[PATCH v6 06/21] arm64: Move VHE-specific SPE setup to mutate_to_vhe()

Will Deacon will at kernel.org
Thu Feb 4 04:34:28 EST 2021


On Thu, Feb 04, 2021 at 09:30:08AM +0000, Marc Zyngier wrote:
> Hi Will,
> 
> On 2021-02-03 21:13, Will Deacon wrote:
> > Hi Marc,
> > 
> > On Mon, Feb 01, 2021 at 11:56:22AM +0000, Marc Zyngier wrote:
> > > There isn't much that a VHE kernel needs on top of whatever has
> > > been done for nVHE, so let's move the little we need to the
> > > VHE stub (the SPE setup), and drop the init_el2_state macro.
> > > 
> > > No expected functional change.
> > > 
> > > Signed-off-by: Marc Zyngier <maz at kernel.org>
> > > Acked-by: David Brazdil <dbrazdil at google.com>
> > > Acked-by: Catalin Marinas <catalin.marinas at arm.com>
> > > ---
> > >  arch/arm64/kernel/hyp-stub.S | 28 +++++++++++++++++++++++++---
> > >  1 file changed, 25 insertions(+), 3 deletions(-)
> > > 
> > > diff --git a/arch/arm64/kernel/hyp-stub.S
> > > b/arch/arm64/kernel/hyp-stub.S
> > > index 373ed2213e1d..6b5c73cf9d52 100644
> > > --- a/arch/arm64/kernel/hyp-stub.S
> > > +++ b/arch/arm64/kernel/hyp-stub.S
> > > @@ -92,9 +92,6 @@ SYM_CODE_START_LOCAL(mutate_to_vhe)
> > >  	msr	hcr_el2, x0
> > >  	isb
> > > 
> > > -	// Doesn't do much on VHE, but still, worth a shot
> > > -	init_el2_state vhe
> > > -
> > >  	// Use the EL1 allocated stack, per-cpu offset
> > >  	mrs	x0, sp_el1
> > >  	mov	sp, x0
> > > @@ -107,6 +104,31 @@ SYM_CODE_START_LOCAL(mutate_to_vhe)
> > >  	mrs_s	x0, SYS_VBAR_EL12
> > >  	msr	vbar_el1, x0
> > > 
> > > +	// Fixup SPE configuration, if supported...
> > > +	mrs	x1, id_aa64dfr0_el1
> > > +	ubfx	x1, x1, #ID_AA64DFR0_PMSVER_SHIFT, #4
> > > +	mov	x2, xzr
> > > +	cbz	x1, skip_spe
> > > +
> > > +	// ... and not owned by EL3
> > > +	mrs_s	x0, SYS_PMBIDR_EL1
> > > +	and	x0, x0, #(1 << SYS_PMBIDR_EL1_P_SHIFT)
> > > +	cbnz	x0, skip_spe
> > > +
> > > +	// Let the SPE driver in control of the sampling
> > > +	mrs_s	x0, SYS_PMSCR_EL1
> > > +	bic	x0, x0, #(1 << SYS_PMSCR_EL2_PCT_SHIFT)
> > > +	bic	x0, x0, #(1 << SYS_PMSCR_EL2_PA_SHIFT)
> > > +	msr_s	SYS_PMSCR_EL1, x0
> > 
> > Why do we need to touch pmscr_el1 at all? The SPE driver should take
> > care of
> > that, no? If you drop the pmscr_el1 accesses, I think you can drop the
> > pmbidr_el1 check as well.
> 
> That's definitely a brain fart, and is what we need on the nVHE path,
> not here. Doing the same thing twice isn't exactly helpful.
> 
> > And actually, then why even check dfr0? Doing the
> > bic for the mdcr_el1.e2pb bits is harmless.
> > 
> > > +	mov	x2, #MDCR_EL2_TPMS
> > > +
> > > +skip_spe:
> > > +	// For VHE, use EL2 translation and disable access from EL1
> > > +	mrs	x0, mdcr_el2
> > > +	bic	x0, x0, #(MDCR_EL2_E2PB_MASK << MDCR_EL2_E2PB_SHIFT)
> > > +	orr	x0, x0, x2
> > > +	msr	mdcr_el2, x0
> > 
> > Doesn't this undo the setting of mdcr_el2.hpmn if SPE is not present or
> > unavailable? (This wouldn't be an issue if we removed the skip_spe paths
> > altogether).
> 
> I don't think it does. We only clear the E2PB bits (harmless, as you point
> out above), and OR something that is either 0 (no SPE) or MDCR_EL2_TPMS.
> In any case, the HPMN bits are preserved (having been set during the nVHE
> setup).

Duh, OR not AND. Yes, sorry.

> I think the following patch addresses the above issue, which I'll squash
> with the original patch. Please shout if I missed anything.
> 
> Thanks,
> 
>         M.
> 
> diff --git a/arch/arm64/kernel/hyp-stub.S b/arch/arm64/kernel/hyp-stub.S
> index aa7e8d592295..3e08dcc924b5 100644
> --- a/arch/arm64/kernel/hyp-stub.S
> +++ b/arch/arm64/kernel/hyp-stub.S
> @@ -115,29 +115,9 @@ SYM_CODE_START_LOCAL(mutate_to_vhe)
>  	mrs_s	x0, SYS_VBAR_EL12
>  	msr	vbar_el1, x0
> 
> -	// Fixup SPE configuration, if supported...
> -	mrs	x1, id_aa64dfr0_el1
> -	ubfx	x1, x1, #ID_AA64DFR0_PMSVER_SHIFT, #4
> -	mov	x2, xzr
> -	cbz	x1, skip_spe
> -
> -	// ... and not owned by EL3
> -	mrs_s	x0, SYS_PMBIDR_EL1
> -	and	x0, x0, #(1 << SYS_PMBIDR_EL1_P_SHIFT)
> -	cbnz	x0, skip_spe
> -
> -	// Let the SPE driver in control of the sampling
> -	mrs_s	x0, SYS_PMSCR_EL1
> -	bic	x0, x0, #(1 << SYS_PMSCR_EL2_PCT_SHIFT)
> -	bic	x0, x0, #(1 << SYS_PMSCR_EL2_PA_SHIFT)
> -	msr_s	SYS_PMSCR_EL1, x0
> -	mov	x2, #MDCR_EL2_TPMS
> -
> -skip_spe:
> -	// For VHE, use EL2 translation and disable access from EL1
> +	// Use EL2 translations for SPE and disable access from EL1
>  	mrs	x0, mdcr_el2
>  	bic	x0, x0, #(MDCR_EL2_E2PB_MASK << MDCR_EL2_E2PB_SHIFT)
> -	orr	x0, x0, x2
>  	msr	mdcr_el2, x0

Looks a tonne better, thanks! Be nice if somebody could test it for us.

Will



More information about the linux-arm-kernel mailing list