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 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 hash = IP_VS_PROTO_HASH(pp->protocol);
70 struct ip_vs_proto_data *pd =
71 kzalloc(sizeof(struct ip_vs_proto_data), GFP_ATOMIC);
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 pp->init_netns(net, pd);
88 * unregister an ipvs protocol
90 static int unregister_ip_vs_protocol(struct ip_vs_protocol *pp)
92 struct ip_vs_protocol **pp_p;
93 unsigned hash = IP_VS_PROTO_HASH(pp->protocol);
95 pp_p = &ip_vs_proto_table[hash];
96 for (; *pp_p; pp_p = &(*pp_p)->next) {
109 * unregister an ipvs protocols netns data
112 unregister_ip_vs_proto_netns(struct net *net, struct ip_vs_proto_data *pd)
114 struct netns_ipvs *ipvs = net_ipvs(net);
115 struct ip_vs_proto_data **pd_p;
116 unsigned hash = IP_VS_PROTO_HASH(pd->pp->protocol);
118 pd_p = &ipvs->proto_data_table[hash];
119 for (; *pd_p; pd_p = &(*pd_p)->next) {
122 if (pd->pp->exit_netns != NULL)
123 pd->pp->exit_netns(net, pd);
133 * get ip_vs_protocol object by its proto.
135 struct ip_vs_protocol * ip_vs_proto_get(unsigned short proto)
137 struct ip_vs_protocol *pp;
138 unsigned hash = IP_VS_PROTO_HASH(proto);
140 for (pp = ip_vs_proto_table[hash]; pp; pp = pp->next) {
141 if (pp->protocol == proto)
147 EXPORT_SYMBOL(ip_vs_proto_get);
150 * get ip_vs_protocol object data by netns and proto
152 struct ip_vs_proto_data *
153 __ipvs_proto_data_get(struct netns_ipvs *ipvs, unsigned short proto)
155 struct ip_vs_proto_data *pd;
156 unsigned hash = IP_VS_PROTO_HASH(proto);
158 for (pd = ipvs->proto_data_table[hash]; pd; pd = pd->next) {
159 if (pd->pp->protocol == proto)
166 struct ip_vs_proto_data *
167 ip_vs_proto_data_get(struct net *net, unsigned short proto)
169 struct netns_ipvs *ipvs = net_ipvs(net);
171 return __ipvs_proto_data_get(ipvs, proto);
173 EXPORT_SYMBOL(ip_vs_proto_data_get);
176 * Propagate event for state change to all protocols
178 void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags)
180 struct ip_vs_proto_data *pd;
183 for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
184 for (pd = ipvs->proto_data_table[i]; pd; pd = pd->next) {
185 if (pd->pp->timeout_change)
186 pd->pp->timeout_change(pd, flags);
193 ip_vs_create_timeout_table(int *table, int size)
195 return kmemdup(table, size, GFP_ATOMIC);
200 * Set timeout value for state specified by name
203 ip_vs_set_state_timeout(int *table, int num, const char *const *names,
204 const char *name, int to)
208 if (!table || !name || !to)
211 for (i = 0; i < num; i++) {
212 if (strcmp(names[i], name))
221 const char * ip_vs_state_name(__u16 proto, int state)
223 struct ip_vs_protocol *pp = ip_vs_proto_get(proto);
225 if (pp == NULL || pp->state_name == NULL)
226 return (IPPROTO_IP == proto) ? "NONE" : "ERR!";
227 return pp->state_name(state);
232 ip_vs_tcpudp_debug_packet_v4(struct ip_vs_protocol *pp,
233 const struct sk_buff *skb,
238 struct iphdr _iph, *ih;
240 ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph);
242 sprintf(buf, "TRUNCATED");
243 else if (ih->frag_off & htons(IP_OFFSET))
