Re: [PATCH v6 28/33] KVM: s390: Add basic arm64 kvm module

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] Integer underflow in `kvm_max_guest_address()` bypasses KVM IPA limit validation for systems with less than 1GB of max address.
--

--- Patch [28]: [PATCH v6 28/33] KVM: s390: Add basic arm64 kvm module ---
commit 57ce89696f07b8212cf2fc85a369b60b1ec4f89b
Author: Steffen Eiden <[email protected]>

KVM: s390: Add basic arm64 kvm module

This commit adds the initial architecture-specific KVM infrastructure for
supporting arm64 guests on s390 hosts. It includes basic module
initialization, tracepoint scaffolding, and API skeleton functions.

> diff --git a/arch/s390/kvm/arm64/arm.c b/arch/s390/kvm/arm64/arm.c
> new file mode 100644
> index 0000000000000..8e17b7eef8131
> --- /dev/null
> +++ b/arch/s390/kvm/arm64/arm.c

[ ... ]

> +static u64 kvm_max_guest_address(void)
> +{
> +	u64 max_addr;
> +
> +	if (sclp.hamax == U64_MAX)
> +		max_addr = TASK_SIZE_MAX;
> +	else
> +		max_addr = min_t(u64, TASK_SIZE_MAX, sclp.hamax);
> +	return ALIGN_DOWN(max_addr + 1, 1 << 30) - 1;

[Severity: High]
Can this underflow if max_addr + 1 is less than 1GB (1 << 30)?

On a system with a small maximum address limit, ALIGN_DOWN() will evaluate 
to 0. Subtracting 1 from 0 will result in a U64_MAX return value, bypassing 
the intended limit.

> +}

[ ... ]

> +u32 get_kvm_ipa_limit(void)
> +{
> +	return fls64(kvm_max_guest_address() + 1) - 1;

[Severity: High]
If kvm_max_guest_address() underflows to U64_MAX as mentioned above, does 
this cause get_kvm_ipa_limit() to return U32_MAX?

Adding 1 to U64_MAX wraps to 0. Calling fls64(0) - 1 would then evaluate
to U32_MAX. This effectively disables the IPA size limit validation in
kvm_vm_type_ipa_size_shift(), allowing userspace to request unsupported
IPA sizes during VM creation.

> +}

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=28