Re: Building the 2.6.17 kernel

From: Rtp <arnaud.patard_at_rtp-net.org>
Date: Wed, 13 Dec 2006 09:11:13 +0100

tmtalpey_at_rcn.com writes:

Hi !

> I've been running the 2.6.17 kernel from the project downloads
> on my h1940 with pretty good success. (Thanks!) Mostly I'm
> using GPE, because it seems to support the environment a little
> better (for example, the buttons mostly work), but frankly I like
> Opie better because it's smaller and faster.
>
> Anyway, I am trying to customize the kernel and running into
> problems. After patching and then using the config file from
> <http://anymore.nl/ipaq/files/kernels/2.6.17-1/defconfig> and
> the arm toolchain from handhelds.org, I get a link error on the
> kernel build. It complains about a relocation error in fbcon.c
> If I switch to the codesourcery toolchain, it fails to link with a
> very odd reference to a "discarded" eth_unbind routine.

It's not very odd reference. The eth_unbind method on 2.6.17 in
marked as __exit and it's called from eth_bind which is marked as
__init, causing build failure. You need this patch :
http://git.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=a353678d3136306c1d00f0d2319de1dac8a6b1db

>
> The second one is avoidable by turning off the USB ethernet
> gadget, then I get a good kernel that boots fine and works well
> with GPE, except that the bluetooth is nonfunctional. Not useful,
> since I'm building a bluetooth app!

Are you using a home made image ?
If so, you need to enable the bt chip with :

echo 1>/sys/bus/platform/devices/h1940-bt/enable

and then to use hciattach to get the hci device.

>
> I am assuming this is pilot error somehow, but I'm not seeing it.
> Any suggestions? Thanks!
>
> BTW, do you need a developer? I'm a volunteer, especially for
> getting suspend/resume working, and/or a bootloader. Battery

We got our first suspend/resume last week :)
Stay tuned !

> reading would be nice too. Right now I'm trying to figure out why

We know how to read the battery too.

> the RTC isn't being probed. It seems to successfully make it to

Check that the rtc device is listed in the device list in mach-h1940.c
(iirc it should be something like ' &s3c_device_rtc,')

> platform_driver_register() but the probe routine doesn't get called
> back. As a result it always comes up in March 2006. Known issue?
>
> Oh, I also have minicom 2.2 built and working. Might be a worthwhile
> addition to the packages.

Ask Arjan but you have to wait for him. He's atm on vacations afaik.

Arnaud
Received on Wed Dec 13 2006 - 03:38:47 EST

This archive was generated by hypermail 2.2.0 : Wed Dec 13 2006 - 03:39:04 EST