244 sprintf(buf, "%pI4->%pI4 frag", &ih->saddr, &ih->daddr);
246 __be16 _ports[2], *pptr;
248 pptr = skb_header_pointer(skb, offset + ih->ihl*4,
249 sizeof(_ports), _ports);
251 sprintf(buf, "TRUNCATED %pI4->%pI4",
252 &ih->saddr, &ih->daddr);
254 sprintf(buf, "%pI4:%u->%pI4:%u",
255 &ih->saddr, ntohs(pptr[0]),
256 &ih->daddr, ntohs(pptr[1]));
259 pr_debug("%s: %s %s\n", msg, pp->name, buf);
262 #ifdef CONFIG_IP_VS_IPV6
264 ip_vs_tcpudp_debug_packet_v6(struct ip_vs_protocol *pp,
265 const struct sk_buff *skb,
270 struct ipv6hdr _iph, *ih;
272 ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph);
274 sprintf(buf, "TRUNCATED");
275 else if (ih->nexthdr == IPPROTO_FRAGMENT)
276 sprintf(buf, "%pI6->%pI6 frag", &ih->saddr, &ih->daddr);
278 __be16 _ports[2], *pptr;
280 pptr = skb_header_pointer(skb, offset + sizeof(struct ipv6hdr),
281 sizeof(_ports), _ports);
283 sprintf(buf, "TRUNCATED %pI6->%pI6",
284 &ih->saddr, &ih->daddr);
286 sprintf(buf, "%pI6:%u->%pI6:%u",
287 &ih->saddr, ntohs(pptr[0]),
288 &ih->daddr, ntohs(pptr[1]));
291 pr_debug("%s: %s %s\n", msg, pp->name, buf);
297 ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
298 const struct sk_buff *skb,
302 #ifdef CONFIG_IP_VS_IPV6
304 ip_vs_tcpudp_debug_packet_v6(pp, skb, offset, msg);
307 ip_vs_tcpudp_debug_packet_v4(pp, skb, offset, msg);
311 * per network name-space init
313 int __net_init ip_vs_protocol_net_init(struct net *net)
316 static struct ip_vs_protocol *protos[] = {
317 #ifdef CONFIG_IP_VS_PROTO_TCP
320 #ifdef CONFIG_IP_VS_PROTO_UDP
323 #ifdef CONFIG_IP_VS_PROTO_SCTP
324 &ip_vs_protocol_sctp,
326 #ifdef CONFIG_IP_VS_PROTO_AH
329 #ifdef CONFIG_IP_VS_PROTO_ESP
334 for (i = 0; i < ARRAY_SIZE(protos); i++) {
335 ret = register_ip_vs_proto_netns(net, protos[i]);
342 ip_vs_protocol_net_cleanup(net);
346 void __net_exit ip_vs_protocol_net_cleanup(struct net *net)
348 struct netns_ipvs *ipvs = net_ipvs(net);
349 struct ip_vs_proto_data *pd;
352 /* unregister all the ipvs proto data for this netns */
353 for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
354 while ((pd = ipvs->proto_data_table[i]) != NULL)
355 unregister_ip_vs_proto_netns(net, pd);
359 int __init ip_vs_protocol_init(void)
362 #define REGISTER_PROTOCOL(p) \
364 register_ip_vs_protocol(p); \
365 strcat(protocols, ", "); \
366 strcat(protocols, (p)->name); \
371 #ifdef CONFIG_IP_VS_PROTO_TCP
372 REGISTER_PROTOCOL(&ip_vs_protocol_tcp);
374 #ifdef CONFIG_IP_VS_PROTO_UDP
375 REGISTER_PROTOCOL(&ip_vs_protocol_udp);
377 #ifdef CONFIG_IP_VS_PROTO_SCTP
378 REGISTER_PROTOCOL(&ip_vs_protocol_sctp);
380 #ifdef CONFIG_IP_VS_PROTO_AH
381 REGISTER_PROTOCOL(&ip_vs_protocol_ah);
383 #ifdef CONFIG_IP_VS_PROTO_ESP
384 REGISTER_PROTOCOL(&ip_vs_protocol_esp);
386 pr_info("Registered protocols (%s)\n", &protocols[2]);
392 void ip_vs_protocol_cleanup(void)
394 struct ip_vs_protocol *pp;
397 /* unregister all the ipvs protocols */
398 for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
399 while ((pp = ip_vs_proto_table[i]) != NULL)
400 unregister_ip_vs_protocol(pp);