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 | ||
f71499aa | 179 | static inline void |
85999283 SH |
180 | ip_vs_conn_fill_param_persist(const struct ip_vs_service *svc, |
181 | struct sk_buff *skb, int protocol, | |
182 | const union nf_inet_addr *caddr, __be16 cport, | |
183 | const union nf_inet_addr *vaddr, __be16 vport, | |
184 | struct ip_vs_conn_param *p) | |
185 | { | |
186 | ip_vs_conn_fill_param(svc->af, protocol, caddr, cport, vaddr, vport, p); | |
187 | p->pe = svc->pe; | |
188 | if (p->pe && p->pe->fill_param) | |
f71499aa | 189 | p->pe->fill_param(p, skb); |
85999283 | 190 | } |
1da177e4 | 191 | |
1da177e4 LT |
192 | /* |
193 | * IPVS persistent scheduling function | |
194 | * It creates a connection entry according to its template if exists, | |
195 | * or selects a server and creates a connection entry plus a template. | |
196 | * Locking: we are svc user (svc->refcnt), so we hold all dests too | |
197 | * Protocols supported: TCP, UDP | |
198 | */ | |
199 | static struct ip_vs_conn * | |
200 | ip_vs_sched_persist(struct ip_vs_service *svc, | |
85999283 | 201 | struct sk_buff *skb, |
014d730d | 202 | __be16 ports[2]) |
1da177e4 LT |
203 | { |
204 | struct ip_vs_conn *cp = NULL; | |
28364a59 | 205 | struct ip_vs_iphdr iph; |
1da177e4 LT |
206 | struct ip_vs_dest *dest; |
207 | struct ip_vs_conn *ct; | |
5b57a98c | 208 | __be16 dport = 0; /* destination port to forward */ |
3575792e | 209 | unsigned int flags; |
f11017ec | 210 | struct ip_vs_conn_param param; |
28364a59 JV |
211 | union nf_inet_addr snet; /* source network of the client, |
212 | after masking */ | |
cd17f9ed JV |
213 | |
214 | ip_vs_fill_iphdr(svc->af, skb_network_header(skb), &iph); | |
1da177e4 LT |
215 | |
216 | /* Mask saddr with the netmask to adjust template granularity */ | |
cd17f9ed JV |
217 | #ifdef CONFIG_IP_VS_IPV6 |
218 | if (svc->af == AF_INET6) | |
219 | ipv6_addr_prefix(&snet.in6, &iph.saddr.in6, svc->netmask); | |
220 | else | |
221 | #endif | |
222 | snet.ip = iph.saddr.ip & svc->netmask; | |
1da177e4 | 223 | |
cd17f9ed JV |
224 | IP_VS_DBG_BUF(6, "p-schedule: src %s:%u dest %s:%u " |
225 | "mnet %s\n", | |
226 | IP_VS_DBG_ADDR(svc->af, &iph.saddr), ntohs(ports[0]), | |
227 | IP_VS_DBG_ADDR(svc->af, &iph.daddr), ntohs(ports[1]), | |
228 | IP_VS_DBG_ADDR(svc->af, &snet)); | |
1da177e4 LT |
229 | |
230 | /* | |
231 | * As far as we know, FTP is a very complicated network protocol, and | |
232 | * it uses control connection and data connections. For active FTP, | |
233 | * FTP server initialize data connection to the client, its source port | |
234 | * is often 20. For passive FTP, FTP server tells the clients the port | |
235 | * that it passively listens to, and the client issues the data | |
236 | * connection. In the tunneling or direct routing mode, the load | |
237 | * balancer is on the client-to-server half of connection, the port | |
238 | * number is unknown to the load balancer. So, a conn template like | |
239 | * <caddr, 0, vaddr, 0, daddr, 0> is created for persistent FTP | |
240 | * service, and a template like <caddr, 0, vaddr, vport, daddr, dport> | |
241 | * is created for other persistent services. | |
242 | */ | |
5b57a98c | 243 | { |
f11017ec SH |
244 | int protocol = iph.protocol; |
245 | const union nf_inet_addr *vaddr = &iph.daddr; | |
246 | const union nf_inet_addr fwmark = { .ip = htonl(svc->fwmark) }; | |
247 | __be16 vport = 0; | |
248 | ||
5b57a98c SH |
249 | if (ports[1] == svc->port) { |
250 | /* non-FTP template: | |
251 | * <protocol, caddr, 0, vaddr, vport, daddr, dport> | |
252 | * FTP template: | |
253 | * <protocol, caddr, 0, vaddr, 0, daddr, 0> | |
1da177e4 LT |
254 | */ |
255 | if (svc->port != FTPPORT) | |
5b57a98c | 256 | vport = ports[1]; |
1da177e4 | 257 | } else { |
5b57a98c SH |
258 | /* Note: persistent fwmark-based services and |
259 | * persistent port zero service are handled here. | |
260 | * fwmark template: | |
261 | * <IPPROTO_IP,caddr,0,fwmark,0,daddr,0> | |
262 | * port zero template: | |
263 | * <protocol,caddr,0,vaddr,0,daddr,0> | |
1da177e4 | 264 | */ |
28364a59 | 265 | if (svc->fwmark) { |
5b57a98c SH |
266 | protocol = IPPROTO_IP; |
267 | vaddr = &fwmark; | |
268 | } | |
1da177e4 | 269 | } |
f71499aa SH |
270 | ip_vs_conn_fill_param_persist(svc, skb, protocol, &snet, 0, |
271 | vaddr, vport, ¶m); | |
1da177e4 LT |
272 | } |
273 | ||
5b57a98c | 274 | /* Check if a template already exists */ |
f11017ec | 275 | ct = ip_vs_ct_in_get(¶m); |
5b57a98c SH |
276 | if (!ct || !ip_vs_check_template(ct)) { |
277 | /* No template found or the dest of the connection | |
278 | * template is not available. | |
279 | */ | |
280 | dest = svc->scheduler->schedule(svc, skb); | |
281 | if (!dest) { | |
282 | IP_VS_DBG(1, "p-schedule: no dest found.\n"); | |
85999283 | 283 | kfree(param.pe_data); |
5b57a98c SH |
284 | return NULL; |
285 | } | |
286 | ||
287 | if (ports[1] == svc->port && svc->port != FTPPORT) | |
288 | dport = dest->port; | |
289 | ||
85999283 SH |
290 | /* Create a template |
291 | * This adds param.pe_data to the template, | |
292 | * and thus param.pe_data will be destroyed | |
293 | * when the template expires */ | |
f11017ec | 294 | ct = ip_vs_conn_new(¶m, &dest->addr, dport, |
5b57a98c | 295 | IP_VS_CONN_F_TEMPLATE, dest); |
85999283 SH |
296 | if (ct == NULL) { |
297 | kfree(param.pe_data); | |
5b57a98c | 298 | return NULL; |
85999283 | 299 | } |
5b57a98c SH |
300 | |
301 | ct->timeout = svc->timeout; | |
85999283 | 302 | } else { |
5b57a98c SH |
303 | /* set destination with the found template */ |
304 | dest = ct->dest; | |
85999283 SH |
305 | kfree(param.pe_data); |
306 | } | |
5b57a98c SH |
307 | |
308 | dport = ports[1]; | |
309 | if (dport == svc->port && dest->port) | |
310 | dport = dest->port; | |
311 | ||
26ec037f NC |
312 | flags = (svc->flags & IP_VS_SVC_F_ONEPACKET |
313 | && iph.protocol == IPPROTO_UDP)? | |
314 | IP_VS_CONN_F_ONE_PACKET : 0; | |
315 | ||
1da177e4 LT |
316 | /* |
317 | * Create a new connection according to the template | |
318 | */ | |
f11017ec SH |
319 | ip_vs_conn_fill_param(svc->af, iph.protocol, &iph.saddr, ports[0], |
320 | &iph.daddr, ports[1], ¶m); | |
321 | cp = ip_vs_conn_new(¶m, &dest->addr, dport, flags, dest); | |
1da177e4 LT |
322 | if (cp == NULL) { |
323 | ip_vs_conn_put(ct); | |
324 | return NULL; | |
325 | } | |
326 | ||
327 | /* | |
328 | * Add its control | |
329 | */ | |
330 | ip_vs_control_add(cp, ct); | |
331 | ip_vs_conn_put(ct); | |
332 | ||
333 | ip_vs_conn_stats(cp, svc); | |
334 | return cp; | |
335 | } | |
336 | ||
337 | ||
338 | /* | |
339 | * IPVS main scheduling function | |
340 | * It selects a server according to the virtual service, and | |
341 | * creates a connection entry. | |
342 | * Protocols supported: TCP, UDP | |
343 | */ | |
344 | struct ip_vs_conn * | |
190ecd27 JA |
345 | ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb, |
