1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * IPVS An implementation of the IP virtual server support for the
4 * LINUX operating system. IPVS is now implemented as a module
5 * over the NetFilter framework. IPVS can be used to build a
6 * high-performance and highly available server based on a
7 * cluster of servers.
8 *
9 * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
10 * Peter Kese <peter.kese@ijs.si>
11 * Julian Anastasov <ja@ssi.bg>
12 *
13 * Changes:
14 */
15
16 #define KMSG_COMPONENT "IPVS"
17 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
18
19 #include <linux/module.h>
20 #include <linux/init.h>
21 #include <linux/types.h>
22 #include <linux/capability.h>
23 #include <linux/fs.h>
24 #include <linux/sysctl.h>
25 #include <linux/proc_fs.h>
26 #include <linux/workqueue.h>
27 #include <linux/seq_file.h>
28 #include <linux/slab.h>
29
30 #include <linux/netfilter.h>
31 #include <linux/netfilter_ipv4.h>
32 #include <linux/mutex.h>
33
34 #include <net/net_namespace.h>
35 #include <linux/nsproxy.h>
36 #include <net/ip.h>
37 #ifdef CONFIG_IP_VS_IPV6
38 #include <net/ipv6.h>
39 #include <net/ip6_route.h>
40 #include <net/netfilter/ipv6/nf_defrag_ipv6.h>
41 #endif
42 #include <net/route.h>
43 #include <net/sock.h>
44 #include <net/genetlink.h>
45
46 #include <linux/uaccess.h>
47
48 #include <net/ip_vs.h>
49
50 MODULE_ALIAS_GENL_FAMILY(IPVS_GENL_NAME);
51
52 DEFINE_MUTEX(__ip_vs_mutex); /* Serialize configuration with sockopt/netlink */
53
54 /* sysctl variables */
55
56 #ifdef CONFIG_IP_VS_DEBUG
57 static int sysctl_ip_vs_debug_level = 0;
58
ip_vs_get_debug_level(void)59 int ip_vs_get_debug_level(void)
60 {
61 return sysctl_ip_vs_debug_level;
62 }
63 #endif
64
65
66 /* Protos */
67 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup);
68
69
70 #ifdef CONFIG_IP_VS_IPV6
71 /* Taken from rt6_fill_node() in net/ipv6/route.c, is there a better way? */
__ip_vs_addr_is_local_v6(struct net * net,const struct in6_addr * addr)72 static bool __ip_vs_addr_is_local_v6(struct net *net,
73 const struct in6_addr *addr)
74 {
75 struct flowi6 fl6 = {
76 .daddr = *addr,
77 };
78 struct dst_entry *dst = ip6_route_output(net, NULL, &fl6);
79 bool is_local;
80
81 is_local = !dst->error && dst->dev && (dst->dev->flags & IFF_LOOPBACK);
82
83 dst_release(dst);
84 return is_local;
85 }
86 #endif
87
88 #ifdef CONFIG_SYSCTL
89 /*
90 * update_defense_level is called from keventd and from sysctl,
91 * so it needs to protect itself from softirqs
92 */
update_defense_level(struct netns_ipvs * ipvs)93 static void update_defense_level(struct netns_ipvs *ipvs)
94 {
95 struct sysinfo i;
96 int availmem;
97 int amemthresh;
98 int nomem;
99 int to_change = -1;
100
101 /* we only count free and buffered memory (in pages) */
102 si_meminfo(&i);
103 availmem = i.freeram + i.bufferram;
104 /* however in linux 2.5 the i.bufferram is total page cache size,
105 we need adjust it */
106 /* si_swapinfo(&i); */
107 /* availmem = availmem - (i.totalswap - i.freeswap); */
108
109 amemthresh = max(READ_ONCE(ipvs->sysctl_amemthresh), 0);
110 nomem = (availmem < amemthresh);
111
112 local_bh_disable();
113
114 /* drop_entry */
115 spin_lock(&ipvs->dropentry_lock);
116 switch (ipvs->sysctl_drop_entry) {
117 case 0:
118 atomic_set(&ipvs->dropentry, 0);
119 break;
120 case 1:
121 if (nomem) {
122 atomic_set(&ipvs->dropentry, 1);
123 ipvs->sysctl_drop_entry = 2;
124 } else {
125 atomic_set(&ipvs->dropentry, 0);
126 }
127 break;
128 case 2:
129 if (nomem) {
130 atomic_set(&ipvs->dropentry, 1);
131 } else {
132 atomic_set(&ipvs->dropentry, 0);
133 ipvs->sysctl_drop_entry = 1;
134 }
135 break;
136 case 3:
137 atomic_set(&ipvs->dropentry, 1);
138 break;
139 }
140 spin_unlock(&ipvs->dropentry_lock);
141
142 /* drop_packet */
143 spin_lock(&ipvs->droppacket_lock);
144 switch (ipvs->sysctl_drop_packet) {
145 case 0:
146 ipvs->drop_rate = 0;
147 break;
148 case 1:
149 if (nomem) {
150 ipvs->drop_counter = amemthresh / (amemthresh - availmem);
151 ipvs->drop_rate = ipvs->drop_counter;
152 ipvs->sysctl_drop_packet = 2;
153 } else {
154 ipvs->drop_rate = 0;
155 }
156 break;
157 case 2:
158 if (nomem) {
159 ipvs->drop_counter = amemthresh / (amemthresh - availmem);
160 ipvs->drop_rate = ipvs->drop_counter;
161 } else {
162 ipvs->drop_rate = 0;
163 ipvs->sysctl_drop_packet = 1;
164 }
165 break;
166 case 3:
167 ipvs->drop_rate = ipvs->sysctl_am_droprate;
168 break;
169 }
170 spin_unlock(&ipvs->droppacket_lock);
171
172 /* secure_tcp */
173 spin_lock(&ipvs->securetcp_lock);
174 switch (ipvs->sysctl_secure_tcp) {
175 case 0:
176 if (ipvs->old_secure_tcp >= 2)
177 to_change = 0;
178 break;
179 case 1:
180 if (nomem) {
181 if (ipvs->old_secure_tcp < 2)
182 to_change = 1;
183 ipvs->sysctl_secure_tcp = 2;
184 } else {
185 if (ipvs->old_secure_tcp >= 2)
186 to_change = 0;
187 }
188 break;
189 case 2:
190 if (nomem) {
191 if (ipvs->old_secure_tcp < 2)
192 to_change = 1;
193 } else {
194 if (ipvs->old_secure_tcp >= 2)
195 to_change = 0;
196 ipvs->sysctl_secure_tcp = 1;
197 }
198 break;
199 case 3:
200 if (ipvs->old_secure_tcp < 2)
201 to_change = 1;
202 break;
203 }
204 ipvs->old_secure_tcp = ipvs->sysctl_secure_tcp;
205 if (to_change >= 0)
206 ip_vs_protocol_timeout_change(ipvs,
207 ipvs->sysctl_secure_tcp > 1);
208 spin_unlock(&ipvs->securetcp_lock);
209
210 local_bh_enable();
211 }
212
213 /* Handler for delayed work for expiring no
214 * destination connections
215 */
expire_nodest_conn_handler(struct work_struct * work)216 static void expire_nodest_conn_handler(struct work_struct *work)
217 {
218 struct netns_ipvs *ipvs;
219
220 ipvs = container_of(work, struct netns_ipvs,
221 expire_nodest_conn_work.work);
222 ip_vs_expire_nodest_conn_flush(ipvs);
223 }
224
225 /*
226 * Timer for checking the defense
227 */
228 #define DEFENSE_TIMER_PERIOD 1*HZ
229
defense_work_handler(struct work_struct * work)230 static void defense_work_handler(struct work_struct *work)
231 {
232 struct netns_ipvs *ipvs =
233 container_of(work, struct netns_ipvs, defense_work.work);
234
235 update_defense_level(ipvs);
236 if (atomic_read(&ipvs->dropentry))
237 ip_vs_random_dropentry(ipvs);
238 queue_delayed_work(system_long_wq, &ipvs->defense_work,
239 DEFENSE_TIMER_PERIOD);
240 }
241 #endif
242
est_reload_work_handler(struct work_struct * work)243 static void est_reload_work_handler(struct work_struct *work)
244 {
245 struct netns_ipvs *ipvs =
246 container_of(work, struct netns_ipvs, est_reload_work.work);
247 int genid_done = atomic_read(&ipvs->est_genid_done);
248 unsigned long delay = HZ / 10; /* repeat startups after failure */
249 bool repeat = false;
250 int genid;
251 int id;
252
253 mutex_lock(&ipvs->est_mutex);
254 genid = atomic_read(&ipvs->est_genid);
255 for (id = 0; id < ipvs->est_kt_count; id++) {
256 struct ip_vs_est_kt_data *kd = ipvs->est_kt_arr[id];
257
258 /* netns clean up started, abort delayed work */
259 if (!ipvs->enable)
260 goto unlock;
261 if (!kd)
262 continue;
263 /* New config ? Stop kthread tasks */
264 if (genid != genid_done)
265 ip_vs_est_kthread_stop(kd);
266 if (!kd->task && !ip_vs_est_stopped(ipvs)) {
267 /* Do not start kthreads above 0 in calc phase */
268 if ((!id || !ipvs->est_calc_phase) &&
269 ip_vs_est_kthread_start(ipvs, kd) < 0)
270 repeat = true;
271 }
272 }
273
274 atomic_set(&ipvs->est_genid_done, genid);
275
276 if (repeat)
277 queue_delayed_work(system_long_wq, &ipvs->est_reload_work,
278 delay);
279
280 unlock:
281 mutex_unlock(&ipvs->est_mutex);
282 }
283
284 int
ip_vs_use_count_inc(void)285 ip_vs_use_count_inc(void)
286 {
287 return try_module_get(THIS_MODULE);
288 }
289
290 void
ip_vs_use_count_dec(void)291 ip_vs_use_count_dec(void)
292 {
293 module_put(THIS_MODULE);
294 }
295
296
297 /*
298 * Hash table: for virtual service lookups
299 */
300 #define IP_VS_SVC_TAB_BITS 8
301 #define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
302 #define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)
303
304 /* the service table hashed by <protocol, addr, port> */
305 static struct hlist_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE];
306 /* the service table hashed by fwmark */
307 static struct hlist_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE];
308
309
310 /*
311 * Returns hash value for virtual service
312 */
313 static inline unsigned int
ip_vs_svc_hashkey(struct netns_ipvs * ipvs,int af,unsigned int proto,const union nf_inet_addr * addr,__be16 port)314 ip_vs_svc_hashkey(struct netns_ipvs *ipvs, int af, unsigned int proto,
315 const union nf_inet_addr *addr, __be16 port)
316 {
317 unsigned int porth = ntohs(port);
318 __be32 addr_fold = addr->ip;
319 __u32 ahash;
320
321 #ifdef CONFIG_IP_VS_IPV6
322 if (af == AF_INET6)
323 addr_fold = addr->ip6[0]^addr->ip6[1]^
324 addr->ip6[2]^addr->ip6[3];
325 #endif
326 ahash = ntohl(addr_fold);
327 ahash ^= ((size_t) ipvs >> 8);
328
329 return (proto ^ ahash ^ (porth >> IP_VS_SVC_TAB_BITS) ^ porth) &
330 IP_VS_SVC_TAB_MASK;
331 }
332
333 /*
334 * Returns hash value of fwmark for virtual service lookup
335 */
ip_vs_svc_fwm_hashkey(struct netns_ipvs * ipvs,__u32 fwmark)336 static inline unsigned int ip_vs_svc_fwm_hashkey(struct netns_ipvs *ipvs, __u32 fwmark)
337 {
338 return (((size_t)ipvs>>8) ^ fwmark) & IP_VS_SVC_TAB_MASK;
339 }
340
341 /*
342 * Hashes a service in the ip_vs_svc_table by <netns,proto,addr,port>
343 * or in the ip_vs_svc_fwm_table by fwmark.
344 * Should be called with locked tables.
345 */
ip_vs_svc_hash(struct ip_vs_service * svc)346 static int ip_vs_svc_hash(struct ip_vs_service *svc)
347 {
348 unsigned int hash;
349
350 if (svc->flags & IP_VS_SVC_F_HASHED) {
351 pr_err("%s(): request for already hashed, called from %pS\n",
352 __func__, __builtin_return_address(0));
353 return 0;
354 }
355
356 if (svc->fwmark == 0) {
357 /*
358 * Hash it by <netns,protocol,addr,port> in ip_vs_svc_table
359 */
360 hash = ip_vs_svc_hashkey(svc->ipvs, svc->af, svc->protocol,
361 &svc->addr, svc->port);
362 hlist_add_head_rcu(&svc->s_list, &ip_vs_svc_table[hash]);
363 } else {
364 /*
365 * Hash it by fwmark in svc_fwm_table
366 */
367 hash = ip_vs_svc_fwm_hashkey(svc->ipvs, svc->fwmark);
368 hlist_add_head_rcu(&svc->f_list, &ip_vs_svc_fwm_table[hash]);
369 }
370
371 svc->flags |= IP_VS_SVC_F_HASHED;
372 /* increase its refcnt because it is referenced by the svc table */
373 atomic_inc(&svc->refcnt);
374 return 1;
375 }
376
377
378 /*
379 * Unhashes a service from svc_table / svc_fwm_table.
380 * Should be called with locked tables.
381 */
ip_vs_svc_unhash(struct ip_vs_service * svc)382 static int ip_vs_svc_unhash(struct ip_vs_service *svc)
383 {
384 if (!(svc->flags & IP_VS_SVC_F_HASHED)) {
385 pr_err("%s(): request for unhash flagged, called from %pS\n",
386 __func__, __builtin_return_address(0));
387 return 0;
388 }
389
390 if (svc->fwmark == 0) {
391 /* Remove it from the svc_table table */
392 hlist_del_rcu(&svc->s_list);
393 } else {
394 /* Remove it from the svc_fwm_table table */
395 hlist_del_rcu(&svc->f_list);
396 }
397
398 svc->flags &= ~IP_VS_SVC_F_HASHED;
399 atomic_dec(&svc->refcnt);
400 return 1;
401 }
402
403
404 /*
405 * Get service by {netns, proto,addr,port} in the service table.
406 */
407 static inline struct ip_vs_service *
__ip_vs_service_find(struct netns_ipvs * ipvs,int af,__u16 protocol,const union nf_inet_addr * vaddr,__be16 vport)408 __ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u16 protocol,
409 const union nf_inet_addr *vaddr, __be16 vport)
410 {
411 unsigned int hash;
412 struct ip_vs_service *svc;
413
414 /* Check for "full" addressed entries */
415 hash = ip_vs_svc_hashkey(ipvs, af, protocol, vaddr, vport);
416
417 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[hash], s_list) {
418 if ((svc->af == af)
419 && ip_vs_addr_equal(af, &svc->addr, vaddr)
420 && (svc->port == vport)
421 && (svc->protocol == protocol)
422 && (svc->ipvs == ipvs)) {
423 /* HIT */
424 return svc;
425 }
426 }
427
428 return NULL;
429 }
430
431
432 /*
433 * Get service by {fwmark} in the service table.
434 */
435 static inline struct ip_vs_service *
__ip_vs_svc_fwm_find(struct netns_ipvs * ipvs,int af,__u32 fwmark)436 __ip_vs_svc_fwm_find(struct netns_ipvs *ipvs, int af, __u32 fwmark)
437 {
438 unsigned int hash;
439 struct ip_vs_service *svc;
440
441 /* Check for fwmark addressed entries */
442 hash = ip_vs_svc_fwm_hashkey(ipvs, fwmark);
443
444 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[hash], f_list) {
445 if (svc->fwmark == fwmark && svc->af == af
446 && (svc->ipvs == ipvs)) {
447 /* HIT */
448 return svc;
449 }
450 }
451
452 return NULL;
453 }
454
455 /* Find service, called under RCU lock */
456 struct ip_vs_service *
ip_vs_service_find(struct netns_ipvs * ipvs,int af,__u32 fwmark,__u16 protocol,const union nf_inet_addr * vaddr,__be16 vport)457 ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
458 const union nf_inet_addr *vaddr, __be16 vport)
459 {
460 struct ip_vs_service *svc;
461
462 /*
463 * Check the table hashed by fwmark first
464 */
465 if (fwmark) {
466 svc = __ip_vs_svc_fwm_find(ipvs, af, fwmark);
467 if (svc)
468 goto out;
469 }
470
471 /*
472 * Check the table hashed by <protocol,addr,port>
473 * for "full" addressed entries
474 */
475 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, vport);
476
477 if (!svc && protocol == IPPROTO_TCP &&
478 atomic_read(&ipvs->ftpsvc_counter) &&
479 (vport == FTPDATA || !inet_port_requires_bind_service(ipvs->net, ntohs(vport)))) {
480 /*
481 * Check if ftp service entry exists, the packet
482 * might belong to FTP data connections.
483 */
484 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, FTPPORT);
485 }
486
487 if (svc == NULL
488 && atomic_read(&ipvs->nullsvc_counter)) {
489 /*
490 * Check if the catch-all port (port zero) exists
491 */
492 svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, 0);
493 }
494
495 out:
496 IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
497 fwmark, ip_vs_proto_name(protocol),
498 IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
499 svc ? "hit" : "not hit");
500
501 return svc;
502 }
503
504
505 static inline void
__ip_vs_bind_svc(struct ip_vs_dest * dest,struct ip_vs_service * svc)506 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
507 {
508 atomic_inc(&svc->refcnt);
509 rcu_assign_pointer(dest->svc, svc);
510 }
511
ip_vs_service_free(struct ip_vs_service * svc)512 static void ip_vs_service_free(struct ip_vs_service *svc)
513 {
514 ip_vs_stats_release(&svc->stats);
515 kfree(svc);
516 }
517
ip_vs_service_rcu_free(struct rcu_head * head)518 static void ip_vs_service_rcu_free(struct rcu_head *head)
519 {
520 struct ip_vs_service *svc;
521
522 svc = container_of(head, struct ip_vs_service, rcu_head);
523 ip_vs_service_free(svc);
524 }
525
__ip_vs_svc_put(struct ip_vs_service * svc)526 static void __ip_vs_svc_put(struct ip_vs_service *svc)
527 {
528 if (atomic_dec_and_test(&svc->refcnt)) {
529 IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n",
530 svc->fwmark,
531 IP_VS_DBG_ADDR(svc->af, &svc->addr),
532 ntohs(svc->port));
533 call_rcu(&svc->rcu_head, ip_vs_service_rcu_free);
534 }
535 }
536
537
538 /*
539 * Returns hash value for real service
540 */
ip_vs_rs_hashkey(int af,const union nf_inet_addr * addr,__be16 port)541 static inline unsigned int ip_vs_rs_hashkey(int af,
542 const union nf_inet_addr *addr,
543 __be16 port)
544 {
545 unsigned int porth = ntohs(port);
546 __be32 addr_fold = addr->ip;
547
548 #ifdef CONFIG_IP_VS_IPV6
549 if (af == AF_INET6)
550 addr_fold = addr->ip6[0]^addr->ip6[1]^
551 addr->ip6[2]^addr->ip6[3];
552 #endif
553
554 return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
555 & IP_VS_RTAB_MASK;
556 }
557
558 /* Hash ip_vs_dest in rs_table by <proto,addr,port>. */
ip_vs_rs_hash(struct netns_ipvs * ipvs,struct ip_vs_dest * dest)559 static void ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
560 {
561 unsigned int hash;
562 __be16 port;
563
564 if (dest->in_rs_table)
565 return;
566
567 switch (IP_VS_DFWD_METHOD(dest)) {
568 case IP_VS_CONN_F_MASQ:
569 port = dest->port;
570 break;
571 case IP_VS_CONN_F_TUNNEL:
572 switch (dest->tun_type) {
573 case IP_VS_CONN_F_TUNNEL_TYPE_GUE:
574 port = dest->tun_port;
575 break;
576 case IP_VS_CONN_F_TUNNEL_TYPE_IPIP:
577 case IP_VS_CONN_F_TUNNEL_TYPE_GRE:
578 port = 0;
579 break;
580 default:
581 return;
582 }
583 break;
584 default:
585 return;
586 }
587
588 /*
589 * Hash by proto,addr,port,
590 * which are the parameters of the real service.
591 */
592 hash = ip_vs_rs_hashkey(dest->af, &dest->addr, port);
593
594 hlist_add_head_rcu(&dest->d_list, &ipvs->rs_table[hash]);
595 dest->in_rs_table = 1;
596 }
597
598 /* Unhash ip_vs_dest from rs_table. */
ip_vs_rs_unhash(struct ip_vs_dest * dest)599 static void ip_vs_rs_unhash(struct ip_vs_dest *dest)
600 {
601 /*
602 * Remove it from the rs_table table.
