Re: ext3: bogus i_mode errors with 2.6.18.1

[Andreas Dilger <adilger@clusterfs.com>]

On Oct 25, 2006  11:44 +0200, Andre Noll wrote:
> Are you saying that ext3_set_bit() should simply be called with
> "ret_block" as its first argument? If yes, that is what the revised
> patch below does.

You might need to call ext3_set_bit_atomic() (as claim_block() does,
not sure.

> @@ -1372,12 +1370,21 @@ allocated:
>  	    in_range(ret_block, le32_to_cpu(gdp->bg_inode_table),
>  		      EXT3_SB(sb)->s_itb_per_group) ||
>  	    in_range(ret_block + num - 1, le32_to_cpu(gdp->bg_inode_table),
> +		      EXT3_SB(sb)->s_itb_per_group)) {
> +		ext3_error(sb, __FUNCTION__,
>  			    "Allocating block in system zone - "
>  			    "blocks from "E3FSBLK", length %lu",
>  			     ret_block, num);
> +		/* Note: This will potentially use up one of the handle's
> +		 * buffer credits.  Normally we have way too many credits,
> +		 * so that is OK.  In _very_ rare cases it might not be OK.
> +		 * We will trigger an assertion if we run out of credits,
> +		 * and we will have to do a full fsck of the filesystem -
> +		 * better than randomly corrupting filesystem metadata.
> +		 */
> +		ext3_set_bit(ret_block, gdp_bh->b_data);
> +		goto repeat;
> +	}

The other issue is that you need to potentially set "num" bits in the
bitmap here, if those all overlap metadata.  In fact, it might just
make more sense at this stage to walk all of the bits in the bitmaps,
the inode table and the backup superblock and group descriptor to see
if they need fixing also.

Cheers, Andreas
--
Andreas Dilger
Principal Software Engineer
Cluster File Systems, Inc.