346 | struct ip_vs_protocol *pp, int *ignored) | |
1da177e4 LT |
347 | { |
348 | struct ip_vs_conn *cp = NULL; | |
28364a59 | 349 | struct ip_vs_iphdr iph; |
1da177e4 | 350 | struct ip_vs_dest *dest; |
3575792e JA |
351 | __be16 _ports[2], *pptr; |
352 | unsigned int flags; | |
1da177e4 | 353 | |
190ecd27 | 354 | *ignored = 1; |
28364a59 JV |
355 | ip_vs_fill_iphdr(svc->af, skb_network_header(skb), &iph); |
356 | pptr = skb_header_pointer(skb, iph.len, sizeof(_ports), _ports); | |
1da177e4 LT |
357 | if (pptr == NULL) |
358 | return NULL; | |
359 | ||
190ecd27 JA |
360 | /* |
361 | * FTPDATA needs this check when using local real server. | |
362 | * Never schedule Active FTPDATA connections from real server. | |
363 | * For LVS-NAT they must be already created. For other methods | |
364 | * with persistence the connection is created on SYN+ACK. | |
365 | */ | |
366 | if (pptr[0] == FTPDATA) { | |
367 | IP_VS_DBG_PKT(12, pp, skb, 0, "Not scheduling FTPDATA"); | |
368 | return NULL; | |
369 | } | |
370 | ||
371 | /* | |
372 | * Do not schedule replies from local real server. It is risky | |
373 | * for fwmark services but mostly for persistent services. | |
374 | */ | |
375 | if ((!skb->dev || skb->dev->flags & IFF_LOOPBACK) && | |
376 | (svc->flags & IP_VS_SVC_F_PERSISTENT || svc->fwmark) && | |
377 | (cp = pp->conn_in_get(svc->af, skb, pp, &iph, iph.len, 1))) { | |
378 | IP_VS_DBG_PKT(12, pp, skb, 0, | |
379 | "Not scheduling reply for existing connection"); | |
380 | __ip_vs_conn_put(cp); | |
381 | return NULL; | |
382 | } | |
383 | ||
1da177e4 LT |
384 | /* |
385 | * Persistent service | |
386 | */ | |
190ecd27 JA |
387 | if (svc->flags & IP_VS_SVC_F_PERSISTENT) { |
388 | *ignored = 0; | |
1da177e4 | 389 | return ip_vs_sched_persist(svc, skb, pptr); |
190ecd27 | 390 | } |
1da177e4 LT |
391 | |
392 | /* | |
393 | * Non-persistent service | |
394 | */ | |
395 | if (!svc->fwmark && pptr[1] != svc->port) { | |
396 | if (!svc->port) | |
1e3e238e HE |
397 | pr_err("Schedule: port zero only supported " |
398 | "in persistent services, " | |
399 | "check your ipvs configuration\n"); | |
1da177e4 LT |
400 | return NULL; |
401 | } | |
402 | ||
190ecd27 JA |
403 | *ignored = 0; |
404 | ||
1da177e4 LT |
405 | dest = svc->scheduler->schedule(svc, skb); |
406 | if (dest == NULL) { | |
407 | IP_VS_DBG(1, "Schedule: no dest found.\n"); | |
408 | return NULL; | |
409 | } | |
410 | ||
26ec037f NC |
411 | flags = (svc->flags & IP_VS_SVC_F_ONEPACKET |
412 | && iph.protocol == IPPROTO_UDP)? | |
413 | IP_VS_CONN_F_ONE_PACKET : 0; | |
414 | ||
1da177e4 LT |
415 | /* |
416 | * Create a connection entry. | |
417 | */ | |
f11017ec SH |
418 | { |
419 | struct ip_vs_conn_param p; | |
420 | ip_vs_conn_fill_param(svc->af, iph.protocol, &iph.saddr, | |
421 | pptr[0], &iph.daddr, pptr[1], &p); | |
422 | cp = ip_vs_conn_new(&p, &dest->addr, | |
423 | dest->port ? dest->port : pptr[1], | |
424 | flags, dest); | |
425 | if (!cp) | |
426 | return NULL; | |
427 | } | |
1da177e4 | 428 | |
cd17f9ed JV |
429 | IP_VS_DBG_BUF(6, "Schedule fwd:%c c:%s:%u v:%s:%u " |
430 | "d:%s:%u conn->flags:%X conn->refcnt:%d\n", | |
431 | ip_vs_fwd_tag(cp), | |
432 | IP_VS_DBG_ADDR(svc->af, &cp->caddr), ntohs(cp->cport), | |
433 | IP_VS_DBG_ADDR(svc->af, &cp->vaddr), ntohs(cp->vport), | |
434 | IP_VS_DBG_ADDR(svc->af, &cp->daddr), ntohs(cp->dport), | |
435 | cp->flags, atomic_read(&cp->refcnt)); | |
1da177e4 LT |
436 | |
437 | ip_vs_conn_stats(cp, svc); | |
438 | return cp; | |
439 | } | |
440 | ||
441 | ||
442 | /* | |
443 | * Pass or drop the packet. | |
444 | * Called by ip_vs_in, when the virtual service is available but | |
445 | * no destination is available for a new connection. | |
446 | */ | |
447 | int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb, | |
448 | struct ip_vs_protocol *pp) | |
449 | { | |
014d730d | 450 | __be16 _ports[2], *pptr; |
28364a59 | 451 | struct ip_vs_iphdr iph; |
2a3b791e JV |
452 | int unicast; |
453 | ip_vs_fill_iphdr(svc->af, skb_network_header(skb), &iph); | |
1da177e4 | 454 | |
28364a59 | 455 | pptr = skb_header_pointer(skb, iph.len, sizeof(_ports), _ports); |
1da177e4 LT |
456 | if (pptr == NULL) { |
457 | ip_vs_service_put(svc); | |
458 | return NF_DROP; | |
459 | } | |
460 | ||
2a3b791e JV |
461 | #ifdef CONFIG_IP_VS_IPV6 |
462 | if (svc->af == AF_INET6) | |
463 | unicast = ipv6_addr_type(&iph.daddr.in6) & IPV6_ADDR_UNICAST; | |
464 | else | |
465 | #endif | |
466 | unicast = (inet_addr_type(&init_net, iph.daddr.ip) == RTN_UNICAST); | |
467 | ||
1da177e4 | 468 | /* if it is fwmark-based service, the cache_bypass sysctl is up |
2a3b791e | 469 | and the destination is a non-local unicast, then create |
1da177e4 | 470 | a cache_bypass connection entry */ |
2a3b791e | 471 | if (sysctl_ip_vs_cache_bypass && svc->fwmark && unicast) { |
1da177e4 LT |
472 | int ret, cs; |
473 | struct ip_vs_conn *cp; | |
3575792e JA |
474 | unsigned int flags = (svc->flags & IP_VS_SVC_F_ONEPACKET && |
475 | iph.protocol == IPPROTO_UDP)? | |
476 | IP_VS_CONN_F_ONE_PACKET : 0; | |
dff630dd | 477 | union nf_inet_addr daddr = { .all = { 0, 0, 0, 0 } }; |
1da177e4 LT |
478 | |
479 | ip_vs_service_put(svc); | |
480 | ||
481 | /* create a new connection entry */ | |
1e3e238e | 482 | IP_VS_DBG(6, "%s(): create a cache_bypass entry\n", __func__); |
f11017ec SH |
483 | { |
484 | struct ip_vs_conn_param p; | |
485 | ip_vs_conn_fill_param(svc->af, iph.protocol, | |
486 | &iph.saddr, pptr[0], | |
487 | &iph.daddr, pptr[1], &p); | |
488 | cp = ip_vs_conn_new(&p, &daddr, 0, | |
489 | IP_VS_CONN_F_BYPASS | flags, | |
490 | NULL); | |
491 | if (!cp) | |
492 | return NF_DROP; | |
493 | } | |
1da177e4 LT |
494 | |
495 | /* statistics */ | |
496 | ip_vs_in_stats(cp, skb); | |
497 | ||
498 | /* set state */ | |
499 | cs = ip_vs_set_state(cp, IP_VS_DIR_INPUT, skb, pp); | |
500 | ||
501 | /* transmit the first SYN packet */ | |
502 | ret = cp->packet_xmit(skb, cp, pp); | |
503 | /* do not touch skb anymore */ | |
504 | ||
505 | atomic_inc(&cp->in_pkts); | |
506 | ip_vs_conn_put(cp); | |
507 | return ret; | |
508 | } | |
509 | ||
510 | /* | |
511 | * When the virtual ftp service is presented, packets destined | |
512 | * for other services on the VIP may get here (except services | |
513 | * listed in the ipvs table), pass the packets, because it is | |
514 | * not ipvs job to decide to drop the packets. | |
515 | */ | |
516 | if ((svc->port == FTPPORT) && (pptr[1] != FTPPORT)) { | |
517 | ip_vs_service_put(svc); | |
518 | return NF_ACCEPT; | |
519 | } | |
520 | ||
521 | ip_vs_service_put(svc); | |
522 | ||
523 | /* | |
524 | * Notify the client that the destination is unreachable, and | |
525 | * release the socket buffer. | |
526 | * Since it is in IP layer, the TCP socket is not actually | |
527 | * created, the TCP RST packet cannot be sent, instead that | |
