[net-next-2.6.git] / fs / ext4 / block_validity.c

/*
 *  linux/fs/ext4/block_validity.c
 *
 * Copyright (C) 2009
 * Theodore Ts'o (tytso@mit.edu)
 *
 * Track which blocks in the filesystem are metadata blocks that
 * should never be used as data blocks by files or directories.
 */

#include <linux/time.h>
#include <linux/fs.h>
#include <linux/namei.h>
#include <linux/quotaops.h>
#include <linux/buffer_head.h>
#include <linux/module.h>
#include <linux/swap.h>
#include <linux/pagemap.h>
#include <linux/blkdev.h>
#include <linux/mutex.h>
#include <linux/slab.h>
#include "ext4.h"

struct ext4_system_zone {
	struct rb_node	node;
	ext4_fsblk_t	start_blk;
	unsigned int	count;
};

static struct kmem_cache *ext4_system_zone_cachep;

int __init ext4_init_system_zone(void)
{
	ext4_system_zone_cachep = KMEM_CACHE(ext4_system_zone, 0);
	if (ext4_system_zone_cachep == NULL)
		return -ENOMEM;
	return 0;
}

void ext4_exit_system_zone(void)
{
	kmem_cache_destroy(ext4_system_zone_cachep);
}

static inline int can_merge(struct ext4_system_zone *entry1,
		     struct ext4_system_zone *entry2)
{
	if ((entry1->start_blk + entry1->count) == entry2->start_blk)
		return 1;
	return 0;
}

/*
 * Mark a range of blocks as belonging to the "system zone" --- that
 * is, filesystem metadata blocks which should never be used by
 * inodes.
 */
static int add_system_zone(struct ext4_sb_info *sbi,
			   ext4_fsblk_t start_blk,
			   unsigned int count)
{
	struct ext4_system_zone *new_entry = NULL, *entry;
	struct rb_node **n = &sbi->system_blks.rb_node, *node;
	struct rb_node *parent = NULL, *new_node = NULL;

	while (*n) {
		parent = *n;
		entry = rb_entry(parent, struct ext4_system_zone, node);
		if (start_blk < entry->start_blk)
			n = &(*n)->rb_left;
		else if (start_blk >= (entry->start_blk + entry->count))
			n = &(*n)->rb_right;
		else {
			if (start_blk + count > (entry->start_blk +
						 entry->count))
				entry->count = (start_blk + count -
						entry->start_blk);
			new_node = *n;
			new_entry = rb_entry(new_node, struct ext4_system_zone,
					     node);
			break;
		}
	}

	if (!new_entry) {
		new_entry = kmem_cache_alloc(ext4_system_zone_cachep,
					     GFP_KERNEL);
		if (!new_entry)
			return -ENOMEM;
		new_entry->start_blk = start_blk;
		new_entry->count = count;
		new_node = &new_entry->node;

		rb_link_node(new_node, parent, n);
		rb_insert_color(new_node, &sbi->system_blks);
	}

	/* Can we merge to the left? */
	node = rb_prev(new_node);
	if (node) {
		entry = rb_entry(node, struct ext4_system_zone, node);
		if (can_merge(entry, new_entry)) {
			new_entry->start_blk = entry->start_blk;
			new_entry->count += entry->count;
			rb_erase(node, &sbi->system_blks);
			kmem_cache_free(ext4_system_zone_cachep, entry);
		}
	}

	/* Can we merge to the right? */
	node = rb_next(new_node);
	if (node) {
		entry = rb_entry(node, struct ext4_system_zone, node);
		if (can_merge(new_entry, entry)) {
			new_entry->count += entry->count;
			rb_erase(node, &sbi->system_blks);
			kmem_cache_free(ext4_system_zone_cachep, entry);
		}
	}
	return 0;
}

static void debug_print_tree(struct ext4_sb_info *sbi)
{
	struct rb_node *node;
	struct ext4_system_zone *entry;
	int first = 1;

	printk(KERN_INFO "System zones: ");
	node = rb_first(&sbi->system_blks);
	while (node) {
		entry = rb_entry(node, struct ext4_system_zone, node);
		printk("%s%llu-%llu", first ? "" : ", ",
		       entry->start_blk, entry->start_blk + entry->count - 1);
		first = 0;
		node = rb_next(node);
	}
	printk("\n");
}

int ext4_setup_system_zone(struct super_block *sb)
{
	ext4_group_t ngroups = ext4_get_groups_count(sb);
	struct ext4_sb_info *sbi = EXT4_SB(sb);
	struct ext4_group_desc *gdp;
	ext4_group_t i;
	int flex_size = ext4_flex_bg_size(sbi);
	int ret;

