xfs_da_btree.c

来自「Linux Kernel 2.6.9 for OMAP1710」· C语言 代码 · 共 2,203 行 · 第 1/5 页

	 * For version 1 directories, adjust the file size if it changed.
	 */
	if (w == XFS_DATA_FORK && XFS_DIR_IS_V1(mp)) {
		ASSERT(mapi == 1);
		if ((error = xfs_bmap_last_offset(tp, dp, &bno, w)))
			return error;
		size = XFS_FSB_TO_B(mp, bno);
		if (size != dp->i_d.di_size) {
			dp->i_d.di_size = size;
			xfs_trans_log_inode(tp, dp, XFS_ILOG_CORE);
		}
	}
	return 0;
}

/*
 * Ick.  We need to always be able to remove a btree block, even
 * if there's no space reservation because the filesystem is full.
 * This is called if xfs_bunmapi on a btree block fails due to ENOSPC.
 * It swaps the target block with the last block in the file.  The
 * last block in the file can always be removed since it can't cause
 * a bmap btree split to do that.
 */
STATIC int
xfs_da_swap_lastblock(xfs_da_args_t *args, xfs_dablk_t *dead_blknop,
		      xfs_dabuf_t **dead_bufp)
{
	xfs_dablk_t dead_blkno, last_blkno, sib_blkno, par_blkno;
	xfs_dabuf_t *dead_buf, *last_buf, *sib_buf, *par_buf;
	xfs_fileoff_t lastoff;
	xfs_inode_t *ip;
	xfs_trans_t *tp;
	xfs_mount_t *mp;
	int error, w, entno, level, dead_level;
	xfs_da_blkinfo_t *dead_info, *sib_info;
	xfs_da_intnode_t *par_node, *dead_node;
	xfs_dir_leafblock_t *dead_leaf;
	xfs_dir2_leaf_t *dead_leaf2;
	xfs_dahash_t dead_hash;

