[linux.git] / fs / ocfs2 / slot_map.c

/* -*- mode: c; c-basic-offset: 8; -*-
 * vim: noexpandtab sw=8 ts=8 sts=0:
 *
 * slot_map.c
 *
 *
 *
 * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
 *
 * 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.
 *
 * 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 for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 */

#include <linux/types.h>
#include <linux/slab.h>
#include <linux/highmem.h>

#define MLOG_MASK_PREFIX ML_SUPER
#include <cluster/masklog.h>

#include "ocfs2.h"

#include "dlmglue.h"
#include "extent_map.h"
#include "heartbeat.h"
#include "inode.h"
#include "slot_map.h"
#include "super.h"
#include "sysfile.h"

#include "buffer_head_io.h"

static s16 __ocfs2_node_num_to_slot(struct ocfs2_slot_info *si,
				    s16 global);
static void __ocfs2_fill_slot(struct ocfs2_slot_info *si,
			      s16 slot_num,
			      s16 node_num);

/* Use the slot information we've collected to create a map of mounted
 * nodes. Should be holding an EX on super block. assumes slot info is
 * up to date. Note that we call this *after* we find a slot, so our
 * own node should be set in the map too... */
void ocfs2_populate_mounted_map(struct ocfs2_super *osb)
{
	int i;
	struct ocfs2_slot_info *si = osb->slot_info;

	spin_lock(&si->si_lock);

	for (i = 0; i < si->si_size; i++)
		if (si->si_global_node_nums[i] != OCFS2_INVALID_SLOT)
			ocfs2_node_map_set_bit(osb, &osb->mounted_map,
					      si->si_global_node_nums[i]);

	spin_unlock(&si->si_lock);
}

/* post the slot information on disk into our slot_info struct. */
void ocfs2_update_slot_info(struct ocfs2_slot_info *si)
{
	int i;
	__le16 *disk_info;

	/* we don't read the slot block here as ocfs2_super_lock
	 * should've made sure we have the most recent copy. */
	spin_lock(&si->si_lock);
	disk_info = (__le16 *) si->si_bh->b_data;

	for (i = 0; i < si->si_size; i++)
		si->si_global_node_nums[i] = le16_to_cpu(disk_info[i]);

	spin_unlock(&si->si_lock);
}

/* post the our slot info stuff into it's destination bh and write it
 * out. */
int ocfs2_update_disk_slots(struct ocfs2_super *osb,
			    struct ocfs2_slot_info *si)
{
	int status, i;
	__le16 *disk_info = (__le16 *) si->si_bh->b_data;

	spin_lock(&si->si_lock);
	for (i = 0; i < si->si_size; i++)
		disk_info[i] = cpu_to_le16(si->si_global_node_nums[i]);
	spin_unlock(&si->si_lock);

	status = ocfs2_write_block(osb, si->si_bh, si->si_inode);
	if (status < 0)
		mlog_errno(status);

	return status;
}

/* try to find global node in the slot info. Returns
 * OCFS2_INVALID_SLOT if nothing is found. */
static s16 __ocfs2_node_num_to_slot(struct ocfs2_slot_info *si,
				    s16 global)
{
	int i;
	s16 ret = OCFS2_INVALID_SLOT;

	for(i = 0; i < si->si_num_slots; i++) {
		if (global == si->si_global_node_nums[i]) {
			ret = (s16) i;
			break;
		}
	}
	return ret;
}

static s16 __ocfs2_find_empty_slot(struct ocfs2_slot_info *si, s16 preferred)
{
	int i;
	s16 ret = OCFS2_INVALID_SLOT;

	if (preferred >= 0 && preferred < si->si_num_slots) {
		if (OCFS2_INVALID_SLOT == si->si_global_node_nums[preferred]) {
			ret = preferred;
			goto out;
		}
	}

	for(i = 0; i < si->si_num_slots; i++) {
		if (OCFS2_INVALID_SLOT == si->si_global_node_nums[i]) {
			ret = (s16) i;
			break;
		}
	}
out:
	return ret;
}

