# Racklet Community Meeting Notes 2025

This document contains the notes of the [Racklet](https://github.com/racklet/) community meeting. The meeting occurs every other Tuesday at [6 PM CET/CEST](https://dateful.com/convert/berlin-germany?t=6pm) (on even weeks). Check the [#racklet](https://osfw.slack.com/messages/racklet/) channel in the OSFW Slack for more info.

This document is best viewed and edited online: [](https://hackmd.io/PRLxctjyT8OUb0zvsMpgLw)

[TOC]

# November 11th, 2025 6:00 PM CET/CEST

:::info
- **Location:** https://jitsi.hamburg.ccc.de/racklet-community
- **Date:** November 11th, 2025 6:00 PM CET/CEST
- **Host:** @twelho
- **Participants:**
- Dennis Marttinen, @twelho
- Daniel Maslowski, @orangecms
:::

## Biweekly Recap

### Dennis

- Quite a lot has changed in the past month...
- Founded a new startup! :tada:
- Got tickets for 39C3
- Discussion on the future of Racklet
- 10" form factor got popular
- The spirit of Racklet also lives within other projects
- An end for now, repos will be archived :zipper_mouth_face:

### Daniel

- Made tons of progress with [intel_fw](https://github.com/platform-system-interface/intel_fw/)
- Should also benefit Racklet, given that there are also Intel based SBCs

# October 28th, 2025 6:00 PM CET/CEST

:::info
No meeting
:::

# October 14th, 2025 6:00 PM CET/CEST

:::info
No meeting
:::

# September 30th, 2025 6:00 PM CET/CEST

:::info
- **Location:** https://jitsi.hamburg.ccc.de/racklet-community
- **Date:** September 30th, 2025 6:00 PM CET/CEST
- **Host:** @twelho
- **Participants:**
- Dennis Marttinen, @twelho
- Daniel Maslowski, @orangecms
:::

## Biweekly Recap

### Dennis

- Did some hacking with friends on the SiFive X280 cores in a [Tenstorrent Blackhole P100a](https://tenstorrent.com/hardware/blackhole) AI accelerator card
- Booted up Linux using [tt-bh-linux](https://github.com/tenstorrent/tt-bh-linux)
- Also, mainline patches have [just been submitted](https://lore.kernel.org/lkml/[e-mail]/)
- Ran some [terminal-doom](https://github.com/cryptocode/terminal-doom)
- Also did a [Coremark run](https://openbenchmarking.org/result/2509165-NE-MUSTAAUKK18) with the Phoronix Test suite

### Daniel

- Gave a [talk on the status of RISC-V in mainline Linux](https://metaspora.org/riscv-mainline-linux-2025.pdf) at LinuxDay.at
- Reworking the oreboot build system and tooling
- Bootstrapping Allwinner H616 (we can run and print!)

# September 16th, 2025 6:00 PM CET/CEST

:::info
No meeting
:::

# September 2nd, 2025 6:00 PM CET/CEST

:::info
No meeting
:::

# August 19th, 2025 6:00 PM CET/CEST

:::info
No meeting
:::

# August 5th, 2025 6:00 PM CET/CEST

:::info
- **Location:** https://jitsi.hamburg.ccc.de/racklet-community
- **Date:** August 5th, 2025 6:00 PM CET/CEST
- **Host:** @twelho
- **Participants:**
- Dennis Marttinen, @twelho
- Daniel Maslowski, @orangecms
:::

## Biweekly Recap

### Dennis

- MSc thesis finally submitted :tada:
- Will be working on Racklet software over the next couple of weeks on the x86 test cluster
- First "production" tests, esp. for the filesystem stuff
- Will try to find time to boot the Rock Pi 5B
- Capability evaluation, tests with both [ubuntu-rockchip](https://github.com/Joshua-Riek/ubuntu-rockchip) and mainline
- DRAM PHY reverse engineering on the horizon (quite different compared to RK3566...)

