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); | |
ceec4c38 | 206 | p->pe = rcu_dereference(svc->pe); |
85999283 | 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) | |
0a925864 JA |
238 | ipv6_addr_prefix(&snet.in6, &iph->saddr.in6, |
239 | (__force __u32) svc->netmask); | |
cd17f9ed JV |
240 | else |
241 | #endif | |
d4383f04 | 242 | snet.ip = iph->saddr.ip & svc->netmask; |
1da177e4 | 243 | |
cd17f9ed JV |
244 | IP_VS_DBG_BUF(6, "p-schedule: src %s:%u dest %s:%u " |
245 | "mnet %s\n", | |
d4383f04 JDB |
246 | IP_VS_DBG_ADDR(svc->af, &iph->saddr), ntohs(src_port), |
247 | IP_VS_DBG_ADDR(svc->af, &iph->daddr), ntohs(dst_port), | |
cd17f9ed | 248 | IP_VS_DBG_ADDR(svc->af, &snet)); |
1da177e4 LT |
249 | |
250 | /* | |
251 | * As far as we know, FTP is a very complicated network protocol, and | |
252 | * it uses control connection and data connections. For active FTP, | |
253 | * FTP server initialize data connection to the client, its source port | |
254 | * is often 20. For passive FTP, FTP server tells the clients the port | |
255 | * that it passively listens to, and the client issues the data | |
256 | * connection. In the tunneling or direct routing mode, the load | |
257 | * balancer is on the client-to-server half of connection, the port | |
258 | * number is unknown to the load balancer. So, a conn template like | |
259 | * <caddr, 0, vaddr, 0, daddr, 0> is created for persistent FTP | |
260 | * service, and a template like <caddr, 0, vaddr, vport, daddr, dport> | |
261 | * is created for other persistent services. | |
262 | */ | |
5b57a98c | 263 | { |
d4383f04 JDB |
264 | int protocol = iph->protocol; |
265 | const union nf_inet_addr *vaddr = &iph->daddr; | |
f11017ec SH |
266 | __be16 vport = 0; |
267 | ||
ce144f24 | 268 | if (dst_port == svc->port) { |
5b57a98c SH |
269 | /* non-FTP template: |
270 | * <protocol, caddr, 0, vaddr, vport, daddr, dport> | |
271 | * FTP template: | |
272 | * <protocol, caddr, 0, vaddr, 0, daddr, 0> | |
1da177e4 LT |
273 | */ |
274 | if (svc->port != FTPPORT) | |
ce144f24 | 275 | vport = dst_port; |
1da177e4 | 276 | } else { |
5b57a98c SH |
277 | /* Note: persistent fwmark-based services and |
278 | * persistent port zero service are handled here. | |
279 | * fwmark template: | |
280 | * <IPPROTO_IP,caddr,0,fwmark,0,daddr,0> | |
281 | * port zero template: | |
282 | * <protocol,caddr,0,vaddr,0,daddr,0> | |
1da177e4 | 283 | */ |
28364a59 | 284 | if (svc->fwmark) { |
5b57a98c SH |
285 | protocol = IPPROTO_IP; |
286 | vaddr = &fwmark; | |
287 | } | |
1da177e4 | 288 | } |
a5959d53 HS |
289 | /* return *ignored = -1 so NF_DROP can be used */ |
290 | if (ip_vs_conn_fill_param_persist(svc, skb, protocol, &snet, 0, | |
291 | vaddr, vport, ¶m) < 0) { | |
292 | *ignored = -1; | |
293 | return NULL; | |
294 | } | |
1da177e4 LT |
295 | } |
296 | ||
5b57a98c | 297 | /* Check if a template already exists */ |
f11017ec | 298 | ct = ip_vs_ct_in_get(¶m); |
5b57a98c | 299 | if (!ct || !ip_vs_check_template(ct)) { |
ceec4c38 JA |
300 | struct ip_vs_scheduler *sched; |
301 | ||
a5959d53 HS |
302 | /* |
303 | * No template found or the dest of the connection | |
5b57a98c | 304 | * template is not available. |
a5959d53 | 305 | * return *ignored=0 i.e. ICMP and NF_DROP |
5b57a98c | 306 | */ |
ceec4c38 JA |
307 | sched = rcu_dereference(svc->scheduler); |
308 | dest = sched->schedule(svc, skb); | |
5b57a98c SH |
309 | if (!dest) { |
310 | IP_VS_DBG(1, "p-schedule: no dest found.\n"); | |
85999283 | 311 | kfree(param.pe_data); |
a5959d53 | 312 | *ignored = 0; |
5b57a98c SH |
313 | return NULL; |
314 | } | |
315 | ||
ce144f24 | 316 | if (dst_port == svc->port && svc->port != FTPPORT) |
5b57a98c SH |
317 | dport = dest->port; |
318 | ||
85999283 SH |
319 | /* Create a template |
320 | * This adds param.pe_data to the template, | |
321 | * and thus param.pe_data will be destroyed | |
322 | * when the template expires */ | |
f11017ec | 323 | ct = ip_vs_conn_new(¶m, &dest->addr, dport, |
0e051e68 | 324 | IP_VS_CONN_F_TEMPLATE, dest, skb->mark); |
85999283 SH |
325 | if (ct == NULL) { |
326 | kfree(param.pe_data); | |
a5959d53 | 327 | *ignored = -1; |
5b57a98c | 328 | return NULL; |
85999283 | 329 | } |
5b57a98c SH |
330 | |
331 | ct->timeout = svc->timeout; | |
85999283 | 332 | } else { |
5b57a98c SH |
333 | /* set destination with the found template */ |
334 | dest = ct->dest; | |
85999283 SH |
335 | kfree(param.pe_data); |
336 | } | |
5b57a98c | 337 | |
ce144f24 | 338 | dport = dst_port; |
5b57a98c SH |
339 | if (dport == svc->port && dest->port) |
340 | dport = dest->port; | |
341 | ||
26ec037f | 342 | flags = (svc->flags & IP_VS_SVC_F_ONEPACKET |
d4383f04 | 343 | && iph->protocol == IPPROTO_UDP) ? |
26ec037f NC |
344 | IP_VS_CONN_F_ONE_PACKET : 0; |
345 | ||
1da177e4 LT |
346 | /* |
347 | * Create a new connection according to the template | |
348 | */ | |
d4383f04 JDB |
349 | ip_vs_conn_fill_param(svc->net, svc->af, iph->protocol, &iph->saddr, |
350 | src_port, &iph->daddr, dst_port, ¶m); | |
ce144f24 | 351 | |
0e051e68 | 352 | cp = ip_vs_conn_new(¶m, &dest->addr, dport, flags, dest, skb->mark); |
1da177e4 LT |
353 | if (cp == NULL) { |
354 | ip_vs_conn_put(ct); | |
a5959d53 | 355 | *ignored = -1; |
1da177e4 LT |
356 | return NULL; |
357 | } | |
358 | ||
359 | /* | |
360 | * Add its control | |
361 | */ | |
362 | ip_vs_control_add(cp, ct); | |
363 | ip_vs_conn_put(ct); | |
364 | ||
365 | ip_vs_conn_stats(cp, svc); | |
366 | return cp; | |
367 | } | |
368 | ||
369 | ||
370 | /* | |
371 | * IPVS main scheduling function | |
372 | * It selects a server according to the virtual service, and | |
373 | * creates a connection entry. | |
374 | * Protocols supported: TCP, UDP | |
a5959d53 HS |
375 | * |
376 | * Usage of *ignored | |
377 | * | |
378 | * 1 : protocol tried to schedule (eg. on SYN), found svc but the | |
379 | * svc/scheduler decides that this packet should be accepted with | |
380 | * NF_ACCEPT because it must not be scheduled. | |
381 | * | |
382 | * 0 : scheduler can not find destination, so try bypass or | |
383 | * return ICMP and then NF_DROP (ip_vs_leave). | |
384 | * | |
385 | * -1 : scheduler tried to schedule but fatal error occurred, eg. | |
386 | * ip_vs_conn_new failure (ENOMEM) or ip_vs_sip_fill_param | |
387 | * failure such as missing Call-ID, ENOMEM on skb_linearize | |
388 | * or pe_data. In this case we should return NF_DROP without | |
389 | * any attempts to send ICMP with ip_vs_leave. | |
1da177e4 LT |
390 | */ |
391 | struct ip_vs_conn * | |
190ecd27 | 392 | ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb, |
d4383f04 JDB |
393 | struct ip_vs_proto_data *pd, int *ignored, |
394 | struct ip_vs_iphdr *iph) | |
1da177e4 | 395 | { |
9330419d | 396 | struct ip_vs_protocol *pp = pd->pp; |
1da177e4 | 397 | struct ip_vs_conn *cp = NULL; |
ceec4c38 | 398 | struct ip_vs_scheduler *sched; |
1da177e4 | 399 | struct ip_vs_dest *dest; |
3575792e JA |
400 | __be16 _ports[2], *pptr; |
401 | unsigned int flags; | |
1da177e4 | 402 | |
190ecd27 | 403 | *ignored = 1; |
2f74713d JDB |
404 | /* |
405 | * IPv6 frags, only the first hit here. | |
406 | */ | |
d4383f04 | 407 | pptr = frag_safe_skb_hp(skb, iph->len, sizeof(_ports), _ports, iph); |
1da177e4 LT |
408 | if (pptr == NULL) |
409 | return NULL; | |
410 | ||
190ecd27 JA |
411 | /* |
412 | * FTPDATA needs this check when using local real server. | |
413 | * Never schedule Active FTPDATA connections from real server. | |
414 | * For LVS-NAT they must be already created. For other methods | |
415 | * with persistence the connection is created on SYN+ACK. | |
416 | */ | |
417 | if (pptr[0] == FTPDATA) { | |
0d79641a JA |
418 | IP_VS_DBG_PKT(12, svc->af, pp, skb, 0, |
419 | "Not scheduling FTPDATA"); | |
190ecd27 JA |
420 | return NULL; |
421 | } | |
422 | ||
423 | /* | |
a5959d53 | 424 | * Do not schedule replies from local real server. |
190ecd27 JA |
425 | */ |
426 | if ((!skb->dev || skb->dev->flags & IFF_LOOPBACK) && | |
d4383f04 | 427 | (cp = pp->conn_in_get(svc->af, skb, iph, 1))) { |
0d79641a | 428 | IP_VS_DBG_PKT(12, svc->af, pp, skb, 0, |
190ecd27 JA |
429 | "Not scheduling reply for existing connection"); |
430 | __ip_vs_conn_put(cp); | |
431 | return NULL; | |
432 | } | |
433 | ||
1da177e4 LT |
434 | /* |
435 | * Persistent service | |
436 | */ | |
a5959d53 | 437 | if (svc->flags & IP_VS_SVC_F_PERSISTENT) |
d4383f04 JDB |
438 | return ip_vs_sched_persist(svc, skb, pptr[0], pptr[1], ignored, |
439 | iph); | |
a5959d53 HS |
440 | |
441 | *ignored = 0; | |
1da177e4 LT |
442 | |
443 | /* | |
444 | * Non-persistent service | |
445 | */ | |
446 | if (!svc->fwmark && pptr[1] != svc->port) { | |
447 | if (!svc->port) | |
1e3e238e HE |
448 | pr_err("Schedule: port zero only supported " |
449 | "in persistent services, " | |
450 | "check your ipvs configuration\n"); | |
1da177e4 LT |
451 | return NULL; |
452 | } | |
453 | ||
ceec4c38 JA |
454 | sched = rcu_dereference(svc->scheduler); |
455 | dest = sched->schedule(svc, skb); | |
1da177e4 LT |
456 | if (dest == NULL) { |
457 | IP_VS_DBG(1, "Schedule: no dest found.\n"); | |
458 | return NULL; | |
459 | } | |
460 | ||
26ec037f | 461 | flags = (svc->flags & IP_VS_SVC_F_ONEPACKET |
d4383f04 | 462 | && iph->protocol == IPPROTO_UDP) ? |
26ec037f NC |
463 | IP_VS_CONN_F_ONE_PACKET : 0; |
464 | ||
1da177e4 LT |
465 | /* |
466 | * Create a connection entry. | |
