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; | |
5b57a98c SH |
196 | int protocol = iph.protocol; |
197 | __be16 dport = 0; /* destination port to forward */ | |
198 | __be16 vport = 0; /* virtual service port */ | |
3575792e | 199 | unsigned int flags; |
28364a59 JV |
200 | union nf_inet_addr snet; /* source network of the client, |
201 | after masking */ | |
5b57a98c SH |
202 | const union nf_inet_addr fwmark = { .ip = htonl(svc->fwmark) }; |
203 | const union nf_inet_addr *vaddr = &iph.daddr; | |
cd17f9ed JV |
204 | |
205 | ip_vs_fill_iphdr(svc->af, skb_network_header(skb), &iph); | |
1da177e4 LT |
206 | |
207 | /* Mask saddr with the netmask to adjust template granularity */ | |
cd17f9ed JV |
208 | #ifdef CONFIG_IP_VS_IPV6 |
209 | if (svc->af == AF_INET6) | |
210 | ipv6_addr_prefix(&snet.in6, &iph.saddr.in6, svc->netmask); | |
211 | else | |
212 | #endif | |
213 | snet.ip = iph.saddr.ip & svc->netmask; | |
1da177e4 | 214 | |
cd17f9ed JV |
215 | IP_VS_DBG_BUF(6, "p-schedule: src %s:%u dest %s:%u " |
216 | "mnet %s\n", | |
217 | IP_VS_DBG_ADDR(svc->af, &iph.saddr), ntohs(ports[0]), | |
218 | IP_VS_DBG_ADDR(svc->af, &iph.daddr), ntohs(ports[1]), | |
219 | IP_VS_DBG_ADDR(svc->af, &snet)); | |
1da177e4 LT |
220 | |
221 | /* | |
222 | * As far as we know, FTP is a very complicated network protocol, and | |
223 | * it uses control connection and data connections. For active FTP, | |
224 | * FTP server initialize data connection to the client, its source port | |
225 | * is often 20. For passive FTP, FTP server tells the clients the port | |
226 | * that it passively listens to, and the client issues the data | |
227 | * connection. In the tunneling or direct routing mode, the load | |
228 | * balancer is on the client-to-server half of connection, the port | |
229 | * number is unknown to the load balancer. So, a conn template like | |
230 | * <caddr, 0, vaddr, 0, daddr, 0> is created for persistent FTP | |
231 | * service, and a template like <caddr, 0, vaddr, vport, daddr, dport> | |
232 | * is created for other persistent services. | |
233 | */ | |
5b57a98c SH |
234 | { |
235 | if (ports[1] == svc->port) { | |
236 | /* non-FTP template: | |
237 | * <protocol, caddr, 0, vaddr, vport, daddr, dport> | |
238 | * FTP template: | |
239 | * <protocol, caddr, 0, vaddr, 0, daddr, 0> | |
1da177e4 LT |
240 | */ |
241 | if (svc->port != FTPPORT) | |
5b57a98c | 242 | vport = ports[1]; |
1da177e4 | 243 | } else { |
5b57a98c SH |
244 | /* Note: persistent fwmark-based services and |
245 | * persistent port zero service are handled here. | |
246 | * fwmark template: | |
247 | * <IPPROTO_IP,caddr,0,fwmark,0,daddr,0> | |
248 | * port zero template: | |
249 | * <protocol,caddr,0,vaddr,0,daddr,0> | |
1da177e4 | 250 | */ |
28364a59 | 251 | if (svc->fwmark) { |
5b57a98c SH |
252 | protocol = IPPROTO_IP; |
253 | vaddr = &fwmark; | |
254 | } | |
1da177e4 | 255 | } |
1da177e4 LT |
256 | } |
257 | ||
5b57a98c SH |
258 | /* Check if a template already exists */ |
259 | ct = ip_vs_ct_in_get(svc->af, protocol, &snet, 0, vaddr, vport); | |
260 | ||
261 | if (!ct || !ip_vs_check_template(ct)) { | |
262 | /* No template found or the dest of the connection | |
263 | * template is not available. | |
264 | */ | |
265 | dest = svc->scheduler->schedule(svc, skb); | |
266 | if (!dest) { | |
267 | IP_VS_DBG(1, "p-schedule: no dest found.\n"); | |
268 | return NULL; | |
269 | } | |
270 | ||
271 | if (ports[1] == svc->port && svc->port != FTPPORT) | |
272 | dport = dest->port; | |
273 | ||
274 | /* Create a template */ | |
275 | ct = ip_vs_conn_new(svc->af, protocol, &snet, 0,vaddr, vport, | |
276 | &dest->addr, dport, | |
277 | IP_VS_CONN_F_TEMPLATE, dest); | |
278 | if (ct == NULL) | |
279 | return NULL; | |
280 | ||
281 | ct->timeout = svc->timeout; | |
282 | } else | |
283 | /* set destination with the found template */ | |
284 | dest = ct->dest; | |
285 | ||
286 | dport = ports[1]; | |
287 | if (dport == svc->port && dest->port) | |
288 | dport = dest->port; | |
289 | ||
26ec037f NC |
290 | flags = (svc->flags & IP_VS_SVC_F_ONEPACKET |
291 | && iph.protocol == IPPROTO_UDP)? | |
292 | IP_VS_CONN_F_ONE_PACKET : 0; | |
293 | ||
1da177e4 LT |
294 | /* |
295 | * Create a new connection according to the template | |
296 | */ | |
cd17f9ed | 297 | cp = ip_vs_conn_new(svc->af, iph.protocol, |
28364a59 JV |
298 | &iph.saddr, ports[0], |
299 | &iph.daddr, ports[1], | |
300 | &dest->addr, dport, | |
26ec037f | 301 | flags, |
1da177e4 LT |
302 | dest); |
303 | if (cp == NULL) { | |
304 | ip_vs_conn_put(ct); | |
305 | return NULL; | |
306 | } | |
307 | ||
308 | /* | |
309 | * Add its control | |
310 | */ | |
311 | ip_vs_control_add(cp, ct); | |
312 | ip_vs_conn_put(ct); | |
313 | ||
314 | ip_vs_conn_stats(cp, svc); | |
315 | return cp; | |
316 | } | |
317 | ||
318 | ||
319 | /* | |
320 | * IPVS main scheduling function | |
321 | * It selects a server according to the virtual service, and | |
322 | * creates a connection entry. | |
323 | * Protocols supported: TCP, UDP | |
324 | */ | |
325 | struct ip_vs_conn * | |
326 | ip_vs_schedule(struct ip_vs_service *svc, const struct sk_buff *skb) | |
327 | { | |
328 | struct ip_vs_conn *cp = NULL; | |
28364a59 | 329 | struct ip_vs_iphdr iph; |
1da177e4 | 330 | struct ip_vs_dest *dest; |
3575792e JA |
331 | __be16 _ports[2], *pptr; |
332 | unsigned int flags; | |
1da177e4 | 333 | |
28364a59 JV |
334 | ip_vs_fill_iphdr(svc->af, skb_network_header(skb), &iph); |
335 | pptr = skb_header_pointer(skb, iph.len, sizeof(_ports), _ports); | |
1da177e4 LT |
336 | if (pptr == NULL) |
337 | return NULL; | |
338 | ||
339 | /* | |
340 | * Persistent service | |
341 | */ | |
342 | if (svc->flags & IP_VS_SVC_F_PERSISTENT) | |
343 | return ip_vs_sched_persist(svc, skb, pptr); | |
344 | ||
345 | /* | |
346 | * Non-persistent service | |
347 | */ | |
348 | if (!svc->fwmark && pptr[1] != svc->port) { | |
349 | if (!svc->port) | |
1e3e238e HE |
350 | pr_err("Schedule: port zero only supported " |
351 | "in persistent services, " | |
352 | "check your ipvs configuration\n"); | |
1da177e4 LT |
353 | return NULL; |
354 | } | |
355 | ||
356 | dest = svc->scheduler->schedule(svc, skb); | |
357 | if (dest == NULL) { | |
358 | IP_VS_DBG(1, "Schedule: no dest found.\n"); | |
359 | return NULL; | |
360 | } | |
361 | ||
26ec037f NC |
362 | flags = (svc->flags & IP_VS_SVC_F_ONEPACKET |
363 | && iph.protocol == IPPROTO_UDP)? | |
364 | IP_VS_CONN_F_ONE_PACKET : 0; | |
365 | ||
1da177e4 LT |
366 | /* |
