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 | ||
27 | #include <linux/module.h> | |
28 | #include <linux/kernel.h> | |
29 | #include <linux/ip.h> | |
30 | #include <linux/tcp.h> | |
31 | #include <linux/icmp.h> | |
32 | ||
33 | #include <net/ip.h> | |
34 | #include <net/tcp.h> | |
35 | #include <net/udp.h> | |
36 | #include <net/icmp.h> /* for icmp_send */ | |
37 | #include <net/route.h> | |
38 | ||
39 | #include <linux/netfilter.h> | |
40 | #include <linux/netfilter_ipv4.h> | |
41 | ||
42 | #include <net/ip_vs.h> | |
43 | ||
44 | ||
45 | EXPORT_SYMBOL(register_ip_vs_scheduler); | |
46 | EXPORT_SYMBOL(unregister_ip_vs_scheduler); | |
47 | EXPORT_SYMBOL(ip_vs_skb_replace); | |
48 | EXPORT_SYMBOL(ip_vs_proto_name); | |
49 | EXPORT_SYMBOL(ip_vs_conn_new); | |
50 | EXPORT_SYMBOL(ip_vs_conn_in_get); | |
51 | EXPORT_SYMBOL(ip_vs_conn_out_get); | |
52 | #ifdef CONFIG_IP_VS_PROTO_TCP | |
53 | EXPORT_SYMBOL(ip_vs_tcp_conn_listen); | |
54 | #endif | |
55 | EXPORT_SYMBOL(ip_vs_conn_put); | |
56 | #ifdef CONFIG_IP_VS_DEBUG | |
57 | EXPORT_SYMBOL(ip_vs_get_debug_level); | |
58 | #endif | |
1da177e4 LT |
59 | |
60 | ||
61 | /* ID used in ICMP lookups */ | |
62 | #define icmp_id(icmph) (((icmph)->un).echo.id) | |
63 | ||
64 | const char *ip_vs_proto_name(unsigned proto) | |
65 | { | |
66 | static char buf[20]; | |
67 | ||
68 | switch (proto) { | |
69 | case IPPROTO_IP: | |
70 | return "IP"; | |
71 | case IPPROTO_UDP: | |
72 | return "UDP"; | |
73 | case IPPROTO_TCP: | |
74 | return "TCP"; | |
75 | case IPPROTO_ICMP: | |
76 | return "ICMP"; | |
77 | default: | |
78 | sprintf(buf, "IP_%d", proto); | |
79 | return buf; | |
80 | } | |
81 | } | |
82 | ||
83 | void ip_vs_init_hash_table(struct list_head *table, int rows) | |
84 | { | |
85 | while (--rows >= 0) | |
86 | INIT_LIST_HEAD(&table[rows]); | |
87 | } | |
88 | ||
89 | static inline void | |
90 | ip_vs_in_stats(struct ip_vs_conn *cp, struct sk_buff *skb) | |
91 | { | |
92 | struct ip_vs_dest *dest = cp->dest; | |
93 | if (dest && (dest->flags & IP_VS_DEST_F_AVAILABLE)) { | |
94 | spin_lock(&dest->stats.lock); | |
95 | dest->stats.inpkts++; | |
96 | dest->stats.inbytes += skb->len; | |
97 | spin_unlock(&dest->stats.lock); | |
98 | ||
99 | spin_lock(&dest->svc->stats.lock); | |
100 | dest->svc->stats.inpkts++; | |
101 | dest->svc->stats.inbytes += skb->len; | |
102 | spin_unlock(&dest->svc->stats.lock); | |
103 | ||
104 | spin_lock(&ip_vs_stats.lock); | |
105 | ip_vs_stats.inpkts++; | |
106 | ip_vs_stats.inbytes += skb->len; | |
107 | spin_unlock(&ip_vs_stats.lock); | |
108 | } | |
109 | } | |
110 | ||
111 | ||
112 | static inline void | |
113 | ip_vs_out_stats(struct ip_vs_conn *cp, struct sk_buff *skb) | |
114 | { | |
115 | struct ip_vs_dest *dest = cp->dest; | |
116 | if (dest && (dest->flags & IP_VS_DEST_F_AVAILABLE)) { | |
117 | spin_lock(&dest->stats.lock); | |
118 | dest->stats.outpkts++; | |
119 | dest->stats.outbytes += skb->len; | |
120 | spin_unlock(&dest->stats.lock); | |
121 | ||
122 | spin_lock(&dest->svc->stats.lock); | |
123 | dest->svc->stats.outpkts++; | |
124 | dest->svc->stats.outbytes += skb->len; | |
125 | spin_unlock(&dest->svc->stats.lock); | |
126 | ||
127 | spin_lock(&ip_vs_stats.lock); | |
128 | ip_vs_stats.outpkts++; | |
129 | ip_vs_stats.outbytes += skb->len; | |
130 | spin_unlock(&ip_vs_stats.lock); | |
131 | } | |
132 | } | |
133 | ||
134 | ||
135 | static inline void | |
136 | ip_vs_conn_stats(struct ip_vs_conn *cp, struct ip_vs_service *svc) | |
137 | { | |
138 | spin_lock(&cp->dest->stats.lock); | |
139 | cp->dest->stats.conns++; | |
140 | spin_unlock(&cp->dest->stats.lock); | |
141 | ||
142 | spin_lock(&svc->stats.lock); | |
143 | svc->stats.conns++; | |
144 | spin_unlock(&svc->stats.lock); | |
145 | ||
146 | spin_lock(&ip_vs_stats.lock); | |
147 | ip_vs_stats.conns++; | |
148 | spin_unlock(&ip_vs_stats.lock); | |
149 | } | |
150 | ||
151 | ||
152 | static inline int | |
153 | ip_vs_set_state(struct ip_vs_conn *cp, int direction, | |
154 | const struct sk_buff *skb, | |
155 | struct ip_vs_protocol *pp) | |
156 | { | |
157 | if (unlikely(!pp->state_transition)) | |
158 | return 0; | |
159 | return pp->state_transition(cp, direction, skb, pp); | |
160 | } | |
161 | ||
162 | ||
1da177e4 LT |
163 | /* |
164 | * IPVS persistent scheduling function | |
165 | * It creates a connection entry according to its template if exists, | |
166 | * or selects a server and creates a connection entry plus a template. | |
167 | * Locking: we are svc user (svc->refcnt), so we hold all dests too | |
168 | * Protocols supported: TCP, UDP | |
169 | */ | |
170 | static struct ip_vs_conn * | |
171 | ip_vs_sched_persist(struct ip_vs_service *svc, | |
172 | const struct sk_buff *skb, | |
014d730d | 173 | __be16 ports[2]) |
1da177e4 LT |
174 | { |
175 | struct ip_vs_conn *cp = NULL; | |
