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 | |
1da177e4 | 71 | |
b962abdc | 72 | static int ip_vs_net_id __read_mostly; |
61b1ab45 HS |
73 | /* netns cnt used for uniqueness */ |
74 | static atomic_t ipvs_netns_cnt = ATOMIC_INIT(0); | |
1da177e4 LT |
75 | |
76 | /* ID used in ICMP lookups */ | |
77 | #define icmp_id(icmph) (((icmph)->un).echo.id) | |
2a3b791e | 78 | #define icmpv6_id(icmph) (icmph->icmp6_dataun.u_echo.identifier) |
1da177e4 | 79 | |
95c96174 | 80 | const char *ip_vs_proto_name(unsigned int proto) |
1da177e4 LT |
81 | { |
82 | static char buf[20]; | |
83 | ||
84 | switch (proto) { | |
85 | case IPPROTO_IP: | |
86 | return "IP"; | |
87 | case IPPROTO_UDP: | |
88 | return "UDP"; | |
89 | case IPPROTO_TCP: | |
90 | return "TCP"; | |
2906f66a VMR |
91 | case IPPROTO_SCTP: |
92 | return "SCTP"; | |
1da177e4 LT |
93 | case IPPROTO_ICMP: |
94 | return "ICMP"; | |
2a3b791e JV |
95 | #ifdef CONFIG_IP_VS_IPV6 |
96 | case IPPROTO_ICMPV6: | |
97 | return "ICMPv6"; | |
98 | #endif | |
1da177e4 LT |
99 | default: |
100 | sprintf(buf, "IP_%d", proto); | |
101 | return buf; | |
102 | } | |
103 | } | |
104 | ||
105 | void ip_vs_init_hash_table(struct list_head *table, int rows) | |
106 | { | |
107 | while (--rows >= 0) | |
108 | INIT_LIST_HEAD(&table[rows]); | |
109 | } | |
110 | ||
111 | static inline void | |
112 | ip_vs_in_stats(struct ip_vs_conn *cp, struct sk_buff *skb) | |
113 | { | |
114 | struct ip_vs_dest *dest = cp->dest; | |
b17fc996 HS |
115 | struct netns_ipvs *ipvs = net_ipvs(skb_net(skb)); |
116 | ||
1da177e4 | 117 | if (dest && (dest->flags & IP_VS_DEST_F_AVAILABLE)) { |
b17fc996 HS |
118 | struct ip_vs_cpu_stats *s; |
119 | ||
120 | s = this_cpu_ptr(dest->stats.cpustats); | |
121 | s->ustats.inpkts++; | |
122 | u64_stats_update_begin(&s->syncp); | |
123 | s->ustats.inbytes += skb->len; | |
124 | u64_stats_update_end(&s->syncp); | |
125 | ||
126 | s = this_cpu_ptr(dest->svc->stats.cpustats); | |
127 | s->ustats.inpkts++; | |
128 | u64_stats_update_begin(&s->syncp); | |
129 | s->ustats.inbytes += skb->len; | |
130 | u64_stats_update_end(&s->syncp); | |
131 | ||
2a0751af | 132 | s = this_cpu_ptr(ipvs->tot_stats.cpustats); |
b17fc996 HS |
133 | s->ustats.inpkts++; |
134 | u64_stats_update_begin(&s->syncp); | |
135 | s->ustats.inbytes += skb->len; | |
136 | u64_stats_update_end(&s->syncp); | |
1da177e4 LT |
137 | } |
138 | } | |
139 | ||
140 | ||
141 | static inline void | |
142 | ip_vs_out_stats(struct ip_vs_conn *cp, struct sk_buff *skb) | |
143 | { | |
144 | struct ip_vs_dest *dest = cp->dest; | |
b17fc996 HS |
145 | struct netns_ipvs *ipvs = net_ipvs(skb_net(skb)); |
146 | ||
1da177e4 | 147 | if (dest && (dest->flags & IP_VS_DEST_F_AVAILABLE)) { |
b17fc996 HS |
148 | struct ip_vs_cpu_stats *s; |
149 | ||
150 | s = this_cpu_ptr(dest->stats.cpustats); | |
151 | s->ustats.outpkts++; | |
152 | u64_stats_update_begin(&s->syncp); | |
153 | s->ustats.outbytes += skb->len; | |
154 | u64_stats_update_end(&s->syncp); | |
155 | ||
156 | s = this_cpu_ptr(dest->svc->stats.cpustats); | |
157 | s->ustats.outpkts++; | |
158 | u64_stats_update_begin(&s->syncp); | |
159 | s->ustats.outbytes += skb->len; | |
160 | u64_stats_update_end(&s->syncp); | |
161 | ||
2a0751af | 162 | s = this_cpu_ptr(ipvs->tot_stats.cpustats); |
b17fc996 HS |
163 | s->ustats.outpkts++; |
164 | u64_stats_update_begin(&s->syncp); | |
165 | s->ustats.outbytes += skb->len; | |
166 | u64_stats_update_end(&s->syncp); | |
1da177e4 LT |
167 | } |
168 | } | |
169 | ||
170 | ||
171 | static inline void | |
172 | ip_vs_conn_stats(struct ip_vs_conn *cp, struct ip_vs_service *svc) | |
173 | { | |
b17fc996 HS |
174 | struct netns_ipvs *ipvs = net_ipvs(svc->net); |
175 | struct ip_vs_cpu_stats *s; | |
1da177e4 | 176 | |
b17fc996 HS |
177 | s = this_cpu_ptr(cp->dest->stats.cpustats); |
178 | s->ustats.conns++; | |
1da177e4 | 179 | |
b17fc996 HS |
180 | s = this_cpu_ptr(svc->stats.cpustats); |
181 | s->ustats.conns++; | |
182 | ||
2a0751af | 183 | s = this_cpu_ptr(ipvs->tot_stats.cpustats); |
b17fc996 | 184 | s->ustats.conns++; |
1da177e4 LT |
185 | } |
186 | ||
187 | ||
4a516f11 | 188 | static inline void |
1da177e4 LT |
189 | ip_vs_set_state(struct ip_vs_conn *cp, int direction, |
190 | const struct sk_buff *skb, | |
9330419d | 191 | struct ip_vs_proto_data *pd) |
1da177e4 | 192 | { |
4a516f11 SH |
193 | if (likely(pd->pp->state_transition)) |
194 | pd->pp->state_transition(cp, direction, skb, pd); | |
1da177e4 LT |
195 | } |
196 | ||
a5959d53 | 197 | static inline int |
85999283 SH |
198 | ip_vs_conn_fill_param_persist(const struct ip_vs_service *svc, |
199 | struct sk_buff *skb, int protocol, | |
200 | const union nf_inet_addr *caddr, __be16 cport, | |
201 | const union nf_inet_addr *vaddr, __be16 vport, | |
202 | struct ip_vs_conn_param *p) | |
203 | { | |
6e67e586 HS |
204 | ip_vs_conn_fill_param(svc->net, svc->af, protocol, caddr, cport, vaddr, |
205 | vport, p); | |
85999283 SH |
206 | p->pe = svc->pe; |
207 | if (p->pe && p->pe->fill_param) | |
a5959d53 HS |
208 | return p->pe->fill_param(p, skb); |
209 | ||
210 | return 0; | |
85999283 | 211 | } |
1da177e4 | 212 | |
1da177e4 LT |
213 | /* |
214 | * IPVS persistent scheduling function | |
215 | * It creates a connection entry according to its template if exists, | |
216 | * or selects a server and creates a connection entry plus a template. | |
217 | * Locking: we are svc user (svc->refcnt), so we hold all dests too | |
218 | * Protocols supported: TCP, UDP | |
219 | */ | |
220 | static struct ip_vs_conn * | |
221 | ip_vs_sched_persist(struct ip_vs_service *svc, | |
d4383f04 JDB |
222 | struct sk_buff *skb, __be16 src_port, __be16 dst_port, |
223 | int *ignored, struct ip_vs_iphdr *iph) | |
1da177e4 LT |
224 | { |
225 | struct ip_vs_conn *cp = NULL; | |
1da177e4 LT |
226 | struct ip_vs_dest *dest; |
227 | struct ip_vs_conn *ct; | |
5b57a98c | 228 | __be16 dport = 0; /* destination port to forward */ |
3575792e | 229 | unsigned int flags; |
f11017ec | 230 | struct ip_vs_conn_param param; |
e0aac52e | 231 | const union nf_inet_addr fwmark = { .ip = htonl(svc->fwmark) }; |
28364a59 JV |
232 | union nf_inet_addr snet; /* source network of the client, |
233 | after masking */ | |
cd17f9ed | 234 | |
1da177e4 | 235 | /* Mask saddr with the netmask to adjust template granularity */ |
cd17f9ed JV |
236 | #ifdef CONFIG_IP_VS_IPV6 |
237 | if (svc->af == AF_INET6) | |
d4383f04 | 238 | ipv6_addr_prefix(&snet.in6, &iph->saddr.in6, svc->netmask); |
cd17f9ed JV |
239 | else |
240 | #endif | |
d4383f04 | 241 | snet.ip = iph->saddr.ip & svc->netmask; |
1da177e4 | 242 | |
cd17f9ed JV |
243 | IP_VS_DBG_BUF(6, "p-schedule: src %s:%u dest %s:%u " |
244 | "mnet %s\n", | |
d4383f04 JDB |
245 | IP_VS_DBG_ADDR(svc->af, &iph->saddr), ntohs(src_port), |
246 | IP_VS_DBG_ADDR(svc->af, &iph->daddr), ntohs(dst_port), | |
cd17f9ed | 247 | IP_VS_DBG_ADDR(svc->af, &snet)); |
1da177e4 LT |
248 | |
249 | /* | |
250 | * As far as we know, FTP is a very complicated network protocol, and | |
251 | * it uses control connection and data connections. For active FTP, | |
252 | * FTP server initialize data connection to the client, its source port | |
253 | * is often 20. For passive FTP, FTP server tells the clients the port | |
254 | * that it passively listens to, and the client issues the data | |
255 | * connection. In the tunneling or direct routing mode, the load | |
256 | * balancer is on the client-to-server half of connection, the port | |
257 | * number is unknown to the load balancer. So, a conn template like | |
258 | * <caddr, 0, vaddr, 0, daddr, 0> is created for persistent FTP | |
259 | * service, and a template like <caddr, 0, vaddr, vport, daddr, dport> | |
260 | * is created for other persistent services. | |
261 | */ | |
5b57a98c | 262 | { |
d4383f04 JDB |
263 | int protocol = iph->protocol; |
264 | const union nf_inet_addr *vaddr = &iph->daddr; | |
f11017ec SH |
265 | __be16 vport = 0; |
266 | ||
ce144f24 | 267 | if (dst_port == svc->port) { |
5b57a98c SH |
268 | /* non-FTP template: |
269 | * <protocol, caddr, 0, vaddr, vport, daddr, dport> | |
270 | * FTP template: | |
271 | * <protocol, caddr, 0, vaddr, 0, daddr, 0> | |
1da177e4 LT |
272 | */ |
273 | if (svc->port != FTPPORT) | |
ce144f24 | 274 | vport = dst_port; |
1da177e4 | 275 | } else { |
5b57a98c SH |
276 | /* Note: persistent fwmark-based services and |
277 | * persistent port zero service are handled here. | |
278 | * fwmark template: | |
279 | * <IPPROTO_IP,caddr,0,fwmark,0,daddr,0> | |
280 | * port zero template: | |
281 | * <protocol,caddr,0,vaddr,0,daddr,0> | |
1da177e4 | 282 | */ |
28364a59 | 283 | if (svc->fwmark) { |
5b57a98c SH |
284 | protocol = IPPROTO_IP; |
285 | vaddr = &fwmark; | |
286 | } | |
1da177e4 | 287 | } |
a5959d53 HS |
288 | /* return *ignored = -1 so NF_DROP can be used */ |
289 | if (ip_vs_conn_fill_param_persist(svc, skb, protocol, &snet, 0, | |
290 | vaddr, vport, ¶m) < 0) { | |
291 | *ignored = -1; | |
292 | return NULL; | |
293 | } | |
1da177e4 LT |
294 | } |
295 | ||
5b57a98c | 296 | /* Check if a template already exists */ |
f11017ec | 297 | ct = ip_vs_ct_in_get(¶m); |
5b57a98c | 298 | if (!ct || !ip_vs_check_template(ct)) { |
a5959d53 HS |
299 | /* |
300 | * No template found or the dest of the connection | |
5b57a98c | 301 | * template is not available. |
a5959d53 | 302 | * return *ignored=0 i.e. ICMP and NF_DROP |
5b57a98c SH |
303 | */ |
304 | dest = svc->scheduler->schedule(svc, skb); | |
305 | if (!dest) { | |
306 | IP_VS_DBG(1, "p-schedule: no dest found.\n"); | |
85999283 | 307 | kfree(param.pe_data); |
a5959d53 | 308 | *ignored = 0; |
5b57a98c SH |
309 | return NULL; |
310 | } | |
311 | ||
ce144f24 | 312 | if (dst_port == svc->port && svc->port != FTPPORT) |
5b57a98c SH |
313 | dport = dest->port; |
314 | ||
85999283 SH |
315 | /* Create a template |
316 | * This adds param.pe_data to the template, | |
317 | * and thus param.pe_data will be destroyed | |
318 | * when the template expires */ | |
f11017ec | 319 | ct = ip_vs_conn_new(¶m, &dest->addr, dport, |
0e051e68 | 320 | IP_VS_CONN_F_TEMPLATE, dest, skb->mark); |
85999283 SH |
321 | if (ct == NULL) { |
322 | kfree(param.pe_data); | |
