = sizeof(int),
1760 .mode = 0644,
1761 .proc_handler = &proc_dointvec,
1762 },
b880c1f0
HS		 1763 {
1764 .procname = "sync_version",
b880c1f0
HS		 1765 .maxlen = sizeof(int),
1766 .mode = 0644,
1767 .proc_handler = &proc_do_sync_mode,
1768 },
f73181c8
PNA		 1769 {
1770 .procname = "sync_ports",
1771 .maxlen = sizeof(int),
1772 .mode = 0644,
1773 .proc_handler = &proc_do_sync_ports,
1774 },
1c003b15
PNA		 1775 {
1776 .procname = "sync_qlen_max",
1777 .maxlen = sizeof(int),
1778 .mode = 0644,
1779 .proc_handler = proc_dointvec,
1780 },
1781 {
1782 .procname = "sync_sock_size",
1783 .maxlen = sizeof(int),
1784 .mode = 0644,
1785 .proc_handler = proc_dointvec,
1786 },
a0840e2e
HS		 1787 {
1788 .procname = "cache_bypass",
1789 .maxlen = sizeof(int),
1790 .mode = 0644,
1791 .proc_handler = proc_dointvec,
1792 },
1793 {
1794 .procname = "expire_nodest_conn",
1795 .maxlen = sizeof(int),
1796 .mode = 0644,
1797 .proc_handler = proc_dointvec,
1798 },
1799 {
1800 .procname = "expire_quiescent_template",
1801 .maxlen = sizeof(int),
1802 .mode = 0644,
1803 .proc_handler = proc_dointvec,
1804 },
1805 {
1806 .procname = "sync_threshold",
1807 .maxlen =
1808 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1809 .mode = 0644,
1810 .proc_handler = proc_do_sync_threshold,
1811 },
749c42b6
JA		 1812 {
1813 .procname = "sync_refresh_period",
1814 .maxlen = sizeof(int),
1815 .mode = 0644,
1816 .proc_handler = proc_dointvec_jiffies,
1817 },
1818 {
1819 .procname = "sync_retries",
1820 .maxlen = sizeof(int),
1821 .mode = 0644,
1822 .proc_handler = proc_dointvec_minmax,
1823 .extra1 = &zero,
1824 .extra2 = &three,
1825 },
a0840e2e
HS		 1826 {
1827 .procname = "nat_icmp_send",
1828 .maxlen = sizeof(int),
1829 .mode = 0644,
1830 .proc_handler = proc_dointvec,
1831 },
3654e611
JA		 1832 {
1833 .procname = "pmtu_disc",
1834 .maxlen = sizeof(int),
1835 .mode = 0644,
1836 .proc_handler = proc_dointvec,
1837 },
0c12582f
JA		 1838 {
1839 .procname = "backup_only",
1840 .maxlen = sizeof(int),
1841 .mode = 0644,
1842 .proc_handler = proc_dointvec,
1843 },
a0840e2e
HS		 1844#ifdef CONFIG_IP_VS_DEBUG
1845 {
1846 .procname = "debug_level",
1847 .data = &sysctl_ip_vs_debug_level,
1848 .maxlen = sizeof(int),
1849 .mode = 0644,
1850 .proc_handler = proc_dointvec,
1851 },
1852#endif
1da177e4
LT		 1853#if 0
1854 {
1da177e4
LT		 1855 .procname = "timeout_established",
1856 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ESTABLISHED],
1857 .maxlen = sizeof(int),
1858 .mode = 0644,
6d9f239a 1859 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1860 },
1861 {
1da177e4
LT		 1862 .procname = "timeout_synsent",
1863 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_SENT],
1864 .maxlen = sizeof(int),
1865 .mode = 0644,
6d9f239a 1866 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1867 },
1868 {
1da177e4
LT		 1869 .procname = "timeout_synrecv",
1870 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYN_RECV],
1871 .maxlen = sizeof(int),
1872 .mode = 0644,
6d9f239a 1873 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1874 },
1875 {
1da177e4
LT		 1876 .procname = "timeout_finwait",
1877 .data = &vs_timeout_table_dos.timeout[IP_VS_S_FIN_WAIT],
1878 .maxlen = sizeof(int),
1879 .mode = 0644,
6d9f239a 1880 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1881 },
1882 {
1da177e4
LT		 1883 .procname = "timeout_timewait",
1884 .data = &vs_timeout_table_dos.timeout[IP_VS_S_TIME_WAIT],
1885 .maxlen = sizeof(int),
1886 .mode = 0644,
6d9f239a 1887 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1888 },
1889 {
1da177e4
LT		 1890 .procname = "timeout_close",
1891 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE],
1892 .maxlen = sizeof(int),
1893 .mode = 0644,
6d9f239a 1894 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1895 },
1896 {
1da177e4
LT		 1897 .procname = "timeout_closewait",
1898 .data = &vs_timeout_table_dos.timeout[IP_VS_S_CLOSE_WAIT],
1899 .maxlen = sizeof(int),
1900 .mode = 0644,
6d9f239a 1901 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1902 },
1903 {
1da177e4
LT		 1904 .procname = "timeout_lastack",
1905 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LAST_ACK],
1906 .maxlen = sizeof(int),
1907 .mode = 0644,
6d9f239a 1908 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1909 },
1910 {
1da177e4
LT		 1911 .procname = "timeout_listen",
1912 .data = &vs_timeout_table_dos.timeout[IP_VS_S_LISTEN],
1913 .maxlen = sizeof(int),
1914 .mode = 0644,
6d9f239a 1915 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1916 },
1917 {
1da177e4
LT		 1918 .procname = "timeout_synack",
1919 .data = &vs_timeout_table_dos.timeout[IP_VS_S_SYNACK],
1920 .maxlen = sizeof(int),
1921 .mode = 0644,
6d9f239a 1922 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1923 },
1924 {
1da177e4
LT		 1925 .procname = "timeout_udp",
1926 .data = &vs_timeout_table_dos.timeout[IP_VS_S_UDP],
1927 .maxlen = sizeof(int),
1928 .mode = 0644,
6d9f239a 1929 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1930 },
1931 {
1da177e4
LT		 1932 .procname = "timeout_icmp",
1933 .data = &vs_timeout_table_dos.timeout[IP_VS_S_ICMP],
1934 .maxlen = sizeof(int),
1935 .mode = 0644,
6d9f239a 1936 .proc_handler = proc_dointvec_jiffies,
1da177e4
LT		 1937 },
1938#endif
f8572d8f 1939 { }
1da177e4
LT		 1940};
1941
14e40546 1942#endif
1da177e4 1943
1da177e4
LT		 1944#ifdef CONFIG_PROC_FS
1945
1946struct ip_vs_iter {
fc723250 1947 struct seq_net_private p; /* Do not move this, netns depends upon it*/
1da177e4
LT		 1948 struct list_head *table;
1949 int bucket;
1950};
1951
1952/*
1953 * Write the contents of the VS rule table to a PROCfs file.
1954 * (It is kept just for backward compatibility)
1955 */
95c96174 1956static inline const char *ip_vs_fwd_name(unsigned int flags)
1da177e4
LT		 1957{
1958 switch (flags & IP_VS_CONN_F_FWD_MASK) {
1959 case IP_VS_CONN_F_LOCALNODE:
1960 return "Local";
1961 case IP_VS_CONN_F_TUNNEL:
1962 return "Tunnel";
1963 case IP_VS_CONN_F_DROUTE:
1964 return "Route";
1965 default:
1966 return "Masq";
1967 }
1968}
1969
1970
1971/* Get the Nth entry in the two lists */
1972static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
1973{
fc723250 1974 struct net *net = seq_file_net(seq);
1da177e4
LT		 1975 struct ip_vs_iter *iter = seq->private;
1976 int idx;
1977 struct ip_vs_service *svc;
1978
1979 /* look in hash by protocol */
1980 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1981 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
fc723250 1982 if (net_eq(svc->net, net) && pos-- == 0) {
1da177e4
LT		 1983 iter->table = ip_vs_svc_table;
1984 iter->bucket = idx;
1985 return svc;
1986 }
1987 }
1988 }
1989
1990 /* keep looking in fwmark */
1991 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1992 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
fc723250 1993 if (net_eq(svc->net, net) && pos-- == 0) {
1da177e4
LT		 1994 iter->table = ip_vs_svc_fwm_table;
1995 iter->bucket = idx;
1996 return svc;
1997 }
1998 }
1999 }
2000
2001 return NULL;
2002}
2003
2004static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
563e94f0 2005__acquires(__ip_vs_svc_lock)
1da177e4
LT		 2006{
2007
2008 read_lock_bh(&__ip_vs_svc_lock);
2009 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
2010}
2011
2012
2013static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2014{
2015 struct list_head *e;
2016 struct ip_vs_iter *iter;
2017 struct ip_vs_service *svc;
2018
2019 ++*pos;
2020 if (v == SEQ_START_TOKEN)
2021 return ip_vs_info_array(seq,0);
2022
2023 svc = v;
2024 iter = seq->private;
2025
2026 if (iter->table == ip_vs_svc_table) {
2027 /* next service in table hashed by protocol */
2028 if ((e = svc->s_list.next) != &ip_vs_svc_table[iter->bucket])
2029 return list_entry(e, struct ip_vs_service, s_list);
2030
2031
