These patches are incorrect, see below:



On 1/8/25 3:50 AM, Hecht, Martin (Avnet Silica) via lists.yoctoproject.org 
wrote:
> Hi Tony, hi maintainers,
>
> I see a lot of good progress since a while in your yocto engagement.
>
> Nevertheless I just stumbled about some obviosly wrong assignments in some 
board
> config files for AMD/Xilinx boards. There is a fixed coded assignment of
> HDF_URI and checksum what should be replaced with ?= as default only.

These assignements are not incorrect. This was a design decision, and we won't be changing it.

The decision directly stems from the issue that people don't understand that when the XSA changes, it may or may not be compatible with a previously generated machine file. There is simply no way for the hardware (Vivado) engineers to know if their changes are going to break an existing Yocto Project machine configuration file.

*** If you change the XSA, it's no longer the same machine. ***

> This simplifies usage of your boards for customer evaluation like I'm 
preparing
> right now to connect camera arrays on zcu104 and zcu106.

You need to run gen-machine-conf for your custom design(s). It's the only safe way to make sure what you generate will work, more below.

The way we generate these files is scripted, and lives in the 'scripts' directory of the layer that holds the machine configuration files. I recommend you script it for your own usage as well.

Note, further, each customer evaluation with a modified XSA is a custom configuration. Best practice would be to use DIFFERENT machine names for each of these configurations to make it clear exactly which configuration is being targeted.

Further, this allows your Yocto Project engineers and support staff to be able to run diffs between what we (AMD) ship, and what you are using. If the only difference is the HDF_URI, then you can be pretty confident that your changes are limited in functionality to something 'new'. If it starts to change other lines, then your changes are creeping quickly into custom system changes which may lead to problems or other integration challenges, especially for evaluation designs.

> Please check attached patches and hopefully take over that approach for other
> AMD/Xilinx development boards.
>
> BR Martin


Subject: [PATCH 1/2] change assignemt operator for HDF_URI and
 HDF_URI[sha256sum] to default assignment

---
 conf/machine/zcu106-zynqmp-2024.2.conf | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/conf/machine/zcu106-zynqmp-2024.2.conf 
b/conf/machine/zcu106-zynqmp-2024.2.conf
index fede881..f9de9f0 100644
--- a/conf/machine/zcu106-zynqmp-2024.2.conf
+++ b/conf/machine/zcu106-zynqmp-2024.2.conf
@@ -56,8 +56,8 @@ XILINX_XSCT_VERSION = "2024.2"
 XILINX_WITH_ESW = "xsct"
# Add system XSA
-HDF_URI = 
"https://petalinux.xilinx.com/sswreleases/rel-v2024.2/hdf-examples/2024.2/2024.2_1111_1_12031048/external/zcu106-zynqmp/zcu106-zynqmp-2024.2_1111_1.xsa";
-HDF_URI[sha256sum] = 
"f41824dcfbb8faeec9f1ea3f0c5751a94ccd807d648aa3b9cb8efff78d820091"
+HDF_URI ?= 
"https://petalinux.xilinx.com/sswreleases/rel-v2024.2/hdf-examples/2024.2/2024.2_1111_1_12031048/external/zcu106-zynqmp/zcu106-zynqmp-2024.2_1111_1.xsa";
+HDF_URI[sha256sum] ?= 
"f41824dcfbb8faeec9f1ea3f0c5751a94ccd807d648aa3b9cb8efff78d820091"

Again, when the XSA changes, it is NOT the same board any longer. You _MUST_ re-run gen-machine-conf for YOUR custom board.

The conf/machine entries are specific to a _SINGLE_ hardware configuration which includes the physical board and associated SoC/FPGA configuration. If the hardware configuration changes, it is no longer the same machine.

Your workflow should expect to be:

1. Someone creates/modifies a board in Vivado. XSA is stored somewhere you can access it in your organization.

2. The integration engineer runs gen-machine-conf, you can look at the meta-xilinx-tools/scripts directory for how _WE_ run gen-machine-conf (including specific command line arguments.)

3. Integration engineer verifies that the new machine.conf works via some sort of build/validation.

4. Publish the new machine .conf in your own layer for your users (if 
applicable).

5. Go to 1.

 #### No additional settings should be after the Postamble
 #### Postamble
--
2.34.1


*Dipl. Inform. Martin Hecht*

Senior Embedded Specialist /
Functional Safety Engineer
TA1130871952

Avnet SILICA

http://digital.avnet.com/signature/images/silica/Avnet_silica_logo.png

Avnet SILICA
Kaiserin-Augusta-Allee 14

10553 Berlin

[email protected] <mailto:%[email protected]>
O +49 (0) 30 214 88 227

M +49 (0) 172 890 60 19

avnet-silica.com <http://avnet-silica.eu/>

We continuously commit to comply with the applicable data protection laws and ensure fair and transparent processing of your personal data. Please read our privacy statement including an information notice and data protection policy for detailed information on our website.





-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#5518): 
https://lists.yoctoproject.org/g/meta-xilinx/message/5518
Mute This Topic: https://lists.yoctoproject.org/mt/110494172/21656
Group Owner: [email protected]
Unsubscribe: https://lists.yoctoproject.org/g/meta-xilinx/unsub 
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

  • [meta-xilinx] wron... Hecht, Martin (Avnet Silica) via lists.yoctoproject.org
    • [meta-xilinx]... Hecht, Martin (Avnet Silica) via lists.yoctoproject.org
    • Re: [meta-xil... Mark Hatle
      • Re: [meta... Hecht, Martin (Avnet Silica) via lists.yoctoproject.org
      • [meta-xil... Hecht, Martin (Avnet Silica) via lists.yoctoproject.org
        • Re: [... Mark Hatle
          • R... Hecht, Martin (Avnet Silica) via lists.yoctoproject.org
            • ... Mark Hatle
            • ... Mark Hatle
              • ... Hecht, Martin (Avnet Silica) via lists.yoctoproject.org
              • ... Hecht, Martin (Avnet Silica) via lists.yoctoproject.org
                • ... Mark Hatle
                • ... Hecht, Martin (Avnet Silica) via lists.yoctoproject.org

Reply via email to