Subject: Re: shmem_fill_super(): WARNING: kmemcheck: Caught 32-bit read
 from uninitialized memory
From: Pekka Enberg <penberg@cs.helsinki.fi>
To: Hugh Dickins <hugh.dickins@tiscali.co.uk>
Cc: Andrew Morton <akpm@linux-foundation.org>,
       Vegard Nossum <vegard.nossum@gmail.com>, Ingo Molnar <mingo@elte.hu>,
       linux-kernel@vger.kernel.org, Eric Paris <eparis@redhat.com>
In-Reply-To: <Pine.LNX.4.64.0909211126080.32504@sister.anvils>
References: <20090915080953.GA24958@elte.hu>
	 <20090920072210.GA23787@elte.hu>
	 <19f34abd0909201035t3157948amee532a3a5e96dbab@mail.gmail.com>
	 <alpine.DEB.2.00.0909202046460.7394@melkki.cs.helsinki.fi>
	 <Pine.LNX.4.64.0909211126080.32504@sister.anvils>
Date: Mon, 21 Sep 2009 14:07:19 +0300
Message-Id: <1253531239.5216.27.camel@penberg-laptop>
Mime-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 972
Lines: 21

Hi Hugh,

On Mon, 2009-09-21 at 11:49 +0100, Hugh Dickins wrote:
> I got a bit anxious when I saw that the mode arg to shmem_get_inode()
> is declared as an int: was afraid that compiler was then passing a bad
> upper half down, which in fact would cause no trouble, but how could it
> be sure of that?  However, it looks okay: after doing the 32-bit load,
> it goes on to do a movzwl %ax,%eax - seems an odd way to proceed to me,
> but I bet it knows a lot more about efficiency of memory loads than I do.

Yeah, that's a common cause of kmemcheck false positives. I guess GCC
wants to avoid a partial register stall in the memory load and expects
CPU register renaming to work for "movzwl %ax, %eax" or something.

			Pekka

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