603 */
604 if (dest->in_rs_table) {
605 hlist_del_rcu(&dest->d_list);
606 dest->in_rs_table = 0;
607 }
608 }
609
610 /* Check if real service by <proto,addr,port> is present */
ip_vs_has_real_service(struct netns_ipvs * ipvs,int af,__u16 protocol,const union nf_inet_addr * daddr,__be16 dport)611 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
612 const union nf_inet_addr *daddr, __be16 dport)
613 {
614 unsigned int hash;
615 struct ip_vs_dest *dest;
616
617 /* Check for "full" addressed entries */
618 hash = ip_vs_rs_hashkey(af, daddr, dport);
619
620 hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
621 if (dest->port == dport &&
622 dest->af == af &&
623 ip_vs_addr_equal(af, &dest->addr, daddr) &&
624 (dest->protocol == protocol || dest->vfwmark) &&
625 IP_VS_DFWD_METHOD(dest) == IP_VS_CONN_F_MASQ) {
626 /* HIT */
627 return true;
628 }
629 }
630
631 return false;
632 }
633
634 /* Find real service record by <proto,addr,port>.
635 * In case of multiple records with the same <proto,addr,port>, only
636 * the first found record is returned.
637 *
638 * To be called under RCU lock.
639 */
ip_vs_find_real_service(struct netns_ipvs * ipvs,int af,__u16 protocol,const union nf_inet_addr * daddr,__be16 dport)640 struct ip_vs_dest *ip_vs_find_real_service(struct netns_ipvs *ipvs, int af,
641 __u16 protocol,
642 const union nf_inet_addr *daddr,
643 __be16 dport)
644 {
645 unsigned int hash;
646 struct ip_vs_dest *dest;
647
648 /* Check for "full" addressed entries */
649 hash = ip_vs_rs_hashkey(af, daddr, dport);
650
651 hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
652 if (dest->port == dport &&
653 dest->af == af &&
654 ip_vs_addr_equal(af, &dest->addr, daddr) &&
655 (dest->protocol == protocol || dest->vfwmark) &&
656 IP_VS_DFWD_METHOD(dest) == IP_VS_CONN_F_MASQ) {
657 /* HIT */
658 return dest;
659 }
660 }
661
662 return NULL;
663 }
664
665 /* Find real service record by <af,addr,tun_port>.
666 * In case of multiple records with the same <af,addr,tun_port>, only
667 * the first found record is returned.
668 *
669 * To be called under RCU lock.
670 */
ip_vs_find_tunnel(struct netns_ipvs * ipvs,int af,const union nf_inet_addr * daddr,__be16 tun_port)671 struct ip_vs_dest *ip_vs_find_tunnel(struct netns_ipvs *ipvs, int af,
672 const union nf_inet_addr *daddr,
673 __be16 tun_port)
674 {
675 struct ip_vs_dest *dest;
676 unsigned int hash;
677
678 /* Check for "full" addressed entries */
679 hash = ip_vs_rs_hashkey(af, daddr, tun_port);
680
681 hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
682 if (dest->tun_port == tun_port &&
683 dest->af == af &&
684 ip_vs_addr_equal(af, &dest->addr, daddr) &&
685 IP_VS_DFWD_METHOD(dest) == IP_VS_CONN_F_TUNNEL) {
686 /* HIT */
687 return dest;
688 }
689 }
690
691 return NULL;
692 }
693
694 /* Lookup destination by {addr,port} in the given service
695 * Called under RCU lock.
696 */
697 static struct ip_vs_dest *
ip_vs_lookup_dest(struct ip_vs_service * svc,int dest_af,const union nf_inet_addr * daddr,__be16 dport)698 ip_vs_lookup_dest(struct ip_vs_service *svc, int dest_af,
699 const union nf_inet_addr *daddr, __be16 dport)
700 {
701 struct ip_vs_dest *dest;
702
703 /*
704 * Find the destination for the given service
705 */
706 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
707 if ((dest->af == dest_af) &&
708 ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
709 (dest->port == dport)) {
710 /* HIT */
711 return dest;
712 }
713 }
714
715 return NULL;
716 }
717
718 /*
719 * Find destination by {daddr,dport,vaddr,protocol}
720 * Created to be used in ip_vs_process_message() in
721 * the backup synchronization daemon. It finds the
722 * destination to be bound to the received connection
723 * on the backup.
724 * Called under RCU lock, no refcnt is returned.
725 */
ip_vs_find_dest(struct netns_ipvs * ipvs,int svc_af,int dest_af,const union nf_inet_addr * daddr,__be16 dport,const union nf_inet_addr * vaddr,__be16 vport,__u16 protocol,__u32 fwmark,__u32 flags)726 struct ip_vs_dest *ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
727 const union nf_inet_addr *daddr,
728 __be16 dport,
729 const union nf_inet_addr *vaddr,
730 __be16 vport, __u16 protocol, __u32 fwmark,
731 __u32 flags)
732 {
733 struct ip_vs_dest *dest;
734 struct ip_vs_service *svc;
735 __be16 port = dport;
736
737 svc = ip_vs_service_find(ipvs, svc_af, fwmark, protocol, vaddr, vport);
738 if (!svc)
739 return NULL;
740 if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
741 port = 0;
742 dest = ip_vs_lookup_dest(svc, dest_af, daddr, port);
743 if (!dest)
744 dest = ip_vs_lookup_dest(svc, dest_af, daddr, port ^ dport);
745 return dest;
746 }
747
ip_vs_dest_dst_rcu_free(struct rcu_head * head)748 void ip_vs_dest_dst_rcu_free(struct rcu_head *head)
749 {
750 struct ip_vs_dest_dst *dest_dst = container_of(head,
751 struct ip_vs_dest_dst,
752 rcu_head);
753
754 dst_release(dest_dst->dst_cache);
755 kfree(dest_dst);
756 }
757
758 /* Release dest_dst and dst_cache for dest in user context */
__ip_vs_dst_cache_reset(struct ip_vs_dest * dest)759 static void __ip_vs_dst_cache_reset(struct ip_vs_dest *dest)
760 {
761 struct ip_vs_dest_dst *old;
762
763 old = rcu_dereference_protected(dest->dest_dst, 1);
764 if (old) {
765 RCU_INIT_POINTER(dest->dest_dst, NULL);
766 call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
767 }
768 }
769
770 /*
771 * Lookup dest by {svc,addr,port} in the destination trash.
772 * The destination trash is used to hold the destinations that are removed
773 * from the service table but are still referenced by some conn entries.
774 * The reason to add the destination trash is when the dest is temporary
775 * down (either by administrator or by monitor program), the dest can be
776 * picked back from the trash, the remaining connections to the dest can
777 * continue, and the counting information of the dest is also useful for
778 * scheduling.
779 */
780 static struct ip_vs_dest *
ip_vs_trash_get_dest(struct ip_vs_service * svc,int dest_af,const union nf_inet_addr * daddr,__be16 dport)781 ip_vs_trash_get_dest(struct ip_vs_service *svc, int dest_af,
782 const union nf_inet_addr *daddr, __be16 dport)
783 {
784 struct ip_vs_dest *dest;
785 struct netns_ipvs *ipvs = svc->ipvs;
786
787 /*
788 * Find the destination in trash
789 */
790 spin_lock_bh(&ipvs->dest_trash_lock);
791 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
792 IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
793 "dest->refcnt=%d\n",
794 dest->vfwmark,
795 IP_VS_DBG_ADDR(dest->af, &dest->addr),
796 ntohs(dest->port),
797 refcount_read(&dest->refcnt));
798 if (dest->af == dest_af &&
799 ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
800 dest->port == dport &&
801 dest->vfwmark == svc->fwmark &&
802 dest->protocol == svc->protocol &&
803 (svc->fwmark ||
804 (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
805 dest->vport == svc->port))) {
806 /* HIT */
807 list_del(&dest->t_list);
808 goto out;
809 }
810 }
811
812 dest = NULL;
813
814 out:
815 spin_unlock_bh(&ipvs->dest_trash_lock);
816
817 return dest;
818 }
819
ip_vs_dest_rcu_free(struct rcu_head * head)820 static void ip_vs_dest_rcu_free(struct rcu_head *head)
821 {
822 struct ip_vs_dest *dest;
823
824 dest = container_of(head, struct ip_vs_dest, rcu_head);
825 ip_vs_stats_release(&dest->stats);
826 ip_vs_dest_put_and_free(dest);
827 }
828
ip_vs_dest_free(struct ip_vs_dest * dest)829 static void ip_vs_dest_free(struct ip_vs_dest *dest)
830 {
831 struct ip_vs_service *svc = rcu_dereference_protected(dest->svc, 1);
832
833 __ip_vs_dst_cache_reset(dest);
834 __ip_vs_svc_put(svc);
835 call_rcu(&dest->rcu_head, ip_vs_dest_rcu_free);
836 }
837
838 /*
839 * Clean up all the destinations in the trash
840 * Called by the ip_vs_control_cleanup()
841 *
842 * When the ip_vs_control_clearup is activated by ipvs module exit,
843 * the service tables must have been flushed and all the connections
844 * are expired, and the refcnt of each destination in the trash must
845 * be 1, so we simply release them here.
846 */
ip_vs_trash_cleanup(struct netns_ipvs * ipvs)847 static void ip_vs_trash_cleanup(struct netns_ipvs *ipvs)
848 {
849 struct ip_vs_dest *dest, *nxt;
850
851 del_timer_sync(&ipvs->dest_trash_timer);
852 /* No need to use dest_trash_lock */
853 list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, t_list) {
854 list_del(&dest->t_list);
855 ip_vs_dest_free(dest);
856 }
857 }
858
ip_vs_stats_rcu_free(struct rcu_head * head)859 static void ip_vs_stats_rcu_free(struct rcu_head *head)
860 {
861 struct ip_vs_stats_rcu *rs = container_of(head,
862 struct ip_vs_stats_rcu,
863 rcu_head);
864
865 ip_vs_stats_release(&rs->s);
866 kfree(rs);
867 }
868
869 static void
ip_vs_copy_stats(struct ip_vs_kstats * dst,struct ip_vs_stats * src)870 ip_vs_copy_stats(struct ip_vs_kstats *dst, struct ip_vs_stats *src)
871 {
872 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->kstats.c - src->kstats0.c
873
874 spin_lock(&src->lock);
875
876 IP_VS_SHOW_STATS_COUNTER(conns);
877 IP_VS_SHOW_STATS_COUNTER(inpkts);
878 IP_VS_SHOW_STATS_COUNTER(outpkts);
879 IP_VS_SHOW_STATS_COUNTER(inbytes);
880 IP_VS_SHOW_STATS_COUNTER(outbytes);
881
882 ip_vs_read_estimator(dst, src);
883
884 spin_unlock(&src->lock);
885 }
886
887 static void
ip_vs_export_stats_user(struct ip_vs_stats_user * dst,struct ip_vs_kstats * src)888 ip_vs_export_stats_user(struct ip_vs_stats_user *dst, struct ip_vs_kstats *src)
889 {
890 dst->conns = (u32)src->conns;
891 dst->inpkts = (u32)src->inpkts;
892 dst->outpkts = (u32)src->outpkts;
893 dst->inbytes = src->inbytes;
894 dst->outbytes = src->outbytes;
895 dst->cps = (u32)src->cps;
896 dst->inpps = (u32)src->inpps;
897 dst->outpps = (u32)src->outpps;
898 dst->inbps = (u32)src->inbps;
899 dst->outbps = (u32)src->outbps;
900 }
901
902 static void
ip_vs_zero_stats(struct ip_vs_stats * stats)903 ip_vs_zero_stats(struct ip_vs_stats *stats)
904 {
905 spin_lock(&stats->lock);
906
907 /* get current counters as zero point, rates are zeroed */
908
909 #define IP_VS_ZERO_STATS_COUNTER(c) stats->kstats0.c = stats->kstats.c
910
911 IP_VS_ZERO_STATS_COUNTER(conns);
912 IP_VS_ZERO_STATS_COUNTER(inpkts);
913 IP_VS_ZERO_STATS_COUNTER(outpkts);
914 IP_VS_ZERO_STATS_COUNTER(inbytes);
915 IP_VS_ZERO_STATS_COUNTER(outbytes);
916
917 ip_vs_zero_estimator(stats);
918
919 spin_unlock(&stats->lock);
920 }
921
922 /* Allocate fields after kzalloc */
ip_vs_stats_init_alloc(struct ip_vs_stats * s)923 int ip_vs_stats_init_alloc(struct ip_vs_stats *s)
924 {
925 int i;
926
927 spin_lock_init(&s->lock);
928 s->cpustats = alloc_percpu(struct ip_vs_cpu_stats);
929 if (!s->cpustats)
930 return -ENOMEM;
931
932 for_each_possible_cpu(i) {
933 struct ip_vs_cpu_stats *cs = per_cpu_ptr(s->cpustats, i);
934
935 u64_stats_init(&cs->syncp);
936 }
937 return 0;
938 }
939
ip_vs_stats_alloc(void)940 struct ip_vs_stats *ip_vs_stats_alloc(void)
941 {
942 struct ip_vs_stats *s = kzalloc(sizeof(*s), GFP_KERNEL);
943
944 if (s && ip_vs_stats_init_alloc(s) >= 0)
945 return s;
946 kfree(s);
947 return NULL;
948 }
949
ip_vs_stats_release(struct ip_vs_stats * stats)950 void ip_vs_stats_release(struct ip_vs_stats *stats)
951 {
952 free_percpu(stats->cpustats);
953 }
954
ip_vs_stats_free(struct ip_vs_stats * stats)955 void ip_vs_stats_free(struct ip_vs_stats *stats)
956 {
957 if (stats) {
958 ip_vs_stats_release(stats);
959 kfree(stats);
960 }
961 }
962
963 /*
964 * Update a destination in the given service
965 */
966 static void
__ip_vs_update_dest(struct ip_vs_service * svc,struct ip_vs_dest * dest,struct ip_vs_dest_user_kern * udest,int add)967 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
968 struct ip_vs_dest_user_kern *udest, int add)
969 {
970 struct netns_ipvs *ipvs = svc->ipvs;
971 struct ip_vs_service *old_svc;
972 struct ip_vs_scheduler *sched;
973 int conn_flags;
974
975 /* We cannot modify an address and change the address family */
976 BUG_ON(!add && udest->af != dest->af);
977
978 if (add && udest->af != svc->af)
979 ipvs->mixed_address_family_dests++;
980
981 /* keep the last_weight with latest non-0 weight */
982 if (add || udest->weight != 0)
983 atomic_set(&dest->last_weight, udest->weight);
984
985 /* set the weight and the flags */
986 atomic_set(&dest->weight, udest->weight);
987 conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
988 conn_flags |= IP_VS_CONN_F_INACTIVE;
989
990 /* Need to rehash? */
991 if ((udest->conn_flags & IP_VS_CONN_F_FWD_MASK) !=
992 IP_VS_DFWD_METHOD(dest) ||
993 udest->tun_type != dest->tun_type ||
994 udest->tun_port != dest->tun_port)
995 ip_vs_rs_unhash(dest);
996
997 /* set the tunnel info */
998 dest->tun_type = udest->tun_type;
999 dest->tun_port = udest->tun_port;
1000 dest->tun_flags = udest->tun_flags;
1001
1002 /* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
1003 if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
1004 conn_flags |= IP_VS_CONN_F_NOOUTPUT;
1005 } else {
1006 /* FTP-NAT requires conntrack for mangling */
1007 if (svc->port == FTPPORT)
1008 ip_vs_register_conntrack(svc);
1009 }
1010 atomic_set(&dest->conn_flags, conn_flags);
1011 /* Put the real service in rs_table if not present. */
1012 ip_vs_rs_hash(ipvs, dest);
1013
1014 /* bind the service */
1015 old_svc = rcu_dereference_protected(dest->svc, 1);
1016 if (!old_svc) {
1017 __ip_vs_bind_svc(dest, svc);
1018 } else {
1019 if (old_svc != svc) {
1020 ip_vs_zero_stats(&dest->stats);
1021 __ip_vs_bind_svc(dest, svc);
1022 __ip_vs_svc_put(old_svc);
1023 }
1024 }
1025
1026 /* set the dest status flags */
1027 dest->flags |= IP_VS_DEST_F_AVAILABLE;
1028
1029 if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
1030 dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
1031 dest->u_threshold = udest->u_threshold;
1032 dest->l_threshold = udest->l_threshold;
1033
1034 dest->af = udest->af;
1035
1036 spin_lock_bh(&dest->dst_lock);
1037 __ip_vs_dst_cache_reset(dest);
1038 spin_unlock_bh(&dest->dst_lock);
1039
1040 if (add) {
1041 list_add_rcu(&dest->n_list, &svc->destinations);
1042 svc->num_dests++;
1043 sched = rcu_dereference_protected(svc->scheduler, 1);
1044 if (sched && sched->add_dest)
1045 sched->add_dest(svc, dest);
1046 } else {
1047 sched = rcu_dereference_protected(svc->scheduler, 1);
1048 if (sched && sched->upd_dest)
1049 sched->upd_dest(svc, dest);
1050 }
1051 }
1052
1053
1054 /*
1055 * Create a destination for the given service
1056 */
1057 static int
ip_vs_new_dest(struct ip_vs_service * svc,struct ip_vs_dest_user_kern * udest)1058 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1059 {
1060 struct ip_vs_dest *dest;
1061 unsigned int atype;
1062 int ret;
1063
1064 #ifdef CONFIG_IP_VS_IPV6
1065 if (udest->af == AF_INET6) {
1066 atype = ipv6_addr_type(&udest->addr.in6);
1067 if ((!(atype & IPV6_ADDR_UNICAST) ||
1068 atype & IPV6_ADDR_LINKLOCAL) &&
1069 !__ip_vs_addr_is_local_v6(svc->ipvs->net, &udest->addr.in6))
1070 return -EINVAL;
1071
1072 ret = nf_defrag_ipv6_enable(svc->ipvs->net);
1073 if (ret)
1074 return ret;
1075 } else
1076 #endif
1077 {
1078 atype = inet_addr_type(svc->ipvs->net, udest->addr.ip);
1079 if (atype != RTN_LOCAL && atype != RTN_UNICAST)
1080 return -EINVAL;
1081 }
1082
1083 dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
1084 if (dest == NULL)
1085 return -ENOMEM;
1086
1087 ret = ip_vs_stats_init_alloc(&dest->stats);
1088 if (ret < 0)
1089 goto err_alloc;
1090
1091 ret = ip_vs_start_estimator(svc->ipvs, &dest->stats);
1092 if (ret < 0)
1093 goto err_stats;
1094
1095 dest->af = udest->af;
1096 dest->protocol = svc->protocol;
1097 dest->vaddr = svc->addr;
1098 dest->vport = svc->port;
1099 dest->vfwmark = svc->fwmark;
1100 ip_vs_addr_copy(udest->af, &dest->addr, &udest->addr);
1101 dest->port = udest->port;
1102
1103 atomic_set(&dest->activeconns, 0);
1104 atomic_set(&dest->inactconns, 0);
1105 atomic_set(&dest->persistconns, 0);
1106 refcount_set(&dest->refcnt, 1);
1107
1108 INIT_HLIST_NODE(&dest->d_list);
1109 spin_lock_init(&dest->dst_lock);
1110 __ip_vs_update_dest(svc, dest, udest, 1);
1111
1112 return 0;
1113
1114 err_stats:
1115 ip_vs_stats_release(&dest->stats);
1116
1117 err_alloc:
1118 kfree(dest);
1119 return ret;
1120 }
1121
1122
1123 /*
1124 * Add a destination into an existing service
1125 */
1126 static int
ip_vs_add_dest(struct ip_vs_service * svc,struct ip_vs_dest_user_kern * udest)1127 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1128 {
1129 struct ip_vs_dest *dest;
1130 union nf_inet_addr daddr;
1131 __be16 dport = udest->port;
1132 int ret;
1133
1134 if (udest->weight < 0) {
1135 pr_err("%s(): server weight less than zero\n", __func__);
1136 return -ERANGE;
1137 }
1138
1139 if (udest->l_threshold > udest->u_threshold) {
1140 pr_err("%s(): lower threshold is higher than upper threshold\n",
1141 __func__);
1142 return -ERANGE;
1143 }
1144
1145 if (udest->tun_type == IP_VS_CONN_F_TUNNEL_TYPE_GUE) {
1146 if (udest->tun_port == 0) {
1147 pr_err("%s(): tunnel port is zero\n", __func__);
1148 return -EINVAL;
1149 }
1150 }
1151
1152 ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
1153
1154 /* We use function that requires RCU lock */
1155 rcu_read_lock();
1156 dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
1157 rcu_read_unlock();
1158
1159 if (dest != NULL) {
1160 IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
1161 return -EEXIST;
1162 }
1163
1164 /*
1165 * Check if the dest already exists in the trash and
1166 * is from the same service
1167 */
1168 dest = ip_vs_trash_get_dest(svc, udest->af, &daddr, dport);
1169
1170 if (dest != NULL) {
1171 IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
1172 "dest->refcnt=%d, service %u/%s:%u\n",
1173 IP_VS_DBG_ADDR(udest->af, &daddr), ntohs(dport),
1174 refcount_read(&dest->refcnt),
1175 dest->vfwmark,
1176 IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
1177 ntohs(dest->vport));
1178
1179 ret = ip_vs_start_estimator(svc->ipvs, &dest->stats);
1180 if (ret < 0)
1181 return ret;
1182 __ip_vs_update_dest(svc, dest, udest, 1);
1183 } else {
1184 /*
1185 * Allocate and initialize the dest structure
1186 */
1187 ret = ip_vs_new_dest(svc, udest);
1188 }
1189
1190 return ret;
1191 }
1192
1193
1194 /*
1195 * Edit a destination in the given service
1196 */
1197 static int
ip_vs_edit_dest(struct ip_vs_service * svc,struct ip_vs_dest_user_kern * udest)1198 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1199 {
1200 struct ip_vs_dest *dest;
1201 union nf_inet_addr daddr;
1202 __be16 dport = udest->port;
1203
1204 if (udest->weight < 0) {
1205 pr_err("%s(): server weight less than zero\n", __func__);
1206 return -ERANGE;
1207 }
1208
1209 if (udest->l_threshold > udest->u_threshold) {
1210 pr_err("%s(): lower threshold is higher than upper threshold\n",
1211 __func__);
1212 return -ERANGE;
1213 }
1214
1215 if (udest->tun_type == IP_VS_CONN_F_TUNNEL_TYPE_GUE) {
1216 if (udest->tun_port == 0) {
1217 pr_err("%s(): tunnel port is zero\n", __func__);
1218 return -EINVAL;
1219 }
1220 }
1221
1222 ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
1223
1224 /* We use function that requires RCU lock */
1225 rcu_read_lock();
1226 dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
1227 rcu_read_unlock();
1228
1229 if (dest == NULL) {
1230 IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1231 return -ENOENT;
1232 }
1233
1234 __ip_vs_update_dest(svc, dest, udest, 0);
1235
1236 return 0;
1237 }
1238
1239 /*
1240 * Delete a destination (must be already unlinked from the service)
1241 */
__ip_vs_del_dest(struct netns_ipvs * ipvs,struct ip_vs_dest * dest,bool cleanup)1242 static void __ip_vs_del_dest(struct netns_ipvs *ipvs, struct ip_vs_dest *dest,
1243 bool cleanup)
1244 {
1245 ip_vs_stop_estimator(ipvs, &dest->stats);
1246
1247 /*
1248 * Remove it from the d-linked list with the real services.