367 | * Create a connection entry. | |
368 | */ | |
cd17f9ed | 369 | cp = ip_vs_conn_new(svc->af, iph.protocol, |
28364a59 JV |
370 | &iph.saddr, pptr[0], |
371 | &iph.daddr, pptr[1], | |
372 | &dest->addr, dest->port ? dest->port : pptr[1], | |
26ec037f | 373 | flags, |
1da177e4 LT |
374 | dest); |
375 | if (cp == NULL) | |
376 | return NULL; | |
377 | ||
cd17f9ed JV |
378 | IP_VS_DBG_BUF(6, "Schedule fwd:%c c:%s:%u v:%s:%u " |
379 | "d:%s:%u conn->flags:%X conn->refcnt:%d\n", | |
380 | ip_vs_fwd_tag(cp), | |
381 | IP_VS_DBG_ADDR(svc->af, &cp->caddr), ntohs(cp->cport), | |
382 | IP_VS_DBG_ADDR(svc->af, &cp->vaddr), ntohs(cp->vport), | |
383 | IP_VS_DBG_ADDR(svc->af, &cp->daddr), ntohs(cp->dport), | |
384 | cp->flags, atomic_read(&cp->refcnt)); | |
1da177e4 LT |
385 | |
386 | ip_vs_conn_stats(cp, svc); | |
387 | return cp; | |
388 | } | |
389 | ||
390 | ||
391 | /* | |
392 | * Pass or drop the packet. | |
393 | * Called by ip_vs_in, when the virtual service is available but | |
394 | * no destination is available for a new connection. | |
395 | */ | |
396 | int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb, | |
397 | struct ip_vs_protocol *pp) | |
398 | { | |
014d730d | 399 | __be16 _ports[2], *pptr; |
28364a59 | 400 | struct ip_vs_iphdr iph; |
2a3b791e JV |
401 | int unicast; |
402 | ip_vs_fill_iphdr(svc->af, skb_network_header(skb), &iph); | |
1da177e4 | 403 | |
28364a59 | 404 | pptr = skb_header_pointer(skb, iph.len, sizeof(_ports), _ports); |
1da177e4 LT |
405 | if (pptr == NULL) { |
406 | ip_vs_service_put(svc); | |
407 | return NF_DROP; | |
408 | } | |
409 | ||
2a3b791e JV |
410 | #ifdef CONFIG_IP_VS_IPV6 |
411 | if (svc->af == AF_INET6) | |
412 | unicast = ipv6_addr_type(&iph.daddr.in6) & IPV6_ADDR_UNICAST; | |
413 | else | |
414 | #endif | |
415 | unicast = (inet_addr_type(&init_net, iph.daddr.ip) == RTN_UNICAST); | |
416 | ||
1da177e4 | 417 | /* if it is fwmark-based service, the cache_bypass sysctl is up |
2a3b791e | 418 | and the destination is a non-local unicast, then create |
1da177e4 | 419 | a cache_bypass connection entry */ |
2a3b791e | 420 | if (sysctl_ip_vs_cache_bypass && svc->fwmark && unicast) { |
1da177e4 LT |
421 | int ret, cs; |
422 | struct ip_vs_conn *cp; | |
3575792e JA |
423 | unsigned int flags = (svc->flags & IP_VS_SVC_F_ONEPACKET && |
424 | iph.protocol == IPPROTO_UDP)? | |
425 | IP_VS_CONN_F_ONE_PACKET : 0; | |
dff630dd | 426 | union nf_inet_addr daddr = { .all = { 0, 0, 0, 0 } }; |
1da177e4 LT |
427 | |
428 | ip_vs_service_put(svc); | |
429 | ||
430 | /* create a new connection entry */ | |
1e3e238e | 431 | IP_VS_DBG(6, "%s(): create a cache_bypass entry\n", __func__); |
2a3b791e | 432 | cp = ip_vs_conn_new(svc->af, iph.protocol, |
28364a59 JV |
433 | &iph.saddr, pptr[0], |
434 | &iph.daddr, pptr[1], | |
dff630dd | 435 | &daddr, 0, |
26ec037f | 436 | IP_VS_CONN_F_BYPASS | flags, |
1da177e4 LT |
437 | NULL); |
438 | if (cp == NULL) | |
439 | return NF_DROP; | |
440 | ||
441 | /* statistics */ | |
442 | ip_vs_in_stats(cp, skb); | |
443 | ||
444 | /* set state */ | |
445 | cs = ip_vs_set_state(cp, IP_VS_DIR_INPUT, skb, pp); | |
446 | ||
447 | /* transmit the first SYN packet */ | |
448 | ret = cp->packet_xmit(skb, cp, pp); | |
449 | /* do not touch skb anymore */ | |
450 | ||
451 | atomic_inc(&cp->in_pkts); | |
452 | ip_vs_conn_put(cp); | |
453 | return ret; | |
454 | } | |
455 | ||
456 | /* | |
457 | * When the virtual ftp service is presented, packets destined | |
458 | * for other services on the VIP may get here (except services | |
459 | * listed in the ipvs table), pass the packets, because it is | |
460 | * not ipvs job to decide to drop the packets. | |
461 | */ | |
462 | if ((svc->port == FTPPORT) && (pptr[1] != FTPPORT)) { | |
463 | ip_vs_service_put(svc); | |
464 | return NF_ACCEPT; | |
465 | } | |
466 | ||
467 | ip_vs_service_put(svc); | |
468 | ||
469 | /* | |
470 | * Notify the client that the destination is unreachable, and | |
471 | * release the socket buffer. | |
472 | * Since it is in IP layer, the TCP socket is not actually | |
473 | * created, the TCP RST packet cannot be sent, instead that | |
474 | * ICMP_PORT_UNREACH is sent here no matter it is TCP/UDP. --WZ | |
475 | */ | |
2a3b791e JV |
476 | #ifdef CONFIG_IP_VS_IPV6 |
477 | if (svc->af == AF_INET6) | |
3ffe533c | 478 | icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); |
2a3b791e JV |
479 | else |
480 | #endif | |
481 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); | |
482 | ||
1da177e4 LT |
483 | return NF_DROP; |
484 | } | |
485 | ||
f4bc17cd JA |
486 | /* |
487 | * It is hooked before NF_IP_PRI_NAT_SRC at the NF_INET_POST_ROUTING | |
488 | * chain and is used to avoid double NAT and confirmation when we do | |
489 | * not want to keep the conntrack structure | |
490 | */ | |
491 | static unsigned int ip_vs_post_routing(unsigned int hooknum, | |
492 | struct sk_buff *skb, | |
493 | const struct net_device *in, | |
494 | const struct net_device *out, | |
495 | int (*okfn)(struct sk_buff *)) | |
496 | { | |
497 | if (!skb->ipvs_property) | |
498 | return NF_ACCEPT; | |
499 | /* The packet was sent from IPVS, exit this chain */ | |
500 | return NF_STOP; | |
501 | } | |
502 | ||
b1550f22 | 503 | __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset) |
1da177e4 | 504 | { |
d3bc23e7 | 505 | return csum_fold(skb_checksum(skb, offset, skb->len - offset, 0)); |
1da177e4 LT |
506 | } |
507 | ||
776c729e | 508 | static inline int ip_vs_gather_frags(struct sk_buff *skb, u_int32_t user) |
1da177e4 | 509 | { |
776c729e HX |
510 | int err = ip_defrag(skb, user); |
511 | ||
512 | if (!err) | |
eddc9ec5 | 513 | ip_send_check(ip_hdr(skb)); |
776c729e HX |
514 | |
515 | return err; | |
1da177e4 LT |
516 | } |
517 | ||
2a3b791e JV |
518 | #ifdef CONFIG_IP_VS_IPV6 |
519 | static inline int ip_vs_gather_frags_v6(struct sk_buff *skb, u_int32_t user) | |
520 | { | |
521 | /* TODO IPv6: Find out what to do here for IPv6 */ | |
522 | return 0; | |
523 | } | |
524 | #endif | |
525 | ||
1da177e4 LT |
526 | /* |
527 | * Packet has been made sufficiently writable in caller | |
528 | * - inout: 1=in->out, 0=out->in | |
529 | */ | |
530 | void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp, | |
531 | struct ip_vs_conn *cp, int inout) | |
532 | { | |
eddc9ec5 | 533 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 | 534 | unsigned int icmp_offset = iph->ihl*4; |
d56f90a7 ACM |
535 | struct icmphdr *icmph = (struct icmphdr *)(skb_network_header(skb) + |
536 | icmp_offset); | |