eddc9ec5 | 176 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 LT |
177 | struct ip_vs_dest *dest; |
178 | struct ip_vs_conn *ct; | |
014d730d AV |
179 | __be16 dport; /* destination port to forward */ |
180 | __be32 snet; /* source network of the client, after masking */ | |
1da177e4 LT |
181 | |
182 | /* Mask saddr with the netmask to adjust template granularity */ | |
183 | snet = iph->saddr & svc->netmask; | |
184 | ||
185 | IP_VS_DBG(6, "p-schedule: src %u.%u.%u.%u:%u dest %u.%u.%u.%u:%u " | |
186 | "mnet %u.%u.%u.%u\n", | |
187 | NIPQUAD(iph->saddr), ntohs(ports[0]), | |
188 | NIPQUAD(iph->daddr), ntohs(ports[1]), | |
189 | NIPQUAD(snet)); | |
190 | ||
191 | /* | |
192 | * As far as we know, FTP is a very complicated network protocol, and | |
193 | * it uses control connection and data connections. For active FTP, | |
194 | * FTP server initialize data connection to the client, its source port | |
195 | * is often 20. For passive FTP, FTP server tells the clients the port | |
196 | * that it passively listens to, and the client issues the data | |
197 | * connection. In the tunneling or direct routing mode, the load | |
198 | * balancer is on the client-to-server half of connection, the port | |
199 | * number is unknown to the load balancer. So, a conn template like | |
200 | * <caddr, 0, vaddr, 0, daddr, 0> is created for persistent FTP | |
201 | * service, and a template like <caddr, 0, vaddr, vport, daddr, dport> | |
202 | * is created for other persistent services. | |
203 | */ | |
204 | if (ports[1] == svc->port) { | |
205 | /* Check if a template already exists */ | |
206 | if (svc->port != FTPPORT) | |
87375ab4 | 207 | ct = ip_vs_ct_in_get(iph->protocol, snet, 0, |
1da177e4 LT |
208 | iph->daddr, ports[1]); |
209 | else | |
87375ab4 | 210 | ct = ip_vs_ct_in_get(iph->protocol, snet, 0, |
1da177e4 LT |
211 | iph->daddr, 0); |
212 | ||
213 | if (!ct || !ip_vs_check_template(ct)) { | |
214 | /* | |
215 | * No template found or the dest of the connection | |
216 | * template is not available. | |
217 | */ | |
218 | dest = svc->scheduler->schedule(svc, skb); | |
219 | if (dest == NULL) { | |
220 | IP_VS_DBG(1, "p-schedule: no dest found.\n"); | |
221 | return NULL; | |
222 | } | |
223 | ||
224 | /* | |
225 | * Create a template like <protocol,caddr,0, | |
226 | * vaddr,vport,daddr,dport> for non-ftp service, | |
227 | * and <protocol,caddr,0,vaddr,0,daddr,0> | |
228 | * for ftp service. | |
229 | */ | |
230 | if (svc->port != FTPPORT) | |
231 | ct = ip_vs_conn_new(iph->protocol, | |
232 | snet, 0, | |
233 | iph->daddr, | |
234 | ports[1], | |
e7ade46a | 235 | dest->addr.ip, dest->port, |
87375ab4 | 236 | IP_VS_CONN_F_TEMPLATE, |
1da177e4 LT |
237 | dest); |
238 | else | |
239 | ct = ip_vs_conn_new(iph->protocol, | |
240 | snet, 0, | |
241 | iph->daddr, 0, | |
e7ade46a | 242 | dest->addr.ip, 0, |
87375ab4 | 243 | IP_VS_CONN_F_TEMPLATE, |
1da177e4 LT |
244 | dest); |
245 | if (ct == NULL) | |
246 | return NULL; | |
247 | ||
248 | ct->timeout = svc->timeout; | |
249 | } else { | |
250 | /* set destination with the found template */ | |
251 | dest = ct->dest; | |
252 | } | |
253 | dport = dest->port; | |
254 | } else { | |
255 | /* | |
256 | * Note: persistent fwmark-based services and persistent | |
257 | * port zero service are handled here. | |
258 | * fwmark template: <IPPROTO_IP,caddr,0,fwmark,0,daddr,0> | |
259 | * port zero template: <protocol,caddr,0,vaddr,0,daddr,0> | |
260 | */ | |
261 | if (svc->fwmark) | |
87375ab4 | 262 | ct = ip_vs_ct_in_get(IPPROTO_IP, snet, 0, |
1da177e4 LT |
263 | htonl(svc->fwmark), 0); |
264 | else | |
87375ab4 | 265 | ct = ip_vs_ct_in_get(iph->protocol, snet, 0, |
1da177e4 LT |
266 | iph->daddr, 0); |
267 | ||
268 | if (!ct || !ip_vs_check_template(ct)) { | |
269 | /* | |
270 | * If it is not persistent port zero, return NULL, | |
271 | * otherwise create a connection template. | |
272 | */ | |
273 | if (svc->port) | |
274 | return NULL; | |
275 | ||
276 | dest = svc->scheduler->schedule(svc, skb); | |
277 | if (dest == NULL) { | |
278 | IP_VS_DBG(1, "p-schedule: no dest found.\n"); | |
279 | return NULL; | |
280 | } | |
281 | ||
282 | /* | |
283 | * Create a template according to the service | |
284 | */ | |
285 | if (svc->fwmark) | |
286 | ct = ip_vs_conn_new(IPPROTO_IP, | |
287 | snet, 0, | |
288 | htonl(svc->fwmark), 0, | |
e7ade46a | 289 | dest->addr.ip, 0, |
87375ab4 | 290 | IP_VS_CONN_F_TEMPLATE, |
1da177e4 LT |
291 | dest); |
292 | else | |
293 | ct = ip_vs_conn_new(iph->protocol, | |
294 | snet, 0, | |
295 | iph->daddr, 0, | |
e7ade46a | 296 | dest->addr.ip, 0, |
87375ab4 | 297 | IP_VS_CONN_F_TEMPLATE, |
1da177e4 LT |
298 | dest); |
299 | if (ct == NULL) | |
300 | return NULL; | |
301 | ||
302 | ct->timeout = svc->timeout; | |
303 | } else { | |
304 | /* set destination with the found template */ | |
305 | dest = ct->dest; | |
306 | } | |
307 | dport = ports[1]; | |
308 | } | |
309 | ||
310 | /* | |
311 | * Create a new connection according to the template | |
312 | */ | |
313 | cp = ip_vs_conn_new(iph->protocol, | |
314 | iph->saddr, ports[0], | |
315 | iph->daddr, ports[1], | |
e7ade46a | 316 | dest->addr.ip, dport, |
1da177e4 LT |
317 | 0, |
318 | dest); | |
319 | if (cp == NULL) { | |
320 | ip_vs_conn_put(ct); | |
321 | return NULL; | |
322 | } | |
323 | ||
324 | /* | |
325 | * Add its control | |
326 | */ | |
327 | ip_vs_control_add(cp, ct); | |
328 | ip_vs_conn_put(ct); | |
329 | ||
330 | ip_vs_conn_stats(cp, svc); | |
331 | return cp; | |
332 | } | |
333 | ||
334 | ||
335 | /* | |
336 | * IPVS main scheduling function | |
337 | * It selects a server according to the virtual service, and | |
338 | * creates a connection entry. | |
339 | * Protocols supported: TCP, UDP | |
340 | */ | |
341 | struct ip_vs_conn * | |
342 | ip_vs_schedule(struct ip_vs_service *svc, const struct sk_buff *skb) | |
343 | { | |
344 | struct ip_vs_conn *cp = NULL; | |
eddc9ec5 | 345 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 | 346 | struct ip_vs_dest *dest; |
014d730d | 347 | __be16 _ports[2], *pptr; |
1da177e4 LT |
348 | |
349 | pptr = skb_header_pointer(skb, iph->ihl*4, | |
350 | sizeof(_ports), _ports); | |
351 | if (pptr == NULL) | |
352 | return NULL; | |
353 | ||
354 | /* | |
355 | * Persistent service | |
356 | */ | |
357 | if (svc->flags & IP_VS_SVC_F_PERSISTENT) | |
358 | return ip_vs_sched_persist(svc, skb, pptr); | |
359 | ||
360 | /* | |
361 | * Non-persistent service | |
362 | */ | |
363 | if (!svc->fwmark && pptr[1] != svc->port) { | |
364 | if (!svc->port) | |
365 | IP_VS_ERR("Schedule: port zero only supported " | |
366 | "in persistent services, " | |
367 | "check your ipvs configuration\n"); | |
368 | return NULL; | |
369 | } | |
370 | ||
371 | dest = svc->scheduler->schedule(svc, skb); | |
372 | if (dest == NULL) { | |
373 | IP_VS_DBG(1, "Schedule: no dest found.\n"); | |
374 | return NULL; | |
375 | } | |
376 | ||
377 | /* | |
378 | * Create a connection entry. | |
379 | */ | |
380 | cp = ip_vs_conn_new(iph->protocol, | |
381 | iph->saddr, pptr[0], | |
382 | iph->daddr, pptr[1], | |
e7ade46a | 383 | dest->addr.ip, dest->port ? dest->port : pptr[1], |
1da177e4 LT |
384 | 0, |
385 | dest); | |
386 | if (cp == NULL) | |
387 | return NULL; | |
388 | ||
389 | IP_VS_DBG(6, "Schedule fwd:%c c:%u.%u.%u.%u:%u v:%u.%u.%u.%u:%u " | |
4b5bdf5c | 390 | "d:%u.%u.%u.%u:%u conn->flags:%X conn->refcnt:%d\n", |
1da177e4 | 391 | ip_vs_fwd_tag(cp), |
e7ade46a JV |
392 | NIPQUAD(cp->caddr.ip), ntohs(cp->cport), |
393 | NIPQUAD(cp->vaddr.ip), ntohs(cp->vport), | |
394 | NIPQUAD(cp->daddr.ip), ntohs(cp->dport), | |
1da177e4 LT |
395 | cp->flags, atomic_read(&cp->refcnt)); |
396 | ||
397 | ip_vs_conn_stats(cp, svc); | |
398 | return cp; | |
399 | } | |
400 | ||
401 | ||
402 | /* | |
403 | * Pass or drop the packet. | |
404 | * Called by ip_vs_in, when the virtual service is available but | |
405 | * no destination is available for a new connection. | |
406 | */ | |
407 | int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb, | |
408 | struct ip_vs_protocol *pp) | |
409 | { | |
014d730d | 410 | __be16 _ports[2], *pptr; |
eddc9ec5 | 411 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 LT |
412 | |
413 | pptr = skb_header_pointer(skb, iph->ihl*4, | |
414 | sizeof(_ports), _ports); | |
415 | if (pptr == NULL) { | |
416 | ip_vs_service_put(svc); | |
417 | return NF_DROP; | |
418 | } | |
419 | ||
420 | /* if it is fwmark-based service, the cache_bypass sysctl is up | |
421 | and the destination is RTN_UNICAST (and not local), then create | |
422 | a cache_bypass connection entry */ | |
423 | if (sysctl_ip_vs_cache_bypass && svc->fwmark | |
6b175b26 | 424 | && (inet_addr_type(&init_net, iph->daddr) == RTN_UNICAST)) { |
1da177e4 LT |
425 | int ret, cs; |
426 | struct ip_vs_conn *cp; | |
427 | ||
428 | ip_vs_service_put(svc); | |
429 | ||
430 | /* create a new connection entry */ | |
431 | IP_VS_DBG(6, "ip_vs_leave: create a cache_bypass entry\n"); | |
432 | cp = ip_vs_conn_new(iph->protocol, | |
433 | iph->saddr, pptr[0], | |
434 | iph->daddr, pptr[1], | |
435 | 0, 0, | |
436 | IP_VS_CONN_F_BYPASS, | |
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 | */ | |
476 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); | |
477 | return NF_DROP; | |
478 | } | |
479 | ||
480 | ||
481 | /* | |
6e23ae2a | 482 | * It is hooked before NF_IP_PRI_NAT_SRC at the NF_INET_POST_ROUTING |
1da177e4 LT |
483 | * chain, and is used for VS/NAT. |
484 | * It detects packets for VS/NAT connections and sends the packets | |
485 | * immediately. This can avoid that iptable_nat mangles the packets | |
486 | * for VS/NAT. | |
487 | */ | |
