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 /* semaphore for IPVS sockopts. And, [gs]etsockopt may sleep. */
53 static DEFINE_MUTEX(__ip_vs_mutex);
54 
55 /* sysctl variables */
56 
57 #ifdef CONFIG_IP_VS_DEBUG
58 static int sysctl_ip_vs_debug_level = 0;
59 
ip_vs_get_debug_level(void)60 int ip_vs_get_debug_level(void)
61 {
62 	return sysctl_ip_vs_debug_level;
63 }
64 #endif
65 
66 
67 /*  Protos */
68 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup);
69 
70 
71 #ifdef CONFIG_IP_VS_IPV6
72 /* 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)73 static bool __ip_vs_addr_is_local_v6(struct net *net,
74 				     const struct in6_addr *addr)
75 {
76 	struct flowi6 fl6 = {
77 		.daddr = *addr,
78 	};
79 	struct dst_entry *dst = ip6_route_output(net, NULL, &fl6);
80 	bool is_local;
81 
82 	is_local = !dst->error && dst->dev && (dst->dev->flags & IFF_LOOPBACK);
83 
84 	dst_release(dst);
85 	return is_local;
86 }
87 #endif
88 
89 #ifdef CONFIG_SYSCTL
90 /*
91  *	update_defense_level is called from keventd and from sysctl,
92  *	so it needs to protect itself from softirqs
93  */
update_defense_level(struct netns_ipvs * ipvs)94 static void update_defense_level(struct netns_ipvs *ipvs)
95 {
96 	struct sysinfo i;
97 	int availmem;
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 	nomem = (availmem < ipvs->sysctl_amemthresh);
110 
111 	local_bh_disable();
112 
113 	/* drop_entry */
114 	spin_lock(&ipvs->dropentry_lock);
115 	switch (ipvs->sysctl_drop_entry) {
116 	case 0:
117 		atomic_set(&ipvs->dropentry, 0);
118 		break;
119 	case 1:
120 		if (nomem) {
121 			atomic_set(&ipvs->dropentry, 1);
122 			ipvs->sysctl_drop_entry = 2;
123 		} else {
124 			atomic_set(&ipvs->dropentry, 0);
125 		}
126 		break;
127 	case 2:
128 		if (nomem) {
129 			atomic_set(&ipvs->dropentry, 1);
130 		} else {
131 			atomic_set(&ipvs->dropentry, 0);
132 			ipvs->sysctl_drop_entry = 1;
133 		}
134 		break;
135 	case 3:
136 		atomic_set(&ipvs->dropentry, 1);
137 		break;
138 	}
139 	spin_unlock(&ipvs->dropentry_lock);
140 
141 	/* drop_packet */
142 	spin_lock(&ipvs->droppacket_lock);
143 	switch (ipvs->sysctl_drop_packet) {
144 	case 0:
145 		ipvs->drop_rate = 0;
146 		break;
147 	case 1:
148 		if (nomem) {
149 			ipvs->drop_rate = ipvs->drop_counter
150 				= ipvs->sysctl_amemthresh /
151 				(ipvs->sysctl_amemthresh-availmem);
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_rate = ipvs->drop_counter
160 				= ipvs->sysctl_amemthresh /
161 				(ipvs->sysctl_amemthresh-availmem);
162 		} else {
163 			ipvs->drop_rate = 0;
164 			ipvs->sysctl_drop_packet = 1;
165 		}
166 		break;
167 	case 3:
168 		ipvs->drop_rate = ipvs->sysctl_am_droprate;
169 		break;
170 	}
171 	spin_unlock(&ipvs->droppacket_lock);
172 
173 	/* secure_tcp */
174 	spin_lock(&ipvs->securetcp_lock);
175 	switch (ipvs->sysctl_secure_tcp) {
176 	case 0:
177 		if (ipvs->old_secure_tcp >= 2)
178 			to_change = 0;
179 		break;
180 	case 1:
181 		if (nomem) {
182 			if (ipvs->old_secure_tcp < 2)
183 				to_change = 1;
184 			ipvs->sysctl_secure_tcp = 2;
185 		} else {
186 			if (ipvs->old_secure_tcp >= 2)
187 				to_change = 0;
188 		}
189 		break;
190 	case 2:
191 		if (nomem) {
192 			if (ipvs->old_secure_tcp < 2)
193 				to_change = 1;
194 		} else {
195 			if (ipvs->old_secure_tcp >= 2)
196 				to_change = 0;
197 			ipvs->sysctl_secure_tcp = 1;
198 		}
199 		break;
200 	case 3:
201 		if (ipvs->old_secure_tcp < 2)
202 			to_change = 1;
203 		break;
204 	}
205 	ipvs->old_secure_tcp = ipvs->sysctl_secure_tcp;
206 	if (to_change >= 0)
207 		ip_vs_protocol_timeout_change(ipvs,
208 					      ipvs->sysctl_secure_tcp > 1);
209 	spin_unlock(&ipvs->securetcp_lock);
210 
211 	local_bh_enable();
212 }
213 
214 /* Handler for delayed work for expiring no
215  * destination connections
216  */
expire_nodest_conn_handler(struct work_struct * work)217 static void expire_nodest_conn_handler(struct work_struct *work)
218 {
219 	struct netns_ipvs *ipvs;
220 
221 	ipvs = container_of(work, struct netns_ipvs,
222 			    expire_nodest_conn_work.work);
223 	ip_vs_expire_nodest_conn_flush(ipvs);
224 }
225 
226 /*
227  *	Timer for checking the defense
228  */
229 #define DEFENSE_TIMER_PERIOD	1*HZ
230 
defense_work_handler(struct work_struct * work)231 static void defense_work_handler(struct work_struct *work)
232 {
233 	struct netns_ipvs *ipvs =
234 		container_of(work, struct netns_ipvs, defense_work.work);
235 
236 	update_defense_level(ipvs);
237 	if (atomic_read(&ipvs->dropentry))
238 		ip_vs_random_dropentry(ipvs);
239 	queue_delayed_work(system_long_wq, &ipvs->defense_work,
240 			   DEFENSE_TIMER_PERIOD);
241 }
242 #endif
243 
244 int
ip_vs_use_count_inc(void)245 ip_vs_use_count_inc(void)
246 {
247 	return try_module_get(THIS_MODULE);
248 }
249 
250 void
ip_vs_use_count_dec(void)251 ip_vs_use_count_dec(void)
252 {
253 	module_put(THIS_MODULE);
254 }
255 
256 
257 /*
258  *	Hash table: for virtual service lookups
259  */
260 #define IP_VS_SVC_TAB_BITS 8
261 #define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
262 #define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)
263 
264 /* the service table hashed by <protocol, addr, port> */
265 static struct hlist_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE];
266 /* the service table hashed by fwmark */
267 static struct hlist_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE];
268 
269 
270 /*
271  *	Returns hash value for virtual service
272  */
273 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)274 ip_vs_svc_hashkey(struct netns_ipvs *ipvs, int af, unsigned int proto,
275 		  const union nf_inet_addr *addr, __be16 port)
276 {
277 	unsigned int porth = ntohs(port);
278 	__be32 addr_fold = addr->ip;
279 	__u32 ahash;
280 
281 #ifdef CONFIG_IP_VS_IPV6
282 	if (af == AF_INET6)
283 		addr_fold = addr->ip6[0]^addr->ip6[1]^
284 			    addr->ip6[2]^addr->ip6[3];
285 #endif
286 	ahash = ntohl(addr_fold);
287 	ahash ^= ((size_t) ipvs >> 8);
288 
289 	return (proto ^ ahash ^ (porth >> IP_VS_SVC_TAB_BITS) ^ porth) &
290 	       IP_VS_SVC_TAB_MASK;
291 }
292 
293 /*
294  *	Returns hash value of fwmark for virtual service lookup
295  */
ip_vs_svc_fwm_hashkey(struct netns_ipvs * ipvs,__u32 fwmark)296 static inline unsigned int ip_vs_svc_fwm_hashkey(struct netns_ipvs *ipvs, __u32 fwmark)
297 {
298 	return (((size_t)ipvs>>8) ^ fwmark) & IP_VS_SVC_TAB_MASK;
299 }
300 
301 /*
302  *	Hashes a service in the ip_vs_svc_table by <netns,proto,addr,port>
303  *	or in the ip_vs_svc_fwm_table by fwmark.
304  *	Should be called with locked tables.
305  */
ip_vs_svc_hash(struct ip_vs_service * svc)306 static int ip_vs_svc_hash(struct ip_vs_service *svc)
307 {
308 	unsigned int hash;
309 
310 	if (svc->flags & IP_VS_SVC_F_HASHED) {
311 		pr_err("%s(): request for already hashed, called from %pS\n",
312 		       __func__, __builtin_return_address(0));
313 		return 0;
314 	}
315 
316 	if (svc->fwmark == 0) {
317 		/*
318 		 *  Hash it by <netns,protocol,addr,port> in ip_vs_svc_table
319 		 */
320 		hash = ip_vs_svc_hashkey(svc->ipvs, svc->af, svc->protocol,
321 					 &svc->addr, svc->port);
322 		hlist_add_head_rcu(&svc->s_list, &ip_vs_svc_table[hash]);
323 	} else {
324 		/*
325 		 *  Hash it by fwmark in svc_fwm_table
326 		 */
327 		hash = ip_vs_svc_fwm_hashkey(svc->ipvs, svc->fwmark);
328 		hlist_add_head_rcu(&svc->f_list, &ip_vs_svc_fwm_table[hash]);
329 	}
330 
331 	svc->flags |= IP_VS_SVC_F_HASHED;
332 	/* increase its refcnt because it is referenced by the svc table */
333 	atomic_inc(&svc->refcnt);
334 	return 1;
335 }
336 
337 
338 /*
339  *	Unhashes a service from svc_table / svc_fwm_table.
340  *	Should be called with locked tables.
341  */
ip_vs_svc_unhash(struct ip_vs_service * svc)342 static int ip_vs_svc_unhash(struct ip_vs_service *svc)
343 {
344 	if (!(svc->flags & IP_VS_SVC_F_HASHED)) {
345 		pr_err("%s(): request for unhash flagged, called from %pS\n",
346 		       __func__, __builtin_return_address(0));
347 		return 0;
348 	}
349 
350 	if (svc->fwmark == 0) {
351 		/* Remove it from the svc_table table */
352 		hlist_del_rcu(&svc->s_list);
353 	} else {
354 		/* Remove it from the svc_fwm_table table */
355 		hlist_del_rcu(&svc->f_list);
356 	}
357 
358 	svc->flags &= ~IP_VS_SVC_F_HASHED;
359 	atomic_dec(&svc->refcnt);
360 	return 1;
361 }
362 
363 
364 /*
365  *	Get service by {netns, proto,addr,port} in the service table.
366  */
367 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)368 __ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u16 protocol,
369 		     const union nf_inet_addr *vaddr, __be16 vport)
370 {
371 	unsigned int hash;
372 	struct ip_vs_service *svc;
373 
374 	/* Check for "full" addressed entries */
375 	hash = ip_vs_svc_hashkey(ipvs, af, protocol, vaddr, vport);
376 
377 	hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[hash], s_list) {
378 		if ((svc->af == af)
379 		    && ip_vs_addr_equal(af, &svc->addr, vaddr)
380 		    && (svc->port == vport)
381 		    && (svc->protocol == protocol)
382 		    && (svc->ipvs == ipvs)) {
383 			/* HIT */
384 			return svc;
385 		}
386 	}
387 
388 	return NULL;
389 }
390 
391 
392 /*
393  *	Get service by {fwmark} in the service table.
394  */
395 static inline struct ip_vs_service *
__ip_vs_svc_fwm_find(struct netns_ipvs * ipvs,int af,__u32 fwmark)396 __ip_vs_svc_fwm_find(struct netns_ipvs *ipvs, int af, __u32 fwmark)
397 {
398 	unsigned int hash;
399 	struct ip_vs_service *svc;
400 
401 	/* Check for fwmark addressed entries */
402 	hash = ip_vs_svc_fwm_hashkey(ipvs, fwmark);
403 
404 	hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[hash], f_list) {
405 		if (svc->fwmark == fwmark && svc->af == af
406 		    && (svc->ipvs == ipvs)) {
407 			/* HIT */
408 			return svc;
409 		}
410 	}
411 
412 	return NULL;
413 }
414 
415 /* Find service, called under RCU lock */
416 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)417 ip_vs_service_find(struct netns_ipvs *ipvs, int af, __u32 fwmark, __u16 protocol,
418 		   const union nf_inet_addr *vaddr, __be16 vport)
419 {
420 	struct ip_vs_service *svc;
421 
422 	/*
423 	 *	Check the table hashed by fwmark first
424 	 */
425 	if (fwmark) {
426 		svc = __ip_vs_svc_fwm_find(ipvs, af, fwmark);
427 		if (svc)
428 			goto out;
429 	}
430 
431 	/*
432 	 *	Check the table hashed by <protocol,addr,port>
433 	 *	for "full" addressed entries
434 	 */
435 	svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, vport);
436 
437 	if (!svc && protocol == IPPROTO_TCP &&
438 	    atomic_read(&ipvs->ftpsvc_counter) &&
439 	    (vport == FTPDATA || !inet_port_requires_bind_service(ipvs->net, ntohs(vport)))) {
440 		/*
441 		 * Check if ftp service entry exists, the packet
442 		 * might belong to FTP data connections.
443 		 */
444 		svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, FTPPORT);
445 	}
446 
447 	if (svc == NULL
448 	    && atomic_read(&ipvs->nullsvc_counter)) {
449 		/*
450 		 * Check if the catch-all port (port zero) exists
451 		 */
452 		svc = __ip_vs_service_find(ipvs, af, protocol, vaddr, 0);
453 	}
454 
455   out:
456 	IP_VS_DBG_BUF(9, "lookup service: fwm %u %s %s:%u %s\n",
457 		      fwmark, ip_vs_proto_name(protocol),
458 		      IP_VS_DBG_ADDR(af, vaddr), ntohs(vport),
459 		      svc ? "hit" : "not hit");
460 
461 	return svc;
462 }
463 
464 
465 static inline void
__ip_vs_bind_svc(struct ip_vs_dest * dest,struct ip_vs_service * svc)466 __ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
467 {
468 	atomic_inc(&svc->refcnt);
469 	rcu_assign_pointer(dest->svc, svc);
470 }
471 
ip_vs_service_free(struct ip_vs_service * svc)472 static void ip_vs_service_free(struct ip_vs_service *svc)
473 {
474 	free_percpu(svc->stats.cpustats);
475 	kfree(svc);
476 }
477 
ip_vs_service_rcu_free(struct rcu_head * head)478 static void ip_vs_service_rcu_free(struct rcu_head *head)
479 {
480 	struct ip_vs_service *svc;
481 
482 	svc = container_of(head, struct ip_vs_service, rcu_head);
483 	ip_vs_service_free(svc);
484 }
485 
__ip_vs_svc_put(struct ip_vs_service * svc,bool do_delay)486 static void __ip_vs_svc_put(struct ip_vs_service *svc, bool do_delay)
487 {
488 	if (atomic_dec_and_test(&svc->refcnt)) {
489 		IP_VS_DBG_BUF(3, "Removing service %u/%s:%u\n",
490 			      svc->fwmark,
491 			      IP_VS_DBG_ADDR(svc->af, &svc->addr),
492 			      ntohs(svc->port));
493 		if (do_delay)
494 			call_rcu(&svc->rcu_head, ip_vs_service_rcu_free);
495 		else
496 			ip_vs_service_free(svc);
497 	}
498 }
499 
500 
501 /*
502  *	Returns hash value for real service
503  */
ip_vs_rs_hashkey(int af,const union nf_inet_addr * addr,__be16 port)504 static inline unsigned int ip_vs_rs_hashkey(int af,
505 					    const union nf_inet_addr *addr,
506 					    __be16 port)
507 {
508 	unsigned int porth = ntohs(port);
509 	__be32 addr_fold = addr->ip;
510 
511 #ifdef CONFIG_IP_VS_IPV6
512 	if (af == AF_INET6)
513 		addr_fold = addr->ip6[0]^addr->ip6[1]^
514 			    addr->ip6[2]^addr->ip6[3];
515 #endif
516 
517 	return (ntohl(addr_fold)^(porth>>IP_VS_RTAB_BITS)^porth)
518 		& IP_VS_RTAB_MASK;
519 }
520 
521 /* Hash ip_vs_dest in rs_table by <proto,addr,port>. */
ip_vs_rs_hash(struct netns_ipvs * ipvs,struct ip_vs_dest * dest)522 static void ip_vs_rs_hash(struct netns_ipvs *ipvs, struct ip_vs_dest *dest)
523 {
524 	unsigned int hash;
525 	__be16 port;
526 
527 	if (dest->in_rs_table)
528 		return;
529 
530 	switch (IP_VS_DFWD_METHOD(dest)) {
531 	case IP_VS_CONN_F_MASQ:
532 		port = dest->port;
533 		break;
534 	case IP_VS_CONN_F_TUNNEL:
535 		switch (dest->tun_type) {
536 		case IP_VS_CONN_F_TUNNEL_TYPE_GUE:
537 			port = dest->tun_port;
538 			break;
539 		case IP_VS_CONN_F_TUNNEL_TYPE_IPIP:
540 		case IP_VS_CONN_F_TUNNEL_TYPE_GRE:
541 			port = 0;
542 			break;
543 		default:
544 			return;
545 		}
546 		break;
547 	default:
548 		return;
549 	}
550 
551 	/*
552 	 *	Hash by proto,addr,port,
553 	 *	which are the parameters of the real service.
554 	 */
555 	hash = ip_vs_rs_hashkey(dest->af, &dest->addr, port);
556 
557 	hlist_add_head_rcu(&dest->d_list, &ipvs->rs_table[hash]);
558 	dest->in_rs_table = 1;
559 }
560 
561 /* Unhash ip_vs_dest from rs_table. */
ip_vs_rs_unhash(struct ip_vs_dest * dest)562 static void ip_vs_rs_unhash(struct ip_vs_dest *dest)
563 {
564 	/*
565 	 * Remove it from the rs_table table.
566 	 */
567 	if (dest->in_rs_table) {
568 		hlist_del_rcu(&dest->d_list);
569 		dest->in_rs_table = 0;
570 	}
571 }
572 
573 /* 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)574 bool ip_vs_has_real_service(struct netns_ipvs *ipvs, int af, __u16 protocol,
575 			    const union nf_inet_addr *daddr, __be16 dport)
576 {
577 	unsigned int hash;
578 	struct ip_vs_dest *dest;
579 
580 	/* Check for "full" addressed entries */
581 	hash = ip_vs_rs_hashkey(af, daddr, dport);
582 
583 	hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
584 		if (dest->port == dport &&
585 		    dest->af == af &&
586 		    ip_vs_addr_equal(af, &dest->addr, daddr) &&
587 		    (dest->protocol == protocol || dest->vfwmark) &&
588 		    IP_VS_DFWD_METHOD(dest) == IP_VS_CONN_F_MASQ) {
589 			/* HIT */
590 			return true;
591 		}
592 	}
593 
594 	return false;
595 }
596 
597 /* Find real service record by <proto,addr,port>.
598  * In case of multiple records with the same <proto,addr,port>, only
599  * the first found record is returned.
600  *
601  * To be called under RCU lock.
602  */
ip_vs_find_real_service(struct netns_ipvs * ipvs,int af,__u16 protocol,const union nf_inet_addr * daddr,__be16 dport)603 struct ip_vs_dest *ip_vs_find_real_service(struct netns_ipvs *ipvs, int af,
604 					   __u16 protocol,
605 					   const union nf_inet_addr *daddr,
606 					   __be16 dport)
607 {
608 	unsigned int hash;
609 	struct ip_vs_dest *dest;
610 
611 	/* Check for "full" addressed entries */
612 	hash = ip_vs_rs_hashkey(af, daddr, dport);
613 
614 	hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
615 		if (dest->port == dport &&
616 		    dest->af == af &&
617 		    ip_vs_addr_equal(af, &dest->addr, daddr) &&
618 		    (dest->protocol == protocol || dest->vfwmark) &&
619 		    IP_VS_DFWD_METHOD(dest) == IP_VS_CONN_F_MASQ) {
620 			/* HIT */
621 			return dest;
622 		}
623 	}
624 
625 	return NULL;
626 }
627 
628 /* Find real service record by <af,addr,tun_port>.
629  * In case of multiple records with the same <af,addr,tun_port>, only
630  * the first found record is returned.
631  *
632  * To be called under RCU lock.
633  */
ip_vs_find_tunnel(struct netns_ipvs * ipvs,int af,const union nf_inet_addr * daddr,__be16 tun_port)634 struct ip_vs_dest *ip_vs_find_tunnel(struct netns_ipvs *ipvs, int af,
635 				     const union nf_inet_addr *daddr,
636 				     __be16 tun_port)
637 {
638 	struct ip_vs_dest *dest;
639 	unsigned int hash;
640 
641 	/* Check for "full" addressed entries */
642 	hash = ip_vs_rs_hashkey(af, daddr, tun_port);
643 
644 	hlist_for_each_entry_rcu(dest, &ipvs->rs_table[hash], d_list) {
645 		if (dest->tun_port == tun_port &&
646 		    dest->af == af &&
647 		    ip_vs_addr_equal(af, &dest->addr, daddr) &&
648 		    IP_VS_DFWD_METHOD(dest) == IP_VS_CONN_F_TUNNEL) {
649 			/* HIT */
650 			return dest;
651 		}
652 	}
653 
654 	return NULL;
655 }
656 
657 /* Lookup destination by {addr,port} in the given service
658  * Called under RCU lock.
659  */
660 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)661 ip_vs_lookup_dest(struct ip_vs_service *svc, int dest_af,
662 		  const union nf_inet_addr *daddr, __be16 dport)
663 {
664 	struct ip_vs_dest *dest;
665 
666 	/*
667 	 * Find the destination for the given service
668 	 */
669 	list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
670 		if ((dest->af == dest_af) &&
671 		    ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
672 		    (dest->port == dport)) {
673 			/* HIT */
674 			return dest;
675 		}
676 	}
677 
678 	return NULL;
679 }
680 
681 /*
682  * Find destination by {daddr,dport,vaddr,protocol}
683  * Created to be used in ip_vs_process_message() in
684  * the backup synchronization daemon. It finds the
685  * destination to be bound to the received connection
686  * on the backup.
687  * Called under RCU lock, no refcnt is returned.
