TDA4VM: PSDKLA Build output of Arago / Yocto and boot flow of linux

Hi Karthik,

thank you very much for answering my question!

Some more came up after digging into the topic:

Q5: How can I avoid that vmlinx.gz is deployed on target rootfs/boot? I want to make rootfs is tiny as possible for production. For example, is it sufficient to remove in "meta-ti/conf/machine/include/k3.inc" vmlinux.gz from line "KERNEL_IMAGETYPES = "Image vmlinux.gz"" ?
Or should I rather overwrite the KERNEL_IMAGETYPES variable in my own layer?

For the next question, I need to provide some additional information:

I created a own image configuration which is based on recipes-core/images/arago-image.inc. I tried this to get a reduced rootfs without GUI related stuff.
I added following packagegroups to that image:

IMAGE_INSTALL += "\
        packagegroup-arago-base \
        packagegroup-arago-console \
        packagegroup-arago-base-tisdk \
        packagegroup-arago-tisdk-addons \
"
This basically creates a functional kernel and rootfs. However, when "bitbaking" this image, the additional boot binaries like R5 SPL, DMSC Firmware, A72 SPL and u-boot for A72 do not appear in temp/deploy/images/j7-evm.

Q7: Which recipes are responsible for building and moving tiboot3.bin, sysfw.itb, tispl.bin and u-boot.img to temp/deploy/images/j7-evm ?

"Grepping" through sstate-control and arago recipes, I think the recipe "u-boot-ti-staging" should generate these files.

However, by building this recipe (MACHINE=j7-evm bitbake u-boot-ti-staging) only these files are created:

lrwxrwxrwx 2 ewdt ewdt        50 Oct 13 09:40 tispl.bin -> tispl.bin-j7-evm-2020.01+gitAUTOINC+f9b0d030d3-r12
lrwxrwxrwx 2 ewdt ewdt        50 Oct 13 09:40 tispl.bin-j7-evm -> tispl.bin-j7-evm-2020.01+gitAUTOINC+f9b0d030d3-r12
-rw-r--r-- 2 ewdt ewdt    687160 Oct 13 09:40 tispl.bin-j7-evm-2020.01+gitAUTOINC+f9b0d030d3-r12
-rw-r--r-- 2 ewdt ewdt   1025720 Oct 13 09:40 u-boot-j7-evm-2020.01+gitAUTOINC+f9b0d030d3-r12.img
lrwxrwxrwx 2 ewdt ewdt        51 Oct 13 09:40 u-boot-j7-evm.img -> u-boot-j7-evm-2020.01+gitAUTOINC+f9b0d030d3-r12.img
lrwxrwxrwx 2 ewdt ewdt        51 Oct 13 09:40 u-boot.img -> u-boot-j7-evm-2020.01+gitAUTOINC+f9b0d030d3-r12.img

It seems for me, that only binaries for the A72 were built, the binaries for R5F and M3 (DMSC firmware) are missing.

Sorry for the inconvenience, but I am a little stuck here.

Best regards!
Thomas

Hello Karthik,

thank you for your answers!

Regarding Q6:

I just added a additional image definition in my own layer:

# Arago TI based image which targets the "VIOP"
SUMMARY = "VIOP testing image"

#require recipes-core/images/tisdk-tiny-image.bb
require recipes-core/images/arago-image.inc

# Remove unwanted DISTRO and MACHINE FEATURES here
# TODO: Consider moving these into DISTRO and MACHINE configuration files
DISTRO_FEATURES_remove = "wayland"
DISTRO_FEATURES_append = "systemd"
MACHINE_FEATURES_remove = "alsa"
MACHINE_FEATURES_remove = "wayland"
MACHINE_FEATURES_remove = "gpu"


IMAGE_INSTALL += "\
        packagegroup-arago-base \
        packagegroup-arago-console \
        packagegroup-arago-base-tisdk \
        packagegroup-arago-tisdk-addons \
"
export IMAGE_BASENAME = "viop-image"

Nothing really special here. Just as a starting point to get a reduced linux rootfs in the end. I will check if ti-sci-fw_2020.04a.bb and k3r5.inc is included somewhere in the packagegroup definition.

Regarding Q7:
Could you please confirm, that "MACHINE=j7-evm bitbake u-boot-ti-staging" produce tiboot3.bin, tispl.bin and u-boot.img at your side?
On my machine, only tispl.bin and u-boot.img are created.

Again, many thanks!
Best regards,

Thomas

Hi Karthik,

thanks!

with "MACHINE=j7-evm-k3r5 bitbake u-boot-ti-staging" I get finally tiboot3.bin :)

Maybe I have a misunderstanding of the concept here: I thought, given by TI documentation, with the command "MACHINE=j7-evm bitbake tisdk-default-image" everything get built which I need to boot linux on A72. Everything includes for me tiboot3.bin, sysfw.itb, tispl.bin and u-boot.img.

Is the assumption wrong? Is it intended to execute "MACHINE=j7-evm-k3r5 bitbake u-boot-ti-staging" separately, to get the boot binaries for the R5?

I still work on building DMSC firmware. Got some errors, but I try to solve it myself first before I ask you again.

Best regards

Thomas

Hi Thomas,

When you execute the command "MACHINE=j7-evm bitbake tisdk-default-image" it takes care of this machine requirement. This is actually called multiconfig and the pointer to the recipe that I provided to you earlier does the trick in code.

When you are building the recipe independently like MACHINE=j7-evm-k3r5 bitbake u-boot-ti-staging, I have found this to work. Not sure if there is any other way but this works! :)

Thanks for confirming.

Regards

Karthik

Hi Karthik,

I would like to wrap up this thread and share my conclusions:

After executing "MACHINE=j7-evm bitbake tisdk-default-image" for the very first time, everything gets built and deployed in /temp/deploy/image .

When I manually delete /temp/deploy/image, and run bitbake tisdk-default-image for the second time, /temp/deploy/image will miss some boot binaries (especially for the R5).

I think some deploy tasks are not getting re-executed and are not cached. So when I delete the whole temp directory, and run bitbake tisdk-default-image again, everything is fine.

The same for my custom image configuration. That cost me 2 days, since I made false assumptions about the possible error, why my custom image misses some of the boot binaries in /temp/deploy/image  ;)

Second, my uncompressed/unstripped kernel image was so big, since a kernel-configuration fragment enabled debug symbols for the kernel. This is also fixed now.

Thank you for your support! I'm sure I will run into more problems, but I will create new threads for them.

Best Regards,
Thomas

Hi Thomas,

Thanks for a very nice summary, I'm glad that we could solve a majority of the issues in this discussion.

I will continue to chase down the issue of deploy tasks not getting executed during a rebuild. I haven't tried it myself yet. This may take some time to revert back and I will record on this thread for sake of future reference.

And yes, please create new threads if you have any other topics to discuss.

Regards
Karthik