Re: [android-internals] Re: Android on Nokia n810 (OMAP2420)

Re: [android-internals] Re: Android on Nokia n810 (OMAP2420)

[Nishanth Menon]

Looping in OMAP ML to the good news..
Congrats!
sim0nx said the following on 04/05/2008 06:35 AM:
> I can't believe it !!!
> Android runs on the maemo n810 !!!
>
> After reading this:
> http://groups.google.lu/group/android-internals/browse_thread/thread/93570c41bce07f16?hl=en
> for the 100th time, and following the instructions on the mentioned
> link:
> http://nemustech.blogspot.com/2007/12/android-porting-to-real-target-hw.html
> it works.
>
> Here are the specifics I've implemented:
> * patch /etc/init.rc
> ---- begin : init.rc
> ....
> ## qemu-init {
> ## exec /etc/qemu-init.sh
> ## }
> network-property {
> exec /etc/set_network.sh
> }
>
>
> /bin/chmod -R a+rw /data /tmp
> /bin/chmod a+rw .
> /bin/chmod -R a+rw data
>
> # rm /dev/log (if you have)
> # mkdir /dev/log;touch /dev/log/{main,events,radio}
> * you can see usefull infos from strace log and /dev/log/main
>
> and instead of:
> chroot . /a.sh
> I execute the android init process:
> chroot . /init
>
> Not sure yet what exactly of those steps solved the problem (maybe the
> qemu one), but it does work.
> After a quick try I see that the touchscreen doesn't work (the
> keyboard does though), but that shouldn't be too hard to fix.
>
> BTW, this is using an m5 kernel, and binder.c
> --~--~---------~--~----~------------~-------~--~----~
> You received this message because you are subscribed to the Google Groups "Android Internals" group.
> To post to this group, send email to android-internals@googlegroups.com
> To unsubscribe from this group, send email to android-internals-unsubscribe@googlegroups.com
> For more options, visit this group at http://groups.google.com/group/android-internals?hl=en
> -~----------~----~----~----~------~----~------~--~---
>
>
>   

Re: [android-internals] Re: Android on Nokia n810 (OMAP2420)

[Dirk Behme]

Nishanth Menon wrote:
> Looping in OMAP ML to the good news..
> Congrats!
> sim0nx said the following on 04/05/2008 06:35 AM:
> 
>>I can't believe it !!!
>>Android runs on the maemo n810 !!!

Great!

Do you have some screenshots we can add to

http://elinux.org/Android_on_OMAP#Screenshots

?

Anything else we should add/correct at

http://elinux.org/Android_on_OMAP

?

Thanks

Dirk

Btw: Do you have any power management issues like

http://elinux.org/Android_on_OMAP#Power_management

? We still don't know anything about them :(

>>After reading this:
>>http://groups.google.lu/group/android-internals/browse_thread/thread/93570c41bce07f16?hl=en
>>for the 100th time, and following the instructions on the mentioned
>>link:
>>http://nemustech.blogspot.com/2007/12/android-porting-to-real-target-hw.html
>>it works.
>>
>>Here are the specifics I've implemented:
>>* patch /etc/init.rc
>>---- begin : init.rc
>>....
>>## qemu-init {
>>## exec /etc/qemu-init.sh
>>## }
>>network-property {
>>exec /etc/set_network.sh
>>}
>>
>>
>>/bin/chmod -R a+rw /data /tmp
>>/bin/chmod a+rw .
>>/bin/chmod -R a+rw data
>>
>># rm /dev/log (if you have)
>># mkdir /dev/log;touch /dev/log/{main,events,radio}
>>* you can see usefull infos from strace log and /dev/log/main
>>
>>and instead of:
>>chroot . /a.sh
>>I execute the android init process:
>>chroot . /init
>>
>>Not sure yet what exactly of those steps solved the problem (maybe the
>>qemu one), but it does work.
>>After a quick try I see that the touchscreen doesn't work (the
>>keyboard does though), but that shouldn't be too hard to fix.
>>
>>BTW, this is using an m5 kernel, and binder.c
>>--~--~---------~--~----~------------~-------~--~----~
>>You received this message because you are subscribed to the Google Groups "Android Internals" group.
>>To post to this group, send email to android-internals@googlegroups.com
>>To unsubscribe from this group, send email to android-internals-unsubscribe@googlegroups.com
>>For more options, visit this group at http://groups.google.com/group/android-internals?hl=en
>>-~----------~----~----~----~------~----~------~--~---
>>
>>
>>  
> 
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-omap" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 

Re: [android-internals] Re: Android on Nokia n810 (OMAP2420)

[Georges Toth]

Hi,
> Great!
> Do you have some screenshots we can add to
> http://elinux.org/Android_on_OMAP#Screenshots
> ?
I've added 3 screenshots on that page :-)

> Anything else we should add/correct at
> http://elinux.org/Android_on_OMAP
> ?
I wanted to add some instructions for the maemo once I got it fully 
working :-).
Basically the differences from the wiki I had to use are:
- m3 or m5 kernel, but using binder.c in any case (not the original 
binder code from m3)
- backport the android patch, to the nokia kernel sources (2.6.21)
- use the android init instead of the mentioned a.sh script
- edit the init configuration and comment the qemu parts
- only the m3 userspace files currently work

The touchscreen doesn't work yet, but I hope we can figure that out too.
-> The driver works, there is output when one does a cat on 
/dev/input/event3, but android doesn't recognize it.

Some buttons don't work ... wrong mapping.
If anybody has an idea how to see what keycodes certain buttons are 
bound to and how to edit the mapping in android that would be great :-).


> Btw: Do you have any power management issues like
> http://elinux.org/Android_on_OMAP#Power_management
> ? We still don't know anything about them :(
No. It's just that the display rather quickly dims and shuts down (2-4s 
to dim, and another 2 secs to shutdown).
It doesn't work to shut the display down via the android menu, and 
android doesn't display the battery status.

Besides that there are no problems.


-- 
regards,

Georges Toth

Re: [android-internals] Re: Android on Nokia n810 (OMAP2420)

[Brian Swetland]

[Georges Toth <georges.toth@gmail.com>]
>
> The touchscreen doesn't work yet, but I hope we can figure that out too.
> -> The driver works, there is output when one does a cat on 
> /dev/input/event3, but android doesn't recognize it.
>
> Some buttons don't work ... wrong mapping.
> If anybody has an idea how to see what keycodes certain buttons are bound 
> to and how to edit the mapping in android that would be great :-).

Here's a brute force approach to sorting out input events: run getevent
on the emulator and on the target hardware and compare the results.  It's
in /system/bin.

I believe the touchpad issue has to do with sending events that the
system recognizes as "pen down" and "pen up" events.  Haven't looked at
this in a while.

Keylayouts live in /system/usr/keylayout/*.kl and are used to translate
from the raw input event codes to android keycodes.

Keymaps live in /system/user/keychars/*.kcm.bin (undocumented binary
format right now, sorry) and are used to describe how the key events
and modifiers and such are related.

Brian

Re: [android-internals] Re: Android on Nokia n810 (OMAP2420)

[Georges Toth]

Brian Swetland wrote:
> [Georges Toth <georges.toth@gmail.com>]
>   
>> The touchscreen doesn't work yet, but I hope we can figure that out too.
>> -> The driver works, there is output when one does a cat on 
>> /dev/input/event3, but android doesn't recognize it.
>>
>> Some buttons don't work ... wrong mapping.
>> If anybody has an idea how to see what keycodes certain buttons are bound 
>> to and how to edit the mapping in android that would be great :-).
>>     
>
> Here's a brute force approach to sorting out input events: run getevent
> on the emulator and on the target hardware and compare the results.  It's
> in /system/bin.
>
> I believe the touchpad issue has to do with sending events that the
> system recognizes as "pen down" and "pen up" events.  Haven't looked at
> this in a while.
>   
Thanks for the tip, I'll have a look at that.

