linux-fslc: add CONFIG_KPROBES=y to defconfig on linux-fslc #668
Conversation
Jiteshprm
force-pushed
the
master
branch
4 times, most recently
from
f5a2750
to
dcc6c9d
Compare
There was a problem hiding this comment.
First of, this commit is missing Signed-off-by: tag, please add it.
Also, I do not see this failure in the build... Do you have any additional fragments or recipe appends that enables the LTT-ng?
I do not see this config parameter also to be enabled upstream... As we try to keep the defconfig here as close to upstream as possible - it would be better if you can use this config variable in your own layer and apply it as a fragment.
-- andrey
|
Hi zandrey, To retest the fix, today I tried to rebuild with a brand new workspace from master but unfortunately and I got the error below. Probably I got lucky and could fix the issue with this simple fix when I checked out master the other day and the issues were only related with LTTng at that point in time? :) Error new workspace from 28 Feb: Repo info from old workspace build: Repo info from new workspace build: |
|
Hello @Jiteshprm
This shows a lot of warnings about almost all
So this means: this PR is not required anymore?
This has been already discussed on the meta-freescale ML: https://lists.yoctoproject.org/g/meta-freescale/message/24559 -- andrey |
|
Hi andrey,
This only started to happen in this new workspace, only upon the first checkout and when doing the "Parsing recipes" steps. I assure you that with the old workspace it did not happen.
It seems that it won't be needed as there are new issues happening now
As suggested in the post I have added the entry However a new issue when building the kernel appeared, which I show below. |
|
Hello @Jiteshprm
Can you try to do a clean repo clone in this case? Please do not re-use any of your previous working folder.
Can you close it then? I believe what is discussed here is not related to the PR itself.
I believe there was a suggestion in the entire topic on the ML about how can CODA VPU can be used with mainline BSP on i.MX6 derivative. What you've currently tried out is not considered as a valid combination of Community and NXP BSP flavors, it was also discussed on the ML in the same topic.
-- andrey |
|
We seem to reproduce it in other machine using mainline Linux kernel. You can see the error: Full log at: https://ci.ossystems.com.br/job/fsl-community-bsp-master_framebuffer/1030/console |
Yesterday I started a new clean folder checking out the available master branch versions.
Also I have read thoroughly and I understood that what I was doing was not correct. This time I did something which I don't know if it's very correct either (I'm not an expert on bitbake) and I changed the file:
to
Afterwards the build started correctly.
After leaving the build overnight running, the same build error appeared again, as Oscar has reported too. Full log from yesterday's build is attached. riotboard-master-build-new-build.txt
|
|
Those are two different issues. One is the lttng-modules and the other is the providers. Please open a issue for the providers one. |
|
I've just reproduced the issue with missing I guess the whole issue was discovered when The question is: should Adding In addition, I guess it would be better to perform either -- andrey |
|
We ought to enable it and apply this to our Linux kernel for the defconfig files we use as base. So we can later send it upstream. |
OK, understood. I've tested the @Jiteshprm Can you work on this PR further?
-- andrey |
Jiteshprm
force-pushed
the
master
branch
2 times, most recently
from
775f6cd
to
b8702a2
Compare
|
I have followed all the steps you have indicated on a new workspace that I have cloned today. riotboard-master-build-3March.txt Thanks, |
|
There are still some style problems with your patch - your When you commit your changes - please use You can find more details on it in kernel documentation section Submitting patches: the essential guide to getting your code into the kernel. I encourage you to follow it, since this patch is aimed at kernel source base, and perhaps would be reported as-is upstream, where those guidelines applies. Also, can you provide a content of your -- andrey |
|
Hi @zandrey Thank you very much for all the guidance and tips, they are very helpful, as it's my first time doing something like this. Also, I send attached the files you requested. I took a look and they seem unrelated with the kprobes activation. Anything else, please do let me know. Kind Regards |
There was a problem hiding this comment.
Thanks for updates and log files. From what I see there is no regression in the config items with your addition here. Those invalid items reported as the same as currently reported in upstream.
As for the commit message styling, I do not want to be too pedantic, but would give you a hint: if this commit would be reported as-is upstream to mainline kernel - it would get rejected, because it was not run through scripts/checkpatch.pl in the kernel. It does have long lines, which would be rejected by the script and hence - to accepted in the upstream.
I would leave it up to you and @otavio to decide whether commit message would require more fixing or can be merged as it is.
From my side here for the content: OK.
-- andrey
|
Hi, @zandrey - Thanks for the hint, I have googled it and I will run the patch through checkpatch.pl to make the commit message compatible with it. @otavio - Happy to do the change but I have one doubt. These are the the defconfig files I found within the sources folder: I believe that the ones that need updating are these below only. Kind Regards |
|
In fact, the linux-fslc ones. This is the only one that is newer than 5.7 ;) |
|
First, thank you both I've been learning a lot about these frameworks :) Then, as Andrey requested, I've build tested the configurations. For mxs I used the machine MACHINE=imx28evk and it also built successfully as you can see from the log: For mx25 I used the machine MACHINE=imx25pdk and it also built successfully as you can see from the log: But for armv8a, I'm struggling to find a compatible machine. I think it needs to be for iMX8 but it seems that these devices, at least for imx8mq-evk it seems to use the linux-imx and not linux-fslc kernel. Kind Regards |
Community BSP (containing You can pick it up locally and have a test build of the modified kernel containing your option additions. -- andrey |
|
Hi @zandrey As you said, I checked out your branch, added my changes manually and the build was good for armv8a with the kprobes on. Kind Regards |
|
Hello @Jiteshprm
Thanks for testing it out!
From the attached log, I see that It looks like that the config option you've introduced in the commit is actually not being set, and this needs to be investigated further on why it is not taken. I guess there is a dependency of this config option is not met, therefore it is not set proper. We should try to avoid such situations, therefore I need to ask you to have a look at it further.
-- andrey |
Fix compile error when building lttng using bitbake by: Enable kprobes configuration enabled from menuconfig Run diffconfig Copy the resulting diff contents into all the linux-fslc defconfig files Add end of line in defconfig Built the kernel with no warnings from do_kernel_configcheck task Add correct defconfig configuration for armv8a Tested on master branch from 3rd March 2021 Error: riotboard/fsl-community-bsp/build/tmp/work/ imx6dl_riotboard-fslc-linux-gnueabi/lttng-modules/2.12.4-r0/ lttng-modules-2.12.4/wrapper/kallsyms.c:20:3: error: 20 | # error "LTTng-modules requires CONFIG_KPROBES on kernels >= 5.7.0" | ^~~~~ make[3]: *** [riotboard/fsl-community-bsp/build/tmp/work-shared/ imx6dl-riotboard/kernel-source/scripts/Makefile.build:279: riotboard/fsl-community-bsp/build/tmp/work/ imx6dl_riotboard-fslc-linux-gnueabi/lttng-modules/2.12.4-r0/ lttng-modules-2.12.4/wrapper/kallsyms.o] Error 1 Compile flags: cd ~/riotboard/fsl-community-bsp MACHINE=imx6dl-riotboard DISTRO=fslc-framebuffer source setup-environment build bitbake fsl-image-machine-test Signed-off-by: Jitesh Pramodray <jiteshprm@gmail.com>
|
Hi @zandrey, I believe I found the issue. Kind Regards |
|
Hello @Jiteshprm Yeap, now it looks good! -- andrey |
Relevant changes: - 9203cfabdfb6 Merge pull request Freescale#668 from DiogoSilva014/6.6-1.0.x-imx-fix-lpcg - d6110c170fb9 Revert lpcg indice changes - 2df26cdfbbbe Merge pull request Freescale#667 from MrCry0/6.6-1.0.x-imx-fix-imx8-usdhc - a744d880dbb4 arch: arm64: freescale: imx8: fix sdhci clocks Signed-off-by: Oleksandr Suvorov <cryosay@gmail.com>
Relevant changes: - a1f3157034fe4 Merge pull request Freescale#670 from DiogoSilva014/6.6-1.0.x-imx-lpcg - 55720fcb81c3e arch: arm64: freescale: imx8qm: fix flexcan clocks - 9203cfabdfb68 Merge pull request Freescale#668 from DiogoSilva014/6.6-1.0.x-imx-fix-lpcg - d6110c170fb98 Revert lpcg indice changes - 2df26cdfbbbe0 Merge pull request Freescale#667 from MrCry0/6.6-1.0.x-imx-fix-imx8-usdhc - a744d880dbb46 arch: arm64: freescale: imx8: fix sdhci clocks Signed-off-by: Oleksandr Suvorov <cryosay@gmail.com>
Fix compile error when building lttng with bitbake, adding CONFIG_KPROBES=y to defconfig on linux-fslc
log.do_compile.2261270.txt
Error:
riotboard/fsl-community-bsp/build/tmp/work/imx6dl_riotboard-fslc-linux-gnueabi/lttng-modules/2.12.4-r0/lttng-modules-2.12.4/wrapper/kallsyms.c:20:3: error: #error "LTTng-modules requires CONFIG_KPROBES on kernels >= 5.7.0"
20 | # error "LTTng-modules requires CONFIG_KPROBES on kernels >= 5.7.0"
| ^~~~~
make[3]: *** [riotboard/fsl-community-bsp/build/tmp/work-shared/imx6dl-riotboard/kernel-source/scripts/Makefile.build:279: riotboard/fsl-community-bsp/build/tmp/work/imx6dl_riotboard-fslc-linux-gnueabi/lttng-modules/2.12.4-r0/lttng-modules-2.12.4/wrapper/kallsyms.o] Error 1
Compile flags:
cd ~/riotboard/fsl-community-bsp
MACHINE=imx6dl-riotboard
DISTRO=fslc-framebuffer
source setup-environment build
bitbake fsl-image-machine-test