From mboxrd@z Thu Jan  1 00:00:00 1970
From: Al Viro <viro@ZenIV.linux.org.uk>
Subject: Re: [PATCH v3 105/110] namei: make unlazy_walk and terminate_walk
 handle nd->stack, add unlazy_link
Date: Tue, 12 May 2015 00:43:33 +0100
Message-ID: <20150511234333.GA7232@ZenIV.linux.org.uk>
References: <20150511180650.GA4147@ZenIV.linux.org.uk>
 <1431367690-5223-105-git-send-email-viro@ZenIV.linux.org.uk>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Cc: Neil Brown <neilb@suse.de>, Christoph Hellwig <hch@infradead.org>,
	linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org
To: Linus Torvalds <torvalds@linux-foundation.org>
Return-path: <linux-kernel-owner@vger.kernel.org>
Content-Disposition: inline
In-Reply-To: <1431367690-5223-105-git-send-email-viro@ZenIV.linux.org.uk>
Sender: linux-kernel-owner@vger.kernel.org
List-Id: linux-fsdevel.vger.kernel.org

On Mon, May 11, 2015 at 07:08:05PM +0100, Al Viro wrote:
> +static bool legitimize_links(struct nameidata *nd)
> +{
> +	int i;
> +	for (i = 0; i < nd->depth; i++) {
> +		struct saved *last = nd->stack + i;
> +		if (unlikely(!legitimize_path(nd, &last->link, last->seq))) {
> +			drop_links(nd);
> +			nd->depth = i;

Broken, actually - it should be i + 1.  What happens is that we attempt to
grab references on nd->stack[...].link; if everything succeeds, we'd won.
If legitimizing nd->stack[i].link fails (e.g. ->d_seq has changed on us),
we
	* put_link everything in stack and clear nd->stack[...].cookie, making
sure that nobody will call ->put_link() on it later.
	* leave the things for terminate_walk() so that it would do
path_put() on everything we have grabbed and ignored everything we hadn't
even got around to.

But this failed legitimize_path() requires path_put() - we *can't* block
there (we wouldn't be able to do ->put_link() afterwards if we did), so
we just zero what we didn't grab and leave what we had for subsequent
path_put().  Which may be anything from "nothing" (mount_lock has been
touched) to "both vfsmount and dentry" (->d_seq mismatch).

So we need to set nd->depth to i + 1 here, not i.  As it is, we are risking
a vfsmount (and possibly dentry) leak.  Fixed and force-pushed...