]>
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> | |
2c70b519 | 43 | #include <net/ip6_checksum.h> |
1da177e4 LT |
44 | |
45 | #include <linux/netfilter.h> | |
46 | #include <linux/netfilter_ipv4.h> | |
47 | ||
2a3b791e JV |
48 | #ifdef CONFIG_IP_VS_IPV6 |
49 | #include <net/ipv6.h> | |
50 | #include <linux/netfilter_ipv6.h> | |
51 | #endif | |
52 | ||
1da177e4 LT |
53 | #include <net/ip_vs.h> |
54 | ||
55 | ||
56 | EXPORT_SYMBOL(register_ip_vs_scheduler); | |
57 | EXPORT_SYMBOL(unregister_ip_vs_scheduler); | |
1da177e4 LT |
58 | EXPORT_SYMBOL(ip_vs_proto_name); |
59 | EXPORT_SYMBOL(ip_vs_conn_new); | |
60 | EXPORT_SYMBOL(ip_vs_conn_in_get); | |
61 | EXPORT_SYMBOL(ip_vs_conn_out_get); | |
62 | #ifdef CONFIG_IP_VS_PROTO_TCP | |
63 | EXPORT_SYMBOL(ip_vs_tcp_conn_listen); | |
64 | #endif | |
65 | EXPORT_SYMBOL(ip_vs_conn_put); | |
66 | #ifdef CONFIG_IP_VS_DEBUG | |
67 | EXPORT_SYMBOL(ip_vs_get_debug_level); | |
68 | #endif | |
1da177e4 LT |
69 | |
70 | ||
71 | /* ID used in ICMP lookups */ | |
72 | #define icmp_id(icmph) (((icmph)->un).echo.id) | |
2a3b791e | 73 | #define icmpv6_id(icmph) (icmph->icmp6_dataun.u_echo.identifier) |
1da177e4 LT |
74 | |
75 | const char *ip_vs_proto_name(unsigned proto) | |
76 | { | |
77 | static char buf[20]; | |
78 | ||
79 | switch (proto) { | |
80 | case IPPROTO_IP: | |
81 | return "IP"; | |
82 | case IPPROTO_UDP: | |
83 | return "UDP"; | |
84 | case IPPROTO_TCP: | |
85 | return "TCP"; | |
2906f66a VMR |
86 | case IPPROTO_SCTP: |
87 | return "SCTP"; | |
1da177e4 LT |
88 | case IPPROTO_ICMP: |
89 | return "ICMP"; | |
2a3b791e JV |
90 | #ifdef CONFIG_IP_VS_IPV6 |
91 | case IPPROTO_ICMPV6: | |
92 | return "ICMPv6"; | |
93 | #endif | |
1da177e4 LT |
94 | default: |
95 | sprintf(buf, "IP_%d", proto); | |
96 | return buf; | |
97 | } | |
98 | } | |
99 | ||
100 | void ip_vs_init_hash_table(struct list_head *table, int rows) | |
101 | { | |
102 | while (--rows >= 0) | |
103 | INIT_LIST_HEAD(&table[rows]); | |
104 | } | |
105 | ||
106 | static inline void | |
107 | ip_vs_in_stats(struct ip_vs_conn *cp, struct sk_buff *skb) | |
108 | { | |
109 | struct ip_vs_dest *dest = cp->dest; | |
110 | if (dest && (dest->flags & IP_VS_DEST_F_AVAILABLE)) { | |
111 | spin_lock(&dest->stats.lock); | |
e9c0ce23 SW |
112 | dest->stats.ustats.inpkts++; |
113 | dest->stats.ustats.inbytes += skb->len; | |
1da177e4 LT |
114 | spin_unlock(&dest->stats.lock); |
115 | ||
116 | spin_lock(&dest->svc->stats.lock); | |
e9c0ce23 SW |
117 | dest->svc->stats.ustats.inpkts++; |
118 | dest->svc->stats.ustats.inbytes += skb->len; | |
1da177e4 LT |
119 | spin_unlock(&dest->svc->stats.lock); |
120 | ||
121 | spin_lock(&ip_vs_stats.lock); | |
e9c0ce23 SW |
122 | ip_vs_stats.ustats.inpkts++; |
123 | ip_vs_stats.ustats.inbytes += skb->len; | |
1da177e4 LT |
124 | spin_unlock(&ip_vs_stats.lock); |
125 | } | |
126 | } | |
127 | ||
128 | ||
129 | static inline void | |
130 | ip_vs_out_stats(struct ip_vs_conn *cp, struct sk_buff *skb) | |
131 | { | |
132 | struct ip_vs_dest *dest = cp->dest; | |
133 | if (dest && (dest->flags & IP_VS_DEST_F_AVAILABLE)) { | |
134 | spin_lock(&dest->stats.lock); | |
e9c0ce23 SW |
135 | dest->stats.ustats.outpkts++; |
136 | dest->stats.ustats.outbytes += skb->len; | |
1da177e4 LT |
137 | spin_unlock(&dest->stats.lock); |
138 | ||
139 | spin_lock(&dest->svc->stats.lock); | |
e9c0ce23 SW |
140 | dest->svc->stats.ustats.outpkts++; |
141 | dest->svc->stats.ustats.outbytes += skb->len; | |
1da177e4 LT |
142 | spin_unlock(&dest->svc->stats.lock); |
143 | ||
144 | spin_lock(&ip_vs_stats.lock); | |
e9c0ce23 SW |
145 | ip_vs_stats.ustats.outpkts++; |
146 | ip_vs_stats.ustats.outbytes += skb->len; | |
1da177e4 LT |
147 | spin_unlock(&ip_vs_stats.lock); |
148 | } | |
149 | } | |
150 | ||
151 | ||
152 | static inline void | |
153 | ip_vs_conn_stats(struct ip_vs_conn *cp, struct ip_vs_service *svc) | |
154 | { | |
155 | spin_lock(&cp->dest->stats.lock); | |
e9c0ce23 | 156 | cp->dest->stats.ustats.conns++; |
1da177e4 LT |
157 | spin_unlock(&cp->dest->stats.lock); |
158 | ||
159 | spin_lock(&svc->stats.lock); | |
e9c0ce23 | 160 | svc->stats.ustats.conns++; |
1da177e4 LT |
161 | spin_unlock(&svc->stats.lock); |
162 | ||
163 | spin_lock(&ip_vs_stats.lock); | |
e9c0ce23 | 164 | ip_vs_stats.ustats.conns++; |
1da177e4 LT |
165 | spin_unlock(&ip_vs_stats.lock); |
166 | } | |
167 | ||
168 | ||
169 | static inline int | |
170 | ip_vs_set_state(struct ip_vs_conn *cp, int direction, | |
171 | const struct sk_buff *skb, | |
172 | struct ip_vs_protocol *pp) | |
173 | { | |
174 | if (unlikely(!pp->state_transition)) | |
175 | return 0; | |
176 | return pp->state_transition(cp, direction, skb, pp); | |
177 | } | |
178 | ||
179 | ||
1da177e4 LT |
180 | /* |
181 | * IPVS persistent scheduling function | |
182 | * It creates a connection entry according to its template if exists, | |
183 | * or selects a server and creates a connection entry plus a template. | |
184 | * Locking: we are svc user (svc->refcnt), so we hold all dests too | |
185 | * Protocols supported: TCP, UDP | |
186 | */ | |
187 | static struct ip_vs_conn * | |
188 | ip_vs_sched_persist(struct ip_vs_service *svc, | |
189 | const struct sk_buff *skb, | |
014d730d | 190 | __be16 ports[2]) |
1da177e4 LT |
191 | { |
192 | struct ip_vs_conn *cp = NULL; | |
28364a59 | 193 | struct ip_vs_iphdr iph; |
1da177e4 LT |
194 | struct ip_vs_dest *dest; |
195 | struct ip_vs_conn *ct; | |
cd17f9ed | 196 | __be16 dport; /* destination port to forward */ |
26ec037f | 197 | __be16 flags; |
28364a59 JV |
198 | union nf_inet_addr snet; /* source network of the client, |
199 | after masking */ | |
cd17f9ed JV |
200 | |
201 | ip_vs_fill_iphdr(svc->af, skb_network_header(skb), &iph); | |
1da177e4 LT |
202 | |
203 | /* Mask saddr with the netmask to adjust template granularity */ | |
cd17f9ed JV |
204 | #ifdef CONFIG_IP_VS_IPV6 |
205 | if (svc->af == AF_INET6) | |
206 | ipv6_addr_prefix(&snet.in6, &iph.saddr.in6, svc->netmask); | |
207 | else | |
208 | #endif | |
209 | snet.ip = iph.saddr.ip & svc->netmask; | |
1da177e4 | 210 | |
cd17f9ed JV |
211 | IP_VS_DBG_BUF(6, "p-schedule: src %s:%u dest %s:%u " |
212 | "mnet %s\n", | |
213 | IP_VS_DBG_ADDR(svc->af, &iph.saddr), ntohs(ports[0]), | |
214 | IP_VS_DBG_ADDR(svc->af, &iph.daddr), ntohs(ports[1]), | |
215 | IP_VS_DBG_ADDR(svc->af, &snet)); | |
1da177e4 LT |
216 | |
217 | /* | |
218 | * As far as we know, FTP is a very complicated network protocol, and | |
219 | * it uses control connection and data connections. For active FTP, | |
220 | * FTP server initialize data connection to the client, its source port | |
221 | * is often 20. For passive FTP, FTP server tells the clients the port | |
222 | * that it passively listens to, and the client issues the data | |
223 | * connection. In the tunneling or direct routing mode, the load | |
224 | * balancer is on the client-to-server half of connection, the port | |
225 | * number is unknown to the load balancer. So, a conn template like | |
226 | * <caddr, 0, vaddr, 0, daddr, 0> is created for persistent FTP | |
227 | * service, and a template like <caddr, 0, vaddr, vport, daddr, dport> | |