488 | static unsigned int ip_vs_post_routing(unsigned int hooknum, | |
3db05fea | 489 | struct sk_buff *skb, |
1da177e4 LT |
490 | const struct net_device *in, |
491 | const struct net_device *out, | |
492 | int (*okfn)(struct sk_buff *)) | |
493 | { | |
3db05fea | 494 | if (!skb->ipvs_property) |
1da177e4 | 495 | return NF_ACCEPT; |
1da177e4 | 496 | /* The packet was sent from IPVS, exit this chain */ |
abbcc739 | 497 | return NF_STOP; |
1da177e4 LT |
498 | } |
499 | ||
b1550f22 | 500 | __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset) |
1da177e4 | 501 | { |
d3bc23e7 | 502 | return csum_fold(skb_checksum(skb, offset, skb->len - offset, 0)); |
1da177e4 LT |
503 | } |
504 | ||
776c729e | 505 | static inline int ip_vs_gather_frags(struct sk_buff *skb, u_int32_t user) |
1da177e4 | 506 | { |
776c729e HX |
507 | int err = ip_defrag(skb, user); |
508 | ||
509 | if (!err) | |
eddc9ec5 | 510 | ip_send_check(ip_hdr(skb)); |
776c729e HX |
511 | |
512 | return err; | |
1da177e4 LT |
513 | } |
514 | ||
515 | /* | |
516 | * Packet has been made sufficiently writable in caller | |
517 | * - inout: 1=in->out, 0=out->in | |
518 | */ | |
519 | void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp, | |
520 | struct ip_vs_conn *cp, int inout) | |
521 | { | |
eddc9ec5 | 522 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 | 523 | unsigned int icmp_offset = iph->ihl*4; |
d56f90a7 ACM |
524 | struct icmphdr *icmph = (struct icmphdr *)(skb_network_header(skb) + |
525 | icmp_offset); | |
1da177e4 LT |
526 | struct iphdr *ciph = (struct iphdr *)(icmph + 1); |
527 | ||
528 | if (inout) { | |
e7ade46a | 529 | iph->saddr = cp->vaddr.ip; |
1da177e4 | 530 | ip_send_check(iph); |
e7ade46a | 531 | ciph->daddr = cp->vaddr.ip; |
1da177e4 LT |
532 | ip_send_check(ciph); |
533 | } else { | |
e7ade46a | 534 | iph->daddr = cp->daddr.ip; |
1da177e4 | 535 | ip_send_check(iph); |
e7ade46a | 536 | ciph->saddr = cp->daddr.ip; |
1da177e4 LT |
537 | ip_send_check(ciph); |
538 | } | |
539 | ||
540 | /* the TCP/UDP port */ | |
541 | if (IPPROTO_TCP == ciph->protocol || IPPROTO_UDP == ciph->protocol) { | |
014d730d | 542 | __be16 *ports = (void *)ciph + ciph->ihl*4; |
1da177e4 LT |
543 | |
544 | if (inout) | |
545 | ports[1] = cp->vport; | |
546 | else | |
547 | ports[0] = cp->dport; | |
548 | } | |
549 | ||
550 | /* And finally the ICMP checksum */ | |
551 | icmph->checksum = 0; | |
552 | icmph->checksum = ip_vs_checksum_complete(skb, icmp_offset); | |
553 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
554 | ||
555 | if (inout) | |
556 | IP_VS_DBG_PKT(11, pp, skb, (void *)ciph - (void *)iph, | |
557 | "Forwarding altered outgoing ICMP"); | |
558 | else | |
559 | IP_VS_DBG_PKT(11, pp, skb, (void *)ciph - (void *)iph, | |
560 | "Forwarding altered incoming ICMP"); | |
561 | } | |
562 | ||
563 | /* | |
564 | * Handle ICMP messages in the inside-to-outside direction (outgoing). | |
565 | * Find any that might be relevant, check against existing connections, | |
566 | * forward to the right destination host if relevant. | |
567 | * Currently handles error types - unreachable, quench, ttl exceeded. | |
568 | * (Only used in VS/NAT) | |
569 | */ | |
3db05fea | 570 | static int ip_vs_out_icmp(struct sk_buff *skb, int *related) |
1da177e4 | 571 | { |
1da177e4 LT |
572 | struct iphdr *iph; |
573 | struct icmphdr _icmph, *ic; | |
574 | struct iphdr _ciph, *cih; /* The ip header contained within the ICMP */ | |
51ef348b | 575 | struct ip_vs_iphdr ciph; |
1da177e4 LT |
576 | struct ip_vs_conn *cp; |
577 | struct ip_vs_protocol *pp; | |
578 | unsigned int offset, ihl, verdict; | |
579 | ||
580 | *related = 1; | |
581 | ||
582 | /* reassemble IP fragments */ | |
eddc9ec5 | 583 | if (ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)) { |
776c729e | 584 | if (ip_vs_gather_frags(skb, IP_DEFRAG_VS_OUT)) |
1da177e4 | 585 | return NF_STOLEN; |
1da177e4 LT |
586 | } |
587 | ||
eddc9ec5 | 588 | iph = ip_hdr(skb); |
1da177e4 LT |
589 | offset = ihl = iph->ihl * 4; |
590 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
591 | if (ic == NULL) | |
592 | return NF_DROP; | |
593 | ||
594 | IP_VS_DBG(12, "Outgoing ICMP (%d,%d) %u.%u.%u.%u->%u.%u.%u.%u\n", | |
595 | ic->type, ntohs(icmp_id(ic)), | |
596 | NIPQUAD(iph->saddr), NIPQUAD(iph->daddr)); | |
597 | ||
598 | /* | |
599 | * Work through seeing if this is for us. | |
600 | * These checks are supposed to be in an order that means easy | |
601 | * things are checked first to speed up processing.... however | |
602 | * this means that some packets will manage to get a long way | |
603 | * down this stack and then be rejected, but that's life. | |
604 | */ | |
605 | if ((ic->type != ICMP_DEST_UNREACH) && | |
606 | (ic->type != ICMP_SOURCE_QUENCH) && | |
607 | (ic->type != ICMP_TIME_EXCEEDED)) { | |
608 | *related = 0; | |
