On 07.07.23 17:35, Simon Glass wrote:
Hi Jan,
On Fri, 7 Jul 2023 at 14:56, Jan Kiszka jan.kiszka@siemens.com wrote:
On 07.07.23 14:42, Simon Glass wrote:
Hi Jan,
On Fri, 7 Jul 2023 at 11:05, Jan Kiszka jan.kiszka@siemens.com wrote:
On 05.07.23 00:14, Simon Glass wrote:
Hi Jan,
On Mon, 3 Jul 2023 at 20:34, Jan Kiszka jan.kiszka@siemens.com wrote:
Hi Simon,
On 28.06.23 13:41, Simon Glass wrote: > Collections can used to collect the contents of other entries into a > single entry, but they result in a single entry, with the original entries > 'left behind' in their old place. > > It is useful to be able to specific a set of entries ones and have it used > in multiple images, or parts of an image. > > Implement this mechanism. > > Signed-off-by: Simon Glass sjg@chromium.org > --- > > tools/binman/binman.rst | 80 ++++++++++++++++++++++++ > tools/binman/control.py | 9 +++ > tools/binman/etype/section.py | 3 +- > tools/binman/ftest.py | 8 +++ > tools/binman/test/286_entry_template.dts | 42 +++++++++++++ > 5 files changed, 141 insertions(+), 1 deletion(-) > create mode 100644 tools/binman/test/286_entry_template.dts > > diff --git a/tools/binman/binman.rst b/tools/binman/binman.rst > index a4b31fe5397b..9be979ae1c5b 100644 > --- a/tools/binman/binman.rst > +++ b/tools/binman/binman.rst > @@ -727,6 +727,13 @@ optional: > Note that missing, optional blobs do not produce a non-zero exit code from > binman, although it does show a warning about the missing external blob. > > +insert-template: > + This is not strictly speaking an entry property, since it is processed early > + in Binman before the entries are read. It is a list of phandles of nodes to > + include in the current (target) node. For each node, its subnodes and their > + properties are brought into the target node. See Templates_ below for > + more information. > + > The attributes supported for images and sections are described below. Several > are similar to those for entries. > > @@ -1172,6 +1179,79 @@ If you are having trouble figuring out what is going on, you can use > arch/arm/dts/u-boot.dtsi ... found: "arch/arm/dts/juno-r2-u-boot.dtsi" > > > +Templates > +========= > + > +Sometimes multiple images need to be created which have all have a common > +part. For example, a board may generate SPI and eMMC images which both include > +a FIT. Since the FIT includes many entries, it is tedious to repeat them twice > +in the image description. > + > +Templates provide a simple way to handle this:: > + > + binman { > + multiple-images; > + common_part: template-1 { > + fit { > + ... lots of entries in here > + }; > + > + text { > + text = "base image"; > + }; > + }; > + > + spi-image { > + filename = "image-spi.bin"; > + insert-template = <&fit>; > + > + /* things specific to SPI follow */ > + header { > + ]; > + > + text { > + text = "SPI image"; > + }; > + }; > + > + mmc-image { > + filename = "image-mmc.bin"; > + insert-template = <&fit>; > + > + /* things specific to MMC follow */ > + header { > + ]; > + > + text { > + text = "MMC image"; > + }; > + }; > + }; > + > +The template node name must start with 'template', so it is not considered to be > +an image itself. > + > +The mechanism is very simple. For each phandle in the 'insert-templates' > +property, the source node is looked up. Then the subnodes of that source node > +are copied into the target node, i.e. the one containing the `insert-template` > +property. > + > +If the target node has a node with the same name as a template, its properties > +override corresponding properties in the template. This allows the template to > +be uses as a base, with the node providing updates to the properties as needed. > +The overriding happens recursively. > + > +At present there is an unpleasant limitation on this feature: it works by > +appending the template nodes after any existing subnodes to the target node. > +This means that if the target node includes any subnodes, these appear in order > +before the template node. In the above example, 'header' becomes the first > +subnode of each image, followed by `fit` and `text`. If this is not what is > +desired, there is no way to adjust it. > + > +Note: The above limitation will likely be removed in future, so that the > +template subnodes appear before the target subnodes. > + > + > Updating an ELF file > ==================== > > diff --git a/tools/binman/control.py b/tools/binman/control.py > index 68597c4e7792..e7faca78e9aa 100644 > --- a/tools/binman/control.py > +++ b/tools/binman/control.py > @@ -22,6 +22,7 @@ from binman import bintool > from binman import cbfs_util > from binman import elf > from binman import entry > +from dtoc import fdt_util > from u_boot_pylib import command > from u_boot_pylib import tools > from u_boot_pylib import tout > @@ -478,6 +479,12 @@ def SignEntries(image_fname, input_fname, privatekey_fname, algo, entry_paths, > > AfterReplace(image, allow_resize=True, write_map=write_map) > > +def _ProcessTemplates(parent): > + for node in parent.subnodes: > + tmpl = fdt_util.GetPhandleList(node, 'insert-template') > + if tmpl: > + node.copy_subnodes_from_phandles(tmpl) > + > def PrepareImagesAndDtbs(dtb_fname, select_images, update_fdt, use_expanded): > """Prepare the images to be processed and select the device tree > > @@ -520,6 +527,8 @@ def PrepareImagesAndDtbs(dtb_fname, select_images, update_fdt, use_expanded): > raise ValueError("Device tree '%s' does not have a 'binman' " > "node" % dtb_fname) > > + _ProcessTemplates(node) > + > images = _ReadImageDesc(node, use_expanded) > > if select_images: > diff --git a/tools/binman/etype/section.py b/tools/binman/etype/section.py > index d56cc11d1023..adac2ff7fa87 100644 > --- a/tools/binman/etype/section.py > +++ b/tools/binman/etype/section.py > @@ -179,7 +179,8 @@ class Entry_section(Entry): > Returns: > bool: True if the node is a special one, else False > """ > - return node.name.startswith('hash') or node.name.startswith('signature') > + start_list = ('hash', 'signature', 'template') > + return any(node.name.startswith(name) for name in start_list) > > def ReadNode(self): > """Read properties from the section node""" > diff --git a/tools/binman/ftest.py b/tools/binman/ftest.py > index 4db54c69682c..9d9e47ce26b0 100644 > --- a/tools/binman/ftest.py > +++ b/tools/binman/ftest.py > @@ -6763,6 +6763,14 @@ fdt fdtmap Extract the devicetree blob from the fdtmap > data = self._DoReadFileDtb('285_spl_expand.dts', > use_expanded=True, entry_args=entry_args)[0] > > + def testEntryTemplate(self): > + """Test using a template""" > + TestFunctional._MakeInputFile('vga2.bin', b'#' + VGA_DATA) > + data = self._DoReadFile('286_entry_template.dts') > + first = U_BOOT_DTB_DATA + U_BOOT_DATA + VGA_DATA > + second = U_BOOT_DTB_DATA + b'#' + VGA_DATA + U_BOOT_DATA > + self.assertEqual(U_BOOT_IMG_DATA + first + second, data) > + > > if __name__ == "__main__": > unittest.main() > diff --git a/tools/binman/test/286_entry_template.dts b/tools/binman/test/286_entry_template.dts > new file mode 100644 > index 000000000000..6980dbfafcc6 > --- /dev/null > +++ b/tools/binman/test/286_entry_template.dts > @@ -0,0 +1,42 @@ > +// SPDX-License-Identifier: GPL-2.0+ > + > +/dts-v1/; > + > +/ { > + #address-cells = <1>; > + #size-cells = <1>; > + > + binman { > + u-boot-img { > + }; > + > + common_part: template { > + u-boot { > + }; > + > + intel-vga { > + filename = "vga.bin"; > + }; > + }; > + > + first { > + type = "section"; > + insert-template = <&common_part>; > + > + u-boot-dtb { > + }; > + }; > + > + second { > + type = "section"; > + insert-template = <&common_part>; > + > + u-boot-dtb { > + }; > + > + intel-vga { > + filename = "vga2.bin"; > + }; > + }; > + }; > +};
I tried to apply this on the IOT2050 use case, but it failed in two cases:
- We need a template where nodes can be manipulated afterwards, thus we
need anchors into those nodes. Naturally, those anchors need to get unique names so that something like this is possible:
common: template { anchor: some-node { value-1 = <1>; }; };
user-a { insert-template = <&common>; };
user-b { insert-template = <&common>; };
&anchor-a { value-2 = <2>; };
&anchor-b { value-2 = <3>; };
Rather than using a phandle, can you not use the node name in the template? For example:
user-a { insert-template = <&common>; some-node { value2 = <2>; }
user-b { insert-template = <&common>; some-node { value2 = <3>; }
I don't think this is working already. This is what I tried:
/dts-v1/; / { binman: binman { multiple-images;
my_template: template { blob-ext@0 { filename = "my-blob.bin"; offset = <0>; }; blob-ext@100 { filename = "/dev/null"; offset = <0x100>; }; }; my-image { filename = "my-image.bin"; insert-template = <&my_template>; blob-ext@100 { filename = "my-blob2.bin"; }; }; };};
First, I had to apply that null filename workaround, and the overridden filename is taken for the final image (if I leave out blob-ext@0), but this is at least ugly.
However, the file above as-is does not build:
binman: Node '/binman/my-image/blob-ext@0': Offset 0x0 (0) overlaps with previous entry '/binman/my-image/blob-ext@100' ending at 0x107 (263)
Probably for the same reason, leaving fit,fdt-list-val out in a template also generates an error during build. Putting an empty string there and overriding it later does not work.
I think it is just that it doesn't support templating of multiple-images nodes property. I sent a patch for this [1]
There may be other cases too, but if you have a failure, please see if you can adjust that test (287) to fail for you.
Hmm, already the test pattern from [1] still gives me
binman: Node '/binman/image/blob-ext@0': Offset 0x0 (0) overlaps with previous entry '/binman/image/blob-ext@8' ending at 0xf (15)
I've applied this series and then [1] on top.
Can you please list the commits so I can check it? I have:
88a31cccc20 (HEAD -> mkim2, dm/mkim-working) binman: Support templating with multiple images 71f1ef46d0c (dm-public/mkim-working) binman: Support simple templates 8841829b4c8 dtoc: Allow inserting a list of nodes into another fd2fb91c616 dtoc: Support copying the contents of a node into another 288f9e7c3b6 binman: Correct handling of zero bss size 4725a43d0ef binman: Drop __bss_size variable in bss_data.c c4bfe4917c9 binman: Provide a way to specific the fdt-list directly 6e571e4d260 binman: Convert mkimage to Entry_section 5fd2892283f stm32mp15: Avoid writing symbols in SPL 24c60031f48 binman: Allow disabling symbol writing 5c6a660333c binman: Read _multiple_data_files in the correct place 24a273113bf binman: Use GetEntries() to obtain section contents 1e0659321d8 binman: Init align_default in entry_Section cc3699d1d04 binman: Correct coverage gap in control 67d8b46e6ef (us/WIP/28Jun2023-next) Merge tag 'u-boot-amlogic-next-20230628' of https://source.denx.de/u-boot/custodians/u-boot-amlogic into next
I was using the dtoc patches from this series, not the updated but apparently not yet sent ones from your branch. Those seem to make a difference. Continuing to test.
Jan