2032 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2033 list_for_each_entry(svc,&ip_vs_svc_table[iter->bucket],
2034 s_list) {
2035 return svc;
2036 }
2037 }
2038
2039 iter->table = ip_vs_svc_fwm_table;
2040 iter->bucket = -1;
2041 goto scan_fwmark;
2042 }
2043
2044 /* next service in hashed by fwmark */
2045 if ((e = svc->f_list.next) != &ip_vs_svc_fwm_table[iter->bucket])
2046 return list_entry(e, struct ip_vs_service, f_list);
2047
2048 scan_fwmark:
2049 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2050 list_for_each_entry(svc, &ip_vs_svc_fwm_table[iter->bucket],
2051 f_list)
2052 return svc;
2053 }
2054
2055 return NULL;
2056}
2057
2058static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
563e94f0 2059__releases(__ip_vs_svc_lock)
1da177e4
LT		 2060{
2061 read_unlock_bh(&__ip_vs_svc_lock);
2062}
2063
2064
2065static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
2066{
2067 if (v == SEQ_START_TOKEN) {
2068 seq_printf(seq,
2069 "IP Virtual Server version %d.%d.%d (size=%d)\n",
6f7edb48 2070 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
1da177e4
LT		 2071 seq_puts(seq,
2072 "Prot LocalAddress:Port Scheduler Flags\n");
2073 seq_puts(seq,
2074 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2075 } else {
2076 const struct ip_vs_service *svc = v;
2077 const struct ip_vs_iter *iter = seq->private;
2078 const struct ip_vs_dest *dest;
2079
667a5f18
VB		 2080 if (iter->table == ip_vs_svc_table) {
2081#ifdef CONFIG_IP_VS_IPV6
2082 if (svc->af == AF_INET6)
5b095d98 2083 seq_printf(seq, "%s [%pI6]:%04X %s ",
667a5f18 2084 ip_vs_proto_name(svc->protocol),
38ff4fa4 2085 &svc->addr.in6,
667a5f18
VB		 2086 ntohs(svc->port),
2087 svc->scheduler->name);
2088 else
2089#endif
26ec037f 2090 seq_printf(seq, "%s %08X:%04X %s %s ",
667a5f18
VB		 2091 ip_vs_proto_name(svc->protocol),
2092 ntohl(svc->addr.ip),
2093 ntohs(svc->port),
26ec037f
NC		 2094 svc->scheduler->name,
2095 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
667a5f18 2096 } else {
26ec037f
NC		 2097 seq_printf(seq, "FWM %08X %s %s",
2098 svc->fwmark, svc->scheduler->name,
2099 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
667a5f18 2100 }
1da177e4
LT		 2101
2102 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2103 seq_printf(seq, "persistent %d %08X\n",
2104 svc->timeout,
2105 ntohl(svc->netmask));
2106 else
2107 seq_putc(seq, '\n');
2108
413c2d04 2109 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
667a5f18
VB		 2110#ifdef CONFIG_IP_VS_IPV6
2111 if (dest->af == AF_INET6)
2112 seq_printf(seq,
5b095d98 2113 " -> [%pI6]:%04X"
667a5f18 2114 " %-7s %-6d %-10d %-10d\n",
38ff4fa4 2115 &dest->addr.in6,
667a5f18
VB		 2116 ntohs(dest->port),
2117 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2118 atomic_read(&dest->weight),
2119 atomic_read(&dest->activeconns),
2120 atomic_read(&dest->inactconns));
2121 else
2122#endif
2123 seq_printf(seq,
2124 " -> %08X:%04X "
2125 "%-7s %-6d %-10d %-10d\n",
2126 ntohl(dest->addr.ip),
2127 ntohs(dest->port),
2128 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2129 atomic_read(&dest->weight),
2130 atomic_read(&dest->activeconns),
2131 atomic_read(&dest->inactconns));
2132
1da177e4
LT		 2133 }
2134 }
2135 return 0;
2136}
2137
56b3d975 2138static const struct seq_operations ip_vs_info_seq_ops = {
1da177e4
LT		 2139 .start = ip_vs_info_seq_start,
2140 .next = ip_vs_info_seq_next,
2141 .stop = ip_vs_info_seq_stop,
2142 .show = ip_vs_info_seq_show,
2143};
2144
2145static int ip_vs_info_open(struct inode *inode, struct file *file)
2146{
fc723250 2147 return seq_open_net(inode, file, &ip_vs_info_seq_ops,
cf7732e4 2148 sizeof(struct ip_vs_iter));
1da177e4
LT		 2149}
2150
9a32144e 2151static const struct file_operations ip_vs_info_fops = {
1da177e4
LT		 2152 .owner = THIS_MODULE,
2153 .open = ip_vs_info_open,
2154 .read = seq_read,
2155 .llseek = seq_lseek,
0f08190f 2156 .release = seq_release_net,
1da177e4
LT		 2157};
2158
1da177e4
LT		 2159static int ip_vs_stats_show(struct seq_file *seq, void *v)
2160{
b17fc996 2161 struct net *net = seq_file_single_net(seq);
55a3d4e1 2162 struct ip_vs_stats_user show;
1da177e4
LT		 2163
2164/* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2165 seq_puts(seq,
2166 " Total Incoming Outgoing Incoming Outgoing\n");
2167 seq_printf(seq,
2168 " Conns Packets Packets Bytes Bytes\n");
2169
55a3d4e1
JA		 2170 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2171 seq_printf(seq, "%8X %8X %8X %16LX %16LX\n\n", show.conns,
2172 show.inpkts, show.outpkts,
2173 (unsigned long long) show.inbytes,
2174 (unsigned long long) show.outbytes);
1da177e4
LT		 2175
2176/* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2177 seq_puts(seq,
2178 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
55a3d4e1
JA		 2179 seq_printf(seq, "%8X %8X %8X %16X %16X\n",
2180 show.cps, show.inpps, show.outpps,
2181 show.inbps, show.outbps);
1da177e4
LT		 2182
2183 return 0;
2184}
2185
2186static int ip_vs_stats_seq_open(struct inode *inode, struct file *file)
2187{
fc723250 2188 return single_open_net(inode, file, ip_vs_stats_show);
1da177e4
LT		 2189}
2190
9a32144e 2191static const struct file_operations ip_vs_stats_fops = {
1da177e4
LT		 2192 .owner = THIS_MODULE,
2193 .open = ip_vs_stats_seq_open,
2194 .read = seq_read,
2195 .llseek = seq_lseek,
0f08190f 2196 .release = single_release_net,
1da177e4
LT		 2197};
2198
b17fc996
HS		 2199static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2200{
2201 struct net *net = seq_file_single_net(seq);
2a0751af
JA		 2202 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2203 struct ip_vs_cpu_stats *cpustats = tot_stats->cpustats;
ea9f22cc 2204 struct ip_vs_stats_user rates;
b17fc996
HS		 2205 int i;
2206
2207/* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2208 seq_puts(seq,
2209 " Total Incoming Outgoing Incoming Outgoing\n");
2210 seq_printf(seq,
2211 "CPU Conns Packets Packets Bytes Bytes\n");
2212
2213 for_each_possible_cpu(i) {
2a0751af
JA		 2214 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2215 unsigned int start;
2216 __u64 inbytes, outbytes;
2217
2218 do {
2219 start = u64_stats_fetch_begin_bh(&u->syncp);
2220 inbytes = u->ustats.inbytes;
2221 outbytes = u->ustats.outbytes;
2222 } while (u64_stats_fetch_retry_bh(&u->syncp, start));
2223
b17fc996 2224 seq_printf(seq, "%3X %8X %8X %8X %16LX %16LX\n",
2a0751af
JA		 2225 i, u->ustats.conns, u->ustats.inpkts,
2226 u->ustats.outpkts, (__u64)inbytes,
2227 (__u64)outbytes);
b17fc996
HS		 2228 }
2229
2230 spin_lock_bh(&tot_stats->lock);
ea9f22cc 2231
b17fc996
HS		 2232 seq_printf(seq, " ~ %8X %8X %8X %16LX %16LX\n\n",
2233 tot_stats->ustats.conns, tot_stats->ustats.inpkts,
2234 tot_stats->ustats.outpkts,
2235 (unsigned long long) tot_stats->ustats.inbytes,
2236 (unsigned long long) tot_stats->ustats.outbytes);
2237
ea9f22cc
JA		 2238 ip_vs_read_estimator(&rates, tot_stats);
2239
2240 spin_unlock_bh(&tot_stats->lock);
2241
b17fc996
HS		 2242/* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2243 seq_puts(seq,
2244 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2245 seq_printf(seq, " %8X %8X %8X %16X %16X\n",
ea9f22cc
JA		 2246 rates.cps,
2247 rates.inpps,
2248 rates.outpps,
2249 rates.inbps,
2250 rates.outbps);
b17fc996
HS		 2251
2252 return 0;
2253}
2254
2255static int ip_vs_stats_percpu_seq_open(struct inode *inode, struct file *file)
2256{
2257 return single_open_net(inode, file, ip_vs_stats_percpu_show);
2258}
2259
2260static const struct file_operations ip_vs_stats_percpu_fops = {
2261 .owner = THIS_MODULE,
2262 .open = ip_vs_stats_percpu_seq_open,
2263 .read = seq_read,
2264 .llseek = seq_lseek,
0f08190f 2265 .release = single_release_net,
b17fc996 2266};
1da177e4
LT		 2267#endif
2268
2269/*
2270 * Set timeout values for tcp tcpfin udp in the timeout_table.
