Re: [PATCH] ima: add a new CONFIG for loading arch-specific policies

[Heiko Carstens]

On Mon, Mar 02, 2020 at 09:56:58AM -0500, Mimi Zohar wrote:
> On Mon, 2020-03-02 at 15:52 +0100, Ard Biesheuvel wrote:
> > On Mon, 2 Mar 2020 at 15:48, Mimi Zohar <zohar@xxxxxxxxxxxxx> wrote:
> > > > diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
> > > > index beea77046f9b..cafa66313fe2 100644
> > > > --- a/arch/x86/Kconfig
> > > > +++ b/arch/x86/Kconfig
> > > > @@ -230,6 +230,7 @@ config X86
> > > >       select VIRT_TO_BUS
> > > >       select X86_FEATURE_NAMES                if PROC_FS
> > > >       select PROC_PID_ARCH_STATUS             if PROC_FS
> > > > +     select IMA_SECURE_AND_OR_TRUSTED_BOOT   if EFI
> > >
> > > Not everyone is interested in enabling IMA or requiring IMA runtime
> > > policies.  With this patch, enabling IMA_ARCH_POLICY is therefore
> > > still left up to the person building the kernel.  As a result, I'm
> > > seeing the following warning, which is kind of cool.
> > >
> > > WARNING: unmet direct dependencies detected for
> > > IMA_SECURE_AND_OR_TRUSTED_BOOT
> > >   Depends on [n]: INTEGRITY [=y] && IMA [=y] && IMA_ARCH_POLICY [=n]
> > >   Selected by [y]:
> > >   - X86 [=y] && EFI [=y]
> > >
> > > Ard, Michael, Martin, just making sure this type of warning is
> > > acceptable before upstreaming this patch.  I would appreciate your
> > > tags.
> > >
> > 
> > Ehm, no, warnings like these are not really acceptable. It means there
> > is an inconsistency in the way the Kconfig dependencies are defined.
> > 
> > Does this help:
> > 
> >   select IMA_SECURE_AND_OR_TRUSTED_BOOT   if EFI && IMA_ARCH_POLICY
> > 
> > ?
> 
> Yes, that's fine for x86.  Michael, Martin, do you want something
> similar or would you prefer actually selecting IMA_ARCH_POLICY?

For s390 something like

	select IMA_SECURE_AND_OR_TRUSTED_BOOT if IMA_ARCH_POLICY

should be fine.

Thanks,
Heiko