Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * IPVS An implementation of the IP virtual server support for the | |
3 | * LINUX operating system. IPVS is now implemented as a module | |
4 | * over the Netfilter framework. IPVS can be used to build a | |
5 | * high-performance and highly available server based on a | |
6 | * cluster of servers. | |
7 | * | |
1da177e4 LT |
8 | * Authors: Wensong Zhang <wensong@linuxvirtualserver.org> |
9 | * Peter Kese <peter.kese@ijs.si> | |
10 | * Julian Anastasov <ja@ssi.bg> | |
11 | * | |
12 | * This program is free software; you can redistribute it and/or | |
13 | * modify it under the terms of the GNU General Public License | |
14 | * as published by the Free Software Foundation; either version | |
15 | * 2 of the License, or (at your option) any later version. | |
16 | * | |
17 | * The IPVS code for kernel 2.2 was done by Wensong Zhang and Peter Kese, | |
18 | * with changes/fixes from Julian Anastasov, Lars Marowsky-Bree, Horms | |
19 | * and others. | |
20 | * | |
21 | * Changes: | |
22 | * Paul `Rusty' Russell properly handle non-linear skbs | |
6869c4d8 | 23 | * Harald Welte don't use nfcache |
1da177e4 LT |
24 | * |
25 | */ | |
26 | ||
9aada7ac HE |
27 | #define KMSG_COMPONENT "IPVS" |
28 | #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt | |
29 | ||
1da177e4 LT |
30 | #include <linux/module.h> |
31 | #include <linux/kernel.h> | |
32 | #include <linux/ip.h> | |
33 | #include <linux/tcp.h> | |
2906f66a | 34 | #include <linux/sctp.h> |
1da177e4 | 35 | #include <linux/icmp.h> |
5a0e3ad6 | 36 | #include <linux/slab.h> |
1da177e4 LT |
37 | |
38 | #include <net/ip.h> | |
39 | #include <net/tcp.h> | |
40 | #include <net/udp.h> | |
41 | #include <net/icmp.h> /* for icmp_send */ | |
42 | #include <net/route.h> | |
43 | ||
44 | #include <linux/netfilter.h> | |
45 | #include <linux/netfilter_ipv4.h> | |
46 | ||
2a3b791e JV |
47 | #ifdef CONFIG_IP_VS_IPV6 |
48 | #include <net/ipv6.h> | |
49 | #include <linux/netfilter_ipv6.h> | |
50 | #endif | |
51 | ||
1da177e4 LT |
52 | #include <net/ip_vs.h> |
53 | ||
54 | ||
55 | EXPORT_SYMBOL(register_ip_vs_scheduler); | |
56 | EXPORT_SYMBOL(unregister_ip_vs_scheduler); | |
1da177e4 LT |
57 | EXPORT_SYMBOL(ip_vs_proto_name); |
58 | EXPORT_SYMBOL(ip_vs_conn_new); | |
59 | EXPORT_SYMBOL(ip_vs_conn_in_get); | |
60 | EXPORT_SYMBOL(ip_vs_conn_out_get); | |
61 | #ifdef CONFIG_IP_VS_PROTO_TCP | |
62 | EXPORT_SYMBOL(ip_vs_tcp_conn_listen); | |
63 | #endif | |
64 | EXPORT_SYMBOL(ip_vs_conn_put); | |
65 | #ifdef CONFIG_IP_VS_DEBUG | |
66 | EXPORT_SYMBOL(ip_vs_get_debug_level); | |
67 | #endif | |
1da177e4 LT |
68 | |
69 | ||
70 | /* ID used in ICMP lookups */ | |
71 | #define icmp_id(icmph) (((icmph)->un).echo.id) | |
2a3b791e | 72 | #define icmpv6_id(icmph) (icmph->icmp6_dataun.u_echo.identifier) |
1da177e4 LT |
73 | |
74 | const char *ip_vs_proto_name(unsigned proto) | |
75 | { | |
76 | static char buf[20]; | |
77 | ||
78 | switch (proto) { | |
79 | case IPPROTO_IP: | |
80 | return "IP"; | |
81 | case IPPROTO_UDP: | |
82 | return "UDP"; | |
83 | case IPPROTO_TCP: | |
84 | return "TCP"; | |
2906f66a VMR |
85 | case IPPROTO_SCTP: |
86 | return "SCTP"; | |
1da177e4 LT |
87 | case IPPROTO_ICMP: |
88 | return "ICMP"; | |
2a3b791e JV |
89 | #ifdef CONFIG_IP_VS_IPV6 |
90 | case IPPROTO_ICMPV6: | |
91 | return "ICMPv6"; | |
92 | #endif | |
1da177e4 LT |
93 | default: |
94 | sprintf(buf, "IP_%d", proto); | |
95 | return buf; | |
96 | } | |
97 | } | |
98 | ||
99 | void ip_vs_init_hash_table(struct list_head *table, int rows) | |
100 | { | |
101 | while (--rows >= 0) | |
102 | INIT_LIST_HEAD(&table[rows]); | |
103 | } | |
104 | ||
105 | static inline void | |
106 | ip_vs_in_stats(struct ip_vs_conn *cp, struct sk_buff *skb) | |
107 | { | |
108 | struct ip_vs_dest *dest = cp->dest; | |
109 | if (dest && (dest->flags & IP_VS_DEST_F_AVAILABLE)) { | |
110 | spin_lock(&dest->stats.lock); | |
e9c0ce23 SW |
111 | dest->stats.ustats.inpkts++; |
112 | dest->stats.ustats.inbytes += skb->len; | |
1da177e4 LT |
113 | spin_unlock(&dest->stats.lock); |
114 | ||
115 | spin_lock(&dest->svc->stats.lock); | |
e9c0ce23 SW |
116 | dest->svc->stats.ustats.inpkts++; |
117 | dest->svc->stats.ustats.inbytes += skb->len; | |
1da177e4 LT |
118 | spin_unlock(&dest->svc->stats.lock); |
119 | ||
120 | spin_lock(&ip_vs_stats.lock); | |
e9c0ce23 SW |
121 | ip_vs_stats.ustats.inpkts++; |
122 | ip_vs_stats.ustats.inbytes += skb->len; | |
1da177e4 LT |
123 | spin_unlock(&ip_vs_stats.lock); |
124 | } | |
125 | } | |
126 | ||
127 | ||
128 | static inline void | |
129 | ip_vs_out_stats(struct ip_vs_conn *cp, struct sk_buff *skb) | |
130 | { | |
131 | struct ip_vs_dest *dest = cp->dest; | |
132 | if (dest && (dest->flags & IP_VS_DEST_F_AVAILABLE)) { | |
133 | spin_lock(&dest->stats.lock); | |
e9c0ce23 SW |
134 | dest->stats.ustats.outpkts++; |
135 | dest->stats.ustats.outbytes += skb->len; | |
1da177e4 LT |
136 | spin_unlock(&dest->stats.lock); |
137 | ||
138 | spin_lock(&dest->svc->stats.lock); | |
e9c0ce23 SW |
139 | dest->svc->stats.ustats.outpkts++; |
140 | dest->svc->stats.ustats.outbytes += skb->len; | |
1da177e4 LT |
141 | spin_unlock(&dest->svc->stats.lock); |
142 | ||
143 | spin_lock(&ip_vs_stats.lock); | |
e9c0ce23 SW |
144 | ip_vs_stats.ustats.outpkts++; |
145 | ip_vs_stats.ustats.outbytes += skb->len; | |
1da177e4 LT |
146 | spin_unlock(&ip_vs_stats.lock); |
147 | } | |
148 | } | |
149 | ||
150 | ||
151 | static inline void | |
152 | ip_vs_conn_stats(struct ip_vs_conn *cp, struct ip_vs_service *svc) | |
153 | { | |
154 | spin_lock(&cp->dest->stats.lock); | |
e9c0ce23 | 155 | cp->dest->stats.ustats.conns++; |
1da177e4 LT |
156 | spin_unlock(&cp->dest->stats.lock); |
157 | ||
158 | spin_lock(&svc->stats.lock); | |
e9c0ce23 | 159 | svc->stats.ustats.conns++; |
1da177e4 LT |
160 | spin_unlock(&svc->stats.lock); |
161 | ||
162 | spin_lock(&ip_vs_stats.lock); | |
e9c0ce23 | 163 | ip_vs_stats.ustats.conns++; |
1da177e4 LT |
164 | spin_unlock(&ip_vs_stats.lock); |
165 | } | |
166 | ||
167 | ||
168 | static inline int | |
169 | ip_vs_set_state(struct ip_vs_conn *cp, int direction, | |
170 | const struct sk_buff *skb, | |
171 | struct ip_vs_protocol *pp) | |
172 | { | |
173 | if (unlikely(!pp->state_transition)) | |
174 | return 0; | |
175 | return pp->state_transition(cp, direction, skb, pp); | |
176 | } | |
177 | ||
178 | ||
1da177e4 LT |
179 | /* |
180 | * IPVS persistent scheduling function | |
181 | * It creates a connection entry according to its template if exists, | |
182 | * or selects a server and creates a connection entry plus a template. | |
183 | * Locking: we are svc user (svc->refcnt), so we hold all dests too | |
184 | * Protocols supported: TCP, UDP | |
185 | */ | |
186 | static struct ip_vs_conn * | |
187 | ip_vs_sched_persist(struct ip_vs_service *svc, | |
188 | const struct sk_buff *skb, | |
014d730d | 189 | __be16 ports[2]) |
1da177e4 LT |
190 | { |
191 | struct ip_vs_conn *cp = NULL; | |
28364a59 | 192 | struct ip_vs_iphdr iph; |
1da177e4 LT |
193 | struct ip_vs_dest *dest; |
194 | struct ip_vs_conn *ct; | |
cd17f9ed | 195 | __be16 dport; /* destination port to forward */ |
26ec037f | 196 | __be16 flags; |
28364a59 JV |
197 | union nf_inet_addr snet; /* source network of the client, |
198 | after masking */ | |
cd17f9ed JV |
199 | |
200 | ip_vs_fill_iphdr(svc->af, skb_network_header(skb), &iph); | |
