[PATCH] Revert "efi_capsule: Move signature from DTB to .rodata"

[Simon Glass]

Hi,

On Mon, 2 Aug 2021 at 01:15, KASHI Takahiro <takahiro.akashi at linaro.org> wrote:
>
> On Sun, Aug 01, 2021 at 08:47:15PM -0600, Simon Glass wrote:
> > Hi Ilias,
> >
> > On Sun, 1 Aug 2021 at 20:28, Ilias Apalodimas
> > <ilias.apalodimas at linaro.org> wrote:
> > >
> > > Hi Simon,
> > >
> > > On Sun, Aug 01, 2021 at 07:46:21PM -0600, Simon Glass wrote:
> > > > This was unfortunately applied despite much discussion about it being
> > > > the wrong way to implement this feature.
> > >
> > > No this was applied *before* the discussion, not despite.
> >
> > Oh sorry...I didn't notice either way. Normally there is an email on
> > the patch saying it was applied. Perhaps I missed it.
> >
> > >
> > > >
> > > > Revert it before too many other things are built on top of it.
> > >
> > > I don't really mind if this gets reverted but there's things that haven't
> > > been answered on that discussion [1] and my concern is what happens if
> > > CONFIG_OF_EMBED is not selected.
> >
> > Can we start a new discussion perhaps? Or use one of the contributor
> > calls to talk about it?
> >
> > We should not be using OF_EMBED except for testing.
> >
> > >
> > > Also you need to revert the entire series, not just one of the patches,
> > > as it changes the QEMU documentation for enabling authenticated capsule
> > > updates, as well as the mkeficapsule app.
> >
> > Heinrich, do you have any thoughts on this?
>
> # I'm not Heinrich :)

Perhaps you could impersonate him :-) I ask because he had been doing
a lot of EFI work.

>
> As far as the authentication logic itself is concerned,
> it is utterly generic except how and from where a public key is
> retrieved. (It can potentially be platform-specific.)
> Moreover, mkeficapsule really doesn't care where the key is.
>
> So I don't think we need revert all those changes.

I agree. Having another look, I think perhaps three patches is enough.
I will try again.

>
> For testing, we can run a test on sandbox by having sandbox-specific
> efi_get_public_key_data() function, i.e. we may want to contain
> the key in a file on ESP or just in a specific flash partition.
>
> Obviously, it's not safe, but it's just a test to verify that the logic
> is sane.
>
> If the discussion goes on for an unexpected spell of time,
> I would like to take this workaround for now.

I think this effort should go back to before the change to putting
things in rodata. That was when things went really off the rails.

With things back in the DT, you should be able to write a test with
the existing sandbox build without any special-case code.

[..]

Regards,
Simon