uACPI - a portable and easy-to-integrate ACPI implementation | Operating System Development | Page 70

fiery ice Jun 9, 2026, 12:15 PM

#

arduino uses 5V io

#

rpi 3.3V

#

just choose the correct one

sharp lichen Jun 9, 2026, 12:15 PM

#

arent motherboard com headers on 12V?

fiery turtle Jun 9, 2026, 12:15 PM

#

db9-female-to-idc-10-pin-serial-ribbon-cable57247101161.png

fiery ice Jun 9, 2026, 12:15 PM

#

huuh, i don't think they are 12V

fiery turtle Jun 9, 2026, 12:15 PM

#

its called idc10 to db9

gentle peak Jun 9, 2026, 12:16 PM

#

fiery turtle

is there a variant of this that ends in a male usb type a

#

I don't have any machines with rs232 ports

fiery ice Jun 9, 2026, 12:16 PM

#

there are usb to rs232 if you want

fiery turtle Jun 9, 2026, 12:16 PM

#

gentle peak is there a variant of this that ends in a male usb type a

yep there is

#

just search idc10 to usb

#

or similar

#

alternatively u can have 2 separate cables, idc10 to db9, and db9 to usb serial

calm latch Jun 9, 2026, 2:28 PM

#

fiery turtle

I have 3 PCs with this...

#

And a modem cable

fiery turtle Jun 9, 2026, 2:29 PM

#

would be nice if u could test it out

calm latch Jun 9, 2026, 2:29 PM

#

(Including my B550 Ryzen system)

fiery turtle Jun 9, 2026, 2:29 PM

#

and check logs if it crashes as well on your hw

full nimbus Jun 9, 2026, 2:30 PM

#

fiery turtle

my desktop has ine of these i installed

#

i used it to test my gdb stub on real hardware

fiery turtle Jun 9, 2026, 2:30 PM

#

nice

calm latch Jun 9, 2026, 2:30 PM

#

calm latch I have 3 PCs with this...

Or actually 2.5 trl

#

I've bought 3 adapters, but lost one of them

fiery turtle Jun 9, 2026, 2:33 PM

#

https://github.com/acpica/acpica/pull/1198

GitHub

iASL: Fix Printf/Fprintf broken by AslKeywordMapping bounds check b...

Commit eff423c ("Add AslKeywordMappingCount for validation in OpcGenerateAmlOpcode") introduced a bounds check to prevent global- buffer-overflow when accessing AslKeywordMapping[...

#

more people noticing vibecoding regressions

#

its been almost a month and a half

#

but the maintainer merges typo fixes instead

#

anyone wants to make an ASL compiler for the uacpi org? trl

#

tbh a clanker could probably oneshot that but

#

maybe someone wants to do it by hand

pallid lava Jun 9, 2026, 2:37 PM

#

apart from the interpreter, it's an assembler, disassembler, and anything else?

fiery turtle Jun 9, 2026, 2:37 PM

#

an interactive debugger

#

but that can be tackled on top of the uacpi test runner

pallid lava Jun 9, 2026, 2:38 PM

#

yes

#

for the assembler, would it be possible to make something quick and good using flex and bison?

fiery turtle Jun 9, 2026, 2:38 PM

#

acpica uses bison i think so yes

#

or flex

#

idk

pallid lava Jun 9, 2026, 2:39 PM

#

i don't think the grammar is too complicated

fiery turtle Jun 9, 2026, 2:39 PM

#

yeah its pretty simple honestly

#

they designed it so that all the complexity goes into the kernel and not the compiler

#

which is very smart ofc

pallid lava Jun 9, 2026, 2:40 PM

#

99.9999% of the complexity is in interpreting the garbage gnuslashtroll

fiery turtle Jun 9, 2026, 2:40 PM

#

yeah

#

they could make a proper bytecode with virtual registers, simple non recursive logical ops etc

#

but instead its ASL but translated 1 for 1 into AML

pallid lava Jun 9, 2026, 2:41 PM

#

i was going to ask whether the assembler is also at fault for being too lax

fiery turtle Jun 9, 2026, 2:41 PM

#

yeah partially

pallid lava Jun 9, 2026, 2:41 PM

#

maybe if the compiler wasn't so garbage and firmware devs were a bit smarter aml wouldn't be crazy

fiery turtle Jun 9, 2026, 2:41 PM

#

the NT ASL compiler does 0 checking

#

u can put any garbage consisting of valid keywords

pallid lava Jun 9, 2026, 2:41 PM

#

bruh

fiery turtle Jun 9, 2026, 2:41 PM

#

thats partially why the AML is such a mess yeah

#

very poor underdeveloped tooling

#

these days acpica produces more warnings and stuff

#

but sometimes that does more harm as well

#

because they work around the warnings by doing even stupider stuff

fiery ice Jun 9, 2026, 2:53 PM

#

I mean, i could try to write one when i'm at home

#

Or well, let clanker loose

#

I have 5.5 for free trl

fiery turtle Jun 9, 2026, 2:54 PM

#

the acpi dump repo has a lot of decompiled ASL you could throw at it

hallow marten Jun 9, 2026, 3:01 PM

#

@fiery turtle so what do i need to work on for the uacpi linux stuff

#

i missed a few things here and too lazy so tldr trl

gentle peak Jun 9, 2026, 3:03 PM

#

well there's the suspend stuff

#

I've ordered the serial stuff so I should be able to get logs on that soon

hallow marten Jun 9, 2026, 3:03 PM

#

great

#

@gentle peak it suspended but didnt wake right?

gentle peak Jun 9, 2026, 3:15 PM

#

no it didn't suspend in the first place

hallow marten Jun 9, 2026, 3:15 PM

#

huh

#

well I fixed a wake path

#

the clanker stubbed it lol

gentle peak Jun 9, 2026, 3:15 PM

#

power led stayed on and fans kept spinning

#

is there a repo for this btw? or is it just sharing patch files

hallow marten Jun 9, 2026, 3:16 PM

#

for now it's just sharing patches but I'll update the uacpi repo eventually

#

@fiery ice you had issues with wake right?

#

can you try this

#

📎 uacpi-linux.iso

fiery ice Jun 9, 2026, 3:18 PM

#

hallow marten <@573903151472836609> you had issues with wake right?

Yes

fiery ice Jun 9, 2026, 3:18 PM

#

hallow marten can you try this

When i get home

fiery turtle Jun 9, 2026, 3:18 PM

#

hallow marten the clanker stubbed it lol

bro

fiery turtle Jun 9, 2026, 3:19 PM

#

hallow marten for now it's just sharing patches but I'll update the uacpi repo eventually

you know monkuous never asks such things for no reason ultrameme

#

but no srsly push it somewhere

hallow marten Jun 9, 2026, 3:19 PM

#

it stubbed acpi_any_fixed_event_status_set acpi_any_gpe_status_set

hallow marten Jun 9, 2026, 3:19 PM

#

fiery turtle you know monkuous never asks such things for no reason <:ultrameme:7290268986669...

I'm too ashamed that it's like 95% clanker

fiery turtle Jun 9, 2026, 3:19 PM

#

its fiiiine

gentle peak Jun 9, 2026, 3:19 PM

#

I'm mostly asking because the patches thing is inconvenient lol

hallow marten Jun 9, 2026, 3:20 PM

#

how should we do this

#

do we delete the repo and make a new linux fork

#

or what

gentle peak Jun 9, 2026, 3:21 PM

#

especially since testing it on the affected machine requires a custom kernel config (for amdgpu) and a custom initrd (for the firmware binaries)

gentle peak Jun 9, 2026, 3:21 PM

#

hallow marten do we delete the repo and make a new linux fork

that's what I'd do

fiery turtle Jun 9, 2026, 3:21 PM

#

hallow marten do we delete the repo and make a new linux fork

yes

hallow marten Jun 9, 2026, 3:21 PM

#

can you do it infy

#

i have org rights but not delete rights (and I shouldnt have them)

fiery turtle Jun 9, 2026, 3:24 PM

#

do u want to import it with the name uacpi-linux?

hallow marten Jun 9, 2026, 3:25 PM

#

yeah i guess

#

wait wdym

#

I make the repo on my profile then move it?

fiery turtle Jun 9, 2026, 3:25 PM

#

exactly yeah

#

alright ive deleted the uacpi-linux repo

#

u can move now

hallow marten Jun 9, 2026, 3:28 PM

#

fiery turtle exactly yeah

do we do proper linux conventions or not

#

for commits

fiery turtle Jun 9, 2026, 3:29 PM

#

that would be nice i suppose

hallow marten Jun 9, 2026, 3:29 PM

#

or if this does end up being patches we would just redo the patches anyway

#

with good conventions

fiery turtle Jun 9, 2026, 3:29 PM

#

i mean the conventions are pretty simple no?

hallow marten Jun 9, 2026, 3:29 PM

#

Signed-Off-By: is annoying meme

fiery turtle Jun 9, 2026, 3:29 PM

#

component: msg

S-o-b

#

wdym why

hallow marten Jun 9, 2026, 3:30 PM

#

no reason im just lazy trl

fiery turtle Jun 9, 2026, 3:30 PM

#

git commit -s?

#

have u been typing it manually or smth lol

hallow marten Jun 9, 2026, 3:30 PM

#

no

#

but then i have to dox myself

#

nooo

fiery turtle Jun 9, 2026, 3:30 PM

#

make up a fake name if u want

hallow marten Jun 9, 2026, 3:31 PM

#

nah it's fine ive already doxxed myself on the lkml anyway

fiery turtle Jun 9, 2026, 3:31 PM

#

i mean especially if we consider the remote possibility of ever upstreaming it

pallid lava Jun 9, 2026, 3:31 PM

#

linux doesn't allow anonymous contributors anyways

fiery turtle Jun 9, 2026, 3:31 PM

#

thatd be useful

hallow marten Jun 9, 2026, 3:31 PM

#

pallid lava linux doesn't allow anonymous contributors anyways

yeah I know I'm a contributor

#

I have made very important contributions

#

https://tenor.com/view/mundo-feliz-onetreehsll-world-if-gif-14071235670792471304

Tenor

#

society using linux after the patch I sent ^

steel inlet Jun 9, 2026, 3:32 PM

#

Add files via upload

full nimbus Jun 9, 2026, 3:32 PM

#

hallow marten https://tenor.com/view/mundo-feliz-onetreehsll-world-if-gif-14071235670792471304

Linux after uACPI becomes the default AML engine

#

Also I spent like 2 hours trying to make a minimal and decent kernel config

#

Just for some bullshit to nuke it

#

💀

#

I'm doing ts tomorrow

hallow marten Jun 9, 2026, 3:33 PM

#

@fiery turtle I'll just name the repo "linux", it's cooler

fiery turtle Jun 9, 2026, 3:34 PM

#

sure, whatever u want lol

hallow marten Jun 9, 2026, 3:34 PM

#

https://github.com/uACPI/linux

#

you need to clone uACPI in the source tree as well

hollow scaffold Jun 9, 2026, 3:35 PM

#

first

hallow marten Jun 9, 2026, 3:35 PM

#

https://github.com/uACPI/linux/tree/master/drivers/acpi/uacpi clanker readme that could be useful

#

~~ugh the commit message is shit~~ forced push lol

full nimbus Jun 9, 2026, 3:39 PM

#

hallow marten you need to clone uACPI in the source tree as well

Sub module it?

hallow marten Jun 9, 2026, 3:39 PM

#

not sure that's a thing linux does

fiery turtle Jun 9, 2026, 3:40 PM

#

yeah no

hallow marten Jun 9, 2026, 3:40 PM

#

in the future the source code will just be inside linux

#

if it gets merged

fiery ice Jun 9, 2026, 4:21 PM

#

@fiery turtle could you create a repo for the compiler / disassembler pls?

fiery turtle Jun 9, 2026, 4:28 PM

#

Why lol

#

make something working first

fiery ice Jun 9, 2026, 4:28 PM

#

k

#

clanker is doing stuff

#

@hallow marten using this cmd it still cannot wakeup

fiery turtle Jun 9, 2026, 4:51 PM

#

does it like not react to anything at all

#

or hangs

fiery ice Jun 9, 2026, 4:53 PM

#

fiery turtle or hangs

screen goes black and power led stays blue

#

then if i press anything, nothing happens

#

not even the power button

fiery turtle Jun 9, 2026, 4:54 PM

#

Yeah that sounds like _PTS and stuff is not getting called for some reason

fiery turtle Jun 9, 2026, 4:54 PM

#

fiery ice screen goes black and power led stays blue

What color is it supposed to be usually on suspend

fiery ice Jun 9, 2026, 4:56 PM

#

huuh idk

#

tho i don't think blue

gentle peak Jun 9, 2026, 4:56 PM

#

fiery ice screen goes black and power led stays blue

That's what's happening on my PC as well

fiery turtle Jun 9, 2026, 4:58 PM

#

Really could use those logs

#

like it could literally be panicking in the kernel and we wouldnt know

fiery ice Jun 9, 2026, 5:00 PM

#

i don't have serial there

fiery turtle Jun 9, 2026, 5:01 PM

#

Ye I know

#

@ mishakov nooo

fiery ice Jun 9, 2026, 5:10 PM

#

@fiery turtle how did i dump the acpi to test?