609 | return NF_ACCEPT; | |
610 | } | |
611 | ||
612 | /* Now find the contained IP header */ | |
613 | offset += sizeof(_icmph); | |
614 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
615 | if (cih == NULL) | |
616 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
617 | ||
618 | pp = ip_vs_proto_get(cih->protocol); | |
619 | if (!pp) | |
620 | return NF_ACCEPT; | |
621 | ||
622 | /* Is the embedded protocol header present? */ | |
4412ec49 | 623 | if (unlikely(cih->frag_off & htons(IP_OFFSET) && |
1da177e4 LT |
624 | pp->dont_defrag)) |
625 | return NF_ACCEPT; | |
626 | ||
627 | IP_VS_DBG_PKT(11, pp, skb, offset, "Checking outgoing ICMP for"); | |
628 | ||
629 | offset += cih->ihl * 4; | |
630 | ||
51ef348b | 631 | ip_vs_fill_iphdr(AF_INET, cih, &ciph); |
1da177e4 | 632 | /* The embedded headers contain source and dest in reverse order */ |
51ef348b | 633 | cp = pp->conn_out_get(AF_INET, skb, pp, &ciph, offset, 1); |
1da177e4 LT |
634 | if (!cp) |
635 | return NF_ACCEPT; | |
636 | ||
637 | verdict = NF_DROP; | |
638 | ||
639 | if (IP_VS_FWD_METHOD(cp) != 0) { | |
464c4f18 | 640 | IP_VS_ERR("shouldn't reach here, because the box is on the " |
1da177e4 LT |
641 | "half connection in the tun/dr module.\n"); |
642 | } | |
643 | ||
644 | /* Ensure the checksum is correct */ | |
60476372 | 645 | if (!skb_csum_unnecessary(skb) && ip_vs_checksum_complete(skb, ihl)) { |
1da177e4 LT |
646 | /* Failed checksum! */ |
647 | IP_VS_DBG(1, "Forward ICMP: failed checksum from %d.%d.%d.%d!\n", | |
648 | NIPQUAD(iph->saddr)); | |
649 | goto out; | |
650 | } | |
651 | ||
652 | if (IPPROTO_TCP == cih->protocol || IPPROTO_UDP == cih->protocol) | |
653 | offset += 2 * sizeof(__u16); | |
af1e1cf0 | 654 | if (!skb_make_writable(skb, offset)) |
1da177e4 | 655 | goto out; |
1da177e4 LT |
656 | |
657 | ip_vs_nat_icmp(skb, pp, cp, 1); | |
658 | ||
659 | /* do the statistics and put it back */ | |
660 | ip_vs_out_stats(cp, skb); | |
661 | ||
6869c4d8 | 662 | skb->ipvs_property = 1; |
1da177e4 LT |
663 | verdict = NF_ACCEPT; |
664 | ||
665 | out: | |
666 | __ip_vs_conn_put(cp); | |
667 | ||
668 | return verdict; | |
669 | } | |
670 | ||
671 | static inline int is_tcp_reset(const struct sk_buff *skb) | |
672 | { | |
673 | struct tcphdr _tcph, *th; | |
674 | ||
c9bdd4b5 | 675 | th = skb_header_pointer(skb, ip_hdrlen(skb), sizeof(_tcph), &_tcph); |
1da177e4 LT |
676 | if (th == NULL) |
677 | return 0; | |
678 | return th->rst; | |
679 | } | |
680 | ||
681 | /* | |
6e23ae2a | 682 | * It is hooked at the NF_INET_FORWARD chain, used only for VS/NAT. |
1da177e4 LT |
683 | * Check if outgoing packet belongs to the established ip_vs_conn, |
684 | * rewrite addresses of the packet and send it on its way... | |
685 | */ | |
686 | static unsigned int | |
3db05fea | 687 | ip_vs_out(unsigned int hooknum, struct sk_buff *skb, |
1da177e4 LT |
688 | const struct net_device *in, const struct net_device *out, |
689 | int (*okfn)(struct sk_buff *)) | |
690 | { | |
51ef348b | 691 | struct ip_vs_iphdr iph; |
1da177e4 LT |
692 | struct ip_vs_protocol *pp; |
693 | struct ip_vs_conn *cp; | |
1da177e4 LT |
694 | |
695 | EnterFunction(11); | |
696 | ||
6869c4d8 | 697 | if (skb->ipvs_property) |
1da177e4 LT |
698 | return NF_ACCEPT; |
699 | ||
51ef348b JV |
700 | ip_vs_fill_iphdr(AF_INET, skb_network_header(skb), &iph); |
701 | if (unlikely(iph.protocol == IPPROTO_ICMP)) { | |
3db05fea | 702 | int related, verdict = ip_vs_out_icmp(skb, &related); |
1da177e4 LT |
703 | |
704 | if (related) | |
705 | return verdict; | |
51ef348b | 706 | ip_vs_fill_iphdr(AF_INET, skb_network_header(skb), &iph); |
1da177e4 LT |
707 | } |
708 | ||
51ef348b | 709 | pp = ip_vs_proto_get(iph.protocol); |
1da177e4 LT |
710 | if (unlikely(!pp)) |
711 | return NF_ACCEPT; | |
712 | ||
713 | /* reassemble IP fragments */ | |
51ef348b | 714 | if (unlikely(ip_hdr(skb)->frag_off & htons(IP_MF|IP_OFFSET) && |
1da177e4 | 715 | !pp->dont_defrag)) { |
776c729e | 716 | if (ip_vs_gather_frags(skb, IP_DEFRAG_VS_OUT)) |
1da177e4 | 717 | return NF_STOLEN; |
1da177e4 | 718 | |
51ef348b JV |
719 | ip_vs_fill_iphdr(AF_INET, skb_network_header(skb), &iph); |
720 | } | |
1da177e4 LT |
721 | |
722 | /* | |
723 | * Check if the packet belongs to an existing entry | |
724 | */ | |
51ef348b | 725 | cp = pp->conn_out_get(AF_INET, skb, pp, &iph, iph.len, 0); |
1da177e4 LT |
726 | |
727 | if (unlikely(!cp)) { | |
728 | if (sysctl_ip_vs_nat_icmp_send && | |
729 | (pp->protocol == IPPROTO_TCP || | |
730 | pp->protocol == IPPROTO_UDP)) { | |
014d730d | 731 | __be16 _ports[2], *pptr; |
1da177e4 | 732 | |
51ef348b | 733 | pptr = skb_header_pointer(skb, iph.len, |
1da177e4 LT |
734 | sizeof(_ports), _ports); |
735 | if (pptr == NULL) | |
736 | return NF_ACCEPT; /* Not for me */ | |
51ef348b JV |
737 | if (ip_vs_lookup_real_service(iph.protocol, |
