Re: [PATCH 1/2] RISC-V: Cache SBI vendor values

From: Andrew Jones
Date: Wed Oct 05 2022 - 13:07:10 EST


On Tue, Oct 04, 2022 at 10:37:23PM +0200, Heiko Stuebner wrote:
> sbi_get_mvendorid(), sbi_get_marchid() and sbi_get_mimpid() might get
> called multiple times, though the values of these CSRs should not change
> during the runtime of a specific machine.
>
> So cache the values in the functions and prevent multiple ecalls
> to read these values.
>
> Suggested-by: Atish Patra <atishp@xxxxxxxxxxxxxx>
> Signed-off-by: Heiko Stuebner <heiko@xxxxxxxxx>
> ---
> arch/riscv/kernel/sbi.c | 21 ++++++++++++++++++---
> 1 file changed, 18 insertions(+), 3 deletions(-)
>
> diff --git a/arch/riscv/kernel/sbi.c b/arch/riscv/kernel/sbi.c
> index 775d3322b422..5be8f90f325e 100644
> --- a/arch/riscv/kernel/sbi.c
> +++ b/arch/riscv/kernel/sbi.c
> @@ -625,17 +625,32 @@ static inline long sbi_get_firmware_version(void)
>
> long sbi_get_mvendorid(void)
> {
> - return __sbi_base_ecall(SBI_EXT_BASE_GET_MVENDORID);
> + static long id = -1;
> +
> + if (id < 0)
> + id = __sbi_base_ecall(SBI_EXT_BASE_GET_MVENDORID);
> +
> + return id;
> }
>
> long sbi_get_marchid(void)
> {
> - return __sbi_base_ecall(SBI_EXT_BASE_GET_MARCHID);
> + static long id = -1;
> +
> + if (id < 0)
> + id = __sbi_base_ecall(SBI_EXT_BASE_GET_MARCHID);

The marchid register will be negative for commercial architecture ids
because the MSB must be set.

> +
> + return id;
> }
>
> long sbi_get_mimpid(void)
> {
> - return __sbi_base_ecall(SBI_EXT_BASE_GET_MIMPID);
> + static long id = -1;
> +
> + if (id < 0)
> + id = __sbi_base_ecall(SBI_EXT_BASE_GET_MIMPID);

The spec says this register is "left to the provider" and may be
left-justified. I don't think we can be sure the MSB will not be set.

For both cases I guess we need an extra bit to determine if we've cached
or not

static bool cached;
static long id;

if (!cached) {
id = ecall();
cached = true;
}

return id;

> +
> + return id;
> }
>
> static void sbi_send_cpumask_ipi(const struct cpumask *target)
> --
> 2.35.1
>

Thanks,
drew