> Keylayouts live in /system/usr/keylayout/*.kl and are used to translate
> from the raw input event codes to android keycodes.
>
> Keymaps live in /system/user/keychars/*.kcm.bin (undocumented binary
> format right now, sorry) and are used to describe how the key events
> and modifiers and such are related.
>   
I would need to add/change the event code for the DPAD_CENTER event.
Do you perhaps know of a simple way of seeing what code a certain button 
generates ?


-- 
regards,

Georges Toth

Re: [android-internals] Re: Android on Nokia n810 (OMAP2420)

[Brian Swetland]

Congrats.

Has anyone had any luck getting a 2.6.23 or 2.6.24 kernel working
with the 800 or 810?  When I was poking around with it a while back,
I had to backport our stuff to .21 due to binary-only wifi drivers
being incompatible with newer kernels.

Also, if anyone knows of work done to figure out how the battery
charge control, etc works on the 800/810 hardware, I'm highly curious.  
Have not had the spare time to reverse engineer that stuff and if
somebody's already done the hard work, so much the better.

Brian

[Nishanth Menon <menon.nishanth@gmail.com>]
> Looping in OMAP ML to the good news..
> Congrats!
> sim0nx said the following on 04/05/2008 06:35 AM:
> > I can't believe it !!!
> > Android runs on the maemo n810 !!!
> >
> > After reading this:
> > http://groups.google.lu/group/android-internals/browse_thread/thread/93570c41bce07f16?hl=en
> > for the 100th time, and following the instructions on the mentioned
> > link:
> > http://nemustech.blogspot.com/2007/12/android-porting-to-real-target-hw.html
> > it works.

Re: [android-internals] Re: Android on Nokia n810 (OMAP2420)

[Georges Toth]

> Congrats.
>   
Thnx :-)

> Has anyone had any luck getting a 2.6.23 or 2.6.24 kernel working
> with the 800 or 810?  When I was poking around with it a while back,
> I had to backport our stuff to .21 due to binary-only wifi drivers
> being incompatible with newer kernels.
>   
AFAIK, no.

I've just searched through some mailinglists and apparently it is 
possible to use a more recent kernel with the n800/n810.
You basically compile a module called cx3110x which is open source [1], 
and force to load a binary-only module (umac); in order to get wifi to work.

I'll probably look into this over the next few days as time permits.


[1]: http://repository.maemo.org/pool/chinook/free/c/cx3110x-module-src/

> Also, if anyone knows of work done to figure out how the battery
> charge control, etc works on the 800/810 hardware, I'm highly curious.  
> Have not had the spare time to reverse engineer that stuff and if
> somebody's already done the hard work, so much the better.
I've only started to work on this last week, but AFAIK the only work that has been done on the n800/n810 so far is to actually get the android patches to work and get it to run.

Next would be to try to get the framebuffer driver hacked to work with the m5 image ... the framebuffer driver doesn't support double buffering thus we only get a black screen :-(

Sebastian (android-internals ml) is currently looking into that.

And of course the touchscreen and buttons issue.

-- 
regards,

Georges Toth

Re: [android-internals] Re: Android on Nokia n810 (OMAP2420)

[andrzej zaborowski]

Hi,

On 06/04/2008, Georges Toth <georges.toth@gmail.com> wrote:
>
>
> > Congrats.
> >
> >
>  Thnx :-)
>
>
> > Has anyone had any luck getting a 2.6.23 or 2.6.24 kernel working
> > with the 800 or 810?  When I was poking around with it a while back,
> > I had to backport our stuff to .21 due to binary-only wifi drivers
> > being incompatible with newer kernels.
> >
> >
>  AFAIK, no.
>
>  I've just searched through some mailinglists and apparently it is possible
> to use a more recent kernel with the n800/n810.
>  You basically compile a module called cx3110x which is open source [1], and
> force to load a binary-only module (umac); in order to get wifi to work.
>
>  I'll probably look into this over the next few days as time permits.
>
>
>  [1]:
> http://repository.maemo.org/pool/chinook/free/c/cx3110x-module-src/
>
>
> > Also, if anyone knows of work done to figure out how the battery
> > charge control, etc works on the 800/810 hardware, I'm highly curious.
> Have not had the spare time to reverse engineer that stuff and if
> > somebody's already done the hard work, so much the better.
> >
>  I've only started to work on this last week, but AFAIK the only work that
> has been done on the n800/n810 so far is to actually get the android patches
> to work and get it to run.