1da177e4 LT |
201 | |
202 | /* Mask saddr with the netmask to adjust template granularity */ | |
cd17f9ed JV |
203 | #ifdef CONFIG_IP_VS_IPV6 |
204 | if (svc->af == AF_INET6) | |
205 | ipv6_addr_prefix(&snet.in6, &iph.saddr.in6, svc->netmask); | |
206 | else | |
207 | #endif | |
208 | snet.ip = iph.saddr.ip & svc->netmask; | |
1da177e4 | 209 | |
cd17f9ed JV |
210 | IP_VS_DBG_BUF(6, "p-schedule: src %s:%u dest %s:%u " |
211 | "mnet %s\n", | |
212 | IP_VS_DBG_ADDR(svc->af, &iph.saddr), ntohs(ports[0]), | |
213 | IP_VS_DBG_ADDR(svc->af, &iph.daddr), ntohs(ports[1]), | |
214 | IP_VS_DBG_ADDR(svc->af, &snet)); | |
1da177e4 LT |
215 | |
216 | /* | |
217 | * As far as we know, FTP is a very complicated network protocol, and | |
218 | * it uses control connection and data connections. For active FTP, | |
219 | * FTP server initialize data connection to the client, its source port | |
220 | * is often 20. For passive FTP, FTP server tells the clients the port | |
221 | * that it passively listens to, and the client issues the data | |
222 | * connection. In the tunneling or direct routing mode, the load | |
223 | * balancer is on the client-to-server half of connection, the port | |
224 | * number is unknown to the load balancer. So, a conn template like | |
225 | * <caddr, 0, vaddr, 0, daddr, 0> is created for persistent FTP | |
226 | * service, and a template like <caddr, 0, vaddr, vport, daddr, dport> | |
227 | * is created for other persistent services. | |
228 | */ | |
229 | if (ports[1] == svc->port) { | |
230 | /* Check if a template already exists */ | |
231 | if (svc->port != FTPPORT) | |
cd17f9ed | 232 | ct = ip_vs_ct_in_get(svc->af, iph.protocol, &snet, 0, |
28364a59 | 233 | &iph.daddr, ports[1]); |
1da177e4 | 234 | else |
cd17f9ed | 235 | ct = ip_vs_ct_in_get(svc->af, iph.protocol, &snet, 0, |
28364a59 | 236 | &iph.daddr, 0); |
1da177e4 LT |
237 | |
238 | if (!ct || !ip_vs_check_template(ct)) { | |
239 | /* | |
240 | * No template found or the dest of the connection | |
241 | * template is not available. | |
242 | */ | |
243 | dest = svc->scheduler->schedule(svc, skb); | |
244 | if (dest == NULL) { | |
245 | IP_VS_DBG(1, "p-schedule: no dest found.\n"); | |
246 | return NULL; | |
247 | } | |
248 | ||
249 | /* | |
250 | * Create a template like <protocol,caddr,0, | |
251 | * vaddr,vport,daddr,dport> for non-ftp service, | |
252 | * and <protocol,caddr,0,vaddr,0,daddr,0> | |
253 | * for ftp service. | |
254 | */ | |
255 | if (svc->port != FTPPORT) | |
cd17f9ed | 256 | ct = ip_vs_conn_new(svc->af, iph.protocol, |
28364a59 JV |
257 | &snet, 0, |
258 | &iph.daddr, | |
1da177e4 | 259 | ports[1], |
28364a59 | 260 | &dest->addr, dest->port, |
87375ab4 | 261 | IP_VS_CONN_F_TEMPLATE, |
1da177e4 LT |
262 | dest); |
263 | else | |
cd17f9ed | 264 | ct = ip_vs_conn_new(svc->af, iph.protocol, |
28364a59 JV |
265 | &snet, 0, |
266 | &iph.daddr, 0, | |
267 | &dest->addr, 0, | |
87375ab4 | 268 | IP_VS_CONN_F_TEMPLATE, |
1da177e4 LT |
269 | dest); |
270 | if (ct == NULL) | |
271 | return NULL; | |
272 | ||
273 | ct->timeout = svc->timeout; | |
274 | } else { | |
275 | /* set destination with the found template */ | |
276 | dest = ct->dest; | |
277 | } | |
278 | dport = dest->port; | |
279 | } else { | |
280 | /* | |
281 | * Note: persistent fwmark-based services and persistent | |
282 | * port zero service are handled here. | |
283 | * fwmark template: <IPPROTO_IP,caddr,0,fwmark,0,daddr,0> | |
284 | * port zero template: <protocol,caddr,0,vaddr,0,daddr,0> | |
285 | */ | |
28364a59 JV |
286 | if (svc->fwmark) { |
287 | union nf_inet_addr fwmark = { | |
be8be9ec | 288 | .ip = htonl(svc->fwmark) |
28364a59 JV |
289 | }; |
290 | ||
cd17f9ed | 291 | ct = ip_vs_ct_in_get(svc->af, IPPROTO_IP, &snet, 0, |
28364a59 JV |
292 | &fwmark, 0); |
293 | } else | |
cd17f9ed | 294 | ct = ip_vs_ct_in_get(svc->af, iph.protocol, &snet, 0, |
28364a59 | 295 | &iph.daddr, 0); |
1da177e4 LT |
296 | |
297 | if (!ct || !ip_vs_check_template(ct)) { | |
298 | /* | |
299 | * If it is not persistent port zero, return NULL, | |
300 | * otherwise create a connection template. | |
301 | */ | |
302 | if (svc->port) | |
303 | return NULL; | |
304 | ||
305 | dest = svc->scheduler->schedule(svc, skb); | |
306 | if (dest == NULL) { | |
307 | IP_VS_DBG(1, "p-schedule: no dest found.\n"); | |
308 | return NULL; | |
309 | } | |
310 | ||
311 | /* | |
312 | * Create a template according to the service | |
313 | */ | |
28364a59 JV |
314 | if (svc->fwmark) { |
315 | union nf_inet_addr fwmark = { | |
be8be9ec | 316 | .ip = htonl(svc->fwmark) |
28364a59 JV |
317 | }; |
318 | ||
cd17f9ed | 319 | ct = ip_vs_conn_new(svc->af, IPPROTO_IP, |
28364a59 JV |
320 | &snet, 0, |
321 | &fwmark, 0, | |
322 | &dest->addr, 0, | |
87375ab4 | 323 | IP_VS_CONN_F_TEMPLATE, |
1da177e4 | 324 | dest); |
28364a59 | 325 | } else |
cd17f9ed | 326 | ct = ip_vs_conn_new(svc->af, iph.protocol, |
28364a59 JV |
327 | &snet, 0, |
328 | &iph.daddr, 0, | |
329 | &dest->addr, 0, | |
87375ab4 | 330 | IP_VS_CONN_F_TEMPLATE, |
1da177e4 LT |
331 | dest); |
332 | if (ct == NULL) | |
333 | return NULL; | |
334 | ||
335 | ct->timeout = svc->timeout; | |
336 | } else { | |
337 | /* set destination with the found template */ | |
338 | dest = ct->dest; | |
339 | } | |
340 | dport = ports[1]; | |
341 | } | |
342 | ||
26ec037f NC |
343 | flags = (svc->flags & IP_VS_SVC_F_ONEPACKET |
344 | && iph.protocol == IPPROTO_UDP)? | |
345 | IP_VS_CONN_F_ONE_PACKET : 0; | |
346 | ||
1da177e4 LT |
347 | /* |
348 | * Create a new connection according to the template | |
349 | */ | |
cd17f9ed | 350 | cp = ip_vs_conn_new(svc->af, iph.protocol, |
28364a59 JV |
351 | &iph.saddr, ports[0], |
352 | &iph.daddr, ports[1], | |
353 | &dest->addr, dport, | |
26ec037f | 354 | flags, |
1da177e4 LT |
355 | dest); |
356 | if (cp == NULL) { | |
357 | ip_vs_conn_put(ct); | |
358 | return NULL; | |
359 | } | |
360 | ||
361 | /* | |
362 | * Add its control | |
363 | */ | |
364 | ip_vs_control_add(cp, ct); | |
365 | ip_vs_conn_put(ct); | |
366 | ||
367 | ip_vs_conn_stats(cp, svc); | |
368 | return cp; | |
369 | } | |
370 | ||
371 | ||
372 | /* | |
373 | * IPVS main scheduling function | |
374 | * It selects a server according to the virtual service, and | |
375 | * creates a connection entry. | |
376 | * Protocols supported: TCP, UDP | |
377 | */ | |
378 | struct ip_vs_conn * | |
379 | ip_vs_schedule(struct ip_vs_service *svc, const struct sk_buff *skb) | |
380 | { | |
381 | struct ip_vs_conn *cp = NULL; | |
28364a59 | 382 | struct ip_vs_iphdr iph; |
1da177e4 | 383 | struct ip_vs_dest *dest; |
26ec037f | 384 | __be16 _ports[2], *pptr, flags; |
1da177e4 | 385 | |
28364a59 JV |
386 | ip_vs_fill_iphdr(svc->af, skb_network_header(skb), &iph); |
387 | pptr = skb_header_pointer(skb, iph.len, sizeof(_ports), _ports); | |
1da177e4 LT |
388 | if (pptr == NULL) |
389 | return NULL; | |
390 | ||
391 | /* | |
392 | * Persistent service | |
393 | */ | |
394 | if (svc->flags & IP_VS_SVC_F_PERSISTENT) | |
395 | return ip_vs_sched_persist(svc, skb, pptr); | |
396 | ||
397 | /* | |
398 | * Non-persistent service | |
399 | */ | |
400 | if (!svc->fwmark && pptr[1] != svc->port) { | |
401 | if (!svc->port) | |
1e3e238e HE |
402 | pr_err("Schedule: port zero only supported " |
403 | "in persistent services, " | |
404 | "check your ipvs configuration\n"); | |
1da177e4 LT |
