Sunday, December 1, 2013

OpenSUSE for Rockchip ARM devices

      Going on with the earlier post on creating your own MicroSD with the Linux RFS of your preference, here is the howto for installing OpenSUSE using the usual kernel and modules+firmware.
      Just in case, whatever your stick/box, you can use these very generic 3.0.36+ kernels for RK3066 or RK3188 (only missing internal Wifi).

       In my case I used the excellent Radxa Rock board, equipped with a quad-core RK3188, for the tests and found no problem booting it:

OpenSUSE on Rockchip RK3188 (Radxa Rock board)

     Of course, this has been made possible by OpenSUSE team's commitment to the ARM architecture, since they are one of the few that have put all the necessary efforts to have their distribution compiled for ARM CPUs.

     OpenSUSE's official wiki has one page dedicated to installing on a RK3066 device (MK808 [1]), though it is outdated since, among other things, it uses the old 3.0.8 kernel. However the guide was useful and I wanted to update it here.

Now the procedure, with the same captions as in the original post (for Linaro Ubuntu), for comparison:

Getting an up-to-date official distribution, for ARM
     You can check the latest ARM distribution at OpenSUSE by heading to this url:

and downloading the "rootfs" suffixed .tbz file, which at the time of writing (Dec 2013) is this one:

Getting the RFS into the MicroSD card (or flash)

Exactly the very same procedure as in original post, except that steps 4 and 6 are NOT to be done, and step 3 becomes:

3) Extract the downloaded ARM RFS in it:
tar xjfv /home/username/Downloads/openSUSE-12.3-ARM-XFCE-rootfs.armv7l-1.12.1-Build49.1.tbz 

And this last step is necessary (or else the root file system is mounted as read-only, preventing correct boot), open the /mnt/whatever_folder/etc/fstab in your linuxroot and add the following first line:
/dev/root / ext4 defaults,noatime 0 0

Getting a desktop environment

    You already have it. However I've found a problem: upon opening the "Install/Remove Software" you will likely stumble upon this fatal error:

"UI Syntax error. Couldn't load plug-in gtk-pkg. Check the logfile"
Solution is simple [2], open a terminal and type:
su -c "zypper in libyui-gtk-pkg4"
You'll be asked the root's password and the missing package installed, so now you can open the Install Software app.



Wednesday, November 27, 2013

Your own official Linux distro in a SD card (for ARM)

Currently for us, owners of ARM devices of the Rockchip family (though this post applies to any ARM CPU), when wanting to boot Linux we have to do two things:

1) Flash a Linux kernel, usually to recovery, in order to be able to boot to Linux
2) Copy a Linux Root File System (RFS) into a MicroSD card for the OS to boot from

This post is about Step 2.

Why can't we use official distributions?

Basically because most distros are [easily] available for x86 CPUs (x86, amd64, ...), not ARM CPUs (armel, armhf). And then, when you find it compiled for ARM, there are some little extras to be done, like copying your kernel's modules into it, and throwing it all into a MicroSD card.

Hence, we are currently forced to choose among static RFS that fellow developers have made available and can update only on their free time, which isn't much. These are:

- Picuntu, from Alok Sinha
- Ubuntu 12.10 RFS, from linuxium
-, from JustinTime4Tea

These RFS are all just the Ubuntu distribution, sometimes with a different name (Xubuntu in the case of Picuntu, Ubuntu in the case of, though they usually include extras like:

- The modules (drivers) for many devices (needed for many, but not all, USB gadgets)
- The Flash video ARM support solution I posted here

However let's face it, you are forced to use Ubuntu, and then you depend on the maintainer's free time to grab the latest version of Ubuntu, add the above extras and publish it so you can use it.

Your own Linux distribution's RFS

Here I am going to show how to create your own RFS by grabbing whatever version of Ubuntu (and with few changes any other distro, like Arch Linux, OpenSUSE, CentOS, just click on the link to be taken to the specific distro's instructions), and create a RFS to use on a MicroSD card to boot Linux your ARM stick/box. Or to flash the same RFS into your device's Nand Flash chip and avoid the SD card altogether.

Getting an up-to-date official distribution, for ARM

The first thing is getting the Linux distribution (all the programs that make up your booted up Linux: from the "ls" commandline, to the GNOME desktop and the LibreOffice suite) already compiled for ARM devices, instead of the usual x86 for PCs.

This is not difficult since, actually, Linaro is doing just that!
If you want a desktop environment (XFCE, Gnome, KDE, etc.) you have two options:

Option 1) Easy way: (Older version but everything done) Head over to:
Look for "Ubuntu Desktop" in the "Developers and Community Builds" section and click on any of the boards that support it on the right (Origen, Panda, etc. doesn't matter). At the time of writing (Nov 2013) this will redirect you to this page:
So the latest version with the desktop already installed is Ubuntu 12.11, you'll have to install this one in your SD and upgrade to the latest version after booting. Download the biggest .tar.gz file, in my case 473 MB of:
Option 2) Manual way (Newer version, but starts with a commandline Linux): You can get the cutting edge latest version of Ubuntu, it will be small and fast but you'll have to type a few more commands, after booting it, in order to install a desktop. Go here:
And click in the folder link of the latest version number ("13.11" for me). A new page appears where you have to select "ubuntu", in the next page select the link that ends in "-images" (prefix depends on Ubuntu's version name, in this case it's "raring-images"), and finally select the version link that you want (nano / developer / server) to get its .tar.gz file.

All three are the same Linux but each one with different sets of packages already installed. They are:

- nano (around 50 MB) gives you a command-line with very very few things (not even a vi/vim/nano editor!), with this version you'll need a network connection (care for /etc/network/interfaces and such) soon enough, to start cranking "apt-get install" for all the little apps.

- developer (around 140 MB) a usable command-line with most things you'll need to comfortably start working and changing whatever configuration. However, IMHO, it includes many packages for software development, while lacking some others more typical of a generic user system.

- server (around 130 MB) a usable command-line with all you need for a quick setup of a desktop environment resembling a PC with Ubuntu. Even if you are a developer you can later install all the packages you need. This is the one I recommend.

N.B.: Compare the *.packages files to get a glimpse for yourself of what's inside each version.

So, following the recommendation we would download this file for the latest Ubuntu 13.11 (Nov 2013, or else go into the page above and get yourself the latest and very best.

Getting the RFS into the MicroSD card (or flash)

First of all introduce the MicroSD card into your PC to partition it in the way that the kernel expects it (please don't use partition utilities, they won't do it!).

Let's say that the MicroSD appears at the folder "/mnt/whatever_folder", then type the following command to know the device name:
df -h | grep whatever_folder
You should see a line somewhat like this:
/dev/sdg1   7,3G   612M  6,4G   9% /mnt/whatever_folder
Now that we know the MicroSD is in "/dev/sdg" (remove the number!!) unmount the MicroSD:
sudo umount /mnt/whatever_folder
And the most important step to avoid later problems: partition & format in this one step:
mkfs.ext4 -F -L linuxroot /dev/sdg
Which labels it "linuxroot" and at "/dev/mmcblk0", where the kernel expects it (CMDLINE), instead of where a partition tool would place it (at "/dev/mmcblk0p1").

