Re: [PATCH v2 05/13] hw/riscv/fdt_common, virt.c: add create_fdt_imsic()

On 26/8/26 18:48, Daniel Henrique Barboza wrote:
> The FDT related to interrupt controllers should definitely be put into
> helpers, even if only 'virt' benefits from it, due to the amount of code
> that is relieved from the board alone.  We'll start with 'imsic'.
> 
> For this endeavor we're going to use a helper struct that will carry all
> the extra FDT arguments in a human format.  The alternative would be to
> add 8 more function arguments, on top of 4 phandle arguments that we're
> passing around in all these controllers, and that's not only a poor
> reading experience but it's quite error prone too.
> 
> We do not want to crowd the fdt-helper with non-FDT logic, hence we're
> handling the !kvm_enabled() cond in virt.c using the imsic_m_base
> argument.  This is a pattern that we'll fall back on in the next
> patches too.
> 
> No FDT changes made.
> 
> Signed-off-by: Daniel Henrique Barboza <[email protected]>
> ---
>   hw/riscv/fdt-common.c         |  94 ++++++++++++++++++++++++++++++
>   hw/riscv/virt.c               | 104 ++++------------------------------
>   include/hw/riscv/fdt-common.h |  22 +++++++
>   3 files changed, 128 insertions(+), 92 deletions(-)


> @@ -583,8 +492,19 @@ static void create_fdt_sockets(RISCVVirtState *s,
>       }
>   
>       if (s->aia_type == VIRT_AIA_TYPE_APLIC_IMSIC) {
> -        create_fdt_imsic(s, phandle, intc_phandles,
> +        IMSICFdtProps props = {
> +            .soc = &s->soc,

New line please.

> .socket_count = riscv_socket_count(ms),
> +            .smp_cpus = ms->smp.cpus,
> +            .imsic_m_base = !kvm_enabled() ? s->memmap[VIRT_IMSIC_M].base : 0,
> +            .imsic_s_base = s->memmap[VIRT_IMSIC_S].base,
> +            .imsic_group_max_size = VIRT_IMSIC_GROUP_MAX_SIZE,
> +            .irqchip_num_msis = VIRT_IRQCHIP_NUM_MSIS,
> +            .aia_guests = s->aia_guests
> +        };
> +
> +        create_fdt_imsic(ms->fdt, &props, phandle, intc_phandles,
>                            &msi_m_phandle, &msi_s_phandle);
> +
>           *msi_pcie_phandle = msi_s_phandle;
>       }