balloc.c

linux 内核源代码

/*
 *  linux/fs/ext4/balloc.c
 *
 * Copyright (C) 1992, 1993, 1994, 1995
 * Remy Card ([e-mail])
 * Laboratoire MASI - Institut Blaise Pascal
 * Universite Pierre et Marie Curie (Paris VI)
 *
 *  Enhanced block allocation by Stephen Tweedie ([e-mail]), 1993
 *  Big-endian to little-endian byte-swapping/bitmaps by
 *        David S. Miller ([e-mail]), 1995
 */

#include <linux/time.h>
#include <linux/capability.h>
#include <linux/fs.h>
#include <linux/jbd2.h>
#include <linux/ext4_fs.h>
#include <linux/ext4_jbd2.h>
#include <linux/quotaops.h>
#include <linux/buffer_head.h>

#include "group.h"
/*
 * balloc.c contains the blocks allocation and deallocation routines
 */

/*
 * Calculate the block group number and offset, given a block number
 */
void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
		unsigned long *blockgrpp, ext4_grpblk_t *offsetp)
{
	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
	ext4_grpblk_t offset;

	blocknr = blocknr - le32_to_cpu(es->s_first_data_block);
	offset = do_div(blocknr, EXT4_BLOCKS_PER_GROUP(sb));
	if (offsetp)
		*offsetp = offset;
	if (blockgrpp)
		*blockgrpp = blocknr;

}

/* Initializes an uninitialized block bitmap if given, and returns the
 * number of blocks free in the group. */
unsigned ext4_init_block_bitmap(struct super_block *sb, struct buffer_head *bh,
				int block_group, struct ext4_group_desc *gdp)
{
	unsigned long start;
	int bit, bit_max;
	unsigned free_blocks, group_blocks;
	struct ext4_sb_info *sbi = EXT4_SB(sb);

	if (bh) {
		J_ASSERT_BH(bh, buffer_locked(bh));

		/* If checksum is bad mark all blocks used to prevent allocation
		 * essentially implementing a per-group read-only flag. */
		if (!ext4_group_desc_csum_verify(sbi, block_group, gdp)) {
			ext4_error(sb, __FUNCTION__,
				   "Checksum bad for group %u\n", block_group);
			gdp->bg_free_blocks_count = 0;
			gdp->bg_free_inodes_count = 0;
			gdp->bg_itable_unused = 0;
			memset(bh->b_data, 0xff, sb->s_blocksize);
			return 0;
		}
		memset(bh->b_data, 0, sb->s_blocksize);
	}

	/* Check for superblock and gdt backups in this group */
	bit_max = ext4_bg_has_super(sb, block_group);

	if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
	    block_group < le32_to_cpu(sbi->s_es->s_first_meta_bg) *
			  sbi->s_desc_per_block) {
		if (bit_max) {
			bit_max += ext4_bg_num_gdb(sb, block_group);
			bit_max +=
				le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks);
		}
	} else { /* For META_BG_BLOCK_GROUPS */
		int group_rel = (block_group -
				 le32_to_cpu(sbi->s_es->s_first_meta_bg)) %
				EXT4_DESC_PER_BLOCK(sb);
		if (group_rel == 0 || group_rel == 1 ||
		    (group_rel == EXT4_DESC_PER_BLOCK(sb) - 1))
			bit_max += 1;
	}

	if (block_group == sbi->s_groups_count - 1) {
		/*
		 * Even though mke2fs always initialize first and last group
		 * if some other tool enabled the EXT4_BG_BLOCK_UNINIT we need
		 * to make sure we calculate the right free blocks
		 */
		group_blocks = ext4_blocks_count(sbi->s_es) -
			le32_to_cpu(sbi->s_es->s_first_data_block) -
			(EXT4_BLOCKS_PER_GROUP(sb) * (sbi->s_groups_count -1));
	} else {
		group_blocks = EXT4_BLOCKS_PER_GROUP(sb);
	}

	free_blocks = group_blocks - bit_max;

	if (bh) {
		for (bit = 0; bit < bit_max; bit++)
			ext4_set_bit(bit, bh->b_data);

		start = block_group * EXT4_BLOCKS_PER_GROUP(sb) +
			le32_to_cpu(sbi->s_es->s_first_data_block);

