[deliverable/linux.git] / drivers / staging / lustre / lustre / lov / lov_pool.c

/*
 * GPL HEADER START
 *
 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 only,
 * as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License version 2 for more details (a copy is included
 * in the LICENSE file that accompanied this code).
 *
 * You should have received a copy of the GNU General Public License
 * version 2 along with this program; If not, see [sun.com URL with a
 * copy of GPLv2].
 *
 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
 * CA 95054 USA or visit www.sun.com if you need additional information or
 * have any questions.
 *
 * GPL HEADER END
 */
/*
 * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
 * Use is subject to license terms.
 *
 * Copyright (c) 2012, Intel Corporation.
 */
/*
 * This file is part of Lustre, http://www.lustre.org/
 * Lustre is a trademark of Sun Microsystems, Inc.
 *
 * lustre/lov/lov_pool.c
 *
 * OST pool methods
 *
 * Author: Jacques-Charles LAFOUCRIERE <jc.lafoucriere@cea.fr>
 * Author: Alex Lyashkov <Alexey.Lyashkov@Sun.COM>
 * Author: Nathaniel Rutman <Nathan.Rutman@Sun.COM>
 */

#define DEBUG_SUBSYSTEM S_LOV

#include <linux/libcfs/libcfs.h>

#include <obd.h>
#include "lov_internal.h"

#define pool_tgt(_p, _i) \
		_p->pool_lobd->u.lov.lov_tgts[_p->pool_obds.op_array[_i]]

static void lov_pool_getref(struct pool_desc *pool)
{
	CDEBUG(D_INFO, "pool %p\n", pool);
	atomic_inc(&pool->pool_refcount);
}

void lov_pool_putref(struct pool_desc *pool)
{
	CDEBUG(D_INFO, "pool %p\n", pool);
	if (atomic_dec_and_test(&pool->pool_refcount)) {
		LASSERT(hlist_unhashed(&pool->pool_hash));
		LASSERT(list_empty(&pool->pool_list));
		LASSERT(pool->pool_proc_entry == NULL);
		lov_ost_pool_free(&(pool->pool_rr.lqr_pool));
		lov_ost_pool_free(&(pool->pool_obds));
		OBD_FREE_PTR(pool);
	}
}

void lov_pool_putref_locked(struct pool_desc *pool)
{
	CDEBUG(D_INFO, "pool %p\n", pool);
	LASSERT(atomic_read(&pool->pool_refcount) > 1);

	atomic_dec(&pool->pool_refcount);
}

/*
 * hash function using a Rotating Hash algorithm
 * Knuth, D. The Art of Computer Programming,
 * Volume 3: Sorting and Searching,
 * Chapter 6.4.
 * Addison Wesley, 1973
 */
static __u32 pool_hashfn(cfs_hash_t *hash_body, const void *key, unsigned mask)
{
	int i;
	__u32 result;
	char *poolname;

	result = 0;
	poolname = (char *)key;
	for (i = 0; i < LOV_MAXPOOLNAME; i++) {
		if (poolname[i] == '\0')
			break;
		result = (result << 4)^(result >> 28) ^  poolname[i];
	}
	return (result % mask);
}

static void *pool_key(struct hlist_node *hnode)
{
	struct pool_desc *pool;

	pool = hlist_entry(hnode, struct pool_desc, pool_hash);
	return (pool->pool_name);
}

static int pool_hashkey_keycmp(const void *key, struct hlist_node *compared_hnode)
{
	char *pool_name;
	struct pool_desc *pool;

	pool_name = (char *)key;
	pool = hlist_entry(compared_hnode, struct pool_desc, pool_hash);
	return !strncmp(pool_name, pool->pool_name, LOV_MAXPOOLNAME);
}

static void *pool_hashobject(struct hlist_node *hnode)
{
	return hlist_entry(hnode, struct pool_desc, pool_hash);
}

static void pool_hashrefcount_get(cfs_hash_t *hs, struct hlist_node *hnode)
{
	struct pool_desc *pool;

	pool = hlist_entry(hnode, struct pool_desc, pool_hash);
	lov_pool_getref(pool);
}

static void pool_hashrefcount_put_locked(cfs_hash_t *hs,
					 struct hlist_node *hnode)
{
	struct pool_desc *pool;

	pool = hlist_entry(hnode, struct pool_desc, pool_hash);
	lov_pool_putref_locked(pool);
}

cfs_hash_ops_t pool_hash_operations = {
	.hs_hash	= pool_hashfn,
	.hs_key	 = pool_key,
	.hs_keycmp      = pool_hashkey_keycmp,
	.hs_object      = pool_hashobject,
	.hs_get	 = pool_hashrefcount_get,
	.hs_put_locked  = pool_hashrefcount_put_locked,

};

#ifdef LPROCFS
/* ifdef needed for liblustre support */
/*
 * pool /proc seq_file methods
 */
/*
 * iterator is used to go through the target pool entries
 * index is the current entry index in the lp_array[] array
 * index >= pos returned to the seq_file interface
 * pos is from 0 to (pool->pool_obds.op_count - 1)
 */
#define POOL_IT_MAGIC 0xB001CEA0
struct pool_iterator {
	int magic;
	struct pool_desc *pool;
	int idx;	/* from 0 to pool_tgt_size - 1 */
};

static void *pool_proc_next(struct seq_file *s, void *v, loff_t *pos)
{
	struct pool_iterator *iter = (struct pool_iterator *)s->private;
	int prev_idx;

	LASSERTF(iter->magic == POOL_IT_MAGIC, "%08X", iter->magic);

	/* test if end of file */
	if (*pos >= pool_tgt_count(iter->pool))
		return NULL;

