2 * ip_vs_proto.c: transport protocol load balancing support for IPVS
4 * Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
5 * Julian Anastasov <ja@ssi.bg>
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.
16 #define KMSG_COMPONENT "IPVS"
17 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
19 #include <linux/module.h>
20 #include <linux/kernel.h>
21 #include <linux/skbuff.h>
22 #include <linux/gfp.h>
25 #include <net/protocol.h>
28 #include <linux/stat.h>
29 #include <linux/proc_fs.h>
31 #include <net/ip_vs.h>
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.
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))
43 static struct ip_vs_protocol *ip_vs_proto_table[IP_VS_PROTO_TAB_SIZE];
47 * register an ipvs protocol
49 static int __used __init register_ip_vs_protocol(struct ip_vs_protocol *pp)
51 unsigned int hash = IP_VS_PROTO_HASH(pp->protocol);
53 pp->next = ip_vs_proto_table[hash];
54 ip_vs_proto_table[hash] = pp;
63 * register an ipvs protocols netns related data
66 register_ip_vs_proto_netns(struct net *net, struct ip_vs_protocol *pp)
68 struct netns_ipvs *ipvs = net_ipvs(net);
69 unsigned int hash = IP_VS_PROTO_HASH(pp->protocol);
70 struct ip_vs_proto_data *pd =
71 kzalloc(sizeof(struct ip_vs_proto_data), GFP_KERNEL);
76 pd->pp = pp; /* For speed issues */
77 pd->next = ipvs->proto_data_table[hash];
78 ipvs->proto_data_table[hash] = pd;
79 atomic_set(&pd->appcnt, 0); /* Init app counter */
81 if (pp->init_netns != NULL) {
82 int ret = pp->init_netns(net, pd);
84 /* unlink an free proto data */
85 ipvs->proto_data_table[hash] = pd->next;
95 * unregister an ipvs protocol
97 static int unregister_ip_vs_protocol(struct ip_vs_protocol *pp)
99 struct ip_vs_protocol **pp_p;
100 unsigned int hash = IP_VS_PROTO_HASH(pp->protocol);
102 pp_p = &ip_vs_proto_table[hash];
103 for (; *pp_p; pp_p = &(*pp_p)->next) {
106 if (pp->exit != NULL)
116 * unregister an ipvs protocols netns data
119 unregister_ip_vs_proto_netns(struct net *net, struct ip_vs_proto_data *pd)
121 struct netns_ipvs *ipvs = net_ipvs(net);
122 struct ip_vs_proto_data **pd_p;
123 unsigned int hash = IP_VS_PROTO_HASH(pd->pp->protocol);
125 pd_p = &ipvs->proto_data_table[hash];
126 for (; *pd_p; pd_p = &(*pd_p)->next) {
129 if (pd->pp->exit_netns != NULL)
130 pd->pp->exit_netns(net, pd);
140 * get ip_vs_protocol object by its proto.
142 struct ip_vs_protocol * ip_vs_proto_get(unsigned short proto)
144 struct ip_vs_protocol *pp;
145 unsigned int hash = IP_VS_PROTO_HASH(proto);
147 for (pp = ip_vs_proto_table[hash]; pp; pp = pp->next) {
148 if (pp->protocol == proto)
154 EXPORT_SYMBOL(ip_vs_proto_get);
157 * get ip_vs_protocol object data by netns and proto
159 static struct ip_vs_proto_data *
160 __ipvs_proto_data_get(struct netns_ipvs *ipvs, unsigned short proto)
162 struct ip_vs_proto_data *pd;
163 unsigned int hash = IP_VS_PROTO_HASH(proto);
165 for (pd = ipvs->proto_data_table[hash]; pd; pd = pd->next) {
166 if (pd->pp->protocol == proto)
173 struct ip_vs_proto_data *
174 ip_vs_proto_data_get(struct net *net, unsigned short proto)
176 struct netns_ipvs *ipvs = net_ipvs(net);
178 return __ipvs_proto_data_get(ipvs, proto);
180 EXPORT_SYMBOL(ip_vs_proto_data_get);
183 * Propagate event for state change to all protocols
185 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags)
187 struct ip_vs_proto_data *pd;
190 for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
191 for (pd = ipvs->proto_data_table[i]; pd; pd = pd->next) {
192 if (pd->pp->timeout_change)
193 pd->pp->timeout_change(pd, flags);
200 ip_vs_create_timeout_table(int *table, int size)
202 return kmemdup(table, size, GFP_KERNEL);
207 * Set timeout value for state specified by name