If you want your RFS to be in your device's Flash chip, and 1) have flashed the right parms, 2) have a kernel with the right CMDLINE for those parms, 3) the kernel has an initramfs.cpio with rknand...ko, then the you have to format it with this command: "mkfs.ext4 /dev/mtdblock0" (mtdblock number is that of your [big] partition).

It may take a couple minutes and when finished you should mount it and start following these steps:

1) Become root (necessary to keep file permissions in the RFS!)
sudo su -
2) Go to the folder where the SD card (or flash) is mounted:
cd /mnt/whatever_folder
3) Extract the downloaded ARM RFS in it:
tar xvfz /home/username/Downloads/linaro-raring-server-20131124-562.tar.gz
4) Since the Linaro .tar.gz contains the RFS inside a folder named "binary", we have to move its contents to the real root of the SD card and then remove the, now empty, "binary" folder:
mv binary/* .
rmdir binary
5) If you are doing this process on a PC, you have to uncompress the modules+firmware file (that came with the kernel you flashed into your stick) into the "/mnt/whatever_folder/lib/" folder.

Or, if you are on an ARM stick, you can just copy your own modules (drivers) and firmware to the new RFS so you have everything from the beginning:
mkdir ./lib/modules
cp -R /lib/modules/* ./lib/modules
cp -R /lib/firmware/* ./lib/firmware
And it won't hurt to copy the library for playing Adobe Flash videos (think YouTube) in ARM (and all the Flash ads and websites, of course):
cp /usr/lib/ ./usr/lib/
If you didn't have it, follow this quick post.

6) You may also want to "touch" certain files to make sure networking works out of the box. For example, if you have an Ethernet connection in your ARM device, then modify the /mnt/whatever_folder/etc/network/interfaces file to add these lines:
auto eth0
iface eth0 inet dhcp
7) It's always a good idea to end operations with this command (to finish all write operations):

You can unmount that MicroSD, because it's ready to boot. If you installed the desktop version the easy way, that's it.
However, if you installed a trimmed commandline version the manual way, you should continue reading the following section.

Getting a desktop environment

Once you boot your device with the created MicroSD card, your first task is to check the network connection. For example type:
apt-get update
If the software repositories start updating, you're good to go. If there is no connectivity: Internet is full of Q&A about networking in Linux :)

Now, with Internet on, in order to install a desktop environment, in my case XFCE (the one used in Xubuntu), I would just type this:
apt-get install xubuntu-desktop
And when asked, say Yes to downloading >400 MB of packages that will use up >1 GB of space when installed, but will land me on the very latest desktop environment for an ARM PC, whatever your CPU (Rockchip, Allwinner, Samsung, ...).

Sorry if it took too long to explain something that is actually simple, but everybody is welcome!
The more users discover that a tiny and cheap stick/TV-box is actually a full-fledged low to mid-end PC (without the noise, size, and the Watts!), the better for all!

Tuesday, November 19, 2013

Linux on UG008 TV Box (RK3066)

I was looking for the cheapest Rockchip device with external Wifi antenna and/or Ethernet connection, just to find both capabilities in the UG008 TV Box, a RK3066 (dual core) device at just 55 USD (shipping included).

After purchasing it and a quick delivery of 2 weeks to Europe, I set out to test it, since I worried it may be lagging behind the newer 4 core experience with the RK3188 CPUs.

The TV box surprised me being very small, at 7.5cm x 7.5cm and looking really slick and solid. I was glad to see Android feels zippy.

Small and cheap UG008 TV Box (wifi antenna is also just 7 cm)

So, hacking started:

1) Backup internal flash: The recovery button is at a tiny hole in one side, near the front. The only quirk here is that you insert a clip there and stay pressing the recovery button, plug the MicroUSB from the PC to the TV box and then: you also have to press a couple of seconds the power button, before you can release the recovery.

A red light appears around the MicroUSB slot, and on your [Linux] PC you can type "lsusb" and find a nameless device (for my PC's buses/devices, it is: "Bus 002 Device 004: ID 2207:300a  "), the 2207:300a ID is what matters, that is our UG008.

First's first, the "partition table", extracted with:
sudo ./rkflashtool r 0 1 | head -n 11
looks like:
CMDLINE: console=ttyFIQ0 androidboot.console=ttyFIQ0 init=/init initrd=0x62000000,0x00800000 mtdparts=rk29xxnand:0x00002000@0x00002000(misc),0x00006000@0x00004000(kernel),0x00006000@0x0000A000(boot),0x00008000@0x00010000(recovery),0x00120000@0x00018000(backup),0x00040000@0x00138000(cache),0x00300000@0x00178000(userdata),0x00002000@0x00478000(kpanic),0x00120000@0x0047A000(system),-@0x0059A000(user)
We go one by one backing them up into our PC (you can safely leave out some, like user, cache, kpanic), just remember for the rkflashtool command you have to swap the hex numbers above.

For example to backup recovery (to flash our Linux kernel on it), which is listed above as "0x00008000@0x00010000(recovery)", you'd do:

sudo ./rkflashtool r 0x10000 0x8000 > UG008_recovery.img

2) Rooting UG008: Once we are done with that, we should root the device (even if just for the sake of being able to dual boot Android or Linux), the rooting procedure is fairly standard and described here. If you only want Linux on the box, you don't need to root it, keep reading,

3) Flash Linux kernel: You have two options depending on what you want:

a) Dual Boot Android/Linux: This is the usual way, you want to flash the recovery partition. I've found this box to be a close relative of the Measy U2C, for which I published several recovery kernels that are compatible with UG008, you'd be missing Wifi (RK901) and Ethernet. Flashing being just:

sudo ./rkflashtool w 0x10000 0x8000 < recovery.img
sudo ./rkflashtool b 

      If you're into working your way around, it should be easy to add the connectivity. Wifi with RK901 has been supported in Linux sticks for some time already (though I never bothered to check myself, being used to wired LAN), and Ethernet should require only to enable this .config options before building your kernel: CONFIG_NET_ETHERNET and CONFIG_RK29_VMAC.

Generic Linux Kernel building for RK devices is described here. Just remember this is a RK3066 device, so don't compile it for RK3188! The audio codec is the well known RK1000.

Now, in order to boot to Linux from Android you can use this excellent "Autorun Linux" app, or install "Android Terminal Emulator" get inside type "su" and then "reboot recovery" every time.

b) Only Linux (remove Android): Very legitimate. Using the same recovery kernel as above, the flashing procedure is all explained here (but follow the instructions for flashing "boot" partition, instead of "kernel" partition, since you are using a recovery kernel). The boot writing (after you back up that partition!!) can be done with this command:

sudo ./rkflashtool w 0xA000 0x6000 < recovery.img
sudo ./rkflashtool b 

Note: I haven't tested the "Only Linux" approach on my UG008, though it should work, or else you can always flash again your backed up stock partition and be back where you started.

4) Booting Linux: Insert your usual MicroSD card with Linux root file system (rfs) and you're good to go.

