#
#asahi-dev · 2026-10-11 13:56
sventhough i guess then they'd need to do two-stage translations if they want to protect SPTM
#asahi-dev · 2026-10-11 13:55
sveni'm not sure what secure kernel does, it might just built its own pagetables
#asahi-dev · 2026-10-11 13:55
sventhe nice part about SPTM and XNU is that they both share pagetables and VBAR
#asahi-dev · 2026-10-11 13:55
svenyeah, i need to swap VBAR when switching between EL and GL to the right shadow tables
#asahi-dev · 2026-10-11 13:54
nickchanam sure you are doing that with things like VBAR already
#asahi-dev · 2026-10-11 13:54
nickchanif you are running more than one VM than you do need to keep track of register states
#asahi-dev · 2026-10-11 13:44
sventhe other concern i had with that hack is that EL1 can write to CPACR_EL1 as well
#asahi-dev · 2026-10-11 13:43
svenoh, nice!
#asahi-dev · 2026-10-11 13:43
nickchanit has all the bits on FEAT_NV2p1, but it seems not even M6 has that
#asahi-dev · 2026-10-11 13:43
nickchanhttps://support.arm.com/documentation/111107/2026-09/AArch64-Registers/CPACR-EL1--Architectural-Feature-Access-Control-Register?lang=en
#asahi-dev · 2026-10-11 13:41
svenit felt like a big hack to me because only a few fields line up iirc
#asahi-dev · 2026-10-11 13:40
nickchan(though they messed it up because some bits are not stateful in CPACR_EL1 but are in CPTR_EL2, and had to make FEAT_NV2p1 to fix that)
#asahi-dev · 2026-10-11 13:40
svenhuh, really?
#asahi-dev · 2026-10-11 13:40
nickchanit is what ARM does with FEAT_NV2
#asahi-dev · 2026-10-11 13:40
nickchannit: the CPTR_EL2 -> CPACR_EL1 redirection is not a hack
#asahi-dev · 2026-10-11 13:36
sventhanks, and yup, that's a typo!
#asahi-dev · 2026-10-11 13:22
jannau#695 looks ok except for a typo in a comment
#asahi-dev · 2026-10-11 12:56
svenalso, yay, looks like my m5 air will arrive tomorrow a week and a half ahead of schedule \o/
#asahi-dev · 2026-10-11 12:48
svenany chance i could get https://github.com/AsahiLinux/m1n1/pull/695 reviewed (or just blindly merged because it won't break anything for end-users :-P). kinda want to continue building the rest of the SPTM crap on top of that
#asahi-dev · 2026-10-11 07:14
sveni'll try to get the rest ready in the next days now that i'm back from vacation
#asahi-dev · 2026-10-11 07:14
svenyuka: don't think that's enough for m4, lukas' PR should be enough to run SPTM on M2/M3 though
#asahi-dev · 2026-10-11 06:35
MichaelReeves[m]cropped the screenshots now and selected just the best ones (imo). fine with any tbh, cant decide how to narrow further: https://drive.google.com/drive/folders/1UDeoxf9VtCyi-lpMrXROD2epa9Q27ZFk?usp=sharing
#asahi-dev · 2026-10-11 03:47
kitlithprobably not going to try that right *now*, or even in the next month or two, since everything seems to be happening all at once right now. but this does help a usecase I'm interested in.
#asahi-dev · 2026-10-11 03:38
kitlithI've been occasionally poking at monado and wivrn on my M1 Pro, having an environment to run games that does not require muvm would make running most VR games with monado in particular actually possible. (Writing a monado bridge for muvm is (i think?) more or less blocked on the same thing that explicit sync is)
#asahi-dev · 2026-10-11 03:36
kitlithGot it. I thought I remembered something along those lines, but it's been awhile so I wanted to double check.
#asahi-dev · 2026-10-11 03:35
JamesCalligeros[m]So yes, p3s fixes these in the sense that fex now works at all
#asahi-dev · 2026-10-11 03:35
JamesCalligeros[m]arm64ec or otherwise
#asahi-dev · 2026-10-11 03:35
JamesCalligeros[m]fex straight up does not work in 90% of cases with 16k anymore
#asahi-dev · 2026-10-11 03:30
kitlithQ regarding p3s: I take it there are issues occurring with wine arm64ec+FEX on 16k page size that do not occur with p3s?
#asahi-dev · 2026-10-11 03:26
JamesCalligeros[m]Nah just whatever you can stomach looking at. Window decorations would be nice at least though
#asahi-dev · 2026-10-11 03:25
MichaelReeves[m]anything specific you'd like in the window?
#asahi-dev · 2026-10-11 03:25
MichaelReeves[m]ah true
#asahi-dev · 2026-10-11 03:25
JamesCalligeros[m]Screenshots with all the extra trimmings dont really look very good inline with a blog post
#asahi-dev · 2026-10-11 03:25
JamesCalligeros[m]Just the game window please
#asahi-dev · 2026-10-11 03:23
MichaelReeves[m]or if you'd like different formatting i.e. bigger window, no info screen on side etc.
#asahi-dev · 2026-10-11 03:22
MichaelReeves[m]lmk if that's a good variety or if you'd like anything else
#asahi-dev · 2026-10-11 03:22
MichaelReeves[m]chaos_princess: took a bunch of screenshots in game just now, https://drive.google.com/drive/folders/1BgtLoM6Na1SIASag5Ask42OobC5QGfTd?usp=share_link
#asahi-dev · 2026-10-11 01:29
yukawill wait with nvme until t604x is in, or at least has more acks
#asahi-dev · 2026-10-11 01:29
yukaanyways, tomorrow I can send the t8140 pinctrl + i2c nodes at least
#asahi-dev · 2026-10-11 01:27
yukaI feel like I'm filling in a sudoku
#asahi-dev · 2026-10-11 01:25
yukapad without crypt ;)
#asahi-dev · 2026-10-11 01:25
JamesCalligeros[m]oh sick a self hosted cryptpad-like thingy
#asahi-dev · 2026-10-11 01:24
yukamade my own, less fancy, overview of what device tree patches I still need to write or send out... https://pad.yuka.dev/Iu1II_aFRRy8G1p4hr2qnA
#asahi-dev · 2026-10-11 01:22
JamesCalligeros[m]lmao
#asahi-dev · 2026-10-11 01:22
yuka(and read the first letters: s-i-e-m-e-n-s)
#asahi-dev · 2026-10-11 01:22
yuka(translation: one thing is for sure; you only get garbage)
#asahi-dev · 2026-10-11 01:21
JamesCalligeros[m]she did the whole "make a switftui translink app make no mistakes" thing and it kinda worked but performance was terrible and the backend was... well, a lot of the compute that should happen on device was happening on the server and performance was not great
#asahi-dev · 2026-10-11 01:21
yukasicher ist eins; man erhält nur schrott
#asahi-dev · 2026-10-11 01:20
myriaI wrote a FastAPI frontend for a compiler, compiling student code on demand (for a compilers course) worked really well with me calling out to random subprocesses and handling many requests at once.
#asahi-dev · 2026-10-11 01:20
JamesCalligeros[m]translink ditched their rather good in-house app for some white labelled piece of shit thing from siemens with less than half the features and a stupid amount of bugs. everyone was suuuper pissed off at the time so a friend and i started making a better translink set of apps out of spite

plus ›