688  */
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)689 struct ip_vs_dest *ip_vs_find_dest(struct netns_ipvs *ipvs, int svc_af, int dest_af,
690 				   const union nf_inet_addr *daddr,
691 				   __be16 dport,
692 				   const union nf_inet_addr *vaddr,
693 				   __be16 vport, __u16 protocol, __u32 fwmark,
694 				   __u32 flags)
695 {
696 	struct ip_vs_dest *dest;
697 	struct ip_vs_service *svc;
698 	__be16 port = dport;
699 
700 	svc = ip_vs_service_find(ipvs, svc_af, fwmark, protocol, vaddr, vport);
701 	if (!svc)
702 		return NULL;
703 	if (fwmark && (flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ)
704 		port = 0;
705 	dest = ip_vs_lookup_dest(svc, dest_af, daddr, port);
706 	if (!dest)
707 		dest = ip_vs_lookup_dest(svc, dest_af, daddr, port ^ dport);
708 	return dest;
709 }
710 
ip_vs_dest_dst_rcu_free(struct rcu_head * head)711 void ip_vs_dest_dst_rcu_free(struct rcu_head *head)
712 {
713 	struct ip_vs_dest_dst *dest_dst = container_of(head,
714 						       struct ip_vs_dest_dst,
715 						       rcu_head);
716 
717 	dst_release(dest_dst->dst_cache);
718 	kfree(dest_dst);
719 }
720 
721 /* Release dest_dst and dst_cache for dest in user context */
__ip_vs_dst_cache_reset(struct ip_vs_dest * dest)722 static void __ip_vs_dst_cache_reset(struct ip_vs_dest *dest)
723 {
724 	struct ip_vs_dest_dst *old;
725 
726 	old = rcu_dereference_protected(dest->dest_dst, 1);
727 	if (old) {
728 		RCU_INIT_POINTER(dest->dest_dst, NULL);
729 		call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
730 	}
731 }
732 
733 /*
734  *  Lookup dest by {svc,addr,port} in the destination trash.
735  *  The destination trash is used to hold the destinations that are removed
736  *  from the service table but are still referenced by some conn entries.
737  *  The reason to add the destination trash is when the dest is temporary
738  *  down (either by administrator or by monitor program), the dest can be
739  *  picked back from the trash, the remaining connections to the dest can
740  *  continue, and the counting information of the dest is also useful for
741  *  scheduling.
742  */
743 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)744 ip_vs_trash_get_dest(struct ip_vs_service *svc, int dest_af,
745 		     const union nf_inet_addr *daddr, __be16 dport)
746 {
747 	struct ip_vs_dest *dest;
748 	struct netns_ipvs *ipvs = svc->ipvs;
749 
750 	/*
751 	 * Find the destination in trash
752 	 */
753 	spin_lock_bh(&ipvs->dest_trash_lock);
754 	list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
755 		IP_VS_DBG_BUF(3, "Destination %u/%s:%u still in trash, "
756 			      "dest->refcnt=%d\n",
757 			      dest->vfwmark,
758 			      IP_VS_DBG_ADDR(dest->af, &dest->addr),
759 			      ntohs(dest->port),
760 			      refcount_read(&dest->refcnt));
761 		if (dest->af == dest_af &&
762 		    ip_vs_addr_equal(dest_af, &dest->addr, daddr) &&
763 		    dest->port == dport &&
764 		    dest->vfwmark == svc->fwmark &&
765 		    dest->protocol == svc->protocol &&
766 		    (svc->fwmark ||
767 		     (ip_vs_addr_equal(svc->af, &dest->vaddr, &svc->addr) &&
768 		      dest->vport == svc->port))) {
769 			/* HIT */
770 			list_del(&dest->t_list);
771 			goto out;
772 		}
773 	}
774 
775 	dest = NULL;
776 
777 out:
778 	spin_unlock_bh(&ipvs->dest_trash_lock);
779 
780 	return dest;
781 }
782 
ip_vs_dest_free(struct ip_vs_dest * dest)783 static void ip_vs_dest_free(struct ip_vs_dest *dest)
784 {
785 	struct ip_vs_service *svc = rcu_dereference_protected(dest->svc, 1);
786 
787 	__ip_vs_dst_cache_reset(dest);
788 	__ip_vs_svc_put(svc, false);
789 	free_percpu(dest->stats.cpustats);
790 	ip_vs_dest_put_and_free(dest);
791 }
792 
793 /*
794  *  Clean up all the destinations in the trash
795  *  Called by the ip_vs_control_cleanup()
796  *
797  *  When the ip_vs_control_clearup is activated by ipvs module exit,
798  *  the service tables must have been flushed and all the connections
799  *  are expired, and the refcnt of each destination in the trash must
800  *  be 1, so we simply release them here.
801  */
ip_vs_trash_cleanup(struct netns_ipvs * ipvs)802 static void ip_vs_trash_cleanup(struct netns_ipvs *ipvs)
803 {
804 	struct ip_vs_dest *dest, *nxt;
805 
806 	del_timer_sync(&ipvs->dest_trash_timer);
807 	/* No need to use dest_trash_lock */
808 	list_for_each_entry_safe(dest, nxt, &ipvs->dest_trash, t_list) {
809 		list_del(&dest->t_list);
810 		ip_vs_dest_free(dest);
811 	}
812 }
813 
814 static void
ip_vs_copy_stats(struct ip_vs_kstats * dst,struct ip_vs_stats * src)815 ip_vs_copy_stats(struct ip_vs_kstats *dst, struct ip_vs_stats *src)
816 {
817 #define IP_VS_SHOW_STATS_COUNTER(c) dst->c = src->kstats.c - src->kstats0.c
818 
819 	spin_lock_bh(&src->lock);
820 
821 	IP_VS_SHOW_STATS_COUNTER(conns);
822 	IP_VS_SHOW_STATS_COUNTER(inpkts);
823 	IP_VS_SHOW_STATS_COUNTER(outpkts);
824 	IP_VS_SHOW_STATS_COUNTER(inbytes);
825 	IP_VS_SHOW_STATS_COUNTER(outbytes);
826 
827 	ip_vs_read_estimator(dst, src);
828 
829 	spin_unlock_bh(&src->lock);
830 }
831 
832 static void
ip_vs_export_stats_user(struct ip_vs_stats_user * dst,struct ip_vs_kstats * src)833 ip_vs_export_stats_user(struct ip_vs_stats_user *dst, struct ip_vs_kstats *src)
834 {
835 	dst->conns = (u32)src->conns;
836 	dst->inpkts = (u32)src->inpkts;
837 	dst->outpkts = (u32)src->outpkts;
838 	dst->inbytes = src->inbytes;
839 	dst->outbytes = src->outbytes;
840 	dst->cps = (u32)src->cps;
841 	dst->inpps = (u32)src->inpps;
842 	dst->outpps = (u32)src->outpps;
843 	dst->inbps = (u32)src->inbps;
844 	dst->outbps = (u32)src->outbps;
845 }
846 
847 static void
ip_vs_zero_stats(struct ip_vs_stats * stats)848 ip_vs_zero_stats(struct ip_vs_stats *stats)
849 {
850 	spin_lock_bh(&stats->lock);
851 
852 	/* get current counters as zero point, rates are zeroed */
853 
854 #define IP_VS_ZERO_STATS_COUNTER(c) stats->kstats0.c = stats->kstats.c
855 
856 	IP_VS_ZERO_STATS_COUNTER(conns);
857 	IP_VS_ZERO_STATS_COUNTER(inpkts);
858 	IP_VS_ZERO_STATS_COUNTER(outpkts);
859 	IP_VS_ZERO_STATS_COUNTER(inbytes);
860 	IP_VS_ZERO_STATS_COUNTER(outbytes);
861 
862 	ip_vs_zero_estimator(stats);
863 
864 	spin_unlock_bh(&stats->lock);
865 }
866 
867 /*
868  *	Update a destination in the given service
869  */
870 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)871 __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
872 		    struct ip_vs_dest_user_kern *udest, int add)
873 {
874 	struct netns_ipvs *ipvs = svc->ipvs;
875 	struct ip_vs_service *old_svc;
876 	struct ip_vs_scheduler *sched;
877 	int conn_flags;
878 
879 	/* We cannot modify an address and change the address family */
880 	BUG_ON(!add && udest->af != dest->af);
881 
882 	if (add && udest->af != svc->af)
883 		ipvs->mixed_address_family_dests++;
884 
885 	/* keep the last_weight with latest non-0 weight */
886 	if (add || udest->weight != 0)
887 		atomic_set(&dest->last_weight, udest->weight);
888 
889 	/* set the weight and the flags */
890 	atomic_set(&dest->weight, udest->weight);
891 	conn_flags = udest->conn_flags & IP_VS_CONN_F_DEST_MASK;
892 	conn_flags |= IP_VS_CONN_F_INACTIVE;
893 
894 	/* Need to rehash? */
895 	if ((udest->conn_flags & IP_VS_CONN_F_FWD_MASK) !=
896 	    IP_VS_DFWD_METHOD(dest) ||
897 	    udest->tun_type != dest->tun_type ||
898 	    udest->tun_port != dest->tun_port)
899 		ip_vs_rs_unhash(dest);
900 
901 	/* set the tunnel info */
902 	dest->tun_type = udest->tun_type;
903 	dest->tun_port = udest->tun_port;
904 	dest->tun_flags = udest->tun_flags;
905 
906 	/* set the IP_VS_CONN_F_NOOUTPUT flag if not masquerading/NAT */
907 	if ((conn_flags & IP_VS_CONN_F_FWD_MASK) != IP_VS_CONN_F_MASQ) {
908 		conn_flags |= IP_VS_CONN_F_NOOUTPUT;
909 	} else {
910 		/* FTP-NAT requires conntrack for mangling */
911 		if (svc->port == FTPPORT)
912 			ip_vs_register_conntrack(svc);
913 	}
914 	atomic_set(&dest->conn_flags, conn_flags);
915 	/* Put the real service in rs_table if not present. */
916 	ip_vs_rs_hash(ipvs, dest);
917 
918 	/* bind the service */
919 	old_svc = rcu_dereference_protected(dest->svc, 1);
920 	if (!old_svc) {
921 		__ip_vs_bind_svc(dest, svc);
922 	} else {
923 		if (old_svc != svc) {
924 			ip_vs_zero_stats(&dest->stats);
925 			__ip_vs_bind_svc(dest, svc);
926 			__ip_vs_svc_put(old_svc, true);
927 		}
928 	}
929 
930 	/* set the dest status flags */
931 	dest->flags |= IP_VS_DEST_F_AVAILABLE;
932 
933 	if (udest->u_threshold == 0 || udest->u_threshold > dest->u_threshold)
934 		dest->flags &= ~IP_VS_DEST_F_OVERLOAD;
935 	dest->u_threshold = udest->u_threshold;
936 	dest->l_threshold = udest->l_threshold;
937 
938 	dest->af = udest->af;
939 
940 	spin_lock_bh(&dest->dst_lock);
941 	__ip_vs_dst_cache_reset(dest);
942 	spin_unlock_bh(&dest->dst_lock);
943 
944 	if (add) {
945 		ip_vs_start_estimator(svc->ipvs, &dest->stats);
946 		list_add_rcu(&dest->n_list, &svc->destinations);
947 		svc->num_dests++;
948 		sched = rcu_dereference_protected(svc->scheduler, 1);
949 		if (sched && sched->add_dest)
950 			sched->add_dest(svc, dest);
951 	} else {
952 		sched = rcu_dereference_protected(svc->scheduler, 1);
953 		if (sched && sched->upd_dest)
954 			sched->upd_dest(svc, dest);
955 	}
956 }
957 
958 
959 /*
960  *	Create a destination for the given service
961  */
962 static int
ip_vs_new_dest(struct ip_vs_service * svc,struct ip_vs_dest_user_kern * udest)963 ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
964 {
965 	struct ip_vs_dest *dest;
966 	unsigned int atype, i;
967 
968 	EnterFunction(2);
969 
970 #ifdef CONFIG_IP_VS_IPV6
971 	if (udest->af == AF_INET6) {
972 		int ret;
973 
974 		atype = ipv6_addr_type(&udest->addr.in6);
975 		if ((!(atype & IPV6_ADDR_UNICAST) ||
976 			atype & IPV6_ADDR_LINKLOCAL) &&
977 			!__ip_vs_addr_is_local_v6(svc->ipvs->net, &udest->addr.in6))
978 			return -EINVAL;
979 
980 		ret = nf_defrag_ipv6_enable(svc->ipvs->net);
981 		if (ret)
982 			return ret;
983 	} else
984 #endif
985 	{
986 		atype = inet_addr_type(svc->ipvs->net, udest->addr.ip);
987 		if (atype != RTN_LOCAL && atype != RTN_UNICAST)
988 			return -EINVAL;
989 	}
990 
991 	dest = kzalloc(sizeof(struct ip_vs_dest), GFP_KERNEL);
992 	if (dest == NULL)
993 		return -ENOMEM;
994 
995 	dest->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
996 	if (!dest->stats.cpustats)
997 		goto err_alloc;
998 
999 	for_each_possible_cpu(i) {
1000 		struct ip_vs_cpu_stats *ip_vs_dest_stats;
1001 		ip_vs_dest_stats = per_cpu_ptr(dest->stats.cpustats, i);
1002 		u64_stats_init(&ip_vs_dest_stats->syncp);
1003 	}
1004 
1005 	dest->af = udest->af;
1006 	dest->protocol = svc->protocol;
1007 	dest->vaddr = svc->addr;
1008 	dest->vport = svc->port;
1009 	dest->vfwmark = svc->fwmark;
1010 	ip_vs_addr_copy(udest->af, &dest->addr, &udest->addr);
1011 	dest->port = udest->port;
1012 
1013 	atomic_set(&dest->activeconns, 0);
1014 	atomic_set(&dest->inactconns, 0);
1015 	atomic_set(&dest->persistconns, 0);
1016 	refcount_set(&dest->refcnt, 1);
1017 
1018 	INIT_HLIST_NODE(&dest->d_list);
1019 	spin_lock_init(&dest->dst_lock);
1020 	spin_lock_init(&dest->stats.lock);
1021 	__ip_vs_update_dest(svc, dest, udest, 1);
1022 
1023 	LeaveFunction(2);
1024 	return 0;
1025 
1026 err_alloc:
1027 	kfree(dest);
1028 	return -ENOMEM;
1029 }
1030 
1031 
1032 /*
1033  *	Add a destination into an existing service
1034  */
1035 static int
ip_vs_add_dest(struct ip_vs_service * svc,struct ip_vs_dest_user_kern * udest)1036 ip_vs_add_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1037 {
1038 	struct ip_vs_dest *dest;
1039 	union nf_inet_addr daddr;
1040 	__be16 dport = udest->port;
1041 	int ret;
1042 
1043 	EnterFunction(2);
1044 
1045 	if (udest->weight < 0) {
1046 		pr_err("%s(): server weight less than zero\n", __func__);
1047 		return -ERANGE;
1048 	}
1049 
1050 	if (udest->l_threshold > udest->u_threshold) {
1051 		pr_err("%s(): lower threshold is higher than upper threshold\n",
1052 			__func__);
1053 		return -ERANGE;
1054 	}
1055 
1056 	if (udest->tun_type == IP_VS_CONN_F_TUNNEL_TYPE_GUE) {
1057 		if (udest->tun_port == 0) {
1058 			pr_err("%s(): tunnel port is zero\n", __func__);
1059 			return -EINVAL;
1060 		}
1061 	}
1062 
1063 	ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
1064 
1065 	/* We use function that requires RCU lock */
1066 	rcu_read_lock();
1067 	dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
1068 	rcu_read_unlock();
1069 
1070 	if (dest != NULL) {
1071 		IP_VS_DBG(1, "%s(): dest already exists\n", __func__);
1072 		return -EEXIST;
1073 	}
1074 
1075 	/*
1076 	 * Check if the dest already exists in the trash and
1077 	 * is from the same service
1078 	 */
1079 	dest = ip_vs_trash_get_dest(svc, udest->af, &daddr, dport);
1080 
1081 	if (dest != NULL) {
1082 		IP_VS_DBG_BUF(3, "Get destination %s:%u from trash, "
1083 			      "dest->refcnt=%d, service %u/%s:%u\n",
1084 			      IP_VS_DBG_ADDR(udest->af, &daddr), ntohs(dport),
1085 			      refcount_read(&dest->refcnt),
1086 			      dest->vfwmark,
1087 			      IP_VS_DBG_ADDR(svc->af, &dest->vaddr),
1088 			      ntohs(dest->vport));
1089 
1090 		__ip_vs_update_dest(svc, dest, udest, 1);
1091 		ret = 0;
1092 	} else {
1093 		/*
1094 		 * Allocate and initialize the dest structure
1095 		 */
1096 		ret = ip_vs_new_dest(svc, udest);
1097 	}
1098 	LeaveFunction(2);
1099 
1100 	return ret;
1101 }
1102 
1103 
1104 /*
1105  *	Edit a destination in the given service
1106  */
1107 static int
ip_vs_edit_dest(struct ip_vs_service * svc,struct ip_vs_dest_user_kern * udest)1108 ip_vs_edit_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1109 {
1110 	struct ip_vs_dest *dest;
1111 	union nf_inet_addr daddr;
1112 	__be16 dport = udest->port;
1113 
1114 	EnterFunction(2);
1115 
1116 	if (udest->weight < 0) {
1117 		pr_err("%s(): server weight less than zero\n", __func__);
1118 		return -ERANGE;
1119 	}
1120 
1121 	if (udest->l_threshold > udest->u_threshold) {
1122 		pr_err("%s(): lower threshold is higher than upper threshold\n",
1123 			__func__);
1124 		return -ERANGE;
1125 	}
1126 
1127 	if (udest->tun_type == IP_VS_CONN_F_TUNNEL_TYPE_GUE) {
1128 		if (udest->tun_port == 0) {
1129 			pr_err("%s(): tunnel port is zero\n", __func__);
1130 			return -EINVAL;
1131 		}
1132 	}
1133 
1134 	ip_vs_addr_copy(udest->af, &daddr, &udest->addr);
1135 
1136 	/* We use function that requires RCU lock */
1137 	rcu_read_lock();
1138 	dest = ip_vs_lookup_dest(svc, udest->af, &daddr, dport);
1139 	rcu_read_unlock();
1140 
1141 	if (dest == NULL) {
1142 		IP_VS_DBG(1, "%s(): dest doesn't exist\n", __func__);
1143 		return -ENOENT;
1144 	}
1145 
1146 	__ip_vs_update_dest(svc, dest, udest, 0);
1147 	LeaveFunction(2);
1148 
1149 	return 0;
1150 }
1151 
1152 /*
1153  *	Delete a destination (must be already unlinked from the service)
1154  */
__ip_vs_del_dest(struct netns_ipvs * ipvs,struct ip_vs_dest * dest,bool cleanup)1155 static void __ip_vs_del_dest(struct netns_ipvs *ipvs, struct ip_vs_dest *dest,
1156 			     bool cleanup)
1157 {
1158 	ip_vs_stop_estimator(ipvs, &dest->stats);
1159 
1160 	/*
1161 	 *  Remove it from the d-linked list with the real services.
1162 	 */
1163 	ip_vs_rs_unhash(dest);
1164 
1165 	spin_lock_bh(&ipvs->dest_trash_lock);
1166 	IP_VS_DBG_BUF(3, "Moving dest %s:%u into trash, dest->refcnt=%d\n",
1167 		      IP_VS_DBG_ADDR(dest->af, &dest->addr), ntohs(dest->port),
1168 		      refcount_read(&dest->refcnt));
1169 	if (list_empty(&ipvs->dest_trash) && !cleanup)
1170 		mod_timer(&ipvs->dest_trash_timer,
1171 			  jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1172 	/* dest lives in trash with reference */
1173 	list_add(&dest->t_list, &ipvs->dest_trash);
1174 	dest->idle_start = 0;
1175 	spin_unlock_bh(&ipvs->dest_trash_lock);
1176 
1177 	/* Queue up delayed work to expire all no destination connections.
1178 	 * No-op when CONFIG_SYSCTL is disabled.
1179 	 */
1180 	if (!cleanup)
1181 		ip_vs_enqueue_expire_nodest_conns(ipvs);
1182 }
1183 
1184 
1185 /*
1186  *	Unlink a destination from the given service
1187  */
__ip_vs_unlink_dest(struct ip_vs_service * svc,struct ip_vs_dest * dest,int svcupd)1188 static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
1189 				struct ip_vs_dest *dest,
1190 				int svcupd)
1191 {
1192 	dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
1193 
1194 	/*
1195 	 *  Remove it from the d-linked destination list.
1196 	 */
1197 	list_del_rcu(&dest->n_list);
1198 	svc->num_dests--;
1199 
1200 	if (dest->af != svc->af)
1201 		svc->ipvs->mixed_address_family_dests--;
1202 
1203 	if (svcupd) {
1204 		struct ip_vs_scheduler *sched;
1205 
1206 		sched = rcu_dereference_protected(svc->scheduler, 1);
1207 		if (sched && sched->del_dest)
1208 			sched->del_dest(svc, dest);
1209 	}
1210 }
1211 
1212 
1213 /*
1214  *	Delete a destination server in the given service
1215  */
1216 static int
ip_vs_del_dest(struct ip_vs_service * svc,struct ip_vs_dest_user_kern * udest)1217 ip_vs_del_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest)
1218 {
1219 	struct ip_vs_dest *dest;
1220 	__be16 dport = udest->port;
1221 
1222 	EnterFunction(2);
1223 
1224 	/* We use function that requires RCU lock */
1225 	rcu_read_lock();
1226 	dest = ip_vs_lookup_dest(svc, udest->af, &udest->addr, dport);
1227 	rcu_read_unlock();
1228 
1229 	if (dest == NULL) {
1230 		IP_VS_DBG(1, "%s(): destination not found!\n", __func__);
1231 		return -ENOENT;
1232 	}
1233 
1234 	/*
1235 	 *	Unlink dest from the service
1236 	 */
1237 	__ip_vs_unlink_dest(svc, dest, 1);
1238 
1239 	/*
1240 	 *	Delete the destination
1241 	 */
1242 	__ip_vs_del_dest(svc->ipvs, dest, false);
1243 
1244 	LeaveFunction(2);
1245 
1246 	return 0;
1247 }
1248 
ip_vs_dest_trash_expire(struct timer_list * t)1249 static void ip_vs_dest_trash_expire(struct timer_list *t)
1250 {
1251 	struct netns_ipvs *ipvs = from_timer(ipvs, t, dest_trash_timer);
1252 	struct ip_vs_dest *dest, *next;
1253 	unsigned long now = jiffies;
1254 
1255 	spin_lock(&ipvs->dest_trash_lock);
1256 	list_for_each_entry_safe(dest, next, &ipvs->dest_trash, t_list) {
1257 		if (refcount_read(&dest->refcnt) > 1)
1258 			continue;
1259 		if (dest->idle_start) {
1260 			if (time_before(now, dest->idle_start +
1261 					     IP_VS_DEST_TRASH_PERIOD))
1262 				continue;
1263 		} else {
1264 			dest->idle_start = max(1UL, now);
1265 			continue;
1266 		}
1267 		IP_VS_DBG_BUF(3, "Removing destination %u/%s:%u from trash\n",
1268 			      dest->vfwmark,
1269 			      IP_VS_DBG_ADDR(dest->af, &dest->addr),
1270 			      ntohs(dest->port));
1271 		list_del(&dest->t_list);
1272 		ip_vs_dest_free(dest);
1273 	}
1274 	if (!list_empty(&ipvs->dest_trash))
1275 		mod_timer(&ipvs->dest_trash_timer,
1276 			  jiffies + (IP_VS_DEST_TRASH_PERIOD >> 1));
1277 	spin_unlock(&ipvs->dest_trash_lock);
1278 }
1279 
1280 /*
1281  *	Add a service into the service hash table
1282  */
1283 static int
ip_vs_add_service(struct netns_ipvs * ipvs,struct ip_vs_service_user_kern * u,struct ip_vs_service ** svc_p)1284 ip_vs_add_service(struct netns_ipvs *ipvs, struct ip_vs_service_user_kern *u,
1285 		  struct ip_vs_service **svc_p)
1286 {
1287 	int ret = 0, i;
1288 	struct ip_vs_scheduler *sched = NULL;
1289 	struct ip_vs_pe *pe = NULL;
1290 	struct ip_vs_service *svc = NULL;
1291 	int ret_hooks = -1;
1292 
1293 	/* increase the module use count */
1294 	if (!ip_vs_use_count_inc())
1295 		return -ENOPROTOOPT;
1296 
1297 	/* Lookup the scheduler by 'u->sched_name' */
1298 	if (strcmp(u->sched_name, "none")) {
1299 		sched = ip_vs_scheduler_get(u->sched_name);
1300 		if (!sched) {
1301 			pr_info("Scheduler module ip_vs_%s not found\n",
1302 				u->sched_name);
1303 			ret = -ENOENT;
1304 			goto out_err;
1305 		}
1306 	}
1307 
1308 	if (u->pe_name && *u->pe_name) {
1309 		pe = ip_vs_pe_getbyname(u->pe_name);
1310 		if (pe == NULL) {
1311 			pr_info("persistence engine module ip_vs_pe_%s "
1312 				"not found\n", u->pe_name);
1313 			ret = -ENOENT;
1314 			goto out_err;
1315 		}
1316 	}
1317 
1318 #ifdef CONFIG_IP_VS_IPV6
1319 	if (u->af == AF_INET6) {
1320 		__u32 plen = (__force __u32) u->netmask;
1321 
1322 		if (plen < 1 || plen > 128) {
1323 			ret = -EINVAL;
1324 			goto out_err;
1325 		}
1326 
1327 		ret = nf_defrag_ipv6_enable(ipvs->net);
1328 		if (ret)
1329 			goto out_err;
1330 	}
1331 #endif
1332 
1333 	if ((u->af == AF_INET && !ipvs->num_services) ||
1334 	    (u->af == AF_INET6 && !ipvs->num_services6)) {
1335 		ret = ip_vs_register_hooks(ipvs, u->af);
1336 		if (ret < 0)
1337 			goto out_err;
1338 		ret_hooks = ret;
1339 	}
1340 
1341 	svc = kzalloc(sizeof(struct ip_vs_service), GFP_KERNEL);
1342 	if (svc == NULL) {
1343 		IP_VS_DBG(1, "%s(): no memory\n", __func__);
1344 		ret = -ENOMEM;
1345 		goto out_err;
1346 	}
1347 	svc->stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
1348 	if (!svc->stats.cpustats) {
1349 		ret = -ENOMEM;
1350 		goto out_err;
1351 	}
1352 
1353 	for_each_possible_cpu(i) {
1354 		struct ip_vs_cpu_stats *ip_vs_stats;
1355 		ip_vs_stats = per_cpu_ptr(svc->stats.cpustats, i);
1356 		u64_stats_init(&ip_vs_stats->syncp);
1357 	}
1358 
1359 
1360 	/* I'm the first user of the service */
1361 	atomic_set(&svc->refcnt, 0);
1362 
1363 	svc->af = u->af;
1364 	svc->protocol = u->protocol;
1365 	ip_vs_addr_copy(svc->af, &svc->addr, &u->addr);
1366 	svc->port = u->port;
1367 	svc->fwmark = u->fwmark;
1368 	svc->flags = u->flags & ~IP_VS_SVC_F_HASHED;
1369 	svc->timeout = u->timeout * HZ;
1370 	svc->netmask = u->netmask;
1371 	svc->ipvs = ipvs;
1372 
1373 	INIT_LIST_HEAD(&svc->destinations);
1374 	spin_lock_init(&svc->sched_lock);
1375 	spin_lock_init(&svc->stats.lock);
1376 
1377 	/* Bind the scheduler */
1378 	if (sched) {
1379 		ret = ip_vs_bind_scheduler(svc, sched);
1380 		if (ret)
1381 			goto out_err;
1382 		sched = NULL;
1383 	}
1384 
1385 	/* Bind the ct retriever */
1386 	RCU_INIT_POINTER(svc->pe, pe);
1387 	pe = NULL;
1388 
1389 	/* Update the virtual service counters */
1390 	if (svc->port == FTPPORT)
1391 		atomic_inc(&ipvs->ftpsvc_counter);
1392 	else if (svc->port == 0)
1393 		atomic_inc(&ipvs->nullsvc_counter);
1394 	if (svc->pe && svc->pe->conn_out)
1395 		atomic_inc(&ipvs->conn_out_counter);
1396 
1397 	ip_vs_start_estimator(ipvs, &svc->stats);
1398 
1399 	/* Count only IPv4 services for old get/setsockopt interface */
1400 	if (svc->af == AF_INET)
1401 		ipvs->num_services++;
1402 	else if (svc->af == AF_INET6)
1403 		ipvs->num_services6++;
1404 
1405 	/* Hash the service into the service table */
1406 	ip_vs_svc_hash(svc);
1407 
1408 	*svc_p = svc;
1409 	/* Now there is a service - full throttle */
1410 	ipvs->enable = 1;
1411 	return 0;
1412 
1413 
1414  out_err:
1415 	if (ret_hooks >= 0)
1416 		ip_vs_unregister_hooks(ipvs, u->af);
1417 	if (svc != NULL) {
1418 		ip_vs_unbind_scheduler(svc, sched);
1419 		ip_vs_service_free(svc);
1420 	}
1421 	ip_vs_scheduler_put(sched);
1422 	ip_vs_pe_put(pe);
1423 
1424 	/* decrease the module use count */
1425 	ip_vs_use_count_dec();
1426 
1427 	return ret;
1428 }
1429 
1430 
1431 /*
1432  *	Edit a service and bind it with a new scheduler
1433  */
1434 static int
ip_vs_edit_service(struct ip_vs_service * svc,struct ip_vs_service_user_kern * u)1435 ip_vs_edit_service(struct ip_vs_service *svc, struct ip_vs_service_user_kern *u)
1436 {
1437 	struct ip_vs_scheduler *sched = NULL, *old_sched;
1438 	struct ip_vs_pe *pe = NULL, *old_pe = NULL;
1439 	int ret = 0;
1440 	bool new_pe_conn_out, old_pe_conn_out;
1441 
1442 	/*
1443 	 * Lookup the scheduler, by 'u->sched_name'
1444 	 */
1445 	if (strcmp(u->sched_name, "none")) {
1446 		sched = ip_vs_scheduler_get(u->sched_name);
1447 		if (!sched) {
1448 			pr_info("Scheduler module ip_vs_%s not found\n",
1449 				u->sched_name);
1450 			return -ENOENT;
1451 		}
1452 	}
1453 	old_sched = sched;
1454 
1455 	if (u->pe_name && *u->pe_name) {
1456 		pe = ip_vs_pe_getbyname(u->pe_name);
1457 		if (pe == NULL) {
1458 			pr_info("persistence engine module ip_vs_pe_%s "
1459 				"not found\n", u->pe_name);
1460 			ret = -ENOENT;
1461 			goto out;
1462 		}
1463 		old_pe = pe;
1464 	}
1465 
1466 #ifdef CONFIG_IP_VS_IPV6
1467 	if (u->af == AF_INET6) {
1468 		__u32 plen = (__force __u32) u->netmask;
1469 
1470 		if (plen < 1 || plen > 128) {
1471 			ret = -EINVAL;
1472 			goto out;
1473 		}
1474 	}
1475 #endif
1476 
1477 	old_sched = rcu_dereference_protected(svc->scheduler, 1);
1478 	if (sched != old_sched) {
1479 		if (old_sched) {
1480 			ip_vs_unbind_scheduler(svc, old_sched);
1481 			RCU_INIT_POINTER(svc->scheduler, NULL);
1482 			/* Wait all svc->sched_data users */
1483 			synchronize_rcu();
1484 		}
1485 		/* Bind the new scheduler */
1486 		if (sched) {
1487 			ret = ip_vs_bind_scheduler(svc, sched);
1488 			if (ret) {
1489 				ip_vs_scheduler_put(sched);
1490 				goto out;
1491 			}
1492 		}
1493 	}
1494 
1495 	/*
1496 	 * Set the flags and timeout value
1497 	 */
1498 	svc->flags = u->flags | IP_VS_SVC_F_HASHED;
1499 	svc->timeout = u->timeout * HZ;
1500 	svc->netmask = u->netmask;
1501 
1502 	old_pe = rcu_dereference_protected(svc->pe, 1);
1503 	if (pe != old_pe) {
1504 		rcu_assign_pointer(svc->pe, pe);
1505 		/* check for optional methods in new pe */
1506 		new_pe_conn_out = (pe && pe->conn_out) ? true : false;
1507 		old_pe_conn_out = (old_pe && old_pe->conn_out) ? true : false;
1508 		if (new_pe_conn_out && !old_pe_conn_out)
1509 			atomic_inc(&svc->ipvs->conn_out_counter);
1510 		if (old_pe_conn_out && !new_pe_conn_out)
1511 			atomic_dec(&svc->ipvs->conn_out_counter);
1512 	}
1513 
1514 out:
1515 	ip_vs_scheduler_put(old_sched);
1516 	ip_vs_pe_put(old_pe);
1517 	return ret;
1518 }
1519 
1520 /*
1521  *	Delete a service from the service list
1522  *	- The service must be unlinked, unlocked and not referenced!
1523  *	- We are called under _bh lock
1524  */
__ip_vs_del_service(struct ip_vs_service * svc,bool cleanup)1525 static void __ip_vs_del_service(struct ip_vs_service *svc, bool cleanup)
1526 {
1527 	struct ip_vs_dest *dest, *nxt;
1528 	struct ip_vs_scheduler *old_sched;
1529 	struct ip_vs_pe *old_pe;
1530 	struct netns_ipvs *ipvs = svc->ipvs;
1531 
1532 	if (svc->af == AF_INET) {
1533 		ipvs->num_services--;
1534 		if (!ipvs->num_services)
1535 			ip_vs_unregister_hooks(ipvs, svc->af);
1536 	} else if (svc->af == AF_INET6) {
1537 		ipvs->num_services6--;
1538 		if (!ipvs->num_services6)
1539 			ip_vs_unregister_hooks(ipvs, svc->af);
1540 	}
1541 
1542 	ip_vs_stop_estimator(svc->ipvs, &svc->stats);
1543 
1544 	/* Unbind scheduler */
1545 	old_sched = rcu_dereference_protected(svc->scheduler, 1);
1546 	ip_vs_unbind_scheduler(svc, old_sched);
1547 	ip_vs_scheduler_put(old_sched);
1548 
1549 	/* Unbind persistence engine, keep svc->pe */
1550 	old_pe = rcu_dereference_protected(svc->pe, 1);
1551 	if (old_pe && old_pe->conn_out)
1552 		atomic_dec(&ipvs->conn_out_counter);
1553 	ip_vs_pe_put(old_pe);
1554 
1555 	/*
1556 	 *    Unlink the whole destination list
1557 	 */
1558 	list_for_each_entry_safe(dest, nxt, &svc->destinations, n_list) {
1559 		__ip_vs_unlink_dest(svc, dest, 0);
1560 		__ip_vs_del_dest(svc->ipvs, dest, cleanup);
1561 	}
1562 
1563 	/*
1564 	 *    Update the virtual service counters
1565 	 */
1566 	if (svc->port == FTPPORT)
1567 		atomic_dec(&ipvs->ftpsvc_counter);
1568 	else if (svc->port == 0)
1569 		atomic_dec(&ipvs->nullsvc_counter);
1570 
1571 	/*
1572 	 *    Free the service if nobody refers to it
1573 	 */
1574 	__ip_vs_svc_put(svc, true);
1575 
1576 	/* decrease the module use count */
1577 	ip_vs_use_count_dec();
1578 }
1579 
1580 /*
1581  * Unlink a service from list and try to delete it if its refcnt reached 0
1582  */
ip_vs_unlink_service(struct ip_vs_service * svc,bool cleanup)1583 static void ip_vs_unlink_service(struct ip_vs_service *svc, bool cleanup)
1584 {
1585 	ip_vs_unregister_conntrack(svc);
1586 	/* Hold svc to avoid double release from dest_trash */
1587 	atomic_inc(&svc->refcnt);
1588 	/*
1589 	 * Unhash it from the service table
1590 	 */
1591 	ip_vs_svc_unhash(svc);
1592 
1593 	__ip_vs_del_service(svc, cleanup);
1594 }
1595 
1596 /*
1597  *	Delete a service from the service list
1598  */
ip_vs_del_service(struct ip_vs_service * svc)1599 static int ip_vs_del_service(struct ip_vs_service *svc)
1600 {
1601 	if (svc == NULL)
1602 		return -EEXIST;
1603 	ip_vs_unlink_service(svc, false);
1604 
1605 	return 0;
1606 }
1607 
1608 
1609 /*
1610  *	Flush all the virtual services
1611  */
ip_vs_flush(struct netns_ipvs * ipvs,bool cleanup)1612 static int ip_vs_flush(struct netns_ipvs *ipvs, bool cleanup)
1613 {
1614 	int idx;
1615 	struct ip_vs_service *svc;
1616 	struct hlist_node *n;
1617 
1618 	/*
1619 	 * Flush the service table hashed by <netns,protocol,addr,port>
1620 	 */
1621 	for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1622 		hlist_for_each_entry_safe(svc, n, &ip_vs_svc_table[idx],
1623 					  s_list) {
1624 			if (svc->ipvs == ipvs)
1625 				ip_vs_unlink_service(svc, cleanup);
1626 		}
1627 	}
1628 
1629 	/*
1630 	 * Flush the service table hashed by fwmark
1631 	 */
1632 	for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1633 		hlist_for_each_entry_safe(svc, n, &ip_vs_svc_fwm_table[idx],
1634 					  f_list) {
1635 			if (svc->ipvs == ipvs)
1636 				ip_vs_unlink_service(svc, cleanup);
1637 		}
1638 	}
1639 
1640 	return 0;
1641 }
1642 
1643 /*
1644  *	Delete service by {netns} in the service table.
1645  *	Called by __ip_vs_batch_cleanup()
1646  */
ip_vs_service_nets_cleanup(struct list_head * net_list)1647 void ip_vs_service_nets_cleanup(struct list_head *net_list)
1648 {
1649 	struct netns_ipvs *ipvs;
1650 	struct net *net;
1651 
1652 	EnterFunction(2);
1653 	/* Check for "full" addressed entries */
1654 	mutex_lock(&__ip_vs_mutex);
1655 	list_for_each_entry(net, net_list, exit_list) {
1656 		ipvs = net_ipvs(net);
1657 		ip_vs_flush(ipvs, true);
1658 	}
1659 	mutex_unlock(&__ip_vs_mutex);
1660 	LeaveFunction(2);
1661 }
1662 
1663 /* Put all references for device (dst_cache) */
1664 static inline void
ip_vs_forget_dev(struct ip_vs_dest * dest,struct net_device * dev)1665 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
1666 {
1667 	struct ip_vs_dest_dst *dest_dst;
1668 
1669 	spin_lock_bh(&dest->dst_lock);
1670 	dest_dst = rcu_dereference_protected(dest->dest_dst, 1);
1671 	if (dest_dst && dest_dst->dst_cache->dev == dev) {
1672 		IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1673 			      dev->name,
1674 			      IP_VS_DBG_ADDR(dest->af, &dest->addr),
1675 			      ntohs(dest->port),
1676 			      refcount_read(&dest->refcnt));
1677 		__ip_vs_dst_cache_reset(dest);
1678 	}
1679 	spin_unlock_bh(&dest->dst_lock);
1680 
1681 }
1682 /* Netdev event receiver
1683  * Currently only NETDEV_DOWN is handled to release refs to cached dsts
1684  */
ip_vs_dst_event(struct notifier_block * this,unsigned long event,void * ptr)1685 static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
1686 			   void *ptr)
1687 {
1688 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1689 	struct net *net = dev_net(dev);
1690 	struct netns_ipvs *ipvs = net_ipvs(net);
1691 	struct ip_vs_service *svc;
1692 	struct ip_vs_dest *dest;
1693 	unsigned int idx;
1694 
1695 	if (event != NETDEV_DOWN || !ipvs)
1696 		return NOTIFY_DONE;
1697 	IP_VS_DBG(3, "%s() dev=%s\n", __func__, dev->name);
1698 	EnterFunction(2);
1699 	mutex_lock(&__ip_vs_mutex);
1700 	for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1701 		hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1702 			if (svc->ipvs == ipvs) {
1703 				list_for_each_entry(dest, &svc->destinations,
1704 						    n_list) {
1705 					ip_vs_forget_dev(dest, dev);
1706 				}
1707 			}
1708 		}
1709 
1710 		hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1711 			if (svc->ipvs == ipvs) {
1712 				list_for_each_entry(dest, &svc->destinations,
1713 						    n_list) {
1714 					ip_vs_forget_dev(dest, dev);
1715 				}
1716 			}
1717 
1718 		}
1719 	}
1720 
1721 	spin_lock_bh(&ipvs->dest_trash_lock);
1722 	list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
1723 		ip_vs_forget_dev(dest, dev);
1724 	}
1725 	spin_unlock_bh(&ipvs->dest_trash_lock);
1726 	mutex_unlock(&__ip_vs_mutex);
1727 	LeaveFunction(2);
1728 	return NOTIFY_DONE;
1729 }
1730 
1731 /*
1732  *	Zero counters in a service or all services
1733  */
ip_vs_zero_service(struct ip_vs_service * svc)1734 static int ip_vs_zero_service(struct ip_vs_service *svc)
1735 {
1736 	struct ip_vs_dest *dest;
1737 
1738 	list_for_each_entry(dest, &svc->destinations, n_list) {
1739 		ip_vs_zero_stats(&dest->stats);
1740 	}
1741 	ip_vs_zero_stats(&svc->stats);
1742 	return 0;
1743 }
1744 
ip_vs_zero_all(struct netns_ipvs * ipvs)1745 static int ip_vs_zero_all(struct netns_ipvs *ipvs)
1746 {
1747 	int idx;
1748 	struct ip_vs_service *svc;
1749 
1750 	for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1751 		hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
1752 			if (svc->ipvs == ipvs)
1753 				ip_vs_zero_service(svc);
1754 		}
1755 	}
1756 
1757 	for(idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
1758 		hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
1759 			if (svc->ipvs == ipvs)
1760 				ip_vs_zero_service(svc);
1761 		}
1762 	}
1763 
1764 	ip_vs_zero_stats(&ipvs->tot_stats);
1765 	return 0;
1766 }
1767 
1768 #ifdef CONFIG_SYSCTL
1769 
1770 static int
proc_do_defense_mode(struct ctl_table * table,int write,void * buffer,size_t * lenp,loff_t * ppos)1771 proc_do_defense_mode(struct ctl_table *table, int write,
1772 		     void *buffer, size_t *lenp, loff_t *ppos)
1773 {
1774 	struct netns_ipvs *ipvs = table->extra2;
1775 	int *valp = table->data;
1776 	int val = *valp;
1777 	int rc;
1778 
1779 	struct ctl_table tmp = {
1780 		.data = &val,
1781 		.maxlen = sizeof(int),
1782 		.mode = table->mode,
1783 	};
1784 
1785 	rc = proc_dointvec(&tmp, write, buffer, lenp, ppos);
1786 	if (write && (*valp != val)) {
1787 		if (val < 0 || val > 3) {
1788 			rc = -EINVAL;
1789 		} else {
1790 			*valp = val;
1791 			update_defense_level(ipvs);
1792 		}
1793 	}
1794 	return rc;
1795 }
1796 
1797 static int
proc_do_sync_threshold(struct ctl_table * table,int write,void * buffer,size_t * lenp,loff_t * ppos)1798 proc_do_sync_threshold(struct ctl_table *table, int write,
1799 		       void *buffer, size_t *lenp, loff_t *ppos)
1800 {
1801 	int *valp = table->data;
1802 	int val[2];
1803 	int rc;
1804 	struct ctl_table tmp = {
1805 		.data = &val,
1806 		.maxlen = table->maxlen,
1807 		.mode = table->mode,
1808 	};
1809 
1810 	memcpy(val, valp, sizeof(val));
1811 	rc = proc_dointvec(&tmp, write, buffer, lenp, ppos);
1812 	if (write) {
1813 		if (val[0] < 0 || val[1] < 0 ||
1814 		    (val[0] >= val[1] && val[1]))
1815 			rc = -EINVAL;
1816 		else
1817 			memcpy(valp, val, sizeof(val));
1818 	}
1819 	return rc;
1820 }
1821 
1822 static int
proc_do_sync_ports(struct ctl_table * table,int write,void * buffer,size_t * lenp,loff_t * ppos)1823 proc_do_sync_ports(struct ctl_table *table, int write,
1824 		   void *buffer, size_t *lenp, loff_t *ppos)
1825 {
1826 	int *valp = table->data;
1827 	int val = *valp;
1828 	int rc;
1829 
1830 	struct ctl_table tmp = {
1831 		.data = &val,
1832 		.maxlen = sizeof(int),
1833 		.mode = table->mode,
1834 	};
1835 
1836 	rc = proc_dointvec(&tmp, write, buffer, lenp, ppos);
1837 	if (write && (*valp != val)) {
1838 		if (val < 1 || !is_power_of_2(val))
1839 			rc = -EINVAL;
1840 		else
1841 			*valp = val;
1842 	}
1843 	return rc;
1844 }
1845 
1846 /*
1847  *	IPVS sysctl table (under the /proc/sys/net/ipv4/vs/)
1848  *	Do not change order or insert new entries without
1849  *	align with netns init in ip_vs_control_net_init()
1850  */
1851 
1852 static struct ctl_table vs_vars[] = {
1853 	{
1854 		.procname	= "amemthresh",
1855 		.maxlen		= sizeof(int),
1856 		.mode		= 0644,
1857 		.proc_handler	= proc_dointvec,
1858 	},
1859 	{
1860 		.procname	= "am_droprate",
1861 		.maxlen		= sizeof(int),
1862 		.mode		= 0644,
1863 		.proc_handler	= proc_dointvec,
1864 	},
1865 	{
1866 		.procname	= "drop_entry",
1867 		.maxlen		= sizeof(int),
1868 		.mode		= 0644,
1869 		.proc_handler	= proc_do_defense_mode,
1870 	},
1871 	{
1872 		.procname	= "drop_packet",
1873 		.maxlen		= sizeof(int),
1874 		.mode		= 0644,
1875 		.proc_handler	= proc_do_defense_mode,
1876 	},
1877 #ifdef CONFIG_IP_VS_NFCT
1878 	{
1879 		.procname	= "conntrack",
1880 		.maxlen		= sizeof(int),
1881 		.mode		= 0644,
1882 		.proc_handler	= &proc_dointvec,
1883 	},
1884 #endif
1885 	{
1886 		.procname	= "secure_tcp",
1887 		.maxlen		= sizeof(int),
1888 		.mode		= 0644,
1889 		.proc_handler	= proc_do_defense_mode,
1890 	},
1891 	{
1892 		.procname	= "snat_reroute",
1893 		.maxlen		= sizeof(int),
1894 		.mode		= 0644,
1895 		.proc_handler	= &proc_dointvec,
1896 	},
1897 	{
1898 		.procname	= "sync_version",
1899 		.maxlen		= sizeof(int),
1900 		.mode		= 0644,
1901 		.proc_handler	= proc_dointvec_minmax,
1902 		.extra1		= SYSCTL_ZERO,
1903 		.extra2		= SYSCTL_ONE,
1904 	},
1905 	{
1906 		.procname	= "sync_ports",
1907 		.maxlen		= sizeof(int),
1908 		.mode		= 0644,
1909 		.proc_handler	= proc_do_sync_ports,
1910 	},
1911 	{
1912 		.procname	= "sync_persist_mode",
1913 		.maxlen		= sizeof(int),
1914 		.mode		= 0644,
1915 		.proc_handler	= proc_dointvec,
1916 	},
1917 	{
1918 		.procname	= "sync_qlen_max",
1919 		.maxlen		= sizeof(unsigned long),
1920 		.mode		= 0644,
1921 		.proc_handler	= proc_doulongvec_minmax,
1922 	},
1923 	{
1924 		.procname	= "sync_sock_size",
1925 		.maxlen		= sizeof(int),
1926 		.mode		= 0644,
1927 		.proc_handler	= proc_dointvec,
1928 	},
1929 	{
1930 		.procname	= "cache_bypass",
1931 		.maxlen		= sizeof(int),
1932 		.mode		= 0644,
1933 		.proc_handler	= proc_dointvec,
1934 	},
1935 	{
1936 		.procname	= "expire_nodest_conn",
1937 		.maxlen		= sizeof(int),
1938 		.mode		= 0644,
1939 		.proc_handler	= proc_dointvec,
1940 	},
1941 	{
1942 		.procname	= "sloppy_tcp",
1943 		.maxlen		= sizeof(int),
1944 		.mode		= 0644,
1945 		.proc_handler	= proc_dointvec,
1946 	},
1947 	{
1948 		.procname	= "sloppy_sctp",
1949 		.maxlen		= sizeof(int),
1950 		.mode		= 0644,
1951 		.proc_handler	= proc_dointvec,
1952 	},
1953 	{
1954 		.procname	= "expire_quiescent_template",
1955 		.maxlen		= sizeof(int),
1956 		.mode		= 0644,
1957 		.proc_handler	= proc_dointvec,
1958 	},
1959 	{
1960 		.procname	= "sync_threshold",
1961 		.maxlen		=
1962 			sizeof(((struct netns_ipvs *)0)->sysctl_sync_threshold),
1963 		.mode		= 0644,
1964 		.proc_handler	= proc_do_sync_threshold,
1965 	},
1966 	{
1967 		.procname	= "sync_refresh_period",
1968 		.maxlen		= sizeof(int),
1969 		.mode		= 0644,
1970 		.proc_handler	= proc_dointvec_jiffies,
1971 	},
1972 	{
1973 		.procname	= "sync_retries",
1974 		.maxlen		= sizeof(int),
1975 		.mode		= 0644,
1976 		.proc_handler	= proc_dointvec_minmax,
1977 		.extra1		= SYSCTL_ZERO,
1978 		.extra2		= SYSCTL_THREE,
1979 	},
1980 	{
1981 		.procname	= "nat_icmp_send",
1982 		.maxlen		= sizeof(int),
1983 		.mode		= 0644,
1984 		.proc_handler	= proc_dointvec,
1985 	},
1986 	{
1987 		.procname	= "pmtu_disc",
1988 		.maxlen		= sizeof(int),
1989 		.mode		= 0644,
1990 		.proc_handler	= proc_dointvec,
1991 	},
1992 	{
1993 		.procname	= "backup_only",
1994 		.maxlen		= sizeof(int),
1995 		.mode		= 0644,
1996 		.proc_handler	= proc_dointvec,
1997 	},
1998 	{
1999 		.procname	= "conn_reuse_mode",
2000 		.maxlen		= sizeof(int),
2001 		.mode		= 0644,
2002 		.proc_handler	= proc_dointvec,
2003 	},
2004 	{
2005 		.procname	= "schedule_icmp",
2006 		.maxlen		= sizeof(int),
2007 		.mode		= 0644,
2008 		.proc_handler	= proc_dointvec,
2009 	},
2010 	{
2011 		.procname	= "ignore_tunneled",
2012 		.maxlen		= sizeof(int),
2013 		.mode		= 0644,
2014 		.proc_handler	= proc_dointvec,
2015 	},
2016 	{
2017 		.procname	= "run_estimation",
2018 		.maxlen		= sizeof(int),
2019 		.mode		= 0644,
2020 		.proc_handler	= proc_dointvec,
2021 	},
2022 #ifdef CONFIG_IP_VS_DEBUG
2023 	{
2024 		.procname	= "debug_level",
2025 		.data		= &sysctl_ip_vs_debug_level,
2026 		.maxlen		= sizeof(int),
2027 		.mode		= 0644,
2028 		.proc_handler	= proc_dointvec,
2029 	},
2030 #endif
2031 	{ }
2032 };
2033 
2034 #endif
2035 
2036 #ifdef CONFIG_PROC_FS
2037 
2038 struct ip_vs_iter {
2039 	struct seq_net_private p;  /* Do not move this, netns depends upon it*/
2040 	struct hlist_head *table;
2041 	int bucket;
2042 };
2043 
2044 /*
2045  *	Write the contents of the VS rule table to a PROCfs file.
2046  *	(It is kept just for backward compatibility)
2047  */
ip_vs_fwd_name(unsigned int flags)2048 static inline const char *ip_vs_fwd_name(unsigned int flags)
2049 {
2050 	switch (flags & IP_VS_CONN_F_FWD_MASK) {
2051 	case IP_VS_CONN_F_LOCALNODE:
2052 		return "Local";
2053 	case IP_VS_CONN_F_TUNNEL:
2054 		return "Tunnel";
2055 	case IP_VS_CONN_F_DROUTE:
2056 		return "Route";
2057 	default:
2058 		return "Masq";
2059 	}
2060 }
2061 
2062 
2063 /* Get the Nth entry in the two lists */
ip_vs_info_array(struct seq_file * seq,loff_t pos)2064 static struct ip_vs_service *ip_vs_info_array(struct seq_file *seq, loff_t pos)
2065 {
2066 	struct net *net = seq_file_net(seq);
2067 	struct netns_ipvs *ipvs = net_ipvs(net);
2068 	struct ip_vs_iter *iter = seq->private;
2069 	int idx;
2070 	struct ip_vs_service *svc;
2071 
2072 	/* look in hash by protocol */
2073 	for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2074 		hlist_for_each_entry_rcu(svc, &ip_vs_svc_table[idx], s_list) {
2075 			if ((svc->ipvs == ipvs) && pos-- == 0) {
2076 				iter->table = ip_vs_svc_table;
2077 				iter->bucket = idx;
2078 				return svc;
2079 			}
2080 		}
2081 	}
2082 
2083 	/* keep looking in fwmark */
2084 	for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2085 		hlist_for_each_entry_rcu(svc, &ip_vs_svc_fwm_table[idx],
2086 					 f_list) {
2087 			if ((svc->ipvs == ipvs) && pos-- == 0) {
2088 				iter->table = ip_vs_svc_fwm_table;
2089 				iter->bucket = idx;
2090 				return svc;
2091 			}
2092 		}
2093 	}
2094 
2095 	return NULL;
2096 }
2097 
ip_vs_info_seq_start(struct seq_file * seq,loff_t * pos)2098 static void *ip_vs_info_seq_start(struct seq_file *seq, loff_t *pos)
2099 	__acquires(RCU)
2100 {
2101 	rcu_read_lock();
2102 	return *pos ? ip_vs_info_array(seq, *pos - 1) : SEQ_START_TOKEN;
2103 }
2104 
2105 
ip_vs_info_seq_next(struct seq_file * seq,void * v,loff_t * pos)2106 static void *ip_vs_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2107 {
2108 	struct hlist_node *e;
2109 	struct ip_vs_iter *iter;
2110 	struct ip_vs_service *svc;
2111 
2112 	++*pos;
2113 	if (v == SEQ_START_TOKEN)
2114 		return ip_vs_info_array(seq,0);
2115 
2116 	svc = v;
2117 	iter = seq->private;
2118 
2119 	if (iter->table == ip_vs_svc_table) {
2120 		/* next service in table hashed by protocol */
2121 		e = rcu_dereference(hlist_next_rcu(&svc->s_list));
2122 		if (e)
2123 			return hlist_entry(e, struct ip_vs_service, s_list);
2124 
2125 		while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2126 			hlist_for_each_entry_rcu(svc,
2127 						 &ip_vs_svc_table[iter->bucket],
2128 						 s_list) {
2129 				return svc;
2130 			}
2131 		}
2132 
2133 		iter->table = ip_vs_svc_fwm_table;
2134 		iter->bucket = -1;
2135 		goto scan_fwmark;
2136 	}
2137 
2138 	/* next service in hashed by fwmark */
2139 	e = rcu_dereference(hlist_next_rcu(&svc->f_list));
2140 	if (e)
2141 		return hlist_entry(e, struct ip_vs_service, f_list);
2142 
2143  scan_fwmark:
2144 	while (++iter->bucket < IP_VS_SVC_TAB_SIZE) {
2145 		hlist_for_each_entry_rcu(svc,
2146 					 &ip_vs_svc_fwm_table[iter->bucket],
2147 					 f_list)
2148 			return svc;
2149 	}
2150 
2151 	return NULL;
2152 }
2153 
ip_vs_info_seq_stop(struct seq_file * seq,void * v)2154 static void ip_vs_info_seq_stop(struct seq_file *seq, void *v)
2155 	__releases(RCU)
2156 {
2157 	rcu_read_unlock();
2158 }
2159 
2160 
ip_vs_info_seq_show(struct seq_file * seq,void * v)2161 static int ip_vs_info_seq_show(struct seq_file *seq, void *v)
2162 {
2163 	if (v == SEQ_START_TOKEN) {
2164 		seq_printf(seq,
2165 			"IP Virtual Server version %d.%d.%d (size=%d)\n",
2166 			NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2167 		seq_puts(seq,
2168 			 "Prot LocalAddress:Port Scheduler Flags\n");
2169 		seq_puts(seq,
2170 			 "  -> RemoteAddress:Port Forward Weight ActiveConn InActConn\n");
2171 	} else {
2172 		struct net *net = seq_file_net(seq);
2173 		struct netns_ipvs *ipvs = net_ipvs(net);
2174 		const struct ip_vs_service *svc = v;
2175 		const struct ip_vs_iter *iter = seq->private;
2176 		const struct ip_vs_dest *dest;
2177 		struct ip_vs_scheduler *sched = rcu_dereference(svc->scheduler);
2178 		char *sched_name = sched ? sched->name : "none";
2179 
2180 		if (svc->ipvs != ipvs)
2181 			return 0;
2182 		if (iter->table == ip_vs_svc_table) {
2183 #ifdef CONFIG_IP_VS_IPV6
2184 			if (svc->af == AF_INET6)
2185 				seq_printf(seq, "%s  [%pI6]:%04X %s ",
2186 					   ip_vs_proto_name(svc->protocol),
2187 					   &svc->addr.in6,
2188 					   ntohs(svc->port),
2189 					   sched_name);
2190 			else
2191 #endif
2192 				seq_printf(seq, "%s  %08X:%04X %s %s ",
2193 					   ip_vs_proto_name(svc->protocol),
2194 					   ntohl(svc->addr.ip),
2195 					   ntohs(svc->port),
2196 					   sched_name,
2197 					   (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2198 		} else {
2199 			seq_printf(seq, "FWM  %08X %s %s",
2200 				   svc->fwmark, sched_name,
2201 				   (svc->flags & IP_VS_SVC_F_ONEPACKET)?"ops ":"");
2202 		}
2203 
2204 		if (svc->flags & IP_VS_SVC_F_PERSISTENT)
2205 			seq_printf(seq, "persistent %d %08X\n",
2206 				svc->timeout,
2207 				ntohl(svc->netmask));
2208 		else
2209 			seq_putc(seq, '\n');
2210 
2211 		list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
2212 #ifdef CONFIG_IP_VS_IPV6
2213 			if (dest->af == AF_INET6)
2214 				seq_printf(seq,
2215 					   "  -> [%pI6]:%04X"
2216 					   "      %-7s %-6d %-10d %-10d\n",
2217 					   &dest->addr.in6,
2218 					   ntohs(dest->port),
2219 					   ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2220 					   atomic_read(&dest->weight),
2221 					   atomic_read(&dest->activeconns),
2222 					   atomic_read(&dest->inactconns));
2223 			else
2224 #endif
2225 				seq_printf(seq,
2226 					   "  -> %08X:%04X      "
2227 					   "%-7s %-6d %-10d %-10d\n",
2228 					   ntohl(dest->addr.ip),
2229 					   ntohs(dest->port),
2230 					   ip_vs_fwd_name(atomic_read(&dest->conn_flags)),
2231 					   atomic_read(&dest->weight),
2232 					   atomic_read(&dest->activeconns),
2233 					   atomic_read(&dest->inactconns));
2234 
2235 		}
2236 	}
2237 	return 0;
2238 }
2239 
2240 static const struct seq_operations ip_vs_info_seq_ops = {
2241 	.start = ip_vs_info_seq_start,
2242 	.next  = ip_vs_info_seq_next,
2243 	.stop  = ip_vs_info_seq_stop,
2244 	.show  = ip_vs_info_seq_show,
2245 };
2246 
ip_vs_stats_show(struct seq_file * seq,void * v)2247 static int ip_vs_stats_show(struct seq_file *seq, void *v)
2248 {
2249 	struct net *net = seq_file_single_net(seq);
2250 	struct ip_vs_kstats show;
2251 
2252 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
2253 	seq_puts(seq,
2254 		 "   Total Incoming Outgoing         Incoming         Outgoing\n");
2255 	seq_puts(seq,
2256 		 "   Conns  Packets  Packets            Bytes            Bytes\n");
2257 
2258 	ip_vs_copy_stats(&show, &net_ipvs(net)->tot_stats);
2259 	seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n\n",
2260 		   (unsigned long long)show.conns,
2261 		   (unsigned long long)show.inpkts,
2262 		   (unsigned long long)show.outpkts,
2263 		   (unsigned long long)show.inbytes,
2264 		   (unsigned long long)show.outbytes);
2265 
2266 /*                01234567 01234567 01234567 0123456701234567 0123456701234567*/
2267 	seq_puts(seq,
2268 		 " Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
2269 	seq_printf(seq, "%8LX %8LX %8LX %16LX %16LX\n",
2270 		   (unsigned long long)show.cps,
2271 		   (unsigned long long)show.inpps,
2272 		   (unsigned long long)show.outpps,
2273 		   (unsigned long long)show.inbps,
2274 		   (unsigned long long)show.outbps);
2275 
2276 	return 0;
2277 }
2278 
ip_vs_stats_percpu_show(struct seq_file * seq,void * v)2279 static int ip_vs_stats_percpu_show(struct seq_file *seq, void *v)
2280 {
2281 	struct net *net = seq_file_single_net(seq);
2282 	struct ip_vs_stats *tot_stats = &net_ipvs(net)->tot_stats;
2283 	struct ip_vs_cpu_stats __percpu *cpustats = tot_stats->cpustats;
2284 	struct ip_vs_kstats kstats;
2285 	int i;
2286 
2287 /*               01234567 01234567 01234567 0123456701234567 0123456701234567 */
2288 	seq_puts(seq,
2289 		 "       Total Incoming Outgoing         Incoming         Outgoing\n");
2290 	seq_puts(seq,
2291 		 "CPU    Conns  Packets  Packets            Bytes            Bytes\n");
2292 
2293 	for_each_possible_cpu(i) {
2294 		struct ip_vs_cpu_stats *u = per_cpu_ptr(cpustats, i);
2295 		unsigned int start;
2296 		u64 conns, inpkts, outpkts, inbytes, outbytes;
2297 
2298 		do {
2299 			start = u64_stats_fetch_begin_irq(&u->syncp);
2300 			conns = u64_stats_read(&u->cnt.conns);
2301 			inpkts = u64_stats_read(&u->cnt.inpkts);
2302 			outpkts = u64_stats_read(&u->cnt.outpkts);
2303 			inbytes = u64_stats_read(&u->cnt.inbytes);
2304 			outbytes = u64_stats_read(&u->cnt.outbytes);
2305 		} while (u64_stats_fetch_retry_irq(&u->syncp, start));
2306 
2307 		seq_printf(seq, "%3X %8LX %8LX %8LX %16LX %16LX\n",
2308 			   i, (u64)conns, (u64)inpkts,
2309 			   (u64)outpkts, (u64)inbytes,
2310 			   (u64)outbytes);
2311 	}
2312 
2313 	ip_vs_copy_stats(&kstats, tot_stats);
2314 
2315 	seq_printf(seq, "  ~ %8LX %8LX %8LX %16LX %16LX\n\n",
2316 		   (unsigned long long)kstats.conns,
2317 		   (unsigned long long)kstats.inpkts,
2318 		   (unsigned long long)kstats.outpkts,
2319 		   (unsigned long long)kstats.inbytes,
2320 		   (unsigned long long)kstats.outbytes);
2321 
2322 /*                ... 01234567 01234567 01234567 0123456701234567 0123456701234567 */
2323 	seq_puts(seq,
2324 		 "     Conns/s   Pkts/s   Pkts/s          Bytes/s          Bytes/s\n");
2325 	seq_printf(seq, "    %8LX %8LX %8LX %16LX %16LX\n",
2326 		   kstats.cps,
2327 		   kstats.inpps,
2328 		   kstats.outpps,
2329 		   kstats.inbps,
2330 		   kstats.outbps);
2331 
2332 	return 0;
2333 }
2334 #endif
2335 
2336 /*
2337  *	Set timeout values for tcp tcpfin udp in the timeout_table.