Any questions, and polite criticism is welcome in the comments below. I hope this post helps!

Wednesday, October 23, 2013

DIY NAS with a RK3188 device (Radxa Rock)

I currently have a full-blown Linux PC noisy tower serving 3 SATA hard disks full of multimedia as a home server... that is some 80 Watts of power consumption, plus the noise, for just a few hours of use per week.

There must be a better solution for my home network file-sharing (also called NAS) needs!

So, come the Radxa Rock board, equipped with a Quad Core ARM CPU, the Rockchip RK3188, and 2 GB RAM, quite a beast actually (I'm currently blogging from it, and use it for almost all my PC needs).

And here is today's proof-of-concept RK3188 NAS happily serving files from my desk:

1st rev.: RK3188 functioning as NAS home fileserver (using Radxa Rock board)
2nd rev.: Radxa Rock as a 3 hard disks NAS (11W idle, 33W with 2 disks streaming)

Now, if you want to bear with me on the technical side... I've been doing the math about data bandwidths for some time already and taken into account the Rockchip datasheet briefs to guess the best I/Os for the task:

Hard disk interface:

Let's face it, Rockchip SoCs don't have SATA ports, and other ARM SoC have at most one SATA port, afaik, and that still means an external power for the SATA device is necessary. That is not a solution for NAS, unless you can fit everything in just one hard disk.

So let's get over it and find the second best solution: USB 2.0 ports are the highest bandwidth inputs (to connect the hard disks). The Radxa Rock board has:

- Two full sized USB 2.0 ports, and a third one in the pins of the expansion connector. All these ports are sharing, through an integrated hub, one single USB bus coming from the RK3188.
- One Micro USB 2.0 OTG  port that is directly wired to the RK3188.

Ideally USB 2.0 would mean up to 480 Mbps raw throughput so, to get the best, we don't want to share it with other devices. Hence, we will be plugging our hard disk/s to the USB OTG port.

Thankfully the market is full of external hard disk enclosures that do a very good job at converting SATA to USB. And let's see what that means:

- SATA is currently doing 3 Gbps
- USB 2.0 is "just" 480 Mbps, here is a bottleneck, certainly, but is it?

Why USB 2.0 is fine instead of SATA

Does this sound like a let down? Really? Let's do the math:

Are you going to be streaming 1080p content over Wifi? That really leaves you with <300 Mbps, and most probably <150 Mbps, if you can maintain that incredible wireless speed.

But, let's see what 150 Mbps really means in today's high compression world:

150 Mbps / 8 bits per byte = ~19 MB/s and let's say your content is a 1h30m long movie...
19 MB/second * 90 minutes * 60 seconds/minute = 100 GB movie!

Is your movie larger than that? Then this solution is clearly not for you. For the rest of the world, let's go on:

You may rightly object: movies have well and bad compressed parts, fast action may mean a bandwidth spike!

Well, you are right, and that's why multimedia players use a technique called buffering! This mostly rids us of the problem.

So once we accept that it's OK to stream things at Megabit speeds, we may stop worrying about using SATA hard disks over USB.

Network interface:

As I/O, the Radxa Rock board has an integrated Fast Ethernet, that is connected directly to the RK3188 (through the usual RMII PHY interface). This means 100 Mbps so... people with movies of >66 GB may stop reading now.  ;-)

For the sake of this proof-of-concept, we will be using this 100 Mbps Ethernet port, although a better solution would be to have a Gigabit Ethernet adapter attached to a full sized USB 2.0 port, which would give us up to 480 Mbps of actual bandwidth.

Do not share the USB OTG port with the hard disk/s! The RK3188 (and the RK3066) has two physical USB 2.0 buses: the OTG and the "full sized", so use both as much as possible, instead of cramming all devices onto just one's bandwidth.

File-sharing Speed Testing:

I am using a bare Xubuntu Linux on the Radxa board so it is possible to connect to the SATA hard disk through FTPS out of the box.

The SATA II hard disk is in a external USB 2.0 enclosure and has a 1.5 TB NTFS partition and we will be using a FTPS server (slower than FTP). Not the best case for speed, but will let us see how well it can do on this not ideal setup.

So, first of all, let's test how fast the RK3188 itself can read files from it, by running this command on the connected [but unmounted] drive:
sudo hdparm -t /dev/sda
 Timing buffered disk reads:  90 MB in  3.03 seconds =  29.74 MB/sec
The repeated tests result always around and very close to 30 MB/second. That is some 240 Mbps, or 50% the theoretical maximum USB 2.0 speed.
Who is to blame? I fear in this case it's my old USB2SATA's fault, since doing the same speed test on an i7 top of the line PC yields some 33 MB/s.

Good, more of a worst case then, so let's do real life tests with actual multimedia files on the SATA disk served over Fast Ethernet by the Radxa Rock RK3188 board:

Streaming an 11 GB action movie (1080p) requires an average of 2 MB/s and is perfectly watchable on the remote PC, as if it were a local file.

Transferring a large file over Ethernet to another PC maintains a speed of 6.7 MB/s, that is a bit over 50 Mbps.

I've also tested transferring a file from the MicroSD card where Xubuntu runs, just to do a test without USB and SATA, the result is a constant speed of 7.2 MB/s.

For the sake of completeness, I too have tried this same raw transfer (FTPS from MicroSD) using a Gigabit Ethernet to USB 2.0 adapter, the result is a constant speed of 8 MB/s.
Better and more than enough for our purposes, but still below my expectations. Next thing will be to test with an FTP server, without the FTPS encryption burden!

Further details about these tests: the "intermediary" is a Gigabit switch that has been thoroughly tested at 1 Gbps speeds.

So it seems the RK3188 is perfectly able to cope with the task, even using a heavy encrypted FTPS server.

Conclusions for a faster NAS file-sharing solution

1) Use a Gigabit Ethernet to USB 2.0 adapter (and from a reliable source or you'll find yourself with a ridiculous USB 1.1 device able of <11 Mbps) and connect it to the full size USB port on your Rockchip board/stick.

2) Set up a simple FTP server for file-sharing, that makes sure your RK3188 can concentrate on reading the hard-disks through USB and the network protocols, instead of the heavyweight encryption for the FTPS. Even if you go wireless, Wifi already encrypts everything on the air!

3) Use Linux-friendly EXT4 partitions. NTFS is not bad, but will require more CPU to read/write (you'll see a process called "mount.ntfs" often). FAT what?

4) If you want to have a 4-disk NAS or even an 8-disks NAS, no problem! Just know you should still plug them all, through a USB hub, to the same USB port.

5) And the final tip to make a nice NAS solution:
sudo hdparm -B 50 /dev/sda
This wonderful command at NAS startup will take care of turning-off your hard disks when they are unused, saving power and hard-disk life, while keeping a 100% silent environment!

UPDATE: If the drive answers with:
HDIO_DRIVE_CMD failed: Input/output error
 APM_level = not supported
Then try this similar command:
sudo hdparm -S 20 -K 1 /dev/sda

(substitute /dev/sda in these commands with your actual /dev/sdX device)

Next step: a quad hard-disk NAS solution based on Rockchip ARMs for all my file serving needs!

Wednesday, October 16, 2013

Radxa Rock communicating with I2C devices

Since I'm a hardware tinkerer at heart, and my background is more on electronics than on software, the Radxa Rock board had a deeper interest for me.

This board has a lot of potential for all kind of low level electronics, thanks to its two expansion headers:

Expansion Headers from Radxa Rock schematics

As you can see there are two UARTs (number 3 is shared with some GPS pins), one I2C bus, two SPI buses (SPI1 on J8 is shared with GPS pins), a full 24 bits LCD display output (!!), one USB port (HOST_D*2), one audio Line In (LINE_L/R), as well as several other GPIO I/Os.

I then set out to test my latest kernel for booting Linux on the Radxa Rock to access one of the, for me, most interesting low level hardware buses: I2C.

This one is heavily used for low bandwidth (control/status, sensor data readings, EEPROMs, etc...) communications inter integrated circuits (I I C... I2C ;). This bus consists of just two lines: SCL (master clock) and SDA (bi-directional data), which makes it extremely handy.

And for this test, a handy old friend of mine: the BMP085 pressure sensor (featured on this blog's first post) inside a little PCB with an accelerometer chip, gyroscope chip, magnetometer compass chip, and our pressure and temperature sensor chip, the BMP085.

This is all called a 10-DOF or Ten Degrees of Freedom sensor board, since it allows for precise movement/position detection through "sensor fusion" (though I really dislike using nerdy "magic words").

So first's first, I wired the 10-DOF board to the Radxa's left expansion header (J8) in the following manner:
- PCB's VCC (5V) wired to J8 exp. header's pin 40 (DC5V)
- PCB's GND wired to J8 exp. header's pin 39 (GND)
- PCB's SDA wired to pin 32 (I2C0_SDA)
- PCB's SCL wired to pin 31 (I2C0_SCL)

The result looks like this:

Radxa Rock board bus I2C #0 connected to a 10 degrees-of-freedom sensor board

So, what to do to start communicating with the board? Well, since my kernel already has all the needed things built-in, you just have to follow these steps (shamelessly taken from this guide):

#install i2c-tools (user space) for talking to I2C buses/devices
sudo apt-get install i2c-tools
#insmod the I2C-dev module to allow user space access to I2C
sudo modprobe i2c-dev
#probe I2C bus 0, the one at expansion header J8
sudo i2cdetect -y 0

The result of this last command is the possibly detected I2C devices on bus 0 (remember the 10-DOF board has 4 devices sharing the same bus). In our case we do know the I2C address of the BMP085 (0x77), so we can talk to it directly.

Since I know the calibration parameters of my device (see AC* at first screenshot in this post) and their addresses, I can just compare to see if this is random data, or actually the BMP085 chip, being talked to from Linux running on the RK3188 of the Radxa Rock board:

Calibration parameter AC1, which for my 10-DOF board has a 2 byte value of 6746 (0x1A5A) is located at I2C registers 0xAA (MSB) and 0xAB (LSB), by reading the sensor's datasheet.

Let's dump the contents of the device at that address (the BMP085) to confirm we are talking to it:

#dump non-interactively (-y) contents of selected device (0x77)
#on given I2C bus (0) in byte read mode (b)
sudo i2cdump -y 0 0x77 b

Not really illustrating, but the result is:
     0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f    0123456789abcdef
00: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00    ................
10: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00    ................
20: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00    ................
30: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00    ................
40: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00    ................
50: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00    ................
60: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00    ................
70: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00    ................
80: a5 94 52 49 a4 27 77 a6 1e 26 1a 5a fb dd c7 bc    ??RI?'w??&?Z????
90: 7c 9e 60 2c 59 11 15 7a 00 37 80 00 d4 bd 09 80    |?`,Y??z.7?.????
a0: a5 94 52 49 a4 27 77 a6 1e 26 1a 5a fb dd c7 bc    ??RI?'w??&?Z????
b0: 7c 9e 60 2c 59 11 15 7a 00 37 80 00 d4 bd 09 80    |?`,Y??z.7?.????
c0: 00 00 bc 33 00 00 00 00 00 00 00 10 00 00 00 03    ..?3.......?...?
d0: 55 01 06 00 00 00 00 00 00 00 00 00 00 00 00 00    U??.............
e0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00    ................
f0: 00 00 00 00 00 00 80 00 00 00 00 00 00 00 00 00    ......?.........

And sure enough! you can find the values (1a and 5a) of this little example at the expected register addresses (0xAA and 0xAB).

RK3188 is communicating with the 10-DOF board through one I2C bus at the expansion header!

Radxa Rock meets the world of electronics!

Saturday, October 12, 2013

Doubts over Rockchip FB Vsync fix

Almost every owner of a RK3188 stick has noticed video problems, like stuttering or a lost frame every now and then, no matter its being a local file or just a small avi.

Some time ago phjanderson (from Freaktab forum) found a solution to this:

It basically eliminated a blocking condition on a function on rk_fb.c, which also happened to apparently solve the stutter problem. This was called phjanderson's Vsync fix and lots of ROMs have added this patch since.

I then set out to test it on Linux along with my framebuffer fix (see ) just to find that after a random amount of minutes after boot, there appeared a new thread called "fb-vsync" that utilized 100% of one of the 4 cores on the RK3188, regardless of the system being idle or never even opening a video.

To me, this is unacceptable since it forces the CPU to run at higher frequencies all the time, as well as consume lots more of power and a 25% performance penalty hit.

I initially thought it had something to do with Linux' screen usage or even with my framebuffer fix. However it did happen without the framebuffer fix, so I just let it out of my kernels.

However, I recently read on a forum ( ) that Android people are seeing also the 25% CPU usage constant (1 core out of 4) due to this thread.

So I heavily suspect this patch is great for fixing the Vsync video stutter problems, but at the cost of losing 25% of the CPU performance, consuming more power all the time (+heat!), and thus reducing the component's life.

If you happen to have flashed a rooted ROM with Vsync fix and have an Android monitor installed (able to track system threads), please write a comment sharing if you've been able to see the thread in action.


I'll try to explain a bit the code related to this problem.
Most of it is located in rk_fb.c where:
1) upon registering fb0 (the main framebuffer) a thread called "fb-vsync" is created.
2) This thread executes the function "rk_fb_wait_for_vsync_thread"
3) This function is an "infinite" loop that emits a "vsync" notification to user-space apps subscribed to it, whenever a Vsync happens, meanwhile it waits idle.

How does it wait for a Vsync?
Through the use of "wait_event_interruptible" with arguments:
   - waitqueue: this is the event queue we're hanging on: dev_drv->vsync_info.wait
   - condition:
!ktime_equal(timestamp, dev_drv->vsync_info.timestamp) && dev_drv->

The 2nd part of the condition (dev_drv-> should always be met since it is set to 1 just after creating this thread.
What about the 1st part? It compares:
a) variable "timestamp": which is constant!! and set to the value of "dev_drv->vsync_info.timestamp" as it was just before calling "wait_event_interruptible".
b) dev_drv->vsync_info.timestamp as it IS any time the waitqueue is awaken

Naturally this condition is meant to become true (and exit the wait event to send the "vsync" notification) at the next Vsync.

