Re: [rfc, patch 0/3] larger block group descriptors

On Jun 07, 2006  15:33 +0200, Alexandre Ratchov wrote:
> as recently pointed out by Andreas, there isn't enough room in the block
> group descriptor structure for all the fields we want to add in order to
> extend the ext3 file system and/or to add new features.
> 
> The attached patch addresses this problem, by allowing larger descriptor
> sizes. The descriptor size is stored in the super-block, it must be a power
> of 2, greater than or equal to 32 bytes (size of the current ext3_group_desc
> structure) and must fit in one block.
> 
> Temporary, i've added a new EXT3_FEATURE_INCOMPAT_DESC_SIZE feature. If
> large descriptors are used by other patches we could just merge them and use
> the same incompat feature.

Patches look good at first examination.  Use of INCOMPAT_DESC_SIZE should
hopefully only be for testing, but possibly not a big deal either way as
this patch will likely be tied to BIG_BG and other 64-bit changes.

I took a quick look bug haven't found a definitive "INCOMPAT_64BIT" patch
in the current BULL/RH/IBM/CFS patchset.  There is INCOMPAT_EXTENTS, which
is part of this, but there should likely be something else to ensure that
64-bit filesystem size in the superblock is flagged as an incompat change.

In the existing Bull patches RO_COMPAT_64BIT is used, but I don't think
this is a safe read-only compat change, since an old kernel will just use
s_blocks_count and not s_blocks_count_hi, truncating potentially a very
large part of the filesystem.

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