2338  */
ip_vs_set_timeout(struct netns_ipvs * ipvs,struct ip_vs_timeout_user * u)2339 static int ip_vs_set_timeout(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2340 {
2341 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2342 	struct ip_vs_proto_data *pd;
2343 #endif
2344 
2345 	IP_VS_DBG(2, "Setting timeout tcp:%d tcpfin:%d udp:%d\n",
2346 		  u->tcp_timeout,
2347 		  u->tcp_fin_timeout,
2348 		  u->udp_timeout);
2349 
2350 #ifdef CONFIG_IP_VS_PROTO_TCP
2351 	if (u->tcp_timeout < 0 || u->tcp_timeout > (INT_MAX / HZ) ||
2352 	    u->tcp_fin_timeout < 0 || u->tcp_fin_timeout > (INT_MAX / HZ)) {
2353 		return -EINVAL;
2354 	}
2355 #endif
2356 
2357 #ifdef CONFIG_IP_VS_PROTO_UDP
2358 	if (u->udp_timeout < 0 || u->udp_timeout > (INT_MAX / HZ))
2359 		return -EINVAL;
2360 #endif
2361 
2362 #ifdef CONFIG_IP_VS_PROTO_TCP
2363 	if (u->tcp_timeout) {
2364 		pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2365 		pd->timeout_table[IP_VS_TCP_S_ESTABLISHED]
2366 			= u->tcp_timeout * HZ;
2367 	}
2368 
2369 	if (u->tcp_fin_timeout) {
2370 		pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2371 		pd->timeout_table[IP_VS_TCP_S_FIN_WAIT]
2372 			= u->tcp_fin_timeout * HZ;
2373 	}
2374 #endif
2375 
2376 #ifdef CONFIG_IP_VS_PROTO_UDP
2377 	if (u->udp_timeout) {
2378 		pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2379 		pd->timeout_table[IP_VS_UDP_S_NORMAL]
2380 			= u->udp_timeout * HZ;
2381 	}
2382 #endif
2383 	return 0;
2384 }
2385 
2386 #define CMDID(cmd)		(cmd - IP_VS_BASE_CTL)
2387 
2388 struct ip_vs_svcdest_user {
2389 	struct ip_vs_service_user	s;
2390 	struct ip_vs_dest_user		d;
2391 };
2392 
2393 static const unsigned char set_arglen[CMDID(IP_VS_SO_SET_MAX) + 1] = {
2394 	[CMDID(IP_VS_SO_SET_ADD)]         = sizeof(struct ip_vs_service_user),
2395 	[CMDID(IP_VS_SO_SET_EDIT)]        = sizeof(struct ip_vs_service_user),
2396 	[CMDID(IP_VS_SO_SET_DEL)]         = sizeof(struct ip_vs_service_user),
2397 	[CMDID(IP_VS_SO_SET_ADDDEST)]     = sizeof(struct ip_vs_svcdest_user),
2398 	[CMDID(IP_VS_SO_SET_DELDEST)]     = sizeof(struct ip_vs_svcdest_user),
2399 	[CMDID(IP_VS_SO_SET_EDITDEST)]    = sizeof(struct ip_vs_svcdest_user),
2400 	[CMDID(IP_VS_SO_SET_TIMEOUT)]     = sizeof(struct ip_vs_timeout_user),
2401 	[CMDID(IP_VS_SO_SET_STARTDAEMON)] = sizeof(struct ip_vs_daemon_user),
2402 	[CMDID(IP_VS_SO_SET_STOPDAEMON)]  = sizeof(struct ip_vs_daemon_user),
2403 	[CMDID(IP_VS_SO_SET_ZERO)]        = sizeof(struct ip_vs_service_user),
2404 };
2405 
2406 union ip_vs_set_arglen {
2407 	struct ip_vs_service_user	field_IP_VS_SO_SET_ADD;
2408 	struct ip_vs_service_user	field_IP_VS_SO_SET_EDIT;
2409 	struct ip_vs_service_user	field_IP_VS_SO_SET_DEL;
2410 	struct ip_vs_svcdest_user	field_IP_VS_SO_SET_ADDDEST;
2411 	struct ip_vs_svcdest_user	field_IP_VS_SO_SET_DELDEST;
2412 	struct ip_vs_svcdest_user	field_IP_VS_SO_SET_EDITDEST;
2413 	struct ip_vs_timeout_user	field_IP_VS_SO_SET_TIMEOUT;
2414 	struct ip_vs_daemon_user	field_IP_VS_SO_SET_STARTDAEMON;
2415 	struct ip_vs_daemon_user	field_IP_VS_SO_SET_STOPDAEMON;
2416 	struct ip_vs_service_user	field_IP_VS_SO_SET_ZERO;
2417 };
2418 
2419 #define MAX_SET_ARGLEN	sizeof(union ip_vs_set_arglen)
2420 
ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern * usvc,struct ip_vs_service_user * usvc_compat)2421 static void ip_vs_copy_usvc_compat(struct ip_vs_service_user_kern *usvc,
2422 				  struct ip_vs_service_user *usvc_compat)
2423 {
2424 	memset(usvc, 0, sizeof(*usvc));
2425 
2426 	usvc->af		= AF_INET;
2427 	usvc->protocol		= usvc_compat->protocol;
2428 	usvc->addr.ip		= usvc_compat->addr;
2429 	usvc->port		= usvc_compat->port;
2430 	usvc->fwmark		= usvc_compat->fwmark;
2431 
2432 	/* Deep copy of sched_name is not needed here */
2433 	usvc->sched_name	= usvc_compat->sched_name;
2434 
2435 	usvc->flags		= usvc_compat->flags;
2436 	usvc->timeout		= usvc_compat->timeout;
2437 	usvc->netmask		= usvc_compat->netmask;
2438 }
2439 
ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern * udest,struct ip_vs_dest_user * udest_compat)2440 static void ip_vs_copy_udest_compat(struct ip_vs_dest_user_kern *udest,
2441 				   struct ip_vs_dest_user *udest_compat)
2442 {
2443 	memset(udest, 0, sizeof(*udest));
2444 
2445 	udest->addr.ip		= udest_compat->addr;
2446 	udest->port		= udest_compat->port;
2447 	udest->conn_flags	= udest_compat->conn_flags;
2448 	udest->weight		= udest_compat->weight;
2449 	udest->u_threshold	= udest_compat->u_threshold;
2450 	udest->l_threshold	= udest_compat->l_threshold;
2451 	udest->af		= AF_INET;
2452 	udest->tun_type		= IP_VS_CONN_F_TUNNEL_TYPE_IPIP;
2453 }
2454 
2455 static int
do_ip_vs_set_ctl(struct sock * sk,int cmd,sockptr_t ptr,unsigned int len)2456 do_ip_vs_set_ctl(struct sock *sk, int cmd, sockptr_t ptr, unsigned int len)
2457 {
2458 	struct net *net = sock_net(sk);
2459 	int ret;
2460 	unsigned char arg[MAX_SET_ARGLEN];
2461 	struct ip_vs_service_user *usvc_compat;
2462 	struct ip_vs_service_user_kern usvc;
2463 	struct ip_vs_service *svc;
2464 	struct ip_vs_dest_user *udest_compat;
2465 	struct ip_vs_dest_user_kern udest;
2466 	struct netns_ipvs *ipvs = net_ipvs(net);
2467 
2468 	BUILD_BUG_ON(sizeof(arg) > 255);
2469 	if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2470 		return -EPERM;
2471 
2472 	if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_SET_MAX)
2473 		return -EINVAL;
2474 	if (len != set_arglen[CMDID(cmd)]) {
2475 		IP_VS_DBG(1, "set_ctl: len %u != %u\n",
2476 			  len, set_arglen[CMDID(cmd)]);
2477 		return -EINVAL;
2478 	}
2479 
2480 	if (copy_from_sockptr(arg, ptr, len) != 0)
2481 		return -EFAULT;
2482 
2483 	/* Handle daemons since they have another lock */
2484 	if (cmd == IP_VS_SO_SET_STARTDAEMON ||
2485 	    cmd == IP_VS_SO_SET_STOPDAEMON) {
2486 		struct ip_vs_daemon_user *dm = (struct ip_vs_daemon_user *)arg;
2487 
2488 		if (cmd == IP_VS_SO_SET_STARTDAEMON) {
2489 			struct ipvs_sync_daemon_cfg cfg;
2490 
2491 			memset(&cfg, 0, sizeof(cfg));
2492 			ret = -EINVAL;
2493 			if (strscpy(cfg.mcast_ifn, dm->mcast_ifn,
2494 				    sizeof(cfg.mcast_ifn)) <= 0)
2495 				return ret;
2496 			cfg.syncid = dm->syncid;
2497 			ret = start_sync_thread(ipvs, &cfg, dm->state);
2498 		} else {
2499 			ret = stop_sync_thread(ipvs, dm->state);
2500 		}
2501 		return ret;
2502 	}
2503 
2504 	mutex_lock(&__ip_vs_mutex);
2505 	if (cmd == IP_VS_SO_SET_FLUSH) {
2506 		/* Flush the virtual service */
2507 		ret = ip_vs_flush(ipvs, false);
2508 		goto out_unlock;
2509 	} else if (cmd == IP_VS_SO_SET_TIMEOUT) {
2510 		/* Set timeout values for (tcp tcpfin udp) */
2511 		ret = ip_vs_set_timeout(ipvs, (struct ip_vs_timeout_user *)arg);
2512 		goto out_unlock;
2513 	} else if (!len) {
2514 		/* No more commands with len == 0 below */
2515 		ret = -EINVAL;
2516 		goto out_unlock;
2517 	}
2518 
2519 	usvc_compat = (struct ip_vs_service_user *)arg;
2520 	udest_compat = (struct ip_vs_dest_user *)(usvc_compat + 1);
2521 
2522 	/* We only use the new structs internally, so copy userspace compat
2523 	 * structs to extended internal versions */
2524 	ip_vs_copy_usvc_compat(&usvc, usvc_compat);
2525 	ip_vs_copy_udest_compat(&udest, udest_compat);
2526 
2527 	if (cmd == IP_VS_SO_SET_ZERO) {
2528 		/* if no service address is set, zero counters in all */
2529 		if (!usvc.fwmark && !usvc.addr.ip && !usvc.port) {
2530 			ret = ip_vs_zero_all(ipvs);
2531 			goto out_unlock;
2532 		}
2533 	}
2534 
2535 	if ((cmd == IP_VS_SO_SET_ADD || cmd == IP_VS_SO_SET_EDIT) &&
2536 	    strnlen(usvc.sched_name, IP_VS_SCHEDNAME_MAXLEN) ==
2537 	    IP_VS_SCHEDNAME_MAXLEN) {
2538 		ret = -EINVAL;
2539 		goto out_unlock;
2540 	}
2541 
2542 	/* Check for valid protocol: TCP or UDP or SCTP, even for fwmark!=0 */
2543 	if (usvc.protocol != IPPROTO_TCP && usvc.protocol != IPPROTO_UDP &&
2544 	    usvc.protocol != IPPROTO_SCTP) {
2545 		pr_err("set_ctl: invalid protocol: %d %pI4:%d\n",
2546 		       usvc.protocol, &usvc.addr.ip,
2547 		       ntohs(usvc.port));
2548 		ret = -EFAULT;
2549 		goto out_unlock;
2550 	}
2551 
2552 	/* Lookup the exact service by <protocol, addr, port> or fwmark */
2553 	rcu_read_lock();
2554 	if (usvc.fwmark == 0)
2555 		svc = __ip_vs_service_find(ipvs, usvc.af, usvc.protocol,
2556 					   &usvc.addr, usvc.port);
2557 	else
2558 		svc = __ip_vs_svc_fwm_find(ipvs, usvc.af, usvc.fwmark);
2559 	rcu_read_unlock();
2560 
2561 	if (cmd != IP_VS_SO_SET_ADD
2562 	    && (svc == NULL || svc->protocol != usvc.protocol)) {
2563 		ret = -ESRCH;
2564 		goto out_unlock;
2565 	}
2566 
2567 	switch (cmd) {
2568 	case IP_VS_SO_SET_ADD:
2569 		if (svc != NULL)
2570 			ret = -EEXIST;
2571 		else
2572 			ret = ip_vs_add_service(ipvs, &usvc, &svc);
2573 		break;
2574 	case IP_VS_SO_SET_EDIT:
2575 		ret = ip_vs_edit_service(svc, &usvc);
2576 		break;
2577 	case IP_VS_SO_SET_DEL:
2578 		ret = ip_vs_del_service(svc);
2579 		if (!ret)
2580 			goto out_unlock;
2581 		break;
2582 	case IP_VS_SO_SET_ZERO:
2583 		ret = ip_vs_zero_service(svc);
2584 		break;
2585 	case IP_VS_SO_SET_ADDDEST:
2586 		ret = ip_vs_add_dest(svc, &udest);
2587 		break;
2588 	case IP_VS_SO_SET_EDITDEST:
2589 		ret = ip_vs_edit_dest(svc, &udest);
2590 		break;
2591 	case IP_VS_SO_SET_DELDEST:
2592 		ret = ip_vs_del_dest(svc, &udest);
2593 	}
2594 
2595   out_unlock:
2596 	mutex_unlock(&__ip_vs_mutex);
2597 	return ret;
2598 }
2599 
2600 
2601 static void
ip_vs_copy_service(struct ip_vs_service_entry * dst,struct ip_vs_service * src)2602 ip_vs_copy_service(struct ip_vs_service_entry *dst, struct ip_vs_service *src)
2603 {
2604 	struct ip_vs_scheduler *sched;
2605 	struct ip_vs_kstats kstats;
2606 	char *sched_name;
2607 
2608 	sched = rcu_dereference_protected(src->scheduler, 1);
2609 	sched_name = sched ? sched->name : "none";
2610 	dst->protocol = src->protocol;
2611 	dst->addr = src->addr.ip;
2612 	dst->port = src->port;
2613 	dst->fwmark = src->fwmark;
2614 	strscpy(dst->sched_name, sched_name, sizeof(dst->sched_name));
2615 	dst->flags = src->flags;
2616 	dst->timeout = src->timeout / HZ;
2617 	dst->netmask = src->netmask;
2618 	dst->num_dests = src->num_dests;
2619 	ip_vs_copy_stats(&kstats, &src->stats);
2620 	ip_vs_export_stats_user(&dst->stats, &kstats);
2621 }
2622 
2623 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)2624 __ip_vs_get_service_entries(struct netns_ipvs *ipvs,
2625 			    const struct ip_vs_get_services *get,
2626 			    struct ip_vs_get_services __user *uptr)
2627 {
2628 	int idx, count=0;
2629 	struct ip_vs_service *svc;
2630 	struct ip_vs_service_entry entry;
2631 	int ret = 0;
2632 
2633 	for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2634 		hlist_for_each_entry(svc, &ip_vs_svc_table[idx], s_list) {
2635 			/* Only expose IPv4 entries to old interface */
2636 			if (svc->af != AF_INET || (svc->ipvs != ipvs))
2637 				continue;
2638 
2639 			if (count >= get->num_services)
2640 				goto out;
2641 			memset(&entry, 0, sizeof(entry));
2642 			ip_vs_copy_service(&entry, svc);
2643 			if (copy_to_user(&uptr->entrytable[count],
2644 					 &entry, sizeof(entry))) {
2645 				ret = -EFAULT;
2646 				goto out;
2647 			}
2648 			count++;
2649 		}
2650 	}
2651 
2652 	for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
2653 		hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[idx], f_list) {
2654 			/* Only expose IPv4 entries to old interface */
2655 			if (svc->af != AF_INET || (svc->ipvs != ipvs))
2656 				continue;
2657 
2658 			if (count >= get->num_services)
2659 				goto out;
2660 			memset(&entry, 0, sizeof(entry));
2661 			ip_vs_copy_service(&entry, svc);
2662 			if (copy_to_user(&uptr->entrytable[count],
2663 					 &entry, sizeof(entry))) {
2664 				ret = -EFAULT;
2665 				goto out;
2666 			}
2667 			count++;
2668 		}
2669 	}
2670 out:
2671 	return ret;
2672 }
2673 
2674 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)2675 __ip_vs_get_dest_entries(struct netns_ipvs *ipvs, const struct ip_vs_get_dests *get,
2676 			 struct ip_vs_get_dests __user *uptr)
2677 {
2678 	struct ip_vs_service *svc;
2679 	union nf_inet_addr addr = { .ip = get->addr };
2680 	int ret = 0;
2681 
2682 	rcu_read_lock();
2683 	if (get->fwmark)
2684 		svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, get->fwmark);
2685 	else
2686 		svc = __ip_vs_service_find(ipvs, AF_INET, get->protocol, &addr,
2687 					   get->port);
2688 	rcu_read_unlock();
2689 
2690 	if (svc) {
2691 		int count = 0;
2692 		struct ip_vs_dest *dest;
2693 		struct ip_vs_dest_entry entry;
2694 		struct ip_vs_kstats kstats;
2695 
2696 		memset(&entry, 0, sizeof(entry));
2697 		list_for_each_entry(dest, &svc->destinations, n_list) {
2698 			if (count >= get->num_dests)
2699 				break;
2700 
2701 			/* Cannot expose heterogeneous members via sockopt
2702 			 * interface
2703 			 */
2704 			if (dest->af != svc->af)
2705 				continue;
2706 
2707 			entry.addr = dest->addr.ip;
2708 			entry.port = dest->port;
2709 			entry.conn_flags = atomic_read(&dest->conn_flags);
2710 			entry.weight = atomic_read(&dest->weight);
2711 			entry.u_threshold = dest->u_threshold;
2712 			entry.l_threshold = dest->l_threshold;
2713 			entry.activeconns = atomic_read(&dest->activeconns);
2714 			entry.inactconns = atomic_read(&dest->inactconns);
2715 			entry.persistconns = atomic_read(&dest->persistconns);
2716 			ip_vs_copy_stats(&kstats, &dest->stats);
2717 			ip_vs_export_stats_user(&entry.stats, &kstats);
2718 			if (copy_to_user(&uptr->entrytable[count],
2719 					 &entry, sizeof(entry))) {
2720 				ret = -EFAULT;
2721 				break;
2722 			}
2723 			count++;
2724 		}
2725 	} else
2726 		ret = -ESRCH;
2727 	return ret;
2728 }
2729 
2730 static inline void
__ip_vs_get_timeouts(struct netns_ipvs * ipvs,struct ip_vs_timeout_user * u)2731 __ip_vs_get_timeouts(struct netns_ipvs *ipvs, struct ip_vs_timeout_user *u)
2732 {
2733 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP)
2734 	struct ip_vs_proto_data *pd;
2735 #endif
2736 
2737 	memset(u, 0, sizeof (*u));
2738 
2739 #ifdef CONFIG_IP_VS_PROTO_TCP
2740 	pd = ip_vs_proto_data_get(ipvs, IPPROTO_TCP);
2741 	u->tcp_timeout = pd->timeout_table[IP_VS_TCP_S_ESTABLISHED] / HZ;
2742 	u->tcp_fin_timeout = pd->timeout_table[IP_VS_TCP_S_FIN_WAIT] / HZ;
2743 #endif
2744 #ifdef CONFIG_IP_VS_PROTO_UDP
2745 	pd = ip_vs_proto_data_get(ipvs, IPPROTO_UDP);
2746 	u->udp_timeout =
2747 			pd->timeout_table[IP_VS_UDP_S_NORMAL] / HZ;
2748 #endif
2749 }
2750 
2751 static const unsigned char get_arglen[CMDID(IP_VS_SO_GET_MAX) + 1] = {
2752 	[CMDID(IP_VS_SO_GET_VERSION)]  = 64,
2753 	[CMDID(IP_VS_SO_GET_INFO)]     = sizeof(struct ip_vs_getinfo),
2754 	[CMDID(IP_VS_SO_GET_SERVICES)] = sizeof(struct ip_vs_get_services),
2755 	[CMDID(IP_VS_SO_GET_SERVICE)]  = sizeof(struct ip_vs_service_entry),
2756 	[CMDID(IP_VS_SO_GET_DESTS)]    = sizeof(struct ip_vs_get_dests),
2757 	[CMDID(IP_VS_SO_GET_TIMEOUT)]  = sizeof(struct ip_vs_timeout_user),
2758 	[CMDID(IP_VS_SO_GET_DAEMON)]   = 2 * sizeof(struct ip_vs_daemon_user),
2759 };
2760 
2761 union ip_vs_get_arglen {
2762 	char				field_IP_VS_SO_GET_VERSION[64];
2763 	struct ip_vs_getinfo		field_IP_VS_SO_GET_INFO;
2764 	struct ip_vs_get_services	field_IP_VS_SO_GET_SERVICES;
2765 	struct ip_vs_service_entry	field_IP_VS_SO_GET_SERVICE;
2766 	struct ip_vs_get_dests		field_IP_VS_SO_GET_DESTS;
2767 	struct ip_vs_timeout_user	field_IP_VS_SO_GET_TIMEOUT;
2768 	struct ip_vs_daemon_user	field_IP_VS_SO_GET_DAEMON[2];
2769 };
2770 
2771 #define MAX_GET_ARGLEN	sizeof(union ip_vs_get_arglen)
2772 
2773 static int
do_ip_vs_get_ctl(struct sock * sk,int cmd,void __user * user,int * len)2774 do_ip_vs_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2775 {
2776 	unsigned char arg[MAX_GET_ARGLEN];
2777 	int ret = 0;
2778 	unsigned int copylen;
2779 	struct net *net = sock_net(sk);
2780 	struct netns_ipvs *ipvs = net_ipvs(net);
2781 
2782 	BUG_ON(!net);
2783 	BUILD_BUG_ON(sizeof(arg) > 255);
2784 	if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2785 		return -EPERM;
2786 
2787 	if (cmd < IP_VS_BASE_CTL || cmd > IP_VS_SO_GET_MAX)
2788 		return -EINVAL;
2789 
2790 	copylen = get_arglen[CMDID(cmd)];
2791 	if (*len < (int) copylen) {
2792 		IP_VS_DBG(1, "get_ctl: len %d < %u\n", *len, copylen);
2793 		return -EINVAL;
2794 	}
2795 
2796 	if (copy_from_user(arg, user, copylen) != 0)
2797 		return -EFAULT;
2798 	/*
2799 	 * Handle daemons first since it has its own locking
2800 	 */
2801 	if (cmd == IP_VS_SO_GET_DAEMON) {
2802 		struct ip_vs_daemon_user d[2];
2803 