1da177e4 LT |
537 | struct iphdr *ciph = (struct iphdr *)(icmph + 1); |
538 | ||
539 | if (inout) { | |
e7ade46a | 540 | iph->saddr = cp->vaddr.ip; |
1da177e4 | 541 | ip_send_check(iph); |
e7ade46a | 542 | ciph->daddr = cp->vaddr.ip; |
1da177e4 LT |
543 | ip_send_check(ciph); |
544 | } else { | |
e7ade46a | 545 | iph->daddr = cp->daddr.ip; |
1da177e4 | 546 | ip_send_check(iph); |
e7ade46a | 547 | ciph->saddr = cp->daddr.ip; |
1da177e4 LT |
548 | ip_send_check(ciph); |
549 | } | |
550 | ||
2906f66a VMR |
551 | /* the TCP/UDP/SCTP port */ |
552 | if (IPPROTO_TCP == ciph->protocol || IPPROTO_UDP == ciph->protocol || | |
553 | IPPROTO_SCTP == ciph->protocol) { | |
014d730d | 554 | __be16 *ports = (void *)ciph + ciph->ihl*4; |
1da177e4 LT |
555 | |
556 | if (inout) | |
557 | ports[1] = cp->vport; | |
558 | else | |
559 | ports[0] = cp->dport; | |
560 | } | |
561 | ||
562 | /* And finally the ICMP checksum */ | |
563 | icmph->checksum = 0; | |
564 | icmph->checksum = ip_vs_checksum_complete(skb, icmp_offset); | |
565 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
566 | ||
567 | if (inout) | |
568 | IP_VS_DBG_PKT(11, pp, skb, (void *)ciph - (void *)iph, | |
569 | "Forwarding altered outgoing ICMP"); | |
570 | else | |
571 | IP_VS_DBG_PKT(11, pp, skb, (void *)ciph - (void *)iph, | |
572 | "Forwarding altered incoming ICMP"); | |
573 | } | |
574 | ||
b3cdd2a7 JV |
575 | #ifdef CONFIG_IP_VS_IPV6 |
576 | void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp, | |
577 | struct ip_vs_conn *cp, int inout) | |
578 | { | |
579 | struct ipv6hdr *iph = ipv6_hdr(skb); | |
580 | unsigned int icmp_offset = sizeof(struct ipv6hdr); | |
581 | struct icmp6hdr *icmph = (struct icmp6hdr *)(skb_network_header(skb) + | |
582 | icmp_offset); | |
583 | struct ipv6hdr *ciph = (struct ipv6hdr *)(icmph + 1); | |
584 | ||
585 | if (inout) { | |
586 | iph->saddr = cp->vaddr.in6; | |
587 | ciph->daddr = cp->vaddr.in6; | |
588 | } else { | |
589 | iph->daddr = cp->daddr.in6; | |
590 | ciph->saddr = cp->daddr.in6; | |
591 | } | |
592 | ||
2906f66a VMR |
593 | /* the TCP/UDP/SCTP port */ |
594 | if (IPPROTO_TCP == ciph->nexthdr || IPPROTO_UDP == ciph->nexthdr || | |
595 | IPPROTO_SCTP == ciph->nexthdr) { | |
b3cdd2a7 JV |
596 | __be16 *ports = (void *)ciph + sizeof(struct ipv6hdr); |
597 | ||
598 | if (inout) | |
599 | ports[1] = cp->vport; | |
600 | else | |
601 | ports[0] = cp->dport; | |
602 | } | |
603 | ||
604 | /* And finally the ICMP checksum */ | |
8870f842 SH |
605 | icmph->icmp6_cksum = ~csum_ipv6_magic(&iph->saddr, &iph->daddr, |
606 | skb->len - icmp_offset, | |
607 | IPPROTO_ICMPV6, 0); | |
608 | skb->csum_start = skb_network_header(skb) - skb->head + icmp_offset; | |
609 | skb->csum_offset = offsetof(struct icmp6hdr, icmp6_cksum); | |
610 | skb->ip_summed = CHECKSUM_PARTIAL; | |
b3cdd2a7 JV |
611 | |
612 | if (inout) | |
613 | IP_VS_DBG_PKT(11, pp, skb, (void *)ciph - (void *)iph, | |
614 | "Forwarding altered outgoing ICMPv6"); | |
615 | else | |
616 | IP_VS_DBG_PKT(11, pp, skb, (void *)ciph - (void *)iph, | |
617 | "Forwarding altered incoming ICMPv6"); | |
618 | } | |
619 | #endif | |
620 | ||
4856c84c MT |
621 | /* Handle relevant response ICMP messages - forward to the right |
622 | * destination host. Used for NAT and local client. | |
623 | */ | |
f2428ed5 SH |
624 | static int handle_response_icmp(int af, struct sk_buff *skb, |
625 | union nf_inet_addr *snet, | |
626 | __u8 protocol, struct ip_vs_conn *cp, | |
4856c84c MT |
627 | struct ip_vs_protocol *pp, |
628 | unsigned int offset, unsigned int ihl) | |
629 | { | |
630 | unsigned int verdict = NF_DROP; | |
631 | ||
632 | if (IP_VS_FWD_METHOD(cp) != 0) { | |
1e3e238e HE |
633 | pr_err("shouldn't reach here, because the box is on the " |
634 | "half connection in the tun/dr module.\n"); | |
4856c84c MT |
635 | } |
636 | ||
637 | /* Ensure the checksum is correct */ | |
638 | if (!skb_csum_unnecessary(skb) && ip_vs_checksum_complete(skb, ihl)) { | |
639 | /* Failed checksum! */ | |
f2428ed5 SH |
640 | IP_VS_DBG_BUF(1, "Forward ICMP: failed checksum from %s!\n", |
641 | IP_VS_DBG_ADDR(af, snet)); | |
4856c84c MT |
642 | goto out; |
643 | } | |
644 | ||
2906f66a VMR |
645 | if (IPPROTO_TCP == protocol || IPPROTO_UDP == protocol || |
646 | IPPROTO_SCTP == protocol) | |
4856c84c MT |
647 | offset += 2 * sizeof(__u16); |
648 | if (!skb_make_writable(skb, offset)) | |
649 | goto out; | |
650 | ||
f2428ed5 SH |
651 | #ifdef CONFIG_IP_VS_IPV6 |
652 | if (af == AF_INET6) | |
653 | ip_vs_nat_icmp_v6(skb, pp, cp, 1); | |
654 | else | |
655 | #endif | |
656 | ip_vs_nat_icmp(skb, pp, cp, 1); | |
4856c84c MT |
657 | |
658 | /* do the statistics and put it back */ | |
659 | ip_vs_out_stats(cp, skb); | |
660 | ||
f4bc17cd JA |
661 | if (!(cp->flags & IP_VS_CONN_F_NFCT)) |
662 | skb->ipvs_property = 1; | |
663 | else | |
664 | ip_vs_update_conntrack(skb, cp, 0); | |
4856c84c MT |
665 | verdict = NF_ACCEPT; |
666 | ||
667 | out: | |
668 | __ip_vs_conn_put(cp); | |
669 | ||
670 | return verdict; | |
671 | } | |
672 | ||
1da177e4 LT |
673 | /* |
674 | * Handle ICMP messages in the inside-to-outside direction (outgoing). | |
4856c84c | 675 | * Find any that might be relevant, check against existing connections. |
1da177e4 | 676 | * Currently handles error types - unreachable, quench, ttl exceeded. |
1da177e4 | 677 | */ |
3db05fea | 678 | static int ip_vs_out_icmp(struct sk_buff *skb, int *related) |
1da177e4 | 679 | { |
1da177e4 LT |
680 | struct iphdr *iph; |
681 | struct icmphdr _icmph, *ic; | |
682 | struct iphdr _ciph, *cih; /* The ip header contained within the ICMP */ | |
51ef348b | 683 | struct ip_vs_iphdr ciph; |
1da177e4 LT |
684 | struct ip_vs_conn *cp; |
685 | struct ip_vs_protocol *pp; | |
4856c84c | 686 | unsigned int offset, ihl; |
f2428ed5 | 687 | union nf_inet_addr snet; |
1da177e4 LT |
688 | |
689 | *related = 1; | |
690 | ||
691 | /* reassemble IP fragments */ | |
eddc9ec5 | 692 | if (ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)) { |
776c729e | 693 | if (ip_vs_gather_frags(skb, IP_DEFRAG_VS_OUT)) |
1da177e4 | 694 | return NF_STOLEN; |
1da177e4 LT |
695 | } |
696 | ||
eddc9ec5 | 697 | iph = ip_hdr(skb); |
1da177e4 LT |
698 | offset = ihl = iph->ihl * 4; |
699 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
700 | if (ic == NULL) | |
701 | return NF_DROP; | |
702 | ||
14d5e834 | 703 | IP_VS_DBG(12, "Outgoing ICMP (%d,%d) %pI4->%pI4\n", |
1da177e4 | 704 | ic->type, ntohs(icmp_id(ic)), |
14d5e834 | 705 | &iph->saddr, &iph->daddr); |
