From mboxrd@z Thu Jan 1 00:00:00 1970 Return-path: Received: from perceval.ideasonboard.com ([95.142.166.194]:40036 "EHLO perceval.ideasonboard.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752078Ab1IZTCr convert rfc822-to-8bit (ORCPT ); Mon, 26 Sep 2011 15:02:47 -0400 From: Laurent Pinchart To: Mauro Carvalho Chehab Subject: Re: cx231xx: DMA problem on ARM Date: Mon, 26 Sep 2011 21:02:48 +0200 Cc: Thomas Petazzoni , Devin Heitmueller , linux-media@vger.kernel.org, srinivasa.deevi@conexant.com, Maxime Ripard References: <20110921135604.64363a2e@skate> <20110926101323.41708d64@skate> <4E80B73F.5020804@redhat.com> In-Reply-To: <4E80B73F.5020804@redhat.com> MIME-Version: 1.0 Content-Type: Text/Plain; charset="utf-8" Content-Transfer-Encoding: 8BIT Message-Id: <201109262102.49056.laurent.pinchart@ideasonboard.com> Sender: linux-media-owner@vger.kernel.org List-ID: On Monday 26 September 2011 19:32:47 Mauro Carvalho Chehab wrote: > Em 26-09-2011 05:13, Thomas Petazzoni escreveu: > > Le Fri, 23 Sep 2011 23:15:54 -0300, Mauro Carvalho Chehab a écrit: > >>> And still the result is the same: we get a first frame, and then > >>> nothing more, and we have a large number of error messages in the > >>> kernel logs. > >> > >> I don't think that this is related to the power manager anymore. It can > >> be related to cache coherency and/or to iommu support. > > > > As you suspected, increasing PWR_SLEEP_INTERVAL didn't change anything. > > What do you suggest to track down the potential cache coherency issues ? Are you using the MMAP or USERPTR capture method ? If using MMAP, can you try (as a test only) to unmap the buffer before queueing it and to remap it after dequeuing it ? > Take a look at the ML. The SoC people discussed a lot about cache > coherency problems and how to solve it. Videobuf2 has a better support > on embedded world. I would take a look on it and see what it does different > than other drivers. Maybe Jonathan Corbet patches for the ccic driver may > help you. > > It is probably a good idea to change cx231xx to use videobuf2, in order to > fix this issue. -- Regards, Laurent Pinchart