]>
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> |
61b1ab45 | 44 | #include <net/netns/generic.h> /* net_generic() */ |
1da177e4 LT |
45 | |
46 | #include <linux/netfilter.h> | |
47 | #include <linux/netfilter_ipv4.h> | |
48 | ||
2a3b791e JV |
49 | #ifdef CONFIG_IP_VS_IPV6 |
50 | #include <net/ipv6.h> | |
51 | #include <linux/netfilter_ipv6.h> | |
489fdeda | 52 | #include <net/ip6_route.h> |
2a3b791e JV |
53 | #endif |
54 | ||
1da177e4 LT |
55 | #include <net/ip_vs.h> |
56 | ||
57 | ||
58 | EXPORT_SYMBOL(register_ip_vs_scheduler); | |
59 | EXPORT_SYMBOL(unregister_ip_vs_scheduler); | |
1da177e4 LT |
60 | EXPORT_SYMBOL(ip_vs_proto_name); |
61 | EXPORT_SYMBOL(ip_vs_conn_new); | |
62 | EXPORT_SYMBOL(ip_vs_conn_in_get); | |
63 | EXPORT_SYMBOL(ip_vs_conn_out_get); | |
64 | #ifdef CONFIG_IP_VS_PROTO_TCP | |
65 | EXPORT_SYMBOL(ip_vs_tcp_conn_listen); | |
66 | #endif | |
67 | EXPORT_SYMBOL(ip_vs_conn_put); | |
68 | #ifdef CONFIG_IP_VS_DEBUG | |
69 | EXPORT_SYMBOL(ip_vs_get_debug_level); | |
70 | #endif | |
39b97223 | 71 | EXPORT_SYMBOL(ip_vs_new_conn_out); |
1da177e4 | 72 | |
c7d03a00 | 73 | static unsigned int ip_vs_net_id __read_mostly; |
61b1ab45 HS |
74 | /* netns cnt used for uniqueness */ |
75 | static atomic_t ipvs_netns_cnt = ATOMIC_INIT(0); | |
1da177e4 LT |
76 | |
77 | /* ID used in ICMP lookups */ | |
78 | #define icmp_id(icmph) (((icmph)->un).echo.id) | |
2a3b791e | 79 | #define icmpv6_id(icmph) (icmph->icmp6_dataun.u_echo.identifier) |
1da177e4 | 80 | |
95c96174 | 81 | const char *ip_vs_proto_name(unsigned int proto) |
1da177e4 LT |
82 | { |
83 | static char buf[20]; | |
84 | ||
85 | switch (proto) { | |
86 | case IPPROTO_IP: | |
87 | return "IP"; | |
88 | case IPPROTO_UDP: | |
89 | return "UDP"; | |
90 | case IPPROTO_TCP: | |
91 | return "TCP"; | |
2906f66a VMR |
92 | case IPPROTO_SCTP: |
93 | return "SCTP"; | |
1da177e4 LT |
94 | case IPPROTO_ICMP: |
95 | return "ICMP"; | |
2a3b791e JV |
96 | #ifdef CONFIG_IP_VS_IPV6 |
97 | case IPPROTO_ICMPV6: | |
98 | return "ICMPv6"; | |
99 | #endif | |
1da177e4 | 100 | default: |
1404c3ab | 101 | sprintf(buf, "IP_%u", proto); |
1da177e4 LT |
102 | return buf; |
103 | } | |
104 | } | |
105 | ||
106 | void ip_vs_init_hash_table(struct list_head *table, int rows) | |
107 | { | |
108 | while (--rows >= 0) | |
109 | INIT_LIST_HEAD(&table[rows]); | |
110 | } | |
111 | ||
112 | static inline void | |
113 | ip_vs_in_stats(struct ip_vs_conn *cp, struct sk_buff *skb) | |
114 | { | |
115 | struct ip_vs_dest *dest = cp->dest; | |
7c08d78e | 116 | struct netns_ipvs *ipvs = cp->ipvs; |
b17fc996 | 117 | |
1da177e4 | 118 | if (dest && (dest->flags & IP_VS_DEST_F_AVAILABLE)) { |
b17fc996 | 119 | struct ip_vs_cpu_stats *s; |
bcbde4c0 | 120 | struct ip_vs_service *svc; |
b17fc996 HS |
121 | |
122 | s = this_cpu_ptr(dest->stats.cpustats); | |
b17fc996 | 123 | u64_stats_update_begin(&s->syncp); |
cd67cd5e JA |
124 | s->cnt.inpkts++; |
125 | s->cnt.inbytes += skb->len; | |
b17fc996 HS |
126 | u64_stats_update_end(&s->syncp); |
127 | ||
bcbde4c0 JA |
128 | rcu_read_lock(); |
129 | svc = rcu_dereference(dest->svc); | |
130 | s = this_cpu_ptr(svc->stats.cpustats); | |
b17fc996 | 131 | u64_stats_update_begin(&s->syncp); |
cd67cd5e JA |
132 | s->cnt.inpkts++; |
133 | s->cnt.inbytes += skb->len; | |
b17fc996 | 134 | u64_stats_update_end(&s->syncp); |
bcbde4c0 | 135 | rcu_read_unlock(); |
b17fc996 | 136 | |
2a0751af | 137 | s = this_cpu_ptr(ipvs->tot_stats.cpustats); |
b17fc996 | 138 | u64_stats_update_begin(&s->syncp); |
cd67cd5e JA |
139 | s->cnt.inpkts++; |
140 | s->cnt.inbytes += skb->len; | |
b17fc996 | 141 | u64_stats_update_end(&s->syncp); |
1da177e4 LT |
142 | } |
143 | } | |
144 | ||
145 | ||
146 | static inline void | |
147 | ip_vs_out_stats(struct ip_vs_conn *cp, struct sk_buff *skb) | |
148 | { | |
149 | struct ip_vs_dest *dest = cp->dest; | |
7c08d78e | 150 | struct netns_ipvs *ipvs = cp->ipvs; |
b17fc996 | 151 | |
1da177e4 | 152 | if (dest && (dest->flags & IP_VS_DEST_F_AVAILABLE)) { |
b17fc996 | 153 | struct ip_vs_cpu_stats *s; |
bcbde4c0 | 154 | struct ip_vs_service *svc; |
b17fc996 HS |
155 | |
156 | s = this_cpu_ptr(dest->stats.cpustats); | |
b17fc996 | 157 | u64_stats_update_begin(&s->syncp); |
cd67cd5e JA |
158 | s->cnt.outpkts++; |
159 | s->cnt.outbytes += skb->len; | |
b17fc996 HS |
160 | u64_stats_update_end(&s->syncp); |
161 | ||
bcbde4c0 JA |
162 | rcu_read_lock(); |
163 | svc = rcu_dereference(dest->svc); | |
164 | s = this_cpu_ptr(svc->stats.cpustats); | |
b17fc996 | 165 | u64_stats_update_begin(&s->syncp); |
cd67cd5e JA |
166 | s->cnt.outpkts++; |
167 | s->cnt.outbytes += skb->len; | |
b17fc996 | 168 | u64_stats_update_end(&s->syncp); |
bcbde4c0 | 169 | rcu_read_unlock(); |
b17fc996 | 170 | |
2a0751af | 171 | s = this_cpu_ptr(ipvs->tot_stats.cpustats); |
b17fc996 | 172 | u64_stats_update_begin(&s->syncp); |
cd67cd5e JA |
173 | s->cnt.outpkts++; |
174 | s->cnt.outbytes += skb->len; | |
b17fc996 | 175 | u64_stats_update_end(&s->syncp); |
1da177e4 LT |
176 | } |
177 | } | |
178 | ||
179 | ||
180 | static inline void | |
181 | ip_vs_conn_stats(struct ip_vs_conn *cp, struct ip_vs_service *svc) | |
182 | { | |
3109d2f2 | 183 | struct netns_ipvs *ipvs = svc->ipvs; |
b17fc996 | 184 | struct ip_vs_cpu_stats *s; |
1da177e4 | 185 | |
b17fc996 | 186 | s = this_cpu_ptr(cp->dest->stats.cpustats); |
cd67cd5e JA |
187 | u64_stats_update_begin(&s->syncp); |
188 | s->cnt.conns++; | |
189 | u64_stats_update_end(&s->syncp); | |
1da177e4 | 190 | |
b17fc996 | 191 | s = this_cpu_ptr(svc->stats.cpustats); |
cd67cd5e JA |
192 | u64_stats_update_begin(&s->syncp); |
193 | s->cnt.conns++; | |
194 | u64_stats_update_end(&s->syncp); | |
b17fc996 | 195 | |
2a0751af | 196 | s = this_cpu_ptr(ipvs->tot_stats.cpustats); |
cd67cd5e JA |
197 | u64_stats_update_begin(&s->syncp); |
198 | s->cnt.conns++; | |
199 | u64_stats_update_end(&s->syncp); | |
1da177e4 LT |
200 | } |
201 | ||
202 | ||
4a516f11 | 203 | static inline void |
1da177e4 LT |
204 | ip_vs_set_state(struct ip_vs_conn *cp, int direction, |
205 | const struct sk_buff *skb, | |
9330419d | 206 | struct ip_vs_proto_data *pd) |
1da177e4 | 207 | { |
4a516f11 SH |
208 | if (likely(pd->pp->state_transition)) |
209 | pd->pp->state_transition(cp, direction, skb, pd); | |
1da177e4 LT |
210 | } |
211 | ||
a5959d53 | 212 | static inline int |
85999283 SH |
213 | ip_vs_conn_fill_param_persist(const struct ip_vs_service *svc, |
214 | struct sk_buff *skb, int protocol, | |
215 | const union nf_inet_addr *caddr, __be16 cport, | |
216 | const union nf_inet_addr *vaddr, __be16 vport, | |
217 | struct ip_vs_conn_param *p) | |
218 | { | |
3109d2f2 | 219 | ip_vs_conn_fill_param(svc->ipvs, svc->af, protocol, caddr, cport, vaddr, |
6e67e586 | 220 | vport, p); |
ceec4c38 | 221 | p->pe = rcu_dereference(svc->pe); |
85999283 | 222 | if (p->pe && p->pe->fill_param) |
a5959d53 HS |
223 | return p->pe->fill_param(p, skb); |
224 | ||
225 | return 0; | |
85999283 | 226 | } |
1da177e4 | 227 | |
1da177e4 LT |
228 | /* |
229 | * IPVS persistent scheduling function | |
230 | * It creates a connection entry according to its template if exists, | |
231 | * or selects a server and creates a connection entry plus a template. | |
232 | * Locking: we are svc user (svc->refcnt), so we hold all dests too | |
233 | * Protocols supported: TCP, UDP | |
234 | */ | |
235 | static struct ip_vs_conn * | |
236 | ip_vs_sched_persist(struct ip_vs_service *svc, | |
d4383f04 JDB |
237 | struct sk_buff *skb, __be16 src_port, __be16 dst_port, |
238 | int *ignored, struct ip_vs_iphdr *iph) | |
1da177e4 LT |
239 | { |
240 | struct ip_vs_conn *cp = NULL; | |
1da177e4 LT |
241 | struct ip_vs_dest *dest; |
242 | struct ip_vs_conn *ct; | |
5b57a98c | 243 | __be16 dport = 0; /* destination port to forward */ |
3575792e | 244 | unsigned int flags; |
f11017ec | 245 | struct ip_vs_conn_param param; |
e0aac52e | 246 | const union nf_inet_addr fwmark = { .ip = htonl(svc->fwmark) }; |
28364a59 JV |
247 | union nf_inet_addr snet; /* source network of the client, |
248 | after masking */ | |
ee78378f AG |
249 | const union nf_inet_addr *src_addr, *dst_addr; |
250 | ||
251 | if (likely(!ip_vs_iph_inverse(iph))) { | |
252 | src_addr = &iph->saddr; | |
253 | dst_addr = &iph->daddr; | |
254 | } else { | |
255 | src_addr = &iph->daddr; | |
256 | dst_addr = &iph->saddr; | |
257 | } | |
258 | ||
cd17f9ed | 259 | |
1da177e4 | 260 | /* Mask saddr with the netmask to adjust template granularity */ |
cd17f9ed JV |
261 | #ifdef CONFIG_IP_VS_IPV6 |
262 | if (svc->af == AF_INET6) | |
ee78378f | 263 | ipv6_addr_prefix(&snet.in6, &src_addr->in6, |
0a925864 | 264 | (__force __u32) svc->netmask); |
cd17f9ed JV |
265 | else |
266 | #endif | |
ee78378f | 267 | snet.ip = src_addr->ip & svc->netmask; |
1da177e4 | 268 | |
cd17f9ed JV |
269 | IP_VS_DBG_BUF(6, "p-schedule: src %s:%u dest %s:%u " |
270 | "mnet %s\n", | |
ee78378f AG |
271 | IP_VS_DBG_ADDR(svc->af, src_addr), ntohs(src_port), |
272 | IP_VS_DBG_ADDR(svc->af, dst_addr), ntohs(dst_port), | |
cd17f9ed | 273 | IP_VS_DBG_ADDR(svc->af, &snet)); |
1da177e4 LT |
274 | |
275 | /* | |
276 | * As far as we know, FTP is a very complicated network protocol, and | |
277 | * it uses control connection and data connections. For active FTP, | |
278 | * FTP server initialize data connection to the client, its source port | |
279 | * is often 20. For passive FTP, FTP server tells the clients the port | |
280 | * that it passively listens to, and the client issues the data | |
281 | * connection. In the tunneling or direct routing mode, the load | |
282 | * balancer is on the client-to-server half of connection, the port | |
283 | * number is unknown to the load balancer. So, a conn template like | |
284 | * <caddr, 0, vaddr, 0, daddr, 0> is created for persistent FTP | |
285 | * service, and a template like <caddr, 0, vaddr, vport, daddr, dport> | |
286 | * is created for other persistent services. | |
287 | */ | |
5b57a98c | 288 | { |
d4383f04 | 289 | int protocol = iph->protocol; |
ee78378f | 290 | const union nf_inet_addr *vaddr = dst_addr; |
f11017ec SH |
291 | __be16 vport = 0; |
292 | ||
ce144f24 | 293 | if (dst_port == svc->port) { |
5b57a98c SH |
294 | /* non-FTP template: |
295 | * <protocol, caddr, 0, vaddr, vport, daddr, dport> | |
296 | * FTP template: | |
297 | * <protocol, caddr, 0, vaddr, 0, daddr, 0> | |
1da177e4 LT |
298 | */ |
299 | if (svc->port != FTPPORT) | |
ce144f24 | 300 | vport = dst_port; |
1da177e4 | 301 | } else { |
5b57a98c SH |
302 | /* Note: persistent fwmark-based services and |
303 | * persistent port zero service are handled here. | |
304 | * fwmark template: | |
305 | * <IPPROTO_IP,caddr,0,fwmark,0,daddr,0> | |
306 | * port zero template: | |
307 | * <protocol,caddr,0,vaddr,0,daddr,0> | |
1da177e4 | 308 | */ |
28364a59 | 309 | if (svc->fwmark) { |
5b57a98c SH |
310 | protocol = IPPROTO_IP; |
311 | vaddr = &fwmark; | |
312 | } | |
1da177e4 | 313 | } |
a5959d53 HS |
314 | /* return *ignored = -1 so NF_DROP can be used */ |
315 | if (ip_vs_conn_fill_param_persist(svc, skb, protocol, &snet, 0, | |
316 | vaddr, vport, ¶m) < 0) { | |
317 | *ignored = -1; | |
318 | return NULL; | |
319 | } | |
1da177e4 LT |
320 | } |
321 | ||
5b57a98c | 322 | /* Check if a template already exists */ |
f11017ec | 323 | ct = ip_vs_ct_in_get(¶m); |
3ec10d3a | 324 | if (!ct || !ip_vs_check_template(ct, NULL)) { |
ceec4c38 JA |
325 | struct ip_vs_scheduler *sched; |
326 | ||
a5959d53 HS |
327 | /* |