228 | * is created for other persistent services. | |
229 | */ | |
230 | if (ports[1] == svc->port) { | |
231 | /* Check if a template already exists */ | |
232 | if (svc->port != FTPPORT) | |
cd17f9ed | 233 | ct = ip_vs_ct_in_get(svc->af, iph.protocol, &snet, 0, |
28364a59 | 234 | &iph.daddr, ports[1]); |
1da177e4 | 235 | else |
cd17f9ed | 236 | ct = ip_vs_ct_in_get(svc->af, iph.protocol, &snet, 0, |
28364a59 | 237 | &iph.daddr, 0); |
1da177e4 LT |
238 | |
239 | if (!ct || !ip_vs_check_template(ct)) { | |
240 | /* | |
241 | * No template found or the dest of the connection | |
242 | * template is not available. | |
243 | */ | |
244 | dest = svc->scheduler->schedule(svc, skb); | |
245 | if (dest == NULL) { | |
246 | IP_VS_DBG(1, "p-schedule: no dest found.\n"); | |
247 | return NULL; | |
248 | } | |
249 | ||
250 | /* | |
251 | * Create a template like <protocol,caddr,0, | |
252 | * vaddr,vport,daddr,dport> for non-ftp service, | |
253 | * and <protocol,caddr,0,vaddr,0,daddr,0> | |
254 | * for ftp service. | |
255 | */ | |
256 | if (svc->port != FTPPORT) | |
cd17f9ed | 257 | ct = ip_vs_conn_new(svc->af, iph.protocol, |
28364a59 JV |
258 | &snet, 0, |
259 | &iph.daddr, | |
1da177e4 | 260 | ports[1], |
28364a59 | 261 | &dest->addr, dest->port, |
87375ab4 | 262 | IP_VS_CONN_F_TEMPLATE, |
1da177e4 LT |
263 | dest); |
264 | else | |
cd17f9ed | 265 | ct = ip_vs_conn_new(svc->af, iph.protocol, |
28364a59 JV |
266 | &snet, 0, |
267 | &iph.daddr, 0, | |
268 | &dest->addr, 0, | |
87375ab4 | 269 | IP_VS_CONN_F_TEMPLATE, |
1da177e4 LT |
270 | dest); |
271 | if (ct == NULL) | |
272 | return NULL; | |
273 | ||
274 | ct->timeout = svc->timeout; | |
275 | } else { | |
276 | /* set destination with the found template */ | |
277 | dest = ct->dest; | |
278 | } | |
279 | dport = dest->port; | |
280 | } else { | |
281 | /* | |
282 | * Note: persistent fwmark-based services and persistent | |
283 | * port zero service are handled here. | |
284 | * fwmark template: <IPPROTO_IP,caddr,0,fwmark,0,daddr,0> | |
285 | * port zero template: <protocol,caddr,0,vaddr,0,daddr,0> | |
286 | */ | |
28364a59 JV |
287 | if (svc->fwmark) { |
288 | union nf_inet_addr fwmark = { | |
be8be9ec | 289 | .ip = htonl(svc->fwmark) |
28364a59 JV |
290 | }; |
291 | ||
cd17f9ed | 292 | ct = ip_vs_ct_in_get(svc->af, IPPROTO_IP, &snet, 0, |
28364a59 JV |
293 | &fwmark, 0); |
294 | } else | |
cd17f9ed | 295 | ct = ip_vs_ct_in_get(svc->af, iph.protocol, &snet, 0, |
28364a59 | 296 | &iph.daddr, 0); |
1da177e4 LT |
297 | |
298 | if (!ct || !ip_vs_check_template(ct)) { | |
299 | /* | |
300 | * If it is not persistent port zero, return NULL, | |
301 | * otherwise create a connection template. | |
302 | */ | |
303 | if (svc->port) | |
304 | return NULL; | |
305 | ||
306 | dest = svc->scheduler->schedule(svc, skb); | |
307 | if (dest == NULL) { | |
308 | IP_VS_DBG(1, "p-schedule: no dest found.\n"); | |
309 | return NULL; | |
310 | } | |
311 | ||
312 | /* | |
313 | * Create a template according to the service | |
314 | */ | |
28364a59 JV |
315 | if (svc->fwmark) { |
316 | union nf_inet_addr fwmark = { | |
be8be9ec | 317 | .ip = htonl(svc->fwmark) |
28364a59 JV |
318 | }; |
319 | ||
cd17f9ed | 320 | ct = ip_vs_conn_new(svc->af, IPPROTO_IP, |
28364a59 JV |
321 | &snet, 0, |
322 | &fwmark, 0, | |
323 | &dest->addr, 0, | |
87375ab4 | 324 | IP_VS_CONN_F_TEMPLATE, |
1da177e4 | 325 | dest); |
28364a59 | 326 | } else |
cd17f9ed | 327 | ct = ip_vs_conn_new(svc->af, iph.protocol, |
28364a59 JV |
328 | &snet, 0, |
329 | &iph.daddr, 0, | |
330 | &dest->addr, 0, | |
87375ab4 | 331 | IP_VS_CONN_F_TEMPLATE, |
1da177e4 LT |
332 | dest); |
333 | if (ct == NULL) | |
334 | return NULL; | |
335 | ||
336 | ct->timeout = svc->timeout; | |
337 | } else { | |
338 | /* set destination with the found template */ | |
339 | dest = ct->dest; | |
340 | } | |
341 | dport = ports[1]; | |
342 | } | |
343 | ||
26ec037f NC |
344 | flags = (svc->flags & IP_VS_SVC_F_ONEPACKET |
345 | && iph.protocol == IPPROTO_UDP)? | |
346 | IP_VS_CONN_F_ONE_PACKET : 0; | |
347 | ||
1da177e4 LT |
348 | /* |
349 | * Create a new connection according to the template | |
350 | */ | |
cd17f9ed | 351 | cp = ip_vs_conn_new(svc->af, iph.protocol, |
28364a59 JV |
352 | &iph.saddr, ports[0], |
353 | &iph.daddr, ports[1], | |
354 | &dest->addr, dport, | |
26ec037f | 355 | flags, |
1da177e4 LT |
356 | dest); |
357 | if (cp == NULL) { | |
358 | ip_vs_conn_put(ct); | |
359 | return NULL; | |
360 | } | |
361 | ||
362 | /* | |
363 | * Add its control | |
364 | */ | |
365 | ip_vs_control_add(cp, ct); | |
366 | ip_vs_conn_put(ct); | |
367 | ||
368 | ip_vs_conn_stats(cp, svc); | |
369 | return cp; | |
370 | } | |
371 | ||
372 | ||
373 | /* | |
374 | * IPVS main scheduling function | |
375 | * It selects a server according to the virtual service, and | |
376 | * creates a connection entry. | |
377 | * Protocols supported: TCP, UDP | |
378 | */ | |
379 | struct ip_vs_conn * | |
380 | ip_vs_schedule(struct ip_vs_service *svc, const struct sk_buff *skb) | |
381 | { | |
382 | struct ip_vs_conn *cp = NULL; | |
28364a59 | 383 | struct ip_vs_iphdr iph; |
1da177e4 | 384 | struct ip_vs_dest *dest; |
26ec037f | 385 | __be16 _ports[2], *pptr, flags; |
1da177e4 | 386 | |
28364a59 JV |
387 | ip_vs_fill_iphdr(svc->af, skb_network_header(skb), &iph); |
388 | pptr = skb_header_pointer(skb, iph.len, sizeof(_ports), _ports); | |
1da177e4 LT |
389 | if (pptr == NULL) |
390 | return NULL; | |
391 | ||
392 | /* | |
393 | * Persistent service | |
394 | */ | |
395 | if (svc->flags & IP_VS_SVC_F_PERSISTENT) | |
396 | return ip_vs_sched_persist(svc, skb, pptr); | |
397 | ||
398 | /* | |
399 | * Non-persistent service | |
400 | */ | |
401 | if (!svc->fwmark && pptr[1] != svc->port) { | |
402 | if (!svc->port) | |
1e3e238e HE |
403 | pr_err("Schedule: port zero only supported " |
404 | "in persistent services, " | |
405 | "check your ipvs configuration\n"); | |
1da177e4 LT |
406 | return NULL; |
407 | } | |
408 | ||
409 | dest = svc->scheduler->schedule(svc, skb); | |
410 | if (dest == NULL) { | |
411 | IP_VS_DBG(1, "Schedule: no dest found.\n"); | |
412 | return NULL; | |
413 | } | |
414 | ||
26ec037f NC |
415 | flags = (svc->flags & IP_VS_SVC_F_ONEPACKET |
416 | && iph.protocol == IPPROTO_UDP)? | |
417 | IP_VS_CONN_F_ONE_PACKET : 0; | |
418 | ||
1da177e4 LT |
419 | /* |
420 | * Create a connection entry. | |
421 | */ | |
cd17f9ed | 422 | cp = ip_vs_conn_new(svc->af, iph.protocol, |
28364a59 JV |
423 | &iph.saddr, pptr[0], |
424 | &iph.daddr, pptr[1], | |
425 | &dest->addr, dest->port ? dest->port : pptr[1], | |
26ec037f | 426 | flags, |
1da177e4 LT |
427 | dest); |
428 | if (cp == NULL) | |
429 | return NULL; | |
430 | ||
cd17f9ed JV |
431 | IP_VS_DBG_BUF(6, "Schedule fwd:%c c:%s:%u v:%s:%u " |
432 | "d:%s:%u conn->flags:%X conn->refcnt:%d\n", | |
433 | ip_vs_fwd_tag(cp), | |
434 | IP_VS_DBG_ADDR(svc->af, &cp->caddr), ntohs(cp->cport), | |
435 | IP_VS_DBG_ADDR(svc->af, &cp->vaddr), ntohs(cp->vport), | |
436 | IP_VS_DBG_ADDR(svc->af, &cp->daddr), ntohs(cp->dport), | |
437 | cp->flags, atomic_read(&cp->refcnt)); | |
1da177e4 LT |
438 | |
439 | ip_vs_conn_stats(cp, svc); | |
440 | return cp; | |
441 | } | |
