[Question] Return value of mii_bus->write()

[Question] Return value of mii_bus->write()

[Buday Csaba]

I am preparing a patch to eliminate kernel-doc warnings in mdio_device.c
and mdio_bus.c

I have ran into an ambiguity: what is mii_bus->write() supposed to
return on success? Documentation/networking/phy.txt does not give any
information about it, neither does the kdoc in include/linux/phy.h.

It is clear that 0 is treated as success, and a negative indicates
failure. The reference implementation also follows this convention.
But the code in mdio_bus.c, for example: __mdiobus_modify_changed(),
seems to also expect positive return values from write().

Is there any other implementation that allows positive return
values for success? Should it be mentioned in kernel-doc?

Thanks,
Csaba

Re: [Question] Return value of mii_bus->write()

[Heiner Kallweit]

On 11/21/2025 7:31 AM, Buday Csaba wrote:
> I am preparing a patch to eliminate kernel-doc warnings in mdio_device.c
> and mdio_bus.c
> 
> I have ran into an ambiguity: what is mii_bus->write() supposed to
> return on success? Documentation/networking/phy.txt does not give any
> information about it, neither does the kdoc in include/linux/phy.h.
> 
> It is clear that 0 is treated as success, and a negative indicates
> failure. The reference implementation also follows this convention.
> But the code in mdio_bus.c, for example: __mdiobus_modify_changed(),
> seems to also expect positive return values from write().
> 
I think you misread the code. __mdiobus_modify_changed() returns
a positive value in case new and old value differ, but __mdiobus_write()
never returns a positive value.

> Is there any other implementation that allows positive return
> values for success? Should it be mentioned in kernel-doc?
> 
> Thanks,
> Csaba
> 
> 

Re: Re: [Question] Return value of mii_bus->write()

[Buday Csaba]

On Fri, Nov 21, 2025 at 07:53:21AM +0100, Heiner Kallweit wrote:
> On 11/21/2025 7:31 AM, Buday Csaba wrote:
> > I am preparing a patch to eliminate kernel-doc warnings in mdio_device.c
> > and mdio_bus.c
> > 
> > I have ran into an ambiguity: what is mii_bus->write() supposed to
> > return on success? Documentation/networking/phy.txt does not give any
> > information about it, neither does the kdoc in include/linux/phy.h.
> > 
> > It is clear that 0 is treated as success, and a negative indicates
> > failure. The reference implementation also follows this convention.
> > But the code in mdio_bus.c, for example: __mdiobus_modify_changed(),
> > seems to also expect positive return values from write().
> > 
> I think you misread the code. __mdiobus_modify_changed() returns
> a positive value in case new and old value differ, but __mdiobus_write()
> never returns a positive value.
> 

That is right, thank you!
Csaba