[linux-2.6-block.git] / fs / ocfs2 / slot_map.c

// SPDX-License-Identifier: GPL-2.0-or-later
/*
 * slot_map.c
 *
 * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
 */

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

#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 "ocfs2_trace.h"

#include "buffer_head_io.h"


struct ocfs2_slot {
	int sl_valid;
	unsigned int sl_node_num;
};

struct ocfs2_slot_info {
	int si_extended;
	int si_slots_per_block;
	struct inode *si_inode;
	unsigned int si_blocks;
	struct buffer_head **si_bh;
	unsigned int si_num_slots;
	struct ocfs2_slot si_slots[] __counted_by(si_num_slots);
};


static int __ocfs2_node_num_to_slot(struct ocfs2_slot_info *si,
				    unsigned int node_num);

static void ocfs2_invalidate_slot(struct ocfs2_slot_info *si,
				  int slot_num)
{
	BUG_ON((slot_num < 0) || (slot_num >= si->si_num_slots));
	si->si_slots[slot_num].sl_valid = 0;
}

static void ocfs2_set_slot(struct ocfs2_slot_info *si,
			   int slot_num, unsigned int node_num)
{
	BUG_ON((slot_num < 0) || (slot_num >= si->si_num_slots));

	si->si_slots[slot_num].sl_valid = 1;
	si->si_slots[slot_num].sl_node_num = node_num;
}

/* This version is for the extended slot map */
static void ocfs2_update_slot_info_extended(struct ocfs2_slot_info *si)
{
	int b, i, slotno;
	struct ocfs2_slot_map_extended *se;

	slotno = 0;
	for (b = 0; b < si->si_blocks; b++) {
		se = (struct ocfs2_slot_map_extended *)si->si_bh[b]->b_data;
		for (i = 0;
		     (i < si->si_slots_per_block) &&
		     (slotno < si->si_num_slots);
		     i++, slotno++) {
			if (se->se_slots[i].es_valid)
				ocfs2_set_slot(si, slotno,
					       le32_to_cpu(se->se_slots[i].es_node_num));
			else
				ocfs2_invalidate_slot(si, slotno);
		}
	}
}

/*
 * Post the slot information on disk into our slot_info struct.
 * Must be protected by osb_lock.
 */
static void ocfs2_update_slot_info_old(struct ocfs2_slot_info *si)
{
	int i;
	struct ocfs2_slot_map *sm;

	sm = (struct ocfs2_slot_map *)si->si_bh[0]->b_data;

	for (i = 0; i < si->si_num_slots; i++) {
		if (le16_to_cpu(sm->sm_slots[i]) == (u16)OCFS2_INVALID_SLOT)
			ocfs2_invalidate_slot(si, i);
		else
			ocfs2_set_slot(si, i, le16_to_cpu(sm->sm_slots[i]));
	}
}

static void ocfs2_update_slot_info(struct ocfs2_slot_info *si)
{
	/*
	 * The slot data will have been refreshed when ocfs2_super_lock
	 * was taken.
	 */
	if (si->si_extended)
		ocfs2_update_slot_info_extended(si);
	else
		ocfs2_update_slot_info_old(si);
}

int ocfs2_refresh_slot_info(struct ocfs2_super *osb)
{
	int ret;
	struct ocfs2_slot_info *si = osb->slot_info;

	if (si == NULL)
		return 0;

	BUG_ON(si->si_blocks == 0);
	BUG_ON(si->si_bh == NULL);

	trace_ocfs2_refresh_slot_info(si->si_blocks);

	/*
	 * We pass -1 as blocknr because we expect all of si->si_bh to
	 * be !NULL.  Thus, ocfs2_read_blocks() will ignore blocknr.  If
	 * this is not true, the read of -1 (UINT64_MAX) will fail.
	 */
	ret = ocfs2_read_blocks(INODE_CACHE(si->si_inode), -1, si->si_blocks,
				si->si_bh, OCFS2_BH_IGNORE_CACHE, NULL);
	if (ret == 0) {
		spin_lock(&osb->osb_lock);
		ocfs2_update_slot_info(si);
		spin_unlock(&osb->osb_lock);
	}

	return ret;
}

/* post the our slot info stuff into it's destination bh and write it
 * out. */
static void ocfs2_update_disk_slot_extended(struct ocfs2_slot_info *si,
					    int slot_num,
					    struct buffer_head **bh)
{
	int blkind = slot_num / si->si_slots_per_block;
	int slotno = slot_num % si->si_slots_per_block;
	struct ocfs2_slot_map_extended *se;