1249 */
1250 ip_vs_rs_unhash(dest);
1251
1252 spin_lock_bh(&ipvs->dest_trash_lock);
1253 IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1254 IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port),
1255 refcount_read(&dest->refcnt));
1256 if (list_empty(&ipvs->dest_trash) && !cleanup)
1257 mod_timer(&ipvs->dest_trash_timer,
1258 jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1259 /* dest lives in trash with reference */
1260 list_add(&dest->t_list, &ipvs->dest_trash);
1261 dest->idle_start = 0;
1262 spin_unlock_bh(&ipvs->dest_trash_lock);
1263
1264 /* Queue up delayed work to expire all no destination connections.
1265 * No-op when CONFIG_SYSCTL is disabled.
1266 */
1267 if (!cleanup)
1268 ip_vs_enqueue_expire_nodest_conns(ipvs);
1269 }
1270
1271
1272 /*
1273 * Unlink a destination from the given service
1274 */
__ip_vs_unlink_dest(struct ip_vs_service * svc,struct ip_vs_dest * dest,int svcupd)1275 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1276 struct ip_vs_dest *dest,
1277 int svcupd)
1278 {
1279 dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1280
1281 /*
1282 * Remove it from the d-linked destination list.
1283 */
1284 list_del_rcu(&dest->n_list);
1285 svc->num_dests--;
1286
1287 if (dest->af != svc->af)
1288 svc->ipvs->mixed_address_family_dests--;
1289
1290 if (svcupd) {
1291 struct ip_vs_scheduler *sched;
1292
1293 sched = rcu_dereference_protected(svc->scheduler, 1);
1294 if (sched && sched->del_dest)
1295 sched->del_dest(svc, dest);
1296 }
1297 }
1298
1299
1300 /*
1301 * Delete a destination server in the given service
1302 */
1303 static int
ip_vs_del_dest(struct ip_vs_service * svc,struct ip_vs_dest_user_kern * udest)1304 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1305 {
1306 struct ip_vs_dest *dest;
1307 __be16 dport = udest->port;
1308
1309 /* We use function that requires RCU lock */
1310 rcu_read_lock();
1311 dest = ip_vs_lookup_dest(svc, udest->af, &udest->addr, dport);
1312 rcu_read_unlock();
1313
1314 if (dest == NULL) {
1315 IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1316 return -ENOENT;
1317 }
1318
1319 /*
1320 * Unlink dest from the service
1321 */
1322 __ip_vs_unlink_dest(svc, dest, 1);
1323
1324 /*
1325 * Delete the destination
1326 */
1327 __ip_vs_del_dest(svc->ipvs, dest, false);
1328
1329 return 0;
1330 }
1331
ip_vs_dest_trash_expire(struct timer_list * t)1332 static void ip_vs_dest_trash_expire(struct timer_list *t)
1333 {
1334 struct netns_ipvs *ipvs = from_timer(ipvs, t, dest_trash_timer);
1335 struct ip_vs_dest *dest, *next;
1336 unsigned long now = jiffies;
1337
1338 spin_lock(&ipvs->dest_trash_lock);
1339 list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) {
1340 if (refcount_read(&dest->refcnt) > 1)
1341 continue;
1342 if (dest->idle_start) {
1343 if (time_before(now, dest->idle_start +
1344 IP_VS_DEST_TRASH_PERIOD))
1345 continue;
1346 } else {
1347 dest->idle_start = max(1UL, now);
1348 continue;
1349 }
1350 IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1351 dest->vfwmark,
1352 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1353 ntohs(dest->port));
1354 list_del(&dest->t_list);
1355 ip_vs_dest_free(dest);
1356 }
1357 if (!list_empty(&ipvs->dest_trash))
1358 mod_timer(&ipvs->dest_trash_timer,
1359 jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1360 spin_unlock(&ipvs->dest_trash_lock);
1361 }
1362
1363 /*
1364 * Add a service into the service hash table
1365 */
1366 static int
ip_vs_add_service(struct netns_ipvs * ipvs,struct ip_vs_service_user_kern * u,struct ip_vs_service ** svc_p)1367 ip_vs_add_service(struct netns_ipvs *ipvs, struct ip_vs_service_user_kern *u,
1368 struct ip_vs_service **svc_p)
1369 {
1370 int ret = 0;
1371 struct ip_vs_scheduler *sched = NULL;
1372 struct ip_vs_pe *pe = NULL;
1373 struct ip_vs_service *svc = NULL;
1374 int ret_hooks = -1;
1375
1376 /* increase the module use count */
1377 if (!ip_vs_use_count_inc())
1378 return -ENOPROTOOPT;
1379
1380 /* Lookup the scheduler by 'u->sched_name' */
1381 if (strcmp(u->sched_name, "none")) {
1382 sched = ip_vs_scheduler_get(u->sched_name);
1383 if (!sched) {
1384 pr_info("Scheduler module ip_vs_%s not found\n",
1385 u->sched_name);
1386 ret = -ENOENT;
1387 goto out_err;
1388 }
1389 }
1390
1391 if (u->pe_name && *u->pe_name) {
1392 pe = ip_vs_pe_getbyname(u->pe_name);
1393 if (pe == NULL) {
1394 pr_info("persistence engine module ip_vs_pe_%s "
1395 "not found\n", u->pe_name);
1396 ret = -ENOENT;
1397 goto out_err;
1398 }
1399 }
1400
1401 #ifdef CONFIG_IP_VS_IPV6
1402 if (u->af == AF_INET6) {
1403 __u32 plen = (__force __u32) u->netmask;
1404
1405 if (plen < 1 || plen > 128) {
1406 ret = -EINVAL;
1407 goto out_err;
1408 }
1409
1410 ret = nf_defrag_ipv6_enable(ipvs->net);
1411 if (ret)
1412 goto out_err;
1413 }
1414 #endif
1415
1416 if ((u->af == AF_INET && !ipvs->num_services) ||
1417 (u->af == AF_INET6 && !ipvs->num_services6)) {
1418 ret = ip_vs_register_hooks(ipvs, u->af);
1419 if (ret < 0)
1420 goto out_err;
1421 ret_hooks = ret;
1422 }
1423
1424 svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1425 if (svc == NULL) {
1426 IP_VS_DBG(1, "%s(): no memory\n", __func__);
1427 ret = -ENOMEM;
1428 goto out_err;
1429 }
1430 ret = ip_vs_stats_init_alloc(&svc->stats);
1431 if (ret < 0)
1432 goto out_err;
1433
1434 /* I'm the first user of the service */
1435 atomic_set(&svc->refcnt, 0);
1436
1437 svc->af = u->af;
1438 svc->protocol = u->protocol;
1439 ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1440 svc->port = u->port;
1441 svc->fwmark = u->fwmark;
1442 svc->flags = u->flags & ~IP_VS_SVC_F_HASHED;
1443 svc->timeout = u->timeout * HZ;
1444 svc->netmask = u->netmask;
1445 svc->ipvs = ipvs;
1446
1447 INIT_LIST_HEAD(&svc->destinations);
1448 spin_lock_init(&svc->sched_lock);
1449
1450 /* Bind the scheduler */
1451 if (sched) {
1452 ret = ip_vs_bind_scheduler(svc, sched);
1453 if (ret)
1454 goto out_err;
1455 sched = NULL;
1456 }
1457
1458 ret = ip_vs_start_estimator(ipvs, &svc->stats);
1459 if (ret < 0)
1460 goto out_err;
1461
1462 /* Bind the ct retriever */
1463 RCU_INIT_POINTER(svc->pe, pe);
1464 pe = NULL;
1465
1466 /* Update the virtual service counters */
1467 if (svc->port == FTPPORT)
1468 atomic_inc(&ipvs->ftpsvc_counter);
1469 else if (svc->port == 0)
1470 atomic_inc(&ipvs->nullsvc_counter);
1471 if (svc->pe && svc->pe->conn_out)
1472 atomic_inc(&ipvs->conn_out_counter);
1473
1474 /* Count only IPv4 services for old get/setsockopt interface */
1475 if (svc->af == AF_INET)
1476 ipvs->num_services++;
1477 else if (svc->af == AF_INET6)
1478 ipvs->num_services6++;
1479
1480 /* Hash the service into the service table */
1481 ip_vs_svc_hash(svc);
1482
1483 *svc_p = svc;
1484
1485 if (!ipvs->enable) {
1486 /* Now there is a service - full throttle */
1487 ipvs->enable = 1;
1488
1489 /* Start estimation for first time */
1490 ip_vs_est_reload_start(ipvs);
1491 }
1492
1493 return 0;
1494
1495
1496 out_err:
1497 if (ret_hooks >= 0)
1498 ip_vs_unregister_hooks(ipvs, u->af);
1499 if (svc != NULL) {
1500 ip_vs_unbind_scheduler(svc, sched);
1501 ip_vs_service_free(svc);
1502 }
1503 ip_vs_scheduler_put(sched);
1504 ip_vs_pe_put(pe);
1505
1506 /* decrease the module use count */
1507 ip_vs_use_count_dec();
1508
1509 return ret;
1510 }
1511
1512
1513 /*
1514 * Edit a service and bind it with a new scheduler
1515 */
1516 static int
ip_vs_edit_service(struct ip_vs_service * svc,struct ip_vs_service_user_kern * u)1517 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1518 {
1519 struct ip_vs_scheduler *sched = NULL, *old_sched;
1520 struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1521 int ret = 0;
1522 bool new_pe_conn_out, old_pe_conn_out;
1523
1524 /*
1525 * Lookup the scheduler, by 'u->sched_name'
1526 */
1527 if (strcmp(u->sched_name, "none")) {
1528 sched = ip_vs_scheduler_get(u->sched_name);
1529 if (!sched) {
1530 pr_info("Scheduler module ip_vs_%s not found\n",
1531 u->sched_name);
1532 return -ENOENT;
1533 }
1534 }
1535 old_sched = sched;
1536
1537 if (u->pe_name && *u->pe_name) {
1538 pe = ip_vs_pe_getbyname(u->pe_name);
1539 if (pe == NULL) {
1540 pr_info("persistence engine module ip_vs_pe_%s "
1541 "not found\n", u->pe_name);
1542 ret = -ENOENT;
1543 goto out;
1544 }
1545 old_pe = pe;
1546 }
1547
1548 #ifdef CONFIG_IP_VS_IPV6
1549 if (u->af == AF_INET6) {
1550 __u32 plen = (__force __u32) u->netmask;
1551
1552 if (plen < 1 || plen > 128) {
1553 ret = -EINVAL;
1554 goto out;
1555 }
1556 }
1557 #endif
1558
1559 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1560 if (sched != old_sched) {
1561 if (old_sched) {
1562 ip_vs_unbind_scheduler(svc, old_sched);
1563 RCU_INIT_POINTER(svc->scheduler, NULL);
1564 /* Wait all svc->sched_data users */
1565 synchronize_rcu();
1566 }
1567 /* Bind the new scheduler */
1568 if (sched) {
1569 ret = ip_vs_bind_scheduler(svc, sched);
1570 if (ret) {
1571 ip_vs_scheduler_put(sched);
1572 goto out;
1573 }
1574 }
1575 }
1576
1577 /*
1578 * Set the flags and timeout value
1579 */
1580 svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1581 svc->timeout = u->timeout * HZ;
1582 svc->netmask = u->netmask;
1583
1584 old_pe = rcu_dereference_protected(svc->pe, 1);
1585 if (pe != old_pe) {
1586 rcu_assign_pointer(svc->pe, pe);
1587 /* check for optional methods in new pe */
1588 new_pe_conn_out = (pe && pe->conn_out) ? true : false;
1589 old_pe_conn_out = (old_pe && old_pe->conn_out) ? true : false;
1590 if (new_pe_conn_out && !old_pe_conn_out)
1591 atomic_inc(&svc->ipvs->conn_out_counter);
1592 if (old_pe_conn_out && !new_pe_conn_out)
1593 atomic_dec(&svc->ipvs->conn_out_counter);
1594 }
1595
1596 out:
1597 ip_vs_scheduler_put(old_sched);
1598 ip_vs_pe_put(old_pe);
1599 return ret;
1600 }
1601
1602 /*
1603 * Delete a service from the service list
1604 * - The service must be unlinked, unlocked and not referenced!
1605 * - We are called under _bh lock
1606 */
__ip_vs_del_service(struct ip_vs_service * svc,bool cleanup)1607 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1608 {
1609 struct ip_vs_dest *dest, *nxt;
1610 struct ip_vs_scheduler *old_sched;
1611 struct ip_vs_pe *old_pe;
1612 struct netns_ipvs *ipvs = svc->ipvs;
1613
1614 if (svc->af == AF_INET) {
1615 ipvs->num_services--;
1616 if (!ipvs->num_services)
1617 ip_vs_unregister_hooks(ipvs, svc->af);
1618 } else if (svc->af == AF_INET6) {
1619 ipvs->num_services6--;
1620 if (!ipvs->num_services6)
1621 ip_vs_unregister_hooks(ipvs, svc->af);
1622 }
1623
1624 ip_vs_stop_estimator(svc->ipvs, &svc->stats);
1625
1626 /* Unbind scheduler */
1627 old_sched = rcu_dereference_protected(svc->scheduler, 1);
1628 ip_vs_unbind_scheduler(svc, old_sched);
1629 ip_vs_scheduler_put(old_sched);
1630
1631 /* Unbind persistence engine, keep svc->pe */
1632 old_pe = rcu_dereference_protected(svc->pe, 1);
1633 if (old_pe && old_pe->conn_out)
1634 atomic_dec(&ipvs->conn_out_counter);
1635 ip_vs_pe_put(old_pe);
1636
1637 /*
1638 * Unlink the whole destination list
1639 */
1640 list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1641 __ip_vs_unlink_dest(svc, dest, 0);
1642 __ip_vs_del_dest(svc->ipvs, dest, cleanup);
1643 }
1644
1645 /*
1646 * Update the virtual service counters
1647 */
1648 if (svc->port == FTPPORT)
1649 atomic_dec(&ipvs->ftpsvc_counter);
1650 else if (svc->port == 0)
1651 atomic_dec(&ipvs->nullsvc_counter);
1652
1653 /*
1654 * Free the service if nobody refers to it
1655 */
1656 __ip_vs_svc_put(svc);
1657
1658 /* decrease the module use count */
1659 ip_vs_use_count_dec();
1660 }
1661
1662 /*
1663 * Unlink a service from list and try to delete it if its refcnt reached 0
1664 */
ip_vs_unlink_service(struct ip_vs_service * svc,bool cleanup)1665 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1666 {
1667 ip_vs_unregister_conntrack(svc);
1668 /* Hold svc to avoid double release from dest_trash */
1669 atomic_inc(&svc->refcnt);
1670 /*
1671 * Unhash it from the service table
1672 */
1673 ip_vs_svc_unhash(svc);
1674
1675 __ip_vs_del_service(svc, cleanup);
1676 }
1677
1678 /*
1679 * Delete a service from the service list
1680 */
ip_vs_del_service(struct ip_vs_service * svc)1681 static int ip_vs_del_service(struct ip_vs_service *svc)
1682 {
1683 if (svc == NULL)
1684 return -EEXIST;
1685 ip_vs_unlink_service(svc, false);
1686
1687 return 0;
1688 }
1689
1690
1691 /*
1692 * Flush all the virtual services
1693 */
ip_vs_flush(struct netns_ipvs * ipvs,bool cleanup)1694 static int ip_vs_flush(struct netns_ipvs *ipvs, bool cleanup)
1695 {
1696 int idx;
1697 struct ip_vs_service *svc;
1698 struct hlist_node *n;
1699
1700 /*
1701 * Flush the service table hashed by <netns,protocol,addr,port>
1702 */
1703 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1704 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1705 s_list) {
1706 if (svc->ipvs == ipvs)
1707 ip_vs_unlink_service(svc, cleanup);
1708 }
1709 }
1710
1711 /*
1712 * Flush the service table hashed by fwmark
1713 */
1714 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1715 hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1716 f_list) {
1717 if (svc->ipvs == ipvs)
1718 ip_vs_unlink_service(svc, cleanup);
1719 }
1720 }
1721
1722 return 0;
1723 }
1724
1725 /*
1726 * Delete service by {netns} in the service table.