2271 */
9330419d 2272static int ip_vs_set_timeout(struct net *net, struct ip_vs_timeout_user *u)
1da177e4 2273{
091bb34c 2274#if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
9330419d 2275 struct ip_vs_proto_data *pd;
091bb34c 2276#endif
9330419d 2277
1da177e4
LT		 2278 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2279 u->tcp_timeout,
2280 u->tcp_fin_timeout,
2281 u->udp_timeout);
2282
2283#ifdef CONFIG_IP_VS_PROTO_TCP
2284 if (u->tcp_timeout) {
9330419d
HS		 2285 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2286 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
1da177e4
LT		 2287 = u->tcp_timeout * HZ;
2288 }
2289
2290 if (u->tcp_fin_timeout) {
9330419d
HS		 2291 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2292 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
1da177e4
LT		 2293 = u->tcp_fin_timeout * HZ;
2294 }
2295#endif
2296
2297#ifdef CONFIG_IP_VS_PROTO_UDP
2298 if (u->udp_timeout) {
9330419d
HS		 2299 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
2300 pd->timeout_table[IP_VS_UDP_S_NORMAL]
1da177e4
LT		 2301 = u->udp_timeout * HZ;
2302 }
2303#endif
2304 return 0;
2305}
2306
2307
2308#define SET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2309#define SERVICE_ARG_LEN (sizeof(struct ip_vs_service_user))
2310#define SVCDEST_ARG_LEN (sizeof(struct ip_vs_service_user) + \
2311 sizeof(struct ip_vs_dest_user))
2312#define TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2313#define DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user))
2314#define MAX_ARG_LEN SVCDEST_ARG_LEN
2315
9b5b5cff 2316static const unsigned char set_arglen[SET_CMDID(IP_VS_SO_SET_MAX)+1] = {
1da177e4
LT		 2317 [SET_CMDID(IP_VS_SO_SET_ADD)] = SERVICE_ARG_LEN,
2318 [SET_CMDID(IP_VS_SO_SET_EDIT)] = SERVICE_ARG_LEN,
2319 [SET_CMDID(IP_VS_SO_SET_DEL)] = SERVICE_ARG_LEN,
2320 [SET_CMDID(IP_VS_SO_SET_FLUSH)] = 0,
2321 [SET_CMDID(IP_VS_SO_SET_ADDDEST)] = SVCDEST_ARG_LEN,
2322 [SET_CMDID(IP_VS_SO_SET_DELDEST)] = SVCDEST_ARG_LEN,
2323 [SET_CMDID(IP_VS_SO_SET_EDITDEST)] = SVCDEST_ARG_LEN,
2324 [SET_CMDID(IP_VS_SO_SET_TIMEOUT)] = TIMEOUT_ARG_LEN,
2325 [SET_CMDID(IP_VS_SO_SET_STARTDAEMON)] = DAEMON_ARG_LEN,
2326 [SET_CMDID(IP_VS_SO_SET_STOPDAEMON)] = DAEMON_ARG_LEN,
2327 [SET_CMDID(IP_VS_SO_SET_ZERO)] = SERVICE_ARG_LEN,
2328};
2329
c860c6b1
JV		 2330static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2331 struct ip_vs_service_user *usvc_compat)
2332{
0d1e71b0
SH		 2333 memset(usvc, 0, sizeof(*usvc));
2334
c860c6b1
JV		 2335 usvc->af = AF_INET;
2336 usvc->protocol = usvc_compat->protocol;
2337 usvc->addr.ip = usvc_compat->addr;
2338 usvc->port = usvc_compat->port;
2339 usvc->fwmark = usvc_compat->fwmark;
2340
2341 /* Deep copy of sched_name is not needed here */
2342 usvc->sched_name = usvc_compat->sched_name;
2343
2344 usvc->flags = usvc_compat->flags;
2345 usvc->timeout = usvc_compat->timeout;
2346 usvc->netmask = usvc_compat->netmask;
2347}
2348
2349static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2350 struct ip_vs_dest_user *udest_compat)
2351{
0d1e71b0
SH		 2352 memset(udest, 0, sizeof(*udest));
2353
c860c6b1
JV		 2354 udest->addr.ip = udest_compat->addr;
2355 udest->port = udest_compat->port;
2356 udest->conn_flags = udest_compat->conn_flags;
2357 udest->weight = udest_compat->weight;
2358 udest->u_threshold = udest_compat->u_threshold;
2359 udest->l_threshold = udest_compat->l_threshold;
2360}
2361
1da177e4
LT		 2362static int
2363do_ip_vs_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2364{
fc723250 2365 struct net *net = sock_net(sk);
1da177e4
LT		 2366 int ret;
2367 unsigned char arg[MAX_ARG_LEN];
c860c6b1
JV		 2368 struct ip_vs_service_user *usvc_compat;
2369 struct ip_vs_service_user_kern usvc;
1da177e4 2370 struct ip_vs_service *svc;
c860c6b1
JV		 2371 struct ip_vs_dest_user *udest_compat;
2372 struct ip_vs_dest_user_kern udest;
ae1d48b2 2373 struct netns_ipvs *ipvs = net_ipvs(net);
1da177e4 2374
df008c91 2375 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1da177e4
LT		 2376 return -EPERM;
2377
04bcef2a
AV		 2378 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2379 return -EINVAL;
2380 if (len < 0 || len > MAX_ARG_LEN)
2381 return -EINVAL;
1da177e4 2382 if (len != set_arglen[SET_CMDID(cmd)]) {
1e3e238e
HE		 2383 pr_err("set_ctl: len %u != %u\n",
2384 len, set_arglen[SET_CMDID(cmd)]);
1da177e4
LT		 2385 return -EINVAL;
2386 }
2387
2388 if (copy_from_user(arg, user, len) != 0)
2389 return -EFAULT;
2390
2391 /* increase the module use count */
2392 ip_vs_use_count_inc();
2393
ae1d48b2
HS		 2394 /* Handle daemons since they have another lock */
2395 if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2396 cmd == IP_VS_SO_SET_STOPDAEMON) {
2397 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2398
2399 if (mutex_lock_interruptible(&ipvs->sync_mutex)) {
2400 ret = -ERESTARTSYS;
2401 goto out_dec;
2402 }
2403 if (cmd == IP_VS_SO_SET_STARTDAEMON)
2404 ret = start_sync_thread(net, dm->state, dm->mcast_ifn,
2405 dm->syncid);
2406 else
2407 ret = stop_sync_thread(net, dm->state);
2408 mutex_unlock(&ipvs->sync_mutex);
2409 goto out_dec;
2410 }
2411
14cc3e2b 2412 if (mutex_lock_interruptible(&__ip_vs_mutex)) {
1da177e4
LT		 2413 ret = -ERESTARTSYS;
2414 goto out_dec;
2415 }
2416
2417 if (cmd == IP_VS_SO_SET_FLUSH) {
2418 /* Flush the virtual service */
578bc3ef 2419 ret = ip_vs_flush(net, false);
1da177e4
LT		 2420 goto out_unlock;
2421 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2422 /* Set timeout values for (tcp tcpfin udp) */
9330419d 2423 ret = ip_vs_set_timeout(net, (struct ip_vs_timeout_user *)arg);
1da177e4 2424 goto out_unlock;
1da177e4
LT		 2425 }
2426
c860c6b1
JV		 2427 usvc_compat = (struct ip_vs_service_user *)arg;
2428 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2429
2430 /* We only use the new structs internally, so copy userspace compat
2431 * structs to extended internal versions */
2432 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2433 ip_vs_copy_udest_compat(&udest, udest_compat);
1da177e4
LT		 2434
2435 if (cmd == IP_VS_SO_SET_ZERO) {
2436 /* if no service address is set, zero counters in all */
c860c6b1 2437 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
fc723250 2438 ret = ip_vs_zero_all(net);
1da177e4
LT		 2439 goto out_unlock;
2440 }
2441 }
2442
2906f66a
VMR		 2443 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2444 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2445 usvc.protocol != IPPROTO_SCTP) {
1e3e238e
HE		 2446 pr_err("set_ctl: invalid protocol: %d %pI4:%d %s\n",
2447 usvc.protocol, &usvc.addr.ip,
2448 ntohs(usvc.port), usvc.sched_name);
1da177e4
LT		 2449 ret = -EFAULT;
2450 goto out_unlock;
2451 }
2452
2453 /* Lookup the exact service by <protocol, addr, port> or fwmark */
c860c6b1 2454 if (usvc.fwmark == 0)
fc723250 2455 svc = __ip_vs_service_find(net, usvc.af, usvc.protocol,
26c15cfd 2456 &usvc.addr, usvc.port);
1da177e4 2457 else
fc723250 2458 svc = __ip_vs_svc_fwm_find(net, usvc.af, usvc.fwmark);
1da177e4
LT		 2459
2460 if (cmd != IP_VS_SO_SET_ADD
c860c6b1 2461 && (svc == NULL || svc->protocol != usvc.protocol)) {
1da177e4 2462 ret = -ESRCH;
26c15cfd 2463 goto out_unlock;
1da177e4
LT		 2464 }
2465
2466 switch (cmd) {
2467 case IP_VS_SO_SET_ADD:
2468 if (svc != NULL)
2469 ret = -EEXIST;
2470 else
fc723250 2471 ret = ip_vs_add_service(net, &usvc, &svc);
1da177e4
LT		 2472 break;
2473 case IP_VS_SO_SET_EDIT:
c860c6b1 2474 ret = ip_vs_edit_service(svc, &usvc);
1da177e4
LT		 2475 break;
2476 case IP_VS_SO_SET_DEL:
2477 ret = ip_vs_del_service(svc);
2478 if (!ret)
2479 goto out_unlock;
2480 break;
2481 case IP_VS_SO_SET_ZERO:
2482 ret = ip_vs_zero_service(svc);
2483 break;
2484 case IP_VS_SO_SET_ADDDEST:
c860c6b1 2485 ret = ip_vs_add_dest(svc, &udest);
1da177e4
LT		 2486 break;
2487 case IP_VS_SO_SET_EDITDEST:
c860c6b1 2488 ret = ip_vs_edit_dest(svc, &udest);
1da177e4
LT		 2489 break;
2490 case IP_VS_SO_SET_DELDEST:
c860c6b1 2491 ret = ip_vs_del_dest(svc, &udest);
1da177e4
LT		 2492 break;
2493 default:
2494 ret = -EINVAL;
2495 }
2496
1da177e4 2497 out_unlock:
14cc3e2b 2498 mutex_unlock(&__ip_vs_mutex);
1da177e4
LT		 2499 out_dec:
2500 /* decrease the module use count */
2501 ip_vs_use_count_dec();
2502
2503 return ret;
2504}
2505
2506
1da177e4
LT		 2507static void
2508ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2509{
2510 dst->protocol = src->protocol;
e7ade46a 2511 dst->addr = src->addr.ip;
1da177e4
LT		 2512 dst->port = src->port;
2513 dst->fwmark = src->fwmark;
4da62fc7 2514 strlcpy(dst->sched_name, src->scheduler->name, sizeof(dst->sched_name));
1da177e4
LT		 2515 dst->flags = src->flags;
2516 dst->timeout = src->timeout / HZ;
2517 dst->netmask = src->netmask;
2518 dst->num_dests = src->num_dests;
2519 ip_vs_copy_stats(&dst->stats, &src->stats);
2520}
2521
2522static inline int
fc723250
HS		 2523__ip_vs_get_service_entries(struct net *net,
2524 const struct ip_vs_get_services *get,
1da177e4
LT		 2525 struct ip_vs_get_services __user *uptr)
2526{
2527 int idx, count=0;
2528 struct ip_vs_service *svc;
2529 struct ip_vs_service_entry entry;
2530 int ret = 0;
2531
2532 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2533 list_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
f94fd041 2534 /* Only expose IPv4 entries to old interface */
fc723250 2535 if (svc->af != AF_INET || !net_eq(svc->net, net))
f94fd041
JV		 2536 continue;
2537
1da177e4