405 | return NULL; |
406 | } | |
407 | ||
408 | dest = svc->scheduler->schedule(svc, skb); | |
409 | if (dest == NULL) { | |
410 | IP_VS_DBG(1, "Schedule: no dest found.\n"); | |
411 | return NULL; | |
412 | } | |
413 | ||
26ec037f NC |
414 | flags = (svc->flags & IP_VS_SVC_F_ONEPACKET |
415 | && iph.protocol == IPPROTO_UDP)? | |
416 | IP_VS_CONN_F_ONE_PACKET : 0; | |
417 | ||
1da177e4 LT |
418 | /* |
419 | * Create a connection entry. | |
420 | */ | |
cd17f9ed | 421 | cp = ip_vs_conn_new(svc->af, iph.protocol, |
28364a59 JV |
422 | &iph.saddr, pptr[0], |
423 | &iph.daddr, pptr[1], | |
424 | &dest->addr, dest->port ? dest->port : pptr[1], | |
26ec037f | 425 | flags, |
1da177e4 LT |
426 | dest); |
427 | if (cp == NULL) | |
428 | return NULL; | |
429 | ||
cd17f9ed JV |
430 | IP_VS_DBG_BUF(6, "Schedule fwd:%c c:%s:%u v:%s:%u " |
431 | "d:%s:%u conn->flags:%X conn->refcnt:%d\n", | |
432 | ip_vs_fwd_tag(cp), | |
433 | IP_VS_DBG_ADDR(svc->af, &cp->caddr), ntohs(cp->cport), | |
434 | IP_VS_DBG_ADDR(svc->af, &cp->vaddr), ntohs(cp->vport), | |
435 | IP_VS_DBG_ADDR(svc->af, &cp->daddr), ntohs(cp->dport), | |
436 | cp->flags, atomic_read(&cp->refcnt)); | |
1da177e4 LT |
437 | |
438 | ip_vs_conn_stats(cp, svc); | |
439 | return cp; | |
440 | } | |
441 | ||
442 | ||
443 | /* | |
444 | * Pass or drop the packet. | |
445 | * Called by ip_vs_in, when the virtual service is available but | |
446 | * no destination is available for a new connection. | |
447 | */ | |
448 | int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb, | |
449 | struct ip_vs_protocol *pp) | |
450 | { | |
014d730d | 451 | __be16 _ports[2], *pptr; |
28364a59 | 452 | struct ip_vs_iphdr iph; |
2a3b791e JV |
453 | int unicast; |
454 | ip_vs_fill_iphdr(svc->af, skb_network_header(skb), &iph); | |
1da177e4 | 455 | |
28364a59 | 456 | pptr = skb_header_pointer(skb, iph.len, sizeof(_ports), _ports); |
1da177e4 LT |
457 | if (pptr == NULL) { |
458 | ip_vs_service_put(svc); | |
459 | return NF_DROP; | |
460 | } | |
461 | ||
2a3b791e JV |
462 | #ifdef CONFIG_IP_VS_IPV6 |
463 | if (svc->af == AF_INET6) | |
464 | unicast = ipv6_addr_type(&iph.daddr.in6) & IPV6_ADDR_UNICAST; | |
465 | else | |
466 | #endif | |
467 | unicast = (inet_addr_type(&init_net, iph.daddr.ip) == RTN_UNICAST); | |
468 | ||
1da177e4 | 469 | /* if it is fwmark-based service, the cache_bypass sysctl is up |
2a3b791e | 470 | and the destination is a non-local unicast, then create |
1da177e4 | 471 | a cache_bypass connection entry */ |
2a3b791e | 472 | if (sysctl_ip_vs_cache_bypass && svc->fwmark && unicast) { |
1da177e4 LT |
473 | int ret, cs; |
474 | struct ip_vs_conn *cp; | |
26ec037f NC |
475 | __u16 flags = (svc->flags & IP_VS_SVC_F_ONEPACKET && |
476 | iph.protocol == IPPROTO_UDP)? | |
477 | IP_VS_CONN_F_ONE_PACKET : 0; | |
dff630dd | 478 | union nf_inet_addr daddr = { .all = { 0, 0, 0, 0 } }; |
1da177e4 LT |
479 | |
480 | ip_vs_service_put(svc); | |
481 | ||
482 | /* create a new connection entry */ | |
1e3e238e | 483 | IP_VS_DBG(6, "%s(): create a cache_bypass entry\n", __func__); |
2a3b791e | 484 | cp = ip_vs_conn_new(svc->af, iph.protocol, |
28364a59 JV |
485 | &iph.saddr, pptr[0], |
486 | &iph.daddr, pptr[1], | |
dff630dd | 487 | &daddr, 0, |
26ec037f | 488 | IP_VS_CONN_F_BYPASS | flags, |
1da177e4 LT |
489 | NULL); |
490 | if (cp == NULL) | |
491 | return NF_DROP; | |
492 | ||
493 | /* statistics */ | |
494 | ip_vs_in_stats(cp, skb); | |
495 | ||
496 | /* set state */ | |
497 | cs = ip_vs_set_state(cp, IP_VS_DIR_INPUT, skb, pp); | |
498 | ||
499 | /* transmit the first SYN packet */ | |
500 | ret = cp->packet_xmit(skb, cp, pp); | |
501 | /* do not touch skb anymore */ | |
502 | ||
503 | atomic_inc(&cp->in_pkts); | |
504 | ip_vs_conn_put(cp); | |
505 | return ret; | |
506 | } | |
507 | ||
508 | /* | |
509 | * When the virtual ftp service is presented, packets destined | |
510 | * for other services on the VIP may get here (except services | |
511 | * listed in the ipvs table), pass the packets, because it is | |
512 | * not ipvs job to decide to drop the packets. | |
513 | */ | |
514 | if ((svc->port == FTPPORT) && (pptr[1] != FTPPORT)) { | |
515 | ip_vs_service_put(svc); | |
516 | return NF_ACCEPT; | |
517 | } | |
518 | ||
519 | ip_vs_service_put(svc); | |
520 | ||
521 | /* | |
522 | * Notify the client that the destination is unreachable, and | |
523 | * release the socket buffer. | |
524 | * Since it is in IP layer, the TCP socket is not actually | |
525 | * created, the TCP RST packet cannot be sent, instead that | |
526 | * ICMP_PORT_UNREACH is sent here no matter it is TCP/UDP. --WZ | |
527 | */ | |
2a3b791e JV |
528 | #ifdef CONFIG_IP_VS_IPV6 |
529 | if (svc->af == AF_INET6) | |
3ffe533c | 530 | icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); |
2a3b791e JV |
531 | else |
532 | #endif | |
533 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); | |
534 | ||
1da177e4 LT |
535 | return NF_DROP; |
536 | } | |
537 | ||
b1550f22 | 538 | __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset) |
1da177e4 | 539 | { |
d3bc23e7 | 540 | return csum_fold(skb_checksum(skb, offset, skb->len - offset, 0)); |
1da177e4 LT |
541 | } |
542 | ||
776c729e | 543 | static inline int ip_vs_gather_frags(struct sk_buff *skb, u_int32_t user) |
1da177e4 | 544 | { |
776c729e HX |
545 | int err = ip_defrag(skb, user); |
546 | ||
547 | if (!err) | |
eddc9ec5 | 548 | ip_send_check(ip_hdr(skb)); |
776c729e HX |
549 | |
550 | return err; | |
1da177e4 LT |
551 | } |
552 | ||
2a3b791e JV |
553 | #ifdef CONFIG_IP_VS_IPV6 |
554 | static inline int ip_vs_gather_frags_v6(struct sk_buff *skb, u_int32_t user) | |
555 | { | |
556 | /* TODO IPv6: Find out what to do here for IPv6 */ | |
557 | return 0; | |
558 | } | |
559 | #endif | |
560 | ||
1da177e4 LT |
561 | /* |
562 | * Packet has been made sufficiently writable in caller | |
563 | * - inout: 1=in->out, 0=out->in | |
564 | */ | |
565 | void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp, | |
566 | struct ip_vs_conn *cp, int inout) | |
567 | { | |
eddc9ec5 | 568 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 | 569 | unsigned int icmp_offset = iph->ihl*4; |
d56f90a7 ACM |
570 | struct icmphdr *icmph = (struct icmphdr *)(skb_network_header(skb) + |
571 | icmp_offset); | |
1da177e4 LT |
572 | struct iphdr *ciph = (struct iphdr *)(icmph + 1); |
573 | ||
574 | if (inout) { | |
e7ade46a | 575 | iph->saddr = cp->vaddr.ip; |
1da177e4 | 576 | ip_send_check(iph); |
e7ade46a | 577 | ciph->daddr = cp->vaddr.ip; |
1da177e4 LT |
578 | ip_send_check(ciph); |
579 | } else { | |
e7ade46a | 580 | iph->daddr = cp->daddr.ip; |
1da177e4 | 581 | ip_send_check(iph); |
e7ade46a | 582 | ciph->saddr = cp->daddr.ip; |
1da177e4 LT |
583 | ip_send_check(ciph); |
584 | } | |
585 | ||
2906f66a VMR |
586 | /* the TCP/UDP/SCTP port */ |
587 | if (IPPROTO_TCP == ciph->protocol || IPPROTO_UDP == ciph->protocol || | |
588 | IPPROTO_SCTP == ciph->protocol) { | |
014d730d | 589 | __be16 *ports = (void *)ciph + ciph->ihl*4; |
1da177e4 LT |
590 | |
591 | if (inout) | |
592 | ports[1] = cp->vport; | |
593 | else | |
594 | ports[0] = cp->dport; | |
595 | } | |
596 | ||
597 | /* And finally the ICMP checksum */ | |
598 | icmph->checksum = 0; | |
599 | icmph->checksum = ip_vs_checksum_complete(skb, icmp_offset); | |
600 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
601 | ||
602 | if (inout) | |