a5959d53 | 323 | *ignored = -1; |
5b57a98c | 324 | return NULL; |
85999283 | 325 | } |
5b57a98c SH |
326 | |
327 | ct->timeout = svc->timeout; | |
85999283 | 328 | } else { |
5b57a98c SH |
329 | /* set destination with the found template */ |
330 | dest = ct->dest; | |
85999283 SH |
331 | kfree(param.pe_data); |
332 | } | |
5b57a98c | 333 | |
ce144f24 | 334 | dport = dst_port; |
5b57a98c SH |
335 | if (dport == svc->port && dest->port) |
336 | dport = dest->port; | |
337 | ||
26ec037f | 338 | flags = (svc->flags & IP_VS_SVC_F_ONEPACKET |
d4383f04 | 339 | && iph->protocol == IPPROTO_UDP) ? |
26ec037f NC |
340 | IP_VS_CONN_F_ONE_PACKET : 0; |
341 | ||
1da177e4 LT |
342 | /* |
343 | * Create a new connection according to the template | |
344 | */ | |
d4383f04 JDB |
345 | ip_vs_conn_fill_param(svc->net, svc->af, iph->protocol, &iph->saddr, |
346 | src_port, &iph->daddr, dst_port, ¶m); | |
ce144f24 | 347 | |
0e051e68 | 348 | cp = ip_vs_conn_new(¶m, &dest->addr, dport, flags, dest, skb->mark); |
1da177e4 LT |
349 | if (cp == NULL) { |
350 | ip_vs_conn_put(ct); | |
a5959d53 | 351 | *ignored = -1; |
1da177e4 LT |
352 | return NULL; |
353 | } | |
354 | ||
355 | /* | |
356 | * Add its control | |
357 | */ | |
358 | ip_vs_control_add(cp, ct); | |
359 | ip_vs_conn_put(ct); | |
360 | ||
361 | ip_vs_conn_stats(cp, svc); | |
362 | return cp; | |
363 | } | |
364 | ||
365 | ||
366 | /* | |
367 | * IPVS main scheduling function | |
368 | * It selects a server according to the virtual service, and | |
369 | * creates a connection entry. | |
370 | * Protocols supported: TCP, UDP | |
a5959d53 HS |
371 | * |
372 | * Usage of *ignored | |
373 | * | |
374 | * 1 : protocol tried to schedule (eg. on SYN), found svc but the | |
375 | * svc/scheduler decides that this packet should be accepted with | |
376 | * NF_ACCEPT because it must not be scheduled. | |
377 | * | |
378 | * 0 : scheduler can not find destination, so try bypass or | |
379 | * return ICMP and then NF_DROP (ip_vs_leave). | |
380 | * | |
381 | * -1 : scheduler tried to schedule but fatal error occurred, eg. | |
382 | * ip_vs_conn_new failure (ENOMEM) or ip_vs_sip_fill_param | |
383 | * failure such as missing Call-ID, ENOMEM on skb_linearize | |
384 | * or pe_data. In this case we should return NF_DROP without | |
385 | * any attempts to send ICMP with ip_vs_leave. | |
1da177e4 LT |
386 | */ |
387 | struct ip_vs_conn * | |
190ecd27 | 388 | ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb, |
d4383f04 JDB |
389 | struct ip_vs_proto_data *pd, int *ignored, |
390 | struct ip_vs_iphdr *iph) | |
1da177e4 | 391 | { |
9330419d | 392 | struct ip_vs_protocol *pp = pd->pp; |
1da177e4 | 393 | struct ip_vs_conn *cp = NULL; |
1da177e4 | 394 | struct ip_vs_dest *dest; |
3575792e JA |
395 | __be16 _ports[2], *pptr; |
396 | unsigned int flags; | |
1da177e4 | 397 | |
190ecd27 | 398 | *ignored = 1; |
2f74713d JDB |
399 | /* |
400 | * IPv6 frags, only the first hit here. | |
401 | */ | |
d4383f04 | 402 | pptr = frag_safe_skb_hp(skb, iph->len, sizeof(_ports), _ports, iph); |
1da177e4 LT |
403 | if (pptr == NULL) |
404 | return NULL; | |
405 | ||
190ecd27 JA |
406 | /* |
407 | * FTPDATA needs this check when using local real server. | |
408 | * Never schedule Active FTPDATA connections from real server. | |
409 | * For LVS-NAT they must be already created. For other methods | |
410 | * with persistence the connection is created on SYN+ACK. | |
411 | */ | |
412 | if (pptr[0] == FTPDATA) { | |
0d79641a JA |
413 | IP_VS_DBG_PKT(12, svc->af, pp, skb, 0, |
414 | "Not scheduling FTPDATA"); | |
190ecd27 JA |
415 | return NULL; |
416 | } | |
417 | ||
418 | /* | |
a5959d53 | 419 | * Do not schedule replies from local real server. |
190ecd27 JA |
420 | */ |
421 | if ((!skb->dev || skb->dev->flags & IFF_LOOPBACK) && | |
d4383f04 | 422 | (cp = pp->conn_in_get(svc->af, skb, iph, 1))) { |
0d79641a | 423 | IP_VS_DBG_PKT(12, svc->af, pp, skb, 0, |
190ecd27 JA |
424 | "Not scheduling reply for existing connection"); |
425 | __ip_vs_conn_put(cp); | |
426 | return NULL; | |
427 | } | |
428 | ||
1da177e4 LT |
429 | /* |
430 | * Persistent service | |
431 | */ | |
a5959d53 | 432 | if (svc->flags & IP_VS_SVC_F_PERSISTENT) |
d4383f04 JDB |
433 | return ip_vs_sched_persist(svc, skb, pptr[0], pptr[1], ignored, |
434 | iph); | |
a5959d53 HS |
435 | |
436 | *ignored = 0; | |
1da177e4 LT |
437 | |
438 | /* | |
439 | * Non-persistent service | |
440 | */ | |
441 | if (!svc->fwmark && pptr[1] != svc->port) { | |
442 | if (!svc->port) | |
1e3e238e HE |
443 | pr_err("Schedule: port zero only supported " |
444 | "in persistent services, " | |
445 | "check your ipvs configuration\n"); | |
1da177e4 LT |
446 | return NULL; |
447 | } | |
448 | ||
449 | dest = svc->scheduler->schedule(svc, skb); | |
450 | if (dest == NULL) { | |
451 | IP_VS_DBG(1, "Schedule: no dest found.\n"); | |
452 | return NULL; | |
453 | } | |
454 | ||
26ec037f | 455 | flags = (svc->flags & IP_VS_SVC_F_ONEPACKET |
d4383f04 | 456 | && iph->protocol == IPPROTO_UDP) ? |
26ec037f NC |
457 | IP_VS_CONN_F_ONE_PACKET : 0; |
458 | ||
1da177e4 LT |
459 | /* |
460 | * Create a connection entry. | |
461 | */ | |
f11017ec SH |
462 | { |
463 | struct ip_vs_conn_param p; | |
6e67e586 | 464 | |
d4383f04 JDB |
465 | ip_vs_conn_fill_param(svc->net, svc->af, iph->protocol, |
466 | &iph->saddr, pptr[0], &iph->daddr, | |
467 | pptr[1], &p); | |
f11017ec SH |
468 | cp = ip_vs_conn_new(&p, &dest->addr, |
469 | dest->port ? dest->port : pptr[1], | |
0e051e68 | 470 | flags, dest, skb->mark); |
a5959d53 HS |
471 | if (!cp) { |
472 | *ignored = -1; | |
f11017ec | 473 | return NULL; |
a5959d53 | 474 | } |
f11017ec | 475 | } |
1da177e4 | 476 | |
cd17f9ed JV |
477 | IP_VS_DBG_BUF(6, "Schedule fwd:%c c:%s:%u v:%s:%u " |
478 | "d:%s:%u conn->flags:%X conn->refcnt:%d\n", | |
479 | ip_vs_fwd_tag(cp), | |
480 | IP_VS_DBG_ADDR(svc->af, &cp->caddr), ntohs(cp->cport), | |
481 | IP_VS_DBG_ADDR(svc->af, &cp->vaddr), ntohs(cp->vport), | |
482 | IP_VS_DBG_ADDR(svc->af, &cp->daddr), ntohs(cp->dport), | |
483 | cp->flags, atomic_read(&cp->refcnt)); | |
1da177e4 LT |
484 | |
485 | ip_vs_conn_stats(cp, svc); | |
486 | return cp; | |
487 | } | |
488 | ||
489 | ||
490 | /* | |
491 | * Pass or drop the packet. | |
492 | * Called by ip_vs_in, when the virtual service is available but | |
493 | * no destination is available for a new connection. | |
494 | */ | |
495 | int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb, | |
d4383f04 | 496 | struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph) |
1da177e4 | 497 | { |
014d730d | 498 | __be16 _ports[2], *pptr; |
a7a86b86 SH |
499 | #ifdef CONFIG_SYSCTL |
500 | struct net *net; | |
501 | struct netns_ipvs *ipvs; | |
2a3b791e | 502 | int unicast; |
a7a86b86 | 503 | #endif |
9330419d | 504 | |
d4383f04 | 505 | pptr = frag_safe_skb_hp(skb, iph->len, sizeof(_ports), _ports, iph); |
1da177e4 LT |
506 | if (pptr == NULL) { |
507 | ip_vs_service_put(svc); | |
508 | return NF_DROP; | |
509 | } | |
a7a86b86 SH |
510 | |
511 | #ifdef CONFIG_SYSCTL | |
4a98480b | 512 | net = skb_net(skb); |
1da177e4 | 513 | |
2a3b791e JV |
514 | #ifdef CONFIG_IP_VS_IPV6 |
515 | if (svc->af == AF_INET6) | |
d4383f04 | 516 | unicast = ipv6_addr_type(&iph->daddr.in6) & IPV6_ADDR_UNICAST; |
2a3b791e JV |
517 | else |
518 | #endif | |
d4383f04 | 519 | unicast = (inet_addr_type(net, iph->daddr.ip) == RTN_UNICAST); |
2a3b791e | 520 | |
1da177e4 | 521 | /* if it is fwmark-based service, the cache_bypass sysctl is up |
2a3b791e | 522 | and the destination is a non-local unicast, then create |
1da177e4 | 523 | a cache_bypass connection entry */ |
4a98480b | 524 | ipvs = net_ipvs(net); |
a0840e2e | 525 | if (ipvs->sysctl_cache_bypass && svc->fwmark && unicast) { |
ad542ced | 526 | int ret; |
1da177e4 | 527 | struct ip_vs_conn *cp; |
3575792e | 528 | unsigned int flags = (svc->flags & IP_VS_SVC_F_ONEPACKET && |
d4383f04 | 529 | iph->protocol == IPPROTO_UDP) ? |
3575792e | 530 | IP_VS_CONN_F_ONE_PACKET : 0; |
dff630dd | 531 | union nf_inet_addr daddr = { .all = { 0, 0, 0, 0 } }; |
1da177e4 LT |
532 | |
533 | ip_vs_service_put(svc); | |
534 | ||
535 | /* create a new connection entry */ | |
1e3e238e | 536 | IP_VS_DBG(6, "%s(): create a cache_bypass entry\n", __func__); |
f11017ec SH |
537 | { |
538 | struct ip_vs_conn_param p; | |
d4383f04 JDB |
539 | ip_vs_conn_fill_param(svc->net, svc->af, iph->protocol, |
540 | &iph->saddr, pptr[0], | |
541 | &iph->daddr, pptr[1], &p); | |
f11017ec SH |
542 | cp = ip_vs_conn_new(&p, &daddr, 0, |
543 | IP_VS_CONN_F_BYPASS | flags, | |
0e051e68 | 544 | NULL, skb->mark); |
f11017ec SH |
545 | if (!cp) |
546 | return NF_DROP; | |
547 | } | |
1da177e4 LT |
548 | |
549 | /* statistics */ | |
550 | ip_vs_in_stats(cp, skb); | |
551 | ||
552 | /* set state */ | |
4a516f11 | 553 | ip_vs_set_state(cp, IP_VS_DIR_INPUT, skb, pd); |
1da177e4 LT |
554 | |
555 | /* transmit the first SYN packet */ | |
d4383f04 | 556 | ret = cp->packet_xmit(skb, cp, pd->pp, iph); |
1da177e4 LT |
557 | /* do not touch skb anymore */ |
558 | ||
559 | atomic_inc(&cp->in_pkts); | |
560 | ip_vs_conn_put(cp); | |
561 | return ret; | |
562 | } | |
a7a86b86 | 563 | #endif |
1da177e4 LT |
564 | |
565 | /* | |
566 | * When the virtual ftp service is presented, packets destined | |
567 | * for other services on the VIP may get here (except services | |
568 | * listed in the ipvs table), pass the packets, because it is | |
569 | * not ipvs job to decide to drop the packets. | |
570 | */ | |
571 | if ((svc->port == FTPPORT) && (pptr[1] != FTPPORT)) { | |
572 | ip_vs_service_put(svc); | |
573 | return NF_ACCEPT; | |
574 | } | |
575 | ||
576 | ip_vs_service_put(svc); | |
577 | ||
578 | /* | |
579 | * Notify the client that the destination is unreachable, and | |
580 | * release the socket buffer. | |
581 | * Since it is in IP layer, the TCP socket is not actually | |
582 | * created, the TCP RST packet cannot be sent, instead that | |
583 | * ICMP_PORT_UNREACH is sent here no matter it is TCP/UDP. --WZ | |
584 | */ | |
2a3b791e | 585 | #ifdef CONFIG_IP_VS_IPV6 |
cb59155f JA |
586 | if (svc->af == AF_INET6) { |
587 | if (!skb->dev) { | |
588 | struct net *net = dev_net(skb_dst(skb)->dev); | |
