Re: [edk2-devel] managing memory attributes in PEI

["Michael D Kinney" <michael.d.kinney@intel.com>]

Hi Ard,

Thanks.  I agree with your plan.

In the future, if we think there is value in enabling paging in 32-bit PEI, we could add the PPI at that time.

Mike

> -----Original Message-----
> From: devel@edk2.groups.io <devel@edk2.groups.io> On Behalf Of Ard
> Biesheuvel
> Sent: Tuesday, May 23, 2023 8:51 AM
> To: devel@edk2.groups.io; Kinney, Michael D
> <michael.d.kinney@intel.com>
> Cc: lersek@redhat.com; Ni, Ray <ray.ni@intel.com>; Yao, Jiewen
> <jiewen.yao@intel.com>; Gerd Hoffmann <kraxel@redhat.com>; Taylor
> Beebe <t@taylorbeebe.com>; Oliver Smith-Denny <osd@smith-denny.com>
> Subject: Re: [edk2-devel] managing memory attributes in PEI
> 
> On Tue, 23 May 2023 at 17:15, Michael D Kinney
> <michael.d.kinney@intel.com> wrote:
> >
> >
> >
> > > -----Original Message-----
> > > From: Ard Biesheuvel <ardb@kernel.org>
> > > Sent: Tuesday, May 23, 2023 7:59 AM
> > > To: Kinney, Michael D <michael.d.kinney@intel.com>
> > > Cc: devel@edk2.groups.io; lersek@redhat.com; Ni, Ray
> <ray.ni@intel.com>;
> > > Yao, Jiewen <jiewen.yao@intel.com>; Gerd Hoffmann
> > > <kraxel@redhat.com>; Taylor Beebe <t@taylorbeebe.com>; Oliver
> Smith-
> > > Denny <osd@smith-denny.com>
> > > Subject: Re: [edk2-devel] managing memory attributes in PEI
> > >
> > > On Tue, 23 May 2023 at 16:49, Kinney, Michael D
> > > <michael.d.kinney@intel.com> wrote:
> > > >
> > > > Ard,
> > > >
> > > > I would prefer to keep the IA32 PEI support for OVMF.
> > > >
> > >
> > > Sure. But does that imply that all enhancements regarding memory
> > > protections should be introduced there as well?
> >
> > I would prefer to not support these protections in IA32 PEI.  Same
> > for IA32 DXE.  Can the proposed PPI do nothing for IA32?
> >
> 
> Absolutely. I was just trying to narrow down whether your 'keeping
> IA32' meant just keeping it in working order, or have it keep up with
> future enhancements.
> 
> My intent is to implement an optional PPI that will be used by the PEI
> image loader to map PE code and data sections with the appropriate
> permissions if they are suitably aligned. Only the DXE core would
> generally fit this description, but there is no reason to disallow
> this for shadowed PEIMs that happen to be built as PE32 binaries with
> 4k section alignment (although I'm not convinced of the value add
> there)
> 
> If the PPI is not exposed (for any reason) things should just keep
> working as they do today.
> 
> Given that OVMF no longer functionally depends on IA32 PEI, we simply
> won't bother to implement the PPI at all for IA32.
> 
> 
> 
>