The LPC32xx Project - introduction

For the last while I've been working on an extremely exciting project at work; certainly one of the best jobs to come my way. I've been lucky in my career to have worked on some really exciting things, and this project is certainly one of them.

We have a customer who has a range of hardware from some very old stuff to some very new stuff. They want the same versions of bootloader, kernel, and userspace running on all of them. Additionally they want, for each device, A/B updates, each of which is to run in its own container, and all built with OpenEmbedded/Yocto.

I'm not at all worried about finding support for the newish stuff. It's actually the oldest hardware that needs the most attention. For example, look at the mailing lists for the Linux kernel, U-Boot, or anything graphics-related and you'll find hundreds of patches from many developers every day all working on hardware so cutting edge, some of it isn't even available for purchase yet. But many fewer people are making sure the old stuff is still working. At best a developer will make sure their changes won't cause support for an older device to stop compiling, but sometimes that's not enough.

Earlier in my career I had an opportunity to do board bring-up. The company I was working for had just created their own custom board. It was based around a variant of the AMCC PowerPC 440 SoC. Our board, thankfully, was very closely modelled on the reference board for the specific SoC we were using, but with two key differences:

1. we were using a brand of SDRAM that differed from the one on the reference board (and therefore the timing parameters needed tweaking), and
2. whereas the reference board was running at a middle-of-the-road clocking, we wanted to run our board at the highest clock rate possible

Although upstream U-Boot already had support for the reference board, it was my job to update it to work on our board by completing these two tasks. Before starting this job, I hadn't even heard the phrase "board bring-up", but I was hooked! If I could have, I would have plotted a career from that time on that would have included a lot more board bring-up activities! But you take what you can get, and there are some other exciting things to do other than board bring-up.

What I enjoy most about board bring-up is it provides an opportunity to get deep down into the details of how an SoC works. We're all aware of SDRAM, and flash, and DMA, and all of the dozens of other pieces that fit together to make an SoC. But it's not every day one gets the opportunity to examine these things at the register level. Getting these two tasks working was very exciting for me; definitely a highlight.

Here I am, years later, and I find myself doing board bring-up again. However this time it's not with cutting-edge hardware, but rather with really old hardware: the NXP LPC32xx SoC. I realize it's strange for me to claim to be doing "board bring-up" for an SoC that clearly already has support in U-Boot and the Linux kernel. However the fact is the support for this device was added decades ago and (especially in the case of U-Boot) has bitrotted quite badly in the interim.

U-Boot is a very current and exciting project! It's not unusual for the daily patch count to run well into the hundreds. There are always new boards and SoCs in need of support, plus there are ongoing projects to improve the underlying structure of the code, its build system, and its test/ci infrastructure. With a code base moving this fast, an older device can quickly fall out of step with the rest of the project.

I've been busily working on this project for a while and enjoying every minute. I'll be blogging more about it!