LT		 2538 if (count >= get->num_services)
2539 goto out;
4da62fc7 2540 memset(&entry, 0, sizeof(entry));
1da177e4
LT		 2541 ip_vs_copy_service(&entry, svc);
2542 if (copy_to_user(&uptr->entrytable[count],
2543 &entry, sizeof(entry))) {
2544 ret = -EFAULT;
2545 goto out;
2546 }
2547 count++;
2548 }
2549 }
2550
2551 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2552 list_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
f94fd041 2553 /* Only expose IPv4 entries to old interface */
fc723250 2554 if (svc->af != AF_INET || !net_eq(svc->net, net))
f94fd041
JV		 2555 continue;
2556
1da177e4
LT		 2557 if (count >= get->num_services)
2558 goto out;
4da62fc7 2559 memset(&entry, 0, sizeof(entry));
1da177e4
LT		 2560 ip_vs_copy_service(&entry, svc);
2561 if (copy_to_user(&uptr->entrytable[count],
2562 &entry, sizeof(entry))) {
2563 ret = -EFAULT;
2564 goto out;
2565 }
2566 count++;
2567 }
2568 }
552ad65a 2569out:
1da177e4
LT		 2570 return ret;
2571}
2572
2573static inline int
fc723250 2574__ip_vs_get_dest_entries(struct net *net, const struct ip_vs_get_dests *get,
1da177e4
LT		 2575 struct ip_vs_get_dests __user *uptr)
2576{
2577 struct ip_vs_service *svc;
b18610de 2578 union nf_inet_addr addr = { .ip = get->addr };
1da177e4
LT		 2579 int ret = 0;
2580
2581 if (get->fwmark)
fc723250 2582 svc = __ip_vs_svc_fwm_find(net, AF_INET, get->fwmark);
1da177e4 2583 else
fc723250 2584 svc = __ip_vs_service_find(net, AF_INET, get->protocol, &addr,
26c15cfd 2585 get->port);
b18610de 2586
1da177e4
LT		 2587 if (svc) {
2588 int count = 0;
2589 struct ip_vs_dest *dest;
2590 struct ip_vs_dest_entry entry;
2591
2592 list_for_each_entry(dest, &svc->destinations, n_list) {
2593 if (count >= get->num_dests)
2594 break;
2595
e7ade46a 2596 entry.addr = dest->addr.ip;
1da177e4
LT		 2597 entry.port = dest->port;
2598 entry.conn_flags = atomic_read(&dest->conn_flags);
2599 entry.weight = atomic_read(&dest->weight);
2600 entry.u_threshold = dest->u_threshold;
2601 entry.l_threshold = dest->l_threshold;
2602 entry.activeconns = atomic_read(&dest->activeconns);
2603 entry.inactconns = atomic_read(&dest->inactconns);
2604 entry.persistconns = atomic_read(&dest->persistconns);
2605 ip_vs_copy_stats(&entry.stats, &dest->stats);
2606 if (copy_to_user(&uptr->entrytable[count],
2607 &entry, sizeof(entry))) {
2608 ret = -EFAULT;
2609 break;
2610 }
2611 count++;
2612 }
1da177e4
LT		 2613 } else
2614 ret = -ESRCH;
2615 return ret;
2616}
2617
2618static inline void
9330419d 2619__ip_vs_get_timeouts(struct net *net, struct ip_vs_timeout_user *u)
1da177e4 2620{
091bb34c 2621#if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
9330419d 2622 struct ip_vs_proto_data *pd;
091bb34c 2623#endif
9330419d 2624
b61a602e
AB		 2625 memset(u, 0, sizeof (*u));
2626
1da177e4 2627#ifdef CONFIG_IP_VS_PROTO_TCP
9330419d
HS		 2628 pd = ip_vs_proto_data_get(net, IPPROTO_TCP);
2629 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2630 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
1da177e4
LT		 2631#endif
2632#ifdef CONFIG_IP_VS_PROTO_UDP
9330419d 2633 pd = ip_vs_proto_data_get(net, IPPROTO_UDP);
1da177e4 2634 u->udp_timeout =
9330419d 2635 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
1da177e4
LT		 2636#endif
2637}
2638
2639
2640#define GET_CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2641#define GET_INFO_ARG_LEN (sizeof(struct ip_vs_getinfo))
2642#define GET_SERVICES_ARG_LEN (sizeof(struct ip_vs_get_services))
2643#define GET_SERVICE_ARG_LEN (sizeof(struct ip_vs_service_entry))
2644#define GET_DESTS_ARG_LEN (sizeof(struct ip_vs_get_dests))
2645#define GET_TIMEOUT_ARG_LEN (sizeof(struct ip_vs_timeout_user))
2646#define GET_DAEMON_ARG_LEN (sizeof(struct ip_vs_daemon_user) * 2)
2647
9b5b5cff 2648static const unsigned char get_arglen[GET_CMDID(IP_VS_SO_GET_MAX)+1] = {
1da177e4
LT		 2649 [GET_CMDID(IP_VS_SO_GET_VERSION)] = 64,
2650 [GET_CMDID(IP_VS_SO_GET_INFO)] = GET_INFO_ARG_LEN,
2651 [GET_CMDID(IP_VS_SO_GET_SERVICES)] = GET_SERVICES_ARG_LEN,
2652 [GET_CMDID(IP_VS_SO_GET_SERVICE)] = GET_SERVICE_ARG_LEN,
2653 [GET_CMDID(IP_VS_SO_GET_DESTS)] = GET_DESTS_ARG_LEN,
2654 [GET_CMDID(IP_VS_SO_GET_TIMEOUT)] = GET_TIMEOUT_ARG_LEN,
2655 [GET_CMDID(IP_VS_SO_GET_DAEMON)] = GET_DAEMON_ARG_LEN,
2656};
2657
2658static int
2659do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2660{
2661 unsigned char arg[128];
2662 int ret = 0;
04bcef2a 2663 unsigned int copylen;
fc723250 2664 struct net *net = sock_net(sk);
f131315f 2665 struct netns_ipvs *ipvs = net_ipvs(net);
1da177e4 2666
fc723250 2667 BUG_ON(!net);
df008c91 2668 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1da177e4
LT		 2669 return -EPERM;
2670
04bcef2a
AV		 2671 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2672 return -EINVAL;
2673
1da177e4 2674 if (*len < get_arglen[GET_CMDID(cmd)]) {
1e3e238e
HE		 2675 pr_err("get_ctl: len %u < %u\n",
2676 *len, get_arglen[GET_CMDID(cmd)]);
1da177e4
LT		 2677 return -EINVAL;
2678 }
2679
04bcef2a
AV		 2680 copylen = get_arglen[GET_CMDID(cmd)];
2681 if (copylen > 128)
2682 return -EINVAL;
2683
2684 if (copy_from_user(arg, user, copylen) != 0)
1da177e4 2685 return -EFAULT;
ae1d48b2
HS		 2686 /*
2687 * Handle daemons first since it has its own locking
2688 */
2689 if (cmd == IP_VS_SO_GET_DAEMON) {
2690 struct ip_vs_daemon_user d[2];
2691
2692 memset(&d, 0, sizeof(d));
2693 if (mutex_lock_interruptible(&ipvs->sync_mutex))
2694 return -ERESTARTSYS;
2695
2696 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2697 d[0].state = IP_VS_STATE_MASTER;
2698 strlcpy(d[0].mcast_ifn, ipvs->master_mcast_ifn,
2699 sizeof(d[0].mcast_ifn));
2700 d[0].syncid = ipvs->master_syncid;
2701 }
2702 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2703 d[1].state = IP_VS_STATE_BACKUP;
2704 strlcpy(d[1].mcast_ifn, ipvs->backup_mcast_ifn,
2705 sizeof(d[1].mcast_ifn));
2706 d[1].syncid = ipvs->backup_syncid;
2707 }
2708 if (copy_to_user(user, &d, sizeof(d)) != 0)
2709 ret = -EFAULT;
2710 mutex_unlock(&ipvs->sync_mutex);
2711 return ret;
2712 }
1da177e4 2713
14cc3e2b 2714 if (mutex_lock_interruptible(&__ip_vs_mutex))
1da177e4
LT		 2715 return -ERESTARTSYS;
2716
2717 switch (cmd) {
2718 case IP_VS_SO_GET_VERSION:
2719 {
2720 char buf[64];
2721
2722 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
6f7edb48 2723 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
1da177e4
LT		 2724 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2725 ret = -EFAULT;
2726 goto out;
2727 }
2728 *len = strlen(buf)+1;
2729 }
2730 break;
2731
2732 case IP_VS_SO_GET_INFO:
2733 {
2734 struct ip_vs_getinfo info;
2735 info.version = IP_VS_VERSION_CODE;
6f7edb48 2736 info.size = ip_vs_conn_tab_size;
a0840e2e 2737 info.num_services = ipvs->num_services;
1da177e4
LT		 2738 if (copy_to_user(user, &info, sizeof(info)) != 0)
2739 ret = -EFAULT;
2740 }
2741 break;
2742
2743 case IP_VS_SO_GET_SERVICES:
2744 {
2745 struct ip_vs_get_services *get;
2746 int size;
2747
2748 get = (struct ip_vs_get_services *)arg;
2749 size = sizeof(*get) +
2750 sizeof(struct ip_vs_service_entry) * get->num_services;
2751 if (*len != size) {
1e3e238e 2752 pr_err("length: %u != %u\n", *len, size);
1da177e4
LT		 2753 ret = -EINVAL;
2754 goto out;
2755 }
fc723250 2756 ret = __ip_vs_get_service_entries(net, get, user);
1da177e4
LT		 2757 }
2758 break;
2759
2760 case IP_VS_SO_GET_SERVICE:
2761 {
2762 struct ip_vs_service_entry *entry;
2763 struct ip_vs_service *svc;
b18610de 2764 union nf_inet_addr addr;
1da177e4
LT		 2765
2766 entry = (struct ip_vs_service_entry *)arg;
b18610de 2767 addr.ip = entry->addr;
1da177e4 2768 if (entry->fwmark)
fc723250 2769 svc = __ip_vs_svc_fwm_find(net, AF_INET, entry->fwmark);
1da177e4 2770 else
fc723250
HS		 2771 svc = __ip_vs_service_find(net, AF_INET,
2772 entry->protocol, &addr,
2773 entry->port);
1da177e4
LT		 2774 if (svc) {
2775 ip_vs_copy_service(entry, svc);
2776 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2777 ret = -EFAULT;
1da177e4
LT		 2778 } else
2779 ret = -ESRCH;
2780 }
2781 break;
2782
2783 case IP_VS_SO_GET_DESTS:
2784 {
2785 struct ip_vs_get_dests *get;
2786 int size;
2787
2788 get = (struct ip_vs_get_dests *)arg;
2789 size = sizeof(*get) +
2790 sizeof(struct ip_vs_dest_entry) * get->num_dests;
2791 if (*len != size) {
1e3e238e 2792 pr_err("length: %u != %u\n", *len, size);
1da177e4
LT		 2793 ret = -EINVAL;
2794 goto out;
2795 }
fc723250 2796 ret = __ip_vs_get_dest_entries(net, get, user);
1da177e4
LT		 2797 }
2798 break;
2799
2800 case IP_VS_SO_GET_TIMEOUT:
2801 {
2802 struct ip_vs_timeout_user t;
2803
9330419d 2804 __ip_vs_get_timeouts(net, &t);
1da177e4
LT		 2805 if (copy_to_user(user, &t, sizeof(t)) != 0)
2806 ret = -EFAULT;
2807 }
2808 break;
2809
1da177e4
LT		 2810 default:
2811 ret = -EINVAL;
2812 }
2813
552ad65a 2814out:
14cc3e2b 2815 mutex_unlock(&__ip_vs_mutex);
1da177e4
LT		 2816 return ret;
2817}
2818
2819
2820static struct nf_sockopt_ops ip_vs_sockopts = {
2821 .pf = PF_INET,
2822 .set_optmin = IP_VS_BASE_CTL,
2823 .set_optmax = IP_VS_SO_SET_MAX+1,
2824 .set = do_ip_vs_set_ctl,
2825 .get_optmin = IP_VS_BASE_CTL,
2826 .get_optmax = IP_VS_SO_GET_MAX+1,
2827 .get = do_ip_vs_get_ctl,
16fcec35 2828 .owner = THIS_MODULE,
1da177e4
LT		 2829};
2830
9a812198
JV		 2831/*
2832 * Generic Netlink interface
2833 */
2834
2835/* IPVS genetlink family */
2836static struct genl_family ip_vs_genl_family = {
2837 .id = GENL_ID_GENERATE,
2838 .hdrsize = 0,
2839 .name = IPVS_GENL_NAME,