2804 		memset(&d, 0, sizeof(d));
2805 		mutex_lock(&ipvs->sync_mutex);
2806 		if (ipvs->sync_state & IP_VS_STATE_MASTER) {
2807 			d[0].state = IP_VS_STATE_MASTER;
2808 			strscpy(d[0].mcast_ifn, ipvs->mcfg.mcast_ifn,
2809 				sizeof(d[0].mcast_ifn));
2810 			d[0].syncid = ipvs->mcfg.syncid;
2811 		}
2812 		if (ipvs->sync_state & IP_VS_STATE_BACKUP) {
2813 			d[1].state = IP_VS_STATE_BACKUP;
2814 			strscpy(d[1].mcast_ifn, ipvs->bcfg.mcast_ifn,
2815 				sizeof(d[1].mcast_ifn));
2816 			d[1].syncid = ipvs->bcfg.syncid;
2817 		}
2818 		if (copy_to_user(user, &d, sizeof(d)) != 0)
2819 			ret = -EFAULT;
2820 		mutex_unlock(&ipvs->sync_mutex);
2821 		return ret;
2822 	}
2823 
2824 	mutex_lock(&__ip_vs_mutex);
2825 	switch (cmd) {
2826 	case IP_VS_SO_GET_VERSION:
2827 	{
2828 		char buf[64];
2829 
2830 		sprintf(buf, "IP Virtual Server version %d.%d.%d (size=%d)",
2831 			NVERSION(IP_VS_VERSION_CODE), ip_vs_conn_tab_size);
2832 		if (copy_to_user(user, buf, strlen(buf)+1) != 0) {
2833 			ret = -EFAULT;
2834 			goto out;
2835 		}
2836 		*len = strlen(buf)+1;
2837 	}
2838 	break;
2839 
2840 	case IP_VS_SO_GET_INFO:
2841 	{
2842 		struct ip_vs_getinfo info;
2843 		info.version = IP_VS_VERSION_CODE;
2844 		info.size = ip_vs_conn_tab_size;
2845 		info.num_services = ipvs->num_services;
2846 		if (copy_to_user(user, &info, sizeof(info)) != 0)
2847 			ret = -EFAULT;
2848 	}
2849 	break;
2850 
2851 	case IP_VS_SO_GET_SERVICES:
2852 	{
2853 		struct ip_vs_get_services *get;
2854 		int size;
2855 
2856 		get = (struct ip_vs_get_services *)arg;
2857 		size = struct_size(get, entrytable, get->num_services);
2858 		if (*len != size) {
2859 			pr_err("length: %u != %u\n", *len, size);
2860 			ret = -EINVAL;
2861 			goto out;
2862 		}
2863 		ret = __ip_vs_get_service_entries(ipvs, get, user);
2864 	}
2865 	break;
2866 
2867 	case IP_VS_SO_GET_SERVICE:
2868 	{
2869 		struct ip_vs_service_entry *entry;
2870 		struct ip_vs_service *svc;
2871 		union nf_inet_addr addr;
2872 
2873 		entry = (struct ip_vs_service_entry *)arg;
2874 		addr.ip = entry->addr;
2875 		rcu_read_lock();
2876 		if (entry->fwmark)
2877 			svc = __ip_vs_svc_fwm_find(ipvs, AF_INET, entry->fwmark);
2878 		else
2879 			svc = __ip_vs_service_find(ipvs, AF_INET,
2880 						   entry->protocol, &addr,
2881 						   entry->port);
2882 		rcu_read_unlock();
2883 		if (svc) {
2884 			ip_vs_copy_service(entry, svc);
2885 			if (copy_to_user(user, entry, sizeof(*entry)) != 0)
2886 				ret = -EFAULT;
2887 		} else
2888 			ret = -ESRCH;
2889 	}
2890 	break;
2891 
2892 	case IP_VS_SO_GET_DESTS:
2893 	{
2894 		struct ip_vs_get_dests *get;
2895 		int size;
2896 
2897 		get = (struct ip_vs_get_dests *)arg;
2898 		size = struct_size(get, entrytable, get->num_dests);
2899 		if (*len != size) {
2900 			pr_err("length: %u != %u\n", *len, size);
2901 			ret = -EINVAL;
2902 			goto out;
2903 		}
2904 		ret = __ip_vs_get_dest_entries(ipvs, get, user);
2905 	}
2906 	break;
2907 
2908 	case IP_VS_SO_GET_TIMEOUT:
2909 	{
2910 		struct ip_vs_timeout_user t;
2911 
2912 		__ip_vs_get_timeouts(ipvs, &t);
2913 		if (copy_to_user(user, &t, sizeof(t)) != 0)
2914 			ret = -EFAULT;
2915 	}
2916 	break;
2917 
2918 	default:
2919 		ret = -EINVAL;
2920 	}
2921 
2922 out:
2923 	mutex_unlock(&__ip_vs_mutex);
2924 	return ret;
2925 }
2926 
2927 
2928 static struct nf_sockopt_ops ip_vs_sockopts = {
2929 	.pf		= PF_INET,
2930 	.set_optmin	= IP_VS_BASE_CTL,
2931 	.set_optmax	= IP_VS_SO_SET_MAX+1,
2932 	.set		= do_ip_vs_set_ctl,
2933 	.get_optmin	= IP_VS_BASE_CTL,
2934 	.get_optmax	= IP_VS_SO_GET_MAX+1,
2935 	.get		= do_ip_vs_get_ctl,
2936 	.owner		= THIS_MODULE,
2937 };
2938 
2939 /*
2940  * Generic Netlink interface
2941  */
2942 
2943 /* IPVS genetlink family */
2944 static struct genl_family ip_vs_genl_family;
2945 
2946 /* Policy used for first-level command attributes */
2947 static const struct nla_policy ip_vs_cmd_policy[IPVS_CMD_ATTR_MAX + 1] = {
2948 	[IPVS_CMD_ATTR_SERVICE]		= { .type = NLA_NESTED },
2949 	[IPVS_CMD_ATTR_DEST]		= { .type = NLA_NESTED },
2950 	[IPVS_CMD_ATTR_DAEMON]		= { .type = NLA_NESTED },
2951 	[IPVS_CMD_ATTR_TIMEOUT_TCP]	= { .type = NLA_U32 },
2952 	[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]	= { .type = NLA_U32 },
2953 	[IPVS_CMD_ATTR_TIMEOUT_UDP]	= { .type = NLA_U32 },
2954 };
2955 
2956 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DAEMON */
2957 static const struct nla_policy ip_vs_daemon_policy[IPVS_DAEMON_ATTR_MAX + 1] = {
2958 	[IPVS_DAEMON_ATTR_STATE]	= { .type = NLA_U32 },
2959 	[IPVS_DAEMON_ATTR_MCAST_IFN]	= { .type = NLA_NUL_STRING,
2960 					    .len = IP_VS_IFNAME_MAXLEN - 1 },
2961 	[IPVS_DAEMON_ATTR_SYNC_ID]	= { .type = NLA_U32 },
2962 	[IPVS_DAEMON_ATTR_SYNC_MAXLEN]	= { .type = NLA_U16 },
2963 	[IPVS_DAEMON_ATTR_MCAST_GROUP]	= { .type = NLA_U32 },
2964 	[IPVS_DAEMON_ATTR_MCAST_GROUP6]	= { .len = sizeof(struct in6_addr) },
2965 	[IPVS_DAEMON_ATTR_MCAST_PORT]	= { .type = NLA_U16 },
2966 	[IPVS_DAEMON_ATTR_MCAST_TTL]	= { .type = NLA_U8 },
2967 };
2968 
2969 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_SERVICE */
2970 static const struct nla_policy ip_vs_svc_policy[IPVS_SVC_ATTR_MAX + 1] = {
2971 	[IPVS_SVC_ATTR_AF]		= { .type = NLA_U16 },
2972 	[IPVS_SVC_ATTR_PROTOCOL]	= { .type = NLA_U16 },
2973 	[IPVS_SVC_ATTR_ADDR]		= { .type = NLA_BINARY,
2974 					    .len = sizeof(union nf_inet_addr) },
2975 	[IPVS_SVC_ATTR_PORT]		= { .type = NLA_U16 },
2976 	[IPVS_SVC_ATTR_FWMARK]		= { .type = NLA_U32 },
2977 	[IPVS_SVC_ATTR_SCHED_NAME]	= { .type = NLA_NUL_STRING,
2978 					    .len = IP_VS_SCHEDNAME_MAXLEN - 1 },
2979 	[IPVS_SVC_ATTR_PE_NAME]		= { .type = NLA_NUL_STRING,
2980 					    .len = IP_VS_PENAME_MAXLEN },
2981 	[IPVS_SVC_ATTR_FLAGS]		= { .type = NLA_BINARY,
2982 					    .len = sizeof(struct ip_vs_flags) },
2983 	[IPVS_SVC_ATTR_TIMEOUT]		= { .type = NLA_U32 },
2984 	[IPVS_SVC_ATTR_NETMASK]		= { .type = NLA_U32 },
2985 	[IPVS_SVC_ATTR_STATS]		= { .type = NLA_NESTED },
2986 };
2987 
2988 /* Policy used for attributes in nested attribute IPVS_CMD_ATTR_DEST */
2989 static const struct nla_policy ip_vs_dest_policy[IPVS_DEST_ATTR_MAX + 1] = {
2990 	[IPVS_DEST_ATTR_ADDR]		= { .type = NLA_BINARY,
2991 					    .len = sizeof(union nf_inet_addr) },
2992 	[IPVS_DEST_ATTR_PORT]		= { .type = NLA_U16 },
2993 	[IPVS_DEST_ATTR_FWD_METHOD]	= { .type = NLA_U32 },
2994 	[IPVS_DEST_ATTR_WEIGHT]		= { .type = NLA_U32 },
2995 	[IPVS_DEST_ATTR_U_THRESH]	= { .type = NLA_U32 },
2996 	[IPVS_DEST_ATTR_L_THRESH]	= { .type = NLA_U32 },
2997 	[IPVS_DEST_ATTR_ACTIVE_CONNS]	= { .type = NLA_U32 },
2998 	[IPVS_DEST_ATTR_INACT_CONNS]	= { .type = NLA_U32 },
2999 	[IPVS_DEST_ATTR_PERSIST_CONNS]	= { .type = NLA_U32 },
3000 	[IPVS_DEST_ATTR_STATS]		= { .type = NLA_NESTED },
3001 	[IPVS_DEST_ATTR_ADDR_FAMILY]	= { .type = NLA_U16 },
3002 	[IPVS_DEST_ATTR_TUN_TYPE]	= { .type = NLA_U8 },
3003 	[IPVS_DEST_ATTR_TUN_PORT]	= { .type = NLA_U16 },
3004 	[IPVS_DEST_ATTR_TUN_FLAGS]	= { .type = NLA_U16 },
3005 };
3006 
ip_vs_genl_fill_stats(struct sk_buff * skb,int container_type,struct ip_vs_kstats * kstats)3007 static int ip_vs_genl_fill_stats(struct sk_buff *skb, int container_type,
3008 				 struct ip_vs_kstats *kstats)
3009 {
3010 	struct nlattr *nl_stats = nla_nest_start_noflag(skb, container_type);
3011 
3012 	if (!nl_stats)
3013 		return -EMSGSIZE;
3014 
3015 	if (nla_put_u32(skb, IPVS_STATS_ATTR_CONNS, (u32)kstats->conns) ||
3016 	    nla_put_u32(skb, IPVS_STATS_ATTR_INPKTS, (u32)kstats->inpkts) ||
3017 	    nla_put_u32(skb, IPVS_STATS_ATTR_OUTPKTS, (u32)kstats->outpkts) ||
3018 	    nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes,
3019 			      IPVS_STATS_ATTR_PAD) ||
3020 	    nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes,
3021 			      IPVS_STATS_ATTR_PAD) ||
3022 	    nla_put_u32(skb, IPVS_STATS_ATTR_CPS, (u32)kstats->cps) ||
3023 	    nla_put_u32(skb, IPVS_STATS_ATTR_INPPS, (u32)kstats->inpps) ||
3024 	    nla_put_u32(skb, IPVS_STATS_ATTR_OUTPPS, (u32)kstats->outpps) ||
3025 	    nla_put_u32(skb, IPVS_STATS_ATTR_INBPS, (u32)kstats->inbps) ||
3026 	    nla_put_u32(skb, IPVS_STATS_ATTR_OUTBPS, (u32)kstats->outbps))
3027 		goto nla_put_failure;
3028 	nla_nest_end(skb, nl_stats);
3029 
3030 	return 0;
3031 
3032 nla_put_failure:
3033 	nla_nest_cancel(skb, nl_stats);
3034 	return -EMSGSIZE;
3035 }
3036 
ip_vs_genl_fill_stats64(struct sk_buff * skb,int container_type,struct ip_vs_kstats * kstats)3037 static int ip_vs_genl_fill_stats64(struct sk_buff *skb, int container_type,
3038 				   struct ip_vs_kstats *kstats)
3039 {
3040 	struct nlattr *nl_stats = nla_nest_start_noflag(skb, container_type);
3041 
3042 	if (!nl_stats)
3043 		return -EMSGSIZE;
3044 
3045 	if (nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CONNS, kstats->conns,
3046 			      IPVS_STATS_ATTR_PAD) ||
3047 	    nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPKTS, kstats->inpkts,
3048 			      IPVS_STATS_ATTR_PAD) ||
3049 	    nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPKTS, kstats->outpkts,
3050 			      IPVS_STATS_ATTR_PAD) ||
3051 	    nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBYTES, kstats->inbytes,
3052 			      IPVS_STATS_ATTR_PAD) ||
3053 	    nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBYTES, kstats->outbytes,
3054 			      IPVS_STATS_ATTR_PAD) ||
3055 	    nla_put_u64_64bit(skb, IPVS_STATS_ATTR_CPS, kstats->cps,
3056 			      IPVS_STATS_ATTR_PAD) ||
3057 	    nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INPPS, kstats->inpps,
3058 			      IPVS_STATS_ATTR_PAD) ||
3059 	    nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTPPS, kstats->outpps,
3060 			      IPVS_STATS_ATTR_PAD) ||
3061 	    nla_put_u64_64bit(skb, IPVS_STATS_ATTR_INBPS, kstats->inbps,
3062 			      IPVS_STATS_ATTR_PAD) ||
3063 	    nla_put_u64_64bit(skb, IPVS_STATS_ATTR_OUTBPS, kstats->outbps,
3064 			      IPVS_STATS_ATTR_PAD))
3065 		goto nla_put_failure;
3066 	nla_nest_end(skb, nl_stats);
3067 
3068 	return 0;
3069 
3070 nla_put_failure:
3071 	nla_nest_cancel(skb, nl_stats);
3072 	return -EMSGSIZE;
3073 }
3074 
ip_vs_genl_fill_service(struct sk_buff * skb,struct ip_vs_service * svc)3075 static int ip_vs_genl_fill_service(struct sk_buff *skb,
3076 				   struct ip_vs_service *svc)
3077 {
3078 	struct ip_vs_scheduler *sched;
3079 	struct ip_vs_pe *pe;
3080 	struct nlattr *nl_service;
3081 	struct ip_vs_flags flags = { .flags = svc->flags,
3082 				     .mask = ~0 };
3083 	struct ip_vs_kstats kstats;
3084 	char *sched_name;
3085 
3086 	nl_service = nla_nest_start_noflag(skb, IPVS_CMD_ATTR_SERVICE);
3087 	if (!nl_service)
3088 		return -EMSGSIZE;
3089 
3090 	if (nla_put_u16(skb, IPVS_SVC_ATTR_AF, svc->af))
3091 		goto nla_put_failure;
3092 	if (svc->fwmark) {
3093 		if (nla_put_u32(skb, IPVS_SVC_ATTR_FWMARK, svc->fwmark))
3094 			goto nla_put_failure;
3095 	} else {
3096 		if (nla_put_u16(skb, IPVS_SVC_ATTR_PROTOCOL, svc->protocol) ||
3097 		    nla_put(skb, IPVS_SVC_ATTR_ADDR, sizeof(svc->addr), &svc->addr) ||
3098 		    nla_put_be16(skb, IPVS_SVC_ATTR_PORT, svc->port))
3099 			goto nla_put_failure;
3100 	}
3101 
3102 	sched = rcu_dereference_protected(svc->scheduler, 1);
3103 	sched_name = sched ? sched->name : "none";
3104 	pe = rcu_dereference_protected(svc->pe, 1);
3105 	if (nla_put_string(skb, IPVS_SVC_ATTR_SCHED_NAME, sched_name) ||
3106 	    (pe && nla_put_string(skb, IPVS_SVC_ATTR_PE_NAME, pe->name)) ||
3107 	    nla_put(skb, IPVS_SVC_ATTR_FLAGS, sizeof(flags), &flags) ||
3108 	    nla_put_u32(skb, IPVS_SVC_ATTR_TIMEOUT, svc->timeout / HZ) ||
3109 	    nla_put_be32(skb, IPVS_SVC_ATTR_NETMASK, svc->netmask))
3110 		goto nla_put_failure;
3111 	ip_vs_copy_stats(&kstats, &svc->stats);
3112 	if (ip_vs_genl_fill_stats(skb, IPVS_SVC_ATTR_STATS, &kstats))
3113 		goto nla_put_failure;
3114 	if (ip_vs_genl_fill_stats64(skb, IPVS_SVC_ATTR_STATS64, &kstats))
3115 		goto nla_put_failure;
3116 
3117 	nla_nest_end(skb, nl_service);
3118 
3119 	return 0;
3120 
3121 nla_put_failure:
3122 	nla_nest_cancel(skb, nl_service);
3123 	return -EMSGSIZE;
3124 }
3125 
ip_vs_genl_dump_service(struct sk_buff * skb,struct ip_vs_service * svc,struct netlink_callback * cb)3126 static int ip_vs_genl_dump_service(struct sk_buff *skb,
3127 				   struct ip_vs_service *svc,
3128 				   struct netlink_callback *cb)
3129 {
3130 	void *hdr;
3131 
3132 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3133 			  &ip_vs_genl_family, NLM_F_MULTI,
3134 			  IPVS_CMD_NEW_SERVICE);
3135 	if (!hdr)
3136 		return -EMSGSIZE;
3137 
3138 	if (ip_vs_genl_fill_service(skb, svc) < 0)
3139 		goto nla_put_failure;
3140 
3141 	genlmsg_end(skb, hdr);
3142 	return 0;
3143 
3144 nla_put_failure:
3145 	genlmsg_cancel(skb, hdr);
3146 	return -EMSGSIZE;
3147 }
3148 
ip_vs_genl_dump_services(struct sk_buff * skb,struct netlink_callback * cb)3149 static int ip_vs_genl_dump_services(struct sk_buff *skb,
3150 				    struct netlink_callback *cb)
3151 {
3152 	int idx = 0, i;
3153 	int start = cb->args[0];
3154 	struct ip_vs_service *svc;
3155 	struct net *net = sock_net(skb->sk);
3156 	struct netns_ipvs *ipvs = net_ipvs(net);
3157 
3158 	mutex_lock(&__ip_vs_mutex);
3159 	for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3160 		hlist_for_each_entry(svc, &ip_vs_svc_table[i], s_list) {
3161 			if (++idx <= start || (svc->ipvs != ipvs))
3162 				continue;
3163 			if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3164 				idx--;
3165 				goto nla_put_failure;
3166 			}
3167 		}
3168 	}
3169 
3170 	for (i = 0; i < IP_VS_SVC_TAB_SIZE; i++) {
3171 		hlist_for_each_entry(svc, &ip_vs_svc_fwm_table[i], f_list) {
3172 			if (++idx <= start || (svc->ipvs != ipvs))
3173 				continue;
3174 			if (ip_vs_genl_dump_service(skb, svc, cb) < 0) {
3175 				idx--;
3176 				goto nla_put_failure;
3177 			}
3178 		}
3179 	}
3180 
3181 nla_put_failure:
3182 	mutex_unlock(&__ip_vs_mutex);
3183 	cb->args[0] = idx;
3184 
3185 	return skb->len;
3186 }
3187 
ip_vs_is_af_valid(int af)3188 static bool ip_vs_is_af_valid(int af)
3189 {
3190 	if (af == AF_INET)
3191 		return true;
3192 #ifdef CONFIG_IP_VS_IPV6
3193 	if (af == AF_INET6 && ipv6_mod_enabled())
3194 		return true;
3195 #endif
3196 	return false;
3197 }
3198 
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)3199 static int ip_vs_genl_parse_service(struct netns_ipvs *ipvs,
3200 				    struct ip_vs_service_user_kern *usvc,
3201 				    struct nlattr *nla, bool full_entry,
3202 				    struct ip_vs_service **ret_svc)
3203 {
3204 	struct nlattr *attrs[IPVS_SVC_ATTR_MAX + 1];
3205 	struct nlattr *nla_af, *nla_port, *nla_fwmark, *nla_protocol, *nla_addr;
3206 	struct ip_vs_service *svc;
3207 
3208 	/* Parse mandatory identifying service fields first */
3209 	if (nla == NULL ||
3210 	    nla_parse_nested_deprecated(attrs, IPVS_SVC_ATTR_MAX, nla, ip_vs_svc_policy, NULL))
3211 		return -EINVAL;
3212 
3213 	nla_af		= attrs[IPVS_SVC_ATTR_AF];
3214 	nla_protocol	= attrs[IPVS_SVC_ATTR_PROTOCOL];
3215 	nla_addr	= attrs[IPVS_SVC_ATTR_ADDR];
3216 	nla_port	= attrs[IPVS_SVC_ATTR_PORT];
3217 	nla_fwmark	= attrs[IPVS_SVC_ATTR_FWMARK];
3218 
3219 	if (!(nla_af && (nla_fwmark || (nla_port && nla_protocol && nla_addr))))
3220 		return -EINVAL;
3221 
3222 	memset(usvc, 0, sizeof(*usvc));
3223 
3224 	usvc->af = nla_get_u16(nla_af);
3225 	if (!ip_vs_is_af_valid(usvc->af))
3226 		return -EAFNOSUPPORT;
3227 
3228 	if (nla_fwmark) {
3229 		usvc->protocol = IPPROTO_TCP;
3230 		usvc->fwmark = nla_get_u32(nla_fwmark);
3231 	} else {
3232 		usvc->protocol = nla_get_u16(nla_protocol);
3233 		nla_memcpy(&usvc->addr, nla_addr, sizeof(usvc->addr));
3234 		usvc->port = nla_get_be16(nla_port);
3235 		usvc->fwmark = 0;
3236 	}
3237 
3238 	rcu_read_lock();
3239 	if (usvc->fwmark)
3240 		svc = __ip_vs_svc_fwm_find(ipvs, usvc->af, usvc->fwmark);
3241 	else
3242 		svc = __ip_vs_service_find(ipvs, usvc->af, usvc->protocol,
3243 					   &usvc->addr, usvc->port);
3244 	rcu_read_unlock();
3245 	*ret_svc = svc;
3246 
3247 	/* If a full entry was requested, check for the additional fields */
3248 	if (full_entry) {
3249 		struct nlattr *nla_sched, *nla_flags, *nla_pe, *nla_timeout,
3250 			      *nla_netmask;
3251 		struct ip_vs_flags flags;
3252 
3253 		nla_sched = attrs[IPVS_SVC_ATTR_SCHED_NAME];
3254 		nla_pe = attrs[IPVS_SVC_ATTR_PE_NAME];
3255 		nla_flags = attrs[IPVS_SVC_ATTR_FLAGS];
3256 		nla_timeout = attrs[IPVS_SVC_ATTR_TIMEOUT];
3257 		nla_netmask = attrs[IPVS_SVC_ATTR_NETMASK];
3258 
3259 		if (!(nla_sched && nla_flags && nla_timeout && nla_netmask))
