On 03/27/2012 04:48 PM, Somebody in the thread at some point said:

Is there a recommended way of generating a config fragment for a board?

Currently, the vexpress config is quite large and its origin is unknown to the LT. If I remove the linaro-base.conf entries from our config, we still end up with a config that is 793 lines long.

One way would be "make oldconfig" their starting-point then add it your arch-specific needs.

But we shouldn't be patching configs, but defconfigs.

If you do make savedefconfig, it'll create you a file "defconfig" that just contains the delta from your .config and "all defaults".

If you use that to create patches against the defconfigs, you'll find you're dealing with much shorter deltas that 793 lines.

-Andy

On 03/27/2012 01:48 AM, Ryan Harkin wrote:

Hi John,

On 26 March 2012 20:20, John Stultz <john.stultz@linaro.org mailto:john.stultz@linaro.org> wrote:

   So after talking about it at the last Linaro Connect, I've
finally gotten around to making a first pass at providing config
fragments for the linaro kernel.  I'd like to propose merging this
for 12.04, and doing so early so we can make sure that all the
desired config options are present in the fragments and to allow
the vairous linaro build systems to begin migrating their config
generation over.

The current tree is here:

http://git.linaro.org/gitweb?p=people/jstultz/android.git;a=shortlog;h=refs/heads/linaro-configs-3.3


The most relevant commit being:

http://git.linaro.org/gitweb?p=people/jstultz/android.git;a=commitdiff;h=da8f6d20e1a768cb486005c5ec62582b6f92990d


This includes config fragments for a linaro-base, an
android-ization fragment, and  board configs for panda, origen and
imx53.

I suspect we'll need an ubuntu specific fragment as well as all
the other board fragments.

There is likely to be quite a bit of churn as we decide what sort
of configs are really common and which are board specific. But
that's ok.

Configs are generated from the config fragments, as follows:

./scripts/kconfig/merge_config.sh ./configs/linaro-base.conf
./configs/android.conf ./configs/panda.conf


You may see warnings, which are not fatal, but should be reported
so they can be properly cleaned up.

I'm asking for Build folks to take a look at the above and
consider how they might merge in fragment assembly into their
system replacing their current config generation.

I'd ask Landing teams to take a look at this, and report any
missing config options or fragment chunks they'd like to see.

thanks
-john

Is there a recommended way of generating a config fragment for a board?

Hey, sorry I missed this email yesterday.

Currently, the vexpress config is quite large and its origin is unknown to the LT. If I remove the linaro-base.conf entries from our config, we still end up with a config that is 793 lines long.

In that set, there are things like CONFIG_WLAN=y. We don't have WLAN on vexpress. That's an easy case, but what about other more obscure things like CONFIG_XZ_DEC_BCJ, or CONFIG_FS_POSIX_ACL. I don't know what they are, where they've come from or if we need them. I know that the other platforms don't set them.

Or is it simpler to start from scratch and add in what we know we need?

As Andy already pointed out, savedefconfig is a good starting point. That gets rid of the config options that are already "default"

From there, I do something like the following:

$ ./scripts/kconfig/merge_config.sh -r ./configs/linaro-base.conf ./configs/ubuntu.conf ./mysaveddefconfig

The -r option will warn on options that are redundantly set. So you can then go through and eliminate them from the mysaveddefconfig, and then copy it to the config dir.

There will be some messiness here as we really haven't finalized how the generic/distro/board split works out. There will likely be a few special cases, but we'll get it sorted.

thanks -john

On 03/27/2012 06:40 AM, Jon Medhurst (Tixy) wrote:

On Mon, 2012-03-26 at 12:20 -0700, John Stultz wrote:

...

I suspect we'll need an ubuntu specific fragment as well as all the other board fragments.

The Ubuntu packaging scripts check that various config options are set, using these I have created an unbuntu.conf (attached) - I doubt this is what could be called a final version.

Very cool! Thanks so much for providing this! I'll add it to my tree.

The packaging scripts also have various checks for Linaro config options which I believe are missing from linaro-base.conf, these are:

CONFIG_HW_PERF_EVENTS=y CONFIG_ENABLE_DEFAULT_TRACERS=y || CONFIG_GENERIC_TRACER=y

I'll add these as well, thanks so much!