328 | * No template found or the dest of the connection | |
5b57a98c | 329 | * template is not available. |
a5959d53 | 330 | * return *ignored=0 i.e. ICMP and NF_DROP |
5b57a98c | 331 | */ |
ceec4c38 | 332 | sched = rcu_dereference(svc->scheduler); |
05f00505 JA |
333 | if (sched) { |
334 | /* read svc->sched_data after svc->scheduler */ | |
335 | smp_rmb(); | |
336 | dest = sched->schedule(svc, skb, iph); | |
337 | } else { | |
338 | dest = NULL; | |
339 | } | |
5b57a98c SH |
340 | if (!dest) { |
341 | IP_VS_DBG(1, "p-schedule: no dest found.\n"); | |
85999283 | 342 | kfree(param.pe_data); |
a5959d53 | 343 | *ignored = 0; |
5b57a98c SH |
344 | return NULL; |
345 | } | |
346 | ||
ce144f24 | 347 | if (dst_port == svc->port && svc->port != FTPPORT) |
5b57a98c SH |
348 | dport = dest->port; |
349 | ||
85999283 SH |
350 | /* Create a template |
351 | * This adds param.pe_data to the template, | |
352 | * and thus param.pe_data will be destroyed | |
353 | * when the template expires */ | |
ba38528a | 354 | ct = ip_vs_conn_new(¶m, dest->af, &dest->addr, dport, |
0e051e68 | 355 | IP_VS_CONN_F_TEMPLATE, dest, skb->mark); |
85999283 SH |
356 | if (ct == NULL) { |
357 | kfree(param.pe_data); | |
a5959d53 | 358 | *ignored = -1; |
5b57a98c | 359 | return NULL; |
85999283 | 360 | } |
5b57a98c SH |
361 | |
362 | ct->timeout = svc->timeout; | |
85999283 | 363 | } else { |
5b57a98c SH |
364 | /* set destination with the found template */ |
365 | dest = ct->dest; | |
85999283 SH |
366 | kfree(param.pe_data); |
367 | } | |
5b57a98c | 368 | |
ce144f24 | 369 | dport = dst_port; |
5b57a98c SH |
370 | if (dport == svc->port && dest->port) |
371 | dport = dest->port; | |
372 | ||
26ec037f | 373 | flags = (svc->flags & IP_VS_SVC_F_ONEPACKET |
d4383f04 | 374 | && iph->protocol == IPPROTO_UDP) ? |
26ec037f NC |
375 | IP_VS_CONN_F_ONE_PACKET : 0; |
376 | ||
1da177e4 LT |
377 | /* |
378 | * Create a new connection according to the template | |
379 | */ | |
3109d2f2 | 380 | ip_vs_conn_fill_param(svc->ipvs, svc->af, iph->protocol, src_addr, |
ee78378f | 381 | src_port, dst_addr, dst_port, ¶m); |
ce144f24 | 382 | |
ba38528a AG |
383 | cp = ip_vs_conn_new(¶m, dest->af, &dest->addr, dport, flags, dest, |
384 | skb->mark); | |
1da177e4 LT |
385 | if (cp == NULL) { |
386 | ip_vs_conn_put(ct); | |
a5959d53 | 387 | *ignored = -1; |
1da177e4 LT |
388 | return NULL; |
389 | } | |
390 | ||
391 | /* | |
392 | * Add its control | |
393 | */ | |
394 | ip_vs_control_add(cp, ct); | |
395 | ip_vs_conn_put(ct); | |
396 | ||
397 | ip_vs_conn_stats(cp, svc); | |
398 | return cp; | |
399 | } | |
400 | ||
401 | ||
402 | /* | |
403 | * IPVS main scheduling function | |
404 | * It selects a server according to the virtual service, and | |
405 | * creates a connection entry. | |
406 | * Protocols supported: TCP, UDP | |
a5959d53 HS |
407 | * |
408 | * Usage of *ignored | |
409 | * | |
410 | * 1 : protocol tried to schedule (eg. on SYN), found svc but the | |
411 | * svc/scheduler decides that this packet should be accepted with | |
412 | * NF_ACCEPT because it must not be scheduled. | |
413 | * | |
414 | * 0 : scheduler can not find destination, so try bypass or | |
415 | * return ICMP and then NF_DROP (ip_vs_leave). | |
416 | * | |
417 | * -1 : scheduler tried to schedule but fatal error occurred, eg. | |
418 | * ip_vs_conn_new failure (ENOMEM) or ip_vs_sip_fill_param | |
419 | * failure such as missing Call-ID, ENOMEM on skb_linearize | |
420 | * or pe_data. In this case we should return NF_DROP without | |
421 | * any attempts to send ICMP with ip_vs_leave. | |
1da177e4 LT |
422 | */ |
423 | struct ip_vs_conn * | |
190ecd27 | 424 | ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb, |
d4383f04 JDB |
425 | struct ip_vs_proto_data *pd, int *ignored, |
426 | struct ip_vs_iphdr *iph) | |
1da177e4 | 427 | { |
9330419d | 428 | struct ip_vs_protocol *pp = pd->pp; |
1da177e4 | 429 | struct ip_vs_conn *cp = NULL; |
ceec4c38 | 430 | struct ip_vs_scheduler *sched; |
1da177e4 | 431 | struct ip_vs_dest *dest; |
ee78378f AG |
432 | __be16 _ports[2], *pptr, cport, vport; |
433 | const void *caddr, *vaddr; | |
3575792e | 434 | unsigned int flags; |
1da177e4 | 435 | |
190ecd27 | 436 | *ignored = 1; |
2f74713d JDB |
437 | /* |
438 | * IPv6 frags, only the first hit here. | |
439 | */ | |
d4383f04 | 440 | pptr = frag_safe_skb_hp(skb, iph->len, sizeof(_ports), _ports, iph); |
1da177e4 LT |
441 | if (pptr == NULL) |
442 | return NULL; | |
443 | ||
ee78378f AG |
444 | if (likely(!ip_vs_iph_inverse(iph))) { |
445 | cport = pptr[0]; | |
446 | caddr = &iph->saddr; | |
447 | vport = pptr[1]; | |
448 | vaddr = &iph->daddr; | |
449 | } else { | |
450 | cport = pptr[1]; | |
451 | caddr = &iph->daddr; | |
452 | vport = pptr[0]; | |
453 | vaddr = &iph->saddr; | |
454 | } | |
455 | ||
190ecd27 JA |
456 | /* |
457 | * FTPDATA needs this check when using local real server. | |
458 | * Never schedule Active FTPDATA connections from real server. | |
459 | * For LVS-NAT they must be already created. For other methods | |
460 | * with persistence the connection is created on SYN+ACK. | |
461 | */ | |
ee78378f | 462 | if (cport == FTPDATA) { |
b0e010c5 | 463 | IP_VS_DBG_PKT(12, svc->af, pp, skb, iph->off, |
0d79641a | 464 | "Not scheduling FTPDATA"); |
190ecd27 JA |
465 | return NULL; |
466 | } | |
467 | ||
468 | /* | |
a5959d53 | 469 | * Do not schedule replies from local real server. |
190ecd27 | 470 | */ |
802c41ad AG |
471 | if ((!skb->dev || skb->dev->flags & IFF_LOOPBACK)) { |
472 | iph->hdr_flags ^= IP_VS_HDR_INVERSE; | |
ab161976 | 473 | cp = pp->conn_in_get(svc->ipvs, svc->af, skb, iph); |
802c41ad AG |
474 | iph->hdr_flags ^= IP_VS_HDR_INVERSE; |
475 | ||
476 | if (cp) { | |
477 | IP_VS_DBG_PKT(12, svc->af, pp, skb, iph->off, | |
478 | "Not scheduling reply for existing" | |
479 | " connection"); | |
480 | __ip_vs_conn_put(cp); | |
481 | return NULL; | |
482 | } | |
190ecd27 JA |
483 | } |
484 | ||
1da177e4 LT |
485 | /* |
486 | * Persistent service | |
487 | */ | |
a5959d53 | 488 | if (svc->flags & IP_VS_SVC_F_PERSISTENT) |
ee78378f | 489 | return ip_vs_sched_persist(svc, skb, cport, vport, ignored, |
d4383f04 | 490 | iph); |
a5959d53 HS |
491 | |
492 | *ignored = 0; | |
1da177e4 LT |
493 | |
494 | /* | |
495 | * Non-persistent service | |
496 | */ | |
ee78378f | 497 | if (!svc->fwmark && vport != svc->port) { |
1da177e4 | 498 | if (!svc->port) |
1e3e238e HE |
499 | pr_err("Schedule: port zero only supported " |
500 | "in persistent services, " | |
501 | "check your ipvs configuration\n"); | |
1da177e4 LT |
502 | return NULL; |
503 | } | |
504 | ||
ceec4c38 | 505 | sched = rcu_dereference(svc->scheduler); |
05f00505 JA |
506 | if (sched) { |
507 | /* read svc->sched_data after svc->scheduler */ | |
508 | smp_rmb(); | |
509 | dest = sched->schedule(svc, skb, iph); | |
510 | } else { | |
511 | dest = NULL; | |
512 | } | |
1da177e4 LT |
513 | if (dest == NULL) { |
514 | IP_VS_DBG(1, "Schedule: no dest found.\n"); | |
515 | return NULL; | |
516 | } | |
517 | ||
26ec037f | 518 | flags = (svc->flags & IP_VS_SVC_F_ONEPACKET |
d4383f04 | 519 | && iph->protocol == IPPROTO_UDP) ? |
26ec037f NC |
520 | IP_VS_CONN_F_ONE_PACKET : 0; |
521 | ||
1da177e4 LT |
522 | /* |
523 | * Create a connection entry. | |
524 | */ | |
f11017ec SH |
525 | { |
526 | struct ip_vs_conn_param p; | |
6e67e586 | 527 | |
3109d2f2 | 528 | ip_vs_conn_fill_param(svc->ipvs, svc->af, iph->protocol, |
ee78378f | 529 | caddr, cport, vaddr, vport, &p); |
ba38528a | 530 | cp = ip_vs_conn_new(&p, dest->af, &dest->addr, |
ee78378f | 531 | dest->port ? dest->port : vport, |
0e051e68 | 532 | flags, dest, skb->mark); |
a5959d53 HS |
533 | if (!cp) { |
534 | *ignored = -1; | |
f11017ec | 535 | return NULL; |
a5959d53 | 536 | } |
f11017ec | 537 | } |
1da177e4 | 538 | |
cd17f9ed JV |
539 | IP_VS_DBG_BUF(6, "Schedule fwd:%c c:%s:%u v:%s:%u " |
540 | "d:%s:%u conn->flags:%X conn->refcnt:%d\n", | |
541 | ip_vs_fwd_tag(cp), | |
f18ae720 JA |
542 | IP_VS_DBG_ADDR(cp->af, &cp->caddr), ntohs(cp->cport), |
543 | IP_VS_DBG_ADDR(cp->af, &cp->vaddr), ntohs(cp->vport), | |
544 | IP_VS_DBG_ADDR(cp->daf, &cp->daddr), ntohs(cp->dport), | |
b54ab92b | 545 | cp->flags, refcount_read(&cp->refcnt)); |
1da177e4 LT |
546 | |
547 | ip_vs_conn_stats(cp, svc); | |
548 | return cp; | |
549 | } | |
550 | ||
0b729021 AG |
551 | static inline int ip_vs_addr_is_unicast(struct net *net, int af, |
552 | union nf_inet_addr *addr) | |
553 | { | |
554 | #ifdef CONFIG_IP_VS_IPV6 | |
555 | if (af == AF_INET6) | |
556 | return ipv6_addr_type(&addr->in6) & IPV6_ADDR_UNICAST; | |
557 | #endif | |
558 | return (inet_addr_type(net, addr->ip) == RTN_UNICAST); | |
559 | } | |
1da177e4 LT |
560 | |
561 | /* | |
562 | * Pass or drop the packet. | |
563 | * Called by ip_vs_in, when the virtual service is available but | |
564 | * no destination is available for a new connection. | |
565 | */ | |
566 | int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb, | |
d4383f04 | 567 | struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph) |
1da177e4 | 568 | { |
0b729021 | 569 | __be16 _ports[2], *pptr, dport; |
51efbcbb EB |
570 | struct netns_ipvs *ipvs = svc->ipvs; |
571 | struct net *net = ipvs->net; | |
9330419d | 572 | |
d4383f04 | 573 | pptr = frag_safe_skb_hp(skb, iph->len, sizeof(_ports), _ports, iph); |
0b729021 | 574 | if (!pptr) |
1da177e4 | 575 | return NF_DROP; |
0b729021 | 576 | dport = likely(!ip_vs_iph_inverse(iph)) ? pptr[1] : pptr[0]; |
a7a86b86 | 577 | |
1da177e4 | 578 | /* if it is fwmark-based service, the cache_bypass sysctl is up |
2a3b791e | 579 | and the destination is a non-local unicast, then create |
1da177e4 | 580 | a cache_bypass connection entry */ |
57032948 | 581 | if (sysctl_cache_bypass(ipvs) && svc->fwmark && |
0b729021 AG |
582 | !(iph->hdr_flags & (IP_VS_HDR_INVERSE | IP_VS_HDR_ICMP)) && |
583 | ip_vs_addr_is_unicast(net, svc->af, &iph->daddr)) { | |
ad542ced | 584 | int ret; |
1da177e4 | 585 | struct ip_vs_conn *cp; |
3575792e | 586 | unsigned int flags = (svc->flags & IP_VS_SVC_F_ONEPACKET && |
d4383f04 | 587 | iph->protocol == IPPROTO_UDP) ? |
3575792e | 588 | IP_VS_CONN_F_ONE_PACKET : 0; |
dff630dd | 589 | union nf_inet_addr daddr = { .all = { 0, 0, 0, 0 } }; |
1da177e4 | 590 | |
1da177e4 | 591 | /* create a new connection entry */ |
1e3e238e | 592 | IP_VS_DBG(6, "%s(): create a cache_bypass entry\n", __func__); |
f11017ec SH |
593 | { |
594 | struct ip_vs_conn_param p; | |
3109d2f2 | 595 | ip_vs_conn_fill_param(svc->ipvs, svc->af, iph->protocol, |
d4383f04 JDB |
596 | &iph->saddr, pptr[0], |
597 | &iph->daddr, pptr[1], &p); | |
ba38528a | 598 | cp = ip_vs_conn_new(&p, svc->af, &daddr, 0, |
f11017ec | 599 | IP_VS_CONN_F_BYPASS | flags, |
0e051e68 | 600 | NULL, skb->mark); |
f11017ec SH |
601 | if (!cp) |
602 | return NF_DROP; | |
603 | } | |
1da177e4 LT |
604 | |
605 | /* statistics */ | |
606 | ip_vs_in_stats(cp, skb); | |
607 | ||
608 | /* set state */ | |
4a516f11 | 609 | ip_vs_set_state(cp, IP_VS_DIR_INPUT, skb, pd); |
1da177e4 LT |
610 | |
611 | /* transmit the first SYN packet */ | |
d4383f04 | 612 | ret = cp->packet_xmit(skb, cp, pd->pp, iph); |
1da177e4 LT |
613 | /* do not touch skb anymore */ |
614 | ||
698e2a8d MA |
615 | if ((cp->flags & IP_VS_CONN_F_ONE_PACKET) && cp->control) |
616 | atomic_inc(&cp->control->in_pkts); | |
617 | else | |
618 | atomic_inc(&cp->in_pkts); | |
1da177e4 LT |
619 | ip_vs_conn_put(cp); |
620 | return ret; | |
621 | } | |
622 | ||
623 | /* | |
624 | * When the virtual ftp service is presented, packets destined | |
625 | * for other services on the VIP may get here (except services | |
626 | * listed in the ipvs table), pass the packets, because it is | |
627 | * not ipvs job to decide to drop the packets. | |
628 | */ | |
0b729021 | 629 | if (svc->port == FTPPORT && dport != FTPPORT) |
1da177e4 | 630 | return NF_ACCEPT; |
0b729021 AG |
631 | |
632 | if (unlikely(ip_vs_iph_icmp(iph))) | |
633 | return NF_DROP; | |
1da177e4 LT |
634 | |
635 | /* | |
636 | * Notify the client that the destination is unreachable, and | |
637 | * release the socket buffer. | |
638 | * Since it is in IP layer, the TCP socket is not actually | |
639 | * created, the TCP RST packet cannot be sent, instead that | |
640 | * ICMP_PORT_UNREACH is sent here no matter it is TCP/UDP. --WZ | |
641 | */ | |
2a3b791e | 642 | #ifdef CONFIG_IP_VS_IPV6 |
cb59155f | 643 | if (svc->af == AF_INET6) { |
57032948 EB |
644 | if (!skb->dev) |
645 | skb->dev = net->loopback_dev; | |
3ffe533c | 646 | icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); |
cb59155f | 647 | } else |
2a3b791e JV |
648 | #endif |
649 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); | |
650 | ||
1da177e4 LT |
651 | return NF_DROP; |
652 | } | |
653 | ||
84b3cee3 SH |
654 | #ifdef CONFIG_SYSCTL |
655 | ||
a43d1a6b | 656 | static int sysctl_snat_reroute(struct netns_ipvs *ipvs) |
84b3cee3 | 657 | { |
84b3cee3 SH |
658 | return ipvs->sysctl_snat_reroute; |
659 | } | |
660 | ||
2300f045 | 661 | static int sysctl_nat_icmp_send(struct netns_ipvs *ipvs) |
0cfa558e | 662 | { |
0cfa558e SH |
663 | return ipvs->sysctl_nat_icmp_send; |
664 | } | |
665 | ||
71a8ab6c SH |
666 | static int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs) |
667 | { | |
668 | return ipvs->sysctl_expire_nodest_conn; | |
669 | } | |
670 | ||
84b3cee3 SH |
671 | #else |
672 | ||
a43d1a6b | 673 | static int sysctl_snat_reroute(struct netns_ipvs *ipvs) { return 0; } |
2300f045 | 674 | static int sysctl_nat_icmp_send(struct netns_ipvs *ipvs) { return 0; } |
71a8ab6c | 675 | static int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs) { return 0; } |
84b3cee3 SH |
676 | |
677 | #endif | |
678 | ||
b1550f22 | 679 | __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset) |
1da177e4 | 680 | { |
d3bc23e7 | 681 | return csum_fold(skb_checksum(skb, offset, skb->len - offset, 0)); |
1da177e4 LT |
682 | } |
683 | ||
1ca5bb54 JA |
684 | static inline enum ip_defrag_users ip_vs_defrag_user(unsigned int hooknum) |
685 | { | |
686 | if (NF_INET_LOCAL_IN == hooknum) | |
687 | return IP_DEFRAG_VS_IN; | |
688 | if (NF_INET_FORWARD == hooknum) | |
689 | return IP_DEFRAG_VS_FWD; | |
690 | return IP_DEFRAG_VS_OUT; | |
691 | } | |
692 | ||
57781c1c EB |
693 | static inline int ip_vs_gather_frags(struct netns_ipvs *ipvs, |
694 | struct sk_buff *skb, u_int32_t user) | |
1da177e4 | 695 | { |
ac69269a | 696 | int err; |
776c729e | 697 | |
ac69269a | 698 | local_bh_disable(); |
19bcf9f2 | 699 | err = ip_defrag(ipvs->net, skb, user); |
ac69269a | 700 | local_bh_enable(); |
776c729e | 701 | if (!err) |
eddc9ec5 | 702 | ip_send_check(ip_hdr(skb)); |
776c729e HX |
703 | |
704 | return err; | |
1da177e4 LT |
705 | } |
706 | ||
a43d1a6b EB |
707 | static int ip_vs_route_me_harder(struct netns_ipvs *ipvs, int af, |
708 | struct sk_buff *skb, unsigned int hooknum) | |
ba4fd7e9 | 709 | { |
a43d1a6b | 710 | if (!sysctl_snat_reroute(ipvs)) |
579eb62a JA |
711 | return 0; |
712 | /* Reroute replies only to remote clients (FORWARD and LOCAL_OUT) */ | |
713 | if (NF_INET_LOCAL_IN == hooknum) | |
714 | return 0; | |
ba4fd7e9 SH |
715 | #ifdef CONFIG_IP_VS_IPV6 |
716 | if (af == AF_INET6) { | |
579eb62a JA |
717 | struct dst_entry *dst = skb_dst(skb); |
718 | ||
719 | if (dst->dev && !(dst->dev->flags & IFF_LOOPBACK) && | |
5f5d74d7 | 720 | ip6_route_me_harder(ipvs->net, skb) != 0) |
ba4fd7e9 SH |
721 | return 1; |
722 | } else | |
723 | #endif | |
579eb62a | 724 | if (!(skb_rtable(skb)->rt_flags & RTCF_LOCAL) && |
e45f5066 | 725 | ip_route_me_harder(ipvs->net, skb, RTN_LOCAL) != 0) |
ba4fd7e9 SH |
726 | return 1; |
727 | ||
728 | return 0; | |
729 | } | |
730 | ||
1da177e4 LT |
731 | /* |
732 | * Packet has been made sufficiently writable in caller | |
733 | * - inout: 1=in->out, 0=out->in | |
734 | */ | |
735 | void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp, | |
736 | struct ip_vs_conn *cp, int inout) | |
737 | { | |
eddc9ec5 | 738 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 | 739 | unsigned int icmp_offset = iph->ihl*4; |
d56f90a7 ACM |
740 | struct icmphdr *icmph = (struct icmphdr *)(skb_network_header(skb) + |
741 | icmp_offset); | |
1da177e4 LT |
742 | struct iphdr *ciph = (struct iphdr *)(icmph + 1); |
743 | ||
744 | if (inout) { | |
e7ade46a | 745 | iph->saddr = cp->vaddr.ip; |
1da177e4 | 746 | ip_send_check(iph); |
e7ade46a | 747 | ciph->daddr = cp->vaddr.ip; |
1da177e4 LT |
748 | ip_send_check(ciph); |
749 | } else { | |
e7ade46a | 750 | iph->daddr = cp->daddr.ip; |
1da177e4 | 751 | ip_send_check(iph); |
e7ade46a | 752 | ciph->saddr = cp->daddr.ip; |
1da177e4 LT |
753 | ip_send_check(ciph); |
754 | } | |
755 | ||
2906f66a VMR |
756 | /* the TCP/UDP/SCTP port */ |
757 | if (IPPROTO_TCP == ciph->protocol || IPPROTO_UDP == ciph->protocol || | |
758 | IPPROTO_SCTP == ciph->protocol) { | |
014d730d | 759 | __be16 *ports = (void *)ciph + ciph->ihl*4; |
1da177e4 LT |
760 | |
761 | if (inout) | |
762 | ports[1] = cp->vport; | |
763 | else | |
764 | ports[0] = cp->dport; | |
765 | } | |
766 | ||
767 | /* And finally the ICMP checksum */ | |
768 | icmph->checksum = 0; | |
769 | icmph->checksum = ip_vs_checksum_complete(skb, icmp_offset); | |
770 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
771 | ||
772 | if (inout) | |
0d79641a | 773 | IP_VS_DBG_PKT(11, AF_INET, pp, skb, (void *)ciph - (void *)iph, |
1da177e4 LT |
774 | "Forwarding altered outgoing ICMP"); |
775 | else | |
0d79641a | 776 | IP_VS_DBG_PKT(11, AF_INET, pp, skb, (void *)ciph - (void *)iph, |
1da177e4 LT |
777 | "Forwarding altered incoming ICMP"); |
778 | } | |
779 | ||
b3cdd2a7 JV |
780 | #ifdef CONFIG_IP_VS_IPV6 |
781 | void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp, | |
782 | struct ip_vs_conn *cp, int inout) | |
783 | { | |
784 | struct ipv6hdr *iph = ipv6_hdr(skb); | |
63dca2c0 JDB |
785 | unsigned int icmp_offset = 0; |
786 | unsigned int offs = 0; /* header offset*/ | |
787 | int protocol; | |
788 | struct icmp6hdr *icmph; | |
789 | struct ipv6hdr *ciph; | |
790 | unsigned short fragoffs; | |
791 | ||
792 | ipv6_find_hdr(skb, &icmp_offset, IPPROTO_ICMPV6, &fragoffs, NULL); | |
793 | icmph = (struct icmp6hdr *)(skb_network_header(skb) + icmp_offset); | |
794 | offs = icmp_offset + sizeof(struct icmp6hdr); | |
795 | ciph = (struct ipv6hdr *)(skb_network_header(skb) + offs); | |
796 | ||
797 | protocol = ipv6_find_hdr(skb, &offs, -1, &fragoffs, NULL); | |
b3cdd2a7 JV |
798 | |
799 | if (inout) { | |
800 | iph->saddr = cp->vaddr.in6; | |
801 | ciph->daddr = cp->vaddr.in6; | |
802 | } else { | |
803 | iph->daddr = cp->daddr.in6; | |
804 | ciph->saddr = cp->daddr.in6; | |
805 | } | |
806 | ||
2906f66a | 807 | /* the TCP/UDP/SCTP port */ |
63dca2c0 JDB |
808 | if (!fragoffs && (IPPROTO_TCP == protocol || IPPROTO_UDP == protocol || |
809 | IPPROTO_SCTP == protocol)) { | |
810 | __be16 *ports = (void *)(skb_network_header(skb) + offs); | |
b3cdd2a7 | 811 | |
63dca2c0 JDB |
812 | IP_VS_DBG(11, "%s() changed port %d to %d\n", __func__, |
813 | ntohs(inout ? ports[1] : ports[0]), | |
814 | ntohs(inout ? cp->vport : cp->dport)); | |
b3cdd2a7 JV |
815 | if (inout) |
816 | ports[1] = cp->vport; | |
817 | else | |
818 | ports[0] = cp->dport; | |
819 | } | |
820 | ||
821 | /* And finally the ICMP checksum */ | |
8870f842 SH |
822 | icmph->icmp6_cksum = ~csum_ipv6_magic(&iph->saddr, &iph->daddr, |
823 | skb->len - icmp_offset, | |
824 | IPPROTO_ICMPV6, 0); | |
825 | skb->csum_start = skb_network_header(skb) - skb->head + icmp_offset; | |
826 | skb->csum_offset = offsetof(struct icmp6hdr, icmp6_cksum); | |
827 | skb->ip_summed = CHECKSUM_PARTIAL; | |
b3cdd2a7 JV |
828 | |
829 | if (inout) | |
0d79641a JA |
830 | IP_VS_DBG_PKT(11, AF_INET6, pp, skb, |
831 | (void *)ciph - (void *)iph, | |
832 | "Forwarding altered outgoing ICMPv6"); | |
b3cdd2a7 | 833 | else |
0d79641a JA |
834 | IP_VS_DBG_PKT(11, AF_INET6, pp, skb, |
835 | (void *)ciph - (void *)iph, | |
836 | "Forwarding altered incoming ICMPv6"); | |
b3cdd2a7 JV |
837 | } |
838 | #endif | |
839 | ||
4856c84c | 840 | /* Handle relevant response ICMP messages - forward to the right |
6cb90db5 | 841 | * destination host. |
4856c84c | 842 | */ |
f2428ed5 SH |
843 | static int handle_response_icmp(int af, struct sk_buff *skb, |
844 | union nf_inet_addr *snet, | |
845 | __u8 protocol, struct ip_vs_conn *cp, | |
4856c84c | 846 | struct ip_vs_protocol *pp, |
579eb62a JA |
847 | unsigned int offset, unsigned int ihl, |
848 | unsigned int hooknum) | |
4856c84c MT |
849 | { |
850 | unsigned int verdict = NF_DROP; | |
851 | ||
852 | if (IP_VS_FWD_METHOD(cp) != 0) { | |
1e3e238e HE |
853 | pr_err("shouldn't reach here, because the box is on the " |
854 | "half connection in the tun/dr module.\n"); | |
4856c84c MT |
855 | } |
856 | ||
857 | /* Ensure the checksum is correct */ | |
858 | if (!skb_csum_unnecessary(skb) && ip_vs_checksum_complete(skb, ihl)) { | |
859 | /* Failed checksum! */ | |
f2428ed5 SH |
860 | IP_VS_DBG_BUF(1, "Forward ICMP: failed checksum from %s!\n", |
861 | IP_VS_DBG_ADDR(af, snet)); | |
4856c84c MT |
862 | goto out; |
863 | } | |
864 | ||
2906f66a VMR |
865 | if (IPPROTO_TCP == protocol || IPPROTO_UDP == protocol || |
866 | IPPROTO_SCTP == protocol) | |
4856c84c MT |
867 | offset += 2 * sizeof(__u16); |
868 | if (!skb_make_writable(skb, offset)) | |
869 | goto out; | |
870 | ||
f2428ed5 SH |
871 | #ifdef CONFIG_IP_VS_IPV6 |
872 | if (af == AF_INET6) | |
873 | ip_vs_nat_icmp_v6(skb, pp, cp, 1); | |
874 | else | |
875 | #endif | |
876 | ip_vs_nat_icmp(skb, pp, cp, 1); | |
4856c84c | 877 | |
a43d1a6b | 878 | if (ip_vs_route_me_harder(cp->ipvs, af, skb, hooknum)) |
ba4fd7e9 | 879 | goto out; |
f5a41847 | 880 | |
4856c84c MT |
881 | /* do the statistics and put it back */ |
882 | ip_vs_out_stats(cp, skb); | |
883 | ||
cf356d69 | 884 | skb->ipvs_property = 1; |
f4bc17cd | 885 | if (!(cp->flags & IP_VS_CONN_F_NFCT)) |
cf356d69 | 886 | ip_vs_notrack(skb); |
f4bc17cd JA |
887 | else |
888 | ip_vs_update_conntrack(skb, cp, 0); | |
4856c84c MT |
889 | verdict = NF_ACCEPT; |
890 | ||
891 | out: | |
892 | __ip_vs_conn_put(cp); | |
893 | ||
894 | return verdict; | |
895 | } | |
896 | ||
1da177e4 LT |
897 | /* |
898 | * Handle ICMP messages in the inside-to-outside direction (outgoing). | |
4856c84c | 899 | * Find any that might be relevant, check against existing connections. |
1da177e4 | 900 | * Currently handles error types - unreachable, quench, ttl exceeded. |
1da177e4 | 901 | */ |
7b5f689a EB |
902 | static int ip_vs_out_icmp(struct netns_ipvs *ipvs, struct sk_buff *skb, |
903 | int *related, unsigned int hooknum) | |
1da177e4 | 904 | { |
1da177e4 LT |
905 | struct iphdr *iph; |
906 | struct icmphdr _icmph, *ic; | |
907 | struct iphdr _ciph, *cih; /* The ip header contained within the ICMP */ | |
51ef348b | 908 | struct ip_vs_iphdr ciph; |
1da177e4 LT |
909 | struct ip_vs_conn *cp; |
910 | struct ip_vs_protocol *pp; | |
4856c84c | 911 | unsigned int offset, ihl; |
f2428ed5 | 912 | union nf_inet_addr snet; |
