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

/*
 * IPVS:        Source Hashing scheduling module
 *
 * Authors:     Wensong Zhang <wensong@gnuchina.org>
 *
 *              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:
 *
 */

/*
 * The sh algorithm is to select server by the hash key of source IP
 * address. The pseudo code is as follows:
 *
 *       n <- servernode[src_ip];
 *       if (n is dead) OR
 *          (n is overloaded) or (n.weight <= 0) then
 *                 return NULL;
 *
 *       return n;
 *
 * Notes that servernode is a 256-bucket hash table that maps the hash
 * index derived from packet source IP address to the current server
 * array. If the sh scheduler is used in cache cluster, it is good to
 * combine it with cache_bypass feature. When the statically assigned
 * server is dead or overloaded, the load balancer can bypass the cache
 * server and send requests to the original server directly.
 *
 */

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

#include <linux/ip.h>
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/skbuff.h>

#include <net/ip_vs.h>


/*
 *      IPVS SH bucket
 */
struct ip_vs_sh_bucket {
	struct ip_vs_dest       *dest;          /* real server (cache) */
};

/*
 *     for IPVS SH entry hash table
 */
#ifndef CONFIG_IP_VS_SH_TAB_BITS
#define CONFIG_IP_VS_SH_TAB_BITS        8
#endif
#define IP_VS_SH_TAB_BITS               CONFIG_IP_VS_SH_TAB_BITS
#define IP_VS_SH_TAB_SIZE               (1 << IP_VS_SH_TAB_BITS)
#define IP_VS_SH_TAB_MASK               (IP_VS_SH_TAB_SIZE - 1)


/*
 *	Returns hash value for IPVS SH entry
 */
static inline unsigned ip_vs_sh_hashkey(int af, const union nf_inet_addr *addr)
{
	__be32 addr_fold = addr->ip;

#ifdef CONFIG_IP_VS_IPV6
	if (af == AF_INET6)
		addr_fold = addr->ip6[0]^addr->ip6[1]^
			    addr->ip6[2]^addr->ip6[3];
#endif
	return (ntohl(addr_fold)*2654435761UL) & IP_VS_SH_TAB_MASK;
}


/*
 *      Get ip_vs_dest associated with supplied parameters.
 */
static inline struct ip_vs_dest *
ip_vs_sh_get(int af, struct ip_vs_sh_bucket *tbl,
	     const union nf_inet_addr *addr)
{
	return (tbl[ip_vs_sh_hashkey(af, addr)]).dest;
}


/*
 *      Assign all the hash buckets of the specified table with the service.
 */
static int
ip_vs_sh_assign(struct ip_vs_sh_bucket *tbl, struct ip_vs_service *svc)
{
	int i;
	struct ip_vs_sh_bucket *b;
	struct list_head *p;
	struct ip_vs_dest *dest;

	b = tbl;
	p = &svc->destinations;
	for (i=0; i<IP_VS_SH_TAB_SIZE; i++) {
		if (list_empty(p)) {
			b->dest = NULL;
		} else {
			if (p == &svc->destinations)
				p = p->next;

			dest = list_entry(p, struct ip_vs_dest, n_list);
			atomic_inc(&dest->refcnt);
			b->dest = dest;

			p = p->next;
		}
		b++;
	}
	return 0;
}


/*
 *      Flush all the hash buckets of the specified table.
 */
static void ip_vs_sh_flush(struct ip_vs_sh_bucket *tbl)
{
	int i;
	struct ip_vs_sh_bucket *b;

	b = tbl;
	for (i=0; i<IP_VS_SH_TAB_SIZE; i++) {
		if (b->dest) {
			atomic_dec(&b->dest->refcnt);
			b->dest = NULL;
		}
		b++;
	}
}


static int ip_vs_sh_init_svc(struct ip_vs_service *svc)
{
	struct ip_vs_sh_bucket *tbl;

	/* allocate the SH table for this service */
	tbl = kmalloc(sizeof(struct ip_vs_sh_bucket)*IP_VS_SH_TAB_SIZE,
		      GFP_ATOMIC);
	if (tbl == NULL) {
		IP_VS_ERR("ip_vs_sh_init_svc(): no memory\n");
		return -ENOMEM;
	}
	svc->sched_data = tbl;
	IP_VS_DBG(6, "SH hash table (memory=%Zdbytes) allocated for "
		  "current service\n",
		  sizeof(struct ip_vs_sh_bucket)*IP_VS_SH_TAB_SIZE);

