From mboxrd@z Thu Jan  1 00:00:00 1970
From: Eric Dumazet <eric.dumazet@gmail.com>
Subject: Re: [Bugme-new] [Bug 38102] New: BUG kmalloc-2048: Poison
 overwritten
Date: Mon, 04 Jul 2011 16:45:05 +0200
Message-ID: <1309790705.2247.22.camel@edumazet-HP-Compaq-6005-Pro-SFF-PC>
References: <bug-38102-10286@https.bugzilla.kernel.org/>
	 <20110629145135.5179fb0c.akpm@linux-foundation.org>
	 <BANLkTi=PxH8V58k-VycsM=1ZiE9hPqCN4Q@mail.gmail.com>
	 <CAB9v_DHXB6n07sEY_=HZ8S4zMO01Vs8B78dGyEWrm-hyRP82XA@mail.gmail.com>
	 <1309707971.2523.20.camel@edumazet-laptop>
	 <CAB9v_DG7e8vDE+4PDwuOR2DYG1FEvUM1fxa+e4a=swwTYGJ9nQ@mail.gmail.com>
	 <1309788049.2247.9.camel@edumazet-HP-Compaq-6005-Pro-SFF-PC>
	 <20110704163134.7f2180b2@maggie>
Mime-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: QUOTED-PRINTABLE
Cc: Alexey Zaytsev <alexey.zaytsev@gmail.com>,
	Andrew Morton <akpm@linux-foundation.org>,
	netdev@vger.kernel.org, Gary Zambrano <zambrano@broadcom.com>,
	bugme-daemon@bugzilla.kernel.org,
	"David S. Miller" <davem@davemloft.net>,
	Pekka Pietikainen <pp@ee.oulu.fi>,
	Florian Schirmer <jolt@tuxbox.org>,
	Felix Fietkau <nbd@openwrt.org>, Michael Buesch <mb@bu3sch.de>
To: Michael =?ISO-8859-1?Q?B=FCsch?= <m@bues.ch>
Return-path: <netdev-owner@vger.kernel.org>
Received: from mail-wy0-f174.google.com ([74.125.82.174]:35184 "EHLO
	mail-wy0-f174.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S1753974Ab1GDOpV (ORCPT
	<rfc822;netdev@vger.kernel.org>); Mon, 4 Jul 2011 10:45:21 -0400
Received: by wyg8 with SMTP id 8so3447947wyg.19
        for <netdev@vger.kernel.org>; Mon, 04 Jul 2011 07:45:19 -0700 (PDT)
In-Reply-To: <20110704163134.7f2180b2@maggie>
Sender: netdev-owner@vger.kernel.org
List-ID: <netdev.vger.kernel.org>

Le lundi 04 juillet 2011 =C3=A0 16:31 +0200, Michael B=C3=BCsch a =C3=A9=
crit :
> On Mon, 04 Jul 2011 16:00:49 +0200
> Eric Dumazet <eric.dumazet@gmail.com> wrote:
> > > And an other question. Why so we have the logic to work-around th=
e 1Gb
> > > DMA limit instead of just setting the dma mask?
> >=20
> > Your problem is in RX side : NIC actually writes to a buffer that i=
s
> > supposedly not its property.
>=20
> The problem is on both sides, because some Linux architectures simply
> do not support any DMA mask less than 32. This applied to i386 (IA32)=
 last
> time I looked.
> The b44 DMA engine can only address 30-bits.


Michael, traces provided by Alexey are in the RX path.

NIC does a DMA  (Receives an UDP frame) into a 2048 bytes buffers that
was freed.