On Wed, Jun 19, 2013 at 02:20:03PM +0100, Marc Zyngier wrote:
> When performing a Stage-2 TLB invalidation, it is necessary to
> make sure the write to the page tables is observable by all CPUs.
>
> For this purpose, add a dsb instruction to __kvm_tlb_flush_vmid_ipa
> before doing the TLB invalidation itself.
>
> Signed-off-by: Marc Zyngier <marc.zyngier@xxxxxxx>
> ---
> arch/arm/kvm/interrupts.S | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/arch/arm/kvm/interrupts.S b/arch/arm/kvm/interrupts.S
> index d0a8fa3..afa6c04 100644
> --- a/arch/arm/kvm/interrupts.S
> +++ b/arch/arm/kvm/interrupts.S
> @@ -49,6 +49,7 @@ __kvm_hyp_code_start:
> ENTRY(__kvm_tlb_flush_vmid_ipa)
> push {r2, r3}
>
> + dsb
shouldn't this be a dmb then?
thinking that a dsb only ensures completion between load/stores and not
between the preceeding store and the actual invalidate operation?
or are we relying on the fact that the store must complete before the
ldrd below, which must happen before the mcrr, which must happen before
the invalidate (?), and therefore it's all good?
> add r0, r0, #KVM_VTTBR
> ldrd r2, r3, [r0]
> mcrr p15, 6, r2, r3, c2 @ Write VTTBR
> --
> 1.8.2.3
>
>
>
> _______________________________________________
> kvmarm mailing list
> kvmarm@xxxxxxxxxxxxxxxxxxxxx
> https://lists.cs.columbia.edu/cucslists/listinfo/kvmarm
_______________________________________________
kvmarm mailing list
kvmarm@xxxxxxxxxxxxxxxxxxxxx
https://lists.cs.columbia.edu/cucslists/listinfo/kvmarm
