Re: [PATCH 3/3] vm_event: vmx: Allow intercepting hypervisor MSRs in debug builds

[Jan Beulich <jbeulich@xxxxxxxx>]

On 24.09.2026 13:57, Teddy Astie wrote:
> MSRs that are not covered by MSR bitmap (e.g hypervisor range) are
> unconditionally intercepted by hardware and don't need any specific handling.
> 
> However, vmx_set_msr_intercept() wrongly ASSERT() if the MSR is a part of the
> hypervisor range. Fix it by skipping the ASSERT() in this case.
> 
> This is only reachable with CONFIG_VM_EVENT using 
> XEN_DOMCTL_MONITOR_EVENT_MOV_TO_MSR
> to configure a MSR intercept in the hypervisor range (0x40000000–0x40001fff).

I don't follow here: Why's the hypervisor range special? Any MSR outside
the two ranges covered by the bitmap would trigger that assertion,
wouldn't it?

Even independent of that, perhaps it may be better to use ...

> --- a/xen/arch/x86/hvm/vmx/vmcs.c
> +++ b/xen/arch/x86/hvm/vmx/vmcs.c
> @@ -986,7 +986,7 @@ void vmx_set_msr_intercept(struct vcpu *v, unsigned int 
> msr,
>          if ( type & VMX_MSR_W )
>              set_bit(msr, msr_bitmap->write_high);
>      }
> -    else
> +    else if ( !((msr >= 0x40000000U) && (msr <= 0x40001fffU)) )
>          ASSERT(!"MSR out of range for interception\n");
>  }

...

    else if ( !monitored_msr(msr) )

here?

Jan