	/* iterate to find a non empty entry */
	prev_idx = iter->idx;
	down_read(&pool_tgt_rw_sem(iter->pool));
	iter->idx++;
	if (iter->idx == pool_tgt_count(iter->pool)) {
		iter->idx = prev_idx; /* we stay on the last entry */
		up_read(&pool_tgt_rw_sem(iter->pool));
		return NULL;
	}
	up_read(&pool_tgt_rw_sem(iter->pool));
	(*pos)++;
	/* return != NULL to continue */
	return iter;
}

static void *pool_proc_start(struct seq_file *s, loff_t *pos)
{
	struct pool_desc *pool = (struct pool_desc *)s->private;
	struct pool_iterator *iter;

	lov_pool_getref(pool);
	if ((pool_tgt_count(pool) == 0) ||
	    (*pos >= pool_tgt_count(pool))) {
		/* iter is not created, so stop() has no way to
		 * find pool to dec ref */
		lov_pool_putref(pool);
		return NULL;
	}

	OBD_ALLOC_PTR(iter);
	if (!iter)
		return ERR_PTR(-ENOMEM);
	iter->magic = POOL_IT_MAGIC;
	iter->pool = pool;
	iter->idx = 0;

	/* we use seq_file private field to memorized iterator so
	 * we can free it at stop() */
	/* /!\ do not forget to restore it to pool before freeing it */
	s->private = iter;
	if (*pos > 0) {
		loff_t i;
		void *ptr;

		i = 0;
		do {
		     ptr = pool_proc_next(s, &iter, &i);
		} while ((i < *pos) && (ptr != NULL));
		return ptr;
	}
	return iter;
}

static void pool_proc_stop(struct seq_file *s, void *v)
{
	struct pool_iterator *iter = (struct pool_iterator *)s->private;

	/* in some cases stop() method is called 2 times, without
	 * calling start() method (see seq_read() from fs/seq_file.c)
	 * we have to free only if s->private is an iterator */
	if ((iter) && (iter->magic == POOL_IT_MAGIC)) {
		/* we restore s->private so next call to pool_proc_start()
		 * will work */
		s->private = iter->pool;
		lov_pool_putref(iter->pool);
		OBD_FREE_PTR(iter);
	}
	return;
}

static int pool_proc_show(struct seq_file *s, void *v)
{
	struct pool_iterator *iter = (struct pool_iterator *)v;
	struct lov_tgt_desc *tgt;

	LASSERTF(iter->magic == POOL_IT_MAGIC, "%08X", iter->magic);
	LASSERT(iter->pool != NULL);
	LASSERT(iter->idx <= pool_tgt_count(iter->pool));

	down_read(&pool_tgt_rw_sem(iter->pool));
	tgt = pool_tgt(iter->pool, iter->idx);
	up_read(&pool_tgt_rw_sem(iter->pool));
	if (tgt)
		seq_printf(s, "%s\n", obd_uuid2str(&(tgt->ltd_uuid)));

	return 0;
}

static struct seq_operations pool_proc_ops = {
	.start	  = pool_proc_start,
	.next	   = pool_proc_next,
	.stop	   = pool_proc_stop,
	.show	   = pool_proc_show,
};

static int pool_proc_open(struct inode *inode, struct file *file)
{
	int rc;

	rc = seq_open(file, &pool_proc_ops);
	if (!rc) {
		struct seq_file *s = file->private_data;
		s->private = PDE_DATA(inode);
	}
	return rc;
}

static struct file_operations pool_proc_operations = {
	.open	   = pool_proc_open,
	.read	   = seq_read,
	.llseek	 = seq_lseek,
	.release	= seq_release,
};
#endif /* LPROCFS */

void lov_dump_pool(int level, struct pool_desc *pool)
{
	int i;

	lov_pool_getref(pool);

	CDEBUG(level, "pool "LOV_POOLNAMEF" has %d members\n",
	       pool->pool_name, pool->pool_obds.op_count);
	down_read(&pool_tgt_rw_sem(pool));

	for (i = 0; i < pool_tgt_count(pool) ; i++) {
		if (!pool_tgt(pool, i) || !(pool_tgt(pool, i))->ltd_exp)
			continue;
		CDEBUG(level, "pool "LOV_POOLNAMEF"[%d] = %s\n",
		       pool->pool_name, i,
		       obd_uuid2str(&((pool_tgt(pool, i))->ltd_uuid)));
	}

	up_read(&pool_tgt_rw_sem(pool));
	lov_pool_putref(pool);
}

#define LOV_POOL_INIT_COUNT 2
int lov_ost_pool_init(struct ost_pool *op, unsigned int count)
{
	if (count == 0)
		count = LOV_POOL_INIT_COUNT;
	op->op_array = NULL;
	op->op_count = 0;
	init_rwsem(&op->op_rw_sem);
	op->op_size = count;
	OBD_ALLOC(op->op_array, op->op_size * sizeof(op->op_array[0]));
	if (op->op_array == NULL) {
		op->op_size = 0;
		return -ENOMEM;
	}
	return 0;
}

/* Caller must hold write op_rwlock */
int lov_ost_pool_extend(struct ost_pool *op, unsigned int min_count)
{
	__u32 *new;
	int new_size;

	LASSERT(min_count != 0);

	if (op->op_count < op->op_size)
		return 0;

	new_size = max(min_count, 2 * op->op_size);
	OBD_ALLOC(new, new_size * sizeof(op->op_array[0]));
	if (new == NULL)
		return -ENOMEM;

	/* copy old array to new one */
	memcpy(new, op->op_array, op->op_size * sizeof(op->op_array[0]));
	OBD_FREE(op->op_array, op->op_size * sizeof(op->op_array[0]));
	op->op_array = new;
	op->op_size = new_size;
	return 0;
}

int lov_ost_pool_add(struct ost_pool *op, __u32 idx, unsigned int min_count)
{
	int rc = 0, i;

	down_write(&op->op_rw_sem);

	rc = lov_ost_pool_extend(op, min_count);
	if (rc)
		GOTO(out, rc);