442 | ||
443 | ||
444 | /* | |
445 | * Pass or drop the packet. | |
446 | * Called by ip_vs_in, when the virtual service is available but | |
447 | * no destination is available for a new connection. | |
448 | */ | |
449 | int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb, | |
450 | struct ip_vs_protocol *pp) | |
451 | { | |
014d730d | 452 | __be16 _ports[2], *pptr; |
28364a59 | 453 | struct ip_vs_iphdr iph; |
2a3b791e JV |
454 | int unicast; |
455 | ip_vs_fill_iphdr(svc->af, skb_network_header(skb), &iph); | |
1da177e4 | 456 | |
28364a59 | 457 | pptr = skb_header_pointer(skb, iph.len, sizeof(_ports), _ports); |
1da177e4 LT |
458 | if (pptr == NULL) { |
459 | ip_vs_service_put(svc); | |
460 | return NF_DROP; | |
461 | } | |
462 | ||
2a3b791e JV |
463 | #ifdef CONFIG_IP_VS_IPV6 |
464 | if (svc->af == AF_INET6) | |
465 | unicast = ipv6_addr_type(&iph.daddr.in6) & IPV6_ADDR_UNICAST; | |
466 | else | |
467 | #endif | |
468 | unicast = (inet_addr_type(&init_net, iph.daddr.ip) == RTN_UNICAST); | |
469 | ||
1da177e4 | 470 | /* if it is fwmark-based service, the cache_bypass sysctl is up |
2a3b791e | 471 | and the destination is a non-local unicast, then create |
1da177e4 | 472 | a cache_bypass connection entry */ |
2a3b791e | 473 | if (sysctl_ip_vs_cache_bypass && svc->fwmark && unicast) { |
1da177e4 LT |
474 | int ret, cs; |
475 | struct ip_vs_conn *cp; | |
26ec037f NC |
476 | __u16 flags = (svc->flags & IP_VS_SVC_F_ONEPACKET && |
477 | iph.protocol == IPPROTO_UDP)? | |
478 | IP_VS_CONN_F_ONE_PACKET : 0; | |
dff630dd | 479 | union nf_inet_addr daddr = { .all = { 0, 0, 0, 0 } }; |
1da177e4 LT |
480 | |
481 | ip_vs_service_put(svc); | |
482 | ||
483 | /* create a new connection entry */ | |
1e3e238e | 484 | IP_VS_DBG(6, "%s(): create a cache_bypass entry\n", __func__); |
2a3b791e | 485 | cp = ip_vs_conn_new(svc->af, iph.protocol, |
28364a59 JV |
486 | &iph.saddr, pptr[0], |
487 | &iph.daddr, pptr[1], | |
dff630dd | 488 | &daddr, 0, |
26ec037f | 489 | IP_VS_CONN_F_BYPASS | flags, |
1da177e4 LT |
490 | NULL); |
491 | if (cp == NULL) | |
492 | return NF_DROP; | |
493 | ||
494 | /* statistics */ | |
495 | ip_vs_in_stats(cp, skb); | |
496 | ||
497 | /* set state */ | |
498 | cs = ip_vs_set_state(cp, IP_VS_DIR_INPUT, skb, pp); | |
499 | ||
500 | /* transmit the first SYN packet */ | |
501 | ret = cp->packet_xmit(skb, cp, pp); | |
502 | /* do not touch skb anymore */ | |
503 | ||
504 | atomic_inc(&cp->in_pkts); | |
505 | ip_vs_conn_put(cp); | |
506 | return ret; | |
507 | } | |
508 | ||
509 | /* | |
510 | * When the virtual ftp service is presented, packets destined | |
511 | * for other services on the VIP may get here (except services | |
512 | * listed in the ipvs table), pass the packets, because it is | |
513 | * not ipvs job to decide to drop the packets. | |
514 | */ | |
515 | if ((svc->port == FTPPORT) && (pptr[1] != FTPPORT)) { | |
516 | ip_vs_service_put(svc); | |
517 | return NF_ACCEPT; | |
518 | } | |
519 | ||
520 | ip_vs_service_put(svc); | |
521 | ||
522 | /* | |
523 | * Notify the client that the destination is unreachable, and | |
524 | * release the socket buffer. | |
525 | * Since it is in IP layer, the TCP socket is not actually | |
526 | * created, the TCP RST packet cannot be sent, instead that | |
527 | * ICMP_PORT_UNREACH is sent here no matter it is TCP/UDP. --WZ | |
528 | */ | |
2a3b791e JV |
529 | #ifdef CONFIG_IP_VS_IPV6 |
530 | if (svc->af == AF_INET6) | |
3ffe533c | 531 | icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); |
2a3b791e JV |
532 | else |
533 | #endif | |
534 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); | |
535 | ||
1da177e4 LT |
536 | return NF_DROP; |
537 | } | |
538 | ||
b1550f22 | 539 | __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset) |
1da177e4 | 540 | { |
d3bc23e7 | 541 | return csum_fold(skb_checksum(skb, offset, skb->len - offset, 0)); |
1da177e4 LT |
542 | } |
543 | ||
776c729e | 544 | static inline int ip_vs_gather_frags(struct sk_buff *skb, u_int32_t user) |
1da177e4 | 545 | { |
776c729e HX |
546 | int err = ip_defrag(skb, user); |
547 | ||
548 | if (!err) | |
eddc9ec5 | 549 | ip_send_check(ip_hdr(skb)); |
776c729e HX |
550 | |
551 | return err; | |
1da177e4 LT |
552 | } |
553 | ||
2a3b791e JV |
554 | #ifdef CONFIG_IP_VS_IPV6 |
555 | static inline int ip_vs_gather_frags_v6(struct sk_buff *skb, u_int32_t user) | |
556 | { | |
557 | /* TODO IPv6: Find out what to do here for IPv6 */ | |
558 | return 0; | |
559 | } | |
560 | #endif | |
561 | ||
1da177e4 LT |
562 | /* |
563 | * Packet has been made sufficiently writable in caller | |
564 | * - inout: 1=in->out, 0=out->in | |
565 | */ | |
566 | void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp, | |
567 | struct ip_vs_conn *cp, int inout) | |
568 | { | |
eddc9ec5 | 569 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 | 570 | unsigned int icmp_offset = iph->ihl*4; |
d56f90a7 ACM |
571 | struct icmphdr *icmph = (struct icmphdr *)(skb_network_header(skb) + |
572 | icmp_offset); | |
1da177e4 LT |
573 | struct iphdr *ciph = (struct iphdr *)(icmph + 1); |
574 | ||
575 | if (inout) { | |
e7ade46a | 576 | iph->saddr = cp->vaddr.ip; |
1da177e4 | 577 | ip_send_check(iph); |
e7ade46a | 578 | ciph->daddr = cp->vaddr.ip; |
1da177e4 LT |
579 | ip_send_check(ciph); |
580 | } else { | |
e7ade46a | 581 | iph->daddr = cp->daddr.ip; |
1da177e4 | 582 | ip_send_check(iph); |
e7ade46a | 583 | ciph->saddr = cp->daddr.ip; |
1da177e4 LT |
584 | ip_send_check(ciph); |
585 | } | |
586 | ||
2906f66a VMR |
587 | /* the TCP/UDP/SCTP port */ |
588 | if (IPPROTO_TCP == ciph->protocol || IPPROTO_UDP == ciph->protocol || | |
589 | IPPROTO_SCTP == ciph->protocol) { | |
014d730d | 590 | __be16 *ports = (void *)ciph + ciph->ihl*4; |
1da177e4 LT |
591 | |
592 | if (inout) | |
593 | ports[1] = cp->vport; | |
594 | else | |
595 | ports[0] = cp->dport; | |
596 | } | |
597 | ||
598 | /* And finally the ICMP checksum */ | |
599 | icmph->checksum = 0; | |
600 | icmph->checksum = ip_vs_checksum_complete(skb, icmp_offset); | |
601 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
602 | ||
603 | if (inout) | |
604 | IP_VS_DBG_PKT(11, pp, skb, (void *)ciph - (void *)iph, | |
605 | "Forwarding altered outgoing ICMP"); | |
606 | else | |
607 | IP_VS_DBG_PKT(11, pp, skb, (void *)ciph - (void *)iph, | |
608 | "Forwarding altered incoming ICMP"); | |
609 | } | |
610 | ||
b3cdd2a7 JV |
611 | #ifdef CONFIG_IP_VS_IPV6 |
612 | void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp, | |
613 | struct ip_vs_conn *cp, int inout) | |
614 | { | |
615 | struct ipv6hdr *iph = ipv6_hdr(skb); | |
616 | unsigned int icmp_offset = sizeof(struct ipv6hdr); | |
617 | struct icmp6hdr *icmph = (struct icmp6hdr *)(skb_network_header(skb) + | |
618 | icmp_offset); | |
619 | struct ipv6hdr *ciph = (struct ipv6hdr *)(icmph + 1); | |
620 | ||
621 | if (inout) { | |
622 | iph->saddr = cp->vaddr.in6; | |
623 | ciph->daddr = cp->vaddr.in6; | |
624 | } else { | |
625 | iph->daddr = cp->daddr.in6; | |
626 | ciph->saddr = cp->daddr.in6; | |
627 | } | |
628 | ||
2906f66a VMR |
629 | /* the TCP/UDP/SCTP port */ |
630 | if (IPPROTO_TCP == ciph->nexthdr || IPPROTO_UDP == ciph->nexthdr || | |