	/* assign the hash buckets with the updated service */
	ip_vs_sh_assign(tbl, svc);

	return 0;
}


static int ip_vs_sh_done_svc(struct ip_vs_service *svc)
{
	struct ip_vs_sh_bucket *tbl = svc->sched_data;

	/* got to clean up hash buckets here */
	ip_vs_sh_flush(tbl);

	/* release the table itself */
	kfree(svc->sched_data);
	IP_VS_DBG(6, "SH hash table (memory=%Zdbytes) released\n",
		  sizeof(struct ip_vs_sh_bucket)*IP_VS_SH_TAB_SIZE);

	return 0;
}


static int ip_vs_sh_update_svc(struct ip_vs_service *svc)
{
	struct ip_vs_sh_bucket *tbl = svc->sched_data;

	/* got to clean up hash buckets here */
	ip_vs_sh_flush(tbl);

	/* assign the hash buckets with the updated service */
	ip_vs_sh_assign(tbl, svc);

	return 0;
}


/*
 *      If the dest flags is set with IP_VS_DEST_F_OVERLOAD,
 *      consider that the server is overloaded here.
 */
static inline int is_overloaded(struct ip_vs_dest *dest)
{
	return dest->flags & IP_VS_DEST_F_OVERLOAD;
}


/*
 *      Source Hashing scheduling
 */
static struct ip_vs_dest *
ip_vs_sh_schedule(struct ip_vs_service *svc, const struct sk_buff *skb)
{
	struct ip_vs_dest *dest;
	struct ip_vs_sh_bucket *tbl;
	struct ip_vs_iphdr iph;

	ip_vs_fill_iphdr(svc->af, skb_network_header(skb), &iph);

	IP_VS_DBG(6, "ip_vs_sh_schedule(): Scheduling...\n");

	tbl = (struct ip_vs_sh_bucket *)svc->sched_data;
	dest = ip_vs_sh_get(svc->af, tbl, &iph.saddr);
	if (!dest
	    || !(dest->flags & IP_VS_DEST_F_AVAILABLE)
	    || atomic_read(&dest->weight) <= 0
	    || is_overloaded(dest)) {
		IP_VS_ERR_RL("SH: no destination available\n");
		return NULL;
	}

	IP_VS_DBG_BUF(6, "SH: source IP address %s --> server %s:%d\n",
		      IP_VS_DBG_ADDR(svc->af, &iph.saddr),
		      IP_VS_DBG_ADDR(svc->af, &dest->addr),
		      ntohs(dest->port));

	return dest;
}


/*
 *      IPVS SH Scheduler structure
 */
static struct ip_vs_scheduler ip_vs_sh_scheduler =
{
	.name =			"sh",
	.refcnt =		ATOMIC_INIT(0),
	.module =		THIS_MODULE,
	.n_list	 =		LIST_HEAD_INIT(ip_vs_sh_scheduler.n_list),
	.init_service =		ip_vs_sh_init_svc,
	.done_service =		ip_vs_sh_done_svc,
	.update_service =	ip_vs_sh_update_svc,
	.schedule =		ip_vs_sh_schedule,
};


static int __init ip_vs_sh_init(void)
{
	return register_ip_vs_scheduler(&ip_vs_sh_scheduler);
}


static void __exit ip_vs_sh_cleanup(void)
{
	unregister_ip_vs_scheduler(&ip_vs_sh_scheduler);
}