	/* search ost in pool array */
	for (i = 0; i < op->op_count; i++) {
		if (op->op_array[i] == idx)
			GOTO(out, rc = -EEXIST);
	}
	/* ost not found we add it */
	op->op_array[op->op_count] = idx;
	op->op_count++;
out:
	up_write(&op->op_rw_sem);
	return rc;
}

int lov_ost_pool_remove(struct ost_pool *op, __u32 idx)
{
	int i;

	down_write(&op->op_rw_sem);

	for (i = 0; i < op->op_count; i++) {
		if (op->op_array[i] == idx) {
			memmove(&op->op_array[i], &op->op_array[i + 1],
				(op->op_count - i - 1) * sizeof(op->op_array[0]));
			op->op_count--;
			up_write(&op->op_rw_sem);
			return 0;
		}
	}

	up_write(&op->op_rw_sem);
	return -EINVAL;
}

int lov_ost_pool_free(struct ost_pool *op)
{
	if (op->op_size == 0)
		return 0;

	down_write(&op->op_rw_sem);

	OBD_FREE(op->op_array, op->op_size * sizeof(op->op_array[0]));
	op->op_array = NULL;
	op->op_count = 0;
	op->op_size = 0;

	up_write(&op->op_rw_sem);
	return 0;
}


int lov_pool_new(struct obd_device *obd, char *poolname)
{
	struct lov_obd *lov;
	struct pool_desc *new_pool;
	int rc;

	lov = &(obd->u.lov);

	if (strlen(poolname) > LOV_MAXPOOLNAME)
		return -ENAMETOOLONG;

	OBD_ALLOC_PTR(new_pool);
	if (new_pool == NULL)
		return -ENOMEM;

	strncpy(new_pool->pool_name, poolname, LOV_MAXPOOLNAME);
	new_pool->pool_name[LOV_MAXPOOLNAME] = '\0';
	new_pool->pool_lobd = obd;
	/* ref count init to 1 because when created a pool is always used
	 * up to deletion
	 */
	atomic_set(&new_pool->pool_refcount, 1);
	rc = lov_ost_pool_init(&new_pool->pool_obds, 0);
	if (rc)
	       GOTO(out_err, rc);

	memset(&(new_pool->pool_rr), 0, sizeof(struct lov_qos_rr));
	rc = lov_ost_pool_init(&new_pool->pool_rr.lqr_pool, 0);
	if (rc)
		GOTO(out_free_pool_obds, rc);

	INIT_HLIST_NODE(&new_pool->pool_hash);

#ifdef LPROCFS
	/* we need this assert seq_file is not implementated for liblustre */
	/* get ref for /proc file */
	lov_pool_getref(new_pool);
	new_pool->pool_proc_entry = lprocfs_add_simple(lov->lov_pool_proc_entry,
						       poolname, new_pool,
						       &pool_proc_operations);
	if (IS_ERR(new_pool->pool_proc_entry)) {
		CWARN("Cannot add proc pool entry "LOV_POOLNAMEF"\n", poolname);
		new_pool->pool_proc_entry = NULL;
		lov_pool_putref(new_pool);
	}
	CDEBUG(D_INFO, "pool %p - proc %p\n", new_pool, new_pool->pool_proc_entry);
#endif

	spin_lock(&obd->obd_dev_lock);
	list_add_tail(&new_pool->pool_list, &lov->lov_pool_list);
	lov->lov_pool_count++;
	spin_unlock(&obd->obd_dev_lock);

	/* add to find only when it fully ready  */
	rc = cfs_hash_add_unique(lov->lov_pools_hash_body, poolname,
				 &new_pool->pool_hash);
	if (rc)
		GOTO(out_err, rc = -EEXIST);

	CDEBUG(D_CONFIG, LOV_POOLNAMEF" is pool #%d\n",
	       poolname, lov->lov_pool_count);

	return 0;

out_err:
	spin_lock(&obd->obd_dev_lock);
	list_del_init(&new_pool->pool_list);
	lov->lov_pool_count--;
	spin_unlock(&obd->obd_dev_lock);

	lprocfs_remove(&new_pool->pool_proc_entry);

	lov_ost_pool_free(&new_pool->pool_rr.lqr_pool);
out_free_pool_obds:
	lov_ost_pool_free(&new_pool->pool_obds);
	OBD_FREE_PTR(new_pool);
	return rc;
}

int lov_pool_del(struct obd_device *obd, char *poolname)
{
	struct lov_obd *lov;
	struct pool_desc *pool;

	lov = &(obd->u.lov);

	/* lookup and kill hash reference */
	pool = cfs_hash_del_key(lov->lov_pools_hash_body, poolname);
	if (pool == NULL)
		return -ENOENT;

	if (pool->pool_proc_entry != NULL) {
		CDEBUG(D_INFO, "proc entry %p\n", pool->pool_proc_entry);
		lprocfs_remove(&pool->pool_proc_entry);
		lov_pool_putref(pool);
	}

	spin_lock(&obd->obd_dev_lock);
	list_del_init(&pool->pool_list);
	lov->lov_pool_count--;
	spin_unlock(&obd->obd_dev_lock);

	/* release last reference */
	lov_pool_putref(pool);

	return 0;
}


int lov_pool_add(struct obd_device *obd, char *poolname, char *ostname)
{
	struct obd_uuid ost_uuid;
	struct lov_obd *lov;
	struct pool_desc *pool;
	unsigned int lov_idx;
	int rc;

	lov = &(obd->u.lov);

	pool = cfs_hash_lookup(lov->lov_pools_hash_body, poolname);
	if (pool == NULL)
		return -ENOENT;

	obd_str2uuid(&ost_uuid, ostname);


