Re: [RFD] gpiolib: gpiochip is always dangling after remove

[Linus Walleij <linus.walleij@xxxxxxxxxx>]

On Fri, Feb 26, 2016 at 8:06 PM, Bamvor Jian Zhang <bamv2005@xxxxxxxxx> wrote:

> From: Bamvor Jian Zhang <bamvor.zhangjian@xxxxxxxxxx>
>
> In my gpiochip mockup driver, after call gpiochip_remove, the
> gpipchip is not actually removed even if there is no clients in
> userspace. It could be removed if remove the corresonding reference
> count in this patch.
>
> But such gpiochip will be removed even if user space open the
> corresponding chardev. I am not sure If I do not right thing.
> Please correct me if I am wrong.
>
> Signed-off-by: Bamvor Jian Zhang <bamvor.zhangjian@xxxxxxxxxx>

I think you're onto the right thing, thank you for looking into this.

My idea was that the gdev->dev and gdev->cdev should stay around
until the last userspace user is gone, even if the gpiochip_remove()
was called.

Maybe I just got the refcounts wrong :(

It'd be great if you can come up with a patch that has the following
properties:

gpiochip_add_data(), no userspace users, gpiochip_remove()
-> refcount goes to zero and gpiodevice_release() gets called

gpiochip_add_data(), add some userspace users, gpiochip_remove()
-> refcount does not go to zero and gpiodevice_release() is not called
Then remove the userspace users:
-> refcount goes to zero and gpiodevice_release() gets called

It might be that using device_add(), device_del() directly like I'm
doing cannot really solve this :( then we need to rethink the mechanism
in terms of device_register()/device_unregister(), but I couldn't
get that to work with the way chardevs are added.

Yours,
Linus Walleij
--
To unsubscribe from this list: send the line "unsubscribe linux-gpio" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html