[net-next-2.6.git] / net / netfilter / ipvs / ip_vs_proto.c

/*
 * ip_vs_proto.c: transport protocol load balancing support for IPVS
 *
 * Authors:     Wensong Zhang <wensong@linuxvirtualserver.org>
 *              Julian Anastasov <ja@ssi.bg>
 *
 *              This program is free software; you can redistribute it and/or
 *              modify it under the terms of the GNU General Public License
 *              as published by the Free Software Foundation; either version
 *              2 of the License, or (at your option) any later version.
 *
 * Changes:
 *
 */

#define KMSG_COMPONENT "IPVS"
#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt

#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/skbuff.h>
#include <linux/in.h>
#include <linux/ip.h>
#include <net/protocol.h>
#include <net/tcp.h>
#include <net/udp.h>
#include <asm/system.h>
#include <linux/stat.h>
#include <linux/proc_fs.h>

#include <net/ip_vs.h>


/*
 * IPVS protocols can only be registered/unregistered when the ipvs
 * module is loaded/unloaded, so no lock is needed in accessing the
 * ipvs protocol table.
 */

#define IP_VS_PROTO_TAB_SIZE		32	/* must be power of 2 */
#define IP_VS_PROTO_HASH(proto)		((proto) & (IP_VS_PROTO_TAB_SIZE-1))

static struct ip_vs_protocol *ip_vs_proto_table[IP_VS_PROTO_TAB_SIZE];


/*
 *	register an ipvs protocol
 */
static int __used __init register_ip_vs_protocol(struct ip_vs_protocol *pp)
{
	unsigned hash = IP_VS_PROTO_HASH(pp->protocol);

	pp->next = ip_vs_proto_table[hash];
	ip_vs_proto_table[hash] = pp;

	if (pp->init != NULL)
		pp->init(pp);

	return 0;
}


/*
 *	unregister an ipvs protocol
 */
static int unregister_ip_vs_protocol(struct ip_vs_protocol *pp)
{
	struct ip_vs_protocol **pp_p;
	unsigned hash = IP_VS_PROTO_HASH(pp->protocol);

	pp_p = &ip_vs_proto_table[hash];
	for (; *pp_p; pp_p = &(*pp_p)->next) {
		if (*pp_p == pp) {
			*pp_p = pp->next;
			if (pp->exit != NULL)
				pp->exit(pp);
			return 0;
		}
	}

	return -ESRCH;
}


/*
 *	get ip_vs_protocol object by its proto.
 */
struct ip_vs_protocol * ip_vs_proto_get(unsigned short proto)
{
	struct ip_vs_protocol *pp;
	unsigned hash = IP_VS_PROTO_HASH(proto);

	for (pp = ip_vs_proto_table[hash]; pp; pp = pp->next) {
		if (pp->protocol == proto)
			return pp;
	}

	return NULL;
}


/*
 *	Propagate event for state change to all protocols
 */
void ip_vs_protocol_timeout_change(int flags)
{
	struct ip_vs_protocol *pp;
	int i;

	for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
		for (pp = ip_vs_proto_table[i]; pp; pp = pp->next) {
			if (pp->timeout_change)
				pp->timeout_change(pp, flags);
		}
	}
}


int *
ip_vs_create_timeout_table(int *table, int size)
{
	return kmemdup(table, size, GFP_ATOMIC);
}


/*
 *	Set timeout value for state specified by name
 */
int
ip_vs_set_state_timeout(int *table, int num, char **names, char *name, int to)
{
	int i;

	if (!table || !name || !to)
		return -EINVAL;

	for (i = 0; i < num; i++) {
		if (strcmp(names[i], name))
			continue;
		table[i] = to * HZ;
		return 0;
	}
	return -ENOENT;
}


const char * ip_vs_state_name(__u16 proto, int state)
{
	struct ip_vs_protocol *pp = ip_vs_proto_get(proto);

	if (pp == NULL || pp->state_name == NULL)
		return (IPPROTO_IP == proto) ? "NONE" : "ERR!";
	return pp->state_name(state);
}


static void
ip_vs_tcpudp_debug_packet_v4(struct ip_vs_protocol *pp,
			     const struct sk_buff *skb,
			     int offset,
			     const char *msg)
{
	char buf[128];
	struct iphdr _iph, *ih;

	ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph);
	if (ih == NULL)
		sprintf(buf, "%s TRUNCATED", pp->name);
	else if (ih->frag_off & htons(IP_OFFSET))
		sprintf(buf, "%s %pI4->%pI4 frag",
			pp->name, &ih->saddr, &ih->daddr);
	else {
		__be16 _ports[2], *pptr
;
		pptr = skb_header_pointer(skb, offset + ih->ihl*4,
					  sizeof(_ports), _ports);
		if (pptr == NULL)
			sprintf(buf, "%s TRUNCATED %pI4->%pI4",
				pp->name, &ih->saddr, &ih->daddr);
		else
			sprintf(buf, "%s %pI4:%u->%pI4:%u",
				pp->name,
				&ih->saddr, ntohs(pptr[0]),
				&ih->daddr, ntohs(pptr[1]));
	}

	pr_debug("%s: %s\n", msg, buf);
}

#ifdef CONFIG_IP_VS_IPV6
static void
ip_vs_tcpudp_debug_packet_v6(struct ip_vs_protocol *pp,
			     const struct sk_buff *skb,
			     int offset,
			     const char *msg)
{
	char buf[192];
	struct ipv6hdr _iph, *ih;

	ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph);
	if (ih == NULL)
		sprintf(buf, "%s TRUNCATED", pp->name);
	else if (ih->nexthdr == IPPROTO_FRAGMENT)
		sprintf(buf, "%s %pI6->%pI6 frag",
			pp->name, &ih->saddr, &ih->daddr);
	else {
		__be16 _ports[2], *pptr;

		pptr = skb_header_pointer(skb, offset + sizeof(struct ipv6hdr),
					  sizeof(_ports), _ports);
		if (pptr == NULL)
			sprintf(buf, "%s TRUNCATED %pI6->%pI6",
				pp->name, &ih->saddr, &ih->daddr);
		else
			sprintf(buf, "%s %pI6:%u->%pI6:%u",
				pp->name,
				&ih->saddr, ntohs(pptr[0]),
				&ih->daddr, ntohs(pptr[1]));
	}

	pr_debug("%s: %s\n", msg, buf);
}
#endif


void
ip_vs_tcpudp_debug_packet(struct ip_vs_protocol *pp,
			  const struct sk_buff *skb,
			  int offset,
			  const char *msg)
{
#ifdef CONFIG_IP_VS_IPV6
	if (skb->protocol == htons(ETH_P_IPV6))
		ip_vs_tcpudp_debug_packet_v6(pp, skb, offset, msg);
	else
#endif
		ip_vs_tcpudp_debug_packet_v4(pp, skb, offset, msg);
}


int __init ip_vs_protocol_init(void)
{
	char protocols[64];
#define REGISTER_PROTOCOL(p)			\
	do {					\
		register_ip_vs_protocol(p);	\
		strcat(protocols, ", ");	\
		strcat(protocols, (p)->name);	\
	} while (0)

	protocols[0] = '\0';
	protocols[2] = '\0';
#ifdef CONFIG_IP_VS_PROTO_TCP
	REGISTER_PROTOCOL(&ip_vs_protocol_tcp);
#endif
#ifdef CONFIG_IP_VS_PROTO_UDP
	REGISTER_PROTOCOL(&ip_vs_protocol_udp);
#endif
#ifdef CONFIG_IP_VS_PROTO_AH
	REGISTER_PROTOCOL(&ip_vs_protocol_ah);
#endif
#ifdef CONFIG_IP_VS_PROTO_ESP
	REGISTER_PROTOCOL(&ip_vs_protocol_esp);
#endif
	pr_info("Registered protocols (%s)\n", &protocols[2]);

	return 0;
}


void ip_vs_protocol_cleanup(void)
{
	struct ip_vs_protocol *pp;
	int i;

