Re: [PATCH v2 2/4] x86: qemu: Wire support for applying EFI capsule updates to pflash

Hi Elliot,

On 2026-08-07T18:07:09, Elliot Berman <[email protected]> wrote:
> x86: qemu: Wire support for applying EFI capsule updates to pflash
>
> Add new UUIDs for QEMU x86/x86-64 EFI capsules and register them with
> update_info. Enable the config options to bind the pflash with dfu,
> expecting to follow the typical convention of the first bank (nor0)
> containing u-boot.
>
> Signed-off-by: Elliot Berman <[email protected]>
>
> board/emulation/common/Kconfig         |  2 ++
>  board/emulation/common/qemu_dfu.c      | 16 ++++-----------
>  board/emulation/common/qemu_mtdparts.c | 36 +++++++++++++++------------------
>  board/emulation/qemu-x86/Kconfig       |  3 +++
>  board/emulation/qemu-x86/Makefile      |  3 +++
>  board/emulation/qemu-x86/qemu-x86.c    | 37 ++++++++++++++++++++++++++++++++++
>  6 files changed, 65 insertions(+), 32 deletions(-)

> diff --git a/board/emulation/common/qemu_mtdparts.c b/board/emulation/common/qemu_mtdparts.c
> @@ -50,28 +50,24 @@ void board_mtdparts_default(const char **mtdids, const char **mtdparts)
> -     /* Currently mtdparts is needed on Qemu ARM64 for capsule updates */
> -     if (IS_ENABLED(CONFIG_EFI_CAPSULE_FIRMWARE_MANAGEMENT) &&
> -         IS_ENABLED(CONFIG_TARGET_QEMU_ARM_64BIT)) {

Just to check the knock-on effect of dropping the target check:
qemu-arm-sbsa also builds this file (it implies SYS_MTDPARTS_RUNTIME
and sources the common Kconfig), so it now looks for nor0 too, and
MTDPARTS_NOR0 has no default there so it resolves to an empty string.
I suspect this is harmless today, since sbsa does not enable
FLASH_CFI_MTD so no nor0 MTD device is registered, but if that changes
the result would be a malformed 'mtdparts=nor0:' - perhaps
board_get_mtdparts() should skip a device whose partition string is
empty. Fine as a follow-up if you agree it is latent.

Reviewed-by: Simon Glass <[email protected]>

Regards,
Simon