Yoctoproject is a solid framework for creating Linux distributions for embedded devices, and its layering mechanism makes it easy to add Linux OS to new target devices; With highly customization for a particular device it can include custom start-up scripts, software packages built with a high degree of optimization for a particular architecture and different user graphical interfaces from full Gnome desktop to a simple a serial console.
YP has 2 basic layers, board support packages layer and core layer; In the BSP layer is where we include all the custom software and configuration tweaks for a particular platform; The core layer provides the common software stack to provide from a simple command line interface to Sato desktop interface (Matchbox based and Gnome mobile stack). A third layer could be added to provide additional user interfaces such as LXDE, XFCE, Active plasma and more; YP is quite flexible.
One of the problems with Freescale i.MX based platforms is the lack of openness, getting the BSP provided by FSL requires to download a large tarball which bundles the kernel, boot loader, user space applications and Ltib an ancient build system which generates a file system so 2004, literary the software stack is around those years probably older. And that’s why I started working on YP BSP support for i.MX5 targets, and now the project has it’s on life with active members helping on getting the best support possible.
Freescale iMX53 quick start board is an Arm cortex-a8 low cost development board, with interesting peripheral interfaces like Sata II, hdmi and GPU based on AMD Z160. And YP BSP layer has a decent support for it, at github Freescale site there are instructions on how to get started.
But with newest i.MX6 processor a Wandboard should be more appealing.
Many people had contact me asking for references about my thesis and decide to post them online so any one could take a look at the work I did last year, hope this helps to every one interested; The general goal was:
Hardware/software co-design approach for re-configurable architectures based on the Xilinx ML507 PowerPC 440 platform.
This thesis aims to find a reliable solution for bringing the Linux operating system on embedded devices, focusing on Xilinx re-configurable architectures as a base platform.
I will continue to maintain Yoctoproject Xilinx meta layer  as soon I get some free time, hope in the next week’s I will publish some patch’s, I’m going to set a personal goal, meta-xilinx should get official BSP layer by Yoctoproject release 1.2 target for April 2012.
Wow another year ended and I’m being just to quite busy concluding with my master degree and doing part time open source projects. From March I begin working on Linux board bring up for several FPGA platforms under the umbrella of the Linux foundation and Google, as a GSoC project. By concluding my participation on GSoC represented the first of my goals for 2010 being a GSoC student; The second goal was to conclude with my studies which I did just a few weeks a go. The third one was getting back to professional labor market and GSoC counts, and from September I’ve been collaborating for a secret company which have many associates around the world and is so secrecy that we don’t know each other, well not exactly.
Working on both projects was quite challenging, in my school GSoC was perceived as a distraction, at some point it reach a critical point since part time activities are not allowed. Some school polices are obsolete and prevent collaboration between universities and companies, far from any other countries where participation in programs similar to GSoC are encorage and students can receive credits for their participation. On top of that in the first semester I was taking 3 courses, GSoC and thesis wow I was with just to much activities to do, I quit with my sport activities just to recover some time and increase my workdays to weekends, by the second semester I was concluding with GSoC and starting doing the heavy work of my thesis, still no weekends free until late September when I decide to celebrate with some friends and went to a nice place to have a few drinks which ended in nothing good I was victim of alcohol check points mafia in Mexico city and end up in jail for the weekend :(.
In November I attend to Linux Plumbers conference at Boston, where I meet a lot of people like some Ubuntu developers, Yocto project contributors and kernel hackers, well I need to continually attend this type of events since is where I can learn from others, working from home and living in a large city without a decent Linux user group does not help.
The only negative aspect in this year is that I don’t plan to continue participating in open source conferences in Mexico city, is hard when you have to deal with people that thinks that are free software contributors because they use open office. Imagine Debian community polices and apply it to a group with some technology knowledge but can’t write a line of code, um not exactly better wording is people that use open software but in any form they contribute to its development. Most Mexico city Lugs are just to elitist, and the “elite” group just sucks! So what can be done to form a decent LUG? That’s another blog story.
I’m a music lover all types indy rock, reggae, hiphop even k-pop; So hanging around with an broken ipod isn’t fun; I was tempted on getting a new media player but just looking that I have a few more media players around I decide to bring them back to live with Rockbox starting with Go-gear mp3 player that had a broken micro drive which I replace it with compact flash card increasing the storage space from 6Gb to 32Gb :).
