Re: [PATCH RFC v5 5/7] ipmi: ls2k: Relax the dependency to its mfd driver

[Miao Wang <shankerwangmiao@gmail.com>]

Hi,

> 2026年8月4日 19:39，Corey Minyard <corey@minyard.net> 写道：
> 
> On Tue, Aug 04, 2026 at 05:40:14PM +0800, Miao Wang wrote:
>> Hi,
>> 
>>> 2026年8月4日 04:46，Corey Minyard <corey@minyard.net> 写道：
>>> 
>>> On Tue, Aug 04, 2026 at 12:55:53AM +0800, Miao Wang via B4 Relay wrote:
>>>> From: Miao Wang <shankerwangmiao@gmail.com>
>>>> 
>>>> There is no strong dependency between the IPMI driver and its mfd
>>>> driver. Although the IPMI driver will not work without the mfd driver,
>>>> it is not a hard dependency. The IPMI driver can actually be compiled
>>>> without the mfd driver, and it will just fail to probe. When the mfd
>>>> driver is loaded, the IPMI driver will probe successfully. Therefore,
>>>> the dependency of the IPMI driver on its mfd driver should be relaxed
>>>> to "imply" from "select". This will allow the mfd driver to be compiled
>>>> as a module and the IPMI driver to be compiled as a part of the ipmi_si
>>>> module. The adjustment to Kconfig for the mfd driver will be introduced
>>>> in the later patch in this series.
>>> 
>>> I don't think that's what "imply" is for.  Imply seems to be for if
>>> there is another subsystem that can use this subsystem, but doesn't
>>> require it to exist.
>>> 
>>> For instance:
>>> 
>>> config SENSORS_NPCM7XX
>>>       tristate "Nuvoton NPCM750 and compatible PWM and Fan controllers"
>>>       imply THERMAL
>>> 
>>> The fan controller will work fine without the thermal subsystem; you
>>> can control the fan speed without it.  But the thermal subsystem is the
>>> logical user of this.  I looked at many of these things like this.
>>> 
>>> In the IPMI case, the IPMI driver is useless without the mfd part.  So
>>> there's no point in compiling the IPMI part of this if the mfd part is
>>> not there.
>>> 
>>> I could be wrong, but I can't see why you would want to do this.
>> 
>> The mfd part and the IPMI part loosely depend on each other. Without the
>> IPMI part, the mfd part can still work to handle the display part.
>> Without the mfd part, the IPMI part is indeed useless, but it will not
>> generate compiling errors or other runtime errors. In the runtime,
>> the IPMI part can be actually loaded earlier than the mfd part. As a
>> result, their dependency is not that strong.
> 
> The operational dependency is strong, which is what I think you want to
> convey here.
> 
>> 
>> The reason why I want to change this is that "select" here requires the
>> mfd part should also be compiled as built-in (i.e. = y), since the
>> type of the configure entry IPMI_LS2K is bool. However, I cannot see
>> there is no other reason preventing the mfd driver from compiling as
>> a module. This patch series will introduce a minor fix, after which
>> the mfd driver will be capable to be compiled as a module.
>> 
>> When the type of MFD_LS2K_BMC_CORE is changed to tristate, the "select"
>> here will prevent selecting =m for MFD_LS2K_BMC_CORE. I thus believe
>> that "select" here should be also changed.
>> 
>> Any suggestions on declaring the dependency of the both parts?
> 
> Ok, I understand now.
> 
> Why can't the IPMI part be compiled as a module?  Making that
> module-capable would be the right fix, I think.  I can't see
> why that wouldn't work.  IIRC, it was bool because the mfd part
> was bool.

I would have agree with you if the IPMI part was a normal driver.
However, the IPMI part is actually a part of ipmi_si. As a result,
the IPMI part can actually be compiled as a module, but as a part
of ipmi_si. That is why the configure entry IPMI_LS2K is bool rather
than tristate. I don't think it is trivial to convert IPMI_LS2K from
bool to tristate.

> If you changed to imply, you would need a depends on the mfd core, BTW.

If imply is used here, and there is "select MFD_CORE" in MFD_LS2K_BMC_CORE,
I think there is no need to add "depends on MFD_CORE" to IPMI_LS2K, if I
understand it correctly.

I don't insist using imply here, if there is a better way to express the
relation between both drivers.

> Also, IPMI_LS2K needs a "depends on IPMI_SI" either way.  I missed that
> earlier.  Could you add that?

If I understand it correctly, there seems to be no need to add the
dependency from the MFD driver to the driver for the actual sub-functions.
Examples are:

- MFD_MENF21BMC and SENSORS_MENF21BMC_HWMON, LEDS_MENF21BMC, 
  MENF21BMC_WATCHDOG: The later three declared depends on the former MFD
  driver
- MFD_INTEL_M10_BMC_CORE and FPGA_M10_BMC_SEC_UPDATE, 
  SENSORS_INTEL_M10_BMC_HWMON: The later two declared depends on the
  former MFD driver.
- MFD_STMFX and PINCTRL_STMFX: The later declared select the former
  MFD driver.

To summarize, as shown in the current mfd drivers, there is no need to
declare dependency to the drivers providing the sub-functions in the
mfd drivers, and the drivers of the sub functions should "depends on" or
"select" the mfd driver. So if IPMI_LS2K could be tristate, then it would
be reasonable to select MFD_LS2K_BMC_CORE or depends on MFD_LS2K_BMC_CORE.


Regarding the fact that IPMI_LS2K is bool and it is not an independent
module but a part of ipmi_si, it comes into my mind that we may choose
to adjust the dependency from ipmi_si. That is to remove "select
MFD_LS2K_BMC_CORE" from IPMI_LS2K, and add "select MFD_LS2K_BMC_CORE
if IPMI_LS2K" to IPMI_SI.

I wonder if such a change would be appropriate.

Cheers,

Miao Wang