Re: [RFC] virtio-pmem: PMEM device spec

[David Hildenbrand <david@redhat.com>]

On 11.04.19 11:12, Pankaj Gupta wrote:
> 
>>>> new file mode 100644
>>>> index 0000000..04e07bb
>>>> --- /dev/null
>>>> +++ b/virtio-pmem.tex
>>>> @@ -0,0 +1,134 @@
>>>> +\section{PMEM Device}\label{sec:Device Types / PMEM Device}
>>>> +
>>>> +The virtio pmem is a fake persistent memory (NVDIMM) device
>>>> +used to bypass the guest page cache and provide a virtio
>>>> +based asynchronous flush mechanism. This avoids the need
>>>
>>> Is there anything "fake" about virtio-pmem from the perspective of the
>>> device interface?
>>>
>>> What you are describing is one use case.  But on a platform that doesn't
>>> have existing physical NVDIMM interfaces, ACPI, etc maybe virtio-pmem
>>> would be used to pass through physical NVDIMMs from the host?  If you
>>> agree, then it might make sense to describe the device simply as a
>>> persistent memory device and give "fake NVDIMM" as an example use case
>>> in a separate paragraph.  This would make the text more future-proof.
>>
>> Very good point. E.g. on x86-64, we might want to pass a real NVDIMM to
>> the guest using virtio-pmem, if our guest VM e.g. has no ACPI configured.
>>
>> In this case, it would be helpful to indicate if the flushing interface
>> is to be used, or if other flushing (e.g. like NVDIMMs) is to be used.
>>
> 
> Yes. Agree. This is a very good point, will consider pass-through of 
> real NVDIMM device for architectures don't support ACPI.
> 
> Yes, corresponding flushing mechanism should be mentioned but I am not sure
> if it should be mentioned in device specification?

Well there has to be a way to indicate how to flush writes, no?

> 
> Thanks,
> Pankaj   
> 


-- 

Thanks,

David / dhildenb