I had a lot of trouble restoring the firmware even using Phillips recovering tools until I found good references at Rockbox wiki ; The next step is to actually replace vendor firmware with Rockbox, there are some driver issues that I can look as a side pet project but firstly I need a battery replacement and can find it in Mexico so I need to order one from a global battery vendor.
Gsoc midterm evaluations pass last week, and there’s some good news about the status of my project, since the close relation between Openembedded and Poky I extend the support for Xilinx platforms to Poky which is faster than OE to build the rootfs for the reason that are far less recipes to parse and some others features. While Poky support for virtex5 platforms resolves my thesis problem to provide a graphical desktop, it also makes easier the software development for new users by using the Poky Anjuta plug-in although in can be used with OE.
Microblaze support still needs some work, but I’m close to complete it there’s no major changes
that I need to add to OE just resolve some issues with gcc and binutils related on shared libraries
linkage, there was reported a similar problem in OE mailing list, and going to start looking on that
And since Gsoc program is approaching to the end, I’m considering to start looking on labor
market, but first I need to complete my thesis and unfortunately this means a two month slip
:(, at the moment I’m having some hard to track issues on a hardware bridge from the system
bus to a simple peripheral and requires to learn how to use chip-scope (virtual logic analyzer)
to track the signals in order to validate its behavior. Thank fully there are some channels on
YouTube where I can learn in a simple and pain less way thanks to several Indian Technology
Institutes for having lectures online.
In this lasts weeks I’have been working on Openembedded
for generic support for Xilinx virtex4, virtex5; Only u-boot and the
Linux kernel needs attention to be configured for the final target arch
since it can be Powerpc 405, 440 or Microblaze. Final arch settings needs
some user configuration in local.conf file it should be set the target
The machine can be xilinx-virtex4, xilinx-virtex5 or xilinx-mb-gen (generic microblaze)
the final arch can be override by setting TARGET_* values, for example for virtex5
family board targeting microblaze
TARGET_ARCH = “microblaze”
TARGET_CPU = “microblaze”
TARGET_TUNE = “microblaze”
I know that this is kind of odd but until I fix how to override this values inside conf
machine files this would be useful for testing stage.
The logic for setting the target options for u-boot and the kernel are handled in
xilinx-bsp.class where I add some logic for the configure settings based on the
TARGET_ARCH and XILINX_BOARD values, a similar approach is needed in config machine
files to do the black magic on setting the default options for a particular target
I had problems with running any ace file (fpga bitstream + elf executable) from compact
flash on ML405 target board, on power-up the board was just flashing a led indicating an
error on compact flash read, and to solve the problem I had to add an option on
I’m not sure but it seems that my compact flash is the problem;
There’s a lot still to be done specially for Microblaze support, where I have to still
deal with glibc or uClibc to be integrated in OE, or to get working an external toolchain
Also fully test Angstrom on ml405 board since it’s failing when running init scripts it
complains about permissions errors.
I start working on OE microblaze support and is kind of tricky,
there are several efforts on updating gcc, binutils, glibc, and
I’m integrating from Xilinx repos since they are the most updated
versions available for microblaze; Still glibc build is failing
due to a mis configuration in binutils; So I have to options,
add a prebuild toolchain in OE or fix glibc; the exact error occurs at
linking the final objects ld complains about “unsupported emulation” to
workaround this there’s an run-time variable that can be used to set the
emulation mode `LDEMULATION` but fails in a later stage, so the error
roots are in binutils configuration.
OE changes are stored at http://gitorious.org/oe-microblaze/oe-microblaze
Xilinx provide me some target platform (ML405, ML507), for testing/validation
one of my concerns was import taxes from Mexican customs, but I provide the
paperwork to claim professional development usage to be extent of pay taxes,
well that’s one good thing of Mexican custom regulations, and I didn’t pay a
single dollar :). Thanks John and Grant for making possible to have some
hardware at home, that would free the only single ML507 board from my
From Google every student is still waiting on welcome package and payment
cards, they are estimated to be delivered in two weeks, well lets wait and
see if Google is generous and send us a cool nexus one ;).