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 <linux/stat.h>
29 #include <linux/proc_fs.h>
30
31 #include <net/ip_vs.h>
32
33
34 /*
35 * IPVS protocols can only be registered/unregistered when the ipvs
36 * module is loaded/unloaded, so no lock is needed in accessing the
37 * ipvs protocol table.
38 */
39
40 #define IP_VS_PROTO_TAB_SIZE 32 /* must be power of 2 */
41 #define IP_VS_PROTO_HASH(proto) ((proto) & (IP_VS_PROTO_TAB_SIZE-1))
42
43 static struct ip_vs_protocol *ip_vs_proto_table[IP_VS_PROTO_TAB_SIZE];
44
45
46 /*
47 * register an ipvs protocol
48 */
49 static int __used __init register_ip_vs_protocol(struct ip_vs_protocol *pp)
50 {
51 unsigned int hash = IP_VS_PROTO_HASH(pp->protocol);
52
53 pp->next = ip_vs_proto_table[hash];
54 ip_vs_proto_table[hash] = pp;
55
56 if (pp->init != NULL)
57 pp->init(pp);
58
59 return 0;
60 }
61
62 /*
63 * register an ipvs protocols netns related data
64 */
65 static int
66 register_ip_vs_proto_netns(struct netns_ipvs *ipvs, struct ip_vs_protocol *pp)
67 {
68 unsigned int hash = IP_VS_PROTO_HASH(pp->protocol);
69 struct ip_vs_proto_data *pd =
70 kzalloc(sizeof(struct ip_vs_proto_data), GFP_KERNEL);
71
72 if (!pd)
73 return -ENOMEM;
74
75 pd->pp = pp; /* For speed issues */
76 pd->next = ipvs->proto_data_table[hash];
77 ipvs->proto_data_table[hash] = pd;
78 atomic_set(&pd->appcnt, 0); /* Init app counter */
79
80 if (pp->init_netns != NULL) {
81 int ret = pp->init_netns(ipvs, pd);
82 if (ret) {
83 /* unlink an free proto data */
84 ipvs->proto_data_table[hash] = pd->next;
85 kfree(pd);
86 return ret;
87 }
88 }
89
90 return 0;
91 }
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 int 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 netns_ipvs *ipvs, struct ip_vs_proto_data *pd)
119 {
120 struct ip_vs_proto_data **pd_p;
121 unsigned int hash = IP_VS_PROTO_HASH(pd->pp->protocol);
122
123 pd_p = &ipvs->proto_data_table[hash];
124 for (; *pd_p; pd_p = &(*pd_p)->next) {
125 if (*pd_p == pd) {
126 *pd_p = pd->next;
127 if (pd->pp->exit_netns != NULL)
128 pd->pp->exit_netns(ipvs, pd);
129 kfree(pd);
130 return 0;
131 }
132 }
133
134 return -ESRCH;
135 }
136
137 /*
138 * get ip_vs_protocol object by its proto.
139 */
140 struct ip_vs_protocol * ip_vs_proto_get(unsigned short proto)
141 {
142 struct ip_vs_protocol *pp;
143 unsigned int hash = IP_VS_PROTO_HASH(proto);
144
145 for (pp = ip_vs_proto_table[hash]; pp; pp = pp->next) {
146 if (pp->protocol == proto)
147 return pp;
148 }
149
150 return NULL;
151 }
152 EXPORT_SYMBOL(ip_vs_proto_get);
153
154 /*
155 * get ip_vs_protocol object data by netns and proto
156 */
157 struct ip_vs_proto_data *
158 ip_vs_proto_data_get(struct netns_ipvs *ipvs, unsigned short proto)
159 {
160 struct ip_vs_proto_data *pd;
161 unsigned int hash = IP_VS_PROTO_HASH(proto);
162
163 for (pd = ipvs->proto_data_table[hash]; pd; pd = pd->next) {
164 if (pd->pp->protocol == proto)
165 return pd;
166 }
167
168 return NULL;
169 }
170 EXPORT_SYMBOL(ip_vs_proto_data_get);
171
172 /*
173 * Propagate event for state change to all protocols
174 */
175 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags)
176 {
177 struct ip_vs_proto_data *pd;
178 int i;
179
180 for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
181 for (pd = ipvs->proto_data_table[i]; pd; pd = pd->next) {
182 if (pd->pp->timeout_change)
183 pd->pp->timeout_change(pd, flags);
184 }
185 }
186 }
187
188
189 int *
190 ip_vs_create_timeout_table(int *table, int size)
191 {
192 return kmemdup(table, size, GFP_KERNEL);
193 }
194
195
196 const char * ip_vs_state_name(__u16 proto, int state)
197 {
198 struct ip_vs_protocol *pp = ip_vs_proto_get(proto);
199
200 if (pp == NULL || pp->state_name == NULL)
201 return (IPPROTO_IP == proto) ? "NONE" : "ERR!";
202 return pp->state_name(state);
203 }
204
205
206 static void
207 ip_vs_tcpudp_debug_packet_v4(struct ip_vs_protocol *pp,
208 const struct sk_buff *skb,
209 int offset,
210 const char *msg)
211 {
212 char buf[128];
213 struct iphdr _iph, *ih;
214
215 ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph);
216 if (ih == NULL)
