balloc.c

linux 内核源代码

		 */
		if (!test_opt(sb, RESERVATION))
			rsv->rsv_goal_size = 0;
		else
			rsv->rsv_goal_size = EXT4_DEFAULT_RESERVE_BLOCKS;
		rsv->rsv_alloc_hit = 0;
		block_i->last_alloc_logical_block = 0;
		block_i->last_alloc_physical_block = 0;
	}
	ei->i_block_alloc_info = block_i;
}

/**
 * ext4_discard_reservation()
 * @inode:		inode
 *
 * Discard(free) block reservation window on last file close, or truncate
 * or at last iput().
 *
 * It is being called in three cases:
 *	ext4_release_file(): last writer close the file
 *	ext4_clear_inode(): last iput(), when nobody link to this file.
 *	ext4_truncate(): when the block indirect map is about to change.
 *
 */
void ext4_discard_reservation(struct inode *inode)
{
	struct ext4_inode_info *ei = EXT4_I(inode);
	struct ext4_block_alloc_info *block_i = ei->i_block_alloc_info;
	struct ext4_reserve_window_node *rsv;
	spinlock_t *rsv_lock = &EXT4_SB(inode->i_sb)->s_rsv_window_lock;

	if (!block_i)
		return;

	rsv = &block_i->rsv_window_node;
	if (!rsv_is_empty(&rsv->rsv_window)) {
		spin_lock(rsv_lock);
		if (!rsv_is_empty(&rsv->rsv_window))
			rsv_window_remove(inode->i_sb, rsv);
		spin_unlock(rsv_lock);
	}
}

/**
 * ext4_free_blocks_sb() -- Free given blocks and update quota
 * @handle:			handle to this transaction
 * @sb:				super block
 * @block:			start physcial block to free
 * @count:			number of blocks to free
 * @pdquot_freed_blocks:	pointer to quota
 */
void ext4_free_blocks_sb(handle_t *handle, struct super_block *sb,
			 ext4_fsblk_t block, unsigned long count,
			 unsigned long *pdquot_freed_blocks)
{
	struct buffer_head *bitmap_bh = NULL;
	struct buffer_head *gd_bh;
	unsigned long block_group;
	ext4_grpblk_t bit;
	unsigned long i;
	unsigned long overflow;
	struct ext4_group_desc * desc;
	struct ext4_super_block * es;
	struct ext4_sb_info *sbi;
	int err = 0, ret;
	ext4_grpblk_t group_freed;

	*pdquot_freed_blocks = 0;
	sbi = EXT4_SB(sb);
	es = sbi->s_es;
	if (block < le32_to_cpu(es->s_first_data_block) ||
	    block + count < block ||
	    block + count > ext4_blocks_count(es)) {
		ext4_error (sb, "ext4_free_blocks",
			    "Freeing blocks not in datazone - "
			    "block = %llu, count = %lu", block, count);
		goto error_return;
	}

	ext4_debug ("freeing block(s) %llu-%llu\n", block, block + count - 1);

do_more:
	overflow = 0;
	ext4_get_group_no_and_offset(sb, block, &block_group, &bit);
	/*
	 * Check to see if we are freeing blocks across a group
	 * boundary.
	 */
	if (bit + count > EXT4_BLOCKS_PER_GROUP(sb)) {
		overflow = bit + count - EXT4_BLOCKS_PER_GROUP(sb);
		count -= overflow;
	}
	brelse(bitmap_bh);
	bitmap_bh = read_block_bitmap(sb, block_group);
	if (!bitmap_bh)
		goto error_return;
	desc = ext4_get_group_desc (sb, block_group, &gd_bh);
	if (!desc)
		goto error_return;

	if (in_range(ext4_block_bitmap(sb, desc), block, count) ||
	    in_range(ext4_inode_bitmap(sb, desc), block, count) ||
	    in_range(block, ext4_inode_table(sb, desc), sbi->s_itb_per_group) ||
	    in_range(block + count - 1, ext4_inode_table(sb, desc),
		     sbi->s_itb_per_group))
		ext4_error (sb, "ext4_free_blocks",
			    "Freeing blocks in system zones - "
			    "Block = %llu, count = %lu",
			    block, count);

	/*
	 * We are about to start releasing blocks in the bitmap,
	 * so we need undo access.
	 */
	/* @@@ check errors */
	BUFFER_TRACE(bitmap_bh, "getting undo access");
	err = ext4_journal_get_undo_access(handle, bitmap_bh);
	if (err)
		goto error_return;