603 | IP_VS_DBG_PKT(11, pp, skb, (void *)ciph - (void *)iph, | |
604 | "Forwarding altered outgoing ICMP"); | |
605 | else | |
606 | IP_VS_DBG_PKT(11, pp, skb, (void *)ciph - (void *)iph, | |
607 | "Forwarding altered incoming ICMP"); | |
608 | } | |
609 | ||
b3cdd2a7 JV |
610 | #ifdef CONFIG_IP_VS_IPV6 |
611 | void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp, | |
612 | struct ip_vs_conn *cp, int inout) | |
613 | { | |
614 | struct ipv6hdr *iph = ipv6_hdr(skb); | |
615 | unsigned int icmp_offset = sizeof(struct ipv6hdr); | |
616 | struct icmp6hdr *icmph = (struct icmp6hdr *)(skb_network_header(skb) + | |
617 | icmp_offset); | |
618 | struct ipv6hdr *ciph = (struct ipv6hdr *)(icmph + 1); | |
619 | ||
620 | if (inout) { | |
621 | iph->saddr = cp->vaddr.in6; | |
622 | ciph->daddr = cp->vaddr.in6; | |
623 | } else { | |
624 | iph->daddr = cp->daddr.in6; | |
625 | ciph->saddr = cp->daddr.in6; | |
626 | } | |
627 | ||
2906f66a VMR |
628 | /* the TCP/UDP/SCTP port */ |
629 | if (IPPROTO_TCP == ciph->nexthdr || IPPROTO_UDP == ciph->nexthdr || | |
630 | IPPROTO_SCTP == ciph->nexthdr) { | |
b3cdd2a7 JV |
631 | __be16 *ports = (void *)ciph + sizeof(struct ipv6hdr); |
632 | ||
633 | if (inout) | |
634 | ports[1] = cp->vport; | |
635 | else | |
636 | ports[0] = cp->dport; | |
637 | } | |
638 | ||
639 | /* And finally the ICMP checksum */ | |
8870f842 SH |
640 | icmph->icmp6_cksum = ~csum_ipv6_magic(&iph->saddr, &iph->daddr, |
641 | skb->len - icmp_offset, | |
642 | IPPROTO_ICMPV6, 0); | |
643 | skb->csum_start = skb_network_header(skb) - skb->head + icmp_offset; | |
644 | skb->csum_offset = offsetof(struct icmp6hdr, icmp6_cksum); | |
645 | skb->ip_summed = CHECKSUM_PARTIAL; | |
b3cdd2a7 JV |
646 | |
647 | if (inout) | |
648 | IP_VS_DBG_PKT(11, pp, skb, (void *)ciph - (void *)iph, | |
649 | "Forwarding altered outgoing ICMPv6"); | |
650 | else | |
651 | IP_VS_DBG_PKT(11, pp, skb, (void *)ciph - (void *)iph, | |
652 | "Forwarding altered incoming ICMPv6"); | |
653 | } | |
654 | #endif | |
655 | ||
4856c84c MT |
656 | /* Handle relevant response ICMP messages - forward to the right |
657 | * destination host. Used for NAT and local client. | |
658 | */ | |
f2428ed5 SH |
659 | static int handle_response_icmp(int af, struct sk_buff *skb, |
660 | union nf_inet_addr *snet, | |
661 | __u8 protocol, struct ip_vs_conn *cp, | |
4856c84c MT |
662 | struct ip_vs_protocol *pp, |
663 | unsigned int offset, unsigned int ihl) | |
664 | { | |
665 | unsigned int verdict = NF_DROP; | |
666 | ||
667 | if (IP_VS_FWD_METHOD(cp) != 0) { | |
1e3e238e HE |
668 | pr_err("shouldn't reach here, because the box is on the " |
669 | "half connection in the tun/dr module.\n"); | |
4856c84c MT |
670 | } |
671 | ||
672 | /* Ensure the checksum is correct */ | |
673 | if (!skb_csum_unnecessary(skb) && ip_vs_checksum_complete(skb, ihl)) { | |
674 | /* Failed checksum! */ | |
f2428ed5 SH |
675 | IP_VS_DBG_BUF(1, "Forward ICMP: failed checksum from %s!\n", |
676 | IP_VS_DBG_ADDR(af, snet)); | |
4856c84c MT |
677 | goto out; |
678 | } | |
679 | ||
2906f66a VMR |
680 | if (IPPROTO_TCP == protocol || IPPROTO_UDP == protocol || |
681 | IPPROTO_SCTP == protocol) | |
4856c84c MT |
682 | offset += 2 * sizeof(__u16); |
683 | if (!skb_make_writable(skb, offset)) | |
684 | goto out; | |
685 | ||
f2428ed5 SH |
686 | #ifdef CONFIG_IP_VS_IPV6 |
687 | if (af == AF_INET6) | |
688 | ip_vs_nat_icmp_v6(skb, pp, cp, 1); | |
689 | else | |
690 | #endif | |
691 | ip_vs_nat_icmp(skb, pp, cp, 1); | |
4856c84c MT |
692 | |
693 | /* do the statistics and put it back */ | |
694 | ip_vs_out_stats(cp, skb); | |
695 | ||
696 | skb->ipvs_property = 1; | |
697 | verdict = NF_ACCEPT; | |
698 | ||
699 | out: | |
700 | __ip_vs_conn_put(cp); | |
701 | ||
702 | return verdict; | |
703 | } | |
704 | ||
1da177e4 LT |
705 | /* |
706 | * Handle ICMP messages in the inside-to-outside direction (outgoing). | |
4856c84c | 707 | * Find any that might be relevant, check against existing connections. |
1da177e4 | 708 | * Currently handles error types - unreachable, quench, ttl exceeded. |
1da177e4 | 709 | */ |
3db05fea | 710 | static int ip_vs_out_icmp(struct sk_buff *skb, int *related) |
1da177e4 | 711 | { |
1da177e4 LT |
712 | struct iphdr *iph; |
713 | struct icmphdr _icmph, *ic; | |
714 | struct iphdr _ciph, *cih; /* The ip header contained within the ICMP */ | |
51ef348b | 715 | struct ip_vs_iphdr ciph; |
1da177e4 LT |
716 | struct ip_vs_conn *cp; |
717 | struct ip_vs_protocol *pp; | |
4856c84c | 718 | unsigned int offset, ihl; |
f2428ed5 | 719 | union nf_inet_addr snet; |
1da177e4 LT |
720 | |
721 | *related = 1; | |
722 | ||
723 | /* reassemble IP fragments */ | |
eddc9ec5 | 724 | if (ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)) { |
776c729e | 725 | if (ip_vs_gather_frags(skb, IP_DEFRAG_VS_OUT)) |
1da177e4 | 726 | return NF_STOLEN; |
1da177e4 LT |
727 | } |
728 | ||
eddc9ec5 | 729 | iph = ip_hdr(skb); |
1da177e4 LT |
730 | offset = ihl = iph->ihl * 4; |
731 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
732 | if (ic == NULL) | |
733 | return NF_DROP; | |
734 | ||
14d5e834 | 735 | IP_VS_DBG(12, "Outgoing ICMP (%d,%d) %pI4->%pI4\n", |
1da177e4 | 736 | ic->type, ntohs(icmp_id(ic)), |
14d5e834 | 737 | &iph->saddr, &iph->daddr); |
1da177e4 LT |
738 | |
739 | /* | |
740 | * Work through seeing if this is for us. | |
741 | * These checks are supposed to be in an order that means easy | |
742 | * things are checked first to speed up processing.... however | |
743 | * this means that some packets will manage to get a long way | |
744 | * down this stack and then be rejected, but that's life. | |
745 | */ | |
746 | if ((ic->type != ICMP_DEST_UNREACH) && | |
747 | (ic->type != ICMP_SOURCE_QUENCH) && | |
748 | (ic->type != ICMP_TIME_EXCEEDED)) { | |
749 | *related = 0; | |
750 | return NF_ACCEPT; | |
751 | } | |
752 | ||
753 | /* Now find the contained IP header */ | |
754 | offset += sizeof(_icmph); | |
755 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
756 | if (cih == NULL) | |
757 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
758 | ||
759 | pp = ip_vs_proto_get(cih->protocol); | |
760 | if (!pp) | |
761 | return NF_ACCEPT; | |
762 | ||
763 | /* Is the embedded protocol header present? */ | |
4412ec49 | 764 | if (unlikely(cih->frag_off & htons(IP_OFFSET) && |
1da177e4 LT |
765 | pp->dont_defrag)) |
766 | return NF_ACCEPT; | |
767 | ||
768 | IP_VS_DBG_PKT(11, pp, skb, offset, "Checking outgoing ICMP for"); | |
769 | ||
770 | offset += cih->ihl * 4; | |
771 | ||
51ef348b | 772 | ip_vs_fill_iphdr(AF_INET, cih, &ciph); |
1da177e4 | 773 | /* The embedded headers contain source and dest in reverse order */ |
51ef348b | 774 | cp = pp->conn_out_get(AF_INET, skb, pp, &ciph, offset, 1); |
1da177e4 LT |
775 | if (!cp) |
776 | return NF_ACCEPT; | |
777 | ||
f2428ed5 SH |
778 | snet.ip = iph->saddr; |
779 | return handle_response_icmp(AF_INET, skb, &snet, cih->protocol, cp, | |
780 | pp, offset, ihl); | |
1da177e4 LT |
781 | } |
782 | ||
2a3b791e JV |
783 | #ifdef CONFIG_IP_VS_IPV6 |
784 | static int ip_vs_out_icmp_v6(struct sk_buff *skb, int *related) | |
785 | { | |
786 | struct ipv6hdr *iph; | |
787 | struct icmp6hdr _icmph, *ic; | |
788 | struct ipv6hdr _ciph, *cih; /* The ip header contained | |
789 | within the ICMP */ | |
790 | struct ip_vs_iphdr ciph; | |
791 | struct ip_vs_conn *cp; | |
792 | struct ip_vs_protocol *pp; | |