	/* search ost in lov array */
	obd_getref(obd);
	for (lov_idx = 0; lov_idx < lov->desc.ld_tgt_count; lov_idx++) {
		if (!lov->lov_tgts[lov_idx])
			continue;
		if (obd_uuid_equals(&ost_uuid,
				    &(lov->lov_tgts[lov_idx]->ltd_uuid)))
			break;
	}
	/* test if ost found in lov */
	if (lov_idx == lov->desc.ld_tgt_count)
		GOTO(out, rc = -EINVAL);

	rc = lov_ost_pool_add(&pool->pool_obds, lov_idx, lov->lov_tgt_size);
	if (rc)
		GOTO(out, rc);

	pool->pool_rr.lqr_dirty = 1;

	CDEBUG(D_CONFIG, "Added %s to "LOV_POOLNAMEF" as member %d\n",
	       ostname, poolname,  pool_tgt_count(pool));

out:
	obd_putref(obd);
	lov_pool_putref(pool);
	return rc;
}

int lov_pool_remove(struct obd_device *obd, char *poolname, char *ostname)
{
	struct obd_uuid ost_uuid;
	struct lov_obd *lov;
	struct pool_desc *pool;
	unsigned int lov_idx;
	int rc = 0;

	lov = &(obd->u.lov);

	pool = cfs_hash_lookup(lov->lov_pools_hash_body, poolname);
	if (pool == NULL)
		return -ENOENT;

	obd_str2uuid(&ost_uuid, ostname);

	obd_getref(obd);
	/* search ost in lov array, to get index */
	for (lov_idx = 0; lov_idx < lov->desc.ld_tgt_count; lov_idx++) {
		if (!lov->lov_tgts[lov_idx])
			continue;

		if (obd_uuid_equals(&ost_uuid,
				    &(lov->lov_tgts[lov_idx]->ltd_uuid)))
			break;
	}

	/* test if ost found in lov */
	if (lov_idx == lov->desc.ld_tgt_count)
		GOTO(out, rc = -EINVAL);

	lov_ost_pool_remove(&pool->pool_obds, lov_idx);

	pool->pool_rr.lqr_dirty = 1;

	CDEBUG(D_CONFIG, "%s removed from "LOV_POOLNAMEF"\n", ostname,
	       poolname);

out:
	obd_putref(obd);
	lov_pool_putref(pool);
	return rc;
}

int lov_check_index_in_pool(__u32 idx, struct pool_desc *pool)
{
	int i, rc;

	/* caller may no have a ref on pool if it got the pool
	 * without calling lov_find_pool() (e.g. go through the lov pool
	 * list)
	 */
	lov_pool_getref(pool);

	down_read(&pool_tgt_rw_sem(pool));

	for (i = 0; i < pool_tgt_count(pool); i++) {
		if (pool_tgt_array(pool)[i] == idx)
			GOTO(out, rc = 0);
	}
	rc = -ENOENT;
out:
	up_read(&pool_tgt_rw_sem(pool));

	lov_pool_putref(pool);
	return rc;
}

struct pool_desc *lov_find_pool(struct lov_obd *lov, char *poolname)
{
	struct pool_desc *pool;