1da177e4 LT |
913 | |
914 | *related = 1; | |
915 | ||
916 | /* reassemble IP fragments */ | |
56f8a75c | 917 | if (ip_is_fragment(ip_hdr(skb))) { |
57781c1c | 918 | if (ip_vs_gather_frags(ipvs, skb, ip_vs_defrag_user(hooknum))) |
1da177e4 | 919 | return NF_STOLEN; |
1da177e4 LT |
920 | } |
921 | ||
eddc9ec5 | 922 | iph = ip_hdr(skb); |
1da177e4 LT |
923 | offset = ihl = iph->ihl * 4; |
924 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
925 | if (ic == NULL) | |
926 | return NF_DROP; | |
927 | ||
14d5e834 | 928 | IP_VS_DBG(12, "Outgoing ICMP (%d,%d) %pI4->%pI4\n", |
1da177e4 | 929 | ic->type, ntohs(icmp_id(ic)), |
14d5e834 | 930 | &iph->saddr, &iph->daddr); |
1da177e4 LT |
931 | |
932 | /* | |
933 | * Work through seeing if this is for us. | |
934 | * These checks are supposed to be in an order that means easy | |
935 | * things are checked first to speed up processing.... however | |
936 | * this means that some packets will manage to get a long way | |
937 | * down this stack and then be rejected, but that's life. | |
938 | */ | |
939 | if ((ic->type != ICMP_DEST_UNREACH) && | |
940 | (ic->type != ICMP_SOURCE_QUENCH) && | |
941 | (ic->type != ICMP_TIME_EXCEEDED)) { | |
942 | *related = 0; | |
943 | return NF_ACCEPT; | |
944 | } | |
945 | ||
946 | /* Now find the contained IP header */ | |
947 | offset += sizeof(_icmph); | |
948 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
949 | if (cih == NULL) | |
950 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
951 | ||
952 | pp = ip_vs_proto_get(cih->protocol); | |
953 | if (!pp) | |
954 | return NF_ACCEPT; | |
955 | ||
956 | /* Is the embedded protocol header present? */ | |
4412ec49 | 957 | if (unlikely(cih->frag_off & htons(IP_OFFSET) && |
1da177e4 LT |
958 | pp->dont_defrag)) |
959 | return NF_ACCEPT; | |
960 | ||
0d79641a JA |
961 | IP_VS_DBG_PKT(11, AF_INET, pp, skb, offset, |
962 | "Checking outgoing ICMP for"); | |
1da177e4 | 963 | |
4fd9beef | 964 | ip_vs_fill_iph_skb_icmp(AF_INET, skb, offset, true, &ciph); |
b0e010c5 | 965 | |
1da177e4 | 966 | /* The embedded headers contain source and dest in reverse order */ |
0cf705c8 | 967 | cp = pp->conn_out_get(ipvs, AF_INET, skb, &ciph); |
1da177e4 LT |
968 | if (!cp) |
969 | return NF_ACCEPT; | |
970 | ||
f2428ed5 SH |
971 | snet.ip = iph->saddr; |
972 | return handle_response_icmp(AF_INET, skb, &snet, cih->protocol, cp, | |
579eb62a | 973 | pp, ciph.len, ihl, hooknum); |
1da177e4 LT |
974 | } |
975 | ||
2a3b791e | 976 | #ifdef CONFIG_IP_VS_IPV6 |
7b5f689a EB |
977 | static int ip_vs_out_icmp_v6(struct netns_ipvs *ipvs, struct sk_buff *skb, |
978 | int *related, unsigned int hooknum, | |
979 | struct ip_vs_iphdr *ipvsh) | |
2a3b791e | 980 | { |
2a3b791e | 981 | struct icmp6hdr _icmph, *ic; |
63dca2c0 | 982 | struct ip_vs_iphdr ciph = {.flags = 0, .fragoffs = 0};/*Contained IP */ |
2a3b791e JV |
983 | struct ip_vs_conn *cp; |
984 | struct ip_vs_protocol *pp; | |
f2428ed5 | 985 | union nf_inet_addr snet; |
b0e010c5 | 986 | unsigned int offset; |
2a3b791e | 987 | |
63dca2c0 | 988 | *related = 1; |
2f74713d | 989 | ic = frag_safe_skb_hp(skb, ipvsh->len, sizeof(_icmph), &_icmph, ipvsh); |
2a3b791e JV |
990 | if (ic == NULL) |
991 | return NF_DROP; | |
992 | ||
2a3b791e JV |
993 | /* |
994 | * Work through seeing if this is for us. | |
995 | * These checks are supposed to be in an order that means easy | |
996 | * things are checked first to speed up processing.... however | |
997 | * this means that some packets will manage to get a long way | |
998 | * down this stack and then be rejected, but that's life. | |
999 | */ | |
2fab8917 | 1000 | if (ic->icmp6_type & ICMPV6_INFOMSG_MASK) { |
2a3b791e JV |
1001 | *related = 0; |
1002 | return NF_ACCEPT; | |
1003 | } | |
2f74713d JDB |
1004 | /* Fragment header that is before ICMP header tells us that: |
1005 | * it's not an error message since they can't be fragmented. | |
1006 | */ | |
e7165030 | 1007 | if (ipvsh->flags & IP6_FH_F_FRAG) |
2f74713d | 1008 | return NF_DROP; |
2a3b791e | 1009 | |
63dca2c0 JDB |
1010 | IP_VS_DBG(8, "Outgoing ICMPv6 (%d,%d) %pI6c->%pI6c\n", |
1011 | ic->icmp6_type, ntohs(icmpv6_id(ic)), | |
1012 | &ipvsh->saddr, &ipvsh->daddr); | |
1013 | ||
4fd9beef AG |
1014 | if (!ip_vs_fill_iph_skb_icmp(AF_INET6, skb, ipvsh->len + sizeof(_icmph), |
1015 | true, &ciph)) | |
2a3b791e | 1016 | return NF_ACCEPT; /* The packet looks wrong, ignore */ |
63dca2c0 JDB |
1017 | |
1018 | pp = ip_vs_proto_get(ciph.protocol); | |
2a3b791e JV |
1019 | if (!pp) |
1020 | return NF_ACCEPT; | |
1021 | ||
2a3b791e | 1022 | /* The embedded headers contain source and dest in reverse order */ |
0cf705c8 | 1023 | cp = pp->conn_out_get(ipvs, AF_INET6, skb, &ciph); |
2a3b791e JV |
1024 | if (!cp) |
1025 | return NF_ACCEPT; | |
1026 | ||
63dca2c0 | 1027 | snet.in6 = ciph.saddr.in6; |
b0e010c5 | 1028 | offset = ciph.len; |
63dca2c0 | 1029 | return handle_response_icmp(AF_INET6, skb, &snet, ciph.protocol, cp, |
b0e010c5 | 1030 | pp, offset, sizeof(struct ipv6hdr), |
579eb62a | 1031 | hooknum); |
2a3b791e JV |
1032 | } |
1033 | #endif | |
1034 | ||
2906f66a VMR |
1035 | /* |
1036 | * Check if sctp chunc is ABORT chunk | |
1037 | */ | |
1038 | static inline int is_sctp_abort(const struct sk_buff *skb, int nh_len) | |
1039 | { | |
1040 | sctp_chunkhdr_t *sch, schunk; | |
1041 | sch = skb_header_pointer(skb, nh_len + sizeof(sctp_sctphdr_t), | |
1042 | sizeof(schunk), &schunk); | |
1043 | if (sch == NULL) | |
1044 | return 0; | |
1045 | if (sch->type == SCTP_CID_ABORT) | |
1046 | return 1; | |
1047 | return 0; | |
1048 | } | |
1049 | ||
2a3b791e | 1050 | static inline int is_tcp_reset(const struct sk_buff *skb, int nh_len) |
1da177e4 LT |
1051 | { |
1052 | struct tcphdr _tcph, *th; | |
1053 | ||
2a3b791e | 1054 | th = skb_header_pointer(skb, nh_len, sizeof(_tcph), &_tcph); |
1da177e4 LT |
1055 | if (th == NULL) |
1056 | return 0; | |
1057 | return th->rst; | |
1058 | } | |
1059 | ||
dc7b3eb9 GL |
1060 | static inline bool is_new_conn(const struct sk_buff *skb, |
1061 | struct ip_vs_iphdr *iph) | |
1062 | { | |
1063 | switch (iph->protocol) { | |
1064 | case IPPROTO_TCP: { | |
1065 | struct tcphdr _tcph, *th; | |
1066 | ||
1067 | th = skb_header_pointer(skb, iph->len, sizeof(_tcph), &_tcph); | |
1068 | if (th == NULL) | |
1069 | return false; | |
1070 | return th->syn; | |
1071 | } | |
1072 | case IPPROTO_SCTP: { | |
1073 | sctp_chunkhdr_t *sch, schunk; | |
1074 | ||
1075 | sch = skb_header_pointer(skb, iph->len + sizeof(sctp_sctphdr_t), | |
1076 | sizeof(schunk), &schunk); | |
1077 | if (sch == NULL) | |
1078 | return false; | |
1079 | return sch->type == SCTP_CID_INIT; | |
1080 | } | |
1081 | default: | |
1082 | return false; | |
1083 | } | |
1084 | } | |
1085 | ||
d752c364 MRL |
1086 | static inline bool is_new_conn_expected(const struct ip_vs_conn *cp, |
1087 | int conn_reuse_mode) | |
1088 | { | |
1089 | /* Controlled (FTP DATA or persistence)? */ | |
1090 | if (cp->control) | |
1091 | return false; | |
1092 | ||
1093 | switch (cp->protocol) { | |
1094 | case IPPROTO_TCP: | |
1095 | return (cp->state == IP_VS_TCP_S_TIME_WAIT) || | |
f911b675 | 1096 | (cp->state == IP_VS_TCP_S_CLOSE) || |
d752c364 MRL |
1097 | ((conn_reuse_mode & 2) && |
1098 | (cp->state == IP_VS_TCP_S_FIN_WAIT) && | |
1099 | (cp->flags & IP_VS_CONN_F_NOOUTPUT)); | |
1100 | case IPPROTO_SCTP: | |
1101 | return cp->state == IP_VS_SCTP_S_CLOSED; | |
1102 | default: | |
1103 | return false; | |
1104 | } | |
1105 | } | |
1106 | ||
39b97223 MA |
1107 | /* Generic function to create new connections for outgoing RS packets |
1108 | * | |
1109 | * Pre-requisites for successful connection creation: | |
1110 | * 1) Virtual Service is NOT fwmark based: | |
1111 | * In fwmark-VS actual vaddr and vport are unknown to IPVS | |
1112 | * 2) Real Server and Virtual Service were NOT configured without port: | |
1113 | * This is to allow match of different VS to the same RS ip-addr | |
1114 | */ | |
1115 | struct ip_vs_conn *ip_vs_new_conn_out(struct ip_vs_service *svc, | |
1116 | struct ip_vs_dest *dest, | |
1117 | struct sk_buff *skb, | |
1118 | const struct ip_vs_iphdr *iph, | |
1119 | __be16 dport, | |
1120 | __be16 cport) | |
1121 | { | |
1122 | struct ip_vs_conn_param param; | |
1123 | struct ip_vs_conn *ct = NULL, *cp = NULL; | |
1124 | const union nf_inet_addr *vaddr, *daddr, *caddr; | |
1125 | union nf_inet_addr snet; | |
1126 | __be16 vport; | |
1127 | unsigned int flags; | |
1128 | ||
1129 | EnterFunction(12); | |
1130 | vaddr = &svc->addr; | |
1131 | vport = svc->port; | |
1132 | daddr = &iph->saddr; | |
1133 | caddr = &iph->daddr; | |
1134 | ||
1135 | /* check pre-requisites are satisfied */ | |
1136 | if (svc->fwmark) | |
1137 | return NULL; | |
1138 | if (!vport || !dport) | |
1139 | return NULL; | |
1140 | ||
1141 | /* for persistent service first create connection template */ | |
1142 | if (svc->flags & IP_VS_SVC_F_PERSISTENT) { | |
1143 | /* apply netmask the same way ingress-side does */ | |
1144 | #ifdef CONFIG_IP_VS_IPV6 | |
1145 | if (svc->af == AF_INET6) | |
1146 | ipv6_addr_prefix(&snet.in6, &caddr->in6, | |
1147 | (__force __u32)svc->netmask); | |
1148 | else | |
1149 | #endif | |
1150 | snet.ip = caddr->ip & svc->netmask; | |
1151 | /* fill params and create template if not existent */ | |
1152 | if (ip_vs_conn_fill_param_persist(svc, skb, iph->protocol, | |
1153 | &snet, 0, vaddr, | |
1154 | vport, ¶m) < 0) | |
1155 | return NULL; | |
1156 | ct = ip_vs_ct_in_get(¶m); | |
3ec10d3a MA |
1157 | /* check if template exists and points to the same dest */ |
1158 | if (!ct || !ip_vs_check_template(ct, dest)) { | |
39b97223 MA |
1159 | ct = ip_vs_conn_new(¶m, dest->af, daddr, dport, |
1160 | IP_VS_CONN_F_TEMPLATE, dest, 0); | |
1161 | if (!ct) { | |
1162 | kfree(param.pe_data); | |
1163 | return NULL; | |
1164 | } | |
1165 | ct->timeout = svc->timeout; | |
1166 | } else { | |
1167 | kfree(param.pe_data); | |
1168 | } | |
1169 | } | |
1170 | ||
1171 | /* connection flags */ | |
1172 | flags = ((svc->flags & IP_VS_SVC_F_ONEPACKET) && | |
1173 | iph->protocol == IPPROTO_UDP) ? IP_VS_CONN_F_ONE_PACKET : 0; | |
1174 | /* create connection */ | |
1175 | ip_vs_conn_fill_param(svc->ipvs, svc->af, iph->protocol, | |
1176 | caddr, cport, vaddr, vport, ¶m); | |
1177 | cp = ip_vs_conn_new(¶m, dest->af, daddr, dport, flags, dest, 0); | |
1178 | if (!cp) { | |
1179 | if (ct) | |
1180 | ip_vs_conn_put(ct); | |
1181 | return NULL; | |
1182 | } | |
1183 | if (ct) { | |
1184 | ip_vs_control_add(cp, ct); | |
1185 | ip_vs_conn_put(ct); | |
1186 | } | |
1187 | ip_vs_conn_stats(cp, svc); | |
1188 | ||
1189 | /* return connection (will be used to handle outgoing packet) */ | |
1190 | IP_VS_DBG_BUF(6, "New connection RS-initiated:%c c:%s:%u v:%s:%u " | |
1191 | "d:%s:%u conn->flags:%X conn->refcnt:%d\n", | |
1192 | ip_vs_fwd_tag(cp), | |
1193 | IP_VS_DBG_ADDR(cp->af, &cp->caddr), ntohs(cp->cport), | |
1194 | IP_VS_DBG_ADDR(cp->af, &cp->vaddr), ntohs(cp->vport), | |
1195 | IP_VS_DBG_ADDR(cp->af, &cp->daddr), ntohs(cp->dport), | |
b54ab92b | 1196 | cp->flags, refcount_read(&cp->refcnt)); |
39b97223 MA |
1197 | LeaveFunction(12); |
1198 | return cp; | |
1199 | } | |
1200 | ||
1201 | /* Handle outgoing packets which are considered requests initiated by | |
1202 | * real servers, so that subsequent responses from external client can be | |
1203 | * routed to the right real server. | |
1204 | * Used also for outgoing responses in OPS mode. | |
1205 | * | |
