Skip to contentSkip to conversation

KVM: arm64: Fix architecturally invalid reset value for FPEXC32_EL2

PATCH Applied Dave Martin 2 participants 4 messages Feb 21, 2019 – Feb 22, 2019

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 KVM: arm64: Fix architecturally invalid reset value for FPEXC32_EL2

Due to what looks like a typo dating back to the original addition of FPEXC32_EL2 handling, KVM currently initialises this register to an architecturally invalid value.

As a result, the VECITR field (RES1) in bits [10:8] is initialised with 0, and the two reserved (RES0) bits [6:5] are initialised with
1. (In the Common VFP Subarchitecture as specified by ARMv7-A, these two bits were IMP DEF. ARMv8-A removes them.)

This patch changes the reset value from 0x70 to 0x700, which reflects the architectural constraints and is presumably what was originally intended.

Cc <[email protected]> # 4.12.x-
Cc Marc Zyngier <[email protected]>
Cc Christoffer Dall <[email protected]>
Fixes 62a89c44954f ("arm64: KVM: 32bit handling of coprocessor traps")

---

For AArch32 the situation is more complicated. FPEXC[29:0] is subarchitecture-defined, so the reset value may need to be per-CPU- implementation. The machinery for this is incomplete today In fact, I haven't currently found any reset logic for the VFP registers in arch/arm/kvm.

Currently I don't attempt to address this, but if anyone has ideas about the best way to address it, please shout.

I'll try to figure out something in the meantime, but it's not my top priority. We have no evidence I'm aware of that this is causing problems for anybody today.

---
 arch/arm64/kvm/sys_regs.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/arm64/kvm/sys_regs.c +1 -1
index c936aa4..b6dac3a 100644
@@ -1476,7 +1476,7 @@ static const struct sys_reg_desc sys_reg_descs[] = {
 
 	{ SYS_DESC(SYS_DACR32_EL2), NULL, reset_unknown, DACR32_EL2 },
 	{ SYS_DESC(SYS_IFSR32_EL2), NULL, reset_unknown, IFSR32_EL2 },
-	{ SYS_DESC(SYS_FPEXC32_EL2), NULL, reset_val, FPEXC32_EL2, 0x70 },
+	{ SYS_DESC(SYS_FPEXC32_EL2), NULL, reset_val, FPEXC32_EL2, 0x700 },
 };
 
 static bool trap_dbgidr(struct kvm_vcpu *vcpu,
Signature · 1 line
2.1.4
Applied
#1 lore raw
Quoted 13 lines · On Thu, 21 Feb 2019 11:42:32 +0000 Dave Martin <[email protected]> wrote:

Due to what looks like a typo dating back to the original addition of FPEXC32_EL2 handling, KVM currently initialises this register to an architecturally invalid value.

As a result, the VECITR field (RES1) in bits [10:8] is initialised with 0, and the two reserved (RES0) bits [6:5] are initialised with
1. (In the Common VFP Subarchitecture as specified by ARMv7-A, these two bits were IMP DEF. ARMv8-A removes them.)

This patch changes the reset value from 0x70 to 0x700, which reflects the architectural constraints and is presumably what was originally intended.

Quoted 5 lines
Cc <[email protected]> # 4.12.x-
Cc Marc Zyngier <[email protected]>
Cc Christoffer Dall <[email protected]>
Fixes 62a89c44954f ("arm64: KVM: 32bit handling of coprocessor traps")

Applied to -next.

Thanks,

	M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
#2 lore raw
Quoted 22 lines · Marc Zyngier
Quoted 13 lines · On Thu, 21 Feb 2019 11:42:32 +0000 Dave Martin <[email protected]> wrote:

Due to what looks like a typo dating back to the original addition of FPEXC32_EL2 handling, KVM currently initialises this register to an architecturally invalid value.

As a result, the VECITR field (RES1) in bits [10:8] is initialised with 0, and the two reserved (RES0) bits [6:5] are initialised with
1. (In the Common VFP Subarchitecture as specified by ARMv7-A, these two bits were IMP DEF. ARMv8-A removes them.)

This patch changes the reset value from 0x70 to 0x700, which reflects the architectural constraints and is presumably what was originally intended.

Quoted 5 lines
Cc <[email protected]> # 4.12.x-
Cc Marc Zyngier <[email protected]>
Cc Christoffer Dall <[email protected]>
Fixes 62a89c44954f ("arm64: KVM: 32bit handling of coprocessor traps")
Quoted 3 lines

Applied to -next.

Thanks,

Thanks.

I have a patch for <= v4.11 which I can send to stable once this patch reaches torvalds/master.

Is it worth it though, in your opinion?

Cheers
---Dave

#3 lore raw
Quoted 25 lines · On Fri, 22 Feb 2019 10:56:07 +0000 Dave Martin <[email protected]> wrote:
Quoted 24 lines · Marc Zyngier
Quoted 13 lines · On Thu, 21 Feb 2019 11:42:32 +0000 Dave Martin <[email protected]> wrote:

Due to what looks like a typo dating back to the original addition of FPEXC32_EL2 handling, KVM currently initialises this register to an architecturally invalid value.

As a result, the VECITR field (RES1) in bits [10:8] is initialised with 0, and the two reserved (RES0) bits [6:5] are initialised with
1. (In the Common VFP Subarchitecture as specified by ARMv7-A, these two bits were IMP DEF. ARMv8-A removes them.)

This patch changes the reset value from 0x70 to 0x700, which reflects the architectural constraints and is presumably what was originally intended.

Quoted 5 lines
Cc <[email protected]> # 4.12.x-
Cc Marc Zyngier <[email protected]>
Cc Christoffer Dall <[email protected]>
Fixes 62a89c44954f ("arm64: KVM: 32bit handling of coprocessor traps")

Applied to -next.

Quoted 8 lines
Quoted 1 line

Thanks,

Thanks.

I have a patch for <= v4.11 which I can send to stable once this patch reaches torvalds/master.

Is it worth it though, in your opinion?

I don't think it has much value. To be honest, this patch mostly serves as documentation, as any implementation worth its salt will honour the RES0/RES1 properties. I've kept the cc stable in your original patch, but arguably it doesn't fix anything real.

Thanks,

	M.
Signature · 1 line
Without deviation from the norm, progress is not possible.