467 | */ | |
f11017ec SH |
468 | { |
469 | struct ip_vs_conn_param p; | |
6e67e586 | 470 | |
d4383f04 JDB |
471 | ip_vs_conn_fill_param(svc->net, svc->af, iph->protocol, |
472 | &iph->saddr, pptr[0], &iph->daddr, | |
473 | pptr[1], &p); | |
f11017ec SH |
474 | cp = ip_vs_conn_new(&p, &dest->addr, |
475 | dest->port ? dest->port : pptr[1], | |
0e051e68 | 476 | flags, dest, skb->mark); |
a5959d53 HS |
477 | if (!cp) { |
478 | *ignored = -1; | |
f11017ec | 479 | return NULL; |
a5959d53 | 480 | } |
f11017ec | 481 | } |
1da177e4 | 482 | |
cd17f9ed JV |
483 | IP_VS_DBG_BUF(6, "Schedule fwd:%c c:%s:%u v:%s:%u " |
484 | "d:%s:%u conn->flags:%X conn->refcnt:%d\n", | |
485 | ip_vs_fwd_tag(cp), | |
486 | IP_VS_DBG_ADDR(svc->af, &cp->caddr), ntohs(cp->cport), | |
487 | IP_VS_DBG_ADDR(svc->af, &cp->vaddr), ntohs(cp->vport), | |
488 | IP_VS_DBG_ADDR(svc->af, &cp->daddr), ntohs(cp->dport), | |
489 | cp->flags, atomic_read(&cp->refcnt)); | |
1da177e4 LT |
490 | |
491 | ip_vs_conn_stats(cp, svc); | |
492 | return cp; | |
493 | } | |
494 | ||
495 | ||
496 | /* | |
497 | * Pass or drop the packet. | |
498 | * Called by ip_vs_in, when the virtual service is available but | |
499 | * no destination is available for a new connection. | |
500 | */ | |
501 | int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb, | |
d4383f04 | 502 | struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph) |
1da177e4 | 503 | { |
014d730d | 504 | __be16 _ports[2], *pptr; |
a7a86b86 SH |
505 | #ifdef CONFIG_SYSCTL |
506 | struct net *net; | |
507 | struct netns_ipvs *ipvs; | |
2a3b791e | 508 | int unicast; |
a7a86b86 | 509 | #endif |
9330419d | 510 | |
d4383f04 | 511 | pptr = frag_safe_skb_hp(skb, iph->len, sizeof(_ports), _ports, iph); |
1da177e4 | 512 | if (pptr == NULL) { |
1da177e4 LT |
513 | return NF_DROP; |
514 | } | |
a7a86b86 SH |
515 | |
516 | #ifdef CONFIG_SYSCTL | |
4a98480b | 517 | net = skb_net(skb); |
1da177e4 | 518 | |
2a3b791e JV |
519 | #ifdef CONFIG_IP_VS_IPV6 |
520 | if (svc->af == AF_INET6) | |
d4383f04 | 521 | unicast = ipv6_addr_type(&iph->daddr.in6) & IPV6_ADDR_UNICAST; |
2a3b791e JV |
522 | else |
523 | #endif | |
d4383f04 | 524 | unicast = (inet_addr_type(net, iph->daddr.ip) == RTN_UNICAST); |
2a3b791e | 525 | |
1da177e4 | 526 | /* if it is fwmark-based service, the cache_bypass sysctl is up |
2a3b791e | 527 | and the destination is a non-local unicast, then create |
1da177e4 | 528 | a cache_bypass connection entry */ |
4a98480b | 529 | ipvs = net_ipvs(net); |
a0840e2e | 530 | if (ipvs->sysctl_cache_bypass && svc->fwmark && unicast) { |
ad542ced | 531 | int ret; |
1da177e4 | 532 | struct ip_vs_conn *cp; |
3575792e | 533 | unsigned int flags = (svc->flags & IP_VS_SVC_F_ONEPACKET && |
d4383f04 | 534 | iph->protocol == IPPROTO_UDP) ? |
3575792e | 535 | IP_VS_CONN_F_ONE_PACKET : 0; |
dff630dd | 536 | union nf_inet_addr daddr = { .all = { 0, 0, 0, 0 } }; |
1da177e4 | 537 | |
1da177e4 | 538 | /* create a new connection entry */ |
1e3e238e | 539 | IP_VS_DBG(6, "%s(): create a cache_bypass entry\n", __func__); |
f11017ec SH |
540 | { |
541 | struct ip_vs_conn_param p; | |
d4383f04 JDB |
542 | ip_vs_conn_fill_param(svc->net, svc->af, iph->protocol, |
543 | &iph->saddr, pptr[0], | |
544 | &iph->daddr, pptr[1], &p); | |
f11017ec SH |
545 | cp = ip_vs_conn_new(&p, &daddr, 0, |
546 | IP_VS_CONN_F_BYPASS | flags, | |
0e051e68 | 547 | NULL, skb->mark); |
f11017ec SH |
548 | if (!cp) |
549 | return NF_DROP; | |
550 | } | |
1da177e4 LT |
551 | |
552 | /* statistics */ | |
553 | ip_vs_in_stats(cp, skb); | |
554 | ||
555 | /* set state */ | |
4a516f11 | 556 | ip_vs_set_state(cp, IP_VS_DIR_INPUT, skb, pd); |
1da177e4 LT |
557 | |
558 | /* transmit the first SYN packet */ | |
d4383f04 | 559 | ret = cp->packet_xmit(skb, cp, pd->pp, iph); |
1da177e4 LT |
560 | /* do not touch skb anymore */ |
561 | ||
562 | atomic_inc(&cp->in_pkts); | |
563 | ip_vs_conn_put(cp); | |
564 | return ret; | |
565 | } | |
a7a86b86 | 566 | #endif |
1da177e4 LT |
567 | |
568 | /* | |
569 | * When the virtual ftp service is presented, packets destined | |
570 | * for other services on the VIP may get here (except services | |
571 | * listed in the ipvs table), pass the packets, because it is | |
572 | * not ipvs job to decide to drop the packets. | |
573 | */ | |
ceec4c38 | 574 | if ((svc->port == FTPPORT) && (pptr[1] != FTPPORT)) |
1da177e4 | 575 | return NF_ACCEPT; |
1da177e4 LT |
576 | |
577 | /* | |
578 | * Notify the client that the destination is unreachable, and | |
579 | * release the socket buffer. | |
580 | * Since it is in IP layer, the TCP socket is not actually | |
581 | * created, the TCP RST packet cannot be sent, instead that | |
582 | * ICMP_PORT_UNREACH is sent here no matter it is TCP/UDP. --WZ | |
583 | */ | |
2a3b791e | 584 | #ifdef CONFIG_IP_VS_IPV6 |
cb59155f JA |
585 | if (svc->af == AF_INET6) { |
586 | if (!skb->dev) { | |
9fd0fa7a | 587 | struct net *net_ = dev_net(skb_dst(skb)->dev); |
cb59155f | 588 | |
9fd0fa7a | 589 | skb->dev = net_->loopback_dev; |
cb59155f | 590 | } |
3ffe533c | 591 | icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); |
cb59155f | 592 | } else |
2a3b791e JV |
593 | #endif |
594 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); | |
595 | ||
1da177e4 LT |
596 | return NF_DROP; |
597 | } | |
598 | ||
84b3cee3 SH |
599 | #ifdef CONFIG_SYSCTL |
600 | ||
601 | static int sysctl_snat_reroute(struct sk_buff *skb) | |
602 | { | |
603 | struct netns_ipvs *ipvs = net_ipvs(skb_net(skb)); | |
604 | return ipvs->sysctl_snat_reroute; | |
605 | } | |
606 | ||
0cfa558e SH |
607 | static int sysctl_nat_icmp_send(struct net *net) |
608 | { | |
609 | struct netns_ipvs *ipvs = net_ipvs(net); | |
610 | return ipvs->sysctl_nat_icmp_send; | |
611 | } | |
612 | ||
71a8ab6c SH |
613 | static int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs) |
614 | { | |
615 | return ipvs->sysctl_expire_nodest_conn; | |
616 | } | |
617 | ||
84b3cee3 SH |
618 | #else |
619 | ||
620 | static int sysctl_snat_reroute(struct sk_buff *skb) { return 0; } | |
0cfa558e | 621 | static int sysctl_nat_icmp_send(struct net *net) { return 0; } |
71a8ab6c | 622 | static int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs) { return 0; } |
84b3cee3 SH |
623 | |
624 | #endif | |
625 | ||
b1550f22 | 626 | __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset) |
1da177e4 | 627 | { |
d3bc23e7 | 628 | return csum_fold(skb_checksum(skb, offset, skb->len - offset, 0)); |
1da177e4 LT |
629 | } |
630 | ||
1ca5bb54 JA |
631 | static inline enum ip_defrag_users ip_vs_defrag_user(unsigned int hooknum) |
632 | { | |
633 | if (NF_INET_LOCAL_IN == hooknum) | |
634 | return IP_DEFRAG_VS_IN; | |
635 | if (NF_INET_FORWARD == hooknum) | |
636 | return IP_DEFRAG_VS_FWD; | |
637 | return IP_DEFRAG_VS_OUT; | |
638 | } | |
639 | ||
776c729e | 640 | static inline int ip_vs_gather_frags(struct sk_buff *skb, u_int32_t user) |
1da177e4 | 641 | { |
ac69269a | 642 | int err; |
776c729e | 643 | |
ac69269a JA |
644 | local_bh_disable(); |
645 | err = ip_defrag(skb, user); | |
646 | local_bh_enable(); | |
776c729e | 647 | if (!err) |
eddc9ec5 | 648 | ip_send_check(ip_hdr(skb)); |
776c729e HX |
649 | |
650 | return err; | |
1da177e4 LT |
651 | } |
652 | ||
96b077b4 JA |
653 | static int ip_vs_route_me_harder(int af, struct sk_buff *skb, |
654 | unsigned int hooknum) | |
ba4fd7e9 | 655 | { |
96b077b4 JA |
656 | if (!sysctl_snat_reroute(skb)) |
657 | return 0; | |
658 | /* Reroute replies only to remote clients (FORWARD and LOCAL_OUT) */ | |
659 | if (NF_INET_LOCAL_IN == hooknum) | |
660 | return 0; | |
ba4fd7e9 SH |
661 | #ifdef CONFIG_IP_VS_IPV6 |
662 | if (af == AF_INET6) { | |
96b077b4 JA |
663 | struct dst_entry *dst = skb_dst(skb); |
664 | ||
665 | if (dst->dev && !(dst->dev->flags & IFF_LOOPBACK) && | |
666 | ip6_route_me_harder(skb) != 0) | |
ba4fd7e9 SH |
667 | return 1; |
668 | } else | |
669 | #endif | |
96b077b4 | 670 | if (!(skb_rtable(skb)->rt_flags & RTCF_LOCAL) && |
ba4fd7e9 SH |
671 | ip_route_me_harder(skb, RTN_LOCAL) != 0) |
672 | return 1; | |
673 | ||
674 | return 0; | |
675 | } | |
676 | ||
1da177e4 LT |
677 | /* |
678 | * Packet has been made sufficiently writable in caller | |
679 | * - inout: 1=in->out, 0=out->in | |
680 | */ | |
681 | void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp, | |
682 | struct ip_vs_conn *cp, int inout) | |
683 | { | |
eddc9ec5 | 684 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 | 685 | unsigned int icmp_offset = iph->ihl*4; |
d56f90a7 ACM |
686 | struct icmphdr *icmph = (struct icmphdr *)(skb_network_header(skb) + |
687 | icmp_offset); | |
1da177e4 LT |
688 | struct iphdr *ciph = (struct iphdr *)(icmph + 1); |
689 | ||
690 | if (inout) { | |
e7ade46a | 691 | iph->saddr = cp->vaddr.ip; |
1da177e4 | 692 | ip_send_check(iph); |
e7ade46a | 693 | ciph->daddr = cp->vaddr.ip; |
1da177e4 LT |
694 | ip_send_check(ciph); |
695 | } else { | |
e7ade46a | 696 | iph->daddr = cp->daddr.ip; |
1da177e4 | 697 | ip_send_check(iph); |
e7ade46a | 698 | ciph->saddr = cp->daddr.ip; |
1da177e4 LT |
699 | ip_send_check(ciph); |
700 | } | |
701 | ||
2906f66a VMR |
702 | /* the TCP/UDP/SCTP port */ |
703 | if (IPPROTO_TCP == ciph->protocol || IPPROTO_UDP == ciph->protocol || | |
704 | IPPROTO_SCTP == ciph->protocol) { | |
014d730d | 705 | __be16 *ports = (void *)ciph + ciph->ihl*4; |
1da177e4 LT |
706 | |