1206 | * Connection management is handled by persistent-engine specific callback. | |
1207 | */ | |
1208 | static struct ip_vs_conn *__ip_vs_rs_conn_out(unsigned int hooknum, | |
1209 | struct netns_ipvs *ipvs, | |
1210 | int af, struct sk_buff *skb, | |
1211 | const struct ip_vs_iphdr *iph) | |
1212 | { | |
1213 | struct ip_vs_dest *dest; | |
1214 | struct ip_vs_conn *cp = NULL; | |
1215 | __be16 _ports[2], *pptr; | |
1216 | ||
1217 | if (hooknum == NF_INET_LOCAL_IN) | |
1218 | return NULL; | |
1219 | ||
1220 | pptr = frag_safe_skb_hp(skb, iph->len, | |
1221 | sizeof(_ports), _ports, iph); | |
1222 | if (!pptr) | |
1223 | return NULL; | |
1224 | ||
1225 | rcu_read_lock(); | |
1226 | dest = ip_vs_find_real_service(ipvs, af, iph->protocol, | |
1227 | &iph->saddr, pptr[0]); | |
1228 | if (dest) { | |
1229 | struct ip_vs_service *svc; | |
1230 | struct ip_vs_pe *pe; | |
1231 | ||
1232 | svc = rcu_dereference(dest->svc); | |
1233 | if (svc) { | |
1234 | pe = rcu_dereference(svc->pe); | |
1235 | if (pe && pe->conn_out) | |
1236 | cp = pe->conn_out(svc, dest, skb, iph, | |
1237 | pptr[0], pptr[1]); | |
1238 | } | |
1239 | } | |
1240 | rcu_read_unlock(); | |
1241 | ||
1242 | return cp; | |
1243 | } | |
1244 | ||
4856c84c | 1245 | /* Handle response packets: rewrite addresses and send away... |
4856c84c MT |
1246 | */ |
1247 | static unsigned int | |
9330419d | 1248 | handle_response(int af, struct sk_buff *skb, struct ip_vs_proto_data *pd, |
579eb62a JA |
1249 | struct ip_vs_conn *cp, struct ip_vs_iphdr *iph, |
1250 | unsigned int hooknum) | |
4856c84c | 1251 | { |
9330419d HS |
1252 | struct ip_vs_protocol *pp = pd->pp; |
1253 | ||
b0e010c5 | 1254 | IP_VS_DBG_PKT(11, af, pp, skb, iph->off, "Outgoing packet"); |
4856c84c | 1255 | |
d4383f04 | 1256 | if (!skb_make_writable(skb, iph->len)) |
4856c84c MT |
1257 | goto drop; |
1258 | ||
1259 | /* mangle the packet */ | |
d4383f04 | 1260 | if (pp->snat_handler && !pp->snat_handler(skb, pp, cp, iph)) |
4856c84c MT |
1261 | goto drop; |
1262 | ||
1263 | #ifdef CONFIG_IP_VS_IPV6 | |
1264 | if (af == AF_INET6) | |
1265 | ipv6_hdr(skb)->saddr = cp->vaddr.in6; | |
1266 | else | |
1267 | #endif | |
1268 | { | |
1269 | ip_hdr(skb)->saddr = cp->vaddr.ip; | |
1270 | ip_send_check(ip_hdr(skb)); | |
1271 | } | |
1272 | ||
8a803040 JA |
1273 | /* |
1274 | * nf_iterate does not expect change in the skb->dst->dev. | |
1275 | * It looks like it is not fatal to enable this code for hooks | |
1276 | * where our handlers are at the end of the chain list and | |
1277 | * when all next handlers use skb->dst->dev and not outdev. | |
1278 | * It will definitely route properly the inout NAT traffic | |
1279 | * when multiple paths are used. | |
1280 | */ | |
1281 | ||
4856c84c MT |
1282 | /* For policy routing, packets originating from this |
1283 | * machine itself may be routed differently to packets | |
1284 | * passing through. We want this packet to be routed as | |
1285 | * if it came from this machine itself. So re-compute | |
1286 | * the routing information. | |
1287 | */ | |
a43d1a6b | 1288 | if (ip_vs_route_me_harder(cp->ipvs, af, skb, hooknum)) |
ba4fd7e9 | 1289 | goto drop; |
4856c84c | 1290 | |
b0e010c5 | 1291 | IP_VS_DBG_PKT(10, af, pp, skb, iph->off, "After SNAT"); |
4856c84c MT |
1292 | |
1293 | ip_vs_out_stats(cp, skb); | |
9330419d | 1294 | ip_vs_set_state(cp, IP_VS_DIR_OUTPUT, skb, pd); |
cf356d69 | 1295 | skb->ipvs_property = 1; |
f4bc17cd | 1296 | if (!(cp->flags & IP_VS_CONN_F_NFCT)) |
cf356d69 | 1297 | ip_vs_notrack(skb); |
f4bc17cd JA |
1298 | else |
1299 | ip_vs_update_conntrack(skb, cp, 0); | |
4856c84c MT |
1300 | ip_vs_conn_put(cp); |
1301 | ||
4856c84c MT |
1302 | LeaveFunction(11); |
1303 | return NF_ACCEPT; | |
1304 | ||
1305 | drop: | |
1306 | ip_vs_conn_put(cp); | |
1307 | kfree_skb(skb); | |
f4bc17cd | 1308 | LeaveFunction(11); |
4856c84c MT |
1309 | return NF_STOLEN; |
1310 | } | |
1311 | ||
1da177e4 | 1312 | /* |
4856c84c | 1313 | * Check if outgoing packet belongs to the established ip_vs_conn. |
1da177e4 LT |
1314 | */ |
1315 | static unsigned int | |
1b75097d | 1316 | ip_vs_out(struct netns_ipvs *ipvs, unsigned int hooknum, struct sk_buff *skb, int af) |
1da177e4 | 1317 | { |
51ef348b | 1318 | struct ip_vs_iphdr iph; |
1da177e4 | 1319 | struct ip_vs_protocol *pp; |
9330419d | 1320 | struct ip_vs_proto_data *pd; |
1da177e4 | 1321 | struct ip_vs_conn *cp; |
340c78e5 | 1322 | struct sock *sk; |
1da177e4 LT |
1323 | |
1324 | EnterFunction(11); | |
1325 | ||
fc604767 | 1326 | /* Already marked as IPVS request or reply? */ |
6869c4d8 | 1327 | if (skb->ipvs_property) |
1da177e4 LT |
1328 | return NF_ACCEPT; |
1329 | ||
340c78e5 | 1330 | sk = skb_to_full_sk(skb); |
fc604767 | 1331 | /* Bad... Do not break raw sockets */ |
340c78e5 | 1332 | if (unlikely(sk && hooknum == NF_INET_LOCAL_OUT && |
fc604767 | 1333 | af == AF_INET)) { |
fc604767 | 1334 | |
340c78e5 | 1335 | if (sk->sk_family == PF_INET && inet_sk(sk)->nodefrag) |
fc604767 JA |
1336 | return NF_ACCEPT; |
1337 | } | |
1338 | ||
1339 | if (unlikely(!skb_dst(skb))) | |
1340 | return NF_ACCEPT; | |
1341 | ||
48aed1b0 | 1342 | if (!ipvs->enable) |
7a4f0761 HS |
1343 | return NF_ACCEPT; |
1344 | ||
4fd9beef | 1345 | ip_vs_fill_iph_skb(af, skb, false, &iph); |
2a3b791e JV |
1346 | #ifdef CONFIG_IP_VS_IPV6 |
1347 | if (af == AF_INET6) { | |
1348 | if (unlikely(iph.protocol == IPPROTO_ICMPV6)) { | |
1ca5bb54 | 1349 | int related; |
7b5f689a | 1350 | int verdict = ip_vs_out_icmp_v6(ipvs, skb, &related, |
d4383f04 | 1351 | hooknum, &iph); |
1da177e4 | 1352 | |
f5a41847 | 1353 | if (related) |
2a3b791e | 1354 | return verdict; |
2a3b791e JV |
1355 | } |
1356 | } else | |
1357 | #endif | |
1358 | if (unlikely(iph.protocol == IPPROTO_ICMP)) { | |
1ca5bb54 | 1359 | int related; |
7b5f689a | 1360 | int verdict = ip_vs_out_icmp(ipvs, skb, &related, hooknum); |
2a3b791e | 1361 | |
f5a41847 | 1362 | if (related) |
2a3b791e | 1363 | return verdict; |
2a3b791e | 1364 | } |
1da177e4 | 1365 | |
18d6ade6 | 1366 | pd = ip_vs_proto_data_get(ipvs, iph.protocol); |
9330419d | 1367 | if (unlikely(!pd)) |
1da177e4 | 1368 | return NF_ACCEPT; |
9330419d | 1369 | pp = pd->pp; |
1da177e4 LT |
1370 | |
1371 | /* reassemble IP fragments */ | |
2a3b791e | 1372 | #ifdef CONFIG_IP_VS_IPV6 |
63dca2c0 | 1373 | if (af == AF_INET) |
2a3b791e | 1374 | #endif |
56f8a75c | 1375 | if (unlikely(ip_is_fragment(ip_hdr(skb)) && !pp->dont_defrag)) { |
57781c1c | 1376 | if (ip_vs_gather_frags(ipvs, skb, |
1ca5bb54 | 1377 | ip_vs_defrag_user(hooknum))) |
2a3b791e JV |
1378 | return NF_STOLEN; |
1379 | ||
4fd9beef | 1380 | ip_vs_fill_iph_skb(AF_INET, skb, false, &iph); |
2a3b791e | 1381 | } |
1da177e4 LT |
1382 | |
1383 | /* | |
1384 | * Check if the packet belongs to an existing entry | |
1385 | */ | |
0cf705c8 | 1386 | cp = pp->conn_out_get(ipvs, af, skb, &iph); |
1da177e4 | 1387 | |
cb59155f | 1388 | if (likely(cp)) |
579eb62a | 1389 | return handle_response(af, skb, pd, cp, &iph, hooknum); |
39b97223 MA |
1390 | |
1391 | /* Check for real-server-started requests */ | |
1392 | if (atomic_read(&ipvs->conn_out_counter)) { | |
1393 | /* Currently only for UDP: | |
1394 | * connection oriented protocols typically use | |
1395 | * ephemeral ports for outgoing connections, so | |
1396 | * related incoming responses would not match any VS | |
1397 | */ | |
1398 | if (pp->protocol == IPPROTO_UDP) { | |
1399 | cp = __ip_vs_rs_conn_out(hooknum, ipvs, af, skb, &iph); | |
1400 | if (likely(cp)) | |
1401 | return handle_response(af, skb, pd, cp, &iph, | |
1402 | hooknum); | |
1403 | } | |
1404 | } | |
1405 | ||
2300f045 | 1406 | if (sysctl_nat_icmp_send(ipvs) && |
cb59155f JA |
1407 | (pp->protocol == IPPROTO_TCP || |
1408 | pp->protocol == IPPROTO_UDP || | |
1409 | pp->protocol == IPPROTO_SCTP)) { | |
1410 | __be16 _ports[2], *pptr; | |
1411 | ||
2f74713d JDB |
1412 | pptr = frag_safe_skb_hp(skb, iph.len, |
1413 | sizeof(_ports), _ports, &iph); | |
cb59155f JA |
1414 | if (pptr == NULL) |
1415 | return NF_ACCEPT; /* Not for me */ | |
48aed1b0 | 1416 | if (ip_vs_has_real_service(ipvs, af, iph.protocol, &iph.saddr, |
276472ea | 1417 | pptr[0])) { |
cb59155f JA |
1418 | /* |
1419 | * Notify the real server: there is no | |
1420 | * existing entry if it is not RST | |
1421 | * packet or not TCP packet. | |
1422 | */ | |
1423 | if ((iph.protocol != IPPROTO_TCP && | |
1424 | iph.protocol != IPPROTO_SCTP) | |
1425 | || ((iph.protocol == IPPROTO_TCP | |
1426 | && !is_tcp_reset(skb, iph.len)) | |
1427 | || (iph.protocol == IPPROTO_SCTP | |
1428 | && !is_sctp_abort(skb, | |
1429 | iph.len)))) { | |
2a3b791e | 1430 | #ifdef CONFIG_IP_VS_IPV6 |
cb59155f | 1431 | if (af == AF_INET6) { |
cb59155f | 1432 | if (!skb->dev) |
ed1c9f0e | 1433 | skb->dev = ipvs->net->loopback_dev; |
cb59155f JA |
1434 | icmpv6_send(skb, |
1435 | ICMPV6_DEST_UNREACH, | |
1436 | ICMPV6_PORT_UNREACH, | |
1437 | 0); | |
1438 | } else | |
2a3b791e | 1439 | #endif |
cb59155f JA |
1440 | icmp_send(skb, |
1441 | ICMP_DEST_UNREACH, | |
1442 | ICMP_PORT_UNREACH, 0); | |
1443 | return NF_DROP; | |
1da177e4 LT |
1444 | } |
1445 | } | |
1da177e4 | 1446 | } |
b0e010c5 | 1447 | IP_VS_DBG_PKT(12, af, pp, skb, iph.off, |
cb59155f JA |
1448 | "ip_vs_out: packet continues traversal as normal"); |
1449 | return NF_ACCEPT; | |
1da177e4 LT |
1450 | } |
1451 | ||
fc604767 | 1452 | /* |
cb59155f JA |
1453 | * It is hooked at the NF_INET_FORWARD and NF_INET_LOCAL_IN chain, |
1454 | * used only for VS/NAT. | |
fc604767 JA |
1455 | * Check if packet is reply for established ip_vs_conn. |
1456 | */ | |
1457 | static unsigned int | |
06198b34 | 1458 | ip_vs_reply4(void *priv, struct sk_buff *skb, |
238e54c9 | 1459 | const struct nf_hook_state *state) |
fc604767 | 1460 | { |
1b75097d | 1461 | return ip_vs_out(net_ipvs(state->net), state->hook, skb, AF_INET); |
fc604767 JA |
1462 | } |
1463 | ||
1464 | /* | |
1465 | * It is hooked at the NF_INET_LOCAL_OUT chain, used only for VS/NAT. | |
1466 | * Check if packet is reply for established ip_vs_conn. | |
1467 | */ | |
1468 | static unsigned int | |
06198b34 | 1469 | ip_vs_local_reply4(void *priv, struct sk_buff *skb, |
238e54c9 | 1470 | const struct nf_hook_state *state) |
fc604767 | 1471 | { |
1b75097d | 1472 | return ip_vs_out(net_ipvs(state->net), state->hook, skb, AF_INET); |
fc604767 JA |
1473 | } |
1474 | ||
1475 | #ifdef CONFIG_IP_VS_IPV6 | |
1476 | ||
1477 | /* | |
cb59155f JA |
1478 | * It is hooked at the NF_INET_FORWARD and NF_INET_LOCAL_IN chain, |
1479 | * used only for VS/NAT. | |
fc604767 JA |
1480 | * Check if packet is reply for established ip_vs_conn. |
1481 | */ | |
1482 | static unsigned int | |
06198b34 | 1483 | ip_vs_reply6(void *priv, struct sk_buff *skb, |
238e54c9 | 1484 | const struct nf_hook_state *state) |
fc604767 | 1485 | { |
1b75097d | 1486 | return ip_vs_out(net_ipvs(state->net), state->hook, skb, AF_INET6); |
fc604767 JA |
1487 | } |
1488 | ||
1489 | /* | |
1490 | * It is hooked at the NF_INET_LOCAL_OUT chain, used only for VS/NAT. | |
1491 | * Check if packet is reply for established ip_vs_conn. | |
1492 | */ | |
1493 | static unsigned int | |
06198b34 | 1494 | ip_vs_local_reply6(void *priv, struct sk_buff *skb, |
238e54c9 | 1495 | const struct nf_hook_state *state) |
fc604767 | 1496 | { |
1b75097d | 1497 | return ip_vs_out(net_ipvs(state->net), state->hook, skb, AF_INET6); |
fc604767 JA |
1498 | } |
1499 | ||
1500 | #endif | |
1da177e4 | 1501 | |
3b5ca617 | 1502 | static unsigned int |
d8f44c33 EB |
1503 | ip_vs_try_to_schedule(struct netns_ipvs *ipvs, int af, struct sk_buff *skb, |
1504 | struct ip_vs_proto_data *pd, | |
3b5ca617 AG |
1505 | int *verdict, struct ip_vs_conn **cpp, |
1506 | struct ip_vs_iphdr *iph) | |
1507 | { | |
1508 | struct ip_vs_protocol *pp = pd->pp; | |
1509 | ||
1510 | if (!iph->fragoffs) { | |
1511 | /* No (second) fragments need to enter here, as nf_defrag_ipv6 | |
1512 | * replayed fragment zero will already have created the cp | |
1513 | */ | |
1514 | ||
1515 | /* Schedule and create new connection entry into cpp */ | |
d8f44c33 | 1516 | if (!pp->conn_schedule(ipvs, af, skb, pd, verdict, cpp, iph)) |
3b5ca617 AG |
1517 | return 0; |
1518 | } | |
1519 | ||
1520 | if (unlikely(!*cpp)) { | |
1521 | /* sorry, all this trouble for a no-hit :) */ | |
1522 | IP_VS_DBG_PKT(12, af, pp, skb, iph->off, | |
1523 | "ip_vs_in: packet continues traversal as normal"); | |
1524 | if (iph->fragoffs) { | |
1525 | /* Fragment that couldn't be mapped to a conn entry | |
1526 | * is missing module nf_defrag_ipv6 | |
1527 | */ | |
1528 | IP_VS_DBG_RL("Unhandled frag, load nf_defrag_ipv6\n"); | |
1529 | IP_VS_DBG_PKT(7, af, pp, skb, iph->off, | |
1530 | "unhandled fragment"); | |
1531 | } | |
1532 | *verdict = NF_ACCEPT; | |
1533 | return 0; | |
1534 | } | |
1535 | ||
1536 | return 1; | |
1537 | } | |
1538 | ||
1da177e4 LT |
1539 | /* |
1540 | * Handle ICMP messages in the outside-to-inside direction (incoming). | |
1541 | * Find any that might be relevant, check against existing connections, | |
1542 | * forward to the right destination host if relevant. | |
1543 | * Currently handles error types - unreachable, quench, ttl exceeded. | |
1544 | */ | |
e905a9ed | 1545 | static int |
6f2bcea9 EB |
1546 | ip_vs_in_icmp(struct netns_ipvs *ipvs, struct sk_buff *skb, int *related, |
1547 | unsigned int hooknum) | |
1da177e4 | 1548 | { |
1da177e4 LT |
1549 | struct iphdr *iph; |
1550 | struct icmphdr _icmph, *ic; | |
1551 | struct iphdr _ciph, *cih; /* The ip header contained within the ICMP */ | |
51ef348b | 1552 | struct ip_vs_iphdr ciph; |
1da177e4 LT |
1553 | struct ip_vs_conn *cp; |
1554 | struct ip_vs_protocol *pp; | |
9330419d | 1555 | struct ip_vs_proto_data *pd; |
f2edb9f7 | 1556 | unsigned int offset, offset2, ihl, verdict; |
6044eeff | 1557 | bool ipip, new_cp = false; |
1da177e4 LT |
1558 | |
1559 | *related = 1; | |
1560 | ||
1561 | /* reassemble IP fragments */ | |
56f8a75c | 1562 | if (ip_is_fragment(ip_hdr(skb))) { |
57781c1c | 1563 | if (ip_vs_gather_frags(ipvs, skb, ip_vs_defrag_user(hooknum))) |
1da177e4 | 1564 | return NF_STOLEN; |
1da177e4 LT |
1565 | } |
1566 | ||
eddc9ec5 | 1567 | iph = ip_hdr(skb); |
1da177e4 LT |
1568 | offset = ihl = iph->ihl * 4; |
1569 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
1570 | if (ic == NULL) | |
1571 | return NF_DROP; | |
1572 | ||
14d5e834 | 1573 | IP_VS_DBG(12, "Incoming ICMP (%d,%d) %pI4->%pI4\n", |
1da177e4 | 1574 | ic->type, ntohs(icmp_id(ic)), |
14d5e834 | 1575 | &iph->saddr, &iph->daddr); |
1da177e4 LT |
1576 | |
1577 | /* | |
1578 | * Work through seeing if this is for us. | |
1579 | * These checks are supposed to be in an order that means easy | |
1580 | * things are checked first to speed up processing.... however | |
1581 | * this means that some packets will manage to get a long way | |
1582 | * down this stack and then be rejected, but that's life. | |
1583 | */ | |
1584 | if ((ic->type != ICMP_DEST_UNREACH) && | |
1585 | (ic->type != ICMP_SOURCE_QUENCH) && | |
1586 | (ic->type != ICMP_TIME_EXCEEDED)) { | |
1587 | *related = 0; | |
1588 | return NF_ACCEPT; | |
1589 | } | |
1590 | ||
1591 | /* Now find the contained IP header */ | |
1592 | offset += sizeof(_icmph); | |
1593 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
1594 | if (cih == NULL) | |
1595 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
1596 | ||
f2edb9f7 JA |
1597 | /* Special case for errors for IPIP packets */ |
1598 | ipip = false; | |
1599 | if (cih->protocol == IPPROTO_IPIP) { | |
1600 | if (unlikely(cih->frag_off & htons(IP_OFFSET))) | |
1601 | return NF_ACCEPT; | |
1602 | /* Error for our IPIP must arrive at LOCAL_IN */ | |
1603 | if (!(skb_rtable(skb)->rt_flags & RTCF_LOCAL)) | |
1604 | return NF_ACCEPT; | |
1605 | offset += cih->ihl * 4; | |
1606 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
1607 | if (cih == NULL) | |
1608 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
1609 | ipip = true; | |
1610 | } | |
1611 | ||
18d6ade6 | 1612 | pd = ip_vs_proto_data_get(ipvs, cih->protocol); |
9330419d | 1613 | if (!pd) |
1da177e4 | 1614 | return NF_ACCEPT; |
9330419d | 1615 | pp = pd->pp; |
1da177e4 LT |
1616 | |
1617 | /* Is the embedded protocol header present? */ | |
4412ec49 | 1618 | if (unlikely(cih->frag_off & htons(IP_OFFSET) && |
1da177e4 LT |
1619 | pp->dont_defrag)) |
1620 | return NF_ACCEPT; | |
1621 | ||
0d79641a JA |
1622 | IP_VS_DBG_PKT(11, AF_INET, pp, skb, offset, |
1623 | "Checking incoming ICMP for"); | |
1da177e4 | 1624 | |
f2edb9f7 | 1625 | offset2 = offset; |
4fd9beef | 1626 | ip_vs_fill_iph_skb_icmp(AF_INET, skb, offset, !ipip, &ciph); |
63dca2c0 | 1627 | offset = ciph.len; |
b0e010c5 | 1628 | |
f2edb9f7 JA |
1629 | /* The embedded headers contain source and dest in reverse order. |
1630 | * For IPIP this is error for request, not for reply. | |
1631 | */ | |
ab161976 | 1632 | cp = pp->conn_in_get(ipvs, AF_INET, skb, &ciph); |
6044eeff AG |
1633 | |
1634 | if (!cp) { | |
1635 | int v; | |
1636 | ||
a47b4300 | 1637 | if (!sysctl_schedule_icmp(ipvs)) |
6044eeff AG |
1638 | return NF_ACCEPT; |
1639 | ||
d8f44c33 | 1640 | if (!ip_vs_try_to_schedule(ipvs, AF_INET, skb, pd, &v, &cp, &ciph)) |
6044eeff AG |
1641 | return v; |
1642 | new_cp = true; | |
1643 | } | |
1da177e4 LT |
1644 | |
1645 | verdict = NF_DROP; | |
1646 | ||
1647 | /* Ensure the checksum is correct */ | |
60476372 | 1648 | if (!skb_csum_unnecessary(skb) && ip_vs_checksum_complete(skb, ihl)) { |
1da177e4 | 1649 | /* Failed checksum! */ |
14d5e834 HH |
1650 | IP_VS_DBG(1, "Incoming ICMP: failed checksum from %pI4!\n", |
1651 | &iph->saddr); | |
1da177e4 LT |
1652 | goto out; |
1653 | } | |
1654 | ||
f2edb9f7 JA |
1655 | if (ipip) { |
1656 | __be32 info = ic->un.gateway; | |
f44a5f45 PC |
1657 | __u8 type = ic->type; |
1658 | __u8 code = ic->code; | |
f2edb9f7 JA |
1659 | |
1660 | /* Update the MTU */ | |
1661 | if (ic->type == ICMP_DEST_UNREACH && | |
1662 | ic->code == ICMP_FRAG_NEEDED) { | |
1663 | struct ip_vs_dest *dest = cp->dest; | |
1664 | u32 mtu = ntohs(ic->un.frag.mtu); | |
f44a5f45 | 1665 | __be16 frag_off = cih->frag_off; |
f2edb9f7 JA |
1666 | |
1667 | /* Strip outer IP and ICMP, go to IPIP header */ | |
f44a5f45 PC |
1668 | if (pskb_pull(skb, ihl + sizeof(_icmph)) == NULL) |
1669 | goto ignore_ipip; | |
f2edb9f7 JA |
1670 | offset2 -= ihl + sizeof(_icmph); |
1671 | skb_reset_network_header(skb); | |
1672 | IP_VS_DBG(12, "ICMP for IPIP %pI4->%pI4: mtu=%u\n", | |
1673 | &ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr, mtu); | |
6f2bcea9 | 1674 | ipv4_update_pmtu(skb, ipvs->net, |
f2edb9f7 | 1675 | mtu, 0, 0, 0, 0); |
f2edb9f7 | 1676 | /* Client uses PMTUD? */ |
f44a5f45 | 1677 | if (!(frag_off & htons(IP_DF))) |
f2edb9f7 JA |
1678 | goto ignore_ipip; |
1679 | /* Prefer the resulting PMTU */ | |
1680 | if (dest) { | |
026ace06 JA |
1681 | struct ip_vs_dest_dst *dest_dst; |
1682 | ||
1683 | rcu_read_lock(); | |
1684 | dest_dst = rcu_dereference(dest->dest_dst); | |
1685 | if (dest_dst) | |
1686 | mtu = dst_mtu(dest_dst->dst_cache); | |
1687 | rcu_read_unlock(); | |
f2edb9f7 JA |
1688 | } |
1689 | if (mtu > 68 + sizeof(struct iphdr)) | |
1690 | mtu -= sizeof(struct iphdr); | |
1691 | info = htonl(mtu); | |
1692 | } | |
1693 | /* Strip outer IP, ICMP and IPIP, go to IP header of | |
1694 | * original request. | |
1695 | */ | |
f44a5f45 PC |
1696 | if (pskb_pull(skb, offset2) == NULL) |
1697 | goto ignore_ipip; | |
f2edb9f7 JA |
1698 | skb_reset_network_header(skb); |
1699 | IP_VS_DBG(12, "Sending ICMP for %pI4->%pI4: t=%u, c=%u, i=%u\n", | |
1700 | &ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr, | |
f44a5f45 PC |
1701 | type, code, ntohl(info)); |
1702 | icmp_send(skb, type, code, info); | |
f2edb9f7 JA |
1703 | /* ICMP can be shorter but anyways, account it */ |
1704 | ip_vs_out_stats(cp, skb); | |
1705 | ||
1706 | ignore_ipip: | |
1707 | consume_skb(skb); | |
1708 | verdict = NF_STOLEN; | |
1709 | goto out; | |
1710 | } | |
1711 | ||
1da177e4 LT |
1712 | /* do the statistics and put it back */ |
1713 | ip_vs_in_stats(cp, skb); | |
06f3d7f9 JA |
1714 | if (IPPROTO_TCP == cih->protocol || IPPROTO_UDP == cih->protocol || |
1715 | IPPROTO_SCTP == cih->protocol) | |
1da177e4 | 1716 | offset += 2 * sizeof(__u16); |
d4383f04 | 1717 | verdict = ip_vs_icmp_xmit(skb, cp, pp, offset, hooknum, &ciph); |
1da177e4 | 1718 | |
552ad65a | 1719 | out: |
6044eeff AG |
1720 | if (likely(!new_cp)) |
1721 | __ip_vs_conn_put(cp); | |
1722 | else | |
1723 | ip_vs_conn_put(cp); | |
1da177e4 LT |
1724 | |
1725 | return verdict; | |
1726 | } | |
1727 | ||
2a3b791e | 1728 | #ifdef CONFIG_IP_VS_IPV6 |
6f2bcea9 EB |
1729 | static int ip_vs_in_icmp_v6(struct netns_ipvs *ipvs, struct sk_buff *skb, |
1730 | int *related, unsigned int hooknum, | |
1731 | struct ip_vs_iphdr *iph) | |
2a3b791e | 1732 | { |
2a3b791e | 1733 | struct icmp6hdr _icmph, *ic; |
63dca2c0 | 1734 | struct ip_vs_iphdr ciph = {.flags = 0, .fragoffs = 0};/*Contained IP */ |
2a3b791e JV |
1735 | struct ip_vs_conn *cp; |
1736 | struct ip_vs_protocol *pp; | |
9330419d | 1737 | struct ip_vs_proto_data *pd; |
b0e010c5 | 1738 | unsigned int offset, verdict; |
6044eeff | 1739 | bool new_cp = false; |
2a3b791e | 1740 | |
63dca2c0 | 1741 | *related = 1; |
2a3b791e | 1742 | |
2f74713d | 1743 | ic = frag_safe_skb_hp(skb, iph->len, sizeof(_icmph), &_icmph, iph); |
2a3b791e JV |
1744 | if (ic == NULL) |
1745 | return NF_DROP; | |
1746 | ||
2a3b791e JV |
1747 | /* |
1748 | * Work through seeing if this is for us. | |
1749 | * These checks are supposed to be in an order that means easy | |
1750 | * things are checked first to speed up processing.... however | |
1751 | * this means that some packets will manage to get a long way | |
1752 | * down this stack and then be rejected, but that's life. | |
1753 | */ | |
2fab8917 | 1754 | if (ic->icmp6_type & ICMPV6_INFOMSG_MASK) { |
2a3b791e JV |
1755 | *related = 0; |
1756 | return NF_ACCEPT; | |
1757 | } | |
2f74713d JDB |
1758 | /* Fragment header that is before ICMP header tells us that: |
1759 | * it's not an error message since they can't be fragmented. | |
1760 | */ | |
e7165030 | 1761 | if (iph->flags & IP6_FH_F_FRAG) |
2f74713d | 1762 | return NF_DROP; |
2a3b791e | 1763 | |
63dca2c0 JDB |
1764 | IP_VS_DBG(8, "Incoming ICMPv6 (%d,%d) %pI6c->%pI6c\n", |
1765 | ic->icmp6_type, ntohs(icmpv6_id(ic)), | |
1766 | &iph->saddr, &iph->daddr); | |
1767 | ||
b0e010c5 | 1768 | offset = iph->len + sizeof(_icmph); |
4fd9beef | 1769 | if (!ip_vs_fill_iph_skb_icmp(AF_INET6, skb, offset, true, &ciph)) |
b0e010c5 | 1770 | return NF_ACCEPT; |
2a3b791e | 1771 | |
18d6ade6 | 1772 | pd = ip_vs_proto_data_get(ipvs, ciph.protocol); |
9330419d | 1773 | if (!pd) |
2a3b791e | 1774 | return NF_ACCEPT; |
9330419d | 1775 | pp = pd->pp; |
2a3b791e | 1776 | |
63dca2c0 JDB |
1777 | /* Cannot handle fragmented embedded protocol */ |
1778 | if (ciph.fragoffs) | |
2a3b791e JV |
1779 | return NF_ACCEPT; |
1780 | ||
b0e010c5 | 1781 | IP_VS_DBG_PKT(11, AF_INET6, pp, skb, offset, |
