Re: [External] Re: [PATCH] firmware: added a firmware information passing method FFI

[运辉崔]

Hi Ard,

1. Regarding the definition of DTS FFI nodes, according to your
suggestion, we plan to make the following modifications：

/ {
...
    ffi_cfg {
        acpi_tbl {
            root_pinter = ; //u64
            ...
        };
        smbios_tbl {
            root_pinter = ; //u64
            ...
        };
    };
    ...
};

2. Let's move on to the discussion: should we put code under arch/risc-v/?

On Mon, Jun 26, 2023 at 6:19 PM 运辉崔 <cuiyunhui@xxxxxxxxxxxxx> wrote:
>
> Hi Ard, Mark,
>
> On Mon, Jun 26, 2023 at 4:23 PM Ard Biesheuvel <ardb@xxxxxxxxxx> wrote:
>
> > DT support for SMBIOS can live in generic code, but the binding has to
> > be sane. As I suggested before, it probably makes sense to supplant
> > the entrypoint rather than just carry its address - this means a 'reg'
> > property with base and size to describe the physical region, and at
> > least major/minor/docrev fields to describe the version.
>
> Regarding dts node binding, our current definition is as follows:
> /dts
> {
> ...
> cfgtables {
> acpi_phy_ptr = 0000000000000000; //u64
> smbios_phy_ptr = 0000000000000000; //u64
> ...
> }
> ...
> }
>
> x86 only gave a root_pointer entry address
> u64 x86_default_get_root_pointer(void)
> {
>        return boot_params.acpi_rsdp_addr;
> }
>
> Regarding the naming of the binding above, Mark,  do you have any suggestions?
>
>
> > For the ACPI side, you should just implement
> > acpi_arch_get_root_pointer() under arch/riscv, and wire it up in
> > whichever way you want. But please check with the RISC-V maintainers
> > if they are up for this, and whether they want to see this mechanism
> > contributed to one of the pertinent specifications.
>
> You suggest putting SMBIOS in general code instead of ACPI, why?
> From the perspective of firmware information passing, they are a class.
>
> SMBIOS and ACPI are not related to ARCH, nor is DTS to obtain firmware
> information,
>
> Why do you have to put part of the ACPI code under arch/risc-v/?
> The scope of the previous discussion was limited to RISC-V because of
> historical reasons such as the binding with EFI on ARM64. We will only
> enable this function on RISC-V in subsequent patches.
>
> The realization of the FFI scheme itself is irrelevant to the arch.
>
> Thanks，
> Yunhui