	/* unregister all the ipvs protocols */
	for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
		while ((pp = ip_vs_proto_table[i]) != NULL)
			unregister_ip_vs_protocol(pp);
	}
}
Commit	Line	Data
1da177e4 LT	1	/*
	2	* ip_vs_proto.c: transport protocol load balancing support for IPVS
	3	*
1da177e4 LT	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
9aada7ac HE	16	#define KMSG_COMPONENT "IPVS"
	17	#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
	18
1da177e4 LT	19	#include <linux/module.h>
	20	#include <linux/kernel.h>
	21	#include <linux/skbuff.h>
	22	#include <linux/in.h>
	23	#include <linux/ip.h>
	24	#include <net/protocol.h>
	25	#include <net/tcp.h>
	26	#include <net/udp.h>
	27	#include <asm/system.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
	43	static struct ip_vs_protocol *ip_vs_proto_table[IP_VS_PROTO_TAB_SIZE];
	44
	45
	46	/*
	47	* register an ipvs protocol
	48	*/
048cf48b	49	static int __used __init register_ip_vs_protocol(struct ip_vs_protocol *pp)
1da177e4 LT	50	{
	51	unsigned 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	/*
	64	* unregister an ipvs protocol
	65	*/
	66	static int unregister_ip_vs_protocol(struct ip_vs_protocol *pp)
	67	{
	68	struct ip_vs_protocol **pp_p;
	69	unsigned hash = IP_VS_PROTO_HASH(pp->protocol);
	70
	71	pp_p = &ip_vs_proto_table[hash];
	72	for (; pp_p; pp_p = &(pp_p)->next) {
	73	if (*pp_p == pp) {
	74	*pp_p = pp->next;
	75	if (pp->exit != NULL)
	76	pp->exit(pp);
	77	return 0;
	78	}
	79	}
	80
	81	return -ESRCH;
	82	}
	83
	84
	85	/*
	86	* get ip_vs_protocol object by its proto.
	87	*/
	88	struct ip_vs_protocol * ip_vs_proto_get(unsigned short proto)
	89	{
	90	struct ip_vs_protocol *pp;
	91	unsigned hash = IP_VS_PROTO_HASH(proto);
	92
	93	for (pp = ip_vs_proto_table[hash]; pp; pp = pp->next) {
	94	if (pp->protocol == proto)
	95	return pp;
	96	}
	97
	98	return NULL;
	99	}
	100
	101
	102	/*
	103	* Propagate event for state change to all protocols
	104	*/
	105	void ip_vs_protocol_timeout_change(int flags)
	106	{
	107	struct ip_vs_protocol *pp;
	108	int i;
	109
	110	for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
	111	for (pp = ip_vs_proto_table[i]; pp; pp = pp->next) {
	112	if (pp->timeout_change)
	113	pp->timeout_change(pp, flags);
114	}
115	}
116	}
117
118
119	int *
120	ip_vs_create_timeout_table(int *table, int size)
121	{
8b2ed4bb	122	return kmemdup(table, size, GFP_ATOMIC);
1da177e4 LT	123	}
	124
	125
	126	/*
	127	* Set timeout value for state specified by name
	128	*/
	129	int
	130	ip_vs_set_state_timeout(int table, int num, char names, char name, int to)
	131	{
	132	int i;
	133
	134	if (!table \|\| !name \|\| !to)
	135	return -EINVAL;
	136
	137	for (i = 0; i < num; i++) {
	138	if (strcmp(names[i], name))
	139	continue;
	140	table[i] = to * HZ;
	141	return 0;
	142	}
	143	return -ENOENT;
	144	}
	145
	146
	147	const char * ip_vs_state_name(__u16 proto, int state)
	148	{
	149	struct ip_vs_protocol *pp = ip_vs_proto_get(proto);
	150
	151	if (pp == NULL \|\| pp->state_name == NULL)
2ad17def	152	return (IPPROTO_IP == proto) ? "NONE" : "ERR!";
1da177e4 LT	153	return pp->state_name(state);
	154	}
	155
	156
77eb8516	157	static void
3b047d9d JV	158	ip_vs_tcpudp_debug_packet_v4(struct ip_vs_protocol *pp,
	159	const struct sk_buff *skb,
	160	int offset,
	161	const char *msg)
1da177e4 LT	162	{
	163	char buf[128];
	164	struct iphdr _iph, *ih;
	165
	166	ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph);