	pool = NULL;
	if (poolname[0] != '\0') {
		pool = cfs_hash_lookup(lov->lov_pools_hash_body, poolname);
		if (pool == NULL)
			CWARN("Request for an unknown pool ("LOV_POOLNAMEF")\n",
			      poolname);
		if ((pool != NULL) && (pool_tgt_count(pool) == 0)) {
			CWARN("Request for an empty pool ("LOV_POOLNAMEF")\n",
			       poolname);
			/* pool is ignored, so we remove ref on it */
			lov_pool_putref(pool);
			pool = NULL;
		}
	}
	return pool;
}
Commit	Line	Data
d7e09d03 PT	1	/*
	2	* GPL HEADER START
	3	*
	4	* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
	5	*
	6	* This program is free software; you can redistribute it and/or modify
	7	* it under the terms of the GNU General Public License version 2 only,
	8	* as published by the Free Software Foundation.
	9	*
	10	* This program is distributed in the hope that it will be useful, but
	11	* WITHOUT ANY WARRANTY; without even the implied warranty of
	12	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	13	* General Public License version 2 for more details (a copy is included
	14	* in the LICENSE file that accompanied this code).
	15	*
	16	* You should have received a copy of the GNU General Public License
	17	* version 2 along with this program; If not, see [sun.com URL with a
	18	* copy of GPLv2].
	19	*
	20	* Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
	21	* CA 95054 USA or visit www.sun.com if you need additional information or
	22	* have any questions.
	23	*
	24	* GPL HEADER END
	25	*/
	26	/*
	27	* Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
	28	* Use is subject to license terms.
	29	*
	30	* Copyright (c) 2012, Intel Corporation.
	31	*/
	32	/*
	33	* This file is part of Lustre, http://www.lustre.org/
	34	* Lustre is a trademark of Sun Microsystems, Inc.
	35	*
	36	* lustre/lov/lov_pool.c
	37	*
	38	* OST pool methods
	39	*
	40	* Author: Jacques-Charles LAFOUCRIERE <jc.lafoucriere@cea.fr>
	41	* Author: Alex Lyashkov <Alexey.Lyashkov@Sun.COM>
	42	* Author: Nathaniel Rutman <Nathan.Rutman@Sun.COM>
	43	*/
	44
	45	#define DEBUG_SUBSYSTEM S_LOV
	46
	47	#include <linux/libcfs/libcfs.h>
	48
	49	#include <obd.h>
	50	#include "lov_internal.h"
	51
	52	#define pool_tgt(_p, _i) \
	53	_p->pool_lobd->u.lov.lov_tgts[_p->pool_obds.op_array[_i]]
	54
	55	static void lov_pool_getref(struct pool_desc *pool)
	56	{
	57	CDEBUG(D_INFO, "pool %p\n", pool);
	58	atomic_inc(&pool->pool_refcount);
	59	}
	60
	61	void lov_pool_putref(struct pool_desc *pool)
	62	{
	63	CDEBUG(D_INFO, "pool %p\n", pool);
	64	if (atomic_dec_and_test(&pool->pool_refcount)) {
65	LASSERT(hlist_unhashed(&pool->pool_hash));
66	LASSERT(list_empty(&pool->pool_list));
67	LASSERT(pool->pool_proc_entry == NULL);
68	lov_ost_pool_free(&(pool->pool_rr.lqr_pool));
69	lov_ost_pool_free(&(pool->pool_obds));
70	OBD_FREE_PTR(pool);
d7e09d03 PT	71	}
	72	}
	73
	74	void lov_pool_putref_locked(struct pool_desc *pool)
	75	{
	76	CDEBUG(D_INFO, "pool %p\n", pool);
	77	LASSERT(atomic_read(&pool->pool_refcount) > 1);
	78
	79	atomic_dec(&pool->pool_refcount);
	80	}
	81
	82	/*
	83	* hash function using a Rotating Hash algorithm
	84	* Knuth, D. The Art of Computer Programming,
	85	* Volume 3: Sorting and Searching,
	86	* Chapter 6.4.
	87	* Addison Wesley, 1973
	88	*/
	89	static __u32 pool_hashfn(cfs_hash_t hash_body, const void key, unsigned mask)
	90	{
	91	int i;
	92	__u32 result;
	93	char *poolname;
	94
	95	result = 0;
	96	poolname = (char *)key;
	97	for (i = 0; i < LOV_MAXPOOLNAME; i++) {
	98	if (poolname[i] == '\0')
	99	break;
	100	result = (result << 4)^(result >> 28) ^ poolname[i];
	101	}
	102	return (result % mask);
	103	}
	104
	105	static void pool_key(struct hlist_node hnode)
	106	{
	107	struct pool_desc *pool;
	108
	109	pool = hlist_entry(hnode, struct pool_desc, pool_hash);
	110	return (pool->pool_name);
	111	}
	112
	113	static int pool_hashkey_keycmp(const void key, struct hlist_node compared_hnode)
	114	{
	115	char *pool_name;
	116	struct pool_desc *pool;
	117
	118	pool_name = (char *)key;
	119	pool = hlist_entry(compared_hnode, struct pool_desc, pool_hash);
	120	return !strncmp(pool_name, pool->pool_name, LOV_MAXPOOLNAME);
	121	}
	122
	123	static void pool_hashobject(struct hlist_node hnode)
	124	{
	125	return hlist_entry(hnode, struct pool_desc, pool_hash);
	126	}
	127
	128	static void pool_hashrefcount_get(cfs_hash_t hs, struct hlist_node hnode)
	129	{
	130	struct pool_desc *pool;
	131
	132	pool = hlist_entry(hnode, struct pool_desc, pool_hash);
	133	lov_pool_getref(pool);
	134	}
135
136	static void pool_hashrefcount_put_locked(cfs_hash_t *hs,
137	struct hlist_node *hnode)
138	{
139	struct pool_desc *pool;
140
141	pool = hlist_entry(hnode, struct pool_desc, pool_hash);
142	lov_pool_putref_locked(pool);
143	}
144
145	cfs_hash_ops_t pool_hash_operations = {
