Re: [PATCH v2 2/7] Documentation: s390-diag.rst: make diag500 a generic KVM hypercall

[Heiko Carstens <hca@linux.ibm.com>]

On Tue, Oct 15, 2024 at 10:32:43AM +0200, David Hildenbrand wrote:
> On 15.10.24 10:21, Heiko Carstens wrote:
> > On Tue, Oct 15, 2024 at 10:16:20AM +0200, David Hildenbrand wrote:
> > > On 15.10.24 10:12, Heiko Carstens wrote:
> > > > On Mon, Oct 14, 2024 at 09:35:27PM +0200, David Hildenbrand wrote:
> > > > > On 14.10.24 20:04, Heiko Carstens wrote:
> > > > "If only there would be a query subcode available, so that the program
> > > > check handling would not be necessary; but in particular my new subcode
> > > > is not worth adding it" :)
> > > > 
> > > > Anyway, I do not care too much.
> > > > 
> > > 
> > > Okay, I see your point: it would allow for removing the program check
> > > handling from the STORAGE LIMIT invocation.
> > > 
> > > ... if only we wouldn't need the exact same program check handling for the
> > > new query subfunction :P
> > 
> > Yeah yeah, but I think you got that this might help in the future.
> 
> Right. Adding it later also doesn't quite help to get rid of the checks
> here, because some user space might implement STORAGE LIMIT without QUERY.

This would only help if the diag500 documentation would state that
implementation of the QUERY subcode is mandatory. That is: for every
new subcode larger than the QUERY subcode QUERY must also exist.

That way we only would have to implement program check handling once,
if a program check happens on QUERY none of the newer subcodes is
available, otherwise the return value would indicate that.

Otherwise this whole excercise would be pointless.