589 | ||
590 | skb->dev = net->loopback_dev; | |
591 | } | |
3ffe533c | 592 | icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); |
cb59155f | 593 | } else |
2a3b791e JV |
594 | #endif |
595 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); | |
596 | ||
1da177e4 LT |
597 | return NF_DROP; |
598 | } | |
599 | ||
84b3cee3 SH |
600 | #ifdef CONFIG_SYSCTL |
601 | ||
602 | static int sysctl_snat_reroute(struct sk_buff *skb) | |
603 | { | |
604 | struct netns_ipvs *ipvs = net_ipvs(skb_net(skb)); | |
605 | return ipvs->sysctl_snat_reroute; | |
606 | } | |
607 | ||
0cfa558e SH |
608 | static int sysctl_nat_icmp_send(struct net *net) |
609 | { | |
610 | struct netns_ipvs *ipvs = net_ipvs(net); | |
611 | return ipvs->sysctl_nat_icmp_send; | |
612 | } | |
613 | ||
71a8ab6c SH |
614 | static int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs) |
615 | { | |
616 | return ipvs->sysctl_expire_nodest_conn; | |
617 | } | |
618 | ||
84b3cee3 SH |
619 | #else |
620 | ||
621 | static int sysctl_snat_reroute(struct sk_buff *skb) { return 0; } | |
0cfa558e | 622 | static int sysctl_nat_icmp_send(struct net *net) { return 0; } |
71a8ab6c | 623 | static int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs) { return 0; } |
84b3cee3 SH |
624 | |
625 | #endif | |
626 | ||
b1550f22 | 627 | __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset) |
1da177e4 | 628 | { |
d3bc23e7 | 629 | return csum_fold(skb_checksum(skb, offset, skb->len - offset, 0)); |
1da177e4 LT |
630 | } |
631 | ||
1ca5bb54 JA |
632 | static inline enum ip_defrag_users ip_vs_defrag_user(unsigned int hooknum) |
633 | { | |
634 | if (NF_INET_LOCAL_IN == hooknum) | |
635 | return IP_DEFRAG_VS_IN; | |
636 | if (NF_INET_FORWARD == hooknum) | |
637 | return IP_DEFRAG_VS_FWD; | |
638 | return IP_DEFRAG_VS_OUT; | |
639 | } | |
640 | ||
776c729e | 641 | static inline int ip_vs_gather_frags(struct sk_buff *skb, u_int32_t user) |
1da177e4 | 642 | { |
776c729e HX |
643 | int err = ip_defrag(skb, user); |
644 | ||
645 | if (!err) | |
eddc9ec5 | 646 | ip_send_check(ip_hdr(skb)); |
776c729e HX |
647 | |
648 | return err; | |
1da177e4 LT |
649 | } |
650 | ||
ba4fd7e9 SH |
651 | static int ip_vs_route_me_harder(int af, struct sk_buff *skb) |
652 | { | |
ba4fd7e9 SH |
653 | #ifdef CONFIG_IP_VS_IPV6 |
654 | if (af == AF_INET6) { | |
84b3cee3 | 655 | if (sysctl_snat_reroute(skb) && ip6_route_me_harder(skb) != 0) |
ba4fd7e9 SH |
656 | return 1; |
657 | } else | |
658 | #endif | |
84b3cee3 | 659 | if ((sysctl_snat_reroute(skb) || |
ba4fd7e9 SH |
660 | skb_rtable(skb)->rt_flags & RTCF_LOCAL) && |
661 | ip_route_me_harder(skb, RTN_LOCAL) != 0) | |
662 | return 1; | |
663 | ||
664 | return 0; | |
665 | } | |
666 | ||
1da177e4 LT |
667 | /* |
668 | * Packet has been made sufficiently writable in caller | |
669 | * - inout: 1=in->out, 0=out->in | |
670 | */ | |
671 | void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp, | |
672 | struct ip_vs_conn *cp, int inout) | |
673 | { | |
eddc9ec5 | 674 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 | 675 | unsigned int icmp_offset = iph->ihl*4; |
d56f90a7 ACM |
676 | struct icmphdr *icmph = (struct icmphdr *)(skb_network_header(skb) + |
677 | icmp_offset); | |
1da177e4 LT |
678 | struct iphdr *ciph = (struct iphdr *)(icmph + 1); |
679 | ||
680 | if (inout) { | |
e7ade46a | 681 | iph->saddr = cp->vaddr.ip; |
1da177e4 | 682 | ip_send_check(iph); |
e7ade46a | 683 | ciph->daddr = cp->vaddr.ip; |
1da177e4 LT |
684 | ip_send_check(ciph); |
685 | } else { | |
e7ade46a | 686 | iph->daddr = cp->daddr.ip; |
1da177e4 | 687 | ip_send_check(iph); |
e7ade46a | 688 | ciph->saddr = cp->daddr.ip; |
1da177e4 LT |
689 | ip_send_check(ciph); |
690 | } | |
691 | ||
2906f66a VMR |
692 | /* the TCP/UDP/SCTP port */ |
693 | if (IPPROTO_TCP == ciph->protocol || IPPROTO_UDP == ciph->protocol || | |
694 | IPPROTO_SCTP == ciph->protocol) { | |
014d730d | 695 | __be16 *ports = (void *)ciph + ciph->ihl*4; |
1da177e4 LT |
696 | |
697 | if (inout) | |
698 | ports[1] = cp->vport; | |
699 | else | |
700 | ports[0] = cp->dport; | |
701 | } | |
702 | ||
703 | /* And finally the ICMP checksum */ | |
704 | icmph->checksum = 0; | |
705 | icmph->checksum = ip_vs_checksum_complete(skb, icmp_offset); | |
706 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
707 | ||
708 | if (inout) | |
0d79641a | 709 | IP_VS_DBG_PKT(11, AF_INET, pp, skb, (void *)ciph - (void *)iph, |
1da177e4 LT |
710 | "Forwarding altered outgoing ICMP"); |
711 | else | |
0d79641a | 712 | IP_VS_DBG_PKT(11, AF_INET, pp, skb, (void *)ciph - (void *)iph, |
1da177e4 LT |
713 | "Forwarding altered incoming ICMP"); |
714 | } | |
715 | ||
b3cdd2a7 JV |
716 | #ifdef CONFIG_IP_VS_IPV6 |
717 | void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp, | |
718 | struct ip_vs_conn *cp, int inout) | |
719 | { | |
720 | struct ipv6hdr *iph = ipv6_hdr(skb); | |
63dca2c0 JDB |
721 | unsigned int icmp_offset = 0; |
722 | unsigned int offs = 0; /* header offset*/ | |
723 | int protocol; | |
724 | struct icmp6hdr *icmph; | |
725 | struct ipv6hdr *ciph; | |
726 | unsigned short fragoffs; | |
727 | ||
728 | ipv6_find_hdr(skb, &icmp_offset, IPPROTO_ICMPV6, &fragoffs, NULL); | |
729 | icmph = (struct icmp6hdr *)(skb_network_header(skb) + icmp_offset); | |
730 | offs = icmp_offset + sizeof(struct icmp6hdr); | |
731 | ciph = (struct ipv6hdr *)(skb_network_header(skb) + offs); | |
732 | ||
733 | protocol = ipv6_find_hdr(skb, &offs, -1, &fragoffs, NULL); | |
b3cdd2a7 JV |
734 | |
735 | if (inout) { | |
736 | iph->saddr = cp->vaddr.in6; | |
737 | ciph->daddr = cp->vaddr.in6; | |
738 | } else { | |
739 | iph->daddr = cp->daddr.in6; | |
740 | ciph->saddr = cp->daddr.in6; | |
741 | } | |
742 | ||
2906f66a | 743 | /* the TCP/UDP/SCTP port */ |
63dca2c0 JDB |
744 | if (!fragoffs && (IPPROTO_TCP == protocol || IPPROTO_UDP == protocol || |
745 | IPPROTO_SCTP == protocol)) { | |
746 | __be16 *ports = (void *)(skb_network_header(skb) + offs); | |
b3cdd2a7 | 747 | |
63dca2c0 JDB |
748 | IP_VS_DBG(11, "%s() changed port %d to %d\n", __func__, |
749 | ntohs(inout ? ports[1] : ports[0]), | |
750 | ntohs(inout ? cp->vport : cp->dport)); | |
b3cdd2a7 JV |
751 | if (inout) |
752 | ports[1] = cp->vport; | |
753 | else | |
754 | ports[0] = cp->dport; | |
755 | } | |
756 | ||
757 | /* And finally the ICMP checksum */ | |
8870f842 SH |
758 | icmph->icmp6_cksum = ~csum_ipv6_magic(&iph->saddr, &iph->daddr, |
759 | skb->len - icmp_offset, | |
760 | IPPROTO_ICMPV6, 0); | |
761 | skb->csum_start = skb_network_header(skb) - skb->head + icmp_offset; | |
762 | skb->csum_offset = offsetof(struct icmp6hdr, icmp6_cksum); | |
763 | skb->ip_summed = CHECKSUM_PARTIAL; | |
b3cdd2a7 JV |
764 | |
765 | if (inout) | |
0d79641a JA |
766 | IP_VS_DBG_PKT(11, AF_INET6, pp, skb, |
767 | (void *)ciph - (void *)iph, | |
768 | "Forwarding altered outgoing ICMPv6"); | |
b3cdd2a7 | 769 | else |
0d79641a JA |
770 | IP_VS_DBG_PKT(11, AF_INET6, pp, skb, |
771 | (void *)ciph - (void *)iph, | |
772 | "Forwarding altered incoming ICMPv6"); | |
b3cdd2a7 JV |
773 | } |
774 | #endif | |
775 | ||
4856c84c | 776 | /* Handle relevant response ICMP messages - forward to the right |
6cb90db5 | 777 | * destination host. |
4856c84c | 778 | */ |
f2428ed5 SH |
779 | static int handle_response_icmp(int af, struct sk_buff *skb, |
780 | union nf_inet_addr *snet, | |
781 | __u8 protocol, struct ip_vs_conn *cp, | |
4856c84c MT |
782 | struct ip_vs_protocol *pp, |
783 | unsigned int offset, unsigned int ihl) | |
784 | { | |
785 | unsigned int verdict = NF_DROP; | |
786 | ||
787 | if (IP_VS_FWD_METHOD(cp) != 0) { | |
1e3e238e HE |
788 | pr_err("shouldn't reach here, because the box is on the " |
789 | "half connection in the tun/dr module.\n"); | |
4856c84c MT |
790 | } |
791 | ||
792 | /* Ensure the checksum is correct */ | |
793 | if (!skb_csum_unnecessary(skb) && ip_vs_checksum_complete(skb, ihl)) { | |
794 | /* Failed checksum! */ | |
f2428ed5 SH |
795 | IP_VS_DBG_BUF(1, "Forward ICMP: failed checksum from %s!\n", |
796 | IP_VS_DBG_ADDR(af, snet)); | |
4856c84c MT |
797 | goto out; |
798 | } | |
799 | ||
2906f66a VMR |
800 | if (IPPROTO_TCP == protocol || IPPROTO_UDP == protocol || |
801 | IPPROTO_SCTP == protocol) | |
4856c84c MT |
802 | offset += 2 * sizeof(__u16); |
803 | if (!skb_make_writable(skb, offset)) | |
804 | goto out; | |
805 | ||
f2428ed5 SH |
806 | #ifdef CONFIG_IP_VS_IPV6 |
807 | if (af == AF_INET6) | |
808 | ip_vs_nat_icmp_v6(skb, pp, cp, 1); | |
809 | else | |
810 | #endif | |
811 | ip_vs_nat_icmp(skb, pp, cp, 1); | |
4856c84c | 812 | |
ba4fd7e9 SH |
813 | if (ip_vs_route_me_harder(af, skb)) |
814 | goto out; | |
f5a41847 | 815 | |
4856c84c MT |
816 | /* do the statistics and put it back */ |
817 | ip_vs_out_stats(cp, skb); | |
818 | ||
cf356d69 | 819 | skb->ipvs_property = 1; |
f4bc17cd | 820 | if (!(cp->flags & IP_VS_CONN_F_NFCT)) |
cf356d69 | 821 | ip_vs_notrack(skb); |
f4bc17cd JA |
822 | else |
823 | ip_vs_update_conntrack(skb, cp, 0); | |
4856c84c MT |
824 | verdict = NF_ACCEPT; |
825 | ||
826 | out: | |
827 | __ip_vs_conn_put(cp); | |
828 | ||
829 | return verdict; | |
830 | } | |
831 | ||
1da177e4 LT |
832 | /* |
833 | * Handle ICMP messages in the inside-to-outside direction (outgoing). | |
4856c84c | 834 | * Find any that might be relevant, check against existing connections. |
1da177e4 | 835 | * Currently handles error types - unreachable, quench, ttl exceeded. |
1da177e4 | 836 | */ |
1ca5bb54 JA |
837 | static int ip_vs_out_icmp(struct sk_buff *skb, int *related, |
838 | unsigned int hooknum) | |
1da177e4 | 839 | { |
1da177e4 LT |
840 | struct iphdr *iph; |
841 | struct icmphdr _icmph, *ic; | |
842 | struct iphdr _ciph, *cih; /* The ip header contained within the ICMP */ | |
51ef348b | 843 | struct ip_vs_iphdr ciph; |
1da177e4 LT |
844 | struct ip_vs_conn *cp; |
845 | struct ip_vs_protocol *pp; | |
4856c84c | 846 | unsigned int offset, ihl; |
f2428ed5 | 847 | union nf_inet_addr snet; |
1da177e4 LT |
848 | |
849 | *related = 1; | |
850 | ||
851 | /* reassemble IP fragments */ | |
56f8a75c | 852 | if (ip_is_fragment(ip_hdr(skb))) { |
1ca5bb54 | 853 | if (ip_vs_gather_frags(skb, ip_vs_defrag_user(hooknum))) |