3260 			return -EINVAL;
3261 
3262 		nla_memcpy(&flags, nla_flags, sizeof(flags));
3263 
3264 		/* prefill flags from service if it already exists */
3265 		if (svc)
3266 			usvc->flags = svc->flags;
3267 
3268 		/* set new flags from userland */
3269 		usvc->flags = (usvc->flags & ~flags.mask) |
3270 			      (flags.flags & flags.mask);
3271 		usvc->sched_name = nla_data(nla_sched);
3272 		usvc->pe_name = nla_pe ? nla_data(nla_pe) : NULL;
3273 		usvc->timeout = nla_get_u32(nla_timeout);
3274 		usvc->netmask = nla_get_be32(nla_netmask);
3275 	}
3276 
3277 	return 0;
3278 }
3279 
ip_vs_genl_find_service(struct netns_ipvs * ipvs,struct nlattr * nla)3280 static struct ip_vs_service *ip_vs_genl_find_service(struct netns_ipvs *ipvs,
3281 						     struct nlattr *nla)
3282 {
3283 	struct ip_vs_service_user_kern usvc;
3284 	struct ip_vs_service *svc;
3285 	int ret;
3286 
3287 	ret = ip_vs_genl_parse_service(ipvs, &usvc, nla, false, &svc);
3288 	return ret ? ERR_PTR(ret) : svc;
3289 }
3290 
ip_vs_genl_fill_dest(struct sk_buff * skb,struct ip_vs_dest * dest)3291 static int ip_vs_genl_fill_dest(struct sk_buff *skb, struct ip_vs_dest *dest)
3292 {
3293 	struct nlattr *nl_dest;
3294 	struct ip_vs_kstats kstats;
3295 
3296 	nl_dest = nla_nest_start_noflag(skb, IPVS_CMD_ATTR_DEST);
3297 	if (!nl_dest)
3298 		return -EMSGSIZE;
3299 
3300 	if (nla_put(skb, IPVS_DEST_ATTR_ADDR, sizeof(dest->addr), &dest->addr) ||
3301 	    nla_put_be16(skb, IPVS_DEST_ATTR_PORT, dest->port) ||
3302 	    nla_put_u32(skb, IPVS_DEST_ATTR_FWD_METHOD,
3303 			(atomic_read(&dest->conn_flags) &
3304 			 IP_VS_CONN_F_FWD_MASK)) ||
3305 	    nla_put_u32(skb, IPVS_DEST_ATTR_WEIGHT,
3306 			atomic_read(&dest->weight)) ||
3307 	    nla_put_u8(skb, IPVS_DEST_ATTR_TUN_TYPE,
3308 		       dest->tun_type) ||
3309 	    nla_put_be16(skb, IPVS_DEST_ATTR_TUN_PORT,
3310 			 dest->tun_port) ||
3311 	    nla_put_u16(skb, IPVS_DEST_ATTR_TUN_FLAGS,
3312 			dest->tun_flags) ||
3313 	    nla_put_u32(skb, IPVS_DEST_ATTR_U_THRESH, dest->u_threshold) ||
3314 	    nla_put_u32(skb, IPVS_DEST_ATTR_L_THRESH, dest->l_threshold) ||
3315 	    nla_put_u32(skb, IPVS_DEST_ATTR_ACTIVE_CONNS,
3316 			atomic_read(&dest->activeconns)) ||
3317 	    nla_put_u32(skb, IPVS_DEST_ATTR_INACT_CONNS,
3318 			atomic_read(&dest->inactconns)) ||
3319 	    nla_put_u32(skb, IPVS_DEST_ATTR_PERSIST_CONNS,
3320 			atomic_read(&dest->persistconns)) ||
3321 	    nla_put_u16(skb, IPVS_DEST_ATTR_ADDR_FAMILY, dest->af))
3322 		goto nla_put_failure;
3323 	ip_vs_copy_stats(&kstats, &dest->stats);
3324 	if (ip_vs_genl_fill_stats(skb, IPVS_DEST_ATTR_STATS, &kstats))
3325 		goto nla_put_failure;
3326 	if (ip_vs_genl_fill_stats64(skb, IPVS_DEST_ATTR_STATS64, &kstats))
3327 		goto nla_put_failure;
3328 
3329 	nla_nest_end(skb, nl_dest);
3330 
3331 	return 0;
3332 
3333 nla_put_failure:
3334 	nla_nest_cancel(skb, nl_dest);
3335 	return -EMSGSIZE;
3336 }
3337 
ip_vs_genl_dump_dest(struct sk_buff * skb,struct ip_vs_dest * dest,struct netlink_callback * cb)3338 static int ip_vs_genl_dump_dest(struct sk_buff *skb, struct ip_vs_dest *dest,
3339 				struct netlink_callback *cb)
3340 {
3341 	void *hdr;
3342 
3343 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3344 			  &ip_vs_genl_family, NLM_F_MULTI,
3345 			  IPVS_CMD_NEW_DEST);
3346 	if (!hdr)
3347 		return -EMSGSIZE;
3348 
3349 	if (ip_vs_genl_fill_dest(skb, dest) < 0)
3350 		goto nla_put_failure;
3351 
3352 	genlmsg_end(skb, hdr);
3353 	return 0;
3354 
3355 nla_put_failure:
3356 	genlmsg_cancel(skb, hdr);
3357 	return -EMSGSIZE;
3358 }
3359 
ip_vs_genl_dump_dests(struct sk_buff * skb,struct netlink_callback * cb)3360 static int ip_vs_genl_dump_dests(struct sk_buff *skb,
3361 				 struct netlink_callback *cb)
3362 {
3363 	int idx = 0;
3364 	int start = cb->args[0];
3365 	struct ip_vs_service *svc;
3366 	struct ip_vs_dest *dest;
3367 	struct nlattr *attrs[IPVS_CMD_ATTR_MAX + 1];
3368 	struct net *net = sock_net(skb->sk);
3369 	struct netns_ipvs *ipvs = net_ipvs(net);
3370 
3371 	mutex_lock(&__ip_vs_mutex);
3372 
3373 	/* Try to find the service for which to dump destinations */
3374 	if (nlmsg_parse_deprecated(cb->nlh, GENL_HDRLEN, attrs, IPVS_CMD_ATTR_MAX, ip_vs_cmd_policy, cb->extack))
3375 		goto out_err;
3376 
3377 
3378 	svc = ip_vs_genl_find_service(ipvs, attrs[IPVS_CMD_ATTR_SERVICE]);
3379 	if (IS_ERR_OR_NULL(svc))
3380 		goto out_err;
3381 
3382 	/* Dump the destinations */
3383 	list_for_each_entry(dest, &svc->destinations, n_list) {
3384 		if (++idx <= start)
3385 			continue;
3386 		if (ip_vs_genl_dump_dest(skb, dest, cb) < 0) {
3387 			idx--;
3388 			goto nla_put_failure;
3389 		}
3390 	}
3391 
3392 nla_put_failure:
3393 	cb->args[0] = idx;
3394 
3395 out_err:
3396 	mutex_unlock(&__ip_vs_mutex);
3397 
3398 	return skb->len;
3399 }
3400 
ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern * udest,struct nlattr * nla,bool full_entry)3401 static int ip_vs_genl_parse_dest(struct ip_vs_dest_user_kern *udest,
3402 				 struct nlattr *nla, bool full_entry)
3403 {
3404 	struct nlattr *attrs[IPVS_DEST_ATTR_MAX + 1];
3405 	struct nlattr *nla_addr, *nla_port;
3406 	struct nlattr *nla_addr_family;
3407 
3408 	/* Parse mandatory identifying destination fields first */
3409 	if (nla == NULL ||
3410 	    nla_parse_nested_deprecated(attrs, IPVS_DEST_ATTR_MAX, nla, ip_vs_dest_policy, NULL))
3411 		return -EINVAL;
3412 
3413 	nla_addr	= attrs[IPVS_DEST_ATTR_ADDR];
3414 	nla_port	= attrs[IPVS_DEST_ATTR_PORT];
3415 	nla_addr_family	= attrs[IPVS_DEST_ATTR_ADDR_FAMILY];
3416 
3417 	if (!(nla_addr && nla_port))
3418 		return -EINVAL;
3419 
3420 	memset(udest, 0, sizeof(*udest));
3421 
3422 	nla_memcpy(&udest->addr, nla_addr, sizeof(udest->addr));
3423 	udest->port = nla_get_be16(nla_port);
3424 
3425 	if (nla_addr_family)
3426 		udest->af = nla_get_u16(nla_addr_family);
3427 	else
3428 		udest->af = 0;
3429 
3430 	/* If a full entry was requested, check for the additional fields */
3431 	if (full_entry) {
3432 		struct nlattr *nla_fwd, *nla_weight, *nla_u_thresh,
3433 			      *nla_l_thresh, *nla_tun_type, *nla_tun_port,
3434 			      *nla_tun_flags;
3435 
3436 		nla_fwd		= attrs[IPVS_DEST_ATTR_FWD_METHOD];
3437 		nla_weight	= attrs[IPVS_DEST_ATTR_WEIGHT];
3438 		nla_u_thresh	= attrs[IPVS_DEST_ATTR_U_THRESH];
3439 		nla_l_thresh	= attrs[IPVS_DEST_ATTR_L_THRESH];
3440 		nla_tun_type	= attrs[IPVS_DEST_ATTR_TUN_TYPE];
3441 		nla_tun_port	= attrs[IPVS_DEST_ATTR_TUN_PORT];
3442 		nla_tun_flags	= attrs[IPVS_DEST_ATTR_TUN_FLAGS];
3443 
3444 		if (!(nla_fwd && nla_weight && nla_u_thresh && nla_l_thresh))
3445 			return -EINVAL;
3446 
3447 		udest->conn_flags = nla_get_u32(nla_fwd)
3448 				    & IP_VS_CONN_F_FWD_MASK;
3449 		udest->weight = nla_get_u32(nla_weight);
3450 		udest->u_threshold = nla_get_u32(nla_u_thresh);
3451 		udest->l_threshold = nla_get_u32(nla_l_thresh);
3452 
3453 		if (nla_tun_type)
3454 			udest->tun_type = nla_get_u8(nla_tun_type);
3455 
3456 		if (nla_tun_port)
3457 			udest->tun_port = nla_get_be16(nla_tun_port);
3458 
3459 		if (nla_tun_flags)
3460 			udest->tun_flags = nla_get_u16(nla_tun_flags);
3461 	}
3462 
3463 	return 0;
3464 }
3465 
ip_vs_genl_fill_daemon(struct sk_buff * skb,__u32 state,struct ipvs_sync_daemon_cfg * c)3466 static int ip_vs_genl_fill_daemon(struct sk_buff *skb, __u32 state,
3467 				  struct ipvs_sync_daemon_cfg *c)
3468 {
3469 	struct nlattr *nl_daemon;
3470 
3471 	nl_daemon = nla_nest_start_noflag(skb, IPVS_CMD_ATTR_DAEMON);
3472 	if (!nl_daemon)
3473 		return -EMSGSIZE;
3474 
3475 	if (nla_put_u32(skb, IPVS_DAEMON_ATTR_STATE, state) ||
3476 	    nla_put_string(skb, IPVS_DAEMON_ATTR_MCAST_IFN, c->mcast_ifn) ||
3477 	    nla_put_u32(skb, IPVS_DAEMON_ATTR_SYNC_ID, c->syncid) ||
3478 	    nla_put_u16(skb, IPVS_DAEMON_ATTR_SYNC_MAXLEN, c->sync_maxlen) ||
3479 	    nla_put_u16(skb, IPVS_DAEMON_ATTR_MCAST_PORT, c->mcast_port) ||
3480 	    nla_put_u8(skb, IPVS_DAEMON_ATTR_MCAST_TTL, c->mcast_ttl))
3481 		goto nla_put_failure;
3482 #ifdef CONFIG_IP_VS_IPV6
3483 	if (c->mcast_af == AF_INET6) {
3484 		if (nla_put_in6_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP6,
3485 				     &c->mcast_group.in6))
3486 			goto nla_put_failure;
3487 	} else
3488 #endif
3489 		if (c->mcast_af == AF_INET &&
3490 		    nla_put_in_addr(skb, IPVS_DAEMON_ATTR_MCAST_GROUP,
3491 				    c->mcast_group.ip))
3492 			goto nla_put_failure;
3493 	nla_nest_end(skb, nl_daemon);
3494 
3495 	return 0;
3496 
3497 nla_put_failure:
3498 	nla_nest_cancel(skb, nl_daemon);
3499 	return -EMSGSIZE;
3500 }
3501 
ip_vs_genl_dump_daemon(struct sk_buff * skb,__u32 state,struct ipvs_sync_daemon_cfg * c,struct netlink_callback * cb)3502 static int ip_vs_genl_dump_daemon(struct sk_buff *skb, __u32 state,
3503 				  struct ipvs_sync_daemon_cfg *c,
3504 				  struct netlink_callback *cb)
3505 {
3506 	void *hdr;
3507 	hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3508 			  &ip_vs_genl_family, NLM_F_MULTI,
3509 			  IPVS_CMD_NEW_DAEMON);
3510 	if (!hdr)
3511 		return -EMSGSIZE;
3512 
3513 	if (ip_vs_genl_fill_daemon(skb, state, c))
3514 		goto nla_put_failure;
3515 
3516 	genlmsg_end(skb, hdr);
3517 	return 0;
3518 
3519 nla_put_failure:
3520 	genlmsg_cancel(skb, hdr);
3521 	return -EMSGSIZE;
3522 }
3523 
ip_vs_genl_dump_daemons(struct sk_buff * skb,struct netlink_callback * cb)3524 static int ip_vs_genl_dump_daemons(struct sk_buff *skb,
3525 				   struct netlink_callback *cb)
3526 {
3527 	struct net *net = sock_net(skb->sk);
3528 	struct netns_ipvs *ipvs = net_ipvs(net);
3529 
3530 	mutex_lock(&ipvs->sync_mutex);
3531 	if ((ipvs->sync_state & IP_VS_STATE_MASTER) && !cb->args[0]) {
3532 		if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_MASTER,
3533 					   &ipvs->mcfg, cb) < 0)
3534 			goto nla_put_failure;
3535 
3536 		cb->args[0] = 1;
3537 	}
3538 
3539 	if ((ipvs->sync_state & IP_VS_STATE_BACKUP) && !cb->args[1]) {
3540 		if (ip_vs_genl_dump_daemon(skb, IP_VS_STATE_BACKUP,
3541 					   &ipvs->bcfg, cb) < 0)
3542 			goto nla_put_failure;
3543 
3544 		cb->args[1] = 1;
3545 	}
3546 
3547 nla_put_failure:
3548 	mutex_unlock(&ipvs->sync_mutex);
3549 
3550 	return skb->len;
3551 }
3552 
ip_vs_genl_new_daemon(struct netns_ipvs * ipvs,struct nlattr ** attrs)3553 static int ip_vs_genl_new_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3554 {
3555 	struct ipvs_sync_daemon_cfg c;
3556 	struct nlattr *a;
3557 	int ret;
3558 
3559 	memset(&c, 0, sizeof(c));
3560 	if (!(attrs[IPVS_DAEMON_ATTR_STATE] &&
3561 	      attrs[IPVS_DAEMON_ATTR_MCAST_IFN] &&
3562 	      attrs[IPVS_DAEMON_ATTR_SYNC_ID]))
3563 		return -EINVAL;
3564 	strscpy(c.mcast_ifn, nla_data(attrs[IPVS_DAEMON_ATTR_MCAST_IFN]),
3565 		sizeof(c.mcast_ifn));
3566 	c.syncid = nla_get_u32(attrs[IPVS_DAEMON_ATTR_SYNC_ID]);
3567 
3568 	a = attrs[IPVS_DAEMON_ATTR_SYNC_MAXLEN];
3569 	if (a)
3570 		c.sync_maxlen = nla_get_u16(a);
3571 
3572 	a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP];
3573 	if (a) {
3574 		c.mcast_af = AF_INET;
3575 		c.mcast_group.ip = nla_get_in_addr(a);
3576 		if (!ipv4_is_multicast(c.mcast_group.ip))
3577 			return -EINVAL;
3578 	} else {
3579 		a = attrs[IPVS_DAEMON_ATTR_MCAST_GROUP6];
3580 		if (a) {
3581 #ifdef CONFIG_IP_VS_IPV6
3582 			int addr_type;
3583 
3584 			c.mcast_af = AF_INET6;
3585 			c.mcast_group.in6 = nla_get_in6_addr(a);
3586 			addr_type = ipv6_addr_type(&c.mcast_group.in6);
3587 			if (!(addr_type & IPV6_ADDR_MULTICAST))
3588 				return -EINVAL;
3589 #else
3590 			return -EAFNOSUPPORT;
3591 #endif
3592 		}
3593 	}
3594 
3595 	a = attrs[IPVS_DAEMON_ATTR_MCAST_PORT];
3596 	if (a)
3597 		c.mcast_port = nla_get_u16(a);
3598 
3599 	a = attrs[IPVS_DAEMON_ATTR_MCAST_TTL];
3600 	if (a)
3601 		c.mcast_ttl = nla_get_u8(a);
3602 
3603 	/* The synchronization protocol is incompatible with mixed family
3604 	 * services
3605 	 */
3606 	if (ipvs->mixed_address_family_dests > 0)
3607 		return -EINVAL;
3608 
3609 	ret = start_sync_thread(ipvs, &c,
3610 				nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3611 	return ret;
3612 }
3613 
ip_vs_genl_del_daemon(struct netns_ipvs * ipvs,struct nlattr ** attrs)3614 static int ip_vs_genl_del_daemon(struct netns_ipvs *ipvs, struct nlattr **attrs)
3615 {
3616 	int ret;
3617 
3618 	if (!attrs[IPVS_DAEMON_ATTR_STATE])
3619 		return -EINVAL;
3620 
3621 	ret = stop_sync_thread(ipvs,
3622 			       nla_get_u32(attrs[IPVS_DAEMON_ATTR_STATE]));
3623 	return ret;
3624 }
3625 
ip_vs_genl_set_config(struct netns_ipvs * ipvs,struct nlattr ** attrs)3626 static int ip_vs_genl_set_config(struct netns_ipvs *ipvs, struct nlattr **attrs)
3627 {
3628 	struct ip_vs_timeout_user t;
3629 
3630 	__ip_vs_get_timeouts(ipvs, &t);
3631 
3632 	if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP])
3633 		t.tcp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP]);
3634 
3635 	if (attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN])
3636 		t.tcp_fin_timeout =
3637 			nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_TCP_FIN]);
3638 
3639 	if (attrs[IPVS_CMD_ATTR_TIMEOUT_UDP])
3640 		t.udp_timeout = nla_get_u32(attrs[IPVS_CMD_ATTR_TIMEOUT_UDP]);
3641 
3642 	return ip_vs_set_timeout(ipvs, &t);
3643 }
3644 
ip_vs_genl_set_daemon(struct sk_buff * skb,struct genl_info * info)3645 static int ip_vs_genl_set_daemon(struct sk_buff *skb, struct genl_info *info)
3646 {
3647 	int ret = -EINVAL, cmd;
3648 	struct net *net = sock_net(skb->sk);
3649 	struct netns_ipvs *ipvs = net_ipvs(net);
3650 
3651 	cmd = info->genlhdr->cmd;
3652 
3653 	if (cmd == IPVS_CMD_NEW_DAEMON || cmd == IPVS_CMD_DEL_DAEMON) {
3654 		struct nlattr *daemon_attrs[IPVS_DAEMON_ATTR_MAX + 1];
3655 
3656 		if (!info->attrs[IPVS_CMD_ATTR_DAEMON] ||
3657 		    nla_parse_nested_deprecated(daemon_attrs, IPVS_DAEMON_ATTR_MAX, info->attrs[IPVS_CMD_ATTR_DAEMON], ip_vs_daemon_policy, info->extack))
3658 			goto out;
3659 
3660 		if (cmd == IPVS_CMD_NEW_DAEMON)
3661 			ret = ip_vs_genl_new_daemon(ipvs, daemon_attrs);
3662 		else
3663 			ret = ip_vs_genl_del_daemon(ipvs, daemon_attrs);
3664 	}
3665 
3666 out:
3667 	return ret;
3668 }
3669 
ip_vs_genl_set_cmd(struct sk_buff * skb,struct genl_info * info)3670 static int ip_vs_genl_set_cmd(struct sk_buff *skb, struct genl_info *info)
3671 {
3672 	bool need_full_svc = false, need_full_dest = false;
3673 	struct ip_vs_service *svc = NULL;
3674 	struct ip_vs_service_user_kern usvc;
3675 	struct ip_vs_dest_user_kern udest;
3676 	int ret = 0, cmd;
3677 	struct net *net = sock_net(skb->sk);
3678 	struct netns_ipvs *ipvs = net_ipvs(net);
3679 
3680 	cmd = info->genlhdr->cmd;
3681 
3682 	mutex_lock(&__ip_vs_mutex);
3683 
3684 	if (cmd == IPVS_CMD_FLUSH) {
3685 		ret = ip_vs_flush(ipvs, false);
3686 		goto out;
3687 	} else if (cmd == IPVS_CMD_SET_CONFIG) {
3688 		ret = ip_vs_genl_set_config(ipvs, info->attrs);
3689 		goto out;
3690 	} else if (cmd == IPVS_CMD_ZERO &&
3691 		   !info->attrs[IPVS_CMD_ATTR_SERVICE]) {
3692 		ret = ip_vs_zero_all(ipvs);
3693 		goto out;
3694 	}
3695 
3696 	/* All following commands require a service argument, so check if we
3697 	 * received a valid one. We need a full service specification when
3698 	 * adding / editing a service. Only identifying members otherwise. */
3699 	if (cmd == IPVS_CMD_NEW_SERVICE || cmd == IPVS_CMD_SET_SERVICE)
3700 		need_full_svc = true;
3701 
3702 	ret = ip_vs_genl_parse_service(ipvs, &usvc,
3703 				       info->attrs[IPVS_CMD_ATTR_SERVICE],
3704 				       need_full_svc, &svc);
3705 	if (ret)
3706 		goto out;
3707 
3708 	/* Unless we're adding a new service, the service must already exist */
3709 	if ((cmd != IPVS_CMD_NEW_SERVICE) && (svc == NULL)) {
3710 		ret = -ESRCH;
3711 		goto out;
3712 	}
3713 
3714 	/* Destination commands require a valid destination argument. For
3715 	 * adding / editing a destination, we need a full destination
3716 	 * specification. */
3717 	if (cmd == IPVS_CMD_NEW_DEST || cmd == IPVS_CMD_SET_DEST ||
3718 	    cmd == IPVS_CMD_DEL_DEST) {
3719 		if (cmd != IPVS_CMD_DEL_DEST)
3720 			need_full_dest = true;
3721 
3722 		ret = ip_vs_genl_parse_dest(&udest,
3723 					    info->attrs[IPVS_CMD_ATTR_DEST],
3724 					    need_full_dest);
3725 		if (ret)
3726 			goto out;
3727 
3728 		/* Old protocols did not allow the user to specify address
3729 		 * family, so we set it to zero instead.  We also didn't
3730 		 * allow heterogeneous pools in the old code, so it's safe
3731 		 * to assume that this will have the same address family as
3732 		 * the service.