Poky (pokylinux.org) and mamona both work pretty well for everyday use
on the nseries.  Fortunately the userspace can partially ignore the
charging issue because the RETU (now Betty) chip takes care of that in
the hardware.  Charging / charger-cable / power-button / battery-life
states are readable from /sys and /dev.

For more advanced power management such as that in maemo (battery type
detection, battery quality, alarm wake-ups, watchdogs etc) more work
would be needed as these are done in userspace on Maemo by services
called DSME and BME and libcal (for retrieving information from OTP)
that are closed-source and I believe not redistributable separately
from Maemo. They are effectively the driver for the RETU and TAHVO
chips whose registers / interrupts are accessed through /dev/retu and
/dev/tahvo nodes.

The best bet for a distro is probably to leave the maemo "initfs",
which contains closed-source DSME / BME / flasher, in place and let
the custom rootfs run on top of it.  If you need some details about
retu / tahvo driving / bootloading, I may be able to tell because I
recently had to poke on it for a different project.

>
>  Next would be to try to get the framebuffer driver hacked to work with the
> m5 image ... the framebuffer driver doesn't support double buffering thus we
> only get a black screen :-(

I understand the userspace is going to be open-source at some point,
which would render this work a bit useless.  If it doesn't get
open-sourced then it's a waste of time either way I think?

Regards
-- 
Please do not print this email unless absolutely necessary. Spread
environmental awareness.

Re: [android-internals] Re: Android on Nokia n810 (OMAP2420)

[Georges Toth]

Hi,

>  Fortunately the userspace can partially ignore the
> charging issue because the RETU (now Betty) chip takes care of that in
> the hardware.  Charging / charger-cable / power-button / battery-life
> states are readable from /sys and /dev.
>   
Hmm where exactly can you read that info from ?
I quickly searched through sysfs, proc and dev but without much luck  :-)

> The best bet for a distro is probably to leave the maemo "initfs",
> which contains closed-source DSME / BME / flasher, in place and let
> the custom rootfs run on top of it.  If you need some details about
> retu / tahvo driving / bootloading, I may be able to tell because I
> recently had to poke on it for a different project.
>   
Should I get android to work correctly it would be nice to automatically 
start android after booting the n810.
Thus if you have any info on setting up wifi automatically from a script 
and what exactly to disable to prevent the system from starting Xomap, 
dbus, hal etc... that would be nice :-)

>>  Next would be to try to get the framebuffer driver hacked to work with the
>> m5 image ... the framebuffer driver doesn't support double buffering thus we
>> only get a black screen :-(
>>     
> I understand the userspace is going to be open-source at some point,
> which would render this work a bit useless.  If it doesn't get
> open-sourced then it's a waste of time either way I think?
>   
According to what I've read so far it will be open-sourced as soon as 
the first devices start to appear...which probably won't be anytime soon

But even then that work wouldn't be useless because Google won't release 
a framebuffer driver for Nokia n800/810 ... so if one wants to be able 
to run m5 and future releases, somebody has to modify the current driver 
to be compatible with android...if that is even possible.


-- 
regards,

Georges Toth

Re: [android-internals] Re: Android on Nokia n810 (OMAP2420)

[andrzej zaborowski]

On 06/04/2008, Georges Toth <georges.toth@gmail.com> wrote:
> >  Fortunately the userspace can partially ignore the
> > charging issue because the RETU (now Betty) chip takes care of that in
> > the hardware.  Charging / charger-cable / power-button / battery-life
> > states are readable from /sys and /dev.
> >
> >
>  Hmm where exactly can you read that info from ?
>  I quickly searched through sysfs, proc and dev but without much luck  :-)