528 | * ICMP_PORT_UNREACH is sent here no matter it is TCP/UDP. --WZ | |
529 | */ | |
2a3b791e JV |
530 | #ifdef CONFIG_IP_VS_IPV6 |
531 | if (svc->af == AF_INET6) | |
3ffe533c | 532 | icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); |
2a3b791e JV |
533 | else |
534 | #endif | |
535 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); | |
536 | ||
1da177e4 LT |
537 | return NF_DROP; |
538 | } | |
539 | ||
b1550f22 | 540 | __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset) |
1da177e4 | 541 | { |
d3bc23e7 | 542 | return csum_fold(skb_checksum(skb, offset, skb->len - offset, 0)); |
1da177e4 LT |
543 | } |
544 | ||
776c729e | 545 | static inline int ip_vs_gather_frags(struct sk_buff *skb, u_int32_t user) |
1da177e4 | 546 | { |
776c729e HX |
547 | int err = ip_defrag(skb, user); |
548 | ||
549 | if (!err) | |
eddc9ec5 | 550 | ip_send_check(ip_hdr(skb)); |
776c729e HX |
551 | |
552 | return err; | |
1da177e4 LT |
553 | } |
554 | ||
2a3b791e JV |
555 | #ifdef CONFIG_IP_VS_IPV6 |
556 | static inline int ip_vs_gather_frags_v6(struct sk_buff *skb, u_int32_t user) | |
557 | { | |
558 | /* TODO IPv6: Find out what to do here for IPv6 */ | |
559 | return 0; | |
560 | } | |
561 | #endif | |
562 | ||
1da177e4 LT |
563 | /* |
564 | * Packet has been made sufficiently writable in caller | |
565 | * - inout: 1=in->out, 0=out->in | |
566 | */ | |
567 | void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp, | |
568 | struct ip_vs_conn *cp, int inout) | |
569 | { | |
eddc9ec5 | 570 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 | 571 | unsigned int icmp_offset = iph->ihl*4; |
d56f90a7 ACM |
572 | struct icmphdr *icmph = (struct icmphdr *)(skb_network_header(skb) + |
573 | icmp_offset); | |
1da177e4 LT |
574 | struct iphdr *ciph = (struct iphdr *)(icmph + 1); |
575 | ||
576 | if (inout) { | |
e7ade46a | 577 | iph->saddr = cp->vaddr.ip; |
1da177e4 | 578 | ip_send_check(iph); |
e7ade46a | 579 | ciph->daddr = cp->vaddr.ip; |
1da177e4 LT |
580 | ip_send_check(ciph); |
581 | } else { | |
e7ade46a | 582 | iph->daddr = cp->daddr.ip; |
1da177e4 | 583 | ip_send_check(iph); |
e7ade46a | 584 | ciph->saddr = cp->daddr.ip; |
1da177e4 LT |
585 | ip_send_check(ciph); |
586 | } | |
587 | ||
2906f66a VMR |
588 | /* the TCP/UDP/SCTP port */ |
589 | if (IPPROTO_TCP == ciph->protocol || IPPROTO_UDP == ciph->protocol || | |
590 | IPPROTO_SCTP == ciph->protocol) { | |
014d730d | 591 | __be16 *ports = (void *)ciph + ciph->ihl*4; |
1da177e4 LT |
592 | |
593 | if (inout) | |
594 | ports[1] = cp->vport; | |
595 | else | |
596 | ports[0] = cp->dport; | |
597 | } | |
598 | ||
599 | /* And finally the ICMP checksum */ | |
600 | icmph->checksum = 0; | |
601 | icmph->checksum = ip_vs_checksum_complete(skb, icmp_offset); | |
602 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
603 | ||
604 | if (inout) | |
605 | IP_VS_DBG_PKT(11, pp, skb, (void *)ciph - (void *)iph, | |
606 | "Forwarding altered outgoing ICMP"); | |
607 | else | |
608 | IP_VS_DBG_PKT(11, pp, skb, (void *)ciph - (void *)iph, | |
609 | "Forwarding altered incoming ICMP"); | |
610 | } | |
611 | ||
b3cdd2a7 JV |
612 | #ifdef CONFIG_IP_VS_IPV6 |
613 | void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp, | |
614 | struct ip_vs_conn *cp, int inout) | |
615 | { | |
616 | struct ipv6hdr *iph = ipv6_hdr(skb); | |
617 | unsigned int icmp_offset = sizeof(struct ipv6hdr); | |
618 | struct icmp6hdr *icmph = (struct icmp6hdr *)(skb_network_header(skb) + | |
619 | icmp_offset); | |
620 | struct ipv6hdr *ciph = (struct ipv6hdr *)(icmph + 1); | |
621 | ||
622 | if (inout) { | |
623 | iph->saddr = cp->vaddr.in6; | |
624 | ciph->daddr = cp->vaddr.in6; | |
625 | } else { | |
626 | iph->daddr = cp->daddr.in6; | |
627 | ciph->saddr = cp->daddr.in6; | |
628 | } | |
629 | ||
2906f66a VMR |
630 | /* the TCP/UDP/SCTP port */ |
631 | if (IPPROTO_TCP == ciph->nexthdr || IPPROTO_UDP == ciph->nexthdr || | |
632 | IPPROTO_SCTP == ciph->nexthdr) { | |
b3cdd2a7 JV |
633 | __be16 *ports = (void *)ciph + sizeof(struct ipv6hdr); |
634 | ||
635 | if (inout) | |
636 | ports[1] = cp->vport; | |
637 | else | |
638 | ports[0] = cp->dport; | |
639 | } | |
640 | ||
641 | /* And finally the ICMP checksum */ | |
8870f842 SH |
642 | icmph->icmp6_cksum = ~csum_ipv6_magic(&iph->saddr, &iph->daddr, |
643 | skb->len - icmp_offset, | |
644 | IPPROTO_ICMPV6, 0); | |
645 | skb->csum_start = skb_network_header(skb) - skb->head + icmp_offset; | |
646 | skb->csum_offset = offsetof(struct icmp6hdr, icmp6_cksum); | |
647 | skb->ip_summed = CHECKSUM_PARTIAL; | |
b3cdd2a7 JV |
648 | |
649 | if (inout) | |
650 | IP_VS_DBG_PKT(11, pp, skb, (void *)ciph - (void *)iph, | |
651 | "Forwarding altered outgoing ICMPv6"); | |
652 | else | |
653 | IP_VS_DBG_PKT(11, pp, skb, (void *)ciph - (void *)iph, | |
654 | "Forwarding altered incoming ICMPv6"); | |
655 | } | |
656 | #endif | |
657 | ||
4856c84c MT |
658 | /* Handle relevant response ICMP messages - forward to the right |
659 | * destination host. Used for NAT and local client. | |
660 | */ | |
f2428ed5 SH |
661 | static int handle_response_icmp(int af, struct sk_buff *skb, |
662 | union nf_inet_addr *snet, | |
663 | __u8 protocol, struct ip_vs_conn *cp, | |
4856c84c MT |
664 | struct ip_vs_protocol *pp, |
665 | unsigned int offset, unsigned int ihl) | |
666 | { | |
667 | unsigned int verdict = NF_DROP; | |
668 | ||
669 | if (IP_VS_FWD_METHOD(cp) != 0) { | |
1e3e238e HE |
670 | pr_err("shouldn't reach here, because the box is on the " |
671 | "half connection in the tun/dr module.\n"); | |
4856c84c MT |
672 | } |
673 | ||
674 | /* Ensure the checksum is correct */ | |
675 | if (!skb_csum_unnecessary(skb) && ip_vs_checksum_complete(skb, ihl)) { | |
676 | /* Failed checksum! */ | |
f2428ed5 SH |
677 | IP_VS_DBG_BUF(1, "Forward ICMP: failed checksum from %s!\n", |
678 | IP_VS_DBG_ADDR(af, snet)); | |
4856c84c MT |
679 | goto out; |
680 | } | |
681 | ||
2906f66a VMR |
682 | if (IPPROTO_TCP == protocol || IPPROTO_UDP == protocol || |
683 | IPPROTO_SCTP == protocol) | |
4856c84c MT |
684 | offset += 2 * sizeof(__u16); |
685 | if (!skb_make_writable(skb, offset)) | |
686 | goto out; | |
687 | ||
f2428ed5 SH |
688 | #ifdef CONFIG_IP_VS_IPV6 |
689 | if (af == AF_INET6) | |
690 | ip_vs_nat_icmp_v6(skb, pp, cp, 1); | |
691 | else | |
692 | #endif | |
693 | ip_vs_nat_icmp(skb, pp, cp, 1); | |
4856c84c MT |
694 | |
695 | /* do the statistics and put it back */ | |
696 | ip_vs_out_stats(cp, skb); | |
697 | ||
cf356d69 | 698 | skb->ipvs_property = 1; |
f4bc17cd | 699 | if (!(cp->flags & IP_VS_CONN_F_NFCT)) |
cf356d69 | 700 | ip_vs_notrack(skb); |
f4bc17cd JA |
701 | else |
702 | ip_vs_update_conntrack(skb, cp, 0); | |
4856c84c MT |
703 | verdict = NF_ACCEPT; |
704 | ||
705 | out: | |
706 | __ip_vs_conn_put(cp); | |
707 | ||
708 | return verdict; | |
709 | } | |
710 | ||
1da177e4 LT |
711 | /* |
