Re: [PATCH] gpiolib: fix allocation of mixed dynamic/static GPIOs

[Christophe Leroy <christophe.leroy@xxxxxxxxxx>]

Le 27/04/2023 à 12:46, Andy Shevchenko a écrit :
> On Thu, Apr 27, 2023 at 1:37 PM Andreas Kemnade <andreas@xxxxxxxxxxxx> wrote:
>>
>> On Thu, 27 Apr 2023 06:20:34 +0000
>> Christophe Leroy <christophe.leroy@xxxxxxxxxx> wrote:
>>
>>> Le 27/04/2023 à 08:00, Andy Shevchenko a écrit :
>>>> On Thu, Apr 27, 2023 at 8:40 AM Christophe Leroy
>>>> <christophe.leroy@xxxxxxxxxx> wrote:
>>>>>
>>>>>
>>>>>
>>>>> Le 27/04/2023 à 00:03, Andreas Kemnade a écrit :
>>>>>> [Vous ne recevez pas souvent de courriers de andreas@xxxxxxxxxxxx. Découvrez pourquoi ceci est important à https://aka.ms/LearnAboutSenderIdentification ]
>>>>>>
>>>>>> If static allocation and dynamic allocation GPIOs are present,
>>>>>> dynamic allocation pollutes the numberspace for static allocation,
>>>>>> causing static allocation to fail.
>>>>>> Enfore dynamic allocation above GPIO_DYNAMIC_BASE.
>>>>>
>>>>> Hum ....
>>>>>
>>>>> Commit 7b61212f2a07 ("gpiolib: Get rid of ARCH_NR_GPIOS") was supposed
>>>>> to enforce dynamic allocation above GPIO_DYNAMIC_BASE already.
>>>>>
>>>>> Can you describe what is going wrong exactly with the above commit ?
>>>>
>>>> Above commit only works to the first dynamic allocation, if you need
>>>> more than one with static ones present it mistakenly will give you a
>>>> base _below_ DYNAMIC_BASE.
>>>
>>> Ah right, that needs to be fixed.
>>>
>>>>
>>>> However, this change is just PoC I proposed, the conditional and
>>>> action should be slightly different to cover a corner case, when
>>>> statically allocated chip overlaps the DYNAMIC_BASE, i.e. gdev->base <
>>>> DYNAMIC_BASE, while gdev->base + gdev->ngpio >= DYNAMIC_BASE.
>>>>
>>>
>>> Yes you are right, that's gdev->base + gdev->ngpio that should be checked.
>>>
>> and that not with simple continue or base might simply stay at DYNAMIC_BASE.
>>
>> I will send a v2 of this patch with refined logic.
> 
> Actually it would be nice to integrate a warning (if we don't have it
> yet) when adding a GPIO chip with a static allocation and which will
> overlap the dynamic base. Can you add that into your v2?
> 

At the time being we have a warning for all static allocations, 
allthough their has been some discussion about reverting it, see commit 
502df79b8605 ("gpiolib: Warn on drivers still using static gpiobase 
allocation")

Christophe