Re: [PATCH 02/10] xfs: remove the i_mode check in xfs_release

["Darrick J. Wong" <djwong@kernel.org>]

On Mon, Jun 24, 2024 at 05:50:11PM +0200, Christoph Hellwig wrote:
> On Mon, Jun 24, 2024 at 08:34:59AM -0700, Darrick J. Wong wrote:
> > > -	if (!S_ISREG(VFS_I(ip)->i_mode) || (VFS_I(ip)->i_mode == 0))
> > 
> > How would we encounter !i_mode regular files being released?
> 
> We can't.  If that code ever made any sense than in ancient pre-history
> in IRIX.
> 
> > If an open file's link count is incorrectly low, it can't get freed
> > until after all the open file descriptors have been released, right?
> > Or is there some other vector for this?
> 
> No.
> 
> > I'm wondering if this ought to be:
> > 
> > 	if (XFS_IS_CORRUPT(mp, !VFS_I(ip)->i_mode)) {
> > 		xfs_inode_mark_sick(ip);
> > 		return -EFSCORRUPTED;
> > 	}
> 
> I wouldn't even bother with that.

Oh, right, because xfs_iget checks that for us and bails out before the
vfs can even get its hands on the file.

Reviewed-by: Darrick J. Wong <djwong@kernel.org>

--D