This is the 18th post in a series tracking the development of Tock, a safe multi-tasking operating system for microcontrollers.

1. Tutorial at SenSys
   1. What will the tutorial cover?
   2. Is this tutorial right for me?
2. Pull Requests
   1. Merged
   2. Proposed

We bombarded ourselves this week with two continuous-sampling ADC implementations for the SAM4L (OK, @brghena has been working on one of them for a while), a SPI-slave implementation, MPU bugs and bug-fixes, a Lua runtime and merging of @bradjc’s long-standing accelerometer virtualization pull-request. We officially welcomed @petarpenkov, @mog96, @bbbert and @ptcrews to the team. They are from Stanford CURIS students doing research on Tock this summer and have already proposed two really high quality pull requests. And, as the last update mentioned, the SenSys tutorial is happening this fall, and we’re starting to hammer out the details.

@ppannuto, @brghena and @alevy found and finally addressed (almost completely) a long standing issue with how process memory had to be organized due to MPU constraints on the Cortex-M (PRs #375 and #384). The problem had been that the MPU requires regions to be aligned to their size, which has to be padded to a power-of-two. So, if you had 18kB process, the region covering it had to be padded to 32kB and it’s base address had to be 32kB aligned. This resulted in really large padding between processes and basically put a relatively low limit on how large processes could be. The fix uses MPU subregions when necessary to allow the base to be aligned at a finer granularity (basically 1/8th of the next largest power of two), which in practice means that a 18kB process only needs to be padded to 20kB (instead of 32kB) and the base address aligned to 8kB.

Tutorial at SenSys

Our proposal was accepted to host a tutorial at SenSys 2017 (the ACM Conference on Embedded Networked Sensor Systems) in Delft, The Netherlands on November 5th! You can sign up for updates and to be notified when registration opens here. But, meanwhile, the cliffnotes.

What will the tutorial cover?

The agenda for the tutorial is not yet finalized, but it will be 4-hours in the afternoon, and will cover:

• Introduction to Tock goals and design
• Introduction to hail and imix hardware platforms
• Hands-on Writing processes
• Introduction to Rust
• A hands-on tour of the kernel
• Hands-on excercise Writing new drivers and system calls

Is this tutorial right for me?

That’s for you (or whoever is paying your registration fee) to decide.

The tutorial is geared towards the research community (particularly the sensor networks community) and will therefore focus on using Tock in that context. This means that the excercises we’ll use will involve radios and sensors and we’ll tailor them to what we think the sensor network community would find most useful.

Pull Requests

Merged

• (#359) @bradjc upstreamed the LTC294X driver from Signpost

• (#238) @bradjc’s long-standing 9dof accelerometer virtualization pull-request was merged! There is still some discussion around the exact best way to virtualize drivers more generally, but this is definitely a step in the right direction.

• (#380) @bradjc renamed fxos8700cq to be consistent with other driver naming.

• (#378) @alevy added Lua userland support and a minimal example Lua app.

• (#375) @alevy added MPU subregions for the text segment, allowing the text segment to be aligned at finer granularity than exactly its size.

• (#377) @ppannuto updated MPU tests and added system calls to gather process memory information from userland.

Proposed

• (#384) @alevy generalized MPU subregions support to data, GOT and IPC sections.

• (#383) @petarpenkov implemented ADC continuous sampling for the SAM4L with variable clocks.

• (#381) @bbbert implented a SPI slave with DMA for the SAM4L.