[Ocfs2-devel] [patch 6/8] ocfs2: manually do the iput once ocfs2_add_entry failed in ocfs2_symlink and ocfs2_mknod

[Mark Fasheh]

On Wed, Mar 19, 2014 at 02:10:04PM -0700, Andrew Morton wrote:
> From: jiangyiwen <jiangyiwen at huawei.com>
> Subject: ocfs2: manually do the iput once ocfs2_add_entry failed in ocfs2_symlink and ocfs2_mknod
> 
> When the call to ocfs2_add_entry() failed in ocfs2_symlink() and
> ocfs2_mknod(), iput() will not be called during dput(dentry) because no
> d_instantiate(), and this will lead to umount hung.

This is probably the heaviest way to do this. If you're getting a hang on
mount, there's something wrong with the cleanup code in fs/ocfs2/dcache.c.
See the following comment just above ocfs2_add_entry()

	/*
	 * Do this before adding the entry to the directory. We add
	 * also set d_op after success so that ->d_iput() will cleanup
	 * the dentry lock even if ocfs2_add_entry() fails below.
	 */
	status = ocfs2_dentry_attach_lock(dentry, inode,
					  OCFS2_I(dir)->ip_blkno);
	if (status) {
		mlog_errno(status);
		goto leave;
	}

	status = ocfs2_add_entry(handle, dentry, inode,
				 OCFS2_I(inode)->ip_blkno, parent_fe_bh,
				 &lookup);
	if (status < 0) {
		mlog_errno(status);
		goto leave;
	}

So if you hung during unmount on a dentry lock, I think we need to
investigate that path of recovery first. If there's something that it simply
can't handle, then the large block of code you copied into namei.c needs to
go into it's own function (or be a function broken out of the dentry lock
cleanup code).

Thanks,
	--Mark

--
Mark Fasheh