#
✕ nick:maz
#asahi-dev · 2023-05-04 15:43
mazI'll get to try it then.,
#asahi-dev · 2023-05-04 15:43
mazmarcan: fair enough, thanks for having investigated it. eventually, I need to rebuild my t6002 anyway, as it still runs on the stuff I built when I got it last year (it hasn't rebooted yet).
#asahi-dev · 2023-05-04 14:21
mazOK, so it all edges on this magic sysreg?
#asahi-dev · 2023-05-04 14:08
mazhuh.
#asahi-dev · 2023-05-04 13:25
mazbecause if you're interrupting while running a guest, it takes significantly longer to handle the interrupt than if you were running userspace.
#asahi-dev · 2023-05-04 13:24
mazI wonder if that has something to do with how long it takes to ack the interrupt.
#asahi-dev · 2023-05-04 13:24
mazhmmm. I positively hate magic.
#asahi-dev · 2023-05-04 13:08
mazmarcan: what latency do you measure? from what point to what point? m1n1 or KVM?
#asahi-dev · 2023-05-04 11:28
mazknow*
#asahi-dev · 2023-05-04 11:27
mazanyway, if you find a KVM bug, you now where to shout!
#asahi-dev · 2023-05-04 11:27
mazwell, we set IMO+FMO+AMO, so *everything* has to take us to EL2. If that's not obeyed, that's pretty bad.
#asahi-dev · 2023-05-04 11:24
mazmarcan: interesting. this would mean that a host interrupt doesn't kick us out of a guest? that'd be terminally broken...
#asahi-dev · 2023-05-03 19:12
mazfixed the NV bug that was affecting M2-pro. funny how tiny IPA spaces can hide amusing bugs.
#asahi-dev · 2023-05-02 15:53
mazkettenis: wouldn't that mean adding USB+whatever support to the stage-1 m1n1? seems difficult.
#asahi-dev · 2023-05-02 14:08
mazhuh, looks like some IPA space issue. limiting it to 32bit makes it work.
#asahi-dev · 2023-05-02 14:06
mazhmmm. EDK2 in a NV guest works fine on M2, but not M2 Pro (random exception). Linux itself works fine on both... something is amiss.
#asahi-dev · 2023-05-02 13:33
mazbut yes, this is *very* niche.
#asahi-dev · 2023-05-02 13:33
mazmarcan: only if you go the UEFI way, AFAICT. if you just ask for a raw m1n1 install (proxy only), you don't get that.
#asahi-dev · 2023-05-02 13:27
mazit is just a bit harder to bootstrap the system if you have to depend on an initial ESP.
#asahi-dev · 2023-05-02 13:26
mazoh, of course it is better!
#asahi-dev · 2023-05-02 13:24
mazcreating*
#asahi-dev · 2023-05-02 13:24
mazbut isn't that similar to the current situation where creaking a m1n1+fw+u-boot image isn't distributable?
#asahi-dev · 2023-05-02 13:16
mazkettenis: do you still plan to support the initrd method for u-boot firmware? it's pretty handy for tethered setups without an ESP.
#asahi-dev · 2023-05-01 17:51
mazah, m1n1 of course have them. sorted.
#asahi-dev · 2023-05-01 17:45
mazdoes someone have the part numbers of M2 max CPUs? I'm adding the -pro versions to the vgic SError shit-list, and I might as well do the full M2 range, since it is likely to be the same old story.
#asahi-dev · 2023-05-01 16:08
mazmarcan: in any case, I'd love the person with the problem to show up on the list. if they are experiencing issues, they probable aren't the only ones (and there's a long list of details I need to know about...)
#asahi-dev · 2023-05-01 16:01
mazcould be cpufreq, as I'm not using any PM on my boxes (frequencies set at boot time).
#asahi-dev · 2023-05-01 16:00
mazeven if an IRQ ends up on the "wrong" core, we can service the interrupt in about 1000 cycles, which wouldn't register at all, so something else must be at odds.
#asahi-dev · 2023-05-01 15:58
mazmarcan: it'd be good to have some details. I'm running my desktop in a VM on the Studio, and otherwise compiling on the host. the host definitely doesn't feel sluggish...
#asahi-dev · 2023-05-01 11:04
maz... and I thought my 12MB mini+firmware+dt+u-boot was big... :D
#asahi-dev · 2023-05-01 10:35
mazmarcan: my scripts are based on a m1n1 version that is about 2 year old... that might explain why.
#asahi-dev · 2023-05-01 10:01
maz+1.
#asahi-dev · 2023-05-01 09:54
mazagreed. this is more specific to my own use case where I'm not quite in a position to boot the OS yet, and having to fish for keyboard/mouse every 20 minutes is driving me potty.
#asahi-dev · 2023-05-01 09:50
mazlife*
#asahi-dev · 2023-05-01 09:50
mazthis machine is likely to be tethered for the rest of its like anyway.
#asahi-dev · 2023-05-01 09:49
mazyeah, figured out that it is way easier to add this to my chainload script and forget about it.
#asahi-dev · 2023-05-01 09:05
mazawesome, thanks!
#asahi-dev · 2023-05-01 08:57
maz(if only I could find what it is)
#asahi-dev · 2023-05-01 08:56
mazI'm getting the itch to replicate this in m1n1 or u-boot...
#asahi-dev · 2023-05-01 08:56
mazdoes anyone remember which bit of the kernel is responsible for resetting the boot count that invalidates the m1n1 install after 10 or so failed boots?
#asahi-dev · 2023-04-30 13:37
maz(hopefully I can now consider turning the M2 MBA into a real laptop and not just a dev board... ;-)
#asahi-dev · 2023-04-30 13:37
mazwe're in business! a lot of thanks to everyone!
#asahi-dev · 2023-04-30 13:36
mazjannau: "U-Boot 2023.04-00021-g1b30640649-dirty (Apr 30 2023 - 14:28:03 +0100)"
#asahi-dev · 2023-04-30 13:06
mazexcellent. m1n1 is alive on the machine, thanks a bunch!
#asahi-dev · 2023-04-30 13:01
mazjannau: brilliant! I seem to remember that u-boot works with the same constraints?
#asahi-dev · 2023-04-30 12:54
mazyup, much happier now. I'll report back once I'm done with it (I'll probably need a pointer to a DT and such...)!
#asahi-dev · 2023-04-30 12:48
mazjannau: right, let me try that now.
#asahi-dev · 2023-04-30 12:46
mazhttps://paste.debian.net/1278918/
#asahi-dev · 2023-04-30 12:41
maz"This device is not supported yet!"
#asahi-dev · 2023-04-30 12:37
mazI'll give the pre-build installer a go, which is likely the best option! ;-)

plus ›