1da177e4 LT |
706 | |
707 | /* | |
708 | * Work through seeing if this is for us. | |
709 | * These checks are supposed to be in an order that means easy | |
710 | * things are checked first to speed up processing.... however | |
711 | * this means that some packets will manage to get a long way | |
712 | * down this stack and then be rejected, but that's life. | |
713 | */ | |
714 | if ((ic->type != ICMP_DEST_UNREACH) && | |
715 | (ic->type != ICMP_SOURCE_QUENCH) && | |
716 | (ic->type != ICMP_TIME_EXCEEDED)) { | |
717 | *related = 0; | |
718 | return NF_ACCEPT; | |
719 | } | |
720 | ||
721 | /* Now find the contained IP header */ | |
722 | offset += sizeof(_icmph); | |
723 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
724 | if (cih == NULL) | |
725 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
726 | ||
727 | pp = ip_vs_proto_get(cih->protocol); | |
728 | if (!pp) | |
729 | return NF_ACCEPT; | |
730 | ||
731 | /* Is the embedded protocol header present? */ | |
4412ec49 | 732 | if (unlikely(cih->frag_off & htons(IP_OFFSET) && |
1da177e4 LT |
733 | pp->dont_defrag)) |
734 | return NF_ACCEPT; | |
735 | ||
736 | IP_VS_DBG_PKT(11, pp, skb, offset, "Checking outgoing ICMP for"); | |
737 | ||
738 | offset += cih->ihl * 4; | |
739 | ||
51ef348b | 740 | ip_vs_fill_iphdr(AF_INET, cih, &ciph); |
1da177e4 | 741 | /* The embedded headers contain source and dest in reverse order */ |
51ef348b | 742 | cp = pp->conn_out_get(AF_INET, skb, pp, &ciph, offset, 1); |
1da177e4 LT |
743 | if (!cp) |
744 | return NF_ACCEPT; | |
745 | ||
f2428ed5 SH |
746 | snet.ip = iph->saddr; |
747 | return handle_response_icmp(AF_INET, skb, &snet, cih->protocol, cp, | |
748 | pp, offset, ihl); | |
1da177e4 LT |
749 | } |
750 | ||
2a3b791e JV |
751 | #ifdef CONFIG_IP_VS_IPV6 |
752 | static int ip_vs_out_icmp_v6(struct sk_buff *skb, int *related) | |
753 | { | |
754 | struct ipv6hdr *iph; | |
755 | struct icmp6hdr _icmph, *ic; | |
756 | struct ipv6hdr _ciph, *cih; /* The ip header contained | |
757 | within the ICMP */ | |
758 | struct ip_vs_iphdr ciph; | |
759 | struct ip_vs_conn *cp; | |
760 | struct ip_vs_protocol *pp; | |
f2428ed5 SH |
761 | unsigned int offset; |
762 | union nf_inet_addr snet; | |
2a3b791e JV |
763 | |
764 | *related = 1; | |
765 | ||
766 | /* reassemble IP fragments */ | |
767 | if (ipv6_hdr(skb)->nexthdr == IPPROTO_FRAGMENT) { | |
768 | if (ip_vs_gather_frags_v6(skb, IP_DEFRAG_VS_OUT)) | |
769 | return NF_STOLEN; | |
770 | } | |
771 | ||
772 | iph = ipv6_hdr(skb); | |
773 | offset = sizeof(struct ipv6hdr); | |
774 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
775 | if (ic == NULL) | |
776 | return NF_DROP; | |
777 | ||
5b095d98 | 778 | IP_VS_DBG(12, "Outgoing ICMPv6 (%d,%d) %pI6->%pI6\n", |
2a3b791e | 779 | ic->icmp6_type, ntohs(icmpv6_id(ic)), |
38ff4fa4 | 780 | &iph->saddr, &iph->daddr); |
2a3b791e JV |
781 | |
782 | /* | |
783 | * Work through seeing if this is for us. | |
784 | * These checks are supposed to be in an order that means easy | |
785 | * things are checked first to speed up processing.... however | |
786 | * this means that some packets will manage to get a long way | |
787 | * down this stack and then be rejected, but that's life. | |
788 | */ | |
789 | if ((ic->icmp6_type != ICMPV6_DEST_UNREACH) && | |
790 | (ic->icmp6_type != ICMPV6_PKT_TOOBIG) && | |
791 | (ic->icmp6_type != ICMPV6_TIME_EXCEED)) { | |
792 | *related = 0; | |
793 | return NF_ACCEPT; | |
794 | } | |
795 | ||
796 | /* Now find the contained IP header */ | |
797 | offset += sizeof(_icmph); | |
798 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
799 | if (cih == NULL) | |
800 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
801 | ||
802 | pp = ip_vs_proto_get(cih->nexthdr); | |
803 | if (!pp) | |
804 | return NF_ACCEPT; | |
805 | ||
806 | /* Is the embedded protocol header present? */ | |
807 | /* TODO: we don't support fragmentation at the moment anyways */ | |
808 | if (unlikely(cih->nexthdr == IPPROTO_FRAGMENT && pp->dont_defrag)) | |
809 | return NF_ACCEPT; | |
810 | ||
811 | IP_VS_DBG_PKT(11, pp, skb, offset, "Checking outgoing ICMPv6 for"); | |
812 | ||
813 | offset += sizeof(struct ipv6hdr); | |
814 | ||
815 | ip_vs_fill_iphdr(AF_INET6, cih, &ciph); | |
816 | /* The embedded headers contain source and dest in reverse order */ | |
817 | cp = pp->conn_out_get(AF_INET6, skb, pp, &ciph, offset, 1); | |
818 | if (!cp) | |
819 | return NF_ACCEPT; | |
820 | ||
178f5e49 | 821 | ipv6_addr_copy(&snet.in6, &iph->saddr); |
f2428ed5 SH |
822 | return handle_response_icmp(AF_INET6, skb, &snet, cih->nexthdr, cp, |
823 | pp, offset, sizeof(struct ipv6hdr)); | |
2a3b791e JV |
824 | } |
825 | #endif | |
826 | ||
2906f66a VMR |
827 | /* |
828 | * Check if sctp chunc is ABORT chunk | |
829 | */ | |
830 | static inline int is_sctp_abort(const struct sk_buff *skb, int nh_len) | |
831 | { | |
832 | sctp_chunkhdr_t *sch, schunk; | |
833 | sch = skb_header_pointer(skb, nh_len + sizeof(sctp_sctphdr_t), | |
834 | sizeof(schunk), &schunk); | |
835 | if (sch == NULL) | |
836 | return 0; | |
837 | if (sch->type == SCTP_CID_ABORT) | |
838 | return 1; | |
839 | return 0; | |
840 | } | |
841 | ||
2a3b791e | 842 | static inline int is_tcp_reset(const struct sk_buff *skb, int nh_len) |
1da177e4 LT |
843 | { |
844 | struct tcphdr _tcph, *th; | |
845 | ||
2a3b791e | 846 | th = skb_header_pointer(skb, nh_len, sizeof(_tcph), &_tcph); |
1da177e4 LT |
847 | if (th == NULL) |
848 | return 0; | |
849 | return th->rst; | |
850 | } | |
851 | ||
4856c84c MT |
852 | /* Handle response packets: rewrite addresses and send away... |
853 | * Used for NAT and local client. | |
854 | */ | |
855 | static unsigned int | |
856 | handle_response(int af, struct sk_buff *skb, struct ip_vs_protocol *pp, | |
857 | struct ip_vs_conn *cp, int ihl) | |
858 | { | |
859 | IP_VS_DBG_PKT(11, pp, skb, 0, "Outgoing packet"); | |
860 | ||
861 | if (!skb_make_writable(skb, ihl)) | |
862 | goto drop; | |
863 | ||
864 | /* mangle the packet */ | |
865 | if (pp->snat_handler && !pp->snat_handler(skb, pp, cp)) | |
866 | goto drop; | |
867 | ||
868 | #ifdef CONFIG_IP_VS_IPV6 | |
869 | if (af == AF_INET6) | |
870 | ipv6_hdr(skb)->saddr = cp->vaddr.in6; | |
871 | else | |
872 | #endif | |
873 | { | |
874 | ip_hdr(skb)->saddr = cp->vaddr.ip; | |
875 | ip_send_check(ip_hdr(skb)); | |
876 | } | |
877 | ||
8a803040 JA |
878 | /* |
879 | * nf_iterate does not expect change in the skb->dst->dev. | |
880 | * It looks like it is not fatal to enable this code for hooks | |