707 | if (inout) | |
708 | ports[1] = cp->vport; | |
709 | else | |
710 | ports[0] = cp->dport; | |
711 | } | |
712 | ||
713 | /* And finally the ICMP checksum */ | |
714 | icmph->checksum = 0; | |
715 | icmph->checksum = ip_vs_checksum_complete(skb, icmp_offset); | |
716 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
717 | ||
718 | if (inout) | |
0d79641a | 719 | IP_VS_DBG_PKT(11, AF_INET, pp, skb, (void *)ciph - (void *)iph, |
1da177e4 LT |
720 | "Forwarding altered outgoing ICMP"); |
721 | else | |
0d79641a | 722 | IP_VS_DBG_PKT(11, AF_INET, pp, skb, (void *)ciph - (void *)iph, |
1da177e4 LT |
723 | "Forwarding altered incoming ICMP"); |
724 | } | |
725 | ||
b3cdd2a7 JV |
726 | #ifdef CONFIG_IP_VS_IPV6 |
727 | void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp, | |
728 | struct ip_vs_conn *cp, int inout) | |
729 | { | |
730 | struct ipv6hdr *iph = ipv6_hdr(skb); | |
63dca2c0 JDB |
731 | unsigned int icmp_offset = 0; |
732 | unsigned int offs = 0; /* header offset*/ | |
733 | int protocol; | |
734 | struct icmp6hdr *icmph; | |
735 | struct ipv6hdr *ciph; | |
736 | unsigned short fragoffs; | |
737 | ||
738 | ipv6_find_hdr(skb, &icmp_offset, IPPROTO_ICMPV6, &fragoffs, NULL); | |
739 | icmph = (struct icmp6hdr *)(skb_network_header(skb) + icmp_offset); | |
740 | offs = icmp_offset + sizeof(struct icmp6hdr); | |
741 | ciph = (struct ipv6hdr *)(skb_network_header(skb) + offs); | |
742 | ||
743 | protocol = ipv6_find_hdr(skb, &offs, -1, &fragoffs, NULL); | |
b3cdd2a7 JV |
744 | |
745 | if (inout) { | |
746 | iph->saddr = cp->vaddr.in6; | |
747 | ciph->daddr = cp->vaddr.in6; | |
748 | } else { | |
749 | iph->daddr = cp->daddr.in6; | |
750 | ciph->saddr = cp->daddr.in6; | |
751 | } | |
752 | ||
2906f66a | 753 | /* the TCP/UDP/SCTP port */ |
63dca2c0 JDB |
754 | if (!fragoffs && (IPPROTO_TCP == protocol || IPPROTO_UDP == protocol || |
755 | IPPROTO_SCTP == protocol)) { | |
756 | __be16 *ports = (void *)(skb_network_header(skb) + offs); | |
b3cdd2a7 | 757 | |
63dca2c0 JDB |
758 | IP_VS_DBG(11, "%s() changed port %d to %d\n", __func__, |
759 | ntohs(inout ? ports[1] : ports[0]), | |
760 | ntohs(inout ? cp->vport : cp->dport)); | |
b3cdd2a7 JV |
761 | if (inout) |
762 | ports[1] = cp->vport; | |
763 | else | |
764 | ports[0] = cp->dport; | |
765 | } | |
766 | ||
767 | /* And finally the ICMP checksum */ | |
8870f842 SH |
768 | icmph->icmp6_cksum = ~csum_ipv6_magic(&iph->saddr, &iph->daddr, |
769 | skb->len - icmp_offset, | |
770 | IPPROTO_ICMPV6, 0); | |
771 | skb->csum_start = skb_network_header(skb) - skb->head + icmp_offset; | |
772 | skb->csum_offset = offsetof(struct icmp6hdr, icmp6_cksum); | |
773 | skb->ip_summed = CHECKSUM_PARTIAL; | |
b3cdd2a7 JV |
774 | |
775 | if (inout) | |
0d79641a JA |
776 | IP_VS_DBG_PKT(11, AF_INET6, pp, skb, |
777 | (void *)ciph - (void *)iph, | |
778 | "Forwarding altered outgoing ICMPv6"); | |
b3cdd2a7 | 779 | else |
0d79641a JA |
780 | IP_VS_DBG_PKT(11, AF_INET6, pp, skb, |
781 | (void *)ciph - (void *)iph, | |
782 | "Forwarding altered incoming ICMPv6"); | |
b3cdd2a7 JV |
783 | } |
784 | #endif | |
785 | ||
4856c84c | 786 | /* Handle relevant response ICMP messages - forward to the right |
6cb90db5 | 787 | * destination host. |
4856c84c | 788 | */ |
f2428ed5 SH |
789 | static int handle_response_icmp(int af, struct sk_buff *skb, |
790 | union nf_inet_addr *snet, | |
791 | __u8 protocol, struct ip_vs_conn *cp, | |
4856c84c | 792 | struct ip_vs_protocol *pp, |
96b077b4 JA |
793 | unsigned int offset, unsigned int ihl, |
794 | unsigned int hooknum) | |
4856c84c MT |
795 | { |
796 | unsigned int verdict = NF_DROP; | |
797 | ||
f0dce905 JA |
798 | if (IP_VS_FWD_METHOD(cp) != IP_VS_CONN_F_MASQ) |
799 | goto ignore_cp; | |
4856c84c MT |
800 | |
801 | /* Ensure the checksum is correct */ | |
802 | if (!skb_csum_unnecessary(skb) && ip_vs_checksum_complete(skb, ihl)) { | |
803 | /* Failed checksum! */ | |
f2428ed5 SH |
804 | IP_VS_DBG_BUF(1, "Forward ICMP: failed checksum from %s!\n", |
805 | IP_VS_DBG_ADDR(af, snet)); | |
4856c84c MT |
806 | goto out; |
807 | } | |
808 | ||
2906f66a VMR |
809 | if (IPPROTO_TCP == protocol || IPPROTO_UDP == protocol || |
810 | IPPROTO_SCTP == protocol) | |
4856c84c MT |
811 | offset += 2 * sizeof(__u16); |
812 | if (!skb_make_writable(skb, offset)) | |
813 | goto out; | |
814 | ||
f2428ed5 SH |
815 | #ifdef CONFIG_IP_VS_IPV6 |
816 | if (af == AF_INET6) | |
817 | ip_vs_nat_icmp_v6(skb, pp, cp, 1); | |
818 | else | |
819 | #endif | |
820 | ip_vs_nat_icmp(skb, pp, cp, 1); | |
4856c84c | 821 | |
96b077b4 | 822 | if (ip_vs_route_me_harder(af, skb, hooknum)) |
ba4fd7e9 | 823 | goto out; |
f5a41847 | 824 | |
4856c84c MT |
825 | /* do the statistics and put it back */ |
826 | ip_vs_out_stats(cp, skb); | |
827 | ||
cf356d69 | 828 | skb->ipvs_property = 1; |
f4bc17cd | 829 | if (!(cp->flags & IP_VS_CONN_F_NFCT)) |
cf356d69 | 830 | ip_vs_notrack(skb); |
f4bc17cd JA |
831 | else |
832 | ip_vs_update_conntrack(skb, cp, 0); | |
f0dce905 JA |
833 | |
834 | ignore_cp: | |
4856c84c MT |
835 | verdict = NF_ACCEPT; |
836 | ||
837 | out: | |
838 | __ip_vs_conn_put(cp); | |
839 | ||
840 | return verdict; | |
841 | } | |
842 | ||
1da177e4 LT |
843 | /* |
844 | * Handle ICMP messages in the inside-to-outside direction (outgoing). | |
4856c84c | 845 | * Find any that might be relevant, check against existing connections. |
1da177e4 | 846 | * Currently handles error types - unreachable, quench, ttl exceeded. |
1da177e4 | 847 | */ |
1ca5bb54 JA |
848 | static int ip_vs_out_icmp(struct sk_buff *skb, int *related, |
849 | unsigned int hooknum) | |
1da177e4 | 850 | { |
1da177e4 LT |
851 | struct iphdr *iph; |
852 | struct icmphdr _icmph, *ic; | |
853 | struct iphdr _ciph, *cih; /* The ip header contained within the ICMP */ | |
51ef348b | 854 | struct ip_vs_iphdr ciph; |
1da177e4 LT |
855 | struct ip_vs_conn *cp; |
856 | struct ip_vs_protocol *pp; | |
4856c84c | 857 | unsigned int offset, ihl; |
f2428ed5 | 858 | union nf_inet_addr snet; |
1da177e4 LT |
859 | |
860 | *related = 1; | |
861 | ||
862 | /* reassemble IP fragments */ | |
56f8a75c | 863 | if (ip_is_fragment(ip_hdr(skb))) { |
1ca5bb54 | 864 | if (ip_vs_gather_frags(skb, ip_vs_defrag_user(hooknum))) |
1da177e4 | 865 | return NF_STOLEN; |
1da177e4 LT |
866 | } |
867 | ||
eddc9ec5 | 868 | iph = ip_hdr(skb); |
1da177e4 LT |
869 | offset = ihl = iph->ihl * 4; |
870 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
871 | if (ic == NULL) | |
872 | return NF_DROP; | |
873 | ||
14d5e834 | 874 | IP_VS_DBG(12, "Outgoing ICMP (%d,%d) %pI4->%pI4\n", |
1da177e4 | 875 | ic->type, ntohs(icmp_id(ic)), |
14d5e834 | 876 | &iph->saddr, &iph->daddr); |
1da177e4 LT |
877 | |
878 | /* | |
879 | * Work through seeing if this is for us. | |
880 | * These checks are supposed to be in an order that means easy | |
881 | * things are checked first to speed up processing.... however | |
882 | * this means that some packets will manage to get a long way | |
883 | * down this stack and then be rejected, but that's life. | |
884 | */ | |
885 | if ((ic->type != ICMP_DEST_UNREACH) && | |
886 | (ic->type != ICMP_SOURCE_QUENCH) && | |
887 | (ic->type != ICMP_TIME_EXCEEDED)) { | |
888 | *related = 0; | |
889 | return NF_ACCEPT; | |
890 | } | |
891 | ||
892 | /* Now find the contained IP header */ | |
893 | offset += sizeof(_icmph); | |
894 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
895 | if (cih == NULL) | |
896 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
897 | ||
898 | pp = ip_vs_proto_get(cih->protocol); | |
899 | if (!pp) | |
900 | return NF_ACCEPT; | |
901 | ||
902 | /* Is the embedded protocol header present? */ | |
4412ec49 | 903 | if (unlikely(cih->frag_off & htons(IP_OFFSET) && |
1da177e4 LT |
904 | pp->dont_defrag)) |
905 | return NF_ACCEPT; | |
906 | ||
0d79641a JA |
907 | IP_VS_DBG_PKT(11, AF_INET, pp, skb, offset, |
908 | "Checking outgoing ICMP for"); | |
1da177e4 | 909 | |
63dca2c0 JDB |
910 | ip_vs_fill_ip4hdr(cih, &ciph); |
911 | ciph.len += offset; | |
1da177e4 | 912 | /* The embedded headers contain source and dest in reverse order */ |
d4383f04 | 913 | cp = pp->conn_out_get(AF_INET, skb, &ciph, 1); |
1da177e4 LT |
914 | if (!cp) |
915 | return NF_ACCEPT; | |
916 | ||
f2428ed5 SH |
917 | snet.ip = iph->saddr; |
918 | return handle_response_icmp(AF_INET, skb, &snet, cih->protocol, cp, | |
96b077b4 | 919 | pp, ciph.len, ihl, hooknum); |
1da177e4 LT |
920 | } |
921 | ||
2a3b791e | 922 | #ifdef CONFIG_IP_VS_IPV6 |
1ca5bb54 | 923 | static int ip_vs_out_icmp_v6(struct sk_buff *skb, int *related, |
d4383f04 | 924 | unsigned int hooknum, struct ip_vs_iphdr *ipvsh) |
2a3b791e | 925 | { |
2a3b791e | 926 | struct icmp6hdr _icmph, *ic; |
63dca2c0 JDB |
927 | struct ipv6hdr _ip6h, *ip6h; /* The ip header contained within ICMP */ |
928 | struct ip_vs_iphdr ciph = {.flags = 0, .fragoffs = 0};/*Contained IP */ | |
2a3b791e JV |
929 | struct ip_vs_conn *cp; |
930 | struct ip_vs_protocol *pp; | |
f2428ed5 | 931 | union nf_inet_addr snet; |
63dca2c0 | 932 | unsigned int writable; |
2a3b791e | 933 | |
63dca2c0 | 934 | *related = 1; |
2f74713d | 935 | ic = frag_safe_skb_hp(skb, ipvsh->len, sizeof(_icmph), &_icmph, ipvsh); |
2a3b791e JV |
936 | if (ic == NULL) |
937 | return NF_DROP; | |
938 | ||
2a3b791e JV |
939 | /* |
940 | * Work through seeing if this is for us. | |
941 | * These checks are supposed to be in an order that means easy | |