1da177e4 | 854 | return NF_STOLEN; |
1da177e4 LT |
855 | } |
856 | ||
eddc9ec5 | 857 | iph = ip_hdr(skb); |
1da177e4 LT |
858 | offset = ihl = iph->ihl * 4; |
859 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
860 | if (ic == NULL) | |
861 | return NF_DROP; | |
862 | ||
14d5e834 | 863 | IP_VS_DBG(12, "Outgoing ICMP (%d,%d) %pI4->%pI4\n", |
1da177e4 | 864 | ic->type, ntohs(icmp_id(ic)), |
14d5e834 | 865 | &iph->saddr, &iph->daddr); |
1da177e4 LT |
866 | |
867 | /* | |
868 | * Work through seeing if this is for us. | |
869 | * These checks are supposed to be in an order that means easy | |
870 | * things are checked first to speed up processing.... however | |
871 | * this means that some packets will manage to get a long way | |
872 | * down this stack and then be rejected, but that's life. | |
873 | */ | |
874 | if ((ic->type != ICMP_DEST_UNREACH) && | |
875 | (ic->type != ICMP_SOURCE_QUENCH) && | |
876 | (ic->type != ICMP_TIME_EXCEEDED)) { | |
877 | *related = 0; | |
878 | return NF_ACCEPT; | |
879 | } | |
880 | ||
881 | /* Now find the contained IP header */ | |
882 | offset += sizeof(_icmph); | |
883 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
884 | if (cih == NULL) | |
885 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
886 | ||
887 | pp = ip_vs_proto_get(cih->protocol); | |
888 | if (!pp) | |
889 | return NF_ACCEPT; | |
890 | ||
891 | /* Is the embedded protocol header present? */ | |
4412ec49 | 892 | if (unlikely(cih->frag_off & htons(IP_OFFSET) && |
1da177e4 LT |
893 | pp->dont_defrag)) |
894 | return NF_ACCEPT; | |
895 | ||
0d79641a JA |
896 | IP_VS_DBG_PKT(11, AF_INET, pp, skb, offset, |
897 | "Checking outgoing ICMP for"); | |
1da177e4 | 898 | |
63dca2c0 JDB |
899 | ip_vs_fill_ip4hdr(cih, &ciph); |
900 | ciph.len += offset; | |
1da177e4 | 901 | /* The embedded headers contain source and dest in reverse order */ |
d4383f04 | 902 | cp = pp->conn_out_get(AF_INET, skb, &ciph, 1); |
1da177e4 LT |
903 | if (!cp) |
904 | return NF_ACCEPT; | |
905 | ||
f2428ed5 SH |
906 | snet.ip = iph->saddr; |
907 | return handle_response_icmp(AF_INET, skb, &snet, cih->protocol, cp, | |
63dca2c0 | 908 | pp, ciph.len, ihl); |
1da177e4 LT |
909 | } |
910 | ||
2a3b791e | 911 | #ifdef CONFIG_IP_VS_IPV6 |
1ca5bb54 | 912 | static int ip_vs_out_icmp_v6(struct sk_buff *skb, int *related, |
d4383f04 | 913 | unsigned int hooknum, struct ip_vs_iphdr *ipvsh) |
2a3b791e | 914 | { |
2a3b791e | 915 | struct icmp6hdr _icmph, *ic; |
63dca2c0 JDB |
916 | struct ipv6hdr _ip6h, *ip6h; /* The ip header contained within ICMP */ |
917 | struct ip_vs_iphdr ciph = {.flags = 0, .fragoffs = 0};/*Contained IP */ | |
2a3b791e JV |
918 | struct ip_vs_conn *cp; |
919 | struct ip_vs_protocol *pp; | |
f2428ed5 | 920 | union nf_inet_addr snet; |
63dca2c0 | 921 | unsigned int writable; |
2a3b791e | 922 | |
63dca2c0 | 923 | *related = 1; |
2f74713d | 924 | ic = frag_safe_skb_hp(skb, ipvsh->len, sizeof(_icmph), &_icmph, ipvsh); |
2a3b791e JV |
925 | if (ic == NULL) |
926 | return NF_DROP; | |
927 | ||
2a3b791e JV |
928 | /* |
929 | * Work through seeing if this is for us. | |
930 | * These checks are supposed to be in an order that means easy | |
931 | * things are checked first to speed up processing.... however | |
932 | * this means that some packets will manage to get a long way | |
933 | * down this stack and then be rejected, but that's life. | |
934 | */ | |
2fab8917 | 935 | if (ic->icmp6_type & ICMPV6_INFOMSG_MASK) { |
2a3b791e JV |
936 | *related = 0; |
937 | return NF_ACCEPT; | |
938 | } | |
2f74713d JDB |
939 | /* Fragment header that is before ICMP header tells us that: |
940 | * it's not an error message since they can't be fragmented. | |
941 | */ | |
e7165030 | 942 | if (ipvsh->flags & IP6_FH_F_FRAG) |
2f74713d | 943 | return NF_DROP; |
2a3b791e | 944 | |
63dca2c0 JDB |
945 | IP_VS_DBG(8, "Outgoing ICMPv6 (%d,%d) %pI6c->%pI6c\n", |
946 | ic->icmp6_type, ntohs(icmpv6_id(ic)), | |
947 | &ipvsh->saddr, &ipvsh->daddr); | |
948 | ||
2a3b791e | 949 | /* Now find the contained IP header */ |
63dca2c0 JDB |
950 | ciph.len = ipvsh->len + sizeof(_icmph); |
951 | ip6h = skb_header_pointer(skb, ciph.len, sizeof(_ip6h), &_ip6h); | |
952 | if (ip6h == NULL) | |
2a3b791e | 953 | return NF_ACCEPT; /* The packet looks wrong, ignore */ |
63dca2c0 JDB |
954 | ciph.saddr.in6 = ip6h->saddr; /* conn_out_get() handles reverse order */ |
955 | ciph.daddr.in6 = ip6h->daddr; | |
956 | /* skip possible IPv6 exthdrs of contained IPv6 packet */ | |
957 | ciph.protocol = ipv6_find_hdr(skb, &ciph.len, -1, &ciph.fragoffs, NULL); | |
958 | if (ciph.protocol < 0) | |
959 | return NF_ACCEPT; /* Contained IPv6 hdr looks wrong, ignore */ | |
960 | ||
961 | pp = ip_vs_proto_get(ciph.protocol); | |
2a3b791e JV |
962 | if (!pp) |
963 | return NF_ACCEPT; | |
964 | ||
2a3b791e | 965 | /* The embedded headers contain source and dest in reverse order */ |
d4383f04 | 966 | cp = pp->conn_out_get(AF_INET6, skb, &ciph, 1); |
2a3b791e JV |
967 | if (!cp) |
968 | return NF_ACCEPT; | |
969 | ||
63dca2c0 JDB |
970 | snet.in6 = ciph.saddr.in6; |
971 | writable = ciph.len; | |
972 | return handle_response_icmp(AF_INET6, skb, &snet, ciph.protocol, cp, | |
973 | pp, writable, sizeof(struct ipv6hdr)); | |
2a3b791e JV |
974 | } |
975 | #endif | |
976 | ||
2906f66a VMR |
977 | /* |
978 | * Check if sctp chunc is ABORT chunk | |
979 | */ | |
980 | static inline int is_sctp_abort(const struct sk_buff *skb, int nh_len) | |
981 | { | |
982 | sctp_chunkhdr_t *sch, schunk; | |
983 | sch = skb_header_pointer(skb, nh_len + sizeof(sctp_sctphdr_t), | |
984 | sizeof(schunk), &schunk); | |
985 | if (sch == NULL) | |
986 | return 0; | |
987 | if (sch->type == SCTP_CID_ABORT) | |
988 | return 1; | |
989 | return 0; | |
990 | } | |
991 | ||
2a3b791e | 992 | static inline int is_tcp_reset(const struct sk_buff *skb, int nh_len) |
1da177e4 LT |
993 | { |
994 | struct tcphdr _tcph, *th; | |
995 | ||
2a3b791e | 996 | th = skb_header_pointer(skb, nh_len, sizeof(_tcph), &_tcph); |
1da177e4 LT |
997 | if (th == NULL) |
998 | return 0; | |
999 | return th->rst; | |
1000 | } | |
1001 | ||
4856c84c | 1002 | /* Handle response packets: rewrite addresses and send away... |
4856c84c MT |
1003 | */ |
1004 | static unsigned int | |
9330419d | 1005 | handle_response(int af, struct sk_buff *skb, struct ip_vs_proto_data *pd, |
d4383f04 | 1006 | struct ip_vs_conn *cp, struct ip_vs_iphdr *iph) |
4856c84c | 1007 | { |
9330419d HS |
1008 | struct ip_vs_protocol *pp = pd->pp; |
1009 | ||
0d79641a | 1010 | IP_VS_DBG_PKT(11, af, pp, skb, 0, "Outgoing packet"); |
4856c84c | 1011 | |
d4383f04 | 1012 | if (!skb_make_writable(skb, iph->len)) |
4856c84c MT |
1013 | goto drop; |
1014 | ||
1015 | /* mangle the packet */ | |
d4383f04 | 1016 | if (pp->snat_handler && !pp->snat_handler(skb, pp, cp, iph)) |
4856c84c MT |
1017 | goto drop; |
1018 | ||
1019 | #ifdef CONFIG_IP_VS_IPV6 | |
1020 | if (af == AF_INET6) | |
1021 | ipv6_hdr(skb)->saddr = cp->vaddr.in6; | |
1022 | else | |
1023 | #endif | |
1024 | { | |
1025 | ip_hdr(skb)->saddr = cp->vaddr.ip; | |
1026 | ip_send_check(ip_hdr(skb)); | |
1027 | } | |
1028 | ||
8a803040 JA |
1029 | /* |
1030 | * nf_iterate does not expect change in the skb->dst->dev. | |
1031 | * It looks like it is not fatal to enable this code for hooks | |
1032 | * where our handlers are at the end of the chain list and | |
1033 | * when all next handlers use skb->dst->dev and not outdev. | |
1034 | * It will definitely route properly the inout NAT traffic | |
1035 | * when multiple paths are used. | |
1036 | */ | |
1037 | ||
4856c84c MT |
1038 | /* For policy routing, packets originating from this |
1039 | * machine itself may be routed differently to packets | |
1040 | * passing through. We want this packet to be routed as | |
1041 | * if it came from this machine itself. So re-compute | |
1042 | * the routing information. | |
1043 | */ | |
ba4fd7e9 SH |
1044 | if (ip_vs_route_me_harder(af, skb)) |
1045 | goto drop; | |
4856c84c | 1046 | |
0d79641a | 1047 | IP_VS_DBG_PKT(10, af, pp, skb, 0, "After SNAT"); |
4856c84c MT |
1048 | |
1049 | ip_vs_out_stats(cp, skb); | |
9330419d | 1050 | ip_vs_set_state(cp, IP_VS_DIR_OUTPUT, skb, pd); |
cf356d69 | 1051 | skb->ipvs_property = 1; |
f4bc17cd | 1052 | if (!(cp->flags & IP_VS_CONN_F_NFCT)) |
cf356d69 | 1053 | ip_vs_notrack(skb); |
f4bc17cd JA |
1054 | else |
1055 | ip_vs_update_conntrack(skb, cp, 0); | |
4856c84c MT |
1056 | ip_vs_conn_put(cp); |
1057 | ||
4856c84c MT |
1058 | LeaveFunction(11); |
1059 | return NF_ACCEPT; | |
1060 | ||
1061 | drop: | |
1062 | ip_vs_conn_put(cp); | |
1063 | kfree_skb(skb); | |
f4bc17cd | 1064 | LeaveFunction(11); |
4856c84c MT |
1065 | return NF_STOLEN; |
1066 | } | |
1067 | ||
1da177e4 | 1068 | /* |
4856c84c | 1069 | * Check if outgoing packet belongs to the established ip_vs_conn. |
1da177e4 LT |
1070 | */ |
1071 | static unsigned int | |
fc604767 | 1072 | ip_vs_out(unsigned int hooknum, struct sk_buff *skb, int af) |
1da177e4 | 1073 | { |
fc723250 | 1074 | struct net *net = NULL; |
51ef348b | 1075 | struct ip_vs_iphdr iph; |
1da177e4 | 1076 | struct ip_vs_protocol *pp; |
9330419d | 1077 | struct ip_vs_proto_data *pd; |
1da177e4 | 1078 | struct ip_vs_conn *cp; |
1da177e4 LT |
1079 | |
1080 | EnterFunction(11); | |
1081 | ||
fc604767 | 1082 | /* Already marked as IPVS request or reply? */ |
6869c4d8 | 1083 | if (skb->ipvs_property) |
1da177e4 LT |
1084 | return NF_ACCEPT; |
1085 | ||
fc604767 JA |
1086 | /* Bad... Do not break raw sockets */ |
1087 | if (unlikely(skb->sk != NULL && hooknum == NF_INET_LOCAL_OUT && | |
1088 | af == AF_INET)) { | |
1089 | struct sock *sk = skb->sk; | |
1090 | struct inet_sock *inet = inet_sk(skb->sk); | |
1091 | ||
1092 | if (inet && sk->sk_family == PF_INET && inet->nodefrag) | |
1093 | return NF_ACCEPT; | |
1094 | } | |
1095 | ||
1096 | if (unlikely(!skb_dst(skb))) | |
1097 | return NF_ACCEPT; | |
1098 | ||
fc723250 | 1099 | net = skb_net(skb); |
7a4f0761 HS |
1100 | if (!net_ipvs(net)->enable) |
1101 | return NF_ACCEPT; | |
1102 | ||
63dca2c0 | 1103 | ip_vs_fill_iph_skb(af, skb, &iph); |
2a3b791e JV |
1104 | #ifdef CONFIG_IP_VS_IPV6 |
1105 | if (af == AF_INET6) { | |
