#
✕ nick:maz
#asahi · 2023-10-05 08:26
mazkidplayer666: you have a very restrictive view of what a hypervisor does. direct access to HW from a VM is very common.
#asahi · 2023-10-05 08:04
mazeh, even my SO says "the wifi is down" when I reboot the fibre gateway (and the APs are up...). I pretend I don't understand...
#asahi · 2023-09-30 15:14
mazj`ey: (just got back from hols) wow, they desoldered most of the stuff from the board. at this stage, they would have been better off building their own...
#asahi · 2023-09-21 11:08
mazar: 10G is supported out of the box (it's a PCI device).
#asahi · 2023-09-04 10:28
mazor they use 10bit addressing, which is pretty rare...
#asahi · 2023-08-24 08:33
mazdsharshakov: adding high-speed switches to Central Scrutinizer would be a bit of a challenge -- the price isn't the issue ($1.6 at DigiKey), but the PCB design is. I'll have a look in my copious spare time.
#asahi · 2023-08-21 08:26
mazsort of. only without any HW garantee.
#asahi · 2023-08-15 13:18
maz(for the monitors)
#asahi · 2023-08-15 13:17
mazand the USB-C connection is very likely to be DP, not TB...
#asahi · 2023-08-11 07:47
mazyup, that's the one.
#asahi · 2023-08-11 07:44
mazM1 studio.
#asahi · 2023-08-11 07:44
mazalexander: "03:00.0 Ethernet controller: Aquantia Corp. Device 04c0 (rev 03)"
#asahi · 2023-08-06 18:25
mazlena6: amusingly, variable length instructions *are* part arm64 in the form of the SVE move prefix and the memory copy instructions (the latter always come in group of 3 instructions -- 12 bytes).
#asahi · 2023-08-04 22:31
mazI hadn't read the ML much this week (travelling), but it looks like I didn't miss much.
#asahi · 2023-07-24 14:17
mazplus, on non-arm64 architectures, people still believe that KVM actually emulates the CPU that QEMU asks for... which isn't true, but hey...
#asahi · 2023-07-24 14:15
mazwaldi: because QEMU is about emulation. the KVM backend is only a side hack.
#asahi · 2023-07-24 09:42
mazit's pretty funny how far you can go with just a couple of shell scripts and screen...
#asahi · 2023-07-24 09:40
mazdottedmag: I understand that people want a "universal" virt framework, but it really feels that libvirt is a horrible ball of tangled wires... I tried to use it twice, and wrote my own, KVM-specific ball of wires.
#asahi · 2023-07-24 09:37
mazwaldi: "-accel kvm". old style QEMU command line.
#asahi · 2023-07-24 08:39
mazrycloud: "ls -l /dev/kvm"?
#asahi · 2023-07-24 08:38
maz(or set permissive permissions on /dev/kvm)
#asahi · 2023-07-24 08:38
mazyeah, that's a potential issue. burn libvirt.
#asahi · 2023-07-24 08:38
mazGuest6702: ah, I see what you mean now.
#asahi · 2023-07-24 08:37
mazGuest6702: no. we already know /dev/kvm accessible, since QEMU is able to open it (and rycloud is running as root...)
#asahi · 2023-07-24 08:36
mazrycloud: let's not create additional problems by switching distro. we want to understand the issue, not paper over it.
#asahi · 2023-07-24 08:35
mazfor a start, you should remove this "kvmclock" from your XML. it has no purpose on arm64, and may well be the cause of the error you're seeing
#asahi · 2023-07-24 08:31
mazOK, so you QEMU is OK. this looks like a libvirt problem, for which I'm utterly incompetent.
#asahi · 2023-07-24 08:29
mazrycloud: try "taskset -c 0-3 qemu-system-aarch64 -machine virt,accel=kvm -cpu host" and post the resulting error
#asahi · 2023-07-24 08:13
mazj`ey: thanks. I nominate you as my official decoder! :D
#asahi · 2023-07-24 08:13
mazthe rpi stuff is probably asking for 40bit IPA, which will not work on a mini.
#asahi · 2023-07-24 08:12
mazalso, you need to pin QEMU on a single cluster, it will get upsset if vcpu threads cross cluster boundaries.
#asahi · 2023-07-24 08:06
mazwaldi: can you *please* describe the error your seeing with 8.0.3?
#asahi · 2023-07-24 08:04
mazDebian carries 7.2.2, which works fine on M1 mini, M1 Ultra, M2 and M2 Pro.
#asahi · 2023-07-24 08:03
mazrycloud: see above. at least 7.0.
#asahi · 2023-07-24 08:02
mazrycloud: you need to upgrade your QEMU. this version is too buggy to support the small IPA limit on M1 mini (it would work on an M1 pro though).
#asahi · 2023-07-24 08:01
mazanything before that is busted.
#asahi · 2023-07-24 08:01
mazfor the record, the first version of QEMU that has a chance to run with KVM on M1 is 7.0.
#asahi · 2023-07-24 08:01
mazI run KVM 24/7 on M1/M2, and if things don't work, I really want to know.
#asahi · 2023-07-24 08:00
mazwaldi: if you get errors, please give me a detailed bug report.
#asahi · 2023-07-24 07:58
mazthere is no way this is going to run on M1.
#asahi · 2023-07-24 07:58
mazOMG: "QEMU 6.2.0"
#asahi · 2023-07-24 07:56
mazor "qemu-system-aarch64 --version", which hopefully returns the same thing.
#asahi · 2023-07-24 07:55
maz(run "qemu-system-arm --version")
#asahi · 2023-07-24 07:54
mazrycloud: what is the version of your QEMU?
#asahi · 2023-07-24 07:54
mazit could also be an old QEMU build that doesn't grok the IPA limitations of M1.
#asahi · 2023-07-13 11:47
mazinvalid format smells of another architecture.
#asahi · 2023-07-11 17:35
mazmartinr: yeah, I've been told... :-/
#asahi · 2023-07-11 17:29
mazmartinr: more seriously, I'd be happy to run things by you if you're interested. You can have a look at the two repos jannau__ pointed you to (and I'll try to push the latest state to a branch).
#asahi · 2023-07-11 17:28
mazmartinr: spare time? how I envy you! :D
#asahi · 2023-07-11 17:21
mazmartinr: if you need a board, get in touch.

plus ›