1727 * Called by __ip_vs_batch_cleanup()
1728 */
ip_vs_service_nets_cleanup(struct list_head * net_list)1729 void ip_vs_service_nets_cleanup(struct list_head *net_list)
1730 {
1731 struct netns_ipvs *ipvs;
1732 struct net *net;
1733
1734 /* Check for "full" addressed entries */
1735 mutex_lock(&__ip_vs_mutex);
1736 list_for_each_entry(net, net_list, exit_list) {
1737 ipvs = net_ipvs(net);
1738 ip_vs_flush(ipvs, true);
1739 }
1740 mutex_unlock(&__ip_vs_mutex);
1741 }
1742
1743 /* Put all references for device (dst_cache) */
1744 static inline void
ip_vs_forget_dev(struct ip_vs_dest * dest,struct net_device * dev)1745 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1746 {
1747 struct ip_vs_dest_dst *dest_dst;
1748
1749 spin_lock_bh(&dest->dst_lock);
1750 dest_dst = rcu_dereference_protected(dest->dest_dst, 1);
1751 if (dest_dst && dest_dst->dst_cache->dev == dev) {
1752 IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1753 dev->name,
1754 IP_VS_DBG_ADDR(dest->af, &dest->addr),
1755 ntohs(dest->port),
1756 refcount_read(&dest->refcnt));
1757 __ip_vs_dst_cache_reset(dest);
1758 }
1759 spin_unlock_bh(&dest->dst_lock);
1760
1761 }
1762 /* Netdev event receiver
1763 * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1764 */
ip_vs_dst_event(struct notifier_block * this,unsigned long event,void * ptr)1765 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1766 void *ptr)
1767 {
1768 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1769 struct net *net = dev_net(dev);
1770 struct netns_ipvs *ipvs = net_ipvs(net);
1771 struct ip_vs_service *svc;
1772 struct ip_vs_dest *dest;
1773 unsigned int idx;
1774
1775 if (event != NETDEV_DOWN || !ipvs)
1776 return NOTIFY_DONE;
1777 IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1778 mutex_lock(&__ip_vs_mutex);
1779 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1780 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1781 if (svc->ipvs == ipvs) {
1782 list_for_each_entry(dest, &svc->destinations,
1783 n_list) {
1784 ip_vs_forget_dev(dest, dev);
1785 }
1786 }
1787 }
1788
1789 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1790 if (svc->ipvs == ipvs) {
1791 list_for_each_entry(dest, &svc->destinations,
1792 n_list) {
1793 ip_vs_forget_dev(dest, dev);
1794 }
1795 }
1796
1797 }
1798 }
1799
1800 spin_lock_bh(&ipvs->dest_trash_lock);
1801 list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1802 ip_vs_forget_dev(dest, dev);
1803 }
1804 spin_unlock_bh(&ipvs->dest_trash_lock);
1805 mutex_unlock(&__ip_vs_mutex);
1806 return NOTIFY_DONE;
1807 }
1808
1809 /*
1810 * Zero counters in a service or all services
1811 */
ip_vs_zero_service(struct ip_vs_service * svc)1812 static int ip_vs_zero_service(struct ip_vs_service *svc)
1813 {
1814 struct ip_vs_dest *dest;
1815
1816 list_for_each_entry(dest, &svc->destinations, n_list) {
1817 ip_vs_zero_stats(&dest->stats);
1818 }
1819 ip_vs_zero_stats(&svc->stats);
1820 return 0;
1821 }
1822
ip_vs_zero_all(struct netns_ipvs * ipvs)1823 static int ip_vs_zero_all(struct netns_ipvs *ipvs)
1824 {
1825 int idx;
1826 struct ip_vs_service *svc;
1827
1828 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1829 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1830 if (svc->ipvs == ipvs)
1831 ip_vs_zero_service(svc);
1832 }
1833 }
1834
1835 for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1836 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1837 if (svc->ipvs == ipvs)
1838 ip_vs_zero_service(svc);
1839 }
1840 }
1841
1842 ip_vs_zero_stats(&ipvs->tot_stats->s);
1843 return 0;
1844 }
1845
1846 #ifdef CONFIG_SYSCTL
1847
1848 static int
proc_do_defense_mode(struct ctl_table * table,int write,void * buffer,size_t * lenp,loff_t * ppos)1849 proc_do_defense_mode(struct ctl_table *table, int write,
1850 void *buffer, size_t *lenp, loff_t *ppos)
1851 {
1852 struct netns_ipvs *ipvs = table->extra2;
1853 int *valp = table->data;
1854 int val = *valp;
1855 int rc;
1856
1857 struct ctl_table tmp = {
1858 .data = &val,
1859 .maxlen = sizeof(int),
1860 .mode = table->mode,
1861 };
1862
1863 rc = proc_dointvec(&tmp, write, buffer, lenp, ppos);
1864 if (write && (*valp != val)) {
1865 if (val < 0 || val > 3) {
1866 rc = -EINVAL;
1867 } else {
1868 *valp = val;
1869 update_defense_level(ipvs);
1870 }
1871 }
1872 return rc;
1873 }
1874
1875 static int
proc_do_sync_threshold(struct ctl_table * table,int write,void * buffer,size_t * lenp,loff_t * ppos)1876 proc_do_sync_threshold(struct ctl_table *table, int write,
1877 void *buffer, size_t *lenp, loff_t *ppos)
1878 {
1879 struct netns_ipvs *ipvs = table->extra2;
1880 int *valp = table->data;
1881 int val[2];
1882 int rc;
1883 struct ctl_table tmp = {
1884 .data = &val,
1885 .maxlen = table->maxlen,
1886 .mode = table->mode,
1887 };
1888
1889 mutex_lock(&ipvs->sync_mutex);
1890 memcpy(val, valp, sizeof(val));
1891 rc = proc_dointvec(&tmp, write, buffer, lenp, ppos);
1892 if (write) {
1893 if (val[0] < 0 || val[1] < 0 ||
1894 (val[0] >= val[1] && val[1]))
1895 rc = -EINVAL;
1896 else
1897 memcpy(valp, val, sizeof(val));
1898 }
1899 mutex_unlock(&ipvs->sync_mutex);
1900 return rc;
1901 }
1902
1903 static int
proc_do_sync_ports(struct ctl_table * table,int write,void * buffer,size_t * lenp,loff_t * ppos)1904 proc_do_sync_ports(struct ctl_table *table, int write,
1905 void *buffer, size_t *lenp, loff_t *ppos)
1906 {
1907 int *valp = table->data;
1908 int val = *valp;
1909 int rc;
1910
1911 struct ctl_table tmp = {
1912 .data = &val,
1913 .maxlen = sizeof(int),
1914 .mode = table->mode,
1915 };
1916
1917 rc = proc_dointvec(&tmp, write, buffer, lenp, ppos);
1918 if (write && (*valp != val)) {
1919 if (val < 1 || !is_power_of_2(val))
1920 rc = -EINVAL;
1921 else
1922 *valp = val;
1923 }
1924 return rc;
1925 }
1926
ipvs_proc_est_cpumask_set(struct ctl_table * table,void * buffer)1927 static int ipvs_proc_est_cpumask_set(struct ctl_table *table, void *buffer)
1928 {
1929 struct netns_ipvs *ipvs = table->extra2;
1930 cpumask_var_t *valp = table->data;
1931 cpumask_var_t newmask;
1932 int ret;
1933
1934 if (!zalloc_cpumask_var(&newmask, GFP_KERNEL))
1935 return -ENOMEM;
1936
1937 ret = cpulist_parse(buffer, newmask);
1938 if (ret)
1939 goto out;
1940
1941 mutex_lock(&ipvs->est_mutex);
1942
1943 if (!ipvs->est_cpulist_valid) {
1944 if (!zalloc_cpumask_var(valp, GFP_KERNEL)) {
1945 ret = -ENOMEM;
1946 goto unlock;
1947 }
1948 ipvs->est_cpulist_valid = 1;
1949 }
1950 cpumask_and(newmask, newmask, ¤t->cpus_mask);
1951 cpumask_copy(*valp, newmask);
1952 /* est_max_threads may depend on cpulist size */
1953 ipvs->est_max_threads = ip_vs_est_max_threads(ipvs);
1954 ipvs->est_calc_phase = 1;
1955 ip_vs_est_reload_start(ipvs);
1956
1957 unlock:
1958 mutex_unlock(&ipvs->est_mutex);
1959
1960 out:
1961 free_cpumask_var(newmask);
1962 return ret;
1963 }
1964
ipvs_proc_est_cpumask_get(struct ctl_table * table,void * buffer,size_t size)1965 static int ipvs_proc_est_cpumask_get(struct ctl_table *table, void *buffer,
1966 size_t size)
1967 {
1968 struct netns_ipvs *ipvs = table->extra2;
1969 cpumask_var_t *valp = table->data;
1970 struct cpumask *mask;
1971 int ret;
1972
1973 mutex_lock(&ipvs->est_mutex);
1974
1975 if (ipvs->est_cpulist_valid)
1976 mask = *valp;
1977 else
1978 mask = (struct cpumask *)housekeeping_cpumask(HK_TYPE_KTHREAD);
1979 ret = scnprintf(buffer, size, "%*pbl\n", cpumask_pr_args(mask));
1980
1981 mutex_unlock(&ipvs->est_mutex);
1982
1983 return ret;
1984 }
1985
ipvs_proc_est_cpulist(struct ctl_table * table,int write,void * buffer,size_t * lenp,loff_t * ppos)1986 static int ipvs_proc_est_cpulist(struct ctl_table *table, int write,
1987 void *buffer, size_t *lenp, loff_t *ppos)
1988 {
1989 int ret;
1990
1991 /* Ignore both read and write(append) if *ppos not 0 */
1992 if (*ppos || !*lenp) {
1993 *lenp = 0;
1994 return 0;
1995 }
1996 if (write) {
1997 /* proc_sys_call_handler() appends terminator */
1998 ret = ipvs_proc_est_cpumask_set(table, buffer);
1999 if (ret >= 0)
2000 *ppos += *lenp;
2001 } else {
2002 /* proc_sys_call_handler() allocates 1 byte for terminator */
2003 ret = ipvs_proc_est_cpumask_get(table, buffer, *lenp + 1);
2004 if (ret >= 0) {
2005 *lenp = ret;
2006 *ppos += *lenp;
2007 ret = 0;
2008 }
2009 }
2010 return ret;
2011 }
2012
ipvs_proc_est_nice(struct ctl_table * table,int write,void * buffer,size_t * lenp,loff_t * ppos)2013 static int ipvs_proc_est_nice(struct ctl_table *table, int write,
2014 void *buffer, size_t *lenp, loff_t *ppos)
2015 {
2016 struct netns_ipvs *ipvs = table->extra2;
2017 int *valp = table->data;
2018 int val = *valp;
2019 int ret;
2020
2021 struct ctl_table tmp_table = {
2022 .data = &val,
2023 .maxlen = sizeof(int),
2024 .mode = table->mode,
2025 };
2026
2027 ret = proc_dointvec(&tmp_table, write, buffer, lenp, ppos);
2028 if (write && ret >= 0) {
2029 if (val < MIN_NICE || val > MAX_NICE) {
2030 ret = -EINVAL;
2031 } else {
2032 mutex_lock(&ipvs->est_mutex);
2033 if (*valp != val) {
2034 *valp = val;
2035 ip_vs_est_reload_start(ipvs);
2036 }
2037 mutex_unlock(&ipvs->est_mutex);
2038 }
2039 }
2040 return ret;
2041 }
2042
ipvs_proc_run_estimation(struct ctl_table * table,int write,void * buffer,size_t * lenp,loff_t * ppos)2043 static int ipvs_proc_run_estimation(struct ctl_table *table, int write,
2044 void *buffer, size_t *lenp, loff_t *ppos)
2045 {
2046 struct netns_ipvs *ipvs = table->extra2;
2047 int *valp = table->data;
2048 int val = *valp;
2049 int ret;
2050
2051 struct ctl_table tmp_table = {
2052 .data = &val,
2053 .maxlen = sizeof(int),
2054 .mode = table->mode,
2055 };
2056
2057 ret = proc_dointvec(&tmp_table, write, buffer, lenp, ppos);
2058 if (write && ret >= 0) {
2059 mutex_lock(&ipvs->est_mutex);
2060 if (*valp != val) {
2061 *valp = val;
2062 ip_vs_est_reload_start(ipvs);
2063 }
2064 mutex_unlock(&ipvs->est_mutex);
2065 }
2066 return ret;
2067 }
2068
2069 /*
2070 * IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
2071 * Do not change order or insert new entries without
2072 * align with netns init in ip_vs_control_net_init()
2073 */
2074
2075 static struct ctl_table vs_vars[] = {
2076 {
2077 .procname = "amemthresh",
2078 .maxlen = sizeof(int),
2079 .mode = 0644,
2080 .proc_handler = proc_dointvec,
2081 },
2082 {
2083 .procname = "am_droprate",
2084 .maxlen = sizeof(int),
2085 .mode = 0644,
2086 .proc_handler = proc_dointvec,
2087 },
2088 {
2089 .procname = "drop_entry",
2090 .maxlen = sizeof(int),
2091 .mode = 0644,
2092 .proc_handler = proc_do_defense_mode,
2093 },
2094 {
2095 .procname = "drop_packet",
2096 .maxlen = sizeof(int),
2097 .mode = 0644,
2098 .proc_handler = proc_do_defense_mode,
2099 },
2100 #ifdef CONFIG_IP_VS_NFCT
2101 {
2102 .procname = "conntrack",
2103 .maxlen = sizeof(int),
2104 .mode = 0644,
2105 .proc_handler = &proc_dointvec,
2106 },
2107 #endif
2108 {
2109 .procname = "secure_tcp",
2110 .maxlen = sizeof(int),
2111 .mode = 0644,
2112 .proc_handler = proc_do_defense_mode,
2113 },
2114 {
2115 .procname = "snat_reroute",
2116 .maxlen = sizeof(int),
2117 .mode = 0644,
2118 .proc_handler = &proc_dointvec,
2119 },
2120 {
2121 .procname = "sync_version",
2122 .maxlen = sizeof(int),
2123 .mode = 0644,
2124 .proc_handler = proc_dointvec_minmax,
2125 .extra1 = SYSCTL_ZERO,
2126 .extra2 = SYSCTL_ONE,
2127 },
2128 {
2129 .procname = "sync_ports",
2130 .maxlen = sizeof(int),
2131 .mode = 0644,
2132 .proc_handler = proc_do_sync_ports,
2133 },
2134 {
2135 .procname = "sync_persist_mode",
2136 .maxlen = sizeof(int),
2137 .mode = 0644,
2138 .proc_handler = proc_dointvec,
2139 },
2140 {
2141 .procname = "sync_qlen_max",
2142 .maxlen = sizeof(unsigned long),
2143 .mode = 0644,
2144 .proc_handler = proc_doulongvec_minmax,
2145 },
2146 {
2147 .procname = "sync_sock_size",
2148 .maxlen = sizeof(int),
2149 .mode = 0644,
2150 .proc_handler = proc_dointvec,
2151 },
2152 {
2153 .procname = "cache_bypass",
2154 .maxlen = sizeof(int),
2155 .mode = 0644,
2156 .proc_handler = proc_dointvec,
2157 },
2158 {
2159 .procname = "expire_nodest_conn",
2160 .maxlen = sizeof(int),
2161 .mode = 0644,
2162 .proc_handler = proc_dointvec,
2163 },
2164 {
2165 .procname = "sloppy_tcp",
2166 .maxlen = sizeof(int),
2167 .mode = 0644,
2168 .proc_handler = proc_dointvec,
2169 },
2170 {
2171 .procname = "sloppy_sctp",
2172 .maxlen = sizeof(int),
2173 .mode = 0644,
2174 .proc_handler = proc_dointvec,
2175 },
2176 {
2177 .procname = "expire_quiescent_template",
2178 .maxlen = sizeof(int),
2179 .mode = 0644,
2180 .proc_handler = proc_dointvec,
2181 },
2182 {
2183 .procname = "sync_threshold",
2184 .maxlen =
2185 sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
2186 .mode = 0644,
2187 .proc_handler = proc_do_sync_threshold,
2188 },
2189 {
2190 .procname = "sync_refresh_period",
2191 .maxlen = sizeof(int),
2192 .mode = 0644,
2193 .proc_handler = proc_dointvec_jiffies,
2194 },
2195 {
2196 .procname = "sync_retries",
2197 .maxlen = sizeof(int),
2198 .mode = 0644,
2199 .proc_handler = proc_dointvec_minmax,
2200 .extra1 = SYSCTL_ZERO,
2201 .extra2 = SYSCTL_THREE,
2202 },
2203 {
2204 .procname = "nat_icmp_send",
2205 .maxlen = sizeof(int),
2206 .mode = 0644,
2207 .proc_handler = proc_dointvec,
2208 },
2209 {
2210 .procname = "pmtu_disc",
2211 .maxlen = sizeof(int),
2212 .mode = 0644,
2213 .proc_handler = proc_dointvec,
2214 },
2215 {
2216 .procname = "backup_only",
2217 .maxlen = sizeof(int),
2218 .mode = 0644,
2219 .proc_handler = proc_dointvec,
2220 },
2221 {
2222 .procname = "conn_reuse_mode",
2223 .maxlen = sizeof(int),
2224 .mode = 0644,
2225 .proc_handler = proc_dointvec,
2226 },
2227 {
2228 .procname = "schedule_icmp",
2229 .maxlen = sizeof(int),
2230 .mode = 0644,
2231 .proc_handler = proc_dointvec,
2232 },
2233 {
2234 .procname = "ignore_tunneled",
2235 .maxlen = sizeof(int),
2236 .mode = 0644,
2237 .proc_handler = proc_dointvec,
2238 },
2239 {
2240 .procname = "run_estimation",
2241 .maxlen = sizeof(int),
2242 .mode = 0644,
2243 .proc_handler = ipvs_proc_run_estimation,
2244 },
2245 {
2246 .procname = "est_cpulist",
2247 .maxlen = NR_CPUS, /* unused */
2248 .mode = 0644,
2249 .proc_handler = ipvs_proc_est_cpulist,
2250 },
2251 {
2252 .procname = "est_nice",
2253 .maxlen = sizeof(int),
2254 .mode = 0644,
2255 .proc_handler = ipvs_proc_est_nice,
2256 },
2257 #ifdef CONFIG_IP_VS_DEBUG
2258 {
2259 .procname = "debug_level",
2260 .data = &sysctl_ip_vs_debug_level,
2261 .maxlen = sizeof(int),
2262 .mode = 0644,
2263 .proc_handler = proc_dointvec,
2264 },
2265 #endif
2266 };
2267
2268 #endif
2269
2270 #ifdef CONFIG_PROC_FS
2271
2272 struct ip_vs_iter {
2273 struct seq_net_private p; /* Do not move this, netns depends upon it*/
2274 struct hlist_head *table;
2275 int bucket;
2276 };
2277
2278 /*
2279 * Write the contents of the VS rule table to a PROCfs file.