712 | * Handle ICMP messages in the inside-to-outside direction (outgoing). | |
4856c84c | 713 | * Find any that might be relevant, check against existing connections. |
1da177e4 | 714 | * Currently handles error types - unreachable, quench, ttl exceeded. |
1da177e4 | 715 | */ |
3db05fea | 716 | static int ip_vs_out_icmp(struct sk_buff *skb, int *related) |
1da177e4 | 717 | { |
1da177e4 LT |
718 | struct iphdr *iph; |
719 | struct icmphdr _icmph, *ic; | |
720 | struct iphdr _ciph, *cih; /* The ip header contained within the ICMP */ | |
51ef348b | 721 | struct ip_vs_iphdr ciph; |
1da177e4 LT |
722 | struct ip_vs_conn *cp; |
723 | struct ip_vs_protocol *pp; | |
4856c84c | 724 | unsigned int offset, ihl; |
f2428ed5 | 725 | union nf_inet_addr snet; |
1da177e4 LT |
726 | |
727 | *related = 1; | |
728 | ||
729 | /* reassemble IP fragments */ | |
eddc9ec5 | 730 | if (ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)) { |
776c729e | 731 | if (ip_vs_gather_frags(skb, IP_DEFRAG_VS_OUT)) |
1da177e4 | 732 | return NF_STOLEN; |
1da177e4 LT |
733 | } |
734 | ||
eddc9ec5 | 735 | iph = ip_hdr(skb); |
1da177e4 LT |
736 | offset = ihl = iph->ihl * 4; |
737 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
738 | if (ic == NULL) | |
739 | return NF_DROP; | |
740 | ||
14d5e834 | 741 | IP_VS_DBG(12, "Outgoing ICMP (%d,%d) %pI4->%pI4\n", |
1da177e4 | 742 | ic->type, ntohs(icmp_id(ic)), |
14d5e834 | 743 | &iph->saddr, &iph->daddr); |
1da177e4 LT |
744 | |
745 | /* | |
746 | * Work through seeing if this is for us. | |
747 | * These checks are supposed to be in an order that means easy | |
748 | * things are checked first to speed up processing.... however | |
749 | * this means that some packets will manage to get a long way | |
750 | * down this stack and then be rejected, but that's life. | |
751 | */ | |
752 | if ((ic->type != ICMP_DEST_UNREACH) && | |
753 | (ic->type != ICMP_SOURCE_QUENCH) && | |
754 | (ic->type != ICMP_TIME_EXCEEDED)) { | |
755 | *related = 0; | |
756 | return NF_ACCEPT; | |
757 | } | |
758 | ||
759 | /* Now find the contained IP header */ | |
760 | offset += sizeof(_icmph); | |
761 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
762 | if (cih == NULL) | |
763 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
764 | ||
765 | pp = ip_vs_proto_get(cih->protocol); | |
766 | if (!pp) | |
767 | return NF_ACCEPT; | |
768 | ||
769 | /* Is the embedded protocol header present? */ | |
4412ec49 | 770 | if (unlikely(cih->frag_off & htons(IP_OFFSET) && |
1da177e4 LT |
771 | pp->dont_defrag)) |
772 | return NF_ACCEPT; | |
773 | ||
774 | IP_VS_DBG_PKT(11, pp, skb, offset, "Checking outgoing ICMP for"); | |
775 | ||
776 | offset += cih->ihl * 4; | |
777 | ||
51ef348b | 778 | ip_vs_fill_iphdr(AF_INET, cih, &ciph); |
1da177e4 | 779 | /* The embedded headers contain source and dest in reverse order */ |
51ef348b | 780 | cp = pp->conn_out_get(AF_INET, skb, pp, &ciph, offset, 1); |
1da177e4 LT |
781 | if (!cp) |
782 | return NF_ACCEPT; | |
783 | ||
f2428ed5 SH |
784 | snet.ip = iph->saddr; |
785 | return handle_response_icmp(AF_INET, skb, &snet, cih->protocol, cp, | |
786 | pp, offset, ihl); | |
1da177e4 LT |
787 | } |
788 | ||
2a3b791e JV |
789 | #ifdef CONFIG_IP_VS_IPV6 |
790 | static int ip_vs_out_icmp_v6(struct sk_buff *skb, int *related) | |
791 | { | |
792 | struct ipv6hdr *iph; | |
793 | struct icmp6hdr _icmph, *ic; | |
794 | struct ipv6hdr _ciph, *cih; /* The ip header contained | |
795 | within the ICMP */ | |
796 | struct ip_vs_iphdr ciph; | |
797 | struct ip_vs_conn *cp; | |
798 | struct ip_vs_protocol *pp; | |
f2428ed5 SH |
799 | unsigned int offset; |
800 | union nf_inet_addr snet; | |
2a3b791e JV |
801 | |
802 | *related = 1; | |
803 | ||
804 | /* reassemble IP fragments */ | |
805 | if (ipv6_hdr(skb)->nexthdr == IPPROTO_FRAGMENT) { | |
806 | if (ip_vs_gather_frags_v6(skb, IP_DEFRAG_VS_OUT)) | |
807 | return NF_STOLEN; | |
808 | } | |
809 | ||
810 | iph = ipv6_hdr(skb); | |
811 | offset = sizeof(struct ipv6hdr); | |
812 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
813 | if (ic == NULL) | |
814 | return NF_DROP; | |
815 | ||
5b095d98 | 816 | IP_VS_DBG(12, "Outgoing ICMPv6 (%d,%d) %pI6->%pI6\n", |
2a3b791e | 817 | ic->icmp6_type, ntohs(icmpv6_id(ic)), |
38ff4fa4 | 818 | &iph->saddr, &iph->daddr); |
2a3b791e JV |
819 | |
820 | /* | |
821 | * Work through seeing if this is for us. | |
822 | * These checks are supposed to be in an order that means easy | |
823 | * things are checked first to speed up processing.... however | |
824 | * this means that some packets will manage to get a long way | |
825 | * down this stack and then be rejected, but that's life. | |
826 | */ | |
827 | if ((ic->icmp6_type != ICMPV6_DEST_UNREACH) && | |
828 | (ic->icmp6_type != ICMPV6_PKT_TOOBIG) && | |
829 | (ic->icmp6_type != ICMPV6_TIME_EXCEED)) { | |
830 | *related = 0; | |
831 | return NF_ACCEPT; | |
832 | } | |
833 | ||
834 | /* Now find the contained IP header */ | |
835 | offset += sizeof(_icmph); | |
836 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
837 | if (cih == NULL) | |
838 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
839 | ||
840 | pp = ip_vs_proto_get(cih->nexthdr); | |
841 | if (!pp) | |
842 | return NF_ACCEPT; | |
843 | ||
844 | /* Is the embedded protocol header present? */ | |
845 | /* TODO: we don't support fragmentation at the moment anyways */ | |
846 | if (unlikely(cih->nexthdr == IPPROTO_FRAGMENT && pp->dont_defrag)) | |
847 | return NF_ACCEPT; | |
848 | ||
849 | IP_VS_DBG_PKT(11, pp, skb, offset, "Checking outgoing ICMPv6 for"); | |
850 | ||
851 | offset += sizeof(struct ipv6hdr); | |
852 | ||
853 | ip_vs_fill_iphdr(AF_INET6, cih, &ciph); | |
854 | /* The embedded headers contain source and dest in reverse order */ | |
855 | cp = pp->conn_out_get(AF_INET6, skb, pp, &ciph, offset, 1); | |
856 | if (!cp) | |
857 | return NF_ACCEPT; | |
858 | ||
178f5e49 | 859 | ipv6_addr_copy(&snet.in6, &iph->saddr); |
f2428ed5 SH |
860 | return handle_response_icmp(AF_INET6, skb, &snet, cih->nexthdr, cp, |
861 | pp, offset, sizeof(struct ipv6hdr)); | |
2a3b791e JV |
862 | } |
863 | #endif | |
864 | ||
2906f66a VMR |
865 | /* |
866 | * Check if sctp chunc is ABORT chunk | |
867 | */ | |
868 | static inline int is_sctp_abort(const struct sk_buff *skb, int nh_len) | |
869 | { | |
870 | sctp_chunkhdr_t *sch, schunk; | |
871 | sch = skb_header_pointer(skb, nh_len + sizeof(sctp_sctphdr_t), | |
872 | sizeof(schunk), &schunk); | |
873 | if (sch == NULL) | |
874 | return 0; | |
875 | if (sch->type == SCTP_CID_ABORT) | |
876 | return 1; | |
877 | return 0; | |
878 | } | |
879 | ||
2a3b791e | 880 | static inline int is_tcp_reset(const struct sk_buff *skb, int nh_len) |
1da177e4 LT |
881 | { |
882 | struct tcphdr _tcph, *th; | |
883 | ||
2a3b791e | 884 | th = skb_header_pointer(skb, nh_len, sizeof(_tcph), &_tcph); |
1da177e4 LT |
885 | if (th == NULL) |
886 | return 0; | |
887 | return th->rst; | |
888 | } | |
889 | ||
4856c84c MT |
890 | /* Handle response packets: rewrite addresses and send away... |
891 | * Used for NAT and local client. | |
892 | */ | |
893 | static unsigned int | |
894 | handle_response(int af, struct sk_buff *skb, struct ip_vs_protocol *pp, | |
895 | struct ip_vs_conn *cp, int ihl) | |
896 | { | |
897 | IP_VS_DBG_PKT(11, pp, skb, 0, "Outgoing packet"); | |
898 | ||
899 | if (!skb_make_writable(skb, ihl)) | |
900 | goto drop; | |
901 | ||
902 | /* mangle the packet */ | |
903 | if (pp->snat_handler && !pp->snat_handler(skb, pp, cp)) | |
904 | goto drop; | |
905 | ||
906 | #ifdef CONFIG_IP_VS_IPV6 | |
907 | if (af == AF_INET6) | |
908 | ipv6_hdr(skb)->saddr = cp->vaddr.in6; | |
909 | else | |
910 | #endif | |
911 | { | |
912 | ip_hdr(skb)->saddr = cp->vaddr.ip; | |
913 | ip_send_check(ip_hdr(skb)); | |
914 | } | |
915 | ||
8a803040 JA |
916 | /* |
917 | * nf_iterate does not expect change in the skb->dst->dev. | |
918 | * It looks like it is not fatal to enable this code for hooks | |
919 | * where our handlers are at the end of the chain list and | |
920 | * when all next handlers use skb->dst->dev and not outdev. | |
921 | * It will definitely route properly the inout NAT traffic | |
922 | * when multiple paths are used. | |
923 | */ | |
924 | ||
4856c84c MT |
925 | /* For policy routing, packets originating from this |
926 | * machine itself may be routed differently to packets | |
927 | * passing through. We want this packet to be routed as | |
928 | * if it came from this machine itself. So re-compute | |
929 | * the routing information. | |
930 | */ | |
8a803040 | 931 | if (sysctl_ip_vs_snat_reroute) { |
4856c84c | 932 | #ifdef CONFIG_IP_VS_IPV6 |
8a803040 JA |
933 | if (af == AF_INET6) { |
934 | if (ip6_route_me_harder(skb) != 0) | |
935 | goto drop; | |
936 | } else | |
4856c84c | 937 | #endif |
8a803040 JA |
938 | if (ip_route_me_harder(skb, RTN_LOCAL) != 0) |
939 | goto drop; | |
940 | } | |
4856c84c | 941 | |
4856c84c MT |
942 | IP_VS_DBG_PKT(10, pp, skb, 0, "After SNAT"); |
943 | ||
944 | ip_vs_out_stats(cp, skb); | |
945 | ip_vs_set_state(cp, IP_VS_DIR_OUTPUT, skb, pp); | |
cf356d69 | 946 | skb->ipvs_property = 1; |
f4bc17cd | 947 | if (!(cp->flags & IP_VS_CONN_F_NFCT)) |
cf356d69 | 948 | ip_vs_notrack(skb); |
f4bc17cd JA |
949 | else |
950 | ip_vs_update_conntrack(skb, cp, 0); | |
4856c84c MT |
951 | ip_vs_conn_put(cp); |
952 | ||
4856c84c MT |
953 | LeaveFunction(11); |
954 | return NF_ACCEPT; | |
955 | ||
956 | drop: | |
957 | ip_vs_conn_put(cp); | |
958 | kfree_skb(skb); | |
f4bc17cd | 959 | LeaveFunction(11); |
4856c84c MT |
960 | return NF_STOLEN; |
961 | } | |
962 | ||
1da177e4 | 963 | /* |
6e23ae2a | 964 | * It is hooked at the NF_INET_FORWARD chain, used only for VS/NAT. |
4856c84c | 965 | * Check if outgoing packet belongs to the established ip_vs_conn. |
1da177e4 LT |
966 | */ |
967 | static unsigned int | |
3db05fea | 968 | ip_vs_out(unsigned int hooknum, struct sk_buff *skb, |
1da177e4 LT |
969 | const struct net_device *in, const struct net_device *out, |
970 | int (*okfn)(struct sk_buff *)) | |
971 | { | |
51ef348b | 972 | struct ip_vs_iphdr iph; |
1da177e4 LT |
973 | struct ip_vs_protocol *pp; |
974 | struct ip_vs_conn *cp; | |
2a3b791e | 975 | int af; |
1da177e4 LT |
976 | |
977 | EnterFunction(11); | |
978 | ||
60678040 | 979 | af = (skb->protocol == htons(ETH_P_IP)) ? AF_INET : AF_INET6; |
2a3b791e | 980 | |
6869c4d8 | 981 | if (skb->ipvs_property) |
1da177e4 LT |
982 | return NF_ACCEPT; |
983 | ||
2a3b791e JV |
984 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); |
985 | #ifdef CONFIG_IP_VS_IPV6 | |
986 | if (af == AF_INET6) { | |
987 | if (unlikely(iph.protocol == IPPROTO_ICMPV6)) { | |
988 | int related, verdict = ip_vs_out_icmp_v6(skb, &related); | |
1da177e4 | 989 | |
8a803040 JA |
990 | if (related) { |
991 | if (sysctl_ip_vs_snat_reroute && | |
992 | NF_ACCEPT == verdict && | |
993 | ip6_route_me_harder(skb)) | |
994 | verdict = NF_DROP; | |
2a3b791e | 995 | return verdict; |
8a803040 | 996 | } |
2a3b791e JV |
997 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); |
998 | } | |
999 | } else | |
1000 | #endif | |
1001 | if (unlikely(iph.protocol == IPPROTO_ICMP)) { | |
1002 | int related, verdict = ip_vs_out_icmp(skb, &related); | |
1003 | ||
8a803040 JA |
1004 | if (related) { |
1005 | if (sysctl_ip_vs_snat_reroute && | |
1006 | NF_ACCEPT == verdict && | |
1007 | ip_route_me_harder(skb, RTN_LOCAL)) | |
1008 | verdict = NF_DROP; | |
2a3b791e | 1009 | return verdict; |
8a803040 | 1010 | } |
2a3b791e JV |
1011 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); |
1012 | } | |
1da177e4 | 1013 | |
51ef348b | 1014 | pp = ip_vs_proto_get(iph.protocol); |
1da177e4 LT |
1015 | if (unlikely(!pp)) |
1016 | return NF_ACCEPT; | |
1017 | ||
1018 | /* reassemble IP fragments */ | |
2a3b791e JV |
1019 | #ifdef CONFIG_IP_VS_IPV6 |
1020 | if (af == AF_INET6) { | |
1021 | if (unlikely(iph.protocol == IPPROTO_ICMPV6)) { | |
1022 | int related, verdict = ip_vs_out_icmp_v6(skb, &related); | |
1da177e4 | 1023 | |
2a3b791e JV |
1024 | if (related) |
1025 | return verdict; | |
1026 | ||
1027 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); | |
1028 | } | |
1029 | } else | |
1030 | #endif | |
1031 | if (unlikely(ip_hdr(skb)->frag_off & htons(IP_MF|IP_OFFSET) && | |
1032 | !pp->dont_defrag)) { | |
1033 | if (ip_vs_gather_frags(skb, IP_DEFRAG_VS_OUT)) | |
1034 | return NF_STOLEN; | |
1035 | ||
1036 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); | |
1037 | } | |
1da177e4 LT |
1038 | |
1039 | /* | |
1040 | * Check if the packet belongs to an existing entry | |
1041 | */ | |
2a3b791e | 1042 | cp = pp->conn_out_get(af, skb, pp, &iph, iph.len, 0); |
1da177e4 LT |
1043 | |
1044 | if (unlikely(!cp)) { | |
1045 | if (sysctl_ip_vs_nat_icmp_send && | |
1046 | (pp->protocol == IPPROTO_TCP || | |
2906f66a VMR |
1047 | pp->protocol == IPPROTO_UDP || |
1048 | pp->protocol == IPPROTO_SCTP)) { | |
014d730d | 1049 | __be16 _ports[2], *pptr; |
1da177e4 | 1050 | |
51ef348b | 1051 | pptr = skb_header_pointer(skb, iph.len, |
1da177e4 LT |
1052 | sizeof(_ports), _ports); |
1053 | if (pptr == NULL) | |
1054 | return NF_ACCEPT; /* Not for me */ | |
7937df15 JV |
1055 | if (ip_vs_lookup_real_service(af, iph.protocol, |
1056 | &iph.saddr, | |
2a3b791e | 1057 | pptr[0])) { |
1da177e4 LT |
1058 | /* |
1059 | * Notify the real server: there is no | |
1060 | * existing entry if it is not RST | |
1061 | * packet or not TCP packet. | |
1062 | */ | |
2906f66a VMR |
1063 | if ((iph.protocol != IPPROTO_TCP && |
1064 | iph.protocol != IPPROTO_SCTP) | |
1065 | || ((iph.protocol == IPPROTO_TCP | |
1066 | && !is_tcp_reset(skb, iph.len)) | |
1067 | || (iph.protocol == IPPROTO_SCTP | |
1068 | && !is_sctp_abort(skb, | |
1069 | iph.len)))) { | |
2a3b791e JV |
1070 | #ifdef CONFIG_IP_VS_IPV6 |
1071 | if (af == AF_INET6) | |
1072 | icmpv6_send(skb, | |
1073 | ICMPV6_DEST_UNREACH, | |
1074 | ICMPV6_PORT_UNREACH, | |
3ffe533c | 1075 | 0); |
2a3b791e JV |
1076 | else |
1077 | #endif | |
1078 | icmp_send(skb, | |
1079 | ICMP_DEST_UNREACH, | |
1080 | ICMP_PORT_UNREACH, 0); | |
1da177e4 | 1081 | return NF_DROP; |
5af149cc | 1082 | } |
1da177e4 LT |
1083 | } |
1084 | } | |
1085 | IP_VS_DBG_PKT(12, pp, skb, 0, | |
1086 | "packet continues traversal as normal"); | |
1087 | return NF_ACCEPT; | |
1088 | } | |
1089 | ||
4856c84c | 1090 | return handle_response(af, skb, pp, cp, iph.len); |
1da177e4 LT |
1091 | } |
1092 | ||
1093 | ||
1094 | /* | |
1095 | * Handle ICMP messages in the outside-to-inside direction (incoming). | |
1096 | * Find any that might be relevant, check against existing connections, | |
1097 | * forward to the right destination host if relevant. | |
1098 | * Currently handles error types - unreachable, quench, ttl exceeded. | |
1099 | */ | |
e905a9ed | 1100 | static int |
3db05fea | 1101 | ip_vs_in_icmp(struct sk_buff *skb, int *related, unsigned int hooknum) |
1da177e4 | 1102 | { |
1da177e4 LT |
1103 | struct iphdr *iph; |
1104 | struct icmphdr _icmph, *ic; | |
1105 | struct iphdr _ciph, *cih; /* The ip header contained within the ICMP */ | |
51ef348b | 1106 | struct ip_vs_iphdr ciph; |
1da177e4 LT |
1107 | struct ip_vs_conn *cp; |
1108 | struct ip_vs_protocol *pp; | |
1109 | unsigned int offset, ihl, verdict; | |
f2428ed5 | 1110 | union nf_inet_addr snet; |
1da177e4 LT |
1111 | |
1112 | *related = 1; | |
1113 | ||
1114 | /* reassemble IP fragments */ | |
eddc9ec5 | 1115 | if (ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)) { |
6e23ae2a | 1116 | if (ip_vs_gather_frags(skb, hooknum == NF_INET_LOCAL_IN ? |
776c729e | 1117 | IP_DEFRAG_VS_IN : IP_DEFRAG_VS_FWD)) |
1da177e4 | 1118 | return NF_STOLEN; |
1da177e4 LT |
1119 | } |
1120 | ||
eddc9ec5 | 1121 | iph = ip_hdr(skb); |
1da177e4 LT |
1122 | offset = ihl = iph->ihl * 4; |
1123 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
1124 | if (ic == NULL) | |
1125 | return NF_DROP; | |
1126 | ||
14d5e834 | 1127 | IP_VS_DBG(12, "Incoming ICMP (%d,%d) %pI4->%pI4\n", |
1da177e4 | 1128 | ic->type, ntohs(icmp_id(ic)), |
14d5e834 | 1129 | &iph->saddr, &iph->daddr); |
1da177e4 LT |
1130 | |
1131 | /* | |
1132 | * Work through seeing if this is for us. | |
1133 | * These checks are supposed to be in an order that means easy | |
1134 | * things are checked first to speed up processing.... however | |
1135 | * this means that some packets will manage to get a long way | |
1136 | * down this stack and then be rejected, but that's life. | |
1137 | */ | |
1138 | if ((ic->type != ICMP_DEST_UNREACH) && | |
1139 | (ic->type != ICMP_SOURCE_QUENCH) && | |
1140 | (ic->type != ICMP_TIME_EXCEEDED)) { | |
1141 | *related = 0; | |
1142 | return NF_ACCEPT; | |
1143 | } | |
1144 | ||
1145 | /* Now find the contained IP header */ | |
1146 | offset += sizeof(_icmph); | |
1147 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
1148 | if (cih == NULL) | |
1149 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
1150 | ||
1151 | pp = ip_vs_proto_get(cih->protocol); | |
1152 | if (!pp) | |
1153 | return NF_ACCEPT; | |
1154 | ||
1155 | /* Is the embedded protocol header present? */ | |
4412ec49 | 1156 | if (unlikely(cih->frag_off & htons(IP_OFFSET) && |
1da177e4 LT |
1157 | pp->dont_defrag)) |
1158 | return NF_ACCEPT; | |
1159 | ||
1160 | IP_VS_DBG_PKT(11, pp, skb, offset, "Checking incoming ICMP for"); | |
1161 | ||
1162 | offset += cih->ihl * 4; | |
1163 | ||
51ef348b | 1164 | ip_vs_fill_iphdr(AF_INET, cih, &ciph); |
1da177e4 | 1165 | /* The embedded headers contain source and dest in reverse order */ |
51ef348b | 1166 | cp = pp->conn_in_get(AF_INET, skb, pp, &ciph, offset, 1); |
4856c84c MT |
1167 | if (!cp) { |
1168 | /* The packet could also belong to a local client */ | |
1169 | cp = pp->conn_out_get(AF_INET, skb, pp, &ciph, offset, 1); | |
f2428ed5 SH |
1170 | if (cp) { |
1171 | snet.ip = iph->saddr; | |
1172 | return handle_response_icmp(AF_INET, skb, &snet, | |
1173 | cih->protocol, cp, pp, | |
4856c84c | 1174 | offset, ihl); |
f2428ed5 | 1175 | } |
1da177e4 | 1176 | return NF_ACCEPT; |
4856c84c | 1177 | } |
1da177e4 LT |
1178 | |
1179 | verdict = NF_DROP; | |
1180 | ||
1181 | /* Ensure the checksum is correct */ | |
60476372 | 1182 | if (!skb_csum_unnecessary(skb) && ip_vs_checksum_complete(skb, ihl)) { |
1da177e4 | 1183 | /* Failed checksum! */ |
14d5e834 HH |
1184 | IP_VS_DBG(1, "Incoming ICMP: failed checksum from %pI4!\n", |
1185 | &iph->saddr); | |
1da177e4 LT |
1186 | goto out; |
1187 | } | |
1188 | ||
1189 | /* do the statistics and put it back */ | |
1190 | ip_vs_in_stats(cp, skb); | |
1191 | if (IPPROTO_TCP == cih->protocol || IPPROTO_UDP == cih->protocol) | |
1192 | offset += 2 * sizeof(__u16); | |
1193 | verdict = ip_vs_icmp_xmit(skb, cp, pp, offset); | |
1194 | /* do not touch skb anymore */ | |
1195 | ||
1196 | out: | |
1197 | __ip_vs_conn_put(cp); | |
1198 | ||
1199 | return verdict; | |
1200 | } | |
1201 | ||
2a3b791e JV |
1202 | #ifdef CONFIG_IP_VS_IPV6 |
1203 | static int | |
1204 | ip_vs_in_icmp_v6(struct sk_buff *skb, int *related, unsigned int hooknum) | |
1205 | { | |
1206 | struct ipv6hdr *iph; | |
1207 | struct icmp6hdr _icmph, *ic; | |
1208 | struct ipv6hdr _ciph, *cih; /* The ip header contained | |
1209 | within the ICMP */ | |
1210 | struct ip_vs_iphdr ciph; | |
1211 | struct ip_vs_conn *cp; | |
1212 | struct ip_vs_protocol *pp; | |
1213 | unsigned int offset, verdict; | |
f2428ed5 | 1214 | union nf_inet_addr snet; |
2a3b791e JV |
1215 | |
1216 | *related = 1; | |
1217 | ||
1218 | /* reassemble IP fragments */ | |
1219 | if (ipv6_hdr(skb)->nexthdr == IPPROTO_FRAGMENT) { | |
1220 | if (ip_vs_gather_frags_v6(skb, hooknum == NF_INET_LOCAL_IN ? | |
1221 | IP_DEFRAG_VS_IN : | |
1222 | IP_DEFRAG_VS_FWD)) | |
1223 | return NF_STOLEN; | |
1224 | } | |
1225 | ||
1226 | iph = ipv6_hdr(skb); | |
1227 | offset = sizeof(struct ipv6hdr); | |
1228 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
1229 | if (ic == NULL) | |
1230 | return NF_DROP; | |
1231 | ||
5b095d98 | 1232 | IP_VS_DBG(12, "Incoming ICMPv6 (%d,%d) %pI6->%pI6\n", |
2a3b791e | 1233 | ic->icmp6_type, ntohs(icmpv6_id(ic)), |
38ff4fa4 | 1234 | &iph->saddr, &iph->daddr); |
2a3b791e JV |
1235 | |
1236 | /* | |
1237 | * Work through seeing if this is for us. | |
1238 | * These checks are supposed to be in an order that means easy | |
1239 | * things are checked first to speed up processing.... however | |
1240 | * this means that some packets will manage to get a long way | |
1241 | * down this stack and then be rejected, but that's life. | |
1242 | */ | |
1243 | if ((ic->icmp6_type != ICMPV6_DEST_UNREACH) && | |
1244 | (ic->icmp6_type != ICMPV6_PKT_TOOBIG) && | |
1245 | (ic->icmp6_type != ICMPV6_TIME_EXCEED)) { | |
