From mboxrd@z Thu Jan 1 00:00:00 1970 From: Thomas Monjalon Subject: Re: [PATCH 04/13] pci: use scan_mode configuration Date: Thu, 13 Jul 2017 22:48:34 +0200 Message-ID: <6818817.jrHzc3MJ66@xps> References: <20170711232512.54641-1-jblunck@infradead.org> <20170713175900.GL11154@bidouze.vm.6wind.com> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Cc: dev@dpdk.org To: jblunck@infradead.org, gaetan.rivet@6wind.com Return-path: Received: from out1-smtp.messagingengine.com (out1-smtp.messagingengine.com [66.111.4.25]) by dpdk.org (Postfix) with ESMTP id 037202BAA for ; Thu, 13 Jul 2017 22:48:38 +0200 (CEST) In-Reply-To: List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" 13/07/2017 21:42, Jan Blunck: > On Thu, Jul 13, 2017 at 1:59 PM, Ga=EBtan Rivet = wrote: > > On Tue, Jul 11, 2017 at 07:25:03PM -0400, Jan Blunck wrote: > >> When scanning/probing devices the bus should use its configuration ins= tead > >> of looking at the devargs->type field. > >> > > > > With this patch, how do you probe a device that was previously > > blacklisted? > > > > The answers I see to this question are pretty bad, maybe you have a good > > solution. > > > > On the other hand, can you explain why you want this limitation? What > > problem does this solve? You have one view of the hotplug API, I would > > like to understand why you hold this view. > > > > Regarding the rte_devargs API, it can be fixed without making the > > hotplug needlessly complicated I think. > > > > I must point out that the scan_mode (incorrectly named) is something > > that will be removed next release. The probe policies will be reworked > > and I don't think that the solution should be proposed as a fix a few > > days before the RC2. >=20 > You just introduced them in 17.08-rc1 and you want to remove them > again for 17.11?! Please revert these changes in this case for > 17.08-rc2. Please remember that it has been introduced to prepare the move of the bus drivers and allow some kind of hotplug as used in failsafe, without breaking the devargs syntax. This is a step in an incremental process, and experimental functions and deprecated API can be dropped for a better replacement in 17.11. > Thomas, what is your take on this? I did not change my mind: We must deprecate the syntax in devargs for 17.11. Then, with a new syntax, it will be possible to simplify a lot of things, including the probe policies. =46or now, we must work on 17.08-rc2 with two goals: - fix the API break introduced in devargs API - fix the reworked hotplug to make it work in basic PCI cases - make sure the new failsafe PMD can be integrated I will integrate only the patches which clearly fix something. Patches with justification "it would be better" will wait for 17.11. Are these rules clear enough to let us progress together in the 17.08 timeframe, and 17.11 cycle? Thanks for your efforts and making 17.08 release possible.