2840 .version = IPVS_GENL_VERSION,
2841 .maxattr = IPVS_CMD_MAX,
c6d2d445 2842 .netnsok = true, /* Make ipvsadm to work on netns */
9a812198
JV		 2843};
2844
2845/* Policy used for first-level command attributes */
2846static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2847 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
2848 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
2849 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
2850 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
2851 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
2852 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
2853};
2854
2855/* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2856static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2857 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
2858 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
2859 .len = IP_VS_IFNAME_MAXLEN },
2860 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
2861};
2862
2863/* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2864static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2865 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
2866 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
2867 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
2868 .len = sizeof(union nf_inet_addr) },
2869 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
2870 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
2871 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
2872 .len = IP_VS_SCHEDNAME_MAXLEN },
0d1e71b0
SH		 2873 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
2874 .len = IP_VS_PENAME_MAXLEN },
9a812198
JV		 2875 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
2876 .len = sizeof(struct ip_vs_flags) },
2877 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
2878 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
2879 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
2880};
2881
2882/* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2883static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2884 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
2885 .len = sizeof(union nf_inet_addr) },
2886 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
2887 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
2888 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
2889 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
2890 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
2891 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
2892 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
2893 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
2894 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
2895};
2896
2897static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
2898 struct ip_vs_stats *stats)
2899{
55a3d4e1 2900 struct ip_vs_stats_user ustats;
9a812198
JV		 2901 struct nlattr *nl_stats = nla_nest_start(skb, container_type);
2902 if (!nl_stats)
2903 return -EMSGSIZE;
2904
55a3d4e1 2905 ip_vs_copy_stats(&ustats, stats);
9a812198 2906
969e8e25
DM		 2907 if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, ustats.conns) ||
2908 nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, ustats.inpkts) ||
2909 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, ustats.outpkts) ||
2910 nla_put_u64(skb, IPVS_STATS_ATTR_INBYTES, ustats.inbytes) ||
2911 nla_put_u64(skb, IPVS_STATS_ATTR_OUTBYTES, ustats.outbytes) ||
2912 nla_put_u32(skb, IPVS_STATS_ATTR_CPS, ustats.cps) ||
2913 nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, ustats.inpps) ||
2914 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, ustats.outpps) ||
2915 nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, ustats.inbps) ||
2916 nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, ustats.outbps))
2917 goto nla_put_failure;
9a812198
JV		 2918 nla_nest_end(skb, nl_stats);
2919
2920 return 0;
2921
2922nla_put_failure:
9a812198
JV		 2923 nla_nest_cancel(skb, nl_stats);
2924 return -EMSGSIZE;
2925}
2926
2927static int ip_vs_genl_fill_service(struct sk_buff *skb,
2928 struct ip_vs_service *svc)
2929{
2930 struct nlattr *nl_service;
2931 struct ip_vs_flags flags = { .flags = svc->flags,
2932 .mask = ~0 };
2933
2934 nl_service = nla_nest_start(skb, IPVS_CMD_ATTR_SERVICE);
2935 if (!nl_service)
2936 return -EMSGSIZE;
2937
969e8e25
DM		 2938 if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
2939 goto nla_put_failure;
9a812198 2940 if (svc->fwmark) {
969e8e25
DM		 2941 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
2942 goto nla_put_failure;
9a812198 2943 } else {
969e8e25
DM		 2944 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
2945 nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
2946 nla_put_u16(skb, IPVS_SVC_ATTR_PORT, svc->port))
2947 goto nla_put_failure;
9a812198
JV		 2948 }
2949
969e8e25
DM		 2950 if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, svc->scheduler->name) ||
2951 (svc->pe &&
2952 nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, svc->pe->name)) ||
2953 nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
2954 nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
2955 nla_put_u32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
2956 goto nla_put_failure;
9a812198
JV		 2957 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &svc->stats))
2958 goto nla_put_failure;
2959
2960 nla_nest_end(skb, nl_service);
2961
2962 return 0;
2963
2964nla_put_failure:
2965 nla_nest_cancel(skb, nl_service);
2966 return -EMSGSIZE;
2967}
2968
2969static int ip_vs_genl_dump_service(struct sk_buff *skb,
2970 struct ip_vs_service *svc,
2971 struct netlink_callback *cb)
2972{
2973 void *hdr;
2974
15e47304 2975 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
9a812198
JV		 2976 &ip_vs_genl_family, NLM_F_MULTI,
2977 IPVS_CMD_NEW_SERVICE);
2978 if (!hdr)
2979 return -EMSGSIZE;
2980
2981 if (ip_vs_genl_fill_service(skb, svc) < 0)
2982 goto nla_put_failure;
2983
2984 return genlmsg_end(skb, hdr);
2985
2986nla_put_failure:
2987 genlmsg_cancel(skb, hdr);
2988 return -EMSGSIZE;
2989}
2990
2991static int ip_vs_genl_dump_services(struct sk_buff *skb,
2992 struct netlink_callback *cb)
2993{
2994 int idx = 0, i;
2995 int start = cb->args[0];
2996 struct ip_vs_service *svc;
fc723250 2997 struct net *net = skb_sknet(skb);
9a812198
JV		 2998
2999 mutex_lock(&__ip_vs_mutex);
3000 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3001 list_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
fc723250 3002 if (++idx <= start || !net_eq(svc->net, net))
9a812198
JV		 3003 continue;
3004 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3005 idx--;
3006 goto nla_put_failure;
3007 }
3008 }
3009 }
3010
3011 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3012 list_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
fc723250 3013 if (++idx <= start || !net_eq(svc->net, net))
9a812198
JV		 3014 continue;
3015 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3016 idx--;
3017 goto nla_put_failure;
3018 }
3019 }
3020 }
3021
3022nla_put_failure:
3023 mutex_unlock(&__ip_vs_mutex);
3024 cb->args[0] = idx;
3025
3026 return skb->len;
3027}
3028
fc723250
HS		 3029static int ip_vs_genl_parse_service(struct net *net,
3030 struct ip_vs_service_user_kern *usvc,
26c15cfd
JA		 3031 struct nlattr *nla, int full_entry,
3032 struct ip_vs_service **ret_svc)
9a812198
JV		 3033{
3034 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
3035 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
26c15cfd 3036 struct ip_vs_service *svc;
9a812198
JV		 3037
3038 /* Parse mandatory identifying service fields first */
3039 if (nla == NULL ||
3040 nla_parse_nested(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy))
3041 return -EINVAL;
3042
3043 nla_af = attrs[IPVS_SVC_ATTR_AF];
3044 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
3045 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
3046 nla_port = attrs[IPVS_SVC_ATTR_PORT];
3047 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
3048
3049 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3050 return -EINVAL;
3051
258c8893
SH		 3052 memset(usvc, 0, sizeof(*usvc));
3053
c860c6b1 3054 usvc->af = nla_get_u16(nla_af);
f94fd041
JV		 3055#ifdef CONFIG_IP_VS_IPV6
3056 if (usvc->af != AF_INET && usvc->af != AF_INET6)
3057#else
3058 if (usvc->af != AF_INET)
3059#endif
9a812198
JV		 3060 return -EAFNOSUPPORT;
3061
3062 if (nla_fwmark) {
3063 usvc->protocol = IPPROTO_TCP;
3064 usvc->fwmark = nla_get_u32(nla_fwmark);
3065 } else {
3066 usvc->protocol = nla_get_u16(nla_protocol);
3067 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3068 usvc->port = nla_get_u16(nla_port);
3069 usvc->fwmark = 0;
3070 }
3071
26c15cfd 3072 if (usvc->fwmark)
fc723250 3073 svc = __ip_vs_svc_fwm_find(net, usvc->af, usvc->fwmark);
26c15cfd 3074 else
fc723250 3075 svc = __ip_vs_service_find(net, usvc->af, usvc->protocol,
26c15cfd
JA		 3076 &usvc->addr, usvc->port);
3077 *ret_svc = svc;
3078
9a812198
JV		 3079 /* If a full entry was requested, check for the additional fields */
3080 if (full_entry) {
0d1e71b0 3081 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
9a812198
JV		 3082 *nla_netmask;
3083 struct ip_vs_flags flags;
9a812198
JV		 3084
3085 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
0d1e71b0 3086 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
9a812198
