Re: [PATCH v3] xen/domain: introduce DOMID_ANY

[Roger Pau Monné <roger.pau@xxxxxxxxxx>]

On Wed, Dec 10, 2025 at 08:36:37AM +0100, Jan Beulich wrote:
> On 10.12.2025 02:04, Stefano Stabellini wrote:
> > On Tue, 23 Sep 2025, dmukhin@xxxxxxx wrote:
> >> From: Denis Mukhin <dmukhin@xxxxxxxx> 
> >>
> >> Add a new symbol DOMID_ANY aliasing DOMID_INVALID to improve the 
> >> readability
> >> of the code.
> >>
> >> Update all relevant domid_alloc() call sites.
> >>
> >> Amends: 2d5065060710 ("xen/domain: unify domain ID allocation")
> >> Signed-off-by: Denis Mukhin <dmukhin@xxxxxxxx>
> > 
> > Reviewed-by: Stefano Stabellini <sstabellini@xxxxxxxxxx>
> 
> The other day concern was voiced over aliasing DOMID_ANY with DOMID_INVALID.
> I don't recall though who it was or where.

I'm afraid it was me (at least) that voiced such concern.  But then I
completely forgot to reply to the patch.  I don't think this is a good
idea, aliasing DOMID_ANY with DOMID_INVALID is likely to be dangerous
in the long run.  In the example here it's fine, because the function
itself doesn't use DOMID_INVALID (iow: all usages of DOMID_INVALID are
replaced with DOMID_ANY).

However I could see a function wanting to use both DOMID_INVALID and
DOMID_ANY for different purposes.  Having both aliased to the same
value is not going to work as expected.  If we have to introduce
DOMID_ANY it must use a different value than DOMID_INVALID.  And given
the context here I would be fine leaving domid_alloc() to handle
getting passed DOMID_INVALID as a signal to search for an empty domid
to use, I don't see a compelling reason to introduce DOMID_ANY.

Thanks, Roger.