[PATCH 9/9] repair: detect and handle attribute tree CRC errors

[Brian Foster]

On Tue, Apr 29, 2014 at 07:04:59AM +1000, Dave Chinner wrote:
> From: Dave Chinner <dchinner at redhat.com>
> 
> Currently the attribute code will not detect and correct errors in
> the attribute tree. It also fails to validate the CRCs and headers
> on remote attribute blocks. Ensure that all the attribute blocks are
> CRC checked and that the processing functions understand the correct
> block formats for decoding.
> 
> Signed-off-by: Dave Chinner <dchinner at redhat.com>
> ---
>  repair/attr_repair.c | 45 +++++++++++++++++++++++++++++++++++++--------
>  1 file changed, 37 insertions(+), 8 deletions(-)
> 
> diff --git a/repair/attr_repair.c b/repair/attr_repair.c
> index ba85ac2..9b57960 100644
> --- a/repair/attr_repair.c
> +++ b/repair/attr_repair.c
> @@ -604,6 +604,7 @@ verify_da_path(xfs_mount_t	*mp,
>  			libxfs_putbuf(bp);
>  			return(1);
>  		}
> +
>  		/*
>  		 * update cursor, write out the *current* level if
>  		 * required.  don't write out the descendant level
> @@ -615,6 +616,8 @@ verify_da_path(xfs_mount_t	*mp,
>  			libxfs_writebuf(cursor->level[this_level].bp, 0);
>  		else
>  			libxfs_putbuf(cursor->level[this_level].bp);
> +
> +		/* switch cursor to point at the new buffer we just read */
>  		cursor->level[this_level].bp = bp;
>  		cursor->level[this_level].dirty = 0;
>  		cursor->level[this_level].bno = dabno;
> @@ -624,6 +627,14 @@ verify_da_path(xfs_mount_t	*mp,
>  		cursor->level[this_level].n = newnode;
>  #endif
>  		entry = cursor->level[this_level].index = 0;
> +
> +		/*
> +		 * We want to rewrite the buffer a CRC error seeing as it
Nit:						^ "on a CRC error ..." ?

> +		 * contains what appears to be a valid node block, but only if
> +		 * we are fixing errors.
> +		 */
> +		if (bp->b_error == EFSBADCRC && !no_modify)
> +			cursor->level[this_level].dirty++;
>  	}
>  	/*
>  	 * ditto for block numbers
> @@ -974,6 +985,10 @@ rmtval_get(xfs_mount_t *mp, xfs_ino_t ino, blkmap_t *blkmap,
>  	xfs_dfsbno_t	bno;
>  	xfs_buf_t	*bp;
>  	int		clearit = 0, i = 0, length = 0, amountdone = 0;
> +	int		hdrsize = 0;
> +
> +	if (xfs_sb_version_hascrc(&mp->m_sb))
> +		hdrsize = sizeof(struct xfs_attr3_rmt_hdr);
>  
>  	/* ASSUMPTION: valuelen is a valid number, so use it for looping */
>  	/* Note that valuelen is not a multiple of blocksize */
> @@ -986,16 +1001,26 @@ rmtval_get(xfs_mount_t *mp, xfs_ino_t ino, blkmap_t *blkmap,
>  			break;
>  		}
>  		bp = libxfs_readbuf(mp->m_dev, XFS_FSB_TO_DADDR(mp, bno),
> -				XFS_FSB_TO_BB(mp, 1), 0, NULL);
> +				    XFS_FSB_TO_BB(mp, 1), 0,
> +				    &xfs_attr3_rmt_buf_ops);
>  		if (!bp) {
>  			do_warn(
>  	_("can't read remote block for attributes of inode %" PRIu64 "\n"), ino);
>  			clearit = 1;
>  			break;
>  		}
> +
> +		if (bp->b_error == EFSBADCRC || bp->b_error == EFSCORRUPTED) {
> +			do_warn(
> +	_("Corrupt remote block for attributes of inode %" PRIu64 "\n"), ino);
> +			clearit = 1;
> +			break;
> +		}
> +
>  		ASSERT(mp->m_sb.sb_blocksize == XFS_BUF_COUNT(bp));
> -		length = MIN(XFS_BUF_COUNT(bp), valuelen - amountdone);
> -		memmove(value, XFS_BUF_PTR(bp), length);
> +
> +		length = MIN(XFS_BUF_COUNT(bp) - hdrsize, valuelen - amountdone);
> +		memmove(value, XFS_BUF_PTR(bp) + hdrsize, length);
>  		amountdone += length;
>  		value += length;
>  		i++;
> @@ -1143,7 +1168,6 @@ process_leaf_attr_block(
>  
>  	xfs_attr3_leaf_hdr_from_disk(&leafhdr, leaf);
>  	clearit = usedbs = 0;
> -	*repair = 0;
>  	firstb = mp->m_sb.sb_blocksize;
>  	stop = xfs_attr3_leaf_hdr_size(leaf);
>  
> @@ -1320,13 +1344,16 @@ process_leaf_attr_level(xfs_mount_t	*mp,
>  		}
>  
>  		bp = libxfs_readbuf(mp->m_dev, XFS_FSB_TO_DADDR(mp, dev_bno),
> -					XFS_FSB_TO_BB(mp, 1), 0, NULL);
> +				    XFS_FSB_TO_BB(mp, 1), 0,
> +				    &xfs_attr3_leaf_buf_ops);
>  		if (!bp) {
>  			do_warn(
>  	_("can't read file block %u (fsbno %" PRIu64 ") for attribute fork of inode %" PRIu64 "\n"),
>  				da_bno, dev_bno, ino);
>  			goto error_out;
>  		}
> +		if (bp->b_error == EFSBADCRC)
> +			repair++;
>  
>  		leaf = bp->b_addr;
>  		xfs_attr3_leaf_hdr_from_disk(&leafhdr, leaf);
> @@ -1382,9 +1409,9 @@ process_leaf_attr_level(xfs_mount_t	*mp,
>  
>  		current_hashval = greatest_hashval;
>  
> -		if (repair && !no_modify) 
> +		if (repair && !no_modify)
>  			libxfs_writebuf(bp, 0);
> -		else 
> +		else
>  			libxfs_putbuf(bp);
>  	} while (da_bno != 0);
>  
> @@ -1512,6 +1539,8 @@ process_longform_attr(
>  			ino);
>  		return(1);
>  	}
> +	if (bp->b_error == EFSBADCRC)
> +		(*repair)++;
>  
>  	/* verify leaf block */
>  	leaf = (xfs_attr_leafblock_t *)XFS_BUF_PTR(bp);
> @@ -1555,7 +1584,7 @@ process_longform_attr(
>  	case XFS_DA_NODE_MAGIC:		/* btree-form attribute */
>  	case XFS_DA3_NODE_MAGIC:
>  		/* must do this now, to release block 0 before the traversal */
> -		if (repairlinks) {
> +		if (*repair || repairlinks) {
>  			*repair = 1;
>  			libxfs_writebuf(bp, 0);
>  		} else

repairlinks incorporates a !no_modify check, but *repair does not. It's
incremented unconditionally if we find a CRC error. I suspect this means
we now need a !no_modify check for the writebuf/putbuf check here, as is
done in the alternate path at the end of the function.

Brian

> -- 
> 1.9.0
> 
> _______________________________________________
> xfs mailing list
> xfs at oss.sgi.com
> http://oss.sgi.com/mailman/listinfo/xfs