Finally, where can I find links to any docs on config fragments? I started trying to create a fragment for vexpress which just contained:

CONFIG_ARCH_VEXPRESS=y CONFIG_ARCH_VEXPRESS_CA9X4=y

but merge_config.sh said "Value requested for CONFIG_ARCH_VEXPRESS not in final .config".

So this is likely due to the ARCH_VEXPRESS option having dependencies that aren't met.

They way I usually sort this out is, after building the config. I call make menuconfig and hit "/ARCH_VEXPRESS" and it will show the current state of the dependencies for that option. The ones that aren't set, I then add to the config fragment.

But as the larger point about documentation, its severely lacking right now. My apologies for that. I'll see if I can find some time in the near future to improve this. If you have any specific documentation suggestions, I'd be happy to incorporate them .

thanks -john

On Tue, 2012-03-27 at 09:36 -0700, John Stultz wrote:

On 03/27/2012 06:40 AM, Jon Medhurst (Tixy) wrote:

...

On Mon, 2012-03-26 at 12:20 -0700, John Stultz wrote:

...

I suspect we'll need an ubuntu specific fragment as well as all the other board fragments.

The Ubuntu packaging scripts check that various config options are set, using these I have created an unbuntu.conf (attached) - I doubt this is what could be called a final version.

Very cool! Thanks so much for providing this! I'll add it to my tree.

Sorry, there's a bug in the ubuntu config, CONFIG_DEFAULT_MMAP_MIN_ADDR should be 32768, not 0.

On 03/28/2012 10:17 PM, John Stultz wrote:

On 03/28/2012 05:24 AM, Tushar Behera wrote:

...

On 03/27/2012 12:50 AM, John Stultz wrote:

...

 So after talking about it at the last Linaro Connect, I've finally

gotten around to making a first pass at providing config fragments for the linaro kernel. I'd like to propose merging this for 12.04, and doing so early so we can make sure that all the desired config options are present in the fragments and to allow the vairous linaro build systems to begin migrating their config generation over.

The current tree is here:

http://git.linaro.org/gitweb?p=people/jstultz/android.git%3Ba=shortlog%3Bh=r...

The most relevant commit being:

http://git.linaro.org/gitweb?p=people/jstultz/android.git%3Ba=commitdiff%3Bh...

This includes config fragments for a linaro-base, an android-ization fragment, and board configs for panda, origen and imx53.

I suspect we'll need an ubuntu specific fragment as well as all the other board fragments.

There is likely to be quite a bit of churn as we decide what sort of configs are really common and which are board specific. But that's ok.

Configs are generated from the config fragments, as follows:

./scripts/kconfig/merge_config.sh ./configs/linaro-base.conf ./configs/android.conf ./configs/panda.conf

You may see warnings, which are not fatal, but should be reported so they can be properly cleaned up.

I'm asking for Build folks to take a look at the above and consider how they might merge in fragment assembly into their system replacing their current config generation.

I'd ask Landing teams to take a look at this, and report any missing config options or fragment chunks they'd like to see.

I have updated origen.conf and linaro-base.conf for testing Samsung LT kernel. The results are updated in [1].

I'll take a look at the changes and try to merge them into my tree.

...

I have not validated the changes in android.conf, but by merging linaro-base.conf and origen.conf, I was able to boot my kernel the way I would have expected when using exynos4_defconfig.

One thing I notice is that I find far too many debug messages with this new config.

You mean the debug output from the merge_config.sh script is a bit noisy? Yea. Likely we'll have some extra noise as we settle out what options needs to be generic vs board specific. But it should decrease over time.

I was talking about the console log messages upon booting on a target board.

...

Going forward, would it be better to have linaro-base.conf, android.conf and ubuntu.conf managed centrally and each LT managing their board specific configuration file. That way, we can include the changes in our board specific configurations in respective topic branches.

So yea, I'd like to delegate/give away as much management of the configs as possible. :)

That said, I do think that we'll need someone looking at the entire cross-board fragment picture (since if everyone needs an option, it really isn't board specific). So it might be a good idea to have basic board config fragments that work with upstream. Then any board-specific feature branches can add their config needs in as a patch on top.