881 | * where our handlers are at the end of the chain list and | |
882 | * when all next handlers use skb->dst->dev and not outdev. | |
883 | * It will definitely route properly the inout NAT traffic | |
884 | * when multiple paths are used. | |
885 | */ | |
886 | ||
4856c84c MT |
887 | /* For policy routing, packets originating from this |
888 | * machine itself may be routed differently to packets | |
889 | * passing through. We want this packet to be routed as | |
890 | * if it came from this machine itself. So re-compute | |
891 | * the routing information. | |
892 | */ | |
8a803040 | 893 | if (sysctl_ip_vs_snat_reroute) { |
4856c84c | 894 | #ifdef CONFIG_IP_VS_IPV6 |
8a803040 JA |
895 | if (af == AF_INET6) { |
896 | if (ip6_route_me_harder(skb) != 0) | |
897 | goto drop; | |
898 | } else | |
4856c84c | 899 | #endif |
8a803040 JA |
900 | if (ip_route_me_harder(skb, RTN_LOCAL) != 0) |
901 | goto drop; | |
902 | } | |
4856c84c | 903 | |
4856c84c MT |
904 | IP_VS_DBG_PKT(10, pp, skb, 0, "After SNAT"); |
905 | ||
906 | ip_vs_out_stats(cp, skb); | |
907 | ip_vs_set_state(cp, IP_VS_DIR_OUTPUT, skb, pp); | |
f4bc17cd JA |
908 | if (!(cp->flags & IP_VS_CONN_F_NFCT)) |
909 | skb->ipvs_property = 1; | |
910 | else | |
911 | ip_vs_update_conntrack(skb, cp, 0); | |
4856c84c MT |
912 | ip_vs_conn_put(cp); |
913 | ||
4856c84c MT |
914 | LeaveFunction(11); |
915 | return NF_ACCEPT; | |
916 | ||
917 | drop: | |
918 | ip_vs_conn_put(cp); | |
919 | kfree_skb(skb); | |
f4bc17cd | 920 | LeaveFunction(11); |
4856c84c MT |
921 | return NF_STOLEN; |
922 | } | |
923 | ||
1da177e4 | 924 | /* |
6e23ae2a | 925 | * It is hooked at the NF_INET_FORWARD chain, used only for VS/NAT. |
4856c84c | 926 | * Check if outgoing packet belongs to the established ip_vs_conn. |
1da177e4 LT |
927 | */ |
928 | static unsigned int | |
3db05fea | 929 | ip_vs_out(unsigned int hooknum, struct sk_buff *skb, |
1da177e4 LT |
930 | const struct net_device *in, const struct net_device *out, |
931 | int (*okfn)(struct sk_buff *)) | |
932 | { | |
51ef348b | 933 | struct ip_vs_iphdr iph; |
1da177e4 LT |
934 | struct ip_vs_protocol *pp; |
935 | struct ip_vs_conn *cp; | |
2a3b791e | 936 | int af; |
1da177e4 LT |
937 | |
938 | EnterFunction(11); | |
939 | ||
60678040 | 940 | af = (skb->protocol == htons(ETH_P_IP)) ? AF_INET : AF_INET6; |
2a3b791e | 941 | |
6869c4d8 | 942 | if (skb->ipvs_property) |
1da177e4 LT |
943 | return NF_ACCEPT; |
944 | ||
2a3b791e JV |
945 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); |
946 | #ifdef CONFIG_IP_VS_IPV6 | |
947 | if (af == AF_INET6) { | |
948 | if (unlikely(iph.protocol == IPPROTO_ICMPV6)) { | |
949 | int related, verdict = ip_vs_out_icmp_v6(skb, &related); | |
1da177e4 | 950 | |
8a803040 JA |
951 | if (related) { |
952 | if (sysctl_ip_vs_snat_reroute && | |
953 | NF_ACCEPT == verdict && | |
954 | ip6_route_me_harder(skb)) | |
955 | verdict = NF_DROP; | |
2a3b791e | 956 | return verdict; |
8a803040 | 957 | } |
2a3b791e JV |
958 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); |
959 | } | |
960 | } else | |
961 | #endif | |
962 | if (unlikely(iph.protocol == IPPROTO_ICMP)) { | |
963 | int related, verdict = ip_vs_out_icmp(skb, &related); | |
964 | ||
8a803040 JA |
965 | if (related) { |
966 | if (sysctl_ip_vs_snat_reroute && | |
967 | NF_ACCEPT == verdict && | |
968 | ip_route_me_harder(skb, RTN_LOCAL)) | |
969 | verdict = NF_DROP; | |
2a3b791e | 970 | return verdict; |
8a803040 | 971 | } |
2a3b791e JV |
972 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); |
973 | } | |
1da177e4 | 974 | |
51ef348b | 975 | pp = ip_vs_proto_get(iph.protocol); |
1da177e4 LT |
976 | if (unlikely(!pp)) |
977 | return NF_ACCEPT; | |
978 | ||
979 | /* reassemble IP fragments */ | |
2a3b791e JV |
980 | #ifdef CONFIG_IP_VS_IPV6 |
981 | if (af == AF_INET6) { | |
982 | if (unlikely(iph.protocol == IPPROTO_ICMPV6)) { | |
983 | int related, verdict = ip_vs_out_icmp_v6(skb, &related); | |
1da177e4 | 984 | |
2a3b791e JV |
985 | if (related) |
986 | return verdict; | |
987 | ||
988 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); | |
989 | } | |
990 | } else | |
991 | #endif | |
992 | if (unlikely(ip_hdr(skb)->frag_off & htons(IP_MF|IP_OFFSET) && | |
993 | !pp->dont_defrag)) { | |
994 | if (ip_vs_gather_frags(skb, IP_DEFRAG_VS_OUT)) | |
995 | return NF_STOLEN; | |
996 | ||
997 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); | |
998 | } | |
1da177e4 LT |
999 | |
1000 | /* | |
1001 | * Check if the packet belongs to an existing entry | |
1002 | */ | |
2a3b791e | 1003 | cp = pp->conn_out_get(af, skb, pp, &iph, iph.len, 0); |
1da177e4 LT |
1004 | |
1005 | if (unlikely(!cp)) { | |
1006 | if (sysctl_ip_vs_nat_icmp_send && | |
1007 | (pp->protocol == IPPROTO_TCP || | |
2906f66a VMR |
1008 | pp->protocol == IPPROTO_UDP || |
1009 | pp->protocol == IPPROTO_SCTP)) { | |
014d730d | 1010 | __be16 _ports[2], *pptr; |
1da177e4 | 1011 | |
51ef348b | 1012 | pptr = skb_header_pointer(skb, iph.len, |
1da177e4 LT |
1013 | sizeof(_ports), _ports); |
1014 | if (pptr == NULL) | |
1015 | return NF_ACCEPT; /* Not for me */ | |
7937df15 JV |
1016 | if (ip_vs_lookup_real_service(af, iph.protocol, |
1017 | &iph.saddr, | |
2a3b791e | 1018 | pptr[0])) { |
1da177e4 LT |
1019 | /* |
1020 | * Notify the real server: there is no | |
1021 | * existing entry if it is not RST | |
1022 | * packet or not TCP packet. | |
1023 | */ | |
2906f66a VMR |
1024 | if ((iph.protocol != IPPROTO_TCP && |
1025 | iph.protocol != IPPROTO_SCTP) | |
1026 | || ((iph.protocol == IPPROTO_TCP | |
1027 | && !is_tcp_reset(skb, iph.len)) | |
1028 | || (iph.protocol == IPPROTO_SCTP | |
1029 | && !is_sctp_abort(skb, | |
1030 | iph.len)))) { | |
2a3b791e JV |
1031 | #ifdef CONFIG_IP_VS_IPV6 |
1032 | if (af == AF_INET6) | |
1033 | icmpv6_send(skb, | |
1034 | ICMPV6_DEST_UNREACH, | |
1035 | ICMPV6_PORT_UNREACH, | |
3ffe533c | 1036 | 0); |
2a3b791e JV |
1037 | else |
1038 | #endif | |
1039 | icmp_send(skb, | |
1040 | ICMP_DEST_UNREACH, | |
1041 | ICMP_PORT_UNREACH, 0); | |
1da177e4 | 1042 | return NF_DROP; |
5af149cc | 1043 | } |
1da177e4 LT |
1044 | } |
1045 | } | |
1046 | IP_VS_DBG_PKT(12, pp, skb, 0, | |
1047 | "packet continues traversal as normal"); | |
1048 | return NF_ACCEPT; | |
1049 | } | |
1050 | ||
4856c84c | 1051 | return handle_response(af, skb, pp, cp, iph.len); |
1da177e4 LT |
1052 | } |
1053 | ||
1054 | ||
1055 | /* | |
1056 | * Handle ICMP messages in the outside-to-inside direction (incoming). | |