2f74713d JDB |
1106 | if (!iph.fragoffs && skb_nfct_reasm(skb)) { |
1107 | struct sk_buff *reasm = skb_nfct_reasm(skb); | |
1108 | /* Save fw mark for coming frags */ | |
1109 | reasm->ipvs_property = 1; | |
1110 | reasm->mark = skb->mark; | |
1111 | } | |
2a3b791e | 1112 | if (unlikely(iph.protocol == IPPROTO_ICMPV6)) { |
1ca5bb54 JA |
1113 | int related; |
1114 | int verdict = ip_vs_out_icmp_v6(skb, &related, | |
d4383f04 | 1115 | hooknum, &iph); |
1da177e4 | 1116 | |
f5a41847 | 1117 | if (related) |
2a3b791e | 1118 | return verdict; |
2a3b791e JV |
1119 | } |
1120 | } else | |
1121 | #endif | |
1122 | if (unlikely(iph.protocol == IPPROTO_ICMP)) { | |
1ca5bb54 JA |
1123 | int related; |
1124 | int verdict = ip_vs_out_icmp(skb, &related, hooknum); | |
2a3b791e | 1125 | |
f5a41847 | 1126 | if (related) |
2a3b791e | 1127 | return verdict; |
2a3b791e | 1128 | } |
1da177e4 | 1129 | |
9330419d HS |
1130 | pd = ip_vs_proto_data_get(net, iph.protocol); |
1131 | if (unlikely(!pd)) | |
1da177e4 | 1132 | return NF_ACCEPT; |
9330419d | 1133 | pp = pd->pp; |
1da177e4 LT |
1134 | |
1135 | /* reassemble IP fragments */ | |
2a3b791e | 1136 | #ifdef CONFIG_IP_VS_IPV6 |
63dca2c0 | 1137 | if (af == AF_INET) |
2a3b791e | 1138 | #endif |
56f8a75c | 1139 | if (unlikely(ip_is_fragment(ip_hdr(skb)) && !pp->dont_defrag)) { |
1ca5bb54 JA |
1140 | if (ip_vs_gather_frags(skb, |
1141 | ip_vs_defrag_user(hooknum))) | |
2a3b791e JV |
1142 | return NF_STOLEN; |
1143 | ||
63dca2c0 | 1144 | ip_vs_fill_ip4hdr(skb_network_header(skb), &iph); |
2a3b791e | 1145 | } |
1da177e4 LT |
1146 | |
1147 | /* | |
1148 | * Check if the packet belongs to an existing entry | |
1149 | */ | |
d4383f04 | 1150 | cp = pp->conn_out_get(af, skb, &iph, 0); |
1da177e4 | 1151 | |
cb59155f | 1152 | if (likely(cp)) |
d4383f04 | 1153 | return handle_response(af, skb, pd, cp, &iph); |
0cfa558e | 1154 | if (sysctl_nat_icmp_send(net) && |
cb59155f JA |
1155 | (pp->protocol == IPPROTO_TCP || |
1156 | pp->protocol == IPPROTO_UDP || | |
1157 | pp->protocol == IPPROTO_SCTP)) { | |
1158 | __be16 _ports[2], *pptr; | |
1159 | ||
2f74713d JDB |
1160 | pptr = frag_safe_skb_hp(skb, iph.len, |
1161 | sizeof(_ports), _ports, &iph); | |
cb59155f JA |
1162 | if (pptr == NULL) |
1163 | return NF_ACCEPT; /* Not for me */ | |
fc723250 | 1164 | if (ip_vs_lookup_real_service(net, af, iph.protocol, |
cb59155f JA |
1165 | &iph.saddr, |
1166 | pptr[0])) { | |
1167 | /* | |
1168 | * Notify the real server: there is no | |
1169 | * existing entry if it is not RST | |
1170 | * packet or not TCP packet. | |
1171 | */ | |
1172 | if ((iph.protocol != IPPROTO_TCP && | |
1173 | iph.protocol != IPPROTO_SCTP) | |
1174 | || ((iph.protocol == IPPROTO_TCP | |
1175 | && !is_tcp_reset(skb, iph.len)) | |
1176 | || (iph.protocol == IPPROTO_SCTP | |
1177 | && !is_sctp_abort(skb, | |
1178 | iph.len)))) { | |
2a3b791e | 1179 | #ifdef CONFIG_IP_VS_IPV6 |
cb59155f | 1180 | if (af == AF_INET6) { |
cb59155f JA |
1181 | if (!skb->dev) |
1182 | skb->dev = net->loopback_dev; | |
1183 | icmpv6_send(skb, | |
1184 | ICMPV6_DEST_UNREACH, | |
1185 | ICMPV6_PORT_UNREACH, | |
1186 | 0); | |
1187 | } else | |
2a3b791e | 1188 | #endif |
cb59155f JA |
1189 | icmp_send(skb, |
1190 | ICMP_DEST_UNREACH, | |
1191 | ICMP_PORT_UNREACH, 0); | |
1192 | return NF_DROP; | |
1da177e4 LT |
1193 | } |
1194 | } | |
1da177e4 | 1195 | } |
0d79641a | 1196 | IP_VS_DBG_PKT(12, af, pp, skb, 0, |
cb59155f JA |
1197 | "ip_vs_out: packet continues traversal as normal"); |
1198 | return NF_ACCEPT; | |
1da177e4 LT |
1199 | } |
1200 | ||
fc604767 | 1201 | /* |
cb59155f JA |
1202 | * It is hooked at the NF_INET_FORWARD and NF_INET_LOCAL_IN chain, |
1203 | * used only for VS/NAT. | |
fc604767 JA |
1204 | * Check if packet is reply for established ip_vs_conn. |
1205 | */ | |
1206 | static unsigned int | |
1207 | ip_vs_reply4(unsigned int hooknum, struct sk_buff *skb, | |
1208 | const struct net_device *in, const struct net_device *out, | |
1209 | int (*okfn)(struct sk_buff *)) | |
1210 | { | |
1211 | return ip_vs_out(hooknum, skb, AF_INET); | |
1212 | } | |
1213 | ||
1214 | /* | |
1215 | * It is hooked at the NF_INET_LOCAL_OUT chain, used only for VS/NAT. | |
1216 | * Check if packet is reply for established ip_vs_conn. | |
1217 | */ | |
1218 | static unsigned int | |
1219 | ip_vs_local_reply4(unsigned int hooknum, struct sk_buff *skb, | |
1220 | const struct net_device *in, const struct net_device *out, | |
1221 | int (*okfn)(struct sk_buff *)) | |
1222 | { | |
1223 | unsigned int verdict; | |
1224 | ||
1225 | /* Disable BH in LOCAL_OUT until all places are fixed */ | |
1226 | local_bh_disable(); | |
1227 | verdict = ip_vs_out(hooknum, skb, AF_INET); | |
1228 | local_bh_enable(); | |
1229 | return verdict; | |
1230 | } | |
1231 | ||
1232 | #ifdef CONFIG_IP_VS_IPV6 | |
1233 | ||
1234 | /* | |
cb59155f JA |
1235 | * It is hooked at the NF_INET_FORWARD and NF_INET_LOCAL_IN chain, |
1236 | * used only for VS/NAT. | |
fc604767 JA |
1237 | * Check if packet is reply for established ip_vs_conn. |
1238 | */ | |
1239 | static unsigned int | |
1240 | ip_vs_reply6(unsigned int hooknum, struct sk_buff *skb, | |
1241 | const struct net_device *in, const struct net_device *out, | |
1242 | int (*okfn)(struct sk_buff *)) | |
1243 | { | |
1244 | return ip_vs_out(hooknum, skb, AF_INET6); | |
1245 | } | |
1246 | ||
1247 | /* | |
1248 | * It is hooked at the NF_INET_LOCAL_OUT chain, used only for VS/NAT. | |
1249 | * Check if packet is reply for established ip_vs_conn. | |
1250 | */ | |
1251 | static unsigned int | |
1252 | ip_vs_local_reply6(unsigned int hooknum, struct sk_buff *skb, | |
1253 | const struct net_device *in, const struct net_device *out, | |
1254 | int (*okfn)(struct sk_buff *)) | |
1255 | { | |
1256 | unsigned int verdict; | |
1257 | ||
1258 | /* Disable BH in LOCAL_OUT until all places are fixed */ | |
1259 | local_bh_disable(); | |
1260 | verdict = ip_vs_out(hooknum, skb, AF_INET6); | |
1261 | local_bh_enable(); | |
1262 | return verdict; | |
1263 | } | |
1264 | ||
1265 | #endif | |
1da177e4 LT |
1266 | |
1267 | /* | |
1268 | * Handle ICMP messages in the outside-to-inside direction (incoming). | |
1269 | * Find any that might be relevant, check against existing connections, | |
1270 | * forward to the right destination host if relevant. | |
1271 | * Currently handles error types - unreachable, quench, ttl exceeded. | |
1272 | */ | |
e905a9ed | 1273 | static int |
3db05fea | 1274 | ip_vs_in_icmp(struct sk_buff *skb, int *related, unsigned int hooknum) |
1da177e4 | 1275 | { |
9330419d | 1276 | struct net *net = NULL; |
1da177e4 LT |
1277 | struct iphdr *iph; |
1278 | struct icmphdr _icmph, *ic; | |
1279 | struct iphdr _ciph, *cih; /* The ip header contained within the ICMP */ | |
51ef348b | 1280 | struct ip_vs_iphdr ciph; |
1da177e4 LT |
1281 | struct ip_vs_conn *cp; |
1282 | struct ip_vs_protocol *pp; | |
9330419d | 1283 | struct ip_vs_proto_data *pd; |
f2edb9f7 JA |
1284 | unsigned int offset, offset2, ihl, verdict; |
1285 | bool ipip; | |
1da177e4 LT |
1286 | |
1287 | *related = 1; | |
1288 | ||
1289 | /* reassemble IP fragments */ | |
56f8a75c | 1290 | if (ip_is_fragment(ip_hdr(skb))) { |
1ca5bb54 | 1291 | if (ip_vs_gather_frags(skb, ip_vs_defrag_user(hooknum))) |
1da177e4 | 1292 | return NF_STOLEN; |
1da177e4 LT |
1293 | } |
1294 | ||
eddc9ec5 | 1295 | iph = ip_hdr(skb); |
1da177e4 LT |
1296 | offset = ihl = iph->ihl * 4; |
1297 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
1298 | if (ic == NULL) | |
1299 | return NF_DROP; | |
1300 | ||
14d5e834 | 1301 | IP_VS_DBG(12, "Incoming ICMP (%d,%d) %pI4->%pI4\n", |
1da177e4 | 1302 | ic->type, ntohs(icmp_id(ic)), |
14d5e834 | 1303 | &iph->saddr, &iph->daddr); |
1da177e4 LT |
1304 | |
1305 | /* | |
1306 | * Work through seeing if this is for us. | |
1307 | * These checks are supposed to be in an order that means easy | |
1308 | * things are checked first to speed up processing.... however | |
1309 | * this means that some packets will manage to get a long way | |
1310 | * down this stack and then be rejected, but that's life. | |
1311 | */ | |
1312 | if ((ic->type != ICMP_DEST_UNREACH) && | |
1313 | (ic->type != ICMP_SOURCE_QUENCH) && | |
1314 | (ic->type != ICMP_TIME_EXCEEDED)) { | |
1315 | *related = 0; | |
1316 | return NF_ACCEPT; | |
1317 | } | |
1318 | ||
1319 | /* Now find the contained IP header */ | |
1320 | offset += sizeof(_icmph); | |
1321 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
1322 | if (cih == NULL) | |
1323 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
1324 | ||
9330419d | 1325 | net = skb_net(skb); |
7a4f0761 | 1326 | |
f2edb9f7 JA |
1327 | /* Special case for errors for IPIP packets */ |
1328 | ipip = false; | |
1329 | if (cih->protocol == IPPROTO_IPIP) { | |
1330 | if (unlikely(cih->frag_off & htons(IP_OFFSET))) | |
1331 | return NF_ACCEPT; | |
1332 | /* Error for our IPIP must arrive at LOCAL_IN */ | |
1333 | if (!(skb_rtable(skb)->rt_flags & RTCF_LOCAL)) | |
1334 | return NF_ACCEPT; | |
1335 | offset += cih->ihl * 4; | |
1336 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
1337 | if (cih == NULL) | |
1338 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
1339 | ipip = true; | |
1340 | } | |
1341 | ||
9330419d HS |
1342 | pd = ip_vs_proto_data_get(net, cih->protocol); |
1343 | if (!pd) | |
1da177e4 | 1344 | return NF_ACCEPT; |
9330419d | 1345 | pp = pd->pp; |
1da177e4 LT |
1346 | |
1347 | /* Is the embedded protocol header present? */ | |
4412ec49 | 1348 | if (unlikely(cih->frag_off & htons(IP_OFFSET) && |
1da177e4 LT |
1349 | pp->dont_defrag)) |
1350 | return NF_ACCEPT; | |
1351 | ||
0d79641a JA |
1352 | IP_VS_DBG_PKT(11, AF_INET, pp, skb, offset, |
1353 | "Checking incoming ICMP for"); | |
1da177e4 | 1354 | |
f2edb9f7 | 1355 | offset2 = offset; |
63dca2c0 JDB |
1356 | ip_vs_fill_ip4hdr(cih, &ciph); |
1357 | ciph.len += offset; | |
1358 | offset = ciph.len; | |
f2edb9f7 JA |
1359 | /* The embedded headers contain source and dest in reverse order. |
1360 | * For IPIP this is error for request, not for reply. | |
1361 | */ | |
d4383f04 | 1362 | cp = pp->conn_in_get(AF_INET, skb, &ciph, ipip ? 0 : 1); |
6cb90db5 | 1363 | if (!cp) |