942 | * things are checked first to speed up processing.... however | |
943 | * this means that some packets will manage to get a long way | |
944 | * down this stack and then be rejected, but that's life. | |
945 | */ | |
2fab8917 | 946 | if (ic->icmp6_type & ICMPV6_INFOMSG_MASK) { |
2a3b791e JV |
947 | *related = 0; |
948 | return NF_ACCEPT; | |
949 | } | |
2f74713d JDB |
950 | /* Fragment header that is before ICMP header tells us that: |
951 | * it's not an error message since they can't be fragmented. | |
952 | */ | |
e7165030 | 953 | if (ipvsh->flags & IP6_FH_F_FRAG) |
2f74713d | 954 | return NF_DROP; |
2a3b791e | 955 | |
63dca2c0 JDB |
956 | IP_VS_DBG(8, "Outgoing ICMPv6 (%d,%d) %pI6c->%pI6c\n", |
957 | ic->icmp6_type, ntohs(icmpv6_id(ic)), | |
958 | &ipvsh->saddr, &ipvsh->daddr); | |
959 | ||
2a3b791e | 960 | /* Now find the contained IP header */ |
63dca2c0 JDB |
961 | ciph.len = ipvsh->len + sizeof(_icmph); |
962 | ip6h = skb_header_pointer(skb, ciph.len, sizeof(_ip6h), &_ip6h); | |
963 | if (ip6h == NULL) | |
2a3b791e | 964 | return NF_ACCEPT; /* The packet looks wrong, ignore */ |
63dca2c0 JDB |
965 | ciph.saddr.in6 = ip6h->saddr; /* conn_out_get() handles reverse order */ |
966 | ciph.daddr.in6 = ip6h->daddr; | |
967 | /* skip possible IPv6 exthdrs of contained IPv6 packet */ | |
968 | ciph.protocol = ipv6_find_hdr(skb, &ciph.len, -1, &ciph.fragoffs, NULL); | |
969 | if (ciph.protocol < 0) | |
970 | return NF_ACCEPT; /* Contained IPv6 hdr looks wrong, ignore */ | |
971 | ||
972 | pp = ip_vs_proto_get(ciph.protocol); | |
2a3b791e JV |
973 | if (!pp) |
974 | return NF_ACCEPT; | |
975 | ||
2a3b791e | 976 | /* The embedded headers contain source and dest in reverse order */ |
d4383f04 | 977 | cp = pp->conn_out_get(AF_INET6, skb, &ciph, 1); |
2a3b791e JV |
978 | if (!cp) |
979 | return NF_ACCEPT; | |
980 | ||
63dca2c0 JDB |
981 | snet.in6 = ciph.saddr.in6; |
982 | writable = ciph.len; | |
983 | return handle_response_icmp(AF_INET6, skb, &snet, ciph.protocol, cp, | |
96b077b4 JA |
984 | pp, writable, sizeof(struct ipv6hdr), |
985 | hooknum); | |
2a3b791e JV |
986 | } |
987 | #endif | |
988 | ||
2906f66a VMR |
989 | /* |
990 | * Check if sctp chunc is ABORT chunk | |
991 | */ | |
992 | static inline int is_sctp_abort(const struct sk_buff *skb, int nh_len) | |
993 | { | |
994 | sctp_chunkhdr_t *sch, schunk; | |
995 | sch = skb_header_pointer(skb, nh_len + sizeof(sctp_sctphdr_t), | |
996 | sizeof(schunk), &schunk); | |
997 | if (sch == NULL) | |
998 | return 0; | |
999 | if (sch->type == SCTP_CID_ABORT) | |
1000 | return 1; | |
1001 | return 0; | |
1002 | } | |
1003 | ||
2a3b791e | 1004 | static inline int is_tcp_reset(const struct sk_buff *skb, int nh_len) |
1da177e4 LT |
1005 | { |
1006 | struct tcphdr _tcph, *th; | |
1007 | ||
2a3b791e | 1008 | th = skb_header_pointer(skb, nh_len, sizeof(_tcph), &_tcph); |
1da177e4 LT |
1009 | if (th == NULL) |
1010 | return 0; | |
1011 | return th->rst; | |
1012 | } | |
1013 | ||
dc7b3eb9 GL |
1014 | static inline bool is_new_conn(const struct sk_buff *skb, |
1015 | struct ip_vs_iphdr *iph) | |
1016 | { | |
1017 | switch (iph->protocol) { | |
1018 | case IPPROTO_TCP: { | |
1019 | struct tcphdr _tcph, *th; | |
1020 | ||
1021 | th = skb_header_pointer(skb, iph->len, sizeof(_tcph), &_tcph); | |
1022 | if (th == NULL) | |
1023 | return false; | |
1024 | return th->syn; | |
1025 | } | |
1026 | case IPPROTO_SCTP: { | |
1027 | sctp_chunkhdr_t *sch, schunk; | |
1028 | ||
1029 | sch = skb_header_pointer(skb, iph->len + sizeof(sctp_sctphdr_t), | |
1030 | sizeof(schunk), &schunk); | |
1031 | if (sch == NULL) | |
1032 | return false; | |
1033 | return sch->type == SCTP_CID_INIT; | |
1034 | } | |
1035 | default: | |
1036 | return false; | |
1037 | } | |
1038 | } | |
1039 | ||
4856c84c | 1040 | /* Handle response packets: rewrite addresses and send away... |
4856c84c MT |
1041 | */ |
1042 | static unsigned int | |
9330419d | 1043 | handle_response(int af, struct sk_buff *skb, struct ip_vs_proto_data *pd, |
96b077b4 JA |
1044 | struct ip_vs_conn *cp, struct ip_vs_iphdr *iph, |
1045 | unsigned int hooknum) | |
4856c84c | 1046 | { |
9330419d HS |
1047 | struct ip_vs_protocol *pp = pd->pp; |
1048 | ||
0d79641a | 1049 | IP_VS_DBG_PKT(11, af, pp, skb, 0, "Outgoing packet"); |
4856c84c | 1050 | |
d4383f04 | 1051 | if (!skb_make_writable(skb, iph->len)) |
4856c84c MT |
1052 | goto drop; |
1053 | ||
1054 | /* mangle the packet */ | |
d4383f04 | 1055 | if (pp->snat_handler && !pp->snat_handler(skb, pp, cp, iph)) |
4856c84c MT |
1056 | goto drop; |
1057 | ||
1058 | #ifdef CONFIG_IP_VS_IPV6 | |
1059 | if (af == AF_INET6) | |
1060 | ipv6_hdr(skb)->saddr = cp->vaddr.in6; | |
1061 | else | |
1062 | #endif | |
1063 | { | |
1064 | ip_hdr(skb)->saddr = cp->vaddr.ip; | |
1065 | ip_send_check(ip_hdr(skb)); | |
1066 | } | |
1067 | ||
8a803040 JA |
1068 | /* |
1069 | * nf_iterate does not expect change in the skb->dst->dev. | |
1070 | * It looks like it is not fatal to enable this code for hooks | |
1071 | * where our handlers are at the end of the chain list and | |
1072 | * when all next handlers use skb->dst->dev and not outdev. | |
1073 | * It will definitely route properly the inout NAT traffic | |
1074 | * when multiple paths are used. | |
1075 | */ | |
1076 | ||
4856c84c MT |
1077 | /* For policy routing, packets originating from this |
1078 | * machine itself may be routed differently to packets | |
1079 | * passing through. We want this packet to be routed as | |
1080 | * if it came from this machine itself. So re-compute | |
1081 | * the routing information. | |
1082 | */ | |
96b077b4 | 1083 | if (ip_vs_route_me_harder(af, skb, hooknum)) |
ba4fd7e9 | 1084 | goto drop; |
4856c84c | 1085 | |
0d79641a | 1086 | IP_VS_DBG_PKT(10, af, pp, skb, 0, "After SNAT"); |
4856c84c MT |
1087 | |
1088 | ip_vs_out_stats(cp, skb); | |
9330419d | 1089 | ip_vs_set_state(cp, IP_VS_DIR_OUTPUT, skb, pd); |
cf356d69 | 1090 | skb->ipvs_property = 1; |
f4bc17cd | 1091 | if (!(cp->flags & IP_VS_CONN_F_NFCT)) |
cf356d69 | 1092 | ip_vs_notrack(skb); |
f4bc17cd JA |
1093 | else |
1094 | ip_vs_update_conntrack(skb, cp, 0); | |
4856c84c MT |
1095 | ip_vs_conn_put(cp); |
1096 | ||
4856c84c MT |
1097 | LeaveFunction(11); |
1098 | return NF_ACCEPT; | |
1099 | ||
1100 | drop: | |
1101 | ip_vs_conn_put(cp); | |
1102 | kfree_skb(skb); | |
f4bc17cd | 1103 | LeaveFunction(11); |
4856c84c MT |
1104 | return NF_STOLEN; |
1105 | } | |
1106 | ||
1da177e4 | 1107 | /* |
4856c84c | 1108 | * Check if outgoing packet belongs to the established ip_vs_conn. |
1da177e4 LT |
1109 | */ |
1110 | static unsigned int | |
fc604767 | 1111 | ip_vs_out(unsigned int hooknum, struct sk_buff *skb, int af) |
1da177e4 | 1112 | { |
fc723250 | 1113 | struct net *net = NULL; |
51ef348b | 1114 | struct ip_vs_iphdr iph; |
1da177e4 | 1115 | struct ip_vs_protocol *pp; |
9330419d | 1116 | struct ip_vs_proto_data *pd; |
1da177e4 | 1117 | struct ip_vs_conn *cp; |
1da177e4 LT |
1118 | |
1119 | EnterFunction(11); | |
1120 | ||
fc604767 | 1121 | /* Already marked as IPVS request or reply? */ |
6869c4d8 | 1122 | if (skb->ipvs_property) |
1da177e4 LT |
1123 | return NF_ACCEPT; |
1124 | ||
fc604767 JA |
1125 | /* Bad... Do not break raw sockets */ |
1126 | if (unlikely(skb->sk != NULL && hooknum == NF_INET_LOCAL_OUT && | |
1127 | af == AF_INET)) { | |
1128 | struct sock *sk = skb->sk; | |
1129 | struct inet_sock *inet = inet_sk(skb->sk); | |
1130 | ||
1131 | if (inet && sk->sk_family == PF_INET && inet->nodefrag) | |
1132 | return NF_ACCEPT; | |
1133 | } | |
1134 | ||
1135 | if (unlikely(!skb_dst(skb))) | |
1136 | return NF_ACCEPT; | |
1137 | ||
fc723250 | 1138 | net = skb_net(skb); |
7a4f0761 HS |
1139 | if (!net_ipvs(net)->enable) |
1140 | return NF_ACCEPT; | |
1141 | ||
63dca2c0 | 1142 | ip_vs_fill_iph_skb(af, skb, &iph); |
2a3b791e JV |
1143 | #ifdef CONFIG_IP_VS_IPV6 |
1144 | if (af == AF_INET6) { | |
1145 | if (unlikely(iph.protocol == IPPROTO_ICMPV6)) { | |
1ca5bb54 JA |
1146 | int related; |
1147 | int verdict = ip_vs_out_icmp_v6(skb, &related, | |
d4383f04 | 1148 | hooknum, &iph); |
1da177e4 | 1149 | |
f5a41847 | 1150 | if (related) |
2a3b791e | 1151 | return verdict; |
2a3b791e JV |
1152 | } |
1153 | } else | |
1154 | #endif | |
1155 | if (unlikely(iph.protocol == IPPROTO_ICMP)) { | |
1ca5bb54 JA |
1156 | int related; |
1157 | int verdict = ip_vs_out_icmp(skb, &related, hooknum); | |
2a3b791e | 1158 | |
f5a41847 | 1159 | if (related) |
2a3b791e | 1160 | return verdict; |
2a3b791e | 1161 | } |
1da177e4 | 1162 | |
9330419d HS |
1163 | pd = ip_vs_proto_data_get(net, iph.protocol); |
1164 | if (unlikely(!pd)) | |
1da177e4 | 1165 | return NF_ACCEPT; |
9330419d | 1166 | pp = pd->pp; |
1da177e4 LT |
1167 | |
1168 | /* reassemble IP fragments */ | |
2a3b791e | 1169 | #ifdef CONFIG_IP_VS_IPV6 |
63dca2c0 | 1170 | if (af == AF_INET) |
2a3b791e | 1171 | #endif |
56f8a75c | 1172 | if (unlikely(ip_is_fragment(ip_hdr(skb)) && !pp->dont_defrag)) { |
1ca5bb54 JA |
1173 | if (ip_vs_gather_frags(skb, |
1174 | ip_vs_defrag_user(hooknum))) | |
2a3b791e JV |
1175 | return NF_STOLEN; |
1176 | ||
63dca2c0 | 1177 | ip_vs_fill_ip4hdr(skb_network_header(skb), &iph); |
2a3b791e | 1178 | } |
1da177e4 LT |
1179 | |
1180 | /* | |
1181 | * Check if the packet belongs to an existing entry | |
1182 | */ | |
d4383f04 | 1183 | cp = pp->conn_out_get(af, skb, &iph, 0); |