Does that sound reasonable?

Sounds good. Any config that enables a feature on a topic branch specific to a board should go in a patch in the topic branch itself.

On 03/29/2012 10:50 PM, John Stultz wrote:

On 03/28/2012 09:37 PM, Tushar Behera wrote:

...

On 03/28/2012 10:17 PM, John Stultz wrote:

...

On 03/28/2012 05:24 AM, Tushar Behera wrote:

...

I have not validated the changes in android.conf, but by merging linaro-base.conf and origen.conf, I was able to boot my kernel the way I would have expected when using exynos4_defconfig.

One thing I notice is that I find far too many debug messages with this new config.

You mean the debug output from the merge_config.sh script is a bit noisy? Yea. Likely we'll have some extra noise as we settle out what options needs to be generic vs board specific. But it should decrease over time.

I was talking about the console log messages upon booting on a target board.

Ok. Can you send me a "Before" config where you didn't see all the log messages?

The new log messages are because of this config entry. CONFIG_PROVE_LOCKING=y

But the information is useful, hence better it stays that way.

thanks -john

---

linaro-dev mailing list linaro-dev@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-dev

On 03/28/2012 09:37 PM, Tushar Behera wrote:

On 03/28/2012 10:17 PM, John Stultz wrote:

...

On 03/28/2012 05:24 AM, Tushar Behera wrote:

...

On 03/27/2012 12:50 AM, John Stultz wrote:

...

  So after talking about it at the last Linaro Connect, I've finally

gotten around to making a first pass at providing config fragments for the linaro kernel. I'd like to propose merging this for 12.04, and doing so early so we can make sure that all the desired config options are present in the fragments and to allow the vairous linaro build systems to begin migrating their config generation over.

The current tree is here:

http://git.linaro.org/gitweb?p=people/jstultz/android.git%3Ba=shortlog%3Bh=r...

The most relevant commit being:

http://git.linaro.org/gitweb?p=people/jstultz/android.git%3Ba=commitdiff%3Bh...

This includes config fragments for a linaro-base, an android-ization fragment, and board configs for panda, origen and imx53.

I suspect we'll need an ubuntu specific fragment as well as all the other board fragments.

There is likely to be quite a bit of churn as we decide what sort of configs are really common and which are board specific. But that's ok.

Configs are generated from the config fragments, as follows:

./scripts/kconfig/merge_config.sh ./configs/linaro-base.conf ./configs/android.conf ./configs/panda.conf

You may see warnings, which are not fatal, but should be reported so they can be properly cleaned up.

I'm asking for Build folks to take a look at the above and consider how they might merge in fragment assembly into their system replacing their current config generation.

I'd ask Landing teams to take a look at this, and report any missing config options or fragment chunks they'd like to see.

I have updated origen.conf and linaro-base.conf for testing Samsung LT kernel. The results are updated in [1].

I'll take a look at the changes and try to merge them into my tree.

Ok. I've merged most of what you added, but made some tweaks of my own to quiet some of the warnings at merge time.

Let me know if you see anything badly missing. Some of the items I dropped look like they're from feature branches?

thanks -john

On 03/30/2012 12:00 AM, Tushar Behera wrote:

Right, some of the config options should better move to the feature branches. I have cleaned up origen.conf so that we can now boot linux-linaro-tracking kernel till the console. Code dropped at [1].

With the help of a couple of patches[2], I was able to get Ubuntu booting up till home screen.

I haven't tested Android though.

Following is the list of added patches. 0b066ba ARM: EXYNOS: Increase DMA pool allocator size for framebuffer This patch is not required if CMA patches are added to the kernel.

fb8fa05 ARM: EXYNOS: Add clkdev lookup entry for lcd clock A rebased version of this patch is queued for 3.4-rc1.

[1] git://git.linaro.org/landing-teams/working/samsung/kernel.git (topic/linaro_config_3.3)

So I'm merged in your changes.

Although per the discussion with Tixy, you can go ahead and push any other configs required by out of tree features. Having topic branches based off of the config tree isn't really the right approach. Sorry for causing extra work here.

thanks -john