	/*
	 * We are about to modify some metadata.  Call the journal APIs
	 * to unshare ->b_data if a currently-committing transaction is
	 * using it
	 */
	BUFFER_TRACE(gd_bh, "get_write_access");
	err = ext4_journal_get_write_access(handle, gd_bh);
	if (err)
		goto error_return;

	jbd_lock_bh_state(bitmap_bh);

	for (i = 0, group_freed = 0; i < count; i++) {
		/*
		 * An HJ special.  This is expensive...
		 */
#ifdef CONFIG_JBD2_DEBUG
		jbd_unlock_bh_state(bitmap_bh);
		{
			struct buffer_head *debug_bh;
			debug_bh = sb_find_get_block(sb, block + i);
			if (debug_bh) {
				BUFFER_TRACE(debug_bh, "Deleted!");
				if (!bh2jh(bitmap_bh)->b_committed_data)
					BUFFER_TRACE(debug_bh,
						"No commited data in bitmap");
				BUFFER_TRACE2(debug_bh, bitmap_bh, "bitmap");
				__brelse(debug_bh);
			}
		}
		jbd_lock_bh_state(bitmap_bh);
#endif
		if (need_resched()) {
			jbd_unlock_bh_state(bitmap_bh);
			cond_resched();
			jbd_lock_bh_state(bitmap_bh);
		}
		/* @@@ This prevents newly-allocated data from being
		 * freed and then reallocated within the same
		 * transaction.
		 *
		 * Ideally we would want to allow that to happen, but to
		 * do so requires making jbd2_journal_forget() capable of
		 * revoking the queued write of a data block, which
		 * implies blocking on the journal lock.  *forget()
		 * cannot block due to truncate races.
		 *
		 * Eventually we can fix this by making jbd2_journal_forget()
		 * return a status indicating whether or not it was able
		 * to revoke the buffer.  On successful revoke, it is
		 * safe not to set the allocation bit in the committed
		 * bitmap, because we know that there is no outstanding
		 * activity on the buffer any more and so it is safe to
		 * reallocate it.
		 */
		BUFFER_TRACE(bitmap_bh, "set in b_committed_data");
		J_ASSERT_BH(bitmap_bh,
				bh2jh(bitmap_bh)->b_committed_data != NULL);
		ext4_set_bit_atomic(sb_bgl_lock(sbi, block_group), bit + i,
				bh2jh(bitmap_bh)->b_committed_data);

		/*
		 * We clear the bit in the bitmap after setting the committed
		 * data bit, because this is the reverse order to that which
		 * the allocator uses.
		 */
		BUFFER_TRACE(bitmap_bh, "clear bit");
		if (!ext4_clear_bit_atomic(sb_bgl_lock(sbi, block_group),
						bit + i, bitmap_bh->b_data)) {
			jbd_unlock_bh_state(bitmap_bh);
			ext4_error(sb, __FUNCTION__,
				   "bit already cleared for block %llu",
				   (ext4_fsblk_t)(block + i));
			jbd_lock_bh_state(bitmap_bh);
			BUFFER_TRACE(bitmap_bh, "bit already cleared");
		} else {
			group_freed++;
		}
	}
	jbd_unlock_bh_state(bitmap_bh);

	spin_lock(sb_bgl_lock(sbi, block_group));
	desc->bg_free_blocks_count =
		cpu_to_le16(le16_to_cpu(desc->bg_free_blocks_count) +
			group_freed);
	desc->bg_checksum = ext4_group_desc_csum(sbi, block_group, desc);
	spin_unlock(sb_bgl_lock(sbi, block_group));
	percpu_counter_add(&sbi->s_freeblocks_counter, count);

	/* We dirtied the bitmap block */
	BUFFER_TRACE(bitmap_bh, "dirtied bitmap block");
	err = ext4_journal_dirty_metadata(handle, bitmap_bh);

	/* And the group descriptor block */
	BUFFER_TRACE(gd_bh, "dirtied group descriptor block");
	ret = ext4_journal_dirty_metadata(handle, gd_bh);
	if (!err) err = ret;
	*pdquot_freed_blocks += group_freed;

	if (overflow && !err) {
		block += count;
		count = overflow;
		goto do_more;
	}
	sb->s_dirt = 1;
error_return:
	brelse(bitmap_bh);
	ext4_std_error(sb, err);
	return;
}

