Re: [PATCH v5 01/11] x86/vioapic: Add ioapic_check() to validate IO-APIC state before restore

[Jan Beulich <jbeulich@xxxxxxxx>]

On 03.09.2026 16:13, Julian Vetter wrote:
> @@ -595,6 +596,47 @@ int vioapic_get_trigger_mode(const struct domain *d, 
> unsigned int gsi)
>      return vioapic->redirtbl[pin].fields.trig_mode;
>  }
>  
> +static int cf_check ioapic_check(const struct domain *d, 
> hvm_domain_context_t *h)
> +{
> +    const HVM_SAVE_TYPE(IOAPIC) *s;
> +
> +    if ( !has_vioapic(d) )
> +        return -ENODEV;
> +
> +    s = hvm_get_entry(IOAPIC, h);
> +    if ( !s )
> +        return -ENODATA;
> +
> +    /*
> +     * base_address must be non-zero, page-aligned (hardware constraint), and
> +     * within the guest's physical address space (with room for the full MMIO
> +     * window).
> +     */
> +    if ( !s->base_address ||
> +         !IS_ALIGNED(s->base_address, PAGE_SIZE) ||
> +         s->base_address > (1ULL << hap_paddr_bits) - VIOAPIC_MEM_LENGTH )
> +        return -EINVAL;

->base_address can only ever be VIOAPIC_DEFAULT_BASE_ADDRESS for a DomU (and
that's all we care about here). If we wanted to relax that, further constraints
would need enforcing (like no conflict with the LAPIC page and no conflict
with anywhere that RAM or PCI BARs can live). This not being a constant would
also require changes to the ACPI tables we build for the guest.

I'm sorry for misleading you with a respective v4 comment.

Jan