	if (!test_opt(sb, BLOCK_VALIDITY)) {
		if (EXT4_SB(sb)->system_blks.rb_node)
			ext4_release_system_zone(sb);
		return 0;
	}
	if (EXT4_SB(sb)->system_blks.rb_node)
		return 0;

	for (i=0; i < ngroups; i++) {
		if (ext4_bg_has_super(sb, i) &&
		    ((i < 5) || ((i % flex_size) == 0)))
			add_system_zone(sbi, ext4_group_first_block_no(sb, i),
					ext4_bg_num_gdb(sb, i) + 1);
		gdp = ext4_get_group_desc(sb, i, NULL);
		ret = add_system_zone(sbi, ext4_block_bitmap(sb, gdp), 1);
		if (ret)
			return ret;
		ret = add_system_zone(sbi, ext4_inode_bitmap(sb, gdp), 1);
		if (ret)
			return ret;
		ret = add_system_zone(sbi, ext4_inode_table(sb, gdp),
				sbi->s_itb_per_group);
		if (ret)
			return ret;
	}

	if (test_opt(sb, DEBUG))
		debug_print_tree(EXT4_SB(sb));
	return 0;
}

/* Called when the filesystem is unmounted */
void ext4_release_system_zone(struct super_block *sb)
{
	struct rb_node	*n = EXT4_SB(sb)->system_blks.rb_node;
	struct rb_node	*parent;
	struct ext4_system_zone	*entry;

	while (n) {
		/* Do the node's children first */
		if (n->rb_left) {
			n = n->rb_left;
			continue;
		}
		if (n->rb_right) {
			n = n->rb_right;
			continue;
		}
		/*
		 * The node has no children; free it, and then zero
		 * out parent's link to it.  Finally go to the
		 * beginning of the loop and try to free the parent
		 * node.
		 */
		parent = rb_parent(n);
		entry = rb_entry(n, struct ext4_system_zone, node);
		kmem_cache_free(ext4_system_zone_cachep, entry);
		if (!parent)
			EXT4_SB(sb)->system_blks = RB_ROOT;
		else if (parent->rb_left == n)
			parent->rb_left = NULL;
		else if (parent->rb_right == n)
			parent->rb_right = NULL;
		n = parent;
	}
	EXT4_SB(sb)->system_blks = RB_ROOT;
}

/*
 * Returns 1 if the passed-in block region (start_blk,
 * start_blk+count) is valid; 0 if some part of the block region
 * overlaps with filesystem metadata blocks.
 */
int ext4_data_block_valid(struct ext4_sb_info *sbi, ext4_fsblk_t start_blk,
			  unsigned int count)
{
	struct ext4_system_zone *entry;
	struct rb_node *n = sbi->system_blks.rb_node;