0d79641a | 1782 | "Checking incoming ICMPv6 for"); |
2a3b791e | 1783 | |
2f74713d JDB |
1784 | /* The embedded headers contain source and dest in reverse order |
1785 | * if not from localhost | |
1786 | */ | |
ab161976 | 1787 | cp = pp->conn_in_get(ipvs, AF_INET6, skb, &ciph); |
2f74713d | 1788 | |
6044eeff AG |
1789 | if (!cp) { |
1790 | int v; | |
1791 | ||
a47b4300 | 1792 | if (!sysctl_schedule_icmp(ipvs)) |
6044eeff AG |
1793 | return NF_ACCEPT; |
1794 | ||
d8f44c33 | 1795 | if (!ip_vs_try_to_schedule(ipvs, AF_INET6, skb, pd, &v, &cp, &ciph)) |
6044eeff AG |
1796 | return v; |
1797 | ||
1798 | new_cp = true; | |
1799 | } | |
1800 | ||
2f74713d JDB |
1801 | /* VS/TUN, VS/DR and LOCALNODE just let it go */ |
1802 | if ((hooknum == NF_INET_LOCAL_OUT) && | |
1803 | (IP_VS_FWD_METHOD(cp) != IP_VS_CONN_F_MASQ)) { | |
6044eeff AG |
1804 | verdict = NF_ACCEPT; |
1805 | goto out; | |
2f74713d | 1806 | } |
2a3b791e | 1807 | |
2a3b791e JV |
1808 | /* do the statistics and put it back */ |
1809 | ip_vs_in_stats(cp, skb); | |
63dca2c0 JDB |
1810 | |
1811 | /* Need to mangle contained IPv6 header in ICMPv6 packet */ | |
b0e010c5 | 1812 | offset = ciph.len; |
63dca2c0 JDB |
1813 | if (IPPROTO_TCP == ciph.protocol || IPPROTO_UDP == ciph.protocol || |
1814 | IPPROTO_SCTP == ciph.protocol) | |
b0e010c5 | 1815 | offset += 2 * sizeof(__u16); /* Also mangle ports */ |
63dca2c0 | 1816 | |
b0e010c5 | 1817 | verdict = ip_vs_icmp_xmit_v6(skb, cp, pp, offset, hooknum, &ciph); |
2a3b791e | 1818 | |
6044eeff AG |
1819 | out: |
1820 | if (likely(!new_cp)) | |
1821 | __ip_vs_conn_put(cp); | |
1822 | else | |
1823 | ip_vs_conn_put(cp); | |
2a3b791e JV |
1824 | |
1825 | return verdict; | |
1826 | } | |
1827 | #endif | |
1828 | ||
1829 | ||
1da177e4 LT |
1830 | /* |
1831 | * Check if it's for virtual services, look it up, | |
1832 | * and send it on its way... | |
1833 | */ | |
1834 | static unsigned int | |
6e385bb3 | 1835 | ip_vs_in(struct netns_ipvs *ipvs, unsigned int hooknum, struct sk_buff *skb, int af) |
1da177e4 | 1836 | { |
51ef348b | 1837 | struct ip_vs_iphdr iph; |
1da177e4 | 1838 | struct ip_vs_protocol *pp; |
9330419d | 1839 | struct ip_vs_proto_data *pd; |
1da177e4 | 1840 | struct ip_vs_conn *cp; |
4a516f11 | 1841 | int ret, pkts; |
d752c364 | 1842 | int conn_reuse_mode; |
340c78e5 | 1843 | struct sock *sk; |
2a3b791e | 1844 | |
fc604767 JA |
1845 | /* Already marked as IPVS request or reply? */ |
1846 | if (skb->ipvs_property) | |
1847 | return NF_ACCEPT; | |
1848 | ||
1da177e4 | 1849 | /* |
cb59155f JA |
1850 | * Big tappo: |
1851 | * - remote client: only PACKET_HOST | |
1852 | * - route: used for struct net when skb->dev is unset | |
1da177e4 | 1853 | */ |
cb59155f JA |
1854 | if (unlikely((skb->pkt_type != PACKET_HOST && |
1855 | hooknum != NF_INET_LOCAL_OUT) || | |
1856 | !skb_dst(skb))) { | |
4fd9beef | 1857 | ip_vs_fill_iph_skb(af, skb, false, &iph); |
cb59155f JA |
1858 | IP_VS_DBG_BUF(12, "packet type=%d proto=%d daddr=%s" |
1859 | " ignored in hook %u\n", | |
1860 | skb->pkt_type, iph.protocol, | |
1861 | IP_VS_DBG_ADDR(af, &iph.daddr), hooknum); | |
1da177e4 LT |
1862 | return NF_ACCEPT; |
1863 | } | |
7a4f0761 | 1864 | /* ipvs enabled in this netns ? */ |
0c12582f | 1865 | if (unlikely(sysctl_backup_only(ipvs) || !ipvs->enable)) |
7a4f0761 HS |
1866 | return NF_ACCEPT; |
1867 | ||
4fd9beef | 1868 | ip_vs_fill_iph_skb(af, skb, false, &iph); |
cb59155f JA |
1869 | |
1870 | /* Bad... Do not break raw sockets */ | |
340c78e5 ED |
1871 | sk = skb_to_full_sk(skb); |
1872 | if (unlikely(sk && hooknum == NF_INET_LOCAL_OUT && | |
cb59155f | 1873 | af == AF_INET)) { |
cb59155f | 1874 | |
340c78e5 | 1875 | if (sk->sk_family == PF_INET && inet_sk(sk)->nodefrag) |
cb59155f JA |
1876 | return NF_ACCEPT; |
1877 | } | |
1da177e4 | 1878 | |
94b26551 JV |
1879 | #ifdef CONFIG_IP_VS_IPV6 |
1880 | if (af == AF_INET6) { | |
1881 | if (unlikely(iph.protocol == IPPROTO_ICMPV6)) { | |
1ca5bb54 | 1882 | int related; |
6f2bcea9 EB |
1883 | int verdict = ip_vs_in_icmp_v6(ipvs, skb, &related, |
1884 | hooknum, &iph); | |
1da177e4 | 1885 | |
94b26551 JV |
1886 | if (related) |
1887 | return verdict; | |
94b26551 JV |
1888 | } |
1889 | } else | |
1890 | #endif | |
1891 | if (unlikely(iph.protocol == IPPROTO_ICMP)) { | |
1ca5bb54 | 1892 | int related; |
6f2bcea9 EB |
1893 | int verdict = ip_vs_in_icmp(ipvs, skb, &related, |
1894 | hooknum); | |
94b26551 JV |
1895 | |
1896 | if (related) | |
1897 | return verdict; | |
94b26551 | 1898 | } |
1da177e4 LT |
1899 | |
1900 | /* Protocol supported? */ | |
18d6ade6 | 1901 | pd = ip_vs_proto_data_get(ipvs, iph.protocol); |
4e478098 AG |
1902 | if (unlikely(!pd)) { |
1903 | /* The only way we'll see this packet again is if it's | |
1904 | * encapsulated, so mark it with ipvs_property=1 so we | |
1905 | * skip it if we're ignoring tunneled packets | |
1906 | */ | |
1907 | if (sysctl_ignore_tunneled(ipvs)) | |
1908 | skb->ipvs_property = 1; | |
1909 | ||
1da177e4 | 1910 | return NF_ACCEPT; |
4e478098 | 1911 | } |
9330419d | 1912 | pp = pd->pp; |
1da177e4 LT |
1913 | /* |
1914 | * Check if the packet belongs to an existing connection entry | |
1915 | */ | |
ab161976 | 1916 | cp = pp->conn_in_get(ipvs, af, skb, &iph); |
dc7b3eb9 | 1917 | |
d752c364 | 1918 | conn_reuse_mode = sysctl_conn_reuse_mode(ipvs); |
f719e375 JA |
1919 | if (conn_reuse_mode && !iph.fragoffs && is_new_conn(skb, &iph) && cp) { |
1920 | bool uses_ct = false, resched = false; | |
1921 | ||
1922 | if (unlikely(sysctl_expire_nodest_conn(ipvs)) && cp->dest && | |
1923 | unlikely(!atomic_read(&cp->dest->weight))) { | |
1924 | resched = true; | |
1925 | uses_ct = ip_vs_conn_uses_conntrack(cp, skb); | |
1926 | } else if (is_new_conn_expected(cp, conn_reuse_mode)) { | |
1927 | uses_ct = ip_vs_conn_uses_conntrack(cp, skb); | |
1928 | if (!atomic_read(&cp->n_control)) { | |
1929 | resched = true; | |
1930 | } else { | |
1931 | /* Do not reschedule controlling connection | |
1932 | * that uses conntrack while it is still | |
1933 | * referenced by controlled connection(s). | |
1934 | */ | |
1935 | resched = !uses_ct; | |
1936 | } | |
1937 | } | |
1938 | ||
1939 | if (resched) { | |
1940 | if (!atomic_read(&cp->n_control)) | |
1941 | ip_vs_conn_expire_now(cp); | |
1942 | __ip_vs_conn_put(cp); | |
1943 | if (uses_ct) | |
1944 | return NF_DROP; | |
1945 | cp = NULL; | |
1946 | } | |
dc7b3eb9 GL |
1947 | } |
1948 | ||
3b5ca617 | 1949 | if (unlikely(!cp)) { |
1da177e4 LT |
1950 | int v; |
1951 | ||
d8f44c33 | 1952 | if (!ip_vs_try_to_schedule(ipvs, af, skb, pd, &v, &cp, &iph)) |
1da177e4 LT |
1953 | return v; |
1954 | } | |
1955 | ||
b0e010c5 | 1956 | IP_VS_DBG_PKT(11, af, pp, skb, iph.off, "Incoming packet"); |
3b5ca617 | 1957 | |
1da177e4 LT |
1958 | /* Check the server status */ |
1959 | if (cp->dest && !(cp->dest->flags & IP_VS_DEST_F_AVAILABLE)) { | |
1960 | /* the destination server is not available */ | |
1961 | ||
71a8ab6c | 1962 | if (sysctl_expire_nodest_conn(ipvs)) { |
1da177e4 LT |
1963 | /* try to expire the connection immediately */ |
1964 | ip_vs_conn_expire_now(cp); | |
1da177e4 | 1965 | } |
dc8103f2 JA |
1966 | /* don't restart its timer, and silently |
1967 | drop the packet. */ | |
1968 | __ip_vs_conn_put(cp); | |
1da177e4 LT |
1969 | return NF_DROP; |
1970 | } | |
1971 | ||
1972 | ip_vs_in_stats(cp, skb); | |
4a516f11 | 1973 | ip_vs_set_state(cp, IP_VS_DIR_INPUT, skb, pd); |
1da177e4 | 1974 | if (cp->packet_xmit) |
d4383f04 | 1975 | ret = cp->packet_xmit(skb, cp, pp, &iph); |
1da177e4 LT |
1976 | /* do not touch skb anymore */ |
1977 | else { | |
1978 | IP_VS_DBG_RL("warning: packet_xmit is null"); | |
1979 | ret = NF_ACCEPT; | |
1980 | } | |
1981 | ||
efac5276 RB |
1982 | /* Increase its packet counter and check if it is needed |
1983 | * to be synchronized | |
1984 | * | |
1985 | * Sync connection if it is about to close to | |
1986 | * encorage the standby servers to update the connections timeout | |
986a0757 HS |
1987 | * |
1988 | * For ONE_PKT let ip_vs_sync_conn() do the filter work. | |
efac5276 | 1989 | */ |
f131315f | 1990 | |
986a0757 | 1991 | if (cp->flags & IP_VS_CONN_F_ONE_PACKET) |
59e0350e | 1992 | pkts = sysctl_sync_threshold(ipvs); |
986a0757 HS |
1993 | else |
1994 | pkts = atomic_add_return(1, &cp->in_pkts); | |
1995 | ||
749c42b6 | 1996 | if (ipvs->sync_state & IP_VS_STATE_MASTER) |
b61a8c1a | 1997 | ip_vs_sync_conn(ipvs, cp, pkts); |
698e2a8d MA |
1998 | else if ((cp->flags & IP_VS_CONN_F_ONE_PACKET) && cp->control) |
1999 | /* increment is done inside ip_vs_sync_conn too */ | |
2000 | atomic_inc(&cp->control->in_pkts); | |
1da177e4 LT |
2001 | |
2002 | ip_vs_conn_put(cp); | |
2003 | return ret; | |
2004 | } | |
2005 | ||
cb59155f JA |
2006 | /* |
2007 | * AF_INET handler in NF_INET_LOCAL_IN chain | |
2008 | * Schedule and forward packets from remote clients | |
2009 | */ | |
2010 | static unsigned int | |
06198b34 | 2011 | ip_vs_remote_request4(void *priv, struct sk_buff *skb, |
238e54c9 | 2012 | const struct nf_hook_state *state) |
cb59155f | 2013 | { |
6e385bb3 | 2014 | return ip_vs_in(net_ipvs(state->net), state->hook, skb, AF_INET); |
cb59155f JA |
2015 | } |
2016 | ||
2017 | /* | |
2018 | * AF_INET handler in NF_INET_LOCAL_OUT chain | |
2019 | * Schedule and forward packets from local clients | |
2020 | */ | |
2021 | static unsigned int | |
06198b34 | 2022 | ip_vs_local_request4(void *priv, struct sk_buff *skb, |
238e54c9 | 2023 | const struct nf_hook_state *state) |
cb59155f | 2024 | { |
6e385bb3 | 2025 | return ip_vs_in(net_ipvs(state->net), state->hook, skb, AF_INET); |
cb59155f JA |
2026 | } |
2027 | ||
2028 | #ifdef CONFIG_IP_VS_IPV6 | |
2029 | ||
2030 | /* | |
2031 | * AF_INET6 handler in NF_INET_LOCAL_IN chain | |
2032 | * Schedule and forward packets from remote clients | |
2033 | */ | |
2034 | static unsigned int | |
06198b34 | 2035 | ip_vs_remote_request6(void *priv, struct sk_buff *skb, |
238e54c9 | 2036 | const struct nf_hook_state *state) |
cb59155f | 2037 | { |
6e385bb3 | 2038 | return ip_vs_in(net_ipvs(state->net), state->hook, skb, AF_INET6); |
cb59155f JA |
2039 | } |
2040 | ||
2041 | /* | |
2042 | * AF_INET6 handler in NF_INET_LOCAL_OUT chain | |
2043 | * Schedule and forward packets from local clients | |
2044 | */ | |
2045 | static unsigned int | |
06198b34 | 2046 | ip_vs_local_request6(void *priv, struct sk_buff *skb, |
238e54c9 | 2047 | const struct nf_hook_state *state) |
cb59155f | 2048 | { |
6e385bb3 | 2049 | return ip_vs_in(net_ipvs(state->net), state->hook, skb, AF_INET6); |
cb59155f JA |
2050 | } |
2051 | ||
2052 | #endif | |
2053 | ||
1da177e4 LT |
2054 | |
2055 | /* | |
6e23ae2a | 2056 | * It is hooked at the NF_INET_FORWARD chain, in order to catch ICMP |
1da177e4 LT |
2057 | * related packets destined for 0.0.0.0/0. |
2058 | * When fwmark-based virtual service is used, such as transparent | |
2059 | * cache cluster, TCP packets can be marked and routed to ip_vs_in, | |
2060 | * but ICMP destined for 0.0.0.0/0 cannot not be easily marked and | |
6e23ae2a | 2061 | * sent to ip_vs_in_icmp. So, catch them at the NF_INET_FORWARD chain |
1da177e4 LT |
2062 | * and send them to ip_vs_in_icmp. |
2063 | */ | |
2064 | static unsigned int | |
06198b34 | 2065 | ip_vs_forward_icmp(void *priv, struct sk_buff *skb, |
238e54c9 | 2066 | const struct nf_hook_state *state) |
1da177e4 LT |
2067 | { |
2068 | int r; | |
6f2bcea9 | 2069 | struct netns_ipvs *ipvs = net_ipvs(state->net); |
1da177e4 | 2070 | |
