Ubuntu precise armhf on Beaglebone – Kernel doesn't start

369 views
Skip to first unread message

georg

unread,
Jan 25, 2012, 6:43:03 PM1/25/12
to Beagle Board
Hello there,

I've been trying to get the armhf branch of Ubuntu testing to run on
the Beaglebone (using the demo image from <http://elinux.org/
BeagleBoardUbuntu#Precise_12.04_armhf_testing>), but it hangs after
uncompressing the kernel.

I thought it might be a problem with the serial console config not
being set correctly in the armhf demo image, so the board is actually
booting fine, and I'm only missing out on the console messages over
serial, but then I've compared the uEnv.txt configuration file to the
one included in the 11.10 armel demo image, and it's exactly the same
– the armel image works, though, so I'm a bit at a loss.

Did anybody manage to get the armhf branch to run on the Beaglebone
yet or maybe have some advice about things to try?

Thanks,
Georg

PS: Below are the console messages I see when trying to boot the armhf
image – Nothing out of the ordinary, but maybe I'm missing
something...

---------

U-Boot SPL 2011.09-dirty (Nov 21 2011 - 10:44:32)
Texas Instruments Revision detection unimplemented
No AC power, disabling frequency switch
OMAP SD/MMC: 0
reading u-boot.img
reading u-boot.img


U-Boot 2011.09-dirty (Nov 21 2011 - 10:44:32)

I2C: ready
DRAM: 256 MiB
No daughter card present
NAND: HW ECC Hamming Code selected
nand_get_flash_type: unknown NAND device: Manufacturer ID: 0x10, Chip
ID: 0x10
No NAND device found!!!
0 MiB
MMC: OMAP SD/MMC: 0
*** Warning - readenv() failed, using default environment

Net: cpsw
Hit any key to stop autoboot: 0
SD/MMC found on device 0
reading uEnv.txt

596 bytes read
Loaded environment from uEnv.txt
Importing environment from mmc ...
reading uImage

2957224 bytes read
reading uInitrd

2882782 bytes read
Booting from mmc ...
## Booting kernel from Legacy Image at 80300000 ...
Image Name: 3.1.0-psp3
Image Type: ARM Linux Kernel Image (uncompressed)
Data Size: 2957160 Bytes = 2.8 MiB
Load Address: 80008000
Entry Point: 80008000
Verifying Checksum ... OK
## Loading init Ramdisk from Legacy Image at 81600000 ...
Image Name: initramfs
Image Type: ARM Linux RAMDisk Image (uncompressed)
Data Size: 2882718 Bytes = 2.7 MiB
Load Address: 00000000
Entry Point: 00000000
Verifying Checksum ... OK
Loading Kernel Image ... OK
OK

Starting kernel ...

Uncompressing Linux... done, booting the kernel.

Robert Nelson

unread,
Jan 26, 2012, 11:23:20 AM1/26/12
to beagl...@googlegroups.com
On Wed, Jan 25, 2012 at 5:43 PM, georg <gka...@gmail.com> wrote:
> Hello there,
>
> I've been trying to get the armhf branch of Ubuntu testing to run on
> the Beaglebone (using the demo image from <http://elinux.org/
> BeagleBoardUbuntu#Precise_12.04_armhf_testing>), but it hangs after
> uncompressing the kernel.
>
> I thought it might be a problem with the serial console config not
> being set correctly in the armhf demo image, so the board is actually
> booting fine, and I'm only missing out on the console messages over
> serial, but then I've compared the uEnv.txt configuration file to the
> one included in the 11.10 armel demo image, and it's exactly the same
> – the armel image works, though, so I'm a bit at a loss.
>
> Did anybody manage to get the armhf branch to run on the Beaglebone
> yet or maybe have some advice about things to try?

Confirmed in the lab with my A2/A3 boards.. Not sure why it's failing
now, actually the first 'armhf' failure as the beagle/panda kernel
images haven't shown that problem.. I'm rebuilding both precise/wheezy
armhf kernel deb's incase it was just a gcc problem..

Regards,

--
Robert Nelson
http://www.rcn-ee.com/

Andrew Bradford

unread,
Jan 26, 2012, 1:30:21 PM1/26/12
to Beagle Board
I'm not fully sure that Ubuntu's kernel supports the AM335x chips,
yet. I have had some issues with the mainline 3.1 kernel on the
BeagleBone and ending up with a non-booting situation like you have.