Why? Because "dev_drv->vsync_info.timestamp" is updated by the lcdc interrupt service routine, which at the same time wakes up our "wait_event_interruptible" above!

Then the !ktime_equal compares the "timestamp" variable that holds the PREVIOUSLY set vsync_info.timestamp (before the wait_event is triggered) with the CURRENT, just updated, vsync_info.timestamp.
Hence the condition may indeed become true.

What did phjanderson's fix do? It just made this wait event to always have a true condition, hence the "while" loop is constantly executing (no wait) and notifying a "vsync".

How does it interact with Galland's framebuffer fix? Note that in the lcdc isr, the fb fix excludes the wake_up_... call, so what it does is make the wait event above to never be woken up. Kind of the opposite to phjanderson's.

The union of both fixes is... unfortunate too :) so I just left out phjanderson's fix, since the isr changes are needed for the overall fb fix.

Monday, October 7, 2013

Bitbanging Radxa Rock gpios

This post assumes you're using the latest Linux kernel from here on your Radxa Rock RK3188 board.

As you may already know, by looking at the openly available Radxa Rock schematics, this board has many GPIOs accessible through two expansion headers (along with other buses like LCD, SPI, I2C, etc.) as well as three user LEDs.

In this brief post I'll explain how to access any of these GPIOs to read/write its value. I've chosen one of the user LEDs to easily see the results of our actions.

By checking the schematics, it is easy to see there is a Red LED connected to GPIO0_B7 (gpio 175), an active low Green LED at GPIO0_B4 (gpio 172), and another active low Yellow LED at GPIO0_B6 (gpio 174).

How do I know the gpio number from the pin name? Relatively easy, look for your pin name here, and calculate the gpio number by taking into account that GPIO_BASE = 160 and NUM_GROUP = 32.

In this example we will toggle the Green LED on and off by following these commands in a Terminal within Linux in your Radxa (or through ssh to the board):

#become root
su root
#tell Linux we want to have access to gpio pin 172
echo 172 > /sys/class/gpio/export
#set it as an output pin (though we will be able to readback its value)
echo "out" > /sys/class/gpio/gpio172/direction
#set its value to 1 (turn off led)
echo 1 > /sys/class/gpio/gpio172/value
#set its value to 0 (turn on led)
echo 0 > /sys/class/gpio/gpio172/value
#in any case read the pin value
cat /sys/class/gpio/gpio172/value

Now, I'll be figuring out how to communicate with SPI and I2C devices hooked to the RR's expansion headers.

Sunday, October 6, 2013

Booting Linux on Radxa Rock

For anybody interested, here is a preliminary working kernel for the Radxa Rock board.

UPDATE (2013/10/11): IR and Audio I/O working, updated links to new flash images

Xubuntu Linux on Radxa Rock RK3188 board

This is WORK IN PROGRESS, so few things are supported, even though it's already almost as usable as any stick. This is the current state of affairs (check commits for updated info):

What works?
- Integrated Ethernet 10/100
- USB 2
- Flash
- MicroSD
- RK1000-S sound (integrated mic, audio out jack)
- IR (point a remote at it and press a button while monitoring cat /proc/interrupts ;)

What doesn't work yet?
- Wifi+BT

Not tested?
- Expansion headers
- SPDIF Audio Out / Exp.header Line In

I've based development on this repository. To compile it for yourself, you'll have to clone it and use the arch/arm/configs/rk3188_Radxa_Rock_Linux_galland_defconfig file as your .config (UPDATE: the initramfs file necessary for this defconfig, with rknand support, is initramfs-3.0.36+.cpio from here).

OR, if you have the board and want to go ahead with the flashing (supposing you already have an Ubuntu Linux in a MicroSD card inserted in the Radxa board), you have two options:

A) Boot to Linux only: Flash "kernel" partition with this file, so you boot directly to Linux (warning: it won't boot to Android unless you revert this step). Follow these steps:

//backup the Android "kernel" to be able to revert this
sudo ./rkflashtool r 0x4000 0x5000 > radxa_stock_kernel.img
//flash the downloaded Linux kernel instead
sudo ./rkflashtool w 0x4000 0x5000 < kernel.img
//safely reboot (into Linux)
sudo ./rkflashtool b

Now the Radxa board will boot directly into Linux.
In order to recover Android booting you'd have to reflash the stock kernel you've backed up.

B) Usual way: flash the Linux kernel (decompressing this different file) into the recovery partition to be able to dual boot Android/Linux:

//flash the downloaded and uncompressed recovery.img
sudo ./rkflashtool w 0x10000 0x10000 < recovery.img
//safely reboot (into Android)
sudo ./rkflashtool b

Now to get into Linux you have to boot into Android and have again two options:

- Enable Debug Mode, connect OTG USB port to PC, go to Settings->USB->Connect to PC, and then in the PC issue an "adb shell reboot recovery"

- Or, install an Android Terminal emulator (beware there is no Google Play in the RR!) get into it and, with root, issue a: reboot recovery

Since booting back to recovery partition tends to be cumbersome and since I only care about Linux, I have flashed kernel partition and always just boot to Linux!

PS: Just in case you don't have the latest modules and firmware, get these ones into the root (not "root" folder!) of your MicroSD Linux (I insert the uSD card in the PC, open a file explorer as root and copy them at uSD's root).

Booting to Linux instead of Android (flashing kernel partition) for RK

1) Get the software

You'll first need to fetch the Linux RK flashing tool (ignore the git name, the tool is valid for RK2*, RK30, RK31 devices), and the Linux rkcrc tool (for pre-/post-fixing a kernel image with the right RK values):

git clone rkflashtool
cd rkflashtool

git clone
cd rk-tools
make rkcrc
cd ..

2) Prepare your own kernel

After compiling your kernel, you have to convert/sign the zImage output file with this command:

./rk-tools/rkcrc -k Linux3188/arch/arm/boot/zImage kernel.img

Substitute "Linux3188" with the folder name where your kernel sources are.

3) Getting device into Recovery mode

This is most usually done by keeping pressed a pushbutton while plugging the USB OTG connector to your PC and then releasing it after a >2 seconds. If upon typing Linux command "lsusb" on the PC we see a nameless device (with ID 2207:...), that's our RK device in recovery mode.

Warning: some devices (like the Radxa Rock) have a 2 minutes timeout for the recovery mode after which it will power off, the timeout restarts upon any flash reading/writing.

4) Know your partitions

Before any flashing operation, we must be sure of at what offset and what size does the partition of interest have. To do it, while in recovery mode, issue the following command in your PC:

sudo ./rkflashtool r 0x0 0x1 | head -n 11

That will result in several lines of parameter information from your stick. We are interested in the last text paragraph, from "mtdparts" to the end.

For example, in my MK908 (check yours, since there are several versions), it is:
These are the partitions' parameters, where each partition is described as:
So the "kernel" partition in my MK908 is at offset 0x4000 and has a size of 0x6000  (= 24,576 sectors of 512 bytes each = 12,582,912 bytes).

