Re: [PATCH] ppc/pnv: Introduce common PNV_SETFIELD() and PNV_GETFIELD() macros

[David Gibson]

On Thu, Apr 02, 2020 at 09:18:09AM +0200, Cédric Le Goater wrote:
> On 4/2/20 8:50 AM, David Gibson wrote:
> > On Thu, Apr 02, 2020 at 08:41:24AM +0200, Cédric Le Goater wrote:
> >> On 4/2/20 2:31 AM, David Gibson wrote:
> >>> On Wed, Apr 01, 2020 at 05:26:33PM +0200, Cédric Le Goater wrote:
> >>>> Most of QEMU definitions of the register fields of the PowerNV machine
> >>>> come from skiboot and the models duplicate a set of macros for this
> >>>> purpose. Make them common under the pnv_utils.h file.
> >>>>
> >>>> Signed-off-by: Cédric Le Goater <address@hidden>
> >>>
> >>> Hrm.  If we're touching these, would it make sense to rewrite them in
> >>> terms of the cross-qemu generic extract64() and deposit64()?
> >>
> >> I won't do that because we will loose compatibility with skiboot.
> > 
> > Uh.. how so?
> 
> What would be very nice is to use the QEMU FIELD() from "hw/registerfields.h".
> 
> But that's a complete different approach from what skiboot does and it would
> mean rewriting all the register definitions we include in QEMU for the 
> powernv 
> models and the code using the fields. It is a major change and I would rather 
> have the same files on both side (without tabs on the QEMU side). I think
> it is safer.
> 
> Using extract64() and deposit64() raises the same kind of problem AFAICT. 
> If we find a clean way to keep the register definition files the same, I am
> OK with any changes.

So, I wasn't suggesting changing all the users to use
extract64()/deposit64()/FIELD().  I was suggesting making the skiboot
like versions a wrapper than translates the arguments into the right
form for FIELD() or whatever.

-- 
David Gibson                    | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au  | minimalist, thank you.  NOT _the_ _other_
                                | _way_ _around_!
http://www.ozlabs.org/~dgibson

signature.asc
Description: PGP signature