Re: [PATCH v2 04/11] gpiolib: Clear the gpio_device's fwnode initialized flag before adding

[Sudeep Holla]

On Mon, Jan 30, 2023 at 08:01:17PM -0800, Saravana Kannan wrote:
> On Mon, Jan 30, 2023 at 7:14 AM Andy Shevchenko
> <andriy.shevchenko@xxxxxxxxxxxxxxx> wrote:
> >
> > On Mon, Jan 30, 2023 at 02:31:53PM +0000, Sudeep Holla wrote:
> > > On Thu, Jan 26, 2023 at 04:11:31PM -0800, Saravana Kannan wrote:
> > > > Registering an irqdomain sets the flag for the fwnode. But having the
> > > > flag set when a device is added is interpreted by fw_devlink to mean the
> > > > device has already been initialized and will never probe. This prevents
> > > > fw_devlink from creating device links with the gpio_device as a
> > > > supplier. So, clear the flag before adding the device.
> >
> > ...
> >
> > > > +   /*
> > > > +    * If fwnode doesn't belong to another device, it's safe to clear its
> > > > +    * initialized flag.
> > > > +    */
> > > > +   if (!gdev->dev.fwnode->dev)
> > > > +           fwnode_dev_initialized(gdev->dev.fwnode, false);
> > >
> > > This is the one causing the kernel crash during the boot on FVP which
> > > Naresh has reported. Just reverted this and was able to boot, confirming
> > > the issue with this patch.
> >
> > I'm wondering if
> >
> >         if (!dev_fwnode(&gdev->dev)->dev)
> >                 fwnode_dev_initialized(&dev_fwnode(gdev->dev), false);
> >
> > works.
> 
> No, that won't help. The problem was that with arm32, we have gpio
> devices created without any of_node or fwnode. So I can't assume
> fwnode will always be present.
>

Correct, and this one is not even arm32. But it is just reusing a driver
that needs to be supported even on arm32.

Not sure on how to proceed. As a simple way to check, I added a NULL check
for fwnode building on top of Andy's suggestion[1]. That works.

Also the driver in question on arm64 FVP model is drivers/mfd/vexpress-sysreg.c
mfd_add_device() in drivers/mfd/mfd-core.c allows addition of devices without
of_node/fwnode. I am sure returning error like[2] will break many platforms
but I just wanted to confirm the root cause and [2] fixes the boot without
NULL check for fwnode in gpiochip_setup_dev().

Hope this helps.

--
Regards,
Sudeep

[1]

-->8
diff --git i/drivers/gpio/gpiolib.c w/drivers/gpio/gpiolib.c
index b23140c6485f..e162f13aa2c9 100644
--- i/drivers/gpio/gpiolib.c
+++ w/drivers/gpio/gpiolib.c
@@ -577,13 +577,15 @@ static void gpiodevice_release(struct device *dev)
 static int gpiochip_setup_dev(struct gpio_device *gdev)
 {
        int ret;
+       struct fwnode_handle *fwnode = dev_fwnode(&gdev->dev);

        /*
         * If fwnode doesn't belong to another device, it's safe to clear its
         * initialized flag.
         */
-       if (!gdev->dev.fwnode->dev)
-               fwnode_dev_initialized(gdev->dev.fwnode, false);
+       if (fwnode && !fwnode->dev)
+               fwnode_dev_initialized(fwnode, false);
+
        ret = gcdev_register(gdev, gpio_devt);
        if (ret)
                return ret;

[2]

-->8

diff --git i/drivers/mfd/mfd-core.c w/drivers/mfd/mfd-core.c
index 16d1861e9682..3b2c4b0e9a2a 100644
--- i/drivers/mfd/mfd-core.c
+++ w/drivers/mfd/mfd-core.c
@@ -231,9 +231,11 @@ static int mfd_add_device(struct device *parent, int id,
                        }
                }

-               if (!pdev->dev.of_node)
+               if (!pdev->dev.of_node) {
                        pr_warn("%s: Failed to locate of_node [id: %d]\n",
                                cell->name, platform_id);
+                       goto fail_alias;
+               }
        }

        mfd_acpi_add_device(cell, pdev);