738 | iph.saddr.ip, pptr[0])) { | |
1da177e4 LT |
739 | /* |
740 | * Notify the real server: there is no | |
741 | * existing entry if it is not RST | |
742 | * packet or not TCP packet. | |
743 | */ | |
51ef348b | 744 | if (iph.protocol != IPPROTO_TCP |
1da177e4 LT |
745 | || !is_tcp_reset(skb)) { |
746 | icmp_send(skb,ICMP_DEST_UNREACH, | |
747 | ICMP_PORT_UNREACH, 0); | |
748 | return NF_DROP; | |
749 | } | |
750 | } | |
751 | } | |
752 | IP_VS_DBG_PKT(12, pp, skb, 0, | |
753 | "packet continues traversal as normal"); | |
754 | return NF_ACCEPT; | |
755 | } | |
756 | ||
757 | IP_VS_DBG_PKT(11, pp, skb, 0, "Outgoing packet"); | |
758 | ||
51ef348b | 759 | if (!skb_make_writable(skb, iph.len)) |
1da177e4 LT |
760 | goto drop; |
761 | ||
762 | /* mangle the packet */ | |
3db05fea | 763 | if (pp->snat_handler && !pp->snat_handler(skb, pp, cp)) |
1da177e4 | 764 | goto drop; |
e7ade46a | 765 | ip_hdr(skb)->saddr = cp->vaddr.ip; |
eddc9ec5 | 766 | ip_send_check(ip_hdr(skb)); |
1da177e4 | 767 | |
e905a9ed YH |
768 | /* For policy routing, packets originating from this |
769 | * machine itself may be routed differently to packets | |
770 | * passing through. We want this packet to be routed as | |
771 | * if it came from this machine itself. So re-compute | |
772 | * the routing information. | |
773 | */ | |
3db05fea | 774 | if (ip_route_me_harder(skb, RTN_LOCAL) != 0) |
e905a9ed | 775 | goto drop; |
901eaf6c | 776 | |
1da177e4 LT |
777 | IP_VS_DBG_PKT(10, pp, skb, 0, "After SNAT"); |
778 | ||
779 | ip_vs_out_stats(cp, skb); | |
780 | ip_vs_set_state(cp, IP_VS_DIR_OUTPUT, skb, pp); | |
781 | ip_vs_conn_put(cp); | |
782 | ||
6869c4d8 | 783 | skb->ipvs_property = 1; |
1da177e4 LT |
784 | |
785 | LeaveFunction(11); | |
786 | return NF_ACCEPT; | |
787 | ||
788 | drop: | |
789 | ip_vs_conn_put(cp); | |
3db05fea | 790 | kfree_skb(skb); |
1da177e4 LT |
791 | return NF_STOLEN; |
792 | } | |
793 | ||
794 | ||
795 | /* | |
796 | * Handle ICMP messages in the outside-to-inside direction (incoming). | |
797 | * Find any that might be relevant, check against existing connections, | |
798 | * forward to the right destination host if relevant. | |
799 | * Currently handles error types - unreachable, quench, ttl exceeded. | |
800 | */ | |
e905a9ed | 801 | static int |
3db05fea | 802 | ip_vs_in_icmp(struct sk_buff *skb, int *related, unsigned int hooknum) |
1da177e4 | 803 | { |
1da177e4 LT |
804 | struct iphdr *iph; |
805 | struct icmphdr _icmph, *ic; | |
806 | struct iphdr _ciph, *cih; /* The ip header contained within the ICMP */ | |
51ef348b | 807 | struct ip_vs_iphdr ciph; |
1da177e4 LT |
808 | struct ip_vs_conn *cp; |
809 | struct ip_vs_protocol *pp; | |
810 | unsigned int offset, ihl, verdict; | |
811 | ||
812 | *related = 1; | |
813 | ||
814 | /* reassemble IP fragments */ | |
eddc9ec5 | 815 | if (ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)) { |
6e23ae2a | 816 | if (ip_vs_gather_frags(skb, hooknum == NF_INET_LOCAL_IN ? |
776c729e | 817 | IP_DEFRAG_VS_IN : IP_DEFRAG_VS_FWD)) |
1da177e4 | 818 | return NF_STOLEN; |
1da177e4 LT |
819 | } |
820 | ||
eddc9ec5 | 821 | iph = ip_hdr(skb); |
1da177e4 LT |
822 | offset = ihl = iph->ihl * 4; |
823 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
824 | if (ic == NULL) | |
825 | return NF_DROP; | |
826 | ||
827 | IP_VS_DBG(12, "Incoming ICMP (%d,%d) %u.%u.%u.%u->%u.%u.%u.%u\n", | |
828 | ic->type, ntohs(icmp_id(ic)), | |
829 | NIPQUAD(iph->saddr), NIPQUAD(iph->daddr)); | |
830 | ||
831 | /* | |
832 | * Work through seeing if this is for us. | |
833 | * These checks are supposed to be in an order that means easy | |
834 | * things are checked first to speed up processing.... however | |
835 | * this means that some packets will manage to get a long way | |
836 | * down this stack and then be rejected, but that's life. | |
837 | */ | |
838 | if ((ic->type != ICMP_DEST_UNREACH) && | |
839 | (ic->type != ICMP_SOURCE_QUENCH) && | |
840 | (ic->type != ICMP_TIME_EXCEEDED)) { | |
841 | *related = 0; | |
842 | return NF_ACCEPT; | |
843 | } | |
844 | ||
845 | /* Now find the contained IP header */ | |
846 | offset += sizeof(_icmph); | |
847 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
848 | if (cih == NULL) | |
849 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
850 | ||
851 | pp = ip_vs_proto_get(cih->protocol); | |
852 | if (!pp) | |
853 | return NF_ACCEPT; | |
854 | ||
855 | /* Is the embedded protocol header present? */ | |
4412ec49 | 856 | if (unlikely(cih->frag_off & htons(IP_OFFSET) && |
1da177e4 LT |
857 | pp->dont_defrag)) |
858 | return NF_ACCEPT; | |
859 | ||
860 | IP_VS_DBG_PKT(11, pp, skb, offset, "Checking incoming ICMP for"); | |
861 | ||
862 | offset += cih->ihl * 4; | |
863 | ||
51ef348b | 864 | ip_vs_fill_iphdr(AF_INET, cih, &ciph); |
1da177e4 | 865 | /* The embedded headers contain source and dest in reverse order */ |