631 | IPPROTO_SCTP == ciph->nexthdr) { | |
b3cdd2a7 JV |
632 | __be16 *ports = (void *)ciph + sizeof(struct ipv6hdr); |
633 | ||
634 | if (inout) | |
635 | ports[1] = cp->vport; | |
636 | else | |
637 | ports[0] = cp->dport; | |
638 | } | |
639 | ||
640 | /* And finally the ICMP checksum */ | |
8870f842 SH |
641 | icmph->icmp6_cksum = ~csum_ipv6_magic(&iph->saddr, &iph->daddr, |
642 | skb->len - icmp_offset, | |
643 | IPPROTO_ICMPV6, 0); | |
644 | skb->csum_start = skb_network_header(skb) - skb->head + icmp_offset; | |
645 | skb->csum_offset = offsetof(struct icmp6hdr, icmp6_cksum); | |
646 | skb->ip_summed = CHECKSUM_PARTIAL; | |
b3cdd2a7 JV |
647 | |
648 | if (inout) | |
649 | IP_VS_DBG_PKT(11, pp, skb, (void *)ciph - (void *)iph, | |
650 | "Forwarding altered outgoing ICMPv6"); | |
651 | else | |
652 | IP_VS_DBG_PKT(11, pp, skb, (void *)ciph - (void *)iph, | |
653 | "Forwarding altered incoming ICMPv6"); | |
654 | } | |
655 | #endif | |
656 | ||
4856c84c MT |
657 | /* Handle relevant response ICMP messages - forward to the right |
658 | * destination host. Used for NAT and local client. | |
659 | */ | |
f2428ed5 SH |
660 | static int handle_response_icmp(int af, struct sk_buff *skb, |
661 | union nf_inet_addr *snet, | |
662 | __u8 protocol, struct ip_vs_conn *cp, | |
4856c84c MT |
663 | struct ip_vs_protocol *pp, |
664 | unsigned int offset, unsigned int ihl) | |
665 | { | |
666 | unsigned int verdict = NF_DROP; | |
667 | ||
668 | if (IP_VS_FWD_METHOD(cp) != 0) { | |
1e3e238e HE |
669 | pr_err("shouldn't reach here, because the box is on the " |
670 | "half connection in the tun/dr module.\n"); | |
4856c84c MT |
671 | } |
672 | ||
673 | /* Ensure the checksum is correct */ | |
674 | if (!skb_csum_unnecessary(skb) && ip_vs_checksum_complete(skb, ihl)) { | |
675 | /* Failed checksum! */ | |
f2428ed5 SH |
676 | IP_VS_DBG_BUF(1, "Forward ICMP: failed checksum from %s!\n", |
677 | IP_VS_DBG_ADDR(af, snet)); | |
4856c84c MT |
678 | goto out; |
679 | } | |
680 | ||
2906f66a VMR |
681 | if (IPPROTO_TCP == protocol || IPPROTO_UDP == protocol || |
682 | IPPROTO_SCTP == protocol) | |
4856c84c MT |
683 | offset += 2 * sizeof(__u16); |
684 | if (!skb_make_writable(skb, offset)) | |
685 | goto out; | |
686 | ||
f2428ed5 SH |
687 | #ifdef CONFIG_IP_VS_IPV6 |
688 | if (af == AF_INET6) | |
689 | ip_vs_nat_icmp_v6(skb, pp, cp, 1); | |
690 | else | |
691 | #endif | |
692 | ip_vs_nat_icmp(skb, pp, cp, 1); | |
4856c84c MT |
693 | |
694 | /* do the statistics and put it back */ | |
695 | ip_vs_out_stats(cp, skb); | |
696 | ||
697 | skb->ipvs_property = 1; | |
698 | verdict = NF_ACCEPT; | |
699 | ||
700 | out: | |
701 | __ip_vs_conn_put(cp); | |
702 | ||
703 | return verdict; | |
704 | } | |
705 | ||
1da177e4 LT |
706 | /* |
707 | * Handle ICMP messages in the inside-to-outside direction (outgoing). | |
4856c84c | 708 | * Find any that might be relevant, check against existing connections. |
1da177e4 | 709 | * Currently handles error types - unreachable, quench, ttl exceeded. |
1da177e4 | 710 | */ |
3db05fea | 711 | static int ip_vs_out_icmp(struct sk_buff *skb, int *related) |
1da177e4 | 712 | { |
1da177e4 LT |
713 | struct iphdr *iph; |
714 | struct icmphdr _icmph, *ic; | |
715 | struct iphdr _ciph, *cih; /* The ip header contained within the ICMP */ | |
51ef348b | 716 | struct ip_vs_iphdr ciph; |
1da177e4 LT |
717 | struct ip_vs_conn *cp; |
718 | struct ip_vs_protocol *pp; | |
4856c84c | 719 | unsigned int offset, ihl; |
f2428ed5 | 720 | union nf_inet_addr snet; |
1da177e4 LT |
721 | |
722 | *related = 1; | |
723 | ||
724 | /* reassemble IP fragments */ | |
eddc9ec5 | 725 | if (ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)) { |
776c729e | 726 | if (ip_vs_gather_frags(skb, IP_DEFRAG_VS_OUT)) |
1da177e4 | 727 | return NF_STOLEN; |
1da177e4 LT |
728 | } |
729 | ||
eddc9ec5 | 730 | iph = ip_hdr(skb); |
1da177e4 LT |
731 | offset = ihl = iph->ihl * 4; |
732 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
733 | if (ic == NULL) | |
734 | return NF_DROP; | |
735 | ||
14d5e834 | 736 | IP_VS_DBG(12, "Outgoing ICMP (%d,%d) %pI4->%pI4\n", |
1da177e4 | 737 | ic->type, ntohs(icmp_id(ic)), |
14d5e834 | 738 | &iph->saddr, &iph->daddr); |
1da177e4 LT |
739 | |
740 | /* | |
741 | * Work through seeing if this is for us. | |
742 | * These checks are supposed to be in an order that means easy | |
743 | * things are checked first to speed up processing.... however | |
744 | * this means that some packets will manage to get a long way | |
745 | * down this stack and then be rejected, but that's life. | |
746 | */ | |
747 | if ((ic->type != ICMP_DEST_UNREACH) && | |
748 | (ic->type != ICMP_SOURCE_QUENCH) && | |
749 | (ic->type != ICMP_TIME_EXCEEDED)) { | |
750 | *related = 0; | |
751 | return NF_ACCEPT; | |
752 | } | |
753 | ||
754 | /* Now find the contained IP header */ | |
755 | offset += sizeof(_icmph); | |
756 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
757 | if (cih == NULL) | |
758 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
759 | ||
760 | pp = ip_vs_proto_get(cih->protocol); | |
761 | if (!pp) | |
762 | return NF_ACCEPT; | |
763 | ||
764 | /* Is the embedded protocol header present? */ | |
4412ec49 | 765 | if (unlikely(cih->frag_off & htons(IP_OFFSET) && |
1da177e4 LT |
766 | pp->dont_defrag)) |
767 | return NF_ACCEPT; | |
768 | ||
769 | IP_VS_DBG_PKT(11, pp, skb, offset, "Checking outgoing ICMP for"); | |
770 | ||
771 | offset += cih->ihl * 4; | |
772 | ||
51ef348b | 773 | ip_vs_fill_iphdr(AF_INET, cih, &ciph); |
1da177e4 | 774 | /* The embedded headers contain source and dest in reverse order */ |
51ef348b | 775 | cp = pp->conn_out_get(AF_INET, skb, pp, &ciph, offset, 1); |
1da177e4 LT |
776 | if (!cp) |
777 | return NF_ACCEPT; | |
778 | ||
f2428ed5 SH |
779 | snet.ip = iph->saddr; |
780 | return handle_response_icmp(AF_INET, skb, &snet, cih->protocol, cp, | |
781 | pp, offset, ihl); | |
1da177e4 LT |
782 | } |
783 | ||
2a3b791e JV |
784 | #ifdef CONFIG_IP_VS_IPV6 |
785 | static int ip_vs_out_icmp_v6(struct sk_buff *skb, int *related) | |
786 | { | |
787 | struct ipv6hdr *iph; | |
788 | struct icmp6hdr _icmph, *ic; | |
789 | struct ipv6hdr _ciph, *cih; /* The ip header contained | |
790 | within the ICMP */ | |
791 | struct ip_vs_iphdr ciph; | |
792 | struct ip_vs_conn *cp; | |
793 | struct ip_vs_protocol *pp; | |
f2428ed5 SH |
794 | unsigned int offset; |
795 | union nf_inet_addr snet; | |
2a3b791e JV |
796 | |
797 | *related = 1; | |
798 | ||
799 | /* reassemble IP fragments */ | |
800 | if (ipv6_hdr(skb)->nexthdr == IPPROTO_FRAGMENT) { | |
801 | if (ip_vs_gather_frags_v6(skb, IP_DEFRAG_VS_OUT)) | |
802 | return NF_STOLEN; | |
803 | } | |
804 | ||
805 | iph = ipv6_hdr(skb); | |
806 | offset = sizeof(struct ipv6hdr); | |
807 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
808 | if (ic == NULL) | |
809 | return NF_DROP; | |
810 | ||
5b095d98 | 811 | IP_VS_DBG(12, "Outgoing ICMPv6 (%d,%d) %pI6->%pI6\n", |
2a3b791e | 812 | ic->icmp6_type, ntohs(icmpv6_id(ic)), |
38ff4fa4 | 813 | &iph->saddr, &iph->daddr); |
2a3b791e JV |
814 | |
815 | /* | |
816 | * Work through seeing if this is for us. | |