	BUG_ON(blkind >= si->si_blocks);

	se = (struct ocfs2_slot_map_extended *)si->si_bh[blkind]->b_data;
	se->se_slots[slotno].es_valid = si->si_slots[slot_num].sl_valid;
	if (si->si_slots[slot_num].sl_valid)
		se->se_slots[slotno].es_node_num =
			cpu_to_le32(si->si_slots[slot_num].sl_node_num);
	*bh = si->si_bh[blkind];
}

static void ocfs2_update_disk_slot_old(struct ocfs2_slot_info *si,
				       int slot_num,
				       struct buffer_head **bh)
{
	int i;
	struct ocfs2_slot_map *sm;

	sm = (struct ocfs2_slot_map *)si->si_bh[0]->b_data;
	for (i = 0; i < si->si_num_slots; i++) {
		if (si->si_slots[i].sl_valid)
			sm->sm_slots[i] =
				cpu_to_le16(si->si_slots[i].sl_node_num);
		else
			sm->sm_slots[i] = cpu_to_le16(OCFS2_INVALID_SLOT);
	}
	*bh = si->si_bh[0];
}

static int ocfs2_update_disk_slot(struct ocfs2_super *osb,
				  struct ocfs2_slot_info *si,
				  int slot_num)
{
	int status;
	struct buffer_head *bh;

	spin_lock(&osb->osb_lock);
	if (si->si_extended)
		ocfs2_update_disk_slot_extended(si, slot_num, &bh);
	else
		ocfs2_update_disk_slot_old(si, slot_num, &bh);
	spin_unlock(&osb->osb_lock);

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

	return status;
}

/*
 * Calculate how many bytes are needed by the slot map.  Returns
 * an error if the slot map file is too small.
 */
static int ocfs2_slot_map_physical_size(struct ocfs2_super *osb,
					struct inode *inode,
					unsigned long long *bytes)
{
	unsigned long long bytes_needed;

	if (ocfs2_uses_extended_slot_map(osb)) {
		bytes_needed = osb->max_slots *
			sizeof(struct ocfs2_extended_slot);
	} else {
		bytes_needed = osb->max_slots * sizeof(__le16);
	}
	if (bytes_needed > i_size_read(inode)) {
		mlog(ML_ERROR,
		     "Slot map file is too small!  (size %llu, needed %llu)\n",
		     i_size_read(inode), bytes_needed);
		return -ENOSPC;
	}

	*bytes = bytes_needed;
	return 0;
}

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

	for(i = 0; i < si->si_num_slots; i++) {
		if (si->si_slots[i].sl_valid &&
		    (node_num == si->si_slots[i].sl_node_num)) {
			ret = i;
			break;
		}
	}

	return ret;
}

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

	if ((preferred >= 0) && (preferred < si->si_num_slots)) {
		if (!si->si_slots[preferred].sl_valid) {
			ret = preferred;
			goto out;
		}
	}

	for(i = 0; i < si->si_num_slots; i++) {
		if (!si->si_slots[i].sl_valid) {
			ret = i;
			break;
		}
	}
out:
	return ret;
}

int ocfs2_node_num_to_slot(struct ocfs2_super *osb, unsigned int node_num)
{
	int slot;
	struct ocfs2_slot_info *si = osb->slot_info;

	spin_lock(&osb->osb_lock);
	slot = __ocfs2_node_num_to_slot(si, node_num);
	spin_unlock(&osb->osb_lock);

	return slot;
}

int ocfs2_slot_to_node_num_locked(struct ocfs2_super *osb, int slot_num,
				  unsigned int *node_num)
{
	struct ocfs2_slot_info *si = osb->slot_info;

	assert_spin_locked(&osb->osb_lock);

	BUG_ON(slot_num < 0);
	BUG_ON(slot_num >= osb->max_slots);

	if (!si->si_slots[slot_num].sl_valid)
		return -ENOENT;

	*node_num = si->si_slots[slot_num].sl_node_num;
	return 0;
}

static void __ocfs2_free_slot_info(struct ocfs2_slot_info *si)
{
	unsigned int i;

	if (si == NULL)
		return;

	iput(si->si_inode);
	if (si->si_bh) {
		for (i = 0; i < si->si_blocks; i++) {
			if (si->si_bh[i]) {
				brelse(si->si_bh[i]);
				si->si_bh[i] = NULL;
			}
		}
		kfree(si->si_bh);
	}

	kfree(si);
}

int ocfs2_clear_slot(struct ocfs2_super *osb, int slot_num)
{
	struct ocfs2_slot_info *si = osb->slot_info;

	if (si == NULL)
		return 0;

	spin_lock(&osb->osb_lock);
	ocfs2_invalidate_slot(si, slot_num);
	spin_unlock(&osb->osb_lock);