Warning: If your partition of interest has an offset that is not multiple of 4MB, you MUST follow the "misaligned" instructions below, using your device's offsets.

Always remember that the flashing tool requires the numbers in reversed position compared to what appears in the mtdparts above, that is: first the offset, then the size, like:
sudo ./rkflashtool r (offset) (size)

5) Backup stock partitions!

I strongly recommend to read and store in your PC at least the boot, kernel, and recovery stock partitions of your devices, before you start tinkering with it!

This is very easily done with these commands (substitute the offsets and sizes with the ones from your mtdparts):

sudo ./rkflashtool r 0x4000 0x6000 > stock_kernel.img
sudo ./rkflashtool r 0xA000 0x8000 > stock_boot.img

sudo ./rkflashtool r 0x12000 0x10000 > stock_recovery.img

6) Flash your new kernel

Since my kernel partition's offset (0x4000 = 16384 => *512B = 8 MB) is a multiple of 4MB, flashing it becomes as easy as:
sudo ./rkflashtool w 0x4000 0x6000 < kernel.img

NOTE: If for some unknown reason you wanted to flash a kernel into boot partition, you should flash the same one that is used to flash recovery.

In order to ensure flashing is done correctly, always reboot the stick after flashing with the following command:
sudo ./rkflashtool b

Now your device will always boot to your kernel!

If it is a Linux kernel, then RK will directly boot to it, ignoring Android (until you revert these changes, by flashing back the stock kernel image)

Misaligned partitions (offset not at 4 MB boundary)

If the partition you want to write/flash starts at an offset not multiple of 4 MB then you will find that the first <4 MB that you write are wrong when/if you read them back.

An example of this can be found in Radxa Rock's boot partition, which starts at offset 0x9000 and has a size of 0x7000. If naively flashing boot with the command:

sudo ./rkflashtool w 0x9000 0x7000 < boot.img
That is: flash boot.img from 0x9000 up to, but not including, 0x10000 (= 0x9000 + 0x7000). If we proceed to read back the just written data, in order to verify it:
sudo ./rkflashtool r 0x9000 0x7000 > boot_readback.img

The result will be:
- Wrong, apparently random, data from 0x9000 up to, but not including, 0xA000 (note this is the first 4 MB boundary within boot partition)
- Good data from 0xA000 up to, but not including, 0x10000, that is: just what we've written, as expected.

This will cause the RK device to ignore the bad boot partition and jump on to recovery partition.

As far as I know, this problem arises when writing images, not when reading. It happens for Linux rkflashtool as well as for Windows RKAndroidTool.exe.

Workaround to flash misaligned partitions

The workaround I've found is to flash the previous partition (kernel) at the same time than boot partition. Why? Because kernel partition's offset is aligned to a 4 MB boundary, so there is no problem flashing it. Hence the operation would be:

//read stock kernel (padded to the full size of the partition)
sudo ./rkflashtool r 0x4000 0x5000 > radxa_stock_kernel.img
//concatenate the kernel partition with your own boot partition
cat radxa_stock_kernel.img my_own_boot_partition.img > kernelboot.img
//flash both partitions at the same time (0x5000+0x7000=0xC000)
sudo ./rkflashtool w 0x4000 0xC000 < kernelboot.img

//optionally read back to verify flashing operation
sudo ./rkflashtool r 0x4000 0xC000 > kernelboot_readback.img
//compare what we wanted to write, to what has been written
cmp -b kernelboot.img kernelboot_readback.img

//always safely reboot the RK device after flashing
sudo ./rkflashtool b

Friday, October 4, 2013

Radxa Rock board Linux development starts!

So, finally the Radxa Rock board (early developer sample) arrived.

First of all a big thanks to Tom Cubie and his team at Radxa for kindly donating it to further Linux development!


This tiny board looks fantastic, first and foremost because it has EXPANSION HEADERS, apart from this varied assortment of ports and peripherals:

- CPU RK3188 quad-core @ 1.6 GHz
- 8 GB Flash
- Wifi + BT
- 10/100 Eth
- Audio I/O + integrated microphone + SPDIF Out
- 2 full size USB ports
- 1 OTG USB port
- Power connector (doesn't take up a USB port)

Very nice specs but, to me, what makes this board unique is the expansion headers (along with openly available schematics!) where you can directly plug and access devices through I2C, SPI, UART, ... there is even an LCD output as well as another USB port at those pins!

First boot

After screwing in the protective top/bottom methacrylate plastics you will need a power adapter and an HDMI cable (normal size connector, not mini/micro) to attach it to a display and be ready to boot.

When you power the board for the first time it boots into Android, scaled 720p and in English, version 4.2.2 with kernel 3.0.36+ based on RK3188 R-BOX Android 4.2.2 SDK v1.0.0-130514.
However, beware, there is no Google Play, so you're on your own to install apps.

Hack it!

If you want to tinker with it, nothing easier, turn it off, unplug power and then: while keeping pressed the Recovery push-button nearby the USB ports, plug the OTG USB port to your PC.

A board red LED will light immediately, give it 2 seconds (or it'll boot Android) and, then, you can release the pushbutton and go to your PC.

I'm supposing you're on Linux, so if you go to a Terminal and write "lsusb" you'll see a new nameless device with ID: 2207:310b which corresponds to a RK3188 SoC awaiting flashing instructions.

IMPORTANT TIMEOUT: This recovery mode for flashing powers off after 2 minutes from last flash operation  or from being plugged in (if you do no operations). So it will issue a "power off" command through the serial console and just disappear from "lsusb" output. You'll have to re-plug it again to access it.

Then, through the usual Windows RK-provided tool or with the Linux rkflashtool that you can download and compile with ease from:   (I know, worst name ever, it does handle RK29* and RK31* too).

So download and compile the Linux USB flashing tool:

git clone rkflashtool
cd rkflashtool
sudo apt-get install libusb-1.0-0-dev
gcc -o rkflashtool rkflashtool.c -lusb-1.0 -O2 -W -Wall -s

Ask the Radxa board about its partition sizes and offsets:

sudo ./rkflashtool r 0x0 0x1 | head -n 11

the result should be:

MAGIC: 0x5041524B
ATAG: 0x60000800
KERNEL_IMG: 0x60408000
#RECOVER_KEY: 1,1,0,20,0
CMDLINE:console=ttyFIQ0 androidboot.console=ttyFIQ0 init=/init initrd=0x62000000,0x00800000 mtdparts=rk29xxnand:0x00002000@0x00002000(misc),0x00005000@0x00004000(kernel),0x00007000@0x00009000(boot),0x00010000@0x00010000(recovery),0x00020000@0x00020000(backup),0x00040000@0x00040000(cache),0x00200000@0x00080000(userdata),0x00002000@0x00280000(kpanic),0x00100000@0x00282000(system),-@0x00382000(user)

What does this tell us?

Well, for us Linux developers, used to sharing sticks' space with Android (unless you decide to wipe it and make your Linux take over the flash...) we are bound to flashing our kernels into the "recovery" partition so that we can dual boot Android or Linux by having the latter's filesystem in a MicroSD (also called uSD) card.

