ip_vs_lblcr.c

linux-2.6.15.6

 */
static inline struct ip_vs_lblcr_entry *
ip_vs_lblcr_get(struct ip_vs_lblcr_table *tbl, __u32 addr)
{
	unsigned hash;
	struct ip_vs_lblcr_entry *en;

	hash = ip_vs_lblcr_hashkey(addr);

	read_lock(&tbl->lock);

	list_for_each_entry(en, &tbl->bucket[hash], list) {
		if (en->addr == addr) {
			/* HIT */
			read_unlock(&tbl->lock);
			return en;
		}
	}

	read_unlock(&tbl->lock);

	return NULL;
}


/*
 *      Flush all the entries of the specified table.
 */
static void ip_vs_lblcr_flush(struct ip_vs_lblcr_table *tbl)
{
	int i;
	struct ip_vs_lblcr_entry *en, *nxt;

	for (i=0; i<IP_VS_LBLCR_TAB_SIZE; i++) {
		write_lock(&tbl->lock);
		list_for_each_entry_safe(en, nxt, &tbl->bucket[i], list) {
			ip_vs_lblcr_free(en);
			atomic_dec(&tbl->entries);
		}
		write_unlock(&tbl->lock);
	}
}


static inline void ip_vs_lblcr_full_check(struct ip_vs_lblcr_table *tbl)
{
	unsigned long now = jiffies;
	int i, j;
	struct ip_vs_lblcr_entry *en, *nxt;

	for (i=0, j=tbl->rover; i<IP_VS_LBLCR_TAB_SIZE; i++) {
		j = (j + 1) & IP_VS_LBLCR_TAB_MASK;

		write_lock(&tbl->lock);
		list_for_each_entry_safe(en, nxt, &tbl->bucket[j], list) {
			if (time_after(en->lastuse+sysctl_ip_vs_lblcr_expiration,
				       now))
				continue;

			ip_vs_lblcr_free(en);
			atomic_dec(&tbl->entries);
		}
		write_unlock(&tbl->lock);
	}
	tbl->rover = j;
}


/*
 *      Periodical timer handler for IPVS lblcr table
 *      It is used to collect stale entries when the number of entries
 *      exceeds the maximum size of the table.
 *
 *      Fixme: we probably need more complicated algorithm to collect
 *             entries that have not been used for a long time even
 *             if the number of entries doesn't exceed the maximum size
 *             of the table.
 *      The full expiration check is for this purpose now.
 */
static void ip_vs_lblcr_check_expire(unsigned long data)
{
	struct ip_vs_lblcr_table *tbl;
	unsigned long now = jiffies;
	int goal;
	int i, j;
	struct ip_vs_lblcr_entry *en, *nxt;

	tbl = (struct ip_vs_lblcr_table *)data;

	if ((tbl->counter % COUNT_FOR_FULL_EXPIRATION) == 0) {
		/* do full expiration check */
		ip_vs_lblcr_full_check(tbl);
		tbl->counter = 1;
		goto out;
	}

	if (atomic_read(&tbl->entries) <= tbl->max_size) {
		tbl->counter++;
		goto out;
	}

	goal = (atomic_read(&tbl->entries) - tbl->max_size)*4/3;
	if (goal > tbl->max_size/2)
		goal = tbl->max_size/2;

	for (i=0, j=tbl->rover; i<IP_VS_LBLCR_TAB_SIZE; i++) {
		j = (j + 1) & IP_VS_LBLCR_TAB_MASK;

		write_lock(&tbl->lock);
		list_for_each_entry_safe(en, nxt, &tbl->bucket[j], list) {
			if (time_before(now, en->lastuse+ENTRY_TIMEOUT))
				continue;

			ip_vs_lblcr_free(en);
			atomic_dec(&tbl->entries);
			goal--;
		}
		write_unlock(&tbl->lock);
		if (goal <= 0)
			break;
	}
	tbl->rover = j;

  out:
	mod_timer(&tbl->periodic_timer, jiffies+CHECK_EXPIRE_INTERVAL);
}


#ifdef CONFIG_IP_VS_LBLCR_DEBUG
static struct ip_vs_lblcr_table *lblcr_table_list;

