Re: [Xen-devel] Xen and safety certification, Minutes of the meeting on Apr 4th (added brief meeting report from Genivi AMM)

[Lars Kurth <lars.kurth@xxxxxxxxxx>]

On 20/04/2018, 19:40, "Rich Persaud" <persaur@xxxxxxxxx> wrote:

    On Apr 20, 2018, at 13:25, Stefano Stabellini <sstabellini@xxxxxxxxxx> 
wrote:
    > 
    >> On Fri, 20 Apr 2018, Lars Kurth wrote:
    >> ## Standardisation: virtio
    >> 
    >> Standardization of hypervisor APIs - Matti (Open Synergy)
    >> 
    >> Matti made a case for virtio, which is probably a very bad idea from a 
Xen perspective (and indeed
    >> also from the perspective of QOQOS which is a proprietary hypervisor 
very similar to what a dom0-
    >> less Xen would look like.
    >> 
    >> The main issues are highlighted in 
https://markmail.org/message/gd7gnkpbsdw54mmm, aka brining
    >> in a device emulator and virtio access model requiring full privileges 
over the VM using the virtio
    >> driver. Artem and I briefly discussed whether we should try and raise 
very loud objections at this
    >> stage, and we agreed to just highlight these issues. In response Matti 
admitted these are issues, but
    >> that he would also not want to have to use QEMU (but something much 
simpler and more
    >> lightweight) and that the access model should be resolvable.
    >> 
    >> It was also interesting that Mentor and Windriver could not be made to 
make a statement whether
    >> they would ever support running such drivers in their OSes as guests.
    >> 
    >> I think we should observe for now, and in a few weeks offer to have some 
of our experts (maybe
    >> someone from OpenXT and/or Stefano) to engage and in more detail raise 
our concerns and
    >> convince Matti. Artem will look out for this.
    >> 
    >> **Note:** A meeting to discuss this at a later time that 10:00 is 
possible
    > 
    > I am happy to jump into any discussions or email threads to explain what
    > are the technical issues with virtio in details. See these previous
    > answers for reference:
    > 
    > https://marc.info/?l=xen-devel&m=151986861423752&w=2
    > https://marc.info/?l=xen-devel&m=152047729624095&q=p5
    > 
    > I think we should write a short whitepaper explaining why virtio is not
    > a good idea, and publish it. That way, we can point people to it. I'll
    > add it to my todo list.
    
    Good approach.  This paper can include a list of reasons why OS/hypervisor 
devs *do* choose virtio. 
    
    Paper should be versioned, since the “network effects” of virtio will 
change over time and the paper will need to be updated.
    
Making sure that we record this on xen-devel after I originally replied to the 
wrong thread which didn’t have xen-devel@ on it
Lars

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel