Re: [Minios-devel] [UNIKRAFT early RFC PATCH 07/11] plat/kvm: arm64: Add entry point for APs (secondary cpus)

[Julien Grall <julien.grall@xxxxxxx>]

Hi,

On 21/06/2019 07:57, Jia He wrote:
> Signed-off-by: Jia He <justin.he@xxxxxxx>
> ---
>   plat/kvm/arm/entry64.S | 30 ++++++++++++++++++++++++++++++
>   1 file changed, 30 insertions(+)
>
> diff --git a/plat/kvm/arm/entry64.S b/plat/kvm/arm/entry64.S
> index df59ef6..9c98102 100644
> --- a/plat/kvm/arm/entry64.S
> +++ b/plat/kvm/arm/entry64.S
> @@ -143,3 +143,33 @@ ENTRY(_libkvmplat_newstack)
>         /* Branch to _libkvmplat_entry2 */
>         br x1
>   END(_libkvmplat_newstack)
> +
> +/*
> + * mpentry(unsigned long)
> + *
> + * Called by a core when it is being brought online.
> + * The data in x0 is passed straight to init_secondary.
> + */
> +ENTRY(mpentry)
> +       /* Disable interrupts */
> +       msr     daifset, #2
Technically you should mask all the interrupts (debug, SError, Interrupt, FIQ) 
until you have configured the vector table correctly.
But I am pretty sure you would arrive here with all interrupts masked.

> +
> +       /* Set the context id */
> +       msr     contextidr_el1, xzr
> +
> +       /* Enable the mmu */
> +       mov x7, #1
See my comment about x7 before.

> +       bl      start_mmu
> +
> +       /* Jump to the virtual address space */
> +       ldr     x15, =mp_virtdone
> +       br      x15
I thought Unikraft page tables were configured with virt == phys mapping. So do 
you really need this?
> +
> +mp_virtdone:
> +       ldr     x4, =secondary_stacks
> +       mov     x5, #(4096 * 4)//TODO
What's that for?

> +       mul     x5, x0, x5
> +       add     sp, x4, x5
> +
> +       b       init_secondary
> +END(mpentry)
Cheers,

--
Julien Grall

_______________________________________________
Minios-devel mailing list
Minios-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/minios-devel