Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-3.6
[deliverable/linux.git] / net / netfilter / ipvs / ip_vs_proto.c
1 /*
2 * ip_vs_proto.c: transport protocol load balancing support for IPVS
3 *
4 * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
5 * Julian Anastasov <ja@ssi.bg>
6 *
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License
9 * as published by the Free Software Foundation; either version
10 * 2 of the License, or (at your option) any later version.
11 *
12 * Changes:
13 *
14 */
15
16 #define KMSG_COMPONENT "IPVS"
17 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
18
19 #include <linux/module.h>
20 #include <linux/kernel.h>
21 #include <linux/skbuff.h>
22 #include <linux/gfp.h>
23 #include <linux/in.h>
24 #include <linux/ip.h>
25 #include <net/protocol.h>
26 #include <net/tcp.h>
27 #include <net/udp.h>
28 #include <asm/system.h>
29 #include <linux/stat.h>
30 #include <linux/proc_fs.h>
31
32 #include <net/ip_vs.h>
33
34
35 /*
36 * IPVS protocols can only be registered/unregistered when the ipvs
37 * module is loaded/unloaded, so no lock is needed in accessing the
38 * ipvs protocol table.
39 */
40
41 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */
42 #define IP_VS_PROTO_HASH(proto) ((proto) & (IP_VS_PROTO_TAB_SIZE-1))
43
44 static struct ip_vs_protocol *ip_vs_proto_table[IP_VS_PROTO_TAB_SIZE];
45
46
47 /*
48 * register an ipvs protocol
49 */
50 static int __used __init register_ip_vs_protocol(struct ip_vs_protocol *pp)
51 {
52 unsigned hash = IP_VS_PROTO_HASH(pp->protocol);
53
54 pp->next = ip_vs_proto_table[hash];
55 ip_vs_proto_table[hash] = pp;
56
57 if (pp->init != NULL)
58 pp->init(pp);
59
60 return 0;
61 }
62
63 #if defined(CONFIG_IP_VS_PROTO_TCP) || defined(CONFIG_IP_VS_PROTO_UDP) || \
64 defined(CONFIG_IP_VS_PROTO_SCTP) || defined(CONFIG_IP_VS_PROTO_AH) || \
65 defined(CONFIG_IP_VS_PROTO_ESP)
66 /*
67 * register an ipvs protocols netns related data
68 */
69 static int
70 register_ip_vs_proto_netns(struct net *net, struct ip_vs_protocol *pp)
71 {
72 struct netns_ipvs *ipvs = net_ipvs(net);
73 unsigned hash = IP_VS_PROTO_HASH(pp->protocol);
74 struct ip_vs_proto_data *pd =
75 kzalloc(sizeof(struct ip_vs_proto_data), GFP_ATOMIC);
76
77 if (!pd) {
78 pr_err("%s(): no memory.\n", __func__);
79 return -ENOMEM;
80 }
81 pd->pp = pp; /* For speed issues */
82 pd->next = ipvs->proto_data_table[hash];
83 ipvs->proto_data_table[hash] = pd;
84 atomic_set(&pd->appcnt, 0); /* Init app counter */
85
86 if (pp->init_netns != NULL)
87 pp->init_netns(net, pd);
88
89 return 0;
90 }
91 #endif
92
93 /*
94 * unregister an ipvs protocol
95 */
96 static int unregister_ip_vs_protocol(struct ip_vs_protocol *pp)
97 {
98 struct ip_vs_protocol **pp_p;
99 unsigned hash = IP_VS_PROTO_HASH(pp->protocol);
100
101 pp_p = &ip_vs_proto_table[hash];
102 for (; *pp_p; pp_p = &(*pp_p)->next) {
103 if (*pp_p == pp) {
104 *pp_p = pp->next;
105 if (pp->exit != NULL)
106 pp->exit(pp);
107 return 0;
108 }
109 }
110
111 return -ESRCH;
112 }
113
114 /*
115 * unregister an ipvs protocols netns data
116 */
117 static int
118 unregister_ip_vs_proto_netns(struct net *net, struct ip_vs_proto_data *pd)
119 {
120 struct netns_ipvs *ipvs = net_ipvs(net);
121 struct ip_vs_proto_data **pd_p;
122 unsigned hash = IP_VS_PROTO_HASH(pd->pp->protocol);
123
124 pd_p = &ipvs->proto_data_table[hash];
125 for (; *pd_p; pd_p = &(*pd_p)->next) {
126 if (*pd_p == pd) {
127 *pd_p = pd->next;
128 if (pd->pp->exit_netns != NULL)
129 pd->pp->exit_netns(net, pd);
130 kfree(pd);
131 return 0;
132 }
133 }
134
135 return -ESRCH;
136 }
137
138 /*
139 * get ip_vs_protocol object by its proto.
