Re: [PATCH] xen/x86: allow Dom0 PVH to call XENMEM_exchange

[Jan Beulich <jbeulich@suse.com>]

On 28.04.2025 22:02, Stefano Stabellini wrote:
> On Mon, 28 Apr 2025, Teddy Astie wrote:
>> Hello Stefano,
>>
>> Le 25/04/2025 à 22:21, Stefano Stabellini a écrit :
>>> From: Xenia Ragiadakou <Xenia.Ragiadakou@amd.com>
>>>
>>> Dom0 PVH might need XENMEM_exchange when passing contiguous memory
>>> addresses to firmware or co-processors not behind an IOMMU.
>>>
>>> XENMEM_exchange was blocked for HVM/PVH DomUs, and accidentally it
>>> impacted Dom0 PVH as well.
>>>
>>> Permit Dom0 PVH to call XENMEM_exchange while leaving it blocked for
>>> HVM/PVH DomUs.
>>>
>>
>> In addition to Jan's remarks, I think it wants some additional
>> clarifications on the hypercall interface. public/memory.h indicates
>> "only PV guests can use this operation", so the interface is actually
>> unspecified about HVM guests (e.g what are MFN/GMFN in this case ?).
> 
> That is a new addition from c08a11ab98c.

Later addition or not, I agree with Teddy that it wants amending.

Jan

> If you see fae7d5be8bb, there
> is a statement that "we permitted this operation".