[PATCH v2] gpiolib: handle gpio-hogs only once

[Bartosz Golaszewski]

On Tue, 9 Jun 2026 22:48:03 +0200, Daniel Drake <dan at reactivated.net> said:
> On 09/06/2026 13:39, Bartosz Golaszewski wrote:
>> On Mon, 8 Jun 2026 23:01:08 +0200, Daniel Drake <dan at reactivated.net> said:
>>> diff --git a/drivers/gpio/gpiolib.c b/drivers/gpio/gpiolib.c
>>> index 1e6dce430dca..b02d711289d0 100644
>>> --- a/drivers/gpio/gpiolib.c
>>> +++ b/drivers/gpio/gpiolib.c
>>> @@ -1031,9 +1031,17 @@ static int gpiochip_hog_lines(struct gpio_chip *gc)
>>>   		if (!fwnode_property_present(fwnode, "gpio-hog"))
>>>   			continue;
>>>
>>> +		/* The hog may have been handled by another gpio_chip on the same fwnode */
>>> +		if (is_of_node(fwnode) &&
>>> +		    of_node_check_flag(to_of_node(fwnode), OF_POPULATED))
>>> +			continue;
>>> +
>>>   		ret = gpiochip_add_hog(gc, fwnode);
>>>   		if (ret)
>>>   			return ret;
>>> +
>>> +		if (is_of_node(fwnode))
>>> +			of_node_set_flag(to_of_node(fwnode), OF_POPULATED);
>>
>> Sashiko correctly points out that on errors, the state will be corrupted. We
>> could maybe move the clearing of the flag to gpiochip_free_hogs() and track its
>> state when processing fwnodes in order not to clear it incorrectly?
>
> I guess you are referring to:
>
>> Does setting OF_POPULATED here cause state corruption if a secondary chip on a
>> shared node fails to probe?
>> When multiple gpio_chip instances share a device node, the first chip processes
>> its hogs and sets OF_POPULATED. If a subsequent chip fails probe (for example,
>> returning -EPROBE_DEFER), its cleanup path calls of_gpiochip_remove() which
>> clears the flag for all hogs.
>> If the flag is unconditionally cleared, will the deferred chip attempt to
>> process the first chip's hogs on retry, fail due to a mismatch, and
>> permanently abort probe?
>
>
> I don't think this is actually an issue. If we have two gpio_chips
> sharing a device node, a first one with a hog that probes fine and a
> subsequent one that fails during probe, both of the gpio_chips will
> brought down and the flag is cleared. If it was a EPROBE_DEFER case
> which is then retried later, the first chip's hogs will be set up a 2nd
> time when the probe is retried.
>
> It is true that the teardown of the 2nd gpio_chip would erase the
> OF_POPULATED flag of a gpio-hog node that it does not "own", but the
> first gpio_chip would also be torn down at the same time (and
> OF_POPULATED unset a 2nd time). This is not ideal, but harmless as far
> as I can see.
>
> I don't quite follow the suggestion for doing the clearing better in
> gpiochip_free_hogs(). It would be neat if we could go from a hogged
> gpio_desc back to the fwnode, so that we could only unset OF_POPULATED
> on the fwnode at the point when we are really removing the hog, but I
> don't see a way to derive the gpio-hog fwnode from gpio_desc. Also, this
> would be complicated because one gpio-hog node can hog multiple gpios.
>
> Let me know if I'm missing anything or if you have any preference to
> handle this differently!
>
> Thanks
> Daniel
>
>

Ok, thanks for the clarification. Let's give this a chance then.

Bart