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];
45 /* States for conn templates: NONE or words separated with ",", max 15 chars */
46 static const char *ip_vs_ctpl_state_name_table[IP_VS_CTPL_S_LAST] = {
47 [IP_VS_CTPL_S_NONE] = "NONE",
48 [IP_VS_CTPL_S_ASSURED] = "ASSURED",
52 * register an ipvs protocol
54 static int __used __init register_ip_vs_protocol(struct ip_vs_protocol *pp)
56 unsigned int hash = IP_VS_PROTO_HASH(pp->protocol);
58 pp->next = ip_vs_proto_table[hash];
59 ip_vs_proto_table[hash] = pp;
68 * register an ipvs protocols netns related data
71 register_ip_vs_proto_netns(struct netns_ipvs *ipvs, struct ip_vs_protocol *pp)
73 unsigned int hash = IP_VS_PROTO_HASH(pp->protocol);
74 struct ip_vs_proto_data *pd =
75 kzalloc(sizeof(struct ip_vs_proto_data), GFP_KERNEL);
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 int ret = pp->init_netns(ipvs, pd);
88 /* unlink an free proto data */
89 ipvs->proto_data_table[hash] = pd->next;
99 * unregister an ipvs protocol
101 static int unregister_ip_vs_protocol(struct ip_vs_protocol *pp)
103 struct ip_vs_protocol **pp_p;
104 unsigned int hash = IP_VS_PROTO_HASH(pp->protocol);
106 pp_p = &ip_vs_proto_table[hash];
107 for (; *pp_p; pp_p = &(*pp_p)->next) {
110 if (pp->exit != NULL)
120 * unregister an ipvs protocols netns data
123 unregister_ip_vs_proto_netns(struct netns_ipvs *ipvs, struct ip_vs_proto_data *pd)
125 struct ip_vs_proto_data **pd_p;
126 unsigned int hash = IP_VS_PROTO_HASH(pd->pp->protocol);
128 pd_p = &ipvs->proto_data_table[hash];
129 for (; *pd_p; pd_p = &(*pd_p)->next) {
132 if (pd->pp->exit_netns != NULL)
133 pd->pp->exit_netns(ipvs, pd);
143 * get ip_vs_protocol object by its proto.
145 struct ip_vs_protocol * ip_vs_proto_get(unsigned short proto)
147 struct ip_vs_protocol *pp;
148 unsigned int hash = IP_VS_PROTO_HASH(proto);
150 for (pp = ip_vs_proto_table[hash]; pp; pp = pp->next) {
151 if (pp->protocol == proto)
157 EXPORT_SYMBOL(ip_vs_proto_get);
160 * get ip_vs_protocol object data by netns and proto
162 struct ip_vs_proto_data *
163 ip_vs_proto_data_get(struct netns_ipvs *ipvs, unsigned short proto)
165 struct ip_vs_proto_data *pd;
166 unsigned int hash = IP_VS_PROTO_HASH(proto);
168 for (pd = ipvs->proto_data_table[hash]; pd; pd = pd->next) {
169 if (pd->pp->protocol == proto)
175 EXPORT_SYMBOL(ip_vs_proto_data_get);
178 * Propagate event for state change to all protocols
180 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags)
182 struct ip_vs_proto_data *pd;
185 for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
186 for (pd = ipvs->proto_data_table[i]; pd; pd = pd->next) {
187 if (pd->pp->timeout_change)
188 pd->pp->timeout_change(pd, flags);
195 ip_vs_create_timeout_table(int *table, int size)
197 return kmemdup(table, size, GFP_KERNEL);
201 const char *ip_vs_state_name(const struct ip_vs_conn *cp)
203 unsigned int state = cp->state;
204 struct ip_vs_protocol *pp;
206 if (cp->flags & IP_VS_CONN_F_TEMPLATE) {
208 if (state >= IP_VS_CTPL_S_LAST)
210 return ip_vs_ctpl_state_name_table[state] ? : "?";
212 pp = ip_vs_proto_get(cp->protocol);
213 if (pp == NULL || pp->state_name == NULL)
214 return (cp->protocol == IPPROTO_IP) ? "NONE" : "ERR!";
215 return pp->state_name(state);