s16 ocfs2_node_num_to_slot(struct ocfs2_slot_info *si,
			   s16 global)
{
	s16 ret;

	spin_lock(&si->si_lock);
	ret = __ocfs2_node_num_to_slot(si, global);
	spin_unlock(&si->si_lock);
	return ret;
}

static void __ocfs2_fill_slot(struct ocfs2_slot_info *si,
			      s16 slot_num,
			      s16 node_num)
{
	BUG_ON(slot_num == OCFS2_INVALID_SLOT);
	BUG_ON(slot_num >= si->si_num_slots);
	BUG_ON((node_num != O2NM_INVALID_NODE_NUM) &&
	       (node_num >= O2NM_MAX_NODES));

	si->si_global_node_nums[slot_num] = node_num;
}

void ocfs2_clear_slot(struct ocfs2_slot_info *si,
		      s16 slot_num)
{
	spin_lock(&si->si_lock);
	__ocfs2_fill_slot(si, slot_num, OCFS2_INVALID_SLOT);
	spin_unlock(&si->si_lock);
}

int ocfs2_init_slot_info(struct ocfs2_super *osb)
{
	int status, i;
	u64 blkno;
	struct inode *inode = NULL;
	struct buffer_head *bh = NULL;
	struct ocfs2_slot_info *si;

	si = kzalloc(sizeof(struct ocfs2_slot_info), GFP_KERNEL);
	if (!si) {
		status = -ENOMEM;
		mlog_errno(status);
		goto bail;
	}

	spin_lock_init(&si->si_lock);
	si->si_num_slots = osb->max_slots;
	si->si_size = OCFS2_MAX_SLOTS;

	for(i = 0; i < si->si_num_slots; i++)
		si->si_global_node_nums[i] = OCFS2_INVALID_SLOT;

	inode = ocfs2_get_system_file_inode(osb, SLOT_MAP_SYSTEM_INODE,
					    OCFS2_INVALID_SLOT);
	if (!inode) {
		status = -EINVAL;
		mlog_errno(status);
		goto bail;
	}

	status = ocfs2_extent_map_get_blocks(inode, 0ULL, &blkno, NULL, NULL);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	status = ocfs2_read_block(osb, blkno, &bh, 0, inode);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	si->si_inode = inode;
	si->si_bh = bh;
	osb->slot_info = si;
bail:
	if (status < 0 && si)
		ocfs2_free_slot_info(si);

	return status;
}

void ocfs2_free_slot_info(struct ocfs2_slot_info *si)
{
	if (si->si_inode)
		iput(si->si_inode);
	if (si->si_bh)
		brelse(si->si_bh);
	kfree(si);
}

int ocfs2_find_slot(struct ocfs2_super *osb)
{
	int status;
	s16 slot;
	struct ocfs2_slot_info *si;

	mlog_entry_void();

	si = osb->slot_info;

	ocfs2_update_slot_info(si);

	spin_lock(&si->si_lock);
	/* search for ourselves first and take the slot if it already
	 * exists. Perhaps we need to mark this in a variable for our
	 * own journal recovery? Possibly not, though we certainly
	 * need to warn to the user */
	slot = __ocfs2_node_num_to_slot(si, osb->node_num);
	if (slot == OCFS2_INVALID_SLOT) {
		/* if no slot yet, then just take 1st available
		 * one. */
		slot = __ocfs2_find_empty_slot(si, osb->preferred_slot);
		if (slot == OCFS2_INVALID_SLOT) {
			spin_unlock(&si->si_lock);
			mlog(ML_ERROR, "no free slots available!\n");
			status = -EINVAL;
			goto bail;
		}
	} else
		mlog(ML_NOTICE, "slot %d is already allocated to this node!\n",
		     slot);