JV		 3087 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3088 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3089 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3090
3091 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3092 return -EINVAL;
3093
3094 nla_memcpy(&flags, nla_flags, sizeof(flags));
3095
3096 /* prefill flags from service if it already exists */
26c15cfd 3097 if (svc)
9a812198 3098 usvc->flags = svc->flags;
9a812198
JV		 3099
3100 /* set new flags from userland */
3101 usvc->flags = (usvc->flags & ~flags.mask) |
3102 (flags.flags & flags.mask);
c860c6b1 3103 usvc->sched_name = nla_data(nla_sched);
0d1e71b0 3104 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
9a812198
JV		 3105 usvc->timeout = nla_get_u32(nla_timeout);
3106 usvc->netmask = nla_get_u32(nla_netmask);
3107 }
3108
3109 return 0;
3110}
3111
fc723250
HS		 3112static struct ip_vs_service *ip_vs_genl_find_service(struct net *net,
3113 struct nlattr *nla)
9a812198 3114{
c860c6b1 3115 struct ip_vs_service_user_kern usvc;
26c15cfd 3116 struct ip_vs_service *svc;
9a812198
JV		 3117 int ret;
3118
fc723250 3119 ret = ip_vs_genl_parse_service(net, &usvc, nla, 0, &svc);
26c15cfd 3120 return ret ? ERR_PTR(ret) : svc;
9a812198
JV		 3121}
3122
3123static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3124{
3125 struct nlattr *nl_dest;
3126
3127 nl_dest = nla_nest_start(skb, IPVS_CMD_ATTR_DEST);
3128 if (!nl_dest)
3129 return -EMSGSIZE;
3130
969e8e25
DM		 3131 if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3132 nla_put_u16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3133 nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3134 (atomic_read(&dest->conn_flags) &
3135 IP_VS_CONN_F_FWD_MASK)) ||
3136 nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3137 atomic_read(&dest->weight)) ||
3138 nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3139 nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3140 nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3141 atomic_read(&dest->activeconns)) ||
3142 nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3143 atomic_read(&dest->inactconns)) ||
3144 nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3145 atomic_read(&dest->persistconns)))
3146 goto nla_put_failure;
9a812198
JV		 3147 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &dest->stats))
3148 goto nla_put_failure;
3149
3150 nla_nest_end(skb, nl_dest);
3151
3152 return 0;
3153
3154nla_put_failure:
3155 nla_nest_cancel(skb, nl_dest);
3156 return -EMSGSIZE;
3157}
3158
3159static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3160 struct netlink_callback *cb)
3161{
3162 void *hdr;
3163
15e47304 3164 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
9a812198
JV		 3165 &ip_vs_genl_family, NLM_F_MULTI,
3166 IPVS_CMD_NEW_DEST);
3167 if (!hdr)
3168 return -EMSGSIZE;
3169
3170 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3171 goto nla_put_failure;
3172
3173 return genlmsg_end(skb, hdr);
3174
3175nla_put_failure:
3176 genlmsg_cancel(skb, hdr);
3177 return -EMSGSIZE;
3178}
3179
3180static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3181 struct netlink_callback *cb)
3182{
3183 int idx = 0;
3184 int start = cb->args[0];
3185 struct ip_vs_service *svc;
3186 struct ip_vs_dest *dest;
3187 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
a0840e2e 3188 struct net *net = skb_sknet(skb);
9a812198
JV		 3189
3190 mutex_lock(&__ip_vs_mutex);
3191
3192 /* Try to find the service for which to dump destinations */
3193 if (nlmsg_parse(cb->nlh, GENL_HDRLEN, attrs,
3194 IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy))
3195 goto out_err;
3196
a0840e2e 3197
fc723250 3198 svc = ip_vs_genl_find_service(net, attrs[IPVS_CMD_ATTR_SERVICE]);
9a812198
JV		 3199 if (IS_ERR(svc) || svc == NULL)
3200 goto out_err;
3201
3202 /* Dump the destinations */
3203 list_for_each_entry(dest, &svc->destinations, n_list) {
3204 if (++idx <= start)
3205 continue;
3206 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3207 idx--;
3208 goto nla_put_failure;
3209 }
3210 }
3211
3212nla_put_failure:
3213 cb->args[0] = idx;
9a812198
JV		 3214
3215out_err:
3216 mutex_unlock(&__ip_vs_mutex);
3217
3218 return skb->len;
3219}
3220
c860c6b1 3221static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
9a812198
JV		 3222 struct nlattr *nla, int full_entry)
3223{
3224 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3225 struct nlattr *nla_addr, *nla_port;
3226
3227 /* Parse mandatory identifying destination fields first */
3228 if (nla == NULL ||
3229 nla_parse_nested(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy))
3230 return -EINVAL;
3231
3232 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3233 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3234
3235 if (!(nla_addr && nla_port))
3236 return -EINVAL;
3237
258c8893
SH		 3238 memset(udest, 0, sizeof(*udest));
3239
9a812198
JV		 3240 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3241 udest->port = nla_get_u16(nla_port);
3242
3243 /* If a full entry was requested, check for the additional fields */
3244 if (full_entry) {
3245 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3246 *nla_l_thresh;
3247
3248 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3249 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3250 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3251 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3252
3253 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3254 return -EINVAL;
3255
3256 udest->conn_flags = nla_get_u32(nla_fwd)
3257 & IP_VS_CONN_F_FWD_MASK;
3258 udest->weight = nla_get_u32(nla_weight);
3259 udest->u_threshold = nla_get_u32(nla_u_thresh);
3260 udest->l_threshold = nla_get_u32(nla_l_thresh);
3261 }
3262
3263 return 0;
3264}
3265
3266static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __be32 state,
3267 const char *mcast_ifn, __be32 syncid)
3268{
3269 struct nlattr *nl_daemon;
3270
3271 nl_daemon = nla_nest_start(skb, IPVS_CMD_ATTR_DAEMON);
3272 if (!nl_daemon)
3273 return -EMSGSIZE;
3274
969e8e25
DM		 3275 if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3276 nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, mcast_ifn) ||
3277 nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, syncid))
3278 goto nla_put_failure;
9a812198
JV		 3279 nla_nest_end(skb, nl_daemon);
3280
3281 return 0;
3282
3283nla_put_failure:
3284 nla_nest_cancel(skb, nl_daemon);
3285 return -EMSGSIZE;
3286}
3287
3288static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __be32 state,
3289 const char *mcast_ifn, __be32 syncid,
3290 struct netlink_callback *cb)
3291{
3292 void *hdr;
15e47304 3293 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
9a812198
JV		 3294 &ip_vs_genl_family, NLM_F_MULTI,
3295 IPVS_CMD_NEW_DAEMON);
3296 if (!hdr)
3297 return -EMSGSIZE;
3298
3299 if (ip_vs_genl_fill_daemon(skb, state, mcast_ifn, syncid))
3300 goto nla_put_failure;
3301
3302 return genlmsg_end(skb, hdr);
3303
3304nla_put_failure:
3305 genlmsg_cancel(skb, hdr);
3306 return -EMSGSIZE;
3307}
3308
3309static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3310 struct netlink_callback *cb)
3311{
a09d1977 3312 struct net *net = skb_sknet(skb);
f131315f
HS		 3313 struct netns_ipvs *ipvs = net_ipvs(net);
3314
ae1d48b2 3315 mutex_lock(&ipvs->sync_mutex);
f131315f 3316 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
9a812198 3317 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
f131315f
HS		 3318 ipvs->master_mcast_ifn,
3319 ipvs->master_syncid, cb) < 0)
9a812198
JV		 3320 goto nla_put_failure;
3321
3322 cb->args[0] = 1;
3323 }
3324
f131315f 3325 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
9a812198 3326 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
f131315f
HS		 3327 ipvs->backup_mcast_ifn,
3328 ipvs->backup_syncid, cb) < 0)
9a812198
JV		 3329 goto nla_put_failure;
3330
3331 cb->args[1] = 1;
3332 }
3333
3334nla_put_failure:
ae1d48b2 3335 mutex_unlock(&ipvs->sync_mutex);
9a812198
JV		 3336
3337 return skb->len;
3338}
3339
f131315f 3340static int ip_vs_genl_new_daemon(struct net *net, struct nlattr **attrs)
9a812198
JV		 3341{
3342 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3343 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3344 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3345 return -EINVAL;
3346
f131315f
HS		 3347 return start_sync_thread(net,
3348 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]),
9a812198
JV		 3349 nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3350 nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]));
3351}
3352
f131315f 3353static int ip_vs_genl_del_daemon(struct net *net, struct nlattr **attrs)
9a812198
JV		 3354{
3355 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3356 return -EINVAL;
3357
f131315f
HS		 3358 return stop_sync_thread(net,
3359 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
9a812198
JV		 3360}
3361
9330419d 3362static int ip_vs_genl_set_config(struct net *net, struct nlattr **attrs)
9a812198
JV		 3363{
3364 struct ip_vs_timeout_user t;
3365
9330419d 3366 __ip_vs_get_timeouts(net, &t);
9a812198
JV		 3367
3368 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3369 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3370
3371 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3372 t.tcp_fin_timeout =
3373 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3374
3375 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3376 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3377