		/* Set bits for block and inode bitmaps, and inode table */
		ext4_set_bit(ext4_block_bitmap(sb, gdp) - start, bh->b_data);
		ext4_set_bit(ext4_inode_bitmap(sb, gdp) - start, bh->b_data);
		for (bit = (ext4_inode_table(sb, gdp) - start),
		     bit_max = bit + sbi->s_itb_per_group; bit < bit_max; bit++)
			ext4_set_bit(bit, bh->b_data);

		/*
		 * Also if the number of blocks within the group is
		 * less than the blocksize * 8 ( which is the size
		 * of bitmap ), set rest of the block bitmap to 1
		 */
		mark_bitmap_end(group_blocks, sb->s_blocksize * 8, bh->b_data);
	}

	return free_blocks - sbi->s_itb_per_group - 2;
}


/*
 * The free blocks are managed by bitmaps.  A file system contains several
 * blocks groups.  Each group contains 1 bitmap block for blocks, 1 bitmap
 * block for inodes, N blocks for the inode table and data blocks.
 *
 * The file system contains group descriptors which are located after the
 * super block.  Each descriptor contains the number of the bitmap block and
 * the free blocks count in the block.  The descriptors are loaded in memory
 * when a file system is mounted (see ext4_fill_super).
 */


#define in_range(b, first, len)	((b) >= (first) && (b) <= (first) + (len) - 1)

/**
 * ext4_get_group_desc() -- load group descriptor from disk
 * @sb:			super block
 * @block_group:	given block group
 * @bh:			pointer to the buffer head to store the block
 *			group descriptor
 */
struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
					     unsigned int block_group,
					     struct buffer_head ** bh)
{
	unsigned long group_desc;
	unsigned long offset;
	struct ext4_group_desc * desc;
	struct ext4_sb_info *sbi = EXT4_SB(sb);

	if (block_group >= sbi->s_groups_count) {
		ext4_error (sb, "ext4_get_group_desc",
			    "block_group >= groups_count - "
			    "block_group = %d, groups_count = %lu",
			    block_group, sbi->s_groups_count);

		return NULL;
	}
	smp_rmb();

	group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb);
	offset = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1);
	if (!sbi->s_group_desc[group_desc]) {
		ext4_error (sb, "ext4_get_group_desc",
			    "Group descriptor not loaded - "
			    "block_group = %d, group_desc = %lu, desc = %lu",
			     block_group, group_desc, offset);
		return NULL;
	}

	desc = (struct ext4_group_desc *)(
		(__u8 *)sbi->s_group_desc[group_desc]->b_data +
		offset * EXT4_DESC_SIZE(sb));
	if (bh)
		*bh = sbi->s_group_desc[group_desc];
	return desc;
}

/**
 * read_block_bitmap()
 * @sb:			super block
 * @block_group:	given block group
 *
 * Read the bitmap for a given block_group, reading into the specified
 * slot in the superblock's bitmap cache.
 *
 * Return buffer_head on success or NULL in case of failure.
 */
struct buffer_head *
read_block_bitmap(struct super_block *sb, unsigned int block_group)
{
	struct ext4_group_desc * desc;
	struct buffer_head * bh = NULL;
	ext4_fsblk_t bitmap_blk;

	desc = ext4_get_group_desc(sb, block_group, NULL);
	if (!desc)
		return NULL;
	bitmap_blk = ext4_block_bitmap(sb, desc);
	if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) {
		bh = sb_getblk(sb, bitmap_blk);
		if (!buffer_uptodate(bh)) {
			lock_buffer(bh);
			if (!buffer_uptodate(bh)) {
				ext4_init_block_bitmap(sb, bh, block_group,
						       desc);
				set_buffer_uptodate(bh);
			}
			unlock_buffer(bh);
		}
	} else {
		bh = sb_bread(sb, bitmap_blk);
	}
	if (!bh)
		ext4_error (sb, __FUNCTION__,
			    "Cannot read block bitmap - "
			    "block_group = %d, block_bitmap = %llu",
			    block_group, bitmap_blk);
	return bh;
}
/*
 * The reservation window structure operations
 * --------------------------------------------
 * Operations include:
 * dump, find, add, remove, is_empty, find_next_reservable_window, etc.
 *
 * We use a red-black tree to represent per-filesystem reservation
 * windows.
 *
 */