/**
 * ext4_free_blocks() -- Free given blocks and update quota
 * @handle:		handle for this transaction
 * @inode:		inode
 * @block:		start physical block to free
 * @count:		number of blocks to count
 */
void ext4_free_blocks(handle_t *handle, struct inode *inode,
			ext4_fsblk_t block, unsigned long count)
{
	struct super_block * sb;
	unsigned long dquot_freed_blocks;

	sb = inode->i_sb;
	if (!sb) {
		printk ("ext4_free_blocks: nonexistent device");
		return;
	}
	ext4_free_blocks_sb(handle, sb, block, count, &dquot_freed_blocks);
	if (dquot_freed_blocks)
		DQUOT_FREE_BLOCK(inode, dquot_freed_blocks);
	return;
}

/**
 * ext4_test_allocatable()
 * @nr:			given allocation block group
 * @bh:			bufferhead contains the bitmap of the given block group
 *
 * For ext4 allocations, we must not reuse any blocks which are
 * allocated in the bitmap buffer's "last committed data" copy.  This
 * prevents deletes from freeing up the page for reuse until we have
 * committed the delete transaction.
 *
 * If we didn't do this, then deleting something and reallocating it as
 * data would allow the old block to be overwritten before the
 * transaction committed (because we force data to disk before commit).
 * This would lead to corruption if we crashed between overwriting the
 * data and committing the delete.
 *
 * @@@ We may want to make this allocation behaviour conditional on
 * data-writes at some point, and disable it for metadata allocations or
 * sync-data inodes.
 */
static int ext4_test_allocatable(ext4_grpblk_t nr, struct buffer_head *bh)
{
	int ret;
	struct journal_head *jh = bh2jh(bh);

	if (ext4_test_bit(nr, bh->b_data))
		return 0;

	jbd_lock_bh_state(bh);
	if (!jh->b_committed_data)
		ret = 1;
	else
		ret = !ext4_test_bit(nr, jh->b_committed_data);
	jbd_unlock_bh_state(bh);
	return ret;
}

/**
 * bitmap_search_next_usable_block()
 * @start:		the starting block (group relative) of the search
 * @bh:			bufferhead contains the block group bitmap
 * @maxblocks:		the ending block (group relative) of the reservation
 *
 * The bitmap search --- search forward alternately through the actual
 * bitmap on disk and the last-committed copy in journal, until we find a
 * bit free in both bitmaps.
 */
static ext4_grpblk_t
bitmap_search_next_usable_block(ext4_grpblk_t start, struct buffer_head *bh,
					ext4_grpblk_t maxblocks)
{
	ext4_grpblk_t next;
	struct journal_head *jh = bh2jh(bh);

	while (start < maxblocks) {
		next = ext4_find_next_zero_bit(bh->b_data, maxblocks, start);
		if (next >= maxblocks)
			return -1;
		if (ext4_test_allocatable(next, bh))
			return next;
		jbd_lock_bh_state(bh);
		if (jh->b_committed_data)
			start = ext4_find_next_zero_bit(jh->b_committed_data,
							maxblocks, next);
		jbd_unlock_bh_state(bh);
	}
	return -1;
}

/**
 * find_next_usable_block()
 * @start:		the starting block (group relative) to find next
 *			allocatable block in bitmap.
 * @bh:			bufferhead contains the block group bitmap
 * @maxblocks:		the ending block (group relative) for the search
 *
 * Find an allocatable block in a bitmap.  We honor both the bitmap and
 * its last-committed copy (if that exists), and perform the "most
 * appropriate allocation" algorithm of looking for a free block near
 * the initial goal; then for a free byte somewhere in the bitmap; then
 * for any free bit in the bitmap.
 */
static ext4_grpblk_t
find_next_usable_block(ext4_grpblk_t start, struct buffer_head *bh,
			ext4_grpblk_t maxblocks)
{
	ext4_grpblk_t here, next;
	char *p, *r;

	if (start > 0) {
		/*
		 * The goal was occupied; search forward for a free
		 * block within the next XX blocks.
		 *
		 * end_goal is more or less random, but it has to be
		 * less than EXT4_BLOCKS_PER_GROUP. Aligning up to the
		 * next 64-bit boundary is simple..
		 */
		ext4_grpblk_t end_goal = (start + 63) & ~63;
		if (end_goal > maxblocks)
			end_goal = maxblocks;
		here = ext4_find_next_zero_bit(bh->b_data, end_goal, start);
		if (here < end_goal && ext4_test_allocatable(here, bh))
			return here;
		ext4_debug("Bit not found near goal\n");
	}

	here = start;
	if (here < 0)
		here = 0;

	p = ((char *)bh->b_data) + (here >> 3);
	r = memscan(p, 0, ((maxblocks + 7) >> 3) - (here >> 3));
	next = (r - ((char *)bh->b_data)) << 3;

	if (next < maxblocks && next >= start && ext4_test_allocatable(next, bh))
		return next;

