[U-Boot] [PATCH v3 3/9] dm: uclass: Add uclass_foreach_dev_probe

[Simon Glass]

Hi Patrick,

On 15 October 2018 at 03:01, Patrice CHOTARD <patrice.chotard at st.com> wrote:
> Hi Bin
>
> On 10/12/2018 12:13 PM, Bin Meng wrote:
>> Hi Patrice,
>>
>> On Fri, Oct 12, 2018 at 3:51 PM Patrice CHOTARD <patrice.chotard at st.com> wrote:
>>>
>>> Hi Bin
>>>
>>> On 10/11/2018 11:06 AM, Bin Meng wrote:
>>>> Hi Patrice,
>>>>
>>>> On Tue, Oct 9, 2018 at 9:41 PM Patrice Chotard <patrice.chotard at st.com> wrote:
>>>>>
>>>>> Add uclass_foreach_dev_probe() which iterates through
>>>>> devices of a given uclass. Devices are probed if necessary
>>>>> and are ready to use.
>>>>>
>>>>> Signed-off-by: Patrice Chotard <patrice.chotard at st.com>
>>>>> Reviewed-by: Simon Glass <sjg at chromium.org>
>>>>> ---
>>>>>
>>>>> Changes in v3: None
>>>>> Changes in v2: None
>>>>>
>>>>>  include/dm/uclass.h | 16 ++++++++++++++++
>>>>>  1 file changed, 16 insertions(+)
>>>>>
>>>>> diff --git a/include/dm/uclass.h b/include/dm/uclass.h
>>>>> index 6e7c1cd3e8bc..10ccfdce951e 100644
>>>>> --- a/include/dm/uclass.h
>>>>> +++ b/include/dm/uclass.h
>>>>> @@ -376,4 +376,20 @@ int uclass_resolve_seq(struct udevice *dev);
>>>>>  #define uclass_foreach_dev_safe(pos, next, uc) \
>>>>>         list_for_each_entry_safe(pos, next, &uc->dev_head, uclass_node)
>>>>>
>>>>> +/**
>>>>> + * uclass_foreach_dev_probe() - Helper function to iteration through devices
>>>>> + * of given uclass
>>>>> + *
>>>>> + * This creates a for() loop which works through the available devices in
>>>>> + * a uclass in order from start to end. Devices are probed if necessary,
>>>>> + * and ready for use.
>>>>> + *
>>>>> + * @id: Uclass ID
>>>>> + * @dev: struct udevice * to hold the current device. Set to NULL when there
>>>>> + * are no more devices.
>>>>> + */
>>>>> +#define uclass_foreach_dev_probe(id, dev)      \
>>>>> +       for (uclass_first_device(id, &dev); dev; \
>>>>> +            uclass_next_device(&dev))
>>>>
>>>> Shouldn't we check the return value of uclass_first_device() and
>>>> uclass_next_device()?
>>>
>>> It's not necessary to check the return value of uclass_first_device(id,
>>> &dev) because in any error case, dev is NULL, which is the loop output
>>> condition. This is only for the first iteration.
>>>
>>
>> Please see the comment of uclass_first_device().
>>
>>  * @devp: Returns pointer to the first device in that uclass if no error
>>  * occurred, or NULL if there is no first device, or an error occurred with
>>  * that device.
>>  * @return 0 if OK (found or not found), other -ve on error
>>
>> So it can return error with a valid dev.
>
> I checked the uclass_first_device() implementation and the comment seems
> not aligned with what is coded. When uclass_first_device() returns a
> valid dev, return value is always 0. (see uclass_get_device_tail()
> return value).

I actually don't like this function. Let's use
uclass_first_device_err() when we care about errors. I am wondering if
we should delete uclass_first_device().
>
> In any case uclass_first_device() returns a valid dev with an error.
>
> Simon, Bin do you confirm ?

I don't see how, no.
>
>>
>>> For the other iteration, dev comes from uclass_next_device(&dev),
>>> similarly to uclass_first_device(), in any error case dev is NULL.
>>>
>>
>> Similar please see the comment of uclass_next_device().
>
> I have the same remark regarding uclass_next_device().


Reviewed-by: Simon Glass <sjg at chromium.org>