From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from wes1-so1.wedos.net (wes1-so1.wedos.net [46.28.106.15]) by dpdk.org (Postfix) with ESMTP id 44517199A9 for ; Wed, 20 Sep 2017 22:39:30 +0200 (CEST) Received: from jvn (unknown [83.240.61.19]) by wes1-so1.wedos.net (Postfix) with ESMTPSA id 3xyBS14Vn2z6bK; Wed, 20 Sep 2017 22:39:29 +0200 (CEST) Date: Wed, 20 Sep 2017 22:39:21 +0200 From: Jan Viktorin To: Thomas Monjalon Cc: Ferruh Yigit , Shreyansh Jain , dev@dpdk.org, hemant.agrawal@nxp.com Message-ID: <20170920223921.19afec62@jvn> In-Reply-To: <2610477.kMnjFRTE32@xps> References: <20170823141213.25476-1-shreyansh.jain@nxp.com> <9142ad53-f75b-de33-cb8d-51ae5a781a2e@intel.com> <2610477.kMnjFRTE32@xps> Organization: RehiveTech X-Mailer: Claws Mail 3.15.0-dirty (GTK+ 2.24.31; x86_64-pc-linux-gnu) MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Subject: Re: [dpdk-dev] [PATCH v4 04/41] bus/dpaa: add OF parser for device scanning X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 20 Sep 2017 20:39:30 -0000 On Tue, 19 Sep 2017 22:01:23 +0200 Thomas Monjalon wrote: > 19/09/2017 16:15, Ferruh Yigit: > > On 9/19/2017 2:37 PM, Shreyansh Jain wrote: > > > On Monday 18 September 2017 08:19 PM, Ferruh Yigit wrote: > > >> On 9/9/2017 12:20 PM, Shreyansh Jain wrote: > > >>> This layer is used by Bus driver's scan function. Devices are parsed > > >>> using OF parser and added to DPAA device list. > > >> > > >> So this is device tree parser in DPDK. Do we really want this, and as > > >> long as DPDK target the bare metal why not get device information from > > >> Linux, as done in other cases? > > > As of now I don't prefer to modify the internal framework as much as > > > possible as this is stable DPDK DPAA driver. > > > There is indeed a planned transition from OF to /sys/ parsing, but it is > > > still in pipeline. > > > > > > You see a blocking issue if we go incremental here? > > > That would be probably more of replacing this file with another /sys > > > parser without much changes to the DPDK glue code. > > > > OF parser in DPDK looks weird to me, OS will do this for us already. > > > > If replacing this is in the roadmap, I think this is not showstopper, > > added Thomas in case he thinks otherwise. > > I agree with Ferruh. > > I am interested to know if there are cases where a device tree parser > would be relevant in DPDK. > Cc Jan who already worked on this idea. Hello, I don't know the details here. In general, I think it is better to always use /sys. However, there might be information in the device tree which are not exposed via /sys. This highly depends on the used driver. I was trying to use some generic driver (uio) which is very limited in many ways. I was also dealing with a specific HW configuration for FPGA where the NIC was divided into separate DMA and EMAC components. For DPDK, these would be two separate devices with not information how they are connected to each other. Such information was accessible only via the device tree. Finally, I also needed to control the PHY from DPDK. Again, information about PHY is unavailable via /sys. Regards Jan -- Jan Viktorin E-mail: Viktorin@RehiveTech.com System Architect Web: www.RehiveTech.com RehiveTech Brno, Czech Republic