#
✕ nick:noopwafel
#asahi-dev · 2026-11-01 23:14
noopwafelwill take a week off sometime soon(tm) and try to at least PR the m1n1 pieces
#asahi-dev · 2026-11-01 23:12
noopwafelnot a huge amount of progress since the XDC talk (altho the kernel driver does work now)
#asahi-dev · 2026-11-01 23:09
noopwafelvery nice work :)
#asahi-dev · 2026-11-01 23:05
noopwafelcursed witchery
#asahi-dev · 2026-11-01 12:03
noopwafelshould work now! lmk if you need other traces
#asahi-dev · 2026-11-01 11:53
noopwafelat least for t8122 there are a bunch of people who can test :) looking at the cert horror ..
#asahi-dev · 2026-11-01 11:47
noopwafelI think already we're deviating quite a lot from macos trace with that kind of thing
#asahi-dev · 2026-10-26 16:30
noopwafelsven: :o
#asahi-dev · 2026-10-20 12:23
noopwafelsobek[m]: come and spend ccc hacking on asahi stuff with us \o/
#asahi-dev · 2026-10-06 09:26
noopwafeloh I already said the latter :P but I cannot praise enough
#asahi-dev · 2026-10-06 09:26
noopwafelLarstiQ: yes it's really not that huge a diff in the end on kernel side :) current diff (and I did end up fixing it btw, \o/) is <5k lines and most of that is basically churn like added ::ver. the refactor of much of initdata to m1n1 helps a lot.
#asahi-dev · 2026-10-05 20:41
noopwafel(although tbh I'm assuming it's easier for me to continue just using a script around tuxvdmtool+initramfs, given it'll only be a few machines)
#asahi-dev · 2026-10-05 20:38
noopwafelis the plan still the draft MR using the CS board?
#asahi-dev · 2026-10-05 20:37
noopwafel(for drm/mesa in particular)
#asahi-dev · 2026-10-05 20:37
noopwafelI was wondering about setting something up for the M1 at home, because otherwise it's easy to break things
#asahi-dev · 2026-10-02 13:09
noopwafelLarstiQ: the driver is still broken \o/ arrived home a few hours ago, very jetlagged today, pretty sure I know what's broken.. it really isn't _that_ huge a diff on the kernel side (esp. with initdata-in-m1n1). and the renderer is silly :D but the mesa/m1n1 drmshim is quite nice for bringup.
#asahi-dev · 2026-09-22 07:55
noopwafelintegralpilot: re zircon: nice :)
#asahi-dev · 2026-09-18 07:44
noopwafel\o/
#asahi-dev · 2026-09-06 13:23
noopwafel(yes, that's from m3 with 14.x)
#asahi-dev · 2026-08-28 09:53
noopwafelreal cga vibes :)
#asahi-dev · 2026-05-31 14:21
noopwafelif you don't apply tunables/etc then I think it come up working.. no idea about the region though
#asahi-dev · 2026-05-31 14:20
noopwafeljust ignore atcphy entirely (and make sure XHCI is enabled in your kernel, which was my mistake).
#asahi-dev · 2026-05-31 14:19
noopwafelI think you don't even need it to do that.
#asahi-dev · 2026-04-25 12:47
noopwafel(although diffstat of my local m1n1 tree is almost 10k lines now, so YMMV)
#asahi-dev · 2026-04-25 12:43
noopwafelfor 14.x, my local tree NOPs out one branch (to avoid xnu bugs) in all cases, and then for M3, NOPs out the 5 known register writes; it is just as reliable on M3 as it is on M1 for me, with that.
#asahi-dev · 2026-04-18 08:06
noopwafelthat is at least good :)
#asahi-dev · 2026-04-07 15:04
noopwafel(and that address does seem to be in ane[0], so my first thought would also be to make sure pmgr turned it on..)
#asahi-dev · 2026-04-07 14:53
noopwafelyou can ignore the "wrong PA; not mapped", sorry, that's just noting that it isn't mapped on the host side (to be expected)
#asahi-dev · 2026-04-03 18:01
noopwafeljannau: in the binary I see 80000 only btw, so 🤷. and trivial change, but just in case it helps, https://noopwafel.net/asahi/0001-soc-apple-pmp-Force-inlining-of-pmp_read32.patch
#asahi-dev · 2026-04-03 17:50
noopwafeland the resulting kernel works on my j613 (plasma works, wifi works).
#asahi-dev · 2026-04-03 17:47
noopwafel(I assume because then the offsets are known, and so the alignment check can be statically resolved)
#asahi-dev · 2026-04-03 17:46
noopwafelinlining pmp_read32 itself fixes it
#asahi-dev · 2026-04-03 17:36
noopwafelit's (seemingly?) a io_addr_assert in pmp_read32 (the unaligned check?). io_addr_assert does have the inline(always).
#asahi-dev · 2026-04-03 17:18
noopwafelbut it's entirely possible that it's some build config issue, will work it out..
#asahi-dev · 2026-04-03 17:16
noopwafel( https://lkml.org/lkml/2025/11/28/103 to be clear)
#asahi-dev · 2026-04-03 17:09
noopwafelthere's a linux kernel list post which says it needs to be inlined, and that commit is in the tree
#asahi-dev · 2026-04-03 17:04
noopwafel(and sounds like it's nothing asahi specific, so I will poke at it and see if I can solve)
#asahi-dev · 2026-04-03 17:03
noopwafelthat does not help. I am building a monolithic kernel without modules, though.
#asahi-dev · 2026-04-03 17:02
noopwafelthanks, let me update
#asahi-dev · 2026-04-03 17:00
noopwafel/home/fuzzie/dev/linux/rust/kernel/build_assert.rs:78:(.text+0x27a0): undefined reference to `rust_build_error'
#asahi-dev · 2026-04-03 17:00
noopwafelld: drivers/soc/apple/pmp.o: in function `<kernel::io::Io<1048576>>::io_addr_assert::<u32>':
#asahi-dev · 2026-04-03 17:00
noopwafel1.92.0. bad compiler? it's hitting the rust_build_error thing.
#asahi-dev · 2026-04-03 16:58
noopwafelthanks for all the work, jannau. pmp fails to build for me...
#asahi-dev · 2026-04-03 12:49
noopwafeloh, that's a good point. so the atcphy changes could be upstreamed independently.
#asahi-dev · 2026-04-03 12:39
noopwafelpreliminary SPMI patches on top of mildsunrise's tree are at https://noopwafel.net/asahi/ (the usb-typec-tipd series) which work for tipd, but you also need https://noopwafel.net/asahi/unfinished_t8122_atcphy_bringup.txt
#asahi-dev · 2026-04-03 12:35
noopwafellarger changes
#asahi-dev · 2026-04-03 12:30
noopwafelI do, but note that atcphy is not apple,t8103-atcphy compatible
#asahi-dev · 2026-04-03 12:20
noopwafeland if it's indeed 0x59e00c000 then the other phy reg values are predictable at least. sorry for the chaos, thank you for checking!
#asahi-dev · 2026-04-03 12:18
noopwafel0x59e00c000 works fine, so I would use that..
#asahi-dev · 2026-04-03 12:13
noopwafelmy local dts has 0x59e00c000 for phy0 commented out with a note saying 'nope' :) so building a tree to try on hw

plus ›