[PATCH v5] RISC-V: KVM: Validate SBI STA shmem alignment in kvm_sbi_ext_sta_set_reg()

[Andrew Jones]

On Mon, Feb 02, 2026 at 12:38:57AM +0000, Jiakai Xu wrote:
> The RISC-V SBI Steal-Time Accounting (STA) extension requires the shared
> memory physical address to be 64-byte aligned, and the shared memory size
> to be at least 64 bytes.
> 
> KVM exposes the SBI STA shared memory configuration to userspace via
> KVM_SET_ONE_REG. However, the current implementation of
> kvm_sbi_ext_sta_set_reg() does not validate the alignment of the configured
> shared memory address. As a result, userspace can install a misaligned
> shared memory address that violates the SBI specification.
> 
> Such an invalid configuration may later reach runtime code paths that
> assume a valid and properly aligned shared memory region. In particular,
> KVM_RUN can trigger the following WARN_ON in
> kvm_riscv_vcpu_record_steal_time():
> 
>   WARNING: arch/riscv/kvm/vcpu_sbi_sta.c:49 at
>   kvm_riscv_vcpu_record_steal_time
> 
> WARN_ON paths are not expected to be reachable during normal runtime
> execution, and may result in a kernel panic when panic_on_warn is enabled.
> 
> Fix this by validating the shared memory alignment at the
> KVM_SET_ONE_REG boundary and rejecting misaligned configurations with
> -EINVAL. The validation is performed on a temporary computed address and
> only committed to vcpu->arch.sta.shmem once it is known to be valid, 
> similar to the existing logic in kvm_sbi_sta_steal_time_set_shmem() and
> kvm_sbi_ext_sta_handler().
> 
> With this change, invalid userspace state is rejected early and cannot
> reach runtime code paths that rely on the SBI specification invariants.
> 
> A reproducer triggering the WARN_ON and the complete kernel log are
> available at: https://github.com/j1akai/temp/tree/main/20260124

Any reason not to add these tests to
tools/testing/selftests/kvm/steal_time.c in the linux repo?

> 
> Fixes: f61ce890b1f074 ("RISC-V: KVM: Add support for SBI STA registers")
> Signed-off-by: Jiakai Xu <xujiakai2025 at iscas.ac.cn>
> Signed-off-by: Jiakai Xu <jiakaiPeanut at gmail.com>
> ---
> V4 -> V5: Added parentheses to function name in subject.
> V3 -> V4: Declared new_shmem at the top of kvm_sbi_ext_sta_set_reg().
>           Initialized new_shmem to 0 instead of vcpu->arch.sta.shmem.
>           Added blank lines per review feedback.
> V2 -> V3: Added parentheses to function name in subject.
> V1 -> V2: Added Fixes tag.
> 
> ---
>  arch/riscv/kvm/vcpu_sbi_sta.c | 16 +++++++++++-----
>  1 file changed, 11 insertions(+), 5 deletions(-)
> 
> diff --git a/arch/riscv/kvm/vcpu_sbi_sta.c b/arch/riscv/kvm/vcpu_sbi_sta.c
> index afa0545c3bcfc..bb13aa8eab7ee 100644
> --- a/arch/riscv/kvm/vcpu_sbi_sta.c
> +++ b/arch/riscv/kvm/vcpu_sbi_sta.c
> @@ -181,6 +181,7 @@ static int kvm_sbi_ext_sta_set_reg(struct kvm_vcpu *vcpu, unsigned long reg_num,
>  				   unsigned long reg_size, const void *reg_val)
>  {
>  	unsigned long value;
> +	gpa_t new_shmem = 0;

Sorry I missed this on my first review, but new_shmem should be
initialized to INVALID_GPA, since zero is a valid gpa.

>  
>  	if (reg_size != sizeof(unsigned long))
>  		return -EINVAL;
> @@ -191,18 +192,18 @@ static int kvm_sbi_ext_sta_set_reg(struct kvm_vcpu *vcpu, unsigned long reg_num,
>  		if (IS_ENABLED(CONFIG_32BIT)) {
>  			gpa_t hi = upper_32_bits(vcpu->arch.sta.shmem);
>  
> -			vcpu->arch.sta.shmem = value;
> -			vcpu->arch.sta.shmem |= hi << 32;
> +			new_shmem = value;
> +			new_shmem |= hi << 32;
>  		} else {
> -			vcpu->arch.sta.shmem = value;
> +			new_shmem = value;
>  		}
>  		break;
>  	case KVM_REG_RISCV_SBI_STA_REG(shmem_hi):
>  		if (IS_ENABLED(CONFIG_32BIT)) {
>  			gpa_t lo = lower_32_bits(vcpu->arch.sta.shmem);
>  
> -			vcpu->arch.sta.shmem = ((gpa_t)value << 32);
> -			vcpu->arch.sta.shmem |= lo;
> +			new_shmem = ((gpa_t)value << 32);
> +			new_shmem |= lo;
>  		} else if (value != 0) {
>  			return -EINVAL;
>  		}
> @@ -211,6 +212,11 @@ static int kvm_sbi_ext_sta_set_reg(struct kvm_vcpu *vcpu, unsigned long reg_num,
>  		return -ENOENT;
>  	}
>  
> +	if (new_shmem && !IS_ALIGNED(new_shmem, 64))

And then here check 'new_shmem != INVALID_GPA' since we want to allow the
user to set the "disabled shared memory" value (all-ones). Indeed our
testing should confirm that the value is either all-ones (disabled) or a
64-byte aligned address.

Thanks,
drew