I have made further updates to the ARM Primary Feature page and think it
is now good enough for FESCO. If you would like to give it a look
before it is reviewed, now is the time.
Brendan Conoboy / Red Hat, Inc. / blc(a)redhat.com
Hi ARM list!
I am working with Peter on putting the final touches on the OS that
goes on these 60K laptops . The OS is based on F-14 armv5tel. Old,
I know, but we'll release a build of F17 mid-2012 to attune for our
outmoded ways :-)
One of the side projects we have is to make Skype run on these
XO-1.75s, using SkypeKit  (they give you the runtime, and a
unixsockets API, you build the UI). The SDK contains 3 binaries:
ARMv5, ARMv6, ARMv7.
So on F14-armv5tel, I am trying to run the ARMv5 binary. It starts up
and remains spinning.
On a pre-alpha build of F17-armv7hl, the ARMv5 binary runs (buggy);
the ARMv7 binary spins doing nothing.
I am still debugging this, but I strongly suspect an EABI or compiler
options mismatch. Is there any reasonable approach to test those two
Thanks, and apologies for the off-topic post,
0 - http://blog.laptop.org/2012/03/16/uruguay-is-first-country-to-get-xo-1-75/
1 - http://developer.skype.com/public/skypekit
martin(a)laptop.org -- Software Architect - OLPC
- ask interesting questions
- don't get distracted with shiny stuff - working code first
I am having trouble booting a Fedora-15 ARM userspace on a pico-SAM9G45 with a custom-compiled kernel.
I am able to cross-compile a vendor-supplied kernel source with support for fbcon to see what I am doing. The boot microSD has a FAT boot partition with an uImage and a ext3 partition with Fedora 15. The ext3 filesystem previously booted correctly on a
qemu-system-arm emulation with -M versatilepb . On the LCD I see the penguin logo and the ordinary kernel messages, but stops right after "Freeing init memory: 140K" Then I have no output from userspace.
I can plug in an USB keyboard and I see kernel messages. I can even scroll up and down on the LCD output using the USB kernel. That shows that the kernel is not dead. So I suspect I am having trouble running /sbin/init .However, I see no output that could
point me to what exactly is wrong. Nothing like a SIGSEGV or an incompatibility message. The root filesystem is being mounted, because I see "VFS: Mounted root (ext3 filesystem) on device 179:2"
I have tried adding initcall_debug to the kernel command line but it has no effect in my case.
The kernel is being booted without an initrd. At first I thought it could be the issue, but I tried replicating the problem on a x86 PC (removing the initrd directive from GRUB and adding root=/dev/sdb2) and the system does boot to completion.
I also though maybe the Fedora 15 init requires /dev/ to be populated, so I checked. It has /dev/null , /dev/console, and /dev/zero. I tried adding /dev/mmcblk0, /dev/mmcblk0p1 and /dev/mmcblk0p2 but it also has no effect.
I searched in Google and found a post that described a problem similar to my own, and resolved it to an OABI/EABI incompatibility. However, the .config I used for the kernel has CONFIG_AEABI=y and CONFIG_OABI_COMPAT=y , just the same as
config-188.8.131.52-1.fc15.armv5tel being used to compare.
The config file used to compile the kernel is attached.
How does one change the kernel that uboot loads from mtdflash and
instead boots a kernel from an ext filesystem on an attached (uboot
supported) usb flash drive? Furthermore, is there a way to get uboot
to also load an initial ram file (disk or fs)? The hardware in question
is a guru plug and I've got the jtag and console adapter.
Thanks in advance,
Red Hat, Inc
"It's not whether you get knocked down, it's whether you get up." - Vince Lombardi