[Cluster-devel] [PATCH v3 6/8] gfs2: Implement iomap direct I/O support

[Bob Peterson]

Hi,

----- Original Message -----
> With that, the direct_IO address space operation can be all but
> eliminated: only a dummy remains which indicates that the filesystem
> supports direct I/O.
> 
> Signed-off-by: Andreas Gruenbacher <agruenba at redhat.com>
> ---
>  fs/gfs2/aops.c |  92 +------------------------------
>  fs/gfs2/bmap.c |  14 ++++-
>  fs/gfs2/file.c | 171
>  +++++++++++++++++++++++++++++++++++++++++++++++++++++----
>  3 files changed, 174 insertions(+), 103 deletions(-)
(snip)
> +		pos = iocb->ki_pos;
> +		endbyte = pos + buffered - 1;
> +		ret = filemap_write_and_wait_range(mapping, pos, endbyte);
> +		if (!ret) {
> +			iocb->ki_pos += buffered;
> +			written += buffered;
> +			invalidate_mapping_pages(mapping,
> +						 pos >> PAGE_SHIFT,
> +						 endbyte >> PAGE_SHIFT);
> +		} else {
> +			/*
> +			 * We don't know how much we wrote, so just return
> +			 * the number of bytes which were direct-written
> +			 */

Can't we get in trouble here because we don't update iocb->io_pos?

Regards,

Bob Peterson