Re: [PATCH 5/6] xfs: handle reconnecting metadir subdirectories

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

On Thu, Aug 27, 2026 at 02:53:30PM +0530, Anuj Gupta wrote:
> > +	error = xfs_dir_createname(tp, dp, name, I_INO(ip), resblks);
> > +	if (error)
> > +		return error;
> > +
> > +	xfs_trans_log_inode(tp, dp, XFS_ILOG_CORE);
> 
> should we update dp's mtime/ctime after creating the dirent using
> xfs_trans_ichgtime()?

Metadata directory trees aren't exposed to userspace, why would we care
about timestamps?

> > +		/*
> > +		 * subdir didn't give us a dotdot entry, so we just give up
> > +		 * and let the repair get marked as failed.
> > +		 */
> > +		if (old_dotdot == NULLFSINO)
> > +			return 0;
> > +	}
> 
> The comment says "let the repair get marked as failed" but return 0 here
> looks like it reports success.

The repair reports success, and the subsequent revalidation sees that
the metapath is still broken and reports that corruption to userspace.
I suppose -EFSCORRUPTED here would be clearer.

--D