f2428ed5 SH |
793 | unsigned int offset; |
794 | union nf_inet_addr snet; | |
2a3b791e JV |
795 | |
796 | *related = 1; | |
797 | ||
798 | /* reassemble IP fragments */ | |
799 | if (ipv6_hdr(skb)->nexthdr == IPPROTO_FRAGMENT) { | |
800 | if (ip_vs_gather_frags_v6(skb, IP_DEFRAG_VS_OUT)) | |
801 | return NF_STOLEN; | |
802 | } | |
803 | ||
804 | iph = ipv6_hdr(skb); | |
805 | offset = sizeof(struct ipv6hdr); | |
806 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
807 | if (ic == NULL) | |
808 | return NF_DROP; | |
809 | ||
5b095d98 | 810 | IP_VS_DBG(12, "Outgoing ICMPv6 (%d,%d) %pI6->%pI6\n", |
2a3b791e | 811 | ic->icmp6_type, ntohs(icmpv6_id(ic)), |
38ff4fa4 | 812 | &iph->saddr, &iph->daddr); |
2a3b791e JV |
813 | |
814 | /* | |
815 | * Work through seeing if this is for us. | |
816 | * These checks are supposed to be in an order that means easy | |
817 | * things are checked first to speed up processing.... however | |
818 | * this means that some packets will manage to get a long way | |
819 | * down this stack and then be rejected, but that's life. | |
820 | */ | |
821 | if ((ic->icmp6_type != ICMPV6_DEST_UNREACH) && | |
822 | (ic->icmp6_type != ICMPV6_PKT_TOOBIG) && | |
823 | (ic->icmp6_type != ICMPV6_TIME_EXCEED)) { | |
824 | *related = 0; | |
825 | return NF_ACCEPT; | |
826 | } | |
827 | ||
828 | /* Now find the contained IP header */ | |
829 | offset += sizeof(_icmph); | |
830 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
831 | if (cih == NULL) | |
832 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
833 | ||
834 | pp = ip_vs_proto_get(cih->nexthdr); | |
835 | if (!pp) | |
836 | return NF_ACCEPT; | |
837 | ||
838 | /* Is the embedded protocol header present? */ | |
839 | /* TODO: we don't support fragmentation at the moment anyways */ | |
840 | if (unlikely(cih->nexthdr == IPPROTO_FRAGMENT && pp->dont_defrag)) | |
841 | return NF_ACCEPT; | |
842 | ||
843 | IP_VS_DBG_PKT(11, pp, skb, offset, "Checking outgoing ICMPv6 for"); | |
844 | ||
845 | offset += sizeof(struct ipv6hdr); | |
846 | ||
847 | ip_vs_fill_iphdr(AF_INET6, cih, &ciph); | |
848 | /* The embedded headers contain source and dest in reverse order */ | |
849 | cp = pp->conn_out_get(AF_INET6, skb, pp, &ciph, offset, 1); | |
850 | if (!cp) | |
851 | return NF_ACCEPT; | |
852 | ||
178f5e49 | 853 | ipv6_addr_copy(&snet.in6, &iph->saddr); |
f2428ed5 SH |
854 | return handle_response_icmp(AF_INET6, skb, &snet, cih->nexthdr, cp, |
855 | pp, offset, sizeof(struct ipv6hdr)); | |
2a3b791e JV |
856 | } |
857 | #endif | |
858 | ||
2906f66a VMR |
859 | /* |
860 | * Check if sctp chunc is ABORT chunk | |
861 | */ | |
862 | static inline int is_sctp_abort(const struct sk_buff *skb, int nh_len) | |
863 | { | |
864 | sctp_chunkhdr_t *sch, schunk; | |
865 | sch = skb_header_pointer(skb, nh_len + sizeof(sctp_sctphdr_t), | |
866 | sizeof(schunk), &schunk); | |
867 | if (sch == NULL) | |
868 | return 0; | |
869 | if (sch->type == SCTP_CID_ABORT) | |
870 | return 1; | |
871 | return 0; | |
872 | } | |
873 | ||
2a3b791e | 874 | static inline int is_tcp_reset(const struct sk_buff *skb, int nh_len) |
1da177e4 LT |
875 | { |
876 | struct tcphdr _tcph, *th; | |
877 | ||
2a3b791e | 878 | th = skb_header_pointer(skb, nh_len, sizeof(_tcph), &_tcph); |
1da177e4 LT |
879 | if (th == NULL) |
880 | return 0; | |
881 | return th->rst; | |
882 | } | |
883 | ||
4856c84c MT |
884 | /* Handle response packets: rewrite addresses and send away... |
885 | * Used for NAT and local client. | |
886 | */ | |
887 | static unsigned int | |
888 | handle_response(int af, struct sk_buff *skb, struct ip_vs_protocol *pp, | |
889 | struct ip_vs_conn *cp, int ihl) | |
890 | { | |
891 | IP_VS_DBG_PKT(11, pp, skb, 0, "Outgoing packet"); | |
892 | ||
893 | if (!skb_make_writable(skb, ihl)) | |
894 | goto drop; | |
895 | ||
896 | /* mangle the packet */ | |
897 | if (pp->snat_handler && !pp->snat_handler(skb, pp, cp)) | |
898 | goto drop; | |
899 | ||
900 | #ifdef CONFIG_IP_VS_IPV6 | |
901 | if (af == AF_INET6) | |
902 | ipv6_hdr(skb)->saddr = cp->vaddr.in6; | |
903 | else | |
904 | #endif | |
905 | { | |
906 | ip_hdr(skb)->saddr = cp->vaddr.ip; | |
907 | ip_send_check(ip_hdr(skb)); | |
908 | } | |
909 | ||
910 | /* For policy routing, packets originating from this | |
911 | * machine itself may be routed differently to packets | |
912 | * passing through. We want this packet to be routed as | |
913 | * if it came from this machine itself. So re-compute | |
914 | * the routing information. | |
915 | */ | |
916 | #ifdef CONFIG_IP_VS_IPV6 | |
917 | if (af == AF_INET6) { | |
918 | if (ip6_route_me_harder(skb) != 0) | |
919 | goto drop; | |
920 | } else | |
921 | #endif | |
922 | if (ip_route_me_harder(skb, RTN_LOCAL) != 0) | |
923 | goto drop; | |
924 | ||
4856c84c MT |
925 | IP_VS_DBG_PKT(10, pp, skb, 0, "After SNAT"); |
926 | ||
927 | ip_vs_out_stats(cp, skb); | |
928 | ip_vs_set_state(cp, IP_VS_DIR_OUTPUT, skb, pp); | |
929 | ip_vs_conn_put(cp); | |
930 | ||
931 | skb->ipvs_property = 1; | |
932 | ||
933 | LeaveFunction(11); | |
934 | return NF_ACCEPT; | |
935 | ||
936 | drop: | |
937 | ip_vs_conn_put(cp); | |
938 | kfree_skb(skb); | |
939 | return NF_STOLEN; | |
940 | } | |
941 | ||
1da177e4 | 942 | /* |
6e23ae2a | 943 | * It is hooked at the NF_INET_FORWARD chain, used only for VS/NAT. |
4856c84c | 944 | * Check if outgoing packet belongs to the established ip_vs_conn. |
1da177e4 LT |
945 | */ |
946 | static unsigned int | |
3db05fea | 947 | ip_vs_out(unsigned int hooknum, struct sk_buff *skb, |
1da177e4 LT |
948 | const struct net_device *in, const struct net_device *out, |
949 | int (*okfn)(struct sk_buff *)) | |
950 | { | |
51ef348b | 951 | struct ip_vs_iphdr iph; |
1da177e4 LT |
952 | struct ip_vs_protocol *pp; |
953 | struct ip_vs_conn *cp; | |
2a3b791e | 954 | int af; |
1da177e4 LT |
955 | |
956 | EnterFunction(11); | |
957 | ||
60678040 | 958 | af = (skb->protocol == htons(ETH_P_IP)) ? AF_INET : AF_INET6; |
2a3b791e | 959 | |
6869c4d8 | 960 | if (skb->ipvs_property) |
1da177e4 LT |
961 | return NF_ACCEPT; |
962 | ||
2a3b791e JV |
963 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); |
964 | #ifdef CONFIG_IP_VS_IPV6 | |
965 | if (af == AF_INET6) { | |
966 | if (unlikely(iph.protocol == IPPROTO_ICMPV6)) { | |
967 | int related, verdict = ip_vs_out_icmp_v6(skb, &related); | |
1da177e4 | 968 | |
2a3b791e JV |
969 | if (related) |
970 | return verdict; | |
971 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); | |
972 | } | |
973 | } else | |
974 | #endif | |
975 | if (unlikely(iph.protocol == IPPROTO_ICMP)) { | |
976 | int related, verdict = ip_vs_out_icmp(skb, &related); | |
977 | ||
978 | if (related) | |
979 | return verdict; | |
980 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); | |
981 | } | |
1da177e4 | 982 | |
51ef348b | 983 | pp = ip_vs_proto_get(iph.protocol); |
1da177e4 LT |
984 | if (unlikely(!pp)) |
985 | return NF_ACCEPT; | |
986 | ||
987 | /* reassemble IP fragments */ | |
2a3b791e JV |
988 | #ifdef CONFIG_IP_VS_IPV6 |
989 | if (af == AF_INET6) { | |
990 | if (unlikely(iph.protocol == IPPROTO_ICMPV6)) { | |
991 | int related, verdict = ip_vs_out_icmp_v6(skb, &related); | |
1da177e4 | 992 | |
2a3b791e JV |
993 | if (related) |
994 | return verdict; | |
995 | ||
996 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); | |
997 | } | |
998 | } else | |
