Re: [PATCH 5/7] gpio: ppc44x: Convert GPIO to generic MMIO

Le 29/07/2026 à 20:32, Rosen Penev a écrit :

On Wed, Jul 29, 2026 at 3:55 AM Madhavan Srinivasan [off-list ref] wrote:

quoted

On 6/2/26 2:56 PM, Rosen Penev wrote:

quoted

On Tue Jun 2, 2026 at 12:51 AM PDT, Bartosz Golaszewski wrote:

quoted

On Tue, 2 Jun 2026 07:01:29 +0200, Rosen Penev [off-list ref] said:

quoted

Use gpio_generic_chip_init() to set up the PPC44x GPIO chip
instead of open-coding the basic get, set, locking and state handling.

Keep the PPC44x-specific direction callbacks because they still need to
program ODR and the OSR/TSR registers around the generic data and
direction registers.

Assisted-by: Codex:GPT-5.5
Signed-off-by: Rosen Penev <redacted>
---

...

quoted

@@ -124,10 +102,11 @@ static int
   ppc4xx_gpio_dir_out(struct gpio_chip *gc, unsigned int gpio, int val)
   {
     struct ppc4xx_gpio_chip *chip = gpiochip_get_data(gc);
+   struct gpio_generic_chip *gen_gc = &chip->chip;
     struct ppc4xx_gpio __iomem *regs = chip->regs;
     unsigned long flags;

-   spin_lock_irqsave(&chip->lock, flags);
+   gpio_generic_chip_lock_irqsave(gen_gc, flags);

If you're already doing it, can you use lock guards too?

Sure. btw, I avoided placing

Do you have plan to address this comment in v2

Yeah I have it fixed locally.

You should probably put patch https://patchwork.ozlabs.org/project/linuxppc-dev/patch/20260517063754.21819-1-rosenp@gmail.com/ as first patch in this series in order to enable applying the fix to 6.18 without applying the entire series and to avoid any future conflict.

Christophe

I'm not sure if this has to do through the PPC or GPIO tree.

quoted

Maddy

quoted

https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Flore.kernel.org%2Fall%2F20260517063754.21819-1-rosenp%40gmail.com%2F&data=05%7C02%7Cchristophe.leroy%40csgroup.eu%7C2ac0ec525c7045cd969508deed9fc7e0%7C8b87af7d86474dc78df45f69a2011bb5%7C0%7C0%7C639209467691381845%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=iqGebSy%2BbKMBFZz%2FZwZ9N2Jf8v63FkGPNhrdUf864fE%3D&reserved=0

in the beginning of the series. My thinking is that's for older kernels.
I believe either the generic API or devm_gpiochip_add_data fixes this.

quoted

Bart