On Wed, Feb 10, 2016 at 03:28:00AM -0800, Christoph Hellwig wrote:
> > +static int
> > +xfs_writepage_submit(
> > + struct xfs_ioend *ioend,
> > + struct xfs_ioend *iohead,
> > + struct writeback_control *wbc,
> > + int status)
> > +{
> > + struct blk_plug plug;
> > +
> > + /* Reserve log space if we might write beyond the on-disk inode size. */
> > + if (!status && ioend && ioend->io_type != XFS_IO_UNWRITTEN &&
> > + xfs_ioend_is_append(ioend))
> > + status = xfs_setfilesize_trans_alloc(ioend);
> > +
> > + if (iohead) {
> > + blk_start_plug(&plug);
> > + xfs_submit_ioend(wbc, iohead, status);
> > + blk_finish_plug(&plug);
> > + }
> > + return status;
> > +}
>
> We return the xfs_setfilesize_trans_alloc failure status here,
> but none of the callers pick it up. The way this is handled later
> changes a bit, but even at the end of the series only 1 of the
> three callers handles the error.
I'll propagate it through where it makes sense. If we alrady have an
error, then we aren't going to call xfs_setfilesize_trans_alloc()
anyway, so checking the return value only matters in the non-error
cases.
Cheers,
Dave.
--
Dave Chinner
david@xxxxxxxxxxxxx
|