mail archive of the barebox mailing list
 help / color / mirror / Atom feed
From: Johannes Schneider <johannes.schneider@leica-geosystems.com>
To: barebox@lists.infradead.org
Cc: Marco Felsch <m.felsch@pengutronix.de>,
	Johannes Schneider <johannes.schneider@leica-geosystems.com>
Subject: [PATCH v1 00/14] efi: loader: boot Authenticode-signed UKIs
Date: Sun,  4 Oct 2026 01:19:33 +0000	[thread overview]
Message-ID: <20261004011958.3255011-1-johannes.schneider@leica-geosystems.com> (raw)

Hoi,


motivated by yet another FIT image CVE that "forced" us to roll out an
update (not really, we do that regularly anyway ;-) and Marco's hint
about UKIs on ARM, this series lets the EFI loader boot an
Authenticode-signed UKI with signed images forced, as an alternative
to a signed FIT image.

systemd-stub picks the devicetree from the UKI and hands it back
through EFI_DT_FIXUP_PROTOCOL, so barebox's fixups and overlays apply
as with FIT. With signing forced, LoadImage() verifies as well, as the
stub would otherwise load unsigned addons, and a UKI barebox cannot
boot properly is refused before it starts.

Open for opinions: overlays travel in a barebox-specific ".bbdtbo"
section, as UKIs have no place for them.
Known gap: memory reserved via EFI_DT_FIXUP_PROTOCOL stays reserved if
the payload returns to barebox instead of calling ExitBootServices().

Tested on an i.MX8MM board with rauc A/B updates and bootchooser,
including the refusal cases; the Authenticode parser has a fuzz test.

Beware: AI-generated code ahead. It went through a couple of
iterations, trimming and manual rewording, bench and smoke tests, and
agentic monkey-testing of the links in the verified boot chain.


Gruß
Johannes



Johannes Schneider (14):
  mfd: hgs-efi: do not claim the name of the EFI loader's device
  efi: loader: file: report EFI_UNSUPPORTED for volumes without a
    filesystem
  efi: loader: bootm: free the devicetree after installing it
  efi: loader: provide EFI_DT_FIXUP_PROTOCOL
  efi: loader: pe: add helpers for the image size and a named section
  efi: loader: bootm: load only the PE image, not the whole file
  efi: loader: bootm: pass the kernel command line to UKIs
  efi: loader: verify Authenticode signatures against built-in keys
  efi: loader: authenticode: add a fuzz test
  efi: loader: authenticate LoadImage() images when signing is forced
  efi: loader: file: expose no filesystem when signed images are forced
  bootm: efi: boot signed EFI images when signed images are forced
  efi: loader: bootm: install a devicetree for matching UKI devicetrees
  efi: loader: bootm: apply overlays carried by a UKI

 arch/arm/boards/hgs-gs05/board.c       |   2 +-
 common/boards/configs/libfuzzer.config |   1 +
 common/boards/hgs/common.c             |   2 +-
 common/bootm.c                         |   6 +-
 drivers/mfd/hgs-efi.c                  |   2 +-
 efi/Kconfig                            |  11 +
 efi/loader/Kconfig                     |  17 +
 efi/loader/Makefile                    |   1 +
 efi/loader/authenticode.c              | 448 +++++++++++++++++++++++++
 efi/loader/boot.c                      |   3 +
 efi/loader/bootm.c                     | 275 ++++++++++++++-
 efi/loader/pe.c                        | 123 ++++++-
 efi/loader/protocols/Kconfig           |   9 +
 efi/loader/protocols/Makefile          |   1 +
 efi/loader/protocols/dt_fixup.c        | 185 ++++++++++
 efi/loader/protocols/file.c            |  16 +
 include/efi/loader/authenticode.h      |  20 ++
 include/efi/loader/pe.h                |   7 +
 include/efi/protocol/dt_fixup.h        |  21 ++
 19 files changed, 1138 insertions(+), 12 deletions(-)
 create mode 100644 efi/loader/authenticode.c
 create mode 100644 efi/loader/protocols/dt_fixup.c
 create mode 100644 include/efi/loader/authenticode.h
 create mode 100644 include/efi/protocol/dt_fixup.h


base-commit: 58eda915df7a4e907dc741e26a90c9c7ba59a18b
-- 
2.43.0




             reply	other threads:[~2026-10-04  1:25 UTC|newest]

Thread overview: 19+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-10-04  1:19 Johannes Schneider [this message]
2026-10-04  1:19 ` [PATCH v1 01/14] mfd: hgs-efi: do not claim the name of the EFI loader's device Johannes Schneider
2026-10-04  1:19 ` [PATCH v1 02/14] efi: loader: file: report EFI_UNSUPPORTED for volumes without a filesystem Johannes Schneider
2026-10-04  1:19 ` [PATCH v1 03/14] efi: loader: bootm: free the devicetree after installing it Johannes Schneider
2026-10-04  1:19 ` [PATCH v1 04/14] efi: loader: provide EFI_DT_FIXUP_PROTOCOL Johannes Schneider
2026-10-04  1:19 ` [PATCH v1 05/14] efi: loader: pe: add helpers for the image size and a named section Johannes Schneider
2026-10-04  1:19 ` [PATCH v1 06/14] efi: loader: bootm: load only the PE image, not the whole file Johannes Schneider
2026-10-04  1:19 ` [PATCH v1 07/14] efi: loader: bootm: pass the kernel command line to UKIs Johannes Schneider
2026-10-05 17:18   ` Ahmad Fatoum
2026-10-04  1:19 ` [PATCH v1 08/14] efi: loader: verify Authenticode signatures against built-in keys Johannes Schneider
2026-10-05  5:33   ` Ahmad Fatoum
2026-10-05  5:45     ` SCHNEIDER Johannes
2026-10-09  0:07       ` SCHNEIDER Johannes
2026-10-04  1:19 ` [PATCH v1 09/14] efi: loader: authenticode: add a fuzz test Johannes Schneider
2026-10-04  1:19 ` [PATCH v1 10/14] efi: loader: authenticate LoadImage() images when signing is forced Johannes Schneider
2026-10-04  1:19 ` [PATCH v1 11/14] efi: loader: file: expose no filesystem when signed images are forced Johannes Schneider
2026-10-04  1:19 ` [PATCH v1 12/14] bootm: efi: boot signed EFI images " Johannes Schneider
2026-10-04  1:19 ` [PATCH v1 13/14] efi: loader: bootm: install a devicetree for matching UKI devicetrees Johannes Schneider
2026-10-04  1:19 ` [PATCH v1 14/14] efi: loader: bootm: apply overlays carried by a UKI Johannes Schneider

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20261004011958.3255011-1-johannes.schneider@leica-geosystems.com \
    --to=johannes.schneider@leica-geosystems.com \
    --cc=barebox@lists.infradead.org \
    --cc=m.felsch@pengutronix.de \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox