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 <asm/system.h>
29 #include <linux/stat.h>
30 #include <linux/proc_fs.h>
32 #include <net/ip_vs.h>
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.
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))
44 static struct ip_vs_protocol *ip_vs_proto_table[IP_VS_PROTO_TAB_SIZE];
48 * register an ipvs protocol
50 static int __used __init register_ip_vs_protocol(struct ip_vs_protocol *pp)
52 unsigned hash = IP_VS_PROTO_HASH(pp->protocol);
54 pp->next = ip_vs_proto_table[hash];
55 ip_vs_proto_table[hash] = pp;
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)
67 * register an ipvs protocols netns related data
70 register_ip_vs_proto_netns(struct net *net, struct ip_vs_protocol *pp)
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);
80 pd->pp = pp; /* For speed issues */
81 pd->next = ipvs->proto_data_table[hash];
82 ipvs->proto_data_table[hash] = pd;
83 atomic_set(&pd->appcnt, 0); /* Init app counter */
85 if (pp->init_netns != NULL)
86 pp->init_netns(net, pd);
93 * unregister an ipvs protocol
95 static int unregister_ip_vs_protocol(struct ip_vs_protocol *pp)
97 struct ip_vs_protocol **pp_p;
98 unsigned hash = IP_VS_PROTO_HASH(pp->protocol);
100 pp_p = &ip_vs_proto_table[hash];
101 for (; *pp_p; pp_p = &(*pp_p)->next) {
104 if (pp->exit != NULL)
114 * unregister an ipvs protocols netns data
117 unregister_ip_vs_proto_netns(struct net *net, struct ip_vs_proto_data *pd)
119 struct netns_ipvs *ipvs = net_ipvs(net);
120 struct ip_vs_proto_data **pd_p;
121 unsigned hash = IP_VS_PROTO_HASH(pd->pp->protocol);
123 pd_p = &ipvs->proto_data_table[hash];
124 for (; *pd_p; pd_p = &(*pd_p)->next) {
127 if (pd->pp->exit_netns != NULL)
128 pd->pp->exit_netns(net, pd);
138 * get ip_vs_protocol object by its proto.
140 struct ip_vs_protocol * ip_vs_proto_get(unsigned short proto)
142 struct ip_vs_protocol *pp;
143 unsigned hash = IP_VS_PROTO_HASH(proto);
145 for (pp = ip_vs_proto_table[hash]; pp; pp = pp->next) {
146 if (pp->protocol == proto)
152 EXPORT_SYMBOL(ip_vs_proto_get);
155 * get ip_vs_protocol object data by netns and proto
157 struct ip_vs_proto_data *
158 __ipvs_proto_data_get(struct netns_ipvs *ipvs, unsigned short proto)
160 struct ip_vs_proto_data *pd;
161 unsigned hash = IP_VS_PROTO_HASH(proto);
163 for (pd = ipvs->proto_data_table[hash]; pd; pd = pd->next) {
164 if (pd->pp->protocol == proto)
171 struct ip_vs_proto_data *
172 ip_vs_proto_data_get(struct net *net, unsigned short proto)
174 struct netns_ipvs *ipvs = net_ipvs(net);
176 return __ipvs_proto_data_get(ipvs, proto);
178 EXPORT_SYMBOL(ip_vs_proto_data_get);
181 * Propagate event for state change to all protocols
183 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags)
185 struct ip_vs_proto_data *pd;
188 for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
189 for (pd = ipvs->proto_data_table[i]; pd; pd = pd->next) {
190 if (pd->pp->timeout_change)
191 pd->pp->timeout_change(pd, flags);
198 ip_vs_create_timeout_table(int *table, int size)
200 return kmemdup(table, size, GFP_ATOMIC);
205 * Set timeout value for state specified by name
208 ip_vs_set_state_timeout(int *table, int num, const char *const *names,
209 const char *name, int to)
213 if (!table || !name || !to)
216 for (i = 0; i < num; i++) {
217 if (strcmp(names[i], name))
226 const char * ip_vs_state_name(__u16 proto, int state)
228 struct ip_vs_protocol *pp = ip_vs_proto_get(proto);
230 if (pp == NULL || pp->state_name == NULL)
