Re: [6.9 gpiolib regression] gpiolib: triggers: kobject: 'gpiochipX' is not, initialized, yet kobject_get() errors

[Bartosz Golaszewski <brgl@xxxxxxxx>]

On Fri, 29 Mar 2024 15:11:21 +0100, Hans de Goede <hans@xxxxxxxxx> said:
> Hi All,
>
> I've already tried to fix this, so let me just copy and paste my half finished patch
> to explain the problem.
>
> I was planning on submitting this as a RFC patch at least, but there are also some
> other new issues with 6.9 on this tablet and I'm not sure how this interacts
> with those issues and I don't have time to work on this any further this weekend.
>
> Anyways below is the patch / bug report.
>
> I'm wondering if a better fix would be to add a "ready" flag to gdev
> and may gpiochip_find ignore not yet ready chips (we need them on
> the list before they are ready to reserve the GPIO numbers) ?
>
> Regards,
>
> Hans
>

Hi Hans!

Thanks for the report. I hope I'm not being naive here but would the following
one-liner work?

diff --git a/drivers/gpio/gpiolib.c b/drivers/gpio/gpiolib.c
index ce94e37bcbee..69f365ccbfd8 100644
--- a/drivers/gpio/gpiolib.c
+++ b/drivers/gpio/gpiolib.c
@@ -1179,7 +1179,7 @@ struct gpio_device *gpio_device_find(const void *data,

 		gc = srcu_dereference(gdev->chip, &gdev->srcu);

-		if (gc && match(gc, data))
+		if (device_is_registered(&gdev->dev) && gc && match(gc, data))
 			return gpio_device_get(gdev);
 	}

This would make gpio_device_find() ignore any GPIO device that's not yet
registered on the GPIO bus which is almost the last step of the registration
process right before creating the sysfs attributes.

Bartosz