On Wed, Jul 14, 2021 at 10:56:01AM +0100, Alexandru Elisei wrote:
> According to ARM DDI 0487G.a, page D9-2930, profiling is disabled when
> the PE is executing at a higher exception level than the profiling
> buffer owning exception level. This is also confirmed by the pseudocode
> for the StatisticalProfilingEnabled() function.
> 
> During the world switch and before activating guest traps, KVM executes
> at EL2 with the buffer owning exception level being EL1 (MDCR_EL2.E2PB =
> 0b11). As a result, profiling is already disabled when draining the
> buffer, making the isb() after the write to PMSCR_EL1 unnecessary.
> 
> CC: Will Deacon <w...@kernel.org>
> Signed-off-by: Alexandru Elisei <alexandru.eli...@arm.com>
> ---
>  arch/arm64/kvm/hyp/nvhe/debug-sr.c | 1 -
>  1 file changed, 1 deletion(-)
> 
> diff --git a/arch/arm64/kvm/hyp/nvhe/debug-sr.c 
> b/arch/arm64/kvm/hyp/nvhe/debug-sr.c
> index 7d3f25868cae..fdf0e0ba17e9 100644
> --- a/arch/arm64/kvm/hyp/nvhe/debug-sr.c
> +++ b/arch/arm64/kvm/hyp/nvhe/debug-sr.c
> @@ -33,7 +33,6 @@ static void __debug_save_spe(u64 *pmscr_el1)
>       /* Yes; save the control register and disable data generation */
>       *pmscr_el1 = read_sysreg_s(SYS_PMSCR_EL1);
>       write_sysreg_s(0, SYS_PMSCR_EL1);
> -     isb();

Hmm, but we still need an ISB somewhere between clearing pmscr_el1 and
mdcr_el2.e2pb, right? Where does that occur?

Will
_______________________________________________
kvmarm mailing list
kvmarm@lists.cs.columbia.edu
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm

Reply via email to