	return ocfs2_update_disk_slot(osb, osb->slot_info, slot_num);
}

static int ocfs2_map_slot_buffers(struct ocfs2_super *osb,
				  struct ocfs2_slot_info *si)
{
	int status = 0;
	u64 blkno;
	unsigned long long blocks, bytes = 0;
	unsigned int i;
	struct buffer_head *bh;

	status = ocfs2_slot_map_physical_size(osb, si->si_inode, &bytes);
	if (status)
		goto bail;

	blocks = ocfs2_blocks_for_bytes(si->si_inode->i_sb, bytes);
	BUG_ON(blocks > UINT_MAX);
	si->si_blocks = blocks;
	if (!si->si_blocks)
		goto bail;

	if (si->si_extended)
		si->si_slots_per_block =
			(osb->sb->s_blocksize /
			 sizeof(struct ocfs2_extended_slot));
	else
		si->si_slots_per_block = osb->sb->s_blocksize / sizeof(__le16);

	/* The size checks above should ensure this */
	BUG_ON((osb->max_slots / si->si_slots_per_block) > blocks);

	trace_ocfs2_map_slot_buffers(bytes, si->si_blocks);

	si->si_bh = kcalloc(si->si_blocks, sizeof(struct buffer_head *),
			    GFP_KERNEL);
	if (!si->si_bh) {
		status = -ENOMEM;
		mlog_errno(status);
		goto bail;
	}

	for (i = 0; i < si->si_blocks; i++) {
		status = ocfs2_extent_map_get_blocks(si->si_inode, i,
						     &blkno, NULL, NULL);
		if (status < 0) {
			mlog_errno(status);
			goto bail;
		}

		trace_ocfs2_map_slot_buffers_block((unsigned long long)blkno, i);

		bh = NULL;  /* Acquire a fresh bh */
		status = ocfs2_read_blocks(INODE_CACHE(si->si_inode), blkno,
					   1, &bh, OCFS2_BH_IGNORE_CACHE, NULL);
		if (status < 0) {
			mlog_errno(status);
			goto bail;
		}

		si->si_bh[i] = bh;
	}

bail:
	return status;
}

int ocfs2_init_slot_info(struct ocfs2_super *osb)
{
	int status;
	struct inode *inode = NULL;
	struct ocfs2_slot_info *si;

	si = kzalloc(struct_size(si, si_slots, osb->max_slots), GFP_KERNEL);
	if (!si) {
		status = -ENOMEM;
		mlog_errno(status);
		return status;
	}

	si->si_extended = ocfs2_uses_extended_slot_map(osb);
	si->si_num_slots = osb->max_slots;

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

	si->si_inode = inode;
	status = ocfs2_map_slot_buffers(osb, si);
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}

	osb->slot_info = (struct ocfs2_slot_info *)si;
bail:
	if (status < 0)
		__ocfs2_free_slot_info(si);

	return status;
}

void ocfs2_free_slot_info(struct ocfs2_super *osb)
{
	struct ocfs2_slot_info *si = osb->slot_info;

	osb->slot_info = NULL;
	__ocfs2_free_slot_info(si);
}

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

	si = osb->slot_info;

	spin_lock(&osb->osb_lock);
	ocfs2_update_slot_info(si);

	/* 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 < 0) {
		/* if no slot yet, then just take 1st available
		 * one. */
		slot = __ocfs2_find_empty_slot(si, osb->preferred_slot);
		if (slot < 0) {
			spin_unlock(&osb->osb_lock);
			mlog(ML_ERROR, "no free slots available!\n");
			status = -EINVAL;
			goto bail;
		}
	} else
		printk(KERN_INFO "ocfs2: Slot %d on device (%s) was already "
		       "allocated to this node!\n", slot, osb->dev_str);

	ocfs2_set_slot(si, slot, osb->node_num);
	osb->slot_num = slot;
	spin_unlock(&osb->osb_lock);

	trace_ocfs2_find_slot(osb->slot_num);

	status = ocfs2_update_disk_slot(osb, si, osb->slot_num);
	if (status < 0) {
		mlog_errno(status);
		/*
		 * if write block failed, invalidate slot to avoid overwrite
		 * slot during dismount in case another node rightly has mounted
		 */
		spin_lock(&osb->osb_lock);
		ocfs2_invalidate_slot(si, osb->slot_num);
		osb->slot_num = OCFS2_INVALID_SLOT;
		spin_unlock(&osb->osb_lock);
	}

bail:
	return status;
}

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

	if (!si)
		return;

	spin_lock(&osb->osb_lock);
	ocfs2_update_slot_info(si);

