Re: [PATCH v2 1/2] gpio: loongson: add dts/acpi gpio support

[Bartosz Golaszewski <brgl@xxxxxxxx>]

On Tue, Nov 15, 2022 at 2:07 PM Yinbo Zhu <zhuyinbo@xxxxxxxxxxx> wrote:
>
>
>
> 在 2022/11/15 下午6:17, WANG Xuerui 写道:
> > Sorry for jumping in, but...
> >
> > On 2022/11/15 17:53, Yinbo Zhu wrote:
> >>
> >>
> >> 在 2022/11/15 下午5:05, Bartosz Golaszewski 写道:
> >>> On Mon, Nov 14, 2022 at 10:53 AM Yinbo Zhu <zhuyinbo@xxxxxxxxxxx> wrote:
> >>>>
> >>>> The latest Loongson series platform use dts or acpi framework to
> >>>> register gpio device resources, such as the Loongson-2 series
> >>>> SoC of LOONGARCH architecture. In order to support dts, acpi and
> >>>> compatibility with previous platform device resources in driver,
> >>>> this patch was added.
> >>>>
> >>>> Signed-off-by: lvjianmin <lvjianmin@xxxxxxxxxxx>
> >>>> Signed-off-by: zhanghongchen <zhanghongchen@xxxxxxxxxxx>
> >>>> Signed-off-by: Liu Peibao <liupeibao@xxxxxxxxxxx>
> >>>> Signed-off-by: Juxin Gao <gaojuxin@xxxxxxxxxxx>
> >>>> Signed-off-by: Yinbo Zhu <zhuyinbo@xxxxxxxxxxx>
> >>>> ---
> >>>> Change in v2:
> >>>>                  1. Fixup of_loongson_gpio_get_props and remove the
> >>>> parse logic about
> >>>>                     "loongson,conf_offset", "loongson,out_offset",
> >>>> "loongson,in_offset",
> >>>>                     "loongson,gpio_base", "loongson,support_irq"
> >>>> then kernel driver will
> >>>>                     initial them that depend compatible except
> >>>> "loongson,gpio_base".
> >>>>
> >>>>   arch/loongarch/include/asm/loongson.h         |  13 +
> >>>>   .../include/asm/mach-loongson2ef/loongson.h   |  12 +
> >>>>   .../include/asm/mach-loongson64/loongson.h    |  13 +
> >>>>   drivers/gpio/Kconfig                          |   6 +-
> >>>>   drivers/gpio/gpio-loongson.c                  | 422
> >>>> +++++++++++++++---
> >>>>   5 files changed, 391 insertions(+), 75 deletions(-)
> >>>>
> >>>> diff --git a/arch/loongarch/include/asm/loongson.h
> >>>> b/arch/loongarch/include/asm/loongson.h
> >>>> index 00db93edae1b..383fdda155f0 100644
> >>>> --- a/arch/loongarch/include/asm/loongson.h
> >>>> +++ b/arch/loongarch/include/asm/loongson.h
> >>>> @@ -60,6 +60,19 @@ static inline void xconf_writeq(u64 val64,
> >>>> volatile void __iomem *addr)
> >>>>          );
> >>>>   }
> >>>>
> >>>> +/* ============== Data structrues =============== */
> >>>> +
> >>>> +/* gpio data */
> >>>> +struct platform_gpio_data {
> >>>> +       u32 gpio_conf;
> >>>> +       u32 gpio_out;
> >>>> +       u32 gpio_in;
> >>>> +       u32 support_irq;
> >>>> +       char *label;
> >>>> +       int gpio_base;
> >>>> +       int ngpio;
> >>>> +};
> >>>
> >>> This is a terrible name for an exported structure. You would at least
> >>> need some kind of a namespace prefix. But even then the need to add a
> >>> platform data structure is very questionable. We've moved past the
> >>> need for platform data in the kernel. I don't see anyone setting it up
> >>> in this series either. Could you provide more explanation on why you
> >>> would need it and who would use it?
> >> okay, I will add a namespace prefix, about this platform data was added
> >> that was to compatible with legacy platforms that do not support dts or
> >> acpi, then, the mips loongson platform or loongarch loongson platform
> >
> > Why are you trying to support "legacy" LoongArch platforms when the
> > architecture was just upstreamed *this year*?
> the leagacy gpio driver had support LoongArch, and you can find some
> gpio register defined in arch/loongarch/include
> /asm/loongson.h in legacy gpio driver, such as LOONGSON_GPIODATA,
> The legacy gpio driver is the driver that doesn't include my gpio patch.
> >
> >> can implement the gpio device driver to initialize the
> >> platform_gpio_data structure as needed after exporting the structure.
> >>>
> >>>> +
> >>>>   /* ============== LS7A registers =============== */
> >>>>   #define LS7A_PCH_REG_BASE              0x10000000UL
> >>>>   /* LPC regs */
> >>>> diff --git a/arch/mips/include/asm/mach-loongson2ef/loongson.h
> >>>> b/arch/mips/include/asm/mach-loongson2ef/loongson.h
> >>>> index ca039b8dcde3..b261cea4fee1 100644
> >>>> --- a/arch/mips/include/asm/mach-loongson2ef/loongson.h
> >>>> +++ b/arch/mips/include/asm/mach-loongson2ef/loongson.h
> >>>> @@ -315,4 +315,16 @@ extern unsigned long _loongson_addrwincfg_base;
> >>>>
> >>>>   #endif /* ! CONFIG_CPU_SUPPORTS_ADDRWINCFG */
> >>>>
> >>>> +/* ============== Data structrues =============== */
> >>>> +
> >>>> +/* gpio data */
> >>>> +struct platform_gpio_data {
> >>>> +       u32 gpio_conf;
> >>>> +       u32 gpio_out;
> >>>> +       u32 gpio_in;
> >>>> +       u32 support_irq;
> >>>> +       char *label;
> >>>> +       int gpio_base;
> >>>> +       int ngpio;
> >>>> +};
> >>>
> >>> No idea why you would need to duplicate it like this either. And why
> >>> put it in arch/.
> >> because loongson platform include mips and loongarch, and the gpio
> >> device data was defined in arch/ in leagcy loongson gpio driver.  so the
> >> latest loongson gpio drvier add platform_gpio_data in same dir.
> >
> > I think at this point it's hopefully clear, that the way forward to
> > supporting Loongson IP blocks shared between MIPS/LoongArch SoCs is to
> > start over and do things properly, making the code as platform-agnostic
> > as possible. Just make sure the drivers can get initialized via DT and
> > ACPI then you're all set -- the upstream kernel is guaranteed to use one
> > of the two well-established boot flows for every Loongson chip it
> > supports. Be it hard-coded DT in arch/mips/boot/dts/loongson, or the
> > LoongArch ACPI/upcoming DT, no need for hard-coding things in arch/ in
> > either case.
> Our old platforms are used by customers, but we will not maintain those
> platforms. Adding dts/acpi support to those old platforms not only makes
> no sense, but also affects their use. Because the configuration of
> dts/acpi requires the support of the firmware team, but in fact, we have
> no one to maintain those old platforms.
>
> in a words, My patch to upstream was supposed to consider dts/acpi in
> LoongArch platform  But I have to consider the original legacy gpio
> driver and to compatible with it.

Which legacy driver are you referring to? Neither gpio-loongson nor
gpio-loongson1 define any platform data. If it wasn't needed before
then it's sure we won't be adding it in 2022. If you have board-files
upstream that need to use this driver then you can do fine with
regular device properties.

Bartosz