146	.hs_hash = pool_hashfn,
147	.hs_key = pool_key,
148	.hs_keycmp = pool_hashkey_keycmp,
149	.hs_object = pool_hashobject,
150	.hs_get = pool_hashrefcount_get,
151	.hs_put_locked = pool_hashrefcount_put_locked,
152
153	};
154
155	#ifdef LPROCFS
156	/* ifdef needed for liblustre support */
157	/*
158	* pool /proc seq_file methods
159	*/
160	/*
161	* iterator is used to go through the target pool entries
162	* index is the current entry index in the lp_array[] array
163	* index >= pos returned to the seq_file interface
164	* pos is from 0 to (pool->pool_obds.op_count - 1)
165	*/
166	#define POOL_IT_MAGIC 0xB001CEA0
167	struct pool_iterator {
168	int magic;
169	struct pool_desc *pool;
170	int idx; /* from 0 to pool_tgt_size - 1 */
171	};
172
173	static void pool_proc_next(struct seq_file s, void v, loff_t pos)
174	{
175	struct pool_iterator iter = (struct pool_iterator )s->private;
176	int prev_idx;
177
178	LASSERTF(iter->magic == POOL_IT_MAGIC, "%08X", iter->magic);
179
180	/* test if end of file */
181	if (*pos >= pool_tgt_count(iter->pool))
182	return NULL;
183
184	/* iterate to find a non empty entry */
185	prev_idx = iter->idx;
186	down_read(&pool_tgt_rw_sem(iter->pool));
187	iter->idx++;
188	if (iter->idx == pool_tgt_count(iter->pool)) {
189	iter->idx = prev_idx; /* we stay on the last entry */
190	up_read(&pool_tgt_rw_sem(iter->pool));
191	return NULL;
192	}
193	up_read(&pool_tgt_rw_sem(iter->pool));
194	(*pos)++;
195	/* return != NULL to continue */
196	return iter;
197	}
198
199	static void pool_proc_start(struct seq_file s, loff_t *pos)
200	{
201	struct pool_desc pool = (struct pool_desc )s->private;
202	struct pool_iterator *iter;
203
204	lov_pool_getref(pool);
205	if ((pool_tgt_count(pool) == 0) \|\|
206	(*pos >= pool_tgt_count(pool))) {
207	/* iter is not created, so stop() has no way to
208	* find pool to dec ref */
209	lov_pool_putref(pool);
210	return NULL;
211	}
212
213	OBD_ALLOC_PTR(iter);
214	if (!iter)
215	return ERR_PTR(-ENOMEM);
216	iter->magic = POOL_IT_MAGIC;
217	iter->pool = pool;
218	iter->idx = 0;
219
220	/* we use seq_file private field to memorized iterator so
221	* we can free it at stop() */
222	/* /!\ do not forget to restore it to pool before freeing it */
223	s->private = iter;
224	if (*pos > 0) {
225	loff_t i;
226	void *ptr;
227
228	i = 0;
229	do {
230	ptr = pool_proc_next(s, &iter, &i);
231	} while ((i < *pos) && (ptr != NULL));
232	return ptr;
233	}
234	return iter;
235	}
236
237	static void pool_proc_stop(struct seq_file s, void v)
238	{
239	struct pool_iterator iter = (struct pool_iterator )s->private;
240
241	/* in some cases stop() method is called 2 times, without
242	* calling start() method (see seq_read() from fs/seq_file.c)
243	* we have to free only if s->private is an iterator */
244	if ((iter) && (iter->magic == POOL_IT_MAGIC)) {
245	/* we restore s->private so next call to pool_proc_start()
246	* will work */
247	s->private = iter->pool;
248	lov_pool_putref(iter->pool);
249	OBD_FREE_PTR(iter);
250	}
251	return;
252	}
253
254	static int pool_proc_show(struct seq_file s, void v)
255	{
256	struct pool_iterator iter = (struct pool_iterator )v;
257	struct lov_tgt_desc *tgt;
258
259	LASSERTF(iter->magic == POOL_IT_MAGIC, "%08X", iter->magic);
260	LASSERT(iter->pool != NULL);
261	LASSERT(iter->idx <= pool_tgt_count(iter->pool));
262
263	down_read(&pool_tgt_rw_sem(iter->pool));
264	tgt = pool_tgt(iter->pool, iter->idx);
265	up_read(&pool_tgt_rw_sem(iter->pool));
266	if (tgt)
267	seq_printf(s, "%s\n", obd_uuid2str(&(tgt->ltd_uuid)));
268
269	return 0;
270	}
271
272	static struct seq_operations pool_proc_ops = {
273	.start = pool_proc_start,
274	.next = pool_proc_next,
275	.stop = pool_proc_stop,
276	.show = pool_proc_show,
277	};
278
279	static int pool_proc_open(struct inode inode, struct file file)
280	{
281	int rc;
282
283	rc = seq_open(file, &pool_proc_ops);
284	if (!rc) {
285	struct seq_file *s = file->private_data;
73bb1da6	286	s->private = PDE_DATA(inode);
d7e09d03 PT	287	}
	288	return rc;
	289	}
	290
	291	static struct file_operations pool_proc_operations = {
	292	.open = pool_proc_open,
	293	.read = seq_read,
	294	.llseek = seq_lseek,
	295	.release = seq_release,
	296	};
	297	#endif /* LPROCFS */
	298
	299	void lov_dump_pool(int level, struct pool_desc *pool)
	300	{
	301	int i;
	302
	303	lov_pool_getref(pool);
	304
	305	CDEBUG(level, "pool "LOV_POOLNAMEF" has %d members\n",
	306	pool->pool_name, pool->pool_obds.op_count);
	307	down_read(&pool_tgt_rw_sem(pool));
	308
	309	for (i = 0; i < pool_tgt_count(pool) ; i++) {
	310	if (!pool_tgt(pool, i) \|\| !(pool_tgt(pool, i))->ltd_exp)
	311	continue;
	312	CDEBUG(level, "pool "LOV_POOLNAMEF"[%d] = %s\n",
	313	pool->pool_name, i,
	314	obd_uuid2str(&((pool_tgt(pool, i))->ltd_uuid)));
	315	}
