[anarsoul]

okay cool. I finally got my environment set up properly I think. I will see if it builds Friday evening.

It builds (already tried with codesourcery 2010q1 toolchain), however there're some compile errors in SMP code (they declared inline function in C file and then trying reference it from another file... gcc reports error

), but they're easy to fix...

 

[anarsoul]

working modules @ http://dropcanvas.com/wzyod

* md4.ko and cifs.ko
* pl2303.ko

using:
Darkelf 1.0
android-ndk-r8/toolchains/arm-linux-androideabi-4.4.3

Makefile:
EXTRAVERSION = -g1acfcec-dirty
CFLAGS_MODULE = -fno-pic

Can you try to build&test bluetooth, rfcomm and btusb modules?

 

[anarsoul]

As near as I can tell, the Bluetooth subsystem support must be built into the kernel. This includes L2CAP, SCO and RFCOMM. They cannot be built as modules. BTUSB will not work I think without L2CAP. Someone please correct me if that is not right.

Why? I can easily build bluetooth.ko and rfcomm.ko, but after inserting rfcomm.ko kernel just panics for some reason.

 

[anarsoul]

I can build the modules too, but they will not work without the subsystem, which is a built-in to the kernel (not by default). So you must have a modified kernel for it to work. Just insmod'ing the module will load it, but because the existing kernel doesnt support it, it wont work. We will have to update our kernel to make bluetooth work, I am sure.

However, there are many more issues at the moment, such as the HDMI drivers will not compile.

Well, I think it's possible to fix HDMI driver. Non-bootable kernel and missing elf-2/aurora-2 LCD drivers bother me a bit more

Maybe it's time to ask Ainol for their patches to Amlogic tree?

 

[anarsoul]

sorry not had time to test this yet as booting , but it built with gcc 4.6 (some minor tweaks) and with CS 201203
without a serial console its going to be difficult to work out how to progress without full patches.

I believe that there should be UART pins somewhere on board. Are Aurora-II or Elf-II board photos in high
resolution and good quality were published somewhere?

 

[anarsoul]

Maybe it's time to ask Ainol for their patches to Amlogic tree?

Just sent GPL violation notification to Ainol.
Will publish it and will send a letter to gpl-violations in case of missing response.
I think that since AmLogic released their sources Ainol has nothing to hide.

 

[anarsoul]

As for Elf II, touchscreen is:

I: Bus=0000 Vendor=0000 Product=0000 Version=0000<br />
N: Name="ft5x06"<br />
P: Phys=<br />
S: Sysfs=/devices/virtual/input/input5<br />
U: Uniq=<br />
H: Handlers=event5 <br />
B: PROP=0<br />
B: EV=b<br />
B: KEY=400 0 0 0 0 0 0 0 0 0 0<br />
B: ABS=650000 0<br />

Which is Focaltech I2C capacitive touchscreen in config. Tomorrow I'll check for Aurora, but you can also check with simple:
cat /proc/bus/input/devices

N: Name="ft5x06" for Aurora II

 

[anarsoul]

Just had a look and there only seems very generic amlcd devices inside the rom

3 i2c channels
i2c0 has
0x38 ft5x06 touchscreen
0x03c gt2005-i2c cmos sensor
i2c1 has
1x018 bma250 accelerometer
1x01a wm8960 wolfson audio code.
I2c2 has
2x034 Axp20, multifunction power management

I'd say all these are soc so likely to be in the amlogic common.
The cmos is common enough to have existing code if not. There is a weird one though a device with no hw allocation. ir_kbd_i2c

Now lcd......

Sent from my ELF-II using Tapatalk 2

On aurora there's hi253_i2c cmos sensor instead of gt2005

 

[anarsoul]

Hi Anarsoul,
I also got a few warnings from gcc, could you please share with the rest of us how you fixed the offending code?

I am still trying to get my compilation to finish without fatal errors!
Thanks!

As of SMP part, here's diff: http://dpaste.org/scJnM/
Now trying to fix mali driver, something's broken in its Makefile for sure, I'm getting a log of:
gcc: error: directory: No such file or directory
gcc: error: directory": No such file or directory
gcc: error: directory: No such file or directory
gcc: error: directory": No such file or directory

Also I did not fix HDMI driver, just disabled it.

 

[anarsoul]

Did anyone succeed to build mali driver from Ainol tree?