1da177e4 | 1184 | |
f0dce905 JA |
1185 | if (likely(cp)) { |
1186 | if (IP_VS_FWD_METHOD(cp) != IP_VS_CONN_F_MASQ) | |
1187 | goto ignore_cp; | |
96b077b4 | 1188 | return handle_response(af, skb, pd, cp, &iph, hooknum); |
f0dce905 | 1189 | } |
0cfa558e | 1190 | if (sysctl_nat_icmp_send(net) && |
cb59155f JA |
1191 | (pp->protocol == IPPROTO_TCP || |
1192 | pp->protocol == IPPROTO_UDP || | |
1193 | pp->protocol == IPPROTO_SCTP)) { | |
1194 | __be16 _ports[2], *pptr; | |
1195 | ||
2f74713d JDB |
1196 | pptr = frag_safe_skb_hp(skb, iph.len, |
1197 | sizeof(_ports), _ports, &iph); | |
cb59155f JA |
1198 | if (pptr == NULL) |
1199 | return NF_ACCEPT; /* Not for me */ | |
276472ea JA |
1200 | if (ip_vs_has_real_service(net, af, iph.protocol, &iph.saddr, |
1201 | pptr[0])) { | |
cb59155f JA |
1202 | /* |
1203 | * Notify the real server: there is no | |
1204 | * existing entry if it is not RST | |
1205 | * packet or not TCP packet. | |
1206 | */ | |
1207 | if ((iph.protocol != IPPROTO_TCP && | |
1208 | iph.protocol != IPPROTO_SCTP) | |
1209 | || ((iph.protocol == IPPROTO_TCP | |
1210 | && !is_tcp_reset(skb, iph.len)) | |
1211 | || (iph.protocol == IPPROTO_SCTP | |
1212 | && !is_sctp_abort(skb, | |
1213 | iph.len)))) { | |
2a3b791e | 1214 | #ifdef CONFIG_IP_VS_IPV6 |
cb59155f | 1215 | if (af == AF_INET6) { |
cb59155f JA |
1216 | if (!skb->dev) |
1217 | skb->dev = net->loopback_dev; | |
1218 | icmpv6_send(skb, | |
1219 | ICMPV6_DEST_UNREACH, | |
1220 | ICMPV6_PORT_UNREACH, | |
1221 | 0); | |
1222 | } else | |
2a3b791e | 1223 | #endif |
cb59155f JA |
1224 | icmp_send(skb, |
1225 | ICMP_DEST_UNREACH, | |
1226 | ICMP_PORT_UNREACH, 0); | |
1227 | return NF_DROP; | |
1da177e4 LT |
1228 | } |
1229 | } | |
1da177e4 | 1230 | } |
f0dce905 JA |
1231 | |
1232 | out: | |
0d79641a | 1233 | IP_VS_DBG_PKT(12, af, pp, skb, 0, |
cb59155f JA |
1234 | "ip_vs_out: packet continues traversal as normal"); |
1235 | return NF_ACCEPT; | |
f0dce905 JA |
1236 | |
1237 | ignore_cp: | |
1238 | __ip_vs_conn_put(cp); | |
1239 | goto out; | |
1da177e4 LT |
1240 | } |
1241 | ||
fc604767 | 1242 | /* |
cb59155f JA |
1243 | * It is hooked at the NF_INET_FORWARD and NF_INET_LOCAL_IN chain, |
1244 | * used only for VS/NAT. | |
fc604767 JA |
1245 | * Check if packet is reply for established ip_vs_conn. |
1246 | */ | |
1247 | static unsigned int | |
1248 | ip_vs_reply4(unsigned int hooknum, struct sk_buff *skb, | |
1249 | const struct net_device *in, const struct net_device *out, | |
1250 | int (*okfn)(struct sk_buff *)) | |
1251 | { | |
1252 | return ip_vs_out(hooknum, skb, AF_INET); | |
1253 | } | |
1254 | ||
1255 | /* | |
1256 | * It is hooked at the NF_INET_LOCAL_OUT chain, used only for VS/NAT. | |
1257 | * Check if packet is reply for established ip_vs_conn. | |
1258 | */ | |
1259 | static unsigned int | |
1260 | ip_vs_local_reply4(unsigned int hooknum, struct sk_buff *skb, | |
1261 | const struct net_device *in, const struct net_device *out, | |
1262 | int (*okfn)(struct sk_buff *)) | |
1263 | { | |
ac69269a | 1264 | return ip_vs_out(hooknum, skb, AF_INET); |
fc604767 JA |
1265 | } |
1266 | ||
1267 | #ifdef CONFIG_IP_VS_IPV6 | |
1268 | ||
1269 | /* | |
cb59155f JA |
1270 | * It is hooked at the NF_INET_FORWARD and NF_INET_LOCAL_IN chain, |
1271 | * used only for VS/NAT. | |
fc604767 JA |
1272 | * Check if packet is reply for established ip_vs_conn. |
1273 | */ | |
1274 | static unsigned int | |
1275 | ip_vs_reply6(unsigned int hooknum, struct sk_buff *skb, | |
1276 | const struct net_device *in, const struct net_device *out, | |
1277 | int (*okfn)(struct sk_buff *)) | |
1278 | { | |
1279 | return ip_vs_out(hooknum, skb, AF_INET6); | |
1280 | } | |
1281 | ||
1282 | /* | |
1283 | * It is hooked at the NF_INET_LOCAL_OUT chain, used only for VS/NAT. | |
1284 | * Check if packet is reply for established ip_vs_conn. | |
1285 | */ | |
1286 | static unsigned int | |
1287 | ip_vs_local_reply6(unsigned int hooknum, struct sk_buff *skb, | |
1288 | const struct net_device *in, const struct net_device *out, | |
1289 | int (*okfn)(struct sk_buff *)) | |
1290 | { | |
ac69269a | 1291 | return ip_vs_out(hooknum, skb, AF_INET6); |
fc604767 JA |
1292 | } |
1293 | ||
1294 | #endif | |
1da177e4 LT |
1295 | |
1296 | /* | |
1297 | * Handle ICMP messages in the outside-to-inside direction (incoming). | |
1298 | * Find any that might be relevant, check against existing connections, | |
1299 | * forward to the right destination host if relevant. | |
1300 | * Currently handles error types - unreachable, quench, ttl exceeded. | |
1301 | */ | |
e905a9ed | 1302 | static int |
3db05fea | 1303 | ip_vs_in_icmp(struct sk_buff *skb, int *related, unsigned int hooknum) |
1da177e4 | 1304 | { |
9330419d | 1305 | struct net *net = NULL; |
1da177e4 LT |
1306 | struct iphdr *iph; |
1307 | struct icmphdr _icmph, *ic; | |
1308 | struct iphdr _ciph, *cih; /* The ip header contained within the ICMP */ | |
51ef348b | 1309 | struct ip_vs_iphdr ciph; |
1da177e4 LT |
1310 | struct ip_vs_conn *cp; |
1311 | struct ip_vs_protocol *pp; | |
9330419d | 1312 | struct ip_vs_proto_data *pd; |
f2edb9f7 JA |
1313 | unsigned int offset, offset2, ihl, verdict; |
1314 | bool ipip; | |
1da177e4 LT |
1315 | |
1316 | *related = 1; | |
1317 | ||
1318 | /* reassemble IP fragments */ | |
56f8a75c | 1319 | if (ip_is_fragment(ip_hdr(skb))) { |
1ca5bb54 | 1320 | if (ip_vs_gather_frags(skb, ip_vs_defrag_user(hooknum))) |
1da177e4 | 1321 | return NF_STOLEN; |
1da177e4 LT |
1322 | } |
1323 | ||
eddc9ec5 | 1324 | iph = ip_hdr(skb); |
1da177e4 LT |
1325 | offset = ihl = iph->ihl * 4; |
1326 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
1327 | if (ic == NULL) | |
1328 | return NF_DROP; | |
1329 | ||
14d5e834 | 1330 | IP_VS_DBG(12, "Incoming ICMP (%d,%d) %pI4->%pI4\n", |
1da177e4 | 1331 | ic->type, ntohs(icmp_id(ic)), |
14d5e834 | 1332 | &iph->saddr, &iph->daddr); |
1da177e4 LT |
1333 | |
1334 | /* | |
1335 | * Work through seeing if this is for us. | |
1336 | * These checks are supposed to be in an order that means easy | |
1337 | * things are checked first to speed up processing.... however | |
1338 | * this means that some packets will manage to get a long way | |
1339 | * down this stack and then be rejected, but that's life. | |
1340 | */ | |
1341 | if ((ic->type != ICMP_DEST_UNREACH) && | |
1342 | (ic->type != ICMP_SOURCE_QUENCH) && | |
1343 | (ic->type != ICMP_TIME_EXCEEDED)) { | |
1344 | *related = 0; | |
1345 | return NF_ACCEPT; | |
1346 | } | |
1347 | ||
1348 | /* Now find the contained IP header */ | |
1349 | offset += sizeof(_icmph); | |
1350 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
1351 | if (cih == NULL) | |
1352 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
1353 | ||
9330419d | 1354 | net = skb_net(skb); |
7a4f0761 | 1355 | |
f2edb9f7 JA |
1356 | /* Special case for errors for IPIP packets */ |
1357 | ipip = false; | |
1358 | if (cih->protocol == IPPROTO_IPIP) { | |
1359 | if (unlikely(cih->frag_off & htons(IP_OFFSET))) | |
1360 | return NF_ACCEPT; | |
1361 | /* Error for our IPIP must arrive at LOCAL_IN */ | |
1362 | if (!(skb_rtable(skb)->rt_flags & RTCF_LOCAL)) | |
1363 | return NF_ACCEPT; | |
1364 | offset += cih->ihl * 4; | |
1365 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
1366 | if (cih == NULL) | |
1367 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
1368 | ipip = true; | |
1369 | } | |
1370 | ||
9330419d HS |
1371 | pd = ip_vs_proto_data_get(net, cih->protocol); |
1372 | if (!pd) | |
1da177e4 | 1373 | return NF_ACCEPT; |
9330419d | 1374 | pp = pd->pp; |
1da177e4 LT |
1375 | |
1376 | /* Is the embedded protocol header present? */ | |
4412ec49 | 1377 | if (unlikely(cih->frag_off & htons(IP_OFFSET) && |
1da177e4 LT |
1378 | pp->dont_defrag)) |
1379 | return NF_ACCEPT; | |
1380 | ||
0d79641a JA |
1381 | IP_VS_DBG_PKT(11, AF_INET, pp, skb, offset, |
1382 | "Checking incoming ICMP for"); | |
1da177e4 | 1383 | |
f2edb9f7 | 1384 | offset2 = offset; |
63dca2c0 JDB |
1385 | ip_vs_fill_ip4hdr(cih, &ciph); |
1386 | ciph.len += offset; | |
1387 | offset = ciph.len; | |
f2edb9f7 JA |
1388 | /* The embedded headers contain source and dest in reverse order. |
1389 | * For IPIP this is error for request, not for reply. | |
1390 | */ | |
d4383f04 | 1391 | cp = pp->conn_in_get(AF_INET, skb, &ciph, ipip ? 0 : 1); |
6cb90db5 | 1392 | if (!cp) |
1da177e4 LT |
1393 | return NF_ACCEPT; |
1394 | ||
1395 | verdict = NF_DROP; | |
1396 | ||
1397 | /* Ensure the checksum is correct */ | |
60476372 | 1398 | if (!skb_csum_unnecessary(skb) && ip_vs_checksum_complete(skb, ihl)) { |
1da177e4 | 1399 | /* Failed checksum! */ |
14d5e834 HH |
1400 | IP_VS_DBG(1, "Incoming ICMP: failed checksum from %pI4!\n", |
1401 | &iph->saddr); | |
1da177e4 LT |
1402 | goto out; |
1403 | } | |
1404 | ||
f2edb9f7 JA |
1405 | if (ipip) { |
1406 | __be32 info = ic->un.gateway; | |
25ca1ecb PC |
1407 | __u8 type = ic->type; |
1408 | __u8 code = ic->code; | |
f2edb9f7 JA |
1409 | |
1410 | /* Update the MTU */ | |
1411 | if (ic->type == ICMP_DEST_UNREACH && | |
1412 | ic->code == ICMP_FRAG_NEEDED) { | |
1413 | struct ip_vs_dest *dest = cp->dest; | |
1414 | u32 mtu = ntohs(ic->un.frag.mtu); | |
25ca1ecb | 1415 | __be16 frag_off = cih->frag_off; |
f2edb9f7 JA |
1416 | |
1417 | /* Strip outer IP and ICMP, go to IPIP header */ | |
25ca1ecb PC |