9330419d 3378 return ip_vs_set_timeout(net, &t);
9a812198
JV		 3379}
3380
ae1d48b2 3381static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
9a812198 3382{
9a812198 3383 int ret = 0, cmd;
fc723250 3384 struct net *net;
a0840e2e 3385 struct netns_ipvs *ipvs;
9a812198 3386
fc723250 3387 net = skb_sknet(skb);
a0840e2e 3388 ipvs = net_ipvs(net);
9a812198
JV		 3389 cmd = info->genlhdr->cmd;
3390
ae1d48b2 3391 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
9a812198
JV		 3392 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3393
ae1d48b2 3394 mutex_lock(&ipvs->sync_mutex);
9a812198
JV		 3395 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3396 nla_parse_nested(daemon_attrs, IPVS_DAEMON_ATTR_MAX,
3397 info->attrs[IPVS_CMD_ATTR_DAEMON],
3398 ip_vs_daemon_policy)) {
3399 ret = -EINVAL;
3400 goto out;
3401 }
3402
3403 if (cmd == IPVS_CMD_NEW_DAEMON)
f131315f 3404 ret = ip_vs_genl_new_daemon(net, daemon_attrs);
9a812198 3405 else
f131315f 3406 ret = ip_vs_genl_del_daemon(net, daemon_attrs);
ae1d48b2
HS		 3407out:
3408 mutex_unlock(&ipvs->sync_mutex);
3409 }
3410 return ret;
3411}
3412
3413static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3414{
3415 struct ip_vs_service *svc = NULL;
3416 struct ip_vs_service_user_kern usvc;
3417 struct ip_vs_dest_user_kern udest;
3418 int ret = 0, cmd;
3419 int need_full_svc = 0, need_full_dest = 0;
3420 struct net *net;
ae1d48b2
HS		 3421
3422 net = skb_sknet(skb);
ae1d48b2
HS		 3423 cmd = info->genlhdr->cmd;
3424
3425 mutex_lock(&__ip_vs_mutex);
3426
3427 if (cmd == IPVS_CMD_FLUSH) {
578bc3ef 3428 ret = ip_vs_flush(net, false);
ae1d48b2
HS		 3429 goto out;
3430 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3431 ret = ip_vs_genl_set_config(net, info->attrs);
9a812198
JV		 3432 goto out;
3433 } else if (cmd == IPVS_CMD_ZERO &&
3434 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
fc723250 3435 ret = ip_vs_zero_all(net);
9a812198
JV		 3436 goto out;
3437 }
3438
3439 /* All following commands require a service argument, so check if we
3440 * received a valid one. We need a full service specification when
3441 * adding / editing a service. Only identifying members otherwise. */
3442 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3443 need_full_svc = 1;
3444
fc723250 3445 ret = ip_vs_genl_parse_service(net, &usvc,
9a812198 3446 info->attrs[IPVS_CMD_ATTR_SERVICE],
26c15cfd 3447 need_full_svc, &svc);
9a812198
JV		 3448 if (ret)
3449 goto out;
3450
9a812198
JV		 3451 /* Unless we're adding a new service, the service must already exist */
3452 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3453 ret = -ESRCH;
3454 goto out;
3455 }
3456
3457 /* Destination commands require a valid destination argument. For
3458 * adding / editing a destination, we need a full destination
3459 * specification. */
3460 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3461 cmd == IPVS_CMD_DEL_DEST) {
3462 if (cmd != IPVS_CMD_DEL_DEST)
3463 need_full_dest = 1;
3464
3465 ret = ip_vs_genl_parse_dest(&udest,
3466 info->attrs[IPVS_CMD_ATTR_DEST],
3467 need_full_dest);
3468 if (ret)
3469 goto out;
3470 }
3471
3472 switch (cmd) {
3473 case IPVS_CMD_NEW_SERVICE:
3474 if (svc == NULL)
fc723250 3475 ret = ip_vs_add_service(net, &usvc, &svc);
9a812198
JV		 3476 else
3477 ret = -EEXIST;
3478 break;
3479 case IPVS_CMD_SET_SERVICE:
3480 ret = ip_vs_edit_service(svc, &usvc);
3481 break;
3482 case IPVS_CMD_DEL_SERVICE:
3483 ret = ip_vs_del_service(svc);
26c15cfd 3484 /* do not use svc, it can be freed */
9a812198
JV		 3485 break;
3486 case IPVS_CMD_NEW_DEST:
3487 ret = ip_vs_add_dest(svc, &udest);
3488 break;
3489 case IPVS_CMD_SET_DEST:
3490 ret = ip_vs_edit_dest(svc, &udest);
3491 break;
3492 case IPVS_CMD_DEL_DEST:
3493 ret = ip_vs_del_dest(svc, &udest);
3494 break;
3495 case IPVS_CMD_ZERO:
3496 ret = ip_vs_zero_service(svc);
3497 break;
3498 default:
3499 ret = -EINVAL;
3500 }
3501
3502out:
9a812198
JV		 3503 mutex_unlock(&__ip_vs_mutex);
3504
3505 return ret;
3506}
3507
3508static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3509{
3510 struct sk_buff *msg;
3511 void *reply;
3512 int ret, cmd, reply_cmd;
fc723250 3513 struct net *net;
9a812198 3514
fc723250 3515 net = skb_sknet(skb);
9a812198
JV		 3516 cmd = info->genlhdr->cmd;
3517
3518 if (cmd == IPVS_CMD_GET_SERVICE)
3519 reply_cmd = IPVS_CMD_NEW_SERVICE;
3520 else if (cmd == IPVS_CMD_GET_INFO)
3521 reply_cmd = IPVS_CMD_SET_INFO;
3522 else if (cmd == IPVS_CMD_GET_CONFIG)
3523 reply_cmd = IPVS_CMD_SET_CONFIG;
3524 else {
1e3e238e 3525 pr_err("unknown Generic Netlink command\n");
9a812198
JV		 3526 return -EINVAL;
3527 }
3528
3529 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3530 if (!msg)
3531 return -ENOMEM;
3532
3533 mutex_lock(&__ip_vs_mutex);
3534
3535 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3536 if (reply == NULL)
3537 goto nla_put_failure;
3538
3539 switch (cmd) {
3540 case IPVS_CMD_GET_SERVICE:
3541 {
3542 struct ip_vs_service *svc;
3543
fc723250
HS		 3544 svc = ip_vs_genl_find_service(net,
3545 info->attrs[IPVS_CMD_ATTR_SERVICE]);
9a812198
JV		 3546 if (IS_ERR(svc)) {
3547 ret = PTR_ERR(svc);
3548 goto out_err;
3549 } else if (svc) {
3550 ret = ip_vs_genl_fill_service(msg, svc);
9a812198
JV		 3551 if (ret)
3552 goto nla_put_failure;
3553 } else {
3554 ret = -ESRCH;
3555 goto out_err;
3556 }
3557
3558 break;
3559 }
3560
3561 case IPVS_CMD_GET_CONFIG:
3562 {
3563 struct ip_vs_timeout_user t;
3564
9330419d 3565 __ip_vs_get_timeouts(net, &t);
9a812198 3566#ifdef CONFIG_IP_VS_PROTO_TCP
969e8e25
DM		 3567 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3568 t.tcp_timeout) ||
3569 nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3570 t.tcp_fin_timeout))
3571 goto nla_put_failure;
9a812198
JV		 3572#endif
3573#ifdef CONFIG_IP_VS_PROTO_UDP
969e8e25
DM		 3574 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3575 goto nla_put_failure;
9a812198
JV		 3576#endif
3577
3578 break;
3579 }
3580
3581 case IPVS_CMD_GET_INFO:
969e8e25
DM		 3582 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3583 IP_VS_VERSION_CODE) ||
3584 nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3585 ip_vs_conn_tab_size))
3586 goto nla_put_failure;
9a812198
JV		 3587 break;
3588 }
3589
3590 genlmsg_end(msg, reply);
134e6375 3591 ret = genlmsg_reply(msg, info);
9a812198
JV		 3592 goto out;
3593
3594nla_put_failure:
1e3e238e 3595 pr_err("not enough space in Netlink message\n");
9a812198
JV		 3596 ret = -EMSGSIZE;
3597
3598out_err:
3599 nlmsg_free(msg);
3600out:
3601 mutex_unlock(&__ip_vs_mutex);
3602
3603 return ret;
3604}
3605
3606
3607static struct genl_ops ip_vs_genl_ops[] __read_mostly = {
3608 {
3609 .cmd = IPVS_CMD_NEW_SERVICE,
3610 .flags = GENL_ADMIN_PERM,
3611 .policy = ip_vs_cmd_policy,
3612 .doit = ip_vs_genl_set_cmd,
3613 },
3614 {
3615 .cmd = IPVS_CMD_SET_SERVICE,
3616 .flags = GENL_ADMIN_PERM,
3617 .policy = ip_vs_cmd_policy,
3618 .doit = ip_vs_genl_set_cmd,
3619 },
3620 {
3621 .cmd = IPVS_CMD_DEL_SERVICE,
3622 .flags = GENL_ADMIN_PERM,
3623 .policy = ip_vs_cmd_policy,
3624 .doit = ip_vs_genl_set_cmd,
3625 },
3626 {
3627 .cmd = IPVS_CMD_GET_SERVICE,
3628 .flags = GENL_ADMIN_PERM,
3629 .doit = ip_vs_genl_get_cmd,
3630 .dumpit = ip_vs_genl_dump_services,
3631 .policy = ip_vs_cmd_policy,
3632 },
3633 {
3634 .cmd = IPVS_CMD_NEW_DEST,
3635 .flags = GENL_ADMIN_PERM,
3636 .policy = ip_vs_cmd_policy,
3637 .doit = ip_vs_genl_set_cmd,
3638 },
3639 {
3640 .cmd = IPVS_CMD_SET_DEST,
3641 .flags = GENL_ADMIN_PERM,
3642 .policy = ip_vs_cmd_policy,
3643 .doit = ip_vs_genl_set_cmd,
3644 },
3645 {
3646 .cmd = IPVS_CMD_DEL_DEST,
3647 .flags = GENL_ADMIN_PERM,
3648 .policy = ip_vs_cmd_policy,
3649 .doit = ip_vs_genl_set_cmd,
3650 },
3651 {
3652 .cmd = IPVS_CMD_GET_DEST,
3653 .flags = GENL_ADMIN_PERM,
3654 .policy = ip_vs_cmd_policy,
3655 .dumpit = ip_vs_genl_dump_dests,
3656 },
3657 {
3658 .cmd = IPVS_CMD_NEW_DAEMON,
3659 .flags = GENL_ADMIN_PERM,
3660 .policy = ip_vs_cmd_policy,
ae1d48b2 3661 .doit = ip_vs_genl_set_daemon,
9a812198
JV		 3662 },
3663 {
3664 .cmd = IPVS_CMD_DEL_DAEMON,
3665 .flags = GENL_ADMIN_PERM,
3666 .policy = ip_vs_cmd_policy,
ae1d48b2 3667 .doit = ip_vs_genl_set_daemon,
9a812198
JV		 3668 },
3669 {
3670 .cmd = IPVS_CMD_GET_DAEMON,
3671 .flags = GENL_ADMIN_PERM,
3672 .dumpit = ip_vs_genl_dump_daemons,
3673 },
3674 {
3675 .cmd = IPVS_CMD_SET_CONFIG,
3676 .flags = GENL_ADMIN_PERM,
3677 .policy = ip_vs_cmd_policy,
3678 .doit = ip_vs_genl_set_cmd,
3679 },
3680 {
3681 .cmd = IPVS_CMD_GET_CONFIG,
3682 .flags = GENL_ADMIN_PERM,
3683 .doit = ip_vs_genl_get_cmd,
3684 },
3685 {
3686 .cmd = IPVS_CMD_GET_INFO,
3687 .flags = GENL_ADMIN_PERM,
3688 .doit = ip_vs_genl_get_cmd,
3689 },
3690 {
3691 .cmd = IPVS_CMD_ZERO,
3692 .flags = GENL_ADMIN_PERM,
3693 .policy = ip_vs_cmd_policy,
3694 .doit = ip_vs_genl_set_cmd,
3695 },
3696 {
3697 .cmd = IPVS_CMD_FLUSH,
3698 .flags = GENL_ADMIN_PERM,
3699 .doit = ip_vs_genl_set_cmd,
3700 },
3701};
3702
3703static int __init ip_vs_genl_register(void)
3704{
8f698d54
MM		 3705 return genl_register_family_with_ops(&ip_vs_genl_family,
3706 ip_vs_genl_ops, ARRAY_SIZE(ip_vs_genl_ops));
9a812198
JV		 3707}
3708
3709static void ip_vs_genl_unregister(void)
3710{
3711 genl_unregister_family(&ip_vs_genl_family);
3712}
3713
3714/* End of Generic Netlink interface definitions */
3715
61b1ab45
HS		 3716/*
3717 * per netns intit/exit func.
