Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Wed, 14 Jun 2023 11:27:27 +0200
From:   Jiri Pirko <jiri@resnulli.us>
To:     Jakub Kicinski <kuba@kernel.org>
Cc:     Arkadiusz Kubalewski <arkadiusz.kubalewski@intel.com>,
        vadfed@meta.com, jonathan.lemon@gmail.com, pabeni@redhat.com,
        corbet@lwn.net, davem@davemloft.net, edumazet@google.com,
        vadfed@fb.com, jesse.brandeburg@intel.com,
        anthony.l.nguyen@intel.com, saeedm@nvidia.com, leon@kernel.org,
        richardcochran@gmail.com, sj@kernel.org, javierm@redhat.com,
        ricardo.canuelo@collabora.com, mst@redhat.com, tzimmermann@suse.de,
        michal.michalik@intel.com, gregkh@linuxfoundation.org,
        jacek.lawrynowicz@linux.intel.com, airlied@redhat.com,
        ogabbay@kernel.org, arnd@arndb.de, nipun.gupta@amd.com,
        axboe@kernel.dk, linux@zary.sk, masahiroy@kernel.org,
        benjamin.tissoires@redhat.com, geert+renesas@glider.be,
        milena.olech@intel.com, kuniyu@amazon.com, liuhangbin@gmail.com,
        hkallweit1@gmail.com, andy.ren@getcruise.com, razor@blackwall.org,
        idosch@nvidia.com, lucien.xin@gmail.com, nicolas.dichtel@6wind.com,
        phil@nwl.cc, claudiajkang@gmail.com, linux-doc@vger.kernel.org,
        linux-kernel@vger.kernel.org, netdev@vger.kernel.org,
        intel-wired-lan@lists.osuosl.org, linux-rdma@vger.kernel.org,
        linux-arm-kernel@lists.infradead.org, poros@redhat.com,
        mschmidt@redhat.com, linux-clk@vger.kernel.org,
        vadim.fedorenko@linux.dev
Subject: Re: [RFC PATCH v8 01/10] dpll: documentation on DPLL subsystem
 interface
Message-ID: <ZImH/6GzGdydC3U3@nanopsycho>
References: <20230609121853.3607724-1-arkadiusz.kubalewski@intel.com>
 <20230609121853.3607724-2-arkadiusz.kubalewski@intel.com>
 <20230612154329.7bd2d52f@kernel.org>
 <ZIg8/0UJB9Lbyx2D@nanopsycho>
 <20230613093801.735cd341@kernel.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20230613093801.735cd341@kernel.org>
Precedence: bulk

Tue, Jun 13, 2023 at 06:38:01PM CEST, kuba@kernel.org wrote:
>On Tue, 13 Jun 2023 11:55:11 +0200 Jiri Pirko wrote:
>> >> +``'pin': [{
>> >> + {'clock-id': 282574471561216,
>> >> +  'module-name': 'ice',
>> >> +  'pin-dpll-caps': 4,
>> >> +  'pin-id': 13,
>> >> +  'pin-parent': [{'pin-id': 2, 'pin-state': 'connected'},
>> >> +                 {'pin-id': 3, 'pin-state': 'disconnected'},
>> >> +                 {'id': 0, 'pin-direction': 'input'},
>> >> +                 {'id': 1, 'pin-direction': 'input'}],
>> >> +  'pin-type': 'synce-eth-port'}
>> >> +}]``  
>> >
>> >It seems like pin-parent is overloaded, can we split it into two
>> >different nests?  
>> 
>> Yeah, we had it as two and converged to this one. The thing is, the rest
>> of the attrs are the same for both parent pin and parent device. I link
>> it this way a bit better. No strong feeling.
>
>Do you mean the same attribute enum / "space" / "set"?

Yeah, that is my understanding. Arkadiusz, could you please clarify?


>In the example above the attributes present don't seem to overlap.
>For user space its an extra if to sift thru the objects under
>pin-parent.
>
>> >Also sounds like setting pin attrs and pin-parent attrs should be
>> >different commands.  
>> 
>> Could be, but what't the benefit? Also, you are not configuring
>> pin-parent. You are configuring pin:pin-parent tuple. Basically the pin
>> configuration as a child. So this is mainly config of the pin itsest
>> Therefore does not really make sense to me to split to two comments.
>
>Clarity of the API. If muxing everything thru few calls was the goal
>we should also have very few members in struct dpll_pin_ops, and we
>don't.

How the the internal kernel api related to the uapi? I mean, it's quite
common to have 1:N relationsip between cmd and op. I have to be missing
your point. Could you be more specific please?

Current code presents PIN_SET command with accepts structured set of
attribute to be set. The core-driver api is pretty clear. Squashing to
a single op would be disaster. Having one command per attr looks like an
overkill without any real benefit. How exactly do you propose to change
this?