	/*
	 * The bitmap search --- search forward alternately through the actual
	 * bitmap and the last-committed copy until we find a bit free in
	 * both
	 */
	here = bitmap_search_next_usable_block(here, bh, maxblocks);
	return here;
}

/**
 * claim_block()
 * @block:		the free block (group relative) to allocate
 * @bh:			the bufferhead containts the block group bitmap
 *
 * We think we can allocate this block in this bitmap.  Try to set the bit.
 * If that succeeds then check that nobody has allocated and then freed the
 * block since we saw that is was not marked in b_committed_data.  If it _was_
 * allocated and freed then clear the bit in the bitmap again and return
 * zero (failure).
 */
static inline int
claim_block(spinlock_t *lock, ext4_grpblk_t block, struct buffer_head *bh)
{
	struct journal_head *jh = bh2jh(bh);
	int ret;

	if (ext4_set_bit_atomic(lock, block, bh->b_data))
		return 0;
	jbd_lock_bh_state(bh);
	if (jh->b_committed_data && ext4_test_bit(block,jh->b_committed_data)) {
		ext4_clear_bit_atomic(lock, block, bh->b_data);
		ret = 0;
	} else {
		ret = 1;
	}
	jbd_unlock_bh_state(bh);
	return ret;
}

/**
 * ext4_try_to_allocate()
 * @sb:			superblock
 * @handle:		handle to this transaction
 * @group:		given allocation block group
 * @bitmap_bh:		bufferhead holds the block bitmap
 * @grp_goal:		given target block within the group
 * @count:		target number of blocks to allocate
 * @my_rsv:		reservation window
 *
 * Attempt to allocate blocks within a give range. Set the range of allocation
 * first, then find the first free bit(s) from the bitmap (within the range),
 * and at last, allocate the blocks by claiming the found free bit as allocated.
 *
 * To set the range of this allocation:
 *	if there is a reservation window, only try to allocate block(s) from the
 *	file's own reservation window;
 *	Otherwise, the allocation range starts from the give goal block, ends at
 *	the block group's last block.
 *
 * If we failed to allocate the desired block then we may end up crossing to a
 * new bitmap.  In that case we must release write access to the old one via
 * ext4_journal_release_buffer(), else we'll run out of credits.
 */
static ext4_grpblk_t
ext4_try_to_allocate(struct super_block *sb, handle_t *handle, int group,
			struct buffer_head *bitmap_bh, ext4_grpblk_t grp_goal,
			unsigned long *count, struct ext4_reserve_window *my_rsv)
{
	ext4_fsblk_t group_first_block;
	ext4_grpblk_t start, end;
	unsigned long num = 0;

	/* we do allocation within the reservation window if we have a window */
	if (my_rsv) {
		group_first_block = ext4_group_first_block_no(sb, group);
		if (my_rsv->_rsv_start >= group_first_block)
			start = my_rsv->_rsv_start - group_first_block;
		else
			/* reservation window cross group boundary */
			start = 0;
		end = my_rsv->_rsv_end - group_first_block + 1;
		if (end > EXT4_BLOCKS_PER_GROUP(sb))
			/* reservation window crosses group boundary */
			end = EXT4_BLOCKS_PER_GROUP(sb);
		if ((start <= grp_goal) && (grp_goal < end))
			start = grp_goal;
		else
			grp_goal = -1;
	} else {
		if (grp_goal > 0)
			start = grp_goal;
		else
			start = 0;
		end = EXT4_BLOCKS_PER_GROUP(sb);
	}

	BUG_ON(start > EXT4_BLOCKS_PER_GROUP(sb));

repeat:
	if (grp_goal < 0 || !ext4_test_allocatable(grp_goal, bitmap_bh)) {
		grp_goal = find_next_usable_block(start, bitmap_bh, end);
		if (grp_goal < 0)
			goto fail_access;
		if (!my_rsv) {
			int i;

			for (i = 0; i < 7 && grp_goal > start &&
					ext4_test_allocatable(grp_goal - 1,
								bitmap_bh);
					i++, grp_goal--)