1418 | if (pskb_pull(skb, ihl + sizeof(_icmph)) == NULL) |
1419 | goto ignore_ipip; | |
f2edb9f7 JA |
1420 | offset2 -= ihl + sizeof(_icmph); |
1421 | skb_reset_network_header(skb); | |
1422 | IP_VS_DBG(12, "ICMP for IPIP %pI4->%pI4: mtu=%u\n", | |
1423 | &ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr, mtu); | |
f2edb9f7 JA |
1424 | ipv4_update_pmtu(skb, dev_net(skb->dev), |
1425 | mtu, 0, 0, 0, 0); | |
f2edb9f7 | 1426 | /* Client uses PMTUD? */ |
25ca1ecb | 1427 | if (!(frag_off & htons(IP_DF))) |
f2edb9f7 JA |
1428 | goto ignore_ipip; |
1429 | /* Prefer the resulting PMTU */ | |
1430 | if (dest) { | |
026ace06 JA |
1431 | struct ip_vs_dest_dst *dest_dst; |
1432 | ||
1433 | rcu_read_lock(); | |
1434 | dest_dst = rcu_dereference(dest->dest_dst); | |
1435 | if (dest_dst) | |
1436 | mtu = dst_mtu(dest_dst->dst_cache); | |
1437 | rcu_read_unlock(); | |
f2edb9f7 JA |
1438 | } |
1439 | if (mtu > 68 + sizeof(struct iphdr)) | |
1440 | mtu -= sizeof(struct iphdr); | |
1441 | info = htonl(mtu); | |
1442 | } | |
1443 | /* Strip outer IP, ICMP and IPIP, go to IP header of | |
1444 | * original request. | |
1445 | */ | |
25ca1ecb PC |
1446 | if (pskb_pull(skb, offset2) == NULL) |
1447 | goto ignore_ipip; | |
f2edb9f7 JA |
1448 | skb_reset_network_header(skb); |
1449 | IP_VS_DBG(12, "Sending ICMP for %pI4->%pI4: t=%u, c=%u, i=%u\n", | |
1450 | &ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr, | |
25ca1ecb PC |
1451 | type, code, ntohl(info)); |
1452 | icmp_send(skb, type, code, info); | |
f2edb9f7 JA |
1453 | /* ICMP can be shorter but anyways, account it */ |
1454 | ip_vs_out_stats(cp, skb); | |
1455 | ||
1456 | ignore_ipip: | |
1457 | consume_skb(skb); | |
1458 | verdict = NF_STOLEN; | |
1459 | goto out; | |
1460 | } | |
1461 | ||
1da177e4 LT |
1462 | /* do the statistics and put it back */ |
1463 | ip_vs_in_stats(cp, skb); | |
06f3d7f9 JA |
1464 | if (IPPROTO_TCP == cih->protocol || IPPROTO_UDP == cih->protocol || |
1465 | IPPROTO_SCTP == cih->protocol) | |
1da177e4 | 1466 | offset += 2 * sizeof(__u16); |
d4383f04 | 1467 | verdict = ip_vs_icmp_xmit(skb, cp, pp, offset, hooknum, &ciph); |
1da177e4 | 1468 | |
552ad65a | 1469 | out: |
1da177e4 LT |
1470 | __ip_vs_conn_put(cp); |
1471 | ||
1472 | return verdict; | |
1473 | } | |
1474 | ||
2a3b791e | 1475 | #ifdef CONFIG_IP_VS_IPV6 |
d4383f04 JDB |
1476 | static int ip_vs_in_icmp_v6(struct sk_buff *skb, int *related, |
1477 | unsigned int hooknum, struct ip_vs_iphdr *iph) | |
2a3b791e | 1478 | { |
9330419d | 1479 | struct net *net = NULL; |
63dca2c0 | 1480 | struct ipv6hdr _ip6h, *ip6h; |
2a3b791e | 1481 | struct icmp6hdr _icmph, *ic; |
63dca2c0 | 1482 | struct ip_vs_iphdr ciph = {.flags = 0, .fragoffs = 0};/*Contained IP */ |
2a3b791e JV |
1483 | struct ip_vs_conn *cp; |
1484 | struct ip_vs_protocol *pp; | |
9330419d | 1485 | struct ip_vs_proto_data *pd; |
63dca2c0 | 1486 | unsigned int offs_ciph, writable, verdict; |
2a3b791e | 1487 | |
63dca2c0 | 1488 | *related = 1; |
2a3b791e | 1489 | |
2f74713d | 1490 | ic = frag_safe_skb_hp(skb, iph->len, sizeof(_icmph), &_icmph, iph); |
2a3b791e JV |
1491 | if (ic == NULL) |
1492 | return NF_DROP; | |
1493 | ||
2a3b791e JV |
1494 | /* |
1495 | * Work through seeing if this is for us. | |
1496 | * These checks are supposed to be in an order that means easy | |
1497 | * things are checked first to speed up processing.... however | |
1498 | * this means that some packets will manage to get a long way | |
1499 | * down this stack and then be rejected, but that's life. | |
1500 | */ | |
2fab8917 | 1501 | if (ic->icmp6_type & ICMPV6_INFOMSG_MASK) { |
2a3b791e JV |
1502 | *related = 0; |
1503 | return NF_ACCEPT; | |
1504 | } | |
2f74713d JDB |
1505 | /* Fragment header that is before ICMP header tells us that: |
1506 | * it's not an error message since they can't be fragmented. | |
1507 | */ | |
e7165030 | 1508 | if (iph->flags & IP6_FH_F_FRAG) |
2f74713d | 1509 | return NF_DROP; |
2a3b791e | 1510 | |
63dca2c0 JDB |
1511 | IP_VS_DBG(8, "Incoming ICMPv6 (%d,%d) %pI6c->%pI6c\n", |
1512 | ic->icmp6_type, ntohs(icmpv6_id(ic)), | |
1513 | &iph->saddr, &iph->daddr); | |
1514 | ||
2a3b791e | 1515 | /* Now find the contained IP header */ |
63dca2c0 JDB |
1516 | ciph.len = iph->len + sizeof(_icmph); |
1517 | offs_ciph = ciph.len; /* Save ip header offset */ | |
1518 | ip6h = skb_header_pointer(skb, ciph.len, sizeof(_ip6h), &_ip6h); | |
1519 | if (ip6h == NULL) | |
2a3b791e | 1520 | return NF_ACCEPT; /* The packet looks wrong, ignore */ |
63dca2c0 JDB |
1521 | ciph.saddr.in6 = ip6h->saddr; /* conn_in_get() handles reverse order */ |
1522 | ciph.daddr.in6 = ip6h->daddr; | |
1523 | /* skip possible IPv6 exthdrs of contained IPv6 packet */ | |
1524 | ciph.protocol = ipv6_find_hdr(skb, &ciph.len, -1, &ciph.fragoffs, NULL); | |
1525 | if (ciph.protocol < 0) | |
1526 | return NF_ACCEPT; /* Contained IPv6 hdr looks wrong, ignore */ | |
2a3b791e | 1527 | |
9330419d | 1528 | net = skb_net(skb); |
63dca2c0 | 1529 | pd = ip_vs_proto_data_get(net, ciph.protocol); |
9330419d | 1530 | if (!pd) |
2a3b791e | 1531 | return NF_ACCEPT; |
9330419d | 1532 | pp = pd->pp; |
2a3b791e | 1533 | |
63dca2c0 JDB |
1534 | /* Cannot handle fragmented embedded protocol */ |
1535 | if (ciph.fragoffs) | |
2a3b791e JV |
1536 | return NF_ACCEPT; |
1537 | ||
63dca2c0 | 1538 | IP_VS_DBG_PKT(11, AF_INET6, pp, skb, offs_ciph, |
0d79641a | 1539 | "Checking incoming ICMPv6 for"); |
2a3b791e | 1540 | |
2f74713d JDB |
1541 | /* The embedded headers contain source and dest in reverse order |
1542 | * if not from localhost | |
1543 | */ | |
d4383f04 | 1544 | cp = pp->conn_in_get(AF_INET6, skb, &ciph, |
2f74713d JDB |
1545 | (hooknum == NF_INET_LOCAL_OUT) ? 0 : 1); |
1546 | ||
6cb90db5 | 1547 | if (!cp) |
2a3b791e | 1548 | return NF_ACCEPT; |
2f74713d JDB |
1549 | /* VS/TUN, VS/DR and LOCALNODE just let it go */ |
1550 | if ((hooknum == NF_INET_LOCAL_OUT) && | |
1551 | (IP_VS_FWD_METHOD(cp) != IP_VS_CONN_F_MASQ)) { | |
1552 | __ip_vs_conn_put(cp); | |
1553 | return NF_ACCEPT; | |
1554 | } | |
2a3b791e | 1555 | |
2a3b791e JV |
1556 | /* do the statistics and put it back */ |
1557 | ip_vs_in_stats(cp, skb); | |
63dca2c0 JDB |
1558 | |
1559 | /* Need to mangle contained IPv6 header in ICMPv6 packet */ | |
1560 | writable = ciph.len; | |
1561 | if (IPPROTO_TCP == ciph.protocol || IPPROTO_UDP == ciph.protocol || | |
1562 | IPPROTO_SCTP == ciph.protocol) | |
1563 | writable += 2 * sizeof(__u16); /* Also mangle ports */ | |
1564 | ||
d4383f04 | 1565 | verdict = ip_vs_icmp_xmit_v6(skb, cp, pp, writable, hooknum, &ciph); |
2a3b791e JV |
1566 | |
1567 | __ip_vs_conn_put(cp); | |
1568 | ||
1569 | return verdict; | |
1570 | } | |
1571 | #endif | |
1572 | ||
1573 | ||
1da177e4 LT |
1574 | /* |
1575 | * Check if it's for virtual services, look it up, | |
1576 | * and send it on its way... | |
1577 | */ | |
1578 | static unsigned int | |
cb59155f | 1579 | ip_vs_in(unsigned int hooknum, struct sk_buff *skb, int af) |
1da177e4 | 1580 | { |
f131315f | 1581 | struct net *net; |
51ef348b | 1582 | struct ip_vs_iphdr iph; |
1da177e4 | 1583 | struct ip_vs_protocol *pp; |
9330419d | 1584 | struct ip_vs_proto_data *pd; |
1da177e4 | 1585 | struct ip_vs_conn *cp; |
4a516f11 | 1586 | int ret, pkts; |
f131315f | 1587 | struct netns_ipvs *ipvs; |
2a3b791e | 1588 | |
fc604767 JA |
1589 | /* Already marked as IPVS request or reply? */ |
1590 | if (skb->ipvs_property) | |
1591 | return NF_ACCEPT; | |
1592 | ||
1da177e4 | 1593 | /* |
cb59155f JA |
1594 | * Big tappo: |
1595 | * - remote client: only PACKET_HOST | |
1596 | * - route: used for struct net when skb->dev is unset | |
1da177e4 | 1597 | */ |
cb59155f JA |
1598 | if (unlikely((skb->pkt_type != PACKET_HOST && |
1599 | hooknum != NF_INET_LOCAL_OUT) || | |
1600 | !skb_dst(skb))) { | |
63dca2c0 | 1601 | ip_vs_fill_iph_skb(af, skb, &iph); |
cb59155f JA |
1602 | IP_VS_DBG_BUF(12, "packet type=%d proto=%d daddr=%s" |
1603 | " ignored in hook %u\n", | |
1604 | skb->pkt_type, iph.protocol, | |
1605 | IP_VS_DBG_ADDR(af, &iph.daddr), hooknum); | |
1da177e4 LT |
1606 | return NF_ACCEPT; |
1607 | } | |
7a4f0761 HS |
1608 | /* ipvs enabled in this netns ? */ |
1609 | net = skb_net(skb); | |
0c12582f JA |
1610 | ipvs = net_ipvs(net); |
1611 | if (unlikely(sysctl_backup_only(ipvs) || !ipvs->enable)) | |
7a4f0761 HS |
1612 | return NF_ACCEPT; |
1613 | ||
63dca2c0 | 1614 | ip_vs_fill_iph_skb(af, skb, &iph); |
cb59155f JA |
1615 | |
1616 | /* Bad... Do not break raw sockets */ | |
1617 | if (unlikely(skb->sk != NULL && hooknum == NF_INET_LOCAL_OUT && | |
1618 | af == AF_INET)) { | |
1619 | struct sock *sk = skb->sk; | |
1620 | struct inet_sock *inet = inet_sk(skb->sk); | |
1621 | ||
1622 | if (inet && sk->sk_family == PF_INET && inet->nodefrag) | |
1623 | return NF_ACCEPT; | |
1624 | } | |
1da177e4 | 1625 | |
94b26551 JV |
1626 | #ifdef CONFIG_IP_VS_IPV6 |
1627 | if (af == AF_INET6) { | |
1628 | if (unlikely(iph.protocol == IPPROTO_ICMPV6)) { | |
1ca5bb54 | 1629 | int related; |
d4383f04 JDB |
1630 | int verdict = ip_vs_in_icmp_v6(skb, &related, hooknum, |
1631 | &iph); | |
1da177e4 | 1632 | |
94b26551 JV |
1633 | if (related) |
1634 | return verdict; | |
94b26551 JV |
1635 | } |
1636 | } else | |
1637 | #endif | |
1638 | if (unlikely(iph.protocol == IPPROTO_ICMP)) { | |
1ca5bb54 JA |
1639 | int related; |
1640 | int verdict = ip_vs_in_icmp(skb, &related, hooknum); | |
94b26551 JV |