module_init(ip_vs_sh_init);
module_exit(ip_vs_sh_cleanup);
MODULE_LICENSE("GPL");
Commit	Line	Data
1da177e4 LT	1	/*
	2	* IPVS: Source Hashing scheduling module
	3	*
1da177e4 LT	4	* Authors: Wensong Zhang <wensong@gnuchina.org>
	5	*
	6	* This program is free software; you can redistribute it and/or
	7	* modify it under the terms of the GNU General Public License
	8	* as published by the Free Software Foundation; either version
	9	* 2 of the License, or (at your option) any later version.
	10	*
	11	* Changes:
	12	*
	13	*/
	14
	15	/*
	16	* The sh algorithm is to select server by the hash key of source IP
	17	* address. The pseudo code is as follows:
	18	*
	19	* n <- servernode[src_ip];
	20	* if (n is dead) OR
	21	* (n is overloaded) or (n.weight <= 0) then
	22	* return NULL;
	23	*
	24	* return n;
	25	*
	26	* Notes that servernode is a 256-bucket hash table that maps the hash
	27	* index derived from packet source IP address to the current server
	28	* array. If the sh scheduler is used in cache cluster, it is good to
	29	* combine it with cache_bypass feature. When the statically assigned
	30	* server is dead or overloaded, the load balancer can bypass the cache
	31	* server and send requests to the original server directly.
	32	*
	33	*/
	34
9aada7ac HE	35	#define KMSG_COMPONENT "IPVS"
	36	#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
	37
14c85021	38	#include <linux/ip.h>
1da177e4 LT	39	#include <linux/module.h>
1da177e4 LT	40	#include <linux/kernel.h>
14c85021	41	#include <linux/skbuff.h>
1da177e4 LT	42
	43	#include <net/ip_vs.h>
	44
	45
	46	/*
	47	* IPVS SH bucket
	48	*/
	49	struct ip_vs_sh_bucket {
	50	struct ip_vs_dest dest; / real server (cache) */
	51	};
	52
	53	/*
	54	* for IPVS SH entry hash table
	55	*/
	56	#ifndef CONFIG_IP_VS_SH_TAB_BITS
	57	#define CONFIG_IP_VS_SH_TAB_BITS 8
	58	#endif
	59	#define IP_VS_SH_TAB_BITS CONFIG_IP_VS_SH_TAB_BITS
	60	#define IP_VS_SH_TAB_SIZE (1 << IP_VS_SH_TAB_BITS)
	61	#define IP_VS_SH_TAB_MASK (IP_VS_SH_TAB_SIZE - 1)
	62
	63
	64	/*
	65	* Returns hash value for IPVS SH entry
	66	*/
20971a0a	67	static inline unsigned ip_vs_sh_hashkey(int af, const union nf_inet_addr *addr)
1da177e4	68	{
20971a0a JV	69	__be32 addr_fold = addr->ip;
	70
	71	#ifdef CONFIG_IP_VS_IPV6
	72	if (af == AF_INET6)
	73	addr_fold = addr->ip6[0]^addr->ip6[1]^
	74	addr->ip6[2]^addr->ip6[3];
	75	#endif
	76	return (ntohl(addr_fold)*2654435761UL) & IP_VS_SH_TAB_MASK;
1da177e4 LT	77	}
	78
	79
	80	/*
	81	* Get ip_vs_dest associated with supplied parameters.
	82	*/
	83	static inline struct ip_vs_dest *
20971a0a JV	84	ip_vs_sh_get(int af, struct ip_vs_sh_bucket *tbl,
20971a0a JV	85	const union nf_inet_addr *addr)
1da177e4	86	{
20971a0a	87	return (tbl[ip_vs_sh_hashkey(af, addr)]).dest;
1da177e4 LT	88	}
	89
	90
	91	/*
	92	* Assign all the hash buckets of the specified table with the service.
	93	*/
	94	static int
	95	ip_vs_sh_assign(struct ip_vs_sh_bucket tbl, struct ip_vs_service svc)
	96	{
	97	int i;
	98	struct ip_vs_sh_bucket *b;
	99	struct list_head *p;
	100	struct ip_vs_dest *dest;
	101
	102	b = tbl;
	103	p = &svc->destinations;
	104	for (i=0; i<IP_VS_SH_TAB_SIZE; i++) {
	105	if (list_empty(p)) {
	106	b->dest = NULL;
	107	} else {
	108	if (p == &svc->destinations)
	109	p = p->next;
	110
	111	dest = list_entry(p, struct ip_vs_dest, n_list);
	112	atomic_inc(&dest->refcnt);
	113	b->dest = dest;
	114
	115	p = p->next;
	116	}
	117	b++;
	118	}
	119	return 0;
	120	}
	121
	122
	123	/*
	124	* Flush all the hash buckets of the specified table.
	125	*/
	126	static void ip_vs_sh_flush(struct ip_vs_sh_bucket *tbl)
	127	{
	128	int i;
	129	struct ip_vs_sh_bucket *b;
