Re: [cros-discuss] ChromiumOS on Exynos 4412

[Anush Elangovan(அனுஷ்)]

-chromium-os-discuss +chromium-os-dev

On Sat, Nov 3, 2012 at 7:21 PM, Mauro Ribeiro <mdr...@gmail.com> wrote:

Hey There,

I'm working on porting ChromiumOS to Exynos 4412.

Nice

 

Basicly what I've achieved so far is to boot it.

I'm running a modified overlay of daisy.

If you have a fully functional overlay feel free to submit CLs to gerrit if you would like to add them into the tree. 

 

I'm booting of a sdcard that has the exynos proprietary stuff, uboot and a fat partition that the kernel is on.

I've setted the root to /dev/sda3 (my usb drive)

So far my ChromiumOS image builds and boot.

But here's my problem:

Uppon boot it will keep at "Your system is repairing itself. Please wait" 

and boot loop on that.

Maybe it is looking for your stateful partition in a particular location. You could hack chromeos_startup to debug (or force the use of tmpfs)

http://git.chromium.org/gitweb/?p=chromiumos/platform/init.git;a=blob;f=chromeos_startup;h=3439c4c29e257a019657a1dfb1884cef3caf32b2;hb=HEAD

 

I've did a research around the web and I've found that its something on the state paritition, but the state partition is there and it isn't corrupted, If i put the sdcard on my laptop I can read it.

I've uploaded my clobber.log to www.mdrjr.net/clobber.log

Is there anything that I'm missing?

The integrity of the usb drive looks to be ok.

At first glance I've tought that it was somehow ChromiumOS getting lost and trying to access the sdcard that I use to boot instead of the USB key.

I've even disabled the sdcard support on the kernel to try. Still the same issue.

Any directions or points will be very helpful.

Best Regards,

Mauro Ribeiro

--
Chromium OS discuss mailing list: chromium-...@chromium.org
View archives, change email options, or unsubscribe:
http://groups.google.com/a/chromium.org/group/chromium-os-discuss?hl=en

[Mauro Ribeiro]

I've already did the hack to echo the STATE_DEV and it shows /dev/sda1 that
is suppoused to be correct. Am I right here?

What are the problems or issues to move the stateful partition to tmpfs ?

Could you point me where to change that?

As far as uploading the overlay I'll, but as soon as my vendor provide me
the magical blobs to the MALI GPU. (aka. EGL).
My overlay rightnow is just a daisy overlay buiding with cortex-a9 instead
of a15. And removing part of its gfx stuff.
More likely a arm-generic overlay (I've built that and it works as well).

Right now all I want is a image that will boot and log onto ChromiumOS.
Later after I achieve this goal I'll work on getting certain parts of my
hardware properly supported.

Thank you for your help ;)

2012/11/4 Mauro Ribeiro <mdr...@gmail.com>

I've already did the hack to echo the STATE_DEV and it shows /dev/sda1 that is suppoused to be correct. Am I right here?

What are the problems or issues to move the stateful partition to tmpfs ?

Could you point me where to change that?

As far as uploading the overlay I'll, but as soon as my vendor provide me the magical blobs to the MALI GPU. (aka. EGL).

My overlay rightnow is just a daisy overlay buiding with cortex-a9 instead of a15. And removing part of its gfx stuff.

More likely a arm-generic overlay (I've built that and it works as well).

Right now all I want is a image that will boot and log onto ChromiumOS. Later after I achieve this goal I'll work on getting certain parts of my hardware properly supported.

Thank you for your help ;) 

2012/11/4 Anush Elangovan(அனுஷ்) <an...@chromium.org>

[Mauro Ribeiro]

Just to note:

Its getting that on:

cleanup_mounts()

Even that the STATEFUL partition is getting mounted correctly. (used a mount >> /dev/ttySAC1 to see that) 

[Hung-Te Lin]

Maybe caused by "encrypted stateful partition" / TPM issue, since you probably don't have TPM in system?

Do it boot if you build factory test image (build_image factory_test) instead?

--
--
Chromium OS Developers mailing list: chromiu...@chromium.org

View archives, change email options, or unsubscribe:

http://groups.google.com/a/chromium.org/group/chromium-os-dev?hl=en
 
 
 

[Mauro Ribeiro]

Hung-Te Lin,

Thank you for your answer.

I'll test the factory_test image tonight at home. Is it supoused to work?

And yes, we don't have a TPM. 

2012/11/5 Hung-Te Lin <hun...@chromium.org>

[Richard Barnette]

On Nov 4, 2012, at 11:00 PM, Hung-Te Lin wrote:

> Maybe caused by "encrypted stateful partition" / TPM issue, since you probably don't have TPM in system?
> Do it boot if you build factory test image (build_image factory_test) instead?
>

Trouble with encrypted stateful is a good suspect. AFAIK, a TPM
isn't required, but encrypted stateful *does* require some specific
kernel modules. The fix would be to add those modules, or skip
the encrypted stateful step.

I think specifying 'factory' on the 'mount-encrypted' command line
will skip the encryption; keescook@ or hungte@ will know more
about what that option does. The option is specified automatically
in factory test images, but you could also change chromeos_startup
to pass the argument unconditionally.

Somebody like keescook@ will have to comment on exactly
which kernel modules would be needed.