1da177e4 LT |
1364 | return NF_ACCEPT; |
1365 | ||
1366 | verdict = NF_DROP; | |
1367 | ||
1368 | /* Ensure the checksum is correct */ | |
60476372 | 1369 | if (!skb_csum_unnecessary(skb) && ip_vs_checksum_complete(skb, ihl)) { |
1da177e4 | 1370 | /* Failed checksum! */ |
14d5e834 HH |
1371 | IP_VS_DBG(1, "Incoming ICMP: failed checksum from %pI4!\n", |
1372 | &iph->saddr); | |
1da177e4 LT |
1373 | goto out; |
1374 | } | |
1375 | ||
f2edb9f7 JA |
1376 | if (ipip) { |
1377 | __be32 info = ic->un.gateway; | |
1378 | ||
1379 | /* Update the MTU */ | |
1380 | if (ic->type == ICMP_DEST_UNREACH && | |
1381 | ic->code == ICMP_FRAG_NEEDED) { | |
1382 | struct ip_vs_dest *dest = cp->dest; | |
1383 | u32 mtu = ntohs(ic->un.frag.mtu); | |
1384 | ||
1385 | /* Strip outer IP and ICMP, go to IPIP header */ | |
1386 | __skb_pull(skb, ihl + sizeof(_icmph)); | |
1387 | offset2 -= ihl + sizeof(_icmph); | |
1388 | skb_reset_network_header(skb); | |
1389 | IP_VS_DBG(12, "ICMP for IPIP %pI4->%pI4: mtu=%u\n", | |
1390 | &ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr, mtu); | |
f2edb9f7 JA |
1391 | ipv4_update_pmtu(skb, dev_net(skb->dev), |
1392 | mtu, 0, 0, 0, 0); | |
f2edb9f7 JA |
1393 | /* Client uses PMTUD? */ |
1394 | if (!(cih->frag_off & htons(IP_DF))) | |
1395 | goto ignore_ipip; | |
1396 | /* Prefer the resulting PMTU */ | |
1397 | if (dest) { | |
026ace06 JA |
1398 | struct ip_vs_dest_dst *dest_dst; |
1399 | ||
1400 | rcu_read_lock(); | |
1401 | dest_dst = rcu_dereference(dest->dest_dst); | |
1402 | if (dest_dst) | |
1403 | mtu = dst_mtu(dest_dst->dst_cache); | |
1404 | rcu_read_unlock(); | |
f2edb9f7 JA |
1405 | } |
1406 | if (mtu > 68 + sizeof(struct iphdr)) | |
1407 | mtu -= sizeof(struct iphdr); | |
1408 | info = htonl(mtu); | |
1409 | } | |
1410 | /* Strip outer IP, ICMP and IPIP, go to IP header of | |
1411 | * original request. | |
1412 | */ | |
1413 | __skb_pull(skb, offset2); | |
1414 | skb_reset_network_header(skb); | |
1415 | IP_VS_DBG(12, "Sending ICMP for %pI4->%pI4: t=%u, c=%u, i=%u\n", | |
1416 | &ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr, | |
1417 | ic->type, ic->code, ntohl(info)); | |
1418 | icmp_send(skb, ic->type, ic->code, info); | |
1419 | /* ICMP can be shorter but anyways, account it */ | |
1420 | ip_vs_out_stats(cp, skb); | |
1421 | ||
1422 | ignore_ipip: | |
1423 | consume_skb(skb); | |
1424 | verdict = NF_STOLEN; | |
1425 | goto out; | |
1426 | } | |
1427 | ||
1da177e4 LT |
1428 | /* do the statistics and put it back */ |
1429 | ip_vs_in_stats(cp, skb); | |
1430 | if (IPPROTO_TCP == cih->protocol || IPPROTO_UDP == cih->protocol) | |
1431 | offset += 2 * sizeof(__u16); | |
d4383f04 | 1432 | verdict = ip_vs_icmp_xmit(skb, cp, pp, offset, hooknum, &ciph); |
1da177e4 | 1433 | |
552ad65a | 1434 | out: |
1da177e4 LT |
1435 | __ip_vs_conn_put(cp); |
1436 | ||
1437 | return verdict; | |
1438 | } | |
1439 | ||
2a3b791e | 1440 | #ifdef CONFIG_IP_VS_IPV6 |
d4383f04 JDB |
1441 | static int ip_vs_in_icmp_v6(struct sk_buff *skb, int *related, |
1442 | unsigned int hooknum, struct ip_vs_iphdr *iph) | |
2a3b791e | 1443 | { |
9330419d | 1444 | struct net *net = NULL; |
63dca2c0 | 1445 | struct ipv6hdr _ip6h, *ip6h; |
2a3b791e | 1446 | struct icmp6hdr _icmph, *ic; |
63dca2c0 | 1447 | struct ip_vs_iphdr ciph = {.flags = 0, .fragoffs = 0};/*Contained IP */ |
2a3b791e JV |
1448 | struct ip_vs_conn *cp; |
1449 | struct ip_vs_protocol *pp; | |
9330419d | 1450 | struct ip_vs_proto_data *pd; |
63dca2c0 | 1451 | unsigned int offs_ciph, writable, verdict; |
2a3b791e | 1452 | |
63dca2c0 | 1453 | *related = 1; |
2a3b791e | 1454 | |
2f74713d | 1455 | ic = frag_safe_skb_hp(skb, iph->len, sizeof(_icmph), &_icmph, iph); |
2a3b791e JV |
1456 | if (ic == NULL) |
1457 | return NF_DROP; | |
1458 | ||
2a3b791e JV |
1459 | /* |
1460 | * Work through seeing if this is for us. | |
1461 | * These checks are supposed to be in an order that means easy | |
1462 | * things are checked first to speed up processing.... however | |
1463 | * this means that some packets will manage to get a long way | |
1464 | * down this stack and then be rejected, but that's life. | |
1465 | */ | |
2fab8917 | 1466 | if (ic->icmp6_type & ICMPV6_INFOMSG_MASK) { |
2a3b791e JV |
1467 | *related = 0; |
1468 | return NF_ACCEPT; | |
1469 | } | |
2f74713d JDB |
1470 | /* Fragment header that is before ICMP header tells us that: |
1471 | * it's not an error message since they can't be fragmented. | |
1472 | */ | |
e7165030 | 1473 | if (iph->flags & IP6_FH_F_FRAG) |
2f74713d | 1474 | return NF_DROP; |
2a3b791e | 1475 | |
63dca2c0 JDB |
1476 | IP_VS_DBG(8, "Incoming ICMPv6 (%d,%d) %pI6c->%pI6c\n", |
1477 | ic->icmp6_type, ntohs(icmpv6_id(ic)), | |
1478 | &iph->saddr, &iph->daddr); | |
1479 | ||
2a3b791e | 1480 | /* Now find the contained IP header */ |
63dca2c0 JDB |
1481 | ciph.len = iph->len + sizeof(_icmph); |
1482 | offs_ciph = ciph.len; /* Save ip header offset */ | |
1483 | ip6h = skb_header_pointer(skb, ciph.len, sizeof(_ip6h), &_ip6h); | |
1484 | if (ip6h == NULL) | |
2a3b791e | 1485 | return NF_ACCEPT; /* The packet looks wrong, ignore */ |
63dca2c0 JDB |
1486 | ciph.saddr.in6 = ip6h->saddr; /* conn_in_get() handles reverse order */ |
1487 | ciph.daddr.in6 = ip6h->daddr; | |
1488 | /* skip possible IPv6 exthdrs of contained IPv6 packet */ | |
1489 | ciph.protocol = ipv6_find_hdr(skb, &ciph.len, -1, &ciph.fragoffs, NULL); | |
1490 | if (ciph.protocol < 0) | |
1491 | return NF_ACCEPT; /* Contained IPv6 hdr looks wrong, ignore */ | |
2a3b791e | 1492 | |
9330419d | 1493 | net = skb_net(skb); |
63dca2c0 | 1494 | pd = ip_vs_proto_data_get(net, ciph.protocol); |
9330419d | 1495 | if (!pd) |
2a3b791e | 1496 | return NF_ACCEPT; |
9330419d | 1497 | pp = pd->pp; |
2a3b791e | 1498 | |
63dca2c0 JDB |
1499 | /* Cannot handle fragmented embedded protocol */ |
1500 | if (ciph.fragoffs) | |
2a3b791e JV |
1501 | return NF_ACCEPT; |
1502 | ||
63dca2c0 | 1503 | IP_VS_DBG_PKT(11, AF_INET6, pp, skb, offs_ciph, |
0d79641a | 1504 | "Checking incoming ICMPv6 for"); |
2a3b791e | 1505 | |
2f74713d JDB |
1506 | /* The embedded headers contain source and dest in reverse order |
1507 | * if not from localhost | |
1508 | */ | |
d4383f04 | 1509 | cp = pp->conn_in_get(AF_INET6, skb, &ciph, |
2f74713d JDB |
1510 | (hooknum == NF_INET_LOCAL_OUT) ? 0 : 1); |
1511 | ||
6cb90db5 | 1512 | if (!cp) |
2a3b791e | 1513 | return NF_ACCEPT; |
2f74713d JDB |
1514 | /* VS/TUN, VS/DR and LOCALNODE just let it go */ |
1515 | if ((hooknum == NF_INET_LOCAL_OUT) && | |
1516 | (IP_VS_FWD_METHOD(cp) != IP_VS_CONN_F_MASQ)) { | |
1517 | __ip_vs_conn_put(cp); | |
1518 | return NF_ACCEPT; | |
1519 | } | |
2a3b791e | 1520 | |
2a3b791e JV |
1521 | /* do the statistics and put it back */ |
1522 | ip_vs_in_stats(cp, skb); | |
63dca2c0 JDB |
1523 | |
1524 | /* Need to mangle contained IPv6 header in ICMPv6 packet */ | |
1525 | writable = ciph.len; | |
1526 | if (IPPROTO_TCP == ciph.protocol || IPPROTO_UDP == ciph.protocol || | |
1527 | IPPROTO_SCTP == ciph.protocol) | |
1528 | writable += 2 * sizeof(__u16); /* Also mangle ports */ | |
1529 | ||
d4383f04 | 1530 | verdict = ip_vs_icmp_xmit_v6(skb, cp, pp, writable, hooknum, &ciph); |
2a3b791e JV |
1531 | |
1532 | __ip_vs_conn_put(cp); | |
1533 | ||
1534 | return verdict; | |
1535 | } | |
1536 | #endif | |
1537 | ||
1538 | ||
1da177e4 LT |
1539 | /* |
1540 | * Check if it's for virtual services, look it up, | |
1541 | * and send it on its way... | |
1542 | */ | |
1543 | static unsigned int | |
cb59155f | 1544 | ip_vs_in(unsigned int hooknum, struct sk_buff *skb, int af) |
1da177e4 | 1545 | { |
f131315f | 1546 | struct net *net; |
51ef348b | 1547 | struct ip_vs_iphdr iph; |
1da177e4 | 1548 | struct ip_vs_protocol *pp; |
9330419d | 1549 | struct ip_vs_proto_data *pd; |
1da177e4 | 1550 | struct ip_vs_conn *cp; |
4a516f11 | 1551 | int ret, pkts; |
f131315f | 1552 | struct netns_ipvs *ipvs; |
2a3b791e | 1553 | |
fc604767 JA |
1554 | /* Already marked as IPVS request or reply? */ |
1555 | if (skb->ipvs_property) | |
1556 | return NF_ACCEPT; | |
1557 | ||
1da177e4 | 1558 | /* |
cb59155f JA |
1559 | * Big tappo: |
1560 | * - remote client: only PACKET_HOST | |
1561 | * - route: used for struct net when skb->dev is unset | |
1da177e4 | 1562 | */ |
cb59155f JA |
1563 | if (unlikely((skb->pkt_type != PACKET_HOST && |
1564 | hooknum != NF_INET_LOCAL_OUT) || | |
1565 | !skb_dst(skb))) { | |
63dca2c0 | 1566 | ip_vs_fill_iph_skb(af, skb, &iph); |
cb59155f JA |
1567 | IP_VS_DBG_BUF(12, "packet type=%d proto=%d daddr=%s" |
1568 | " ignored in hook %u\n", | |
1569 | skb->pkt_type, iph.protocol, | |
1570 | IP_VS_DBG_ADDR(af, &iph.daddr), hooknum); | |
1da177e4 LT |
1571 | return NF_ACCEPT; |
1572 | } | |
7a4f0761 HS |
1573 | /* ipvs enabled in this netns ? */ |
1574 | net = skb_net(skb); | |
0c12582f JA |
1575 | ipvs = net_ipvs(net); |
1576 | if (unlikely(sysctl_backup_only(ipvs) || !ipvs->enable)) | |
7a4f0761 HS |
1577 | return NF_ACCEPT; |
1578 | ||
63dca2c0 | 1579 | ip_vs_fill_iph_skb(af, skb, &iph); |
cb59155f JA |
1580 | |
1581 | /* Bad... Do not break raw sockets */ | |
1582 | if (unlikely(skb->sk != NULL && hooknum == NF_INET_LOCAL_OUT && | |
1583 | af == AF_INET)) { | |
1584 | struct sock *sk = skb->sk; | |
1585 | struct inet_sock *inet = inet_sk(skb->sk); | |
1586 | ||
1587 | if (inet && sk->sk_family == PF_INET && inet->nodefrag) | |
1588 | return NF_ACCEPT; | |
1589 | } | |
1da177e4 | 1590 | |
94b26551 JV |
1591 | #ifdef CONFIG_IP_VS_IPV6 |
1592 | if (af == AF_INET6) { | |
2f74713d JDB |
1593 | if (!iph.fragoffs && skb_nfct_reasm(skb)) { |
1594 | struct sk_buff *reasm = skb_nfct_reasm(skb); | |
1595 | /* Save fw mark for coming frags. */ | |
1596 | reasm->ipvs_property = 1; | |
1597 | reasm->mark = skb->mark; | |
1598 | } | |
94b26551 | 1599 | if (unlikely(iph.protocol == IPPROTO_ICMPV6)) { |
1ca5bb54 | 1600 | int related; |
d4383f04 JDB |
1601 | int verdict = ip_vs_in_icmp_v6(skb, &related, hooknum, |
1602 | &iph); | |
1da177e4 | 1603 | |
94b26551 JV |
1604 | if (related) |
1605 | return verdict; | |
94b26551 JV |
1606 | } |
1607 | } else | |
1608 | #endif | |
1609 | if (unlikely(iph.protocol == IPPROTO_ICMP)) { | |
1ca5bb54 JA |
1610 | int related; |