1057 | * Find any that might be relevant, check against existing connections, | |
1058 | * forward to the right destination host if relevant. | |
1059 | * Currently handles error types - unreachable, quench, ttl exceeded. | |
1060 | */ | |
e905a9ed | 1061 | static int |
3db05fea | 1062 | ip_vs_in_icmp(struct sk_buff *skb, int *related, unsigned int hooknum) |
1da177e4 | 1063 | { |
1da177e4 LT |
1064 | struct iphdr *iph; |
1065 | struct icmphdr _icmph, *ic; | |
1066 | struct iphdr _ciph, *cih; /* The ip header contained within the ICMP */ | |
51ef348b | 1067 | struct ip_vs_iphdr ciph; |
1da177e4 LT |
1068 | struct ip_vs_conn *cp; |
1069 | struct ip_vs_protocol *pp; | |
1070 | unsigned int offset, ihl, verdict; | |
f2428ed5 | 1071 | union nf_inet_addr snet; |
1da177e4 LT |
1072 | |
1073 | *related = 1; | |
1074 | ||
1075 | /* reassemble IP fragments */ | |
eddc9ec5 | 1076 | if (ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)) { |
6e23ae2a | 1077 | if (ip_vs_gather_frags(skb, hooknum == NF_INET_LOCAL_IN ? |
776c729e | 1078 | IP_DEFRAG_VS_IN : IP_DEFRAG_VS_FWD)) |
1da177e4 | 1079 | return NF_STOLEN; |
1da177e4 LT |
1080 | } |
1081 | ||
eddc9ec5 | 1082 | iph = ip_hdr(skb); |
1da177e4 LT |
1083 | offset = ihl = iph->ihl * 4; |
1084 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
1085 | if (ic == NULL) | |
1086 | return NF_DROP; | |
1087 | ||
14d5e834 | 1088 | IP_VS_DBG(12, "Incoming ICMP (%d,%d) %pI4->%pI4\n", |
1da177e4 | 1089 | ic->type, ntohs(icmp_id(ic)), |
14d5e834 | 1090 | &iph->saddr, &iph->daddr); |
1da177e4 LT |
1091 | |
1092 | /* | |
1093 | * Work through seeing if this is for us. | |
1094 | * These checks are supposed to be in an order that means easy | |
1095 | * things are checked first to speed up processing.... however | |
1096 | * this means that some packets will manage to get a long way | |
1097 | * down this stack and then be rejected, but that's life. | |
1098 | */ | |
1099 | if ((ic->type != ICMP_DEST_UNREACH) && | |
1100 | (ic->type != ICMP_SOURCE_QUENCH) && | |
1101 | (ic->type != ICMP_TIME_EXCEEDED)) { | |
1102 | *related = 0; | |
1103 | return NF_ACCEPT; | |
1104 | } | |
1105 | ||
1106 | /* Now find the contained IP header */ | |
1107 | offset += sizeof(_icmph); | |
1108 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
1109 | if (cih == NULL) | |
1110 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
1111 | ||
1112 | pp = ip_vs_proto_get(cih->protocol); | |
1113 | if (!pp) | |
1114 | return NF_ACCEPT; | |
1115 | ||
1116 | /* Is the embedded protocol header present? */ | |
4412ec49 | 1117 | if (unlikely(cih->frag_off & htons(IP_OFFSET) && |
1da177e4 LT |
1118 | pp->dont_defrag)) |
1119 | return NF_ACCEPT; | |
1120 | ||
1121 | IP_VS_DBG_PKT(11, pp, skb, offset, "Checking incoming ICMP for"); | |
1122 | ||
1123 | offset += cih->ihl * 4; | |
1124 | ||
51ef348b | 1125 | ip_vs_fill_iphdr(AF_INET, cih, &ciph); |
1da177e4 | 1126 | /* The embedded headers contain source and dest in reverse order */ |
51ef348b | 1127 | cp = pp->conn_in_get(AF_INET, skb, pp, &ciph, offset, 1); |
4856c84c MT |
1128 | if (!cp) { |
1129 | /* The packet could also belong to a local client */ | |
1130 | cp = pp->conn_out_get(AF_INET, skb, pp, &ciph, offset, 1); | |
f2428ed5 SH |
1131 | if (cp) { |
1132 | snet.ip = iph->saddr; | |
1133 | return handle_response_icmp(AF_INET, skb, &snet, | |
1134 | cih->protocol, cp, pp, | |
4856c84c | 1135 | offset, ihl); |
f2428ed5 | 1136 | } |
1da177e4 | 1137 | return NF_ACCEPT; |
4856c84c | 1138 | } |
1da177e4 LT |
1139 | |
1140 | verdict = NF_DROP; | |
1141 | ||
1142 | /* Ensure the checksum is correct */ | |
60476372 | 1143 | if (!skb_csum_unnecessary(skb) && ip_vs_checksum_complete(skb, ihl)) { |
1da177e4 | 1144 | /* Failed checksum! */ |
14d5e834 HH |
1145 | IP_VS_DBG(1, "Incoming ICMP: failed checksum from %pI4!\n", |
1146 | &iph->saddr); | |
1da177e4 LT |
1147 | goto out; |
1148 | } | |
1149 | ||
1150 | /* do the statistics and put it back */ | |
1151 | ip_vs_in_stats(cp, skb); | |
1152 | if (IPPROTO_TCP == cih->protocol || IPPROTO_UDP == cih->protocol) | |
1153 | offset += 2 * sizeof(__u16); | |
1154 | verdict = ip_vs_icmp_xmit(skb, cp, pp, offset); | |
1155 | /* do not touch skb anymore */ | |
1156 | ||
1157 | out: | |
1158 | __ip_vs_conn_put(cp); | |
1159 | ||
1160 | return verdict; | |
1161 | } | |
1162 | ||
2a3b791e JV |
1163 | #ifdef CONFIG_IP_VS_IPV6 |
1164 | static int | |
1165 | ip_vs_in_icmp_v6(struct sk_buff *skb, int *related, unsigned int hooknum) | |
1166 | { | |
1167 | struct ipv6hdr *iph; | |
1168 | struct icmp6hdr _icmph, *ic; | |
1169 | struct ipv6hdr _ciph, *cih; /* The ip header contained | |
1170 | within the ICMP */ | |
1171 | struct ip_vs_iphdr ciph; | |
1172 | struct ip_vs_conn *cp; | |
1173 | struct ip_vs_protocol *pp; | |
1174 | unsigned int offset, verdict; | |
f2428ed5 | 1175 | union nf_inet_addr snet; |
2a3b791e JV |
1176 | |
1177 | *related = 1; | |
1178 | ||
1179 | /* reassemble IP fragments */ | |
1180 | if (ipv6_hdr(skb)->nexthdr == IPPROTO_FRAGMENT) { | |
1181 | if (ip_vs_gather_frags_v6(skb, hooknum == NF_INET_LOCAL_IN ? | |
1182 | IP_DEFRAG_VS_IN : | |
1183 | IP_DEFRAG_VS_FWD)) | |
1184 | return NF_STOLEN; | |
1185 | } | |
1186 | ||
1187 | iph = ipv6_hdr(skb); | |
1188 | offset = sizeof(struct ipv6hdr); | |
1189 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
1190 | if (ic == NULL) | |
1191 | return NF_DROP; | |
1192 | ||
5b095d98 | 1193 | IP_VS_DBG(12, "Incoming ICMPv6 (%d,%d) %pI6->%pI6\n", |
2a3b791e | 1194 | ic->icmp6_type, ntohs(icmpv6_id(ic)), |
38ff4fa4 | 1195 | &iph->saddr, &iph->daddr); |
2a3b791e JV |
1196 | |
1197 | /* | |
1198 | * Work through seeing if this is for us. | |
1199 | * These checks are supposed to be in an order that means easy | |
1200 | * things are checked first to speed up processing.... however | |
1201 | * this means that some packets will manage to get a long way | |
1202 | * down this stack and then be rejected, but that's life. | |
1203 | */ | |
1204 | if ((ic->icmp6_type != ICMPV6_DEST_UNREACH) && | |
1205 | (ic->icmp6_type != ICMPV6_PKT_TOOBIG) && | |
1206 | (ic->icmp6_type != ICMPV6_TIME_EXCEED)) { | |
1207 | *related = 0; | |
1208 | return NF_ACCEPT; | |
1209 | } | |
1210 | ||
1211 | /* Now find the contained IP header */ | |
1212 | offset += sizeof(_icmph); | |
1213 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
1214 | if (cih == NULL) | |
1215 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
1216 | ||
1217 | pp = ip_vs_proto_get(cih->nexthdr); | |