51ef348b | 866 | cp = pp->conn_in_get(AF_INET, skb, pp, &ciph, offset, 1); |
1da177e4 LT |
867 | if (!cp) |
868 | return NF_ACCEPT; | |
869 | ||
870 | verdict = NF_DROP; | |
871 | ||
872 | /* Ensure the checksum is correct */ | |
60476372 | 873 | if (!skb_csum_unnecessary(skb) && ip_vs_checksum_complete(skb, ihl)) { |
1da177e4 LT |
874 | /* Failed checksum! */ |
875 | IP_VS_DBG(1, "Incoming ICMP: failed checksum from %d.%d.%d.%d!\n", | |
876 | NIPQUAD(iph->saddr)); | |
877 | goto out; | |
878 | } | |
879 | ||
880 | /* do the statistics and put it back */ | |
881 | ip_vs_in_stats(cp, skb); | |
882 | if (IPPROTO_TCP == cih->protocol || IPPROTO_UDP == cih->protocol) | |
883 | offset += 2 * sizeof(__u16); | |
884 | verdict = ip_vs_icmp_xmit(skb, cp, pp, offset); | |
885 | /* do not touch skb anymore */ | |
886 | ||
887 | out: | |
888 | __ip_vs_conn_put(cp); | |
889 | ||
890 | return verdict; | |
891 | } | |
892 | ||
893 | /* | |
894 | * Check if it's for virtual services, look it up, | |
895 | * and send it on its way... | |
896 | */ | |
897 | static unsigned int | |
3db05fea | 898 | ip_vs_in(unsigned int hooknum, struct sk_buff *skb, |
1da177e4 LT |
899 | const struct net_device *in, const struct net_device *out, |
900 | int (*okfn)(struct sk_buff *)) | |
901 | { | |
51ef348b | 902 | struct ip_vs_iphdr iph; |
1da177e4 LT |
903 | struct ip_vs_protocol *pp; |
904 | struct ip_vs_conn *cp; | |
905 | int ret, restart; | |
51ef348b JV |
906 | |
907 | ip_vs_fill_iphdr(AF_INET, skb_network_header(skb), &iph); | |
1da177e4 LT |
908 | |
909 | /* | |
910 | * Big tappo: only PACKET_HOST (neither loopback nor mcasts) | |
911 | * ... don't know why 1st test DOES NOT include 2nd (?) | |
912 | */ | |
913 | if (unlikely(skb->pkt_type != PACKET_HOST | |
0cc217e1 | 914 | || skb->dev->flags & IFF_LOOPBACK || skb->sk)) { |
51ef348b JV |
915 | IP_VS_DBG_BUF(12, "packet type=%d proto=%d daddr=%s ignored\n", |
916 | skb->pkt_type, | |
917 | iph.protocol, | |
918 | IP_VS_DBG_ADDR(AF_INET, &iph.daddr)); | |
1da177e4 LT |
919 | return NF_ACCEPT; |
920 | } | |
921 | ||
51ef348b | 922 | if (unlikely(iph.protocol == IPPROTO_ICMP)) { |
3db05fea | 923 | int related, verdict = ip_vs_in_icmp(skb, &related, hooknum); |
1da177e4 LT |
924 | |
925 | if (related) | |
926 | return verdict; | |
51ef348b | 927 | ip_vs_fill_iphdr(AF_INET, skb_network_header(skb), &iph); |
1da177e4 LT |
928 | } |
929 | ||
930 | /* Protocol supported? */ | |
51ef348b | 931 | pp = ip_vs_proto_get(iph.protocol); |
1da177e4 LT |
932 | if (unlikely(!pp)) |
933 | return NF_ACCEPT; | |
934 | ||
1da177e4 LT |
935 | /* |
936 | * Check if the packet belongs to an existing connection entry | |
937 | */ | |
51ef348b | 938 | cp = pp->conn_in_get(AF_INET, skb, pp, &iph, iph.len, 0); |
1da177e4 LT |
939 | |
940 | if (unlikely(!cp)) { | |
941 | int v; | |
942 | ||
51ef348b | 943 | if (!pp->conn_schedule(AF_INET, skb, pp, &v, &cp)) |
1da177e4 LT |
944 | return v; |
945 | } | |
946 | ||
947 | if (unlikely(!cp)) { | |
948 | /* sorry, all this trouble for a no-hit :) */ | |
949 | IP_VS_DBG_PKT(12, pp, skb, 0, | |
950 | "packet continues traversal as normal"); | |
951 | return NF_ACCEPT; | |
952 | } | |
953 | ||
954 | IP_VS_DBG_PKT(11, pp, skb, 0, "Incoming packet"); | |
955 | ||
956 | /* Check the server status */ | |
957 | if (cp->dest && !(cp->dest->flags & IP_VS_DEST_F_AVAILABLE)) { | |
958 | /* the destination server is not available */ | |
959 | ||
960 | if (sysctl_ip_vs_expire_nodest_conn) { | |
961 | /* try to expire the connection immediately */ | |
962 | ip_vs_conn_expire_now(cp); | |
1da177e4 | 963 | } |
dc8103f2 JA |
964 | /* don't restart its timer, and silently |
965 | drop the packet. */ | |
966 | __ip_vs_conn_put(cp); | |
1da177e4 LT |
967 | return NF_DROP; |
968 | } | |
969 | ||
970 | ip_vs_in_stats(cp, skb); | |
971 | restart = ip_vs_set_state(cp, IP_VS_DIR_INPUT, skb, pp); | |
972 | if (cp->packet_xmit) | |
973 | ret = cp->packet_xmit(skb, cp, pp); | |
974 | /* do not touch skb anymore */ | |
975 | else { | |
976 | IP_VS_DBG_RL("warning: packet_xmit is null"); | |
977 | ret = NF_ACCEPT; | |
978 | } | |
979 | ||
efac5276 RB |
980 | /* Increase its packet counter and check if it is needed |
981 | * to be synchronized | |
982 | * | |
983 | * Sync connection if it is about to close to | |
984 | * encorage the standby servers to update the connections timeout | |
985 | */ | |
1da177e4 LT |
986 | atomic_inc(&cp->in_pkts); |
987 | if ((ip_vs_sync_state & IP_VS_STATE_MASTER) && | |
efac5276 RB |
988 | (((cp->protocol != IPPROTO_TCP || |
989 | cp->state == IP_VS_TCP_S_ESTABLISHED) && | |
990 | (atomic_read(&cp->in_pkts) % sysctl_ip_vs_sync_threshold[1] | |
991 | == sysctl_ip_vs_sync_threshold[0])) || | |
992 | ((cp->protocol == IPPROTO_TCP) && (cp->old_state != cp->state) && | |