### Daniel

- Only made tiny steps with oreboot on the RK3566
- Found someone interested from the linux-rockchip community
- We still have a [Matrix channel](https://matrix.to/#/#oreboot-dev:matrix.org) btw, but lost the Slack bridge

# July 22nd, 2025 6:00 PM CET/CEST

:::info
No meeting
:::

# July 8th, 2025 6:00 PM CET/CEST

:::info
- **Location:** https://jitsi.hamburg.ccc.de/racklet-community
- **Date:** July 8th, 2025 6:00 PM CET/CEST
- **Host:** @twelho
- **Participants:**
- Dennis Marttinen, @twelho
- Daniel Maslowski, @orangecms
:::

## Biweekly Recap

### Dennis

- Thesis work is taking all allocatable time at the moment
- Will be submitted at the end of the month
- After that should have more time for Racklet again
- Found this interesting official firmware tool from Qualcomm in Rust: https://github.com/qualcomm/qdlrs
- Looked very briefly more into the TrackPoint firmware, seems to be 8K-bootloader-padded 8051

### Daniel

- Pushed a few more things into [psi-spec](https://platform-system-interface.github.io/psi-spec/application-processors.html) and [LinuxBoot](https://github.com/linuxboot/book/pull/145)

# June 24th, 2025 6:00 PM CET/CEST

:::info
- **Location:** https://jitsi.hamburg.ccc.de/racklet-community
- **Date:** June 24th, 2025 6:00 PM CET/CEST
- **Host:** @twelho
- **Participants:**
- Dennis Marttinen, @twelho
- Daniel Maslowski, @orangecms
:::

## Biweekly Recap

### Dennis

- Focus on thesis work, not much time to work on Racklet/firmware stuff
- Swapped new SSDs into the x86 test cluster
- Pro tip: don't deploy DRAM-less SSDs in a 24/7 application
- My Linux patch for Ceph is getting backported to 6.1, 6.6, 6.12, 6.15 :partying_face:
- Offtopic: dug a little bit into ELAN firmware for TrackPoints: https://github.com/twelho/trackpoint

### Daniel

- Again back to the RK356x, turns out I took the wrong config
- It is LPDDR4 for sure, had some part mistranslated earlier
- Compilers can be very good at optimizing loops btw, e.g., [4 similar ones](https://github.com/u-boot/u-boot/blob/903eb123236ccbd8ef05d43507a2a910b785bd56/drivers/ram/rockchip/sdram_rv1126.c#L944-L978) down to 1 (and another one around it).

# June 10th, 2025 6:00 PM CET/CEST

:::info
- **Location:** https://jitsi.hamburg.ccc.de/racklet-community
- **Date:** June 10th, 2025 6:00 PM CET/CEST
- **Host:** @twelho
- **Participants:**
- Dennis Marttinen, @twelho
- Daniel Maslowski, @orangecms
:::

## Biweekly Recap

### Dennis

- Looked into Radxa Orion O6 and its [upstreaming](https://github.com/System64fumo/linux/blob/main/hardware/devices/arm/radxa/orion/orion.md)
- Geerling's SBC reviews: https://github.com/geerlingguy/sbc-reviews/issues/62
- It's a bit of a mess, the SoC vendor CIX is providing a Device Tree (kind of) with custom kernel
- There's also UEFI (SystemReady) firmware, but it's also very broken and basically only there for booting Windows (disables some of the cores that don't work under Windows)
- This would be a good candidate to include for a universal LinuxBoot/DT "Rackboot" implementation
- [PoePwrNegotiator](https://github.com/orbitrod/PoePwrNegotiator) is a cool thing
- Blog post: https://roderickkhan.com/posts/2025-05-16-poe-uefi-solution

### Daniel

- Continued with the RK3566
- Controller looks very much the same as for earlier SoCs, just the PHY differs a lot
- Generally, the controller appears to be simple vs the PHY being rather complex

# May 27th, 2025 6:00 PM CET/CEST

:::info
- **Location:** https://jitsi.hamburg.ccc.de/racklet-community
- **Date:** May 27th, 2025 6:00 PM CET/CEST
- **Host:** @twelho
- **Participants:**
- Dennis Marttinen, @twelho
- Daniel Maslowski, @orangecms
:::

## Biweekly Recap

### Dennis

- Still working on upstreaming the kernel patch
- Now getting reliable benchmark results from PTS (`fs-mark`)
- Some benchmarks are still failing -> figured out during the meeting, disk quotas are being exceeded

### Daniel

- Getting further with the RK3566
- Recognized parts of the code from the open source U-Boot code for earlier SoCs