The battery and charger are connected to the ADC in retu which has 14
channels.  I thought I saw an interface in the stock n810 kernel to
get their readings through /sys but apparently there isn't any, except
for the "Headset detection" channel, my bad.  In this case, to read
the values you need to open /dev/retu and issue RETU_IOCH_ADC_READ
ioctl (number _IO(0x60, 0x03)) with the channel number as parameter.
It returns the 12-bit ADC value. The channel names are the following:

0 - Ground
1 - BSI
2 - Battery temperature
3 - Charger voltage
4 - Headset detection
5 - Hook detection
6 - RF GP
7 - Wideband Tx detection
8 - Battery voltage
9 - Ground
10 - Light sensor
11 - Light sensor temperature
12 - Backup battery voltage
13 - RETU temperature
(the names come from the NOLO bootloader help messages)

The light sensor is actually connected to a separate ADC whose value
is available in /sys/class/hwmon/... somewhere.

When BME is running the already processed battery state can be read
from hal with:
$ hal-get-property --udi /org/freedesktop/Hal/devices/bme --key
battery.charge_level.percentage

("hal-device /org/freedesktop/Hal/devices/bme" gives more info)

>
>
> > The best bet for a distro is probably to leave the maemo "initfs",
> > which contains closed-source DSME / BME / flasher, in place and let
> > the custom rootfs run on top of it.  If you need some details about
> > retu / tahvo driving / bootloading, I may be able to tell because I
> > recently had to poke on it for a different project.
> >
> >
>  Should I get android to work correctly it would be nice to automatically
> start android after booting the n810.

Poky works that way, the initfs boots first (the stock initfs,
unmodified) and eventually chroot's into the rootfs which is a Poky
image and the system starts automatically but you still get the stock
nokia flasher and charger screen in initfs.

>  Thus if you have any info on setting up wifi automatically from a script
> and what exactly to disable to prevent the system from starting Xomap, dbus,
> hal etc... that would be nice :-)

I haven't looked into the software side much, but I'm sure that if you
replace the rootfs, Xomap won't start.  The initfs has a /linuxrc file
that can be custimsed.

>
>
> >
> > >  Next would be to try to get the framebuffer driver hacked to work with
> the
> > > m5 image ... the framebuffer driver doesn't support double buffering
> thus we
> > > only get a black screen :-(
> > >
> > >
> > I understand the userspace is going to be open-source at some point,
> > which would render this work a bit useless.  If it doesn't get
> > open-sourced then it's a waste of time either way I think?
> >
> >
>  According to what I've read so far it will be open-sourced as soon as the
> first devices start to appear...which probably won't be anytime soon
>
>  But even then that work wouldn't be useless because Google won't release a
> framebuffer driver for Nokia n800/810 ... so if one wants to be able to run
> m5 and future releases, somebody has to modify the current driver to be
> compatible with android...if that is even possible.

What I mean is that this is the kind of issue that takes probably
minutes to fix in the userspace for a person that has the sources, but
can take days to a person trying to work around the userspace bug in
the kernel, time that could be spent on actual free software,
collaborative projects.  The issue as I understand is in the userspace
which is written in a way that's not portable between framebuffers,
probably ignoring existing standards and existing libraries for
accessing framebuffers.
-- 
Please do not print this email unless absolutely necessary. Spread
environmental awareness.

Re: [android-internals] Re: Android on Nokia n810 (OMAP2420)

[Brian Swetland]

[andrzej zaborowski <balrogg@gmail.com>]
> 
> Poky (pokylinux.org) and mamona both work pretty well for everyday use
> on the nseries.  Fortunately the userspace can partially ignore the
> charging issue because the RETU (now Betty) chip takes care of that in
> the hardware.  Charging / charger-cable / power-button / battery-life
> states are readable from /sys and /dev.

Ah, interesting.  When I was looking at the nokia patches to 2.6.21
it seemed like all power control happened from the userspace side
and the retu driver just allowed userspace to read/write the chip.
Does this patches beyond kernel-source-rx-34_2.6.21.0-osso71.diff.gz?

Where would I be looking in /sys or /dev for the charger/battery
state?

