Re: [PATCH 4/9] 48bit support in extents

[Andrew Morton <akpm@osdl.org>]

On Wed, 09 Aug 2006 18:21:13 -0700
Mingming Cao <cmm@us.ibm.com> wrote:

> +/* this macro combines low and hi parts of phys. blocknr into ext4_fsblk_t */

This isn't a macro.

> +static inline ext4_fsblk_t ext_pblock(struct ext4_extent *ex)
> +{
> +	ext4_fsblk_t block;
> +
> +	block = le32_to_cpu(ex->ee_start);
> +	if (sizeof(ext4_fsblk_t) > 4)
> +		block |= ((ext4_fsblk_t) le16_to_cpu(ex->ee_start_hi) << 31) << 1;
> +	return block;
> +}

Oh.  I see the other patch did

	typedef sector_t ext4_fs_block_t;

(except someone misspelled "block" as "blk" ;))

Do we really want to do this?  I guess there's some value, for people with
32-bit machines who want extents and who are too mean to use a 64-bit
sector_t.  But gee that's marginal.  And it introduces interesting
compatibility questions and significantly adds to the testing burden.

I think that requiring 64-bit sector_t is reasonable?

Then again, I see from the other patches that considerable thought and
effort has gone into sustaining this turkey, so I guess I'm missing
something again.