Notices by Manawyrm | Sarah (manawyrm@chaos.social)

@erincandescent built in ethernet with a gdb port just via IP — back in 1999!!

modern debuggers could learn a thing (or two!)

@erincandescent Abatron used to be Swiss, but I guess that almost counts (this is going to make some swiss people mad 🫣😅)

@derf no comment :P

@hennichodernich @whitequark that's basically parallel ATA :D

@caligula9997 😹

@ryanc @gsuberland hmm, sad -- I don't see anything very helpful :/

@gsuberland @ryanc yeah, glasgow can just do that upstream these days.

@ryanc @gsuberland no, those are sadly just fiducial markers:
https://en.wikipedia.org/wiki/Fiducial_marker#Printed_circuit_boards

I don't see any relevant pins/testpoints on the photos on openwrt.org, maybe something below the RF shielding? hmm..

@ryanc yeah, that does require a bit of pratice first... :(

holding the CPU in reset would probably work, but I didn't see anything that looks like JTAG pins on your board.

@gsuberland @ryanc yeah, I've been there... this happened to me on a NXP QorIQ PowerPC platform and then I had to add an inline resistor with the CS line like this:

That was a bit tricky...

@gsuberland @ryanc IPQ40xx (like in Ryan's AP) will go High-Z on the flash chips when in reset, though, I've done this in the past.

@ryanc if all else fails, I do crap like this:

TOSLINK is one of the very few consumer fiber optical standards. It transports digital, uncompressed PCM audio over plastic fibers.

Why not transport IP-over-Toslink instead?
It's a 1.536 Mbit/s data link after all...

https://www.youtube.com/watch?v=9uWi8iWym9c

New blogpost:
https://kittenlabs.de/ip-over-toslink/

/cc @n0emis @benjojo

@dalias @marcan "You're wrong to want that and here's why" sounds good to me as a user. I learned something and know what to expect now.

@mia there is a funny redneck on YouTube who has done exactly that (but I can't remember the channel name now. Guy with a big beard...)

tl;dr: Works, EMI is a nightmare, otherwise possible.

Got a chance to play with new WiFi hardware:
BananaPi BPI-R4, mt76-based WiFi 7 7/802.11be access point.

Support in mainline OpenWRT is now good enough to "just work".
No MLO support yet, WIP.

Client is a Intel BE200 M.2 card, 2x2 MIMO.
Was able to get MCS13 (aka 4096QAM) working at 320 MHz, which results in a 5.7Gbit/s PHY rate and ~3Gbit/s of actual TCP throughput.

BE200 drivers on Linux still seem a bit wonky, "only" 2 Gbit/s on Linux.
Windows 11 works as expected.

https://openwrt.org/inbox/toh/sinovoip/bananapi_bpi-r4

@ryanc fs optics come with pretty nice PDF datasheets, always worth having a look into.

Basically the only thing you don't want to do is to connect 60km optics with a meter of cable, or you'll blow the receiver out of the other end :)

@astrid @ryanc indeed! most transceivers have some protection against operating at high powers without a bidirectional link, but I would _NOT_ recommend going anywhere near the fiber with your eyes on a high power transceiver.

the 0dBm of your transceiver is about 1mW, which is still eye-safe (still don't do it!), anything above that is getting seriously dangerous!

@ryanc Check the sfp-rx-power shown is within the maximum receive levels of your SFP used — otherwise, sure.

For a typical 2km transceiver, those values are perfectly good.

@ryanc bootloader correctly installed as removable? (aka BOOTX64.EFI file present on the FAT partition?)

EFI non volatile storage is a huge, unreliable shitshow...