On Sat, Dec 05, 2020 at 11:21:09PM +0100, Linus Walleij wrote:
> On Tue, Dec 1, 2020 at 4:50 PM Johan Hovold <johan@xxxxxxxxxx> wrote:
> > On Tue, Dec 01, 2020 at 03:37:38PM +0100, Linus Walleij wrote:
> > > On Sun, Nov 22, 2020 at 6:08 PM Manivannan Sadhasivam <mani@xxxxxxxxxx> wrote:
>
> > > You know the names of the pins...
> > >
> > > > + port_priv->gc.ngpio = 6;
> > > > + port_priv->gc.label = "xr_gpios";
> > > > + port_priv->gc.request = xr_gpio_request;
> > > > + port_priv->gc.free = xr_gpio_free;
> > > > + port_priv->gc.get_direction = xr_gpio_direction_get;
> > > > + port_priv->gc.direction_input = xr_gpio_direction_input;
> > > > + port_priv->gc.direction_output = xr_gpio_direction_output;
> > > > + port_priv->gc.get = xr_gpio_get;
> > > > + port_priv->gc.set = xr_gpio_set;
> > > > + port_priv->gc.owner = THIS_MODULE;
> > > > + port_priv->gc.parent = &port->dev;
> > > > + port_priv->gc.base = -1;
> > > > + port_priv->gc.can_sleep = true;
> > >
> > > So assign port_priv->gc.names here as well with an array
> > > of strings with the names ("RI", "CD", ... etc).
> > > This makes it look really nice in userspace if you do
> > > e.g. "lsgpio".
> >
> > Last time we tried that gpiolib still used a flat namespace so that you
> > can't have have more than one device using the same names. Unless that
> > has changed this is a no-go. See
> >
> > https://lore.kernel.org/r/20180930122703.7115-1-johan@xxxxxxxxxx
> >
> > for our previous discussion about this.
>
> Hm hm yeah we actually put in a nasty warning there since:
>
> gpio = gpio_name_to_desc(gc->names[i]);
> if (gpio)
> dev_warn(&gdev->dev,
> "Detected name collision for GPIO name '%s'\n",
> gc->names[i]);
>
>
> A better approach might be to create an array of names
> prepended with something device-unique like the USB
> bus topology? Or do we need a helper to help naming the
> GPIOs? What would be helpful here?
>
> name = kasprintf(GFP_KERNEL, "%s-NAME", topology_str);
Well we started discussing this back when we only had the sysfs
interface which suffered from the same problem. I thought the chardev
interface was supposed to get rid of the assumption of a flat name
space? Perhaps in v3 of the ABI. ;P
If this is too built into the new chardev interface as well to be fixed
up, a unique prefix is the only way to go. Perhaps gpiolib can just
prefix it with the controller name?
gpiochip508-CBUS0
Based on a hotpluggable bus flag? But what about any other non-pluggable
IC, which provides a few named GPIO lines and of which there could be
more than one in a system?
The topology is already encoded in sysfs and it seems backwards to have
each and every gpio driver reconstruct it.
Johan
