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 | 99 | default: |
1404c3ab | 100 | sprintf(buf, "IP_%u", proto); |
1da177e4 LT |
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 | 118 | struct ip_vs_cpu_stats *s; |
bcbde4c0 | 119 | struct ip_vs_service *svc; |
b17fc996 HS |
120 | |
121 | s = this_cpu_ptr(dest->stats.cpustats); | |
b17fc996 | 122 | u64_stats_update_begin(&s->syncp); |
cd67cd5e JA |
123 | s->cnt.inpkts++; |
124 | s->cnt.inbytes += skb->len; | |
b17fc996 HS |
125 | u64_stats_update_end(&s->syncp); |
126 | ||
bcbde4c0 JA |
127 | rcu_read_lock(); |
128 | svc = rcu_dereference(dest->svc); | |
129 | s = this_cpu_ptr(svc->stats.cpustats); | |
b17fc996 | 130 | u64_stats_update_begin(&s->syncp); |
cd67cd5e JA |
131 | s->cnt.inpkts++; |
132 | s->cnt.inbytes += skb->len; | |
b17fc996 | 133 | u64_stats_update_end(&s->syncp); |
bcbde4c0 | 134 | rcu_read_unlock(); |
b17fc996 | 135 | |
2a0751af | 136 | s = this_cpu_ptr(ipvs->tot_stats.cpustats); |
b17fc996 | 137 | u64_stats_update_begin(&s->syncp); |
cd67cd5e JA |
138 | s->cnt.inpkts++; |
139 | s->cnt.inbytes += skb->len; | |
b17fc996 | 140 | u64_stats_update_end(&s->syncp); |
1da177e4 LT |
141 | } |
142 | } | |
143 | ||
144 | ||
145 | static inline void | |
146 | ip_vs_out_stats(struct ip_vs_conn *cp, struct sk_buff *skb) | |
147 | { | |
148 | struct ip_vs_dest *dest = cp->dest; | |
b17fc996 HS |
149 | struct netns_ipvs *ipvs = net_ipvs(skb_net(skb)); |
150 | ||
1da177e4 | 151 | if (dest && (dest->flags & IP_VS_DEST_F_AVAILABLE)) { |
b17fc996 | 152 | struct ip_vs_cpu_stats *s; |
bcbde4c0 | 153 | struct ip_vs_service *svc; |
b17fc996 HS |
154 | |
155 | s = this_cpu_ptr(dest->stats.cpustats); | |
b17fc996 | 156 | u64_stats_update_begin(&s->syncp); |
cd67cd5e JA |
157 | s->cnt.outpkts++; |
158 | s->cnt.outbytes += skb->len; | |
b17fc996 HS |
159 | u64_stats_update_end(&s->syncp); |
160 | ||
bcbde4c0 JA |
161 | rcu_read_lock(); |
162 | svc = rcu_dereference(dest->svc); | |
163 | s = this_cpu_ptr(svc->stats.cpustats); | |
b17fc996 | 164 | u64_stats_update_begin(&s->syncp); |
cd67cd5e JA |
165 | s->cnt.outpkts++; |
166 | s->cnt.outbytes += skb->len; | |
b17fc996 | 167 | u64_stats_update_end(&s->syncp); |
bcbde4c0 | 168 | rcu_read_unlock(); |
b17fc996 | 169 | |
2a0751af | 170 | s = this_cpu_ptr(ipvs->tot_stats.cpustats); |
b17fc996 | 171 | u64_stats_update_begin(&s->syncp); |
cd67cd5e JA |
172 | s->cnt.outpkts++; |
173 | s->cnt.outbytes += skb->len; | |
b17fc996 | 174 | u64_stats_update_end(&s->syncp); |
1da177e4 LT |
175 | } |
176 | } | |
177 | ||
178 | ||
179 | static inline void | |
180 | ip_vs_conn_stats(struct ip_vs_conn *cp, struct ip_vs_service *svc) | |
181 | { | |
b17fc996 HS |
182 | struct netns_ipvs *ipvs = net_ipvs(svc->net); |
183 | struct ip_vs_cpu_stats *s; | |
1da177e4 | 184 | |
b17fc996 | 185 | s = this_cpu_ptr(cp->dest->stats.cpustats); |
cd67cd5e JA |
186 | u64_stats_update_begin(&s->syncp); |
187 | s->cnt.conns++; | |
188 | u64_stats_update_end(&s->syncp); | |
1da177e4 | 189 | |
b17fc996 | 190 | s = this_cpu_ptr(svc->stats.cpustats); |
cd67cd5e JA |
191 | u64_stats_update_begin(&s->syncp); |
192 | s->cnt.conns++; | |
193 | u64_stats_update_end(&s->syncp); | |
b17fc996 | 194 | |
2a0751af | 195 | s = this_cpu_ptr(ipvs->tot_stats.cpustats); |
cd67cd5e JA |
196 | u64_stats_update_begin(&s->syncp); |
197 | s->cnt.conns++; | |
198 | u64_stats_update_end(&s->syncp); | |
1da177e4 LT |
199 | } |
200 | ||
201 | ||
4a516f11 | 202 | static inline void |
1da177e4 LT |
203 | ip_vs_set_state(struct ip_vs_conn *cp, int direction, |
204 | const struct sk_buff *skb, | |
9330419d | 205 | struct ip_vs_proto_data *pd) |
1da177e4 | 206 | { |
4a516f11 SH |
207 | if (likely(pd->pp->state_transition)) |
208 | pd->pp->state_transition(cp, direction, skb, pd); | |
1da177e4 LT |
209 | } |
210 | ||
a5959d53 | 211 | static inline int |
85999283 SH |
212 | ip_vs_conn_fill_param_persist(const struct ip_vs_service *svc, |
213 | struct sk_buff *skb, int protocol, | |
214 | const union nf_inet_addr *caddr, __be16 cport, | |
215 | const union nf_inet_addr *vaddr, __be16 vport, | |
216 | struct ip_vs_conn_param *p) | |
217 | { | |
6e67e586 HS |
218 | ip_vs_conn_fill_param(svc->net, svc->af, protocol, caddr, cport, vaddr, |
219 | vport, p); | |
ceec4c38 | 220 | p->pe = rcu_dereference(svc->pe); |
85999283 | 221 | if (p->pe && p->pe->fill_param) |
a5959d53 HS |
222 | return p->pe->fill_param(p, skb); |
223 | ||
224 | return 0; | |
85999283 | 225 | } |
1da177e4 | 226 | |
1da177e4 LT |
227 | /* |
228 | * IPVS persistent scheduling function | |
229 | * It creates a connection entry according to its template if exists, | |
230 | * or selects a server and creates a connection entry plus a template. | |
231 | * Locking: we are svc user (svc->refcnt), so we hold all dests too | |
232 | * Protocols supported: TCP, UDP | |
233 | */ | |
234 | static struct ip_vs_conn * | |
235 | ip_vs_sched_persist(struct ip_vs_service *svc, | |
d4383f04 JDB |
236 | struct sk_buff *skb, __be16 src_port, __be16 dst_port, |
237 | int *ignored, struct ip_vs_iphdr *iph) | |
1da177e4 LT |
238 | { |
239 | struct ip_vs_conn *cp = NULL; | |
1da177e4 LT |
240 | struct ip_vs_dest *dest; |
241 | struct ip_vs_conn *ct; | |
5b57a98c | 242 | __be16 dport = 0; /* destination port to forward */ |
3575792e | 243 | unsigned int flags; |
f11017ec | 244 | struct ip_vs_conn_param param; |
e0aac52e | 245 | const union nf_inet_addr fwmark = { .ip = htonl(svc->fwmark) }; |
28364a59 JV |
246 | union nf_inet_addr snet; /* source network of the client, |
247 | after masking */ | |
cd17f9ed | 248 | |
1da177e4 | 249 | /* Mask saddr with the netmask to adjust template granularity */ |
cd17f9ed JV |
250 | #ifdef CONFIG_IP_VS_IPV6 |
251 | if (svc->af == AF_INET6) | |
0a925864 JA |
252 | ipv6_addr_prefix(&snet.in6, &iph->saddr.in6, |
253 | (__force __u32) svc->netmask); | |
cd17f9ed JV |
254 | else |
255 | #endif | |
d4383f04 | 256 | snet.ip = iph->saddr.ip & svc->netmask; |
1da177e4 | 257 | |
cd17f9ed JV |
258 | IP_VS_DBG_BUF(6, "p-schedule: src %s:%u dest %s:%u " |
259 | "mnet %s\n", | |
d4383f04 JDB |
260 | IP_VS_DBG_ADDR(svc->af, &iph->saddr), ntohs(src_port), |
261 | IP_VS_DBG_ADDR(svc->af, &iph->daddr), ntohs(dst_port), | |
cd17f9ed | 262 | IP_VS_DBG_ADDR(svc->af, &snet)); |
1da177e4 LT |
263 | |
264 | /* | |
265 | * As far as we know, FTP is a very complicated network protocol, and | |
266 | * it uses control connection and data connections. For active FTP, | |
267 | * FTP server initialize data connection to the client, its source port | |
268 | * is often 20. For passive FTP, FTP server tells the clients the port | |
269 | * that it passively listens to, and the client issues the data | |
270 | * connection. In the tunneling or direct routing mode, the load | |
271 | * balancer is on the client-to-server half of connection, the port | |
272 | * number is unknown to the load balancer. So, a conn template like | |
273 | * <caddr, 0, vaddr, 0, daddr, 0> is created for persistent FTP | |
274 | * service, and a template like <caddr, 0, vaddr, vport, daddr, dport> | |
275 | * is created for other persistent services. | |
276 | */ | |
5b57a98c | 277 | { |
d4383f04 JDB |
278 | int protocol = iph->protocol; |
279 | const union nf_inet_addr *vaddr = &iph->daddr; | |
f11017ec SH |
280 | __be16 vport = 0; |
281 | ||
ce144f24 | 282 | if (dst_port == svc->port) { |
5b57a98c SH |
283 | /* non-FTP template: |
284 | * <protocol, caddr, 0, vaddr, vport, daddr, dport> | |
285 | * FTP template: | |
286 | * <protocol, caddr, 0, vaddr, 0, daddr, 0> | |
1da177e4 LT |
287 | */ |
288 | if (svc->port != FTPPORT) | |
ce144f24 | 289 | vport = dst_port; |
1da177e4 | 290 | } else { |
5b57a98c SH |
291 | /* Note: persistent fwmark-based services and |
292 | * persistent port zero service are handled here. | |
293 | * fwmark template: | |
294 | * <IPPROTO_IP,caddr,0,fwmark,0,daddr,0> | |
295 | * port zero template: | |
296 | * <protocol,caddr,0,vaddr,0,daddr,0> | |
1da177e4 | 297 | */ |
28364a59 | 298 | if (svc->fwmark) { |
5b57a98c SH |
299 | protocol = IPPROTO_IP; |
300 | vaddr = &fwmark; | |
301 | } | |
1da177e4 | 302 | } |
a5959d53 HS |
303 | /* return *ignored = -1 so NF_DROP can be used */ |
304 | if (ip_vs_conn_fill_param_persist(svc, skb, protocol, &snet, 0, | |
305 | vaddr, vport, ¶m) < 0) { | |
306 | *ignored = -1; | |
307 | return NULL; | |
308 | } | |
1da177e4 LT |
309 | } |
310 | ||
5b57a98c | 311 | /* Check if a template already exists */ |
f11017ec | 312 | ct = ip_vs_ct_in_get(¶m); |
5b57a98c | 313 | if (!ct || !ip_vs_check_template(ct)) { |
ceec4c38 JA |
314 | struct ip_vs_scheduler *sched; |
315 | ||
a5959d53 HS |
316 | /* |
317 | * No template found or the dest of the connection | |
5b57a98c | 318 | * template is not available. |
a5959d53 | 319 | * return *ignored=0 i.e. ICMP and NF_DROP |
5b57a98c | 320 | */ |
ceec4c38 | 321 | sched = rcu_dereference(svc->scheduler); |
bba54de5 | 322 | dest = sched->schedule(svc, skb, iph); |
5b57a98c SH |
323 | if (!dest) { |
324 | IP_VS_DBG(1, "p-schedule: no dest found.\n"); | |
85999283 | 325 | kfree(param.pe_data); |
a5959d53 | 326 | *ignored = 0; |
5b57a98c SH |
327 | return NULL; |
328 | } | |
329 | ||
ce144f24 | 330 | if (dst_port == svc->port && svc->port != FTPPORT) |
5b57a98c SH |
331 | dport = dest->port; |
332 | ||
85999283 SH |
333 | /* Create a template |
334 | * This adds param.pe_data to the template, | |
335 | * and thus param.pe_data will be destroyed | |
336 | * when the template expires */ | |
ba38528a | 337 | ct = ip_vs_conn_new(¶m, dest->af, &dest->addr, dport, |
0e051e68 | 338 | IP_VS_CONN_F_TEMPLATE, dest, skb->mark); |
85999283 SH |
339 | if (ct == NULL) { |
340 | kfree(param.pe_data); | |
a5959d53 | 341 | *ignored = -1; |
5b57a98c | 342 | return NULL; |
85999283 | 343 | } |
5b57a98c SH |
344 | |
345 | ct->timeout = svc->timeout; | |
85999283 | 346 | } else { |
5b57a98c SH |
347 | /* set destination with the found template */ |
348 | dest = ct->dest; | |
85999283 SH |
349 | kfree(param.pe_data); |
350 | } | |
5b57a98c | 351 | |
ce144f24 | 352 | dport = dst_port; |
5b57a98c SH |
353 | if (dport == svc->port && dest->port) |
354 | dport = dest->port; | |
355 | ||
26ec037f | 356 | flags = (svc->flags & IP_VS_SVC_F_ONEPACKET |
d4383f04 | 357 | && iph->protocol == IPPROTO_UDP) ? |
26ec037f NC |
358 | IP_VS_CONN_F_ONE_PACKET : 0; |
359 | ||
1da177e4 LT |
360 | /* |
361 | * Create a new connection according to the template | |
362 | */ | |
d4383f04 JDB |
363 | ip_vs_conn_fill_param(svc->net, svc->af, iph->protocol, &iph->saddr, |
364 | src_port, &iph->daddr, dst_port, ¶m); | |
ce144f24 | 365 | |
ba38528a AG |
366 | cp = ip_vs_conn_new(¶m, dest->af, &dest->addr, dport, flags, dest, |
367 | skb->mark); | |
1da177e4 LT |
368 | if (cp == NULL) { |
369 | ip_vs_conn_put(ct); | |
a5959d53 | 370 | *ignored = -1; |
1da177e4 LT |
371 | return NULL; |
372 | } | |
373 | ||
374 | /* | |
375 | * Add its control | |
376 | */ | |
377 | ip_vs_control_add(cp, ct); | |
378 | ip_vs_conn_put(ct); | |
379 | ||
380 | ip_vs_conn_stats(cp, svc); | |
381 | return cp; | |
382 | } | |
383 | ||
384 | ||
385 | /* | |
386 | * IPVS main scheduling function | |
387 | * It selects a server according to the virtual service, and | |
388 | * creates a connection entry. | |
389 | * Protocols supported: TCP, UDP | |
a5959d53 HS |
390 | * |
391 | * Usage of *ignored | |
392 | * | |
393 | * 1 : protocol tried to schedule (eg. on SYN), found svc but the | |
394 | * svc/scheduler decides that this packet should be accepted with | |
395 | * NF_ACCEPT because it must not be scheduled. | |
396 | * | |
397 | * 0 : scheduler can not find destination, so try bypass or | |
398 | * return ICMP and then NF_DROP (ip_vs_leave). | |
399 | * | |
400 | * -1 : scheduler tried to schedule but fatal error occurred, eg. | |
401 | * ip_vs_conn_new failure (ENOMEM) or ip_vs_sip_fill_param | |
402 | * failure such as missing Call-ID, ENOMEM on skb_linearize | |
403 | * or pe_data. In this case we should return NF_DROP without | |
404 | * any attempts to send ICMP with ip_vs_leave. | |
1da177e4 LT |
405 | */ |
406 | struct ip_vs_conn * | |
190ecd27 | 407 | ip_vs_schedule(struct ip_vs_service *svc, struct sk_buff *skb, |
d4383f04 JDB |
408 | struct ip_vs_proto_data *pd, int *ignored, |
409 | struct ip_vs_iphdr *iph) | |
1da177e4 | 410 | { |
9330419d | 411 | struct ip_vs_protocol *pp = pd->pp; |
1da177e4 | 412 | struct ip_vs_conn *cp = NULL; |
ceec4c38 | 413 | struct ip_vs_scheduler *sched; |
1da177e4 | 414 | struct ip_vs_dest *dest; |
3575792e JA |
415 | __be16 _ports[2], *pptr; |
416 | unsigned int flags; | |
1da177e4 | 417 | |
190ecd27 | 418 | *ignored = 1; |
2f74713d JDB |
419 | /* |
420 | * IPv6 frags, only the first hit here. | |
421 | */ | |
d4383f04 | 422 | pptr = frag_safe_skb_hp(skb, iph->len, sizeof(_ports), _ports, iph); |
1da177e4 LT |
423 | if (pptr == NULL) |
424 | return NULL; | |
425 | ||
190ecd27 JA |
426 | /* |
427 | * FTPDATA needs this check when using local real server. | |
428 | * Never schedule Active FTPDATA connections from real server. | |
429 | * For LVS-NAT they must be already created. For other methods | |
430 | * with persistence the connection is created on SYN+ACK. | |
431 | */ | |
432 | if (pptr[0] == FTPDATA) { | |
0d79641a JA |
433 | IP_VS_DBG_PKT(12, svc->af, pp, skb, 0, |
434 | "Not scheduling FTPDATA"); | |
190ecd27 JA |
435 | return NULL; |
436 | } | |
437 | ||
438 | /* | |
a5959d53 | 439 | * Do not schedule replies from local real server. |
190ecd27 JA |
440 | */ |
441 | if ((!skb->dev || skb->dev->flags & IFF_LOOPBACK) && | |
d4383f04 | 442 | (cp = pp->conn_in_get(svc->af, skb, iph, 1))) { |
0d79641a | 443 | IP_VS_DBG_PKT(12, svc->af, pp, skb, 0, |
190ecd27 JA |
444 | "Not scheduling reply for existing connection"); |
445 | __ip_vs_conn_put(cp); | |
446 | return NULL; | |
447 | } | |
448 | ||
1da177e4 LT |
449 | /* |
450 | * Persistent service | |
451 | */ | |
a5959d53 | 452 | if (svc->flags & IP_VS_SVC_F_PERSISTENT) |
d4383f04 JDB |
453 | return ip_vs_sched_persist(svc, skb, pptr[0], pptr[1], ignored, |
454 | iph); | |
a5959d53 HS |
455 | |
456 | *ignored = 0; | |
1da177e4 LT |
457 | |
458 | /* | |
459 | * Non-persistent service | |
460 | */ | |
461 | if (!svc->fwmark && pptr[1] != svc->port) { | |
462 | if (!svc->port) | |
1e3e238e HE |
463 | pr_err("Schedule: port zero only supported " |
464 | "in persistent services, " | |
465 | "check your ipvs configuration\n"); | |
1da177e4 LT |
466 | return NULL; |
467 | } | |
468 | ||
ceec4c38 | 469 | sched = rcu_dereference(svc->scheduler); |
bba54de5 | 470 | dest = sched->schedule(svc, skb, iph); |
1da177e4 LT |
471 | if (dest == NULL) { |
472 | IP_VS_DBG(1, "Schedule: no dest found.\n"); | |
473 | return NULL; | |
474 | } | |
475 | ||
26ec037f | 476 | flags = (svc->flags & IP_VS_SVC_F_ONEPACKET |
d4383f04 | 477 | && iph->protocol == IPPROTO_UDP) ? |
26ec037f NC |
478 | IP_VS_CONN_F_ONE_PACKET : 0; |
479 | ||
1da177e4 LT |
480 | /* |
481 | * Create a connection entry. | |
482 | */ | |
f11017ec SH |
483 | { |
484 | struct ip_vs_conn_param p; | |
6e67e586 | 485 | |
d4383f04 JDB |
486 | ip_vs_conn_fill_param(svc->net, svc->af, iph->protocol, |
487 | &iph->saddr, pptr[0], &iph->daddr, | |
488 | pptr[1], &p); | |
ba38528a | 489 | cp = ip_vs_conn_new(&p, dest->af, &dest->addr, |
f11017ec | 490 | dest->port ? dest->port : pptr[1], |
0e051e68 | 491 | flags, dest, skb->mark); |
a5959d53 HS |
492 | if (!cp) { |
493 | *ignored = -1; | |
f11017ec | 494 | return NULL; |
a5959d53 | 495 | } |
f11017ec | 496 | } |
1da177e4 | 497 | |
cd17f9ed JV |
498 | IP_VS_DBG_BUF(6, "Schedule fwd:%c c:%s:%u v:%s:%u " |
499 | "d:%s:%u conn->flags:%X conn->refcnt:%d\n", | |
500 | ip_vs_fwd_tag(cp), | |
f18ae720 JA |
501 | IP_VS_DBG_ADDR(cp->af, &cp->caddr), ntohs(cp->cport), |
502 | IP_VS_DBG_ADDR(cp->af, &cp->vaddr), ntohs(cp->vport), | |
503 | IP_VS_DBG_ADDR(cp->daf, &cp->daddr), ntohs(cp->dport), | |
cd17f9ed | 504 | cp->flags, atomic_read(&cp->refcnt)); |
1da177e4 LT |
505 | |
506 | ip_vs_conn_stats(cp, svc); | |
507 | return cp; | |
508 | } | |
509 | ||
510 | ||
511 | /* | |
512 | * Pass or drop the packet. | |
513 | * Called by ip_vs_in, when the virtual service is available but | |
514 | * no destination is available for a new connection. | |
515 | */ | |
516 | int ip_vs_leave(struct ip_vs_service *svc, struct sk_buff *skb, | |
d4383f04 | 517 | struct ip_vs_proto_data *pd, struct ip_vs_iphdr *iph) |
1da177e4 | 518 | { |
014d730d | 519 | __be16 _ports[2], *pptr; |
a7a86b86 SH |
520 | #ifdef CONFIG_SYSCTL |
521 | struct net *net; | |
522 | struct netns_ipvs *ipvs; | |
2a3b791e | 523 | int unicast; |
a7a86b86 | 524 | #endif |
9330419d | 525 | |
d4383f04 | 526 | pptr = frag_safe_skb_hp(skb, iph->len, sizeof(_ports), _ports, iph); |
1da177e4 | 527 | if (pptr == NULL) { |
1da177e4 LT |
528 | return NF_DROP; |
529 | } | |
a7a86b86 SH |
530 | |
531 | #ifdef CONFIG_SYSCTL | |
4a98480b | 532 | net = skb_net(skb); |
1da177e4 | 533 | |
2a3b791e JV |
534 | #ifdef CONFIG_IP_VS_IPV6 |
535 | if (svc->af == AF_INET6) | |
d4383f04 | 536 | unicast = ipv6_addr_type(&iph->daddr.in6) & IPV6_ADDR_UNICAST; |
2a3b791e JV |
537 | else |
538 | #endif | |
d4383f04 | 539 | unicast = (inet_addr_type(net, iph->daddr.ip) == RTN_UNICAST); |
2a3b791e | 540 | |
1da177e4 | 541 | /* if it is fwmark-based service, the cache_bypass sysctl is up |
2a3b791e | 542 | and the destination is a non-local unicast, then create |
1da177e4 | 543 | a cache_bypass connection entry */ |
4a98480b | 544 | ipvs = net_ipvs(net); |
a0840e2e | 545 | if (ipvs->sysctl_cache_bypass && svc->fwmark && unicast) { |
ad542ced | 546 | int ret; |
1da177e4 | 547 | struct ip_vs_conn *cp; |
3575792e | 548 | unsigned int flags = (svc->flags & IP_VS_SVC_F_ONEPACKET && |
d4383f04 | 549 | iph->protocol == IPPROTO_UDP) ? |
3575792e | 550 | IP_VS_CONN_F_ONE_PACKET : 0; |
dff630dd | 551 | union nf_inet_addr daddr = { .all = { 0, 0, 0, 0 } }; |
1da177e4 | 552 | |
1da177e4 | 553 | /* create a new connection entry */ |
1e3e238e | 554 | IP_VS_DBG(6, "%s(): create a cache_bypass entry\n", __func__); |
f11017ec SH |
555 | { |
556 | struct ip_vs_conn_param p; | |
d4383f04 JDB |
557 | ip_vs_conn_fill_param(svc->net, svc->af, iph->protocol, |
558 | &iph->saddr, pptr[0], | |
559 | &iph->daddr, pptr[1], &p); | |
ba38528a | 560 | cp = ip_vs_conn_new(&p, svc->af, &daddr, 0, |
f11017ec | 561 | IP_VS_CONN_F_BYPASS | flags, |
0e051e68 | 562 | NULL, skb->mark); |
f11017ec SH |
563 | if (!cp) |
564 | return NF_DROP; | |
565 | } | |
1da177e4 LT |
566 | |
567 | /* statistics */ | |
568 | ip_vs_in_stats(cp, skb); | |
569 | ||
570 | /* set state */ | |
4a516f11 | 571 | ip_vs_set_state(cp, IP_VS_DIR_INPUT, skb, pd); |
1da177e4 LT |
572 | |
573 | /* transmit the first SYN packet */ | |
d4383f04 | 574 | ret = cp->packet_xmit(skb, cp, pd->pp, iph); |
1da177e4 LT |
575 | /* do not touch skb anymore */ |
576 | ||
577 | atomic_inc(&cp->in_pkts); | |
578 | ip_vs_conn_put(cp); | |
579 | return ret; | |
580 | } | |
a7a86b86 | 581 | #endif |
1da177e4 LT |
582 | |
583 | /* | |
584 | * When the virtual ftp service is presented, packets destined | |
585 | * for other services on the VIP may get here (except services | |
586 | * listed in the ipvs table), pass the packets, because it is | |
587 | * not ipvs job to decide to drop the packets. | |
588 | */ | |
ceec4c38 | 589 | if ((svc->port == FTPPORT) && (pptr[1] != FTPPORT)) |
1da177e4 | 590 | return NF_ACCEPT; |
1da177e4 LT |
591 | |
592 | /* | |
593 | * Notify the client that the destination is unreachable, and | |
594 | * release the socket buffer. | |
595 | * Since it is in IP layer, the TCP socket is not actually | |
596 | * created, the TCP RST packet cannot be sent, instead that | |
597 | * ICMP_PORT_UNREACH is sent here no matter it is TCP/UDP. --WZ | |
598 | */ | |
2a3b791e | 599 | #ifdef CONFIG_IP_VS_IPV6 |
cb59155f JA |
600 | if (svc->af == AF_INET6) { |
601 | if (!skb->dev) { | |
9fd0fa7a | 602 | struct net *net_ = dev_net(skb_dst(skb)->dev); |
cb59155f | 603 | |
9fd0fa7a | 604 | skb->dev = net_->loopback_dev; |
cb59155f | 605 | } |
3ffe533c | 606 | icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0); |
cb59155f | 607 | } else |
2a3b791e JV |
608 | #endif |
609 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0); | |
610 | ||
1da177e4 LT |
611 | return NF_DROP; |
612 | } | |
613 | ||
84b3cee3 SH |
614 | #ifdef CONFIG_SYSCTL |
615 | ||
616 | static int sysctl_snat_reroute(struct sk_buff *skb) | |
617 | { | |
618 | struct netns_ipvs *ipvs = net_ipvs(skb_net(skb)); | |
619 | return ipvs->sysctl_snat_reroute; | |
620 | } | |
621 | ||
0cfa558e SH |
622 | static int sysctl_nat_icmp_send(struct net *net) |
623 | { | |
624 | struct netns_ipvs *ipvs = net_ipvs(net); | |
625 | return ipvs->sysctl_nat_icmp_send; | |
626 | } | |
627 | ||
71a8ab6c SH |
628 | static int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs) |
629 | { | |
630 | return ipvs->sysctl_expire_nodest_conn; | |
631 | } | |
632 | ||
84b3cee3 SH |
633 | #else |
634 | ||
635 | static int sysctl_snat_reroute(struct sk_buff *skb) { return 0; } | |
0cfa558e | 636 | static int sysctl_nat_icmp_send(struct net *net) { return 0; } |
71a8ab6c | 637 | static int sysctl_expire_nodest_conn(struct netns_ipvs *ipvs) { return 0; } |
84b3cee3 SH |
638 | |
639 | #endif | |
640 | ||
b1550f22 | 641 | __sum16 ip_vs_checksum_complete(struct sk_buff *skb, int offset) |
1da177e4 | 642 | { |
d3bc23e7 | 643 | return csum_fold(skb_checksum(skb, offset, skb->len - offset, 0)); |
1da177e4 LT |
644 | } |
645 | ||
1ca5bb54 JA |
646 | static inline enum ip_defrag_users ip_vs_defrag_user(unsigned int hooknum) |
647 | { | |
648 | if (NF_INET_LOCAL_IN == hooknum) | |
649 | return IP_DEFRAG_VS_IN; | |
650 | if (NF_INET_FORWARD == hooknum) | |
651 | return IP_DEFRAG_VS_FWD; | |
652 | return IP_DEFRAG_VS_OUT; | |
653 | } | |
654 | ||
776c729e | 655 | static inline int ip_vs_gather_frags(struct sk_buff *skb, u_int32_t user) |
1da177e4 | 656 | { |
ac69269a | 657 | int err; |
776c729e | 658 | |
ac69269a JA |
659 | local_bh_disable(); |
660 | err = ip_defrag(skb, user); | |
661 | local_bh_enable(); | |
776c729e | 662 | if (!err) |
eddc9ec5 | 663 | ip_send_check(ip_hdr(skb)); |
776c729e HX |
664 | |
665 | return err; | |
1da177e4 LT |
666 | } |
667 | ||
579eb62a JA |
668 | static int ip_vs_route_me_harder(int af, struct sk_buff *skb, |
669 | unsigned int hooknum) | |
ba4fd7e9 | 670 | { |
579eb62a JA |
671 | if (!sysctl_snat_reroute(skb)) |
672 | return 0; | |
673 | /* Reroute replies only to remote clients (FORWARD and LOCAL_OUT) */ | |
674 | if (NF_INET_LOCAL_IN == hooknum) | |
675 | return 0; | |
ba4fd7e9 SH |
676 | #ifdef CONFIG_IP_VS_IPV6 |
677 | if (af == AF_INET6) { | |
579eb62a JA |
678 | struct dst_entry *dst = skb_dst(skb); |
679 | ||
680 | if (dst->dev && !(dst->dev->flags & IFF_LOOPBACK) && | |
681 | ip6_route_me_harder(skb) != 0) | |
ba4fd7e9 SH |
682 | return 1; |
683 | } else | |
684 | #endif | |
579eb62a | 685 | if (!(skb_rtable(skb)->rt_flags & RTCF_LOCAL) && |
ba4fd7e9 SH |
686 | ip_route_me_harder(skb, RTN_LOCAL) != 0) |
687 | return 1; | |
688 | ||
689 | return 0; | |
690 | } | |
691 | ||
1da177e4 LT |
692 | /* |
693 | * Packet has been made sufficiently writable in caller | |
694 | * - inout: 1=in->out, 0=out->in | |
695 | */ | |
696 | void ip_vs_nat_icmp(struct sk_buff *skb, struct ip_vs_protocol *pp, | |
697 | struct ip_vs_conn *cp, int inout) | |
698 | { | |
eddc9ec5 | 699 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 | 700 | unsigned int icmp_offset = iph->ihl*4; |
d56f90a7 ACM |
701 | struct icmphdr *icmph = (struct icmphdr *)(skb_network_header(skb) + |
702 | icmp_offset); | |
1da177e4 LT |
703 | struct iphdr *ciph = (struct iphdr *)(icmph + 1); |
704 | ||
705 | if (inout) { | |
e7ade46a | 706 | iph->saddr = cp->vaddr.ip; |
1da177e4 | 707 | ip_send_check(iph); |
e7ade46a | 708 | ciph->daddr = cp->vaddr.ip; |
1da177e4 LT |
709 | ip_send_check(ciph); |
710 | } else { | |
e7ade46a | 711 | iph->daddr = cp->daddr.ip; |
1da177e4 | 712 | ip_send_check(iph); |
e7ade46a | 713 | ciph->saddr = cp->daddr.ip; |
1da177e4 LT |
714 | ip_send_check(ciph); |
715 | } | |
716 | ||
2906f66a VMR |
717 | /* the TCP/UDP/SCTP port */ |
718 | if (IPPROTO_TCP == ciph->protocol || IPPROTO_UDP == ciph->protocol || | |
719 | IPPROTO_SCTP == ciph->protocol) { | |
014d730d | 720 | __be16 *ports = (void *)ciph + ciph->ihl*4; |
1da177e4 LT |
721 | |
722 | if (inout) | |
723 | ports[1] = cp->vport; | |
724 | else | |
725 | ports[0] = cp->dport; | |
726 | } | |
727 | ||
728 | /* And finally the ICMP checksum */ | |
729 | icmph->checksum = 0; | |
730 | icmph->checksum = ip_vs_checksum_complete(skb, icmp_offset); | |
731 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
732 | ||
733 | if (inout) | |
0d79641a | 734 | IP_VS_DBG_PKT(11, AF_INET, pp, skb, (void *)ciph - (void *)iph, |
1da177e4 LT |
735 | "Forwarding altered outgoing ICMP"); |
736 | else | |
0d79641a | 737 | IP_VS_DBG_PKT(11, AF_INET, pp, skb, (void *)ciph - (void *)iph, |
1da177e4 LT |
738 | "Forwarding altered incoming ICMP"); |
739 | } | |
740 | ||
b3cdd2a7 JV |
741 | #ifdef CONFIG_IP_VS_IPV6 |
742 | void ip_vs_nat_icmp_v6(struct sk_buff *skb, struct ip_vs_protocol *pp, | |
743 | struct ip_vs_conn *cp, int inout) | |
744 | { | |
745 | struct ipv6hdr *iph = ipv6_hdr(skb); | |
63dca2c0 JDB |
746 | unsigned int icmp_offset = 0; |
747 | unsigned int offs = 0; /* header offset*/ | |
748 | int protocol; | |
749 | struct icmp6hdr *icmph; | |
750 | struct ipv6hdr *ciph; | |
751 | unsigned short fragoffs; | |
752 | ||
753 | ipv6_find_hdr(skb, &icmp_offset, IPPROTO_ICMPV6, &fragoffs, NULL); | |
754 | icmph = (struct icmp6hdr *)(skb_network_header(skb) + icmp_offset); | |
755 | offs = icmp_offset + sizeof(struct icmp6hdr); | |
756 | ciph = (struct ipv6hdr *)(skb_network_header(skb) + offs); | |
757 | ||
758 | protocol = ipv6_find_hdr(skb, &offs, -1, &fragoffs, NULL); | |
b3cdd2a7 JV |
759 | |
760 | if (inout) { | |
761 | iph->saddr = cp->vaddr.in6; | |
762 | ciph->daddr = cp->vaddr.in6; | |
763 | } else { | |
764 | iph->daddr = cp->daddr.in6; | |
765 | ciph->saddr = cp->daddr.in6; | |
766 | } | |
767 | ||
2906f66a | 768 | /* the TCP/UDP/SCTP port */ |
63dca2c0 JDB |
769 | if (!fragoffs && (IPPROTO_TCP == protocol || IPPROTO_UDP == protocol || |
770 | IPPROTO_SCTP == protocol)) { | |
771 | __be16 *ports = (void *)(skb_network_header(skb) + offs); | |
b3cdd2a7 | 772 | |
63dca2c0 JDB |
773 | IP_VS_DBG(11, "%s() changed port %d to %d\n", __func__, |
774 | ntohs(inout ? ports[1] : ports[0]), | |
775 | ntohs(inout ? cp->vport : cp->dport)); | |
b3cdd2a7 JV |
776 | if (inout) |
777 | ports[1] = cp->vport; | |
778 | else | |
779 | ports[0] = cp->dport; | |
780 | } | |
781 | ||
782 | /* And finally the ICMP checksum */ | |
8870f842 SH |
783 | icmph->icmp6_cksum = ~csum_ipv6_magic(&iph->saddr, &iph->daddr, |
784 | skb->len - icmp_offset, | |
785 | IPPROTO_ICMPV6, 0); | |
786 | skb->csum_start = skb_network_header(skb) - skb->head + icmp_offset; | |
787 | skb->csum_offset = offsetof(struct icmp6hdr, icmp6_cksum); | |
788 | skb->ip_summed = CHECKSUM_PARTIAL; | |
b3cdd2a7 JV |
789 | |
790 | if (inout) | |
0d79641a JA |
791 | IP_VS_DBG_PKT(11, AF_INET6, pp, skb, |
792 | (void *)ciph - (void *)iph, | |
793 | "Forwarding altered outgoing ICMPv6"); | |
b3cdd2a7 | 794 | else |
0d79641a JA |
795 | IP_VS_DBG_PKT(11, AF_INET6, pp, skb, |
796 | (void *)ciph - (void *)iph, | |
797 | "Forwarding altered incoming ICMPv6"); | |
b3cdd2a7 JV |
798 | } |
799 | #endif | |
800 | ||
4856c84c | 801 | /* Handle relevant response ICMP messages - forward to the right |
6cb90db5 | 802 | * destination host. |
4856c84c | 803 | */ |
f2428ed5 SH |
804 | static int handle_response_icmp(int af, struct sk_buff *skb, |
805 | union nf_inet_addr *snet, | |
806 | __u8 protocol, struct ip_vs_conn *cp, | |
4856c84c | 807 | struct ip_vs_protocol *pp, |
579eb62a JA |
808 | unsigned int offset, unsigned int ihl, |
809 | unsigned int hooknum) | |
4856c84c MT |
810 | { |
811 | unsigned int verdict = NF_DROP; | |
812 | ||
813 | if (IP_VS_FWD_METHOD(cp) != 0) { | |
1e3e238e HE |
814 | pr_err("shouldn't reach here, because the box is on the " |
815 | "half connection in the tun/dr module.\n"); | |
4856c84c MT |
816 | } |
817 | ||
818 | /* Ensure the checksum is correct */ | |
819 | if (!skb_csum_unnecessary(skb) && ip_vs_checksum_complete(skb, ihl)) { | |
820 | /* Failed checksum! */ | |
f2428ed5 SH |
821 | IP_VS_DBG_BUF(1, "Forward ICMP: failed checksum from %s!\n", |
822 | IP_VS_DBG_ADDR(af, snet)); | |
4856c84c MT |
823 | goto out; |
824 | } | |
825 | ||
2906f66a VMR |
826 | if (IPPROTO_TCP == protocol || IPPROTO_UDP == protocol || |
827 | IPPROTO_SCTP == protocol) | |
4856c84c MT |
828 | offset += 2 * sizeof(__u16); |
829 | if (!skb_make_writable(skb, offset)) | |
830 | goto out; | |
831 | ||
f2428ed5 SH |
832 | #ifdef CONFIG_IP_VS_IPV6 |
833 | if (af == AF_INET6) | |
834 | ip_vs_nat_icmp_v6(skb, pp, cp, 1); | |
835 | else | |
836 | #endif | |
837 | ip_vs_nat_icmp(skb, pp, cp, 1); | |
4856c84c | 838 | |
579eb62a | 839 | if (ip_vs_route_me_harder(af, skb, hooknum)) |
ba4fd7e9 | 840 | goto out; |
f5a41847 | 841 | |
4856c84c MT |
842 | /* do the statistics and put it back */ |
843 | ip_vs_out_stats(cp, skb); | |
844 | ||
cf356d69 | 845 | skb->ipvs_property = 1; |
f4bc17cd | 846 | if (!(cp->flags & IP_VS_CONN_F_NFCT)) |
cf356d69 | 847 | ip_vs_notrack(skb); |
f4bc17cd JA |
848 | else |
849 | ip_vs_update_conntrack(skb, cp, 0); | |
4856c84c MT |
850 | verdict = NF_ACCEPT; |
851 | ||
852 | out: | |
853 | __ip_vs_conn_put(cp); | |
854 | ||
855 | return verdict; | |
856 | } | |
857 | ||
1da177e4 LT |
858 | /* |
859 | * Handle ICMP messages in the inside-to-outside direction (outgoing). | |
4856c84c | 860 | * Find any that might be relevant, check against existing connections. |
1da177e4 | 861 | * Currently handles error types - unreachable, quench, ttl exceeded. |
1da177e4 | 862 | */ |
1ca5bb54 JA |
863 | static int ip_vs_out_icmp(struct sk_buff *skb, int *related, |
864 | unsigned int hooknum) | |
1da177e4 | 865 | { |
1da177e4 LT |
866 | struct iphdr *iph; |
867 | struct icmphdr _icmph, *ic; | |
868 | struct iphdr _ciph, *cih; /* The ip header contained within the ICMP */ | |
51ef348b | 869 | struct ip_vs_iphdr ciph; |
1da177e4 LT |
870 | struct ip_vs_conn *cp; |
871 | struct ip_vs_protocol *pp; | |
4856c84c | 872 | unsigned int offset, ihl; |
f2428ed5 | 873 | union nf_inet_addr snet; |
