Re: [Xen-devel] [PATCH] passthrough: give XEN_DOMCTL_test_assign_device more sane semantics

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 22.06.17 at 11:56, <george.dunlap@xxxxxxxxxx> wrote:
> On 22/06/17 08:05, Jan Beulich wrote:
>> No - I'm open to any change to it which makes the currently ignored
>> argument no longer ignored, without breaking existing (known and
>> unknown) callers of the libxc wrapper. I.e. I'm in no way opposed to
>> make it work the way you think it was originally meant to work; it is
>> just that given its current use I've come to a different conclusion as
>> to what the original intention may have been.
> 
> Actually, I think the clincher is this:
> 
> test_assign_device, assign_device, and deassign_device all use the same
> structure.
> 
> That makes it pretty obvious that "test_assign_device" was meant to ask
> the question, "If I call this hypercall with assign_device instead, will
> it succeed or fail?"

That's again one possible interpretation, yes, but that would again
collide with the domain left unused in the present implementation.
Anyway, let's see what others would prefer (and in the worst case
we can leave it in the strange state it's in now).

Jan


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