Android Tablets Forum banner
21 - 27 of 27 Posts

·
Registered
Joined
·
353 Posts
I'm not a real expert with GIT and kernel merging, but, I noticed that the Cyanogenmod kernel (CM Kernel build; GIT 2.6.37.6 kernel) is missing at least the the "emxx" machine, under the arch/arm directory. We have that from the Renesas SDK released here: EVBSP-SDK

Is someone here able to push Renesas code upstream in CyanogenMod kernel?

Only after that, it makes sense I think to build CyanogenMod for these devices, though I'm pretty sure it isn't enough yet to have a fully working tablet...
 

·
Registered
Joined
·
353 Posts
Well... I tried myself approaching this huge task. But I had to give up...

CyanogenMod (will call it "CM") 2.6.37 kernel from https://github.com/CyanogenMod/cm-kernel and forked to https://github.com/ffxx68/cm-kernel
Renesas (will call it "R") 2.6.29 kernel from https://github.com/cAPSLOCK7

Renesas-specific code was searched in R with commands like:

find . -type d -name "*emxx*"
find . -type f -name "*emxx*"
find . -type f -exec grep -l -n -i "emxx" /dev/null {} \; | sort
find . -type f -exec grep -l -n -i "emev" /dev/null {} \; | sort

Then, all files found were either moved to CM (if missing) or compared one by one for differences.

Renesas-specific code in R is either included in entirely new directories, or files, or wrapped inside a #define ..._EMXX / #endif.

That made almost straightforward moving entire directories or files from R to CM, and also helped a lot looking for the differences in most of common files. But... devil is well hidden in the details, as usual


The first relatively complex file to merge was ./drivers/media/video/videobuf-core.c. But I manged to do that, somehow.

Merging of ./drivers/mmc/core/*.c and ./drivers/mmc/card/*.c was even trickier. I had to leave some sections unchanged in CM, as they looked like already including logic found in R. But, once again I managed to merge the code, though I'm afraid not all necessary changes are there, or what I've done is entirely correct. Infact, being a newer kernel, non-Renesas code has evolved in CM too, with respect to R.

I finally gave up when I reached ./arch/arm/mm/proc-v7.S, which in my opinion is too complex to merge without specific knowledge of ARM and Renesas EV CPU architecture.

I know many here already knew what I discovered but myself, and I also expected something like that, but... direct experience is everything!

Though incomplete, I think I'm going to commit what I've done until now back into my repository at https://github.com/ffxx68/cm-kernel, just not to throw away what I've done until now. Maybe someone else could be curious to have a look at it. Will let you know when I'm done.

EDIT - Nov 30 - I've done my commit into https://github.com/ffxx68/cm-kernel . See Merge.txt with status.
 

·
Registered
Joined
·
353 Posts
Well... I've carried on a bit my investigations, but I later gave up as I think it isn't feasible. Differences are too big and I'm sure now that the EVBSP-SDK kernel isn't made for this tablet board. It misses too many components, not to tell it's outdated.
So, the only hope is to have the real tablet kernel being made available from Renesas, if we manage to "convince" them it's a GPL stuff...
 

·
Registered
Joined
·
11 Posts
I stopped developing on this because official 2.3 versions appeared, so this project became pointless. Also I faced some problems with SGX libs which we do not have sources of. And nobody does (in public).
As for cm kernel, there's no need to build kernel. I had a Motorola Milestone and happily used CM7 on it with stock moto kernel. Kernel is low level , android does not depend on it. You can run android 4.0 on 2.6.29 kernel.
Right now it should be rather easy to build CM7 for rena using binary blobs from 1220 FW, but it will keep all the issues of 1220 (may be even add some new).

I've tried to build CM9 (4.0.3) but had no luck at all, did not even got it to boot.

I saw in another thread that our chinese friends faced some strange lockups trying to boot 4.0.3. I'm rather noob in all this android stuff, but I've heard B&N Nook Color (which has the very same GPU) recently got updated SGX libs and drivers from TI and now devs have managed to get full HW acceleration in UI and other... May be our chinese friends just need some new functions to be implemented by Imagination Technologies in SGX's drivers?
 

·
Registered
Joined
·
217 Posts
Take a look here in cyanogenmod (CM9) progress update:
http://www.cyanogenmod.com/blog/cm9-progress-update

There are a number of challenges that we are up against. Google has made some pretty major changes to the Android framework that break compatibility with older proprietary camera and graphics drivers in order to achieve some pretty insane performance, but I am confident that the team will be able to overcome these issues like we have in the past.
So reading about what dnkn said about the Nook color... maybe the problem is that rena's development team is trying to build android 4 using the old non-compatible SGX drivers.

PS: be aware I'm just guessing... i really don't know
 

·
Registered
Joined
·
353 Posts
Yes, they were trying to build android 4 using the old non-compatible SGX drivers... but I think they've managed to fix it (that's what the chinese friend told me at least).
 
21 - 27 of 27 Posts
Top