999 | #endif | |
1000 | if (unlikely(ip_hdr(skb)->frag_off & htons(IP_MF|IP_OFFSET) && | |
1001 | !pp->dont_defrag)) { | |
1002 | if (ip_vs_gather_frags(skb, IP_DEFRAG_VS_OUT)) | |
1003 | return NF_STOLEN; | |
1004 | ||
1005 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); | |
1006 | } | |
1da177e4 LT |
1007 | |
1008 | /* | |
1009 | * Check if the packet belongs to an existing entry | |
1010 | */ | |
2a3b791e | 1011 | cp = pp->conn_out_get(af, skb, pp, &iph, iph.len, 0); |
1da177e4 LT |
1012 | |
1013 | if (unlikely(!cp)) { | |
1014 | if (sysctl_ip_vs_nat_icmp_send && | |
1015 | (pp->protocol == IPPROTO_TCP || | |
2906f66a VMR |
1016 | pp->protocol == IPPROTO_UDP || |
1017 | pp->protocol == IPPROTO_SCTP)) { | |
014d730d | 1018 | __be16 _ports[2], *pptr; |
1da177e4 | 1019 | |
51ef348b | 1020 | pptr = skb_header_pointer(skb, iph.len, |
1da177e4 LT |
1021 | sizeof(_ports), _ports); |
1022 | if (pptr == NULL) | |
1023 | return NF_ACCEPT; /* Not for me */ | |
7937df15 JV |
1024 | if (ip_vs_lookup_real_service(af, iph.protocol, |
1025 | &iph.saddr, | |
2a3b791e | 1026 | pptr[0])) { |
1da177e4 LT |
1027 | /* |
1028 | * Notify the real server: there is no | |
1029 | * existing entry if it is not RST | |
1030 | * packet or not TCP packet. | |
1031 | */ | |
2906f66a VMR |
1032 | if ((iph.protocol != IPPROTO_TCP && |
1033 | iph.protocol != IPPROTO_SCTP) | |
1034 | || ((iph.protocol == IPPROTO_TCP | |
1035 | && !is_tcp_reset(skb, iph.len)) | |
1036 | || (iph.protocol == IPPROTO_SCTP | |
1037 | && !is_sctp_abort(skb, | |
1038 | iph.len)))) { | |
2a3b791e JV |
1039 | #ifdef CONFIG_IP_VS_IPV6 |
1040 | if (af == AF_INET6) | |
1041 | icmpv6_send(skb, | |
1042 | ICMPV6_DEST_UNREACH, | |
1043 | ICMPV6_PORT_UNREACH, | |
3ffe533c | 1044 | 0); |
2a3b791e JV |
1045 | else |
1046 | #endif | |
1047 | icmp_send(skb, | |
1048 | ICMP_DEST_UNREACH, | |
1049 | ICMP_PORT_UNREACH, 0); | |
1da177e4 | 1050 | return NF_DROP; |
5af149cc | 1051 | } |
1da177e4 LT |
1052 | } |
1053 | } | |
1054 | IP_VS_DBG_PKT(12, pp, skb, 0, | |
1055 | "packet continues traversal as normal"); | |
1056 | return NF_ACCEPT; | |
1057 | } | |
1058 | ||
4856c84c | 1059 | return handle_response(af, skb, pp, cp, iph.len); |
1da177e4 LT |
1060 | } |
1061 | ||
1062 | ||
1063 | /* | |
1064 | * Handle ICMP messages in the outside-to-inside direction (incoming). | |
1065 | * Find any that might be relevant, check against existing connections, | |
1066 | * forward to the right destination host if relevant. | |
1067 | * Currently handles error types - unreachable, quench, ttl exceeded. | |
1068 | */ | |
e905a9ed | 1069 | static int |
3db05fea | 1070 | ip_vs_in_icmp(struct sk_buff *skb, int *related, unsigned int hooknum) |
1da177e4 | 1071 | { |
1da177e4 LT |
1072 | struct iphdr *iph; |
1073 | struct icmphdr _icmph, *ic; | |
1074 | struct iphdr _ciph, *cih; /* The ip header contained within the ICMP */ | |
51ef348b | 1075 | struct ip_vs_iphdr ciph; |
1da177e4 LT |
1076 | struct ip_vs_conn *cp; |
1077 | struct ip_vs_protocol *pp; | |
1078 | unsigned int offset, ihl, verdict; | |
f2428ed5 | 1079 | union nf_inet_addr snet; |
1da177e4 LT |
1080 | |
1081 | *related = 1; | |
1082 | ||
1083 | /* reassemble IP fragments */ | |
eddc9ec5 | 1084 | if (ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)) { |
6e23ae2a | 1085 | if (ip_vs_gather_frags(skb, hooknum == NF_INET_LOCAL_IN ? |
776c729e | 1086 | IP_DEFRAG_VS_IN : IP_DEFRAG_VS_FWD)) |
1da177e4 | 1087 | return NF_STOLEN; |
1da177e4 LT |
1088 | } |
1089 | ||
eddc9ec5 | 1090 | iph = ip_hdr(skb); |
1da177e4 LT |
1091 | offset = ihl = iph->ihl * 4; |
1092 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
1093 | if (ic == NULL) | |
1094 | return NF_DROP; | |
1095 | ||
14d5e834 | 1096 | IP_VS_DBG(12, "Incoming ICMP (%d,%d) %pI4->%pI4\n", |
1da177e4 | 1097 | ic->type, ntohs(icmp_id(ic)), |
14d5e834 | 1098 | &iph->saddr, &iph->daddr); |
1da177e4 LT |
1099 | |
1100 | /* | |
1101 | * Work through seeing if this is for us. | |
1102 | * These checks are supposed to be in an order that means easy | |
1103 | * things are checked first to speed up processing.... however | |
1104 | * this means that some packets will manage to get a long way | |
1105 | * down this stack and then be rejected, but that's life. | |
1106 | */ | |
1107 | if ((ic->type != ICMP_DEST_UNREACH) && | |
1108 | (ic->type != ICMP_SOURCE_QUENCH) && | |
1109 | (ic->type != ICMP_TIME_EXCEEDED)) { | |
1110 | *related = 0; | |
1111 | return NF_ACCEPT; | |
1112 | } | |
1113 | ||
1114 | /* Now find the contained IP header */ | |
1115 | offset += sizeof(_icmph); | |
1116 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
1117 | if (cih == NULL) | |
1118 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
1119 | ||
1120 | pp = ip_vs_proto_get(cih->protocol); | |
1121 | if (!pp) | |
1122 | return NF_ACCEPT; | |
1123 | ||
1124 | /* Is the embedded protocol header present? */ | |
4412ec49 | 1125 | if (unlikely(cih->frag_off & htons(IP_OFFSET) && |
1da177e4 LT |
1126 | pp->dont_defrag)) |
1127 | return NF_ACCEPT; | |
1128 | ||
1129 | IP_VS_DBG_PKT(11, pp, skb, offset, "Checking incoming ICMP for"); | |
1130 | ||
1131 | offset += cih->ihl * 4; | |
1132 | ||
51ef348b | 1133 | ip_vs_fill_iphdr(AF_INET, cih, &ciph); |
1da177e4 | 1134 | /* The embedded headers contain source and dest in reverse order */ |
51ef348b | 1135 | cp = pp->conn_in_get(AF_INET, skb, pp, &ciph, offset, 1); |
4856c84c MT |
1136 | if (!cp) { |
1137 | /* The packet could also belong to a local client */ | |
1138 | cp = pp->conn_out_get(AF_INET, skb, pp, &ciph, offset, 1); | |
f2428ed5 SH |
1139 | if (cp) { |
1140 | snet.ip = iph->saddr; | |
1141 | return handle_response_icmp(AF_INET, skb, &snet, | |
1142 | cih->protocol, cp, pp, | |
4856c84c | 1143 | offset, ihl); |
f2428ed5 | 1144 | } |
1da177e4 | 1145 | return NF_ACCEPT; |
4856c84c | 1146 | } |
1da177e4 LT |
1147 | |
1148 | verdict = NF_DROP; | |
1149 | ||
1150 | /* Ensure the checksum is correct */ | |
60476372 | 1151 | if (!skb_csum_unnecessary(skb) && ip_vs_checksum_complete(skb, ihl)) { |
1da177e4 | 1152 | /* Failed checksum! */ |
14d5e834 HH |
1153 | IP_VS_DBG(1, "Incoming ICMP: failed checksum from %pI4!\n", |
1154 | &iph->saddr); | |
1da177e4 LT |
1155 | goto out; |
1156 | } | |
1157 | ||
1158 | /* do the statistics and put it back */ | |
1159 | ip_vs_in_stats(cp, skb); | |
1160 | if (IPPROTO_TCP == cih->protocol || IPPROTO_UDP == cih->protocol) | |
1161 | offset += 2 * sizeof(__u16); | |
1162 | verdict = ip_vs_icmp_xmit(skb, cp, pp, offset); | |
1163 | /* do not touch skb anymore */ | |
1164 | ||
1165 | out: | |
1166 | __ip_vs_conn_put(cp); | |
1167 | ||
1168 | return verdict; | |
1169 | } | |
1170 | ||
2a3b791e JV |
1171 | #ifdef CONFIG_IP_VS_IPV6 |
1172 | static int | |
1173 | ip_vs_in_icmp_v6(struct sk_buff *skb, int *related, unsigned int hooknum) | |
1174 | { | |
1175 | struct ipv6hdr *iph; | |
1176 | struct icmp6hdr _icmph, *ic; | |
1177 | struct ipv6hdr _ciph, *cih; /* The ip header contained | |
1178 | within the ICMP */ | |
1179 | struct ip_vs_iphdr ciph; | |
1180 | struct ip_vs_conn *cp; | |
1181 | struct ip_vs_protocol *pp; | |
1182 | unsigned int offset, verdict; | |
f2428ed5 | 1183 | union nf_inet_addr snet; |
2a3b791e JV |
1184 | |
1185 | *related = 1; | |
1186 | ||
1187 | /* reassemble IP fragments */ | |
1188 | if (ipv6_hdr(skb)->nexthdr == IPPROTO_FRAGMENT) { | |
