Re: [PATCH 4/5] gpiolib: simplify notifying user-space about line requests

[Bartosz Golaszewski]

On Sat, Oct 5, 2024 at 5:46 AM Kent Gibson <warthog618@xxxxxxxxx> wrote:
>
> On Fri, Oct 04, 2024 at 04:43:25PM +0200, Bartosz Golaszewski wrote:
> > From: Bartosz Golaszewski <bartosz.golaszewski@xxxxxxxxxx>
> >
> > Instead of emitting the line state change event on request in three
> > different places, just do it once, closer to the source: in
> > gpiod_request_commit().
> >
> > Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@xxxxxxxxxx>
> > ---
> >  drivers/gpio/gpiolib-cdev.c | 6 ------
> >  drivers/gpio/gpiolib.c      | 4 ++--
> >  2 files changed, 2 insertions(+), 8 deletions(-)
> >
> > diff --git a/drivers/gpio/gpiolib-cdev.c b/drivers/gpio/gpiolib-cdev.c
> > index b0050250ac3a..f614a981253d 100644
> > --- a/drivers/gpio/gpiolib-cdev.c
> > +++ b/drivers/gpio/gpiolib-cdev.c
> > @@ -372,8 +372,6 @@ static int linehandle_create(struct gpio_device *gdev, void __user *ip)
> >                               goto out_free_lh;
> >               }
> >
> > -             gpiod_line_state_notify(desc, GPIO_V2_LINE_CHANGED_REQUESTED);
> > -
> >               dev_dbg(&gdev->dev, "registered chardev handle for line %d\n",
> >                       offset);
>
> This moves the notify to before the desc->flags have been set.
> So the notified will now see the flags as previously set, not what they
> have been requested as.
>

Ah, I got fooled by the libgpiod tests passing. I guess we should
cover that first in tests-kernel-uapi.c.

> That might be acceptible if you subsequently issue GPIO_V2_LINE_CHANGED_CONFIG
> when the flags are set, but that is not done here and you explicitly don't
> notify from here in patch 5 when you add notifying to gpiod_direction_output()
> etc.
>

IMO it doesn't make sense to always emit REQUESTED and CONFIG_CHANGED
events together. The initial config should be part of the request
event. I'll get back to the drawing board.

Bart