[PATCH v1 3/4] sysinfo: rcar3: Implement BOARD_ID and BOARD_REVISION

[Marek Vasut]

On 6/20/23 22:40, Detlev Casanova wrote:
> On Tuesday, June 20, 2023 4:03:18 P.M. EDT Marek Vasut wrote:
>> On 6/20/23 19:49, Detlev Casanova wrote:
>>> On Monday, June 19, 2023 5:54:44 P.M. EDT Marek Vasut wrote:
>>>> On 6/19/23 20:27, Detlev Casanova wrote:
>>>>> On Monday, June 19, 2023 12:11:18 P.M. EDT Marek Vasut wrote:
>>>>>> On 6/19/23 16:42, Detlev Casanova wrote:
>>>>>>> On Friday, June 16, 2023 8:43:33 P.M. EDT Marek Vasut wrote:
>>>>>>>> On 6/16/23 17:21, Detlev Casanova wrote:
>>>>>>>>> Expose that information to the command shell to let scripts select
>>>>>>>>> the
>>>>>>>>> correct devicetree name.
>>>>>>>>>
>>>>>>>>> Signed-off-by: Detlev Casanova <detlev.casanova at collabora.com>
>>>>>>>>> ---
>>>>>>>>>
>>>>>>>>>       drivers/sysinfo/rcar3.c | 46
>>>>>>>>>       ++++++++++++++++++++++++++++++++---------
>>>>>>>>>       1 file changed, 36 insertions(+), 10 deletions(-)
>>>>>>>>>
>>>>>>>>> diff --git a/drivers/sysinfo/rcar3.c b/drivers/sysinfo/rcar3.c
>>>>>>>>> index 7b127986da7..89ad46c5422 100644
>>>>>>>>> --- a/drivers/sysinfo/rcar3.c
>>>>>>>>> +++ b/drivers/sysinfo/rcar3.c
>>>>>>>>> @@ -32,6 +32,8 @@
>>>>>>>>>
>>>>>>>>>        */
>>>>>>>>>       
>>>>>>>>>       struct sysinfo_rcar_priv {
>>>>>>>>>       
>>>>>>>>>       	char	boardmodel[64];
>>>>>>>>>
>>>>>>>>> +	u8	id;
>>>>>>>>> +	char	revision[4];
>>>>>>>>>
>>>>>>>>>       	u8	val;
>>>>>>>>>       
>>>>>>>>>       };
>>>>>>>>>
>>>>>>>>> @@ -48,17 +50,37 @@ static int sysinfo_rcar_get_str(struct udevice
>>>>>>>>> *dev,
>>>>>>>>> int id, size_t size, char *>
>>>>>>>>>
>>>>>>>>>       	switch (id) {
>>>>>>>>>
>>>>>>>>>       	case SYSINFO_ID_BOARD_MODEL:
>>>>>>>>> -		strncpy(val, priv->boardmodel, size);
>>>>>>>>> -		val[size - 1] = '\0';
>>>>>>>>> +		strlcpy(val, priv->boardmodel, size);
>>>>>>>>> +		break;
>>>>>>>>> +	case SYSINFO_ID_BOARD_REVISION:
>>>>>>>>> +		strlcpy(val, priv->revision, size);
>>>>>>>>> +		break;
>>>>>>>>> +	default:
>>>>>>>>> +		return -EINVAL;
>>>>>>>>> +	};
>>>>>>>>> +
>>>>>>>>> +	val[size - 1] = '\0';
>>>>>>>>> +	return 0;
>>>>>>>>> +}
>>>>>>>>> +
>>>>>>>>> +static int sysinfo_rcar_get_int(struct udevice *dev, int id, int
>>>>>>>>> *val)
>>>>>>>>> +{
>>>>>>>>> +	struct sysinfo_rcar_priv *priv = dev_get_priv(dev);
>>>>>>>>> +
>>>>>>>>> +	switch (id) {
>>>>>>>>> +	case SYSINFO_ID_BOARD_ID:
>>>>>>>>> +		*val = priv->id;
>>>>>>>>>
>>>>>>>>>       		return 0;
>>>>>>>>
>>>>>>>> Why not return SYSINFO_ID_BOARD_REVISION as integer here ?
>>>>>>>
>>>>>>> Because the revision (on r-car3 boards at least) is in the format X.Y.
>>>>>>> It
>>>>>>> could be returned as "(X << 8) | Y" or split in major/minor. But
>>>>>>> different
>>>>>>> boards will use different revisions and I think that having a str is
>>>>>>> easier to deal with in a shell script.
>>>>>>
>>>>>> With rcar they are numbers, so lets go with major/minor integers
>>>>>> please.
>>>>>
>>>>> Ok for this part, but shouldn't the sysinfo command use a common
>>>>> interface
>>>>> for all boards ? Or should it also have rev_major/rev_minor arguments ?
>>>>
>>>> I would expect other boards to either report rev_major/rev_minor if
>>>> implemented, or errno if those boards don't implement this property.
>>>
>>> Another thing on rcar is that the revision is stored as 2 char values.
>>> Would you oppose a change form using a char (e.g. rev_major = '1') to
>>> using u8 values (e.g. rev_major = 1) instead ?
>>
>> Shouldn't those rev fields just be integer(s) to cover the generic case?
> 
> On rcar, they are chars. I don't really see a reason for this except to show
> the '?.?' on unknown board ids. But that can be managed in other ways.

Yes I know, I am more concerned about other boards which might not have 
such short revision number, so why not just make the revision fields 
integer which covers very much anything ? Also, if you have those fields 
set to integer, then -EINVAL could be printed as '?' .