#
✕ nick:maz
#asahi · 2022-07-27 08:15
mazmps: qemu + hvf?
#asahi · 2022-07-23 09:59
mazthere is a driver that has requested this interrupt, but it is not handling it. the kernel kills it after something like 100000 tries.
#asahi · 2022-07-22 12:56
mazdon't think there is one. maybe we should poll information!
#asahi · 2022-07-22 12:55
mazthere is no structure, that's the very point! ;-)
#asahi · 2022-07-22 12:54
mazboth really a mix of various arch revisions, with stuff ranging from 8.3 to 8.6... ARMv8.x is only a vague indicator these days.
#asahi · 2022-07-22 12:47
maznested virt,
#asahi · 2022-07-22 12:47
mazManouchehri: indeed, as N1 doesn't support NV (nor does KVM for now)
#asahi · 2022-07-22 11:02
mazGathering, Reordering, Early response
#asahi · 2022-07-22 10:59
mazthey certainly were advocating for the S2 mappings to be NC instead of nGnRE when mapping PCIe devices.
#asahi · 2022-07-22 10:58
mazodd. I've had robust discussions with some of the NV folks who were adamant they wanted to implement WC as NC. different groups, I suppose.
#asahi · 2022-07-19 16:35
mazlooks like a loader issue.
#asahi · 2022-07-19 16:33
mazalthough the "not a dynamic executable" is even more worrying...
#asahi · 2022-07-19 16:32
mazManouchehri: you inspect the source code.
#asahi · 2022-07-19 16:30
mazI'm pretty sure cuttlefish has plenty of 4kB-page hardcoded in it, because Android.
#asahi · 2022-07-19 13:18
mazj`ey: train had aircon, and I took the 6:47 this morning... I'll try a find a napping pod later...
#asahi · 2022-07-19 13:13
mazj`ey: I came into the office just to avoid cooking...
#asahi · 2022-07-19 13:08
mazstop bragging! :D
#asahi · 2022-07-19 13:05
mazhttps://developer.arm.com/documentation/den0022/db/?lang=en
#asahi · 2022-07-19 13:05
mazPower State Coordination Interface.
#asahi · 2022-07-17 15:11
mazwookey[m]: awesome, glad you put this M1 to a good use! :)
#asahi · 2022-06-27 13:48
maz(though writing the workaround for TX *was* pretty annoying).
#asahi · 2022-06-27 13:47
mazand that was a pretty easy one to address, given that we already had all the code.
#asahi · 2022-06-27 13:47
mazirrespective of the content of the registers.
#asahi · 2022-06-27 13:47
mazthe bug triggers when the guest does something that violates the GIC state machine.
#asahi · 2022-06-27 13:46
mazamarioguy: no, that's *not* the bug.
#asahi · 2022-06-27 13:45
mazwell, that's an easy one to check. just run an older version of EDK2.
#asahi · 2022-06-27 13:32
mazBTI*
#asahi · 2022-06-27 13:32
mazmore interesting, has PTI.
#asahi · 2022-06-27 13:20
mazTWED too (which is somehow useless, but hey...)
#asahi · 2022-06-27 13:20
mazyeah, and FGT.
#asahi · 2022-06-27 13:04
mazfuck. now I need to spend more money.
#asahi · 2022-06-27 13:03
mazwell, I didn't. or rather didn't want to believe it.
#asahi · 2022-06-27 13:03
mazmarcan: I hope this dump is a lie.. it says NV2...
#asahi · 2022-06-27 13:00
mazmarcan: thanks!
#asahi · 2022-06-27 12:55
mazmarcan: if you get a chance, a dump of the id registers would be good (if only to find out that there is no new stuff).
#asahi · 2022-06-21 13:09
mazgreat for a vertical ecosystem, but pretty bad for something like Linux.
#asahi · 2022-06-21 13:08
maznow userspace has to know "oh, I run on this particular flavour of fruity computer, let's interpret this random syndrome in a particular way".
#asahi · 2022-06-21 13:07
mazthe crap with this is that userspace doesn't really know what sort of braindead extension it has to deal with, which negates the idea of the usrspace ABI being an abstraction layer.
#asahi · 2022-06-21 13:02
mazone thing that *could* be done is to forward the stuff that isn't architectural to userspace and let it deal with it. Not sure you could work around everything, but I guess that's an interesting project for someone willing to mess with that sort of stuff.
#asahi · 2022-06-21 12:55
mazwell, if you think you can do a better job, I don't care.
#asahi · 2022-06-21 12:54
mazzx2c4: hey, feel free to go and maintain KVM/arm64 if that's what you want.
#asahi · 2022-06-21 12:53
mazj`ey: probably. assuming that's only for SMC.
#asahi · 2022-06-21 12:52
mazwe already have to deal with a 10k page spec. adding support for something *that doesn't have a public spec* is definitely not on my radar.
#asahi · 2022-06-21 12:51
mazwhy not make KVM more crap and unmaintainable, you mean?
#asahi · 2022-06-21 12:50
mazI only care about the stuff I maintain upstream. other architecture can pile whatever crap they want, that's none of my business!
#asahi · 2022-06-21 12:49
mazbecause KVM/arm64 is architectural, and only that.
#asahi · 2022-06-21 12:48
mazI'm not interested in supporting that in KVM.
#asahi · 2022-06-21 12:48
mazyou can do all sort of things if you want to sidestep the architecture, which HVF does.
#asahi · 2022-06-21 12:48
mazundefined instruction.
#asahi · 2022-06-21 12:48
maz" ESR : EC 0x00 IL 0x1 ISS 0x00000000 "

plus ›