1611 | int verdict = ip_vs_in_icmp(skb, &related, hooknum); | |
94b26551 JV |
1612 | |
1613 | if (related) | |
1614 | return verdict; | |
94b26551 | 1615 | } |
1da177e4 LT |
1616 | |
1617 | /* Protocol supported? */ | |
9330419d HS |
1618 | pd = ip_vs_proto_data_get(net, iph.protocol); |
1619 | if (unlikely(!pd)) | |
1da177e4 | 1620 | return NF_ACCEPT; |
9330419d | 1621 | pp = pd->pp; |
1da177e4 LT |
1622 | /* |
1623 | * Check if the packet belongs to an existing connection entry | |
1624 | */ | |
d4383f04 | 1625 | cp = pp->conn_in_get(af, skb, &iph, 0); |
63dca2c0 | 1626 | if (unlikely(!cp) && !iph.fragoffs) { |
2f74713d JDB |
1627 | /* No (second) fragments need to enter here, as nf_defrag_ipv6 |
1628 | * replayed fragment zero will already have created the cp | |
1629 | */ | |
1da177e4 LT |
1630 | int v; |
1631 | ||
2f74713d | 1632 | /* Schedule and create new connection entry into &cp */ |
d4383f04 | 1633 | if (!pp->conn_schedule(af, skb, pd, &v, &cp, &iph)) |
1da177e4 LT |
1634 | return v; |
1635 | } | |
1636 | ||
1637 | if (unlikely(!cp)) { | |
1638 | /* sorry, all this trouble for a no-hit :) */ | |
0d79641a | 1639 | IP_VS_DBG_PKT(12, af, pp, skb, 0, |
cb59155f | 1640 | "ip_vs_in: packet continues traversal as normal"); |
2f74713d JDB |
1641 | if (iph.fragoffs && !skb_nfct_reasm(skb)) { |
1642 | /* Fragment that couldn't be mapped to a conn entry | |
1643 | * and don't have any pointer to a reasm skb | |
1644 | * is missing module nf_defrag_ipv6 | |
1645 | */ | |
1646 | IP_VS_DBG_RL("Unhandled frag, load nf_defrag_ipv6\n"); | |
1647 | IP_VS_DBG_PKT(7, af, pp, skb, 0, "unhandled fragment"); | |
1648 | } | |
1da177e4 LT |
1649 | return NF_ACCEPT; |
1650 | } | |
1651 | ||
0d79641a | 1652 | IP_VS_DBG_PKT(11, af, pp, skb, 0, "Incoming packet"); |
1da177e4 LT |
1653 | /* Check the server status */ |
1654 | if (cp->dest && !(cp->dest->flags & IP_VS_DEST_F_AVAILABLE)) { | |
1655 | /* the destination server is not available */ | |
1656 | ||
71a8ab6c | 1657 | if (sysctl_expire_nodest_conn(ipvs)) { |
1da177e4 LT |
1658 | /* try to expire the connection immediately */ |
1659 | ip_vs_conn_expire_now(cp); | |
1da177e4 | 1660 | } |
dc8103f2 JA |
1661 | /* don't restart its timer, and silently |
1662 | drop the packet. */ | |
1663 | __ip_vs_conn_put(cp); | |
1da177e4 LT |
1664 | return NF_DROP; |
1665 | } | |
1666 | ||
1667 | ip_vs_in_stats(cp, skb); | |
4a516f11 | 1668 | ip_vs_set_state(cp, IP_VS_DIR_INPUT, skb, pd); |
1da177e4 | 1669 | if (cp->packet_xmit) |
d4383f04 | 1670 | ret = cp->packet_xmit(skb, cp, pp, &iph); |
1da177e4 LT |
1671 | /* do not touch skb anymore */ |
1672 | else { | |
1673 | IP_VS_DBG_RL("warning: packet_xmit is null"); | |
1674 | ret = NF_ACCEPT; | |
1675 | } | |
1676 | ||
efac5276 RB |
1677 | /* Increase its packet counter and check if it is needed |
1678 | * to be synchronized | |
1679 | * | |
1680 | * Sync connection if it is about to close to | |
1681 | * encorage the standby servers to update the connections timeout | |
986a0757 HS |
1682 | * |
1683 | * For ONE_PKT let ip_vs_sync_conn() do the filter work. | |
efac5276 | 1684 | */ |
f131315f | 1685 | |
986a0757 | 1686 | if (cp->flags & IP_VS_CONN_F_ONE_PACKET) |
59e0350e | 1687 | pkts = sysctl_sync_threshold(ipvs); |
986a0757 HS |
1688 | else |
1689 | pkts = atomic_add_return(1, &cp->in_pkts); | |
1690 | ||
749c42b6 JA |
1691 | if (ipvs->sync_state & IP_VS_STATE_MASTER) |
1692 | ip_vs_sync_conn(net, cp, pkts); | |
1da177e4 LT |
1693 | |
1694 | ip_vs_conn_put(cp); | |
1695 | return ret; | |
1696 | } | |
1697 | ||
cb59155f JA |
1698 | /* |
1699 | * AF_INET handler in NF_INET_LOCAL_IN chain | |
1700 | * Schedule and forward packets from remote clients | |
1701 | */ | |
1702 | static unsigned int | |
1703 | ip_vs_remote_request4(unsigned int hooknum, struct sk_buff *skb, | |
1704 | const struct net_device *in, | |
1705 | const struct net_device *out, | |
1706 | int (*okfn)(struct sk_buff *)) | |
1707 | { | |
1708 | return ip_vs_in(hooknum, skb, AF_INET); | |
1709 | } | |
1710 | ||
1711 | /* | |
1712 | * AF_INET handler in NF_INET_LOCAL_OUT chain | |
1713 | * Schedule and forward packets from local clients | |
1714 | */ | |
1715 | static unsigned int | |
1716 | ip_vs_local_request4(unsigned int hooknum, struct sk_buff *skb, | |
1717 | const struct net_device *in, const struct net_device *out, | |
1718 | int (*okfn)(struct sk_buff *)) | |
1719 | { | |
1720 | unsigned int verdict; | |
1721 | ||
1722 | /* Disable BH in LOCAL_OUT until all places are fixed */ | |
1723 | local_bh_disable(); | |
1724 | verdict = ip_vs_in(hooknum, skb, AF_INET); | |
1725 | local_bh_enable(); | |
1726 | return verdict; | |
1727 | } | |
1728 | ||
1729 | #ifdef CONFIG_IP_VS_IPV6 | |
1730 | ||
2f74713d JDB |
1731 | /* |
1732 | * AF_INET6 fragment handling | |
1733 | * Copy info from first fragment, to the rest of them. | |
1734 | */ | |
1735 | static unsigned int | |
1736 | ip_vs_preroute_frag6(unsigned int hooknum, struct sk_buff *skb, | |
1737 | const struct net_device *in, | |
1738 | const struct net_device *out, | |
1739 | int (*okfn)(struct sk_buff *)) | |
1740 | { | |
1741 | struct sk_buff *reasm = skb_nfct_reasm(skb); | |
1742 | struct net *net; | |
1743 | ||
1744 | /* Skip if not a "replay" from nf_ct_frag6_output or first fragment. | |
1745 | * ipvs_property is set when checking first fragment | |
1746 | * in ip_vs_in() and ip_vs_out(). | |
1747 | */ | |
1748 | if (reasm) | |
1749 | IP_VS_DBG(2, "Fragment recv prop:%d\n", reasm->ipvs_property); | |
1750 | if (!reasm || !reasm->ipvs_property) | |
1751 | return NF_ACCEPT; | |
1752 | ||
1753 | net = skb_net(skb); | |
1754 | if (!net_ipvs(net)->enable) | |
1755 | return NF_ACCEPT; | |
1756 | ||
1757 | /* Copy stored fw mark, saved in ip_vs_{in,out} */ | |
1758 | skb->mark = reasm->mark; | |
1759 | ||
1760 | return NF_ACCEPT; | |
1761 | } | |
1762 | ||
cb59155f JA |
1763 | /* |
1764 | * AF_INET6 handler in NF_INET_LOCAL_IN chain | |
1765 | * Schedule and forward packets from remote clients | |
1766 | */ | |
1767 | static unsigned int | |
1768 | ip_vs_remote_request6(unsigned int hooknum, struct sk_buff *skb, | |
1769 | const struct net_device *in, | |
1770 | const struct net_device *out, | |
1771 | int (*okfn)(struct sk_buff *)) | |
1772 | { | |
1773 | return ip_vs_in(hooknum, skb, AF_INET6); | |
1774 | } | |
1775 | ||
1776 | /* | |
1777 | * AF_INET6 handler in NF_INET_LOCAL_OUT chain | |
1778 | * Schedule and forward packets from local clients | |
1779 | */ | |
1780 | static unsigned int | |
1781 | ip_vs_local_request6(unsigned int hooknum, struct sk_buff *skb, | |
1782 | const struct net_device *in, const struct net_device *out, | |
1783 | int (*okfn)(struct sk_buff *)) | |
1784 | { | |
1785 | unsigned int verdict; | |
1786 | ||
1787 | /* Disable BH in LOCAL_OUT until all places are fixed */ | |
1788 | local_bh_disable(); | |
1789 | verdict = ip_vs_in(hooknum, skb, AF_INET6); | |
1790 | local_bh_enable(); | |
1791 | return verdict; | |
1792 | } | |
1793 | ||
1794 | #endif | |
1795 | ||
1da177e4 LT |
1796 | |
1797 | /* | |
6e23ae2a | 1798 | * It is hooked at the NF_INET_FORWARD chain, in order to catch ICMP |
1da177e4 LT |
1799 | * related packets destined for 0.0.0.0/0. |
1800 | * When fwmark-based virtual service is used, such as transparent | |
1801 | * cache cluster, TCP packets can be marked and routed to ip_vs_in, | |
1802 | * but ICMP destined for 0.0.0.0/0 cannot not be easily marked and | |
6e23ae2a | 1803 | * sent to ip_vs_in_icmp. So, catch them at the NF_INET_FORWARD chain |
1da177e4 LT |
1804 | * and send them to ip_vs_in_icmp. |
1805 | */ | |
1806 | static unsigned int | |
3db05fea | 1807 | ip_vs_forward_icmp(unsigned int hooknum, struct sk_buff *skb, |
1da177e4 LT |
1808 | const struct net_device *in, const struct net_device *out, |
1809 | int (*okfn)(struct sk_buff *)) | |
1810 | { | |
1811 | int r; | |
7a4f0761 | 1812 | struct net *net; |
0c12582f | 1813 | struct netns_ipvs *ipvs; |
1da177e4 | 1814 | |
3db05fea | 1815 | if (ip_hdr(skb)->protocol != IPPROTO_ICMP) |
1da177e4 LT |
1816 | return NF_ACCEPT; |
1817 | ||
7a4f0761 HS |
1818 | /* ipvs enabled in this netns ? */ |
1819 | net = skb_net(skb); | |
0c12582f JA |
1820 | ipvs = net_ipvs(net); |
1821 | if (unlikely(sysctl_backup_only(ipvs) || !ipvs->enable)) | |
7a4f0761 HS |
1822 | return NF_ACCEPT; |
1823 | ||
3db05fea | 1824 | return ip_vs_in_icmp(skb, &r, hooknum); |
1da177e4 LT |
1825 | } |
1826 | ||
2a3b791e JV |
1827 | #ifdef CONFIG_IP_VS_IPV6 |
1828 | static unsigned int | |
1829 | ip_vs_forward_icmp_v6(unsigned int hooknum, struct sk_buff *skb, | |
1830 | const struct net_device *in, const struct net_device *out, | |
1831 | int (*okfn)(struct sk_buff *)) | |
1832 | { | |
1833 | int r; | |
7a4f0761 | 1834 | struct net *net; |
0c12582f | 1835 | struct netns_ipvs *ipvs; |
63dca2c0 | 1836 | struct ip_vs_iphdr iphdr; |
2a3b791e | 1837 | |
63dca2c0 JDB |
1838 | ip_vs_fill_iph_skb(AF_INET6, skb, &iphdr); |
1839 | if (iphdr.protocol != IPPROTO_ICMPV6) | |
2a3b791e JV |
1840 | return NF_ACCEPT; |
1841 | ||
7a4f0761 HS |
1842 | /* ipvs enabled in this netns ? */ |
1843 | net = skb_net(skb); | |
0c12582f JA |
1844 | ipvs = net_ipvs(net); |
1845 | if (unlikely(sysctl_backup_only(ipvs) || !ipvs->enable)) | |
7a4f0761 HS |
1846 | return NF_ACCEPT; |
1847 | ||
d4383f04 | 1848 | return ip_vs_in_icmp_v6(skb, &r, hooknum, &iphdr); |
2a3b791e JV |
1849 | } |
1850 | #endif | |
1851 | ||
1da177e4 | 1852 | |
1999414a | 1853 | static struct nf_hook_ops ip_vs_ops[] __read_mostly = { |
cb59155f JA |
1854 | /* After packet filtering, change source only for VS/NAT */ |
1855 | { | |
1856 | .hook = ip_vs_reply4, | |
1857 | .owner = THIS_MODULE, | |
4c809d63 | 1858 | .pf = NFPROTO_IPV4, |
cb59155f | 1859 | .hooknum = NF_INET_LOCAL_IN, |
afb523c5 | 1860 | .priority = NF_IP_PRI_NAT_SRC - 2, |
cb59155f | 1861 | }, |
41c5b317 PM |
1862 | /* After packet filtering, forward packet through VS/DR, VS/TUN, |
1863 | * or VS/NAT(change destination), so that filtering rules can be | |
1864 | * applied to IPVS. */ | |
1865 | { | |
cb59155f | 1866 | .hook = ip_vs_remote_request4, |
41c5b317 | 1867 | .owner = THIS_MODULE, |
4c809d63 | 1868 | .pf = NFPROTO_IPV4, |