2280 * (It is kept just for backward compatibility)
2281 */
ip_vs_fwd_name(unsigned int flags)2282 static inline const char *ip_vs_fwd_name(unsigned int flags)
2283 {
2284 switch (flags & IP_VS_CONN_F_FWD_MASK) {
2285 case IP_VS_CONN_F_LOCALNODE:
2286 return "Local";
2287 case IP_VS_CONN_F_TUNNEL:
2288 return "Tunnel";
2289 case IP_VS_CONN_F_DROUTE:
2290 return "Route";
2291 default:
2292 return "Masq";
2293 }
2294 }
2295
2296
2297 /* Get the Nth entry in the two lists */
ip_vs_info_array(struct seq_file * seq,loff_t pos)2298 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
2299 {
2300 struct net *net = seq_file_net(seq);
2301 struct netns_ipvs *ipvs = net_ipvs(net);
2302 struct ip_vs_iter *iter = seq->private;
2303 int idx;
2304 struct ip_vs_service *svc;
2305
2306 /* look in hash by protocol */
2307 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2308 hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
2309 if ((svc->ipvs == ipvs) && pos-- == 0) {
2310 iter->table = ip_vs_svc_table;
2311 iter->bucket = idx;
2312 return svc;
2313 }
2314 }
2315 }
2316
2317 /* keep looking in fwmark */
2318 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2319 hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
2320 f_list) {
2321 if ((svc->ipvs == ipvs) && pos-- == 0) {
2322 iter->table = ip_vs_svc_fwm_table;
2323 iter->bucket = idx;
2324 return svc;
2325 }
2326 }
2327 }
2328
2329 return NULL;
2330 }
2331
ip_vs_info_seq_start(struct seq_file * seq,loff_t * pos)2332 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
2333 __acquires(RCU)
2334 {
2335 rcu_read_lock();
2336 return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
2337 }
2338
2339
ip_vs_info_seq_next(struct seq_file * seq,void * v,loff_t * pos)2340 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2341 {
2342 struct hlist_node *e;
2343 struct ip_vs_iter *iter;
2344 struct ip_vs_service *svc;
2345
2346 ++*pos;
2347 if (v == SEQ_START_TOKEN)
2348 return ip_vs_info_array(seq,0);
2349
2350 svc = v;
2351 iter = seq->private;
2352
2353 if (iter->table == ip_vs_svc_table) {
2354 /* next service in table hashed by protocol */
2355 e = rcu_dereference(hlist_next_rcu(&svc->s_list));
2356 if (e)
2357 return hlist_entry(e, struct ip_vs_service, s_list);
2358
2359 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2360 hlist_for_each_entry_rcu(svc,
2361 &ip_vs_svc_table[iter->bucket],
2362 s_list) {
2363 return svc;
2364 }
2365 }
2366
2367 iter->table = ip_vs_svc_fwm_table;
2368 iter->bucket = -1;
2369 goto scan_fwmark;
2370 }
2371
2372 /* next service in hashed by fwmark */
2373 e = rcu_dereference(hlist_next_rcu(&svc->f_list));
2374 if (e)
2375 return hlist_entry(e, struct ip_vs_service, f_list);
2376
2377 scan_fwmark:
2378 while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2379 hlist_for_each_entry_rcu(svc,
2380 &ip_vs_svc_fwm_table[iter->bucket],
2381 f_list)
2382 return svc;
2383 }
2384
2385 return NULL;
2386 }
2387
ip_vs_info_seq_stop(struct seq_file * seq,void * v)2388 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
2389 __releases(RCU)
2390 {
2391 rcu_read_unlock();
2392 }
2393
2394
ip_vs_info_seq_show(struct seq_file * seq,void * v)2395 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
2396 {
2397 if (v == SEQ_START_TOKEN) {
2398 seq_printf(seq,
2399 "IP Virtual Server version %d.%d.%d (size=%d)\n",
2400 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2401 seq_puts(seq,
2402 "Prot LocalAddress:Port Scheduler Flags\n");
2403 seq_puts(seq,
2404 " -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2405 } else {
2406 struct net *net = seq_file_net(seq);
2407 struct netns_ipvs *ipvs = net_ipvs(net);
2408 const struct ip_vs_service *svc = v;
2409 const struct ip_vs_iter *iter = seq->private;
2410 const struct ip_vs_dest *dest;
2411 struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler);
2412 char *sched_name = sched ? sched->name : "none";
2413
2414 if (svc->ipvs != ipvs)
2415 return 0;
2416 if (iter->table == ip_vs_svc_table) {
2417 #ifdef CONFIG_IP_VS_IPV6
2418 if (svc->af == AF_INET6)
2419 seq_printf(seq, "%s [%pI6]:%04X %s ",
2420 ip_vs_proto_name(svc->protocol),
2421 &svc->addr.in6,
2422 ntohs(svc->port),
2423 sched_name);
2424 else
2425 #endif
2426 seq_printf(seq, "%s %08X:%04X %s %s ",
2427 ip_vs_proto_name(svc->protocol),
2428 ntohl(svc->addr.ip),
2429 ntohs(svc->port),
2430 sched_name,
2431 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2432 } else {
2433 seq_printf(seq, "FWM %08X %s %s",
2434 svc->fwmark, sched_name,
2435 (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2436 }
2437
2438 if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2439 seq_printf(seq, "persistent %d %08X\n",
2440 svc->timeout,
2441 ntohl(svc->netmask));
2442 else
2443 seq_putc(seq, '\n');
2444
2445 list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
2446 #ifdef CONFIG_IP_VS_IPV6
2447 if (dest->af == AF_INET6)
2448 seq_printf(seq,
2449 " -> [%pI6]:%04X"
2450 " %-7s %-6d %-10d %-10d\n",
2451 &dest->addr.in6,
2452 ntohs(dest->port),
2453 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2454 atomic_read(&dest->weight),
2455 atomic_read(&dest->activeconns),
2456 atomic_read(&dest->inactconns));
2457 else
2458 #endif
2459 seq_printf(seq,
2460 " -> %08X:%04X "
2461 "%-7s %-6d %-10d %-10d\n",
2462 ntohl(dest->addr.ip),
2463 ntohs(dest->port),
2464 ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2465 atomic_read(&dest->weight),
2466 atomic_read(&dest->activeconns),
2467 atomic_read(&dest->inactconns));
2468
2469 }
2470 }
2471 return 0;
2472 }
2473
2474 static const struct seq_operations ip_vs_info_seq_ops = {
2475 .start = ip_vs_info_seq_start,
2476 .next = ip_vs_info_seq_next,
2477 .stop = ip_vs_info_seq_stop,
2478 .show = ip_vs_info_seq_show,
2479 };
2480
ip_vs_stats_show(struct seq_file * seq,void * v)2481 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2482 {
2483 struct net *net = seq_file_single_net(seq);
2484 struct ip_vs_kstats show;
2485
2486 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2487 seq_puts(seq,
2488 " Total Incoming Outgoing Incoming Outgoing\n");
2489 seq_puts(seq,
2490 " Conns Packets Packets Bytes Bytes\n");
2491
2492 ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats->s);
2493 seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n\n",
2494 (unsigned long long)show.conns,
2495 (unsigned long long)show.inpkts,
2496 (unsigned long long)show.outpkts,
2497 (unsigned long long)show.inbytes,
2498 (unsigned long long)show.outbytes);
2499
2500 /* 01234567 01234567 01234567 0123456701234567 0123456701234567*/
2501 seq_puts(seq,
2502 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2503 seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n",
2504 (unsigned long long)show.cps,
2505 (unsigned long long)show.inpps,
2506 (unsigned long long)show.outpps,
2507 (unsigned long long)show.inbps,
2508 (unsigned long long)show.outbps);
2509
2510 return 0;
2511 }
2512
ip_vs_stats_percpu_show(struct seq_file * seq,void * v)2513 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2514 {
2515 struct net *net = seq_file_single_net(seq);
2516 struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats->s;
2517 struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats;
2518 struct ip_vs_kstats kstats;
2519 int i;
2520
2521 /* 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2522 seq_puts(seq,
2523 " Total Incoming Outgoing Incoming Outgoing\n");
2524 seq_puts(seq,
2525 "CPU Conns Packets Packets Bytes Bytes\n");
2526
2527 for_each_possible_cpu(i) {
2528 struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2529 unsigned int start;
2530 u64 conns, inpkts, outpkts, inbytes, outbytes;
2531
2532 do {
2533 start = u64_stats_fetch_begin(&u->syncp);
2534 conns = u64_stats_read(&u->cnt.conns);
2535 inpkts = u64_stats_read(&u->cnt.inpkts);
2536 outpkts = u64_stats_read(&u->cnt.outpkts);
2537 inbytes = u64_stats_read(&u->cnt.inbytes);
2538 outbytes = u64_stats_read(&u->cnt.outbytes);
2539 } while (u64_stats_fetch_retry(&u->syncp, start));
2540
2541 seq_printf(seq, "%3X %8LX %8LX %8LX %16LX %16LX\n",
2542 i, (u64)conns, (u64)inpkts,
2543 (u64)outpkts, (u64)inbytes,
2544 (u64)outbytes);
2545 }
2546
2547 ip_vs_copy_stats(&kstats, tot_stats);
2548
2549 seq_printf(seq, " ~ %8LX %8LX %8LX %16LX %16LX\n\n",
2550 (unsigned long long)kstats.conns,
2551 (unsigned long long)kstats.inpkts,
2552 (unsigned long long)kstats.outpkts,
2553 (unsigned long long)kstats.inbytes,
2554 (unsigned long long)kstats.outbytes);
2555
2556 /* ... 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2557 seq_puts(seq,
2558 " Conns/s Pkts/s Pkts/s Bytes/s Bytes/s\n");
2559 seq_printf(seq, " %8LX %8LX %8LX %16LX %16LX\n",
2560 kstats.cps,
2561 kstats.inpps,
2562 kstats.outpps,
2563 kstats.inbps,
2564 kstats.outbps);
2565
2566 return 0;
2567 }
2568 #endif
2569
2570 /*
2571 * Set timeout values for tcp tcpfin udp in the timeout_table.
2572 */
ip_vs_set_timeout(struct netns_ipvs * ipvs,struct ip_vs_timeout_user * u)2573 static int ip_vs_set_timeout(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2574 {
2575 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2576 struct ip_vs_proto_data *pd;
2577 #endif
2578
2579 IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2580 u->tcp_timeout,
2581 u->tcp_fin_timeout,
2582 u->udp_timeout);
2583
2584 #ifdef CONFIG_IP_VS_PROTO_TCP
2585 if (u->tcp_timeout < 0 || u->tcp_timeout > (INT_MAX / HZ) ||
2586 u->tcp_fin_timeout < 0 || u->tcp_fin_timeout > (INT_MAX / HZ)) {
2587 return -EINVAL;
2588 }
2589 #endif
2590
2591 #ifdef CONFIG_IP_VS_PROTO_UDP
2592 if (u->udp_timeout < 0 || u->udp_timeout > (INT_MAX / HZ))
2593 return -EINVAL;
2594 #endif
2595
2596 #ifdef CONFIG_IP_VS_PROTO_TCP
2597 if (u->tcp_timeout) {
2598 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2599 pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2600 = u->tcp_timeout * HZ;
2601 }
2602
2603 if (u->tcp_fin_timeout) {
2604 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2605 pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2606 = u->tcp_fin_timeout * HZ;
2607 }
2608 #endif
2609
2610 #ifdef CONFIG_IP_VS_PROTO_UDP
2611 if (u->udp_timeout) {
2612 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2613 pd->timeout_table[IP_VS_UDP_S_NORMAL]
2614 = u->udp_timeout * HZ;
2615 }
2616 #endif
2617 return 0;
2618 }
2619
2620 #define CMDID(cmd) (cmd - IP_VS_BASE_CTL)
2621
2622 struct ip_vs_svcdest_user {
2623 struct ip_vs_service_user s;
2624 struct ip_vs_dest_user d;
2625 };
2626
2627 static const unsigned char set_arglen[CMDID(IP_VS_SO_SET_MAX) + 1] = {
2628 [CMDID(IP_VS_SO_SET_ADD)] = sizeof(struct ip_vs_service_user),
2629 [CMDID(IP_VS_SO_SET_EDIT)] = sizeof(struct ip_vs_service_user),
2630 [CMDID(IP_VS_SO_SET_DEL)] = sizeof(struct ip_vs_service_user),
2631 [CMDID(IP_VS_SO_SET_ADDDEST)] = sizeof(struct ip_vs_svcdest_user),
2632 [CMDID(IP_VS_SO_SET_DELDEST)] = sizeof(struct ip_vs_svcdest_user),
2633 [CMDID(IP_VS_SO_SET_EDITDEST)] = sizeof(struct ip_vs_svcdest_user),
2634 [CMDID(IP_VS_SO_SET_TIMEOUT)] = sizeof(struct ip_vs_timeout_user),
2635 [CMDID(IP_VS_SO_SET_STARTDAEMON)] = sizeof(struct ip_vs_daemon_user),
2636 [CMDID(IP_VS_SO_SET_STOPDAEMON)] = sizeof(struct ip_vs_daemon_user),
2637 [CMDID(IP_VS_SO_SET_ZERO)] = sizeof(struct ip_vs_service_user),
2638 };
2639
2640 union ip_vs_set_arglen {
2641 struct ip_vs_service_user field_IP_VS_SO_SET_ADD;
2642 struct ip_vs_service_user field_IP_VS_SO_SET_EDIT;
2643 struct ip_vs_service_user field_IP_VS_SO_SET_DEL;
2644 struct ip_vs_svcdest_user field_IP_VS_SO_SET_ADDDEST;
2645 struct ip_vs_svcdest_user field_IP_VS_SO_SET_DELDEST;
2646 struct ip_vs_svcdest_user field_IP_VS_SO_SET_EDITDEST;
2647 struct ip_vs_timeout_user field_IP_VS_SO_SET_TIMEOUT;
2648 struct ip_vs_daemon_user field_IP_VS_SO_SET_STARTDAEMON;
2649 struct ip_vs_daemon_user field_IP_VS_SO_SET_STOPDAEMON;
2650 struct ip_vs_service_user field_IP_VS_SO_SET_ZERO;
2651 };
2652
2653 #define MAX_SET_ARGLEN sizeof(union ip_vs_set_arglen)
2654
ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern * usvc,struct ip_vs_service_user * usvc_compat)2655 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2656 struct ip_vs_service_user *usvc_compat)
2657 {
2658 memset(usvc, 0, sizeof(*usvc));
2659
2660 usvc->af = AF_INET;
2661 usvc->protocol = usvc_compat->protocol;
2662 usvc->addr.ip = usvc_compat->addr;
2663 usvc->port = usvc_compat->port;
2664 usvc->fwmark = usvc_compat->fwmark;
2665
2666 /* Deep copy of sched_name is not needed here */
2667 usvc->sched_name = usvc_compat->sched_name;
2668
2669 usvc->flags = usvc_compat->flags;
2670 usvc->timeout = usvc_compat->timeout;
2671 usvc->netmask = usvc_compat->netmask;
2672 }
2673
ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern * udest,struct ip_vs_dest_user * udest_compat)2674 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2675 struct ip_vs_dest_user *udest_compat)
2676 {
2677 memset(udest, 0, sizeof(*udest));
2678
2679 udest->addr.ip = udest_compat->addr;
2680 udest->port = udest_compat->port;
2681 udest->conn_flags = udest_compat->conn_flags;
2682 udest->weight = udest_compat->weight;
2683 udest->u_threshold = udest_compat->u_threshold;
2684 udest->l_threshold = udest_compat->l_threshold;
2685 udest->af = AF_INET;
2686 udest->tun_type = IP_VS_CONN_F_TUNNEL_TYPE_IPIP;
2687 }
2688
2689 static int
do_ip_vs_set_ctl(struct sock * sk,int cmd,sockptr_t ptr,unsigned int len)2690 do_ip_vs_set_ctl(struct sock *sk, int cmd, sockptr_t ptr, unsigned int len)
2691 {
2692 struct net *net = sock_net(sk);
2693 int ret;
2694 unsigned char arg[MAX_SET_ARGLEN];
2695 struct ip_vs_service_user *usvc_compat;
2696 struct ip_vs_service_user_kern usvc;
2697 struct ip_vs_service *svc;
2698 struct ip_vs_dest_user *udest_compat;
2699 struct ip_vs_dest_user_kern udest;
2700 struct netns_ipvs *ipvs = net_ipvs(net);
2701
2702 BUILD_BUG_ON(sizeof(arg) > 255);
2703 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2704 return -EPERM;
2705
2706 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2707 return -EINVAL;
2708 if (len != set_arglen[CMDID(cmd)]) {
2709 IP_VS_DBG(1, "set_ctl: len %u != %u\n",
2710 len, set_arglen[CMDID(cmd)]);
2711 return -EINVAL;
2712 }
2713
2714 if (copy_from_sockptr(arg, ptr, len) != 0)
2715 return -EFAULT;
2716
2717 /* Handle daemons since they have another lock */
2718 if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2719 cmd == IP_VS_SO_SET_STOPDAEMON) {
2720 struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2721
2722 if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2723 struct ipvs_sync_daemon_cfg cfg;
2724
2725 memset(&cfg, 0, sizeof(cfg));
2726 ret = -EINVAL;
2727 if (strscpy(cfg.mcast_ifn, dm->mcast_ifn,
2728 sizeof(cfg.mcast_ifn)) <= 0)
2729 return ret;
2730 cfg.syncid = dm->syncid;
2731 ret = start_sync_thread(ipvs, &cfg, dm->state);
2732 } else {
2733 ret = stop_sync_thread(ipvs, dm->state);
2734 }
2735 return ret;
2736 }
2737
2738 mutex_lock(&__ip_vs_mutex);
2739 if (cmd == IP_VS_SO_SET_FLUSH) {
2740 /* Flush the virtual service */
2741 ret = ip_vs_flush(ipvs, false);
2742 goto out_unlock;
2743 } else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2744 /* Set timeout values for (tcp tcpfin udp) */
2745 ret = ip_vs_set_timeout(ipvs, (struct ip_vs_timeout_user *)arg);
2746 goto out_unlock;
2747 } else if (!len) {
2748 /* No more commands with len == 0 below */
2749 ret = -EINVAL;
2750 goto out_unlock;
2751 }
2752
2753 usvc_compat = (struct ip_vs_service_user *)arg;
2754 udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2755
2756 /* We only use the new structs internally, so copy userspace compat
2757 * structs to extended internal versions */
2758 ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2759 ip_vs_copy_udest_compat(&udest, udest_compat);
2760
2761 if (cmd == IP_VS_SO_SET_ZERO) {
2762 /* if no service address is set, zero counters in all */
2763 if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2764 ret = ip_vs_zero_all(ipvs);
2765 goto out_unlock;
2766 }
2767 }
2768
2769 if ((cmd == IP_VS_SO_SET_ADD || cmd == IP_VS_SO_SET_EDIT) &&
2770 strnlen(usvc.sched_name, IP_VS_SCHEDNAME_MAXLEN) ==
2771 IP_VS_SCHEDNAME_MAXLEN) {
2772 ret = -EINVAL;
2773 goto out_unlock;
2774 }
2775
2776 /* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2777 if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2778 usvc.protocol != IPPROTO_SCTP) {
2779 pr_err("set_ctl: invalid protocol: %d %pI4:%d\n",
2780 usvc.protocol, &usvc.addr.ip,
2781 ntohs(usvc.port));
2782 ret = -EFAULT;
2783 goto out_unlock;
2784 }
2785
2786 /* Lookup the exact service by <protocol, addr, port> or fwmark */
2787 rcu_read_lock();
2788 if (usvc.fwmark == 0)
2789 svc = __ip_vs_service_find(ipvs, usvc.af, usvc.protocol,
2790 &usvc.addr, usvc.port);
2791 else
2792 svc = __ip_vs_svc_fwm_find(ipvs, usvc.af, usvc.fwmark);
2793 rcu_read_unlock();
2794
2795 if (cmd != IP_VS_SO_SET_ADD
2796 && (svc == NULL || svc->protocol != usvc.protocol)) {
2797 ret = -ESRCH;
2798 goto out_unlock;
2799 }
2800
2801 switch (cmd) {
2802 case IP_VS_SO_SET_ADD:
2803 if (svc != NULL)
2804 ret = -EEXIST;
2805 else
2806 ret = ip_vs_add_service(ipvs, &usvc, &svc);
2807 break;
2808 case IP_VS_SO_SET_EDIT:
2809 ret = ip_vs_edit_service(svc, &usvc);
2810 break;
2811 case IP_VS_SO_SET_DEL:
2812 ret = ip_vs_del_service(svc);
2813 if (!ret)
2814 goto out_unlock;
2815 break;
2816 case IP_VS_SO_SET_ZERO:
2817 ret = ip_vs_zero_service(svc);
2818 break;
2819 case IP_VS_SO_SET_ADDDEST:
2820 ret = ip_vs_add_dest(svc, &udest);
2821 break;
2822 case IP_VS_SO_SET_EDITDEST:
2823 ret = ip_vs_edit_dest(svc, &udest);
2824 break;
2825 case IP_VS_SO_SET_DELDEST:
2826 ret = ip_vs_del_dest(svc, &udest);
2827 break;
2828 default:
2829 WARN_ON_ONCE(1);
2830 ret = -EINVAL;
2831 break;
2832 }
2833
2834 out_unlock:
2835 mutex_unlock(&__ip_vs_mutex);
2836 return ret;
2837 }
2838
2839
2840 static void
ip_vs_copy_service(struct ip_vs_service_entry * dst,struct ip_vs_service * src)2841 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2842 {
2843 struct ip_vs_scheduler *sched;
2844 struct ip_vs_kstats kstats;
2845 char *sched_name;
2846
2847 sched = rcu_dereference_protected(src->scheduler, 1);
2848 sched_name = sched ? sched->name : "none";
2849 dst->protocol = src->protocol;
2850 dst->addr = src->addr.ip;
2851 dst->port = src->port;
2852 dst->fwmark = src->fwmark;
2853 strscpy(dst->sched_name, sched_name, sizeof(dst->sched_name));
2854 dst->flags = src->flags;
2855 dst->timeout = src->timeout / HZ;
2856 dst->netmask = src->netmask;
2857 dst->num_dests = src->num_dests;
2858 ip_vs_copy_stats(&kstats, &src->stats);