I've found that the Arago project kernel sources and using the
am335x_evm_defconfig to build the kernel works a bit easier than the
mainline kernel. There's some important patches in the Arago repo
that aren't yet in mainline for supporting the AM335x. Give those a
try if you're still having trouble.

http://arago-project.org/git/projects/?p=linux-am33x.git

If you can get a Ubuntu or mainline unpatched kernel to build for the
Bone, definitely let me know. I'd be very interested to see
that .config file.

-Andrew


On Jan 26, 11:23 am, Robert Nelson <robertcnel...@gmail.com> wrote:

Andrew Bradford

unread,
Jan 26, 2012, 1:35:46 PM1/26/12
to Beagle Board
Actually... I could be completely wrong...
Sorry about that.

Robert Nelson

unread,
Jan 26, 2012, 1:40:18 PM1/26/12
to beagl...@googlegroups.com
On Thu, Jan 26, 2012 at 12:30 PM, Andrew Bradford <bra...@gmail.com> wrote:
> I'm not fully sure that Ubuntu's kernel supports the AM335x chips,
> yet.  I have had some issues with the mainline 3.1 kernel on the
> BeagleBone and ending up with a non-booting situation like you have.
>
> I've found that the Arago project kernel sources and using the
> am335x_evm_defconfig to build the kernel works a bit easier than the
> mainline kernel.  There's some important patches in the Arago repo
> that aren't yet in mainline for supporting the AM335x.  Give those a
> try if you're still having trouble.
>
> http://arago-project.org/git/projects/?p=linux-am33x.git
>
> If you can get a Ubuntu or mainline unpatched kernel to build for the
> Bone, definitely let me know.  I'd be very interested to see
> that .config file.

Correct, except my images don't use anything from ubuntu's kernel tree.. ;)

The kernel images in my demo image for the bone, are using that very
same arago tree.

So for some reason, the precise 'armhf' kernel build is locking up on
bootup.. Where as the oneiric 'armel' is booting fine.. (same exact
kernel source)..

georg

unread,
Jan 28, 2012, 9:45:13 AM1/28/12
to Beagle Board
Robert, I saw that you uploaded a new -psp3 build for precise/armhf –
Does this new build boot successfully? If so, I'd be interested to
know if you had to change anything, or if it was just a gcc fluke. I
intend to develop a kernel module for a SPI networking device on the
Beaglebone, so if there are some special armhf precautions to take
care of in the build process, it'd be great to know.

Also, as a very small sidenote, I noticed that the /lib/modules/XXXX/
build and ./source symlinks always point to /build/buildd/linux-3.1 –
It would be great if they'd correctly point to the /usr/src/ subdir
under which the accompanying kernel-header deb installs the headers.
It's trivial to fix if you are familiar with kernel module Makefiles,
but people just getting started might be confused when their
boilerplate Makefiles fail.

Cheers,
Georg

mre...@google.com

unread,
Jan 29, 2012, 11:17:46 AM1/29/12
to beagl...@googlegroups.com
Yeah, don't use the stock generic kernel, it's going to be missing some stuff and not configured quite like you need.

I used the armhf toolchain to build a kernel from the arago tree:
git://arago-project.org/git/projects/linux-am33x.git

I am now looking for a newer port for those patches to something like 3.2.x so I can experiement with the PREEMPT_RT patch.

Andrew Bradford

unread,
Jan 30, 2012, 1:20:45 PM1/30/12
to Beagle Board
On Jan 29, 11:17 am, mren...@google.com wrote:
> I used the armhf toolchain to build a kernel from the arago tree:
> git://arago-project.org/git/projects/linux-am33x.git
>
> I am now looking for a newer port for those patches to something like 3.2.x so I can experiement with the PREEMPT_RT patch.

Arago has a 3.2-staging branch:
http://arago-project.org/git/projects/?p=linux-am33x.git;a=shortlog;h=refs/heads/v3.2-staging

I've booted the 3.2-staging sources on my Bone but haven't checked
that anything beyond basic boot and networking works. I'm using
Emdebian stable toolchain (GCC 4.4.5) for armel and a Debian 6
filesystem.

-Andrew
Reply all
Reply to author
Forward
0 new messages