-- jrb

[Richard Barnette]

+chromium-os-discuss

(moving the discussion back to the list, so everyone can benefit from
what's learned).

On Nov 5, 2012, at 4:09 PM, Mauro Ribeiro wrote:

> Richard,
>
> Thank you for your notes.
>
> Since the kernel that ChromiumOS uses doesn't support my board I've went ahead to search on ChromiumOS kernel to see what I need.
> I've found some config's needed and added.
> Also I've found a device driver on drivers/platform I've also ported those to my vendor suplied kernel (We use 3.6 series for Linux and 3.0 series for Android). I've ported those to 3.6.
>
> Now the board boots and doesn't goes onto a random loop of repair.
>
> Now it will stop saying that init can't start certain process.
>
> mtp and thermal I've disabled (mtpd as far as I know is related to VPN stuff and thermal is daisy related, I've moved the .conf's files in /etc/init somewhere else)
>
> Now 3 process that can't be started:
> factory
> factorylog
> bootcache
>
To clarify, the three things named are upstart jobs, not processes
as such. The first two are specific to factory test images; once
you're back to the regular developer image, these won't exist.

I don't know why bootcache would fail to start; however, unless it
still won't start once you're booting a developer image, I don't think
it's an interesting problem.


> Any idea?
>
> Right now I'm switching from the factory_test image back to dev.
>
Yeah, retrying with a dev image is the right next step. If it doesn't
fail in a dev image, it doesn't matter (to you, at any rate).


>
>
>
> 2012/11/5 Richard Barnette <jrbar...@chromium.org>

-- jrb

[Jorge Lucangeli Obes]

On Mon, Nov 5, 2012 at 6:35 PM, Richard Barnette
<jrbar...@chromium.org> wrote:
> +chromium-os-discuss
>
> (moving the discussion back to the list, so everyone can benefit from
> what's learned).
>
> On Nov 5, 2012, at 4:09 PM, Mauro Ribeiro wrote:
>
>> Richard,
>>
>> Thank you for your notes.
>>
>> Since the kernel that ChromiumOS uses doesn't support my board I've went ahead to search on ChromiumOS kernel to see what I need.
>> I've found some config's needed and added.
>> Also I've found a device driver on drivers/platform I've also ported those to my vendor suplied kernel (We use 3.6 series for Linux and 3.0 series for Android). I've ported those to 3.6.
>>
>> Now the board boots and doesn't goes onto a random loop of repair.
>>
>> Now it will stop saying that init can't start certain process.
>>
>> mtp and thermal I've disabled (mtpd as far as I know is related to VPN stuff and thermal is daisy related, I've moved the .conf's files in /etc/init somewhere else)
>>

FWIW, mtpd is the daemon responsible for exposing MTP (Media Transfer
Protocol) devices to Chrome.

[Mauro Ribeiro]

Hey there,

Now I've just runned onto a new issue.

I'm getting 

[6653:6653:1031/165421:ERROR:user_manager_impl.cc(105)] UserManager::RegisterPrefs

[6653:6692:1031/165421:ERROR:nss_util.cc(477)] Error initializing NSS with a persistent database (sql:/etc/fake_root_ca/nssdb): NSS error code: -8174

[6653:6653:1031/165421:WARNING:display_change_observer_x11.cc(65)] No display size available

[6697:6697:1031/165421:ERROR:nss_util.cc(477)] Error initializing NSS with a persistent database (sql:/etc/fake_root_ca/nssdb): NSS error code: -8174

[6653:6653:1031/165421:FATAL:image_transport_factory.cc(427)] Failed to initialize UI shared context.

On my /var/log/chrome

Checking /etc/fake_root_ca <- doesn't exists! 

Am I missing something here?

2012/11/6 Jorge Lucangeli Obes <jor...@chromium.org>

[Mauro Ribeiro]

I think that I've found the problem.

Google Chromium the Browser requires GL to work right?

My vendor doesn't provide OpenGL ES (Aka Mali Magical Blobs) for Linux.

Is there any way to get "Software" GL just to run?

I understand the concerts with performance if I do that.

2012/11/6 Mauro Ribeiro <mdr...@gmail.com>

[Simon Hong]

2012. 11. 7. 오전 10:18에 "Mauro Ribeiro" <mdr...@gmail.com>님이 작성:

>
> I think that I've found the problem.
>
> Google Chromium the Browser requires GL to work right?
>

Yes, cros is using gl for ui and rendering.

> My vendor doesn't provide OpenGL ES (Aka Mali Magical Blobs) for Linux.
>
> Is there any way to get "Software" GL just to run?

There is no open source gles driver for mali.
Also, cros is running on X.
I think you can use mesa, and it is already included in chromium.

[Olof Johansson]

Hi,

2012/11/7 Mauro Ribeiro <mdr...@gmail.com>

I think that I've found the problem.

Google Chromium the Browser requires GL to work right?

My vendor doesn't provide OpenGL ES (Aka Mali Magical Blobs) for Linux.

I haven't verified for sure, but I think it's possible to get a hold of the Mali 400 binaries through either Linaro, or ARM's Mali developer community somewhere. I think it's the same drivers as for 4210, i.e. what's used on the Origen board, etc.

(Sorry, I don't have any more detailed pointers than that for you, but hopefully you can find something there without too much searching)

-Olof