	__ocfs2_fill_slot(si, slot, osb->node_num);
	osb->slot_num = slot;
	spin_unlock(&si->si_lock);

	mlog(0, "taking node slot %d\n", osb->slot_num);

	status = ocfs2_update_disk_slots(osb, si);
	if (status < 0)
		mlog_errno(status);

bail:
	mlog_exit(status);
	return status;
}

void ocfs2_put_slot(struct ocfs2_super *osb)
{
	int status;
	struct ocfs2_slot_info *si = osb->slot_info;

	if (!si)
		return;

	ocfs2_update_slot_info(si);

	spin_lock(&si->si_lock);
	__ocfs2_fill_slot(si, osb->slot_num, OCFS2_INVALID_SLOT);
	osb->slot_num = OCFS2_INVALID_SLOT;
	spin_unlock(&si->si_lock);

	status = ocfs2_update_disk_slots(osb, si);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

bail:
	osb->slot_info = NULL;
	ocfs2_free_slot_info(si);
}
Commit	Line	Data
ccd979bd MF	1	/* -- mode: c; c-basic-offset: 8; --
	2	* vim: noexpandtab sw=8 ts=8 sts=0:
	3	*
	4	* slot_map.c
	5	*
	6	*
	7	*
	8	* Copyright (C) 2002, 2004 Oracle. All rights reserved.
	9	*
	10	* This program is free software; you can redistribute it and/or
	11	* modify it under the terms of the GNU General Public
	12	* License as published by the Free Software Foundation; either
	13	* version 2 of the License, or (at your option) any later version.
	14	*
	15	* This program is distributed in the hope that it will be useful,
	16	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	17	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	18	* General Public License for more details.
	19	*
	20	* You should have received a copy of the GNU General Public
	21	* License along with this program; if not, write to the
	22	* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
	23	* Boston, MA 021110-1307, USA.
	24	*/
	25
	26	#include <linux/types.h>
	27	#include <linux/slab.h>
	28	#include <linux/highmem.h>
ccd979bd MF	29
	30	#define MLOG_MASK_PREFIX ML_SUPER
	31	#include <cluster/masklog.h>
	32
	33	#include "ocfs2.h"
	34
	35	#include "dlmglue.h"
	36	#include "extent_map.h"
	37	#include "heartbeat.h"
	38	#include "inode.h"
	39	#include "slot_map.h"
	40	#include "super.h"
	41	#include "sysfile.h"
	42
	43	#include "buffer_head_io.h"
	44
	45	static s16 __ocfs2_node_num_to_slot(struct ocfs2_slot_info *si,
	46	s16 global);
	47	static void __ocfs2_fill_slot(struct ocfs2_slot_info *si,
	48	s16 slot_num,
	49	s16 node_num);
	50
	51	/* Use the slot information we've collected to create a map of mounted
	52	* nodes. Should be holding an EX on super block. assumes slot info is
	53	* up to date. Note that we call this after we find a slot, so our
	54	* own node should be set in the map too... */
	55	void ocfs2_populate_mounted_map(struct ocfs2_super *osb)
	56	{
	57	int i;
	58	struct ocfs2_slot_info *si = osb->slot_info;
	59
	60	spin_lock(&si->si_lock);
	61
	62	for (i = 0; i < si->si_size; i++)
	63	if (si->si_global_node_nums[i] != OCFS2_INVALID_SLOT)
	64	ocfs2_node_map_set_bit(osb, &osb->mounted_map,
	65	si->si_global_node_nums[i]);
	66
	67	spin_unlock(&si->si_lock);
	68	}
	69
	70	/* post the slot information on disk into our slot_info struct. */
	71	void ocfs2_update_slot_info(struct ocfs2_slot_info *si)
