Now, Can the Device That Pass-through to The Guest Linux OS use "iommu" again to map its own device iova?

[tugouxp]

Hello folks:

  on device pass-through scenarios, a device can be pass-throued throud iommu by making a duplicate page table (IOVA, HPA) with the VCPU map(GPA,HPA), but the device in the guest os cant used iommu again

to make its own map like in HOST OS. that can be caculated through two ways:

1. dmesg|grep -i iommu

2. printk the struct busy_type->iommu_ops.

which the first one shows the iommu cant open in vm. and the later shows there are no iommu ops bingding to device driver.

So, is this the general case on present device virtualization? if so , did this means althoug we can do a nest virtuallization on CPU, but we still cant do the device nest virtualzation present?  thank your your kindly help.

BRs

zlcao.