/**
 * __rsv_window_dump() -- Dump the filesystem block allocation reservation map
 * @rb_root:		root of per-filesystem reservation rb tree
 * @verbose:		verbose mode
 * @fn:			function which wishes to dump the reservation map
 *
 * If verbose is turned on, it will print the whole block reservation
 * windows(start, end).	Otherwise, it will only print out the "bad" windows,
 * those windows that overlap with their immediate neighbors.
 */
#if 1
static void __rsv_window_dump(struct rb_root *root, int verbose,
			      const char *fn)
{
	struct rb_node *n;
	struct ext4_reserve_window_node *rsv, *prev;
	int bad;

restart:
	n = rb_first(root);
	bad = 0;
	prev = NULL;

	printk("Block Allocation Reservation Windows Map (%s):\n", fn);
	while (n) {
		rsv = rb_entry(n, struct ext4_reserve_window_node, rsv_node);
		if (verbose)
			printk("reservation window 0x%p "
			       "start:  %llu, end:  %llu\n",
			       rsv, rsv->rsv_start, rsv->rsv_end);
		if (rsv->rsv_start && rsv->rsv_start >= rsv->rsv_end) {
			printk("Bad reservation %p (start >= end)\n",
			       rsv);
			bad = 1;
		}
		if (prev && prev->rsv_end >= rsv->rsv_start) {
			printk("Bad reservation %p (prev->end >= start)\n",
			       rsv);
			bad = 1;
		}
		if (bad) {
			if (!verbose) {
				printk("Restarting reservation walk in verbose mode\n");
				verbose = 1;
				goto restart;
			}
		}
		n = rb_next(n);
		prev = rsv;
	}
	printk("Window map complete.\n");
	if (bad)
		BUG();
}
#define rsv_window_dump(root, verbose) \
	__rsv_window_dump((root), (verbose), __FUNCTION__)
#else
#define rsv_window_dump(root, verbose) do {} while (0)
#endif

/**
 * goal_in_my_reservation()
 * @rsv:		inode's reservation window
 * @grp_goal:		given goal block relative to the allocation block group
 * @group:		the current allocation block group
 * @sb:			filesystem super block
 *
 * Test if the given goal block (group relative) is within the file's
 * own block reservation window range.
 *
 * If the reservation window is outside the goal allocation group, return 0;
 * grp_goal (given goal block) could be -1, which means no specific
 * goal block. In this case, always return 1.
 * If the goal block is within the reservation window, return 1;
 * otherwise, return 0;
 */
static int
goal_in_my_reservation(struct ext4_reserve_window *rsv, ext4_grpblk_t grp_goal,
			unsigned int group, struct super_block * sb)
{
	ext4_fsblk_t group_first_block, group_last_block;

	group_first_block = ext4_group_first_block_no(sb, group);
	group_last_block = group_first_block + (EXT4_BLOCKS_PER_GROUP(sb) - 1);

	if ((rsv->_rsv_start > group_last_block) ||
	    (rsv->_rsv_end < group_first_block))
		return 0;
	if ((grp_goal >= 0) && ((grp_goal + group_first_block < rsv->_rsv_start)
		|| (grp_goal + group_first_block > rsv->_rsv_end)))
		return 0;
	return 1;
}

/**
 * search_reserve_window()
 * @rb_root:		root of reservation tree
 * @goal:		target allocation block
 *
 * Find the reserved window which includes the goal, or the previous one
 * if the goal is not in any window.
 * Returns NULL if there are no windows or if all windows start after the goal.
 */
static struct ext4_reserve_window_node *
search_reserve_window(struct rb_root *root, ext4_fsblk_t goal)
{
	struct rb_node *n = root->rb_node;
	struct ext4_reserve_window_node *rsv;

	if (!n)
		return NULL;

	do {
		rsv = rb_entry(n, struct ext4_reserve_window_node, rsv_node);