3718 */
14e40546 3719#ifdef CONFIG_SYSCTL
2b2d2808 3720static int __net_init ip_vs_control_net_init_sysctl(struct net *net)
61b1ab45 3721{
fc723250
HS		 3722 int idx;
3723 struct netns_ipvs *ipvs = net_ipvs(net);
a0840e2e 3724 struct ctl_table *tbl;
fc723250 3725
a0840e2e
HS		 3726 atomic_set(&ipvs->dropentry, 0);
3727 spin_lock_init(&ipvs->dropentry_lock);
3728 spin_lock_init(&ipvs->droppacket_lock);
3729 spin_lock_init(&ipvs->securetcp_lock);
a0840e2e
HS		 3730
3731 if (!net_eq(net, &init_net)) {
3732 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
3733 if (tbl == NULL)
14e40546 3734 return -ENOMEM;
464dc801
EB		 3735
3736 /* Don't export sysctls to unprivileged users */
3737 if (net->user_ns != &init_user_ns)
3738 tbl[0].procname = NULL;
a0840e2e
HS		 3739 } else
3740 tbl = vs_vars;
3741 /* Initialize sysctl defaults */
3742 idx = 0;
3743 ipvs->sysctl_amemthresh = 1024;
3744 tbl[idx++].data = &ipvs->sysctl_amemthresh;
3745 ipvs->sysctl_am_droprate = 10;
3746 tbl[idx++].data = &ipvs->sysctl_am_droprate;
3747 tbl[idx++].data = &ipvs->sysctl_drop_entry;
3748 tbl[idx++].data = &ipvs->sysctl_drop_packet;
3749#ifdef CONFIG_IP_VS_NFCT
3750 tbl[idx++].data = &ipvs->sysctl_conntrack;
3751#endif
3752 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
3753 ipvs->sysctl_snat_reroute = 1;
3754 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
3755 ipvs->sysctl_sync_ver = 1;
3756 tbl[idx++].data = &ipvs->sysctl_sync_ver;
f73181c8
PNA		 3757 ipvs->sysctl_sync_ports = 1;
3758 tbl[idx++].data = &ipvs->sysctl_sync_ports;
1c003b15
PNA		 3759 ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
3760 tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
3761 ipvs->sysctl_sync_sock_size = 0;
3762 tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
a0840e2e
HS		 3763 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
3764 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
3765 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
59e0350e
SH		 3766 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
3767 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
a0840e2e
HS		 3768 tbl[idx].data = &ipvs->sysctl_sync_threshold;
3769 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
749c42b6
JA		 3770 ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
3771 tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
3772 ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
3773 tbl[idx++].data = &ipvs->sysctl_sync_retries;
a0840e2e 3774 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
3654e611
JA		 3775 ipvs->sysctl_pmtu_disc = 1;
3776 tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
0c12582f 3777 tbl[idx++].data = &ipvs->sysctl_backup_only;
a0840e2e
HS		 3778
3779
ec8f23ce 3780 ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
0443929f
SH		 3781 if (ipvs->sysctl_hdr == NULL) {
3782 if (!net_eq(net, &init_net))
3783 kfree(tbl);
14e40546 3784 return -ENOMEM;
0443929f 3785 }
6ef757f9 3786 ip_vs_start_estimator(net, &ipvs->tot_stats);
a0840e2e 3787 ipvs->sysctl_tbl = tbl;
f6340ee0
HS		 3788 /* Schedule defense work */
3789 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
3790 schedule_delayed_work(&ipvs->defense_work, DEFENSE_TIMER_PERIOD);
61b1ab45 3791
61b1ab45 3792 return 0;
61b1ab45
HS		 3793}
3794
2b2d2808 3795static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net)
61b1ab45 3796{
b17fc996
HS		 3797 struct netns_ipvs *ipvs = net_ipvs(net);
3798
f2431e6e
HS		 3799 cancel_delayed_work_sync(&ipvs->defense_work);
3800 cancel_work_sync(&ipvs->defense_work.work);
a0840e2e 3801 unregister_net_sysctl_table(ipvs->sysctl_hdr);
14e40546
SH		 3802}
3803
3804#else
3805
2b2d2808
CG		 3806static int __net_init ip_vs_control_net_init_sysctl(struct net *net) { return 0; }
3807static void __net_exit ip_vs_control_net_cleanup_sysctl(struct net *net) { }
14e40546 3808
0443929f 3809#endif
14e40546 3810
7a4f0761
HS		 3811static struct notifier_block ip_vs_dst_notifier = {
3812 .notifier_call = ip_vs_dst_event,
3813};
3814
503cf15a 3815int __net_init ip_vs_control_net_init(struct net *net)
14e40546
SH		 3816{
3817 int idx;
3818 struct netns_ipvs *ipvs = net_ipvs(net);
3819
14e40546
SH		 3820 /* Initialize rs_table */
3821 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
276472ea 3822 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
14e40546
SH		 3823
3824 INIT_LIST_HEAD(&ipvs->dest_trash);
578bc3ef
JA		 3825 spin_lock_init(&ipvs->dest_trash_lock);
3826 setup_timer(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire,
3827 (unsigned long) net);
14e40546
SH		 3828 atomic_set(&ipvs->ftpsvc_counter, 0);
3829 atomic_set(&ipvs->nullsvc_counter, 0);
3830
3831 /* procfs stats */
3832 ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
0a9ee813 3833 if (!ipvs->tot_stats.cpustats)
14e40546 3834 return -ENOMEM;
0a9ee813 3835
14e40546
SH		 3836 spin_lock_init(&ipvs->tot_stats.lock);
3837
d4beaa66
G		 3838 proc_create("ip_vs", 0, net->proc_net, &ip_vs_info_fops);
3839 proc_create("ip_vs_stats", 0, net->proc_net, &ip_vs_stats_fops);
3840 proc_create("ip_vs_stats_percpu", 0, net->proc_net,
3841 &ip_vs_stats_percpu_fops);
14e40546 3842
503cf15a 3843 if (ip_vs_control_net_init_sysctl(net))
14e40546
SH		 3844 goto err;
3845
3846 return 0;
3847
3848err:
2a0751af 3849 free_percpu(ipvs->tot_stats.cpustats);
61b1ab45
HS		 3850 return -ENOMEM;
3851}
3852
503cf15a 3853void __net_exit ip_vs_control_net_cleanup(struct net *net)
61b1ab45 3854{
b17fc996
HS		 3855 struct netns_ipvs *ipvs = net_ipvs(net);
3856
578bc3ef
JA		 3857 /* Some dest can be in grace period even before cleanup, we have to
3858 * defer ip_vs_trash_cleanup until ip_vs_dest_wait_readers is called.
3859 */
3860 rcu_barrier();
f2431e6e 3861 ip_vs_trash_cleanup(net);
6ef757f9 3862 ip_vs_stop_estimator(net, &ipvs->tot_stats);
503cf15a 3863 ip_vs_control_net_cleanup_sysctl(net);
ece31ffd
G		 3864 remove_proc_entry("ip_vs_stats_percpu", net->proc_net);
3865 remove_proc_entry("ip_vs_stats", net->proc_net);
3866 remove_proc_entry("ip_vs", net->proc_net);
2a0751af 3867 free_percpu(ipvs->tot_stats.cpustats);
61b1ab45
HS		 3868}
3869
8537de8a 3870int __init ip_vs_register_nl_ioctl(void)
1da177e4 3871{
fc723250 3872 int ret;
1da177e4 3873
1da177e4
LT		 3874 ret = nf_register_sockopt(&ip_vs_sockopts);
3875 if (ret) {
1e3e238e 3876 pr_err("cannot register sockopt.\n");
7a4f0761 3877 goto err_sock;
1da177e4
LT		 3878 }
3879
9a812198
JV		 3880 ret = ip_vs_genl_register();
3881 if (ret) {
1e3e238e 3882 pr_err("cannot register Generic Netlink interface.\n");
7a4f0761 3883 goto err_genl;
9a812198 3884 }
1da177e4 3885 return 0;
fc723250 3886
7a4f0761
HS		 3887err_genl:
3888 nf_unregister_sockopt(&ip_vs_sockopts);
3889err_sock:
fc723250 3890 return ret;
1da177e4
LT		 3891}
3892
8537de8a
HS		 3893void ip_vs_unregister_nl_ioctl(void)
3894{
3895 ip_vs_genl_unregister();
3896 nf_unregister_sockopt(&ip_vs_sockopts);
3897}
3898
3899int __init ip_vs_control_init(void)
3900{
3901 int idx;
3902 int ret;
3903
3904 EnterFunction(2);
3905
276472ea 3906 /* Initialize svc_table, ip_vs_svc_fwm_table */
8537de8a
HS		 3907 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
3908 INIT_LIST_HEAD(&ip_vs_svc_table[idx]);
3909 INIT_LIST_HEAD(&ip_vs_svc_fwm_table[idx]);
3910 }
3911
3912 smp_wmb(); /* Do we really need it now ? */
3913
3914 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
3915 if (ret < 0)
3916 return ret;
3917
3918 LeaveFunction(2);
3919 return 0;
3920}
3921
1da177e4
LT		 3922
3923void ip_vs_control_cleanup(void)
3924{
3925 EnterFunction(2);
7676e345 3926 unregister_netdevice_notifier(&ip_vs_dst_notifier);
1da177e4
LT		 3927 LeaveFunction(2);
3928}
Linux kernel - Deliverables
This page took 1.022909 seconds and 5 git commands to generate.