Thanks,

Brian

Re: [android-internals] Re: Android on Nokia n810 (OMAP2420)

[Jarkko Nikula]

On Sun, 6 Apr 2008 06:23:23 -0700
"ext Brian Swetland" <swetland@google.com> wrote:

> Has anyone had any luck getting a 2.6.23 or 2.6.24 kernel working
> with the 800 or 810?  When I was poking around with it a while back,
> I had to backport our stuff to .21 due to binary-only wifi drivers
> being incompatible with newer kernels.
> 
I would say that N800 has the best support at the moment in linux-omap.
I think almost all the drivers are there. IIRC only temperature sensor
is missing.

N810 has unfortunately only very minimal support. But at least USB and
OneNAND are working so it can run as a block box. However there
shouldn't be any new binary stuff so in principle it should be able to
get into same shape than N800 with help of our kernel sources.


Jarkko

Re: [android-internals] Re: Android on Nokia n810 (OMAP2420)

[Brian Swetland]

[Jarkko Nikula <jarkko.nikula@nokia.com>]
> On Sun, 6 Apr 2008 06:23:23 -0700
> "ext Brian Swetland" <swetland@google.com> wrote:
> 
> > Has anyone had any luck getting a 2.6.23 or 2.6.24 kernel working
> > with the 800 or 810?  When I was poking around with it a while back,
> > I had to backport our stuff to .21 due to binary-only wifi drivers
> > being incompatible with newer kernels.
> > 
> I would say that N800 has the best support at the moment in linux-omap.
> I think almost all the drivers are there. IIRC only temperature sensor
> is missing.

Well I'm looking for two specific things:

- can I use wifi with a 2.6.23 or 2.6.24 kernel?  The proprietary module
  part of the wifi driver did not work with > 2.6.21 kernels when I tried
  (due to network stack structure changes by the look of things).

- can I get the battery to charge without proprietary, binary-only,
  non-redistributable userspace code?  It did not seem like that was
  the case a couple months ago.  Perhaps it's better now?

Brian

Re: [android-internals] Re: Android on Nokia n810 (OMAP2420)

[Jarkko Nikula]

On Sun, 6 Apr 2008 23:51:09 -0700
"ext Brian Swetland" <swetland@google.com> wrote:

> Well I'm looking for two specific things:
> 
> - can I use wifi with a 2.6.23 or 2.6.24 kernel?  The proprietary
> module part of the wifi driver did not work with > 2.6.21 kernels
> when I tried (due to network stack structure changes by the look of
> things).
> 
Most probably not I think. But worth to search the maemo developers
mailing list are there any info about this.

> - can I get the battery to charge without proprietary, binary-only,
>   non-redistributable userspace code?  It did not seem like that was
>   the case a couple months ago.  Perhaps it's better now?
> 
This should work if you boot through the initfs as usual. E.g. I've
run the Debian armel on N810 using up-to-date linux-omap.


Jarkko

Re: [android-internals] Re: Android on Nokia n810 (OMAP2420)

[Tony Lindgren]

* Jarkko Nikula <jarkko.nikula@nokia.com> [080407 07:13]:
> On Sun, 6 Apr 2008 23:51:09 -0700
> "ext Brian Swetland" <swetland@google.com> wrote:
> 
> > Well I'm looking for two specific things:
> > 
> > - can I use wifi with a 2.6.23 or 2.6.24 kernel?  The proprietary
> > module part of the wifi driver did not work with > 2.6.21 kernels
> > when I tried (due to network stack structure changes by the look of
> > things).
> > 
> Most probably not I think. But worth to search the maemo developers
> mailing list are there any info about this.
> 
> > - can I get the battery to charge without proprietary, binary-only,
> >   non-redistributable userspace code?  It did not seem like that was
> >   the case a couple months ago.  Perhaps it's better now?
> > 
> This should work if you boot through the initfs as usual. E.g. I've
> run the Debian armel on N810 using up-to-date linux-omap.

I guess there are still some N800/N810 patches missing from
l-o tree.

Also, you probably need to patch N810 machine id to
5555 in arch/arm/tools/mach-types to boot..

Tony