		if (goal < rsv->rsv_start)
			n = n->rb_left;
		else if (goal > rsv->rsv_end)
			n = n->rb_right;
		else
			return rsv;
	} while (n);
	/*
	 * We've fallen off the end of the tree: the goal wasn't inside
	 * any particular node.  OK, the previous node must be to one
	 * side of the interval containing the goal.  If it's the RHS,
	 * we need to back up one.
	 */
	if (rsv->rsv_start > goal) {
		n = rb_prev(&rsv->rsv_node);
		rsv = rb_entry(n, struct ext4_reserve_window_node, rsv_node);
	}
	return rsv;
}

/**
 * ext4_rsv_window_add() -- Insert a window to the block reservation rb tree.
 * @sb:			super block
 * @rsv:		reservation window to add
 *
 * Must be called with rsv_lock hold.
 */
void ext4_rsv_window_add(struct super_block *sb,
		    struct ext4_reserve_window_node *rsv)
{
	struct rb_root *root = &EXT4_SB(sb)->s_rsv_window_root;
	struct rb_node *node = &rsv->rsv_node;
	ext4_fsblk_t start = rsv->rsv_start;

	struct rb_node ** p = &root->rb_node;
	struct rb_node * parent = NULL;
	struct ext4_reserve_window_node *this;

	while (*p)
	{
		parent = *p;
		this = rb_entry(parent, struct ext4_reserve_window_node, rsv_node);

		if (start < this->rsv_start)
			p = &(*p)->rb_left;
		else if (start > this->rsv_end)
			p = &(*p)->rb_right;
		else {
			rsv_window_dump(root, 1);
			BUG();
		}
	}

	rb_link_node(node, parent, p);
	rb_insert_color(node, root);
}

/**
 * ext4_rsv_window_remove() -- unlink a window from the reservation rb tree
 * @sb:			super block
 * @rsv:		reservation window to remove
 *
 * Mark the block reservation window as not allocated, and unlink it
 * from the filesystem reservation window rb tree. Must be called with
 * rsv_lock hold.
 */
static void rsv_window_remove(struct super_block *sb,
			      struct ext4_reserve_window_node *rsv)
{
	rsv->rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
	rsv->rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
	rsv->rsv_alloc_hit = 0;
	rb_erase(&rsv->rsv_node, &EXT4_SB(sb)->s_rsv_window_root);
}

/*
 * rsv_is_empty() -- Check if the reservation window is allocated.
 * @rsv:		given reservation window to check
 *
 * returns 1 if the end block is EXT4_RESERVE_WINDOW_NOT_ALLOCATED.
 */
static inline int rsv_is_empty(struct ext4_reserve_window *rsv)
{
	/* a valid reservation end block could not be 0 */
	return rsv->_rsv_end == EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
}

/**
 * ext4_init_block_alloc_info()
 * @inode:		file inode structure
 *
 * Allocate and initialize the	reservation window structure, and
 * link the window to the ext4 inode structure at last
 *
 * The reservation window structure is only dynamically allocated
 * and linked to ext4 inode the first time the open file
 * needs a new block. So, before every ext4_new_block(s) call, for
 * regular files, we should check whether the reservation window
 * structure exists or not. In the latter case, this function is called.
 * Fail to do so will result in block reservation being turned off for that
 * open file.
 *
 * This function is called from ext4_get_blocks_handle(), also called
 * when setting the reservation window size through ioctl before the file
 * is open for write (needs block allocation).
 *
 * Needs truncate_mutex protection prior to call this function.
 */
void ext4_init_block_alloc_info(struct inode *inode)
{
	struct ext4_inode_info *ei = EXT4_I(inode);
	struct ext4_block_alloc_info *block_i = ei->i_block_alloc_info;
	struct super_block *sb = inode->i_sb;

	block_i = kmalloc(sizeof(*block_i), GFP_NOFS);
	if (block_i) {
		struct ext4_reserve_window_node *rsv = &block_i->rsv_window_node;

		rsv->rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
		rsv->rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;

		/*
		 * if filesystem is mounted with NORESERVATION, the goal
		 * reservation window size is set to zero to indicate
		 * block reservation is off