	slot_num = osb->slot_num;
	ocfs2_invalidate_slot(si, osb->slot_num);
	osb->slot_num = OCFS2_INVALID_SLOT;
	spin_unlock(&osb->osb_lock);

	status = ocfs2_update_disk_slot(osb, si, slot_num);
	if (status < 0)
		mlog_errno(status);

	ocfs2_free_slot_info(osb);
}
Commit	Line	Data
328970de	1	// SPDX-License-Identifier: GPL-2.0-or-later
fa60ce2c	2	/*
ccd979bd MF	3	* slot_map.c
ccd979bd MF	4	*
ccd979bd	5	* Copyright (C) 2002, 2004 Oracle. All rights reserved.
ccd979bd MF	6	*/
	7
	8	#include <linux/types.h>
	9	#include <linux/slab.h>
	10	#include <linux/highmem.h>
ccd979bd	11
ccd979bd MF	12	#include <cluster/masklog.h>
	13
	14	#include "ocfs2.h"
	15
	16	#include "dlmglue.h"
	17	#include "extent_map.h"
	18	#include "heartbeat.h"
	19	#include "inode.h"
	20	#include "slot_map.h"
	21	#include "super.h"
	22	#include "sysfile.h"
a8731086	23	#include "ocfs2_trace.h"
ccd979bd MF	24
	25	#include "buffer_head_io.h"
	26
fc881fa0 JB	27
	28	struct ocfs2_slot {
	29	int sl_valid;
	30	unsigned int sl_node_num;
	31	};
	32
d85b20e4	33	struct ocfs2_slot_info {
386a2ef8 JB	34	int si_extended;
386a2ef8 JB	35	int si_slots_per_block;
d85b20e4	36	struct inode *si_inode;
1c8d9a6a JB	37	unsigned int si_blocks;
1c8d9a6a JB	38	struct buffer_head **si_bh;
d85b20e4	39	unsigned int si_num_slots;
e91673b8	40	struct ocfs2_slot si_slots[] __counted_by(si_num_slots);
d85b20e4 JB	41	};
	42
	43
fc881fa0 JB	44	static int __ocfs2_node_num_to_slot(struct ocfs2_slot_info *si,
	45	unsigned int node_num);
	46
	47	static void ocfs2_invalidate_slot(struct ocfs2_slot_info *si,
	48	int slot_num)
	49	{
	50	BUG_ON((slot_num < 0) \|\| (slot_num >= si->si_num_slots));
	51	si->si_slots[slot_num].sl_valid = 0;
	52	}
	53
	54	static void ocfs2_set_slot(struct ocfs2_slot_info *si,
	55	int slot_num, unsigned int node_num)
	56	{
	57	BUG_ON((slot_num < 0) \|\| (slot_num >= si->si_num_slots));
fc881fa0 JB	58
	59	si->si_slots[slot_num].sl_valid = 1;
	60	si->si_slots[slot_num].sl_node_num = node_num;
	61	}
ccd979bd	62
386a2ef8 JB	63	/* This version is for the extended slot map */
	64	static void ocfs2_update_slot_info_extended(struct ocfs2_slot_info *si)
	65	{
	66	int b, i, slotno;
	67	struct ocfs2_slot_map_extended *se;
	68
	69	slotno = 0;
	70	for (b = 0; b < si->si_blocks; b++) {
	71	se = (struct ocfs2_slot_map_extended *)si->si_bh[b]->b_data;
	72	for (i = 0;
	73	(i < si->si_slots_per_block) &&
	74	(slotno < si->si_num_slots);
	75	i++, slotno++) {
	76	if (se->se_slots[i].es_valid)
	77	ocfs2_set_slot(si, slotno,
	78	le32_to_cpu(se->se_slots[i].es_node_num));
	79	else
	80	ocfs2_invalidate_slot(si, slotno);
	81	}
	82	}
	83	}
	84
d85b20e4 JB	85	/*
	86	* Post the slot information on disk into our slot_info struct.
	87	* Must be protected by osb_lock.
	88	*/
386a2ef8	89	static void ocfs2_update_slot_info_old(struct ocfs2_slot_info *si)
ccd979bd MF	90	{
ccd979bd MF	91	int i;
fb86b1f0	92	struct ocfs2_slot_map *sm;
ccd979bd	93
fb86b1f0	94	sm = (struct ocfs2_slot_map *)si->si_bh[0]->b_data;
ccd979bd	95
fc881fa0	96	for (i = 0; i < si->si_num_slots; i++) {
fb86b1f0	97	if (le16_to_cpu(sm->sm_slots[i]) == (u16)OCFS2_INVALID_SLOT)
fc881fa0 JB	98	ocfs2_invalidate_slot(si, i);
fc881fa0 JB	99	else