2859 ip_vs_export_stats_user(&dst->stats, &kstats);
2860 }
2861
2862 static inline int
__ip_vs_get_service_entries(struct netns_ipvs * ipvs,const struct ip_vs_get_services * get,struct ip_vs_get_services __user * uptr)2863 __ip_vs_get_service_entries(struct netns_ipvs *ipvs,
2864 const struct ip_vs_get_services *get,
2865 struct ip_vs_get_services __user *uptr)
2866 {
2867 int idx, count=0;
2868 struct ip_vs_service *svc;
2869 struct ip_vs_service_entry entry;
2870 int ret = 0;
2871
2872 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2873 hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2874 /* Only expose IPv4 entries to old interface */
2875 if (svc->af != AF_INET || (svc->ipvs != ipvs))
2876 continue;
2877
2878 if (count >= get->num_services)
2879 goto out;
2880 memset(&entry, 0, sizeof(entry));
2881 ip_vs_copy_service(&entry, svc);
2882 if (copy_to_user(&uptr->entrytable[count],
2883 &entry, sizeof(entry))) {
2884 ret = -EFAULT;
2885 goto out;
2886 }
2887 count++;
2888 }
2889 }
2890
2891 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2892 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2893 /* Only expose IPv4 entries to old interface */
2894 if (svc->af != AF_INET || (svc->ipvs != ipvs))
2895 continue;
2896
2897 if (count >= get->num_services)
2898 goto out;
2899 memset(&entry, 0, sizeof(entry));
2900 ip_vs_copy_service(&entry, svc);
2901 if (copy_to_user(&uptr->entrytable[count],
2902 &entry, sizeof(entry))) {
2903 ret = -EFAULT;
2904 goto out;
2905 }
2906 count++;
2907 }
2908 }
2909 out:
2910 return ret;
2911 }
2912
2913 static inline int
__ip_vs_get_dest_entries(struct netns_ipvs * ipvs,const struct ip_vs_get_dests * get,struct ip_vs_get_dests __user * uptr)2914 __ip_vs_get_dest_entries(struct netns_ipvs *ipvs, const struct ip_vs_get_dests *get,
2915 struct ip_vs_get_dests __user *uptr)
2916 {
2917 struct ip_vs_service *svc;
2918 union nf_inet_addr addr = { .ip = get->addr };
2919 int ret = 0;
2920
2921 rcu_read_lock();
2922 if (get->fwmark)
2923 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, get->fwmark);
2924 else
2925 svc = __ip_vs_service_find(ipvs, AF_INET, get->protocol, &addr,
2926 get->port);
2927 rcu_read_unlock();
2928
2929 if (svc) {
2930 int count = 0;
2931 struct ip_vs_dest *dest;
2932 struct ip_vs_dest_entry entry;
2933 struct ip_vs_kstats kstats;
2934
2935 memset(&entry, 0, sizeof(entry));
2936 list_for_each_entry(dest, &svc->destinations, n_list) {
2937 if (count >= get->num_dests)
2938 break;
2939
2940 /* Cannot expose heterogeneous members via sockopt
2941 * interface
2942 */
2943 if (dest->af != svc->af)
2944 continue;
2945
2946 entry.addr = dest->addr.ip;
2947 entry.port = dest->port;
2948 entry.conn_flags = atomic_read(&dest->conn_flags);
2949 entry.weight = atomic_read(&dest->weight);
2950 entry.u_threshold = dest->u_threshold;
2951 entry.l_threshold = dest->l_threshold;
2952 entry.activeconns = atomic_read(&dest->activeconns);
2953 entry.inactconns = atomic_read(&dest->inactconns);
2954 entry.persistconns = atomic_read(&dest->persistconns);
2955 ip_vs_copy_stats(&kstats, &dest->stats);
2956 ip_vs_export_stats_user(&entry.stats, &kstats);
2957 if (copy_to_user(&uptr->entrytable[count],
2958 &entry, sizeof(entry))) {
2959 ret = -EFAULT;
2960 break;
2961 }
2962 count++;
2963 }
2964 } else
2965 ret = -ESRCH;
2966 return ret;
2967 }
2968
2969 static inline void
__ip_vs_get_timeouts(struct netns_ipvs * ipvs,struct ip_vs_timeout_user * u)2970 __ip_vs_get_timeouts(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2971 {
2972 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2973 struct ip_vs_proto_data *pd;
2974 #endif
2975
2976 memset(u, 0, sizeof (*u));
2977
2978 #ifdef CONFIG_IP_VS_PROTO_TCP
2979 pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2980 u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2981 u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2982 #endif
2983 #ifdef CONFIG_IP_VS_PROTO_UDP
2984 pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2985 u->udp_timeout =
2986 pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2987 #endif
2988 }
2989
2990 static const unsigned char get_arglen[CMDID(IP_VS_SO_GET_MAX) + 1] = {
2991 [CMDID(IP_VS_SO_GET_VERSION)] = 64,
2992 [CMDID(IP_VS_SO_GET_INFO)] = sizeof(struct ip_vs_getinfo),
2993 [CMDID(IP_VS_SO_GET_SERVICES)] = sizeof(struct ip_vs_get_services),
2994 [CMDID(IP_VS_SO_GET_SERVICE)] = sizeof(struct ip_vs_service_entry),
2995 [CMDID(IP_VS_SO_GET_DESTS)] = sizeof(struct ip_vs_get_dests),
2996 [CMDID(IP_VS_SO_GET_TIMEOUT)] = sizeof(struct ip_vs_timeout_user),
2997 [CMDID(IP_VS_SO_GET_DAEMON)] = 2 * sizeof(struct ip_vs_daemon_user),
2998 };
2999
3000 union ip_vs_get_arglen {
3001 char field_IP_VS_SO_GET_VERSION[64];
3002 struct ip_vs_getinfo field_IP_VS_SO_GET_INFO;
3003 struct ip_vs_get_services field_IP_VS_SO_GET_SERVICES;
3004 struct ip_vs_service_entry field_IP_VS_SO_GET_SERVICE;
3005 struct ip_vs_get_dests field_IP_VS_SO_GET_DESTS;
3006 struct ip_vs_timeout_user field_IP_VS_SO_GET_TIMEOUT;
3007 struct ip_vs_daemon_user field_IP_VS_SO_GET_DAEMON[2];
3008 };
3009
3010 #define MAX_GET_ARGLEN sizeof(union ip_vs_get_arglen)
3011
3012 static int
do_ip_vs_get_ctl(struct sock * sk,int cmd,void __user * user,int * len)3013 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
3014 {
3015 unsigned char arg[MAX_GET_ARGLEN];
3016 int ret = 0;
3017 unsigned int copylen;
3018 struct net *net = sock_net(sk);
3019 struct netns_ipvs *ipvs = net_ipvs(net);
3020
3021 BUG_ON(!net);
3022 BUILD_BUG_ON(sizeof(arg) > 255);
3023 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
3024 return -EPERM;
3025
3026 if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
3027 return -EINVAL;
3028
3029 copylen = get_arglen[CMDID(cmd)];
3030 if (*len < (int) copylen) {
3031 IP_VS_DBG(1, "get_ctl: len %d < %u\n", *len, copylen);
3032 return -EINVAL;
3033 }
3034
3035 if (copy_from_user(arg, user, copylen) != 0)
3036 return -EFAULT;
3037 /*
3038 * Handle daemons first since it has its own locking
3039 */
3040 if (cmd == IP_VS_SO_GET_DAEMON) {
3041 struct ip_vs_daemon_user d[2];
3042
3043 memset(&d, 0, sizeof(d));
3044 mutex_lock(&ipvs->sync_mutex);
3045 if (ipvs->sync_state & IP_VS_STATE_MASTER) {
3046 d[0].state = IP_VS_STATE_MASTER;
3047 strscpy(d[0].mcast_ifn, ipvs->mcfg.mcast_ifn,
3048 sizeof(d[0].mcast_ifn));
3049 d[0].syncid = ipvs->mcfg.syncid;
3050 }
3051 if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
3052 d[1].state = IP_VS_STATE_BACKUP;
3053 strscpy(d[1].mcast_ifn, ipvs->bcfg.mcast_ifn,
3054 sizeof(d[1].mcast_ifn));
3055 d[1].syncid = ipvs->bcfg.syncid;
3056 }
3057 if (copy_to_user(user, &d, sizeof(d)) != 0)
3058 ret = -EFAULT;
3059 mutex_unlock(&ipvs->sync_mutex);
3060 return ret;
3061 }
3062
3063 mutex_lock(&__ip_vs_mutex);
3064 switch (cmd) {
3065 case IP_VS_SO_GET_VERSION:
3066 {
3067 char buf[64];
3068
3069 sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
3070 NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
3071 if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
3072 ret = -EFAULT;
3073 goto out;
3074 }
3075 *len = strlen(buf)+1;
3076 }
3077 break;
3078
3079 case IP_VS_SO_GET_INFO:
3080 {
3081 struct ip_vs_getinfo info;
3082 info.version = IP_VS_VERSION_CODE;
3083 info.size = ip_vs_conn_tab_size;
3084 info.num_services = ipvs->num_services;
3085 if (copy_to_user(user, &info, sizeof(info)) != 0)
3086 ret = -EFAULT;
3087 }
3088 break;
3089
3090 case IP_VS_SO_GET_SERVICES:
3091 {
3092 struct ip_vs_get_services *get;
3093 int size;
3094
3095 get = (struct ip_vs_get_services *)arg;
3096 size = struct_size(get, entrytable, get->num_services);
3097 if (*len != size) {
3098 pr_err("length: %u != %u\n", *len, size);
3099 ret = -EINVAL;
3100 goto out;
3101 }
3102 ret = __ip_vs_get_service_entries(ipvs, get, user);
3103 }
3104 break;
3105
3106 case IP_VS_SO_GET_SERVICE:
3107 {
3108 struct ip_vs_service_entry *entry;
3109 struct ip_vs_service *svc;
3110 union nf_inet_addr addr;
3111
3112 entry = (struct ip_vs_service_entry *)arg;
3113 addr.ip = entry->addr;
3114 rcu_read_lock();
3115 if (entry->fwmark)
3116 svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, entry->fwmark);
3117 else
3118 svc = __ip_vs_service_find(ipvs, AF_INET,
3119 entry->protocol, &addr,
3120 entry->port);
3121 rcu_read_unlock();
3122 if (svc) {
3123 ip_vs_copy_service(entry, svc);
3124 if (copy_to_user(user, entry, sizeof(*entry)) != 0)
3125 ret = -EFAULT;
3126 } else
3127 ret = -ESRCH;
3128 }
3129 break;
3130
3131 case IP_VS_SO_GET_DESTS:
3132 {
3133 struct ip_vs_get_dests *get;
3134 int size;
3135
3136 get = (struct ip_vs_get_dests *)arg;
3137 size = struct_size(get, entrytable, get->num_dests);
3138 if (*len != size) {
3139 pr_err("length: %u != %u\n", *len, size);
3140 ret = -EINVAL;
3141 goto out;
3142 }
3143 ret = __ip_vs_get_dest_entries(ipvs, get, user);
3144 }
3145 break;
3146
3147 case IP_VS_SO_GET_TIMEOUT:
3148 {
3149 struct ip_vs_timeout_user t;
3150
3151 __ip_vs_get_timeouts(ipvs, &t);
3152 if (copy_to_user(user, &t, sizeof(t)) != 0)
3153 ret = -EFAULT;
3154 }
3155 break;
3156
3157 default:
3158 ret = -EINVAL;
3159 }
3160
3161 out:
3162 mutex_unlock(&__ip_vs_mutex);
3163 return ret;
3164 }
3165
3166
3167 static struct nf_sockopt_ops ip_vs_sockopts = {
3168 .pf = PF_INET,
3169 .set_optmin = IP_VS_BASE_CTL,
3170 .set_optmax = IP_VS_SO_SET_MAX+1,
3171 .set = do_ip_vs_set_ctl,
3172 .get_optmin = IP_VS_BASE_CTL,
3173 .get_optmax = IP_VS_SO_GET_MAX+1,
3174 .get = do_ip_vs_get_ctl,
3175 .owner = THIS_MODULE,
3176 };
3177
3178 /*
3179 * Generic Netlink interface
3180 */
3181
3182 /* IPVS genetlink family */
3183 static struct genl_family ip_vs_genl_family;
3184
3185 /* Policy used for first-level command attributes */
3186 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
3187 [IPVS_CMD_ATTR_SERVICE] = { .type = NLA_NESTED },
3188 [IPVS_CMD_ATTR_DEST] = { .type = NLA_NESTED },
3189 [IPVS_CMD_ATTR_DAEMON] = { .type = NLA_NESTED },
3190 [IPVS_CMD_ATTR_TIMEOUT_TCP] = { .type = NLA_U32 },
3191 [IPVS_CMD_ATTR_TIMEOUT_TCP_FIN] = { .type = NLA_U32 },
3192 [IPVS_CMD_ATTR_TIMEOUT_UDP] = { .type = NLA_U32 },
3193 };
3194
3195 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
3196 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
3197 [IPVS_DAEMON_ATTR_STATE] = { .type = NLA_U32 },
3198 [IPVS_DAEMON_ATTR_MCAST_IFN] = { .type = NLA_NUL_STRING,
3199 .len = IP_VS_IFNAME_MAXLEN - 1 },
3200 [IPVS_DAEMON_ATTR_SYNC_ID] = { .type = NLA_U32 },
3201 [IPVS_DAEMON_ATTR_SYNC_MAXLEN] = { .type = NLA_U16 },
3202 [IPVS_DAEMON_ATTR_MCAST_GROUP] = { .type = NLA_U32 },
3203 [IPVS_DAEMON_ATTR_MCAST_GROUP6] = { .len = sizeof(struct in6_addr) },
3204 [IPVS_DAEMON_ATTR_MCAST_PORT] = { .type = NLA_U16 },
3205 [IPVS_DAEMON_ATTR_MCAST_TTL] = { .type = NLA_U8 },
3206 };
3207
3208 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
3209 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
3210 [IPVS_SVC_ATTR_AF] = { .type = NLA_U16 },
3211 [IPVS_SVC_ATTR_PROTOCOL] = { .type = NLA_U16 },
3212 [IPVS_SVC_ATTR_ADDR] = { .type = NLA_BINARY,
3213 .len = sizeof(union nf_inet_addr) },
3214 [IPVS_SVC_ATTR_PORT] = { .type = NLA_U16 },
3215 [IPVS_SVC_ATTR_FWMARK] = { .type = NLA_U32 },
3216 [IPVS_SVC_ATTR_SCHED_NAME] = { .type = NLA_NUL_STRING,
3217 .len = IP_VS_SCHEDNAME_MAXLEN - 1 },
3218 [IPVS_SVC_ATTR_PE_NAME] = { .type = NLA_NUL_STRING,
3219 .len = IP_VS_PENAME_MAXLEN },
3220 [IPVS_SVC_ATTR_FLAGS] = { .type = NLA_BINARY,
3221 .len = sizeof(struct ip_vs_flags) },
3222 [IPVS_SVC_ATTR_TIMEOUT] = { .type = NLA_U32 },
3223 [IPVS_SVC_ATTR_NETMASK] = { .type = NLA_U32 },
3224 [IPVS_SVC_ATTR_STATS] = { .type = NLA_NESTED },
3225 };
3226
3227 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
3228 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
3229 [IPVS_DEST_ATTR_ADDR] = { .type = NLA_BINARY,
3230 .len = sizeof(union nf_inet_addr) },
3231 [IPVS_DEST_ATTR_PORT] = { .type = NLA_U16 },
3232 [IPVS_DEST_ATTR_FWD_METHOD] = { .type = NLA_U32 },
3233 [IPVS_DEST_ATTR_WEIGHT] = { .type = NLA_U32 },
3234 [IPVS_DEST_ATTR_U_THRESH] = { .type = NLA_U32 },
3235 [IPVS_DEST_ATTR_L_THRESH] = { .type = NLA_U32 },
3236 [IPVS_DEST_ATTR_ACTIVE_CONNS] = { .type = NLA_U32 },
3237 [IPVS_DEST_ATTR_INACT_CONNS] = { .type = NLA_U32 },
3238 [IPVS_DEST_ATTR_PERSIST_CONNS] = { .type = NLA_U32 },
3239 [IPVS_DEST_ATTR_STATS] = { .type = NLA_NESTED },
3240 [IPVS_DEST_ATTR_ADDR_FAMILY] = { .type = NLA_U16 },
3241 [IPVS_DEST_ATTR_TUN_TYPE] = { .type = NLA_U8 },
3242 [IPVS_DEST_ATTR_TUN_PORT] = { .type = NLA_U16 },
3243 [IPVS_DEST_ATTR_TUN_FLAGS] = { .type = NLA_U16 },
3244 };
3245
ip_vs_genl_fill_stats(struct sk_buff * skb,int container_type,struct ip_vs_kstats * kstats)3246 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
3247 struct ip_vs_kstats *kstats)
3248 {
3249 struct nlattr *nl_stats = nla_nest_start_noflag(skb, container_type);
3250
3251 if (!nl_stats)
3252 return -EMSGSIZE;
3253
3254 if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, (u32)kstats->conns) ||
3255 nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, (u32)kstats->inpkts) ||
3256 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, (u32)kstats->outpkts) ||
3257 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes,
3258 IPVS_STATS_ATTR_PAD) ||
3259 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes,
3260 IPVS_STATS_ATTR_PAD) ||
3261 nla_put_u32(skb, IPVS_STATS_ATTR_CPS, (u32)kstats->cps) ||
3262 nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, (u32)kstats->inpps) ||
3263 nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, (u32)kstats->outpps) ||
3264 nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, (u32)kstats->inbps) ||
3265 nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, (u32)kstats->outbps))
3266 goto nla_put_failure;
3267 nla_nest_end(skb, nl_stats);
3268
3269 return 0;
3270
3271 nla_put_failure:
3272 nla_nest_cancel(skb, nl_stats);
3273 return -EMSGSIZE;
3274 }
3275
ip_vs_genl_fill_stats64(struct sk_buff * skb,int container_type,struct ip_vs_kstats * kstats)3276 static int ip_vs_genl_fill_stats64(struct sk_buff *skb, int container_type,
3277 struct ip_vs_kstats *kstats)
3278 {
3279 struct nlattr *nl_stats = nla_nest_start_noflag(skb, container_type);
3280
3281 if (!nl_stats)
3282 return -EMSGSIZE;
3283
3284 if (nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CONNS, kstats->conns,
3285 IPVS_STATS_ATTR_PAD) ||
3286 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPKTS, kstats->inpkts,
3287 IPVS_STATS_ATTR_PAD) ||
3288 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPKTS, kstats->outpkts,
3289 IPVS_STATS_ATTR_PAD) ||
3290 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes,
3291 IPVS_STATS_ATTR_PAD) ||
3292 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes,
3293 IPVS_STATS_ATTR_PAD) ||
3294 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CPS, kstats->cps,
3295 IPVS_STATS_ATTR_PAD) ||
3296 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPPS, kstats->inpps,
3297 IPVS_STATS_ATTR_PAD) ||
3298 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPPS, kstats->outpps,
3299 IPVS_STATS_ATTR_PAD) ||
3300 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBPS, kstats->inbps,
3301 IPVS_STATS_ATTR_PAD) ||
3302 nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBPS, kstats->outbps,
3303 IPVS_STATS_ATTR_PAD))
3304 goto nla_put_failure;
3305 nla_nest_end(skb, nl_stats);
3306
3307 return 0;
3308
3309 nla_put_failure:
3310 nla_nest_cancel(skb, nl_stats);
3311 return -EMSGSIZE;
3312 }
3313
ip_vs_genl_fill_service(struct sk_buff * skb,struct ip_vs_service * svc)3314 static int ip_vs_genl_fill_service(struct sk_buff *skb,
3315 struct ip_vs_service *svc)
3316 {
3317 struct ip_vs_scheduler *sched;
3318 struct ip_vs_pe *pe;
3319 struct nlattr *nl_service;
3320 struct ip_vs_flags flags = { .flags = svc->flags,
3321 .mask = ~0 };
3322 struct ip_vs_kstats kstats;
3323 char *sched_name;
3324
3325 nl_service = nla_nest_start_noflag(skb, IPVS_CMD_ATTR_SERVICE);
3326 if (!nl_service)
3327 return -EMSGSIZE;
3328
3329 if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
3330 goto nla_put_failure;
3331 if (svc->fwmark) {
3332 if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
3333 goto nla_put_failure;
3334 } else {
3335 if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
3336 nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
3337 nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
3338 goto nla_put_failure;
3339 }
3340
3341 sched = rcu_dereference_protected(svc->scheduler, 1);
3342 sched_name = sched ? sched->name : "none";
3343 pe = rcu_dereference_protected(svc->pe, 1);
3344 if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched_name) ||
3345 (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
3346 nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
3347 nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
3348 nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
3349 goto nla_put_failure;
3350 ip_vs_copy_stats(&kstats, &svc->stats);
3351 if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &kstats))
3352 goto nla_put_failure;
3353 if (ip_vs_genl_fill_stats64(skb, IPVS_SVC_ATTR_STATS64, &kstats))
3354 goto nla_put_failure;
3355
3356 nla_nest_end(skb, nl_service);
3357
3358 return 0;
3359
3360 nla_put_failure:
3361 nla_nest_cancel(skb, nl_service);
3362 return -EMSGSIZE;
3363 }
3364
ip_vs_genl_dump_service(struct sk_buff * skb,struct ip_vs_service * svc,struct netlink_callback * cb)3365 static int ip_vs_genl_dump_service(struct sk_buff *skb,
3366 struct ip_vs_service *svc,
3367 struct netlink_callback *cb)
3368 {
3369 void *hdr;
3370
3371 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3372 &ip_vs_genl_family, NLM_F_MULTI,
3373 IPVS_CMD_NEW_SERVICE);
3374 if (!hdr)
3375 return -EMSGSIZE;
3376
3377 if (ip_vs_genl_fill_service(skb, svc) < 0)
3378 goto nla_put_failure;
3379
3380 genlmsg_end(skb, hdr);
3381 return 0;
3382
3383 nla_put_failure:
3384 genlmsg_cancel(skb, hdr);
3385 return -EMSGSIZE;
3386 }
3387
ip_vs_genl_dump_services(struct sk_buff * skb,struct netlink_callback * cb)3388 static int ip_vs_genl_dump_services(struct sk_buff *skb,
3389 struct netlink_callback *cb)
3390 {
3391 int idx = 0, i;
3392 int start = cb->args[0];
3393 struct ip_vs_service *svc;
3394 struct net *net = sock_net(skb->sk);
3395 struct netns_ipvs *ipvs = net_ipvs(net);
3396
3397 mutex_lock(&__ip_vs_mutex);
3398 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3399 hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
3400 if (++idx <= start || (svc->ipvs != ipvs))
3401 continue;
3402 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3403 idx--;