817 | * These checks are supposed to be in an order that means easy | |
818 | * things are checked first to speed up processing.... however | |
819 | * this means that some packets will manage to get a long way | |
820 | * down this stack and then be rejected, but that's life. | |
821 | */ | |
822 | if ((ic->icmp6_type != ICMPV6_DEST_UNREACH) && | |
823 | (ic->icmp6_type != ICMPV6_PKT_TOOBIG) && | |
824 | (ic->icmp6_type != ICMPV6_TIME_EXCEED)) { | |
825 | *related = 0; | |
826 | return NF_ACCEPT; | |
827 | } | |
828 | ||
829 | /* Now find the contained IP header */ | |
830 | offset += sizeof(_icmph); | |
831 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
832 | if (cih == NULL) | |
833 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
834 | ||
835 | pp = ip_vs_proto_get(cih->nexthdr); | |
836 | if (!pp) | |
837 | return NF_ACCEPT; | |
838 | ||
839 | /* Is the embedded protocol header present? */ | |
840 | /* TODO: we don't support fragmentation at the moment anyways */ | |
841 | if (unlikely(cih->nexthdr == IPPROTO_FRAGMENT && pp->dont_defrag)) | |
842 | return NF_ACCEPT; | |
843 | ||
844 | IP_VS_DBG_PKT(11, pp, skb, offset, "Checking outgoing ICMPv6 for"); | |
845 | ||
846 | offset += sizeof(struct ipv6hdr); | |
847 | ||
848 | ip_vs_fill_iphdr(AF_INET6, cih, &ciph); | |
849 | /* The embedded headers contain source and dest in reverse order */ | |
850 | cp = pp->conn_out_get(AF_INET6, skb, pp, &ciph, offset, 1); | |
851 | if (!cp) | |
852 | return NF_ACCEPT; | |
853 | ||
178f5e49 | 854 | ipv6_addr_copy(&snet.in6, &iph->saddr); |
f2428ed5 SH |
855 | return handle_response_icmp(AF_INET6, skb, &snet, cih->nexthdr, cp, |
856 | pp, offset, sizeof(struct ipv6hdr)); | |
2a3b791e JV |
857 | } |
858 | #endif | |
859 | ||
2906f66a VMR |
860 | /* |
861 | * Check if sctp chunc is ABORT chunk | |
862 | */ | |
863 | static inline int is_sctp_abort(const struct sk_buff *skb, int nh_len) | |
864 | { | |
865 | sctp_chunkhdr_t *sch, schunk; | |
866 | sch = skb_header_pointer(skb, nh_len + sizeof(sctp_sctphdr_t), | |
867 | sizeof(schunk), &schunk); | |
868 | if (sch == NULL) | |
869 | return 0; | |
870 | if (sch->type == SCTP_CID_ABORT) | |
871 | return 1; | |
872 | return 0; | |
873 | } | |
874 | ||
2a3b791e | 875 | static inline int is_tcp_reset(const struct sk_buff *skb, int nh_len) |
1da177e4 LT |
876 | { |
877 | struct tcphdr _tcph, *th; | |
878 | ||
2a3b791e | 879 | th = skb_header_pointer(skb, nh_len, sizeof(_tcph), &_tcph); |
1da177e4 LT |
880 | if (th == NULL) |
881 | return 0; | |
882 | return th->rst; | |
883 | } | |
884 | ||
4856c84c MT |
885 | /* Handle response packets: rewrite addresses and send away... |
886 | * Used for NAT and local client. | |
887 | */ | |
888 | static unsigned int | |
889 | handle_response(int af, struct sk_buff *skb, struct ip_vs_protocol *pp, | |
890 | struct ip_vs_conn *cp, int ihl) | |
891 | { | |
892 | IP_VS_DBG_PKT(11, pp, skb, 0, "Outgoing packet"); | |
893 | ||
894 | if (!skb_make_writable(skb, ihl)) | |
895 | goto drop; | |
896 | ||
897 | /* mangle the packet */ | |
898 | if (pp->snat_handler && !pp->snat_handler(skb, pp, cp)) | |
899 | goto drop; | |
900 | ||
901 | #ifdef CONFIG_IP_VS_IPV6 | |
902 | if (af == AF_INET6) | |
903 | ipv6_hdr(skb)->saddr = cp->vaddr.in6; | |
904 | else | |
905 | #endif | |
906 | { | |
907 | ip_hdr(skb)->saddr = cp->vaddr.ip; | |
908 | ip_send_check(ip_hdr(skb)); | |
909 | } | |
910 | ||
911 | /* For policy routing, packets originating from this | |
912 | * machine itself may be routed differently to packets | |
913 | * passing through. We want this packet to be routed as | |
914 | * if it came from this machine itself. So re-compute | |
915 | * the routing information. | |
916 | */ | |
917 | #ifdef CONFIG_IP_VS_IPV6 | |
918 | if (af == AF_INET6) { | |
919 | if (ip6_route_me_harder(skb) != 0) | |
920 | goto drop; | |
921 | } else | |
922 | #endif | |
923 | if (ip_route_me_harder(skb, RTN_LOCAL) != 0) | |
924 | goto drop; | |
925 | ||
4856c84c MT |
926 | IP_VS_DBG_PKT(10, pp, skb, 0, "After SNAT"); |
927 | ||
928 | ip_vs_out_stats(cp, skb); | |
929 | ip_vs_set_state(cp, IP_VS_DIR_OUTPUT, skb, pp); | |
930 | ip_vs_conn_put(cp); | |
931 | ||
932 | skb->ipvs_property = 1; | |
933 | ||
934 | LeaveFunction(11); | |
935 | return NF_ACCEPT; | |
936 | ||
937 | drop: | |
938 | ip_vs_conn_put(cp); | |
939 | kfree_skb(skb); | |
940 | return NF_STOLEN; | |
941 | } | |
942 | ||
1da177e4 | 943 | /* |
6e23ae2a | 944 | * It is hooked at the NF_INET_FORWARD chain, used only for VS/NAT. |
4856c84c | 945 | * Check if outgoing packet belongs to the established ip_vs_conn. |
1da177e4 LT |
946 | */ |
947 | static unsigned int | |
3db05fea | 948 | ip_vs_out(unsigned int hooknum, struct sk_buff *skb, |
1da177e4 LT |
949 | const struct net_device *in, const struct net_device *out, |
950 | int (*okfn)(struct sk_buff *)) | |
951 | { | |
51ef348b | 952 | struct ip_vs_iphdr iph; |
1da177e4 LT |
953 | struct ip_vs_protocol *pp; |
954 | struct ip_vs_conn *cp; | |
2a3b791e | 955 | int af; |
1da177e4 LT |
956 | |
957 | EnterFunction(11); | |
958 | ||
60678040 | 959 | af = (skb->protocol == htons(ETH_P_IP)) ? AF_INET : AF_INET6; |
2a3b791e | 960 | |
6869c4d8 | 961 | if (skb->ipvs_property) |
1da177e4 LT |
962 | return NF_ACCEPT; |
963 | ||
2a3b791e JV |
964 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); |
965 | #ifdef CONFIG_IP_VS_IPV6 | |
966 | if (af == AF_INET6) { | |
967 | if (unlikely(iph.protocol == IPPROTO_ICMPV6)) { | |
968 | int related, verdict = ip_vs_out_icmp_v6(skb, &related); | |
1da177e4 | 969 | |
2a3b791e JV |
970 | if (related) |
971 | return verdict; | |
972 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); | |
973 | } | |
974 | } else | |
975 | #endif | |
976 | if (unlikely(iph.protocol == IPPROTO_ICMP)) { | |
977 | int related, verdict = ip_vs_out_icmp(skb, &related); | |
978 | ||
979 | if (related) | |
980 | return verdict; | |
981 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); | |
982 | } | |
1da177e4 | 983 | |
51ef348b | 984 | pp = ip_vs_proto_get(iph.protocol); |
1da177e4 LT |
985 | if (unlikely(!pp)) |
986 | return NF_ACCEPT; | |
987 | ||
988 | /* reassemble IP fragments */ | |
2a3b791e JV |
989 | #ifdef CONFIG_IP_VS_IPV6 |
990 | if (af == AF_INET6) { | |
991 | if (unlikely(iph.protocol == IPPROTO_ICMPV6)) { | |
992 | int related, verdict = ip_vs_out_icmp_v6(skb, &related); | |
1da177e4 | 993 | |
2a3b791e JV |
994 | if (related) |
995 | return verdict; | |
996 | ||
997 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); | |
998 | } | |
999 | } else | |
1000 | #endif | |
1001 | if (unlikely(ip_hdr(skb)->frag_off & htons(IP_MF|IP_OFFSET) && | |
1002 | !pp->dont_defrag)) { | |
1003 | if (ip_vs_gather_frags(skb, IP_DEFRAG_VS_OUT)) | |
1004 | return NF_STOLEN; | |
1005 | ||
1006 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); | |
1007 | } | |
1da177e4 LT |
1008 | |
1009 | /* | |
1010 | * Check if the packet belongs to an existing entry | |
1011 | */ | |
2a3b791e | 1012 | cp = pp->conn_out_get(af, skb, pp, &iph, iph.len, 0); |
1da177e4 LT |
1013 | |
1014 | if (unlikely(!cp)) { | |
1015 | if (sysctl_ip_vs_nat_icmp_send && | |
1016 | (pp->protocol == IPPROTO_TCP || | |
