Re: [PATCH 1/1] efi: avoid video ghosting in EFI app

Simon Glass <[email protected]> schrieb am Do., 6. Aug. 2026, 00:00:

> Hi Heinrich,
>
> On 2026-07-31T08:57:24, Heinrich Schuchardt
> <[email protected]> wrote:
> > efi: avoid video ghosting in EFI app
> >
> > When running the EFI app via QEMU with EDK II using
> >
> >   cp /usr/share/OVMF/OVMF_VARS_4M.fd .
> >   qemu-system-x86_64 -enable-kvm -m 2048 \
> >     -machine q35,smm=on \
> >     -drive
> if=pflash,format=raw,readonly=on,file=/usr/share/OVMF/OVMF_CODE_4M.fd \
> >     -drive if=pflash,format=raw,file=OVMF_VARS_4M.fd \
> >     -hda disk
> >
> > every character printed to the console by the U-Boot app appears twice.
> See
> > https://gist.github.com/xypron/7214e785c7d73104f8541bcac363a192 .
> >
> > The reason is that environment variable stdout has serial,vidconsole as
> > value. This is defined by board/efi/efi-x86_app/efi-x86_app.env including
> > include/env/x86.env.
> >
> > Provide a proper efi-x86_app.env file which does not use vidconsole.
> >
> > [...]
> >
> > board/efi/efi-x86_app/efi-x86_app.env | 13 ++++++++++++-
> >  1 file changed, 12 insertions(+), 1 deletion(-)
>
> > diff --git a/board/efi/efi-x86_app/efi-x86_app.env
> b/board/efi/efi-x86_app/efi-x86_app.env
> > @@ -1,6 +1,17 @@
> >  /* SPDX-License-Identifier: GPL-2.0+ */
> > +
> >  /*
> >   * Copyright (c) 2015 Google, Inc
> >   */
> >
> > -#include <env/x86.env>
> > +netdev=eth0
> > +consoledev=ttyS0
> > +scriptaddr=0x7000000
> > +kernel_addr_r=0x1000000
> > +ramdisk_addr_r=0x4000000
> > +ramdiskfile=initramfs.gz
> > +
> > +/* common console settings */
> > +stdin=serial
> > +stdout=serial
> > +stderr=serial
>
> Please can you keep '#include <env/x86.env>' and override just the
> three console variables afterwards?


Hello Simon,

the EFI app should not be x86 specific. You already fixed this in your
private repo.

Do you miss any value?

env2string.awk uses vars[var] =
> env, so a later assignment wins, and we avoid duplicating netdev,
> consoledev, scriptaddr, kernel_addr_r, ramdisk_addr_r and ramdiskfile
> — otherwise any future change to include/env/x86.env will drift away
> from the EFI app.
>
> Also, the previous include pulled in pciconfighost=1 which this
> version silently drops. If that is intentional it deserves a mention
> in the commit message; if not, the include-then-override approach
> fixes it for free.
>

The EFI must not drive any hardware directly. There is no value in
pciconfighost.

Best regards

Heinrich


> BTW an alternative is to leave x86.env alone and make the vidconsole
> entries conditional on something like !defined(CONFIG_EFI_APP), so
> efi-x86_payload.env and the other x86 boards keep sharing one source
> of truth. What do you think?
>
> Regards,
> Simon
>