Re: [PATCH] KVM: RISC-V: Clear HSTATUS.HU on hypervisor initialization

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [Critical] The mitigation for HSTATUS.HU is placed in KVM initialization, leaving systems vulnerable if KVM is not loaded.
--

--- Patch [1]: [PATCH] KVM: RISC-V: Clear HSTATUS.HU on hypervisor initialization ---
commit 57c0e5888211429befb42a79f59ff40cc9adf07a
Author: Yong-Xuan Wang <[email protected]>

KVM: RISC-V: Clear HSTATUS.HU on hypervisor initialization

This commit clears the HSTATUS.HU bit during KVM initialization. This
prevents U-mode code from executing hypervisor instructions (HLV/HLVX/HSV)
to access memory, as the RISC-V specification does not mandate a reset
value for this bit.

> diff --git a/arch/riscv/kvm/main.c b/arch/riscv/kvm/main.c
> index 85e772588fad6..e80a99e93d569 100644
> --- a/arch/riscv/kvm/main.c
> +++ b/arch/riscv/kvm/main.c
> @@ -47,6 +47,9 @@ static void kvm_riscv_csr_init(void)
>  	csr_write(CSR_HCOUNTEREN, 0x02);
>  
>  	csr_write(CSR_HVIP, 0);
> +
> +	/* clear HSTATUS.HU */
> +	csr_clear(CSR_HSTATUS, HSTATUS_HU);

[Severity: Critical]
This is a pre-existing issue, but does placing this mitigation in
kvm_riscv_csr_init() leave the host vulnerable if the KVM module is never
loaded?

If the hardware or firmware leaves HSTATUS.HU=1 after reset, and KVM is
not initialized, could an unprivileged user-space (U-mode) process execute
hypervisor memory-management instructions (HLV/HSV)?

Since guest translation registers might remain in an uninitialized (Bare)
state, this might allow direct read/write access to host physical memory.
Should HSTATUS.HU be sanitized unconditionally during early core
architecture setup (e.g., in arch/riscv/kernel/setup.c) instead of only
when KVM is active?

>  }

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1