Btw, drivers/amlogic/power/aml_power.c tries to include missing linux/aml_power.h,
Also, arch/arm/mach-meson6/board-meson6-ref.c lacks power{on,off} function for
SDIO WiFi...

In general code quality is very poor, AMLogic did not use gpiolib framework,
clk framework, generic rfkill-gpio driver, code does not follow code style guide,
etc, etc...

I think it's too much work to clean up that code and reverse engineer missing parts,
can take up to several months.

I think AMLogic still violates GPL, it's close to impossible to compile _usable_ kernel image
from this crap they released.

 

[anarsoul]

Compiled without errors using arm-linux-gnueabi- toolchain from emdebian repository http://www.emdebian.org/crosstools.html

I'm using codesourcery 2010q1 toolchain on archlinux, I think it's make/shell related issue

 

[anarsoul]

Anarsoul,
I agree with you that the code is crappy and that this kernel code is a very half-hearted attempt by Amlogic to comply with the GPL without disclosing any useful information.

We would also need the source code for the u-boot loader, right now we have to use the binary included in the Ainol firware upgrades.
But this is all we have! Please don't give up, your input here in this thread is extremely valuable!
The guys working on the Allwinner A10 kernel faced a similar situation a few months ago and look at all they have accomplished since then (I also have a couple of A10 devices that I can run Linaro armhf Ubuntu 12.04 on right now). They have a kernel tree that they have managed to upgrade to Linux 3.5 and I believe at one point they'll manage to merge their code back into mainline. The code in their tree is quite clean and compiles with just a few warnings.
I think if enough people pitch and we organize ourselves we can reach similar results here. Or so I hope!

I'm not giving up yet, don't worry

 

[anarsoul]

Why do we need source code for bootloader? First, we can port uboot using drivers from that kernel.

No, we can't. Kernel does not touch DDR controller registers at all, it's up to bootloader to
initialize memory somehow. However, why do we need to change bootloader?

 

[anarsoul]

Why do we need source code for bootloader? First, we can port uboot using drivers from that kernel.
Second, we can launch custom kernels without custom bootloader.
I posted basic instruction on 3 page.
With aml_autoscript we can modify uboot environment, launch kernels, format nand, etc.
uboot environment info
uboot environment info2
uboot environment from 0615

Just tried to boot self-built kernel using manual from 3rd page, it just hangs

Anyone here brave enough to solder wires to serial pins and check what's going wrong?

 

[anarsoul]

Anarsoul, did you specify a load address of 80008000 on the mkimage command line? And are you using an initramfs? The instructions on page 3 are not very clear about how to make a kernel boot.

Just did 'make uImage', however load address is 0x80008000.
I'm not using any initramfs, but kernel at least should initialize framebuffer and print something on it...

 

[anarsoul]

Default kernel ( meson6smp_defconfig ) works, but no readable output to screen ( only random lines at first second, then blank screen ).

Did you set correct screen resolution in kernel config?

CONFIG_FB_OSD2_DEFAULT_BITS_PER_PIXEL=32
CONFIG_FB_OSD2_DEFAULT_WIDTH=1024
CONFIG_FB_OSD2_DEFAULT_HEIGHT=600
CONFIG_FB_OSD2_DEFAULT_WIDTH_VIRTUAL=1024
CONFIG_FB_OSD2_DEFAULT_HEIGHT_VIRTUAL=1200
CONFIG_FB_OSD1_DEFAULT_BITS_PER_PIXEL=32
CONFIG_FB_OSD1_DEFAULT_WIDTH=1024
CONFIG_FB_OSD1_DEFAULT_HEIGHT=600
CONFIG_FB_OSD1_DEFAULT_WIDTH_VIRTUAL=1024
CONFIG_FB_OSD1_DEFAULT_HEIGHT_VIRTUAL=1200

 

[anarsoul]

Tried, still the same.

Then we need serial console to advance (looks like non-invasive ways are over

)

 

[anarsoul]

Or get a config file from a working Elf II kernel? Does anybody have a config.gz file in /proc on any of the alternative ROMS?

Nope, Ainol engineers disabled that option in kernel config, so no config.gz in /proc unfortunately

 

[anarsoul]

Just thinking out ... ain't that the bootloader source code?

http://openlinux.amlogic.com/download/linux/ARM/u-boot/arm-u-boot-2011.03.svn1828.zip

Maybe, but we don't need bootloader code right now. Anyway, I'm not going to mess with bootloaders
without JTAG (or ability to unbrick device without JTAG)