#

think i did it

fiery turtle Jun 9, 2026, 5:18 PM

#

acpidump

fiery ice Jun 9, 2026, 5:18 PM

#

yeah, but it was decompiling it to multiple table files, lol

#

i just kept dsdt and ssdt

#

that had actual aml code

fiery turtle Jun 9, 2026, 5:19 PM

#

Yea

#

U can use the acpi dumps repo in the uacpi org

#

It has a lot of asl

fiery ice Jun 9, 2026, 5:19 PM

#

yeah i saw

#

i'm trying with the one from my pc first

fiery turtle Jun 9, 2026, 5:20 PM

#

Sure

fiery ice Jun 9, 2026, 5:28 PM

#

i hate that AI literally "fixes" the missing operation and replaces them with a No-Op

hallow marten Jun 9, 2026, 5:56 PM

#

Lol I should try the new mythos release

#

on the uACPI thing

#

Claude fix everything make no mistake

#

(I won't, I'm not AI bro enough)

fiery turtle Jun 9, 2026, 6:01 PM

#

oh yeah

fiery ice Jun 9, 2026, 6:06 PM

#

oh, dsdt compiled

#

and my 5 hour limit is finished

#

pain

fiery turtle Jun 9, 2026, 6:38 PM

#

lol

#

disassemble with acpica and see if it works

sharp lichen Jun 9, 2026, 6:47 PM

#

does Intel AMT Serial-over-lan thingy also work for getting logs? If so, then I might be able to test on my laptop

fiery turtle Jun 9, 2026, 6:48 PM

#

depends if yours captures normal output via COM1 or a PCI bar

#

if its a pci bar it might be too late but not sure

fiery ice Jun 9, 2026, 6:49 PM

#

fiery turtle disassemble with acpica and see if it works

yeah, lemme try, lol

#

the disassembler is borked on the app clanker made

#

i need to force it to fix it

#

💀

#

bruh, it was using medium reasoning

fiery turtle Jun 9, 2026, 6:51 PM

#

gg

fiery ice Jun 9, 2026, 7:02 PM

#

fiery turtle gg

it did not fully decompile, but it's still a good result

calm latch Jun 9, 2026, 7:16 PM

#

fiery turtle @ mishakov <:nooo:750466870993223750>

I need to find the second serial lead

sharp lichen Jun 9, 2026, 7:19 PM

#

fiery turtle depends if yours captures normal output via COM1 or a PCI bar

Is there any way to know? Or do I just chuck the ISO at it and see what happens?

fiery turtle Jun 9, 2026, 7:27 PM

#

sharp lichen Is there any way to know? Or do I just chuck the ISO at it and see what happens?

yeah, the latter

#

with the appropriate command line to enable said output

fiery ice Jun 9, 2026, 9:43 PM

#

@hallow marten @fiery turtle i have some logs

#

📎 message.txt

#

full logs from serial

#

it seems to behave the same way

fiery turtle Jun 9, 2026, 9:45 PM

#

so it does wake up but the screen doesnt light up or?

fiery ice Jun 9, 2026, 9:45 PM

#

no, this is from qemu

#

it never reaches full sleep seems like

fiery turtle Jun 9, 2026, 9:45 PM

#

from your log it does, and it even wakes up

#

qemu has a seabios bug that prevents ECAM from working on resume

#

try without q35

fiery ice Jun 9, 2026, 9:46 PM

#

ah

#

what machine then

fiery turtle Jun 9, 2026, 9:46 PM

#

just omit machine

fiery ice Jun 9, 2026, 9:46 PM

#

i did

#

qemu-system-x86_64 -cdrom /home/diamantino/Downloads/uacpi.iso -serial stdi

#

i'm running it like this

fiery turtle Jun 9, 2026, 9:46 PM

#

ah ok

#

then thats the bug

#

But your qemu does wake up [ 33.454310] PM: suspend exit

fiery ice Jun 9, 2026, 9:47 PM

#

i never touch anything

#

it wakes up from itself

#

and also, the terminal gets stuck

#

i can't do anything

fiery turtle Jun 9, 2026, 9:47 PM

#

ah ok then that should be easy to debug by clanker

#

@hallow marten can u tell clanker to run suspend in qemu and see why it doesnt suspend?

fiery ice Jun 9, 2026, 9:48 PM

#

anyway, 5 hour window resetted, time to get clanker back at working, lol

fiery turtle Jun 9, 2026, 9:48 PM

#

lol

steel inlet Jun 9, 2026, 9:48 PM

#

uAICP

fiery ice Jun 9, 2026, 10:13 PM

#

the compiler is almost done i think

calm latch Jun 9, 2026, 10:41 PM

#

steel inlet uAICP

ACPIAI trl

fiery ice Jun 9, 2026, 10:44 PM

#

💀

#

5hr budget already run out

#

but seems like it won't stop the current job

#

it's now trying to fix the decompiler

fiery turtle Jun 9, 2026, 10:45 PM

#

what language are you even doing it in

fiery ice Jun 9, 2026, 10:45 PM

#

c

fiery turtle Jun 9, 2026, 10:46 PM

#

ah

fiery ice Jun 9, 2026, 10:46 PM

#

why?

fiery turtle Jun 9, 2026, 10:46 PM

#

idk

fiery ice Jun 9, 2026, 10:48 PM

#

i think it's done?

#

ok nope

#

oh finally

#

after 40 mins

#

(or well 70 in total)

#

@fiery turtle

fiery turtle Jun 9, 2026, 10:56 PM

#

lol nice

#

u could try shoving it into the uacpi test runner instead of acpica

#

and see if its able to compile stuff

fiery ice Jun 9, 2026, 10:57 PM

#

hmm

#

lemme check

fiery turtle Jun 9, 2026, 11:00 PM

#

theres the --asl-compiler argument

#

but it will obv try to pass acpica options to it

fiery ice Jun 9, 2026, 11:01 PM

#

those will not work, yeah

fiery ice Jun 9, 2026, 11:34 PM

#

bruh

#

@fiery turtle

#

https://github.com/Diamantino-Op/uACPIAslDC

#

if you wanna check

fiery turtle Jun 9, 2026, 11:59 PM

#

fiery ice <@353593948523986945>

The answers it gets from acpiexec

#

Maybe the generated aml is not correct so acpiexec doesn't generate whats expected as the output

fiery ice Jun 9, 2026, 11:59 PM

#

f

#

any way to add more debug and check if that's the case?

fiery turtle Jun 10, 2026, 12:01 AM

#

Compile the test file with iasl and your compiler and then run with acpiexec and compare the outputs ig

fiery ice Jun 10, 2026, 12:02 AM

#

what file is that if you know

#

maybe 2080-buffer-reads?

#

it's the only one without a .aml

#

hmm, 2080 runs fine i think?

#

the writes does not tho

fiery turtle Jun 10, 2026, 12:13 AM

#

Yeah the 2080 is the autogenerated thing

fiery turtle Jun 10, 2026, 12:16 AM

#

fiery ice hmm, 2080 runs fine i think?

Thats not the answers file

#

It failed to parse the output of the answers run

fiery turtle Jun 10, 2026, 12:16 AM

#

fiery ice the writes does not tho

This is tho

fiery ice Jun 10, 2026, 12:16 AM

#

yeah

fiery turtle Jun 10, 2026, 12:16 AM

#

Compiler bug?

fiery ice Jun 10, 2026, 12:16 AM

#

idk, maybe?

#

i'll ask clanker tmr

fiery turtle Jun 10, 2026, 12:17 AM

#

Yeah im pretty sure it is tbh

#

It creates buffer ops incorrectly I think

hallow marten Jun 10, 2026, 12:46 AM

#

fiery ice why?

Should've done rust

fiery ice Jun 10, 2026, 12:47 AM

#

hallow marten Should've done rust

then i would've need to port the uacpi .h files every time it updated

hallow marten Jun 10, 2026, 12:47 AM

#

Maybe but rust is agentic language bro

fiery ice Jun 10, 2026, 12:47 AM

#

XD

hallow marten Jun 10, 2026, 1:43 AM

#

fiery turtle <@379432087620550658> can u tell clanker to run suspend in qemu and see why it d...

it does work in qemu

#

idk what hes on about

gentle peak Jun 10, 2026, 4:20 AM

#

hallow marten it does work in qemu

idk i can reproduce that as well

#

qemu-system-x86_64 -cdrom uacpi-linux-iso -serial stdio, add no_console_suspend to kernel args, echo mem > /sys/power/state

#

interestingly it has different behavior with kvm enabled: it does enter suspend but on wakeup it gets stuck at the same point that it gets to with no interaction on tcg

#

or at least sometimes it does? sometimes it behaves the same as tcg

gentle peak Jun 10, 2026, 4:47 AM

#

nevermind i'm stupid

#

this is just because linux defconfig doesn't have a proper vga driver

#

the "automatic wakeup" is just the interrupt for the enter key being released

#

and it's only "stuck" because userspace uses tty0 instead of ttyS0 and it doesn't know how to resume vga

gentle peak Jun 10, 2026, 5:21 AM

#

fiery turtle qemu has a seabios bug that prevents ECAM from working on resume

works perfectly fine for me on q35

fiery turtle Jun 10, 2026, 8:02 AM

#

gentle peak and it's only "stuck" because userspace uses tty0 instead of ttyS0 and it doesn'...

Ah ok that makes perfect sense

#

That explains why his log contained 0 errors

fiery turtle Jun 10, 2026, 8:02 AM

#

gentle peak works perfectly fine for me on q35

Thats probably linux magic

#

Like ecam shouldn't be usable at that point unless linux fixes the chipset register itself

fiery turtle Jun 10, 2026, 8:25 AM

#

That would also explain how that regression has gone unnoticed

fiery ice Jun 10, 2026, 12:08 PM

#

@fiery turtle is this correct?

📎 message.txt

fiery turtle Jun 10, 2026, 12:09 PM

#

do u think i know all 2080 values it outputs by heart? KEKW

#

cant u just compare to the iasl output?

fiery ice Jun 10, 2026, 12:09 PM

#

that's buffer-writes

#

not 2080

fiery turtle Jun 10, 2026, 12:09 PM

#

thats the 2080

#

just the preparation part

fiery ice Jun 10, 2026, 12:09 PM

#

ah

#

well, at least it doesn't fail XD

fiery turtle Jun 10, 2026, 12:10 PM

#

the test idea is:
generate asl that writes every possbile combination of buffer fields
compile it
run it in acpiexec, get what it outputs
run it in uacpi, get what it outputs
make sure they agree

pallid lava Jun 10, 2026, 12:10 PM

#

fiery turtle do u think i know all 2080 values it outputs by heart? <:KEKW:913708372329644042...

yes

fiery turtle Jun 10, 2026, 12:13 PM

#

fiery ice well, at least it doesn't fail XD

yeah it seems likely that thats correct

fiery ice Jun 10, 2026, 12:13 PM

#

gg

#

this was why btw, lol

fiery turtle Jun 10, 2026, 12:14 PM

#

makes sense

fiery ice Jun 10, 2026, 12:32 PM

#

@fiery turtle good progress

#

lol

fiery turtle Jun 10, 2026, 12:33 PM

#

lol

fiery ice Jun 10, 2026, 12:33 PM

#

well, it is going so hopefully in a bit we'll have a good version

hallow marten Jun 10, 2026, 12:35 PM

#

/goal make all tests run

fiery ice Jun 10, 2026, 12:35 PM

#

hallow marten /goal make all tests run

i think that's what it's going for, lol

#

67/69 tests completed

fiery ice Jun 10, 2026, 1:11 PM

#

@fiery turtle

#

it's done

#

all tests complete successfully

fiery turtle Jun 10, 2026, 1:11 PM

#

Crazy

#

How bad is the compiler code KEKW

fiery ice Jun 10, 2026, 1:12 PM

#

huuh

#

lemme go see

#

not that bad tbh

#

it could def use some refactoring, but it is not very bad

#

and the file with the most LOCs is just the parser at 3.5k LOC

#

all the others are 1.5k or less

fiery turtle Jun 10, 2026, 1:23 PM

#

what about the disassembler?

fiery ice Jun 10, 2026, 1:24 PM

#

lemme test

#

it's broken again, ooooof

#

i think it only updated the compiler for the tests

hallow marten Jun 10, 2026, 3:11 PM

#

yk it's AI code when it has =======

fiery turtle Jun 10, 2026, 3:12 PM

#

In this case its actually handrolled uacpi tests

calm latch Jun 10, 2026, 3:15 PM

#

I think ======= is just very popular in general?

kind mantle Jun 10, 2026, 3:15 PM

#

hallow marten yk it's AI code when it has =======

Nah AI does that but with em-dashes

hallow marten Jun 10, 2026, 3:36 PM

#

kind mantle Nah AI does that but with em-dashes

Not in code

kind mantle Jun 10, 2026, 3:41 PM

#

Yes it does???

