[PATCH 2/2] gpio: spacemit: Add GPIO support for K3 SoC
Bartosz Golaszewski
brgl at kernel.org
Fri Jan 2 05:38:12 PST 2026
On Fri, 2 Jan 2026 13:20:45 +0100, Yixun Lan <dlan at gentoo.org> said:
> Hi bart,
>
> On 19:36 Fri 02 Jan , Yixun Lan wrote:
>> Hi Bart,
>>
>> On 12:10 Fri 02 Jan , Bartosz Golaszewski wrote:
>> > On Mon, Dec 29, 2025 at 1:47 PM Yixun Lan <dlan at gentoo.org> wrote:
>> > >
>> > > SpacemiT K3 SoC has changed gpio register layout while comparing
>> > > with previous generation, the register offset and bank offset
>> > > need to be adjusted, introduce a compatible data to extend the
>> > > driver to support this.
>> > >
>> > > Signed-off-by: Yixun Lan <dlan at gentoo.org>
>> > > ---
>> > > drivers/gpio/gpio-spacemit-k1.c | 150 ++++++++++++++++++++++++++++------------
>> > > 1 file changed, 106 insertions(+), 44 deletions(-)
>> > >
>> > > diff --git a/drivers/gpio/gpio-spacemit-k1.c b/drivers/gpio/gpio-spacemit-k1.c
>> > > index eb66a15c002f..02cc5c11b617 100644
>> > > --- a/drivers/gpio/gpio-spacemit-k1.c
>> > > +++ b/drivers/gpio/gpio-spacemit-k1.c
>> > > @@ -15,28 +15,19 @@
>> > > #include <linux/platform_device.h>
>> > > #include <linux/seq_file.h>
>> > >
> [snip]...
>> > > static u32 spacemit_gpio_bank_index(struct spacemit_gpio_bank *gb)
>> > > {
>> > > return (u32)(gb - gb->sg->sgb);
>> > > @@ -60,13 +70,14 @@ static u32 spacemit_gpio_bank_index(struct spacemit_gpio_bank *gb)
>> > > static irqreturn_t spacemit_gpio_irq_handler(int irq, void *dev_id)
>> > > {
>> > > struct spacemit_gpio_bank *gb = dev_id;
>> > > + struct spacemit_gpio *sg = gb->sg;
>> > > unsigned long pending;
>> > > u32 n, gedr;
>> > >
>> > > - gedr = readl(gb->base + SPACEMIT_GEDR);
>> > > + gedr = readl(gb->base + to_spacemit_gpio_regs(sg)->gedr);
>> >
>> > Since you're already touching all these register accesses - can you
>> > maybe provide dedicated wrapper functions around readl()/writel() and
>> > avoid any file-wide changes in the future if anything requires further
>> > modification?
>> >
>> can you elaborate a bit further on this?
>> I don't get how a wrapper helper could help to avoid file-wide changes..
>>
> here is my attempt to solve this, define a macro to register address:
>
> #define to_spacemit_gpio_regs(gb) ((gb)->sg->data->reg_offsets)
>
> #define SPACEMIT_GEDR(gb) ((gb)->base + to_spacemit_gpio_regs(gb)->gedr)
>
> gedr = readl(SPACEMIT_GEDR(gb));
>
> please let me know if this follow your suggestion or not
>
> --
> Yixun Lan (dlan)
>
I was thinking more of something like this:
enum spacemit_gpio_registers {
SPACEMIT_GPLR,
SPACEMIT_GPDR,
...
};
static const unsigned int spacemit_gpio_k1_offsets = {
[SPACEMIT_GPLR] = 0x00,
[SPACEMIT_GPDR] = 0x0c,
...
};
static const unsigned int spacemit_gpio_k3_offsets = ...
struct spacemit_gpio_data {
const unsigned int *offsets;
u32 bank_offsets[4];
};
static void spacemit_gpio_write(struct spacemit_gpio_bank *gb,
enum spacemit_gpio_registers reg, u32 val)
{
writel(val, gb->base + gb->data->offsets[reg]);
}
Bart
More information about the linux-riscv
mailing list