fb86b1f0	100	ocfs2_set_slot(si, i, le16_to_cpu(sm->sm_slots[i]));
fc881fa0	101	}
ccd979bd MF	102	}
ccd979bd MF	103
386a2ef8 JB	104	static void ocfs2_update_slot_info(struct ocfs2_slot_info *si)
	105	{
	106	/*
	107	* The slot data will have been refreshed when ocfs2_super_lock
	108	* was taken.
	109	*/
	110	if (si->si_extended)
	111	ocfs2_update_slot_info_extended(si);
	112	else
	113	ocfs2_update_slot_info_old(si);
	114	}
	115
8e8a4603 MF	116	int ocfs2_refresh_slot_info(struct ocfs2_super *osb)
	117	{
	118	int ret;
	119	struct ocfs2_slot_info *si = osb->slot_info;
8e8a4603 MF	120
	121	if (si == NULL)
	122	return 0;
	123
1c8d9a6a JB	124	BUG_ON(si->si_blocks == 0);
	125	BUG_ON(si->si_bh == NULL);
	126
a8731086	127	trace_ocfs2_refresh_slot_info(si->si_blocks);
1c8d9a6a JB	128
	129	/*
	130	* We pass -1 as blocknr because we expect all of si->si_bh to
	131	* be !NULL. Thus, ocfs2_read_blocks() will ignore blocknr. If
	132	* this is not true, the read of -1 (UINT64_MAX) will fail.
	133	*/
8cb471e8 JB	134	ret = ocfs2_read_blocks(INODE_CACHE(si->si_inode), -1, si->si_blocks,
8cb471e8 JB	135	si->si_bh, OCFS2_BH_IGNORE_CACHE, NULL);
d85b20e4 JB	136	if (ret == 0) {
d85b20e4 JB	137	spin_lock(&osb->osb_lock);
8e8a4603	138	ocfs2_update_slot_info(si);
d85b20e4 JB	139	spin_unlock(&osb->osb_lock);
d85b20e4 JB	140	}
8e8a4603 MF	141
	142	return ret;
	143	}
	144
ccd979bd MF	145	/* post the our slot info stuff into it's destination bh and write it
ccd979bd MF	146	* out. */
386a2ef8 JB	147	static void ocfs2_update_disk_slot_extended(struct ocfs2_slot_info *si,
	148	int slot_num,
	149	struct buffer_head **bh)
	150	{
	151	int blkind = slot_num / si->si_slots_per_block;
	152	int slotno = slot_num % si->si_slots_per_block;
	153	struct ocfs2_slot_map_extended *se;
	154
	155	BUG_ON(blkind >= si->si_blocks);
	156
	157	se = (struct ocfs2_slot_map_extended *)si->si_bh[blkind]->b_data;
	158	se->se_slots[slotno].es_valid = si->si_slots[slot_num].sl_valid;
	159	if (si->si_slots[slot_num].sl_valid)
	160	se->se_slots[slotno].es_node_num =
	161	cpu_to_le32(si->si_slots[slot_num].sl_node_num);
	162	*bh = si->si_bh[blkind];
	163	}
	164
	165	static void ocfs2_update_disk_slot_old(struct ocfs2_slot_info *si,
	166	int slot_num,
	167	struct buffer_head **bh)
ccd979bd	168	{
386a2ef8	169	int i;
fb86b1f0	170	struct ocfs2_slot_map *sm;
ccd979bd	171
fb86b1f0	172	sm = (struct ocfs2_slot_map *)si->si_bh[0]->b_data;
fc881fa0 JB	173	for (i = 0; i < si->si_num_slots; i++) {
fc881fa0 JB	174	if (si->si_slots[i].sl_valid)
fb86b1f0	175	sm->sm_slots[i] =
fc881fa0 JB	176	cpu_to_le16(si->si_slots[i].sl_node_num);
fc881fa0 JB	177	else
fb86b1f0	178	sm->sm_slots[i] = cpu_to_le16(OCFS2_INVALID_SLOT);
fc881fa0	179	}
386a2ef8 JB	180	*bh = si->si_bh[0];
	181	}
	182
	183	static int ocfs2_update_disk_slot(struct ocfs2_super *osb,
	184	struct ocfs2_slot_info *si,
	185	int slot_num)
	186	{
	187	int status;
	188	struct buffer_head *bh;
	189
	190	spin_lock(&osb->osb_lock);
	191	if (si->si_extended)
	192	ocfs2_update_disk_slot_extended(si, slot_num, &bh);
	193	else
	194	ocfs2_update_disk_slot_old(si, slot_num, &bh);
d85b20e4	195	spin_unlock(&osb->osb_lock);
ccd979bd	196
