Re: [PATCH] hw/arm/virt: Allow additions to the generated device tree

[Peter Maydell <peter.maydell@linaro.org>]

On Mon, 27 Sept 2021 at 21:12, Simon Glass <sjg@chromium.org> wrote:
> I think you are misunderstanding my patch and that may be the problem here.
>
> Where QEMU is provided with a dtb (-dtb) it uses that and passes it
> on. This is absolutely fine and I have tested that this works well
> with U-Boot. No issues.
>
> Where QEMU creates its own dtb on the fly the -dtb parameter is
> actually ignored and there is no way to adjust what QEMU passes on,
> without recompiling QEMU. It is quite inflexible, actually. Even just
> creating a new device for development purposes is not possible. That
> is the problem I am trying to solve.
>
> There is certainly no intent to combine two bits of dtb with my patch.
> We could easily do that externally to QEMU.

So what *is* this patch doing? The subject says "Allow additions to
the generated device tree", and the patch adds an option to
"Merge a device tree binary (dtb) file into the generated device tree".
That sounds exactly like "combine two bits of dtb" -- the one the
user provides to the -dtbi option and the one QEMU generates
internally.

> The only current working option is to just pass the U-Boot dtb through
> and not use QEMU's on-the-fly-generated dtb at all. But I am assuming
> there is a reason why QEMU generates that dtb, so that would not be
> desirable?

QEMU generates the dtb to tell the guest what hardware is
present and where it is. We don't guarantee that that hardware
arrangement is necessarily stable between QEMU versions (in
practice there are generally additions and not deletions or
moves, but in theory there might be). All the guest is supposed
to assume is the location of RAM and flash; everything else it
must look in the dtb to determine.

> One more question...other than dtb, does QEMU typically add support
> for data structures needed by particular projects or groups of
> projects? It looks like dtb was supported for ARM Linux originally? I
> am looking at supporting bloblist as a way of communicating
> information between firmware (basically a simple way of packaging
> multiple blobs).

The answer here is essentially "no, as far as possible". We
ideally would not have special case support for any particular
guest code. We do have special handling for "directly boot the
Linux kernel" for a combination of historical reasons (we've
always supported it) and it being a massive use case. But even
that is painful to maintain (it starts off seeming easy but
gradually requires more weird tweaks and handling as CPU
architectures evolve). I really strongly don't want to add
additional categories of guests that QEMU has special case
support for, which is the main driver behind why I'm so negative
about this patchset.

Guest software running on the "virt" board needs to know it is
running on the "virt" board and deal with the interface and
information that QEMU provides. (For instance, Arm Trusted
Firmware and UEFI both have done this.)

-- PMM