993 | ((cp->state == IP_VS_TCP_S_FIN_WAIT) || | |
9d3a0de7 RB |
994 | (cp->state == IP_VS_TCP_S_CLOSE_WAIT) || |
995 | (cp->state == IP_VS_TCP_S_TIME_WAIT))))) | |
1da177e4 | 996 | ip_vs_sync_conn(cp); |
efac5276 | 997 | cp->old_state = cp->state; |
1da177e4 LT |
998 | |
999 | ip_vs_conn_put(cp); | |
1000 | return ret; | |
1001 | } | |
1002 | ||
1003 | ||
1004 | /* | |
6e23ae2a | 1005 | * It is hooked at the NF_INET_FORWARD chain, in order to catch ICMP |
1da177e4 LT |
1006 | * related packets destined for 0.0.0.0/0. |
1007 | * When fwmark-based virtual service is used, such as transparent | |
1008 | * cache cluster, TCP packets can be marked and routed to ip_vs_in, | |
1009 | * but ICMP destined for 0.0.0.0/0 cannot not be easily marked and | |
6e23ae2a | 1010 | * sent to ip_vs_in_icmp. So, catch them at the NF_INET_FORWARD chain |
1da177e4 LT |
1011 | * and send them to ip_vs_in_icmp. |
1012 | */ | |
1013 | static unsigned int | |
3db05fea | 1014 | ip_vs_forward_icmp(unsigned int hooknum, struct sk_buff *skb, |
1da177e4 LT |
1015 | const struct net_device *in, const struct net_device *out, |
1016 | int (*okfn)(struct sk_buff *)) | |
1017 | { | |
1018 | int r; | |
1019 | ||
3db05fea | 1020 | if (ip_hdr(skb)->protocol != IPPROTO_ICMP) |
1da177e4 LT |
1021 | return NF_ACCEPT; |
1022 | ||
3db05fea | 1023 | return ip_vs_in_icmp(skb, &r, hooknum); |
1da177e4 LT |
1024 | } |
1025 | ||
1026 | ||
1999414a | 1027 | static struct nf_hook_ops ip_vs_ops[] __read_mostly = { |
41c5b317 PM |
1028 | /* After packet filtering, forward packet through VS/DR, VS/TUN, |
1029 | * or VS/NAT(change destination), so that filtering rules can be | |
1030 | * applied to IPVS. */ | |
1031 | { | |
1032 | .hook = ip_vs_in, | |
1033 | .owner = THIS_MODULE, | |
1034 | .pf = PF_INET, | |
1035 | .hooknum = NF_INET_LOCAL_IN, | |
1036 | .priority = 100, | |
1037 | }, | |
1038 | /* After packet filtering, change source only for VS/NAT */ | |
1039 | { | |
1040 | .hook = ip_vs_out, | |
1041 | .owner = THIS_MODULE, | |
1042 | .pf = PF_INET, | |
1043 | .hooknum = NF_INET_FORWARD, | |
1044 | .priority = 100, | |
1045 | }, | |
1046 | /* After packet filtering (but before ip_vs_out_icmp), catch icmp | |
1047 | * destined for 0.0.0.0/0, which is for incoming IPVS connections */ | |
1048 | { | |
1049 | .hook = ip_vs_forward_icmp, | |
1050 | .owner = THIS_MODULE, | |
1051 | .pf = PF_INET, | |
1052 | .hooknum = NF_INET_FORWARD, | |
1053 | .priority = 99, | |
1054 | }, | |
1055 | /* Before the netfilter connection tracking, exit from POST_ROUTING */ | |
1056 | { | |
1057 | .hook = ip_vs_post_routing, | |
1058 | .owner = THIS_MODULE, | |
1059 | .pf = PF_INET, | |
1060 | .hooknum = NF_INET_POST_ROUTING, | |
1061 | .priority = NF_IP_PRI_NAT_SRC-1, | |
1062 | }, | |
1da177e4 LT |
1063 | }; |
1064 | ||
1065 | ||
1066 | /* | |
1067 | * Initialize IP Virtual Server | |
1068 | */ | |
1069 | static int __init ip_vs_init(void) | |
1070 | { | |
1071 | int ret; | |
1072 | ||
a919cf4b SW |
1073 | ip_vs_estimator_init(); |
1074 | ||
1da177e4 LT |
1075 | ret = ip_vs_control_init(); |
1076 | if (ret < 0) { | |
1077 | IP_VS_ERR("can't setup control.\n"); | |
a919cf4b | 1078 | goto cleanup_estimator; |
1da177e4 LT |
1079 | } |
1080 | ||
1081 | ip_vs_protocol_init(); | |
1082 | ||
1083 | ret = ip_vs_app_init(); | |
1084 | if (ret < 0) { | |
1085 | IP_VS_ERR("can't setup application helper.\n"); | |
1086 | goto cleanup_protocol; | |
1087 | } | |
1088 | ||
1089 | ret = ip_vs_conn_init(); | |
1090 | if (ret < 0) { | |
1091 | IP_VS_ERR("can't setup connection table.\n"); | |
1092 | goto cleanup_app; | |
1093 | } | |
1094 | ||
41c5b317 | 1095 | ret = nf_register_hooks(ip_vs_ops, ARRAY_SIZE(ip_vs_ops)); |
1da177e4 | 1096 | if (ret < 0) { |
41c5b317 | 1097 | IP_VS_ERR("can't register hooks.\n"); |
1da177e4 LT |
1098 | goto cleanup_conn; |
1099 | } | |
1100 | ||
1da177e4 LT |
1101 | IP_VS_INFO("ipvs loaded.\n"); |
1102 | return ret; | |
1103 | ||
1da177e4 LT |
1104 | cleanup_conn: |
1105 | ip_vs_conn_cleanup(); | |
1106 | cleanup_app: | |
1107 | ip_vs_app_cleanup(); | |
1108 | cleanup_protocol: | |
1109 | ip_vs_protocol_cleanup(); | |
1110 | ip_vs_control_cleanup(); | |
a919cf4b SW |
1111 | cleanup_estimator: |
1112 | ip_vs_estimator_cleanup(); | |
1da177e4 LT |
1113 | return ret; |
1114 | } | |
1115 | ||
1116 | static void __exit ip_vs_cleanup(void) | |
1117 | { | |
41c5b317 | 1118 | nf_unregister_hooks(ip_vs_ops, ARRAY_SIZE(ip_vs_ops)); |
1da177e4 LT |
1119 | ip_vs_conn_cleanup(); |
1120 | ip_vs_app_cleanup(); | |
1121 | ip_vs_protocol_cleanup(); | |
1122 | ip_vs_control_cleanup(); | |
a919cf4b | 1123 | ip_vs_estimator_cleanup(); |
1da177e4 LT |
1124 | IP_VS_INFO("ipvs unloaded.\n"); |
1125 | } | |
1126 | ||
1127 | module_init(ip_vs_init); | |
1128 | module_exit(ip_vs_cleanup); | |
1129 | MODULE_LICENSE("GPL"); |