2906f66a VMR |
1017 | pp->protocol == IPPROTO_UDP || |
1018 | pp->protocol == IPPROTO_SCTP)) { | |
014d730d | 1019 | __be16 _ports[2], *pptr; |
1da177e4 | 1020 | |
51ef348b | 1021 | pptr = skb_header_pointer(skb, iph.len, |
1da177e4 LT |
1022 | sizeof(_ports), _ports); |
1023 | if (pptr == NULL) | |
1024 | return NF_ACCEPT; /* Not for me */ | |
7937df15 JV |
1025 | if (ip_vs_lookup_real_service(af, iph.protocol, |
1026 | &iph.saddr, | |
2a3b791e | 1027 | pptr[0])) { |
1da177e4 LT |
1028 | /* |
1029 | * Notify the real server: there is no | |
1030 | * existing entry if it is not RST | |
1031 | * packet or not TCP packet. | |
1032 | */ | |
2906f66a VMR |
1033 | if ((iph.protocol != IPPROTO_TCP && |
1034 | iph.protocol != IPPROTO_SCTP) | |
1035 | || ((iph.protocol == IPPROTO_TCP | |
1036 | && !is_tcp_reset(skb, iph.len)) | |
1037 | || (iph.protocol == IPPROTO_SCTP | |
1038 | && !is_sctp_abort(skb, | |
1039 | iph.len)))) { | |
2a3b791e JV |
1040 | #ifdef CONFIG_IP_VS_IPV6 |
1041 | if (af == AF_INET6) | |
1042 | icmpv6_send(skb, | |
1043 | ICMPV6_DEST_UNREACH, | |
1044 | ICMPV6_PORT_UNREACH, | |
3ffe533c | 1045 | 0); |
2a3b791e JV |
1046 | else |
1047 | #endif | |
1048 | icmp_send(skb, | |
1049 | ICMP_DEST_UNREACH, | |
1050 | ICMP_PORT_UNREACH, 0); | |
1da177e4 | 1051 | return NF_DROP; |
5af149cc | 1052 | } |
1da177e4 LT |
1053 | } |
1054 | } | |
1055 | IP_VS_DBG_PKT(12, pp, skb, 0, | |
1056 | "packet continues traversal as normal"); | |
1057 | return NF_ACCEPT; | |
1058 | } | |
1059 | ||
4856c84c | 1060 | return handle_response(af, skb, pp, cp, iph.len); |
1da177e4 LT |
1061 | } |
1062 | ||
1063 | ||
1064 | /* | |
1065 | * Handle ICMP messages in the outside-to-inside direction (incoming). | |
1066 | * Find any that might be relevant, check against existing connections, | |
1067 | * forward to the right destination host if relevant. | |
1068 | * Currently handles error types - unreachable, quench, ttl exceeded. | |
1069 | */ | |
e905a9ed | 1070 | static int |
3db05fea | 1071 | ip_vs_in_icmp(struct sk_buff *skb, int *related, unsigned int hooknum) |
1da177e4 | 1072 | { |
1da177e4 LT |
1073 | struct iphdr *iph; |
1074 | struct icmphdr _icmph, *ic; | |
1075 | struct iphdr _ciph, *cih; /* The ip header contained within the ICMP */ | |
51ef348b | 1076 | struct ip_vs_iphdr ciph; |
1da177e4 LT |
1077 | struct ip_vs_conn *cp; |
1078 | struct ip_vs_protocol *pp; | |
1079 | unsigned int offset, ihl, verdict; | |
f2428ed5 | 1080 | union nf_inet_addr snet; |
1da177e4 LT |
1081 | |
1082 | *related = 1; | |
1083 | ||
1084 | /* reassemble IP fragments */ | |
eddc9ec5 | 1085 | if (ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)) { |
6e23ae2a | 1086 | if (ip_vs_gather_frags(skb, hooknum == NF_INET_LOCAL_IN ? |
776c729e | 1087 | IP_DEFRAG_VS_IN : IP_DEFRAG_VS_FWD)) |
1da177e4 | 1088 | return NF_STOLEN; |
1da177e4 LT |
1089 | } |
1090 | ||
eddc9ec5 | 1091 | iph = ip_hdr(skb); |
1da177e4 LT |
1092 | offset = ihl = iph->ihl * 4; |
1093 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
1094 | if (ic == NULL) | |
1095 | return NF_DROP; | |
1096 | ||
14d5e834 | 1097 | IP_VS_DBG(12, "Incoming ICMP (%d,%d) %pI4->%pI4\n", |
1da177e4 | 1098 | ic->type, ntohs(icmp_id(ic)), |
14d5e834 | 1099 | &iph->saddr, &iph->daddr); |
1da177e4 LT |
1100 | |
1101 | /* | |
1102 | * Work through seeing if this is for us. | |
1103 | * These checks are supposed to be in an order that means easy | |
1104 | * things are checked first to speed up processing.... however | |
1105 | * this means that some packets will manage to get a long way | |
1106 | * down this stack and then be rejected, but that's life. | |
1107 | */ | |
1108 | if ((ic->type != ICMP_DEST_UNREACH) && | |
1109 | (ic->type != ICMP_SOURCE_QUENCH) && | |
1110 | (ic->type != ICMP_TIME_EXCEEDED)) { | |
1111 | *related = 0; | |
1112 | return NF_ACCEPT; | |
1113 | } | |
1114 | ||
1115 | /* Now find the contained IP header */ | |
1116 | offset += sizeof(_icmph); | |
1117 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
1118 | if (cih == NULL) | |
1119 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
1120 | ||
1121 | pp = ip_vs_proto_get(cih->protocol); | |
1122 | if (!pp) | |
1123 | return NF_ACCEPT; | |
1124 | ||
1125 | /* Is the embedded protocol header present? */ | |
4412ec49 | 1126 | if (unlikely(cih->frag_off & htons(IP_OFFSET) && |
1da177e4 LT |
1127 | pp->dont_defrag)) |
1128 | return NF_ACCEPT; | |
1129 | ||
1130 | IP_VS_DBG_PKT(11, pp, skb, offset, "Checking incoming ICMP for"); | |
1131 | ||
1132 | offset += cih->ihl * 4; | |
1133 | ||
51ef348b | 1134 | ip_vs_fill_iphdr(AF_INET, cih, &ciph); |
1da177e4 | 1135 | /* The embedded headers contain source and dest in reverse order */ |
51ef348b | 1136 | cp = pp->conn_in_get(AF_INET, skb, pp, &ciph, offset, 1); |
4856c84c MT |
1137 | if (!cp) { |
1138 | /* The packet could also belong to a local client */ | |
1139 | cp = pp->conn_out_get(AF_INET, skb, pp, &ciph, offset, 1); | |
f2428ed5 SH |
1140 | if (cp) { |
1141 | snet.ip = iph->saddr; | |
1142 | return handle_response_icmp(AF_INET, skb, &snet, | |
1143 | cih->protocol, cp, pp, | |
4856c84c | 1144 | offset, ihl); |
f2428ed5 | 1145 | } |
1da177e4 | 1146 | return NF_ACCEPT; |
4856c84c | 1147 | } |
1da177e4 LT |
1148 | |
1149 | verdict = NF_DROP; | |
1150 | ||
1151 | /* Ensure the checksum is correct */ | |
60476372 | 1152 | if (!skb_csum_unnecessary(skb) && ip_vs_checksum_complete(skb, ihl)) { |
1da177e4 | 1153 | /* Failed checksum! */ |
14d5e834 HH |
1154 | IP_VS_DBG(1, "Incoming ICMP: failed checksum from %pI4!\n", |
1155 | &iph->saddr); | |
1da177e4 LT |
1156 | goto out; |
1157 | } | |
1158 | ||
1159 | /* do the statistics and put it back */ | |
1160 | ip_vs_in_stats(cp, skb); | |
1161 | if (IPPROTO_TCP == cih->protocol || IPPROTO_UDP == cih->protocol) | |
1162 | offset += 2 * sizeof(__u16); | |
1163 | verdict = ip_vs_icmp_xmit(skb, cp, pp, offset); | |
1164 | /* do not touch skb anymore */ | |
1165 | ||
1166 | out: | |
1167 | __ip_vs_conn_put(cp); | |
1168 | ||
1169 | return verdict; | |
1170 | } | |
1171 | ||
2a3b791e JV |
1172 | #ifdef CONFIG_IP_VS_IPV6 |
1173 | static int | |
1174 | ip_vs_in_icmp_v6(struct sk_buff *skb, int *related, unsigned int hooknum) | |
1175 | { | |
1176 | struct ipv6hdr *iph; | |
1177 | struct icmp6hdr _icmph, *ic; | |
1178 | struct ipv6hdr _ciph, *cih; /* The ip header contained | |
1179 | within the ICMP */ | |
1180 | struct ip_vs_iphdr ciph; | |
1181 | struct ip_vs_conn *cp; | |
1182 | struct ip_vs_protocol *pp; | |
1183 | unsigned int offset, verdict; | |
f2428ed5 | 1184 | union nf_inet_addr snet; |
2a3b791e JV |
1185 | |
1186 | *related = 1; | |
1187 | ||
1188 | /* reassemble IP fragments */ | |
1189 | if (ipv6_hdr(skb)->nexthdr == IPPROTO_FRAGMENT) { | |
1190 | if (ip_vs_gather_frags_v6(skb, hooknum == NF_INET_LOCAL_IN ? | |
1191 | IP_DEFRAG_VS_IN : | |
1192 | IP_DEFRAG_VS_FWD)) | |
1193 | return NF_STOLEN; | |
1194 | } | |
1195 | ||
1196 | iph = ipv6_hdr(skb); | |
1197 | offset = sizeof(struct ipv6hdr); | |
1198 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
1199 | if (ic == NULL) | |
1200 | return NF_DROP; | |
1201 | ||