1218 | if (!pp) | |
1219 | return NF_ACCEPT; | |
1220 | ||
1221 | /* Is the embedded protocol header present? */ | |
1222 | /* TODO: we don't support fragmentation at the moment anyways */ | |
1223 | if (unlikely(cih->nexthdr == IPPROTO_FRAGMENT && pp->dont_defrag)) | |
1224 | return NF_ACCEPT; | |
1225 | ||
1226 | IP_VS_DBG_PKT(11, pp, skb, offset, "Checking incoming ICMPv6 for"); | |
1227 | ||
1228 | offset += sizeof(struct ipv6hdr); | |
1229 | ||
1230 | ip_vs_fill_iphdr(AF_INET6, cih, &ciph); | |
1231 | /* The embedded headers contain source and dest in reverse order */ | |
1232 | cp = pp->conn_in_get(AF_INET6, skb, pp, &ciph, offset, 1); | |
f2428ed5 SH |
1233 | if (!cp) { |
1234 | /* The packet could also belong to a local client */ | |
1235 | cp = pp->conn_out_get(AF_INET6, skb, pp, &ciph, offset, 1); | |
1236 | if (cp) { | |
178f5e49 | 1237 | ipv6_addr_copy(&snet.in6, &iph->saddr); |
f2428ed5 SH |
1238 | return handle_response_icmp(AF_INET6, skb, &snet, |
1239 | cih->nexthdr, | |
1240 | cp, pp, offset, | |
1241 | sizeof(struct ipv6hdr)); | |
1242 | } | |
2a3b791e | 1243 | return NF_ACCEPT; |
f2428ed5 | 1244 | } |
2a3b791e JV |
1245 | |
1246 | verdict = NF_DROP; | |
1247 | ||
1248 | /* do the statistics and put it back */ | |
1249 | ip_vs_in_stats(cp, skb); | |
2906f66a VMR |
1250 | if (IPPROTO_TCP == cih->nexthdr || IPPROTO_UDP == cih->nexthdr || |
1251 | IPPROTO_SCTP == cih->nexthdr) | |
2a3b791e JV |
1252 | offset += 2 * sizeof(__u16); |
1253 | verdict = ip_vs_icmp_xmit_v6(skb, cp, pp, offset); | |
1254 | /* do not touch skb anymore */ | |
1255 | ||
1256 | __ip_vs_conn_put(cp); | |
1257 | ||
1258 | return verdict; | |
1259 | } | |
1260 | #endif | |
1261 | ||
1262 | ||
1da177e4 LT |
1263 | /* |
1264 | * Check if it's for virtual services, look it up, | |
1265 | * and send it on its way... | |
1266 | */ | |
1267 | static unsigned int | |
3db05fea | 1268 | ip_vs_in(unsigned int hooknum, struct sk_buff *skb, |
1da177e4 LT |
1269 | const struct net_device *in, const struct net_device *out, |
1270 | int (*okfn)(struct sk_buff *)) | |
1271 | { | |
51ef348b | 1272 | struct ip_vs_iphdr iph; |
1da177e4 LT |
1273 | struct ip_vs_protocol *pp; |
1274 | struct ip_vs_conn *cp; | |
1e66dafc | 1275 | int ret, restart, af, pkts; |
2a3b791e | 1276 | |
60678040 | 1277 | af = (skb->protocol == htons(ETH_P_IP)) ? AF_INET : AF_INET6; |
51ef348b | 1278 | |
2a3b791e | 1279 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); |
1da177e4 LT |
1280 | |
1281 | /* | |
4856c84c MT |
1282 | * Big tappo: only PACKET_HOST, including loopback for local client |
1283 | * Don't handle local packets on IPv6 for now | |
1da177e4 | 1284 | */ |
f2428ed5 | 1285 | if (unlikely(skb->pkt_type != PACKET_HOST)) { |
51ef348b JV |
1286 | IP_VS_DBG_BUF(12, "packet type=%d proto=%d daddr=%s ignored\n", |
1287 | skb->pkt_type, | |
1288 | iph.protocol, | |
2a3b791e | 1289 | IP_VS_DBG_ADDR(af, &iph.daddr)); |
1da177e4 LT |
1290 | return NF_ACCEPT; |
1291 | } | |
1292 | ||
94b26551 JV |
1293 | #ifdef CONFIG_IP_VS_IPV6 |
1294 | if (af == AF_INET6) { | |
1295 | if (unlikely(iph.protocol == IPPROTO_ICMPV6)) { | |
1296 | int related, verdict = ip_vs_in_icmp_v6(skb, &related, hooknum); | |
1da177e4 | 1297 | |
94b26551 JV |
1298 | if (related) |
1299 | return verdict; | |
1300 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); | |
1301 | } | |
1302 | } else | |
1303 | #endif | |
1304 | if (unlikely(iph.protocol == IPPROTO_ICMP)) { | |
1305 | int related, verdict = ip_vs_in_icmp(skb, &related, hooknum); | |
1306 | ||
1307 | if (related) | |
1308 | return verdict; | |
1309 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); | |
1310 | } | |
1da177e4 LT |
1311 | |
1312 | /* Protocol supported? */ | |
51ef348b | 1313 | pp = ip_vs_proto_get(iph.protocol); |
1da177e4 LT |
1314 | if (unlikely(!pp)) |
1315 | return NF_ACCEPT; | |
1316 | ||
1da177e4 LT |
1317 | /* |
1318 | * Check if the packet belongs to an existing connection entry | |
1319 | */ | |
2a3b791e | 1320 | cp = pp->conn_in_get(af, skb, pp, &iph, iph.len, 0); |
1da177e4 LT |
1321 | |
1322 | if (unlikely(!cp)) { | |
1323 | int v; | |
1324 | ||
4856c84c MT |
1325 | /* For local client packets, it could be a response */ |
1326 | cp = pp->conn_out_get(af, skb, pp, &iph, iph.len, 0); | |
1327 | if (cp) | |
1328 | return handle_response(af, skb, pp, cp, iph.len); | |
1329 | ||
2a3b791e | 1330 | if (!pp->conn_schedule(af, skb, pp, &v, &cp)) |
1da177e4 LT |
1331 | return v; |
1332 | } | |
1333 | ||
1334 | if (unlikely(!cp)) { | |
1335 | /* sorry, all this trouble for a no-hit :) */ | |
1336 | IP_VS_DBG_PKT(12, pp, skb, 0, | |
1337 | "packet continues traversal as normal"); | |
1338 | return NF_ACCEPT; | |
1339 | } | |
1340 | ||
1341 | IP_VS_DBG_PKT(11, pp, skb, 0, "Incoming packet"); | |
1342 | ||
1343 | /* Check the server status */ | |
1344 | if (cp->dest && !(cp->dest->flags & IP_VS_DEST_F_AVAILABLE)) { | |
1345 | /* the destination server is not available */ | |
1346 | ||
1347 | if (sysctl_ip_vs_expire_nodest_conn) { | |
1348 | /* try to expire the connection immediately */ | |
1349 | ip_vs_conn_expire_now(cp); | |
1da177e4 | 1350 | } |
dc8103f2 JA |
1351 | /* don't restart its timer, and silently |
1352 | drop the packet. */ | |
1353 | __ip_vs_conn_put(cp); | |
1da177e4 LT |
1354 | return NF_DROP; |
1355 | } | |
1356 | ||
1357 | ip_vs_in_stats(cp, skb); | |
1358 | restart = ip_vs_set_state(cp, IP_VS_DIR_INPUT, skb, pp); | |
1359 | if (cp->packet_xmit) | |
1360 | ret = cp->packet_xmit(skb, cp, pp); | |
1361 | /* do not touch skb anymore */ | |
1362 | else { | |
1363 | IP_VS_DBG_RL("warning: packet_xmit is null"); | |
1364 | ret = NF_ACCEPT; | |
1365 | } | |
1366 | ||
efac5276 RB |
1367 | /* Increase its packet counter and check if it is needed |
1368 | * to be synchronized | |
1369 | * | |
1370 | * Sync connection if it is about to close to | |
1371 | * encorage the standby servers to update the connections timeout | |
1372 | */ | |
1e66dafc | 1373 | pkts = atomic_add_return(1, &cp->in_pkts); |
2906f66a VMR |
1374 | if (af == AF_INET && (ip_vs_sync_state & IP_VS_STATE_MASTER) && |
1375 | cp->protocol == IPPROTO_SCTP) { | |
1376 | if ((cp->state == IP_VS_SCTP_S_ESTABLISHED && | |
8ed2163f | 1377 | (pkts % sysctl_ip_vs_sync_threshold[1] |
2906f66a VMR |
1378 | == sysctl_ip_vs_sync_threshold[0])) || |
1379 | (cp->old_state != cp->state && | |
1380 | ((cp->state == IP_VS_SCTP_S_CLOSED) || | |
1381 | (cp->state == IP_VS_SCTP_S_SHUT_ACK_CLI) || | |
