Re: [PATCH v2 2/6] x86/gnttab: do not implement GNTTABOP_cache_flush

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

On Fri, May 16, 2025 at 12:38:02PM +0200, Jan Beulich wrote:
> On 16.05.2025 12:31, Roger Pau Monné wrote:
> > On Fri, May 16, 2025 at 11:48:48AM +0200, Jan Beulich wrote:
> >> On 16.05.2025 11:45, Roger Pau Monne wrote:
> >>> --- a/xen/common/Kconfig
> >>> +++ b/xen/common/Kconfig
> >>> @@ -35,6 +35,11 @@ config GRANT_TABLE
> >>>  
> >>>     If unsure, say Y.
> >>>  
> >>> +config HAS_GRANT_CACHE_FLUSH
> >>> + bool
> >>> + depends on GRANT_TABLE
> >>> + default ARM
> >>
> >> To keep arch stuff out of common file as much as possible, I think this 
> >> instead
> >> wants to be a "select ..." from ARM. Then:
> >> Reviewed-by: Jan Beulich <jbeulich@xxxxxxxx>
> > 
> > My first attempt was to do it as you suggest, but then if the users
> > disables GRANT_TABLE you get the following warning:
> > 
> > WARNING: unmet direct dependencies detected for HAS_GRANT_CACHE_FLUSH
> >   Depends on [n]: GRANT_TABLE [=n]
> >   Selected by [y]:
> >   - ARM [=y]
> > configuration written to .config
> 
> Right, it needs to be
> 
>       select HAS_GRANT_CACHE_FLUSH if GRANT_TABLE
> 
> (and the "depends on" on the new HAS_* can also go away; HAS_* imo shouldn't
> normally have any dependencies).

Hm, OK, I didn't like it this way because then we force all users of
HAS_GRANT_CACHE_FLUSH to make it depend on GRANT_TABLE, while putting
the dependency in HAS_GRANT_CACHE_FLUSH avoids that.

Anyway, have adjusted according to your suggestion.

Thanks, Roger.