5b095d98 | 1202 | IP_VS_DBG(12, "Incoming ICMPv6 (%d,%d) %pI6->%pI6\n", |
2a3b791e | 1203 | ic->icmp6_type, ntohs(icmpv6_id(ic)), |
38ff4fa4 | 1204 | &iph->saddr, &iph->daddr); |
2a3b791e JV |
1205 | |
1206 | /* | |
1207 | * Work through seeing if this is for us. | |
1208 | * These checks are supposed to be in an order that means easy | |
1209 | * things are checked first to speed up processing.... however | |
1210 | * this means that some packets will manage to get a long way | |
1211 | * down this stack and then be rejected, but that's life. | |
1212 | */ | |
1213 | if ((ic->icmp6_type != ICMPV6_DEST_UNREACH) && | |
1214 | (ic->icmp6_type != ICMPV6_PKT_TOOBIG) && | |
1215 | (ic->icmp6_type != ICMPV6_TIME_EXCEED)) { | |
1216 | *related = 0; | |
1217 | return NF_ACCEPT; | |
1218 | } | |
1219 | ||
1220 | /* Now find the contained IP header */ | |
1221 | offset += sizeof(_icmph); | |
1222 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
1223 | if (cih == NULL) | |
1224 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
1225 | ||
1226 | pp = ip_vs_proto_get(cih->nexthdr); | |
1227 | if (!pp) | |
1228 | return NF_ACCEPT; | |
1229 | ||
1230 | /* Is the embedded protocol header present? */ | |
1231 | /* TODO: we don't support fragmentation at the moment anyways */ | |
1232 | if (unlikely(cih->nexthdr == IPPROTO_FRAGMENT && pp->dont_defrag)) | |
1233 | return NF_ACCEPT; | |
1234 | ||
1235 | IP_VS_DBG_PKT(11, pp, skb, offset, "Checking incoming ICMPv6 for"); | |
1236 | ||
1237 | offset += sizeof(struct ipv6hdr); | |
1238 | ||
1239 | ip_vs_fill_iphdr(AF_INET6, cih, &ciph); | |
1240 | /* The embedded headers contain source and dest in reverse order */ | |
1241 | cp = pp->conn_in_get(AF_INET6, skb, pp, &ciph, offset, 1); | |
f2428ed5 SH |
1242 | if (!cp) { |
1243 | /* The packet could also belong to a local client */ | |
1244 | cp = pp->conn_out_get(AF_INET6, skb, pp, &ciph, offset, 1); | |
1245 | if (cp) { | |
178f5e49 | 1246 | ipv6_addr_copy(&snet.in6, &iph->saddr); |
f2428ed5 SH |
1247 | return handle_response_icmp(AF_INET6, skb, &snet, |
1248 | cih->nexthdr, | |
1249 | cp, pp, offset, | |
1250 | sizeof(struct ipv6hdr)); | |
1251 | } | |
2a3b791e | 1252 | return NF_ACCEPT; |
f2428ed5 | 1253 | } |
2a3b791e JV |
1254 | |
1255 | verdict = NF_DROP; | |
1256 | ||
1257 | /* do the statistics and put it back */ | |
1258 | ip_vs_in_stats(cp, skb); | |
2906f66a VMR |
1259 | if (IPPROTO_TCP == cih->nexthdr || IPPROTO_UDP == cih->nexthdr || |
1260 | IPPROTO_SCTP == cih->nexthdr) | |
2a3b791e JV |
1261 | offset += 2 * sizeof(__u16); |
1262 | verdict = ip_vs_icmp_xmit_v6(skb, cp, pp, offset); | |
1263 | /* do not touch skb anymore */ | |
1264 | ||
1265 | __ip_vs_conn_put(cp); | |
1266 | ||
1267 | return verdict; | |
1268 | } | |
1269 | #endif | |
1270 | ||
1271 | ||
1da177e4 LT |
1272 | /* |
1273 | * Check if it's for virtual services, look it up, | |
1274 | * and send it on its way... | |
1275 | */ | |
1276 | static unsigned int | |
3db05fea | 1277 | ip_vs_in(unsigned int hooknum, struct sk_buff *skb, |
1da177e4 LT |
1278 | const struct net_device *in, const struct net_device *out, |
1279 | int (*okfn)(struct sk_buff *)) | |
1280 | { | |
51ef348b | 1281 | struct ip_vs_iphdr iph; |
1da177e4 LT |
1282 | struct ip_vs_protocol *pp; |
1283 | struct ip_vs_conn *cp; | |
1e66dafc | 1284 | int ret, restart, af, pkts; |
2a3b791e | 1285 | |
60678040 | 1286 | af = (skb->protocol == htons(ETH_P_IP)) ? AF_INET : AF_INET6; |
51ef348b | 1287 | |
2a3b791e | 1288 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); |
1da177e4 LT |
1289 | |
1290 | /* | |
4856c84c MT |
1291 | * Big tappo: only PACKET_HOST, including loopback for local client |
1292 | * Don't handle local packets on IPv6 for now | |
1da177e4 | 1293 | */ |
f2428ed5 | 1294 | if (unlikely(skb->pkt_type != PACKET_HOST)) { |
51ef348b JV |
1295 | IP_VS_DBG_BUF(12, "packet type=%d proto=%d daddr=%s ignored\n", |
1296 | skb->pkt_type, | |
1297 | iph.protocol, | |
2a3b791e | 1298 | IP_VS_DBG_ADDR(af, &iph.daddr)); |
1da177e4 LT |
1299 | return NF_ACCEPT; |
1300 | } | |
1301 | ||
94b26551 JV |
1302 | #ifdef CONFIG_IP_VS_IPV6 |
1303 | if (af == AF_INET6) { | |
1304 | if (unlikely(iph.protocol == IPPROTO_ICMPV6)) { | |
1305 | int related, verdict = ip_vs_in_icmp_v6(skb, &related, hooknum); | |
1da177e4 | 1306 | |
94b26551 JV |
1307 | if (related) |
1308 | return verdict; | |
1309 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); | |
1310 | } | |
1311 | } else | |
1312 | #endif | |
1313 | if (unlikely(iph.protocol == IPPROTO_ICMP)) { | |
1314 | int related, verdict = ip_vs_in_icmp(skb, &related, hooknum); | |
1315 | ||
1316 | if (related) | |
1317 | return verdict; | |
1318 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); | |
1319 | } | |
1da177e4 LT |
1320 | |
1321 | /* Protocol supported? */ | |
51ef348b | 1322 | pp = ip_vs_proto_get(iph.protocol); |
1da177e4 LT |
1323 | if (unlikely(!pp)) |
1324 | return NF_ACCEPT; | |
1325 | ||
1da177e4 LT |
1326 | /* |
1327 | * Check if the packet belongs to an existing connection entry | |
1328 | */ | |
2a3b791e | 1329 | cp = pp->conn_in_get(af, skb, pp, &iph, iph.len, 0); |
1da177e4 LT |
1330 | |
1331 | if (unlikely(!cp)) { | |
1332 | int v; | |
1333 | ||
4856c84c MT |
1334 | /* For local client packets, it could be a response */ |
1335 | cp = pp->conn_out_get(af, skb, pp, &iph, iph.len, 0); | |
1336 | if (cp) | |
1337 | return handle_response(af, skb, pp, cp, iph.len); | |
1338 | ||
2a3b791e | 1339 | if (!pp->conn_schedule(af, skb, pp, &v, &cp)) |
1da177e4 LT |
1340 | return v; |
1341 | } | |
1342 | ||
1343 | if (unlikely(!cp)) { | |
1344 | /* sorry, all this trouble for a no-hit :) */ | |
1345 | IP_VS_DBG_PKT(12, pp, skb, 0, | |
1346 | "packet continues traversal as normal"); | |
1347 | return NF_ACCEPT; | |
1348 | } | |
1349 | ||
1350 | IP_VS_DBG_PKT(11, pp, skb, 0, "Incoming packet"); | |
1351 | ||
1352 | /* Check the server status */ | |
1353 | if (cp->dest && !(cp->dest->flags & IP_VS_DEST_F_AVAILABLE)) { | |
1354 | /* the destination server is not available */ | |
1355 | ||
1356 | if (sysctl_ip_vs_expire_nodest_conn) { | |
1357 | /* try to expire the connection immediately */ | |
1358 | ip_vs_conn_expire_now(cp); | |
1da177e4 | 1359 | } |
dc8103f2 JA |
1360 | /* don't restart its timer, and silently |
1361 | drop the packet. */ | |
1362 | __ip_vs_conn_put(cp); | |
1da177e4 LT |
1363 | return NF_DROP; |
1364 | } | |
1365 | ||
1366 | ip_vs_in_stats(cp, skb); | |
1367 | restart = ip_vs_set_state(cp, IP_VS_DIR_INPUT, skb, pp); | |
1368 | if (cp->packet_xmit) | |
1369 | ret = cp->packet_xmit(skb, cp, pp); | |
1370 | /* do not touch skb anymore */ | |
1371 | else { | |
1372 | IP_VS_DBG_RL("warning: packet_xmit is null"); | |
1373 | ret = NF_ACCEPT; | |
1374 | } | |
1375 | ||
efac5276 RB |
1376 | /* Increase its packet counter and check if it is needed |
1377 | * to be synchronized | |
1378 | * | |
1379 | * Sync connection if it is about to close to | |
1380 | * encorage the standby servers to update the connections timeout | |
1381 | */ | |
1e66dafc | 1382 | pkts = atomic_add_return(1, &cp->in_pkts); |
2906f66a VMR |
1383 | if (af == AF_INET && (ip_vs_sync_state & IP_VS_STATE_MASTER) && |