3404 goto nla_put_failure;
3405 }
3406 }
3407 }
3408
3409 for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3410 hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
3411 if (++idx <= start || (svc->ipvs != ipvs))
3412 continue;
3413 if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3414 idx--;
3415 goto nla_put_failure;
3416 }
3417 }
3418 }
3419
3420 nla_put_failure:
3421 mutex_unlock(&__ip_vs_mutex);
3422 cb->args[0] = idx;
3423
3424 return skb->len;
3425 }
3426
ip_vs_is_af_valid(int af)3427 static bool ip_vs_is_af_valid(int af)
3428 {
3429 if (af == AF_INET)
3430 return true;
3431 #ifdef CONFIG_IP_VS_IPV6
3432 if (af == AF_INET6 && ipv6_mod_enabled())
3433 return true;
3434 #endif
3435 return false;
3436 }
3437
ip_vs_genl_parse_service(struct netns_ipvs * ipvs,struct ip_vs_service_user_kern * usvc,struct nlattr * nla,bool full_entry,struct ip_vs_service ** ret_svc)3438 static int ip_vs_genl_parse_service(struct netns_ipvs *ipvs,
3439 struct ip_vs_service_user_kern *usvc,
3440 struct nlattr *nla, bool full_entry,
3441 struct ip_vs_service **ret_svc)
3442 {
3443 struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
3444 struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
3445 struct ip_vs_service *svc;
3446
3447 /* Parse mandatory identifying service fields first */
3448 if (nla == NULL ||
3449 nla_parse_nested_deprecated(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy, NULL))
3450 return -EINVAL;
3451
3452 nla_af = attrs[IPVS_SVC_ATTR_AF];
3453 nla_protocol = attrs[IPVS_SVC_ATTR_PROTOCOL];
3454 nla_addr = attrs[IPVS_SVC_ATTR_ADDR];
3455 nla_port = attrs[IPVS_SVC_ATTR_PORT];
3456 nla_fwmark = attrs[IPVS_SVC_ATTR_FWMARK];
3457
3458 if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3459 return -EINVAL;
3460
3461 memset(usvc, 0, sizeof(*usvc));
3462
3463 usvc->af = nla_get_u16(nla_af);
3464 if (!ip_vs_is_af_valid(usvc->af))
3465 return -EAFNOSUPPORT;
3466
3467 if (nla_fwmark) {
3468 usvc->protocol = IPPROTO_TCP;
3469 usvc->fwmark = nla_get_u32(nla_fwmark);
3470 } else {
3471 usvc->protocol = nla_get_u16(nla_protocol);
3472 nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3473 usvc->port = nla_get_be16(nla_port);
3474 usvc->fwmark = 0;
3475 }
3476
3477 rcu_read_lock();
3478 if (usvc->fwmark)
3479 svc = __ip_vs_svc_fwm_find(ipvs, usvc->af, usvc->fwmark);
3480 else
3481 svc = __ip_vs_service_find(ipvs, usvc->af, usvc->protocol,
3482 &usvc->addr, usvc->port);
3483 rcu_read_unlock();
3484 *ret_svc = svc;
3485
3486 /* If a full entry was requested, check for the additional fields */
3487 if (full_entry) {
3488 struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3489 *nla_netmask;
3490 struct ip_vs_flags flags;
3491
3492 nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3493 nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3494 nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3495 nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3496 nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3497
3498 if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3499 return -EINVAL;
3500
3501 nla_memcpy(&flags, nla_flags, sizeof(flags));
3502
3503 /* prefill flags from service if it already exists */
3504 if (svc)
3505 usvc->flags = svc->flags;
3506
3507 /* set new flags from userland */
3508 usvc->flags = (usvc->flags & ~flags.mask) |
3509 (flags.flags & flags.mask);
3510 usvc->sched_name = nla_data(nla_sched);
3511 usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3512 usvc->timeout = nla_get_u32(nla_timeout);
3513 usvc->netmask = nla_get_be32(nla_netmask);
3514 }
3515
3516 return 0;
3517 }
3518
ip_vs_genl_find_service(struct netns_ipvs * ipvs,struct nlattr * nla)3519 static struct ip_vs_service *ip_vs_genl_find_service(struct netns_ipvs *ipvs,
3520 struct nlattr *nla)
3521 {
3522 struct ip_vs_service_user_kern usvc;
3523 struct ip_vs_service *svc;
3524 int ret;
3525
3526 ret = ip_vs_genl_parse_service(ipvs, &usvc, nla, false, &svc);
3527 return ret ? ERR_PTR(ret) : svc;
3528 }
3529
ip_vs_genl_fill_dest(struct sk_buff * skb,struct ip_vs_dest * dest)3530 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3531 {
3532 struct nlattr *nl_dest;
3533 struct ip_vs_kstats kstats;
3534
3535 nl_dest = nla_nest_start_noflag(skb, IPVS_CMD_ATTR_DEST);
3536 if (!nl_dest)
3537 return -EMSGSIZE;
3538
3539 if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3540 nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3541 nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3542 (atomic_read(&dest->conn_flags) &
3543 IP_VS_CONN_F_FWD_MASK)) ||
3544 nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3545 atomic_read(&dest->weight)) ||
3546 nla_put_u8(skb, IPVS_DEST_ATTR_TUN_TYPE,
3547 dest->tun_type) ||
3548 nla_put_be16(skb, IPVS_DEST_ATTR_TUN_PORT,
3549 dest->tun_port) ||
3550 nla_put_u16(skb, IPVS_DEST_ATTR_TUN_FLAGS,
3551 dest->tun_flags) ||
3552 nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3553 nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3554 nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3555 atomic_read(&dest->activeconns)) ||
3556 nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3557 atomic_read(&dest->inactconns)) ||
3558 nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3559 atomic_read(&dest->persistconns)) ||
3560 nla_put_u16(skb, IPVS_DEST_ATTR_ADDR_FAMILY, dest->af))
3561 goto nla_put_failure;
3562 ip_vs_copy_stats(&kstats, &dest->stats);
3563 if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &kstats))
3564 goto nla_put_failure;
3565 if (ip_vs_genl_fill_stats64(skb, IPVS_DEST_ATTR_STATS64, &kstats))
3566 goto nla_put_failure;
3567
3568 nla_nest_end(skb, nl_dest);
3569
3570 return 0;
3571
3572 nla_put_failure:
3573 nla_nest_cancel(skb, nl_dest);
3574 return -EMSGSIZE;
3575 }
3576
ip_vs_genl_dump_dest(struct sk_buff * skb,struct ip_vs_dest * dest,struct netlink_callback * cb)3577 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3578 struct netlink_callback *cb)
3579 {
3580 void *hdr;
3581
3582 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3583 &ip_vs_genl_family, NLM_F_MULTI,
3584 IPVS_CMD_NEW_DEST);
3585 if (!hdr)
3586 return -EMSGSIZE;
3587
3588 if (ip_vs_genl_fill_dest(skb, dest) < 0)
3589 goto nla_put_failure;
3590
3591 genlmsg_end(skb, hdr);
3592 return 0;
3593
3594 nla_put_failure:
3595 genlmsg_cancel(skb, hdr);
3596 return -EMSGSIZE;
3597 }
3598
ip_vs_genl_dump_dests(struct sk_buff * skb,struct netlink_callback * cb)3599 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3600 struct netlink_callback *cb)
3601 {
3602 int idx = 0;
3603 int start = cb->args[0];
3604 struct ip_vs_service *svc;
3605 struct ip_vs_dest *dest;
3606 struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3607 struct net *net = sock_net(skb->sk);
3608 struct netns_ipvs *ipvs = net_ipvs(net);
3609
3610 mutex_lock(&__ip_vs_mutex);
3611
3612 /* Try to find the service for which to dump destinations */
3613 if (nlmsg_parse_deprecated(cb->nlh, GENL_HDRLEN, attrs, IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy, cb->extack))
3614 goto out_err;
3615
3616
3617 svc = ip_vs_genl_find_service(ipvs, attrs[IPVS_CMD_ATTR_SERVICE]);
3618 if (IS_ERR_OR_NULL(svc))
3619 goto out_err;
3620
3621 /* Dump the destinations */
3622 list_for_each_entry(dest, &svc->destinations, n_list) {
3623 if (++idx <= start)
3624 continue;
3625 if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3626 idx--;
3627 goto nla_put_failure;
3628 }
3629 }
3630
3631 nla_put_failure:
3632 cb->args[0] = idx;
3633
3634 out_err:
3635 mutex_unlock(&__ip_vs_mutex);
3636
3637 return skb->len;
3638 }
3639
ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern * udest,struct nlattr * nla,bool full_entry)3640 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3641 struct nlattr *nla, bool full_entry)
3642 {
3643 struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3644 struct nlattr *nla_addr, *nla_port;
3645 struct nlattr *nla_addr_family;
3646
3647 /* Parse mandatory identifying destination fields first */
3648 if (nla == NULL ||
3649 nla_parse_nested_deprecated(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy, NULL))
3650 return -EINVAL;
3651
3652 nla_addr = attrs[IPVS_DEST_ATTR_ADDR];
3653 nla_port = attrs[IPVS_DEST_ATTR_PORT];
3654 nla_addr_family = attrs[IPVS_DEST_ATTR_ADDR_FAMILY];
3655
3656 if (!(nla_addr && nla_port))
3657 return -EINVAL;
3658
3659 memset(udest, 0, sizeof(*udest));
3660
3661 nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3662 udest->port = nla_get_be16(nla_port);
3663
3664 if (nla_addr_family)
3665 udest->af = nla_get_u16(nla_addr_family);
3666 else
3667 udest->af = 0;
3668
3669 /* If a full entry was requested, check for the additional fields */
3670 if (full_entry) {
3671 struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3672 *nla_l_thresh, *nla_tun_type, *nla_tun_port,
3673 *nla_tun_flags;
3674
3675 nla_fwd = attrs[IPVS_DEST_ATTR_FWD_METHOD];
3676 nla_weight = attrs[IPVS_DEST_ATTR_WEIGHT];
3677 nla_u_thresh = attrs[IPVS_DEST_ATTR_U_THRESH];
3678 nla_l_thresh = attrs[IPVS_DEST_ATTR_L_THRESH];
3679 nla_tun_type = attrs[IPVS_DEST_ATTR_TUN_TYPE];
3680 nla_tun_port = attrs[IPVS_DEST_ATTR_TUN_PORT];
3681 nla_tun_flags = attrs[IPVS_DEST_ATTR_TUN_FLAGS];
3682
3683 if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3684 return -EINVAL;
3685
3686 udest->conn_flags = nla_get_u32(nla_fwd)
3687 & IP_VS_CONN_F_FWD_MASK;
3688 udest->weight = nla_get_u32(nla_weight);
3689 udest->u_threshold = nla_get_u32(nla_u_thresh);
3690 udest->l_threshold = nla_get_u32(nla_l_thresh);
3691
3692 if (nla_tun_type)
3693 udest->tun_type = nla_get_u8(nla_tun_type);
3694
3695 if (nla_tun_port)
3696 udest->tun_port = nla_get_be16(nla_tun_port);
3697
3698 if (nla_tun_flags)
3699 udest->tun_flags = nla_get_u16(nla_tun_flags);
3700 }
3701
3702 return 0;
3703 }
3704
ip_vs_genl_fill_daemon(struct sk_buff * skb,__u32 state,struct ipvs_sync_daemon_cfg * c)3705 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3706 struct ipvs_sync_daemon_cfg *c)
3707 {
3708 struct nlattr *nl_daemon;
3709
3710 nl_daemon = nla_nest_start_noflag(skb, IPVS_CMD_ATTR_DAEMON);
3711 if (!nl_daemon)
3712 return -EMSGSIZE;
3713
3714 if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3715 nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, c->mcast_ifn) ||
3716 nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, c->syncid) ||
3717 nla_put_u16(skb, IPVS_DAEMON_ATTR_SYNC_MAXLEN, c->sync_maxlen) ||
3718 nla_put_u16(skb, IPVS_DAEMON_ATTR_MCAST_PORT, c->mcast_port) ||
3719 nla_put_u8(skb, IPVS_DAEMON_ATTR_MCAST_TTL, c->mcast_ttl))
3720 goto nla_put_failure;
3721 #ifdef CONFIG_IP_VS_IPV6
3722 if (c->mcast_af == AF_INET6) {
3723 if (nla_put_in6_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP6,
3724 &c->mcast_group.in6))
3725 goto nla_put_failure;
3726 } else
3727 #endif
3728 if (c->mcast_af == AF_INET &&
3729 nla_put_in_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP,
3730 c->mcast_group.ip))
3731 goto nla_put_failure;
3732 nla_nest_end(skb, nl_daemon);
3733
3734 return 0;
3735
3736 nla_put_failure:
3737 nla_nest_cancel(skb, nl_daemon);
3738 return -EMSGSIZE;
3739 }
3740
ip_vs_genl_dump_daemon(struct sk_buff * skb,__u32 state,struct ipvs_sync_daemon_cfg * c,struct netlink_callback * cb)3741 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3742 struct ipvs_sync_daemon_cfg *c,
3743 struct netlink_callback *cb)
3744 {
3745 void *hdr;
3746 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3747 &ip_vs_genl_family, NLM_F_MULTI,
3748 IPVS_CMD_NEW_DAEMON);
3749 if (!hdr)
3750 return -EMSGSIZE;
3751
3752 if (ip_vs_genl_fill_daemon(skb, state, c))
3753 goto nla_put_failure;
3754
3755 genlmsg_end(skb, hdr);
3756 return 0;
3757
3758 nla_put_failure:
3759 genlmsg_cancel(skb, hdr);
3760 return -EMSGSIZE;
3761 }
3762
ip_vs_genl_dump_daemons(struct sk_buff * skb,struct netlink_callback * cb)3763 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3764 struct netlink_callback *cb)
3765 {
3766 struct net *net = sock_net(skb->sk);
3767 struct netns_ipvs *ipvs = net_ipvs(net);
3768
3769 mutex_lock(&ipvs->sync_mutex);
3770 if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3771 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3772 &ipvs->mcfg, cb) < 0)
3773 goto nla_put_failure;
3774
3775 cb->args[0] = 1;
3776 }
3777
3778 if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3779 if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3780 &ipvs->bcfg, cb) < 0)
3781 goto nla_put_failure;
3782
3783 cb->args[1] = 1;
3784 }
3785
3786 nla_put_failure:
3787 mutex_unlock(&ipvs->sync_mutex);
3788
3789 return skb->len;
3790 }
3791
ip_vs_genl_new_daemon(struct netns_ipvs * ipvs,struct nlattr ** attrs)3792 static int ip_vs_genl_new_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3793 {
3794 struct ipvs_sync_daemon_cfg c;
3795 struct nlattr *a;
3796 int ret;
3797
3798 memset(&c, 0, sizeof(c));
3799 if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3800 attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3801 attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3802 return -EINVAL;
3803 strscpy(c.mcast_ifn, nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3804 sizeof(c.mcast_ifn));
3805 c.syncid = nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]);
3806
3807 a = attrs[IPVS_DAEMON_ATTR_SYNC_MAXLEN];
3808 if (a)
3809 c.sync_maxlen = nla_get_u16(a);
3810
3811 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP];
3812 if (a) {
3813 c.mcast_af = AF_INET;
3814 c.mcast_group.ip = nla_get_in_addr(a);
3815 if (!ipv4_is_multicast(c.mcast_group.ip))
3816 return -EINVAL;
3817 } else {
3818 a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP6];
3819 if (a) {
3820 #ifdef CONFIG_IP_VS_IPV6
3821 int addr_type;
3822
3823 c.mcast_af = AF_INET6;
3824 c.mcast_group.in6 = nla_get_in6_addr(a);
3825 addr_type = ipv6_addr_type(&c.mcast_group.in6);
3826 if (!(addr_type & IPV6_ADDR_MULTICAST))
3827 return -EINVAL;
3828 #else
3829 return -EAFNOSUPPORT;
3830 #endif
3831 }
3832 }
3833
3834 a = attrs[IPVS_DAEMON_ATTR_MCAST_PORT];
3835 if (a)
3836 c.mcast_port = nla_get_u16(a);
3837
3838 a = attrs[IPVS_DAEMON_ATTR_MCAST_TTL];
3839 if (a)
3840 c.mcast_ttl = nla_get_u8(a);
3841
3842 /* The synchronization protocol is incompatible with mixed family
3843 * services
3844 */
3845 if (ipvs->mixed_address_family_dests > 0)
3846 return -EINVAL;
3847
3848 ret = start_sync_thread(ipvs, &c,
3849 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3850 return ret;
3851 }
3852
ip_vs_genl_del_daemon(struct netns_ipvs * ipvs,struct nlattr ** attrs)3853 static int ip_vs_genl_del_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3854 {
3855 int ret;
3856
3857 if (!attrs[IPVS_DAEMON_ATTR_STATE])
3858 return -EINVAL;
3859
3860 ret = stop_sync_thread(ipvs,
3861 nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3862 return ret;
3863 }
3864
ip_vs_genl_set_config(struct netns_ipvs * ipvs,struct nlattr ** attrs)3865 static int ip_vs_genl_set_config(struct netns_ipvs *ipvs, struct nlattr **attrs)
3866 {
3867 struct ip_vs_timeout_user t;
3868
3869 __ip_vs_get_timeouts(ipvs, &t);
3870
3871 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3872 t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3873
3874 if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3875 t.tcp_fin_timeout =
3876 nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3877
3878 if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3879 t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3880
3881 return ip_vs_set_timeout(ipvs, &t);
3882 }
3883
ip_vs_genl_set_daemon(struct sk_buff * skb,struct genl_info * info)3884 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3885 {
3886 int ret = -EINVAL, cmd;
3887 struct net *net = sock_net(skb->sk);
3888 struct netns_ipvs *ipvs = net_ipvs(net);
3889
3890 cmd = info->genlhdr->cmd;
3891
3892 if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3893 struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3894
3895 if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3896 nla_parse_nested_deprecated(daemon_attrs, IPVS_DAEMON_ATTR_MAX, info->attrs[IPVS_CMD_ATTR_DAEMON], ip_vs_daemon_policy, info->extack))
3897 goto out;
3898
3899 if (cmd == IPVS_CMD_NEW_DAEMON)
3900 ret = ip_vs_genl_new_daemon(ipvs, daemon_attrs);
3901 else
3902 ret = ip_vs_genl_del_daemon(ipvs, daemon_attrs);
3903 }
3904
3905 out:
3906 return ret;
3907 }
3908
ip_vs_genl_set_cmd(struct sk_buff * skb,struct genl_info * info)3909 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3910 {
3911 bool need_full_svc = false, need_full_dest = false;
3912 struct ip_vs_service *svc = NULL;
3913 struct ip_vs_service_user_kern usvc;
3914 struct ip_vs_dest_user_kern udest;
3915 int ret = 0, cmd;
3916 struct net *net = sock_net(skb->sk);
3917 struct netns_ipvs *ipvs = net_ipvs(net);
3918
3919 cmd = info->genlhdr->cmd;
3920
3921 mutex_lock(&__ip_vs_mutex);
3922
3923 if (cmd == IPVS_CMD_FLUSH) {
3924 ret = ip_vs_flush(ipvs, false);
3925 goto out;
3926 } else if (cmd == IPVS_CMD_SET_CONFIG) {
3927 ret = ip_vs_genl_set_config(ipvs, info->attrs);
3928 goto out;
3929 } else if (cmd == IPVS_CMD_ZERO &&
3930 !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3931 ret = ip_vs_zero_all(ipvs);
3932 goto out;
3933 }
3934
3935 /* All following commands require a service argument, so check if we
3936 * received a valid one. We need a full service specification when
3937 * adding / editing a service. Only identifying members otherwise. */
3938 if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3939 need_full_svc = true;
3940
3941 ret = ip_vs_genl_parse_service(ipvs, &usvc,
3942 info->attrs[IPVS_CMD_ATTR_SERVICE],
3943 need_full_svc, &svc);
3944 if (ret)
3945 goto out;
3946
3947 /* Unless we're adding a new service, the service must already exist */
3948 if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3949 ret = -ESRCH;
3950 goto out;
3951 }
3952
3953 /* Destination commands require a valid destination argument. For
3954 * adding / editing a destination, we need a full destination
3955 * specification. */
3956 if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3957 cmd == IPVS_CMD_DEL_DEST) {
3958 if (cmd != IPVS_CMD_DEL_DEST)
3959 need_full_dest = true;
3960
3961 ret = ip_vs_genl_parse_dest(&udest,
3962 info->attrs[IPVS_CMD_ATTR_DEST],
3963 need_full_dest);
3964 if (ret)
3965 goto out;
3966
3967 /* Old protocols did not allow the user to specify address
3968 * family, so we set it to zero instead. We also didn't
3969 * allow heterogeneous pools in the old code, so it's safe
3970 * to assume that this will have the same address family as
3971 * the service.