220 ip_vs_tcpudp_debug_packet_v4(struct ip_vs_protocol *pp,
221 const struct sk_buff *skb,
226 struct iphdr _iph, *ih;
228 ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph);
230 sprintf(buf, "TRUNCATED");
231 else if (ih->frag_off & htons(IP_OFFSET))
232 sprintf(buf, "%pI4->%pI4 frag", &ih->saddr, &ih->daddr);
234 __be16 _ports[2], *pptr;
236 pptr = skb_header_pointer(skb, offset + ih->ihl*4,
237 sizeof(_ports), _ports);
239 sprintf(buf, "TRUNCATED %pI4->%pI4",
240 &ih->saddr, &ih->daddr);
242 sprintf(buf, "%pI4:%u->%pI4:%u",
243 &ih->saddr, ntohs(pptr[0]),
244 &ih->daddr, ntohs(pptr[1]));
247 pr_debug("%s: %s %s\n", msg, pp->name, buf);
250 #ifdef CONFIG_IP_VS_IPV6
252 ip_vs_tcpudp_debug_packet_v6(struct ip_vs_protocol *pp,
253 const struct sk_buff *skb,
258 struct ipv6hdr _iph, *ih;
260 ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph);
262 sprintf(buf, "TRUNCATED");
263 else if (ih->nexthdr == IPPROTO_FRAGMENT)
264 sprintf(buf, "%pI6c->%pI6c frag", &ih->saddr, &ih->daddr);
266 __be16 _ports[2], *pptr;
268 pptr = skb_header_pointer(skb, offset + sizeof(struct ipv6hdr),
269 sizeof(_ports), _ports);
271 sprintf(buf, "TRUNCATED %pI6c->%pI6c",
272 &ih->saddr, &ih->daddr);
274 sprintf(buf, "%pI6c:%u->%pI6c:%u",
275 &ih->saddr, ntohs(pptr[0]),
276 &ih->daddr, ntohs(pptr[1]));
279 pr_debug("%s: %s %s\n", msg, pp->name, buf);
285 ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
286 const struct sk_buff *skb,
290 #ifdef CONFIG_IP_VS_IPV6
292 ip_vs_tcpudp_debug_packet_v6(pp, skb, offset, msg);
295 ip_vs_tcpudp_debug_packet_v4(pp, skb, offset, msg);
299 * per network name-space init
301 int __net_init ip_vs_protocol_net_init(struct netns_ipvs *ipvs)
304 static struct ip_vs_protocol *protos[] = {
305 #ifdef CONFIG_IP_VS_PROTO_TCP
308 #ifdef CONFIG_IP_VS_PROTO_UDP
311 #ifdef CONFIG_IP_VS_PROTO_SCTP
312 &ip_vs_protocol_sctp,
314 #ifdef CONFIG_IP_VS_PROTO_AH
317 #ifdef CONFIG_IP_VS_PROTO_ESP
322 for (i = 0; i < ARRAY_SIZE(protos); i++) {
323 ret = register_ip_vs_proto_netns(ipvs, protos[i]);
330 ip_vs_protocol_net_cleanup(ipvs);
334 void __net_exit ip_vs_protocol_net_cleanup(struct netns_ipvs *ipvs)
336 struct ip_vs_proto_data *pd;
339 /* unregister all the ipvs proto data for this netns */
340 for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
341 while ((pd = ipvs->proto_data_table[i]) != NULL)
342 unregister_ip_vs_proto_netns(ipvs, pd);
346 int __init ip_vs_protocol_init(void)
349 #define REGISTER_PROTOCOL(p) \
351 register_ip_vs_protocol(p); \
352 strcat(protocols, ", "); \
353 strcat(protocols, (p)->name); \
358 #ifdef CONFIG_IP_VS_PROTO_TCP
359 REGISTER_PROTOCOL(&ip_vs_protocol_tcp);
361 #ifdef CONFIG_IP_VS_PROTO_UDP
362 REGISTER_PROTOCOL(&ip_vs_protocol_udp);
364 #ifdef CONFIG_IP_VS_PROTO_SCTP
365 REGISTER_PROTOCOL(&ip_vs_protocol_sctp);
367 #ifdef CONFIG_IP_VS_PROTO_AH
368 REGISTER_PROTOCOL(&ip_vs_protocol_ah);
370 #ifdef CONFIG_IP_VS_PROTO_ESP
371 REGISTER_PROTOCOL(&ip_vs_protocol_esp);
373 pr_info("Registered protocols (%s)\n", &protocols[2]);
379 void ip_vs_protocol_cleanup(void)
381 struct ip_vs_protocol *pp;
384 /* unregister all the ipvs protocols */
385 for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
386 while ((pp = ip_vs_proto_table[i]) != NULL)
387 unregister_ip_vs_protocol(pp);