1da177e4 LT |
874 | |
875 | *related = 1; | |
876 | ||
877 | /* reassemble IP fragments */ | |
56f8a75c | 878 | if (ip_is_fragment(ip_hdr(skb))) { |
1ca5bb54 | 879 | if (ip_vs_gather_frags(skb, ip_vs_defrag_user(hooknum))) |
1da177e4 | 880 | return NF_STOLEN; |
1da177e4 LT |
881 | } |
882 | ||
eddc9ec5 | 883 | iph = ip_hdr(skb); |
1da177e4 LT |
884 | offset = ihl = iph->ihl * 4; |
885 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
886 | if (ic == NULL) | |
887 | return NF_DROP; | |
888 | ||
14d5e834 | 889 | IP_VS_DBG(12, "Outgoing ICMP (%d,%d) %pI4->%pI4\n", |
1da177e4 | 890 | ic->type, ntohs(icmp_id(ic)), |
14d5e834 | 891 | &iph->saddr, &iph->daddr); |
1da177e4 LT |
892 | |
893 | /* | |
894 | * Work through seeing if this is for us. | |
895 | * These checks are supposed to be in an order that means easy | |
896 | * things are checked first to speed up processing.... however | |
897 | * this means that some packets will manage to get a long way | |
898 | * down this stack and then be rejected, but that's life. | |
899 | */ | |
900 | if ((ic->type != ICMP_DEST_UNREACH) && | |
901 | (ic->type != ICMP_SOURCE_QUENCH) && | |
902 | (ic->type != ICMP_TIME_EXCEEDED)) { | |
903 | *related = 0; | |
904 | return NF_ACCEPT; | |
905 | } | |
906 | ||
907 | /* Now find the contained IP header */ | |
908 | offset += sizeof(_icmph); | |
909 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
910 | if (cih == NULL) | |
911 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
912 | ||
913 | pp = ip_vs_proto_get(cih->protocol); | |
914 | if (!pp) | |
915 | return NF_ACCEPT; | |
916 | ||
917 | /* Is the embedded protocol header present? */ | |
4412ec49 | 918 | if (unlikely(cih->frag_off & htons(IP_OFFSET) && |
1da177e4 LT |
919 | pp->dont_defrag)) |
920 | return NF_ACCEPT; | |
921 | ||
0d79641a JA |
922 | IP_VS_DBG_PKT(11, AF_INET, pp, skb, offset, |
923 | "Checking outgoing ICMP for"); | |
1da177e4 | 924 | |
63dca2c0 JDB |
925 | ip_vs_fill_ip4hdr(cih, &ciph); |
926 | ciph.len += offset; | |
1da177e4 | 927 | /* The embedded headers contain source and dest in reverse order */ |
d4383f04 | 928 | cp = pp->conn_out_get(AF_INET, skb, &ciph, 1); |
1da177e4 LT |
929 | if (!cp) |
930 | return NF_ACCEPT; | |
931 | ||
f2428ed5 SH |
932 | snet.ip = iph->saddr; |
933 | return handle_response_icmp(AF_INET, skb, &snet, cih->protocol, cp, | |
579eb62a | 934 | pp, ciph.len, ihl, hooknum); |
1da177e4 LT |
935 | } |
936 | ||
2a3b791e | 937 | #ifdef CONFIG_IP_VS_IPV6 |
1ca5bb54 | 938 | static int ip_vs_out_icmp_v6(struct sk_buff *skb, int *related, |
d4383f04 | 939 | unsigned int hooknum, struct ip_vs_iphdr *ipvsh) |
2a3b791e | 940 | { |
2a3b791e | 941 | struct icmp6hdr _icmph, *ic; |
63dca2c0 JDB |
942 | struct ipv6hdr _ip6h, *ip6h; /* The ip header contained within ICMP */ |
943 | struct ip_vs_iphdr ciph = {.flags = 0, .fragoffs = 0};/*Contained IP */ | |
2a3b791e JV |
944 | struct ip_vs_conn *cp; |
945 | struct ip_vs_protocol *pp; | |
f2428ed5 | 946 | union nf_inet_addr snet; |
63dca2c0 | 947 | unsigned int writable; |
2a3b791e | 948 | |
63dca2c0 | 949 | *related = 1; |
2f74713d | 950 | ic = frag_safe_skb_hp(skb, ipvsh->len, sizeof(_icmph), &_icmph, ipvsh); |
2a3b791e JV |
951 | if (ic == NULL) |
952 | return NF_DROP; | |
953 | ||
2a3b791e JV |
954 | /* |
955 | * Work through seeing if this is for us. | |
956 | * These checks are supposed to be in an order that means easy | |
957 | * things are checked first to speed up processing.... however | |
958 | * this means that some packets will manage to get a long way | |
959 | * down this stack and then be rejected, but that's life. | |
960 | */ | |
2fab8917 | 961 | if (ic->icmp6_type & ICMPV6_INFOMSG_MASK) { |
2a3b791e JV |
962 | *related = 0; |
963 | return NF_ACCEPT; | |
964 | } | |
2f74713d JDB |
965 | /* Fragment header that is before ICMP header tells us that: |
966 | * it's not an error message since they can't be fragmented. | |
967 | */ | |
e7165030 | 968 | if (ipvsh->flags & IP6_FH_F_FRAG) |
2f74713d | 969 | return NF_DROP; |
2a3b791e | 970 | |
63dca2c0 JDB |
971 | IP_VS_DBG(8, "Outgoing ICMPv6 (%d,%d) %pI6c->%pI6c\n", |
972 | ic->icmp6_type, ntohs(icmpv6_id(ic)), | |
973 | &ipvsh->saddr, &ipvsh->daddr); | |
974 | ||
2a3b791e | 975 | /* Now find the contained IP header */ |
63dca2c0 JDB |
976 | ciph.len = ipvsh->len + sizeof(_icmph); |
977 | ip6h = skb_header_pointer(skb, ciph.len, sizeof(_ip6h), &_ip6h); | |
978 | if (ip6h == NULL) | |
2a3b791e | 979 | return NF_ACCEPT; /* The packet looks wrong, ignore */ |
63dca2c0 JDB |
980 | ciph.saddr.in6 = ip6h->saddr; /* conn_out_get() handles reverse order */ |
981 | ciph.daddr.in6 = ip6h->daddr; | |
982 | /* skip possible IPv6 exthdrs of contained IPv6 packet */ | |
983 | ciph.protocol = ipv6_find_hdr(skb, &ciph.len, -1, &ciph.fragoffs, NULL); | |
984 | if (ciph.protocol < 0) | |
985 | return NF_ACCEPT; /* Contained IPv6 hdr looks wrong, ignore */ | |
986 | ||
987 | pp = ip_vs_proto_get(ciph.protocol); | |
2a3b791e JV |
988 | if (!pp) |
989 | return NF_ACCEPT; | |
990 | ||
2a3b791e | 991 | /* The embedded headers contain source and dest in reverse order */ |
d4383f04 | 992 | cp = pp->conn_out_get(AF_INET6, skb, &ciph, 1); |
2a3b791e JV |
993 | if (!cp) |
994 | return NF_ACCEPT; | |
995 | ||
63dca2c0 JDB |
996 | snet.in6 = ciph.saddr.in6; |
997 | writable = ciph.len; | |
998 | return handle_response_icmp(AF_INET6, skb, &snet, ciph.protocol, cp, | |
579eb62a JA |
999 | pp, writable, sizeof(struct ipv6hdr), |
1000 | hooknum); | |
2a3b791e JV |
1001 | } |
1002 | #endif | |
1003 | ||
2906f66a VMR |
1004 | /* |
1005 | * Check if sctp chunc is ABORT chunk | |
1006 | */ | |
1007 | static inline int is_sctp_abort(const struct sk_buff *skb, int nh_len) | |
1008 | { | |
1009 | sctp_chunkhdr_t *sch, schunk; | |
1010 | sch = skb_header_pointer(skb, nh_len + sizeof(sctp_sctphdr_t), | |
1011 | sizeof(schunk), &schunk); | |
1012 | if (sch == NULL) | |
1013 | return 0; | |
1014 | if (sch->type == SCTP_CID_ABORT) | |
1015 | return 1; | |
1016 | return 0; | |
1017 | } | |
1018 | ||
2a3b791e | 1019 | static inline int is_tcp_reset(const struct sk_buff *skb, int nh_len) |
1da177e4 LT |
1020 | { |
1021 | struct tcphdr _tcph, *th; | |
1022 | ||
2a3b791e | 1023 | th = skb_header_pointer(skb, nh_len, sizeof(_tcph), &_tcph); |
1da177e4 LT |
1024 | if (th == NULL) |
1025 | return 0; | |
1026 | return th->rst; | |
1027 | } | |
1028 | ||
dc7b3eb9 GL |
1029 | static inline bool is_new_conn(const struct sk_buff *skb, |
1030 | struct ip_vs_iphdr *iph) | |
1031 | { | |
1032 | switch (iph->protocol) { | |
1033 | case IPPROTO_TCP: { | |
1034 | struct tcphdr _tcph, *th; | |
1035 | ||
1036 | th = skb_header_pointer(skb, iph->len, sizeof(_tcph), &_tcph); | |
1037 | if (th == NULL) | |
1038 | return false; | |
1039 | return th->syn; | |
1040 | } | |
1041 | case IPPROTO_SCTP: { | |
1042 | sctp_chunkhdr_t *sch, schunk; | |
1043 | ||
1044 | sch = skb_header_pointer(skb, iph->len + sizeof(sctp_sctphdr_t), | |
1045 | sizeof(schunk), &schunk); | |
1046 | if (sch == NULL) | |
1047 | return false; | |
1048 | return sch->type == SCTP_CID_INIT; | |
1049 | } | |
1050 | default: | |
1051 | return false; | |
1052 | } | |
1053 | } | |
1054 | ||
d752c364 MRL |
1055 | static inline bool is_new_conn_expected(const struct ip_vs_conn *cp, |
1056 | int conn_reuse_mode) | |
1057 | { | |
1058 | /* Controlled (FTP DATA or persistence)? */ | |
1059 | if (cp->control) | |
1060 | return false; | |
1061 | ||
1062 | switch (cp->protocol) { | |
1063 | case IPPROTO_TCP: | |
1064 | return (cp->state == IP_VS_TCP_S_TIME_WAIT) || | |
1065 | ((conn_reuse_mode & 2) && | |
1066 | (cp->state == IP_VS_TCP_S_FIN_WAIT) && | |
1067 | (cp->flags & IP_VS_CONN_F_NOOUTPUT)); | |
1068 | case IPPROTO_SCTP: | |
1069 | return cp->state == IP_VS_SCTP_S_CLOSED; | |
1070 | default: | |
1071 | return false; | |
1072 | } | |
1073 | } | |
1074 | ||
4856c84c | 1075 | /* Handle response packets: rewrite addresses and send away... |
4856c84c MT |
1076 | */ |
1077 | static unsigned int | |
9330419d | 1078 | handle_response(int af, struct sk_buff *skb, struct ip_vs_proto_data *pd, |
579eb62a JA |
1079 | struct ip_vs_conn *cp, struct ip_vs_iphdr *iph, |
1080 | unsigned int hooknum) | |
4856c84c | 1081 | { |
9330419d HS |
1082 | struct ip_vs_protocol *pp = pd->pp; |
1083 | ||
0d79641a | 1084 | IP_VS_DBG_PKT(11, af, pp, skb, 0, "Outgoing packet"); |
4856c84c | 1085 | |
d4383f04 | 1086 | if (!skb_make_writable(skb, iph->len)) |
4856c84c MT |
1087 | goto drop; |
1088 | ||
1089 | /* mangle the packet */ | |
d4383f04 | 1090 | if (pp->snat_handler && !pp->snat_handler(skb, pp, cp, iph)) |
4856c84c MT |
1091 | goto drop; |
1092 | ||
1093 | #ifdef CONFIG_IP_VS_IPV6 | |
1094 | if (af == AF_INET6) | |
1095 | ipv6_hdr(skb)->saddr = cp->vaddr.in6; | |
1096 | else | |
1097 | #endif | |
1098 | { | |
1099 | ip_hdr(skb)->saddr = cp->vaddr.ip; | |
1100 | ip_send_check(ip_hdr(skb)); | |
1101 | } | |
1102 | ||
8a803040 JA |
1103 | /* |
1104 | * nf_iterate does not expect change in the skb->dst->dev. | |
1105 | * It looks like it is not fatal to enable this code for hooks | |
1106 | * where our handlers are at the end of the chain list and | |
1107 | * when all next handlers use skb->dst->dev and not outdev. | |
1108 | * It will definitely route properly the inout NAT traffic | |
1109 | * when multiple paths are used. | |
1110 | */ | |
1111 | ||
4856c84c MT |
1112 | /* For policy routing, packets originating from this |
1113 | * machine itself may be routed differently to packets | |
1114 | * passing through. We want this packet to be routed as | |
1115 | * if it came from this machine itself. So re-compute | |
1116 | * the routing information. | |
1117 | */ | |
579eb62a | 1118 | if (ip_vs_route_me_harder(af, skb, hooknum)) |
ba4fd7e9 | 1119 | goto drop; |
4856c84c | 1120 | |
0d79641a | 1121 | IP_VS_DBG_PKT(10, af, pp, skb, 0, "After SNAT"); |
4856c84c MT |
1122 | |
1123 | ip_vs_out_stats(cp, skb); | |
9330419d | 1124 | ip_vs_set_state(cp, IP_VS_DIR_OUTPUT, skb, pd); |
cf356d69 | 1125 | skb->ipvs_property = 1; |
f4bc17cd | 1126 | if (!(cp->flags & IP_VS_CONN_F_NFCT)) |
cf356d69 | 1127 | ip_vs_notrack(skb); |