	167	if (ih == NULL)
	168	sprintf(buf, "%s TRUNCATED", pp->name);
5661df7b	169	else if (ih->frag_off & htons(IP_OFFSET))
14d5e834 HH	170	sprintf(buf, "%s %pI4->%pI4 frag",
14d5e834 HH	171	pp->name, &ih->saddr, &ih->daddr);
1da177e4	172	else {
014d730d	173	__be16 _ports[2], *pptr
1da177e4 LT	174	;
	175	pptr = skb_header_pointer(skb, offset + ih->ihl*4,
	176	sizeof(_ports), _ports);
	177	if (pptr == NULL)
14d5e834 HH	178	sprintf(buf, "%s TRUNCATED %pI4->%pI4",
14d5e834 HH	179	pp->name, &ih->saddr, &ih->daddr);
1da177e4	180	else
14d5e834	181	sprintf(buf, "%s %pI4:%u->%pI4:%u",
1da177e4	182	pp->name,
14d5e834 HH	183	&ih->saddr, ntohs(pptr[0]),
14d5e834 HH	184	&ih->daddr, ntohs(pptr[1]));
1da177e4 LT	185	}
1da177e4 LT	186
9aada7ac	187	pr_debug("%s: %s\n", msg, buf);
1da177e4 LT	188	}
1da177e4 LT	189
3b047d9d	190	#ifdef CONFIG_IP_VS_IPV6
77eb8516	191	static void
3b047d9d JV	192	ip_vs_tcpudp_debug_packet_v6(struct ip_vs_protocol *pp,
	193	const struct sk_buff *skb,
	194	int offset,
	195	const char *msg)
	196	{
	197	char buf[192];
	198	struct ipv6hdr _iph, *ih;
	199
	200	ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph);
	201	if (ih == NULL)
	202	sprintf(buf, "%s TRUNCATED", pp->name);
	203	else if (ih->nexthdr == IPPROTO_FRAGMENT)
5b095d98	204	sprintf(buf, "%s %pI6->%pI6 frag",
38ff4fa4	205	pp->name, &ih->saddr, &ih->daddr);
3b047d9d JV	206	else {
	207	__be16 _ports[2], *pptr;
	208
	209	pptr = skb_header_pointer(skb, offset + sizeof(struct ipv6hdr),
	210	sizeof(_ports), _ports);
	211	if (pptr == NULL)
5b095d98	212	sprintf(buf, "%s TRUNCATED %pI6->%pI6",
38ff4fa4	213	pp->name, &ih->saddr, &ih->daddr);
3b047d9d	214	else
5b095d98	215	sprintf(buf, "%s %pI6:%u->%pI6:%u",
3b047d9d	216	pp->name,
38ff4fa4 HH	217	&ih->saddr, ntohs(pptr[0]),
38ff4fa4 HH	218	&ih->daddr, ntohs(pptr[1]));
3b047d9d JV	219	}
3b047d9d JV	220
9aada7ac	221	pr_debug("%s: %s\n", msg, buf);
3b047d9d JV	222	}
	223	#endif
	224
	225
	226	void
	227	ip_vs_tcpudp_debug_packet(struct ip_vs_protocol *pp,
	228	const struct sk_buff *skb,
	229	int offset,
	230	const char *msg)
	231	{
	232	#ifdef CONFIG_IP_VS_IPV6
d286600e	233	if (skb->protocol == htons(ETH_P_IPV6))
3b047d9d JV	234	ip_vs_tcpudp_debug_packet_v6(pp, skb, offset, msg);
	235	else
	236	#endif
	237	ip_vs_tcpudp_debug_packet_v4(pp, skb, offset, msg);
	238	}
	239
1da177e4	240
048cf48b	241	int __init ip_vs_protocol_init(void)
1da177e4 LT	242	{
	243	char protocols[64];
	244	#define REGISTER_PROTOCOL(p) \
	245	do { \
	246	register_ip_vs_protocol(p); \
	247	strcat(protocols, ", "); \
	248	strcat(protocols, (p)->name); \
	249	} while (0)
	250
	251	protocols[0] = '\0';
	252	protocols[2] = '\0';
	253	#ifdef CONFIG_IP_VS_PROTO_TCP
	254	REGISTER_PROTOCOL(&ip_vs_protocol_tcp);
	255	#endif
	256	#ifdef CONFIG_IP_VS_PROTO_UDP
	257	REGISTER_PROTOCOL(&ip_vs_protocol_udp);
	258	#endif
1da177e4 LT	259	#ifdef CONFIG_IP_VS_PROTO_AH
	260	REGISTER_PROTOCOL(&ip_vs_protocol_ah);
	261	#endif
	262	#ifdef CONFIG_IP_VS_PROTO_ESP
	263	REGISTER_PROTOCOL(&ip_vs_protocol_esp);
	264	#endif
1e3e238e	265	pr_info("Registered protocols (%s)\n", &protocols[2]);
1da177e4 LT	266
	267	return 0;
	268	}
	269
	270
	271	void ip_vs_protocol_cleanup(void)
	272	{
	273	struct ip_vs_protocol *pp;
	274	int i;
	275
	276	/* unregister all the ipvs protocols */
	277	for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
	278	while ((pp = ip_vs_proto_table[i]) != NULL)
	279	unregister_ip_vs_protocol(pp);
	280	}
	281	}