Question about VA fault detection in system both support Sv32 and Sv39

[viktoryou]

In such IOMMU implementation with capabilities.Sv32=1 & capabilities.Sv39=1 and 64-bit addresss width, as the IOMMU model suggests, when iosatp.MODE=Sv32 and DC.SXL=1, 63–32 should all equal to bit 31, or else a page-fault exception will occur.

While I think this is not exactly what the IOMMU spec requires:

When SXL is 1, the following rules apply:

• If the first-stage is not Bare, then a page fault corresponding to the original access type occurs if the IOVA has bits beyond bit 31 set to 1.
• If the second-stage is not Bare, then a guest page fault corresponding to the original access type occurs if the incoming GPA has bits beyond bit 33 set to 1.

I prefer this is a bug in the model (see code), since in a 32-bit system, there is no such canonical rule for Sv32 as Sv39/Sv48/Sv57 in 64-bit system. The check may only detect the [63:32] for IOVA and [63:34] for GPA not to one.

[ved-rivos]

You are right that the code would produce a wrong result for 0xFFFFFFFF_???????? case for SXL=1. Fixing as follows in PR #286 :

if ( (masked_upper_bits != 0 && masked_upper_bits != mask && SXL == 0) ||
     (masked_upper_bits != 0 && SXL == 1) ) goto page_fault;