### Discussion

- Looked a bit into SMM
- https://doc.coreboot.org/drivers/smmstorev2.html
- Exchanged about power/thermal management
- ACPI https://uefi.org/htmlspecs/ACPI_Spec_6_4_html/11_Thermal_Management/thermal-objects.html
- PSCI https://developer.arm.com/documentation/den0022/latest/
- SCMI https://developer.arm.com/documentation/102886/001?lang=en

# May 13th, 2025 6:00 PM CET/CEST

:::info
No meeting
:::

**Note**: Daniel is at RustWeek. :crab: :calendar:

# April 29th, 2025 6:00 PM CET/CEST

:::info
- **Location:** https://jitsi.hamburg.ccc.de/racklet-community
- **Date:** April 29th, 2025 6:00 PM CET/CEST
- **Host:** @twelho
- **Participants:**
- Dennis Marttinen, @twelho
- Daniel Maslowski, @orangecms
:::

## Biweekly Recap

### Dennis

- Finally found the source of the distributed filesystem performance problems I've been debugging for a year now: it's a **kernel bug** (and another one is likely a Talos Linux bug)
- Next steps are to figure out how to contribute patches upstream
- The fix increases file ops/s by 10-20x on my x86 test cluster, making distributed filesystems such as Ceph also viable for Racklet-class SBCs
- It'll probably even run fine on stuff like the Spacemit K1, the overhead is very low now

### Daniel

- Rockchip DRAM init making progress :hourglass_flowing_sand:

# April 15th, 2025 6:00 PM CET/CEST

:::info
- **Location:** https://jitsi.hamburg.ccc.de/racklet-community
- **Date:** April 15th, 2025 6:00 PM CET/CEST
- **Host:** @twelho
- **Participants:**
- Dennis Marttinen, @twelho
- Daniel Maslowski, @orangecms
- Philip Molloy, @pamolloy
- Ron Minnich, @rminnich
:::

## Biweekly Recap

### Daniel

- Got back to Rockchip
- Created https://github.com/platform-system-interface/rk_boot
- Got first prints on RK356x, already [in oreboot](https://github.com/oreboot/oreboot/tree/main/src/mainboard/rockchip/rk3566/bt0)
- Dumped the mask ROM
- Started reversing [vendor binaries](https://github.com/rockchip-linux/rkbin/blob/master/bin/rk35/rk3566_ddr_528MHz_ultra_v1.10.bin)
- Dumped OTP areas, both "secure" and non-secure
- DRAM init will take time, first steps not yet working
- Confirmed that Ky X1 ([Orange Pi RV2](http://www.orangepi.org/html/hardWare/computerAndMicrocontrollers/details/Orange-Pi-RV2.html)) is indeed a rebranded SpacemiT K1x
- oreboot code (still in WIP branch) just works on it
- Went to Embedded World and Hannover Messe
- Philip and I are doing a LinuxBoot revival
- Reworked the website to be more of a landing page
- Historical references moved over to the book
- Planning to shift the focus to a more general approach
- Reevaluating rootfs and kexec implementations
- Started some work on Qualcomm, having a hard time with setup
- Cannot find a UART TX pin on the two devices I have
- Ordered two older [DragonBoard 410C](https://www.arrow.de/products/dragonboard410c/arrow-development-tools)s for further development
- Started https://github.com/platform-system-interface/qc_boot
- https://mainlining.org by folks from CCCAC, pmOS et al

### Dennis

- KubeCon was a blast, quite a lot of people joined my talk
- Recordings *just* uploaded 1h ago, [here's my talk!](https://www.youtube.com/watch?v=QbR908kgk1Y)
- Disclaimer: I haven't even watched it myself yet
- Lots of people contacted me from various national and academic supercomputing settings
- More work on distributed filesystems, learning some hard lessons about latency and overhead
- Acquired some M.2 enterprise PLP (Power Loss Protection) SSDs, now testing `fsync` overhead with different Ceph + JuiceFS combinations
- Still battling with sequential performance, no combination has gotten over 20 files/s in fs-mark yet
- This shouldn't be CPU-bound, but I need to somehow verify that
- This affects the viability of distributed filesystems on Racklet

