Re: [PATCH v4 1/7] binman: Allow writing section contents to a file

From: Simon Glass <sjg@chromium.org>
To: Quentin Schulz <quentin.schulz@theobroma-systems.com>
Cc: U-Boot Mailing List <u-boot@lists.denx.de>,
	Roger Quadros <rogerq@ti.com>,
	 Alper Nebi Yasak <alpernebiyasak@gmail.com>,
	Peter Geis <pgwipeout@gmail.com>,
	 Philippe Reynes <philippe.reynes@softathome.com>,
	 Ivan Mikhaylov <ivan.mikhaylov@siemens.com>,
	Tom Rini <trini@konsulko.com>,  huang lin <hl@rock-chips.com>,
	Jeffy Chen <jeffy.chen@rock-chips.com>,
	 Kever Yang <kever.yang@rock-chips.com>,
	 Philipp Tomsich <philipp.tomsich@theobroma-systems.com>,
	 Heiko Thiery <heiko.thiery@gmail.com>,
	 Stefan Herbrechtsmeier <stefan.herbrechtsmeier@weidmueller.com>
Subject: Re: [PATCH v4 1/7] binman: Allow writing section contents to a file
Date: Mon, 7 Nov 2022 08:28:35 -0700	[thread overview]
Message-ID: <CAPnjgZ2Pv5ppTm8SSmNtAYMB-3raNU_HptGd4h6azjNUqER0HA@mail.gmail.com> (raw)
In-Reply-To: <c54c56b8-4fca-4191-4c5c-1f4ac02752b4@theobroma-systems.com>

Hi Quentin,

On Mon, 7 Nov 2022 at 07:25, Quentin Schulz
<quentin.schulz@theobroma-systems.com> wrote:
>
> Hi Simon,
>
> On 11/6/22 23:40, Simon Glass wrote:
> > At present only the image (which is a section) has a filename. Move this
> > implementation to the entry_Section class so that any section can have a
> > filename. With this, the section data is written to a file.
> >
> > This allows parts of an image to be written, along with the entire image.
> >
> > Signed-off-by: Simon Glass <sjg@chromium.org>
> > ---
> >
> > (no changes since v1)
> >
> >   tools/binman/binman.rst                 |  5 +++++
> >   tools/binman/etype/section.py           | 12 +++++++++-
> >   tools/binman/ftest.py                   | 14 ++++++++++++
> >   tools/binman/image.py                   |  3 ---
> >   tools/binman/test/261_section_fname.dts | 29 +++++++++++++++++++++++++
> >   5 files changed, 59 insertions(+), 4 deletions(-)
> >   create mode 100644 tools/binman/test/261_section_fname.dts
> >
> > diff --git a/tools/binman/binman.rst b/tools/binman/binman.rst
> > index fda16f1992d..79578ff127b 100644
> > --- a/tools/binman/binman.rst
> > +++ b/tools/binman/binman.rst
> > @@ -837,6 +837,11 @@ name-prefix:
> >       renamed to 'ro-u-boot' and 'rw-u-boot'. This can be useful to
> >       distinguish binaries with otherwise identical names.
> >
> > +filename:
> > +    This allows the contents of the section to be written to a file in the
> > +    output directory. This can sometimes be useful to use the data in one
> > +    section in different image, since there is currently no way to share data
> > +    beteen images other than through files.
> >
>
> IIRC, this is currently incorrect until we have inter-image dependencies
> since binman is building images in parallel by default. Suggesting this
> is a possible use-case is at beast misleading. For me, this is only
> useful for archiving embedded binaries, e.g. what we did for
> idbloader.img for Rockchip lately (which is "needed" only to keep the
> 3rd party tutorials/documentation not outdated).
>
> Maybe I missed some recent development that fixes this, lemme know if
> that's the case or if my assumptions are wrong.

Images themselves are built one after the other, so far as I know:

            for image in images.values():
                invalid |= ProcessImage(image, args.update_fdt, args.map,
                                       allow_missing=args.allow_missing,
                                       allow_fake_blobs=args.fake_ext_blobs)

WIthin each image Binman tries to build everything in parallel if
possible, but it is currently possible to use the outputs of one image
in a subsequent one.

I know it would be better to allow cross-image collections, etc. but
that is not implemented at present.

Regards,
SImon