	316
	317	up_read(&pool_tgt_rw_sem(pool));
	318	lov_pool_putref(pool);
	319	}
	320
	321	#define LOV_POOL_INIT_COUNT 2
	322	int lov_ost_pool_init(struct ost_pool *op, unsigned int count)
	323	{
d7e09d03 PT	324	if (count == 0)
	325	count = LOV_POOL_INIT_COUNT;
	326	op->op_array = NULL;
	327	op->op_count = 0;
	328	init_rwsem(&op->op_rw_sem);
	329	op->op_size = count;
	330	OBD_ALLOC(op->op_array, op->op_size * sizeof(op->op_array[0]));
	331	if (op->op_array == NULL) {
	332	op->op_size = 0;
0a3bdb00	333	return -ENOMEM;
d7e09d03	334	}
d7e09d03 PT	335	return 0;
	336	}
	337
	338	/* Caller must hold write op_rwlock */
	339	int lov_ost_pool_extend(struct ost_pool *op, unsigned int min_count)
	340	{
	341	__u32 *new;
	342	int new_size;
	343
	344	LASSERT(min_count != 0);
	345
	346	if (op->op_count < op->op_size)
	347	return 0;
	348
	349	new_size = max(min_count, 2 * op->op_size);
	350	OBD_ALLOC(new, new_size * sizeof(op->op_array[0]));
	351	if (new == NULL)
	352	return -ENOMEM;
	353
	354	/* copy old array to new one */
	355	memcpy(new, op->op_array, op->op_size * sizeof(op->op_array[0]));
	356	OBD_FREE(op->op_array, op->op_size * sizeof(op->op_array[0]));
	357	op->op_array = new;
	358	op->op_size = new_size;
	359	return 0;
	360	}
	361
	362	int lov_ost_pool_add(struct ost_pool *op, __u32 idx, unsigned int min_count)
	363	{
	364	int rc = 0, i;
d7e09d03 PT	365
	366	down_write(&op->op_rw_sem);
	367
	368	rc = lov_ost_pool_extend(op, min_count);
	369	if (rc)
	370	GOTO(out, rc);
	371
	372	/* search ost in pool array */
	373	for (i = 0; i < op->op_count; i++) {
	374	if (op->op_array[i] == idx)
	375	GOTO(out, rc = -EEXIST);
	376	}
	377	/* ost not found we add it */
	378	op->op_array[op->op_count] = idx;
	379	op->op_count++;
d7e09d03 PT	380	out:
	381	up_write(&op->op_rw_sem);
	382	return rc;
	383	}
	384
	385	int lov_ost_pool_remove(struct ost_pool *op, __u32 idx)
	386	{
	387	int i;
d7e09d03 PT	388
	389	down_write(&op->op_rw_sem);
	390
	391	for (i = 0; i < op->op_count; i++) {
	392	if (op->op_array[i] == idx) {
	393	memmove(&op->op_array[i], &op->op_array[i + 1],
	394	(op->op_count - i - 1) * sizeof(op->op_array[0]));
	395	op->op_count--;
	396	up_write(&op->op_rw_sem);
d7e09d03 PT	397	return 0;
	398	}
	399	}
	400
	401	up_write(&op->op_rw_sem);
0a3bdb00	402	return -EINVAL;
d7e09d03 PT	403	}
	404
	405	int lov_ost_pool_free(struct ost_pool *op)
	406	{
d7e09d03	407	if (op->op_size == 0)
0a3bdb00	408	return 0;
d7e09d03 PT	409
	410	down_write(&op->op_rw_sem);
	411
	412	OBD_FREE(op->op_array, op->op_size * sizeof(op->op_array[0]));
	413	op->op_array = NULL;
	414	op->op_count = 0;
	415	op->op_size = 0;
	416
	417	up_write(&op->op_rw_sem);
0a3bdb00	418	return 0;
d7e09d03 PT	419	}
	420
	421
	422	int lov_pool_new(struct obd_device obd, char poolname)
	423	{
	424	struct lov_obd *lov;
	425	struct pool_desc *new_pool;
	426	int rc;
d7e09d03 PT	427
	428	lov = &(obd->u.lov);
	429
	430	if (strlen(poolname) > LOV_MAXPOOLNAME)
0a3bdb00	431	return -ENAMETOOLONG;
d7e09d03 PT	432
	433	OBD_ALLOC_PTR(new_pool);
	434	if (new_pool == NULL)
0a3bdb00	435	return -ENOMEM;
d7e09d03 PT	436
	437	strncpy(new_pool->pool_name, poolname, LOV_MAXPOOLNAME);
	438	new_pool->pool_name[LOV_MAXPOOLNAME] = '\0';
	439	new_pool->pool_lobd = obd;
	440	/* ref count init to 1 because when created a pool is always used
	441	* up to deletion
	442	*/
	443	atomic_set(&new_pool->pool_refcount, 1);
	444	rc = lov_ost_pool_init(&new_pool->pool_obds, 0);
	445	if (rc)
	446	GOTO(out_err, rc);
	447
	448	memset(&(new_pool->pool_rr), 0, sizeof(struct lov_qos_rr));
	449	rc = lov_ost_pool_init(&new_pool->pool_rr.lqr_pool, 0);
	450	if (rc)
	451	GOTO(out_free_pool_obds, rc);
	452
	453	INIT_HLIST_NODE(&new_pool->pool_hash);
	454
	455	#ifdef LPROCFS
	456	/* we need this assert seq_file is not implementated for liblustre */
	457	/* get ref for /proc file */
	458	lov_pool_getref(new_pool);
	459	new_pool->pool_proc_entry = lprocfs_add_simple(lov->lov_pool_proc_entry,
73bb1da6	460	poolname, new_pool,
d7e09d03 PT	461	&pool_proc_operations);
	462	if (IS_ERR(new_pool->pool_proc_entry)) {
	463	CWARN("Cannot add proc pool entry "LOV_POOLNAMEF"\n", poolname);
	464	new_pool->pool_proc_entry = NULL;
	465	lov_pool_putref(new_pool);
	466	}
	467	CDEBUG(D_INFO, "pool %p - proc %p\n", new_pool, new_pool->pool_proc_entry);
	468	#endif
	469
	470	spin_lock(&obd->obd_dev_lock);
	471	list_add_tail(&new_pool->pool_list, &lov->lov_pool_list);
	472	lov->lov_pool_count++;
	473	spin_unlock(&obd->obd_dev_lock);
	474
	475	/* add to find only when it fully ready */
	476	rc = cfs_hash_add_unique(lov->lov_pools_hash_body, poolname,
	477	&new_pool->pool_hash);
	478	if (rc)
	479	GOTO(out_err, rc = -EEXIST);
	480
	481	CDEBUG(D_CONFIG, LOV_POOLNAMEF" is pool #%d\n",
	482	poolname, lov->lov_pool_count);
	483
