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?
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
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..
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.
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.
Actually... I could be completely wrong...
Sorry about that.
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
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.
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:
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:
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