Today a Fedora 13 ARM Beta was released. This makes it obviously time for me to blog about my Fedora adventures from last weekend on my loaned BeagleBoard.
I managed to get kernel 2.6.38-rc3 compiled on a virtualized Qemu ARM system. Compiling the kernel really takes some time. A stripped down version, without compiling loadable modules took just over 880 minutes. An other test with
Booting the kernel, made my serial console unusable. A bug and patch has been filed for
With the new beta release, I would need to rebuild
Furthermore I found out how to boot the board in a more useful state than before. Passing the
With this image, my Fedora 12 boots perfectly when setting the
Some of these steps took quite a bit to figure out, but it looks like it's going to work a little bit more everytime.
I managed to get kernel 2.6.38-rc3 compiled on a virtualized Qemu ARM system. Compiling the kernel really takes some time. A stripped down version, without compiling loadable modules took just over 880 minutes. An other test with
make omap2plus_defconfig uImage took around 1400 minutes. I now believe the people that mentioned that the performance of a virtual system (on x86_64) is not very spectacular.Booting the kernel, made my serial console unusable. A bug and patch has been filed for
initscripts in Fedora 14 and below. Fedora 15 uses systemd and is likely not affected with the same. This kernel (well, actually from 2.6.36 an onwards) have a optimized serial-port implementation for OMAP-boards. This unfortunately does not use ttyS* for serial-ports, but ttyO*. Any application (including initscripts) expecting that serial-ports are starting with ttyS or other familiar identifiers, will need an update for the OMAP serial-port driver.With the new beta release, I would need to rebuild
initscripts with the patch that makes ttyO work as main console (giving console=ttyO2,115200n8 on the kernel command line). When all works fine, the boot process will be displayed on the serial-console and logging in works again.Furthermore I found out how to boot the board in a more useful state than before. Passing the
initramfs via u-boot does not seem to be very common, most sites advise to compile the initramfs into the kernel. I'd prefer to keep the standard Fedora way, and that means keeping kernel and initramfs seperate. It seems that u-boot needs a special kind of initramfs/initrd (possibly just a header added). This format can be created like the following:mkimage -A arm -O linux -T ramdisk -C none -a 0x00000000 -e 0x00000000 -n initramfs -d initramfs-2.6.38-rc3.img uinitramfs-2.6.38-rc3.imgWith this image, my Fedora 12 boots perfectly when setting the
bootcmd to load both kernle + initramfs and passing both addresses via bootmOMAP3 beagleboard.org # printenv
stdin=serial
stdout=serial
stderr=serial
serial=3a5c00020000000004013ef109010017
bootargs=console=ttyO2,115200n8 root=/dev/mmcblk0p2 ro
bootcmd=mmcinit ; fatload mmc 0:1 0x80000000 uimage-2.6.38-rc3.beagleboard_201102131201 ; fatload mmc 0:1 0x81600000 uinitramfs-2.6.38-rc3.img ; bootm 0x80000000 0x81600000
Environment size: 309/131068 bytesSome of these steps took quite a bit to figure out, but it looks like it's going to work a little bit more everytime.
