[people/arne_f/kernel.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/gfp.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, const char *const *names,
			const 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_SCTP
	REGISTER_PROTOCOL(&ip_vs_protocol_sctp);
#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>
5a0e3ad6	22	#include <linux/gfp.h>
1da177e4 LT	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 <asm/system.h>
	29	#include <linux/stat.h>
	30	#include <linux/proc_fs.h>
	31
	32	#include <net/ip_vs.h>
	33
	34
	35	/*
	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.
	39	*/
	40
	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))
	43
	44	static struct ip_vs_protocol *ip_vs_proto_table[IP_VS_PROTO_TAB_SIZE];
	45
	46
	47	/*
	48	* register an ipvs protocol
	49	*/
048cf48b	50	static int __used __init register_ip_vs_protocol(struct ip_vs_protocol *pp)
1da177e4 LT	51	{
	52	unsigned hash = IP_VS_PROTO_HASH(pp->protocol);
	53
	54	pp->next = ip_vs_proto_table[hash];
	55	ip_vs_proto_table[hash] = pp;
	56
	57	if (pp->init != NULL)
	58	pp->init(pp);
	59
	60	return 0;
	61	}
	62
	63
	64	/*
	65	* unregister an ipvs protocol
	66	*/
	67	static int unregister_ip_vs_protocol(struct ip_vs_protocol *pp)
	68	{
	69	struct ip_vs_protocol **pp_p;
	70	unsigned hash = IP_VS_PROTO_HASH(pp->protocol);
	71
	72	pp_p = &ip_vs_proto_table[hash];
	73	for (; pp_p; pp_p = &(pp_p)->next) {
	74	if (*pp_p == pp) {
	75	*pp_p = pp->next;
	76	if (pp->exit != NULL)
	77	pp->exit(pp);
	78	return 0;
	79	}
	80	}
	81
	82	return -ESRCH;
	83	}
	84
	85
	86	/*
	87	* get ip_vs_protocol object by its proto.
	88	*/
	89	struct ip_vs_protocol * ip_vs_proto_get(unsigned short proto)
	90	{
	91	struct ip_vs_protocol *pp;
	92	unsigned hash = IP_VS_PROTO_HASH(proto);
	93
	94	for (pp = ip_vs_proto_table[hash]; pp; pp = pp->next) {
	95	if (pp->protocol == proto)
	96	return pp;
	97	}
	98
	99	return NULL;
	100	}
	101
	102
	103	/*
	104	* Propagate event for state change to all protocols
	105	*/
	106	void ip_vs_protocol_timeout_change(int flags)
	107	{
	108	struct ip_vs_protocol *pp;
	109	int i;
	110
	111	for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
	112	for (pp = ip_vs_proto_table[i]; pp; pp = pp->next) {
	113	if (pp->timeout_change)
	114	pp->timeout_change(pp, flags);
115	}
116	}
117	}
118
119
120	int *
121	ip_vs_create_timeout_table(int *table, int size)
122	{
8b2ed4bb	123	return kmemdup(table, size, GFP_ATOMIC);
1da177e4 LT	124	}
	125
	126
	127	/*
	128	* Set timeout value for state specified by name
	129	*/
	130	int
36cbd3dc JE	131	ip_vs_set_state_timeout(int table, int num, const char const *names,
36cbd3dc JE	132	const char *name, int to)
1da177e4 LT	133	{
	134	int i;
	135
	136	if (!table \|\| !name \|\| !to)
	137	return -EINVAL;
	138
	139	for (i = 0; i < num; i++) {
	140	if (strcmp(names[i], name))
	141	continue;
	142	table[i] = to * HZ;
	143	return 0;
	144	}
	145	return -ENOENT;
	146	}
	147
	148
	149	const char * ip_vs_state_name(__u16 proto, int state)
	150	{
	151	struct ip_vs_protocol *pp = ip_vs_proto_get(proto);
	152
	153	if (pp == NULL \|\| pp->state_name == NULL)
2ad17def	154	return (IPPROTO_IP == proto) ? "NONE" : "ERR!";
1da177e4 LT	155	return pp->state_name(state);
	156	}
	157
	158
77eb8516	159	static void
3b047d9d JV	160	ip_vs_tcpudp_debug_packet_v4(struct ip_vs_protocol *pp,
	161	const struct sk_buff *skb,
	162	int offset,
	163	const char *msg)
1da177e4 LT	164	{
	165	char buf[128];
	166	struct iphdr _iph, *ih;
	167
	168	ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph);