140 */
141 struct ip_vs_protocol * ip_vs_proto_get(unsigned short proto)
142 {
143 struct ip_vs_protocol *pp;
144 unsigned hash = IP_VS_PROTO_HASH(proto);
145
146 for (pp = ip_vs_proto_table[hash]; pp; pp = pp->next) {
147 if (pp->protocol == proto)
148 return pp;
149 }
150
151 return NULL;
152 }
153 EXPORT_SYMBOL(ip_vs_proto_get);
154
155 /*
156 * get ip_vs_protocol object data by netns and proto
157 */
158 struct ip_vs_proto_data *
159 __ipvs_proto_data_get(struct netns_ipvs *ipvs, unsigned short proto)
160 {
161 struct ip_vs_proto_data *pd;
162 unsigned hash = IP_VS_PROTO_HASH(proto);
163
164 for (pd = ipvs->proto_data_table[hash]; pd; pd = pd->next) {
165 if (pd->pp->protocol == proto)
166 return pd;
167 }
168
169 return NULL;
170 }
171
172 struct ip_vs_proto_data *
173 ip_vs_proto_data_get(struct net *net, unsigned short proto)
174 {
175 struct netns_ipvs *ipvs = net_ipvs(net);
176
177 return __ipvs_proto_data_get(ipvs, proto);
178 }
179 EXPORT_SYMBOL(ip_vs_proto_data_get);
180
181 /*
182 * Propagate event for state change to all protocols
183 */
184 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags)
185 {
186 struct ip_vs_proto_data *pd;
187 int i;
188
189 for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
190 for (pd = ipvs->proto_data_table[i]; pd; pd = pd->next) {
191 if (pd->pp->timeout_change)
192 pd->pp->timeout_change(pd, flags);
193 }
194 }
195 }
196
197
198 int *
199 ip_vs_create_timeout_table(int *table, int size)
200 {
201 return kmemdup(table, size, GFP_ATOMIC);
202 }
203
204
205 /*
206 * Set timeout value for state specified by name
207 */
208 int
209 ip_vs_set_state_timeout(int *table, int num, const char *const *names,
210 const char *name, int to)
211 {
212 int i;
213
214 if (!table || !name || !to)
215 return -EINVAL;
216
217 for (i = 0; i < num; i++) {
218 if (strcmp(names[i], name))
219 continue;
220 table[i] = to * HZ;
221 return 0;
222 }
223 return -ENOENT;
224 }
225
226
227 const char * ip_vs_state_name(__u16 proto, int state)
228 {
229 struct ip_vs_protocol *pp = ip_vs_proto_get(proto);
230
231 if (pp == NULL || pp->state_name == NULL)
232 return (IPPROTO_IP == proto) ? "NONE" : "ERR!";
233 return pp->state_name(state);
234 }
235
236
237 static void
238 ip_vs_tcpudp_debug_packet_v4(struct ip_vs_protocol *pp,
239 const struct sk_buff *skb,
240 int offset,
241 const char *msg)
242 {
243 char buf[128];
244 struct iphdr _iph, *ih;
245
246 ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph);
247 if (ih == NULL)
248 sprintf(buf, "TRUNCATED");
249 else if (ih->frag_off & htons(IP_OFFSET))
250 sprintf(buf, "%pI4->%pI4 frag", &ih->saddr, &ih->daddr);
251 else {
252 __be16 _ports[2], *pptr;
253
254 pptr = skb_header_pointer(skb, offset + ih->ihl*4,
255 sizeof(_ports), _ports);
256 if (pptr == NULL)
257 sprintf(buf, "TRUNCATED %pI4->%pI4",
258 &ih->saddr, &ih->daddr);
259 else
260 sprintf(buf, "%pI4:%u->%pI4:%u",
261 &ih->saddr, ntohs(pptr[0]),
262 &ih->daddr, ntohs(pptr[1]));
263 }
264
265 pr_debug("%s: %s %s\n", msg, pp->name, buf);
266 }
267
268 #ifdef CONFIG_IP_VS_IPV6
269 static void
270 ip_vs_tcpudp_debug_packet_v6(struct ip_vs_protocol *pp,
271 const struct sk_buff *skb,
272 int offset,
273 const char *msg)
274 {
275 char buf[192];
276 struct ipv6hdr _iph, *ih;
277
278 ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph);
279 if (ih == NULL)
280 sprintf(buf, "TRUNCATED");
281 else if (ih->nexthdr == IPPROTO_FRAGMENT)
282 sprintf(buf, "%pI6->%pI6 frag", &ih->saddr, &ih->daddr);
283 else {
284 __be16 _ports[2], *pptr;
285
286 pptr = skb_header_pointer(skb, offset + sizeof(struct ipv6hdr),
287 sizeof(_ports), _ports);
288 if (pptr == NULL)
289 sprintf(buf, "TRUNCATED %pI6->%pI6",
290 &ih->saddr, &ih->daddr);
291 else
292 sprintf(buf, "%pI6:%u->%pI6:%u",
293 &ih->saddr, ntohs(pptr[0]),
294 &ih->daddr, ntohs(pptr[1]));
295 }
296
297 pr_debug("%s: %s %s\n", msg, pp->name, buf);
298 }
299 #endif
300
301
302 void
303 ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
304 const struct sk_buff *skb,
305 int offset,
306 const char *msg)
307 {
308 #ifdef CONFIG_IP_VS_IPV6
309 if (af == AF_INET6)
310 ip_vs_tcpudp_debug_packet_v6(pp, skb, offset, msg);
311 else
312 #endif
313 ip_vs_tcpudp_debug_packet_v4(pp, skb, offset, msg);
314 }
315
316 /*
317 * per network name-space init
318 */
319 int __net_init __ip_vs_protocol_init(struct net *net)
320 {
321 #ifdef CONFIG_IP_VS_PROTO_TCP
322 register_ip_vs_proto_netns(net, &ip_vs_protocol_tcp);
323 #endif
324 #ifdef CONFIG_IP_VS_PROTO_UDP
325 register_ip_vs_proto_netns(net, &ip_vs_protocol_udp);
326 #endif
327 #ifdef CONFIG_IP_VS_PROTO_SCTP
328 register_ip_vs_proto_netns(net, &ip_vs_protocol_sctp);
329 #endif
330 #ifdef CONFIG_IP_VS_PROTO_AH
331 register_ip_vs_proto_netns(net, &ip_vs_protocol_ah);
332 #endif
333 #ifdef CONFIG_IP_VS_PROTO_ESP
334 register_ip_vs_proto_netns(net, &ip_vs_protocol_esp);
335 #endif
336 return 0;
337 }
338
339 void __net_exit __ip_vs_protocol_cleanup(struct net *net)
340 {
341 struct netns_ipvs *ipvs = net_ipvs(net);
342 struct ip_vs_proto_data *pd;
343 int i;
344
345 /* unregister all the ipvs proto data for this netns */
346 for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
347 while ((pd = ipvs->proto_data_table[i]) != NULL)
348 unregister_ip_vs_proto_netns(net, pd);
349 }
350 }
351
352 int __init ip_vs_protocol_init(void)
353 {
354 char protocols[64];
355 #define REGISTER_PROTOCOL(p) \
356 do { \
357 register_ip_vs_protocol(p); \
358 strcat(protocols, ", "); \
359 strcat(protocols, (p)->name); \
360 } while (0)
361
362 protocols[0] = '\0';
363 protocols[2] = '\0';
364 #ifdef CONFIG_IP_VS_PROTO_TCP
365 REGISTER_PROTOCOL(&ip_vs_protocol_tcp);
366 #endif
367 #ifdef CONFIG_IP_VS_PROTO_UDP
368 REGISTER_PROTOCOL(&ip_vs_protocol_udp);
369 #endif
370 #ifdef CONFIG_IP_VS_PROTO_SCTP
371 REGISTER_PROTOCOL(&ip_vs_protocol_sctp);
372 #endif
373 #ifdef CONFIG_IP_VS_PROTO_AH
374 REGISTER_PROTOCOL(&ip_vs_protocol_ah);
375 #endif
376 #ifdef CONFIG_IP_VS_PROTO_ESP
377 REGISTER_PROTOCOL(&ip_vs_protocol_esp);
378 #endif
379 pr_info("Registered protocols (%s)\n", &protocols[2]);
380
381 return 0;
382 }
383
384
385 void ip_vs_protocol_cleanup(void)
386 {
387 struct ip_vs_protocol *pp;
388 int i;
389
390 /* unregister all the ipvs protocols */
391 for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
392 while ((pp = ip_vs_proto_table[i]) != NULL)
393 unregister_ip_vs_protocol(pp);
394 }
395 }
This page took 0.038964 seconds and 6 git commands to generate.