	130
	131	b = tbl;
	132	for (i=0; i<IP_VS_SH_TAB_SIZE; i++) {
	133	if (b->dest) {
	134	atomic_dec(&b->dest->refcnt);
	135	b->dest = NULL;
	136	}
	137	b++;
	138	}
	139	}
	140
	141
	142	static int ip_vs_sh_init_svc(struct ip_vs_service *svc)
	143	{
	144	struct ip_vs_sh_bucket *tbl;
	145
	146	/* allocate the SH table for this service */
	147	tbl = kmalloc(sizeof(struct ip_vs_sh_bucket)*IP_VS_SH_TAB_SIZE,
	148	GFP_ATOMIC);
	149	if (tbl == NULL) {
	150	IP_VS_ERR("ip_vs_sh_init_svc(): no memory\n");
	151	return -ENOMEM;
152	}
153	svc->sched_data = tbl;
154	IP_VS_DBG(6, "SH hash table (memory=%Zdbytes) allocated for "
155	"current service\n",
156	sizeof(struct ip_vs_sh_bucket)*IP_VS_SH_TAB_SIZE);
157
158	/* assign the hash buckets with the updated service */
159	ip_vs_sh_assign(tbl, svc);
160
161	return 0;
162	}
163
164
165	static int ip_vs_sh_done_svc(struct ip_vs_service *svc)
166	{
167	struct ip_vs_sh_bucket *tbl = svc->sched_data;
168
169	/* got to clean up hash buckets here */
170	ip_vs_sh_flush(tbl);
171
172	/* release the table itself */
173	kfree(svc->sched_data);
174	IP_VS_DBG(6, "SH hash table (memory=%Zdbytes) released\n",
175	sizeof(struct ip_vs_sh_bucket)*IP_VS_SH_TAB_SIZE);
176
177	return 0;
178	}
179
180
181	static int ip_vs_sh_update_svc(struct ip_vs_service *svc)
182	{
183	struct ip_vs_sh_bucket *tbl = svc->sched_data;
184
185	/* got to clean up hash buckets here */
186	ip_vs_sh_flush(tbl);
187
188	/* assign the hash buckets with the updated service */
189	ip_vs_sh_assign(tbl, svc);
190
191	return 0;
192	}
193
194
195	/*
196	* If the dest flags is set with IP_VS_DEST_F_OVERLOAD,
197	* consider that the server is overloaded here.
198	*/
199	static inline int is_overloaded(struct ip_vs_dest *dest)
200	{
201	return dest->flags & IP_VS_DEST_F_OVERLOAD;
202	}
203
204
205	/*
206	* Source Hashing scheduling
207	*/
208	static struct ip_vs_dest *
209	ip_vs_sh_schedule(struct ip_vs_service svc, const struct sk_buff skb)
210	{
211	struct ip_vs_dest *dest;
212	struct ip_vs_sh_bucket *tbl;
20971a0a JV	213	struct ip_vs_iphdr iph;
	214
	215	ip_vs_fill_iphdr(svc->af, skb_network_header(skb), &iph);
1da177e4 LT	216
	217	IP_VS_DBG(6, "ip_vs_sh_schedule(): Scheduling...\n");
	218
	219	tbl = (struct ip_vs_sh_bucket *)svc->sched_data;
20971a0a	220	dest = ip_vs_sh_get(svc->af, tbl, &iph.saddr);
1da177e4 LT	221	if (!dest
	222	\|\| !(dest->flags & IP_VS_DEST_F_AVAILABLE)
	223	\|\| atomic_read(&dest->weight) <= 0
	224	\|\| is_overloaded(dest)) {
68888d10	225	IP_VS_ERR_RL("SH: no destination available\n");
1da177e4 LT	226	return NULL;
	227	}
	228
20971a0a JV	229	IP_VS_DBG_BUF(6, "SH: source IP address %s --> server %s:%d\n",
	230	IP_VS_DBG_ADDR(svc->af, &iph.saddr),
	231	IP_VS_DBG_ADDR(svc->af, &dest->addr),
	232	ntohs(dest->port));
1da177e4 LT	233
	234	return dest;
	235	}
	236
	237
	238	/*
	239	* IPVS SH Scheduler structure
	240	*/
	241	static struct ip_vs_scheduler ip_vs_sh_scheduler =
	242	{
	243	.name = "sh",
	244	.refcnt = ATOMIC_INIT(0),
	245	.module = THIS_MODULE,
d149ccc9	246	.n_list = LIST_HEAD_INIT(ip_vs_sh_scheduler.n_list),
1da177e4 LT	247	.init_service = ip_vs_sh_init_svc,
	248	.done_service = ip_vs_sh_done_svc,
	249	.update_service = ip_vs_sh_update_svc,
	250	.schedule = ip_vs_sh_schedule,
	251	};
	252
	253
	254	static int __init ip_vs_sh_init(void)
	255	{
1da177e4 LT	256	return register_ip_vs_scheduler(&ip_vs_sh_scheduler);
	257	}
	258
	259
	260	static void __exit ip_vs_sh_cleanup(void)
	261	{
	262	unregister_ip_vs_scheduler(&ip_vs_sh_scheduler);
	263	}
	264
	265
	266	module_init(ip_vs_sh_init);
	267	module_exit(ip_vs_sh_cleanup);
	268	MODULE_LICENSE("GPL");