Re: [PATCH 2/2] xfs_metadump: Zap more stale data

["Darrick J. Wong" <darrick.wong@oracle.com>]

On Thu, Oct 11, 2018 at 09:44:24PM +0200, Stefan Ring wrote:
> I have empirically found and tried to fix some places where stale data was not
> properly zeroed out.
> 
> In the order of the code changes:
> 
> - The "freeindex" blocks in inode directories, from last entry to end of block.
> - XFS_DIR{2,3}_LEAFN_MAGIC, from last entry to end of block.
> - In btree format inodes before as well as after the btree pointers.
> - In dev inodes, everything after the header.
> ---
>  db/metadump.c | 106 +++++++++++++++++++++++++++++++++++++++++++++++++++++-----
>  1 file changed, 98 insertions(+), 8 deletions(-)
> 
> diff --git a/db/metadump.c b/db/metadump.c
> index ff96860d..c8213d41 100644
> --- a/db/metadump.c
> +++ b/db/metadump.c
> @@ -1421,12 +1421,49 @@ process_sf_attr(
>  		memset(asfep, 0, XFS_DFORK_ASIZE(dip, mp) - ino_attr_size);
>  }
>  
> +static void
> +process_dir_free_block(

I think patch 1 references this function in patch 2?

> +	char				*block)
> +{
> +	struct xfs_dir2_free		*free;
> +	struct xfs_dir3_icfree_hdr	freehdr;
> +
> +	if (!zero_stale_data)
> +		return;
> +
> +	free = (struct xfs_dir2_free *)block;
> +	M_DIROPS(mp)->free_hdr_from_disk(&freehdr, free);
> +
> +	switch (freehdr.magic) {
> +	case XFS_DIR2_FREE_MAGIC:
> +	case XFS_DIR3_FREE_MAGIC: {
> +		__be16			*bests;
> +		char			*high;
> +		int			used;
> +
> +		/* Zero out space from end of bests[] to end of block */
> +		bests = M_DIROPS(mp)->free_bests_p(free);
> +		high = (char *)&bests[freehdr.nvalid];
> +		used = high - (char*)free;
> +		memset(high, 0, mp->m_dir_geo->blksize - used);
> +		iocur_top->need_crc = 1;
> +		break;
> +	}
> +	default:
> +		if (show_warnings)
> +			print_warning("invalid magic in dir inode %llu "
> +				      "free block",
> +				      (unsigned long long)cur_ino);
> +		break;
> +	}
> +}
> +
>  static void
>  process_dir_leaf_block(
>  	char				*block)
>  {
>  	struct xfs_dir2_leaf		*leaf;
> -	struct xfs_dir3_icleaf_hdr 	leafhdr;
> +	struct xfs_dir3_icleaf_hdr	leafhdr;
>  
>  	if (!zero_stale_data)
>  		return;
> @@ -1435,20 +1472,39 @@ process_dir_leaf_block(
>  	leaf = (struct xfs_dir2_leaf *)block;
>  	M_DIROPS(mp)->leaf_hdr_from_disk(&leafhdr, leaf);
>  
> -	/* Zero out space from end of ents[] to bests */
> -	if (leafhdr.magic == XFS_DIR2_LEAF1_MAGIC ||
> -	    leafhdr.magic == XFS_DIR3_LEAF1_MAGIC) {
> +	switch (leafhdr.magic) {
> +	case XFS_DIR2_LEAF1_MAGIC:
> +	case XFS_DIR3_LEAF1_MAGIC: {
>  		struct xfs_dir2_leaf_tail	*ltp;
>  		__be16				*lbp;
>  		struct xfs_dir2_leaf_entry	*ents;
>  		char				*free; /* end of ents */
>  
> +		/* Zero out space from end of ents[] to bests */
>  		ents = M_DIROPS(mp)->leaf_ents_p(leaf);
>  		free = (char *)&ents[leafhdr.count];
>  		ltp = xfs_dir2_leaf_tail_p(mp->m_dir_geo, leaf);
>  		lbp = xfs_dir2_leaf_bests_p(ltp);
>  		memset(free, 0, (char *)lbp - free);
>  		iocur_top->need_crc = 1;
> +		break;
> +	}
> +	case XFS_DIR2_LEAFN_MAGIC:
> +	case XFS_DIR3_LEAFN_MAGIC: {
> +		struct xfs_dir2_leaf_entry	*ents;
> +		char				*free;
> +		int				used;
> +
> +		/* Zero out space from end of ents[] to end of block */
> +		ents = M_DIROPS(mp)->leaf_ents_p(leaf);
> +		free = (char *)&ents[leafhdr.count];
> +		used = free - (char*)leaf;
> +		memset(free, 0, mp->m_dir_geo->blksize - used);
> +		iocur_top->need_crc = 1;
> +		break;
> +	}
> +	default:
> +		break;
>  	}
>  }
>  
> @@ -1499,7 +1555,7 @@ process_dir_data_block(
>  		if (show_warnings)
>  			print_warning(
>  		"invalid magic in dir inode %llu block %ld",
> -					(long long)cur_ino, (long)offset);
> +					(unsigned long long)cur_ino, (long)offset);
>  		return;
>  	}
>  
> @@ -1813,8 +1869,7 @@ process_single_fsb_objects(
>  		switch (btype) {
>  		case TYP_DIR2:
>  			if (o >= mp->m_dir_geo->freeblk) {
> -				/* TODO, zap any stale data */
> -				break;
> +				process_dir_free_block(dp);
>  			} else if (o >= mp->m_dir_geo->leafblk) {
>  				process_dir_leaf_block(dp);
>  			} else {
> @@ -2118,6 +2173,21 @@ process_btinode(
>  	}
>  
>  	pp = XFS_BMDR_PTR_ADDR(dib, 1, maxrecs);
> +
> +	if (zero_stale_data) {
> +		char	*top;
> +		int	used;
> +
> +		/* Space before btree pointers */
> +		top = (char*)XFS_BMDR_PTR_ADDR(dib, 1, nrecs);
> +		memset(top, 0, (char*)pp - top);
> +
> +		/* Space after btree pointers */
> +		top = (char*)&pp[nrecs];
> +		used = top - (char*)dip;
> +		memset(top, 0, mp->m_sb.sb_inodesize - used);