8cb471e8	197	status = ocfs2_write_block(osb, bh, INODE_CACHE(si->si_inode));
ccd979bd MF	198	if (status < 0)
	199	mlog_errno(status);
	200
	201	return status;
	202	}
	203
1c8d9a6a JB	204	/*
	205	* Calculate how many bytes are needed by the slot map. Returns
	206	* an error if the slot map file is too small.
	207	*/
	208	static int ocfs2_slot_map_physical_size(struct ocfs2_super *osb,
	209	struct inode *inode,
	210	unsigned long long *bytes)
	211	{
	212	unsigned long long bytes_needed;
	213
386a2ef8 JB	214	if (ocfs2_uses_extended_slot_map(osb)) {
	215	bytes_needed = osb->max_slots *
	216	sizeof(struct ocfs2_extended_slot);
	217	} else {
	218	bytes_needed = osb->max_slots * sizeof(__le16);
	219	}
1c8d9a6a JB	220	if (bytes_needed > i_size_read(inode)) {
	221	mlog(ML_ERROR,
	222	"Slot map file is too small! (size %llu, needed %llu)\n",
	223	i_size_read(inode), bytes_needed);
	224	return -ENOSPC;
	225	}
	226
	227	*bytes = bytes_needed;
	228	return 0;
	229	}
	230
fc881fa0 JB	231	/* try to find global node in the slot info. Returns -ENOENT
	232	* if nothing is found. */
	233	static int __ocfs2_node_num_to_slot(struct ocfs2_slot_info *si,
	234	unsigned int node_num)
ccd979bd	235	{
fc881fa0	236	int i, ret = -ENOENT;
ccd979bd MF	237
ccd979bd MF	238	for(i = 0; i < si->si_num_slots; i++) {
fc881fa0 JB	239	if (si->si_slots[i].sl_valid &&
	240	(node_num == si->si_slots[i].sl_node_num)) {
	241	ret = i;
ccd979bd MF	242	break;
	243	}
	244	}
fc881fa0	245
ccd979bd MF	246	return ret;
	247	}
	248
fc881fa0 JB	249	static int __ocfs2_find_empty_slot(struct ocfs2_slot_info *si,
fc881fa0 JB	250	int preferred)
ccd979bd	251	{
fc881fa0	252	int i, ret = -ENOSPC;
ccd979bd	253
fc881fa0	254	if ((preferred >= 0) && (preferred < si->si_num_slots)) {
c80af0c2	255	if (!si->si_slots[preferred].sl_valid) {
baf4661a SM	256	ret = preferred;
	257	goto out;
	258	}
	259	}
	260
ccd979bd	261	for(i = 0; i < si->si_num_slots; i++) {
c80af0c2	262	if (!si->si_slots[i].sl_valid) {
fc881fa0	263	ret = i;
ccd979bd MF	264	break;
	265	}
	266	}
baf4661a	267	out:
ccd979bd MF	268	return ret;
	269	}
	270
d85b20e4	271	int ocfs2_node_num_to_slot(struct ocfs2_super *osb, unsigned int node_num)
ccd979bd	272	{
fc881fa0	273	int slot;
d85b20e4	274	struct ocfs2_slot_info *si = osb->slot_info;
ccd979bd	275
d85b20e4 JB	276	spin_lock(&osb->osb_lock);
	277	slot = __ocfs2_node_num_to_slot(si, node_num);
	278	spin_unlock(&osb->osb_lock);
	279
d85b20e4 JB	280	return slot;
	281	}
	282
	283	int ocfs2_slot_to_node_num_locked(struct ocfs2_super *osb, int slot_num,
	284	unsigned int *node_num)
	285	{
	286	struct ocfs2_slot_info *si = osb->slot_info;
	287
	288	assert_spin_locked(&osb->osb_lock);
	289
	290	BUG_ON(slot_num < 0);
519a2861	291	BUG_ON(slot_num >= osb->max_slots);
d85b20e4	292
fc881fa0	293	if (!si->si_slots[slot_num].sl_valid)
d85b20e4 JB	294	return -ENOENT;
d85b20e4 JB	295
fc881fa0	296	*node_num = si->si_slots[slot_num].sl_node_num;
d85b20e4	297	return 0;
ccd979bd MF	298	}
ccd979bd MF	299
8e8a4603 MF	300	static void __ocfs2_free_slot_info(struct ocfs2_slot_info *si)
8e8a4603 MF	301	{
1c8d9a6a JB	302	unsigned int i;
1c8d9a6a JB	303
8e8a4603 MF	304	if (si == NULL)
	305	return;
	306
72865d92	307	iput(si->si_inode);
1c8d9a6a JB	308	if (si->si_bh) {
	309	for (i = 0; i < si->si_blocks; i++) {
	310	if (si->si_bh[i]) {
	311	brelse(si->si_bh[i]);
	312	si->si_bh[i] = NULL;
	313	}
	314	}
	315	kfree(si->si_bh);
	316	}
8e8a4603 MF	317
	318	kfree(si);
	319	}
	320
fc881fa0	321	int ocfs2_clear_slot(struct ocfs2_super *osb, int slot_num)
ccd979bd	322	{
8e8a4603 MF	323	struct ocfs2_slot_info *si = osb->slot_info;
	324
	325	if (si == NULL)
	326	return 0;
	327
d85b20e4	328	spin_lock(&osb->osb_lock);
fc881fa0	329	ocfs2_invalidate_slot(si, slot_num);
d85b20e4	330	spin_unlock(&osb->osb_lock);
8e8a4603	331
386a2ef8	332	return ocfs2_update_disk_slot(osb, osb->slot_info, slot_num);
ccd979bd MF	333	}
ccd979bd MF	334
1c8d9a6a JB	335	static int ocfs2_map_slot_buffers(struct ocfs2_super *osb,
	336	struct ocfs2_slot_info *si)
	337	{
	338	int status = 0;
	339	u64 blkno;
f30d44f3	340	unsigned long long blocks, bytes = 0;
1c8d9a6a JB	341	unsigned int i;
	342	struct buffer_head *bh;
	343
	344	status = ocfs2_slot_map_physical_size(osb, si->si_inode, &bytes);
	345	if (status)
	346	goto bail;
	347
	348	blocks = ocfs2_blocks_for_bytes(si->si_inode->i_sb, bytes);
	349	BUG_ON(blocks > UINT_MAX);
	350	si->si_blocks = blocks;
	351	if (!si->si_blocks)
	352	goto bail;
	353
386a2ef8 JB	354	if (si->si_extended)
	355	si->si_slots_per_block =
	356	(osb->sb->s_blocksize /
	357	sizeof(struct ocfs2_extended_slot));
	358	else
	359	si->si_slots_per_block = osb->sb->s_blocksize / sizeof(__le16);
	360
	361	/* The size checks above should ensure this */
	362	BUG_ON((osb->max_slots / si->si_slots_per_block) > blocks);
	363
a8731086	364	trace_ocfs2_map_slot_buffers(bytes, si->si_blocks);
1c8d9a6a	365
1b7f8ba6	366	si->si_bh = kcalloc(si->si_blocks, sizeof(struct buffer_head *),
1c8d9a6a JB	367	GFP_KERNEL);
	368	if (!si->si_bh) {
	369	status = -ENOMEM;
	370	mlog_errno(status);
	371	goto bail;
	372	}
	373
	374	for (i = 0; i < si->si_blocks; i++) {
	375	status = ocfs2_extent_map_get_blocks(si->si_inode, i,
	376	&blkno, NULL, NULL);
	377	if (status < 0) {
	378	mlog_errno(status);
	379	goto bail;
	380	}
	381
a8731086	382	trace_ocfs2_map_slot_buffers_block((unsigned long long)blkno, i);
1c8d9a6a JB	383
1c8d9a6a JB	384	bh = NULL; /* Acquire a fresh bh */
8cb471e8 JB	385	status = ocfs2_read_blocks(INODE_CACHE(si->si_inode), blkno,
8cb471e8 JB	386	1, &bh, OCFS2_BH_IGNORE_CACHE, NULL);
1c8d9a6a JB	387	if (status < 0) {
	388	mlog_errno(status);
	389	goto bail;
	390	}
	391
	392	si->si_bh[i] = bh;
	393	}
	394
	395	bail:
	396	return status;
	397	}
	398
ccd979bd MF	399	int ocfs2_init_slot_info(struct ocfs2_super *osb)
ccd979bd MF	400	{
fc881fa0	401	int status;
ccd979bd	402	struct inode *inode = NULL;
ccd979bd MF	403	struct ocfs2_slot_info *si;
ccd979bd MF	404
f402cf03	405	si = kzalloc(struct_size(si, si_slots, osb->max_slots), GFP_KERNEL);
ccd979bd MF	406	if (!si) {
	407	status = -ENOMEM;
	408	mlog_errno(status);
bb34ed21	409	return status;
ccd979bd MF	410	}
ccd979bd MF	411
386a2ef8	412	si->si_extended = ocfs2_uses_extended_slot_map(osb);
ccd979bd	413	si->si_num_slots = osb->max_slots;
ccd979bd MF	414
	415	inode = ocfs2_get_system_file_inode(osb, SLOT_MAP_SYSTEM_INODE,
	416	OCFS2_INVALID_SLOT);
	417	if (!inode) {
