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
43static struct ip_vs_protocol *ip_vs_proto_table[IP_VS_PROTO_TAB_SIZE];
44
45
46/*
47 *	register an ipvs protocol
48 */
49static 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 */
65static int
66register_ip_vs_proto_netns(struct net *net, struct ip_vs_protocol *pp)
67{
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);
72
73	if (!pd)
74		return -ENOMEM;
75
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 */
80
81	if (pp->init_netns != NULL) {
82		int ret = pp->init_netns(net, pd);
83		if (ret) {
84			/* unlink an free proto data */
85			ipvs->proto_data_table[hash] = pd->next;
86			kfree(pd);
87			return ret;
88		}
89	}
90
91	return 0;
92}
93
94/*
95 *	unregister an ipvs protocol
96 */
97static int unregister_ip_vs_protocol(struct ip_vs_protocol *pp)
98{
99	struct ip_vs_protocol **pp_p;
100	unsigned int hash = IP_VS_PROTO_HASH(pp->protocol);
101
102	pp_p = &ip_vs_proto_table[hash];
103	for (; *pp_p; pp_p = &(*pp_p)->next) {
104		if (*pp_p == pp) {
105			*pp_p = pp->next;
106			if (pp->exit != NULL)
107				pp->exit(pp);
108			return 0;
109		}
110	}
111
112	return -ESRCH;
113}
114
115/*
116 *	unregister an ipvs protocols netns data
117 */
118static int
119unregister_ip_vs_proto_netns(struct net *net, struct ip_vs_proto_data *pd)
120{
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);
124
125	pd_p = &ipvs->proto_data_table[hash];
126	for (; *pd_p; pd_p = &(*pd_p)->next) {
127		if (*pd_p == pd) {
128			*pd_p = pd->next;
129			if (pd->pp->exit_netns != NULL)
130				pd->pp->exit_netns(net, pd);
131			kfree(pd);
132			return 0;
133		}
134	}
135
136	return -ESRCH;
137}
138
139/*
140 *	get ip_vs_protocol object by its proto.
141 */
142struct ip_vs_protocol * ip_vs_proto_get(unsigned short proto)
143{
144	struct ip_vs_protocol *pp;
145	unsigned int hash = IP_VS_PROTO_HASH(proto);
146
147	for (pp = ip_vs_proto_table[hash]; pp; pp = pp->next) {
148		if (pp->protocol == proto)
149			return pp;
150	}
151
152	return NULL;
153}
154EXPORT_SYMBOL(ip_vs_proto_get);
155
156/*
157 *	get ip_vs_protocol object data by netns and proto
158 */
159static struct ip_vs_proto_data *
160__ipvs_proto_data_get(struct netns_ipvs *ipvs, unsigned short proto)
161{
162	struct ip_vs_proto_data *pd;
163	unsigned int hash = IP_VS_PROTO_HASH(proto);
164
165	for (pd = ipvs->proto_data_table[hash]; pd; pd = pd->next) {
166		if (pd->pp->protocol == proto)
167			return pd;
168	}
169
170	return NULL;
171}
172
173struct ip_vs_proto_data *
174ip_vs_proto_data_get(struct net *net, unsigned short proto)
175{
176	struct netns_ipvs *ipvs = net_ipvs(net);
177
178	return __ipvs_proto_data_get(ipvs, proto);
179}
180EXPORT_SYMBOL(ip_vs_proto_data_get);
181
182/*
183 *	Propagate event for state change to all protocols
184 */
185void ip_vs_protocol_timeout_change(struct netns_ipvs *ipvs, int flags)
186{
187	struct ip_vs_proto_data *pd;
188	int i;
189
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);
194		}
195	}
196}
197
198
199int *
200ip_vs_create_timeout_table(int *table, int size)
201{
202	return kmemdup(table, size, GFP_KERNEL);
203}
204
205
206/*
207 *	Set timeout value for state specified by name
208 */
209int
210ip_vs_set_state_timeout(int *table, int num, const char *const *names,
211			const char *name, int to)
212{
213	int i;
214
215	if (!table || !name || !to)
216		return -EINVAL;
217
218	for (i = 0; i < num; i++) {
219		if (strcmp(names[i], name))
220			continue;
221		table[i] = to * HZ;
222		return 0;
223	}
224	return -ENOENT;
225}
226
227
228const char * ip_vs_state_name(__u16 proto, int state)
229{
230	struct ip_vs_protocol *pp = ip_vs_proto_get(proto);
231
232	if (pp == NULL || pp->state_name == NULL)
233		return (IPPROTO_IP == proto) ? "NONE" : "ERR!";
234	return pp->state_name(state);
235}
236
237
238static void
239ip_vs_tcpudp_debug_packet_v4(struct ip_vs_protocol *pp,
