From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757125AbcK3IhD (ORCPT ); Wed, 30 Nov 2016 03:37:03 -0500 Received: from www381.your-server.de ([78.46.137.84]:47359 "EHLO www381.your-server.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756181AbcK3Ig5 (ORCPT ); Wed, 30 Nov 2016 03:36:57 -0500 Subject: Re: [PATCH 1/1] ALSA: SOC: DMA: increment buffer pointer atomically To: Jiada Wang , perex@perex.cz, tiwai@suse.com References: <20161130082221.15399-1-jiada_wang@mentor.com> <36d5025d-d327-878d-7f2e-5bf0c457581d@metafoo.de> Cc: alsa-devel@alsa-project.org, linux-kernel@vger.kernel.org, apape@de.adit-jv.com From: Lars-Peter Clausen Message-ID: Date: Wed, 30 Nov 2016 09:36:16 +0100 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101 Icedove/45.4.0 MIME-Version: 1.0 In-Reply-To: <36d5025d-d327-878d-7f2e-5bf0c457581d@metafoo.de> Content-Type: text/plain; charset=windows-1252 Content-Transfer-Encoding: 7bit X-Authenticated-Sender: lars@metafoo.de Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 11/30/2016 09:30 AM, Lars-Peter Clausen wrote: > On 11/30/2016 09:22 AM, Jiada Wang wrote: >> From: Andreas Pape >> >> Setting pointer and afterwards check for wrap around leads >> to the possibility of returning the inconsistent pointer position. >> This patch increments buffer pointer atomically to avoid this issue. > > Makes sense. > >> >> Signed-off-by: Andreas Pape >> Signed-off-by: Jiada Wang >> --- >> sound/core/pcm_dmaengine.c | 8 +++++--- >> 1 file changed, 5 insertions(+), 3 deletions(-) >> >> diff --git a/sound/core/pcm_dmaengine.c b/sound/core/pcm_dmaengine.c >> index 8eb58c7..6f6da11 100644 >> --- a/sound/core/pcm_dmaengine.c >> +++ b/sound/core/pcm_dmaengine.c >> @@ -139,12 +139,14 @@ EXPORT_SYMBOL_GPL(snd_dmaengine_pcm_set_config_from_dai_data); >> >> static void dmaengine_pcm_dma_complete(void *arg) >> { >> + unsigned int new_pos; >> struct snd_pcm_substream *substream = arg; >> struct dmaengine_pcm_runtime_data *prtd = substream_to_prtd(substream); >> >> - prtd->pos += snd_pcm_lib_period_bytes(substream); >> - if (prtd->pos >= snd_pcm_lib_buffer_bytes(substream)) >> - prtd->pos = 0; >> + new_pos = prtd->pos + snd_pcm_lib_period_bytes(substream); >> + if (new_pos >= snd_pcm_lib_buffer_bytes(substream)) >> + new_pos = 0; >> + prtd->pos = new_pos; > > But to really make it atomic I think this needs READ_ONCE/WRITE_ONCE. And the access to prtd->pos in snd_dmaengine_pcm_pointer_no_residue() should also use READ_ONCE(). It is very unlikely that the code gets mis-compiled to generate more than one access, but having READ_ONCE() acts as a annotation that makes it explicit that this is data that can be updated concurrently without further synchronization.