1382 | (cp->state == IP_VS_SCTP_S_SHUT_ACK_SER)))) { | |
1383 | ip_vs_sync_conn(cp); | |
1384 | goto out; | |
1385 | } | |
1386 | } | |
1387 | ||
8ed2163f JA |
1388 | /* Keep this block last: TCP and others with pp->num_states <= 1 */ |
1389 | else if (af == AF_INET && | |
c6883f58 | 1390 | (ip_vs_sync_state & IP_VS_STATE_MASTER) && |
efac5276 RB |
1391 | (((cp->protocol != IPPROTO_TCP || |
1392 | cp->state == IP_VS_TCP_S_ESTABLISHED) && | |
1e66dafc | 1393 | (pkts % sysctl_ip_vs_sync_threshold[1] |
efac5276 RB |
1394 | == sysctl_ip_vs_sync_threshold[0])) || |
1395 | ((cp->protocol == IPPROTO_TCP) && (cp->old_state != cp->state) && | |
1396 | ((cp->state == IP_VS_TCP_S_FIN_WAIT) || | |
9abfe315 | 1397 | (cp->state == IP_VS_TCP_S_CLOSE) || |
9d3a0de7 RB |
1398 | (cp->state == IP_VS_TCP_S_CLOSE_WAIT) || |
1399 | (cp->state == IP_VS_TCP_S_TIME_WAIT))))) | |
1da177e4 | 1400 | ip_vs_sync_conn(cp); |
2906f66a | 1401 | out: |
efac5276 | 1402 | cp->old_state = cp->state; |
1da177e4 LT |
1403 | |
1404 | ip_vs_conn_put(cp); | |
1405 | return ret; | |
1406 | } | |
1407 | ||
1408 | ||
1409 | /* | |
6e23ae2a | 1410 | * It is hooked at the NF_INET_FORWARD chain, in order to catch ICMP |
1da177e4 LT |
1411 | * related packets destined for 0.0.0.0/0. |
1412 | * When fwmark-based virtual service is used, such as transparent | |
1413 | * cache cluster, TCP packets can be marked and routed to ip_vs_in, | |
1414 | * but ICMP destined for 0.0.0.0/0 cannot not be easily marked and | |
6e23ae2a | 1415 | * sent to ip_vs_in_icmp. So, catch them at the NF_INET_FORWARD chain |
1da177e4 LT |
1416 | * and send them to ip_vs_in_icmp. |
1417 | */ | |
1418 | static unsigned int | |
3db05fea | 1419 | ip_vs_forward_icmp(unsigned int hooknum, struct sk_buff *skb, |
1da177e4 LT |
1420 | const struct net_device *in, const struct net_device *out, |
1421 | int (*okfn)(struct sk_buff *)) | |
1422 | { | |
1423 | int r; | |
1424 | ||
3db05fea | 1425 | if (ip_hdr(skb)->protocol != IPPROTO_ICMP) |
1da177e4 LT |
1426 | return NF_ACCEPT; |
1427 | ||
3db05fea | 1428 | return ip_vs_in_icmp(skb, &r, hooknum); |
1da177e4 LT |
1429 | } |
1430 | ||
2a3b791e JV |
1431 | #ifdef CONFIG_IP_VS_IPV6 |
1432 | static unsigned int | |
1433 | ip_vs_forward_icmp_v6(unsigned int hooknum, struct sk_buff *skb, | |
1434 | const struct net_device *in, const struct net_device *out, | |
1435 | int (*okfn)(struct sk_buff *)) | |
1436 | { | |
1437 | int r; | |
1438 | ||
1439 | if (ipv6_hdr(skb)->nexthdr != IPPROTO_ICMPV6) | |
1440 | return NF_ACCEPT; | |
1441 | ||
1442 | return ip_vs_in_icmp_v6(skb, &r, hooknum); | |
1443 | } | |
1444 | #endif | |
1445 | ||
1da177e4 | 1446 | |
1999414a | 1447 | static struct nf_hook_ops ip_vs_ops[] __read_mostly = { |
41c5b317 PM |
1448 | /* After packet filtering, forward packet through VS/DR, VS/TUN, |
1449 | * or VS/NAT(change destination), so that filtering rules can be | |
1450 | * applied to IPVS. */ | |
1451 | { | |
1452 | .hook = ip_vs_in, | |
1453 | .owner = THIS_MODULE, | |
1454 | .pf = PF_INET, | |
1455 | .hooknum = NF_INET_LOCAL_IN, | |
1456 | .priority = 100, | |
1457 | }, | |
1458 | /* After packet filtering, change source only for VS/NAT */ | |
1459 | { | |
1460 | .hook = ip_vs_out, | |
1461 | .owner = THIS_MODULE, | |
1462 | .pf = PF_INET, | |
1463 | .hooknum = NF_INET_FORWARD, | |
1464 | .priority = 100, | |
1465 | }, | |
1466 | /* After packet filtering (but before ip_vs_out_icmp), catch icmp | |
1467 | * destined for 0.0.0.0/0, which is for incoming IPVS connections */ | |
1468 | { | |
1469 | .hook = ip_vs_forward_icmp, | |
1470 | .owner = THIS_MODULE, | |
1471 | .pf = PF_INET, | |
1472 | .hooknum = NF_INET_FORWARD, | |
1473 | .priority = 99, | |
1474 | }, | |
f4bc17cd JA |
1475 | /* Before the netfilter connection tracking, exit from POST_ROUTING */ |
1476 | { | |
1477 | .hook = ip_vs_post_routing, | |
1478 | .owner = THIS_MODULE, | |
1479 | .pf = PF_INET, | |
1480 | .hooknum = NF_INET_POST_ROUTING, | |
1481 | .priority = NF_IP_PRI_NAT_SRC-1, | |
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 | }, | |
f4bc17cd JA |
1511 | /* Before the netfilter connection tracking, exit from POST_ROUTING */ |
1512 | { | |
1513 | .hook = ip_vs_post_routing, | |
1514 | .owner = THIS_MODULE, | |
1515 | .pf = PF_INET6, | |
1516 | .hooknum = NF_INET_POST_ROUTING, | |
1517 | .priority = NF_IP6_PRI_NAT_SRC-1, | |
1518 | }, | |
473b23d3 | 1519 | #endif |
1da177e4 LT |
1520 | }; |
1521 | ||
1522 | ||
1523 | /* | |
1524 | * Initialize IP Virtual Server | |
1525 | */ | |
1526 | static int __init ip_vs_init(void) | |
1527 | { | |
1528 | int ret; | |
1529 | ||
a919cf4b SW |
1530 | ip_vs_estimator_init(); |
1531 | ||
1da177e4 LT |
1532 | ret = ip_vs_control_init(); |
1533 | if (ret < 0) { | |
1e3e238e | 1534 | pr_err("can't setup control.\n"); |
a919cf4b | 1535 | goto cleanup_estimator; |
1da177e4 LT |
1536 | } |
1537 | ||
1538 | ip_vs_protocol_init(); | |
1539 | ||
1540 | ret = ip_vs_app_init(); | |
1541 | if (ret < 0) { | |
1e3e238e | 1542 | pr_err("can't setup application helper.\n"); |
1da177e4 LT |
1543 | goto cleanup_protocol; |
1544 | } | |
1545 | ||
1546 | ret = ip_vs_conn_init(); | |
1547 | if (ret < 0) { | |
1e3e238e | 1548 | pr_err("can't setup connection table.\n"); |
1da177e4 LT |
1549 | goto cleanup_app; |
1550 | } | |
1551 | ||
41c5b317 | 1552 | ret = nf_register_hooks(ip_vs_ops, ARRAY_SIZE(ip_vs_ops)); |
1da177e4 | 1553 | if (ret < 0) { |
1e3e238e | 1554 | pr_err("can't register hooks.\n"); |
1da177e4 LT |
1555 | goto cleanup_conn; |
1556 | } | |
1557 | ||
1e3e238e | 1558 | pr_info("ipvs loaded.\n"); |
1da177e4 LT |
1559 | return ret; |
1560 | ||
1da177e4 LT |
1561 | cleanup_conn: |
1562 | ip_vs_conn_cleanup(); | |
1563 | cleanup_app: | |
1564 | ip_vs_app_cleanup(); | |
1565 | cleanup_protocol: | |
1566 | ip_vs_protocol_cleanup(); | |
1567 | ip_vs_control_cleanup(); | |
a919cf4b SW |
1568 | cleanup_estimator: |
1569 | ip_vs_estimator_cleanup(); | |
1da177e4 LT |
1570 | return ret; |
1571 | } | |
1572 | ||
1573 | static void __exit ip_vs_cleanup(void) | |
1574 | { | |
41c5b317 | 1575 | nf_unregister_hooks(ip_vs_ops, ARRAY_SIZE(ip_vs_ops)); |
1da177e4 LT |
1576 | ip_vs_conn_cleanup(); |
1577 | ip_vs_app_cleanup(); | |
1578 | ip_vs_protocol_cleanup(); | |
1579 | ip_vs_control_cleanup(); | |
a919cf4b | 1580 | ip_vs_estimator_cleanup(); |
1e3e238e | 1581 | pr_info("ipvs unloaded.\n"); |
1da177e4 LT |
1582 | } |
1583 | ||
1584 | module_init(ip_vs_init); | |
1585 | module_exit(ip_vs_cleanup); | |
1586 | MODULE_LICENSE("GPL"); |