Message-ID: <ae7f8b87-5cf5-493d-9cff-eb76bebda513@redhat.com>
Date: Fri, 25 Apr 2025 12:29:11 +0200
From: Ivan Vecera <ivecera@...hat.com>
To: Lee Jones <lee@...nel.org>
Cc: Andrew Lunn <andrew@...n.ch>, netdev@...r.kernel.org,
 Vadim Fedorenko <vadim.fedorenko@...ux.dev>,
 Arkadiusz Kubalewski <arkadiusz.kubalewski@...el.com>,
 Jiri Pirko <jiri@...nulli.us>, Rob Herring <robh@...nel.org>,
 Krzysztof Kozlowski <krzk+dt@...nel.org>, Conor Dooley
 <conor+dt@...nel.org>, Prathosh Satish <Prathosh.Satish@...rochip.com>,
 Kees Cook <kees@...nel.org>, Andy Shevchenko <andy@...nel.org>,
 Andrew Morton <akpm@...ux-foundation.org>,
 Michal Schmidt <mschmidt@...hat.com>, devicetree@...r.kernel.org,
 linux-kernel@...r.kernel.org, linux-hardening@...r.kernel.org
Subject: Re: [PATCH net-next v4 5/8] mfd: zl3073x: Add functions to work with
 register mailboxes



On 25. 04. 25 12:17 odp., Lee Jones wrote:
> On Fri, 25 Apr 2025, Ivan Vecera wrote:
> 
>>
>>
>> On 25. 04. 25 8:55 dop., Lee Jones wrote:
>>> On Thu, 24 Apr 2025, Ivan Vecera wrote:
>>>
>>>>
>>>>
>>>> On 24. 04. 25 9:29 odp., Andrew Lunn wrote:
>>>>>> Yes, PHC (PTP) sub-driver is using mailboxes as well. Gpio as well for some
>>>>>> initial configuration.
>>>>>
>>>>> O.K, so the mailbox code needs sharing. The question is, where do you
>>>>> put it.
>>>>
>>>> This is crucial question... If I put the MB API into DPLL sub-driver
>>>> then PTP sub-driver will depend on it. Potential GPIO sub-driver as
>>>> well.
>>>>
>>>> There could be some special library module to provide this for
>>>> sub-drivers but is this what we want? And if so where to put it?
>>>
>>> MFD is designed to take potentially large, monolithic devices and split
>>> them up into smaller, more organised chunks, then Linusify them.  This
>>> way, area experts (subsystem maintainers) get to concern themselves only
>>> with the remit to which they are most specialised / knowledgable.  MFD
>>> will handle how each of these areas are divided up and create all of the
>>> shared resources for them.  On the odd occasion it will also provide a
>>> _small_ API that the children can use to talk to the parent device.
>>>
>>> However .... some devices, like yours, demand an API which is too
>>> complex to reside in the MFD subsystem itself.  This is not the first
>>> time this has happened and I doubt it will be the last.  My first
>>> recommendation is usually to place all of the comms in drivers/platform,
>>> since, at least in my own mind, if a complex API is required, then the
>>> device has become almost platform-like.  There are lots of examples of
>>> H/W comm APIs in there already for you to peruse.
>>
>> OK, I will do it differently... Will drop MB API at all from MFD and
>> just expose the additional mutex from MFD for multi-op access.
>> Mailboxes will be handled directly by sub-devices.
>>
>> Short description:
>> MFD exposes:
>> zl3073x_{read,write}_u{8,16,32,48}() & zl3073x_poll_u8()
>> - to read/write/poll registers
>> - they checks that multiop_lock is taken when caller is accessing
>>    registers from Page 10 and above
>>
>> zl3073x_multiop_{lock,unlock}()
>> - to protect operation where multiple reads, writes and poll is required
>>    to be done atomically

Or maybe just define zl3073x_dev.multiop_lock and allow callers to use

guard(mutex)(zldev->multiop_lock)
--or--
scoped_guard(mutex)(zldev_multiop_lock) {
...
}

> Looks sensible.  If this is aligned with the discussions that have been
> taking place between you and Andrew.  Let's see the code before we make
> any binding agreements.  =:)
> 

Powered by blists - more mailing lists