Er... this zaps everything past the end of the bmbt pointers, which
means that this can zero the attr fork root.

> +	}
> +
>  	for (i = 0; i < nrecs; i++) {
>  		xfs_agnumber_t	ag;
>  		xfs_agblock_t	bno;
> @@ -2201,6 +2271,24 @@ process_inode_data(
>  	return 1;
>  }
>  
> +static int
> +process_dev_inode(
> +	xfs_dinode_t		*dip)
> +{
> +	if (XFS_DFORK_NEXTENTS(dip, XFS_ATTR_FORK) ||
> +	    XFS_DFORK_NEXTENTS(dip, XFS_DATA_FORK)) {
> +		if (show_warnings)
> +			print_warning("inode %llu has unexpected extents",
> +				      (unsigned long long)cur_ino);
> +		return 0;
> +	} else {

No need for else here.

> +		int used = XFS_DFORK_DPTR(dip) - (char*)dip;
> +
> +		memset(XFS_DFORK_DPTR(dip), 0, mp->m_sb.sb_inodesize - used);

This zeroes both data fork and attr fork roots for fifo and device
inodes.  Wouldn't we want to preserve the device number information?
And whatever attrs are attached to the inode, if any?

--D

> +		return 1;
> +	}
> +}
> +
>  /*
>   * when we process the inode, we may change the data in the data and/or
>   * attribute fork if they are in short form and we are obfuscating names.
> @@ -2253,7 +2341,9 @@ process_inode(
>  		case S_IFREG:
>  			success = process_inode_data(dip, TYP_DATA);
>  			break;
> -		default: ;
> +		default:
> +			success = process_dev_inode(dip);
> +			break;
>  	}
>  	nametable_clear();
>  
> -- 
> 2.14.4
>