linux-kernel - Re: [PATCH] NFS4: Fix NULL deref in nfsd4_lock() by makeing unhash

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20120130214351.GA1518@fieldses.org>
Date:	Mon, 30 Jan 2012 16:43:51 -0500
From:	"J. Bruce Fields" <bfields@...ldses.org>
To:	Jesper Juhl <jj@...osbits.net>
Cc:	linux-nfs@...r.kernel.org, linux-kernel@...r.kernel.org,
	Kendrick Smith <kmsmith@...ch.edu>,
	Andy Adamson <kandros@...ch.edu>,
	Trond Myklebust <Trond.Myklebust@...app.com>
Subject: Re: [PATCH] NFS4: Fix NULL deref in nfsd4_lock() by makeing
 unhash_lockowner() safe to call with NULL arg

On Sun, Jan 29, 2012 at 10:29:24PM +0100, Jesper Juhl wrote:
> The Coverity checker noticed a path through nfsd4_lock() where we call
> release_lockowner(lock_sop); (at the 'out:' label) where 'lock_sop' is
> NULL.
> That goes bad since release_lockowner() calls unhash_lockowner() which
> dereferences its argument.
> release_lockowner() also calls nfs4_free_lockowner(), but that's not a
> problem since that function just calls kfree() and kmem_cache_free(),
> both of which are safe to call with NULL as argument.
> 
> There are several ways to fix the bug.
>  - rework nfsd4_lock() so the call to release_lockowner(NULL) will never happen.
>  - let release_lockowner() test for NULL and return if given one.
>  - let unhash_lockowner() test for NULL and return if given one (which makes both it and release_lockowner() safe).
> 
> I chose the last option for this patch.
> 
> For information, the path Coverity spotted (in defect report 201504) is this:
> 
...


> [...]
> 4098out:
> At conditional (12): "status" taking the true branch.
> At conditional (13): "new_state" taking the true branch.
> 4099        if (status && new_state)

new_state is initialized false, and referenced otherwise only once,
when a reference ot is is passed here:


> 4010                status = lookup_or_create_lock_state(cstate, open_stp, lock,
> 4011                                                        &lock_stp, &new_state);

so if new_state is true, then lookup_or_create_lock_state() set it to
true.  But it sets that only in one spot, at the end:

	*new = true;
	return nfs_ok;

Note nfs_ok is zero.  Therefore:

> At conditional (11): "status" taking the true branch.
> 4012                if (status)
> 4013                        goto out;

this could not have happened.

So it looks like a Coverity bug.

--b.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/