	418	status = -EINVAL;
	419	mlog_errno(status);
	420	goto bail;
	421	}
	422
1c8d9a6a JB	423	si->si_inode = inode;
1c8d9a6a JB	424	status = ocfs2_map_slot_buffers(osb, si);
ccd979bd MF	425	if (status < 0) {
	426	mlog_errno(status);
	427	goto bail;
	428	}
	429
d85b20e4	430	osb->slot_info = (struct ocfs2_slot_info *)si;
ccd979bd	431	bail:
fd90d4df	432	if (status < 0)
8e8a4603	433	__ocfs2_free_slot_info(si);
ccd979bd MF	434
	435	return status;
	436	}
	437
8e8a4603	438	void ocfs2_free_slot_info(struct ocfs2_super *osb)
ccd979bd	439	{
8e8a4603 MF	440	struct ocfs2_slot_info *si = osb->slot_info;
	441
	442	osb->slot_info = NULL;
	443	__ocfs2_free_slot_info(si);
ccd979bd MF	444	}
	445
	446	int ocfs2_find_slot(struct ocfs2_super *osb)
	447	{
	448	int status;
fc881fa0	449	int slot;
ccd979bd MF	450	struct ocfs2_slot_info *si;
ccd979bd MF	451
ccd979bd MF	452	si = osb->slot_info;
ccd979bd MF	453
d85b20e4	454	spin_lock(&osb->osb_lock);
ccd979bd MF	455	ocfs2_update_slot_info(si);
ccd979bd MF	456
c80af0c2 JB	457	/* search for ourselves first and take the slot if it already
	458	* exists. Perhaps we need to mark this in a variable for our
	459	* own journal recovery? Possibly not, though we certainly
	460	* need to warn to the user */
	461	slot = __ocfs2_node_num_to_slot(si, osb->node_num);
	462	if (slot < 0) {
	463	/* if no slot yet, then just take 1st available
	464	* one. */
	465	slot = __ocfs2_find_empty_slot(si, osb->preferred_slot);
fc881fa0	466	if (slot < 0) {
c80af0c2 JB	467	spin_unlock(&osb->osb_lock);
	468	mlog(ML_ERROR, "no free slots available!\n");
	469	status = -EINVAL;
	470	goto bail;
	471	}
	472	} else
	473	printk(KERN_INFO "ocfs2: Slot %d on device (%s) was already "
	474	"allocated to this node!\n", slot, osb->dev_str);
ccd979bd	475
fc881fa0	476	ocfs2_set_slot(si, slot, osb->node_num);
ccd979bd	477	osb->slot_num = slot;
d85b20e4	478	spin_unlock(&osb->osb_lock);
ccd979bd	479
a8731086	480	trace_ocfs2_find_slot(osb->slot_num);
ccd979bd	481
386a2ef8	482	status = ocfs2_update_disk_slot(osb, si, osb->slot_num);
1247017f	483	if (status < 0) {
ccd979bd	484	mlog_errno(status);
1247017f	485	/*
	486	* if write block failed, invalidate slot to avoid overwrite
	487	* slot during dismount in case another node rightly has mounted
	488	*/
	489	spin_lock(&osb->osb_lock);
	490	ocfs2_invalidate_slot(si, osb->slot_num);
	491	osb->slot_num = OCFS2_INVALID_SLOT;
	492	spin_unlock(&osb->osb_lock);
	493	}
ccd979bd MF	494
ccd979bd MF	495	bail:
ccd979bd MF	496	return status;
	497	}
	498
	499	void ocfs2_put_slot(struct ocfs2_super *osb)
	500	{
386a2ef8	501	int status, slot_num;
ccd979bd MF	502	struct ocfs2_slot_info *si = osb->slot_info;
	503
	504	if (!si)
	505	return;
	506
d85b20e4	507	spin_lock(&osb->osb_lock);
ccd979bd MF	508	ocfs2_update_slot_info(si);
ccd979bd MF	509
386a2ef8	510	slot_num = osb->slot_num;
fc881fa0	511	ocfs2_invalidate_slot(si, osb->slot_num);
ccd979bd	512	osb->slot_num = OCFS2_INVALID_SLOT;
d85b20e4	513	spin_unlock(&osb->osb_lock);
ccd979bd	514
386a2ef8	515	status = ocfs2_update_disk_slot(osb, si, slot_num);
8f9b1802	516	if (status < 0)
ccd979bd	517	mlog_errno(status);
ccd979bd	518
8e8a4603	519	ocfs2_free_slot_info(osb);
ccd979bd	520	}