	if ((start_blk <= le32_to_cpu(sbi->s_es->s_first_data_block)) ||
	    (start_blk + count < start_blk) ||
	    (start_blk + count > ext4_blocks_count(sbi->s_es))) {
		sbi->s_es->s_last_error_block = cpu_to_le64(start_blk);
		return 0;
	}
	while (n) {
		entry = rb_entry(n, struct ext4_system_zone, node);
		if (start_blk + count - 1 < entry->start_blk)
			n = n->rb_left;
		else if (start_blk >= (entry->start_blk + entry->count))
			n = n->rb_right;
		else {
			sbi->s_es->s_last_error_block = cpu_to_le64(start_blk);
			return 0;
		}
	}
	return 1;
}
Commit	Line	Data
6fd058f7 TT	1	/*
	2	* linux/fs/ext4/block_validity.c
	3	*
	4	* Copyright (C) 2009
	5	* Theodore Ts'o (tytso@mit.edu)
	6	*
	7	* Track which blocks in the filesystem are metadata blocks that
	8	* should never be used as data blocks by files or directories.
	9	*/
	10
	11	#include <linux/time.h>
	12	#include <linux/fs.h>
	13	#include <linux/namei.h>
	14	#include <linux/quotaops.h>
	15	#include <linux/buffer_head.h>
	16	#include <linux/module.h>
	17	#include <linux/swap.h>
	18	#include <linux/pagemap.h>
6fd058f7 TT	19	#include <linux/blkdev.h>
6fd058f7 TT	20	#include <linux/mutex.h>
5a0e3ad6	21	#include <linux/slab.h>
6fd058f7 TT	22	#include "ext4.h"
	23
	24	struct ext4_system_zone {
	25	struct rb_node node;
	26	ext4_fsblk_t start_blk;
	27	unsigned int count;
	28	};
	29
	30	static struct kmem_cache *ext4_system_zone_cachep;
	31
5dabfc78	32	int __init ext4_init_system_zone(void)
6fd058f7	33	{
16828088	34	ext4_system_zone_cachep = KMEM_CACHE(ext4_system_zone, 0);
6fd058f7 TT	35	if (ext4_system_zone_cachep == NULL)
	36	return -ENOMEM;
	37	return 0;
	38	}
	39
5dabfc78	40	void ext4_exit_system_zone(void)
6fd058f7 TT	41	{
	42	kmem_cache_destroy(ext4_system_zone_cachep);
	43	}
	44
	45	static inline int can_merge(struct ext4_system_zone *entry1,
	46	struct ext4_system_zone *entry2)
	47	{
	48	if ((entry1->start_blk + entry1->count) == entry2->start_blk)
	49	return 1;
	50	return 0;
	51	}
	52
	53	/*
	54	* Mark a range of blocks as belonging to the "system zone" --- that
	55	* is, filesystem metadata blocks which should never be used by
	56	* inodes.
	57	*/
	58	static int add_system_zone(struct ext4_sb_info *sbi,
	59	ext4_fsblk_t start_blk,
	60	unsigned int count)
	61	{
	62	struct ext4_system_zone new_entry = NULL, entry;
	63	struct rb_node *n = &sbi->system_blks.rb_node, node;
	64	struct rb_node parent = NULL, new_node = NULL;
	65
	66	while (*n) {
	67	parent = *n;
	68	entry = rb_entry(parent, struct ext4_system_zone, node);
	69	if (start_blk < entry->start_blk)
	70	n = &(*n)->rb_left;
	71	else if (start_blk >= (entry->start_blk + entry->count))
	72	n = &(*n)->rb_right;
	73	else {
60e6679e	74	if (start_blk + count > (entry->start_blk +
6fd058f7	75	entry->count))
60e6679e	76	entry->count = (start_blk + count -
6fd058f7 TT	77	entry->start_blk);
	78	new_node = *n;
	79	new_entry = rb_entry(new_node, struct ext4_system_zone,
	80	node);
	81	break;
	82	}
	83	}
	84
	85	if (!new_entry) {
	86	new_entry = kmem_cache_alloc(ext4_system_zone_cachep,
	87	GFP_KERNEL);
	88	if (!new_entry)
	89	return -ENOMEM;
	90	new_entry->start_blk = start_blk;
	91	new_entry->count = count;
	92	new_node = &new_entry->node;
	93
	94	rb_link_node(new_node, parent, n);
	95	rb_insert_color(new_node, &sbi->system_blks);
	96	}
	97
	98	/* Can we merge to the left? */
	99	node = rb_prev(new_node);
	100	if (node) {
	101	entry = rb_entry(node, struct ext4_system_zone, node);
	102	if (can_merge(entry, new_entry)) {
	103	new_entry->start_blk = entry->start_blk;
	104	new_entry->count += entry->count;
	105	rb_erase(node, &sbi->system_blks);
	106	kmem_cache_free(ext4_system_zone_cachep, entry);
	107	}
	108	}
	109
	110	/* Can we merge to the right? */
	111	node = rb_next(new_node);
	112	if (node) {
	113	entry = rb_entry(node, struct ext4_system_zone, node);
	114	if (can_merge(new_entry, entry)) {
	115	new_entry->count += entry->count;
	116	rb_erase(node, &sbi->system_blks);
