The build the kernel procedure fails for me at "git checkout origin/am33x_v3.2 -b am33x_v3.2" with "fatal: git checkout: updating paths in incompatible with switching branches."

Thanks for the instructions. Worked like a charm.

However, the new kernel I've built + rootfs combination (Squeeze) doesn't list eth0 anymore when I do an ifconfig.

Do I need to load a specific kernel module or force a special driver to be included in the kernel?

Hmmm. So for the linux kernel source I'm using the omap branch. your instructions lists 2 kernel sources: Linus' and omap. which one do you use as in the setup process we need to change the config to refer to that directory.

Actually, the Kernel fetching / building process is not clear and that may be the cause of the issue I have.

For U-Boot, if someone follows the instructions as is everything works fine.

Now for the Kernel it is a different story. I if clone you project, checkout the branch and launch build_kernel.sh I get the following error: need to copy system.sh and configure it.

From that file I concluded I had to at least clone the kernel project and update the path in the system.sh file. There are however 3 references to the kernel at the top of your page so it is not clear which one we should clone the kernel from (if it matters). Also the links aren't git:// urls so someone must be minimally knowledgeable to understand how to get that url from the web page you reference.

So I think it would be beneficial if you could provide the exact instructions as for u-boot about how to clone the right kernel, modify the system.sh file before someone attempts to launch a kernel build.

Hope this will solve my problem and help others.

Here is what I'm trying after trashing my previous kernel:git clone git://github.com/RobertCNelson/linux-dev.gitcd linux-dev/
git checkout origin/am33x-v3.2 -b am33x-v3.2

cd ..; git clone git://arago-project.org/git/projects/linux-am33x.git

cd linux-dev; cp system.sh.sample system.sh

modify LINUX_GIT=~/linux-stable/ in system.sh to point to the location of the linux-am3x directory

./build_kernel.sh

Thanks for a great job and clearer than usual instructions on setting things up.  I've got an A6 Beaglebone running your Ubuntu 12.04 system with kernel 3.2.23-psp18, everything seems to work.

But git protocol is blocked here so I'm having a hard time figuring out how to get the kernel source/headers so I can cross compile a driver module to install on the BeagleBone.

Are there any alternatives?

For my purposes I don't much care if I run Angstrom 3.2.25#1 or Ubuntu 12.04 3.2.23-psp18 as with both systems I've been able to compile local sample programs and run cross-compiled sample programs transfered via USB stick.

I thought I could compile my module on the BeagleBone out of tree if I could get the kernel headers,  using the procedure described here (stackoverflow.com/questions/4025320/building-out-of-tree-kernel-modules-depmod-and-why-reboot) but:

sudo apt-get install kernel-headers-$(uname -r)

didn't work: Unable to locate package kernel-headers-3.2.23-psp18

Hi Robert,

when I used "make menuconfig" to enable "device drivers->usb support-> USB Gadget Support-> USB Gadget Drivers -> Compile HID Gadget as Module", then running with "make"  would give me error as

CC [M]  drivers/mtd/nand/nand_base.o
drivers/mtd/nand/nand_base.c: In function ‘nand_command_lp’:
drivers/mtd/nand/nand_base.c:728:3: error: implicit declaration of function ‘dmb’ [-Werror=implicit-function-declaration]
cc1: some warnings being treated as errors
make[3]: *** [drivers/mtd/nand/nand_base.o] Error 1
make[2]: *** [drivers/mtd/nand] Error 2
make[1]: *** [drivers/mtd] Error 2
make: *** [drivers] Error 2

later on, then I fix the problem by defining "#define dmb() _asm_ _volatile_ ("" : : : "memory")" into nand_base.c and make again, it gave me error as

  LD      arch/x86/built-in.o
  CHK     kernel/config_data.h
  CC [M]  drivers/usb/musb/musb_procfs.o
drivers/usb/musb/musb_procfs.c:39:35: fatal error: mach/hardware.h: No such file or directory
compilation terminated.