1246 | *related = 0; | |
1247 | return NF_ACCEPT; | |
1248 | } | |
1249 | ||
1250 | /* Now find the contained IP header */ | |
1251 | offset += sizeof(_icmph); | |
1252 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
1253 | if (cih == NULL) | |
1254 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
1255 | ||
1256 | pp = ip_vs_proto_get(cih->nexthdr); | |
1257 | if (!pp) | |
1258 | return NF_ACCEPT; | |
1259 | ||
1260 | /* Is the embedded protocol header present? */ | |
1261 | /* TODO: we don't support fragmentation at the moment anyways */ | |
1262 | if (unlikely(cih->nexthdr == IPPROTO_FRAGMENT && pp->dont_defrag)) | |
1263 | return NF_ACCEPT; | |
1264 | ||
1265 | IP_VS_DBG_PKT(11, pp, skb, offset, "Checking incoming ICMPv6 for"); | |
1266 | ||
1267 | offset += sizeof(struct ipv6hdr); | |
1268 | ||
1269 | ip_vs_fill_iphdr(AF_INET6, cih, &ciph); | |
1270 | /* The embedded headers contain source and dest in reverse order */ | |
1271 | cp = pp->conn_in_get(AF_INET6, skb, pp, &ciph, offset, 1); | |
f2428ed5 SH |
1272 | if (!cp) { |
1273 | /* The packet could also belong to a local client */ | |
1274 | cp = pp->conn_out_get(AF_INET6, skb, pp, &ciph, offset, 1); | |
1275 | if (cp) { | |
178f5e49 | 1276 | ipv6_addr_copy(&snet.in6, &iph->saddr); |
f2428ed5 SH |
1277 | return handle_response_icmp(AF_INET6, skb, &snet, |
1278 | cih->nexthdr, | |
1279 | cp, pp, offset, | |
1280 | sizeof(struct ipv6hdr)); | |
1281 | } | |
2a3b791e | 1282 | return NF_ACCEPT; |
f2428ed5 | 1283 | } |
2a3b791e JV |
1284 | |
1285 | verdict = NF_DROP; | |
1286 | ||
1287 | /* do the statistics and put it back */ | |
1288 | ip_vs_in_stats(cp, skb); | |
2906f66a VMR |
1289 | if (IPPROTO_TCP == cih->nexthdr || IPPROTO_UDP == cih->nexthdr || |
1290 | IPPROTO_SCTP == cih->nexthdr) | |
2a3b791e JV |
1291 | offset += 2 * sizeof(__u16); |
1292 | verdict = ip_vs_icmp_xmit_v6(skb, cp, pp, offset); | |
1293 | /* do not touch skb anymore */ | |
1294 | ||
1295 | __ip_vs_conn_put(cp); | |
1296 | ||
1297 | return verdict; | |
1298 | } | |
1299 | #endif | |
1300 | ||
1301 | ||
1da177e4 LT |
1302 | /* |
1303 | * Check if it's for virtual services, look it up, | |
1304 | * and send it on its way... | |
1305 | */ | |
1306 | static unsigned int | |
3db05fea | 1307 | ip_vs_in(unsigned int hooknum, struct sk_buff *skb, |
1da177e4 LT |
1308 | const struct net_device *in, const struct net_device *out, |
1309 | int (*okfn)(struct sk_buff *)) | |
1310 | { | |
51ef348b | 1311 | struct ip_vs_iphdr iph; |
1da177e4 LT |
1312 | struct ip_vs_protocol *pp; |
1313 | struct ip_vs_conn *cp; | |
1e66dafc | 1314 | int ret, restart, af, pkts; |
2a3b791e | 1315 | |
60678040 | 1316 | af = (skb->protocol == htons(ETH_P_IP)) ? AF_INET : AF_INET6; |
51ef348b | 1317 | |
2a3b791e | 1318 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); |
1da177e4 LT |
1319 | |
1320 | /* | |
4856c84c MT |
1321 | * Big tappo: only PACKET_HOST, including loopback for local client |
1322 | * Don't handle local packets on IPv6 for now | |
1da177e4 | 1323 | */ |
f2428ed5 | 1324 | if (unlikely(skb->pkt_type != PACKET_HOST)) { |
51ef348b JV |
1325 | IP_VS_DBG_BUF(12, "packet type=%d proto=%d daddr=%s ignored\n", |
1326 | skb->pkt_type, | |
1327 | iph.protocol, | |
2a3b791e | 1328 | IP_VS_DBG_ADDR(af, &iph.daddr)); |
1da177e4 LT |
1329 | return NF_ACCEPT; |
1330 | } | |
1331 | ||
94b26551 JV |
1332 | #ifdef CONFIG_IP_VS_IPV6 |
1333 | if (af == AF_INET6) { | |
1334 | if (unlikely(iph.protocol == IPPROTO_ICMPV6)) { | |
1335 | int related, verdict = ip_vs_in_icmp_v6(skb, &related, hooknum); | |
1da177e4 | 1336 | |
94b26551 JV |
1337 | if (related) |
1338 | return verdict; | |
1339 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); | |
1340 | } | |
1341 | } else | |
1342 | #endif | |
1343 | if (unlikely(iph.protocol == IPPROTO_ICMP)) { | |
1344 | int related, verdict = ip_vs_in_icmp(skb, &related, hooknum); | |
1345 | ||
1346 | if (related) | |
1347 | return verdict; | |
1348 | ip_vs_fill_iphdr(af, skb_network_header(skb), &iph); | |
1349 | } | |
1da177e4 LT |
1350 | |
1351 | /* Protocol supported? */ | |
51ef348b | 1352 | pp = ip_vs_proto_get(iph.protocol); |
1da177e4 LT |
1353 | if (unlikely(!pp)) |
1354 | return NF_ACCEPT; | |
1355 | ||
1da177e4 LT |
1356 | /* |
1357 | * Check if the packet belongs to an existing connection entry | |
1358 | */ | |
2a3b791e | 1359 | cp = pp->conn_in_get(af, skb, pp, &iph, iph.len, 0); |
1da177e4 LT |
1360 | |
1361 | if (unlikely(!cp)) { | |
1362 | int v; | |
1363 | ||
4856c84c MT |
1364 | /* For local client packets, it could be a response */ |
1365 | cp = pp->conn_out_get(af, skb, pp, &iph, iph.len, 0); | |
1366 | if (cp) | |
1367 | return handle_response(af, skb, pp, cp, iph.len); | |
1368 | ||
2a3b791e | 1369 | if (!pp->conn_schedule(af, skb, pp, &v, &cp)) |
1da177e4 LT |
1370 | return v; |
1371 | } | |
1372 | ||
1373 | if (unlikely(!cp)) { | |
1374 | /* sorry, all this trouble for a no-hit :) */ | |
1375 | IP_VS_DBG_PKT(12, pp, skb, 0, | |
1376 | "packet continues traversal as normal"); | |
1377 | return NF_ACCEPT; | |
1378 | } | |
1379 | ||
1380 | IP_VS_DBG_PKT(11, pp, skb, 0, "Incoming packet"); | |
1381 | ||
1382 | /* Check the server status */ | |
1383 | if (cp->dest && !(cp->dest->flags & IP_VS_DEST_F_AVAILABLE)) { | |
1384 | /* the destination server is not available */ | |
1385 | ||
1386 | if (sysctl_ip_vs_expire_nodest_conn) { | |
1387 | /* try to expire the connection immediately */ | |
1388 | ip_vs_conn_expire_now(cp); | |
1da177e4 | 1389 | } |
dc8103f2 JA |
1390 | /* don't restart its timer, and silently |
1391 | drop the packet. */ | |
1392 | __ip_vs_conn_put(cp); | |
1da177e4 LT |
1393 | return NF_DROP; |
1394 | } | |
1395 | ||
1396 | ip_vs_in_stats(cp, skb); | |
1397 | restart = ip_vs_set_state(cp, IP_VS_DIR_INPUT, skb, pp); | |
1398 | if (cp->packet_xmit) | |
1399 | ret = cp->packet_xmit(skb, cp, pp); | |
1400 | /* do not touch skb anymore */ | |
1401 | else { | |
1402 | IP_VS_DBG_RL("warning: packet_xmit is null"); | |
1403 | ret = NF_ACCEPT; | |
1404 | } | |
1405 | ||
efac5276 RB |
1406 | /* Increase its packet counter and check if it is needed |
1407 | * to be synchronized | |
1408 | * | |
1409 | * Sync connection if it is about to close to | |
1410 | * encorage the standby servers to update the connections timeout | |
1411 | */ | |
1e66dafc | 1412 | pkts = atomic_add_return(1, &cp->in_pkts); |
2906f66a VMR |
1413 | if (af == AF_INET && (ip_vs_sync_state & IP_VS_STATE_MASTER) && |
1414 | cp->protocol == IPPROTO_SCTP) { | |
1415 | if ((cp->state == IP_VS_SCTP_S_ESTABLISHED && | |
8ed2163f | 1416 | (pkts % sysctl_ip_vs_sync_threshold[1] |
2906f66a VMR |
1417 | == sysctl_ip_vs_sync_threshold[0])) || |
1418 | (cp->old_state != cp->state && | |
1419 | ((cp->state == IP_VS_SCTP_S_CLOSED) || | |
1420 | (cp->state == IP_VS_SCTP_S_SHUT_ACK_CLI) || | |