/*
 *	/proc/net/ip_vs_lblcr to display the mappings of
 *                  destination IP address <==> its serverSet
 */
static int
ip_vs_lblcr_getinfo(char *buffer, char **start, off_t offset, int length)
{
	off_t pos=0, begin;
	int len=0, size;
	struct ip_vs_lblcr_table *tbl;
	unsigned long now = jiffies;
	int i;
	struct ip_vs_lblcr_entry *en;

	tbl = lblcr_table_list;

	size = sprintf(buffer, "LastTime Dest IP address  Server set\n");
	pos += size;
	len += size;

	for (i=0; i<IP_VS_LBLCR_TAB_SIZE; i++) {
		read_lock_bh(&tbl->lock);
		list_for_each_entry(en, &tbl->bucket[i], list) {
			char tbuf[16];
			struct ip_vs_dest_list *d;

			sprintf(tbuf, "%u.%u.%u.%u", NIPQUAD(en->addr));
			size = sprintf(buffer+len, "%8lu %-16s ",
				       now-en->lastuse, tbuf);

			read_lock(&en->set.lock);
			for (d=en->set.list; d!=NULL; d=d->next) {
				size += sprintf(buffer+len+size,
						"%u.%u.%u.%u ",
						NIPQUAD(d->dest->addr));
			}
			read_unlock(&en->set.lock);
			size += sprintf(buffer+len+size, "\n");
			len += size;
			pos += size;
			if (pos <= offset)
				len=0;
			if (pos >= offset+length) {
				read_unlock_bh(&tbl->lock);
				goto done;
			}
		}
		read_unlock_bh(&tbl->lock);
	}

  done:
	begin = len - (pos - offset);
	*start = buffer + begin;
	len -= begin;
	if(len>length)
		len = length;
	return len;
}
#endif


static int ip_vs_lblcr_init_svc(struct ip_vs_service *svc)
{
	int i;
	struct ip_vs_lblcr_table *tbl;

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

	/*
	 *    Initialize the hash buckets
	 */
	for (i=0; i<IP_VS_LBLCR_TAB_SIZE; i++) {
		INIT_LIST_HEAD(&tbl->bucket[i]);
	}
	rwlock_init(&tbl->lock);
	tbl->max_size = IP_VS_LBLCR_TAB_SIZE*16;
	tbl->rover = 0;
	tbl->counter = 1;

	/*
	 *    Hook periodic timer for garbage collection
	 */
	init_timer(&tbl->periodic_timer);
	tbl->periodic_timer.data = (unsigned long)tbl;
	tbl->periodic_timer.function = ip_vs_lblcr_check_expire;
	tbl->periodic_timer.expires = jiffies+CHECK_EXPIRE_INTERVAL;
	add_timer(&tbl->periodic_timer);

#ifdef CONFIG_IP_VS_LBLCR_DEBUG
	lblcr_table_list = tbl;
#endif
	return 0;
}


static int ip_vs_lblcr_done_svc(struct ip_vs_service *svc)
{
	struct ip_vs_lblcr_table *tbl = svc->sched_data;

	/* remove periodic timer */
	del_timer_sync(&tbl->periodic_timer);

	/* got to clean up table entries here */
	ip_vs_lblcr_flush(tbl);

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

	return 0;
}


static int ip_vs_lblcr_update_svc(struct ip_vs_service *svc)
{
	return 0;
}


static inline struct ip_vs_dest *
__ip_vs_wlc_schedule(struct ip_vs_service *svc, struct iphdr *iph)
{
	struct ip_vs_dest *dest, *least;
	int loh, doh;

	/*
	 * We think the overhead of processing active connections is fifty
	 * times higher than that of inactive connections in average. (This
	 * fifty times might not be accurate, we will change it later.) We
	 * use the following formula to estimate the overhead:
	 *                dest->activeconns*50 + dest->inactconns
	 * and the load:
	 *                (dest overhead) / dest->weight
	 *
	 * Remember -- no floats in kernel mode!!!
	 * The comparison of h1*w2 > h2*w1 is equivalent to that of
	 *                h1/w1 > h2/w2
	 * if every weight is larger than zero.
	 *
	 * The server with weight=0 is quiesced and will not receive any
	 * new connection.
	 */
	list_for_each_entry(dest, &svc->destinations, n_list) {
		if (dest->flags & IP_VS_DEST_F_OVERLOAD)
			continue;

		if (atomic_read(&dest->weight) > 0) {
			least = dest;
			loh = atomic_read(&least->activeconns) * 50
				+ atomic_read(&least->inactconns);
			goto nextstage;
		}
	}
	return NULL;

	/*
	 *    Find the destination with the least load.
	 */
  nextstage:
	list_for_each_entry_continue(dest, &svc->destinations, n_list) {
		if (dest->flags & IP_VS_DEST_F_OVERLOAD)
			continue;

		doh = atomic_read(&dest->activeconns) * 50
			+ atomic_read(&dest->inactconns);
		if (loh * atomic_read(&dest->weight) >
		    doh * atomic_read(&least->weight)) {
			least = dest;
			loh = doh;
		}
	}