f4bc17cd JA |
1128 | else |
1129 | ip_vs_update_conntrack(skb, cp, 0); | |
4856c84c MT |
1130 | ip_vs_conn_put(cp); |
1131 | ||
4856c84c MT |
1132 | LeaveFunction(11); |
1133 | return NF_ACCEPT; | |
1134 | ||
1135 | drop: | |
1136 | ip_vs_conn_put(cp); | |
1137 | kfree_skb(skb); | |
f4bc17cd | 1138 | LeaveFunction(11); |
4856c84c MT |
1139 | return NF_STOLEN; |
1140 | } | |
1141 | ||
1da177e4 | 1142 | /* |
4856c84c | 1143 | * Check if outgoing packet belongs to the established ip_vs_conn. |
1da177e4 LT |
1144 | */ |
1145 | static unsigned int | |
fc604767 | 1146 | ip_vs_out(unsigned int hooknum, struct sk_buff *skb, int af) |
1da177e4 | 1147 | { |
fc723250 | 1148 | struct net *net = NULL; |
51ef348b | 1149 | struct ip_vs_iphdr iph; |
1da177e4 | 1150 | struct ip_vs_protocol *pp; |
9330419d | 1151 | struct ip_vs_proto_data *pd; |
1da177e4 | 1152 | struct ip_vs_conn *cp; |
1da177e4 LT |
1153 | |
1154 | EnterFunction(11); | |
1155 | ||
fc604767 | 1156 | /* Already marked as IPVS request or reply? */ |
6869c4d8 | 1157 | if (skb->ipvs_property) |
1da177e4 LT |
1158 | return NF_ACCEPT; |
1159 | ||
fc604767 JA |
1160 | /* Bad... Do not break raw sockets */ |
1161 | if (unlikely(skb->sk != NULL && hooknum == NF_INET_LOCAL_OUT && | |
1162 | af == AF_INET)) { | |
1163 | struct sock *sk = skb->sk; | |
1164 | struct inet_sock *inet = inet_sk(skb->sk); | |
1165 | ||
1166 | if (inet && sk->sk_family == PF_INET && inet->nodefrag) | |
1167 | return NF_ACCEPT; | |
1168 | } | |
1169 | ||
1170 | if (unlikely(!skb_dst(skb))) | |
1171 | return NF_ACCEPT; | |
1172 | ||
fc723250 | 1173 | net = skb_net(skb); |
7a4f0761 HS |
1174 | if (!net_ipvs(net)->enable) |
1175 | return NF_ACCEPT; | |
1176 | ||
63dca2c0 | 1177 | ip_vs_fill_iph_skb(af, skb, &iph); |
2a3b791e JV |
1178 | #ifdef CONFIG_IP_VS_IPV6 |
1179 | if (af == AF_INET6) { | |
1180 | if (unlikely(iph.protocol == IPPROTO_ICMPV6)) { | |
1ca5bb54 JA |
1181 | int related; |
1182 | int verdict = ip_vs_out_icmp_v6(skb, &related, | |
d4383f04 | 1183 | hooknum, &iph); |
1da177e4 | 1184 | |
f5a41847 | 1185 | if (related) |
2a3b791e | 1186 | return verdict; |
2a3b791e JV |
1187 | } |
1188 | } else | |
1189 | #endif | |
1190 | if (unlikely(iph.protocol == IPPROTO_ICMP)) { | |
1ca5bb54 JA |
1191 | int related; |
1192 | int verdict = ip_vs_out_icmp(skb, &related, hooknum); | |
2a3b791e | 1193 | |
f5a41847 | 1194 | if (related) |
2a3b791e | 1195 | return verdict; |
2a3b791e | 1196 | } |
1da177e4 | 1197 | |
9330419d HS |
1198 | pd = ip_vs_proto_data_get(net, iph.protocol); |
1199 | if (unlikely(!pd)) | |
1da177e4 | 1200 | return NF_ACCEPT; |
9330419d | 1201 | pp = pd->pp; |
1da177e4 LT |
1202 | |
1203 | /* reassemble IP fragments */ | |
2a3b791e | 1204 | #ifdef CONFIG_IP_VS_IPV6 |
63dca2c0 | 1205 | if (af == AF_INET) |
2a3b791e | 1206 | #endif |
56f8a75c | 1207 | if (unlikely(ip_is_fragment(ip_hdr(skb)) && !pp->dont_defrag)) { |
1ca5bb54 JA |
1208 | if (ip_vs_gather_frags(skb, |
1209 | ip_vs_defrag_user(hooknum))) | |
2a3b791e JV |
1210 | return NF_STOLEN; |
1211 | ||
63dca2c0 | 1212 | ip_vs_fill_ip4hdr(skb_network_header(skb), &iph); |
2a3b791e | 1213 | } |
1da177e4 LT |
1214 | |
1215 | /* | |
1216 | * Check if the packet belongs to an existing entry | |
1217 | */ | |
d4383f04 | 1218 | cp = pp->conn_out_get(af, skb, &iph, 0); |
1da177e4 | 1219 | |
cb59155f | 1220 | if (likely(cp)) |
579eb62a | 1221 | return handle_response(af, skb, pd, cp, &iph, hooknum); |
0cfa558e | 1222 | if (sysctl_nat_icmp_send(net) && |
cb59155f JA |
1223 | (pp->protocol == IPPROTO_TCP || |
1224 | pp->protocol == IPPROTO_UDP || | |
1225 | pp->protocol == IPPROTO_SCTP)) { | |
1226 | __be16 _ports[2], *pptr; | |
1227 | ||
2f74713d JDB |
1228 | pptr = frag_safe_skb_hp(skb, iph.len, |
1229 | sizeof(_ports), _ports, &iph); | |
cb59155f JA |
1230 | if (pptr == NULL) |
1231 | return NF_ACCEPT; /* Not for me */ | |
276472ea JA |
1232 | if (ip_vs_has_real_service(net, af, iph.protocol, &iph.saddr, |
1233 | pptr[0])) { | |
cb59155f JA |
1234 | /* |
1235 | * Notify the real server: there is no | |
1236 | * existing entry if it is not RST | |
1237 | * packet or not TCP packet. | |
1238 | */ | |
1239 | if ((iph.protocol != IPPROTO_TCP && | |
1240 | iph.protocol != IPPROTO_SCTP) | |
1241 | || ((iph.protocol == IPPROTO_TCP | |
1242 | && !is_tcp_reset(skb, iph.len)) | |
1243 | || (iph.protocol == IPPROTO_SCTP | |
1244 | && !is_sctp_abort(skb, | |
1245 | iph.len)))) { | |
2a3b791e | 1246 | #ifdef CONFIG_IP_VS_IPV6 |
cb59155f | 1247 | if (af == AF_INET6) { |
cb59155f JA |
1248 | if (!skb->dev) |
1249 | skb->dev = net->loopback_dev; | |
1250 | icmpv6_send(skb, | |
1251 | ICMPV6_DEST_UNREACH, | |
1252 | ICMPV6_PORT_UNREACH, | |
1253 | 0); | |
1254 | } else | |
2a3b791e | 1255 | #endif |
cb59155f JA |
1256 | icmp_send(skb, |
1257 | ICMP_DEST_UNREACH, | |
1258 | ICMP_PORT_UNREACH, 0); | |
1259 | return NF_DROP; | |
1da177e4 LT |
1260 | } |
1261 | } | |
1da177e4 | 1262 | } |
0d79641a | 1263 | IP_VS_DBG_PKT(12, af, pp, skb, 0, |
cb59155f JA |
1264 | "ip_vs_out: packet continues traversal as normal"); |
1265 | return NF_ACCEPT; | |
1da177e4 LT |
1266 | } |
1267 | ||
fc604767 | 1268 | /* |
cb59155f JA |
1269 | * It is hooked at the NF_INET_FORWARD and NF_INET_LOCAL_IN chain, |
1270 | * used only for VS/NAT. | |
fc604767 JA |
1271 | * Check if packet is reply for established ip_vs_conn. |
1272 | */ | |
1273 | static unsigned int | |
795aa6ef | 1274 | ip_vs_reply4(const struct nf_hook_ops *ops, struct sk_buff *skb, |
238e54c9 | 1275 | const struct nf_hook_state *state) |
fc604767 | 1276 | { |
795aa6ef | 1277 | return ip_vs_out(ops->hooknum, skb, AF_INET); |
fc604767 JA |
1278 | } |
1279 | ||
1280 | /* | |
1281 | * It is hooked at the NF_INET_LOCAL_OUT chain, used only for VS/NAT. | |
1282 | * Check if packet is reply for established ip_vs_conn. | |
1283 | */ | |
1284 | static unsigned int | |
795aa6ef | 1285 | ip_vs_local_reply4(const struct nf_hook_ops *ops, struct sk_buff *skb, |
238e54c9 | 1286 | const struct nf_hook_state *state) |
fc604767 | 1287 | { |
795aa6ef | 1288 | return ip_vs_out(ops->hooknum, skb, AF_INET); |
fc604767 JA |
1289 | } |
1290 | ||
1291 | #ifdef CONFIG_IP_VS_IPV6 | |
1292 | ||
1293 | /* | |
cb59155f JA |
1294 | * It is hooked at the NF_INET_FORWARD and NF_INET_LOCAL_IN chain, |
1295 | * used only for VS/NAT. | |
fc604767 JA |
1296 | * Check if packet is reply for established ip_vs_conn. |
1297 | */ | |
1298 | static unsigned int | |
795aa6ef | 1299 | ip_vs_reply6(const struct nf_hook_ops *ops, struct sk_buff *skb, |
238e54c9 | 1300 | const struct nf_hook_state *state) |
fc604767 | 1301 | { |
795aa6ef | 1302 | return ip_vs_out(ops->hooknum, skb, AF_INET6); |
fc604767 JA |
1303 | } |
1304 | ||
1305 | /* | |
1306 | * It is hooked at the NF_INET_LOCAL_OUT chain, used only for VS/NAT. | |
1307 | * Check if packet is reply for established ip_vs_conn. | |
1308 | */ | |
1309 | static unsigned int | |
795aa6ef | 1310 | ip_vs_local_reply6(const struct nf_hook_ops *ops, struct sk_buff *skb, |
238e54c9 | 1311 | const struct nf_hook_state *state) |
fc604767 | 1312 | { |
795aa6ef | 1313 | return ip_vs_out(ops->hooknum, skb, AF_INET6); |
fc604767 JA |
1314 | } |
1315 | ||
1316 | #endif | |
1da177e4 LT |
1317 | |
1318 | /* | |
1319 | * Handle ICMP messages in the outside-to-inside direction (incoming). | |
1320 | * Find any that might be relevant, check against existing connections, | |
1321 | * forward to the right destination host if relevant. | |
1322 | * Currently handles error types - unreachable, quench, ttl exceeded. | |
1323 | */ | |
e905a9ed | 1324 | static int |
3db05fea | 1325 | ip_vs_in_icmp(struct sk_buff *skb, int *related, unsigned int hooknum) |
1da177e4 | 1326 | { |
9330419d | 1327 | struct net *net = NULL; |
1da177e4 LT |
1328 | struct iphdr *iph; |
1329 | struct icmphdr _icmph, *ic; | |
1330 | struct iphdr _ciph, *cih; /* The ip header contained within the ICMP */ | |
51ef348b | 1331 | struct ip_vs_iphdr ciph; |
1da177e4 LT |
1332 | struct ip_vs_conn *cp; |
1333 | struct ip_vs_protocol *pp; | |
9330419d | 1334 | struct ip_vs_proto_data *pd; |
f2edb9f7 JA |
1335 | unsigned int offset, offset2, ihl, verdict; |
1336 | bool ipip; | |
1da177e4 LT |
1337 | |
1338 | *related = 1; | |
1339 | ||
1340 | /* reassemble IP fragments */ | |
56f8a75c | 1341 | if (ip_is_fragment(ip_hdr(skb))) { |
1ca5bb54 | 1342 | if (ip_vs_gather_frags(skb, ip_vs_defrag_user(hooknum))) |
1da177e4 | 1343 | return NF_STOLEN; |
1da177e4 LT |
1344 | } |
1345 | ||
eddc9ec5 | 1346 | iph = ip_hdr(skb); |
1da177e4 LT |
1347 | offset = ihl = iph->ihl * 4; |
1348 | ic = skb_header_pointer(skb, offset, sizeof(_icmph), &_icmph); | |
1349 | if (ic == NULL) | |
1350 | return NF_DROP; | |
1351 | ||
14d5e834 | 1352 | IP_VS_DBG(12, "Incoming ICMP (%d,%d) %pI4->%pI4\n", |
1da177e4 | 1353 | ic->type, ntohs(icmp_id(ic)), |
14d5e834 | 1354 | &iph->saddr, &iph->daddr); |
1da177e4 LT |
1355 | |
1356 | /* | |
1357 | * Work through seeing if this is for us. | |
1358 | * These checks are supposed to be in an order that means easy | |
1359 | * things are checked first to speed up processing.... however | |
1360 | * this means that some packets will manage to get a long way | |
1361 | * down this stack and then be rejected, but that's life. | |
1362 | */ | |
1363 | if ((ic->type != ICMP_DEST_UNREACH) && | |
1364 | (ic->type != ICMP_SOURCE_QUENCH) && | |
1365 | (ic->type != ICMP_TIME_EXCEEDED)) { | |
1366 | *related = 0; | |
1367 | return NF_ACCEPT; | |
1368 | } | |
1369 | ||
1370 | /* Now find the contained IP header */ | |
1371 | offset += sizeof(_icmph); | |
1372 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
1373 | if (cih == NULL) | |
1374 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
1375 | ||
9330419d | 1376 | net = skb_net(skb); |
7a4f0761 | 1377 | |
f2edb9f7 JA |
1378 | /* Special case for errors for IPIP packets */ |
1379 | ipip = false; | |
1380 | if (cih->protocol == IPPROTO_IPIP) { | |
1381 | if (unlikely(cih->frag_off & htons(IP_OFFSET))) | |
1382 | return NF_ACCEPT; | |
1383 | /* Error for our IPIP must arrive at LOCAL_IN */ | |
1384 | if (!(skb_rtable(skb)->rt_flags & RTCF_LOCAL)) | |
1385 | return NF_ACCEPT; | |
1386 | offset += cih->ihl * 4; | |
1387 | cih = skb_header_pointer(skb, offset, sizeof(_ciph), &_ciph); | |
1388 | if (cih == NULL) | |
1389 | return NF_ACCEPT; /* The packet looks wrong, ignore */ | |
1390 | ipip = true; | |
1391 | } | |
1392 | ||
9330419d HS |
1393 | pd = ip_vs_proto_data_get(net, cih->protocol); |
1394 | if (!pd) | |
1da177e4 | 1395 | return NF_ACCEPT; |
9330419d | 1396 | pp = pd->pp; |
1da177e4 LT |
1397 | |
1398 | /* Is the embedded protocol header present? */ | |
4412ec49 | 1399 | if (unlikely(cih->frag_off & htons(IP_OFFSET) && |
1da177e4 LT |
1400 | pp->dont_defrag)) |
1401 | return NF_ACCEPT; | |
1402 | ||
0d79641a JA |
1403 | IP_VS_DBG_PKT(11, AF_INET, pp, skb, offset, |
1404 | "Checking incoming ICMP for"); | |
1da177e4 | 1405 | |
f2edb9f7 | 1406 | offset2 = offset; |
63dca2c0 JDB |
1407 | ip_vs_fill_ip4hdr(cih, &ciph); |
1408 | ciph.len += offset; | |
1409 | offset = ciph.len; | |
f2edb9f7 JA |
1410 | /* The embedded headers contain source and dest in reverse order. |
1411 | * For IPIP this is error for request, not for reply. | |
1412 | */ | |
d4383f04 | 1413 | cp = pp->conn_in_get(AF_INET, skb, &ciph, ipip ? 0 : 1); |
6cb90db5 | 1414 | if (!cp) |
1da177e4 LT |
1415 | return NF_ACCEPT; |
1416 | ||
1417 | verdict = NF_DROP; | |
1418 | ||
1419 | /* Ensure the checksum is correct */ | |
60476372 | 1420 | if (!skb_csum_unnecessary(skb) && ip_vs_checksum_complete(skb, ihl)) { |
1da177e4 | 1421 | /* Failed checksum! */ |
14d5e834 HH |
1422 | IP_VS_DBG(1, "Incoming ICMP: failed checksum from %pI4!\n", |
1423 | &iph->saddr); | |
1da177e4 LT |
1424 | goto out; |
1425 | } | |
1426 | ||
f2edb9f7 JA |
1427 | if (ipip) { |
1428 | __be32 info = ic->un.gateway; | |
f44a5f45 PC |
1429 | __u8 type = ic->type; |
1430 | __u8 code = ic->code; | |
f2edb9f7 JA |
1431 | |
1432 | /* Update the MTU */ | |
1433 | if (ic->type == ICMP_DEST_UNREACH && | |
1434 | ic->code == ICMP_FRAG_NEEDED) { | |
1435 | struct ip_vs_dest *dest = cp->dest; | |
1436 | u32 mtu = ntohs(ic->un.frag.mtu); | |
f44a5f45 | 1437 | __be16 frag_off = cih->frag_off; |
f2edb9f7 JA |
1438 | |
1439 | /* Strip outer IP and ICMP, go to IPIP header */ | |
f44a5f45 PC |
1440 | if (pskb_pull(skb, ihl + sizeof(_icmph)) == NULL) |
1441 | goto ignore_ipip; | |
f2edb9f7 JA |
1442 | offset2 -= ihl + sizeof(_icmph); |
1443 | skb_reset_network_header(skb); | |
1444 | IP_VS_DBG(12, "ICMP for IPIP %pI4->%pI4: mtu=%u\n", | |
1445 | &ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr, mtu); | |
f2edb9f7 JA |
1446 | ipv4_update_pmtu(skb, dev_net(skb->dev), |
1447 | mtu, 0, 0, 0, 0); | |
f2edb9f7 | 1448 | /* Client uses PMTUD? */ |
f44a5f45 | 1449 | if (!(frag_off & htons(IP_DF))) |
f2edb9f7 JA |
1450 | goto ignore_ipip; |
1451 | /* Prefer the resulting PMTU */ | |
1452 | if (dest) { | |
026ace06 JA |
1453 | struct ip_vs_dest_dst *dest_dst; |
1454 | ||
1455 | rcu_read_lock(); | |
1456 | dest_dst = rcu_dereference(dest->dest_dst); | |
1457 | if (dest_dst) | |
1458 | mtu = dst_mtu(dest_dst->dst_cache); | |
1459 | rcu_read_unlock(); | |
f2edb9f7 JA |
1460 | } |
1461 | if (mtu > 68 + sizeof(struct iphdr)) | |
1462 | mtu -= sizeof(struct iphdr); | |
1463 | info = htonl(mtu); | |
1464 | } | |
1465 | /* Strip outer IP, ICMP and IPIP, go to IP header of | |
1466 | * original request. | |
1467 | */ | |
f44a5f45 PC |
1468 | if (pskb_pull(skb, offset2) == NULL) |
1469 | goto ignore_ipip; | |
f2edb9f7 JA |
1470 | skb_reset_network_header(skb); |
1471 | IP_VS_DBG(12, "Sending ICMP for %pI4->%pI4: t=%u, c=%u, i=%u\n", | |
1472 | &ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr, | |
f44a5f45 PC |
1473 | type, code, ntohl(info)); |
1474 | icmp_send(skb, type, code, info); | |
f2edb9f7 JA |
1475 | /* ICMP can be shorter but anyways, account it */ |
1476 | ip_vs_out_stats(cp, skb); | |
1477 | ||
1478 | ignore_ipip: | |
1479 | consume_skb(skb); | |
1480 | verdict = NF_STOLEN; | |
1481 | goto out; | |
1482 | } | |
1483 | ||
1da177e4 LT |
1484 | /* do the statistics and put it back */ |
1485 | ip_vs_in_stats(cp, skb); | |
06f3d7f9 JA |
1486 | if (IPPROTO_TCP == cih->protocol || IPPROTO_UDP == cih->protocol || |
1487 | IPPROTO_SCTP == cih->protocol) | |
1da177e4 | 1488 | offset += 2 * sizeof(__u16); |
d4383f04 | 1489 | verdict = ip_vs_icmp_xmit(skb, cp, pp, offset, hooknum, &ciph); |
1da177e4 | 1490 | |
552ad65a | 1491 | out: |
1da177e4 LT |
1492 | __ip_vs_conn_put(cp); |
1493 | ||
1494 | return verdict; | |
1495 | } | |
1496 | ||
2a3b791e | 1497 | #ifdef CONFIG_IP_VS_IPV6 |
d4383f04 JDB |
1498 | static int ip_vs_in_icmp_v6(struct sk_buff *skb, int *related, |
1499 | unsigned int hooknum, struct ip_vs_iphdr *iph) | |
2a3b791e | 1500 | { |
9330419d | 1501 | struct net *net = NULL; |
63dca2c0 | 1502 | struct ipv6hdr _ip6h, *ip6h; |
2a3b791e | 1503 | struct icmp6hdr _icmph, *ic; |
63dca2c0 | 1504 | struct ip_vs_iphdr ciph = {.flags = 0, .fragoffs = 0};/*Contained IP */ |
2a3b791e JV |
1505 | struct ip_vs_conn *cp; |
1506 | struct ip_vs_protocol *pp; | |
9330419d | 1507 | struct ip_vs_proto_data *pd; |
63dca2c0 | 1508 | unsigned int offs_ciph, writable, verdict; |
2a3b791e | 1509 | |
63dca2c0 | 1510 | *related = 1; |
2a3b791e | 1511 | |
2f74713d | 1512 | ic = frag_safe_skb_hp(skb, iph->len, sizeof(_icmph), &_icmph, iph); |
2a3b791e JV |
1513 | if (ic == NULL) |
1514 | return NF_DROP; | |
1515 | ||
2a3b791e JV |
1516 | /* |
1517 | * Work through seeing if this is for us. | |
1518 | * These checks are supposed to be in an order that means easy | |
1519 | * things are checked first to speed up processing.... however | |
1520 | * this means that some packets will manage to get a long way | |
1521 | * down this stack and then be rejected, but that's life. | |
1522 | */ | |
2fab8917 | 1523 | if (ic->icmp6_type & ICMPV6_INFOMSG_MASK) { |
2a3b791e JV |
1524 | *related = 0; |
1525 | return NF_ACCEPT; | |
1526 | } | |
2f74713d JDB |
1527 | /* Fragment header that is before ICMP header tells us that: |
1528 | * it's not an error message since they can't be fragmented. | |
1529 | */ | |
e7165030 | 1530 | if (iph->flags & IP6_FH_F_FRAG) |
2f74713d | 1531 | return NF_DROP; |
2a3b791e | 1532 | |
63dca2c0 JDB |
1533 | IP_VS_DBG(8, "Incoming ICMPv6 (%d,%d) %pI6c->%pI6c\n", |
1534 | ic->icmp6_type, ntohs(icmpv6_id(ic)), | |
1535 | &iph->saddr, &iph->daddr); | |
1536 | ||
2a3b791e | 1537 | /* Now find the contained IP header */ |
63dca2c0 JDB |
1538 | ciph.len = iph->len + sizeof(_icmph); |
1539 | offs_ciph = ciph.len; /* Save ip header offset */ | |
1540 | ip6h = skb_header_pointer(skb, ciph.len, sizeof(_ip6h), &_ip6h); | |
1541 | if (ip6h == NULL) | |
2a3b791e | 1542 | return NF_ACCEPT; /* The packet looks wrong, ignore */ |
63dca2c0 JDB |
1543 | ciph.saddr.in6 = ip6h->saddr; /* conn_in_get() handles reverse order */ |
1544 | ciph.daddr.in6 = ip6h->daddr; | |
1545 | /* skip possible IPv6 exthdrs of contained IPv6 packet */ | |
1546 | ciph.protocol = ipv6_find_hdr(skb, &ciph.len, -1, &ciph.fragoffs, NULL); | |
1547 | if (ciph.protocol < 0) | |
1548 | return NF_ACCEPT; /* Contained IPv6 hdr looks wrong, ignore */ | |
2a3b791e | 1549 | |
9330419d | 1550 | net = skb_net(skb); |
63dca2c0 | 1551 | pd = ip_vs_proto_data_get(net, ciph.protocol); |
9330419d | 1552 | if (!pd) |
2a3b791e | 1553 | return NF_ACCEPT; |
9330419d | 1554 | pp = pd->pp; |
2a3b791e | 1555 | |
63dca2c0 JDB |
1556 | /* Cannot handle fragmented embedded protocol */ |
1557 | if (ciph.fragoffs) | |
2a3b791e JV |
1558 | return NF_ACCEPT; |
1559 | ||
63dca2c0 | 1560 | IP_VS_DBG_PKT(11, AF_INET6, pp, skb, offs_ciph, |
0d79641a | 1561 | "Checking incoming ICMPv6 for"); |
2a3b791e | 1562 | |
2f74713d JDB |
1563 | /* The embedded headers contain source and dest in reverse order |
1564 | * if not from localhost | |
1565 | */ | |
d4383f04 | 1566 | cp = pp->conn_in_get(AF_INET6, skb, &ciph, |
2f74713d JDB |
1567 | (hooknum == NF_INET_LOCAL_OUT) ? 0 : 1); |
1568 | ||
6cb90db5 | 1569 | if (!cp) |
2a3b791e | 1570 | return NF_ACCEPT; |
2f74713d JDB |
1571 | /* VS/TUN, VS/DR and LOCALNODE just let it go */ |
1572 | if ((hooknum == NF_INET_LOCAL_OUT) && | |
1573 | (IP_VS_FWD_METHOD(cp) != IP_VS_CONN_F_MASQ)) { | |
1574 | __ip_vs_conn_put(cp); | |
1575 | return NF_ACCEPT; | |
1576 | } | |
2a3b791e | 1577 | |
2a3b791e JV |
1578 | /* do the statistics and put it back */ |
1579 | ip_vs_in_stats(cp, skb); | |
63dca2c0 JDB |
1580 | |
1581 | /* Need to mangle contained IPv6 header in ICMPv6 packet */ | |
1582 | writable = ciph.len; | |
1583 | if (IPPROTO_TCP == ciph.protocol || IPPROTO_UDP == ciph.protocol || | |
1584 | IPPROTO_SCTP == ciph.protocol) | |
1585 | writable += 2 * sizeof(__u16); /* Also mangle ports */ | |
1586 | ||
d4383f04 | 1587 | verdict = ip_vs_icmp_xmit_v6(skb, cp, pp, writable, hooknum, &ciph); |
2a3b791e JV |
1588 | |
1589 | __ip_vs_conn_put(cp); | |
1590 | ||
1591 | return verdict; | |
1592 | } | |
1593 | #endif | |
1594 | ||
1595 | ||
1da177e4 LT |
1596 | /* |
1597 | * Check if it's for virtual services, look it up, | |
1598 | * and send it on its way... | |
1599 | */ | |
1600 | static unsigned int | |
cb59155f | 1601 | ip_vs_in(unsigned int hooknum, struct sk_buff *skb, int af) |
1da177e4 | 1602 | { |
f131315f | 1603 | struct net *net; |
51ef348b | 1604 | struct ip_vs_iphdr iph; |
1da177e4 | 1605 | struct ip_vs_protocol *pp; |
9330419d | 1606 | struct ip_vs_proto_data *pd; |
1da177e4 | 1607 | struct ip_vs_conn *cp; |
4a516f11 | 1608 | int ret, pkts; |
f131315f | 1609 | struct netns_ipvs *ipvs; |
d752c364 | 1610 | int conn_reuse_mode; |
2a3b791e | 1611 | |
fc604767 JA |
1612 | /* Already marked as IPVS request or reply? */ |
1613 | if (skb->ipvs_property) | |
1614 | return NF_ACCEPT; | |
1615 | ||
1da177e4 | 1616 | /* |
cb59155f JA |
1617 | * Big tappo: |
1618 | * - remote client: only PACKET_HOST | |
1619 | * - route: used for struct net when skb->dev is unset | |
1da177e4 | 1620 | */ |
cb59155f JA |
1621 | if (unlikely((skb->pkt_type != PACKET_HOST && |
1622 | hooknum != NF_INET_LOCAL_OUT) || | |
1623 | !skb_dst(skb))) { | |
63dca2c0 | 1624 | ip_vs_fill_iph_skb(af, skb, &iph); |
cb59155f JA |
1625 | IP_VS_DBG_BUF(12, "packet type=%d proto=%d daddr=%s" |
1626 | " ignored in hook %u\n", | |
1627 | skb->pkt_type, iph.protocol, | |
1628 | IP_VS_DBG_ADDR(af, &iph.daddr), hooknum); | |
1da177e4 LT |
1629 | return NF_ACCEPT; |
1630 | } | |
7a4f0761 HS |
1631 | /* ipvs enabled in this netns ? */ |
1632 | net = skb_net(skb); | |
0c12582f JA |
1633 | ipvs = net_ipvs(net); |
1634 | if (unlikely(sysctl_backup_only(ipvs) || !ipvs->enable)) | |
7a4f0761 HS |
1635 | return NF_ACCEPT; |
1636 | ||
63dca2c0 | 1637 | ip_vs_fill_iph_skb(af, skb, &iph); |
cb59155f JA |
1638 | |
1639 | /* Bad... Do not break raw sockets */ | |
1640 | if (unlikely(skb->sk != NULL && hooknum == NF_INET_LOCAL_OUT && | |
1641 | af == AF_INET)) { | |
1642 | struct sock *sk = skb->sk; | |
1643 | struct inet_sock *inet = inet_sk(skb->sk); | |
1644 | ||
1645 | if (inet && sk->sk_family == PF_INET && inet->nodefrag) | |
1646 | return NF_ACCEPT; | |
1647 | } | |
1da177e4 | 1648 | |
94b26551 JV |
1649 | #ifdef CONFIG_IP_VS_IPV6 |
1650 | if (af == AF_INET6) { | |
1651 | if (unlikely(iph.protocol == IPPROTO_ICMPV6)) { | |
1ca5bb54 | 1652 | int related; |
d4383f04 JDB |
1653 | int verdict = ip_vs_in_icmp_v6(skb, &related, hooknum, |
1654 | &iph); | |
1da177e4 | 1655 | |
94b26551 JV |
1656 | if (related) |
1657 | return verdict; | |
94b26551 JV |
1658 | } |
1659 | } else | |
1660 | #endif | |
1661 | if (unlikely(iph.protocol == IPPROTO_ICMP)) { | |
1ca5bb54 JA |
1662 | int related; |
1663 | int verdict = ip_vs_in_icmp(skb, &related, hooknum); | |
94b26551 JV |
1664 | |
1665 | if (related) | |
1666 | return verdict; | |
94b26551 | 1667 | } |
1da177e4 LT |
1668 | |
1669 | /* Protocol supported? */ | |
9330419d HS |
1670 | pd = ip_vs_proto_data_get(net, iph.protocol); |
1671 | if (unlikely(!pd)) | |
1da177e4 | 1672 | return NF_ACCEPT; |
9330419d | 1673 | pp = pd->pp; |
1da177e4 LT |
1674 | /* |
1675 | * Check if the packet belongs to an existing connection entry | |
1676 | */ | |
d4383f04 | 1677 | cp = pp->conn_in_get(af, skb, &iph, 0); |
dc7b3eb9 | 1678 | |
d752c364 MRL |
1679 | conn_reuse_mode = sysctl_conn_reuse_mode(ipvs); |
1680 | if (conn_reuse_mode && !iph.fragoffs && | |
1681 | is_new_conn(skb, &iph) && cp && | |
1682 | ((unlikely(sysctl_expire_nodest_conn(ipvs)) && cp->dest && | |
1683 | unlikely(!atomic_read(&cp->dest->weight))) || | |
1684 | unlikely(is_new_conn_expected(cp, conn_reuse_mode)))) { | |
1685 | if (!atomic_read(&cp->n_control)) | |
1686 | ip_vs_conn_expire_now(cp); | |
dc7b3eb9 GL |
1687 | __ip_vs_conn_put(cp); |
1688 | cp = NULL; | |
1689 | } | |
1690 | ||
63dca2c0 | 1691 | if (unlikely(!cp) && !iph.fragoffs) { |
2f74713d JDB |
1692 | /* No (second) fragments need to enter here, as nf_defrag_ipv6 |
1693 | * replayed fragment zero will already have created the cp | |
1694 | */ | |
1da177e4 LT |
1695 | int v; |
1696 | ||
2f74713d | 1697 | /* Schedule and create new connection entry into &cp */ |
d4383f04 | 1698 | if (!pp->conn_schedule(af, skb, pd, &v, &cp, &iph)) |
1da177e4 LT |
1699 | return v; |
1700 | } | |
1701 | ||
1702 | if (unlikely(!cp)) { | |
1703 | /* sorry, all this trouble for a no-hit :) */ | |
0d79641a | 1704 | IP_VS_DBG_PKT(12, af, pp, skb, 0, |
cb59155f | 1705 | "ip_vs_in: packet continues traversal as normal"); |
6aafeef0 | 1706 | if (iph.fragoffs) { |
2f74713d | 1707 | /* Fragment that couldn't be mapped to a conn entry |
2f74713d JDB |
1708 | * is missing module nf_defrag_ipv6 |
1709 | */ | |
1710 | IP_VS_DBG_RL("Unhandled frag, load nf_defrag_ipv6\n"); | |
1711 | IP_VS_DBG_PKT(7, af, pp, skb, 0, "unhandled fragment"); | |
1712 | } | |
1da177e4 LT |
1713 | return NF_ACCEPT; |
1714 | } | |
1715 | ||
0d79641a | 1716 | IP_VS_DBG_PKT(11, af, pp, skb, 0, "Incoming packet"); |
1da177e4 LT |
1717 | /* Check the server status */ |
1718 | if (cp->dest && !(cp->dest->flags & IP_VS_DEST_F_AVAILABLE)) { | |
1719 | /* the destination server is not available */ | |
1720 | ||
71a8ab6c | 1721 | if (sysctl_expire_nodest_conn(ipvs)) { |
1da177e4 LT |
1722 | /* try to expire the connection immediately */ |
1723 | ip_vs_conn_expire_now(cp); | |
1da177e4 | 1724 | } |
dc8103f2 JA |
1725 | /* don't restart its timer, and silently |
1726 | drop the packet. */ | |
1727 | __ip_vs_conn_put(cp); | |
1da177e4 LT |
1728 | return NF_DROP; |
1729 | } | |
1730 | ||
1731 | ip_vs_in_stats(cp, skb); | |
4a516f11 | 1732 | ip_vs_set_state(cp, IP_VS_DIR_INPUT, skb, pd); |
1da177e4 | 1733 | if (cp->packet_xmit) |
d4383f04 | 1734 | ret = cp->packet_xmit(skb, cp, pp, &iph); |
1da177e4 LT |
1735 | /* do not touch skb anymore */ |
1736 | else { | |
1737 | IP_VS_DBG_RL("warning: packet_xmit is null"); | |
1738 | ret = NF_ACCEPT; | |
1739 | } | |
1740 | ||
efac5276 RB |
1741 | /* Increase its packet counter and check if it is needed |
1742 | * to be synchronized | |
1743 | * | |
1744 | * Sync connection if it is about to close to | |
1745 | * encorage the standby servers to update the connections timeout | |
986a0757 HS |
1746 | * |
1747 | * For ONE_PKT let ip_vs_sync_conn() do the filter work. | |
efac5276 | 1748 | */ |
f131315f | 1749 | |
986a0757 | 1750 | if (cp->flags & IP_VS_CONN_F_ONE_PACKET) |
59e0350e | 1751 | pkts = sysctl_sync_threshold(ipvs); |
986a0757 HS |
1752 | else |
1753 | pkts = atomic_add_return(1, &cp->in_pkts); | |
1754 | ||
749c42b6 JA |
1755 | if (ipvs->sync_state & IP_VS_STATE_MASTER) |
1756 | ip_vs_sync_conn(net, cp, pkts); | |
1da177e4 LT |
1757 | |
1758 | ip_vs_conn_put(cp); | |
1759 | return ret; | |
1760 | } | |
1761 | ||
cb59155f JA |
1762 | /* |
1763 | * AF_INET handler in NF_INET_LOCAL_IN chain | |
1764 | * Schedule and forward packets from remote clients | |
1765 | */ | |
1766 | static unsigned int | |
795aa6ef | 1767 | ip_vs_remote_request4(const struct nf_hook_ops *ops, struct sk_buff *skb, |
238e54c9 | 1768 | const struct nf_hook_state *state) |
cb59155f | 1769 | { |
795aa6ef | 1770 | return ip_vs_in(ops->hooknum, skb, AF_INET); |
cb59155f JA |
1771 | } |
1772 | ||
1773 | /* | |
1774 | * AF_INET handler in NF_INET_LOCAL_OUT chain | |
1775 | * Schedule and forward packets from local clients | |
1776 | */ | |
1777 | static unsigned int | |
795aa6ef | 1778 | ip_vs_local_request4(const struct nf_hook_ops *ops, struct sk_buff *skb, |
238e54c9 | 1779 | const struct nf_hook_state *state) |
cb59155f | 1780 | { |
795aa6ef | 1781 | return ip_vs_in(ops->hooknum, skb, AF_INET); |
cb59155f JA |
1782 | } |
1783 | ||
1784 | #ifdef CONFIG_IP_VS_IPV6 | |
1785 | ||
1786 | /* | |
1787 | * AF_INET6 handler in NF_INET_LOCAL_IN chain | |
1788 | * Schedule and forward packets from remote clients | |
1789 | */ | |
1790 | static unsigned int | |
795aa6ef | 1791 | ip_vs_remote_request6(const struct nf_hook_ops *ops, struct sk_buff *skb, |
238e54c9 | 1792 | const struct nf_hook_state *state) |
cb59155f | 1793 | { |
795aa6ef | 1794 | return ip_vs_in(ops->hooknum, skb, AF_INET6); |
cb59155f JA |
1795 | } |
1796 | ||
1797 | /* | |
1798 | * AF_INET6 handler in NF_INET_LOCAL_OUT chain | |
1799 | * Schedule and forward packets from local clients | |
1800 | */ | |
1801 | static unsigned int | |
795aa6ef | 1802 | ip_vs_local_request6(const struct nf_hook_ops *ops, struct sk_buff *skb, |
238e54c9 | 1803 | const struct nf_hook_state *state) |
cb59155f | 1804 | { |
795aa6ef | 1805 | return ip_vs_in(ops->hooknum, skb, AF_INET6); |
cb59155f JA |
1806 | } |
1807 | ||
1808 | #endif | |
1809 | ||
1da177e4 LT |
1810 | |
1811 | /* | |
6e23ae2a | 1812 | * It is hooked at the NF_INET_FORWARD chain, in order to catch ICMP |
1da177e4 LT |
1813 | * related packets destined for 0.0.0.0/0. |
1814 | * When fwmark-based virtual service is used, such as transparent | |
1815 | * cache cluster, TCP packets can be marked and routed to ip_vs_in, | |
1816 | * but ICMP destined for 0.0.0.0/0 cannot not be easily marked and | |
6e23ae2a | 1817 | * sent to ip_vs_in_icmp. So, catch them at the NF_INET_FORWARD chain |
1da177e4 LT |
1818 | * and send them to ip_vs_in_icmp. |
1819 | */ | |
1820 | static unsigned int | |
795aa6ef | 1821 | ip_vs_forward_icmp(const struct nf_hook_ops *ops, struct sk_buff *skb, |
238e54c9 | 1822 | const struct nf_hook_state *state) |
1da177e4 LT |
1823 | { |
1824 | int r; | |
7a4f0761 | 1825 | struct net *net; |
0c12582f | 1826 | struct netns_ipvs *ipvs; |
1da177e4 | 1827 | |
3db05fea | 1828 | if (ip_hdr(skb)->protocol != IPPROTO_ICMP) |
1da177e4 LT |
1829 | return NF_ACCEPT; |
1830 | ||
7a4f0761 HS |
1831 | /* ipvs enabled in this netns ? */ |
1832 | net = skb_net(skb); | |
0c12582f JA |
1833 | ipvs = net_ipvs(net); |
1834 | if (unlikely(sysctl_backup_only(ipvs) || !ipvs->enable)) | |
7a4f0761 HS |
1835 | return NF_ACCEPT; |
1836 | ||
795aa6ef | 1837 | return ip_vs_in_icmp(skb, &r, ops->hooknum); |
1da177e4 LT |
1838 | } |
1839 | ||
2a3b791e JV |
1840 | #ifdef CONFIG_IP_VS_IPV6 |
1841 | static unsigned int | |
795aa6ef | 1842 | ip_vs_forward_icmp_v6(const struct nf_hook_ops *ops, struct sk_buff *skb, |
238e54c9 | 1843 | const struct nf_hook_state *state) |
2a3b791e JV |
1844 | { |
1845 | int r; | |
7a4f0761 | 1846 | struct net *net; |
0c12582f | 1847 | struct netns_ipvs *ipvs; |
63dca2c0 | 1848 | struct ip_vs_iphdr iphdr; |
2a3b791e | 1849 | |
63dca2c0 JDB |
1850 | ip_vs_fill_iph_skb(AF_INET6, skb, &iphdr); |
1851 | if (iphdr.protocol != IPPROTO_ICMPV6) | |
2a3b791e JV |
1852 | return NF_ACCEPT; |
1853 | ||
7a4f0761 HS |
1854 | /* ipvs enabled in this netns ? */ |
1855 | net = skb_net(skb); | |
0c12582f JA |
1856 | ipvs = net_ipvs(net); |
1857 | if (unlikely(sysctl_backup_only(ipvs) || !ipvs->enable)) | |
7a4f0761 HS |
1858 | return NF_ACCEPT; |
1859 | ||
795aa6ef | 1860 | return ip_vs_in_icmp_v6(skb, &r, ops->hooknum, &iphdr); |
2a3b791e JV |
1861 | } |
1862 | #endif | |
1863 | ||
1da177e4 | 1864 | |
1999414a | 1865 | static struct nf_hook_ops ip_vs_ops[] __read_mostly = { |
cb59155f JA |
1866 | /* After packet filtering, change source only for VS/NAT */ |
1867 | { | |
1868 | .hook = ip_vs_reply4, | |
1869 | .owner = THIS_MODULE, | |
4c809d63 | 1870 | .pf = NFPROTO_IPV4, |
cb59155f | 1871 | .hooknum = NF_INET_LOCAL_IN, |
afb523c5 | 1872 | .priority = NF_IP_PRI_NAT_SRC - 2, |
cb59155f | 1873 | }, |
41c5b317 PM |
1874 | /* After packet filtering, forward packet through VS/DR, VS/TUN, |
