From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756678Ab2IFOws (ORCPT ); Thu, 6 Sep 2012 10:52:48 -0400 Received: from eusmtp01.atmel.com ([212.144.249.242]:13412 "EHLO eusmtp01.atmel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756492Ab2IFOwq (ORCPT ); Thu, 6 Sep 2012 10:52:46 -0400 Message-ID: <5048B8BD.3000804@atmel.com> Date: Thu, 6 Sep 2012 16:52:45 +0200 From: Nicolas Ferre Organization: atmel User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:15.0) Gecko/20120827 Thunderbird/15.0 MIME-Version: 1.0 To: David Miller CC: , , , , , , Subject: Re: [PATCH 04/10] net/macb: Fix a race in macb_start_xmit() References: <40f02ee50a29aaec6c949432a1bcf09f4b027181.1346775479.git.nicolas.ferre@atmel.com> <20120905.173023.2235590074897156746.davem@davemloft.net> In-Reply-To: <20120905.173023.2235590074897156746.davem@davemloft.net> X-Enigmail-Version: 1.4.4 Content-Type: text/plain; charset="ISO-8859-1" Content-Transfer-Encoding: 7bit X-Originating-IP: [10.161.30.18] Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 09/05/2012 11:30 PM, David Miller : > From: Nicolas Ferre > Date: Wed, 5 Sep 2012 10:19:11 +0200 > >> From: Havard Skinnemoen >> >> Fix a race in macb_start_xmit() where we unconditionally set the TSTART bit. >> If an underrun just happened (we do this with interrupts disabled, so it might >> not have been handled yet), the controller starts transmitting from the first >> entry in the ring, which is usually wrong. >> Restart the controller after error handling. >> >> Signed-off-by: Havard Skinnemoen >> [nicolas.ferre@atmel.com: split patch in topics] >> Signed-off-by: Nicolas Ferre > > Accumulating special case code and checks into the hot path of TX packet > processing is extremely unwise. > > Instead, when you handle the TX error conditions and reset the chip you > should first ensure that there are no flows of control in the transmit > function of your driver by using the appropriate locking et al. facilities. > > For example, you can quiesce the transmit path by handling the chip error > interrupt as follows: > > 1) Disable chip interrupt generation. > > 2) Schedule a workqueue so you can process the reset outside of hard > interrupt context. > > 3) In the workqueue function, disable NAPI and perform a > netif_tx_disable() to guarentee there are no threads of > execution trying to queue up packets for TX into the driver. > > 4) Perform your chip reset and whatever else is necessary. > > 5) Re-enable NAPI and TX. > > Then you don't need any special checks in your xmit method at all. I see... I will rework the series and try to implement this as part of the "[PATCH 06/10] net/macb: better manage tx errors" So this patch will disappear in future v2 series and patch 06 will be seriously modified. In fact I will also try to stack "cosmetic" patches at the beginning of the series. Thanks, best regards, -- Nicolas Ferre