1384 | cp->protocol == IPPROTO_SCTP) { | |
1385 | if ((cp->state == IP_VS_SCTP_S_ESTABLISHED && | |
1386 | (atomic_read(&cp->in_pkts) % | |
1387 | sysctl_ip_vs_sync_threshold[1] | |
1388 | == sysctl_ip_vs_sync_threshold[0])) || | |
1389 | (cp->old_state != cp->state && | |
1390 | ((cp->state == IP_VS_SCTP_S_CLOSED) || | |
1391 | (cp->state == IP_VS_SCTP_S_SHUT_ACK_CLI) || | |
1392 | (cp->state == IP_VS_SCTP_S_SHUT_ACK_SER)))) { | |
1393 | ip_vs_sync_conn(cp); | |
1394 | goto out; | |
1395 | } | |
1396 | } | |
1397 | ||
c6883f58 JV |
1398 | if (af == AF_INET && |
1399 | (ip_vs_sync_state & IP_VS_STATE_MASTER) && | |
efac5276 RB |
1400 | (((cp->protocol != IPPROTO_TCP || |
1401 | cp->state == IP_VS_TCP_S_ESTABLISHED) && | |
1e66dafc | 1402 | (pkts % sysctl_ip_vs_sync_threshold[1] |
efac5276 RB |
1403 | == sysctl_ip_vs_sync_threshold[0])) || |
1404 | ((cp->protocol == IPPROTO_TCP) && (cp->old_state != cp->state) && | |
1405 | ((cp->state == IP_VS_TCP_S_FIN_WAIT) || | |
9abfe315 | 1406 | (cp->state == IP_VS_TCP_S_CLOSE) || |
9d3a0de7 RB |
1407 | (cp->state == IP_VS_TCP_S_CLOSE_WAIT) || |
1408 | (cp->state == IP_VS_TCP_S_TIME_WAIT))))) | |
1da177e4 | 1409 | ip_vs_sync_conn(cp); |
2906f66a | 1410 | out: |
efac5276 | 1411 | cp->old_state = cp->state; |
1da177e4 LT |
1412 | |
1413 | ip_vs_conn_put(cp); | |
1414 | return ret; | |
1415 | } | |
1416 | ||
1417 | ||
1418 | /* | |
6e23ae2a | 1419 | * It is hooked at the NF_INET_FORWARD chain, in order to catch ICMP |
1da177e4 LT |
1420 | * related packets destined for 0.0.0.0/0. |
1421 | * When fwmark-based virtual service is used, such as transparent | |
1422 | * cache cluster, TCP packets can be marked and routed to ip_vs_in, | |
1423 | * but ICMP destined for 0.0.0.0/0 cannot not be easily marked and | |
6e23ae2a | 1424 | * sent to ip_vs_in_icmp. So, catch them at the NF_INET_FORWARD chain |
1da177e4 LT |
1425 | * and send them to ip_vs_in_icmp. |
1426 | */ | |
1427 | static unsigned int | |
3db05fea | 1428 | ip_vs_forward_icmp(unsigned int hooknum, struct sk_buff *skb, |
1da177e4 LT |
1429 | const struct net_device *in, const struct net_device *out, |
1430 | int (*okfn)(struct sk_buff *)) | |
1431 | { | |
1432 | int r; | |
1433 | ||
3db05fea | 1434 | if (ip_hdr(skb)->protocol != IPPROTO_ICMP) |
1da177e4 LT |
1435 | return NF_ACCEPT; |
1436 | ||
3db05fea | 1437 | return ip_vs_in_icmp(skb, &r, hooknum); |
1da177e4 LT |
1438 | } |
1439 | ||
2a3b791e JV |
1440 | #ifdef CONFIG_IP_VS_IPV6 |
1441 | static unsigned int | |
1442 | ip_vs_forward_icmp_v6(unsigned int hooknum, struct sk_buff *skb, | |
1443 | const struct net_device *in, const struct net_device *out, | |
1444 | int (*okfn)(struct sk_buff *)) | |
1445 | { | |
1446 | int r; | |
1447 | ||
1448 | if (ipv6_hdr(skb)->nexthdr != IPPROTO_ICMPV6) | |
1449 | return NF_ACCEPT; | |
1450 | ||
1451 | return ip_vs_in_icmp_v6(skb, &r, hooknum); | |
1452 | } | |
1453 | #endif | |
1454 | ||
1da177e4 | 1455 | |
1999414a | 1456 | static struct nf_hook_ops ip_vs_ops[] __read_mostly = { |
41c5b317 PM |
1457 | /* After packet filtering, forward packet through VS/DR, VS/TUN, |
1458 | * or VS/NAT(change destination), so that filtering rules can be | |
1459 | * applied to IPVS. */ | |
1460 | { | |
1461 | .hook = ip_vs_in, | |
1462 | .owner = THIS_MODULE, | |
1463 | .pf = PF_INET, | |
1464 | .hooknum = NF_INET_LOCAL_IN, | |
1465 | .priority = 100, | |
1466 | }, | |
1467 | /* After packet filtering, change source only for VS/NAT */ | |
1468 | { | |
1469 | .hook = ip_vs_out, | |
1470 | .owner = THIS_MODULE, | |
1471 | .pf = PF_INET, | |
1472 | .hooknum = NF_INET_FORWARD, | |
1473 | .priority = 100, | |
1474 | }, | |
1475 | /* After packet filtering (but before ip_vs_out_icmp), catch icmp | |
1476 | * destined for 0.0.0.0/0, which is for incoming IPVS connections */ | |
1477 | { | |
1478 | .hook = ip_vs_forward_icmp, | |
1479 | .owner = THIS_MODULE, | |
1480 | .pf = PF_INET, | |
1481 | .hooknum = NF_INET_FORWARD, | |
1482 | .priority = 99, | |
1483 | }, | |
473b23d3 JV |
1484 | #ifdef CONFIG_IP_VS_IPV6 |
1485 | /* After packet filtering, forward packet through VS/DR, VS/TUN, | |
1486 | * or VS/NAT(change destination), so that filtering rules can be | |
1487 | * applied to IPVS. */ | |
1488 | { | |
1489 | .hook = ip_vs_in, | |
1490 | .owner = THIS_MODULE, | |
1491 | .pf = PF_INET6, | |
1492 | .hooknum = NF_INET_LOCAL_IN, | |
1493 | .priority = 100, | |
1494 | }, | |
1495 | /* After packet filtering, change source only for VS/NAT */ | |
1496 | { | |
1497 | .hook = ip_vs_out, | |
1498 | .owner = THIS_MODULE, | |
1499 | .pf = PF_INET6, | |
1500 | .hooknum = NF_INET_FORWARD, | |
1501 | .priority = 100, | |
1502 | }, | |
1503 | /* After packet filtering (but before ip_vs_out_icmp), catch icmp | |
1504 | * destined for 0.0.0.0/0, which is for incoming IPVS connections */ | |
1505 | { | |
1506 | .hook = ip_vs_forward_icmp_v6, | |
1507 | .owner = THIS_MODULE, | |
1508 | .pf = PF_INET6, | |
1509 | .hooknum = NF_INET_FORWARD, | |
1510 | .priority = 99, | |
1511 | }, | |
473b23d3 | 1512 | #endif |
1da177e4 LT |
1513 | }; |
1514 | ||
1515 | ||
1516 | /* | |
1517 | * Initialize IP Virtual Server | |
1518 | */ | |
1519 | static int __init ip_vs_init(void) | |
1520 | { | |
1521 | int ret; | |
1522 | ||
a919cf4b SW |
1523 | ip_vs_estimator_init(); |
1524 | ||
1da177e4 LT |
1525 | ret = ip_vs_control_init(); |
1526 | if (ret < 0) { | |
1e3e238e | 1527 | pr_err("can't setup control.\n"); |
a919cf4b | 1528 | goto cleanup_estimator; |
1da177e4 LT |
1529 | } |
1530 | ||
1531 | ip_vs_protocol_init(); | |
1532 | ||
1533 | ret = ip_vs_app_init(); | |
1534 | if (ret < 0) { | |
1e3e238e | 1535 | pr_err("can't setup application helper.\n"); |
1da177e4 LT |
1536 | goto cleanup_protocol; |
1537 | } | |
1538 | ||
1539 | ret = ip_vs_conn_init(); | |
1540 | if (ret < 0) { | |
1e3e238e | 1541 | pr_err("can't setup connection table.\n"); |
1da177e4 LT |
1542 | goto cleanup_app; |
1543 | } | |
1544 | ||
41c5b317 | 1545 | ret = nf_register_hooks(ip_vs_ops, ARRAY_SIZE(ip_vs_ops)); |
1da177e4 | 1546 | if (ret < 0) { |
1e3e238e | 1547 | pr_err("can't register hooks.\n"); |
1da177e4 LT |
1548 | goto cleanup_conn; |
1549 | } | |
1550 | ||
1e3e238e | 1551 | pr_info("ipvs loaded.\n"); |
1da177e4 LT |
1552 | return ret; |
1553 | ||
1da177e4 LT |
1554 | cleanup_conn: |
1555 | ip_vs_conn_cleanup(); | |
1556 | cleanup_app: | |
1557 | ip_vs_app_cleanup(); | |
1558 | cleanup_protocol: | |
1559 | ip_vs_protocol_cleanup(); | |
1560 | ip_vs_control_cleanup(); | |
a919cf4b SW |
1561 | cleanup_estimator: |
1562 | ip_vs_estimator_cleanup(); | |
1da177e4 LT |
1563 | return ret; |
1564 | } | |
1565 | ||
1566 | static void __exit ip_vs_cleanup(void) | |
1567 | { | |
41c5b317 | 1568 | nf_unregister_hooks(ip_vs_ops, ARRAY_SIZE(ip_vs_ops)); |
1da177e4 LT |
1569 | ip_vs_conn_cleanup(); |
1570 | ip_vs_app_cleanup(); | |
1571 | ip_vs_protocol_cleanup(); | |
1572 | ip_vs_control_cleanup(); | |
a919cf4b | 1573 | ip_vs_estimator_cleanup(); |
1e3e238e | 1574 | pr_info("ipvs unloaded.\n"); |
1da177e4 LT |
1575 | } |
1576 | ||
1577 | module_init(ip_vs_init); | |
1578 | module_exit(ip_vs_cleanup); | |
1579 | MODULE_LICENSE("GPL"); |