	dead_buf = *dead_bufp;
	dead_blkno = *dead_blknop;
	tp = args->trans;
	ip = args->dp;
	w = args->whichfork;
	ASSERT(w == XFS_DATA_FORK);
	mp = ip->i_mount;
	if (XFS_DIR_IS_V2(mp)) {
		lastoff = mp->m_dirfreeblk;
		error = xfs_bmap_last_before(tp, ip, &lastoff, w);
	} else
		error = xfs_bmap_last_offset(tp, ip, &lastoff, w);
	if (error)
		return error;
	if (unlikely(lastoff == 0)) {
		XFS_ERROR_REPORT("xfs_da_swap_lastblock(1)", XFS_ERRLEVEL_LOW,
				 mp);
		return XFS_ERROR(EFSCORRUPTED);
	}
	/*
	 * Read the last block in the btree space.
	 */
	last_blkno = (xfs_dablk_t)lastoff - mp->m_dirblkfsbs;
	if ((error = xfs_da_read_buf(tp, ip, last_blkno, -1, &last_buf, w)))
		return error;
	/*
	 * Copy the last block into the dead buffer and log it.
	 */
	memcpy(dead_buf->data, last_buf->data, mp->m_dirblksize);
	xfs_da_log_buf(tp, dead_buf, 0, mp->m_dirblksize - 1);
	dead_info = dead_buf->data;
	/*
	 * Get values from the moved block.
	 */
	if (INT_GET(dead_info->magic, ARCH_CONVERT) == XFS_DIR_LEAF_MAGIC) {
		ASSERT(XFS_DIR_IS_V1(mp));
		dead_leaf = (xfs_dir_leafblock_t *)dead_info;
		dead_level = 0;
		dead_hash =
			INT_GET(dead_leaf->entries[INT_GET(dead_leaf->hdr.count, ARCH_CONVERT) - 1].hashval, ARCH_CONVERT);
	} else if (INT_GET(dead_info->magic, ARCH_CONVERT) == XFS_DIR2_LEAFN_MAGIC) {
		ASSERT(XFS_DIR_IS_V2(mp));
		dead_leaf2 = (xfs_dir2_leaf_t *)dead_info;
		dead_level = 0;
		dead_hash = INT_GET(dead_leaf2->ents[INT_GET(dead_leaf2->hdr.count, ARCH_CONVERT) - 1].hashval, ARCH_CONVERT);
	} else {
		ASSERT(INT_GET(dead_info->magic, ARCH_CONVERT) == XFS_DA_NODE_MAGIC);
		dead_node = (xfs_da_intnode_t *)dead_info;
		dead_level = INT_GET(dead_node->hdr.level, ARCH_CONVERT);
		dead_hash = INT_GET(dead_node->btree[INT_GET(dead_node->hdr.count, ARCH_CONVERT) - 1].hashval, ARCH_CONVERT);
	}
	sib_buf = par_buf = NULL;
	/*
	 * If the moved block has a left sibling, fix up the pointers.
	 */
	if ((sib_blkno = INT_GET(dead_info->back, ARCH_CONVERT))) {
		if ((error = xfs_da_read_buf(tp, ip, sib_blkno, -1, &sib_buf, w)))
			goto done;
		sib_info = sib_buf->data;
		if (unlikely(
		    INT_GET(sib_info->forw, ARCH_CONVERT) != last_blkno ||
		    INT_GET(sib_info->magic, ARCH_CONVERT) != INT_GET(dead_info->magic, ARCH_CONVERT))) {
			XFS_ERROR_REPORT("xfs_da_swap_lastblock(2)",
					 XFS_ERRLEVEL_LOW, mp);
			error = XFS_ERROR(EFSCORRUPTED);
			goto done;
		}
		INT_SET(sib_info->forw, ARCH_CONVERT, dead_blkno);
		xfs_da_log_buf(tp, sib_buf,
			XFS_DA_LOGRANGE(sib_info, &sib_info->forw,
					sizeof(sib_info->forw)));
		xfs_da_buf_done(sib_buf);
		sib_buf = NULL;
	}
	/*
	 * If the moved block has a right sibling, fix up the pointers.
	 */
	if ((sib_blkno = INT_GET(dead_info->forw, ARCH_CONVERT))) {
		if ((error = xfs_da_read_buf(tp, ip, sib_blkno, -1, &sib_buf, w)))
			goto done;
		sib_info = sib_buf->data;
		if (unlikely(
		       INT_GET(sib_info->back, ARCH_CONVERT) != last_blkno
		    || INT_GET(sib_info->magic, ARCH_CONVERT)
				!= INT_GET(dead_info->magic, ARCH_CONVERT))) {
			XFS_ERROR_REPORT("xfs_da_swap_lastblock(3)",
					 XFS_ERRLEVEL_LOW, mp);
			error = XFS_ERROR(EFSCORRUPTED);
			goto done;
		}
		INT_SET(sib_info->back, ARCH_CONVERT, dead_blkno);
		xfs_da_log_buf(tp, sib_buf,
			XFS_DA_LOGRANGE(sib_info, &sib_info->back,
					sizeof(sib_info->back)));
		xfs_da_buf_done(sib_buf);
		sib_buf = NULL;
	}
	par_blkno = XFS_DIR_IS_V1(mp) ? 0 : mp->m_dirleafblk;
	level = -1;
	/*
	 * Walk down the tree looking for the parent of the moved block.
	 */
	for (;;) {
		if ((error = xfs_da_read_buf(tp, ip, par_blkno, -1, &par_buf, w)))
			goto done;
		par_node = par_buf->data;
		if (unlikely(
		    INT_GET(par_node->hdr.info.magic, ARCH_CONVERT) != XFS_DA_NODE_MAGIC ||
		    (level >= 0 && level != INT_GET(par_node->hdr.level, ARCH_CONVERT) + 1))) {
			XFS_ERROR_REPORT("xfs_da_swap_lastblock(4)",
					 XFS_ERRLEVEL_LOW, mp);
			error = XFS_ERROR(EFSCORRUPTED);
			goto done;
		}
		level = INT_GET(par_node->hdr.level, ARCH_CONVERT);
		for (entno = 0;
		     entno < INT_GET(par_node->hdr.count, ARCH_CONVERT) &&
		     INT_GET(par_node->btree[entno].hashval, ARCH_CONVERT) < dead_hash;
		     entno++)
			continue;
		if (unlikely(entno == INT_GET(par_node->hdr.count, ARCH_CONVERT))) {
			XFS_ERROR_REPORT("xfs_da_swap_lastblock(5)",
					 XFS_ERRLEVEL_LOW, mp);
			error = XFS_ERROR(EFSCORRUPTED);
			goto done;
		}
		par_blkno = INT_GET(par_node->btree[entno].before, ARCH_CONVERT);
		if (level == dead_level + 1)
			break;
		xfs_da_brelse(tp, par_buf);
		par_buf = NULL;
	}
	/*
	 * We're in the right parent block.
	 * Look for the right entry.
	 */
	for (;;) {
		for (;
		     entno < INT_GET(par_node->hdr.count, ARCH_CONVERT) &&
		     INT_GET(par_node->btree[entno].before, ARCH_CONVERT) != last_blkno;
		     entno++)
			continue;
		if (entno < INT_GET(par_node->hdr.count, ARCH_CONVERT))
			break;
		par_blkno = INT_GET(par_node->hdr.info.forw, ARCH_CONVERT);
		xfs_da_brelse(tp, par_buf);
		par_buf = NULL;
		if (unlikely(par_blkno == 0)) {
			XFS_ERROR_REPORT("xfs_da_swap_lastblock(6)",
					 XFS_ERRLEVEL_LOW, mp);
			error = XFS_ERROR(EFSCORRUPTED);
			goto done;
		}
		if ((error = xfs_da_read_buf(tp, ip, par_blkno, -1, &par_buf, w)))
			goto done;
		par_node = par_buf->data;
		if (unlikely(
		    INT_GET(par_node->hdr.level, ARCH_CONVERT) != level ||
		    INT_GET(par_node->hdr.info.magic, ARCH_CONVERT) != XFS_DA_NODE_MAGIC)) {
			XFS_ERROR_REPORT("xfs_da_swap_lastblock(7)",
					 XFS_ERRLEVEL_LOW, mp);
			error = XFS_ERROR(EFSCORRUPTED);
			goto done;
		}
		entno = 0;
	}
	/*
	 * Update the parent entry pointing to the moved block.
	 */
	INT_SET(par_node->btree[entno].before, ARCH_CONVERT, dead_blkno);
	xfs_da_log_buf(tp, par_buf,
		XFS_DA_LOGRANGE(par_node, &par_node->btree[entno].before,
				sizeof(par_node->btree[entno].before)));
	xfs_da_buf_done(par_buf);
	xfs_da_buf_done(dead_buf);
	*dead_blknop = last_blkno;
	*dead_bufp = last_buf;
	return 0;
done:
	if (par_buf)
		xfs_da_brelse(tp, par_buf);
	if (sib_buf)
		xfs_da_brelse(tp, sib_buf);
	xfs_da_brelse(tp, last_buf);
	return error;
}