3db05fea | 2071 | if (ip_hdr(skb)->protocol != IPPROTO_ICMP) |
1da177e4 LT |
2072 | return NF_ACCEPT; |
2073 | ||
7a4f0761 | 2074 | /* ipvs enabled in this netns ? */ |
0c12582f | 2075 | if (unlikely(sysctl_backup_only(ipvs) || !ipvs->enable)) |
7a4f0761 HS |
2076 | return NF_ACCEPT; |
2077 | ||
6f2bcea9 | 2078 | return ip_vs_in_icmp(ipvs, skb, &r, state->hook); |
1da177e4 LT |
2079 | } |
2080 | ||
2a3b791e JV |
2081 | #ifdef CONFIG_IP_VS_IPV6 |
2082 | static unsigned int | |
06198b34 | 2083 | ip_vs_forward_icmp_v6(void *priv, struct sk_buff *skb, |
238e54c9 | 2084 | const struct nf_hook_state *state) |
2a3b791e JV |
2085 | { |
2086 | int r; | |
6f2bcea9 | 2087 | struct netns_ipvs *ipvs = net_ipvs(state->net); |
63dca2c0 | 2088 | struct ip_vs_iphdr iphdr; |
2a3b791e | 2089 | |
4fd9beef | 2090 | ip_vs_fill_iph_skb(AF_INET6, skb, false, &iphdr); |
63dca2c0 | 2091 | if (iphdr.protocol != IPPROTO_ICMPV6) |
2a3b791e JV |
2092 | return NF_ACCEPT; |
2093 | ||
7a4f0761 | 2094 | /* ipvs enabled in this netns ? */ |
0c12582f | 2095 | if (unlikely(sysctl_backup_only(ipvs) || !ipvs->enable)) |
7a4f0761 HS |
2096 | return NF_ACCEPT; |
2097 | ||
6f2bcea9 | 2098 | return ip_vs_in_icmp_v6(ipvs, skb, &r, state->hook, &iphdr); |
2a3b791e JV |
2099 | } |
2100 | #endif | |
2101 | ||
1da177e4 | 2102 | |
1999414a | 2103 | static struct nf_hook_ops ip_vs_ops[] __read_mostly = { |
cb59155f JA |
2104 | /* After packet filtering, change source only for VS/NAT */ |
2105 | { | |
2106 | .hook = ip_vs_reply4, | |
4c809d63 | 2107 | .pf = NFPROTO_IPV4, |
cb59155f | 2108 | .hooknum = NF_INET_LOCAL_IN, |
afb523c5 | 2109 | .priority = NF_IP_PRI_NAT_SRC - 2, |
cb59155f | 2110 | }, |
41c5b317 PM |
2111 | /* After packet filtering, forward packet through VS/DR, VS/TUN, |
2112 | * or VS/NAT(change destination), so that filtering rules can be | |
2113 | * applied to IPVS. */ | |
2114 | { | |
cb59155f | 2115 | .hook = ip_vs_remote_request4, |
4c809d63 | 2116 | .pf = NFPROTO_IPV4, |
cb59155f | 2117 | .hooknum = NF_INET_LOCAL_IN, |
afb523c5 | 2118 | .priority = NF_IP_PRI_NAT_SRC - 1, |
41c5b317 | 2119 | }, |
fc604767 | 2120 | /* Before ip_vs_in, change source only for VS/NAT */ |
41c5b317 | 2121 | { |
fc604767 | 2122 | .hook = ip_vs_local_reply4, |
4c809d63 | 2123 | .pf = NFPROTO_IPV4, |
fc604767 | 2124 | .hooknum = NF_INET_LOCAL_OUT, |
afb523c5 | 2125 | .priority = NF_IP_PRI_NAT_DST + 1, |
41c5b317 | 2126 | }, |
cb59155f JA |
2127 | /* After mangle, schedule and forward local requests */ |
2128 | { | |
2129 | .hook = ip_vs_local_request4, | |
4c809d63 | 2130 | .pf = NFPROTO_IPV4, |
cb59155f | 2131 | .hooknum = NF_INET_LOCAL_OUT, |
afb523c5 | 2132 | .priority = NF_IP_PRI_NAT_DST + 2, |
cb59155f | 2133 | }, |
41c5b317 PM |
2134 | /* After packet filtering (but before ip_vs_out_icmp), catch icmp |
2135 | * destined for 0.0.0.0/0, which is for incoming IPVS connections */ | |
2136 | { | |
2137 | .hook = ip_vs_forward_icmp, | |
4c809d63 | 2138 | .pf = NFPROTO_IPV4, |
cb59155f JA |
2139 | .hooknum = NF_INET_FORWARD, |
2140 | .priority = 99, | |
41c5b317 | 2141 | }, |
fc604767 JA |
2142 | /* After packet filtering, change source only for VS/NAT */ |
2143 | { | |
2144 | .hook = ip_vs_reply4, | |
4c809d63 | 2145 | .pf = NFPROTO_IPV4, |
fc604767 JA |
2146 | .hooknum = NF_INET_FORWARD, |
2147 | .priority = 100, | |
2148 | }, | |
473b23d3 | 2149 | #ifdef CONFIG_IP_VS_IPV6 |
cb59155f JA |
2150 | /* After packet filtering, change source only for VS/NAT */ |
2151 | { | |
2152 | .hook = ip_vs_reply6, | |
4c809d63 | 2153 | .pf = NFPROTO_IPV6, |
cb59155f | 2154 | .hooknum = NF_INET_LOCAL_IN, |
afb523c5 | 2155 | .priority = NF_IP6_PRI_NAT_SRC - 2, |
cb59155f | 2156 | }, |
473b23d3 JV |
2157 | /* After packet filtering, forward packet through VS/DR, VS/TUN, |
2158 | * or VS/NAT(change destination), so that filtering rules can be | |
2159 | * applied to IPVS. */ | |
2160 | { | |
cb59155f | 2161 | .hook = ip_vs_remote_request6, |
4c809d63 | 2162 | .pf = NFPROTO_IPV6, |
cb59155f | 2163 | .hooknum = NF_INET_LOCAL_IN, |
afb523c5 | 2164 | .priority = NF_IP6_PRI_NAT_SRC - 1, |
473b23d3 | 2165 | }, |
fc604767 | 2166 | /* Before ip_vs_in, change source only for VS/NAT */ |
473b23d3 | 2167 | { |
fc604767 | 2168 | .hook = ip_vs_local_reply6, |
eb90b0c7 | 2169 | .pf = NFPROTO_IPV6, |
fc604767 | 2170 | .hooknum = NF_INET_LOCAL_OUT, |
afb523c5 | 2171 | .priority = NF_IP6_PRI_NAT_DST + 1, |
473b23d3 | 2172 | }, |
cb59155f JA |
2173 | /* After mangle, schedule and forward local requests */ |
2174 | { | |
2175 | .hook = ip_vs_local_request6, | |
4c809d63 | 2176 | .pf = NFPROTO_IPV6, |
cb59155f | 2177 | .hooknum = NF_INET_LOCAL_OUT, |
afb523c5 | 2178 | .priority = NF_IP6_PRI_NAT_DST + 2, |
cb59155f | 2179 | }, |
473b23d3 JV |
2180 | /* After packet filtering (but before ip_vs_out_icmp), catch icmp |
2181 | * destined for 0.0.0.0/0, which is for incoming IPVS connections */ | |
2182 | { | |
2183 | .hook = ip_vs_forward_icmp_v6, | |
4c809d63 | 2184 | .pf = NFPROTO_IPV6, |
cb59155f JA |
2185 | .hooknum = NF_INET_FORWARD, |
2186 | .priority = 99, | |
473b23d3 | 2187 | }, |
fc604767 JA |
2188 | /* After packet filtering, change source only for VS/NAT */ |
2189 | { | |
2190 | .hook = ip_vs_reply6, | |
4c809d63 | 2191 | .pf = NFPROTO_IPV6, |
fc604767 JA |
2192 | .hooknum = NF_INET_FORWARD, |
2193 | .priority = 100, | |
2194 | }, | |
473b23d3 | 2195 | #endif |
1da177e4 | 2196 | }; |
61b1ab45 HS |
2197 | /* |
2198 | * Initialize IP Virtual Server netns mem. | |
2199 | */ | |
2200 | static int __net_init __ip_vs_init(struct net *net) | |
2201 | { | |
2202 | struct netns_ipvs *ipvs; | |
efe41606 | 2203 | int ret; |
61b1ab45 | 2204 | |
61b1ab45 | 2205 | ipvs = net_generic(net, ip_vs_net_id); |
0a9ee813 | 2206 | if (ipvs == NULL) |
61b1ab45 | 2207 | return -ENOMEM; |
0a9ee813 | 2208 | |
7a4f0761 HS |
2209 | /* Hold the beast until a service is registerd */ |
2210 | ipvs->enable = 0; | |
f6340ee0 | 2211 | ipvs->net = net; |
61b1ab45 HS |
2212 | /* Counters used for creating unique names */ |
2213 | ipvs->gen = atomic_read(&ipvs_netns_cnt); | |
2214 | atomic_inc(&ipvs_netns_cnt); | |
2215 | net->ipvs = ipvs; | |
7a4f0761 | 2216 | |
a4dd0360 | 2217 | if (ip_vs_estimator_net_init(ipvs) < 0) |
7a4f0761 HS |
2218 | goto estimator_fail; |
2219 | ||
3d993766 | 2220 | if (ip_vs_control_net_init(ipvs) < 0) |
7a4f0761 HS |
2221 | goto control_fail; |
2222 | ||
7d1f88ec | 2223 | if (ip_vs_protocol_net_init(ipvs) < 0) |
7a4f0761 HS |
2224 | goto protocol_fail; |
2225 | ||
b5dd212c | 2226 | if (ip_vs_app_net_init(ipvs) < 0) |
7a4f0761 HS |
2227 | goto app_fail; |
2228 | ||
2f3edc6a | 2229 | if (ip_vs_conn_net_init(ipvs) < 0) |
7a4f0761 HS |
2230 | goto conn_fail; |
2231 | ||
802cb437 | 2232 | if (ip_vs_sync_net_init(ipvs) < 0) |
7a4f0761 HS |
2233 | goto sync_fail; |
2234 | ||
efe41606 FW |
2235 | ret = nf_register_net_hooks(net, ip_vs_ops, ARRAY_SIZE(ip_vs_ops)); |
2236 | if (ret < 0) | |
2237 | goto hook_fail; | |
2238 | ||
61b1ab45 | 2239 | return 0; |
7a4f0761 HS |
2240 | /* |
2241 | * Error handling | |
2242 | */ | |
2243 | ||
efe41606 FW |
2244 | hook_fail: |
2245 | ip_vs_sync_net_cleanup(ipvs); | |
7a4f0761 | 2246 | sync_fail: |
2f3edc6a | 2247 | ip_vs_conn_net_cleanup(ipvs); |
7a4f0761 | 2248 | conn_fail: |
b5dd212c | 2249 | ip_vs_app_net_cleanup(ipvs); |
7a4f0761 | 2250 | app_fail: |
7d1f88ec | 2251 | ip_vs_protocol_net_cleanup(ipvs); |
7a4f0761 | 2252 | protocol_fail: |
3d993766 | 2253 | ip_vs_control_net_cleanup(ipvs); |
7a4f0761 | 2254 | control_fail: |
a4dd0360 | 2255 | ip_vs_estimator_net_cleanup(ipvs); |
7a4f0761 | 2256 | estimator_fail: |
39f618b4 | 2257 | net->ipvs = NULL; |
7a4f0761 | 2258 | return -ENOMEM; |
61b1ab45 HS |
2259 | } |
2260 | ||
2261 | static void __net_exit __ip_vs_cleanup(struct net *net) | |
2262 | { | |
56d2169b EB |
2263 | struct netns_ipvs *ipvs = net_ipvs(net); |
2264 | ||
efe41606 | 2265 | nf_unregister_net_hooks(net, ip_vs_ops, ARRAY_SIZE(ip_vs_ops)); |
56d2169b | 2266 | ip_vs_service_net_cleanup(ipvs); /* ip_vs_flush() with locks */ |
2f3edc6a | 2267 | ip_vs_conn_net_cleanup(ipvs); |
b5dd212c | 2268 | ip_vs_app_net_cleanup(ipvs); |
7d1f88ec | 2269 | ip_vs_protocol_net_cleanup(ipvs); |
3d993766 | 2270 | ip_vs_control_net_cleanup(ipvs); |
a4dd0360 | 2271 | ip_vs_estimator_net_cleanup(ipvs); |
56d2169b | 2272 | IP_VS_DBG(2, "ipvs netns %d released\n", ipvs->gen); |
39f618b4 | 2273 | net->ipvs = NULL; |
61b1ab45 HS |
2274 | } |
2275 | ||
7a4f0761 HS |
2276 | static void __net_exit __ip_vs_dev_cleanup(struct net *net) |
2277 | { | |
ebea1f7c | 2278 | struct netns_ipvs *ipvs = net_ipvs(net); |
7a4f0761 | 2279 | EnterFunction(2); |
ebea1f7c | 2280 | ipvs->enable = 0; /* Disable packet reception */ |
8f4e0a18 | 2281 | smp_wmb(); |
ebea1f7c | 2282 | ip_vs_sync_net_cleanup(ipvs); |
7a4f0761 HS |
2283 | LeaveFunction(2); |
2284 | } | |
2285 | ||
61b1ab45 HS |
2286 | static struct pernet_operations ipvs_core_ops = { |
2287 | .init = __ip_vs_init, | |
2288 | .exit = __ip_vs_cleanup, | |
2289 | .id = &ip_vs_net_id, | |
2290 | .size = sizeof(struct netns_ipvs), | |
2291 | }; | |
1da177e4 | 2292 | |
7a4f0761 HS |
2293 | static struct pernet_operations ipvs_core_dev_ops = { |
2294 | .exit = __ip_vs_dev_cleanup, | |
2295 | }; | |
2296 | ||
1da177e4 LT |
2297 | /* |
2298 | * Initialize IP Virtual Server | |
2299 | */ | |
2300 | static int __init ip_vs_init(void) | |
2301 | { | |
2302 | int ret; | |
2303 | ||
2304 | ret = ip_vs_control_init(); | |
2305 | if (ret < 0) { | |
1e3e238e | 2306 | pr_err("can't setup control.\n"); |
6c8f7949 | 2307 | goto exit; |
1da177e4 LT |
2308 | } |
2309 | ||
2310 | ip_vs_protocol_init(); | |
2311 | ||
1da177e4 LT |
2312 | ret = ip_vs_conn_init(); |
2313 | if (ret < 0) { | |
1e3e238e | 2314 | pr_err("can't setup connection table.\n"); |
6c8f7949 | 2315 | goto cleanup_protocol; |
61b1ab45 HS |
2316 | } |
2317 | ||
7a4f0761 HS |
2318 | ret = register_pernet_subsys(&ipvs_core_ops); /* Alloc ip_vs struct */ |
2319 | if (ret < 0) | |
6c8f7949 | 2320 | goto cleanup_conn; |
7a4f0761 HS |
2321 | |
2322 | ret = register_pernet_device(&ipvs_core_dev_ops); | |
2323 | if (ret < 0) | |
2324 | goto cleanup_sub; | |
2325 | ||
8537de8a HS |
2326 | ret = ip_vs_register_nl_ioctl(); |
2327 | if (ret < 0) { | |
2328 | pr_err("can't register netlink/ioctl.\n"); | |
efe41606 | 2329 | goto cleanup_dev; |
8537de8a HS |
2330 | } |
2331 | ||
1e3e238e | 2332 | pr_info("ipvs loaded.\n"); |
7a4f0761 | 2333 | |
1da177e4 LT |
2334 | return ret; |
2335 | ||
7a4f0761 HS |
2336 | cleanup_dev: |
2337 | unregister_pernet_device(&ipvs_core_dev_ops); | |
2338 | cleanup_sub: | |
2339 | unregister_pernet_subsys(&ipvs_core_ops); | |
552ad65a | 2340 | cleanup_conn: |
1da177e4 | 2341 | ip_vs_conn_cleanup(); |
552ad65a | 2342 | cleanup_protocol: |
1da177e4 LT |
2343 | ip_vs_protocol_cleanup(); |
2344 | ip_vs_control_cleanup(); | |
6c8f7949 | 2345 | exit: |
1da177e4 LT |
2346 | return ret; |
2347 | } | |
2348 | ||
2349 | static void __exit ip_vs_cleanup(void) | |
2350 | { | |
8537de8a | 2351 | ip_vs_unregister_nl_ioctl(); |
7a4f0761 HS |
2352 | unregister_pernet_device(&ipvs_core_dev_ops); |
2353 | unregister_pernet_subsys(&ipvs_core_ops); /* free ip_vs struct */ | |
1da177e4 | 2354 | ip_vs_conn_cleanup(); |
1da177e4 LT |
2355 | ip_vs_protocol_cleanup(); |
2356 | ip_vs_control_cleanup(); | |
1e3e238e | 2357 | pr_info("ipvs unloaded.\n"); |
1da177e4 LT |
2358 | } |
2359 | ||
2360 | module_init(ip_vs_init); | |
2361 | module_exit(ip_vs_cleanup); | |
2362 | MODULE_LICENSE("GPL"); |