And then, from above parameters, where each partition is described as:
we infer that our "recovery" partition is at 0x10000 and has a size of 0x10000
Please remember that the flashing tool requires these numbers in reversed position (first the offset, then the size), like   sudo ./rkflashtool r (offset) (size)

Hence, once you have downloaded/compiled your Linux kernel and have a Ubuntu rfs (root file system) in your uSD, you can go ahead and flash your Linux kernel with these simple commands:

sudo ./rkflashtool w 0x10000 0x10000  <   my_kernel_recovery.img
sudo ./rkflashtool b

The "b" command safely reboots the RK device when the flash is written.

I'll initially be using my most up to date kernel for booting Linux, so you may see some activity there:

Have fun hacking!

Monday, September 16, 2013

DWC USB interrupt spam in Rockchip SoCs

Some time ago, when looking for answers to some USB 1.1 problems on the RK chips, like:
- USB 1.1 to Ethernet dongle not working
- Repeated or missing key presses
I stumbled upon some Raspberry Pi forums complaining about the same things.

This is logical since both the Rasp. Pi SoC and the RK3066/3188 are using the same USB HW IP cores (from DesignWare, that is DWC).

One of the problems they were reporting was an excessive number of interrupts per second coming from the USB while the system was idle (I've confirmed it, even with just a USB 1.1 keyboard attached).

One way to check this is with the console command:
vmstat 1
Looking below label "in" will tell you how many interrupts are there per second. In my case, Linux3188 kernel ( ) on a RK3188 Cozyswan S400 (kindly donated by KSK Electrics) the result is some 10600 interrupts per second.

And a way to see who is interrupting is simply to type:
cat /proc/interrupts
In my case, the "dwc_otg_hcd:usb1, dwc_otg_pcd" is the root cause for some more than 8000 interrupts per second, just as it happens to RPi guys.

The problem seems to stem from the DWC driver, as explained here:
"the Synopsys driver relies on the start of frame interrupt for scheduling transfers if "descriptor DMA" is not implemented (which it isn't, on BCM2835).  8000 is one interrupt per microframe."
and propose the use of a FIQ (ARM fast interrupt handler) for USB IRQs to discard USB microframes that contain no data. More info here...

And the FIQ enabled code lies in this repo (see commits starting in April 2013 by Gordon Hollingworth):

Testing real screen resolution on RK Linux


You think you have a 1080p screen resolution but when you draw a non-aliased one pixel width line and move the window around, it doesn't look one solid line anymore.

You're not on 1080p, but likely on a scaled 720p (even if Linux reports 1080p).


On a console type:

cat /sys/devices/platform/rk-fb/graphics/fb0/disp_info

for a working 1080p setup, the result should be something like:


That is xact and yact (the virtual resolution seen by Linux) must be the same as xdsp and ydsp (the real resolution been sent to the screen), or else you have a scaled output (x_scale and y_scale not 1.0).


If this is the case and you find that xdsp=1280 and ydsp=720, whereas you expected a 1080p display, the solution is in this commit:

BTW, notice the 16 bits color: RGB565
I couldn't tell the difference with normal desktop usage, but it means half the memory operations (32 bits ARGB888 <-> 16 bits RGB565).

Wednesday, September 11, 2013

Root Rockchip sticks/tablets from Linux

Since with every new device it becomes a nightmare (at least for me, that I seldom use Windows) to find the right Windows drivers for ADB access... I noticed that the rooting procedure (TPSparky Vondroid) relies only on the adb command and... that is readily accessible from Linux.

Hence this little guide came up as I put together the necessary little bits and has been tested with a RK3066 based tablet (RK3188 devices should also be rootable with this method, confirmed with Minix Neo X7).

1st Set up Linux ADB access to the device

This step is entirely done from the Linux PC (Ubuntu in this case) where the RK device is connected (through the USB). 

Please note that if your RK device has several USB ports, there will be only one that can be connected to the PC, usually marked as OTG or Slave.

1) Download the lightweight adb tools:
sudo apt-get install android-tools-adb
or, if you don't have Ubuntu, proceed to download and install the whole Android SDK :S

2) Open/create the following text file to let your standard user connect to the RK device:
sudo gedit /etc/udev/rules.d/51-android.rules
and then add a new line with the following text:
SUBSYSTEM=="usb", ATTR{idVendor}=="2207", MODE="0666", GROUP="plugdev"
Of course make sure that your Linux user is part of group "plugdev" with the following terminal command:
groups user

3) Force a reload of the USB access rules to get the latest one on:
sudo udevadm control --reload-rules

4) Add RK devices USB Vendor ID to what adb may expect for your currently logged user:
echo "0x2207" >> ~/.android/adb_usb.ini

2nd Set up the RK device for ADB access

Unplug the RK device from the USB and go to its Settings page to follow this steps:

1) Go to Settings->Storage and click at the top right on the three dots

2) Select USB computer connection in the dropdown

3) Select "Mass Storage" check box and go back to Settings

4) Browse down in Settings and select Security

5) In Security Settings, select "Unknown sources"

6) Back in Settings, select Developer Options

7) In Developer Options, select "USB debugging"

8) Close settings, and any other open apps.

9) Using a known good USB cable, connect the RK device to the PC, you'll get "Connecting to USB" in the notifications area in Android, and a USB icon in the notification area. You Do Not want to select "Connect to Pc to Transfer Files", in the notification menu, just plug in the cable and let it be.

PS: If you don't get a connection you may have to go back into Settings and select USB, then click on "Connect to PC".

3rd Rooting the RK device from Linux

The following steps are just a COPY of those in "TPSparkyRoot.bat" with a minor fix for correct SuperSU installation.

These are to be done on the Linux PC:

1) Download the package with the needed su, SuperSU, busybox, and RootExplorer binaries

2) Extract the package into a folder, open a terminal and cd into that folder

3) Create a new text file (i.e. "gedit") and copy the following contents:

   echo "*---* RK device Root Tool based on work by sunnydavid *---*"
   echo "--- Plug in your device, make sure debugging is enabled in Developer Options"
   echo "--- This script will now copy files over to your RK device"
   adb shell mv /data/local/tmp /data/local/tmp.bak
   adb shell ln -s /data /data/local/tmp
   adb reboot
   echo "--- Reboot 1/3 - Press Enter once the device has rebooted (if USB debugging doesn't appear in the Android bar you may have to click Settings->USB->Connect to PC)"
   read -p "or CTRL-C to exit"
   adb shell rm /data/local.prop > nul
   adb shell "echo \"ro.kernel.qemu=1\" > /data/local.prop"
   adb reboot
   echo "--- Reboot 2/3 - Press Enter once the device has rebooted (if USB debugging doesn't appear in the Android bar you may have to click Settings->USB->Connect to PC)"
   read -p "or CTRL-C to exit"
   adb shell id
   echo "--- If the ID shows as 0(root) then continue, otherwise CTRL+C to cancel and start over"
   read -p "CTRL-C to exit"
   adb remount
   adb push su /system/bin/su
   adb shell chown /system/bin/su
   adb shell chmod 6755 /system/bin/su
   adb push busybox /system/bin/busybox
   adb shell chown /system/bin/busybox
   adb shell chmod 0755 /system/bin/busybox
   echo "--- Installing SuperSU"
   adb push SuperSU.apk /system/app/SuperSU.apk
   adb shell chown root.root /system/app/SuperSU.apk
   adb shell chmod 0644 /system/app/SuperSU.apk
   adb push RootExplorer.apk /system/app/RootExplorer.apk
   adb shell chown root.root /system/app/RootExplorer.apk
   adb shell chmod 0644 /system/app/RootExplorer.apk
   echo "Completing Root ..."
   adb shell rm /data/local.prop
   adb shell rm /data/local/tmp
   adb shell mv /data/local/tmp.bak /data/local/tmp
   adb reboot
   echo "--- Reboot 3/3 - Your RK device should now be rooted!"