3733 		 */
3734 		if (udest.af == 0)
3735 			udest.af = svc->af;
3736 
3737 		if (!ip_vs_is_af_valid(udest.af)) {
3738 			ret = -EAFNOSUPPORT;
3739 			goto out;
3740 		}
3741 
3742 		if (udest.af != svc->af && cmd != IPVS_CMD_DEL_DEST) {
3743 			/* The synchronization protocol is incompatible
3744 			 * with mixed family services
3745 			 */
3746 			if (ipvs->sync_state) {
3747 				ret = -EINVAL;
3748 				goto out;
3749 			}
3750 
3751 			/* Which connection types do we support? */
3752 			switch (udest.conn_flags) {
3753 			case IP_VS_CONN_F_TUNNEL:
3754 				/* We are able to forward this */
3755 				break;
3756 			default:
3757 				ret = -EINVAL;
3758 				goto out;
3759 			}
3760 		}
3761 	}
3762 
3763 	switch (cmd) {
3764 	case IPVS_CMD_NEW_SERVICE:
3765 		if (svc == NULL)
3766 			ret = ip_vs_add_service(ipvs, &usvc, &svc);
3767 		else
3768 			ret = -EEXIST;
3769 		break;
3770 	case IPVS_CMD_SET_SERVICE:
3771 		ret = ip_vs_edit_service(svc, &usvc);
3772 		break;
3773 	case IPVS_CMD_DEL_SERVICE:
3774 		ret = ip_vs_del_service(svc);
3775 		/* do not use svc, it can be freed */
3776 		break;
3777 	case IPVS_CMD_NEW_DEST:
3778 		ret = ip_vs_add_dest(svc, &udest);
3779 		break;
3780 	case IPVS_CMD_SET_DEST:
3781 		ret = ip_vs_edit_dest(svc, &udest);
3782 		break;
3783 	case IPVS_CMD_DEL_DEST:
3784 		ret = ip_vs_del_dest(svc, &udest);
3785 		break;
3786 	case IPVS_CMD_ZERO:
3787 		ret = ip_vs_zero_service(svc);
3788 		break;
3789 	default:
3790 		ret = -EINVAL;
3791 	}
3792 
3793 out:
3794 	mutex_unlock(&__ip_vs_mutex);
3795 
3796 	return ret;
3797 }
3798 
ip_vs_genl_get_cmd(struct sk_buff * skb,struct genl_info * info)3799 static int ip_vs_genl_get_cmd(struct sk_buff *skb, struct genl_info *info)
3800 {
3801 	struct sk_buff *msg;
3802 	void *reply;
3803 	int ret, cmd, reply_cmd;
3804 	struct net *net = sock_net(skb->sk);
3805 	struct netns_ipvs *ipvs = net_ipvs(net);
3806 
3807 	cmd = info->genlhdr->cmd;
3808 
3809 	if (cmd == IPVS_CMD_GET_SERVICE)
3810 		reply_cmd = IPVS_CMD_NEW_SERVICE;
3811 	else if (cmd == IPVS_CMD_GET_INFO)
3812 		reply_cmd = IPVS_CMD_SET_INFO;
3813 	else if (cmd == IPVS_CMD_GET_CONFIG)
3814 		reply_cmd = IPVS_CMD_SET_CONFIG;
3815 	else {
3816 		pr_err("unknown Generic Netlink command\n");
3817 		return -EINVAL;
3818 	}
3819 
3820 	msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
3821 	if (!msg)
3822 		return -ENOMEM;
3823 
3824 	mutex_lock(&__ip_vs_mutex);
3825 
3826 	reply = genlmsg_put_reply(msg, info, &ip_vs_genl_family, 0, reply_cmd);
3827 	if (reply == NULL)
3828 		goto nla_put_failure;
3829 
3830 	switch (cmd) {
3831 	case IPVS_CMD_GET_SERVICE:
3832 	{
3833 		struct ip_vs_service *svc;
3834 
3835 		svc = ip_vs_genl_find_service(ipvs,
3836 					      info->attrs[IPVS_CMD_ATTR_SERVICE]);
3837 		if (IS_ERR(svc)) {
3838 			ret = PTR_ERR(svc);
3839 			goto out_err;
3840 		} else if (svc) {
3841 			ret = ip_vs_genl_fill_service(msg, svc);
3842 			if (ret)
3843 				goto nla_put_failure;
3844 		} else {
3845 			ret = -ESRCH;
3846 			goto out_err;
3847 		}
3848 
3849 		break;
3850 	}
3851 
3852 	case IPVS_CMD_GET_CONFIG:
3853 	{
3854 		struct ip_vs_timeout_user t;
3855 
3856 		__ip_vs_get_timeouts(ipvs, &t);
3857 #ifdef CONFIG_IP_VS_PROTO_TCP
3858 		if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP,
3859 				t.tcp_timeout) ||
3860 		    nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_TCP_FIN,
3861 				t.tcp_fin_timeout))
3862 			goto nla_put_failure;
3863 #endif
3864 #ifdef CONFIG_IP_VS_PROTO_UDP
3865 		if (nla_put_u32(msg, IPVS_CMD_ATTR_TIMEOUT_UDP, t.udp_timeout))
3866 			goto nla_put_failure;
3867 #endif
3868 
3869 		break;
3870 	}
3871 
3872 	case IPVS_CMD_GET_INFO:
3873 		if (nla_put_u32(msg, IPVS_INFO_ATTR_VERSION,
3874 				IP_VS_VERSION_CODE) ||
3875 		    nla_put_u32(msg, IPVS_INFO_ATTR_CONN_TAB_SIZE,
3876 				ip_vs_conn_tab_size))
3877 			goto nla_put_failure;
3878 		break;
3879 	}
3880 
3881 	genlmsg_end(msg, reply);
3882 	ret = genlmsg_reply(msg, info);
3883 	goto out;
3884 
3885 nla_put_failure:
3886 	pr_err("not enough space in Netlink message\n");
3887 	ret = -EMSGSIZE;
3888 
3889 out_err:
3890 	nlmsg_free(msg);
3891 out:
3892 	mutex_unlock(&__ip_vs_mutex);
3893 
3894 	return ret;
3895 }
3896 
3897 
3898 static const struct genl_small_ops ip_vs_genl_ops[] = {
3899 	{
3900 		.cmd	= IPVS_CMD_NEW_SERVICE,
3901 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3902 		.flags	= GENL_ADMIN_PERM,
3903 		.doit	= ip_vs_genl_set_cmd,
3904 	},
3905 	{
3906 		.cmd	= IPVS_CMD_SET_SERVICE,
3907 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3908 		.flags	= GENL_ADMIN_PERM,
3909 		.doit	= ip_vs_genl_set_cmd,
3910 	},
3911 	{
3912 		.cmd	= IPVS_CMD_DEL_SERVICE,
3913 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3914 		.flags	= GENL_ADMIN_PERM,
3915 		.doit	= ip_vs_genl_set_cmd,
3916 	},
3917 	{
3918 		.cmd	= IPVS_CMD_GET_SERVICE,
3919 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3920 		.flags	= GENL_ADMIN_PERM,
3921 		.doit	= ip_vs_genl_get_cmd,
3922 		.dumpit	= ip_vs_genl_dump_services,
3923 	},
3924 	{
3925 		.cmd	= IPVS_CMD_NEW_DEST,
3926 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3927 		.flags	= GENL_ADMIN_PERM,
3928 		.doit	= ip_vs_genl_set_cmd,
3929 	},
3930 	{
3931 		.cmd	= IPVS_CMD_SET_DEST,
3932 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3933 		.flags	= GENL_ADMIN_PERM,
3934 		.doit	= ip_vs_genl_set_cmd,
3935 	},
3936 	{
3937 		.cmd	= IPVS_CMD_DEL_DEST,
3938 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3939 		.flags	= GENL_ADMIN_PERM,
3940 		.doit	= ip_vs_genl_set_cmd,
3941 	},
3942 	{
3943 		.cmd	= IPVS_CMD_GET_DEST,
3944 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3945 		.flags	= GENL_ADMIN_PERM,
3946 		.dumpit	= ip_vs_genl_dump_dests,
3947 	},
3948 	{
3949 		.cmd	= IPVS_CMD_NEW_DAEMON,
3950 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3951 		.flags	= GENL_ADMIN_PERM,
3952 		.doit	= ip_vs_genl_set_daemon,
3953 	},
3954 	{
3955 		.cmd	= IPVS_CMD_DEL_DAEMON,
3956 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3957 		.flags	= GENL_ADMIN_PERM,
3958 		.doit	= ip_vs_genl_set_daemon,
3959 	},
3960 	{
3961 		.cmd	= IPVS_CMD_GET_DAEMON,
3962 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3963 		.flags	= GENL_ADMIN_PERM,
3964 		.dumpit	= ip_vs_genl_dump_daemons,
3965 	},
3966 	{
3967 		.cmd	= IPVS_CMD_SET_CONFIG,
3968 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3969 		.flags	= GENL_ADMIN_PERM,
3970 		.doit	= ip_vs_genl_set_cmd,
3971 	},
3972 	{
3973 		.cmd	= IPVS_CMD_GET_CONFIG,
3974 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3975 		.flags	= GENL_ADMIN_PERM,
3976 		.doit	= ip_vs_genl_get_cmd,
3977 	},
3978 	{
3979 		.cmd	= IPVS_CMD_GET_INFO,
3980 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3981 		.flags	= GENL_ADMIN_PERM,
3982 		.doit	= ip_vs_genl_get_cmd,
3983 	},
3984 	{
3985 		.cmd	= IPVS_CMD_ZERO,
3986 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3987 		.flags	= GENL_ADMIN_PERM,
3988 		.doit	= ip_vs_genl_set_cmd,
3989 	},
3990 	{
3991 		.cmd	= IPVS_CMD_FLUSH,
3992 		.validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
3993 		.flags	= GENL_ADMIN_PERM,
3994 		.doit	= ip_vs_genl_set_cmd,
3995 	},
3996 };
3997 
3998 static struct genl_family ip_vs_genl_family __ro_after_init = {
3999 	.hdrsize	= 0,
4000 	.name		= IPVS_GENL_NAME,
4001 	.version	= IPVS_GENL_VERSION,
4002 	.maxattr	= IPVS_CMD_ATTR_MAX,
4003 	.policy = ip_vs_cmd_policy,
4004 	.netnsok        = true,         /* Make ipvsadm to work on netns */
4005 	.module		= THIS_MODULE,
4006 	.small_ops	= ip_vs_genl_ops,
4007 	.n_small_ops	= ARRAY_SIZE(ip_vs_genl_ops),
4008 	.resv_start_op	= IPVS_CMD_FLUSH + 1,
4009 };
4010 
ip_vs_genl_register(void)4011 static int __init ip_vs_genl_register(void)
4012 {
4013 	return genl_register_family(&ip_vs_genl_family);
4014 }
4015 
ip_vs_genl_unregister(void)4016 static void ip_vs_genl_unregister(void)
4017 {
4018 	genl_unregister_family(&ip_vs_genl_family);
4019 }
4020 
4021 /* End of Generic Netlink interface definitions */
4022 
4023 /*
4024  * per netns intit/exit func.
4025  */
4026 #ifdef CONFIG_SYSCTL
ip_vs_control_net_init_sysctl(struct netns_ipvs * ipvs)4027 static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs)
4028 {
4029 	struct net *net = ipvs->net;
4030 	int idx;
4031 	struct ctl_table *tbl;
4032 
4033 	atomic_set(&ipvs->dropentry, 0);
4034 	spin_lock_init(&ipvs->dropentry_lock);
4035 	spin_lock_init(&ipvs->droppacket_lock);
4036 	spin_lock_init(&ipvs->securetcp_lock);
4037 
4038 	if (!net_eq(net, &init_net)) {
4039 		tbl = kmemdup(vs_vars, sizeof(vs_vars), GFP_KERNEL);
4040 		if (tbl == NULL)
4041 			return -ENOMEM;
4042 
4043 		/* Don't export sysctls to unprivileged users */
4044 		if (net->user_ns != &init_user_ns)
4045 			tbl[0].procname = NULL;
4046 	} else
4047 		tbl = vs_vars;
4048 	/* Initialize sysctl defaults */
4049 	for (idx = 0; idx < ARRAY_SIZE(vs_vars); idx++) {
4050 		if (tbl[idx].proc_handler == proc_do_defense_mode)
4051 			tbl[idx].extra2 = ipvs;
4052 	}
4053 	idx = 0;
4054 	ipvs->sysctl_amemthresh = 1024;
4055 	tbl[idx++].data = &ipvs->sysctl_amemthresh;
4056 	ipvs->sysctl_am_droprate = 10;
4057 	tbl[idx++].data = &ipvs->sysctl_am_droprate;
4058 	tbl[idx++].data = &ipvs->sysctl_drop_entry;
4059 	tbl[idx++].data = &ipvs->sysctl_drop_packet;
4060 #ifdef CONFIG_IP_VS_NFCT
4061 	tbl[idx++].data = &ipvs->sysctl_conntrack;
4062 #endif
4063 	tbl[idx++].data = &ipvs->sysctl_secure_tcp;
4064 	ipvs->sysctl_snat_reroute = 1;
4065 	tbl[idx++].data = &ipvs->sysctl_snat_reroute;
4066 	ipvs->sysctl_sync_ver = 1;
4067 	tbl[idx++].data = &ipvs->sysctl_sync_ver;
4068 	ipvs->sysctl_sync_ports = 1;
4069 	tbl[idx++].data = &ipvs->sysctl_sync_ports;
4070 	tbl[idx++].data = &ipvs->sysctl_sync_persist_mode;
4071 	ipvs->sysctl_sync_qlen_max = nr_free_buffer_pages() / 32;
4072 	tbl[idx++].data = &ipvs->sysctl_sync_qlen_max;
4073 	ipvs->sysctl_sync_sock_size = 0;
4074 	tbl[idx++].data = &ipvs->sysctl_sync_sock_size;
4075 	tbl[idx++].data = &ipvs->sysctl_cache_bypass;
4076 	tbl[idx++].data = &ipvs->sysctl_expire_nodest_conn;
4077 	tbl[idx++].data = &ipvs->sysctl_sloppy_tcp;
4078 	tbl[idx++].data = &ipvs->sysctl_sloppy_sctp;
4079 	tbl[idx++].data = &ipvs->sysctl_expire_quiescent_template;
4080 	ipvs->sysctl_sync_threshold[0] = DEFAULT_SYNC_THRESHOLD;
4081 	ipvs->sysctl_sync_threshold[1] = DEFAULT_SYNC_PERIOD;
4082 	tbl[idx].data = &ipvs->sysctl_sync_threshold;
4083 	tbl[idx++].maxlen = sizeof(ipvs->sysctl_sync_threshold);
4084 	ipvs->sysctl_sync_refresh_period = DEFAULT_SYNC_REFRESH_PERIOD;
4085 	tbl[idx++].data = &ipvs->sysctl_sync_refresh_period;
4086 	ipvs->sysctl_sync_retries = clamp_t(int, DEFAULT_SYNC_RETRIES, 0, 3);
4087 	tbl[idx++].data = &ipvs->sysctl_sync_retries;
4088 	tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
4089 	ipvs->sysctl_pmtu_disc = 1;
4090 	tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
4091 	tbl[idx++].data = &ipvs->sysctl_backup_only;
4092 	ipvs->sysctl_conn_reuse_mode = 1;
4093 	tbl[idx++].data = &ipvs->sysctl_conn_reuse_mode;
4094 	tbl[idx++].data = &ipvs->sysctl_schedule_icmp;
4095 	tbl[idx++].data = &ipvs->sysctl_ignore_tunneled;
4096 	ipvs->sysctl_run_estimation = 1;
4097 	tbl[idx++].data = &ipvs->sysctl_run_estimation;
4098 #ifdef CONFIG_IP_VS_DEBUG
4099 	/* Global sysctls must be ro in non-init netns */
4100 	if (!net_eq(net, &init_net))
4101 		tbl[idx++].mode = 0444;
4102 #endif
4103 
4104 	ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
4105 	if (ipvs->sysctl_hdr == NULL) {
4106 		if (!net_eq(net, &init_net))
4107 			kfree(tbl);
4108 		return -ENOMEM;
4109 	}
4110 	ip_vs_start_estimator(ipvs, &ipvs->tot_stats);
4111 	ipvs->sysctl_tbl = tbl;
4112 	/* Schedule defense work */
4113 	INIT_DELAYED_WORK(&ipvs->defense_work, defense_work_handler);
4114 	queue_delayed_work(system_long_wq, &ipvs->defense_work,
4115 			   DEFENSE_TIMER_PERIOD);
4116 
4117 	/* Init delayed work for expiring no dest conn */
4118 	INIT_DELAYED_WORK(&ipvs->expire_nodest_conn_work,
4119 			  expire_nodest_conn_handler);
4120 
4121 	return 0;
4122 }
4123 
ip_vs_control_net_cleanup_sysctl(struct netns_ipvs * ipvs)4124 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs)
4125 {
4126 	struct net *net = ipvs->net;
4127 
4128 	cancel_delayed_work_sync(&ipvs->expire_nodest_conn_work);
4129 	cancel_delayed_work_sync(&ipvs->defense_work);
4130 	cancel_work_sync(&ipvs->defense_work.work);
4131 	unregister_net_sysctl_table(ipvs->sysctl_hdr);
4132 	ip_vs_stop_estimator(ipvs, &ipvs->tot_stats);
4133 
4134 	if (!net_eq(net, &init_net))
4135 		kfree(ipvs->sysctl_tbl);
4136 }
4137 
4138 #else
4139 
ip_vs_control_net_init_sysctl(struct netns_ipvs * ipvs)4140 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)4141 static void __net_exit ip_vs_control_net_cleanup_sysctl(struct netns_ipvs *ipvs) { }
4142 
4143 #endif
4144 
4145 static struct notifier_block ip_vs_dst_notifier = {
4146 	.notifier_call = ip_vs_dst_event,
4147 #ifdef CONFIG_IP_VS_IPV6
4148 	.priority = ADDRCONF_NOTIFY_PRIORITY + 5,
4149 #endif
4150 };
4151 
ip_vs_control_net_init(struct netns_ipvs * ipvs)4152 int __net_init ip_vs_control_net_init(struct netns_ipvs *ipvs)
4153 {
4154 	int i, idx;
4155 
4156 	/* Initialize rs_table */
4157 	for (idx = 0; idx < IP_VS_RTAB_SIZE; idx++)
4158 		INIT_HLIST_HEAD(&ipvs->rs_table[idx]);
4159 
4160 	INIT_LIST_HEAD(&ipvs->dest_trash);
4161 	spin_lock_init(&ipvs->dest_trash_lock);
4162 	timer_setup(&ipvs->dest_trash_timer, ip_vs_dest_trash_expire, 0);
4163 	atomic_set(&ipvs->ftpsvc_counter, 0);
4164 	atomic_set(&ipvs->nullsvc_counter, 0);
4165 	atomic_set(&ipvs->conn_out_counter, 0);
4166 
4167 	/* procfs stats */
4168 	ipvs->tot_stats.cpustats = alloc_percpu(struct ip_vs_cpu_stats);
4169 	if (!ipvs->tot_stats.cpustats)
4170 		return -ENOMEM;
4171 
4172 	for_each_possible_cpu(i) {
4173 		struct ip_vs_cpu_stats *ipvs_tot_stats;
4174 		ipvs_tot_stats = per_cpu_ptr(ipvs->tot_stats.cpustats, i);
4175 		u64_stats_init(&ipvs_tot_stats->syncp);
4176 	}
4177 
4178 	spin_lock_init(&ipvs->tot_stats.lock);
4179 
4180 #ifdef CONFIG_PROC_FS
4181 	if (!proc_create_net("ip_vs", 0, ipvs->net->proc_net,
4182 			     &ip_vs_info_seq_ops, sizeof(struct ip_vs_iter)))
4183 		goto err_vs;
4184 	if (!proc_create_net_single("ip_vs_stats", 0, ipvs->net->proc_net,
4185 				    ip_vs_stats_show, NULL))
4186 		goto err_stats;
4187 	if (!proc_create_net_single("ip_vs_stats_percpu", 0,
4188 				    ipvs->net->proc_net,
4189 				    ip_vs_stats_percpu_show, NULL))
4190 		goto err_percpu;
4191 #endif
4192 
4193 	if (ip_vs_control_net_init_sysctl(ipvs))
4194 		goto err;
4195 
4196 	return 0;
4197 
4198 err:
4199 #ifdef CONFIG_PROC_FS
4200 	remove_proc_entry("ip_vs_stats_percpu", ipvs->net->proc_net);
4201 
4202 err_percpu:
4203 	remove_proc_entry("ip_vs_stats", ipvs->net->proc_net);
4204 
4205 err_stats:
4206 	remove_proc_entry("ip_vs", ipvs->net->proc_net);
4207 
4208 err_vs:
4209 #endif
4210 	free_percpu(ipvs->tot_stats.cpustats);
4211 	return -ENOMEM;
4212 }
4213 
ip_vs_control_net_cleanup(struct netns_ipvs * ipvs)4214 void __net_exit ip_vs_control_net_cleanup(struct netns_ipvs *ipvs)
4215 {
4216 	ip_vs_trash_cleanup(ipvs);
4217 	ip_vs_control_net_cleanup_sysctl(ipvs);
4218 #ifdef CONFIG_PROC_FS
4219 	remove_proc_entry("ip_vs_stats_percpu", ipvs->net->proc_net);
4220 	remove_proc_entry("ip_vs_stats", ipvs->net->proc_net);
4221 	remove_proc_entry("ip_vs", ipvs->net->proc_net);
4222 #endif
4223 	free_percpu(ipvs->tot_stats.cpustats);
4224 }
4225 
ip_vs_register_nl_ioctl(void)4226 int __init ip_vs_register_nl_ioctl(void)
4227 {
4228 	int ret;
4229 
4230 	ret = nf_register_sockopt(&ip_vs_sockopts);
4231 	if (ret) {
4232 		pr_err("cannot register sockopt.\n");
4233 		goto err_sock;
4234 	}
4235 
4236 	ret = ip_vs_genl_register();
4237 	if (ret) {
4238 		pr_err("cannot register Generic Netlink interface.\n");
4239 		goto err_genl;
4240 	}
4241 	return 0;
4242 
4243 err_genl:
4244 	nf_unregister_sockopt(&ip_vs_sockopts);
4245 err_sock:
4246 	return ret;
4247 }
4248 
ip_vs_unregister_nl_ioctl(void)4249 void ip_vs_unregister_nl_ioctl(void)
4250 {
4251 	ip_vs_genl_unregister();
4252 	nf_unregister_sockopt(&ip_vs_sockopts);
4253 }
4254 
ip_vs_control_init(void)4255 int __init ip_vs_control_init(void)
4256 {
4257 	int idx;
4258 	int ret;
4259 
4260 	EnterFunction(2);
4261 
4262 	/* Initialize svc_table, ip_vs_svc_fwm_table */
4263 	for (idx = 0; idx < IP_VS_SVC_TAB_SIZE; idx++) {
4264 		INIT_HLIST_HEAD(&ip_vs_svc_table[idx]);
4265 		INIT_HLIST_HEAD(&ip_vs_svc_fwm_table[idx]);
4266 	}
4267 
4268 	smp_wmb();	/* Do we really need it now ? */
4269 
4270 	ret = register_netdevice_notifier(&ip_vs_dst_notifier);
4271 	if (ret < 0)
4272 		return ret;
4273 
4274 	LeaveFunction(2);
4275 	return 0;
4276 }
4277 
4278 
ip_vs_control_cleanup(void)4279 void ip_vs_control_cleanup(void)
4280 {
4281 	EnterFunction(2);
4282 	unregister_netdevice_notifier(&ip_vs_dst_notifier);
4283 	LeaveFunction(2);
4284 }
4285