	72	{
	73	int i;
	74	__le16 *disk_info;
	75
	76	/* we don't read the slot block here as ocfs2_super_lock
	77	* should've made sure we have the most recent copy. */
	78	spin_lock(&si->si_lock);
	79	disk_info = (__le16 *) si->si_bh->b_data;
	80
	81	for (i = 0; i < si->si_size; i++)
	82	si->si_global_node_nums[i] = le16_to_cpu(disk_info[i]);
	83
	84	spin_unlock(&si->si_lock);
	85	}
	86
	87	/* post the our slot info stuff into it's destination bh and write it
	88	* out. */
	89	int ocfs2_update_disk_slots(struct ocfs2_super *osb,
	90	struct ocfs2_slot_info *si)
	91	{
	92	int status, i;
93	__le16 disk_info = (__le16 ) si->si_bh->b_data;
94
95	spin_lock(&si->si_lock);
96	for (i = 0; i < si->si_size; i++)
97	disk_info[i] = cpu_to_le16(si->si_global_node_nums[i]);
98	spin_unlock(&si->si_lock);
99
100	status = ocfs2_write_block(osb, si->si_bh, si->si_inode);
101	if (status < 0)
102	mlog_errno(status);
103
104	return status;
105	}
106
107	/* try to find global node in the slot info. Returns
108	* OCFS2_INVALID_SLOT if nothing is found. */
109	static s16 __ocfs2_node_num_to_slot(struct ocfs2_slot_info *si,
110	s16 global)
111	{
112	int i;
113	s16 ret = OCFS2_INVALID_SLOT;
114
115	for(i = 0; i < si->si_num_slots; i++) {
116	if (global == si->si_global_node_nums[i]) {
117	ret = (s16) i;
118	break;
119	}
120	}
121	return ret;
122	}
123
baf4661a	124	static s16 __ocfs2_find_empty_slot(struct ocfs2_slot_info *si, s16 preferred)
ccd979bd MF	125	{
	126	int i;
	127	s16 ret = OCFS2_INVALID_SLOT;
	128
baf4661a SM	129	if (preferred >= 0 && preferred < si->si_num_slots) {
	130	if (OCFS2_INVALID_SLOT == si->si_global_node_nums[preferred]) {
	131	ret = preferred;
	132	goto out;
	133	}
	134	}
	135
ccd979bd MF	136	for(i = 0; i < si->si_num_slots; i++) {
	137	if (OCFS2_INVALID_SLOT == si->si_global_node_nums[i]) {
	138	ret = (s16) i;
	139	break;
	140	}
	141	}
baf4661a	142	out:
ccd979bd MF	143	return ret;
	144	}
	145
	146	s16 ocfs2_node_num_to_slot(struct ocfs2_slot_info *si,
	147	s16 global)
	148	{
	149	s16 ret;
	150
	151	spin_lock(&si->si_lock);
	152	ret = __ocfs2_node_num_to_slot(si, global);
	153	spin_unlock(&si->si_lock);
	154	return ret;
	155	}
	156
	157	static void __ocfs2_fill_slot(struct ocfs2_slot_info *si,
	158	s16 slot_num,
	159	s16 node_num)
	160	{
	161	BUG_ON(slot_num == OCFS2_INVALID_SLOT);
	162	BUG_ON(slot_num >= si->si_num_slots);
	163	BUG_ON((node_num != O2NM_INVALID_NODE_NUM) &&
	164	(node_num >= O2NM_MAX_NODES));
	165
	166	si->si_global_node_nums[slot_num] = node_num;
	167	}
	168
	169	void ocfs2_clear_slot(struct ocfs2_slot_info *si,
	170	s16 slot_num)
	171	{
	172	spin_lock(&si->si_lock);
	173	__ocfs2_fill_slot(si, slot_num, OCFS2_INVALID_SLOT);
	174	spin_unlock(&si->si_lock);
	175	}
	176
	177	int ocfs2_init_slot_info(struct ocfs2_super *osb)
	178	{
	179	int status, i;
	180	u64 blkno;
	181	struct inode *inode = NULL;
	182	struct buffer_head *bh = NULL;
	183	struct ocfs2_slot_info *si;
	184
cd861280	185	si = kzalloc(sizeof(struct ocfs2_slot_info), GFP_KERNEL);
ccd979bd MF	186	if (!si) {
	187	status = -ENOMEM;