1189 | if (ip_vs_gather_frags_v6(skb, hooknum == NF_INET_LOCAL_IN ? | |
1190 | IP_DEFRAG_VS_IN : | |
1191 | IP_DEFRAG_VS_FWD)) | |
1192 | return NF_STOLEN; | |
1193 | } | |
1194 | ||
1195 | iph = ipv6_hdr(skb); | |
1196 | offset = sizeof(struct ipv6hdr); | |
1197 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
1198 | if (ic == NULL) | |
1199 | return NF_DROP; | |
1200 | ||
5b095d98 | 1201 | IP_VS_DBG(12, "Incoming ICMPv6 (%d,%d) %pI6->%pI6\n", |
2a3b791e | 1202 | ic->icmp6_type, ntohs(icmpv6_id(ic)), |
38ff4fa4 | 1203 | &iph->saddr, &iph->daddr); |
2a3b791e JV |
1204 | |
1205 | /* | |
1206 | * Work through seeing if this is for us. | |
1207 | * These checks are supposed to be in an order that means easy | |
1208 | * things are checked first to speed up processing.... however | |
1209 | * this means that some packets will manage to get a long way | |
1210 | * down this stack and then be rejected, but that's life. | |
1211 | */ | |
1212 | if ((ic->icmp6_type != ICMPV6_DEST_UNREACH) && | |
1213 | (ic->icmp6_type != ICMPV6_PKT_TOOBIG) && | |
1214 | (ic->icmp6_type != ICMPV6_TIME_EXCEED)) { | |
1215 | *related = 0; | |
1216 | return NF_ACCEPT; | |
1217 | } | |
1218 | ||
1219 | /* Now find the contained IP header */ | |
1220 | offset += sizeof(_icmph); | |
1221 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
1222 | if (cih == NULL) | |
1223 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
1224 | ||
1225 | pp = ip_vs_proto_get(cih->nexthdr); | |
1226 | if (!pp) | |
1227 | return NF_ACCEPT; | |
1228 | ||
1229 | /* Is the embedded protocol header present? */ | |
1230 | /* TODO: we don't support fragmentation at the moment anyways */ | |
1231 | if (unlikely(cih->nexthdr == IPPROTO_FRAGMENT && pp->dont_defrag)) | |
1232 | return NF_ACCEPT; | |
1233 | ||
1234 | IP_VS_DBG_PKT(11, pp, skb, offset, "Checking incoming ICMPv6 for"); | |
1235 | ||
1236 | offset += sizeof(struct ipv6hdr); | |
1237 | ||
1238 | ip_vs_fill_iphdr(AF_INET6, cih, &ciph); | |
1239 | /* The embedded headers contain source and dest in reverse order */ | |
1240 | cp = pp->conn_in_get(AF_INET6, skb, pp, &ciph, offset, 1); | |
f2428ed5 SH |
1241 | if (!cp) { |
1242 | /* The packet could also belong to a local client */ | |
1243 | cp = pp->conn_out_get(AF_INET6, skb, pp, &ciph, offset, 1); | |
1244 | if (cp) { | |
178f5e49 | 1245 | ipv6_addr_copy(&snet.in6, &iph->saddr); |
f2428ed5 SH |
1246 | return handle_response_icmp(AF_INET6, skb, &snet, |
1247 | cih->nexthdr, | |
1248 | cp, pp, offset, | |
1249 | sizeof(struct ipv6hdr)); | |
1250 | } | |
2a3b791e | 1251 | return NF_ACCEPT; |
f2428ed5 | 1252 | } |
2a3b791e JV |
1253 | |
1254 | verdict = NF_DROP; | |
1255 | ||
1256 | /* do the statistics and put it back */ | |
1257 | ip_vs_in_stats(cp, skb); | |
2906f66a VMR |
1258 | if (IPPROTO_TCP == cih->nexthdr || IPPROTO_UDP == cih->nexthdr || |
1259 | IPPROTO_SCTP == cih->nexthdr) | |
2a3b791e JV |
1260 | offset += 2 * sizeof(__u16); |
1261 | verdict = ip_vs_icmp_xmit_v6(skb, cp, pp, offset); | |
1262 | /* do not touch skb anymore */ | |
1263 | ||
1264 | __ip_vs_conn_put(cp); | |
1265 | ||
1266 | return verdict; | |
1267 | } | |
1268 | #endif | |
1269 | ||
1270 | ||
1da177e4 LT |
1271 | /* |
1272 | * Check if it's for virtual services, look it up, | |
1273 | * and send it on its way... | |
1274 | */ | |
1275 | static unsigned int | |
3db05fea | 1276 | ip_vs_in(unsigned int hooknum, struct sk_buff *skb, |
1da177e4 LT |
1277 | const struct net_device *in, const struct net_device *out, |
1278 | int (*okfn)(struct sk_buff *)) | |
1279 | { | |
51ef348b | 1280 | struct ip_vs_iphdr iph; |
1da177e4 LT |
1281 | struct ip_vs_protocol *pp; |
1282 | struct ip_vs_conn *cp; | |
1e66dafc | 1283 | int ret, restart, af, pkts; |
2a3b791e | 1284 | |
60678040 | 1285 | af = (skb->protocol == htons(ETH_P_IP)) ? AF_INET : AF_INET6; |
51ef348b | 1286 | |
2a3b791e | 1287 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); |
1da177e4 LT |
1288 | |
1289 | /* | |
4856c84c MT |
1290 | * Big tappo: only PACKET_HOST, including loopback for local client |
1291 | * Don't handle local packets on IPv6 for now | |
1da177e4 | 1292 | */ |
f2428ed5 | 1293 | if (unlikely(skb->pkt_type != PACKET_HOST)) { |
51ef348b JV |
1294 | IP_VS_DBG_BUF(12, "packet type=%d proto=%d daddr=%s ignored\n", |
1295 | skb->pkt_type, | |
1296 | iph.protocol, | |
2a3b791e | 1297 | IP_VS_DBG_ADDR(af, &iph.daddr)); |
1da177e4 LT |
1298 | return NF_ACCEPT; |
1299 | } | |
1300 | ||
94b26551 JV |
1301 | #ifdef CONFIG_IP_VS_IPV6 |
1302 | if (af == AF_INET6) { | |
1303 | if (unlikely(iph.protocol == IPPROTO_ICMPV6)) { | |
1304 | int related, verdict = ip_vs_in_icmp_v6(skb, &related, hooknum); | |
1da177e4 | 1305 | |
94b26551 JV |
1306 | if (related) |
1307 | return verdict; | |
1308 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); | |
1309 | } | |
1310 | } else | |
1311 | #endif | |
1312 | if (unlikely(iph.protocol == IPPROTO_ICMP)) { | |
1313 | int related, verdict = ip_vs_in_icmp(skb, &related, hooknum); | |
1314 | ||
1315 | if (related) | |
1316 | return verdict; | |
1317 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); | |
1318 | } | |
1da177e4 LT |
1319 | |
1320 | /* Protocol supported? */ | |
51ef348b | 1321 | pp = ip_vs_proto_get(iph.protocol); |
1da177e4 LT |
1322 | if (unlikely(!pp)) |
1323 | return NF_ACCEPT; | |
1324 | ||
1da177e4 LT |
1325 | /* |
1326 | * Check if the packet belongs to an existing connection entry | |
1327 | */ | |
2a3b791e | 1328 | cp = pp->conn_in_get(af, skb, pp, &iph, iph.len, 0); |
1da177e4 LT |
1329 | |
1330 | if (unlikely(!cp)) { | |
1331 | int v; | |
1332 | ||
4856c84c MT |
1333 | /* For local client packets, it could be a response */ |
1334 | cp = pp->conn_out_get(af, skb, pp, &iph, iph.len, 0); | |
1335 | if (cp) | |
1336 | return handle_response(af, skb, pp, cp, iph.len); | |
1337 | ||
2a3b791e | 1338 | if (!pp->conn_schedule(af, skb, pp, &v, &cp)) |
1da177e4 LT |
1339 | return v; |
1340 | } | |
1341 | ||
1342 | if (unlikely(!cp)) { | |
1343 | /* sorry, all this trouble for a no-hit :) */ | |
1344 | IP_VS_DBG_PKT(12, pp, skb, 0, | |
1345 | "packet continues traversal as normal"); | |
1346 | return NF_ACCEPT; | |
1347 | } | |
1348 | ||
1349 | IP_VS_DBG_PKT(11, pp, skb, 0, "Incoming packet"); | |
1350 | ||
1351 | /* Check the server status */ | |
1352 | if (cp->dest && !(cp->dest->flags & IP_VS_DEST_F_AVAILABLE)) { | |
1353 | /* the destination server is not available */ | |
1354 | ||
1355 | if (sysctl_ip_vs_expire_nodest_conn) { | |
1356 | /* try to expire the connection immediately */ | |
1357 | ip_vs_conn_expire_now(cp); | |
1da177e4 | 1358 | } |
dc8103f2 JA |
1359 | /* don't restart its timer, and silently |
1360 | drop the packet. */ | |
1361 | __ip_vs_conn_put(cp); | |
1da177e4 LT |
1362 | return NF_DROP; |
1363 | } | |
1364 | ||
1365 | ip_vs_in_stats(cp, skb); | |
1366 | restart = ip_vs_set_state(cp, IP_VS_DIR_INPUT, skb, pp); | |
1367 | if (cp->packet_xmit) | |
1368 | ret = cp->packet_xmit(skb, cp, pp); | |
1369 | /* do not touch skb anymore */ | |
1370 | else { | |
1371 | IP_VS_DBG_RL("warning: packet_xmit is null"); | |
1372 | ret = NF_ACCEPT; | |
1373 | } | |
1374 | ||
efac5276 RB |
1375 | /* Increase its packet counter and check if it is needed |
1376 | * to be synchronized | |
1377 | * | |
1378 | * Sync connection if it is about to close to | |