0a3bdb00	484	return 0;
d7e09d03 PT	485
	486	out_err:
	487	spin_lock(&obd->obd_dev_lock);
	488	list_del_init(&new_pool->pool_list);
	489	lov->lov_pool_count--;
	490	spin_unlock(&obd->obd_dev_lock);
	491
	492	lprocfs_remove(&new_pool->pool_proc_entry);
	493
	494	lov_ost_pool_free(&new_pool->pool_rr.lqr_pool);
	495	out_free_pool_obds:
	496	lov_ost_pool_free(&new_pool->pool_obds);
	497	OBD_FREE_PTR(new_pool);
	498	return rc;
	499	}
	500
	501	int lov_pool_del(struct obd_device obd, char poolname)
	502	{
	503	struct lov_obd *lov;
	504	struct pool_desc *pool;
d7e09d03 PT	505
	506	lov = &(obd->u.lov);
	507
	508	/* lookup and kill hash reference */
	509	pool = cfs_hash_del_key(lov->lov_pools_hash_body, poolname);
	510	if (pool == NULL)
0a3bdb00	511	return -ENOENT;
d7e09d03 PT	512
	513	if (pool->pool_proc_entry != NULL) {
	514	CDEBUG(D_INFO, "proc entry %p\n", pool->pool_proc_entry);
	515	lprocfs_remove(&pool->pool_proc_entry);
	516	lov_pool_putref(pool);
	517	}
	518
	519	spin_lock(&obd->obd_dev_lock);
	520	list_del_init(&pool->pool_list);
	521	lov->lov_pool_count--;
	522	spin_unlock(&obd->obd_dev_lock);
	523
	524	/* release last reference */
	525	lov_pool_putref(pool);
	526
0a3bdb00	527	return 0;
d7e09d03 PT	528	}
	529
	530
	531	int lov_pool_add(struct obd_device obd, char poolname, char *ostname)
	532	{
	533	struct obd_uuid ost_uuid;
	534	struct lov_obd *lov;
	535	struct pool_desc *pool;
	536	unsigned int lov_idx;
	537	int rc;
d7e09d03 PT	538
	539	lov = &(obd->u.lov);
	540
	541	pool = cfs_hash_lookup(lov->lov_pools_hash_body, poolname);
	542	if (pool == NULL)
0a3bdb00	543	return -ENOENT;
d7e09d03 PT	544
	545	obd_str2uuid(&ost_uuid, ostname);
	546
	547
	548	/* search ost in lov array */
	549	obd_getref(obd);
	550	for (lov_idx = 0; lov_idx < lov->desc.ld_tgt_count; lov_idx++) {
	551	if (!lov->lov_tgts[lov_idx])
	552	continue;
	553	if (obd_uuid_equals(&ost_uuid,
	554	&(lov->lov_tgts[lov_idx]->ltd_uuid)))
	555	break;
	556	}
	557	/* test if ost found in lov */
	558	if (lov_idx == lov->desc.ld_tgt_count)
	559	GOTO(out, rc = -EINVAL);
	560
	561	rc = lov_ost_pool_add(&pool->pool_obds, lov_idx, lov->lov_tgt_size);
	562	if (rc)
	563	GOTO(out, rc);
	564
	565	pool->pool_rr.lqr_dirty = 1;
	566
	567	CDEBUG(D_CONFIG, "Added %s to "LOV_POOLNAMEF" as member %d\n",
	568	ostname, poolname, pool_tgt_count(pool));
	569
d7e09d03 PT	570	out:
	571	obd_putref(obd);
	572	lov_pool_putref(pool);
	573	return rc;
	574	}
	575
	576	int lov_pool_remove(struct obd_device obd, char poolname, char *ostname)
	577	{
	578	struct obd_uuid ost_uuid;
	579	struct lov_obd *lov;
	580	struct pool_desc *pool;
	581	unsigned int lov_idx;
	582	int rc = 0;
d7e09d03 PT	583
	584	lov = &(obd->u.lov);
	585
	586	pool = cfs_hash_lookup(lov->lov_pools_hash_body, poolname);
	587	if (pool == NULL)
0a3bdb00	588	return -ENOENT;
d7e09d03 PT	589
	590	obd_str2uuid(&ost_uuid, ostname);
	591
	592	obd_getref(obd);
	593	/* search ost in lov array, to get index */
	594	for (lov_idx = 0; lov_idx < lov->desc.ld_tgt_count; lov_idx++) {
	595	if (!lov->lov_tgts[lov_idx])
	596	continue;
	597
	598	if (obd_uuid_equals(&ost_uuid,
	599	&(lov->lov_tgts[lov_idx]->ltd_uuid)))
	600	break;
	601	}
	602
	603	/* test if ost found in lov */
	604	if (lov_idx == lov->desc.ld_tgt_count)
	605	GOTO(out, rc = -EINVAL);
	606
	607	lov_ost_pool_remove(&pool->pool_obds, lov_idx);
	608
	609	pool->pool_rr.lqr_dirty = 1;
	610
	611	CDEBUG(D_CONFIG, "%s removed from "LOV_POOLNAMEF"\n", ostname,
	612	poolname);
	613
d7e09d03 PT	614	out:
	615	obd_putref(obd);
	616	lov_pool_putref(pool);
	617	return rc;
	618	}
	619
	620	int lov_check_index_in_pool(__u32 idx, struct pool_desc *pool)
	621	{
	622	int i, rc;
d7e09d03 PT	623
	624	/* caller may no have a ref on pool if it got the pool
	625	* without calling lov_find_pool() (e.g. go through the lov pool
	626	* list)
	627	*/
	628	lov_pool_getref(pool);
	629
	630	down_read(&pool_tgt_rw_sem(pool));
	631
	632	for (i = 0; i < pool_tgt_count(pool); i++) {
	633	if (pool_tgt_array(pool)[i] == idx)
	634	GOTO(out, rc = 0);
	635	}
	636	rc = -ENOENT;
d7e09d03 PT	637	out:
	638	up_read(&pool_tgt_rw_sem(pool));
	639
	640	lov_pool_putref(pool);
	641	return rc;
	642	}
	643
	644	struct pool_desc lov_find_pool(struct lov_obd lov, char *poolname)
	645	{
	646	struct pool_desc *pool;
	647
	648	pool = NULL;
	649	if (poolname[0] != '\0') {
	650	pool = cfs_hash_lookup(lov->lov_pools_hash_body, poolname);
	651	if (pool == NULL)
	652	CWARN("Request for an unknown pool ("LOV_POOLNAMEF")\n",
	653	poolname);
	654	if ((pool != NULL) && (pool_tgt_count(pool) == 0)) {
	655	CWARN("Request for an empty pool ("LOV_POOLNAMEF")\n",
	656	poolname);
	657	/* pool is ignored, so we remove ref on it */
	658	lov_pool_putref(pool);
	659	pool = NULL;
	660	}
	661	}
	662	return pool;
	663	}