Re: [PATCH 3/3] watchdog: omap_wdt: add device tree support

[Jon Hunter <jon-hunter@ti.com>]

Hi Benoit,

On 05/30/2012 10:30 AM, Cousson, Benoit wrote:
> Hi Jon,
> 
> On 5/30/2012 5:03 PM, Jon Hunter wrote:
>> Hi Benoit,
>>
>> On 05/30/2012 02:54 AM, Cousson, Benoit wrote:
>>> On 5/30/2012 5:18 AM, Xiao Jiang wrote:
>>>> Jon Hunter wrote:
>>>>> On 05/25/2012 05:42 AM, jgq516@gmail.com wrote:
>>>>>> From: Xiao Jiang<jgq516@gmail.com>
>>>>>>
>>>>>> Add device table for omap_wdt to support dt.
>>>>>>
>>>>>> Signed-off-by: Xiao Jiang<jgq516@gmail.com>
>>>>>> ---
>>>>>> drivers/watchdog/omap_wdt.c | 8 ++++++++
>>>>>> 1 files changed, 8 insertions(+), 0 deletions(-)
>>>>>>
>>>>>> diff --git a/drivers/watchdog/omap_wdt.c
>>>>>> b/drivers/watchdog/omap_wdt.c
>>>>>> index 8285d65..d98c615 100644
>>>>>> --- a/drivers/watchdog/omap_wdt.c
>>>>>> +++ b/drivers/watchdog/omap_wdt.c
>>>>>> @@ -430,6 +430,13 @@ static int omap_wdt_resume(struct
>>>>>> platform_device *pdev)
>>>>>> #define omap_wdt_resume NULL
>>>>>> #endif
>>>>>>
>>>>>> +static const struct of_device_id omap_wdt_of_match[] = {
>>>>>> + { .compatible = "ti,omap3-wdt", },
>>>>>> + { .compatible = "ti,omap4-wdt", },
>>>
>>> If there is no difference between the OMAP3 and the OMAP4 WDT IP, just
>>> add one entry "ti,omap3-wdt". And then in the OMAP4 DTS you will just
>>> put : compatible = "ti,omap3-wdt"; or compatible =  "ti,omap4-wdt",
>>> "ti,omap3-wdt";
>>
>> Hmmm ... comparing the omap3 and omap4 wdt registers there are some
>> differences. omap4 seems to have more registers than omap3. May be we
>> are not using these right now, but from a register perspective the wdt
>> in omap2, omap3 and omap4 appear to be slightly different. The revision
>> ID register on omap3 and omap4 have different values too.
>>
>> I guess from a driver perspective there is no difference, but it seemed
>> to me that the IP is not completely the same.
> 
> Well, in that case, and assuming that there is no proper HW_REVISION
> information to detect the IP difference, the proper compatible entries
> will indeed have to be used.

So looking at a 4460 and 3430, the WIDR register (IP revision) can be
used to distinguish between IP revisions. So it appears that we do have
proper HW REV info.

So may be I am not completely up to speed of the intent of the
compatible field. In other words, should this be used to indicate if the
IP is same/compatible or the driver is compatible or both. Technically
right now we could just have "ti-omap2-wdt" for all omap2+ devices as
the driver is compatible for all devices. However, technically, the IP
is not completely the same but it is compatible :-)

>>> I'm still a little bit confused about the real need for the
>>> "ti,omap4-wdt: entry, but it seems to be the way to do it in PPC.
>>>
>>>>>> + {},
>>>>>> +};
>>>>>> +MODULE_DEVICE_TABLE(of, omap_wdt_of_match);
>>>>>> +
>>>>>> static struct platform_driver omap_wdt_driver = {
>>>>>> .probe = omap_wdt_probe,
>>>>>> .remove = __devexit_p(omap_wdt_remove),
>>>>>> @@ -439,6 +446,7 @@ static struct platform_driver omap_wdt_driver = {
>>>>>> .driver = {
>>>>>> .owner = THIS_MODULE,
>>>>>> .name = "omap_wdt",
>>>>>> + .of_match_table = omap_wdt_of_match,
>>>>>> },
>>>>>> };
>>>>>>
>>>>>
>>>>> I think we need to add some code to the probe function that calls
>>>>> of_match_device() and ensures we find a match. For example ...
>>>>>
>>>>> if (of_have_populated_dt())
>>>>> if (!of_match_device(omap_wdt_of_match,&pdev->dev))
>>>>> return -EINVAL;
>>>>>
>>>> Will add it in v2, thanks for suggestion.
>>>
>>> No, in fact this is not needed. We need that mainly when several
>>> instances can match the same driver and thus we select the proper one
>>> using the of_match_device. Otherwise, just check is the device_node is
>>> there.
>>>
>>> In that case, the driver does not even care about any DT node so there
>>> is no need to add extra code for that. Keep it simple.
>>
>> Ok. So are you saying get rid of the match table altogether? In other
>> words, drop this patch?
> 
> No, the match table is used by the LDM to find the proper driver to be
> bound to a device. So we do need it. But we do not have to use the
> of_match_device if we do not want to get the entry in the device table.

Ok, thanks.

>> I agree that it does not really do anything today, but I did not know if
>> in the future you were planning to pass things like, register addresses,
>> via DT.
> 
> Well, yes we will have to, otherwise people will keep complaining that
> our DTS sucks and are not compliant with the DTS standards :-)

Ok.

Jon