### Device Tree discussion

- How do we deal with kernel + DT matching?
- Recompile for the target kernel on demand?
- Device Tree syntax is versioned, but content is not
- Linux has a commitment to keep DTs being a pure hardware description
- https://docs.kernel.org/devicetree/usage-model.html
- https://docs.kernel.org/devicetree/bindings/ABI.html
- DT ABI stability [a myth](https://bootlin.com/pub/conferences/2015/elc/petazzoni-dt-as-stable-abi-fairy-tale/petazzoni-dt-as-stable-abi-fairy-tale.pdf)?
- video was pulled from YouTube :( ([last reference](https://unix.stackexchange.com/questions/265890/is-it-possible-to-get-the-information-for-a-device-tree-using-sys-of-a-running))
- Does the DT belong with the bootloader rather than with the kernel?
- It used to be that way.
- General vs special purpose applications

# April 1st, 2025 6:00 PM CET/CEST

:::info
No meeting
:::

# March 18th, 2025 6:00 PM CET/CEST

:::info
- **Location:** https://jitsi.hamburg.ccc.de/racklet-community
- **Date:** March 18th, 2025 6:00 PM CET/CEST
- **Host:** @twelho
- **Participants:**
- Dennis Marttinen, @twelho
:::

## Biweekly Recap

### Dennis

- Ceph latency diagnosis results: Samsung consumer NVMEs really dislike synchronous I/O
- Ceph's BlueStore does synchronous I/O for the WAL and the RocksDB database
- Ordered some enterprise-grade NVMEs with power loss protection now -> they transform sync writes to async using supercaps
- Need to wait a couple of weeks to get the drives and get to benchmarking again
- EU's new [DARE initiative](https://dare-riscv.eu/) sounds interesting
- 240 M€ funding for a general purpose, vector and AI RISC-V chip design
- The general purpose chip is the most interesting, let's see if they have enough time and funding to execute

# March 4th, 2025 6:00 PM CET/CEST

:::info
- **Location:** https://jitsi.hamburg.ccc.de/racklet-community
- **Date:** March 4th, 2025 6:00 PM CET/CEST
- **Host:** @twelho
- **Participants:**
- Dennis Marttinen, @twelho
- Daniel Maslowski, @orangecms
:::

## Biweekly Recap

### Dennis

- Further Ceph latency diagnosis
- Narrowed it down to [`fs_mark`](https://github.com/josefbacik/fs_mark) and many real-world applications, including `rsync`, issuing `fsync` (or some other kind of sync) and causing the MDS journal to be flushed after every operation
- This takes some time, which is probably proportional to the time it takes to write an object to the OSDs (disks)
- The default object size is 4 MiB, which takes around 50 ms to write -> this matches with the latency I'm observing
- Now running test compiles with different flushing/object size etc. tweaks to try and narrow it down further
- Ceph is nasty parallel C++ code though so it's very hard to profile...

### Daniel

- We got Linux to boot on the K230
- Turns out, the "debug register" for the timer was really the normal timer control register after all (STC = System Timer Clock), I just mixed up the columns in the manual again; though the explanation there is still confusing and inconsistent and it doesn't say anything about being related to the mtimer (RISC-V platform timer)
- Had to add some more helpers, now have shared libraries for XuanTie and RISC-V
- Did another Twitch live development stream, started a new series with the K230: https://www.youtube.com/watch?v=hfz8QBB4M3g
- We are cleaning up a lot in oreboot
- Ripped out tarpaulin (test coverage report)
- Removed obsolete Makefiles and targets
- Upgraded to Rust nightly 2025-03-01
- Discussed [compression in oreboot](https://github.com/oreboot/oreboot/issues/772), we'll try `miniz_oxide`
- Easy now with the QEMU RISC-V port and the loader device \
`qemu-system-riscv64 [...] -device loader,addr=0x82200000,file=Image.gz`
- Will be at Embedded World

## Discussions

We talked about hwmon in Linux, SECO making the UDOO Bolt and other boards, how their Embedded Controller can be accessed, and looked at their closed source library: https://git.seco.com/software/eapi/eapi

# February 18th, 2025 6:00 PM CET/CEST

:::info
- **Location:** https://jitsi.hamburg.ccc.de/racklet-community
- **Date:** February 18th, 2025 6:00 PM CET/CEST
- **Host:** @twelho
- **Participants:**
- Daniel Maslowski, @orangecms
- Dennis Marttinen, @twelho
:::

