#
✕ nick:maz
#asahi · 2026-11-16 09:06
mazchaos_princess: I can't possible comment! ;-)
#asahi · 2026-11-16 08:10
mazthis probably won't go anywhere, but at least I got a chance to (publicly) point out how bad things are...
#asahi · 2026-11-16 08:09
mazchaos_princess: ah, thanks! didn't realise people could watch that already!
#asahi · 2026-05-12 12:32
mazah, interesting. I'll keep an eye on that...
#asahi · 2026-05-12 12:31
mazI though M3+ was doomed for KVM due to not being able to access the timer masks?
#asahi · 2026-05-12 12:29
mazalso, be aware that nested support is extremely experimental, and while I'm quite happy with it, it has a lot of very rough corners...
#asahi · 2026-05-12 12:22
mazLinux has gained that recently, openbsd is still in flux, and I'm not sure win is anywhere near that.
#asahi · 2026-05-12 12:21
mazwanting VHE, certainly. expecting to be booted directly in VHE is unusual.
#asahi · 2026-05-12 12:20
mazthat'd explain why it boots correctly at EL1, and dies at EL2.
#asahi · 2026-05-12 12:20
mazone possibility is that Windows requires nVHE to boot, and M2 cannot provide that -- it is firmly VHE only.
#asahi · 2026-05-12 12:19
mazplayboicarti[m]: well, I assumed that, given that you are using KVM.
#asahi · 2026-05-12 12:18
mazchaos_princess: no idea, I've never booted windows. but it does require PMUv3 indeed, and QEMU will expose one.
#asahi · 2026-05-12 12:16
mazis that clearer?
#asahi · 2026-05-12 12:15
mazto use "virtualization=on" in QEMU you need to be running an M2 with a very recent kernel and pass "kvm-arm.mode=nested" on the kernel command-line.
#asahi · 2026-05-12 12:14
mazplayboicarti[m]: what's that what?
#asahi · 2026-05-12 12:13
maz(and that you use an M2 of course)
#asahi · 2026-05-12 12:13
mazs/s/z/
#asahi · 2026-05-12 12:12
maz virtualisation=on works correctly here, provided that you boot the host with kvm-arm.mode=nested.
#asahi · 2026-03-10 09:56
mazDebian does what I need. might not suit everyone.
#asahi · 2026-02-20 11:41
mazjbe: DB is awesome compared to trains in the UK!
#asahi · 2026-02-09 16:32
mazI'm pretty sure things would go wrong, with USB being re-enumerated.
#asahi · 2026-02-09 16:32
mazno
#asahi · 2026-02-09 16:31
mazand of course, no suspend/resume. definitely not recommended if you don't have prior experience in hacking your way around.
#asahi · 2026-02-09 16:30
mazffx: I'm using such a setup, but only for desktop machines.
#asahi · 2026-01-16 14:47
mazLinux will happily spawn a getty on it, and you can login from there.
#asahi · 2026-01-16 14:46
mazfl0_id: yes. it is a full interactive serial console, as you'd expect from a bidirectional UART.
#asahi · 2026-01-16 09:06
mazfl0_id: awesome, glad you were able to fix it.
#asahi · 2026-01-14 09:52
mazyup. I had to build a testing jig, and ended up with about 15% of the boards needing rework.
#asahi · 2026-01-14 09:50
mazsven: ah, true. these things are rather fiddly, and JLCPCB is mediocre at best...
#asahi · 2026-01-14 09:48
mazjannau: I remember that in your case, it was the micro-USB one. but given that fl0_id doesn't even observe VBUS switching, I'd suspect the USB-C side of the CS to have an issue.
#asahi · 2026-01-14 09:36
mazyou may want to check the USB-C connector for some bad soldering.
#asahi · 2026-01-14 09:35
mazif you have all the lines, then it should work. the problem is that not even VBUS is seen.
#asahi · 2026-01-14 09:32
mazfl0_id: you can hack the firmware to replace ^_ with anything else. but it appears that the FUSB doesn't even see the remote device.
#asahi · 2026-01-13 08:52
mazfl0_id: enable debug from the session, disconnect the cable, reconnect it, dump the log somewhere where I can see it.
#asahi · 2026-01-12 13:27
mazfl0_id: but that's exactly when a UART is useful. you can instrument the kernel to spit something on the UART without requiring anything to work.
#asahi · 2026-01-12 12:57
mazfl0_id: CS doesn't rely on the target having USB working.
#asahi · 2026-01-12 11:00
maznoopwafel: glad it was useful!
#asahi · 2026-01-12 10:47
mazput them on tindie, they should disappear pretty quickly (I still get weekly requests...).
#asahi · 2025-12-17 10:42
mazthat smells like memory corruption, since this VA is not a valid one.
#asahi · 2025-11-17 12:02
mazserial console FTW.
#asahi · 2025-10-30 13:45
mazchaos_princess: it's more subtle than that. IOMMUs usually have multiple page-tables, one per set of devices that share the same "source-id" (stream-id, requester-id...).
#asahi · 2025-09-29 10:03
mazauscyber_: then do you want to trust them with your business?
#asahi · 2025-09-29 06:24
mazif your PCB manufacturer can't work that out, you're talking to the wrong people.
#asahi · 2025-09-29 06:23
mazauscyber_: what do you mean? pin 1 is clearly indicated by a dot. it's not exactly ambiguous.
#asahi · 2025-09-25 14:02
mazm4 mini should be very close to that then.
#asahi · 2025-09-25 14:01
mazauscyber_: you have the normal M4 though, right? not max nor pro?
#asahi · 2025-09-25 13:58
mazeyes so's M4 mini...
#asahi · 2025-09-25 13:58
mazeyes so
#asahi · 2025-09-23 17:54
mazcy8aer: M2 air suspend consumption is what still leaves to be desired. otherwise, it's great!
#asahi · 2025-09-23 08:48
mazyou tell me.

plus ›