3972 */
3973 if (udest.af == 0)
3974 udest.af = svc->af;
3975
3976 if (!ip_vs_is_af_valid(udest.af)) {
3977 ret = -EAFNOSUPPORT;
3978 goto out;
3979 }
3980
3981 if (udest.af != svc->af && cmd != IPVS_CMD_DEL_DEST) {
3982 /* The synchronization protocol is incompatible
3983 * with mixed family services
3984 */
3985 if (ipvs->sync_state) {
3986 ret = -EINVAL;
3987 goto out;
3988 }
3989
3990 /* Which connection types do we support? */
3991 switch (udest.conn_flags) {
3992 case IP_VS_CONN_F_TUNNEL:
3993 /* We are able to forward this */
3994 break;
3995 default:
3996 ret = -EINVAL;
3997 goto out;
3998 }
3999 }
4000 }
4001
4002 switch (cmd) {
4003 case IPVS_CMD_NEW_SERVICE:
4004 if (svc == NULL)
4005 ret = ip_vs_add_service(ipvs, &usvc, &svc);
4006 else
4007 ret = -EEXIST;
4008 break;
4009 case IPVS_CMD_SET_SERVICE:
4010 ret = ip_vs_edit_service(svc, &usvc);
4011 break;
4012 case IPVS_CMD_DEL_SERVICE:
4013 ret = ip_vs_del_service(svc);
4014 /* do not use svc, it can be freed */
4015 break;
4016 case IPVS_CMD_NEW_DEST:
4017 ret = ip_vs_add_dest(svc, &udest);
4018 break;
4019 case IPVS_CMD_SET_DEST:
4020 ret = ip_vs_edit_dest(svc, &udest);
4021 break;
4022 case IPVS_CMD_DEL_DEST:
4023 ret = ip_vs_del_dest(svc, &udest);
4024 break;
4025 case IPVS_CMD_ZERO:
4026 ret = ip_vs_zero_service(svc);
4027 break;
4028 default:
4029 ret = -EINVAL;
4030 }
4031
4032 out:
4033 mutex_unlock(&__ip_vs_mutex);
4034
4035 return ret;
4036 }
4037
ip_vs_genl_get_cmd(struct sk_buff * skb,struct genl_info * info)4038 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
4039 {
4040 struct sk_buff *msg;
4041 void *reply;
4042 int ret, cmd, reply_cmd;
4043 struct net *net = sock_net(skb->sk);
4044 struct netns_ipvs *ipvs = net_ipvs(net);
4045
4046 cmd = info->genlhdr->cmd;
4047
4048 if (cmd == IPVS_CMD_GET_SERVICE)
4049 reply_cmd = IPVS_CMD_NEW_SERVICE;
4050 else if (cmd == IPVS_CMD_GET_INFO)
4051 reply_cmd = IPVS_CMD_SET_INFO;
4052 else if (cmd == IPVS_CMD_GET_CONFIG)
4053 reply_cmd = IPVS_CMD_SET_CONFIG;
4054 else {
4055 pr_err("unknown Generic Netlink command\n");
4056 return -EINVAL;
4057 }
4058
4059 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
4060 if (!msg)
4061 return -ENOMEM;
4062
4063 mutex_lock(&__ip_vs_mutex);
4064
4065 reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
4066 if (reply == NULL)
4067 goto nla_put_failure;
4068
4069 switch (cmd) {
4070 case IPVS_CMD_GET_SERVICE:
4071 {
4072 struct ip_vs_service *svc;
4073
4074 svc = ip_vs_genl_find_service(ipvs,
4075 info->attrs[IPVS_CMD_ATTR_SERVICE]);
4076 if (IS_ERR(svc)) {
4077 ret = PTR_ERR(svc);
4078 goto out_err;
4079 } else if (svc) {
4080 ret = ip_vs_genl_fill_service(msg, svc);
4081 if (ret)
4082 goto nla_put_failure;
4083 } else {
4084 ret = -ESRCH;
4085 goto out_err;
4086 }
4087
4088 break;
4089 }
4090
4091 case IPVS_CMD_GET_CONFIG:
4092 {
4093 struct ip_vs_timeout_user t;
4094
4095 __ip_vs_get_timeouts(ipvs, &t);
4096 #ifdef CONFIG_IP_VS_PROTO_TCP
4097 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
4098 t.tcp_timeout) ||
4099 nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
4100 t.tcp_fin_timeout))
4101 goto nla_put_failure;
4102 #endif
4103 #ifdef CONFIG_IP_VS_PROTO_UDP
4104 if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
4105 goto nla_put_failure;
4106 #endif
4107
4108 break;
4109 }
4110
4111 case IPVS_CMD_GET_INFO:
4112 if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
4113 IP_VS_VERSION_CODE) ||
4114 nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
4115 ip_vs_conn_tab_size))
4116 goto nla_put_failure;
4117 break;
4118 }
4119
4120 genlmsg_end(msg, reply);
4121 ret = genlmsg_reply(msg, info);
4122 goto out;
4123
4124 nla_put_failure:
4125 pr_err("not enough space in Netlink message\n");
4126 ret = -EMSGSIZE;
4127
4128 out_err:
4129 nlmsg_free(msg);
4130 out:
4131 mutex_unlock(&__ip_vs_mutex);
4132
4133 return ret;
4134 }
4135
4136
4137 static const struct genl_small_ops ip_vs_genl_ops[] = {
4138 {
4139 .cmd = IPVS_CMD_NEW_SERVICE,
4140 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
4141 .flags = GENL_ADMIN_PERM,
4142 .doit = ip_vs_genl_set_cmd,
4143 },
4144 {
4145 .cmd = IPVS_CMD_SET_SERVICE,
4146 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
4147 .flags = GENL_ADMIN_PERM,
4148 .doit = ip_vs_genl_set_cmd,
4149 },
4150 {
4151 .cmd = IPVS_CMD_DEL_SERVICE,
4152 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
4153 .flags = GENL_ADMIN_PERM,
4154 .doit = ip_vs_genl_set_cmd,
4155 },
4156 {
4157 .cmd = IPVS_CMD_GET_SERVICE,
4158 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
4159 .flags = GENL_ADMIN_PERM,
4160 .doit = ip_vs_genl_get_cmd,
4161 .dumpit = ip_vs_genl_dump_services,
4162 },
4163 {
4164 .cmd = IPVS_CMD_NEW_DEST,
4165 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
4166 .flags = GENL_ADMIN_PERM,
4167 .doit = ip_vs_genl_set_cmd,
4168 },
4169 {
4170 .cmd = IPVS_CMD_SET_DEST,
4171 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
4172 .flags = GENL_ADMIN_PERM,
4173 .doit = ip_vs_genl_set_cmd,
4174 },
4175 {
4176 .cmd = IPVS_CMD_DEL_DEST,
4177 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
4178 .flags = GENL_ADMIN_PERM,
4179 .doit = ip_vs_genl_set_cmd,
4180 },
4181 {
4182 .cmd = IPVS_CMD_GET_DEST,
4183 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
4184 .flags = GENL_ADMIN_PERM,
4185 .dumpit = ip_vs_genl_dump_dests,
4186 },
4187 {
4188 .cmd = IPVS_CMD_NEW_DAEMON,
4189 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
4190 .flags = GENL_ADMIN_PERM,
4191 .doit = ip_vs_genl_set_daemon,
4192 },
4193 {
4194 .cmd = IPVS_CMD_DEL_DAEMON,
4195 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
4196 .flags = GENL_ADMIN_PERM,
4197 .doit = ip_vs_genl_set_daemon,
4198 },
4199 {
4200 .cmd = IPVS_CMD_GET_DAEMON,
4201 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
4202 .flags = GENL_ADMIN_PERM,
4203 .dumpit = ip_vs_genl_dump_daemons,
4204 },
4205 {
4206 .cmd = IPVS_CMD_SET_CONFIG,
4207 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
4208 .flags = GENL_ADMIN_PERM,
4209 .doit = ip_vs_genl_set_cmd,
4210 },
4211 {
4212 .cmd = IPVS_CMD_GET_CONFIG,
4213 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
4214 .flags = GENL_ADMIN_PERM,
4215 .doit = ip_vs_genl_get_cmd,
4216 },
4217 {
4218 .cmd = IPVS_CMD_GET_INFO,
4219 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
4220 .flags = GENL_ADMIN_PERM,
4221 .doit = ip_vs_genl_get_cmd,
4222 },
4223 {
4224 .cmd = IPVS_CMD_ZERO,
4225 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
4226 .flags = GENL_ADMIN_PERM,
4227 .doit = ip_vs_genl_set_cmd,
4228 },
4229 {
4230 .cmd = IPVS_CMD_FLUSH,
4231 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
4232 .flags = GENL_ADMIN_PERM,
4233 .doit = ip_vs_genl_set_cmd,
4234 },
4235 };
4236
4237 static struct genl_family ip_vs_genl_family __ro_after_init = {
4238 .hdrsize = 0,
4239 .name = IPVS_GENL_NAME,
4240 .version = IPVS_GENL_VERSION,
4241 .maxattr = IPVS_CMD_ATTR_MAX,
4242 .policy = ip_vs_cmd_policy,
4243 .netnsok = true, /* Make ipvsadm to work on netns */
4244 .module = THIS_MODULE,
4245 .small_ops = ip_vs_genl_ops,
4246 .n_small_ops = ARRAY_SIZE(ip_vs_genl_ops),
4247 .resv_start_op = IPVS_CMD_FLUSH + 1,
4248 };
4249
ip_vs_genl_register(void)4250 static int __init ip_vs_genl_register(void)
4251 {
4252 return genl_register_family(&ip_vs_genl_family);
4253 }
4254
ip_vs_genl_unregister(void)4255 static void ip_vs_genl_unregister(void)
4256 {
4257 genl_unregister_family(&ip_vs_genl_family);
4258 }
4259
4260 /* End of Generic Netlink interface definitions */
4261
4262 /*
4263 * per netns intit/exit func.
4264 */
4265 #ifdef CONFIG_SYSCTL
ip_vs_control_net_init_sysctl(struct netns_ipvs * ipvs)4266 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs)
4267 {
4268 struct net *net = ipvs->net;
4269 struct ctl_table *tbl;
4270 int idx, ret;
4271 size_t ctl_table_size = ARRAY_SIZE(vs_vars);
4272 bool unpriv = net->user_ns != &init_user_ns;
4273
4274 atomic_set(&ipvs->dropentry, 0);
4275 spin_lock_init(&ipvs->dropentry_lock);
4276 spin_lock_init(&ipvs->droppacket_lock);
4277 spin_lock_init(&ipvs->securetcp_lock);
4278 INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
4279 INIT_DELAYED_WORK(&ipvs->expire_nodest_conn_work,
4280 expire_nodest_conn_handler);
4281 ipvs->est_stopped = 0;
4282
4283 if (!net_eq(net, &init_net)) {
4284 tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
4285 if (tbl == NULL)
4286 return -ENOMEM;
4287 } else
4288 tbl = vs_vars;
4289 /* Initialize sysctl defaults */
4290 for (idx = 0; idx < ARRAY_SIZE(vs_vars); idx++) {
4291 if (tbl[idx].proc_handler == proc_do_defense_mode)
4292 tbl[idx].extra2 = ipvs;
4293 }
4294 idx = 0;
4295 ipvs->sysctl_amemthresh = 1024;
4296 tbl[idx++].data = &ipvs->sysctl_amemthresh;
4297 ipvs->sysctl_am_droprate = 10;
4298 tbl[idx++].data = &ipvs->sysctl_am_droprate;
4299 tbl[idx++].data = &ipvs->sysctl_drop_entry;
4300 tbl[idx++].data = &ipvs->sysctl_drop_packet;
4301 #ifdef CONFIG_IP_VS_NFCT
4302 tbl[idx++].data = &ipvs->sysctl_conntrack;
4303 #endif
4304 tbl[idx++].data = &ipvs->sysctl_secure_tcp;
4305 ipvs->sysctl_snat_reroute = 1;
4306 tbl[idx++].data = &ipvs->sysctl_snat_reroute;
4307 ipvs->sysctl_sync_ver = 1;
4308 tbl[idx++].data = &ipvs->sysctl_sync_ver;
4309 ipvs->sysctl_sync_ports = 1;
4310 tbl[idx++].data = &ipvs->sysctl_sync_ports;
4311 tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
4312
4313 ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
4314 if (unpriv)
4315 tbl[idx].mode = 0444;
4316 tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
4317
4318 ipvs->sysctl_sync_sock_size = 0;
4319 if (unpriv)
4320 tbl[idx].mode = 0444;
4321 tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
4322
4323 tbl[idx++].data = &ipvs->sysctl_cache_bypass;
4324 tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
4325 tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
4326 tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
4327 tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
4328 ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
4329 ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
4330 tbl[idx].data = &ipvs->sysctl_sync_threshold;
4331 tbl[idx].extra2 = ipvs;
4332 tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
4333 ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
4334 tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
4335 ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
4336 tbl[idx++].data = &ipvs->sysctl_sync_retries;
4337 tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
4338 ipvs->sysctl_pmtu_disc = 1;
4339 tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
4340 tbl[idx++].data = &ipvs->sysctl_backup_only;
4341 ipvs->sysctl_conn_reuse_mode = 1;
4342 tbl[idx++].data = &ipvs->sysctl_conn_reuse_mode;
4343 tbl[idx++].data = &ipvs->sysctl_schedule_icmp;
4344 tbl[idx++].data = &ipvs->sysctl_ignore_tunneled;
4345
4346 ipvs->sysctl_run_estimation = 1;
4347 if (unpriv)
4348 tbl[idx].mode = 0444;
4349 tbl[idx].extra2 = ipvs;
4350 tbl[idx++].data = &ipvs->sysctl_run_estimation;
4351
4352 ipvs->est_cpulist_valid = 0;
4353 if (unpriv)
4354 tbl[idx].mode = 0444;
4355 tbl[idx].extra2 = ipvs;
4356 tbl[idx++].data = &ipvs->sysctl_est_cpulist;
4357
4358 ipvs->sysctl_est_nice = IPVS_EST_NICE;
4359 if (unpriv)
4360 tbl[idx].mode = 0444;
4361 tbl[idx].extra2 = ipvs;
4362 tbl[idx++].data = &ipvs->sysctl_est_nice;
4363
4364 #ifdef CONFIG_IP_VS_DEBUG
4365 /* Global sysctls must be ro in non-init netns */
4366 if (!net_eq(net, &init_net))
4367 tbl[idx++].mode = 0444;
4368 #endif
4369
4370 ret = -ENOMEM;
4371 ipvs->sysctl_hdr = register_net_sysctl_sz(net, "net/ipv4/vs", tbl,
4372 ctl_table_size);
4373 if (!ipvs->sysctl_hdr)
4374 goto err;
4375 ipvs->sysctl_tbl = tbl;
4376
4377 ret = ip_vs_start_estimator(ipvs, &ipvs->tot_stats->s);
4378 if (ret < 0)
4379 goto err;
4380
4381 /* Schedule defense work */
4382 queue_delayed_work(system_long_wq, &ipvs->defense_work,
4383 DEFENSE_TIMER_PERIOD);
4384
4385 return 0;
4386
4387 err:
4388 unregister_net_sysctl_table(ipvs->sysctl_hdr);
4389 if (!net_eq(net, &init_net))
4390 kfree(tbl);
4391 return ret;
4392 }
4393
ip_vs_control_net_cleanup_sysctl(struct netns_ipvs * ipvs)4394 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs)
4395 {
4396 struct net *net = ipvs->net;
4397
4398 cancel_delayed_work_sync(&ipvs->expire_nodest_conn_work);
4399 cancel_delayed_work_sync(&ipvs->defense_work);
4400 cancel_work_sync(&ipvs->defense_work.work);
4401 unregister_net_sysctl_table(ipvs->sysctl_hdr);
4402 ip_vs_stop_estimator(ipvs, &ipvs->tot_stats->s);
4403
4404 if (ipvs->est_cpulist_valid)
4405 free_cpumask_var(ipvs->sysctl_est_cpulist);
4406
4407 if (!net_eq(net, &init_net))
4408 kfree(ipvs->sysctl_tbl);
4409 }
4410
4411 #else
4412
ip_vs_control_net_init_sysctl(struct netns_ipvs * ipvs)4413 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs) { return 0; }
ip_vs_control_net_cleanup_sysctl(struct netns_ipvs * ipvs)4414 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs) { }
4415
4416 #endif
4417
4418 static struct notifier_block ip_vs_dst_notifier = {
4419 .notifier_call = ip_vs_dst_event,
4420 #ifdef CONFIG_IP_VS_IPV6
4421 .priority = ADDRCONF_NOTIFY_PRIORITY + 5,
4422 #endif
4423 };
4424
ip_vs_control_net_init(struct netns_ipvs * ipvs)4425 int __net_init ip_vs_control_net_init(struct netns_ipvs *ipvs)
4426 {
4427 int ret = -ENOMEM;
4428 int idx;
4429
4430 /* Initialize rs_table */
4431 for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
4432 INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
4433
4434 INIT_LIST_HEAD(&ipvs->dest_trash);
4435 spin_lock_init(&ipvs->dest_trash_lock);
4436 timer_setup(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire, 0);
4437 atomic_set(&ipvs->ftpsvc_counter, 0);
4438 atomic_set(&ipvs->nullsvc_counter, 0);
4439 atomic_set(&ipvs->conn_out_counter, 0);
4440
4441 INIT_DELAYED_WORK(&ipvs->est_reload_work, est_reload_work_handler);
4442
4443 /* procfs stats */
4444 ipvs->tot_stats = kzalloc(sizeof(*ipvs->tot_stats), GFP_KERNEL);
4445 if (!ipvs->tot_stats)
4446 goto out;
4447 if (ip_vs_stats_init_alloc(&ipvs->tot_stats->s) < 0)
4448 goto err_tot_stats;
4449
4450 #ifdef CONFIG_PROC_FS
4451 if (!proc_create_net("ip_vs", 0, ipvs->net->proc_net,
4452 &ip_vs_info_seq_ops, sizeof(struct ip_vs_iter)))
4453 goto err_vs;
4454 if (!proc_create_net_single("ip_vs_stats", 0, ipvs->net->proc_net,
4455 ip_vs_stats_show, NULL))
4456 goto err_stats;
4457 if (!proc_create_net_single("ip_vs_stats_percpu", 0,
4458 ipvs->net->proc_net,
4459 ip_vs_stats_percpu_show, NULL))
4460 goto err_percpu;
4461 #endif
4462
4463 ret = ip_vs_control_net_init_sysctl(ipvs);
4464 if (ret < 0)
4465 goto err;
4466
4467 return 0;
4468
4469 err:
4470 #ifdef CONFIG_PROC_FS
4471 remove_proc_entry("ip_vs_stats_percpu", ipvs->net->proc_net);
4472
4473 err_percpu:
4474 remove_proc_entry("ip_vs_stats", ipvs->net->proc_net);
4475
4476 err_stats:
4477 remove_proc_entry("ip_vs", ipvs->net->proc_net);
4478
4479 err_vs:
4480 #endif
4481 ip_vs_stats_release(&ipvs->tot_stats->s);
4482
4483 err_tot_stats:
4484 kfree(ipvs->tot_stats);
4485
4486 out:
4487 return ret;
4488 }
4489
ip_vs_control_net_cleanup(struct netns_ipvs * ipvs)4490 void __net_exit ip_vs_control_net_cleanup(struct netns_ipvs *ipvs)
4491 {
4492 ip_vs_trash_cleanup(ipvs);
4493 ip_vs_control_net_cleanup_sysctl(ipvs);
4494 cancel_delayed_work_sync(&ipvs->est_reload_work);
4495 #ifdef CONFIG_PROC_FS
4496 remove_proc_entry("ip_vs_stats_percpu", ipvs->net->proc_net);
4497 remove_proc_entry("ip_vs_stats", ipvs->net->proc_net);
4498 remove_proc_entry("ip_vs", ipvs->net->proc_net);
4499 #endif
4500 call_rcu(&ipvs->tot_stats->rcu_head, ip_vs_stats_rcu_free);
4501 }
4502
ip_vs_register_nl_ioctl(void)4503 int __init ip_vs_register_nl_ioctl(void)
4504 {
4505 int ret;
4506
4507 ret = nf_register_sockopt(&ip_vs_sockopts);
4508 if (ret) {
4509 pr_err("cannot register sockopt.\n");
4510 goto err_sock;
4511 }
4512
4513 ret = ip_vs_genl_register();
4514 if (ret) {
4515 pr_err("cannot register Generic Netlink interface.\n");
4516 goto err_genl;
4517 }
4518 return 0;
4519
4520 err_genl:
4521 nf_unregister_sockopt(&ip_vs_sockopts);
4522 err_sock:
4523 return ret;
4524 }
4525
ip_vs_unregister_nl_ioctl(void)4526 void ip_vs_unregister_nl_ioctl(void)
4527 {
4528 ip_vs_genl_unregister();
4529 nf_unregister_sockopt(&ip_vs_sockopts);
4530 }
4531
ip_vs_control_init(void)4532 int __init ip_vs_control_init(void)
4533 {
4534 int idx;
4535 int ret;
4536
4537 /* Initialize svc_table, ip_vs_svc_fwm_table */
4538 for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
4539 INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
4540 INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
4541 }
4542
4543 smp_wmb(); /* Do we really need it now ? */
4544
4545 ret = register_netdevice_notifier(&ip_vs_dst_notifier);
4546 if (ret < 0)
4547 return ret;
4548
4549 return 0;
4550 }
4551
4552
ip_vs_control_cleanup(void)4553 void ip_vs_control_cleanup(void)
4554 {
4555 unregister_netdevice_notifier(&ip_vs_dst_notifier);
4556 /* relying on common rcu_barrier() in ip_vs_cleanup() */
4557 }
4558