231 return (IPPROTO_IP == proto) ? "NONE" : "ERR!";
232 return pp->state_name(state);
237 ip_vs_tcpudp_debug_packet_v4(struct ip_vs_protocol *pp,
238 const struct sk_buff *skb,
243 struct iphdr _iph, *ih;
245 ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph);
247 sprintf(buf, "TRUNCATED");
248 else if (ih->frag_off & htons(IP_OFFSET))
249 sprintf(buf, "%pI4->%pI4 frag", &ih->saddr, &ih->daddr);
251 __be16 _ports[2], *pptr;
253 pptr = skb_header_pointer(skb, offset + ih->ihl*4,
254 sizeof(_ports), _ports);
256 sprintf(buf, "TRUNCATED %pI4->%pI4",
257 &ih->saddr, &ih->daddr);
259 sprintf(buf, "%pI4:%u->%pI4:%u",
260 &ih->saddr, ntohs(pptr[0]),
261 &ih->daddr, ntohs(pptr[1]));
264 pr_debug("%s: %s %s\n", msg, pp->name, buf);
267 #ifdef CONFIG_IP_VS_IPV6
269 ip_vs_tcpudp_debug_packet_v6(struct ip_vs_protocol *pp,
270 const struct sk_buff *skb,
275 struct ipv6hdr _iph, *ih;
277 ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph);
279 sprintf(buf, "TRUNCATED");
280 else if (ih->nexthdr == IPPROTO_FRAGMENT)
281 sprintf(buf, "%pI6->%pI6 frag", &ih->saddr, &ih->daddr);
283 __be16 _ports[2], *pptr;
285 pptr = skb_header_pointer(skb, offset + sizeof(struct ipv6hdr),
286 sizeof(_ports), _ports);
288 sprintf(buf, "TRUNCATED %pI6->%pI6",
289 &ih->saddr, &ih->daddr);
291 sprintf(buf, "%pI6:%u->%pI6:%u",
292 &ih->saddr, ntohs(pptr[0]),
293 &ih->daddr, ntohs(pptr[1]));
296 pr_debug("%s: %s %s\n", msg, pp->name, buf);
302 ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
303 const struct sk_buff *skb,
307 #ifdef CONFIG_IP_VS_IPV6
309 ip_vs_tcpudp_debug_packet_v6(pp, skb, offset, msg);
312 ip_vs_tcpudp_debug_packet_v4(pp, skb, offset, msg);
316 * per network name-space init
318 int __net_init ip_vs_protocol_net_init(struct net *net)
320 #ifdef CONFIG_IP_VS_PROTO_TCP
321 register_ip_vs_proto_netns(net, &ip_vs_protocol_tcp);
323 #ifdef CONFIG_IP_VS_PROTO_UDP
324 register_ip_vs_proto_netns(net, &ip_vs_protocol_udp);
326 #ifdef CONFIG_IP_VS_PROTO_SCTP
327 register_ip_vs_proto_netns(net, &ip_vs_protocol_sctp);
329 #ifdef CONFIG_IP_VS_PROTO_AH
330 register_ip_vs_proto_netns(net, &ip_vs_protocol_ah);
332 #ifdef CONFIG_IP_VS_PROTO_ESP
333 register_ip_vs_proto_netns(net, &ip_vs_protocol_esp);
338 void __net_exit ip_vs_protocol_net_cleanup(struct net *net)
340 struct netns_ipvs *ipvs = net_ipvs(net);
341 struct ip_vs_proto_data *pd;
344 /* unregister all the ipvs proto data for this netns */
345 for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
346 while ((pd = ipvs->proto_data_table[i]) != NULL)
347 unregister_ip_vs_proto_netns(net, pd);
351 int __init ip_vs_protocol_init(void)
354 #define REGISTER_PROTOCOL(p) \
356 register_ip_vs_protocol(p); \
357 strcat(protocols, ", "); \
358 strcat(protocols, (p)->name); \
363 #ifdef CONFIG_IP_VS_PROTO_TCP
364 REGISTER_PROTOCOL(&ip_vs_protocol_tcp);
366 #ifdef CONFIG_IP_VS_PROTO_UDP
367 REGISTER_PROTOCOL(&ip_vs_protocol_udp);
369 #ifdef CONFIG_IP_VS_PROTO_SCTP
370 REGISTER_PROTOCOL(&ip_vs_protocol_sctp);
372 #ifdef CONFIG_IP_VS_PROTO_AH
373 REGISTER_PROTOCOL(&ip_vs_protocol_ah);
375 #ifdef CONFIG_IP_VS_PROTO_ESP
376 REGISTER_PROTOCOL(&ip_vs_protocol_esp);
378 pr_info("Registered protocols (%s)\n", &protocols[2]);
384 void ip_vs_protocol_cleanup(void)
386 struct ip_vs_protocol *pp;
389 /* unregister all the ipvs protocols */
390 for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
391 while ((pp = ip_vs_proto_table[i]) != NULL)
392 unregister_ip_vs_protocol(pp);