1875 | * or VS/NAT(change destination), so that filtering rules can be | |
1876 | * applied to IPVS. */ | |
1877 | { | |
cb59155f | 1878 | .hook = ip_vs_remote_request4, |
41c5b317 | 1879 | .owner = THIS_MODULE, |
4c809d63 | 1880 | .pf = NFPROTO_IPV4, |
cb59155f | 1881 | .hooknum = NF_INET_LOCAL_IN, |
afb523c5 | 1882 | .priority = NF_IP_PRI_NAT_SRC - 1, |
41c5b317 | 1883 | }, |
fc604767 | 1884 | /* Before ip_vs_in, change source only for VS/NAT */ |
41c5b317 | 1885 | { |
fc604767 | 1886 | .hook = ip_vs_local_reply4, |
41c5b317 | 1887 | .owner = THIS_MODULE, |
4c809d63 | 1888 | .pf = NFPROTO_IPV4, |
fc604767 | 1889 | .hooknum = NF_INET_LOCAL_OUT, |
afb523c5 | 1890 | .priority = NF_IP_PRI_NAT_DST + 1, |
41c5b317 | 1891 | }, |
cb59155f JA |
1892 | /* After mangle, schedule and forward local requests */ |
1893 | { | |
1894 | .hook = ip_vs_local_request4, | |
1895 | .owner = THIS_MODULE, | |
4c809d63 | 1896 | .pf = NFPROTO_IPV4, |
cb59155f | 1897 | .hooknum = NF_INET_LOCAL_OUT, |
afb523c5 | 1898 | .priority = NF_IP_PRI_NAT_DST + 2, |
cb59155f | 1899 | }, |
41c5b317 PM |
1900 | /* After packet filtering (but before ip_vs_out_icmp), catch icmp |
1901 | * destined for 0.0.0.0/0, which is for incoming IPVS connections */ | |
1902 | { | |
1903 | .hook = ip_vs_forward_icmp, | |
1904 | .owner = THIS_MODULE, | |
4c809d63 | 1905 | .pf = NFPROTO_IPV4, |
cb59155f JA |
1906 | .hooknum = NF_INET_FORWARD, |
1907 | .priority = 99, | |
41c5b317 | 1908 | }, |
fc604767 JA |
1909 | /* After packet filtering, change source only for VS/NAT */ |
1910 | { | |
1911 | .hook = ip_vs_reply4, | |
1912 | .owner = THIS_MODULE, | |
4c809d63 | 1913 | .pf = NFPROTO_IPV4, |
fc604767 JA |
1914 | .hooknum = NF_INET_FORWARD, |
1915 | .priority = 100, | |
1916 | }, | |
473b23d3 | 1917 | #ifdef CONFIG_IP_VS_IPV6 |
cb59155f JA |
1918 | /* After packet filtering, change source only for VS/NAT */ |
1919 | { | |
1920 | .hook = ip_vs_reply6, | |
1921 | .owner = THIS_MODULE, | |
4c809d63 | 1922 | .pf = NFPROTO_IPV6, |
cb59155f | 1923 | .hooknum = NF_INET_LOCAL_IN, |
afb523c5 | 1924 | .priority = NF_IP6_PRI_NAT_SRC - 2, |
cb59155f | 1925 | }, |
473b23d3 JV |
1926 | /* After packet filtering, forward packet through VS/DR, VS/TUN, |
1927 | * or VS/NAT(change destination), so that filtering rules can be | |
1928 | * applied to IPVS. */ | |
1929 | { | |
cb59155f | 1930 | .hook = ip_vs_remote_request6, |
473b23d3 | 1931 | .owner = THIS_MODULE, |
4c809d63 | 1932 | .pf = NFPROTO_IPV6, |
cb59155f | 1933 | .hooknum = NF_INET_LOCAL_IN, |
afb523c5 | 1934 | .priority = NF_IP6_PRI_NAT_SRC - 1, |
473b23d3 | 1935 | }, |
fc604767 | 1936 | /* Before ip_vs_in, change source only for VS/NAT */ |
473b23d3 | 1937 | { |
fc604767 | 1938 | .hook = ip_vs_local_reply6, |
473b23d3 | 1939 | .owner = THIS_MODULE, |
eb90b0c7 | 1940 | .pf = NFPROTO_IPV6, |
fc604767 | 1941 | .hooknum = NF_INET_LOCAL_OUT, |
afb523c5 | 1942 | .priority = NF_IP6_PRI_NAT_DST + 1, |
473b23d3 | 1943 | }, |
cb59155f JA |
1944 | /* After mangle, schedule and forward local requests */ |
1945 | { | |
1946 | .hook = ip_vs_local_request6, | |
1947 | .owner = THIS_MODULE, | |
4c809d63 | 1948 | .pf = NFPROTO_IPV6, |
cb59155f | 1949 | .hooknum = NF_INET_LOCAL_OUT, |
afb523c5 | 1950 | .priority = NF_IP6_PRI_NAT_DST + 2, |
cb59155f | 1951 | }, |
473b23d3 JV |
1952 | /* After packet filtering (but before ip_vs_out_icmp), catch icmp |
1953 | * destined for 0.0.0.0/0, which is for incoming IPVS connections */ | |
1954 | { | |
1955 | .hook = ip_vs_forward_icmp_v6, | |
1956 | .owner = THIS_MODULE, | |
4c809d63 | 1957 | .pf = NFPROTO_IPV6, |
cb59155f JA |
1958 | .hooknum = NF_INET_FORWARD, |
1959 | .priority = 99, | |
473b23d3 | 1960 | }, |
fc604767 JA |
1961 | /* After packet filtering, change source only for VS/NAT */ |
1962 | { | |
1963 | .hook = ip_vs_reply6, | |
1964 | .owner = THIS_MODULE, | |
4c809d63 | 1965 | .pf = NFPROTO_IPV6, |
fc604767 JA |
1966 | .hooknum = NF_INET_FORWARD, |
1967 | .priority = 100, | |
1968 | }, | |
473b23d3 | 1969 | #endif |
1da177e4 | 1970 | }; |
61b1ab45 HS |
1971 | /* |
1972 | * Initialize IP Virtual Server netns mem. | |
1973 | */ | |
1974 | static int __net_init __ip_vs_init(struct net *net) | |
1975 | { | |
1976 | struct netns_ipvs *ipvs; | |
1977 | ||
61b1ab45 | 1978 | ipvs = net_generic(net, ip_vs_net_id); |
0a9ee813 | 1979 | if (ipvs == NULL) |
61b1ab45 | 1980 | return -ENOMEM; |
0a9ee813 | 1981 | |
7a4f0761 HS |
1982 | /* Hold the beast until a service is registerd */ |
1983 | ipvs->enable = 0; | |
f6340ee0 | 1984 | ipvs->net = net; |
61b1ab45 HS |
1985 | /* Counters used for creating unique names */ |
1986 | ipvs->gen = atomic_read(&ipvs_netns_cnt); | |
1987 | atomic_inc(&ipvs_netns_cnt); | |
1988 | net->ipvs = ipvs; | |
7a4f0761 | 1989 | |
503cf15a | 1990 | if (ip_vs_estimator_net_init(net) < 0) |
7a4f0761 HS |
1991 | goto estimator_fail; |
1992 | ||
503cf15a | 1993 | if (ip_vs_control_net_init(net) < 0) |
7a4f0761 HS |
1994 | goto control_fail; |
1995 | ||
503cf15a | 1996 | if (ip_vs_protocol_net_init(net) < 0) |
7a4f0761 HS |
1997 | goto protocol_fail; |
1998 | ||
503cf15a | 1999 | if (ip_vs_app_net_init(net) < 0) |
7a4f0761 HS |
2000 | goto app_fail; |
2001 | ||
503cf15a | 2002 | if (ip_vs_conn_net_init(net) < 0) |
7a4f0761 HS |
2003 | goto conn_fail; |
2004 | ||
503cf15a | 2005 | if (ip_vs_sync_net_init(net) < 0) |
7a4f0761 HS |
2006 | goto sync_fail; |
2007 | ||
a870c8c5 | 2008 | printk(KERN_INFO "IPVS: Creating netns size=%zu id=%d\n", |
61b1ab45 HS |
2009 | sizeof(struct netns_ipvs), ipvs->gen); |
2010 | return 0; | |
7a4f0761 HS |
2011 | /* |
2012 | * Error handling | |
2013 | */ | |
2014 | ||
2015 | sync_fail: | |
503cf15a | 2016 | ip_vs_conn_net_cleanup(net); |
7a4f0761 | 2017 | conn_fail: |
503cf15a | 2018 | ip_vs_app_net_cleanup(net); |
7a4f0761 | 2019 | app_fail: |
503cf15a | 2020 | ip_vs_protocol_net_cleanup(net); |
7a4f0761 | 2021 | protocol_fail: |
503cf15a | 2022 | ip_vs_control_net_cleanup(net); |
7a4f0761 | 2023 | control_fail: |
503cf15a | 2024 | ip_vs_estimator_net_cleanup(net); |
7a4f0761 | 2025 | estimator_fail: |
39f618b4 | 2026 | net->ipvs = NULL; |
7a4f0761 | 2027 | return -ENOMEM; |
61b1ab45 HS |
2028 | } |
2029 | ||
2030 | static void __net_exit __ip_vs_cleanup(struct net *net) | |
2031 | { | |
503cf15a HS |
2032 | ip_vs_service_net_cleanup(net); /* ip_vs_flush() with locks */ |
2033 | ip_vs_conn_net_cleanup(net); | |
2034 | ip_vs_app_net_cleanup(net); | |
2035 | ip_vs_protocol_net_cleanup(net); | |
2036 | ip_vs_control_net_cleanup(net); | |
2037 | ip_vs_estimator_net_cleanup(net); | |
1ae132b0 | 2038 | IP_VS_DBG(2, "ipvs netns %d released\n", net_ipvs(net)->gen); |
39f618b4 | 2039 | net->ipvs = NULL; |
61b1ab45 HS |
2040 | } |
2041 | ||
7a4f0761 HS |
2042 | static void __net_exit __ip_vs_dev_cleanup(struct net *net) |
2043 | { | |
2044 | EnterFunction(2); | |
2045 | net_ipvs(net)->enable = 0; /* Disable packet reception */ | |
8f4e0a18 | 2046 | smp_wmb(); |
503cf15a | 2047 | ip_vs_sync_net_cleanup(net); |
7a4f0761 HS |
2048 | LeaveFunction(2); |
2049 | } | |
2050 | ||
61b1ab45 HS |
2051 | static struct pernet_operations ipvs_core_ops = { |
2052 | .init = __ip_vs_init, | |
2053 | .exit = __ip_vs_cleanup, | |
2054 | .id = &ip_vs_net_id, | |
2055 | .size = sizeof(struct netns_ipvs), | |
2056 | }; | |
1da177e4 | 2057 | |
7a4f0761 HS |
2058 | static struct pernet_operations ipvs_core_dev_ops = { |
2059 | .exit = __ip_vs_dev_cleanup, | |
2060 | }; | |
2061 | ||
1da177e4 LT |
2062 | /* |
2063 | * Initialize IP Virtual Server | |
2064 | */ | |
2065 | static int __init ip_vs_init(void) | |
2066 | { | |
2067 | int ret; | |
2068 | ||
2069 | ret = ip_vs_control_init(); | |
2070 | if (ret < 0) { | |
1e3e238e | 2071 | pr_err("can't setup control.\n"); |
6c8f7949 | 2072 | goto exit; |
1da177e4 LT |
2073 | } |
2074 | ||
2075 | ip_vs_protocol_init(); | |
2076 | ||
1da177e4 LT |
2077 | ret = ip_vs_conn_init(); |
2078 | if (ret < 0) { | |
1e3e238e | 2079 | pr_err("can't setup connection table.\n"); |
6c8f7949 | 2080 | goto cleanup_protocol; |
61b1ab45 HS |
2081 | } |
2082 | ||
7a4f0761 HS |
2083 | ret = register_pernet_subsys(&ipvs_core_ops); /* Alloc ip_vs struct */ |
2084 | if (ret < 0) | |
6c8f7949 | 2085 | goto cleanup_conn; |
7a4f0761 HS |
2086 | |
2087 | ret = register_pernet_device(&ipvs_core_dev_ops); | |
2088 | if (ret < 0) | |
2089 | goto cleanup_sub; | |
2090 | ||
41c5b317 | 2091 | ret = nf_register_hooks(ip_vs_ops, ARRAY_SIZE(ip_vs_ops)); |
1da177e4 | 2092 | if (ret < 0) { |
1e3e238e | 2093 | pr_err("can't register hooks.\n"); |
7a4f0761 | 2094 | goto cleanup_dev; |
1da177e4 LT |
2095 | } |
2096 | ||
8537de8a HS |
2097 | ret = ip_vs_register_nl_ioctl(); |
2098 | if (ret < 0) { | |
2099 | pr_err("can't register netlink/ioctl.\n"); | |
2100 | goto cleanup_hooks; | |
2101 | } | |
2102 | ||
1e3e238e | 2103 | pr_info("ipvs loaded.\n"); |
7a4f0761 | 2104 | |
1da177e4 LT |
2105 | return ret; |
2106 | ||
8537de8a HS |
2107 | cleanup_hooks: |
2108 | nf_unregister_hooks(ip_vs_ops, ARRAY_SIZE(ip_vs_ops)); | |
7a4f0761 HS |
2109 | cleanup_dev: |
2110 | unregister_pernet_device(&ipvs_core_dev_ops); | |
2111 | cleanup_sub: | |
2112 | unregister_pernet_subsys(&ipvs_core_ops); | |
552ad65a | 2113 | cleanup_conn: |
1da177e4 | 2114 | ip_vs_conn_cleanup(); |
552ad65a | 2115 | cleanup_protocol: |
1da177e4 LT |
2116 | ip_vs_protocol_cleanup(); |
2117 | ip_vs_control_cleanup(); | |
6c8f7949 | 2118 | exit: |
1da177e4 LT |
2119 | return ret; |
2120 | } | |
2121 | ||
2122 | static void __exit ip_vs_cleanup(void) | |
2123 | { | |
8537de8a | 2124 | ip_vs_unregister_nl_ioctl(); |
41c5b317 | 2125 | nf_unregister_hooks(ip_vs_ops, ARRAY_SIZE(ip_vs_ops)); |
7a4f0761 HS |
2126 | unregister_pernet_device(&ipvs_core_dev_ops); |
2127 | unregister_pernet_subsys(&ipvs_core_ops); /* free ip_vs struct */ | |
1da177e4 | 2128 | ip_vs_conn_cleanup(); |
1da177e4 LT |
2129 | ip_vs_protocol_cleanup(); |
2130 | ip_vs_control_cleanup(); | |
1e3e238e | 2131 | pr_info("ipvs unloaded.\n"); |
1da177e4 LT |
2132 | } |
2133 | ||
2134 | module_init(ip_vs_init); | |
2135 | module_exit(ip_vs_cleanup); | |
2136 | MODULE_LICENSE("GPL"); |