240			     const struct sk_buff *skb,
241			     int offset,
242			     const char *msg)
243{
244	char buf[128];
245	struct iphdr _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->frag_off & htons(IP_OFFSET))
251		sprintf(buf, "%pI4->%pI4 frag", &ih->saddr, &ih->daddr);
252	else {
253		__be16 _ports[2], *pptr;
254
255		pptr = skb_header_pointer(skb, offset + ih->ihl*4,
256					  sizeof(_ports), _ports);
257		if (pptr == NULL)
258			sprintf(buf, "TRUNCATED %pI4->%pI4",
259				&ih->saddr, &ih->daddr);
260		else
261			sprintf(buf, "%pI4:%u->%pI4:%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
269#ifdef CONFIG_IP_VS_IPV6
270static void
271ip_vs_tcpudp_debug_packet_v6(struct ip_vs_protocol *pp,
272			     const struct sk_buff *skb,
273			     int offset,
274			     const char *msg)
275{
276	char buf[192];
277	struct ipv6hdr _iph, *ih;
278
279	ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph);
280	if (ih == NULL)
281		sprintf(buf, "TRUNCATED");
282	else if (ih->nexthdr == IPPROTO_FRAGMENT)
283		sprintf(buf, "%pI6c->%pI6c frag", &ih->saddr, &ih->daddr);
284	else {
285		__be16 _ports[2], *pptr;
286
287		pptr = skb_header_pointer(skb, offset + sizeof(struct ipv6hdr),
288					  sizeof(_ports), _ports);
289		if (pptr == NULL)
290			sprintf(buf, "TRUNCATED %pI6c->%pI6c",
291				&ih->saddr, &ih->daddr);
292		else
293			sprintf(buf, "%pI6c:%u->%pI6c:%u",
294				&ih->saddr, ntohs(pptr[0]),
295				&ih->daddr, ntohs(pptr[1]));
296	}
297
298	pr_debug("%s: %s %s\n", msg, pp->name, buf);
299}
300#endif
301
302
303void
304ip_vs_tcpudp_debug_packet(int af, struct ip_vs_protocol *pp,
305			  const struct sk_buff *skb,
306			  int offset,
307			  const char *msg)
308{
309#ifdef CONFIG_IP_VS_IPV6
310	if (af == AF_INET6)
311		ip_vs_tcpudp_debug_packet_v6(pp, skb, offset, msg);
312	else
313#endif
314		ip_vs_tcpudp_debug_packet_v4(pp, skb, offset, msg);
315}
316
317/*
318 * per network name-space init
319 */
320int __net_init ip_vs_protocol_net_init(struct net *net)
321{
322	int i, ret;
323	static struct ip_vs_protocol *protos[] = {
324#ifdef CONFIG_IP_VS_PROTO_TCP
325        &ip_vs_protocol_tcp,
326#endif
327#ifdef CONFIG_IP_VS_PROTO_UDP
328	&ip_vs_protocol_udp,
329#endif
330#ifdef CONFIG_IP_VS_PROTO_SCTP
331	&ip_vs_protocol_sctp,
332#endif
333#ifdef CONFIG_IP_VS_PROTO_AH
334	&ip_vs_protocol_ah,
335#endif
336#ifdef CONFIG_IP_VS_PROTO_ESP
337	&ip_vs_protocol_esp,
338#endif
339	};
340
341	for (i = 0; i < ARRAY_SIZE(protos); i++) {
342		ret = register_ip_vs_proto_netns(net, protos[i]);
343		if (ret < 0)
344			goto cleanup;
345	}
346	return 0;
347
348cleanup:
349	ip_vs_protocol_net_cleanup(net);
350	return ret;
351}
352
353void __net_exit ip_vs_protocol_net_cleanup(struct net *net)
354{
355	struct netns_ipvs *ipvs = net_ipvs(net);
356	struct ip_vs_proto_data *pd;
357	int i;
358
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);
363	}
364}
365
366int __init ip_vs_protocol_init(void)
367{
368	char protocols[64];
369#define REGISTER_PROTOCOL(p)			\
370	do {					\
371		register_ip_vs_protocol(p);	\
372		strcat(protocols, ", ");	\
373		strcat(protocols, (p)->name);	\
374	} while (0)
375
376	protocols[0] = '\0';
377	protocols[2] = '\0';
378#ifdef CONFIG_IP_VS_PROTO_TCP
379	REGISTER_PROTOCOL(&ip_vs_protocol_tcp);
380#endif
381#ifdef CONFIG_IP_VS_PROTO_UDP
382	REGISTER_PROTOCOL(&ip_vs_protocol_udp);
383#endif
384#ifdef CONFIG_IP_VS_PROTO_SCTP
385	REGISTER_PROTOCOL(&ip_vs_protocol_sctp);
386#endif
387#ifdef CONFIG_IP_VS_PROTO_AH
388	REGISTER_PROTOCOL(&ip_vs_protocol_ah);
389#endif
390#ifdef CONFIG_IP_VS_PROTO_ESP
391	REGISTER_PROTOCOL(&ip_vs_protocol_esp);
392#endif
393	pr_info("Registered protocols (%s)\n", &protocols[2]);
394
395	return 0;
396}
397
398
399void ip_vs_protocol_cleanup(void)
400{
401	struct ip_vs_protocol *pp;
402	int i;
403
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);
408	}
409}
410