	188	mlog_errno(status);
	189	goto bail;
	190	}
	191
	192	spin_lock_init(&si->si_lock);
	193	si->si_num_slots = osb->max_slots;
	194	si->si_size = OCFS2_MAX_SLOTS;
	195
	196	for(i = 0; i < si->si_num_slots; i++)
	197	si->si_global_node_nums[i] = OCFS2_INVALID_SLOT;
	198
	199	inode = ocfs2_get_system_file_inode(osb, SLOT_MAP_SYSTEM_INODE,
	200	OCFS2_INVALID_SLOT);
	201	if (!inode) {
	202	status = -EINVAL;
	203	mlog_errno(status);
	204	goto bail;
	205	}
	206
49cb8d2d	207	status = ocfs2_extent_map_get_blocks(inode, 0ULL, &blkno, NULL, NULL);
ccd979bd MF	208	if (status < 0) {
	209	mlog_errno(status);
	210	goto bail;
	211	}
	212
	213	status = ocfs2_read_block(osb, blkno, &bh, 0, inode);
	214	if (status < 0) {
	215	mlog_errno(status);
	216	goto bail;
	217	}
	218
	219	si->si_inode = inode;
	220	si->si_bh = bh;
	221	osb->slot_info = si;
	222	bail:
	223	if (status < 0 && si)
	224	ocfs2_free_slot_info(si);
	225
	226	return status;
	227	}
	228
	229	void ocfs2_free_slot_info(struct ocfs2_slot_info *si)
	230	{
	231	if (si->si_inode)
	232	iput(si->si_inode);
	233	if (si->si_bh)
	234	brelse(si->si_bh);
	235	kfree(si);
	236	}
	237
	238	int ocfs2_find_slot(struct ocfs2_super *osb)
	239	{
	240	int status;
	241	s16 slot;
	242	struct ocfs2_slot_info *si;
	243
	244	mlog_entry_void();
	245
	246	si = osb->slot_info;
	247
	248	ocfs2_update_slot_info(si);
	249
	250	spin_lock(&si->si_lock);
	251	/* search for ourselves first and take the slot if it already
	252	* exists. Perhaps we need to mark this in a variable for our
	253	* own journal recovery? Possibly not, though we certainly
	254	* need to warn to the user */
	255	slot = __ocfs2_node_num_to_slot(si, osb->node_num);
	256	if (slot == OCFS2_INVALID_SLOT) {
	257	/* if no slot yet, then just take 1st available
	258	* one. */
baf4661a	259	slot = __ocfs2_find_empty_slot(si, osb->preferred_slot);
ccd979bd MF	260	if (slot == OCFS2_INVALID_SLOT) {
	261	spin_unlock(&si->si_lock);
	262	mlog(ML_ERROR, "no free slots available!\n");
	263	status = -EINVAL;
	264	goto bail;
	265	}
	266	} else
	267	mlog(ML_NOTICE, "slot %d is already allocated to this node!\n",
	268	slot);
	269
	270	__ocfs2_fill_slot(si, slot, osb->node_num);
	271	osb->slot_num = slot;
	272	spin_unlock(&si->si_lock);
	273
e7607ab3	274	mlog(0, "taking node slot %d\n", osb->slot_num);
ccd979bd MF	275
	276	status = ocfs2_update_disk_slots(osb, si);
	277	if (status < 0)
	278	mlog_errno(status);
	279
	280	bail:
	281	mlog_exit(status);
	282	return status;
	283	}
	284
	285	void ocfs2_put_slot(struct ocfs2_super *osb)
	286	{
	287	int status;
	288	struct ocfs2_slot_info *si = osb->slot_info;
	289
	290	if (!si)
	291	return;
	292
	293	ocfs2_update_slot_info(si);
	294
	295	spin_lock(&si->si_lock);
	296	__ocfs2_fill_slot(si, osb->slot_num, OCFS2_INVALID_SLOT);
	297	osb->slot_num = OCFS2_INVALID_SLOT;
	298	spin_unlock(&si->si_lock);
	299
	300	status = ocfs2_update_disk_slots(osb, si);
	301	if (status < 0) {
	302	mlog_errno(status);
	303	goto bail;
	304	}
	305
	306	bail:
	307	osb->slot_info = NULL;
	308	ocfs2_free_slot_info(si);
	309	}
	310