/*
 * Remove a btree block from a directory or attribute.
 */
int
xfs_da_shrink_inode(xfs_da_args_t *args, xfs_dablk_t dead_blkno,
		    xfs_dabuf_t *dead_buf)
{
	xfs_inode_t *dp;
	int done, error, w, count;
	xfs_fileoff_t bno;
	xfs_fsize_t size;
	xfs_trans_t *tp;
	xfs_mount_t *mp;

	dp = args->dp;
	w = args->whichfork;
	tp = args->trans;
	mp = dp->i_mount;
	if (w == XFS_DATA_FORK && XFS_DIR_IS_V2(mp))
		count = mp->m_dirblkfsbs;
	else
		count = 1;
	for (;;) {
		/*
		 * Remove extents.  If we get ENOSPC for a dir we have to move
		 * the last block to the place we want to kill.
		 */
		if ((error = xfs_bunmapi(tp, dp, dead_blkno, count,
				XFS_BMAPI_AFLAG(w)|XFS_BMAPI_METADATA,
				0, args->firstblock, args->flist,
				&done)) == ENOSPC) {
			if (w != XFS_DATA_FORK)
				goto done;
			if ((error = xfs_da_swap_lastblock(args, &dead_blkno,
					&dead_buf)))
				goto done;
		} else if (error)
			goto done;
		else
			break;
	}
	ASSERT(done);
	xfs_da_binval(tp, dead_buf);
	/*
	 * Adjust the directory size for version 1.
	 */
	if (w == XFS_DATA_FORK && XFS_DIR_IS_V1(mp)) {
		if ((error = xfs_bmap_last_offset(tp, dp, &bno, w)))
			return error;
		size = XFS_FSB_TO_B(dp->i_mount, bno);
		if (size != dp->i_d.di_size) {
			dp->i_d.di_size = size;
			xfs_trans_log_inode(tp, dp, XFS_ILOG_CORE);
		}
	}
	return 0;
done:
	xfs_da_binval(tp, dead_buf);
	return error;
}

/*
 * See if the mapping(s) for this btree block are valid, i.e.
 * don't contain holes, are logically contiguous, and cover the whole range.
 */
STATIC int
xfs_da_map_covers_blocks(
	int		nmap,
	xfs_bmbt_irec_t	*mapp,
	xfs_dablk_t	bno,
	int		count)
{
	int		i;
	xfs_fileoff_t	off;

	for (i = 0, off = bno; i < nmap; i++) {
		if (mapp[i].br_startblock == HOLESTARTBLOCK ||
		    mapp[i].br_startblock == DELAYSTARTBLOCK) {
			return 0;
		}
		if (off != mapp[i].br_startoff) {
			return 0;
		}
		off += mapp[i].br_blockcount;
	}
	return off == bno + count;
}