4) Make the script executable with: 
chmod +x

5) Restart the adb server by typing:
adb kill-server; adb start-server
6) And now when typing the following terminal command:
adb devices
you should see the connected device listed (my tablet for example is named 12345678...)

7) Execute the previously created script ( ./ ) and follow the steps to the letter, including patiently waiting for the potential several-minutes-long pauses

PS: After every reboot you MUST check that in the Android status bar the "USB Debugging" message appears (so that there is adb connection with the PC), or else you'll have to go back into Settings and select USB, then click on "Connect to PC". Only then you can hit Enter in the script to let it go on.

That's it, your device should be rooted, all from within Linux!

DISCLAIMER: I've followed this instructions to root my own devices without any trouble. The rooting procedure itself is simply the very widely used TPSparky method. However I take no responsibility for any mishappenings that may arise from your trying it on your devices. That is entirely yours.


Friday, July 12, 2013

Serial console mod for RK3188 stick Cozyswan S400

First of all, I would like to thank KSK Electrics for kindly donating this S400 stick (and a MK908) to support the Linux kernel porting on RK3188 devices.

A serial console for the Cozyswan S400 was needed in order to debug the boot problems encountered with the RK3188 kernel source "freed" by Andy from Rikomagic.

Also, since this kernel removed support for the framebuffer console (the text you [should] see on screen at boot, or in text Linux, before installing any XFCE/Gnome/...), the serial console is the only way to see what's happening when there is something wrong.

Fortunately I was able to successfully backport the framebuffer console so that now text is back on screen for Linux!

In any case here is the mod to get the serial console pins, thanks to Alok Sinha, who found those pads to be the serial output. They looked to me so much like an IR tx/rx connector that I would have tested every other pad before trying those! :)

Left to Right: The R (RX) pad for receiving keystrokes, the T (TX) pad for sending to your
Serial2USB module the text, and the required G (Ground) 

Actually only TX and Gnd are needed to receive the boot text, so I usually solder just those two.

The pads are small, with a tiny bit of solder on the iron tip and
another tiny bit on the cable, you should get good results

Omegamoon, who did the MK808 (RK3066) and MK908 (RK3188) sticks' serial console mods, recommends this nice Serial2USB module.

The three clams, as I call them

You can find here the Serial mod instructions for the Measy U2C RK3066 stick.

I hope it helps!

Tuesday, June 18, 2013

Full RK3066 Technical Reference Manual found!

Thanks to Omegamoon, who found it, here is the full RK3066 Technical Reference Manual, 1142 pages and 26MB of completely necessary data for opensource development on this platform.

It is the same file found in chinese websites and named as "Rockchip RK30xx TRM V2.0.pdf"

And here is the full folder shared with more RK29 and RK30 documents:


Monday, June 17, 2013

Measy U2C HW serial console pins

A hardware serial console is a little module soldered to at least two pin pads of a CPU (in this case the RK3066 of the Measy U2C stick) in order to transmit to a PC the stick's boot log sequence happening long before the screen turns on.

Naturally, if a kernel (like the latest one) refuses to boot and does not even get to turn on the screen, the only way to debug the reason is to use the serial console.

With the indications and the useful blogpost (for MK808) of fellow developer Omegamoon, I set out to find the two pins necessary to watch in my PC the boot sequence, that is:

- TXD: the pin where the CPU outputs the console text
- Ground: the necessary reference for the electrical levels in TXD

These had to be connected to the RXD and GND pins, respectively, of a hardware serial-to-USB module, that is then plugged to a PC's USB port. If anybody wonders about the CPU's RXD pin, it is not necessary unless you want to send commands to the stick once it's booted.

In the PC I myself use Ubuntu's CuteCom console app, connecting to serial port /dev/ttyUSB0 at 115200 bauds 8 bits data, 1 bit stop, no parity, and no handshake.

The way in which I found the RK3066's TXD pin was by trial and error. But since I saw there were small circular pads around the same area where Omegamoon had found his stick's TXD/RXD, I supposed mine should be close.

Hence, I first soldered the ground cable to the micro USB port on top of the RK3066 CPU and started very carefully making the serial-to-USB module's RXD pin's cable touch the different pads while powering on the stick, until I found the one that was transmitting data: TXD and proceeded to solder the cable and make the smallest possible hole in the Measy U2C's enclosure to get the two cables out of it while closed.

For reference, these are the places where the cables to the serial-to-USB are soldered
TXD pad of RK3066 (to RXD on serial-to-USB) and GROUND on the Micro USB connector

The TXD pad looks tiny but with a bit of care and a little bit of solder tin on the cable tip and the solder tip, you'll get good results.

Saturday, June 15, 2013

Linux on RK3188 work in progress

Please note this is not a blog post, it's a my personal pastebin where I can put in writing to myself and other fellow developers the small things I find in the latest 3.0.36 kernel that supports rk3188 devices.

MALI and CONFIG_SYNC ( sync_fence_cancel_async )

Talking about Omegamoon's codebase (but with backported drivers folder from my previous one)

Once MALI is configured in as:

# CONFIG_ION is not set
# CONFIG_MALI400_DEBUG is not set
# CONFIG_UMP_DEBUG is not set

And /include/drm/drm.h is backported to rk3x's version, that is, the first:
#if defined(__linux__)
#if defined(__KERNEL__) || defined(__linux__)

Then compiling throws this new error:
drivers/gpu/mali/mali/common/mali_kernel_core.c:1084:4: error: implicit declaration of function 'sync_fence_cancel_async'

That function is called because now CONFIG_SYNC is forced by PLAT_RK and should be implemented in /drivers/base/sync.c (see but it's missing.
Hence, since MALI has worked without CONFIG_SYNC and a quick grep of the codebase shows it's the only driver talking about CONFIG_SYNC, my fix for this is removing the dependence with PLAT_RK by going to /arch/arm/Kconfig
and in the config for PLAT_RK commenting the following lines:

# select SYNC
# select SW_SYNC
# select SW_SYNC_USER

and then in the .config removing:


Also had to add this line at the top of /drivers/video/rockchip/rk_fb.c due to missing references to 'GET_UMP_SECURE_ID_BUF2':

#include "mali_def.h"