	IP_VS_DBG(6, "LBLCR: server %d.%d.%d.%d:%d "
		  "activeconns %d refcnt %d weight %d overhead %d\n",
		  NIPQUAD(least->addr), ntohs(least->port),
		  atomic_read(&least->activeconns),
		  atomic_read(&least->refcnt),
		  atomic_read(&least->weight), loh);

	return least;
}


/*
 *   If this destination server is overloaded and there is a less loaded
 *   server, then return true.
 */
static inline int
is_overloaded(struct ip_vs_dest *dest, struct ip_vs_service *svc)
{
	if (atomic_read(&dest->activeconns) > atomic_read(&dest->weight)) {
		struct ip_vs_dest *d;

		list_for_each_entry(d, &svc->destinations, n_list) {
			if (atomic_read(&d->activeconns)*2
			    < atomic_read(&d->weight)) {
				return 1;
			}
		}
	}
	return 0;
}


/*
 *    Locality-Based (weighted) Least-Connection scheduling
 */
static struct ip_vs_dest *
ip_vs_lblcr_schedule(struct ip_vs_service *svc, const struct sk_buff *skb)
{
	struct ip_vs_dest *dest;
	struct ip_vs_lblcr_table *tbl;
	struct ip_vs_lblcr_entry *en;
	struct iphdr *iph = skb->nh.iph;

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

	tbl = (struct ip_vs_lblcr_table *)svc->sched_data;
	en = ip_vs_lblcr_get(tbl, iph->daddr);
	if (en == NULL) {
		dest = __ip_vs_wlc_schedule(svc, iph);
		if (dest == NULL) {
			IP_VS_DBG(1, "no destination available\n");
			return NULL;
		}
		en = ip_vs_lblcr_new(iph->daddr);
		if (en == NULL) {
			return NULL;
		}
		ip_vs_dest_set_insert(&en->set, dest);
		ip_vs_lblcr_hash(tbl, en);
	} else {
		dest = ip_vs_dest_set_min(&en->set);
		if (!dest || is_overloaded(dest, svc)) {
			dest = __ip_vs_wlc_schedule(svc, iph);
			if (dest == NULL) {
				IP_VS_DBG(1, "no destination available\n");
				return NULL;
			}
			ip_vs_dest_set_insert(&en->set, dest);
		}
		if (atomic_read(&en->set.size) > 1 &&
		    jiffies-en->set.lastmod > sysctl_ip_vs_lblcr_expiration) {
			struct ip_vs_dest *m;
			m = ip_vs_dest_set_max(&en->set);
			if (m)
				ip_vs_dest_set_erase(&en->set, m);
		}
	}
	en->lastuse = jiffies;

	IP_VS_DBG(6, "LBLCR: destination IP address %u.%u.%u.%u "
		  "--> server %u.%u.%u.%u:%d\n",
		  NIPQUAD(en->addr),
		  NIPQUAD(dest->addr),
		  ntohs(dest->port));

	return dest;
}


/*
 *      IPVS LBLCR Scheduler structure
 */
static struct ip_vs_scheduler ip_vs_lblcr_scheduler =
{
	.name =			"lblcr",
	.refcnt =		ATOMIC_INIT(0),
	.module =		THIS_MODULE,
	.init_service =		ip_vs_lblcr_init_svc,
	.done_service =		ip_vs_lblcr_done_svc,
	.update_service =	ip_vs_lblcr_update_svc,
	.schedule =		ip_vs_lblcr_schedule,
};


static int __init ip_vs_lblcr_init(void)
{
	INIT_LIST_HEAD(&ip_vs_lblcr_scheduler.n_list);
	sysctl_header = register_sysctl_table(lblcr_root_table, 0);
#ifdef CONFIG_IP_VS_LBLCR_DEBUG
	proc_net_create("ip_vs_lblcr", 0, ip_vs_lblcr_getinfo);
#endif
	return register_ip_vs_scheduler(&ip_vs_lblcr_scheduler);
}


static void __exit ip_vs_lblcr_cleanup(void)
{
#ifdef CONFIG_IP_VS_LBLCR_DEBUG
	proc_net_remove("ip_vs_lblcr");
#endif
	unregister_sysctl_table(sysctl_header);
	unregister_ip_vs_scheduler(&ip_vs_lblcr_scheduler);
}


module_init(ip_vs_lblcr_init);
module_exit(ip_vs_lblcr_cleanup);
MODULE_LICENSE("GPL");