1641 | |
1642 | if (related) | |
1643 | return verdict; | |
94b26551 | 1644 | } |
1da177e4 LT |
1645 | |
1646 | /* Protocol supported? */ | |
9330419d HS |
1647 | pd = ip_vs_proto_data_get(net, iph.protocol); |
1648 | if (unlikely(!pd)) | |
1da177e4 | 1649 | return NF_ACCEPT; |
9330419d | 1650 | pp = pd->pp; |
1da177e4 LT |
1651 | /* |
1652 | * Check if the packet belongs to an existing connection entry | |
1653 | */ | |
d4383f04 | 1654 | cp = pp->conn_in_get(af, skb, &iph, 0); |
dc7b3eb9 GL |
1655 | |
1656 | if (unlikely(sysctl_expire_nodest_conn(ipvs)) && cp && cp->dest && | |
1657 | unlikely(!atomic_read(&cp->dest->weight)) && !iph.fragoffs && | |
1658 | is_new_conn(skb, &iph)) { | |
1659 | ip_vs_conn_expire_now(cp); | |
1660 | __ip_vs_conn_put(cp); | |
1661 | cp = NULL; | |
1662 | } | |
1663 | ||
63dca2c0 | 1664 | if (unlikely(!cp) && !iph.fragoffs) { |
2f74713d JDB |
1665 | /* No (second) fragments need to enter here, as nf_defrag_ipv6 |
1666 | * replayed fragment zero will already have created the cp | |
1667 | */ | |
1da177e4 LT |
1668 | int v; |
1669 | ||
2f74713d | 1670 | /* Schedule and create new connection entry into &cp */ |
d4383f04 | 1671 | if (!pp->conn_schedule(af, skb, pd, &v, &cp, &iph)) |
1da177e4 LT |
1672 | return v; |
1673 | } | |
1674 | ||
1675 | if (unlikely(!cp)) { | |
1676 | /* sorry, all this trouble for a no-hit :) */ | |
0d79641a | 1677 | IP_VS_DBG_PKT(12, af, pp, skb, 0, |
cb59155f | 1678 | "ip_vs_in: packet continues traversal as normal"); |
fddd8b50 | 1679 | if (iph.fragoffs) { |
2f74713d | 1680 | /* Fragment that couldn't be mapped to a conn entry |
2f74713d JDB |
1681 | * is missing module nf_defrag_ipv6 |
1682 | */ | |
1683 | IP_VS_DBG_RL("Unhandled frag, load nf_defrag_ipv6\n"); | |
1684 | IP_VS_DBG_PKT(7, af, pp, skb, 0, "unhandled fragment"); | |
1685 | } | |
1da177e4 LT |
1686 | return NF_ACCEPT; |
1687 | } | |
1688 | ||
0d79641a | 1689 | IP_VS_DBG_PKT(11, af, pp, skb, 0, "Incoming packet"); |
1da177e4 LT |
1690 | /* Check the server status */ |
1691 | if (cp->dest && !(cp->dest->flags & IP_VS_DEST_F_AVAILABLE)) { | |
1692 | /* the destination server is not available */ | |
1693 | ||
71a8ab6c | 1694 | if (sysctl_expire_nodest_conn(ipvs)) { |
1da177e4 LT |
1695 | /* try to expire the connection immediately */ |
1696 | ip_vs_conn_expire_now(cp); | |
1da177e4 | 1697 | } |
dc8103f2 JA |
1698 | /* don't restart its timer, and silently |
1699 | drop the packet. */ | |
1700 | __ip_vs_conn_put(cp); | |
1da177e4 LT |
1701 | return NF_DROP; |
1702 | } | |
1703 | ||
1704 | ip_vs_in_stats(cp, skb); | |
4a516f11 | 1705 | ip_vs_set_state(cp, IP_VS_DIR_INPUT, skb, pd); |
1da177e4 | 1706 | if (cp->packet_xmit) |
d4383f04 | 1707 | ret = cp->packet_xmit(skb, cp, pp, &iph); |
1da177e4 LT |
1708 | /* do not touch skb anymore */ |
1709 | else { | |
1710 | IP_VS_DBG_RL("warning: packet_xmit is null"); | |
1711 | ret = NF_ACCEPT; | |
1712 | } | |
1713 | ||
efac5276 RB |
1714 | /* Increase its packet counter and check if it is needed |
1715 | * to be synchronized | |
1716 | * | |
1717 | * Sync connection if it is about to close to | |
1718 | * encorage the standby servers to update the connections timeout | |
986a0757 HS |
1719 | * |
1720 | * For ONE_PKT let ip_vs_sync_conn() do the filter work. | |
efac5276 | 1721 | */ |
f131315f | 1722 | |
986a0757 | 1723 | if (cp->flags & IP_VS_CONN_F_ONE_PACKET) |
59e0350e | 1724 | pkts = sysctl_sync_threshold(ipvs); |
986a0757 HS |
1725 | else |
1726 | pkts = atomic_add_return(1, &cp->in_pkts); | |
1727 | ||
749c42b6 JA |
1728 | if (ipvs->sync_state & IP_VS_STATE_MASTER) |
1729 | ip_vs_sync_conn(net, cp, pkts); | |
1da177e4 LT |
1730 | |
1731 | ip_vs_conn_put(cp); | |
1732 | return ret; | |
1733 | } | |
1734 | ||
cb59155f JA |
1735 | /* |
1736 | * AF_INET handler in NF_INET_LOCAL_IN chain | |
1737 | * Schedule and forward packets from remote clients | |
1738 | */ | |
1739 | static unsigned int | |
1740 | ip_vs_remote_request4(unsigned int hooknum, struct sk_buff *skb, | |
1741 | const struct net_device *in, | |
1742 | const struct net_device *out, | |
1743 | int (*okfn)(struct sk_buff *)) | |
1744 | { | |
1745 | return ip_vs_in(hooknum, skb, AF_INET); | |
1746 | } | |
1747 | ||
1748 | /* | |
1749 | * AF_INET handler in NF_INET_LOCAL_OUT chain | |
1750 | * Schedule and forward packets from local clients | |
1751 | */ | |
1752 | static unsigned int | |
1753 | ip_vs_local_request4(unsigned int hooknum, struct sk_buff *skb, | |
1754 | const struct net_device *in, const struct net_device *out, | |
1755 | int (*okfn)(struct sk_buff *)) | |
1756 | { | |
ac69269a | 1757 | return ip_vs_in(hooknum, skb, AF_INET); |
cb59155f JA |
1758 | } |
1759 | ||
1760 | #ifdef CONFIG_IP_VS_IPV6 | |
1761 | ||
1762 | /* | |
1763 | * AF_INET6 handler in NF_INET_LOCAL_IN chain | |
1764 | * Schedule and forward packets from remote clients | |
1765 | */ | |
1766 | static unsigned int | |
1767 | ip_vs_remote_request6(unsigned int hooknum, struct sk_buff *skb, | |
1768 | const struct net_device *in, | |
1769 | const struct net_device *out, | |
1770 | int (*okfn)(struct sk_buff *)) | |
1771 | { | |
1772 | return ip_vs_in(hooknum, skb, AF_INET6); | |
1773 | } | |
1774 | ||
1775 | /* | |
1776 | * AF_INET6 handler in NF_INET_LOCAL_OUT chain | |
1777 | * Schedule and forward packets from local clients | |
1778 | */ | |
1779 | static unsigned int | |
1780 | ip_vs_local_request6(unsigned int hooknum, struct sk_buff *skb, | |
1781 | const struct net_device *in, const struct net_device *out, | |
1782 | int (*okfn)(struct sk_buff *)) | |
1783 | { | |
ac69269a | 1784 | return ip_vs_in(hooknum, skb, AF_INET6); |
cb59155f JA |
1785 | } |
1786 | ||
1787 | #endif | |
1788 | ||
1da177e4 LT |
1789 | |
1790 | /* | |
6e23ae2a | 1791 | * It is hooked at the NF_INET_FORWARD chain, in order to catch ICMP |
1da177e4 LT |
1792 | * related packets destined for 0.0.0.0/0. |
1793 | * When fwmark-based virtual service is used, such as transparent | |
1794 | * cache cluster, TCP packets can be marked and routed to ip_vs_in, | |
1795 | * but ICMP destined for 0.0.0.0/0 cannot not be easily marked and | |
6e23ae2a | 1796 | * sent to ip_vs_in_icmp. So, catch them at the NF_INET_FORWARD chain |
1da177e4 LT |
1797 | * and send them to ip_vs_in_icmp. |
1798 | */ | |
1799 | static unsigned int | |
3db05fea | 1800 | ip_vs_forward_icmp(unsigned int hooknum, struct sk_buff *skb, |
1da177e4 LT |
1801 | const struct net_device *in, const struct net_device *out, |
1802 | int (*okfn)(struct sk_buff *)) | |
1803 | { | |
1804 | int r; | |
7a4f0761 | 1805 | struct net *net; |
0c12582f | 1806 | struct netns_ipvs *ipvs; |
1da177e4 | 1807 | |
3db05fea | 1808 | if (ip_hdr(skb)->protocol != IPPROTO_ICMP) |
1da177e4 LT |
1809 | return NF_ACCEPT; |
1810 | ||
7a4f0761 HS |
1811 | /* ipvs enabled in this netns ? */ |
1812 | net = skb_net(skb); | |
0c12582f JA |
1813 | ipvs = net_ipvs(net); |
1814 | if (unlikely(sysctl_backup_only(ipvs) || !ipvs->enable)) | |
7a4f0761 HS |
1815 | return NF_ACCEPT; |
1816 | ||
3db05fea | 1817 | return ip_vs_in_icmp(skb, &r, hooknum); |
1da177e4 LT |
1818 | } |
1819 | ||
2a3b791e JV |
1820 | #ifdef CONFIG_IP_VS_IPV6 |
1821 | static unsigned int | |
1822 | ip_vs_forward_icmp_v6(unsigned int hooknum, struct sk_buff *skb, | |
1823 | const struct net_device *in, const struct net_device *out, | |
1824 | int (*okfn)(struct sk_buff *)) | |
1825 | { | |
1826 | int r; | |
7a4f0761 | 1827 | struct net *net; |
0c12582f | 1828 | struct netns_ipvs *ipvs; |
63dca2c0 | 1829 | struct ip_vs_iphdr iphdr; |
2a3b791e | 1830 | |
63dca2c0 JDB |
1831 | ip_vs_fill_iph_skb(AF_INET6, skb, &iphdr); |
1832 | if (iphdr.protocol != IPPROTO_ICMPV6) | |
2a3b791e JV |
1833 | return NF_ACCEPT; |
1834 | ||
7a4f0761 HS |
1835 | /* ipvs enabled in this netns ? */ |
1836 | net = skb_net(skb); | |
0c12582f JA |
1837 | ipvs = net_ipvs(net); |
1838 | if (unlikely(sysctl_backup_only(ipvs) || !ipvs->enable)) | |
7a4f0761 HS |
1839 | return NF_ACCEPT; |
1840 | ||
d4383f04 | 1841 | return ip_vs_in_icmp_v6(skb, &r, hooknum, &iphdr); |
2a3b791e JV |
1842 | } |
1843 | #endif | |
1844 | ||
1da177e4 | 1845 | |
1999414a | 1846 | static struct nf_hook_ops ip_vs_ops[] __read_mostly = { |
cb59155f JA |
1847 | /* After packet filtering, change source only for VS/NAT */ |
1848 | { | |
1849 | .hook = ip_vs_reply4, | |
1850 | .owner = THIS_MODULE, | |
4c809d63 | 1851 | .pf = NFPROTO_IPV4, |
cb59155f | 1852 | .hooknum = NF_INET_LOCAL_IN, |
afb523c5 | 1853 | .priority = NF_IP_PRI_NAT_SRC - 2, |
cb59155f | 1854 | }, |
41c5b317 PM |
1855 | /* After packet filtering, forward packet through VS/DR, VS/TUN, |
1856 | * or VS/NAT(change destination), so that filtering rules can be | |
1857 | * applied to IPVS. */ | |
1858 | { | |
cb59155f | 1859 | .hook = ip_vs_remote_request4, |
41c5b317 | 1860 | .owner = THIS_MODULE, |
4c809d63 | 1861 | .pf = NFPROTO_IPV4, |
cb59155f | 1862 | .hooknum = NF_INET_LOCAL_IN, |
afb523c5 | 1863 | .priority = NF_IP_PRI_NAT_SRC - 1, |
41c5b317 | 1864 | }, |
fc604767 | 1865 | /* Before ip_vs_in, change source only for VS/NAT */ |
41c5b317 | 1866 | { |
fc604767 | 1867 | .hook = ip_vs_local_reply4, |
41c5b317 | 1868 | .owner = THIS_MODULE, |
4c809d63 | 1869 | .pf = NFPROTO_IPV4, |
fc604767 | 1870 | .hooknum = NF_INET_LOCAL_OUT, |
afb523c5 | 1871 | .priority = NF_IP_PRI_NAT_DST + 1, |
41c5b317 | 1872 | }, |
cb59155f JA |