#

I've used LLM monkeys you know

fiery ice Jun 10, 2026, 4:00 PM

#

@fiery turtle i was stupid, i tried to decompile using compile 💀

#

it decompiled correctly i think

hallow marten Jun 10, 2026, 4:06 PM

#

kind mantle Yes it does???

in comments yeah

#

but like

#

i never saw it print("emdashemdashemdash")

#

it's always print("==========")

slim panther Jun 10, 2026, 4:38 PM

#

yeah that's true

lofty dragon Jun 10, 2026, 6:00 PM

#

my 2017 project that does ======== was secretly written by an LLM

calm latch Jun 10, 2026, 6:05 PM

#

Are you secretly a LLM?

fiery turtle Jun 10, 2026, 6:07 PM

#

fiery ice <@353593948523986945> i was stupid, i tried to decompile using compile 💀

Gg

fiery ice Jun 10, 2026, 6:08 PM

#

and i fixed a small bug where strings didn't get copied correctly

fierce kiln Jun 11, 2026, 2:37 AM

#

it's pretty surprising that people are able to find an uptime window long enough to perform github CI cyberattacks

frank canopy Jun 14, 2026, 12:21 AM

#

@fiery turtle out of curiosity, do you actually have numbers for perf benefits for keeping aml microps stored inline? or is that just a code thing where in C its easier to make them like that?

fiery turtle Jun 14, 2026, 12:23 AM

#

frank canopy <@353593948523986945> out of curiosity, do you actually have numbers for perf be...

Oh I have no idea, but it would definitely be way less convenient to have to define them separately, then make a declaration, then reference it etc

#

But in general id assume there's some benefit since the entire op array can be hot

frank canopy Jun 14, 2026, 12:24 AM

#

fiery turtle Oh I have no idea, but it would definitely be way less convenient to have to def...

ok cool. for my zaml project im sticking with out of line only then since i can just take address-of an array literal to get a slice and the compiler will put the array in rodata for me

#

it is probably better for cache true

fiery turtle Jun 14, 2026, 12:25 AM

#

Sure yeah

#

Gonna make a thread btw?

frank canopy Jun 14, 2026, 12:28 AM

#

probably when i have more than two ish lines of code

#

its all in the mind only at this point

fiery turtle Jun 14, 2026, 12:29 AM

#

Fair

frank canopy Jun 14, 2026, 1:00 AM

#

@fiery turtle how does UACPI_UNRESOLVED_NAME_STRING_OP work? im having trouble understanding what the microcode is even doing for those

#

is it just hijacking it into the nameseg parsing or something?

#

ig the resolution thing would skip past the whole nameseg looking at this more

frank canopy Jun 15, 2026, 2:07 PM

#

fiery turtle Gonna make a thread btw?

would an acpi library go under non-os or is it close enough for my thread to go in this category? not sure if uacpi is effectively grandfathered in or if this is the right category for mine too lol

fiery turtle Jun 15, 2026, 2:09 PM

#

imo it belongs in an os category

frank canopy Jun 15, 2026, 2:29 PM

#

👍 I'll make a thread this afternoon sometime then

fiery turtle Jun 15, 2026, 3:39 PM

#

frank canopy <@353593948523986945> how does UACPI_UNRESOLVED_NAME_STRING_OP work? im having t...

Did you figure it out BTW

#

I was asleep

frank canopy Jun 15, 2026, 4:07 PM

#

fiery turtle Did you figure it out BTW

i think so? #1217009725711847465 message is my best guess at how it works rn

gentle peak Jun 15, 2026, 4:13 PM

#

update on the whole uacpi-linux real hardware suspend serial output thing: first I mistakenly bought both idc10-to-male-db9 and male-db9-to-usb adapters, then I bought a female-db9-to-usb adapter but when it arrived today I noticed that my idc10-to-db9 used the wrong idc10 layout so I bought the correct one today

#

hopefully tomorrow I will be able to get output

fiery turtle Jun 15, 2026, 4:14 PM

#

frank canopy i think so? https://discord.com/channels/440442961147199490/1217009725711847465/...

case '0'..'9'
case 'A'..'Z'
case 'a'..'z': // all valid ascii chars, must be a namestring
    aml_pc--;
    namestring = parse_namestring(aml + aml_pc);

    // Is it a method call, a reference..?
    do_what_namestring_wants(namestring);

@frank canopy

fiery turtle Jun 15, 2026, 4:14 PM

#

gentle peak hopefully tomorrow I will be able to get output

lol nice

frank canopy Jun 15, 2026, 4:15 PM

#

fiery turtle ``` case '0'..'9' case 'A'..'Z' case 'a'..'z': // all valid ascii chars, must be...

so it is going oh this opcode is actually an identifier character, back up and parse a namestring

fiery turtle Jun 15, 2026, 4:15 PM

#

yep

fiery turtle Jun 15, 2026, 4:15 PM

#

gentle peak update on the whole uacpi-linux real hardware suspend serial output thing: first...

hopefully those werent expensive?

gentle peak Jun 15, 2026, 4:17 PM

#

nah

frank canopy Jun 15, 2026, 4:17 PM

#

fiery turtle ``` case '0'..'9' case 'A'..'Z' case 'a'..'z': // all valid ascii chars, must be...

your actual code doesnt use the same op declaration for 0..9 or lowercase btw, only A-Z and \^_/.

gentle peak Jun 15, 2026, 4:17 PM

#

even if they were I'm getting a full refund

fiery turtle Jun 15, 2026, 4:19 PM

#

frank canopy your actual code doesnt use the same op declaration for 0..9 or lowercase btw, o...

yeah because 0..9 is not a valid starting char

#

so ig drop that case lol

fiery turtle Jun 15, 2026, 4:20 PM

#

gentle peak even if they were I'm getting a full refund

ah good

frank canopy Jun 15, 2026, 4:20 PM

#

fiery turtle yeah because 0..9 is not a valid starting char

ok so i was right then, just bad at describing what i thought it was lol

fiery turtle Jun 15, 2026, 4:20 PM

#

ye

gentle peak Jun 15, 2026, 4:25 PM

#

gentle peak update on the whole uacpi-linux real hardware suspend serial output thing: first...

also, warning for anyone else that wants to get a serial port header adapter: there are (apparently) 2 different electrical layouts in use for the same physical port

#

so make sure you get the right one

kindred beacon Jun 15, 2026, 4:46 PM

#

wtf

fiery turtle Jun 15, 2026, 4:48 PM

#

once we have fully working uacpi-linux ill make that the reference uacpi test os lol

#

if someone comes in with one of those uacpi bugs again meme

full nimbus Jun 15, 2026, 5:18 PM

#

gentle peak hopefully tomorrow I will be able to get output

my desktop has a working port btw

#

i could test this today

#

also wait there are multible difrent IDC10 things??

gentle peak Jun 15, 2026, 5:18 PM

#

does it have the same issue?

gentle peak Jun 15, 2026, 5:18 PM

#

full nimbus also wait there are multible difrent IDC10 things??

yeah it's awful

full nimbus Jun 15, 2026, 5:19 PM

#

oh fuck right trl

#

i did never get to test it working fully

gentle peak Jun 15, 2026, 5:19 PM

#

I've tested it on two of my three computers and it's worked on one

full nimbus Jun 15, 2026, 5:19 PM

#

gentle peak yeah it's awful

I must have gotten lucky as fuckkkk

gentle peak Jun 15, 2026, 5:20 PM

#

I was troubleshooting for like two hours before I decided to check the pinout of the adapter (I would've done so earlier but the adapter I bought had like 0 documentation so it wasn't immediately obvious where to find the pinout)

#

in fact I never did find any adapter specific pinout documentation

#

I found out about the multiple layouts thing first and then noticed that the packaging said "DTK" (as in "DTK/Intel layout")

gentle peak Jun 16, 2026, 3:57 PM

#

gentle peak hopefully tomorrow I will be able to get output

serial works!

#

now to get the actual suspend output

#

📎 message.txt

fiery turtle Jun 16, 2026, 3:58 PM

#

gentle peak serial works!

Wow

#

Whats that printout at the top?

gentle peak Jun 16, 2026, 3:58 PM

#

thats just the stuff from the initramfs

gentle peak Jun 16, 2026, 4:00 PM

#

gentle peak

@full nimbus @hallow marten

fiery turtle Jun 16, 2026, 4:00 PM

#

[   75.281499] BUG: kernel NULL pointer dereference, address: 0000000000000008
[   75.288456] #PF: supervisor read access in kernel mode
[   75.293585] #PF: error_code(0x0000) - not-present page
[   75.298720] PGD 10905b067 P4D 10905b067 PUD 10905c067 PMD 0
[   75.304372] Oops: Oops: 0000 [#1] SMP NOPTI
[   75.308556] CPU: 12 UID: 0 PID: 1 Comm: sh Not tainted 7.1.0-rc7-geb39c6a26396 #4 PREEMPT(lazy)
[   75.317322] Hardware name: Gigabyte Technology Co., Ltd. B550M S2H/B550M S2H, BIOS F19d 09/02/2024
[   75.326265] RIP: 0010:handle_control_flow+0x32/0xe0
[   75.331136] Code: 48 8b 8f 38 1d 00 00 4c 8b 87 58 1d 00 00 48 8b 91 08 07 00 00 48 85 d2 74 77 48 8b 87 40 1d 00 00 48 39 c2 75 33 49 8b 70 08 <8b> 50 08 66 81 7e 1a a5 00 0f 45 50 04 89 91 1c 07 00 00 48 8b 87
[   75.349859] RSP: 0018:ffffa5c980013bf0 EFLAGS: 00010246
[   75.355077] RAX: 0000000000000000 RBX: ffff90ff898a6000 RCX: ffff90ff8713c000
[   75.362202] RDX: 0000000000000000 RSI: ffffffff9c088c40 RDI: ffff90ff898a6000
[   75.369326] RBP: 0000000000000015 R08: ffff90ff8713c080 R09: 0000000000000001
[   75.376450] R10: 0000000000000005 R11: 0000000000000000 R12: ffff90ff8713c080
[   75.383572] R13: 0000000000000015 R14: 0000000000000000 R15: ffff90ff8713c000
[   75.390697] FS:  0000000000713198(0000) GS:ffff910ee17a2000(0000) knlGS:0000000000000000
[   75.398772] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   75.404510] CR2: 0000000000000008 CR3: 0000000109054000 CR4: 0000000000750ef0
[   75.411634] PKRU: 55555554
[   75.414341] Call Trace:
[   75.416791]  <TASK>
[   75.418892]  uacpi_execute_control_method+0xea4/0x2660

#

whats the addr2line for this rip?

gentle peak Jun 16, 2026, 4:01 PM

#

one sec i messed around with a few things between making the usb and today so i don't have the vmlinux for that build anymore

fiery turtle Jun 16, 2026, 4:01 PM

#

u did disable kaslr right?

#

but the fact that it page faults is good tbh, if it was silently suceeding but still hanging that would be way scarier

gentle peak Jun 16, 2026, 4:02 PM

#

not via the command line and the build i'm using there is just defconfig + amdgpu builtin

fiery turtle Jun 16, 2026, 4:02 PM

#

ah yeah then addr2line wont work

gentle peak Jun 16, 2026, 4:02 PM

#

i'll make a more debugging friendly build and retry

fiery turtle Jun 16, 2026, 4:03 PM

#

whats interesting is it dies in uacpi control flow opcodes handler

#

handle_control_flow

#

which (should) be a pretty well tested code path

gentle peak Jun 16, 2026, 4:05 PM

#

also the uacpi entrypoint is uacpi_prepare_for_sleep_state so it can't just be a simple "oh the shim maps arguments incorrectly"

fiery turtle Jun 16, 2026, 4:05 PM

#

yeah

#

have u sent the dump for this hardware before btw

gentle peak Jun 16, 2026, 4:05 PM

#

i think so yes

fiery turtle Jun 16, 2026, 4:08 PM

#

but tbh yours is not the only one that dies on suspend so I doubt its related to a particular aml blob

gentle peak Jun 16, 2026, 4:08 PM

#

gentle peak one sec i messed around with a few things between making the usb and today so i ...

also, some things i discovered while doing said messing around:

it doesn't compile on 32 bit platforms due to uacpi doing 64-bit division (which requires some special macro on linux instead of just using the operator)
it doesn't compile unless PCI is also compiled in (acpica does)

fiery turtle Jun 16, 2026, 4:09 PM

#

the former is a good point, i should probably add a way to replace that with a kernel specific function, e.g. do_div on linux

#

the latter i guess is also a config option

gentle peak Jun 16, 2026, 4:10 PM

#

yeah it's probably a shim thing that the acpica osl.c takes into account

fiery turtle Jun 16, 2026, 4:10 PM

#

ill add those to my todo list, thanks

fiery turtle Jun 16, 2026, 4:11 PM

#

gentle peak i'll make a more debugging friendly build and retry

that would be nice, maybe nokaslr, ubsan, and asan

gentle peak Jun 16, 2026, 4:27 PM

#

ok this is with the debugging build

📎 message.txt

#

for some reason addr2line is saying it's line number 0 wtf

fiery turtle Jun 16, 2026, 4:28 PM

#

lol it GPFed and page faulted at the same time?

fiery turtle Jun 16, 2026, 4:28 PM

#

gentle peak for some reason addr2line is saying it's line number 0 wtf

how did u get the actual rip since its hidden there?

gentle peak Jun 16, 2026, 4:29 PM

#

i just passed the sym+x to llvm-addr2line directly

fiery turtle Jun 16, 2026, 4:29 PM

#

ooh

gentle peak Jun 16, 2026, 4:29 PM

#

doesn't seem to have worked perfectly though

#

$ llvm-addr2line -e vmlinux 0xffffffff823dce20
/home/monkuous/dev/linux/drivers/acpi/uacpi/lib/../../../../uACPI/source/interpreter.c:4175``` this should be accurate

fiery turtle Jun 16, 2026, 4:29 PM

#

very strange

gentle peak Jun 16, 2026, 4:30 PM

#

(i just did objdump -t vmlinux | grep handle_control_flow and added the offset manually)

fiery turtle Jun 16, 2026, 4:30 PM

#

gentle peak ``` $ llvm-addr2line -e vmlinux 0xffffffff823dce20 /home/monkuous/dev/linux/driv...

Hm but thats a different function handle_code_block

gentle peak Jun 16, 2026, 4:30 PM

#

this should be on uacpi master as of ~30 minutes ago

gentle peak Jun 16, 2026, 4:31 PM

#

gentle peak (i just did `objdump -t vmlinux | grep handle_control_flow` and added the offset...

nevermind i used the wrong value as offset

#

using the correct offset does reproduce the interpreter.c:0 thing (???)

fiery turtle Jun 16, 2026, 4:32 PM

#

wtf

#

did it per chance jump to an invalid offset or smth

#

which misinterpreted the instruction

gentle peak Jun 16, 2026, 4:33 PM

#

;         if (op_ctx->op->code == UACPI_AML_OP_BreakOp)
ffffffff823dcbcc: 49 83 c4 08           addq    $0x8, %r12
ffffffff823dcbd0: 4c 89 e0              movq    %r12, %rax
ffffffff823dcbd3: 48 c1 e8 03           shrq    $0x3, %rax
ffffffff823dcbd7: 42 80 3c 30 00        cmpb    $0x0, (%rax,%r14)
ffffffff823dcbdc: 74 08                 je      0xffffffff823dcbe6 <handle_control_flow+0x146>
ffffffff823dcbde: 4c 89 e7              movq    %r12, %rdi
ffffffff823dcbe1: e8 4a bf 55 ff        callq   0xffffffff81938b30 <__asan_report_load8_noabort>
ffffffff823dcbe6: 4d 8b 24 24           movq    (%r12), %r12
ffffffff823dcbea: 49 83 c4 1a           addq    $0x1a, %r12
ffffffff823dcbee: 4c 89 e0              movq    %r12, %rax
ffffffff823dcbf1: 48 c1 e8 03           shrq    $0x3, %rax
ffffffff823dcbf5: 42 0f b6 04 30        movzbl  (%rax,%r14), %eax
ffffffff823dcbfa: 84 c0                 testb   %al, %al
ffffffff823dcbfc: 0f 85 2d 01 00 00     jne     0xffffffff823dcd2f <handle_control_flow+0x28f>
ffffffff823dcc02: 31 c0                 xorl    %eax, %eax
ffffffff823dcc04: 66 41 81 3c 24 a5 00  cmpw    $0xa5, (%r12)
ffffffff823dcc0b: 0f 94 c0              sete    %al
ffffffff823dcc0e: 4c 8d 24 85 04 00 00 00       leaq    0x4(,%rax,4), %r12
ffffffff823dcc16: 4d 01 ec              addq    %r13, %r12
ffffffff823dcc19: 4c 89 e0              movq    %r12, %rax
ffffffff823dcc1c: 48 c1 e8 03           shrq    $0x3, %rax
ffffffff823dcc20: 42 0f b6 04 30        movzbl  (%rax,%r14), %eax``` this is what `llvm-objdump -dS` has to say about it (last instruction is the faulting one)

fiery turtle Jun 16, 2026, 4:34 PM

#

frame->code_offset = ctx->cur_block->begin; probably this line, because thats the only field at offset 8 there

#

Or rather, this line

        if (op_ctx->op->code == UACPI_AML_OP_BreakOp)
            frame->code_offset = ctx->cur_block->end;

#

Lemme try to look for a dump

#

https://github.com/uACPI/ACPIDumps/pull/3

GitHub

Add Gigabyte B550M S2H and Toshiba Portege R30-A dumps by monkuous ...

#

this i think

gentle peak Jun 16, 2026, 4:38 PM

#

yeah

#

it's the gigabyte one

#

i don't think i've updated the bios since then so that should be the exact one

#

i could try kgdb?

#

afaik that's like an in kernel gdbserver thing

fiery turtle Jun 16, 2026, 4:40 PM

#

Wow it executes just ridiculous amounts of AML when u _PTS with != 5

fiery turtle Jun 16, 2026, 4:40 PM

#

gentle peak i could try kgdb?

yeah that would be nice

#

if u could dump the state of the execution_context when it dies

#

ill try to look at the aml in the meantime

hallow marten Jun 16, 2026, 4:48 PM

#

gentle peak for some reason addr2line is saying it's line number 0 wtf

That happens when it's an instruction but not a line

fiery turtle Jun 16, 2026, 4:53 PM

#

gentle peak i could try kgdb?

Also, it would really help if u added a printf that printed the method name that it tries to call in prepare_method_call, maybe just using uacpi_namespace_node_generate_absolute_path on the node object

gentle peak Jun 16, 2026, 4:54 PM

#

lol kgdb isn't very well integrated with the rest of the kernel it seems

#

i was just setting up my breakpoints and when i continued it triggered the watchdog

fiery turtle Jun 16, 2026, 4:55 PM

#

I have a suspicion its this break https://pastebin.com/DXRb6uqN but i cant repro anything with my synthetic tests so maybe not

Pastebin

Method (MM12, 2, NotSerialized) { L...

Pastebin.com is the number one paste tool since 2002. Pastebin is a website where you can store text online for a set period of time.

fiery turtle Jun 16, 2026, 4:55 PM

#

gentle peak i was just setting up my breakpoints and when i continued it triggered the watch...

bruh

pallid lava Jun 16, 2026, 4:57 PM

#

this problem wouldn't have happened if uacpi were written in rust trollmogus

fiery turtle Jun 16, 2026, 4:58 PM

#

there isnt even a problem yet KEKW

#

or rather, we dont know what its about

gentle peak Jun 16, 2026, 5:00 PM

#

gentle peak i was just setting up my breakpoints and when i continued it triggered the watch...

decided to just disable watchdog, turns out there's more watchdog: amdgpu 0000:07:00.0: [drm] vblank wait timed out on crtc 0

fiery turtle Jun 16, 2026, 5:01 PM

#

lol

gentle peak Jun 16, 2026, 5:01 PM

#

at least this time it's not a panic

pallid lava Jun 16, 2026, 5:01 PM

#

that happens to me

#

rarely though

gentle peak Jun 16, 2026, 5:02 PM

#

ok it seems like kgdb is just bitrot central

full nimbus Jun 16, 2026, 5:02 PM

#

gentle peak lol kgdb isn't very well integrated with the rest of the kernel it seems

I mean I somewhat get why it wouldn't be fully integrated into everything with watchdogs etc

gentle peak Jun 16, 2026, 5:02 PM

#

it triggered the breakpoint but i didn't get a gdb prompt

#

printf debugging it is

pallid lava Jun 16, 2026, 5:02 PM

#

btw, my laptop's nvidia gpu has an _EJ0

fiery turtle Jun 16, 2026, 5:02 PM

#

💀

pallid lava Jun 16, 2026, 5:03 PM

#

i am afraid to call it KEKW

gentle peak Jun 16, 2026, 5:03 PM

#

what is _EJ0

full nimbus Jun 16, 2026, 5:03 PM

#

What's that again?

pallid lava Jun 16, 2026, 5:03 PM

#

to eject it

fiery turtle Jun 16, 2026, 5:03 PM

#

@gentle peak wont be needed

STDERR FOR desktop_gigabyte_technology_b550m_b550m_s2h_a1360a8647f9:
    /home/d-tatianin/Projects/uACPI/source/interpreter.c:4127:48: runtime error: member access within null pointer of type 'struct code_block'
    AddressSanitizer:DEADLYSIGNAL
    =================================================================
    ==310039==ERROR: AddressSanitizer: SEGV on unknown address 0x000000000008 (pc 0x58ac07f76ba5 bp 0x7ffdb5997690 sp 0x7ffdb5997660 T0)
    ==310039==The signal is caused by a READ memory access.
    ==310039==Hint: address points to the zero page.
        #0 0x58ac07f76ba5 in handle_control_flow /home/d-tatianin/Projects/uACPI/source/interpreter.c:4127
        #1 0x58ac07f84b3c in exec_op /home/d-tatianin/Projects/uACPI/source/interpreter.c:5727
        #2 0x58ac07f88aa7 in uacpi_execute_control_method /home/d-tatianin/Projects/uACPI/source/interpreter.c:6136
        #3 0x58ac07f372f1 in uacpi_eval /home/d-tatianin/Projects/uACPI/source/uacpi.c:779
        #4 0x58ac07fae120 in eval_sleep_helper /home/d-tatianin/Projects/uACPI/source/sleep.c:290
        #5 0x58ac07fae267 in eval_pts /home/d-tatianin/Projects/uACPI/source/sleep.c:309
        #6 0x58ac07fae64d in uacpi_prepare_for_sleep_state /home/d-tatianin/Projects/uACPI/source/sleep.c:386
        #7 0x58ac07f08ddb in enumerate_namespace /home/d-tatianin/Projects/uACPI/tests/runner/test_runner.c:212
        #8 0x58ac07f09849 in run_test /home/d-tatianin/Projects/uACPI/tests/runner/test_runner.c:436
        #9 0x58ac07f0a006 in main /home/d-tatianin/Projects/uACPI/tests/runner/test_runner.c:569
        #10 0x752a55c2a600 in __libc_start_call_main ../sysdeps/nptl/libc_start_call_main.h:59
        #11 0x752a55c2a717 in __libc_start_main_impl ../csu/libc-start.c:360
        #12 0x58ac07f049e4 in _start (/home/d-tatianin/Projects/uACPI/tests/runner/build-linux-64bits/test-runner+0x12a9e4) (BuildId: 86cac9384beae0bde0a4b9ac9f3f72e7ac3b01a8)

full nimbus Jun 16, 2026, 5:03 PM

#

😭

gentle peak Jun 16, 2026, 5:03 PM

#

oh nice

full nimbus Jun 16, 2026, 5:03 PM

#

@pallid lava try it

#

For science

pallid lava Jun 16, 2026, 5:03 PM

#

nah

full nimbus Jun 16, 2026, 5:03 PM

#

Or at least decompile it maybe?

pallid lava Jun 16, 2026, 5:04 PM

#

i think it's just calling some off method and that's it

#

powers it down

fiery turtle Jun 16, 2026, 5:04 PM

#

I added uacpi_prepare_for_sleep_state to the test runner

#

SUMMARY: 681/689 in 105.85s (8 FAILED)

gentle peak Jun 16, 2026, 5:04 PM

#

and one of those is my dump?

fiery turtle Jun 16, 2026, 5:04 PM

#

8 segfaults, all on gigabyte B5XX

gentle peak Jun 16, 2026, 5:04 PM

#

gigabyte W

fiery turtle Jun 16, 2026, 5:05 PM

#

gentle peak and one of those is my dump?

indeed, the first segfault is your dump

fiery turtle Jun 16, 2026, 5:05 PM

#

fiery turtle <@938223886439436320> wont be needed ``` STDERR FOR desktop_gigabyte_technology_...

this one is yours I believe

gentle peak Jun 16, 2026, 5:05 PM

#

yeah i think so judging by the name

fiery turtle Jun 16, 2026, 5:06 PM

#

@fiery ice what was your pc where our uacpi linux died on suspend?

fiery ice Jun 16, 2026, 5:06 PM

#

Samsung galaxy book

fiery turtle Jun 16, 2026, 5:07 PM

#

💀

#

did u send a dump of it?

fiery ice Jun 16, 2026, 5:07 PM

#

yes

#

in this chat a while ago

#

for the pci stuff

#

if i remember correctly

#

and the suspend one

fiery turtle Jun 16, 2026, 5:08 PM

#

https://pastebin.com/uvuawc94

Pastebin

[451/689] desktop_gigabyte_technology_b550m_b550m_s2h_a1360 FAILED ...

Pastebin.com is the number one paste tool since 2002. Pastebin is a website where you can store text online for a set period of time.

fiery turtle Jun 16, 2026, 5:08 PM

#

fiery ice and the suspend one

ill check it out a bit later

gentle peak Jun 16, 2026, 5:11 PM

#

weird that it triggered ASAN in the test runner but not in the live kernel

#

nevermind i'm illiterate

fiery turtle Jun 16, 2026, 5:11 PM

#

lol

gentle peak Jun 16, 2026, 5:17 PM

#

fiery turtle Also, it would really help if u added a printf that printed the method name that...

in case it helps (even if just to check whether you're reproducing the same issue): [ 37.987287] ACPI: PM: Preparing to enter system sleep state S3 [ 37.993268] ACPI: prepare_method_call: \_PTS [ 37.999361] ACPI: prepare_method_call -> OK [ 38.005383] ACPI: prepare_method_call: \_SB_.MPTS [ 38.011909] ACPI: prepare_method_call -> OK [ 38.017929] ACPI: prepare_method_call: \GSA1.MPTS [ 38.024454] ACPI: prepare_method_call -> OK [ 38.030459] ACPI: prepare_method_call: \GSA1.EZV8 [ 38.036995] ACPI: prepare_method_call -> OK [ 38.043028] ACPI: prepare_method_call: \GSA1.EZVH [ 38.049560] ACPI: prepare_method_call -> OK [ 38.055614] ACPI: prepare_method_call: \GSA1.MM01 [ 38.062138] ACPI: prepare_method_call -> OK [ 38.069271] ACPI: prepare_method_call: \GSA1.ISMF [ 38.075827] ACPI: prepare_method_call -> OK [ 38.081898] ACPI: prepare_method_call: \GSA1.HWC0 [ 38.088430] ACPI: prepare_method_call -> OK [ 38.199038] ACPI: prepare_method_call: \GSA1.SMM2 [ 38.205594] ACPI: prepare_method_call -> OK [ 38.211636] ACPI: prepare_method_call: \GSA1.SMB2 [ 38.218171] ACPI: prepare_method_call -> OK [ 38.224179] ACPI: prepare_method_call: \GSA1.SRDY [ 38.230711] ACPI: prepare_method_call -> OK [ 38.236877] ACPI: prepare_method_call: \GSA1.CMPL [ 38.243406] ACPI: prepare_method_call -> OK [ 38.249663] ACPI: prepare_method_call: \GSA1.SMM5 [ 38.256201] ACPI: prepare_method_call -> OK [ 38.262235] ACPI: prepare_method_call: \GSA1.SMB5 [ 38.268768] ACPI: prepare_method_call -> OK [ 38.274773] ACPI: prepare_method_call: \GSA1.SRDY [ 38.281291] ACPI: prepare_method_call -> OK [ 38.287467] ACPI: prepare_method_call: \GSA1.CMPL [ 38.293992] ACPI: prepare_method_call -> OK [ 38.300340] ACPI: prepare_method_call: \GSA1.CMN1 [ 38.306879] ACPI: prepare_method_call -> OK [ 38.313070] Oops: general protection fault, probably for non-canonical address 0xdffffc0000000001: 0000 [#1] SMP KASAN NOPTI

fiery turtle Jun 16, 2026, 5:18 PM

#

thanks

fiery turtle Jun 16, 2026, 5:19 PM

#

gentle peak in case it helps (even if just to check whether you're reproducing the same issu...

whats funny is this doesnt trace returns so we technically dont know which one caused it to die exactly

gentle peak Jun 16, 2026, 5:20 PM

#

where do i add a printf to trace returns

fiery turtle Jun 16, 2026, 5:21 PM

#

hmm good Q

#

technically its ctx_reload_post_ret but there isnt a trivial way to extract the top most code block, which is the bounds of the method, which contains the method node

gentle peak Jun 16, 2026, 5:22 PM

#

i mean i don't need to get the method node, if i just make sure to trace every return i can look at the prepare_method_call log to see what it returned from

fiery turtle Jun 16, 2026, 5:22 PM

#

ah true

gentle peak Jun 16, 2026, 5:28 PM

#

manually indented for readability

📎 trace.txt

fiery turtle Jun 16, 2026, 5:28 PM

#

huh ofc msvc doesnt catch it for some reason

uACPI][ERROR] loop time out after running for 3 seconds
[uACPI][ERROR] aborting method invocation due to previous error: hanging AML while loop
[uACPI][ERROR]     #0 in \GSA1.SRDY()
[uACPI][ERROR]     #1 in \GSA1.SMB2()
[uACPI][ERROR]     #2 in \GSA1.SMM2()
[uACPI][ERROR]     #3 in \GSA1.ISMF()
[uACPI][ERROR]     #4 in \GSA1.EZVH()
[uACPI][ERROR]     #5 in \GSA1.EZV8()
[uACPI][ERROR]     #6 in \GSA1.MPTS()
[uACPI][ERROR]     #7 in \_SB_.MPTS()
[uACPI][ERROR]     #8 in \_PTS()
[uACPI][ERROR] error while evaluating _PTS: hanging AML while loop

gentle peak Jun 16, 2026, 5:29 PM

#

so it's after returning from \GSA1.ISMF

#

aka it's in \GSA1.EZVH

fiery turtle Jun 16, 2026, 5:30 PM

#

https://pastebin.com/pVePrxNA

Pastebin

Method (EZVH, 1, Serialized){ Local0 = MM01 (Arg0) If ((Objec...

Pastebin.com is the number one paste tool since 2002. Pastebin is a website where you can store text online for a set period of time.

gentle peak Jun 16, 2026, 5:30 PM

#

that doesn't even have break

#

i might be incorrect then

fiery turtle Jun 16, 2026, 5:30 PM

#

yeah strange

#

maybe u misread the output

#

huh --log-level=debug gets rid of the crash??

calm latch Jun 16, 2026, 5:35 PM

#

Do other operating systems with uACPI and S3 sleep work? trl

gentle peak Jun 16, 2026, 5:36 PM

#

fiery turtle maybe u misread the output

yeah no i'm pretty sure that's correct

#

something really weird is going on here

fiery turtle Jun 16, 2026, 5:36 PM

#

gentle peak something *really* weird is going on here

the fact that the log level change fixes it makes me think its some sort of a corruption

gentle peak Jun 16, 2026, 5:36 PM

#

yeah

fiery turtle Jun 16, 2026, 5:37 PM

#

like its not even a compile time log level change

#

lemme try valgrind maybe

#

Yeah i managed to catch it with break interpreter.c:4118 if strcmp(why, "\GSA1.EZVH") == 0

#

fiery turtle Jun 16, 2026, 5:50 PM

#

gentle peak that doesn't even have break

ah fuck

#

it has Switch/Case

#

those are macros which are compiled down to while and breaks

gentle peak Jun 16, 2026, 5:52 PM

#

ahh

#

does the aml decompiler have a "literal" option?

#

seems like it would be useful for stuff like this

fiery turtle Jun 16, 2026, 5:53 PM

#

it doesnt KEKW

gentle peak Jun 16, 2026, 5:53 PM

#

damn

fiery turtle Jun 16, 2026, 6:11 PM

#

@gentle peak can u try this fix locally

@@ -2103,11 +2107,8 @@ static void frame_reset_post_end_block(struct execution_context *ctx,
     code_block_array_pop(&frame->code_blocks);
     ctx->cur_block = code_block_array_last(&frame->code_blocks);

-    if (type == CODE_BLOCK_WHILE) {
-        frame->last_while = find_last_block(&frame->code_blocks, type);
-    } else if (type == CODE_BLOCK_SCOPE) {
-        update_scope(frame);
-    }
+    frame->last_while = find_last_block(&frame->code_blocks, CODE_BLOCK_WHILE);
+    update_scope(frame);
 }

#

i dont understand entirely why thats needed yet but it seems to fix it

gentle peak Jun 16, 2026, 6:15 PM

#

successfully went to sleep and woke again!

fiery turtle Jun 16, 2026, 6:15 PM

#

crazy stuff

#

@fiery ice can you try on your hardware with this fix applied?

fiery ice Jun 16, 2026, 6:17 PM

#

fiery turtle <@573903151472836609> can you try on your hardware with this fix applied?

send the iso

#

@hallow marten

gentle peak Jun 16, 2026, 6:18 PM

#

gentle peak successfully went to sleep and woke again!

fiery turtle Jun 16, 2026, 6:19 PM

#

very cool

#

its the most complex aml ive seen

#

it calls like 20 methods deep

#

at peak

gentle peak Jun 16, 2026, 6:21 PM

#

fiery ice send the iso

this is what i use, it's not an iso but you can just extract it onto a fat32 partition and it'll work

📎 uacpi-linux.tar

#

note that performance is not representative, this build has kasan and ubsan enabled

fiery ice Jun 16, 2026, 6:22 PM

#

hmm, i may xorriso it then

gentle peak Jun 16, 2026, 6:22 PM

#

why bother unless you actually use a dvd

hallow marten Jun 16, 2026, 6:22 PM

#

so everything works?

#

nice

fiery ice Jun 16, 2026, 6:22 PM

#

i use ventoy

hallow marten Jun 16, 2026, 6:22 PM

#

i am a genius

gentle peak Jun 16, 2026, 6:22 PM

#

hallow marten so everything works?

as far as i can tell yeah

fiery turtle Jun 16, 2026, 6:23 PM

#

yeah it works now

#

theres a genuine uacpi bug with pointer re-hydration after dynamic array realloc

gentle peak Jun 16, 2026, 6:23 PM

#

fiery ice i use ventoy

ah okay, if you do xorriso it you need to follow the limine xorriso instructions

fiery turtle Jun 16, 2026, 6:23 PM

#

it would only manifest in deeply nested call stacks

gentle peak Jun 16, 2026, 6:23 PM

#

gentle peak note that performance is not representative, this build has kasan and ubsan enab...

[ 3.588785] ACPI: successfully loaded 12 AML blobs, 10097 ops in 238ms (avg 42340/s)

fiery turtle Jun 16, 2026, 6:23 PM

#

still faster than nyaux 💀

hallow marten Jun 16, 2026, 6:23 PM

#

i have a script to generate an iso locally but you could ask a clanker if you dont want to figure it out

gentle peak Jun 16, 2026, 6:24 PM

#

gentle peak as far as i can tell yeah

i guess the next test is building it with some distro's .config and booting a proper distro with it

fiery turtle Jun 16, 2026, 6:26 PM

#

thats the real fix i think

@@ -4235,6 +4244,9 @@ static void refresh_ctx_pointers(struct execution_context *ctx)
     ctx->cur_op_ctx = op_context_array_last(&frame->pending_ops);
     ctx->prev_op_ctx = op_context_array_one_before_last(&frame->pending_ops);
     ctx->cur_block = code_block_array_last(&frame->code_blocks);
+
+    frame->last_while = find_last_block(&frame->code_blocks, CODE_BLOCK_WHILE);
+    update_scope(frame);
 }

#

basically pointer hydration was missing

#

ill add a helper for this i think, since theres some duplication going on there

#

and another callsite where hydration was missing but it never manifested

#

i guess normal aml execution is usually not as deeply nested as sleeping/suspend helpers stuff

fiery turtle Jun 16, 2026, 6:29 PM

#

gentle peak i guess the next test is building it with some distro's `.config` and booting a ...

that would be based af

full nimbus Jun 16, 2026, 6:29 PM

#

fiery turtle i guess normal aml execution is usually not as deeply nested as sleeping/suspend...

God this makes me glad uACPI exists

fiery turtle Jun 16, 2026, 6:30 PM

#

lol

#

https://pastebin.com/pVePrxNA this is the method where it was dying

Pastebin

Method (EZVH, 1, Serialized){ Local0 = MM01 (Arg0) If ((Objec...

Pastebin.com is the number one paste tool since 2002. Pastebin is a website where you can store text online for a set period of time.

fiery ice Jun 16, 2026, 6:30 PM

#

@gentle peak what was the cmd to sleep?

gentle peak Jun 16, 2026, 6:30 PM

#

echo deep > /sys/power/mem_sleep && echo mem > /sys/power/state

fiery ice Jun 16, 2026, 6:31 PM

#

thx

fiery turtle Jun 16, 2026, 6:31 PM

#

gentle peak in case it helps (even if just to check whether you're reproducing the same issu...

basically here u can see it calls really deep, which reallocates the call frame array, which then makes the cached context pointers point to incorrect memory

hallow marten Jun 16, 2026, 6:32 PM

#

fiery turtle i guess normal aml execution is usually not as deeply nested as sleeping/suspend...

This is why you should've made it recursive

fiery ice Jun 16, 2026, 6:32 PM

#

ehm, you sure it works?

gentle peak Jun 16, 2026, 6:32 PM

#

on my pc it does

fiery turtle Jun 16, 2026, 6:32 PM

#

hallow marten This is why you should've made it recursive

true trl

gentle peak Jun 16, 2026, 6:32 PM

#

what happens on yours?

fiery ice Jun 16, 2026, 6:32 PM

#

not on mine

hallow marten Jun 16, 2026, 6:33 PM

#

What happens

fiery turtle Jun 16, 2026, 6:33 PM

#

what happens?

#

lol

fiery ice Jun 16, 2026, 6:33 PM

#

still the blue led

#

and no response from the keyboard

hallow marten Jun 16, 2026, 6:33 PM

#

Logs would be nice

fiery ice Jun 16, 2026, 6:33 PM

#

huuh

#

how

gentle peak Jun 16, 2026, 6:33 PM

#

serial

fiery turtle Jun 16, 2026, 6:33 PM

#

it boots but suspend still dies?

fiery ice Jun 16, 2026, 6:33 PM

#

fiery turtle it boots but suspend still dies?

ye

fiery ice Jun 16, 2026, 6:34 PM

#

gentle peak serial

it's a laptop 💀

gentle peak Jun 16, 2026, 6:34 PM

#

if you have a serial port and use the 2nd boot option (Linux uACPI - test shell on serial (for debugging suspend)) you get serial output through suspend

hallow marten Jun 16, 2026, 6:34 PM

#

Can't wait for the uACPI patches to be in Gentoo repos so now people that configure their own kernels can use it

full nimbus Jun 16, 2026, 6:34 PM

#

fiery ice it's a laptop 💀

USB serial potentially may work?????

fiery ice Jun 16, 2026, 6:34 PM

#

i don't have any serial adapter

full nimbus Jun 16, 2026, 6:34 PM

#

If you have a USB C DRD port

fiery turtle Jun 16, 2026, 6:34 PM

#

lemme try your aml blob

fiery turtle Jun 16, 2026, 6:35 PM

#

fiery ice <@353593948523986945> how did i dump the acpi to test?

I dont think u ever attached it

fiery ice Jun 16, 2026, 6:36 PM

#

sec

#

hmm, i tought i did

#

that message was for my main pc

mortal yoke Jun 16, 2026, 6:40 PM

#

ig I can also try this on my pc for fun KEKW though I don't think it has particularly cursed aml (at least in terms of errors)

fiery turtle Jun 16, 2026, 6:41 PM

#

that would be nice, if it dies then at least its a pc with a com

fiery ice Jun 16, 2026, 6:41 PM

#

#osdev-misc-0 message

#

i think it was this one

#

lol

fiery turtle Jun 16, 2026, 6:43 PM

#

hmm ok doesnt segfault here but it doesnt get far in _PTS because my test runner is not real hardware

[uACPI][ERROR] loop time out after running for 3 seconds
[uACPI][ERROR] aborting method invocation due to previous error: hanging AML while loop
[uACPI][ERROR]     #0 in \_SB_.QPTS()
[uACPI][ERROR]     #1 in \_SB_.SPTS()
[uACPI][ERROR]     #2 in \_PTS()
[uACPI][ERROR] error while evaluating _PTS: hanging AML while loop

#

  While (!(PBTN & 0x02))
  {
      Sleep (0x64)
  }

#

genius hardware

fiery ice Jun 16, 2026, 6:46 PM

#

XD

mortal yoke Jun 16, 2026, 6:55 PM

#

mortal yoke ig I can also try this on my pc for fun <:KEKW:913708372329644042> though I don'...

looks like it did work, it suspended and resumed when I pressed a key on the keyboard (gpu didn't resume properly but typing poweroff -f made it shut down afterwards)

gentle peak Jun 16, 2026, 6:56 PM

#

what gpu model do you have? the only drm modules i have compiled in are simpledrm, i915, and amdgpu

#

and the only firmware included in the initramfs is amdgpu

mortal yoke Jun 16, 2026, 6:57 PM

#

rtx 4070 ti s, yeah I guess then that makes sense

fiery turtle Jun 16, 2026, 7:05 PM

#

mortal yoke looks like it did work, it suspended and resumed when I pressed a key on the key...

awesome!

fiery turtle Jun 16, 2026, 7:07 PM

#

fiery turtle hmm ok doesnt segfault here but it doesnt get far in `_PTS` because my test runn...

commented the while loop out and recompiled dsdt, still no segfault, no hangs either

#

@fiery ice whats the model of that laptop?

fiery ice Jun 16, 2026, 7:14 PM

#

fiery turtle <@573903151472836609> whats the model of that laptop?

galaxy book /750XDA-KC2IT

#

i5 1135g7

fiery turtle Jun 16, 2026, 7:15 PM

#

it does have a

Device (UAH0)
{
    Name (_DDN, "SerialIoUart0")  // _DDN: DOS Device Name
    Name (_UID, "SerialIoUart0")  // _UID: Unique ID
    Method (_HID, 0, NotSerialized)  // _HID: Hardware ID

hmm maybe its possible to capture those

fiery turtle Jun 16, 2026, 7:17 PM

#

fiery ice galaxy book /750XDA-KC2IT

Yeah and your logs say [ 0.138648] printk: legacy console [ttyS0] enabled

#

i wonder where the hell that outputs to

fiery ice Jun 16, 2026, 7:17 PM

#

idk

#

i don't have any port

fiery turtle Jun 16, 2026, 7:18 PM

#

its probably like

#

raw motherboard pinouts

fiery ice Jun 16, 2026, 7:18 PM

#

it didn't even have headers

#

from what i could see

fiery turtle Jun 16, 2026, 7:18 PM

#

damn, I guess we'll have to wait for more test subjects that can repro this nooo

fiery ice Jun 16, 2026, 7:19 PM

#

fiery turtle damn, I guess we'll have to wait for more test subjects that can repro this <:no...

maybe evalyn?

fiery turtle Jun 16, 2026, 7:19 PM

#

yeah perhaps

full nimbus Jun 16, 2026, 7:23 PM

#

i do have a serial port

#

il test tomorrow afternoon

#

i gotta finish up riscv/la64 stuff

fiery turtle Jun 16, 2026, 7:26 PM

#

Thanks, hopefully maybe we get a repro

full nimbus Jun 16, 2026, 7:28 PM

#

dm me an iso incase i wakeup at 2am again so i can get out of the way then too

fiery turtle Jun 16, 2026, 7:38 PM

#

gentle peak this is what i use, it's not an iso but you can just extract it onto a fat32 par...

Here

fiery turtle Jun 16, 2026, 9:01 PM

#

I realized the hydration issue was just a stupid design decision, that piece of code doesnt really need to hold a pointer to the array, it just needs to know the index

#

which is what it does now

#

ill make a test case for it, also ill make all large tests run _PTS(3) on top

fiery turtle Jun 16, 2026, 9:23 PM

#

https://github.com/uACPI/uACPI/pull/156/changes/c36eedb870e541ef413079e6d162925446bb9af1

GitHub

Control flow fixes by d-tatianin · Pull Request #156 · uACPI/uACPI

#

lets see if CI likes it

#

me when i thought my code was well tested but then I meet firmware devs

full nimbus Jun 16, 2026, 9:26 PM

#

"I have a firm belief that most firmware developers are not actually humans, but are instead caged rodents fed a solid diet of crack cocaine. Because that would explain a lot."
- Linus Torvalds

hallow marten Jun 16, 2026, 9:36 PM

#

fiery turtle me when i thought my code was well tested but then I meet firmware devs

isn't this your fault tho?

fiery turtle Jun 16, 2026, 9:36 PM

#

yeah this one is completely my fault

#

but to actually repro it you'd need like really complex aml

#

ill see if i can make clanker come up with a minimal repro so i can use it for a standalone test

#

so far the linux experiment has been extremely useful

#

exercising code paths that hobby kernels never touch and uncovering edge cases

hallow marten Jun 16, 2026, 9:40 PM

#

@anthropic sponsor us

fiery turtle Jun 16, 2026, 9:40 PM

#

real

hallow marten Jun 16, 2026, 9:40 PM

#

we could get acquired by spacex

#

@grok acquire us

full nimbus Jun 16, 2026, 9:41 PM

#

hallow marten we could get acquired by spacex

or lockheedmartin

fiery turtle Jun 16, 2026, 9:41 PM

#

rok is this true

hallow marten Jun 16, 2026, 9:41 PM

#

full nimbus or lockheedmartin

why?

#

I was saying spacex cuz they just bought cursor

full nimbus Jun 16, 2026, 9:41 PM

#

They also make a hypervisor

#

Could be useful to buy this

#

trl

lethal trench Jun 16, 2026, 9:41 PM

#

fiery turtle exercising code paths that hobby kernels never touch and uncovering edge cases

Not my fault it’s so easy to stub :p

hallow marten Jun 16, 2026, 9:41 PM

#

spacex is cooler

#

cuz we'd be at the same company as grok

full nimbus Jun 16, 2026, 9:42 PM

#

hallow marten spacex is cooler

Mecha uACPI

hallow marten Jun 16, 2026, 9:42 PM

#

which is the greatest AI ever

full nimbus Jun 16, 2026, 9:43 PM

#

What if we integrated grok into uACPI to dynamically handle complex AML?

hallow marten Jun 16, 2026, 9:43 PM

#

we could just remove the interpreter entirely and just ask grok

lethal trench Jun 16, 2026, 9:43 PM

#

grok what on earth were they thinking when they made this?

full nimbus Jun 16, 2026, 9:43 PM

#

lethal trench grok what on earth were they thinking when they made this?

Money and white genocide

#

trl

fiery turtle Jun 16, 2026, 9:44 PM

#

lethal trench grok what on earth were they thinking when they made this?

outsource workers paid for each loc

lethal trench Jun 16, 2026, 9:44 PM

#

😭

fiery turtle Jun 16, 2026, 10:10 PM

#

lol

#

github knows whats up

pallid lava Jun 16, 2026, 10:11 PM

#

lmO

fiery turtle Jun 16, 2026, 10:54 PM

#

@gentle peak so the fix I gave u was more of a random bullshit go type of fix, I now added a proper one to master + tests, would be nice if you could do a quick re-test once you have the time

gentle peak Jun 17, 2026, 1:21 PM

#

made this quick little project

#

wonder whether it'd work on real hardware

fiery turtle Jun 17, 2026, 1:22 PM

#

Wtf thats crazy

pallid lava Jun 17, 2026, 1:22 PM

#

how long did it take

fiery turtle Jun 17, 2026, 1:22 PM

#

Did you make a bochs gpu driver also

gentle peak Jun 17, 2026, 1:22 PM

#

pallid lava how long did it take

like 8 hours total i think

gentle peak Jun 17, 2026, 1:22 PM

#

fiery turtle Did you make a bochs gpu driver also

nah this is just debugcon and uart

fiery turtle Jun 17, 2026, 1:23 PM

#

Does it do the wake gpe and power resource stuff also?

#

Or PCI bar restoration etc

gentle peak Jun 17, 2026, 1:25 PM

#

i currently do not do anything with acpi on wake

#

the only thing i do on wake is restore cpu state, irq controller state, and reinit serial

fiery turtle Jun 17, 2026, 1:26 PM

#

Yeah but when suspending, how do you configure the devices for wake?

#

That are allowed to wake you

gentle peak Jun 17, 2026, 1:26 PM

#

oh i don't do anything with that

#

this is the entire acpi side of entering suspend

#

(arch_prepare_suspend does the waking vector stuff)

fiery turtle Jun 17, 2026, 1:32 PM

#

Btw uacpi handles tts for you I think

#

So technically u dont need that

fiery turtle Jun 17, 2026, 1:33 PM

#

gentle peak oh i don't do anything with that

Ah, most likely on real hw it will cause it to just not wake up on key press

gentle peak Jun 17, 2026, 1:33 PM

#

fiery turtle Btw uacpi handles tts for you I think

i checked, it doesn't

#

it handles SST but not TTS

fiery turtle Jun 17, 2026, 1:33 PM

#

Ah wait yeah, what was TTS again

gentle peak Jun 17, 2026, 1:33 PM

#

no idea

fiery turtle Jun 17, 2026, 1:34 PM

#

Did u get that from linux?

gentle peak Jun 17, 2026, 1:34 PM

#

linux does do it yes but i got it from the spec

fiery turtle Jun 17, 2026, 1:34 PM

#

Interesting

#

https://uefi.org/specs/ACPI/6.5/16_Waking_and_Sleeping.html#transitioning-from-the-working-to-the-sleeping-state

#

Yeah this mentions it

fiery turtle Jun 17, 2026, 1:41 PM

#

gentle peak made this quick little project

Where repo btw

gentle peak Jun 17, 2026, 1:48 PM

#

https://github.com/monkuous/uacpi-test

GitHub

GitHub - monkuous/uacpi-test

Contribute to monkuous/uacpi-test development by creating an account on GitHub.

#

tried it on real hardware, for some reason serial only works if it's a warm boot (???) but more importantly it doesn't enter any sleep states correctly

#

not even S5

fiery turtle Jun 17, 2026, 1:52 PM

#

F

gentle peak Jun 17, 2026, 1:53 PM

#

oh nevermind it just takes a while

gentle peak Jun 17, 2026, 2:18 PM

#

fiery turtle <@938223886439436320> so the fix I gave u was more of a random bullshit go type ...

worked

fiery turtle Jun 17, 2026, 2:19 PM

#

letsgo

#

thanks for testing!

fiery turtle Jun 17, 2026, 2:19 PM

#

gentle peak oh nevermind it just takes a while

bad allocator or smth?

gentle peak Jun 17, 2026, 2:20 PM

#

the allocator should be fine

fiery turtle Jun 17, 2026, 2:20 PM

#

hmm, usually at least S5 should be instant

gentle peak Jun 17, 2026, 2:20 PM

#

it gets stuck in uacpi_enter_sleep_state_simple before disabling interrupts (since i still get power button! prints)

fiery turtle Jun 17, 2026, 2:20 PM

#

thats quite interesting

gentle peak Jun 17, 2026, 2:21 PM

#

might just be a scheduler bug or something, the way i designed this isn't exactly conventional (it has no idle thread)

fiery turtle Jun 17, 2026, 2:21 PM

#

lol, what does it do if theres nothing to do?

gentle peak Jun 17, 2026, 2:21 PM

#

#

this is fully cooperative scheduling, there's no preemption (whether from priority or timeslice stuff)

#

i could see firmware developers relying on the timeslice of the thread executing aml expiring or whatever

#

like maybe they do Stalls to wait for a software event idk

fiery turtle Jun 17, 2026, 2:24 PM

#

could be

gentle peak Jun 17, 2026, 3:31 PM

#

rn_image_picker_lib_temp_b1c125dd-f1cc-4eb1-9a08-7cc2fee0bd8f.jpg

median crest Jun 17, 2026, 3:32 PM

#

Wtf

#

Insane

kindred beacon Jun 17, 2026, 3:34 PM

#

gentle peak

#

fucking legend

gentle peak Jun 17, 2026, 3:35 PM

#

what the fuck, there's an actual noticeable performance improvement in glxgears fps of all things

median crest Jun 17, 2026, 3:35 PM

#

Year of the uacpi linux desktop

full nimbus Jun 17, 2026, 3:35 PM

#

AUR package for linux-uacpi when?

gentle peak Jun 17, 2026, 3:35 PM

#

uacpi hovers around 20k fps, acpica around 17k

full nimbus Jun 17, 2026, 3:36 PM

#

gentle peak uacpi hovers around 20k fps, acpica around 17k

Isn't glxgears more variable with stuff?

#

I've heard before somewhat

gentle peak Jun 17, 2026, 3:36 PM

#

it's always been very consistent for me

full nimbus Jun 17, 2026, 3:36 PM

#

full nimbus AUR package for linux-uacpi when?

And make sure to add the malware to it too

#

trl

calm latch Jun 17, 2026, 3:38 PM

#

Mainline kernel when trl

fiery turtle Jun 17, 2026, 3:39 PM

#

gentle peak uacpi hovers around 20k fps, acpica around 17k

What the fuck

#

I wonder if there's some actual gpe traffic that uacpi handles faster

fiery turtle Jun 17, 2026, 3:39 PM

#

gentle peak

Crazy work as always what can I say

gentle peak Jun 17, 2026, 3:39 PM

#

it might just be the changes made between the mainline kernel arch linux uses and the git master that uacpi-linux is based on

fiery turtle Jun 17, 2026, 3:40 PM

#

Yeah could be ofc

#

Although mainline arch is usually cutting edge kernel

hallow marten Jun 17, 2026, 3:40 PM

#

It's probably more of a kernel thing than acpi lol

fiery turtle Jun 17, 2026, 3:41 PM

#

Yep

flat badge Jun 17, 2026, 3:41 PM

#

or some PM difference where either uACPI or acpica doesn't work correctly

fiery turtle Jun 17, 2026, 3:41 PM

#

Yeah that could be a reason too

flat badge Jun 17, 2026, 3:42 PM

#

modern intel chipsets cannot run both the CPU and all chipset devices at maximal performance settings at the same time

full nimbus Jun 17, 2026, 3:43 PM

#

fiery turtle Although mainline arch is usually cutting edge kernel

its only ever a week or two behind or so iirc

flat badge Jun 17, 2026, 3:46 PM

#

you can dump cat /sys/bus/pci/devices/*/power/runtime_status

#

or similar to check if devices are properly being suspended

#

check dmsg for PM issues

gentle peak Jun 17, 2026, 3:50 PM

#

flat badge modern intel chipsets cannot run both the CPU and all chipset devices at maximal...

i am on a not-so-modern AMD chipset

#

anyway i did a few more random perf measurements ACPICA: Time-to-boot 18.50 s (started when pressing enter in GRUB, stopped when plasma-login-manager became visible) GLXGears FPS 17k-18k vkmark score 17193 Time-to-suspend 2.55 s (started when clicking sleep button in plasma launcher, stopped when fans stopped spinning) Time-to-wake 6.30 s (started when pressing power button, stopped when plasma lock screen visible) uACPI: Time-to-boot 16.63 s GLXGears FPS 19k-20k vkmark score 17758 Time-to-suspend 2.44 s Time-to-wake 5.78 s

#

the time measurements were done by hand so there's a fairly large margin of error there

quaint lotus Jun 17, 2026, 3:58 PM

#

gentle peak

Wait you ported uACPI to Linux?

gentle peak Jun 17, 2026, 3:58 PM

#

i wasn't the one doing the porting

quaint lotus Jun 17, 2026, 3:58 PM

#

who was?

gentle peak Jun 17, 2026, 3:59 PM

#

abbix with claude

quaint lotus Jun 17, 2026, 3:59 PM

#

hmm, did you all review the code?

#

and like fix it

#

if so I’d honestly like seeing this getting into mainstream kernel

gentle peak Jun 17, 2026, 3:59 PM

#

my involvement was limited to getting serial output from a device where suspend didn't work initially and using the kernel with an actual distro instead of just a busybox initramfs

#

surprisingly i don't think any of the issues we've encountered so far have been with the llm generated code

#

they've all been uacpi bugs that haven't been encountered before because of the limited scope of hobby kernels

fiery turtle Jun 17, 2026, 4:01 PM

#

gentle peak anyway i did a few more random perf measurements ``` ACPICA: Time-to-boot ...

Awesome

quaint lotus Jun 17, 2026, 4:03 PM

#

gentle peak anyway i did a few more random perf measurements ``` ACPICA: Time-to-boot ...

really cool

#

I’d totally use that

hallow marten Jun 17, 2026, 4:23 PM

#

quaint lotus hmm, did you all review the code?

Yes

gentle peak Jun 17, 2026, 10:53 PM

#

@fiery turtle if you want some more stuff for acpidumps, this is from an MSI Venture 14 AI A2HMG-021NL laptop

📎 dump.txt

#

i removed MSDM ofc but still

fiery turtle Jun 17, 2026, 10:54 PM

#

Did you test the distro on it? meme

#

Is that your new laptop or smth?

gentle peak Jun 17, 2026, 10:55 PM

#

yeah it's my new laptop

fiery turtle Jun 17, 2026, 10:55 PM

#

ah nice

#

it probably doesnt have S3, only S2 idle

gentle peak Jun 17, 2026, 10:55 PM

#

I did try booting uacpi-linux but defconfig lacks a lot of the modules needed to be usable so I don't know how functional it really is

fiery turtle Jun 17, 2026, 10:55 PM

#

gentle peak I did try booting uacpi-linux but defconfig lacks a lot of the modules needed to...

did it die on boot?

gentle peak Jun 17, 2026, 10:56 PM

#

fiery turtle it probably doesnt have S3, only S2 idle

/sys/power/mem_sleep contains both s2idle and deep

#

although s2idle is selected by default

fiery turtle Jun 17, 2026, 10:56 PM

#

ah

gentle peak Jun 17, 2026, 10:56 PM

#

fiery turtle did it die on boot?

nah it got to the desktop (plasma) but didn't have working audio, graphics accel, or touchpad

fiery turtle Jun 17, 2026, 10:57 PM

#

ah

#

I guess stealing the config from some arch distro is the way

gentle peak Jun 17, 2026, 10:57 PM

#

worth noting that this laptop isn't perfect for linux regardless

fiery turtle Jun 17, 2026, 10:57 PM

#

really?

gentle peak Jun 17, 2026, 10:58 PM

#

even on the regular kernel installed by the distro I get driver errors from i915, aml errors, etc

#

and I had to disable the Intel vmd controller in the bios before I could really do anything

fiery turtle Jun 17, 2026, 10:58 PM

#

Can you show what the AML errors are btw?

fiery turtle Jun 17, 2026, 10:58 PM

#

gentle peak and I had to disable the Intel vmd controller in the bios before I could really ...

ah yeah that proprietary crap

gentle peak Jun 17, 2026, 10:59 PM

#

gentle peak and I had to disable the Intel vmd controller in the bios before I could really ...

before I did this the entire system behaved very strangely

pallid lava Jun 17, 2026, 11:00 PM

#

omg i was just yesterday reading how linux actually assigns iommu groups, and for vmd there is a special case

gentle peak Jun 17, 2026, 11:00 PM

#

nvme io kept timing out randomly and for some reason the system could only even get into bash if I replaced the default loglevel=3 quiet with loglevel=7

fiery turtle Jun 17, 2026, 11:00 PM

#

lmao

gentle peak Jun 17, 2026, 11:00 PM

#

and no matter what it couldn't get into a graphical environment

pallid lava Jun 17, 2026, 11:01 PM

#

is there any chance your distribution doesn't have compiled in support for vmd?

gentle peak Jun 17, 2026, 11:03 PM

#

fiery turtle Can you show what the AML errors are btw?

$ sudo dmesg | grep 'ACPI Error'
[    0.275562] ACPI Error: Missing argument(s) for method [RUCC] (20251212/dsmethod-486)
[    0.275565] ACPI Error: Aborting method \_SB.PC00.XHCI.RHUB.HS02._PLD due to previous error (AE_AML_TOO_FEW_ARGUMENTS) (20251212/psparse-529)
[    0.289605] ACPI Error: Missing argument(s) for method [RUCC] (20251212/dsmethod-486)
[    0.289608] ACPI Error: Aborting method \_SB.PC00.TXHC.RHUB.SS04._PLD due to previous error (AE_AML_TOO_FEW_ARGUMENTS) (20251212/psparse-529)
[    0.541321] ACPI Error: Missing argument(s) for method [RUCC] (20251212/dsmethod-486)
[    0.541325] ACPI Error: Aborting method \_SB.PC00.TXHC.RHUB.SS04._PLD due to previous error (AE_AML_TOO_FEW_ARGUMENTS) (20251212/psparse-529)
[    0.541334] ACPI Error: Missing argument(s) for method [RUCC] (20251212/dsmethod-486)
[    0.541336] ACPI Error: Aborting method \_SB.PC00.TXHC.RHUB.SS04._UPC due to previous error (AE_AML_TOO_FEW_ARGUMENTS) (20251212/psparse-529)
[    0.541346] ACPI Error: Missing argument(s) for method [RUCC] (20251212/dsmethod-486)
[    0.541347] ACPI Error: Aborting method \_SB.PC00.TXHC.RHUB.SS04._PLD due to previous error (AE_AML_TOO_FEW_ARGUMENTS) (20251212/psparse-529)
[    0.543802] ACPI Error: Missing argument(s) for method [RUCC] (20251212/dsmethod-486)
[    0.543803] ACPI Error: Aborting method \_SB.PC00.XHCI.RHUB.HS02._PLD due to previous error (AE_AML_TOO_FEW_ARGUMENTS) (20251212/psparse-529)
[    0.543809] ACPI Error: Missing argument(s) for method [RUCC] (20251212/dsmethod-486)
[    0.543810] ACPI Error: Aborting method \_SB.PC00.XHCI.RHUB.HS02._UPC due to previous error (AE_AML_TOO_FEW_ARGUMENTS) (20251212/psparse-529)
[    0.543818] ACPI Error: Missing argument(s) for method [RUCC] (20251212/dsmethod-486)
[    0.543819] ACPI Error: Aborting method \_SB.PC00.XHCI.RHUB.HS02._PLD due to previous error (AE_AML_TOO_FEW_ARGUMENTS) (20251212/psparse-529)

gentle peak Jun 17, 2026, 11:03 PM

#

pallid lava is there any chance your distribution doesn't have compiled in support for vmd?

wouldn't it just fail to boot at all in that case?

