Re: kernel BUG in fs/dcache.c running nfs

From: NeilBrown <neilb@suse.de>
To: Al Viro <viro@ZenIV.linux.org.uk>
Cc: Jeff Layton <jeff.layton@primarydata.com>,
	Christoph Hellwig <hch@infradead.org>,
	linux-nfs@vger.kernel.org
Subject: Re: kernel BUG in fs/dcache.c running nfs
Date: Mon, 29 Sep 2014 11:35:19 +1000	[thread overview]
Message-ID: <20140929113519.30292efb@notabene.brown> (raw)
In-Reply-To: <20140926230801.GR7996@ZenIV.linux.org.uk>

[-- Attachment #1: Type: text/plain, Size: 1383 bytes --]

On Sat, 27 Sep 2014 00:08:01 +0100 Al Viro <viro@ZenIV.linux.org.uk> wrote:

> On Wed, Sep 10, 2014 at 01:57:39PM +1000, NeilBrown wrote:
> 
> > The d_drop();d_add(); pattern is used a number of times in NFS, but what I
> > didn't notice before is that it is only used if ->d_inode is NULL.
> 
> And the point of that "pattern" would be...?  If dentry is already negative
> and hashed, it does nothing whatsoever.  If it is negative and unhashed,
> d_drop() is a no-op, obviously.  And if it's positive, we get a nice
> shiny oops.
> 
> Where are we doing anything of that sort, anyway?  I see a bloody odd
> instance in nfs_atomic_open() and that's it.  Note that nfs_instantiate()
> is different (and not necessary nice - that d_drop() in case when this
> sucker is called from mkdir() is asking for races with mount()).  How
> do we get a positive hashed dentry there, anyway?

Apart from nfs_atomic_open() and  nfs_instantiate(), the main one is
in _nfs4_open_and_get_state:
		/* FIXME: Is this d_drop() ever needed? */
		d_drop(dentry);
		dentry = d_add_unique(dentry, igrab(state->inode));
   maybe you can answer the question in the comment

nfs_link also has

	d_drop(dentry);
	error = NFS_PROTO(dir)->link(inode, dir, &dentry->d_name);
	if (error == 0) {
		ihold(inode);
		d_add(dentry, inode);
	}

which is similar.

Thanks,
NeilBrown

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 828 bytes --]