Re: [v3 07/12] x86/fsgsbase/64: Introduce the new FIND_PERCPU_BASE macro

[Andy Lutomirski]

On Tue, Oct 23, 2018 at 11:43 AM Chang S. Bae <chang.seok.bae@xxxxxxxxx> wrote:
>
> GSBASE is used to find per-CPU data in the kernel. But when it is unknown,
> the per-CPU base can be found from the per_cpu_offset table with a CPU NR.
> The CPU NR is extracted from the limit field of the CPUNODE entry in GDT,
> or by the RDPID instruction.
>
> Also, add the GAS-compatible RDPID macro.
>
> The new macro will be used on a following patch.
>
> Suggested-by: H. Peter Anvin <hpa@xxxxxxxxx>
> Signed-off-by: Chang S. Bae <chang.seok.bae@xxxxxxxxx>
> Cc: Andi Kleen <ak@xxxxxxxxxxxxxxx>
> Cc: Andy Lutomirski <luto@xxxxxxxxxx>
> Cc: Dave Hansen <dave.hansen@xxxxxxxxxxxxxxx>
> Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
> Cc: Ingo Molnar <mingo@xxxxxxxxxx>
> ---
>  arch/x86/include/asm/fsgsbase.h | 52 +++++++++++++++++++++++++++++++++
>  arch/x86/include/asm/inst.h     | 15 ++++++++++
>  2 files changed, 67 insertions(+)
>
> diff --git a/arch/x86/include/asm/fsgsbase.h b/arch/x86/include/asm/fsgsbase.h
> index e500d771155f..0c2d7d8a8c01 100644
> --- a/arch/x86/include/asm/fsgsbase.h
> +++ b/arch/x86/include/asm/fsgsbase.h
> @@ -111,6 +111,58 @@ extern void x86_gsbase_write_cpu_inactive(unsigned long gsbase);
>         MODRM 0xd0 wrgsbase_opd 1
>  .endm
>
> +#if CONFIG_SMP
> +
> +/*
> + * Fetch the per-CPU GSBASE value for this processor and put it in @reg.
> + * We normally use %gs for accessing per-CPU data, but we are setting up
> + * %gs here and obviously can not use %gs itself to access per-CPU data.
> + */
> +.macro FIND_PERCPU_BASE_RDPID reg:req
> +       /*
> +        * The CPU/node NR is initialized earlier, directly in cpu_init().
P
> +        */
> +       RDPID   \reg

I would suggest that you instead add a macro LOAD_CPU_AND_NODE \reg
and have that macro contain the alternative.  It can switch between
RDPID and LSL.  This way you avoid duplicating the rest of it.

This should end up in the same header as __getcpu() -- it probably
makes sense to just move __getcpu() for this purpose.

Also, hpa and Nadav, shouldn't asm/inst.h end up in macros.S?

--Andy