	169	if (ih == NULL)
	170	sprintf(buf, "%s TRUNCATED", pp->name);
5661df7b	171	else if (ih->frag_off & htons(IP_OFFSET))
14d5e834 HH	172	sprintf(buf, "%s %pI4->%pI4 frag",
14d5e834 HH	173	pp->name, &ih->saddr, &ih->daddr);
1da177e4	174	else {
014d730d	175	__be16 _ports[2], *pptr
1da177e4 LT	176	;
	177	pptr = skb_header_pointer(skb, offset + ih->ihl*4,
	178	sizeof(_ports), _ports);
	179	if (pptr == NULL)
14d5e834 HH	180	sprintf(buf, "%s TRUNCATED %pI4->%pI4",
14d5e834 HH	181	pp->name, &ih->saddr, &ih->daddr);
1da177e4	182	else
14d5e834	183	sprintf(buf, "%s %pI4:%u->%pI4:%u",
1da177e4	184	pp->name,
14d5e834 HH	185	&ih->saddr, ntohs(pptr[0]),
14d5e834 HH	186	&ih->daddr, ntohs(pptr[1]));
1da177e4 LT	187	}
1da177e4 LT	188
9aada7ac	189	pr_debug("%s: %s\n", msg, buf);
1da177e4 LT	190	}
1da177e4 LT	191
3b047d9d	192	#ifdef CONFIG_IP_VS_IPV6
77eb8516	193	static void
3b047d9d JV	194	ip_vs_tcpudp_debug_packet_v6(struct ip_vs_protocol *pp,
	195	const struct sk_buff *skb,
	196	int offset,
	197	const char *msg)
	198	{
	199	char buf[192];
	200	struct ipv6hdr _iph, *ih;
	201
	202	ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph);
	203	if (ih == NULL)
	204	sprintf(buf, "%s TRUNCATED", pp->name);
	205	else if (ih->nexthdr == IPPROTO_FRAGMENT)
5b095d98	206	sprintf(buf, "%s %pI6->%pI6 frag",
38ff4fa4	207	pp->name, &ih->saddr, &ih->daddr);
3b047d9d JV	208	else {
	209	__be16 _ports[2], *pptr;
	210
	211	pptr = skb_header_pointer(skb, offset + sizeof(struct ipv6hdr),
	212	sizeof(_ports), _ports);
	213	if (pptr == NULL)
5b095d98	214	sprintf(buf, "%s TRUNCATED %pI6->%pI6",
38ff4fa4	215	pp->name, &ih->saddr, &ih->daddr);
3b047d9d	216	else
5b095d98	217	sprintf(buf, "%s %pI6:%u->%pI6:%u",
3b047d9d	218	pp->name,
38ff4fa4 HH	219	&ih->saddr, ntohs(pptr[0]),
38ff4fa4 HH	220	&ih->daddr, ntohs(pptr[1]));
3b047d9d JV	221	}
3b047d9d JV	222
9aada7ac	223	pr_debug("%s: %s\n", msg, buf);
3b047d9d JV	224	}
	225	#endif
	226
	227
	228	void
	229	ip_vs_tcpudp_debug_packet(struct ip_vs_protocol *pp,
	230	const struct sk_buff *skb,
	231	int offset,
	232	const char *msg)
	233	{
	234	#ifdef CONFIG_IP_VS_IPV6
d286600e	235	if (skb->protocol == htons(ETH_P_IPV6))
3b047d9d JV	236	ip_vs_tcpudp_debug_packet_v6(pp, skb, offset, msg);
	237	else
	238	#endif
	239	ip_vs_tcpudp_debug_packet_v4(pp, skb, offset, msg);
	240	}
	241
1da177e4	242
048cf48b	243	int __init ip_vs_protocol_init(void)
1da177e4 LT	244	{
	245	char protocols[64];
	246	#define REGISTER_PROTOCOL(p) \
	247	do { \
	248	register_ip_vs_protocol(p); \
	249	strcat(protocols, ", "); \
	250	strcat(protocols, (p)->name); \
	251	} while (0)
	252
	253	protocols[0] = '\0';
	254	protocols[2] = '\0';
	255	#ifdef CONFIG_IP_VS_PROTO_TCP
	256	REGISTER_PROTOCOL(&ip_vs_protocol_tcp);
	257	#endif
	258	#ifdef CONFIG_IP_VS_PROTO_UDP
	259	REGISTER_PROTOCOL(&ip_vs_protocol_udp);
	260	#endif
2906f66a VMR	261	#ifdef CONFIG_IP_VS_PROTO_SCTP
	262	REGISTER_PROTOCOL(&ip_vs_protocol_sctp);
	263	#endif
1da177e4 LT	264	#ifdef CONFIG_IP_VS_PROTO_AH
	265	REGISTER_PROTOCOL(&ip_vs_protocol_ah);
	266	#endif
	267	#ifdef CONFIG_IP_VS_PROTO_ESP
	268	REGISTER_PROTOCOL(&ip_vs_protocol_esp);
	269	#endif
1e3e238e	270	pr_info("Registered protocols (%s)\n", &protocols[2]);
1da177e4 LT	271
	272	return 0;
	273	}
	274
	275
	276	void ip_vs_protocol_cleanup(void)
	277	{
	278	struct ip_vs_protocol *pp;
	279	int i;
	280
	281	/* unregister all the ipvs protocols */
	282	for (i = 0; i < IP_VS_PROTO_TAB_SIZE; i++) {
	283	while ((pp = ip_vs_proto_table[i]) != NULL)
	284	unregister_ip_vs_protocol(pp);
	285	}
	286	}