217 sprintf(buf, "TRUNCATED");
218 else if (ih->frag_off & htons(IP_OFFSET))
219 sprintf(buf, "%pI4->%pI4 frag", &ih->saddr, &ih->daddr);
220 else {
221 __be16 _ports[2], *pptr;
222
223 pptr = skb_header_pointer(skb, offset + ih->ihl*4,
224 sizeof(_ports), _ports);
225 if (pptr == NULL)
226 sprintf(buf, "TRUNCATED %pI4->%pI4",
227 &ih->saddr, &ih->daddr);
228 else
229 sprintf(buf, "%pI4:%u->%pI4:%u",
230 &ih->saddr, ntohs(pptr[0]),
231 &ih->daddr, ntohs(pptr[1]));
232 }
233
234 pr_debug("%s: %s %s\n", msg, pp->name, buf);
235 }
236
237 #ifdef CONFIG_IP_VS_IPV6
238 static void
239 ip_vs_tcpudp_debug_packet_v6(struct ip_vs_protocol *pp,
240 const struct sk_buff *skb,
241 int offset,
242 const char *msg)
243 {
244 char buf[192];
245 struct ipv6hdr _iph, *ih;
246
247 ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph);
248 if (ih == NULL)
249 sprintf(buf, "TRUNCATED");
250 else if (ih->nexthdr == IPPROTO_FRAGMENT)
251 sprintf(buf, "%pI6c->%pI6c frag", &ih->saddr, &ih->daddr);
252 else {
253 __be16 _ports[2], *pptr;
254
255 pptr = skb_header_pointer(skb, offset + sizeof(struct ipv6hdr),
256 sizeof(_ports), _ports);
257 if (pptr == NULL)
258 sprintf(buf, "TRUNCATED %pI6c->%pI6c",
259 &ih->saddr, &ih->daddr);
260 else
261 sprintf(buf, "%pI6c:%u->%pI6c:%u",
262 &ih->saddr, ntohs(pptr[0]),
263 &ih->daddr, ntohs(pptr[1]));
264 }
265
266 pr_debug("%s: %s %s\n", msg, pp->name, buf);
267 }
268 #endif
269
270
271 void
272 ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
273 const struct sk_buff *skb,
274 int offset,
275 const char *msg)
276 {
277 #ifdef CONFIG_IP_VS_IPV6
278 if (af == AF_INET6)
279 ip_vs_tcpudp_debug_packet_v6(pp, skb, offset, msg);
280 else
281 #endif
282 ip_vs_tcpudp_debug_packet_v4(pp, skb, offset, msg);
283 }
284
285 /*
286 * per network name-space init
287 */
288 int __net_init ip_vs_protocol_net_init(struct netns_ipvs *ipvs)
289 {
290 int i, ret;
291 static struct ip_vs_protocol *protos[] = {
292 #ifdef CONFIG_IP_VS_PROTO_TCP
293 &ip_vs_protocol_tcp,
294 #endif
295 #ifdef CONFIG_IP_VS_PROTO_UDP
296 &ip_vs_protocol_udp,
297 #endif
298 #ifdef CONFIG_IP_VS_PROTO_SCTP
299 &ip_vs_protocol_sctp,
300 #endif
301 #ifdef CONFIG_IP_VS_PROTO_AH
302 &ip_vs_protocol_ah,
303 #endif
304 #ifdef CONFIG_IP_VS_PROTO_ESP
305 &ip_vs_protocol_esp,
306 #endif
307 };
308
309 for (i = 0; i < ARRAY_SIZE(protos); i++) {
310 ret = register_ip_vs_proto_netns(ipvs, protos[i]);
311 if (ret < 0)
312 goto cleanup;
313 }
314 return 0;
315
316 cleanup:
317 ip_vs_protocol_net_cleanup(ipvs);
318 return ret;
319 }
320
321 void __net_exit ip_vs_protocol_net_cleanup(struct netns_ipvs *ipvs)
322 {
323 struct ip_vs_proto_data *pd;
324 int i;
325
326 /* unregister all the ipvs proto data for this netns */
327 for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
328 while ((pd = ipvs->proto_data_table[i]) != NULL)
329 unregister_ip_vs_proto_netns(ipvs, pd);
330 }
331 }
332
333 int __init ip_vs_protocol_init(void)
334 {
335 char protocols[64];
336 #define REGISTER_PROTOCOL(p) \
337 do { \
338 register_ip_vs_protocol(p); \
339 strcat(protocols, ", "); \
340 strcat(protocols, (p)->name); \
341 } while (0)
342
343 protocols[0] = '\0';
344 protocols[2] = '\0';
345 #ifdef CONFIG_IP_VS_PROTO_TCP
346 REGISTER_PROTOCOL(&ip_vs_protocol_tcp);
347 #endif
348 #ifdef CONFIG_IP_VS_PROTO_UDP
349 REGISTER_PROTOCOL(&ip_vs_protocol_udp);
350 #endif
351 #ifdef CONFIG_IP_VS_PROTO_SCTP
352 REGISTER_PROTOCOL(&ip_vs_protocol_sctp);
353 #endif
354 #ifdef CONFIG_IP_VS_PROTO_AH
355 REGISTER_PROTOCOL(&ip_vs_protocol_ah);
356 #endif
357 #ifdef CONFIG_IP_VS_PROTO_ESP
358 REGISTER_PROTOCOL(&ip_vs_protocol_esp);
359 #endif
360 pr_info("Registered protocols (%s)\n", &protocols[2]);
361
362 return 0;
363 }
364
365
366 void ip_vs_protocol_cleanup(void)
367 {
368 struct ip_vs_protocol *pp;
369 int i;
370
371 /* unregister all the ipvs protocols */
372 for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
373 while ((pp = ip_vs_proto_table[i]) != NULL)
374 unregister_ip_vs_protocol(pp);
375 }
376 }
377
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.