Re: [PATCH 4/4] arm64/sysreg: Add VTCR_EL2 register

On Tue, 19 Aug 2025 05:24:29 +0100,
Anshuman Khandual [off-list ref] wrote:

On 18/08/25 2:52 PM, Mark Rutland wrote:

quoted

On Mon, Aug 18, 2025 at 10:27:59AM +0530, Anshuman Khandual wrote:

quoted

diff --git a/arch/arm64/tools/sysreg b/arch/arm64/tools/sysreg
index d2b40105eb41..f5a0a304f844 100644
--- a/arch/arm64/tools/sysreg
+++ b/arch/arm64/tools/sysreg

@@ -4910,6 +4910,63 @@ Field	1	PIE
 Field	0	PnCH
 EndSysreg
 
+Sysreg	VTCR_EL2	3	4	2	1	2
+Res0	63:46
+Field	45	HDBSS
+Field	44	HAFT
+Res0	43:42
+Field	41	TL0
+Field	40	GCSH
+Res0	39
+Field	38	D128
+Field	37	S2POE
+Field	36	S2PIE
+Field	35	TL1
+Field	34	AssuredOnly
+Field	33	SL2
+Field	32	DS
+Res1	31
+Field	30	NSA
+Field	29	NSW
+Field	28	HWU62
+Field	27	HWU61
+Field	26	HWU60
+Field	25	HWU59
+Res0	24:23
+Field	22	HD
+Field	21	HA
+Res0	20
+UnsignedEnum	19	VS
+	0b0	8BIT
+	0b1	16BIT
+EndEnum

You left TCR_EL1.AS as a single-bit 'Field', so please do the same here
for consistency. I don't think there's much gained by making this any
sort of enum.

But actually there is an use case in kvm_get_vtcr().

        /* Set the vmid bits */
        vtcr |= (get_vmid_bits(mmfr1) == 16) ?
                SYS_FIELD_PREP_ENUM(VTCR_EL2, VS, 16BIT) :
                SYS_FIELD_PREP_ENUM(VTCR_EL2, VS, 8BIT);

Here you go (untested):

diff --git a/arch/arm64/kvm/hyp/pgtable.c b/arch/arm64/kvm/hyp/pgtable.c
index c351b4abd5db..49266efc8bab 100644
--- a/arch/arm64/kvm/hyp/pgtable.c
+++ b/arch/arm64/kvm/hyp/pgtable.c

@@ -623,10 +623,7 @@ u64 kvm_get_vtcr(u64 mmfr0, u64 mmfr1, u32 phys_shift)
 	if (kvm_lpa2_is_enabled())
 		vtcr |= VTCR_EL2_DS;
 
-	/* Set the vmid bits */
-	vtcr |= (get_vmid_bits(mmfr1) == 16) ?
-		VTCR_EL2_VS_16BIT :
-		VTCR_EL2_VS_8BIT;
+	vtcr |= FIELD_PREP(BIT(VTCR_EL2_VS_SHIFT), (get_vmid_bits(mmfr1) == 16));
 
 	return vtcr;
 }

	M.

-- 
Jazz isn't dead. It just smells funny.