## Biweekly Recap

### Daniel

- oreboot can run barebox just fine
- Plus, it runs both in QEMU and [RVVM](https://mastodon.social/@CyReVolt/113971038824718149)
- Got the [platform timer on the K230 to work](https://mastodon.social/@CyReVolt/113991917309418454)
- Needed to set a bit in a debug register that is not remotely documented to be related AFAICT, someone else found it in vendor U-Boot code
- Got SG2002 DRAM init to work, with a little hardcoding hack
- Auto detection needs some fixing somewhere, complicated
- [SBI test with timer works](https://mastodon.social/@CyReVolt/114021889908485034) as well
- WIP libraries for oreboot
- Architecture specific helpers for XuanTie platforms
- Utilities for MMIO access dropping the `unsafe`
- Memory test

### Dennis

- Still fighting with Ceph head latency in the x86 test cluster, will probably need to open an issue to get more insights from the experts
- The [Sipeed NanoCluster](https://x.com/SipeedIO/status/1887709604349399413) looks very interesting for Racklet
- 7 SOMs in a Raspberry Pi form factor
- Each SOM with up to [4x Cortex-A76 or 8x Cortex-A55](https://x.com/SipeedIO/status/1891338072656089259) sounds *very* interesting, better than Pi 5 performance (if cooled properly)
- Up to 448 cores in an 8 compute module 2U Racklet tray!
- Gigabit "trunk" from the SOM carrier might be limiting
- RAM amount per SOM is the final decider, hopefully at least 2 GiB per core

## Discussions

We had another look at Arm, its SMM equivalent, discussed how x86 SMM and RISC-V M-mode are similar, and came across

- https://trustedfirmware-a.readthedocs.io/en/latest/design/firmware-design.html
- 
- Making Arm look like PC: https://uefi.org/sites/default/files/resources/UEFI_Plugfest_March_2016_AMI.pdf
- SMM to circumvent OS: https://disobey.fi/2025/profile/disobey2025-284-sounds-from-the-basement-extreme-low-level-game-cheating
- RK3588 trusted firmware: https://gitlab.collabora.com/hardware-enablement/rockchip-3588/trusted-firmware-a
- https://interrupt.memfault.com/blog/arm-cortex-m-exceptions-and-nvic

We also discussed DTB shenanigans, see

- https://fosdem.org/2025/schedule/event/fosdem-2025-6224-woa-laptops-a-quest-for-getting-the-right-dtb/
- https://learn.microsoft.com/en-us/windows-hardware/drivers/dashboard/using-chids

Finally, on European chips, HPC, etc

- https://ecs-brokerage-event.eu/#overview
- https://www.chips-ju.europa.eu/Projects/
- https://eurohpc-ju.europa.eu/document/download/7e21bf39-bb19-43b8-88b6-dd51650d348e_en?filename=pdf%20chips%20parrt%202.pdf
- https://eurohpc-ju.europa.eu/research-innovation/our-projects/european-processor-initiative-epi_en
- https://www.fz-juelich.de/en/ias/jsc/jupiter/tech

# February 4th, 2025 6:00 PM CET/CEST

:::info
- **Location:** https://jitsi.hamburg.ccc.de/racklet-community
- **Date:** February 4th, 2025 6:00 PM CET/CEST
- **Host:** @twelho
- **Participants:**
- Dennis Marttinen, @twelho
- Daniel Maslowski, @orangecms
:::

## Biweekly Recap

### Dennis

- Debugging Ceph performance on the x86 test cluster
- Very fast read IOPS and decently fast write IOPS
- The MDS (file creations/deletions etc.) is for some reason limited to only around 15 files/second, which is very slow
- The MDS journal is probably the culprit, have yet to dig into the profiling details
- Kernel 6.14 has support for C906 vector extension (0.7) [errata](https://www.phoronix.com/news/Linux-6.14-RISC-V)

### Daniel

- Was at FOSDEM, including 2 OSFW hack days
- Ported [DRAM init for K230D to oreboot](https://github.com/oreboot/oreboot/pull/764)
- Extended [`kendryte_boot`](https://github.com/orangecms/kendryte_boot) so you can load additional code to RAM
- Added `main` stage in oreboot and dummy SBI
- Simple [SBI print test works](https://mastodon.social/@CyReVolt/113930859208849486)
- Currently debugging mtimer, `mtime` is stuck at `0`