so I have to remove the line "include <mach/hardware.h>" from musb_procfs.c. But make again, gave me

LD      .tmp_vmlinux1
drivers/built-in.o: In function `d_can_plat_probe':
d_can_platform.c:(.devinit.text+0x4324): undefined reference to `clk_get'
d_can_platform.c:(.devinit.text+0x444c): undefined reference to `clk_get_rate'
d_can_platform.c:(.devinit.text+0x44fd): undefined reference to `clk_put'
make: *** [.tmp_vmlinux1] Error 1

if you can help me with this would be much appreciated.

I am using ubuntu Precise on a virtual box.

Cheers,

Hi Robert,

I got the compilation problem fixed. But now, I changed somthing on the "board-am335xevm.c" file. However, everytime when I run the "build_kernel.sh", my changes would get lost. Is there any way that I can put my change into a patch file and get applied everything when running the "build_kernel.sh".

Please note, you have a patch file to "board-am335xevm.c" already, how can I apply your patch first, then follow with my one?

Cheers,

Hi Robert,

How can I patch this Kernel to Real-Time. If I patch same Kernel to Real-Time and then use your scripts "build_kernel.sh", support RT is disappear. How I must it properly?

Or can you tell me how to compile kernel with RT patch on BeagleBone (with am3359)?

Hi Robert,

when i switch on FULL RT the kernel will not be build. So, what can i do to get a kernel with full RT support ?

Is it possible to change your scripts to get an older version of linux with RT patch running ?

Regards,

Great guide! Worked as expected.

Thank you!

Cannot make /dev/fb1 to appear. :(

config has:

CONFIG_FB_OMAP2_NUM_FBS=3

CONFIG_FB_OMAP2_DEBUG_SUPPORT=y

CONFIG_OMAP2_DSS_DEBUG_SUPPORT=y

Kernel command line:

vram=8M omapfb.vram=0:4M,1:4M,2:4M omapfb.debug=y omapfb.test=y omapdss.debug=y

During booting:

- my display is recognized.

- NO omapfb ot omapdss debug prints.

After boot-up:

/dev/fb0 device is avaiable and I can even play videos on it using "mplayer2"

- no /dev/fb1, /dev/fb2 devices.

How to make /dev/fb1 or /dev/fb1 to appear?

Hi Robert,

I am trying to build a kernel with Xenomai real time support. Compiling the kernel using your procedure works on my virtual machine, however I have a couple of (newbie/simple) questions.

1) The resulting build is of a 3.6.x kernel, is that now supported for the beaglebone (in the comments you mention that only 3.2.x is currently supported)?

2) I need to compile a 3.2.23 kernel since that is the latest version supported by Xenomai. I am unsure exactly how to do that since the build_kernel.sh script seems to update to the latest version automatically. How can I check out that particular version via git and how to prevent the script from updating it?

3) Am I correct in thinking that after the initial build_kernel run, I need to apply the Xenomai patch to the kernel source files located in /linux-dev/KERNEL and then run the rebuild.sh script?

Thanks!

Did anybody try to get the RT_PREEMPT patches working on the beaglebone kernel?

looking at http://www.kernel.org/pub/linux/kernel/projects/rt/3.6/ I think the patch for v3.6.7 is the closest match

now in  https://github.com/RobertCNelson/linux-dev/commits/am33x-v3.6&nbsp; we have am33x-v3.6 which I think stands at  base v3.6.2

how should I go about it - take am33x-v3.6, apply v3.6.2...v3.6.7 and then the RT_PREEMPT 3.6.7 patch?

thanks

- Michael

Hi Robert,

apparently you are maintaining different kernels: 3.2, 3.6, 3.7 and also 3.8, since you said that all the developers are moving there.

Just curious: do you have any priorities in terms of cape support ? 

We know that the most complete is the 3.2, but which one is going to be, in you opinion the next kernel which will have a decent cape support?

