Re: [Xen-devel] [PATCHv1 0/2] xen/grant-table: fix PV guests and remove V2 support

[Paul Durrant <Paul.Durrant@xxxxxxxxxx>]

> -----Original Message-----
> From: David Vrabel
> Sent: 04 July 2014 13:21
> To: Paul Durrant; David Vrabel; xen-devel@xxxxxxxxxxxxxxxxxxxx
> Cc: Boris Ostrovsky
> Subject: Re: [Xen-devel] [PATCHv1 0/2] xen/grant-table: fix PV guests and
> remove V2 support
> 
> On 04/07/14 13:01, Paul Durrant wrote:
> >> -----Original Message-----
> >> From: David Vrabel
> >> Sent: 04 July 2014 12:55
> >> To: Paul Durrant; xen-devel@xxxxxxxxxxxxxxxxxxxx
> >> Cc: Boris Ostrovsky
> >> Subject: Re: [Xen-devel] [PATCHv1 0/2] xen/grant-table: fix PV guests and
> >> remove V2 support
> >>
> >> On 04/07/14 11:59, Paul Durrant wrote:
> >>>> -----Original Message-----
> >>>> From: xen-devel-bounces@xxxxxxxxxxxxx [mailto:xen-devel-
> >>>> bounces@xxxxxxxxxxxxx] On Behalf Of David Vrabel
> >>>> Sent: 02 July 2014 11:25
> >>>> To: xen-devel@xxxxxxxxxxxxxxxxxxxx
> >>>> Cc: Boris Ostrovsky; David Vrabel
> >>>> Subject: [Xen-devel] [PATCHv1 0/2] xen/grant-table: fix PV guests and
> >>>> remove V2 support
> >>>>
> >>>> The first patch fixes a logn standing issue where the map/unmap of
> >>>> grant table frames by PV guests was unsafe and could BUG() or have
> >>>> other bad behaviour.  This was caused by calling function that were
> >>>> not permitted from atomic contexts.
> >>>>
> >>>> The second patch removes all V2 table code since it is unused.
> >>>>
> >>>
> >>> How confident are you of that assertion?
> >>
> >> This is removing the /Linux/ code which has been using V1 only for a
> >> while now.
> >>
> >
> > If you believe that the API is likely to wither then that seems
> > reasonable. I would have thought the code was unlikely to see much
> > change though and it seems a shame to remove an interface to something
> > that it still there in Xen... I guess *someone* may want to use it in
> > future; particularly the copy-only grants could be useful for something
> > like IDC (if the backend was unprivileged).
> 
> It's already withering. For example PVH doesn't support V2 as it doesn't
> handle mapping/populating the status frames.  And fixing a long standing
> bug (see patch 1/2) was made more complicated by having to fix up the
> (unused) code for dealing with the status frames.
> 
> And as an aside, I'm not convinced that copy-only grants are what's
> needed for IDC with untrusted backend.  I think revocable grants are
> required for efficient usage of shared rings.
> 

Ok. Fair enough,

  Paul

> David

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel