Kernels fail to boot on Rogers HTC Dream, but works on other Dream platforms (ADP1, T-Mo G1)

[David B]

The only difference between the kernel config is that the one that
comes on the Rogers HTC Dream has the following options:

CONFIG_MSM_AMSS_SUPPORT_256MB_EBI1=y
CONFIG_CPU_FREQ_GOV_MSM7K=y

Those options don't seem to be in the public Android kernel repo,
there's no mention of either option on Google Code, and so it seems
like there's a patch HTC or Google is applying to get kernels to work
on the later HTC Dream hardware. I already know the boot locations
have changed on the device (yes, I'm accounting for that).

Does anybody have any working kernel source tree for the Rogers HTC
Dream platform, or if it is public, can I be pointed in the right
direction toward it?

[cyanogen]

I contacted HTC about this, and they told me to ask Google :)

Can anyone help? The Rogers device has an engineering bootloader, so
it would make a great development platform.

[PhaseBurn]

I, too, contacted HTC, and got the same reply... Hence, I posted here
(after checking out the public sources, to see if it was hidden
somewhere else)...

[Julien L]

CONFIG_CPU_FREQ_GOV_MSM7K=y is part of the android-msm-htc-2.6.25
branch. (looky here:
http://android.git.kernel.org/?p=kernel/msm.git;a=tree;f=drivers/cpufreq;h=da45b90bdc34cfa89db503492ef5cbde6c19b41b;hb=refs/heads/android-msm-htc-2.6.25)

Since the Rogers device's Kernel is 2.6.27 it's easy to assume it
comes from some kind of android-msm-htc-2.6.27 branch that is not
public.
The msm7k cpu_freq was comitted in by Mike A. Chan from Google, not
HTC however this cpu_freq does not appear in any of the android-msm
branches other than this one...

[cyanogen]

So what's the story here? I think some of the new HTC Magic devices
are in the same boat as the Rogers Dream. What is so different about
this hardware, and is there something missing from the Android source
tree that we need for these devices? Anyone?

[gboddina]

Any news?

[PhaseBurn]

HTC's tech support has been very unforthcoming (the Legal agreement
for the /kernel file says to contact customer service for source
code), so I've been in contact with an HTC VP, who has gotten back to
me basically asking me to wait until Tuesday the 28th for a further
update. What that will be, I'm not sure just yet, but I'll be patient.
Technical support has all but admitting to having it, and has
specifically stated "We have no way to distribute it at this time" so
they're waiting on corporate to make a few decisions I guess. I'm not
sure.

I'll provide updates as I get 'em but right now, while I'm exploring a
few potential avenues that look promising, I still have no news worth
mentioning (last thing I want to do is provide false hope if these
avenues all don't work out)

[Shwan Ciyako]

This is the email I got when I asked them have you contacted the same
number as in this mail ?
"
Thank you for contacting HTC Europe.
We are just an end user support and do not have the ability to handle
developer related questions.
I would either advice you to contact developer.android.com or our
Headquarters in Taiwan for developer related questions: HTC
Headquarters

23 Xinghua Rd., Taoyuan 330,
Taiwan, R. O. C.
Tel: +886-3-3753252

Kind Regards
Dave Montanya
HTC Europe

[Julien L]

Actually HTC Americas was contacted and now they themselves have
contacted HTC Taiwan who then let Phaseburn know that they are working
on making the sources available, more info to come.

[Wysie]

Hi all, are there any updates with regards to this? The last time I
read on xda-developers some HTC rep says the sources will be out by
end July, which obviously didn't happen :(.

On Jul 31, 10:21 am, Julien L <hellfe...@gmail.com> wrote:
> ActuallyHTCAmericas was contacted and now they themselves have
> contactedHTCTaiwan who then let Phaseburn know that they are working

[Bart Trojanowski]

Wysie,

My understanding is that the Rogers Dream boot.img need to be built
with a kernel at a different offset.

I think there is a patch in Donut that allows for mkbootimg utility to
take this as a parameter. After that you should be able to modify
build/core/Makefile to send it kernel offset of 0x19200000. I am not
sure if the kernel itself needs to be told of this... I didn't find
anything about that, so maybe not

I have it on my TODO list to figure this out, but I hadn't found the
time yet. Once I do, I'll be sure to blog it and try to remember to
post it here. Please keep me posted if beat me to it. :)

-Bart

--
http://www.jukie.net/~bart/sig