Davide

Hi Robert,

apparently I had a problem with u-boot but it was my fault. never mind

Davide

I have compiled a 3.2.33 kernel and u-boot and installed the kernel plus modules on a armel-rootfs-201301141020.tar SD card, but the boot hangs after mounting the rootfs partition.

I have used a ext3 file system, but the result is identical with ext4 (the default).

I have also tried to use the other two rootfs but with no success, they also hang during boot.

My Beaglebone has a Breadboard attached and it boots fine from the same SD card if I download the standard images from www angstrom distribution org.

Hi -

I'm trying to get my beaglebone patched with xenomai - however I seem to have some issues with the versions that match.

Any ideas?

Just an update - I've patched the linux 3.2.21 kernel with xenomai. See the following link for (hopefully) thorough details:

http://yapatel.org/wiki/index.php/Installing_Xenomai_on_a_beaglebone

Hi Robert,

I need a couple of helps on the u-boot version 2013.0.0.1.

1) I would like to enable fatwrite, so I defined CONFIG_FAT_WRITE into file am335x_emv.h and rebuild u-boot. Then I added:

to the uEnv.txt file and reboot the machine, I was expecting a file fatwr would be created at the time of bootup, but it doesn't. Is there anything wrong?

2) I have a DVI-D expansion board with my beaglebone. How can I enable u-boot messages being displayed on the monitor that attached to the DVI expansion board?

I had removed definition CONFIG_SYS_CONSOLE_INFO_QUIET from am335x_evm.h, but there is still nothing come up until Linux was up.

Cheers,

Hello Robert,

I am facing issues in booting up the linux kernel v3.8 on my beaglebone setup (Boardname: A335BONE, Board Rev: A6).

I followed the steps mentioned in this link to download and build the uboot, kernel, ubuntu RFS. I created uEnv.txt and copied the uEnv.txt v3.6.x device tree based bootscript into it.

When i power up the board, i get the following messages. I see my zImage and am335x-bone.dtb file placed inside /boot folder of the sd card.

Hi,

Build the kernel (and u-boot) for an vanilla beagle bone and getting this:

reading uEnv.txt
812 bytes read in 4 ms (198.2 KiB/s)
Loaded environment from uEnv.txt
Importing environment from mmc ...
reading zImage
3421280 bytes read in 396 ms (8.2 MiB/s)
reading /dtbs/am335x-bonelt.dtb
20443 bytes read in 10 ms (1.9 MiB/s)
Booting from mmc ...
## Flattened Device Tree blob at 815f0000
   Booting using the fdt blob at 0x815f0000
   Loading Device Tree to 8fe3a000, end 8fe41fda ... OK
Starting kernel ...

Then nothing.

Tried with both am335x-bone and bonelt. Having a real hard time finding info on the differences between these dtbs. Tried on a rev A01 and rev A06, with same results.

Any ideas on how to debug why the kernel hangs??

I should mention that I turned modules off and removed the keyword "modules" target from the build script. But I'm pretty sure I built-in all relevant hardware drivers before un-selecting modules in menuconfig (I'm old fashioned, like to have a kernel with all necessary hardware support and no modules). I suppose the next step would be to go back to the default config (which drives me insane because it compiles as modules just about everything on the face of the planet). Do you need an initrd to make it work with the defconfig+modules (i.e.: are there any essential drivers compiled as modules instead of linked in with the default config?)

Appreciate any input.

Thanks,
Gilles
.

PS: Built from 3.7.10-x9 tag

PS2: Built the rootfs with buildboot but I know it's good as it worked with the (older) kernel built by buildroot (besides, we're not getting anywhere near loading the rootfs). BTW: no blinking LEDs and the console is directed to ttyO0 as it should.

Hello Robert,

thank you for your Git repro for beaglebone everything went fine and i installed a 3.8 kernel.