/*
 * Make a dabuf.
 * Used for get_buf, read_buf, read_bufr, and reada_buf.
 */
STATIC int
xfs_da_do_buf(
	xfs_trans_t	*trans,
	xfs_inode_t	*dp,
	xfs_dablk_t	bno,
	xfs_daddr_t	*mappedbnop,
	xfs_dabuf_t	**bpp,
	int		whichfork,
	int		caller,
	inst_t		*ra)
{
	xfs_buf_t	*bp = NULL;
	xfs_buf_t	**bplist;
	int		error=0;
	int		i;
	xfs_bmbt_irec_t	map;
	xfs_bmbt_irec_t	*mapp;
	xfs_daddr_t	mappedbno;
	xfs_mount_t	*mp;
	int		nbplist=0;
	int		nfsb;
	int		nmap;
	xfs_dabuf_t	*rbp;

	mp = dp->i_mount;
	if (whichfork == XFS_DATA_FORK && XFS_DIR_IS_V2(mp))
		nfsb = mp->m_dirblkfsbs;
	else
		nfsb = 1;
	mappedbno = *mappedbnop;
	/*
	 * Caller doesn't have a mapping.  -2 means don't complain
	 * if we land in a hole.
	 */
	if (mappedbno == -1 || mappedbno == -2) {
		/*
		 * Optimize the one-block case.
		 */
		if (nfsb == 1) {
			xfs_fsblock_t	fsb;

			if ((error =
			    xfs_bmapi_single(trans, dp, whichfork, &fsb,
				    (xfs_fileoff_t)bno))) {
				return error;
			}
			mapp = &map;
			if (fsb == NULLFSBLOCK) {
				nmap = 0;
			} else {
				map.br_startblock = fsb;
				map.br_startoff = (xfs_fileoff_t)bno;
				map.br_blockcount = 1;
				nmap = 1;
			}
		} else {
			mapp = kmem_alloc(sizeof(*mapp) * nfsb, KM_SLEEP);
			nmap = nfsb;
			if ((error = xfs_bmapi(trans, dp, (xfs_fileoff_t)bno,
					nfsb,
					XFS_BMAPI_METADATA |
						XFS_BMAPI_AFLAG(whichfork),
					NULL, 0, mapp, &nmap, NULL)))
				goto exit0;
		}
	} else {
		map.br_startblock = XFS_DADDR_TO_FSB(mp, mappedbno);
		map.br_startoff = (xfs_fileoff_t)bno;
		map.br_blockcount = nfsb;
		mapp = &map;
		nmap = 1;
	}
	if (!xfs_da_map_covers_blocks(nmap, mapp, bno, nfsb)) {
		error = mappedbno == -2 ? 0 : XFS_ERROR(EFSCORRUPTED);
		if (unlikely(error == EFSCORRUPTED)) {
			if (xfs_error_level >= XFS_ERRLEVEL_LOW) {
				int	i;
				cmn_err(CE_ALERT, "xfs_da_do_buf: bno %lld\n",
					(long long)bno);
				cmn_err(CE_ALERT, "dir: inode %lld\n",
					(long long)dp->i_ino);
				for (i = 0; i < nmap; i++) {
					cmn_err(CE_ALERT,
						"[%02d] br_startoff %lld br_startblock %lld br_blockcount %lld br_state %d\n",
						i,
						(long long)mapp[i].br_startoff,
						(long long)mapp[i].br_startblock,
						(long long)mapp[i].br_blockcount,
						mapp[i].br_state);
				}
			}
			XFS_ERROR_REPORT("xfs_da_do_buf(1)",
					 XFS_ERRLEVEL_LOW, mp);
		}
		goto exit0;
	}
	if (caller != 3 && nmap > 1) {
		bplist = kmem_alloc(sizeof(*bplist) * nmap, KM_SLEEP);
		nbplist = 0;
	} else
		bplist = NULL;
	/*
	 * Turn the mapping(s) into buffer(s).
	 */
	for (i = 0; i < nmap; i++) {
		int	nmapped;

		mappedbno = XFS_FSB_TO_DADDR(mp, mapp[i].br_startblock);
		if (i == 0)
			*mappedbnop = mappedbno;
		nmapped = (int)XFS_FSB_TO_BB(mp, mapp[i].br_blockcount);
		switch (caller) {
		case 0:
			bp = xfs_trans_get_buf(trans, mp->m_ddev_targp,
				mappedbno, nmapped, 0);
			error = bp ? XFS_BUF_GETERROR(bp) : XFS_ERROR(EIO);
			break;
		case 1:
#ifndef __KERNEL__
		cas