# January 21st, 2025 6:00 PM CET/CEST

:::info
- **Location:** https://jitsi.hamburg.ccc.de/racklet-community
- **Date:** January 21st, 2025 6:00 PM CET/CEST
- **Host:** @twelho
- **Participants:**
- Dennis Marttinen, @twelho
- Daniel Maslowski, @orangecms
:::

## Biweekly Recap

### Dennis

- Daniel hinted at the newly established [mini rack project](https://mini-rack.jeffgeerling.com/) from Jeff Geerling
- Uses the same 10" rack form factor as Racklet
- 3D-printed plates to mount Pi's etc.
- Other useful shelves/mounts listed there as well
- Power delivery is a challenge – as we've already discovered with Racklet
- PDU discussion [here](https://github.com/geerlingguy/mini-rack/issues/5)
- UPS discussion [here](https://github.com/geerlingguy/mini-rack/issues/1)
- These are mostly 110/230 V though, nothing yet for USB-C PD etc.
- Network switch lists are useful for Racklet
- Looking forward to the community discovering powerful-enough PoE switches for multiple nodes
- Software/firmware side severely lacking though:
- No firmware security/boot integrity/Open Compute -level secure boot
- Very basic (mutable) Kubernetes setups only
- TL;DR: Useful for sourcing parts for the Racklet hardware side, though we want more density and better firmware security + actual Open Compute -level software (immutable Kubernetes)
- Achieved over 14k 4K direct random read QD2 IOPS with Ceph on the x86 test cluster! :chart_with_upwards_trend:
- In layman's terms: distributed filesystems can be made very snappy at a small scale as well
- This proves the viability of Ceph on lower-end hardware, SBCs such as the Rock 5B should not have any issues running this stack
- 8 GiB of RAM, 4 threads and Gigabit ethernet is sufficient

### Daniel

- Made lots of progress on the BL808
- Got the loader tool to the point where it can load any binary for any core to run immediately
- Got [PSRAM working](https://mastodon.social/@CyReVolt/113852399724541310)
- Did another [dev live stream on it](https://www.youtube.com/watch?v=wCvZkpvgpm0)

## Additional notes

We took a closer look at the RISC-V SBI, including the Hart State Management extension and specifically RustSBI, some projects using RustSBI, and how global singletons are initialized in them.

Refs:
- https://github.com/riscv-non-isa/riscv-sbi-doc/releases/tag/v3.0-rc3
- https://github.com/rustsbi/rustsbi/network/dependents
- https://github.com/arceos-hypervisor/riscv_vcpu
- https://github.com/rustsbi/rustsbi-d1/blob/main/see/src/main.rs
- https://docs.rs/rcore-console/latest/src/rcore_console/lib.rs.html
- https://github.com/rustsbi/rustsbi/blob/main/sbi-rt/src/dbcn.rs
- https://github.com/rcore-os/rCore-Tutorial-v3/blob/main/os/src/console.rs
- https://github.com/mvdnes/spin-rs/blob/master/src/once.rs

# January 7th, 2025 6:00 PM CET/CEST

:::info
- **Location:** https://jitsi.hamburg.ccc.de/racklet-community
- **Date:** January 7th, 2025 6:00 PM CET/CEST
- **Host:** @twelho
- **Participants:**
- Dennis Marttinen, @twelho
- Daniel Maslowski, @orangecms
:::

## Biweekly Recap

### Daniel

- More thorough test of the TH1520 showed its DRAM wasn't really working
- Got back on the Bouffalo Lab BL808
- Started a [loader tool](https://github.com/orangecms/bl_boot) as usual
- Slowly figuring out how the boot process and the many efuse bits work
- Lots of [open resources from the community](https://github.com/openbouffalo/) and [the vendor](https://github.com/bouffalolab/)
- 38C3 was a blast :tada: - blog post on OSFF website coming soon:tm:
- We bootstrapped the Canaan K230 SoC during a workshop
- Reversed its mask ROM in a previous one (got started)

### Dennis

- More work on the x86 Racklet software architecture test cluster
- Ceph (distributed storage) is now running very well
- Encountered some bugs in Cilium (networking)
- Talos Linux has an [annoying bug](https://github.com/siderolabs/talos/pull/10060) that prevents upgrades right now
- Taught myself KiCad over the holidays, received first PCB from JLC
- Haven't yet gotten into messing with the RISC-V ESP32C6
- Explored Linux real-time scheduling on a Raspberry Pi 3
- `threadirqs` is very useful for multicore embedded systems