210 ip_vs_set_state_timeout(int *table, int num, const char *const *names,
211 const char *name, int to)
215 if (!table || !name || !to)
218 for (i = 0; i < num; i++) {
219 if (strcmp(names[i], name))
228 const char * ip_vs_state_name(__u16 proto, int state)
230 struct ip_vs_protocol *pp = ip_vs_proto_get(proto);
232 if (pp == NULL || pp->state_name == NULL)
233 return (IPPROTO_IP == proto) ? "NONE" : "ERR!";
234 return pp->state_name(state);
239 ip_vs_tcpudp_debug_packet_v4(struct ip_vs_protocol *pp,
240 const struct sk_buff *skb,
245 struct iphdr _iph, *ih;
247 ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph);
249 sprintf(buf, "TRUNCATED");
250 else if (ih->frag_off & htons(IP_OFFSET))
251 sprintf(buf, "%pI4->%pI4 frag", &ih->saddr, &ih->daddr);
253 __be16 _ports[2], *pptr;
255 pptr = skb_header_pointer(skb, offset + ih->ihl*4,
256 sizeof(_ports), _ports);
258 sprintf(buf, "TRUNCATED %pI4->%pI4",
259 &ih->saddr, &ih->daddr);
261 sprintf(buf, "%pI4:%u->%pI4:%u",
262 &ih->saddr, ntohs(pptr[0]),
263 &ih->daddr, ntohs(pptr[1]));
266 pr_debug("%s: %s %s\n", msg, pp->name, buf);
269 #ifdef CONFIG_IP_VS_IPV6
271 ip_vs_tcpudp_debug_packet_v6(struct ip_vs_protocol *pp,
272 const struct sk_buff *skb,
277 struct ipv6hdr _iph, *ih;
279 ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph);
281 sprintf(buf, "TRUNCATED");
282 else if (ih->nexthdr == IPPROTO_FRAGMENT)
283 sprintf(buf, "%pI6c->%pI6c frag", &ih->saddr, &ih->daddr);
285 __be16 _ports[2], *pptr;
287 pptr = skb_header_pointer(skb, offset + sizeof(struct ipv6hdr),
288 sizeof(_ports), _ports);
290 sprintf(buf, "TRUNCATED %pI6c->%pI6c",
291 &ih->saddr, &ih->daddr);
293 sprintf(buf, "%pI6c:%u->%pI6c:%u",
294 &ih->saddr, ntohs(pptr[0]),
295 &ih->daddr, ntohs(pptr[1]));
298 pr_debug("%s: %s %s\n", msg, pp->name, buf);
304 ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
305 const struct sk_buff *skb,
309 #ifdef CONFIG_IP_VS_IPV6
311 ip_vs_tcpudp_debug_packet_v6(pp, skb, offset, msg);
314 ip_vs_tcpudp_debug_packet_v4(pp, skb, offset, msg);
318 * per network name-space init
320 int __net_init ip_vs_protocol_net_init(struct net *net)
323 static struct ip_vs_protocol *protos[] = {
324 #ifdef CONFIG_IP_VS_PROTO_TCP
327 #ifdef CONFIG_IP_VS_PROTO_UDP
330 #ifdef CONFIG_IP_VS_PROTO_SCTP
331 &ip_vs_protocol_sctp,
333 #ifdef CONFIG_IP_VS_PROTO_AH
336 #ifdef CONFIG_IP_VS_PROTO_ESP
341 for (i = 0; i < ARRAY_SIZE(protos); i++) {
342 ret = register_ip_vs_proto_netns(net, protos[i]);
349 ip_vs_protocol_net_cleanup(net);
353 void __net_exit ip_vs_protocol_net_cleanup(struct net *net)
355 struct netns_ipvs *ipvs = net_ipvs(net);
356 struct ip_vs_proto_data *pd;
359 /* unregister all the ipvs proto data for this netns */
360 for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
361 while ((pd = ipvs->proto_data_table[i]) != NULL)
362 unregister_ip_vs_proto_netns(net, pd);
366 int __init ip_vs_protocol_init(void)
369 #define REGISTER_PROTOCOL(p) \
371 register_ip_vs_protocol(p); \
372 strcat(protocols, ", "); \
373 strcat(protocols, (p)->name); \
378 #ifdef CONFIG_IP_VS_PROTO_TCP
379 REGISTER_PROTOCOL(&ip_vs_protocol_tcp);
381 #ifdef CONFIG_IP_VS_PROTO_UDP
382 REGISTER_PROTOCOL(&ip_vs_protocol_udp);
384 #ifdef CONFIG_IP_VS_PROTO_SCTP
385 REGISTER_PROTOCOL(&ip_vs_protocol_sctp);
387 #ifdef CONFIG_IP_VS_PROTO_AH
388 REGISTER_PROTOCOL(&ip_vs_protocol_ah);
390 #ifdef CONFIG_IP_VS_PROTO_ESP
391 REGISTER_PROTOCOL(&ip_vs_protocol_esp);
393 pr_info("Registered protocols (%s)\n", &protocols[2]);
399 void ip_vs_protocol_cleanup(void)
401 struct ip_vs_protocol *pp;
404 /* unregister all the ipvs protocols */
405 for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
406 while ((pp = ip_vs_proto_table[i]) != NULL)
407 unregister_ip_vs_protocol(pp);