I want to use the uart2 on the beaglebone and don't now where to start. Any idea where to get an example or explanation how to modify the am335x-bone-common.dtsi . Is it the right file and is it the only one to modify to use uart2 on the expansion connector.

Thank you

Pet

Hi Robert,

stable v3.8 is not booting anymore on beaglebone. "It hangs after Uncompressing Linux... done, booting the kernel." ?? 2 days a go everything went fine.

Regards,

Peter

Hi Robert,

I am trying the freshly baked RT Preempt Patch for kernel 3.8.4 and it works (apparently). On the other hand, I am not able to load my can bus on the new kernel 3.8 (I am using Wheezy).

I used to follow the instruction I found on internet, i.e. commands like: 

ip link set can0 up type can bitrate 125000

gets the answer 

Cannot find device "can0".

But if I use the command

 ip link set can up type can

I get something more juicy:

[   51.518629] can: controller area network core (rev 20120528 abi 9)
[   51.529052] NET: Registered protocol family 29
[   51.538679] Loading kernel module for a network device with CAP_SYS_MODULE (deprecated).  Use CAP_NET_ADMIN and alias netdev-can instead.
Cannot find device "can"

Any clue?

Hi Robert,

Congratulations for this fantastic content. I went through the steps using the Linux Build Script (v3.8.x) and the Ubuntu 12.10 as the FS. I did also created the microSD card. Everything worked great so i decided to set up an NFS server to load the kernel and the FS over the network to ease the development (u-boot and other bootloading files are stored on the SD since they won't change for now).

So it was time to get my hands on the kernel hacking issue, i checked the Kernel Configuration under:

And i saw that the following:

Three boards seem to be selected: OMAP3 BEAGLE board, IGEP v2 board and OMAP4 Panda Board. I need to know which code is being executed so i decided to put printks on the three of them to figure out which one was being executed. Well i did it but nothing, the board boots exactly as before, no changes (i'm rebuilding using the tools/rebuild.sh). I thought it was a bit weird so i decided to unset the three boards and see if there was something that i was missing. Once the three boards were deselected i was hoping the the kernel wouldn't boot but surprisingly it does so.

I read that previously somebody mentioned the "board-am335xevm.c" but that's not on my kernel tree (3.8.4). Would you mind telling me if i'm missing something? Which board configuration file is loading?

Regards,

Hi Robert,

Thank you for your scripts. I have compiled and successfully installed kernel 3.8.

As i can see, the cape manager use capes / slots mechanism, so registering a platform device such as tscadc is impossible without standard cape with eeprom on i2c. Would you recommend a way for breaking this mechanism so i can use the touchscreen functionality with custom cape (without eeprom) with display and touchscreen.

Regards,

Thank you very much for the link, it really helped.

Thanks,

Mahanth

Hi Robert,

Can two USB gadget modules be loaded simultaneously? For example, running command "sudo insmod g_hid.ko"  and then followed with "sudo insmod g_ether.ko".

Regards.

Damien

Hi Robert & everybody

I do have a question:

-I do have an BeagleBone for which I generated using Yocto Project and rootfs (kernel 3.8.5)

-my "distribution" is working fine

In my plan is to buy from CircuitCo and LCD4.

For this LCD I don't know what modification I have to do in the kernel.

I have to use an driver and I don't know which one, what are the option from kernel point of view, etc

Can you be so king and guide me ?

Thank you in advance and all the best

Mihai M

I just finished following the instructions to set up an SD card building: kernel_version=3.8.6-bone12.1 with Ubuntu 12.10 rootfs, went as smooth as could be expected.

My only issues with the instructions were that the fdisk EOF steps EOF didn't work with cut and paste (unlike all the other "code" sections) so I did them manually to create the two partitions, not clear at all what I was supposed to do here.  Also the section about shared SD card To always enable the Ethernet interface as eth0.. made no sense, so I skipped it, perhaps a link or an explanation as to what this is about would be helpful.