1379 | * encorage the standby servers to update the connections timeout | |
1380 | */ | |
1e66dafc | 1381 | pkts = atomic_add_return(1, &cp->in_pkts); |
2906f66a VMR |
1382 | if (af == AF_INET && (ip_vs_sync_state & IP_VS_STATE_MASTER) && |
1383 | cp->protocol == IPPROTO_SCTP) { | |
1384 | if ((cp->state == IP_VS_SCTP_S_ESTABLISHED && | |
1385 | (atomic_read(&cp->in_pkts) % | |
1386 | sysctl_ip_vs_sync_threshold[1] | |
1387 | == sysctl_ip_vs_sync_threshold[0])) || | |
1388 | (cp->old_state != cp->state && | |
1389 | ((cp->state == IP_VS_SCTP_S_CLOSED) || | |
1390 | (cp->state == IP_VS_SCTP_S_SHUT_ACK_CLI) || | |
1391 | (cp->state == IP_VS_SCTP_S_SHUT_ACK_SER)))) { | |
1392 | ip_vs_sync_conn(cp); | |
1393 | goto out; | |
1394 | } | |
1395 | } | |
1396 | ||
c6883f58 JV |
1397 | if (af == AF_INET && |
1398 | (ip_vs_sync_state & IP_VS_STATE_MASTER) && | |
efac5276 RB |
1399 | (((cp->protocol != IPPROTO_TCP || |
1400 | cp->state == IP_VS_TCP_S_ESTABLISHED) && | |
1e66dafc | 1401 | (pkts % sysctl_ip_vs_sync_threshold[1] |
efac5276 RB |
1402 | == sysctl_ip_vs_sync_threshold[0])) || |
1403 | ((cp->protocol == IPPROTO_TCP) && (cp->old_state != cp->state) && | |
1404 | ((cp->state == IP_VS_TCP_S_FIN_WAIT) || | |
9abfe315 | 1405 | (cp->state == IP_VS_TCP_S_CLOSE) || |
9d3a0de7 RB |
1406 | (cp->state == IP_VS_TCP_S_CLOSE_WAIT) || |
1407 | (cp->state == IP_VS_TCP_S_TIME_WAIT))))) | |
1da177e4 | 1408 | ip_vs_sync_conn(cp); |
2906f66a | 1409 | out: |
efac5276 | 1410 | cp->old_state = cp->state; |
1da177e4 LT |
1411 | |
1412 | ip_vs_conn_put(cp); | |
1413 | return ret; | |
1414 | } | |
1415 | ||
1416 | ||
1417 | /* | |
6e23ae2a | 1418 | * It is hooked at the NF_INET_FORWARD chain, in order to catch ICMP |
1da177e4 LT |
1419 | * related packets destined for 0.0.0.0/0. |
1420 | * When fwmark-based virtual service is used, such as transparent | |
1421 | * cache cluster, TCP packets can be marked and routed to ip_vs_in, | |
1422 | * but ICMP destined for 0.0.0.0/0 cannot not be easily marked and | |
6e23ae2a | 1423 | * sent to ip_vs_in_icmp. So, catch them at the NF_INET_FORWARD chain |
1da177e4 LT |
1424 | * and send them to ip_vs_in_icmp. |
1425 | */ | |
1426 | static unsigned int | |
3db05fea | 1427 | ip_vs_forward_icmp(unsigned int hooknum, struct sk_buff *skb, |
1da177e4 LT |
1428 | const struct net_device *in, const struct net_device *out, |
1429 | int (*okfn)(struct sk_buff *)) | |
1430 | { | |
1431 | int r; | |
1432 | ||
3db05fea | 1433 | if (ip_hdr(skb)->protocol != IPPROTO_ICMP) |
1da177e4 LT |
1434 | return NF_ACCEPT; |
1435 | ||
3db05fea | 1436 | return ip_vs_in_icmp(skb, &r, hooknum); |
1da177e4 LT |
1437 | } |
1438 | ||
2a3b791e JV |
1439 | #ifdef CONFIG_IP_VS_IPV6 |
1440 | static unsigned int | |
1441 | ip_vs_forward_icmp_v6(unsigned int hooknum, struct sk_buff *skb, | |
1442 | const struct net_device *in, const struct net_device *out, | |
1443 | int (*okfn)(struct sk_buff *)) | |
1444 | { | |
1445 | int r; | |
1446 | ||
1447 | if (ipv6_hdr(skb)->nexthdr != IPPROTO_ICMPV6) | |
1448 | return NF_ACCEPT; | |
1449 | ||
1450 | return ip_vs_in_icmp_v6(skb, &r, hooknum); | |
1451 | } | |
1452 | #endif | |
1453 | ||
1da177e4 | 1454 | |
1999414a | 1455 | static struct nf_hook_ops ip_vs_ops[] __read_mostly = { |
41c5b317 PM |
1456 | /* After packet filtering, forward packet through VS/DR, VS/TUN, |
1457 | * or VS/NAT(change destination), so that filtering rules can be | |
1458 | * applied to IPVS. */ | |
1459 | { | |
1460 | .hook = ip_vs_in, | |
1461 | .owner = THIS_MODULE, | |
1462 | .pf = PF_INET, | |
1463 | .hooknum = NF_INET_LOCAL_IN, | |
1464 | .priority = 100, | |
1465 | }, | |
1466 | /* After packet filtering, change source only for VS/NAT */ | |
1467 | { | |
1468 | .hook = ip_vs_out, | |
1469 | .owner = THIS_MODULE, | |
1470 | .pf = PF_INET, | |
1471 | .hooknum = NF_INET_FORWARD, | |
1472 | .priority = 100, | |
1473 | }, | |
1474 | /* After packet filtering (but before ip_vs_out_icmp), catch icmp | |
1475 | * destined for 0.0.0.0/0, which is for incoming IPVS connections */ | |
1476 | { | |
1477 | .hook = ip_vs_forward_icmp, | |
1478 | .owner = THIS_MODULE, | |
1479 | .pf = PF_INET, | |
1480 | .hooknum = NF_INET_FORWARD, | |
1481 | .priority = 99, | |
1482 | }, | |
473b23d3 JV |
1483 | #ifdef CONFIG_IP_VS_IPV6 |
1484 | /* After packet filtering, forward packet through VS/DR, VS/TUN, | |
1485 | * or VS/NAT(change destination), so that filtering rules can be | |
1486 | * applied to IPVS. */ | |
1487 | { | |
1488 | .hook = ip_vs_in, | |
1489 | .owner = THIS_MODULE, | |
1490 | .pf = PF_INET6, | |
1491 | .hooknum = NF_INET_LOCAL_IN, | |
1492 | .priority = 100, | |
1493 | }, | |
1494 | /* After packet filtering, change source only for VS/NAT */ | |
1495 | { | |
1496 | .hook = ip_vs_out, | |
1497 | .owner = THIS_MODULE, | |
1498 | .pf = PF_INET6, | |
1499 | .hooknum = NF_INET_FORWARD, | |
1500 | .priority = 100, | |
1501 | }, | |
1502 | /* After packet filtering (but before ip_vs_out_icmp), catch icmp | |
1503 | * destined for 0.0.0.0/0, which is for incoming IPVS connections */ | |
1504 | { | |
1505 | .hook = ip_vs_forward_icmp_v6, | |
1506 | .owner = THIS_MODULE, | |
1507 | .pf = PF_INET6, | |
1508 | .hooknum = NF_INET_FORWARD, | |
1509 | .priority = 99, | |
1510 | }, | |
473b23d3 | 1511 | #endif |
1da177e4 LT |
1512 | }; |
1513 | ||
1514 | ||
1515 | /* | |
1516 | * Initialize IP Virtual Server | |
1517 | */ | |
1518 | static int __init ip_vs_init(void) | |
1519 | { | |
1520 | int ret; | |
1521 | ||
a919cf4b SW |
1522 | ip_vs_estimator_init(); |
1523 | ||
1da177e4 LT |
1524 | ret = ip_vs_control_init(); |
1525 | if (ret < 0) { | |
1e3e238e | 1526 | pr_err("can't setup control.\n"); |
a919cf4b | 1527 | goto cleanup_estimator; |
1da177e4 LT |
1528 | } |
1529 | ||
1530 | ip_vs_protocol_init(); | |
1531 | ||
1532 | ret = ip_vs_app_init(); | |
1533 | if (ret < 0) { | |
1e3e238e | 1534 | pr_err("can't setup application helper.\n"); |
1da177e4 LT |
1535 | goto cleanup_protocol; |
1536 | } | |
1537 | ||
1538 | ret = ip_vs_conn_init(); | |
1539 | if (ret < 0) { | |
1e3e238e | 1540 | pr_err("can't setup connection table.\n"); |
1da177e4 LT |
1541 | goto cleanup_app; |
1542 | } | |
1543 | ||
41c5b317 | 1544 | ret = nf_register_hooks(ip_vs_ops, ARRAY_SIZE(ip_vs_ops)); |
1da177e4 | 1545 | if (ret < 0) { |
1e3e238e | 1546 | pr_err("can't register hooks.\n"); |
1da177e4 LT |
1547 | goto cleanup_conn; |
1548 | } | |
1549 | ||
1e3e238e | 1550 | pr_info("ipvs loaded.\n"); |
1da177e4 LT |
1551 | return ret; |
1552 | ||
1da177e4 LT |
1553 | cleanup_conn: |
1554 | ip_vs_conn_cleanup(); | |
1555 | cleanup_app: | |
1556 | ip_vs_app_cleanup(); | |
1557 | cleanup_protocol: | |
1558 | ip_vs_protocol_cleanup(); | |
1559 | ip_vs_control_cleanup(); | |
a919cf4b SW |
1560 | cleanup_estimator: |
1561 | ip_vs_estimator_cleanup(); | |
1da177e4 LT |
1562 | return ret; |
1563 | } | |
1564 | ||
1565 | static void __exit ip_vs_cleanup(void) | |
1566 | { | |
41c5b317 | 1567 | nf_unregister_hooks(ip_vs_ops, ARRAY_SIZE(ip_vs_ops)); |
1da177e4 LT |
1568 | ip_vs_conn_cleanup(); |
1569 | ip_vs_app_cleanup(); | |
1570 | ip_vs_protocol_cleanup(); | |
1571 | ip_vs_control_cleanup(); | |
a919cf4b | 1572 | ip_vs_estimator_cleanup(); |
1e3e238e | 1573 | pr_info("ipvs unloaded.\n"); |
1da177e4 LT |
1574 | } |
1575 | ||
1576 | module_init(ip_vs_init); | |
1577 | module_exit(ip_vs_cleanup); | |
1578 | MODULE_LICENSE("GPL"); |