Re: [PATCH v2 2/2] hwmon: (adt7475) Added attenuator bypass support

[Guenter Roeck <linux@roeck-us.net>]

On Fri, Dec 27, 2019 at 02:53:16AM +0000, Logan Shaw wrote:
> On Wed, 2019-12-18 at 19:53 -0800, Guenter Roeck wrote:
> > On 12/18/19 7:32 PM, Logan Shaw wrote:
> > > Added a new file documenting the adt7475 devicetree and added the
> > > five
> > > new properties to it.
> > > 
> > > Signed-off-by: Logan Shaw <logan.shaw@alliedtelesis.co.nz>
> > > ---
> > > ---
> > >   .../devicetree/bindings/hwmon/adt7475.txt     | 35
> > > +++++++++++++++++++
> > >   1 file changed, 35 insertions(+)
> > >   create mode 100644
> > > Documentation/devicetree/bindings/hwmon/adt7475.txt
> > > 
> > > diff --git a/Documentation/devicetree/bindings/hwmon/adt7475.txt
> > > b/Documentation/devicetree/bindings/hwmon/adt7475.txt
> > > new file mode 100644
> > > index 000000000000..388dd718a246
> > > --- /dev/null
> > > +++ b/Documentation/devicetree/bindings/hwmon/adt7475.txt
> > > @@ -0,0 +1,35 @@
> > > +*ADT7475 hwmon sensor.
> > > +
> > > +Required properties:
> > > +- compatible: One of
> > > +	"adi,adt7473"
> > > +	"adi,adt7475"
> > > +	"adi,adt7476"
> > > +	"adi,adt7490"
> > > +
> > > +- reg: I2C address
> > > +
> > > +optional properties:
> > > +
> > > +- bypass-attenuator-all: Configures bypassing all voltage input
> > > attenuators.
> > > +	This is only supported on the ADT7476 and ADT7490, this
> > > property does
> > > +	nothing on other chips.
> > 
> > Both adt7473 and adt7475 do support configuring an attenuator on VCCP
> > 
> 
> That is true, but as the function of register 0x73 bit 5 differs
> between the two set of hardware ({adt7473, adt7475} and {adt7476,
> adt7490}) a solution which allows bypassing VCCP on both gets more
> complicated which I was trying to avoid.
> 
> Is it acceptable to split the function
> load_individual_bypass_attenuators into two, one for each set of
> chips, and call the appropriate function for the chip? That way adding
> "bypass-attenuator-in1" to any of the four adt chips DTS will result in
> the attenuator for VCCP being bypassed (albeit by setting different
> bits depending on the specific bit).
> 
You could have per-chip functions, or per-chip mask data in the local data
structure.

> > > +		property present: Bit set to bypass all voltage input
> > > attenuators.
> > > +		property absent: Registers left unchanged.
> > > +
> > > +- bypass-attenuator-inx: Configures bypassing individual voltage
> > > input
> > > +	attenuators, where x is an integer from the set {0, 1, 3, 4}.
> > > This
> > > +	is only supported on the ADT7476 and ADT7490, this property
> > > does nothing
> > > +	on other chips.
> > > +		property present: Bit set to bypass specific voltage
> > > input attenuator
> > > +			for voltage input x.
> > > +		property absent: Registers left unchanged.
> > > +
> > 
> > This is interesting. It essentially means "retain configuration from
> > ROM
> > Monitor", but leaves no means to _disable_ the bypass.
> > 
> 
> Only a power cycle (after removing the properties from the DTS) will
> set the affected bits back to 0. Removing the properties, but only
> softly restarting the system (no interrupted power supply to the adt
> chip), will not set the bits back to 0.
> 
> I decided against setting the bits to 0 by default (if no property was
> present) as doing so might break compatibility with systems that expect
> the bits to remain unchanged. 
> 
> A compromise would be to change the "bypass-attenuator-inx" property to
> "bypass-attenuator-inx = <y>", where y = 1 bypasses the attenuator and
> y = 0 does not. If the property is not present then the register is
> left unchanged. It would make sense to do the same to the "bypass-
> attenuator-all" property. Would these changes be acceptable?
> 
That goes into devicetree object details. Rob might have an opinion
on that.

> > > +Example:
> > > +
> > > +hwmon@2e {
> > > +	compatible = "adi,adt7475";
> > > +	reg = <0x2e>;
> > > +	bypass-attenuator-all;
> > > +	bypass-attenuator-in1;
> > 
> > What would be the purpose of specifying both all and in1 ?
> 
> There would be no practical purpose, it was to keep the example
> compact. Instead "bypass-attenuator-in1" could be removed and added to 
> second device hwmon@2d. This would show off the syntax for each set of
> properties in a more practical way.
> 
That would make more sense.

> > 
> > > +};
> > > \ No newline at end of file

Note that you might want to fix that.

Guenter