1873 | /* After mangle, schedule and forward local requests */ |
1874 | { | |
1875 | .hook = ip_vs_local_request4, | |
1876 | .owner = THIS_MODULE, | |
4c809d63 | 1877 | .pf = NFPROTO_IPV4, |
cb59155f | 1878 | .hooknum = NF_INET_LOCAL_OUT, |
afb523c5 | 1879 | .priority = NF_IP_PRI_NAT_DST + 2, |
cb59155f | 1880 | }, |
41c5b317 PM |
1881 | /* After packet filtering (but before ip_vs_out_icmp), catch icmp |
1882 | * destined for 0.0.0.0/0, which is for incoming IPVS connections */ | |
1883 | { | |
1884 | .hook = ip_vs_forward_icmp, | |
1885 | .owner = THIS_MODULE, | |
4c809d63 | 1886 | .pf = NFPROTO_IPV4, |
cb59155f JA |
1887 | .hooknum = NF_INET_FORWARD, |
1888 | .priority = 99, | |
41c5b317 | 1889 | }, |
fc604767 JA |
1890 | /* After packet filtering, change source only for VS/NAT */ |
1891 | { | |
1892 | .hook = ip_vs_reply4, | |
1893 | .owner = THIS_MODULE, | |
4c809d63 | 1894 | .pf = NFPROTO_IPV4, |
fc604767 JA |
1895 | .hooknum = NF_INET_FORWARD, |
1896 | .priority = 100, | |
1897 | }, | |
473b23d3 | 1898 | #ifdef CONFIG_IP_VS_IPV6 |
cb59155f JA |
1899 | /* After packet filtering, change source only for VS/NAT */ |
1900 | { | |
1901 | .hook = ip_vs_reply6, | |
1902 | .owner = THIS_MODULE, | |
4c809d63 | 1903 | .pf = NFPROTO_IPV6, |
cb59155f | 1904 | .hooknum = NF_INET_LOCAL_IN, |
afb523c5 | 1905 | .priority = NF_IP6_PRI_NAT_SRC - 2, |
cb59155f | 1906 | }, |
473b23d3 JV |
1907 | /* After packet filtering, forward packet through VS/DR, VS/TUN, |
1908 | * or VS/NAT(change destination), so that filtering rules can be | |
1909 | * applied to IPVS. */ | |
1910 | { | |
cb59155f | 1911 | .hook = ip_vs_remote_request6, |
473b23d3 | 1912 | .owner = THIS_MODULE, |
4c809d63 | 1913 | .pf = NFPROTO_IPV6, |
cb59155f | 1914 | .hooknum = NF_INET_LOCAL_IN, |
afb523c5 | 1915 | .priority = NF_IP6_PRI_NAT_SRC - 1, |
473b23d3 | 1916 | }, |
fc604767 | 1917 | /* Before ip_vs_in, change source only for VS/NAT */ |
473b23d3 | 1918 | { |
fc604767 | 1919 | .hook = ip_vs_local_reply6, |
473b23d3 | 1920 | .owner = THIS_MODULE, |
a1b7f13b | 1921 | .pf = NFPROTO_IPV6, |
fc604767 | 1922 | .hooknum = NF_INET_LOCAL_OUT, |
afb523c5 | 1923 | .priority = NF_IP6_PRI_NAT_DST + 1, |
473b23d3 | 1924 | }, |
cb59155f JA |
1925 | /* After mangle, schedule and forward local requests */ |
1926 | { | |
1927 | .hook = ip_vs_local_request6, | |
1928 | .owner = THIS_MODULE, | |
4c809d63 | 1929 | .pf = NFPROTO_IPV6, |
cb59155f | 1930 | .hooknum = NF_INET_LOCAL_OUT, |
afb523c5 | 1931 | .priority = NF_IP6_PRI_NAT_DST + 2, |
cb59155f | 1932 | }, |
473b23d3 JV |
1933 | /* After packet filtering (but before ip_vs_out_icmp), catch icmp |
1934 | * destined for 0.0.0.0/0, which is for incoming IPVS connections */ | |
1935 | { | |
1936 | .hook = ip_vs_forward_icmp_v6, | |
1937 | .owner = THIS_MODULE, | |
4c809d63 | 1938 | .pf = NFPROTO_IPV6, |
cb59155f JA |
1939 | .hooknum = NF_INET_FORWARD, |
1940 | .priority = 99, | |
473b23d3 | 1941 | }, |
fc604767 JA |
1942 | /* After packet filtering, change source only for VS/NAT */ |
1943 | { | |
1944 | .hook = ip_vs_reply6, | |
1945 | .owner = THIS_MODULE, | |
4c809d63 | 1946 | .pf = NFPROTO_IPV6, |
fc604767 JA |
1947 | .hooknum = NF_INET_FORWARD, |
1948 | .priority = 100, | |
1949 | }, | |
473b23d3 | 1950 | #endif |
1da177e4 | 1951 | }; |
61b1ab45 HS |
1952 | /* |
1953 | * Initialize IP Virtual Server netns mem. | |
1954 | */ | |
1955 | static int __net_init __ip_vs_init(struct net *net) | |
1956 | { | |
1957 | struct netns_ipvs *ipvs; | |
1958 | ||
61b1ab45 | 1959 | ipvs = net_generic(net, ip_vs_net_id); |
0a9ee813 | 1960 | if (ipvs == NULL) |
61b1ab45 | 1961 | return -ENOMEM; |
0a9ee813 | 1962 | |
7a4f0761 HS |
1963 | /* Hold the beast until a service is registerd */ |
1964 | ipvs->enable = 0; | |
f6340ee0 | 1965 | ipvs->net = net; |
61b1ab45 HS |
1966 | /* Counters used for creating unique names */ |
1967 | ipvs->gen = atomic_read(&ipvs_netns_cnt); | |
1968 | atomic_inc(&ipvs_netns_cnt); | |
1969 | net->ipvs = ipvs; | |
7a4f0761 | 1970 | |
503cf15a | 1971 | if (ip_vs_estimator_net_init(net) < 0) |
7a4f0761 HS |
1972 | goto estimator_fail; |
1973 | ||
503cf15a | 1974 | if (ip_vs_control_net_init(net) < 0) |
7a4f0761 HS |
1975 | goto control_fail; |
1976 | ||
503cf15a | 1977 | if (ip_vs_protocol_net_init(net) < 0) |
7a4f0761 HS |
1978 | goto protocol_fail; |
1979 | ||
503cf15a | 1980 | if (ip_vs_app_net_init(net) < 0) |
7a4f0761 HS |
1981 | goto app_fail; |
1982 | ||
503cf15a | 1983 | if (ip_vs_conn_net_init(net) < 0) |
7a4f0761 HS |
1984 | goto conn_fail; |
1985 | ||
503cf15a | 1986 | if (ip_vs_sync_net_init(net) < 0) |
7a4f0761 HS |
1987 | goto sync_fail; |
1988 | ||
a870c8c5 | 1989 | printk(KERN_INFO "IPVS: Creating netns size=%zu id=%d\n", |
61b1ab45 HS |
1990 | sizeof(struct netns_ipvs), ipvs->gen); |
1991 | return 0; | |
7a4f0761 HS |
1992 | /* |
1993 | * Error handling | |
1994 | */ | |
1995 | ||
1996 | sync_fail: | |
503cf15a | 1997 | ip_vs_conn_net_cleanup(net); |
7a4f0761 | 1998 | conn_fail: |
503cf15a | 1999 | ip_vs_app_net_cleanup(net); |
7a4f0761 | 2000 | app_fail: |
503cf15a | 2001 | ip_vs_protocol_net_cleanup(net); |
7a4f0761 | 2002 | protocol_fail: |
503cf15a | 2003 | ip_vs_control_net_cleanup(net); |
7a4f0761 | 2004 | control_fail: |
503cf15a | 2005 | ip_vs_estimator_net_cleanup(net); |
7a4f0761 | 2006 | estimator_fail: |
39f618b4 | 2007 | net->ipvs = NULL; |
7a4f0761 | 2008 | return -ENOMEM; |
61b1ab45 HS |
2009 | } |
2010 | ||
2011 | static void __net_exit __ip_vs_cleanup(struct net *net) | |
2012 | { | |
503cf15a HS |
2013 | ip_vs_service_net_cleanup(net); /* ip_vs_flush() with locks */ |
2014 | ip_vs_conn_net_cleanup(net); | |
2015 | ip_vs_app_net_cleanup(net); | |
2016 | ip_vs_protocol_net_cleanup(net); | |
2017 | ip_vs_control_net_cleanup(net); | |
2018 | ip_vs_estimator_net_cleanup(net); | |
1ae132b0 | 2019 | IP_VS_DBG(2, "ipvs netns %d released\n", net_ipvs(net)->gen); |
39f618b4 | 2020 | net->ipvs = NULL; |
61b1ab45 HS |
2021 | } |
2022 | ||
7a4f0761 HS |
2023 | static void __net_exit __ip_vs_dev_cleanup(struct net *net) |
2024 | { | |
2025 | EnterFunction(2); | |
2026 | net_ipvs(net)->enable = 0; /* Disable packet reception */ | |
8f4e0a18 | 2027 | smp_wmb(); |
503cf15a | 2028 | ip_vs_sync_net_cleanup(net); |
7a4f0761 HS |
2029 | LeaveFunction(2); |
2030 | } | |
2031 | ||
61b1ab45 HS |
2032 | static struct pernet_operations ipvs_core_ops = { |
2033 | .init = __ip_vs_init, | |
2034 | .exit = __ip_vs_cleanup, | |
2035 | .id = &ip_vs_net_id, | |
2036 | .size = sizeof(struct netns_ipvs), | |
2037 | }; | |
1da177e4 | 2038 | |
7a4f0761 HS |
2039 | static struct pernet_operations ipvs_core_dev_ops = { |
2040 | .exit = __ip_vs_dev_cleanup, | |
2041 | }; | |
2042 | ||
1da177e4 LT |
2043 | /* |
2044 | * Initialize IP Virtual Server | |
2045 | */ | |
2046 | static int __init ip_vs_init(void) | |
2047 | { | |
2048 | int ret; | |
2049 | ||
2050 | ret = ip_vs_control_init(); | |
2051 | if (ret < 0) { | |
1e3e238e | 2052 | pr_err("can't setup control.\n"); |
6c8f7949 | 2053 | goto exit; |
1da177e4 LT |
2054 | } |
2055 | ||
2056 | ip_vs_protocol_init(); | |
2057 | ||
1da177e4 LT |
2058 | ret = ip_vs_conn_init(); |
2059 | if (ret < 0) { | |
1e3e238e | 2060 | pr_err("can't setup connection table.\n"); |
6c8f7949 | 2061 | goto cleanup_protocol; |
61b1ab45 HS |
2062 | } |
2063 | ||
7a4f0761 HS |
2064 | ret = register_pernet_subsys(&ipvs_core_ops); /* Alloc ip_vs struct */ |
2065 | if (ret < 0) | |
6c8f7949 | 2066 | goto cleanup_conn; |
7a4f0761 HS |
2067 | |
2068 | ret = register_pernet_device(&ipvs_core_dev_ops); | |
2069 | if (ret < 0) | |
2070 | goto cleanup_sub; | |
2071 | ||
41c5b317 | 2072 | ret = nf_register_hooks(ip_vs_ops, ARRAY_SIZE(ip_vs_ops)); |
1da177e4 | 2073 | if (ret < 0) { |
1e3e238e | 2074 | pr_err("can't register hooks.\n"); |
7a4f0761 | 2075 | goto cleanup_dev; |
1da177e4 LT |
2076 | } |
2077 | ||
8537de8a HS |
2078 | ret = ip_vs_register_nl_ioctl(); |
2079 | if (ret < 0) { | |
2080 | pr_err("can't register netlink/ioctl.\n"); | |
2081 | goto cleanup_hooks; | |
2082 | } | |
2083 | ||
1e3e238e | 2084 | pr_info("ipvs loaded.\n"); |
7a4f0761 | 2085 | |
1da177e4 LT |
2086 | return ret; |
2087 | ||
8537de8a HS |
2088 | cleanup_hooks: |
2089 | nf_unregister_hooks(ip_vs_ops, ARRAY_SIZE(ip_vs_ops)); | |
7a4f0761 HS |
2090 | cleanup_dev: |
2091 | unregister_pernet_device(&ipvs_core_dev_ops); | |
2092 | cleanup_sub: | |
2093 | unregister_pernet_subsys(&ipvs_core_ops); | |
552ad65a | 2094 | cleanup_conn: |
1da177e4 | 2095 | ip_vs_conn_cleanup(); |
552ad65a | 2096 | cleanup_protocol: |
1da177e4 LT |
2097 | ip_vs_protocol_cleanup(); |
2098 | ip_vs_control_cleanup(); | |
6c8f7949 | 2099 | exit: |
1da177e4 LT |
2100 | return ret; |
2101 | } | |
2102 | ||
2103 | static void __exit ip_vs_cleanup(void) | |
2104 | { | |
8537de8a | 2105 | ip_vs_unregister_nl_ioctl(); |
41c5b317 | 2106 | nf_unregister_hooks(ip_vs_ops, ARRAY_SIZE(ip_vs_ops)); |
7a4f0761 HS |
2107 | unregister_pernet_device(&ipvs_core_dev_ops); |
2108 | unregister_pernet_subsys(&ipvs_core_ops); /* free ip_vs struct */ | |
1da177e4 | 2109 | ip_vs_conn_cleanup(); |
1da177e4 LT |
2110 | ip_vs_protocol_cleanup(); |
2111 | ip_vs_control_cleanup(); | |
1e3e238e | 2112 | pr_info("ipvs unloaded.\n"); |
1da177e4 LT |
2113 | } |
2114 | ||
2115 | module_init(ip_vs_init); | |
2116 | module_exit(ip_vs_cleanup); | |
2117 | MODULE_LICENSE("GPL"); |