Re: [PATCH v2 2/9] ext2: tell DAX the size of allocation holes

[Ross Zwisler]

On Tue, Aug 23, 2016 at 04:04:12PM -0600, Ross Zwisler wrote:
> When DAX calls ext2_get_block() and the file offset points to a hole we
> currently don't set bh_result->b_size.  When we re-enable PMD faults DAX
> will need bh_result->b_size to tell it the size of the hole so it can
> decide whether to fault in a 4 KiB zero page or a 2 MiB zero page.
> 
> Signed-off-by: Ross Zwisler <ross.zwisler@xxxxxxxxxxxxxxx>
> ---
>  fs/ext2/inode.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/fs/ext2/inode.c b/fs/ext2/inode.c
> index d5c7d09..dd55d74 100644
> --- a/fs/ext2/inode.c
> +++ b/fs/ext2/inode.c
> @@ -773,6 +773,9 @@ int ext2_get_block(struct inode *inode, sector_t iblock, struct buffer_head *bh_
>  	if (ret > 0) {
>  		bh_result->b_size = (ret << inode->i_blkbits);
>  		ret = 0;
> +	} else if (ret == 0) {
> +		/* hole case, need to fill in bh_result->b_size */
> +		bh_result->b_size = 1 << inode->i_blkbits;
>  	}
>  	return ret;
>  
> -- 
> 2.9.0
> 

Jan, is it possible for ext2 to return 2 MiB of contiguous space to us via
ext2_get_block()?

I ask because we have all the infrastructure in place for ext2 to handle PMD
faults (ext2_dax_pmd_fault(), etc.), but I don't think in my testing I've ever
seen this actually happen.

ext2 can obviously return multiple blocks from ext2_get_block(), but can it
actually satisfy a whole PMD's worth (512 contiguous blocks)?  If so, what
steps do I need to take to get this to work in my testing?

If it can't happen, we should probably rip out ext2_dax_pmd_fault() so that we
don't have to keep falling back to PTEs via the PMD path.

Thanks,
- Ross