	117	kmem_cache_free(ext4_system_zone_cachep, entry);
	118	}
	119	}
	120	return 0;
	121	}
	122
	123	static void debug_print_tree(struct ext4_sb_info *sbi)
	124	{
	125	struct rb_node *node;
	126	struct ext4_system_zone *entry;
	127	int first = 1;
	128
	129	printk(KERN_INFO "System zones: ");
	130	node = rb_first(&sbi->system_blks);
	131	while (node) {
	132	entry = rb_entry(node, struct ext4_system_zone, node);
	133	printk("%s%llu-%llu", first ? "" : ", ",
	134	entry->start_blk, entry->start_blk + entry->count - 1);
	135	first = 0;
	136	node = rb_next(node);
	137	}
	138	printk("\n");
	139	}
	140
141	int ext4_setup_system_zone(struct super_block *sb)
142	{
143	ext4_group_t ngroups = ext4_get_groups_count(sb);
144	struct ext4_sb_info *sbi = EXT4_SB(sb);
145	struct ext4_group_desc *gdp;
146	ext4_group_t i;
147	int flex_size = ext4_flex_bg_size(sbi);
148	int ret;
149
150	if (!test_opt(sb, BLOCK_VALIDITY)) {
151	if (EXT4_SB(sb)->system_blks.rb_node)
152	ext4_release_system_zone(sb);
153	return 0;
154	}
155	if (EXT4_SB(sb)->system_blks.rb_node)
156	return 0;
157
158	for (i=0; i < ngroups; i++) {
159	if (ext4_bg_has_super(sb, i) &&
160	((i < 5) \|\| ((i % flex_size) == 0)))
161	add_system_zone(sbi, ext4_group_first_block_no(sb, i),
1032988c	162	ext4_bg_num_gdb(sb, i) + 1);
6fd058f7 TT	163	gdp = ext4_get_group_desc(sb, i, NULL);
	164	ret = add_system_zone(sbi, ext4_block_bitmap(sb, gdp), 1);
	165	if (ret)
	166	return ret;
	167	ret = add_system_zone(sbi, ext4_inode_bitmap(sb, gdp), 1);
	168	if (ret)
	169	return ret;
	170	ret = add_system_zone(sbi, ext4_inode_table(sb, gdp),
	171	sbi->s_itb_per_group);
	172	if (ret)
	173	return ret;
	174	}
	175
	176	if (test_opt(sb, DEBUG))
	177	debug_print_tree(EXT4_SB(sb));
	178	return 0;
	179	}
	180
	181	/* Called when the filesystem is unmounted */
	182	void ext4_release_system_zone(struct super_block *sb)
	183	{
	184	struct rb_node *n = EXT4_SB(sb)->system_blks.rb_node;
	185	struct rb_node *parent;
	186	struct ext4_system_zone *entry;
	187
	188	while (n) {
	189	/* Do the node's children first */
	190	if (n->rb_left) {
	191	n = n->rb_left;
	192	continue;
	193	}
	194	if (n->rb_right) {
	195	n = n->rb_right;
	196	continue;
	197	}
	198	/*
	199	* The node has no children; free it, and then zero
	200	* out parent's link to it. Finally go to the
	201	* beginning of the loop and try to free the parent
	202	* node.
	203	*/
	204	parent = rb_parent(n);
	205	entry = rb_entry(n, struct ext4_system_zone, node);
	206	kmem_cache_free(ext4_system_zone_cachep, entry);
	207	if (!parent)
64e290ec	208	EXT4_SB(sb)->system_blks = RB_ROOT;
6fd058f7 TT	209	else if (parent->rb_left == n)
	210	parent->rb_left = NULL;
	211	else if (parent->rb_right == n)
	212	parent->rb_right = NULL;
	213	n = parent;
	214	}
64e290ec	215	EXT4_SB(sb)->system_blks = RB_ROOT;
6fd058f7 TT	216	}
	217
	218	/*
	219	* Returns 1 if the passed-in block region (start_blk,
	220	* start_blk+count) is valid; 0 if some part of the block region
	221	* overlaps with filesystem metadata blocks.
	222	*/
	223	int ext4_data_block_valid(struct ext4_sb_info *sbi, ext4_fsblk_t start_blk,
	224	unsigned int count)
	225	{
	226	struct ext4_system_zone *entry;
	227	struct rb_node *n = sbi->system_blks.rb_node;
	228
	229	if ((start_blk <= le32_to_cpu(sbi->s_es->s_first_data_block)) \|\|
1585d8d8	230	(start_blk + count < start_blk) \|\|
1c13d5c0 TT	231	(start_blk + count > ext4_blocks_count(sbi->s_es))) {
1c13d5c0 TT	232	sbi->s_es->s_last_error_block = cpu_to_le64(start_blk);
6fd058f7	233	return 0;
1c13d5c0	234	}
6fd058f7 TT	235	while (n) {
	236	entry = rb_entry(n, struct ext4_system_zone, node);
	237	if (start_blk + count - 1 < entry->start_blk)
	238	n = n->rb_left;
	239	else if (start_blk >= (entry->start_blk + entry->count))
	240	n = n->rb_right;
1c13d5c0 TT	241	else {
1c13d5c0 TT	242	sbi->s_es->s_last_error_block = cpu_to_le64(start_blk);
6fd058f7	243	return 0;
1c13d5c0	244	}
6fd058f7 TT	245	}
	246	return 1;
	247	}
	248