Received-SPF: pass (google.com: domain of linux-kernel+bounces-48630-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45e3:2400::1 as permitted sender) client-ip=2604:1380:45e3:2400::1;
Date: Thu, 1 Feb 2024 19:00:00 +0100
From: Jiri Pirko <jiri@resnulli.us>
To: "David E. Box" <david.e.box@linux.intel.com>
Cc: netdev@vger.kernel.org, ilpo.jarvinen@linux.intel.com,
	sathyanarayanan.kuppuswamy@linux.intel.com,
	linux-kernel@vger.kernel.org, platform-driver-x86@vger.kernel.org
Subject: Re: [PATCH 4/8] platform/x86/intel/sdsi: Add netlink SPDM transport
Message-ID: <ZbvcICGKGcD3IHTs@nanopsycho>
References: <20240201010747.471141-1-david.e.box@linux.intel.com>
 <20240201010747.471141-5-david.e.box@linux.intel.com>
 <ZbtjyOBHzVKXu_4H@nanopsycho>
 <94a61858ac82ceaac1ef8ae41067ae7356512d7d.camel@linux.intel.com>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <94a61858ac82ceaac1ef8ae41067ae7356512d7d.camel@linux.intel.com>

Thu, Feb 01, 2024 at 05:42:33PM CET, david.e.box@linux.intel.com wrote:
>Hi Jiro,
>
>Thanks for your comments.
>
>On Thu, 2024-02-01 at 10:26 +0100, Jiri Pirko wrote:
>> Thu, Feb 01, 2024 at 02:07:43AM CET, david.e.box@linux.intel.com?wrote:
>> 
>> [...]
>> 
>> 
>> > +????? -
>> > +??????? name: spdm-req
>> > +??????? type: binary
>> > +????? -
>> > +??????? name: spdm-rsp
>> > +??????? type: binary
>> 
>> I don't understand the need to use netlink for this. Basically what you
>> do is you just use it to pass binary blobs to and from FW.
>> Advantages, like well-defined attributes, notifications etc, for which
>> it makes sense to use Netlink are not utilized at all.
>
>SPDM supports the setup of a secure channel between the responder and requestor
>using TLS based encryption algorthms. While this is just a transport for those
>blobs, netlink seemed an appropriate interface for this type of communication.
>The binary blobs can instead be broken out into the SPDM protocol messages,
>right out of the spec. But for our needs this would still just define the
>protocol. The algorithms themselves are not handled by the driver.

If that is a standard, break it from blob into well-defined attributes
and push it out of your driver to some common code.


>
>> Also, I don't thing it is good idea to have hw-driver-specific genl
>> family. I'm not aware of anything like that so far. Leave netlink
>> for use of generic and abstracted APIs.
>
>Sounds like an implied rule. If so should it be documented somewhere?
>
>> 
>> Can't you just have a simple misc device for this?
>
>It wouldn't be too much work to convert it.
>
>David