1421 | (cp->state == IP_VS_SCTP_S_SHUT_ACK_SER)))) { | |
1422 | ip_vs_sync_conn(cp); | |
1423 | goto out; | |
1424 | } | |
1425 | } | |
1426 | ||
8ed2163f JA |
1427 | /* Keep this block last: TCP and others with pp->num_states <= 1 */ |
1428 | else if (af == AF_INET && | |
c6883f58 | 1429 | (ip_vs_sync_state & IP_VS_STATE_MASTER) && |
efac5276 RB |
1430 | (((cp->protocol != IPPROTO_TCP || |
1431 | cp->state == IP_VS_TCP_S_ESTABLISHED) && | |
1e66dafc | 1432 | (pkts % sysctl_ip_vs_sync_threshold[1] |
efac5276 RB |
1433 | == sysctl_ip_vs_sync_threshold[0])) || |
1434 | ((cp->protocol == IPPROTO_TCP) && (cp->old_state != cp->state) && | |
1435 | ((cp->state == IP_VS_TCP_S_FIN_WAIT) || | |
9abfe315 | 1436 | (cp->state == IP_VS_TCP_S_CLOSE) || |
9d3a0de7 RB |
1437 | (cp->state == IP_VS_TCP_S_CLOSE_WAIT) || |
1438 | (cp->state == IP_VS_TCP_S_TIME_WAIT))))) | |
1da177e4 | 1439 | ip_vs_sync_conn(cp); |
2906f66a | 1440 | out: |
efac5276 | 1441 | cp->old_state = cp->state; |
1da177e4 LT |
1442 | |
1443 | ip_vs_conn_put(cp); | |
1444 | return ret; | |
1445 | } | |
1446 | ||
1447 | ||
1448 | /* | |
6e23ae2a | 1449 | * It is hooked at the NF_INET_FORWARD chain, in order to catch ICMP |
1da177e4 LT |
1450 | * related packets destined for 0.0.0.0/0. |
1451 | * When fwmark-based virtual service is used, such as transparent | |
1452 | * cache cluster, TCP packets can be marked and routed to ip_vs_in, | |
1453 | * but ICMP destined for 0.0.0.0/0 cannot not be easily marked and | |
6e23ae2a | 1454 | * sent to ip_vs_in_icmp. So, catch them at the NF_INET_FORWARD chain |
1da177e4 LT |
1455 | * and send them to ip_vs_in_icmp. |
1456 | */ | |
1457 | static unsigned int | |
3db05fea | 1458 | ip_vs_forward_icmp(unsigned int hooknum, struct sk_buff *skb, |
1da177e4 LT |
1459 | const struct net_device *in, const struct net_device *out, |
1460 | int (*okfn)(struct sk_buff *)) | |
1461 | { | |
1462 | int r; | |
1463 | ||
3db05fea | 1464 | if (ip_hdr(skb)->protocol != IPPROTO_ICMP) |
1da177e4 LT |
1465 | return NF_ACCEPT; |
1466 | ||
3db05fea | 1467 | return ip_vs_in_icmp(skb, &r, hooknum); |
1da177e4 LT |
1468 | } |
1469 | ||
2a3b791e JV |
1470 | #ifdef CONFIG_IP_VS_IPV6 |
1471 | static unsigned int | |
1472 | ip_vs_forward_icmp_v6(unsigned int hooknum, struct sk_buff *skb, | |
1473 | const struct net_device *in, const struct net_device *out, | |
1474 | int (*okfn)(struct sk_buff *)) | |
1475 | { | |
1476 | int r; | |
1477 | ||
1478 | if (ipv6_hdr(skb)->nexthdr != IPPROTO_ICMPV6) | |
1479 | return NF_ACCEPT; | |
1480 | ||
1481 | return ip_vs_in_icmp_v6(skb, &r, hooknum); | |
1482 | } | |
1483 | #endif | |
1484 | ||
1da177e4 | 1485 | |
1999414a | 1486 | static struct nf_hook_ops ip_vs_ops[] __read_mostly = { |
41c5b317 PM |
1487 | /* After packet filtering, forward packet through VS/DR, VS/TUN, |
1488 | * or VS/NAT(change destination), so that filtering rules can be | |
1489 | * applied to IPVS. */ | |
1490 | { | |
1491 | .hook = ip_vs_in, | |
1492 | .owner = THIS_MODULE, | |
1493 | .pf = PF_INET, | |
1494 | .hooknum = NF_INET_LOCAL_IN, | |
1495 | .priority = 100, | |
1496 | }, | |
1497 | /* After packet filtering, change source only for VS/NAT */ | |
1498 | { | |
1499 | .hook = ip_vs_out, | |
1500 | .owner = THIS_MODULE, | |
1501 | .pf = PF_INET, | |
1502 | .hooknum = NF_INET_FORWARD, | |
1503 | .priority = 100, | |
1504 | }, | |
1505 | /* After packet filtering (but before ip_vs_out_icmp), catch icmp | |
1506 | * destined for 0.0.0.0/0, which is for incoming IPVS connections */ | |
1507 | { | |
1508 | .hook = ip_vs_forward_icmp, | |
1509 | .owner = THIS_MODULE, | |
1510 | .pf = PF_INET, | |
1511 | .hooknum = NF_INET_FORWARD, | |
1512 | .priority = 99, | |
1513 | }, | |
473b23d3 JV |
1514 | #ifdef CONFIG_IP_VS_IPV6 |
1515 | /* After packet filtering, forward packet through VS/DR, VS/TUN, | |
1516 | * or VS/NAT(change destination), so that filtering rules can be | |
1517 | * applied to IPVS. */ | |
1518 | { | |
1519 | .hook = ip_vs_in, | |
1520 | .owner = THIS_MODULE, | |
1521 | .pf = PF_INET6, | |
1522 | .hooknum = NF_INET_LOCAL_IN, | |
1523 | .priority = 100, | |
1524 | }, | |
1525 | /* After packet filtering, change source only for VS/NAT */ | |
1526 | { | |
1527 | .hook = ip_vs_out, | |
1528 | .owner = THIS_MODULE, | |
1529 | .pf = PF_INET6, | |
1530 | .hooknum = NF_INET_FORWARD, | |
1531 | .priority = 100, | |
1532 | }, | |
1533 | /* After packet filtering (but before ip_vs_out_icmp), catch icmp | |
1534 | * destined for 0.0.0.0/0, which is for incoming IPVS connections */ | |
1535 | { | |
1536 | .hook = ip_vs_forward_icmp_v6, | |
1537 | .owner = THIS_MODULE, | |
1538 | .pf = PF_INET6, | |
1539 | .hooknum = NF_INET_FORWARD, | |
1540 | .priority = 99, | |
1541 | }, | |
473b23d3 | 1542 | #endif |
1da177e4 LT |
1543 | }; |
1544 | ||
1545 | ||
1546 | /* | |
1547 | * Initialize IP Virtual Server | |
1548 | */ | |
1549 | static int __init ip_vs_init(void) | |
1550 | { | |
1551 | int ret; | |
1552 | ||
a919cf4b SW |
1553 | ip_vs_estimator_init(); |
1554 | ||
1da177e4 LT |
1555 | ret = ip_vs_control_init(); |
1556 | if (ret < 0) { | |
1e3e238e | 1557 | pr_err("can't setup control.\n"); |
a919cf4b | 1558 | goto cleanup_estimator; |
1da177e4 LT |
1559 | } |
1560 | ||
1561 | ip_vs_protocol_init(); | |
1562 | ||
1563 | ret = ip_vs_app_init(); | |
1564 | if (ret < 0) { | |
1e3e238e | 1565 | pr_err("can't setup application helper.\n"); |
1da177e4 LT |
1566 | goto cleanup_protocol; |
1567 | } | |
1568 | ||
1569 | ret = ip_vs_conn_init(); | |
1570 | if (ret < 0) { | |
1e3e238e | 1571 | pr_err("can't setup connection table.\n"); |
1da177e4 LT |
1572 | goto cleanup_app; |
1573 | } | |
1574 | ||
41c5b317 | 1575 | ret = nf_register_hooks(ip_vs_ops, ARRAY_SIZE(ip_vs_ops)); |
1da177e4 | 1576 | if (ret < 0) { |
1e3e238e | 1577 | pr_err("can't register hooks.\n"); |
1da177e4 LT |
1578 | goto cleanup_conn; |
1579 | } | |
1580 | ||
1e3e238e | 1581 | pr_info("ipvs loaded.\n"); |
1da177e4 LT |
1582 | return ret; |
1583 | ||
1da177e4 LT |
1584 | cleanup_conn: |
1585 | ip_vs_conn_cleanup(); | |
1586 | cleanup_app: | |
1587 | ip_vs_app_cleanup(); | |
1588 | cleanup_protocol: | |
1589 | ip_vs_protocol_cleanup(); | |
1590 | ip_vs_control_cleanup(); | |
a919cf4b SW |
1591 | cleanup_estimator: |
1592 | ip_vs_estimator_cleanup(); | |
1da177e4 LT |
1593 | return ret; |
1594 | } | |
1595 | ||
1596 | static void __exit ip_vs_cleanup(void) | |
1597 | { | |
41c5b317 | 1598 | nf_unregister_hooks(ip_vs_ops, ARRAY_SIZE(ip_vs_ops)); |
1da177e4 LT |
1599 | ip_vs_conn_cleanup(); |
1600 | ip_vs_app_cleanup(); | |
1601 | ip_vs_protocol_cleanup(); | |
1602 | ip_vs_control_cleanup(); | |
a919cf4b | 1603 | ip_vs_estimator_cleanup(); |
1e3e238e | 1604 | pr_info("ipvs unloaded.\n"); |
1da177e4 LT |
1605 | } |
1606 | ||
1607 | module_init(ip_vs_init); | |
1608 | module_exit(ip_vs_cleanup); | |
1609 | MODULE_LICENSE("GPL"); |