cb59155f | 1869 | .hooknum = NF_INET_LOCAL_IN, |
afb523c5 | 1870 | .priority = NF_IP_PRI_NAT_SRC - 1, |
41c5b317 | 1871 | }, |
fc604767 | 1872 | /* Before ip_vs_in, change source only for VS/NAT */ |
41c5b317 | 1873 | { |
fc604767 | 1874 | .hook = ip_vs_local_reply4, |
41c5b317 | 1875 | .owner = THIS_MODULE, |
4c809d63 | 1876 | .pf = NFPROTO_IPV4, |
fc604767 | 1877 | .hooknum = NF_INET_LOCAL_OUT, |
afb523c5 | 1878 | .priority = NF_IP_PRI_NAT_DST + 1, |
41c5b317 | 1879 | }, |
cb59155f JA |
1880 | /* After mangle, schedule and forward local requests */ |
1881 | { | |
1882 | .hook = ip_vs_local_request4, | |
1883 | .owner = THIS_MODULE, | |
4c809d63 | 1884 | .pf = NFPROTO_IPV4, |
cb59155f | 1885 | .hooknum = NF_INET_LOCAL_OUT, |
afb523c5 | 1886 | .priority = NF_IP_PRI_NAT_DST + 2, |
cb59155f | 1887 | }, |
41c5b317 PM |
1888 | /* After packet filtering (but before ip_vs_out_icmp), catch icmp |
1889 | * destined for 0.0.0.0/0, which is for incoming IPVS connections */ | |
1890 | { | |
1891 | .hook = ip_vs_forward_icmp, | |
1892 | .owner = THIS_MODULE, | |
4c809d63 | 1893 | .pf = NFPROTO_IPV4, |
cb59155f JA |
1894 | .hooknum = NF_INET_FORWARD, |
1895 | .priority = 99, | |
41c5b317 | 1896 | }, |
fc604767 JA |
1897 | /* After packet filtering, change source only for VS/NAT */ |
1898 | { | |
1899 | .hook = ip_vs_reply4, | |
1900 | .owner = THIS_MODULE, | |
4c809d63 | 1901 | .pf = NFPROTO_IPV4, |
fc604767 JA |
1902 | .hooknum = NF_INET_FORWARD, |
1903 | .priority = 100, | |
1904 | }, | |
473b23d3 | 1905 | #ifdef CONFIG_IP_VS_IPV6 |
2f74713d JDB |
1906 | /* After mangle & nat fetch 2:nd fragment and following */ |
1907 | { | |
1908 | .hook = ip_vs_preroute_frag6, | |
1909 | .owner = THIS_MODULE, | |
1910 | .pf = NFPROTO_IPV6, | |
1911 | .hooknum = NF_INET_PRE_ROUTING, | |
1912 | .priority = NF_IP6_PRI_NAT_DST + 1, | |
1913 | }, | |
cb59155f JA |
1914 | /* After packet filtering, change source only for VS/NAT */ |
1915 | { | |
1916 | .hook = ip_vs_reply6, | |
1917 | .owner = THIS_MODULE, | |
4c809d63 | 1918 | .pf = NFPROTO_IPV6, |
cb59155f | 1919 | .hooknum = NF_INET_LOCAL_IN, |
afb523c5 | 1920 | .priority = NF_IP6_PRI_NAT_SRC - 2, |
cb59155f | 1921 | }, |
473b23d3 JV |
1922 | /* After packet filtering, forward packet through VS/DR, VS/TUN, |
1923 | * or VS/NAT(change destination), so that filtering rules can be | |
1924 | * applied to IPVS. */ | |
1925 | { | |
cb59155f | 1926 | .hook = ip_vs_remote_request6, |
473b23d3 | 1927 | .owner = THIS_MODULE, |
4c809d63 | 1928 | .pf = NFPROTO_IPV6, |
cb59155f | 1929 | .hooknum = NF_INET_LOCAL_IN, |
afb523c5 | 1930 | .priority = NF_IP6_PRI_NAT_SRC - 1, |
473b23d3 | 1931 | }, |
fc604767 | 1932 | /* Before ip_vs_in, change source only for VS/NAT */ |
473b23d3 | 1933 | { |
fc604767 | 1934 | .hook = ip_vs_local_reply6, |
473b23d3 | 1935 | .owner = THIS_MODULE, |
4c809d63 | 1936 | .pf = NFPROTO_IPV4, |
fc604767 | 1937 | .hooknum = NF_INET_LOCAL_OUT, |
afb523c5 | 1938 | .priority = NF_IP6_PRI_NAT_DST + 1, |
473b23d3 | 1939 | }, |
cb59155f JA |
1940 | /* After mangle, schedule and forward local requests */ |
1941 | { | |
1942 | .hook = ip_vs_local_request6, | |
1943 | .owner = THIS_MODULE, | |
4c809d63 | 1944 | .pf = NFPROTO_IPV6, |
cb59155f | 1945 | .hooknum = NF_INET_LOCAL_OUT, |
afb523c5 | 1946 | .priority = NF_IP6_PRI_NAT_DST + 2, |
cb59155f | 1947 | }, |
473b23d3 JV |
1948 | /* After packet filtering (but before ip_vs_out_icmp), catch icmp |
1949 | * destined for 0.0.0.0/0, which is for incoming IPVS connections */ | |
1950 | { | |
1951 | .hook = ip_vs_forward_icmp_v6, | |
1952 | .owner = THIS_MODULE, | |
4c809d63 | 1953 | .pf = NFPROTO_IPV6, |
cb59155f JA |
1954 | .hooknum = NF_INET_FORWARD, |
1955 | .priority = 99, | |
473b23d3 | 1956 | }, |
fc604767 JA |
1957 | /* After packet filtering, change source only for VS/NAT */ |
1958 | { | |
1959 | .hook = ip_vs_reply6, | |
1960 | .owner = THIS_MODULE, | |
4c809d63 | 1961 | .pf = NFPROTO_IPV6, |
fc604767 JA |
1962 | .hooknum = NF_INET_FORWARD, |
1963 | .priority = 100, | |
1964 | }, | |
473b23d3 | 1965 | #endif |
1da177e4 | 1966 | }; |
61b1ab45 HS |
1967 | /* |
1968 | * Initialize IP Virtual Server netns mem. | |
1969 | */ | |
1970 | static int __net_init __ip_vs_init(struct net *net) | |
1971 | { | |
1972 | struct netns_ipvs *ipvs; | |
1973 | ||
61b1ab45 | 1974 | ipvs = net_generic(net, ip_vs_net_id); |
0a9ee813 | 1975 | if (ipvs == NULL) |
61b1ab45 | 1976 | return -ENOMEM; |
0a9ee813 | 1977 | |
7a4f0761 HS |
1978 | /* Hold the beast until a service is registerd */ |
1979 | ipvs->enable = 0; | |
f6340ee0 | 1980 | ipvs->net = net; |
61b1ab45 HS |
1981 | /* Counters used for creating unique names */ |
1982 | ipvs->gen = atomic_read(&ipvs_netns_cnt); | |
1983 | atomic_inc(&ipvs_netns_cnt); | |
1984 | net->ipvs = ipvs; | |
7a4f0761 | 1985 | |
503cf15a | 1986 | if (ip_vs_estimator_net_init(net) < 0) |
7a4f0761 HS |
1987 | goto estimator_fail; |
1988 | ||
503cf15a | 1989 | if (ip_vs_control_net_init(net) < 0) |
7a4f0761 HS |
1990 | goto control_fail; |
1991 | ||
503cf15a | 1992 | if (ip_vs_protocol_net_init(net) < 0) |
7a4f0761 HS |
1993 | goto protocol_fail; |
1994 | ||
503cf15a | 1995 | if (ip_vs_app_net_init(net) < 0) |
7a4f0761 HS |
1996 | goto app_fail; |
1997 | ||
503cf15a | 1998 | if (ip_vs_conn_net_init(net) < 0) |
7a4f0761 HS |
1999 | goto conn_fail; |
2000 | ||
503cf15a | 2001 | if (ip_vs_sync_net_init(net) < 0) |
7a4f0761 HS |
2002 | goto sync_fail; |
2003 | ||
a870c8c5 | 2004 | printk(KERN_INFO "IPVS: Creating netns size=%zu id=%d\n", |
61b1ab45 HS |
2005 | sizeof(struct netns_ipvs), ipvs->gen); |
2006 | return 0; | |
7a4f0761 HS |
2007 | /* |
2008 | * Error handling | |
2009 | */ | |
2010 | ||
2011 | sync_fail: | |
503cf15a | 2012 | ip_vs_conn_net_cleanup(net); |
7a4f0761 | 2013 | conn_fail: |
503cf15a | 2014 | ip_vs_app_net_cleanup(net); |
7a4f0761 | 2015 | app_fail: |
503cf15a | 2016 | ip_vs_protocol_net_cleanup(net); |
7a4f0761 | 2017 | protocol_fail: |
503cf15a | 2018 | ip_vs_control_net_cleanup(net); |
7a4f0761 | 2019 | control_fail: |
503cf15a | 2020 | ip_vs_estimator_net_cleanup(net); |
7a4f0761 | 2021 | estimator_fail: |
39f618b4 | 2022 | net->ipvs = NULL; |
7a4f0761 | 2023 | return -ENOMEM; |
61b1ab45 HS |
2024 | } |
2025 | ||
2026 | static void __net_exit __ip_vs_cleanup(struct net *net) | |
2027 | { | |
503cf15a HS |
2028 | ip_vs_service_net_cleanup(net); /* ip_vs_flush() with locks */ |
2029 | ip_vs_conn_net_cleanup(net); | |
2030 | ip_vs_app_net_cleanup(net); | |
2031 | ip_vs_protocol_net_cleanup(net); | |
2032 | ip_vs_control_net_cleanup(net); | |
2033 | ip_vs_estimator_net_cleanup(net); | |
1ae132b0 | 2034 | IP_VS_DBG(2, "ipvs netns %d released\n", net_ipvs(net)->gen); |
39f618b4 | 2035 | net->ipvs = NULL; |
61b1ab45 HS |
2036 | } |
2037 | ||
7a4f0761 HS |
2038 | static void __net_exit __ip_vs_dev_cleanup(struct net *net) |
2039 | { | |
2040 | EnterFunction(2); | |
2041 | net_ipvs(net)->enable = 0; /* Disable packet reception */ | |
8f4e0a18 | 2042 | smp_wmb(); |
503cf15a | 2043 | ip_vs_sync_net_cleanup(net); |
7a4f0761 HS |
2044 | LeaveFunction(2); |
2045 | } | |
2046 | ||
61b1ab45 HS |
2047 | static struct pernet_operations ipvs_core_ops = { |
2048 | .init = __ip_vs_init, | |
2049 | .exit = __ip_vs_cleanup, | |
2050 | .id = &ip_vs_net_id, | |
2051 | .size = sizeof(struct netns_ipvs), | |
2052 | }; | |
1da177e4 | 2053 | |
7a4f0761 HS |
2054 | static struct pernet_operations ipvs_core_dev_ops = { |
2055 | .exit = __ip_vs_dev_cleanup, | |
2056 | }; | |
2057 | ||
1da177e4 LT |
2058 | /* |
2059 | * Initialize IP Virtual Server | |
2060 | */ | |
2061 | static int __init ip_vs_init(void) | |
2062 | { | |
2063 | int ret; | |
2064 | ||
2065 | ret = ip_vs_control_init(); | |
2066 | if (ret < 0) { | |
1e3e238e | 2067 | pr_err("can't setup control.\n"); |
6c8f7949 | 2068 | goto exit; |
1da177e4 LT |
2069 | } |
2070 | ||
2071 | ip_vs_protocol_init(); | |
2072 | ||
1da177e4 LT |
2073 | ret = ip_vs_conn_init(); |
2074 | if (ret < 0) { | |
1e3e238e | 2075 | pr_err("can't setup connection table.\n"); |
6c8f7949 | 2076 | goto cleanup_protocol; |
61b1ab45 HS |
2077 | } |
2078 | ||
7a4f0761 HS |
2079 | ret = register_pernet_subsys(&ipvs_core_ops); /* Alloc ip_vs struct */ |
2080 | if (ret < 0) | |
6c8f7949 | 2081 | goto cleanup_conn; |
7a4f0761 HS |
2082 | |
2083 | ret = register_pernet_device(&ipvs_core_dev_ops); | |
2084 | if (ret < 0) | |
2085 | goto cleanup_sub; | |
2086 | ||
41c5b317 | 2087 | ret = nf_register_hooks(ip_vs_ops, ARRAY_SIZE(ip_vs_ops)); |
1da177e4 | 2088 | if (ret < 0) { |
1e3e238e | 2089 | pr_err("can't register hooks.\n"); |
7a4f0761 | 2090 | goto cleanup_dev; |
1da177e4 LT |
2091 | } |
2092 | ||
8537de8a HS |
2093 | ret = ip_vs_register_nl_ioctl(); |
2094 | if (ret < 0) { | |
2095 | pr_err("can't register netlink/ioctl.\n"); | |
2096 | goto cleanup_hooks; | |
2097 | } | |
2098 | ||
1e3e238e | 2099 | pr_info("ipvs loaded.\n"); |
7a4f0761 | 2100 | |
1da177e4 LT |
2101 | return ret; |
2102 | ||
8537de8a HS |
2103 | cleanup_hooks: |
2104 | nf_unregister_hooks(ip_vs_ops, ARRAY_SIZE(ip_vs_ops)); | |
7a4f0761 HS |
2105 | cleanup_dev: |
2106 | unregister_pernet_device(&ipvs_core_dev_ops); | |
2107 | cleanup_sub: | |
2108 | unregister_pernet_subsys(&ipvs_core_ops); | |
552ad65a | 2109 | cleanup_conn: |
1da177e4 | 2110 | ip_vs_conn_cleanup(); |
552ad65a | 2111 | cleanup_protocol: |
1da177e4 LT |
2112 | ip_vs_protocol_cleanup(); |
2113 | ip_vs_control_cleanup(); | |
6c8f7949 | 2114 | exit: |
1da177e4 LT |
2115 | return ret; |
2116 | } | |
2117 | ||
2118 | static void __exit ip_vs_cleanup(void) | |
2119 | { | |
8537de8a | 2120 | ip_vs_unregister_nl_ioctl(); |
41c5b317 | 2121 | nf_unregister_hooks(ip_vs_ops, ARRAY_SIZE(ip_vs_ops)); |
7a4f0761 HS |
2122 | unregister_pernet_device(&ipvs_core_dev_ops); |
2123 | unregister_pernet_subsys(&ipvs_core_ops); /* free ip_vs struct */ | |
1da177e4 | 2124 | ip_vs_conn_cleanup(); |
1da177e4 LT |
2125 | ip_vs_protocol_cleanup(); |
2126 | ip_vs_control_cleanup(); | |
1e3e238e | 2127 | pr_info("ipvs unloaded.\n"); |
1da177e4 LT |
2128 | } |
2129 | ||
2130 | module_init(ip_vs_init); | |
2131 | module_exit(ip_vs_cleanup); | |
2132 | MODULE_LICENSE("GPL"); |