#

from what i can tell vmd is some storage thing so it shouldn't have been able to find the rootfs without supporting it

pallid lava Jun 17, 2026, 11:03 PM

#

no idea, the thing i was looking was gated behind a config_vmd

fiery turtle Jun 17, 2026, 11:03 PM

#

gentle peak ``` $ sudo dmesg | grep 'ACPI Error' [ 0.275562] ACPI Error: Missing argument...

they keep fucking up _PLD for some reason

#

maybe because those errors never show up in NT

#

but it definitely fails in this case also

gentle peak Jun 17, 2026, 11:04 PM

#

or maybe they do and they just didn't care

fiery turtle Jun 17, 2026, 11:04 PM

#

its not super important

#

im not even sure what it needs the physical location of usb ports for

gentle peak Jun 17, 2026, 11:06 PM

#

pallid lava no idea, the thing i was looking was gated behind a config_vmd

$ gunzip -c /proc/config.gz | grep VMD
CONFIG_VMD=m
CONFIG_MVMDIO=m```

pallid lava Jun 17, 2026, 11:06 PM

#

well trl who knows

fiery turtle Jun 17, 2026, 11:08 PM

#

gentle peak <@353593948523986945> if you want some more stuff for acpidumps, this is from an...

btw if you get the time to do it, a PR into the acpi-dumps repo would be nice (with the real or fake hw id)

gentle peak Jun 17, 2026, 11:09 PM

#

gentle peak and I had to disable the Intel vmd controller in the bios before I could really ...

this was very annoying to do btw, the option was grayed out unless you pressed some secret key combo while in the bios

#

why are firmware devs like this

fiery turtle Jun 17, 2026, 11:09 PM

#

wtf

#

why do they even shove this vmd shit into consumer laptops

#

who even uses that

gentle peak Jun 17, 2026, 11:09 PM

#

i mean i'm not sure this technically is a consumer laptop

#

like i'm a consumer yes but still

fiery turtle Jun 17, 2026, 11:09 PM

#

oh its not?

gentle peak Jun 17, 2026, 11:10 PM

#

like i said i'm not sure

fiery turtle Jun 17, 2026, 11:10 PM

#

is it supposed to be a corporate laptop?

gentle peak Jun 17, 2026, 11:11 PM

#

idk i got the vibe that it's supposed to be suitable as a corporate laptop in any case

fiery turtle Jun 17, 2026, 11:11 PM

#

The MSI Venture 14 AI A2HMG-021NL is a premium, AI-ready business and productivity laptop. Featuring an Intel Core Ultra 7 processor, Copilot+ support, and 32GB of RAM, it easily powers through demanding office workflows, multitasking, and media editing on its sharp 14-inch OLED display

#

clanker opinion KEKW

pallid lava Jun 17, 2026, 11:11 PM

#

copilot+ support

fiery turtle Jun 17, 2026, 11:12 PM

#

since its ai-ready i expect you to vibecode all of your stuff from now on

gentle peak Jun 17, 2026, 11:12 PM

#

i don't think the npu has done a single thing since i got the laptop a month ago

#

the first thing i did was do a clean install and i don't really use any ai ever

kindred beacon Jun 17, 2026, 11:13 PM

#

you forgot

#

NPU == no purpose unit

fiery turtle Jun 17, 2026, 11:13 PM

#

what the hell even is copilot+

#

if i knew my hardware supported it, first thing id do is find out how to uninstall it

gentle peak Jun 17, 2026, 11:16 PM

#

decided to compile uacpi-linux with the arch .config, wonder how long it'll take

fiery turtle Jun 17, 2026, 11:17 PM

#

its a fast laptop, probably not much

#

Btw why did you pick this one specifically

gentle peak Jun 17, 2026, 11:21 PM

#

filter by high refresh rate and >=32gb ram, sort by price low to high, look at cpu benchmarks for everything below like 1300eur

fiery turtle Jun 17, 2026, 11:23 PM

#

Does refresh rate really matter to you that much?

gentle peak Jun 17, 2026, 11:29 PM

#

i got used to 180 Hz and now whenever i try to use 60 Hz i cannot help but notice the roughness of basically all animations

#

this one is 144 Hz which is enough to feel smooth to me

full nimbus Jun 17, 2026, 11:31 PM

#

gentle peak i got used to 180 Hz and now whenever i try to use 60 Hz i cannot help but notic...

This tends to reduce over time

#

You get used to it if you use a 60hz monitor for like a day or two

gentle peak Jun 17, 2026, 11:31 PM

#

oh yeah definitely but i didn't want to be switching between the two all the time

fiery turtle Jun 17, 2026, 11:31 PM

#

yeah

#

one of my monitors is 144, and the other is 60

#

over time i stopped noticing the difference at all lol

full nimbus Jun 17, 2026, 11:32 PM

#

I can't notice the difference between my 75hz desktop and 120hz phone

gentle peak Jun 17, 2026, 11:38 PM

#

...huh ```
ERROR: modpost: "acpi_gbl_FADT" [drivers/gpu/drm/amd/amdgpu/amdgpu.ko] undefined!
ERROR: modpost: "acpi_gbl_FADT" [sound/soc/intel/avs/snd-soc-avs.ko] undefined!

fiery turtle Jun 17, 2026, 11:39 PM

#

is that the missing EXPORT_SYMBOL thing again?

gentle peak Jun 17, 2026, 11:40 PM

#

yeah

#

simple enough to fix

fiery turtle Jun 17, 2026, 11:40 PM

#

yeah should be

flat badge Jun 18, 2026, 7:05 AM

#

Afaict the main use cases for NPUs etc are for example doing speed recognition locally such that you don't have to send audio to a remote server

gentle peak Jun 18, 2026, 7:20 AM

#

gentle peak nah it got to the desktop (plasma) but didn't have working audio, graphics accel...

it actually got less far with all the modules compiled in

#

it failed to mount /boot from fstab due to a timeout

#

journalctl shows a bunch of udevd hung worker warnings

rn_image_picker_lib_temp_d5895237-20ca-47e1-a351-084440f84bfa.jpg

#

then after a while there's this

rn_image_picker_lib_temp_d1e2b994-286e-4973-864e-6da037228aba.jpg

#

then a bunch of kernel hung task warnings

rn_image_picker_lib_temp_b1db0953-0618-4eb6-9b47-ab5253204f6e.jpg

#

there's also this earlier in boot which might be related?

rn_image_picker_lib_temp_0e86e4c7-8911-4786-9842-0a03b9533b2f.jpg

#

for what it's worth though the aml errors I had with acpica disappeared

gentle peak Jun 18, 2026, 7:26 AM

#

gentle peak it failed to mount /boot from fstab due to a timeout

I'm actually very confused why this mount is timing out

#

because root is on the same device and that's already mounted successfully

#

and io on root works without any timeouts

#

if I disable timeout and let it wait forever it boots successfully

fiery turtle Jun 18, 2026, 7:32 AM

#

gentle peak then a bunch of kernel hung task warnings

Looks like thats also in the i2c code

fiery turtle Jun 18, 2026, 7:33 AM

#

gentle peak there's also this earlier in boot which might be related?

The i2c handler error here is most likely the cause, although I have no idea why it would fail

#

Or why it would hang somewhere also

gentle peak Jun 18, 2026, 7:34 AM

#

gentle peak if I disable timeout and let it wait forever it boots successfully

that being said I still don't have working ethernet, audio, or touchpad

#

I'm guessing those all involve i2c in some way

fiery turtle Jun 18, 2026, 7:34 AM

#

gentle peak if I disable timeout and let it wait forever it boots successfully

Huh lol

fiery turtle Jun 18, 2026, 7:34 AM

#

gentle peak I'm guessing those all involve i2c in some way

Yeah likely

#

I guess first thing id investigate is that i2c error message, its possible clanker fucked up the handler installation since there's some differences in api there between acpica and uacpi

#

But the errors it produces are really strange

gentle peak Jun 18, 2026, 7:36 AM

#

fiery turtle Huh lol

my thinking is that after the hung task warnings the kernel does something to free up whatever mount is blocking on but that takes longer than the timeout

fiery turtle Jun 18, 2026, 7:36 AM

#

Yeah maybe

gentle peak Jun 18, 2026, 7:37 AM

#

fiery turtle I guess first thing id investigate is that i2c error message, its possible clank...

would it by any chance involve one of these functions

rn_image_picker_lib_temp_6874cbf4-8f91-4483-a4f2-dd6497f9e818.jpg

fiery turtle Jun 18, 2026, 7:41 AM

#

I think it's acpi_install_address_space_handler family of helpers

fiery turtle Jun 18, 2026, 7:48 AM

#

gentle peak then a bunch of kernel hung task warnings

also whats strange is it hangs on some mutex lock, even outside of any uACPI code 🤔

gentle peak Jun 18, 2026, 7:48 AM

#

id guess that has something to do with the i2c stuff as well

fiery turtle Jun 18, 2026, 7:49 AM

#

yeah probably

gentle peak Jun 18, 2026, 7:49 AM

#

like maybe whatever's holding the mutex is waiting on something that would get done by the i2c space handler

fiery turtle Jun 18, 2026, 7:49 AM

#

could be

#

But also it wasnt hanging before you recompiled with the new config right? Did that have i2c disabled or smth

gentle peak Jun 18, 2026, 7:52 AM

#

well the old config was just defconfig with uapci, amdgpu, and nvme

#

I doubt it had i2c and, even if it did, the specific i2c device used here

fiery turtle Jun 18, 2026, 7:54 AM

#

hm yeah defconfig probably didnt have it

gentle peak Jun 18, 2026, 7:58 AM

#

added a few printfs

rn_image_picker_lib_temp_7fd3f35e-2f2b-4d42-9725-6b4abef73d11.jpg

fiery turtle Jun 18, 2026, 8:04 AM

#

gentle peak added a few printfs

handle = ACPI_HANDLE(adapter->dev.parent); possible that this is NULL so it always tries to install them to root?

gentle peak Jun 18, 2026, 8:12 AM

#

well the very next line below that is a null check

fiery turtle Jun 18, 2026, 8:12 AM

#

ah right

#

    /* Check for an existing internal object */

    obj_desc = acpi_ns_get_attached_object(node);
    if (obj_desc) {
        /*
         * The attached device object already exists. Now make sure
         * the handler is not already installed.
         */
        handler_obj = acpi_ev_find_region_handler(space_id,
                              obj_desc->
                              common_notify.
                              handler);

        if (handler_obj) {
            if (handler_obj->address_space.handler == handler) {
                /*
                 * It is (relatively) OK to attempt to install the SAME
                 * handler twice. This can easily happen with the
                 * PCI_Config space.
                 */
                status = AE_SAME_HANDLER;
                goto unlock_and_exit;
            } else {
                /* A handler is already installed */

                status = AE_ALREADY_EXISTS;
            }

            goto unlock_and_exit;
        }
    } else {

ACPICA has this check also

#

so something about that handle must be wrong or something?

gentle peak Jun 18, 2026, 8:17 AM

#

it does seem to be using the same handle every time

rn_image_picker_lib_temp_e61fe817-c6a9-4553-bb96-4304c7a25477.jpg

fiery turtle Jun 18, 2026, 8:17 AM

#

hmm very interesting

#

like that would definitely fail under acpica

#

maybe the way clanker made parent resolution work is wrong

gentle peak Jun 18, 2026, 8:19 AM

#

I should check whether this happens with the normal kernel

fiery turtle Jun 18, 2026, 8:20 AM

#

oh yeah true, try replacing the uacpi config with normal acpica

#

and keep these logs

gentle peak Jun 18, 2026, 8:20 AM

#

I'll need to replicate the logs in drivers/acpi/acpica but yeah

fiery turtle Jun 18, 2026, 8:20 AM

#

u could probably put them directly in i2c-core-acpi.c?

gentle peak Jun 18, 2026, 8:21 AM

#

oh true

#

but then i'll have to properly recompile the module and regenerate the initramfs instead of just copying bzimage

fiery turtle Jun 18, 2026, 8:23 AM

#

gentle peak but then i'll have to properly recompile the module and regenerate the initramfs...

seems like that file is part of CONFIG_ACPI, which should be y for you

gentle peak Jun 18, 2026, 8:33 AM

#

okay this is weird

#

on acpica there's only one call to acpi_install_address_space_handler

#

oh nevermind they're just spread out more

fiery turtle Jun 18, 2026, 8:35 AM

#

hows the handle value looking

gentle peak Jun 18, 2026, 8:35 AM

#

they are all on different handles though

fiery turtle Jun 18, 2026, 8:35 AM

#

yeah sounds like clanker fucked something up

gentle peak Jun 18, 2026, 8:35 AM

#

and there's still not as many calls as with uacpi

#

there's only 7 where uacpi logged 9

fiery turtle Jun 18, 2026, 8:36 AM

#

could be that it retries some stuff in case it fails from different code paths

gentle peak Jun 18, 2026, 8:42 AM

#

this is very strange

#

the specific i2c device objects that fail to install their space handlers on uacpi never even try to do so on acpica

fiery turtle Jun 18, 2026, 8:43 AM

#

wtf

gentle peak Jun 18, 2026, 8:43 AM

#

$ sudo dmesg | grep 'space handler'
[    1.116434] i2c i2c-0: installing i2c space handler, parent=0xffff8aff01d51410
[    1.121845] i2c i2c-1: installing i2c space handler, parent=0xffff8aff0e781810
[    1.760916] i2c i2c-2: installing i2c space handler, parent=0xffff8aff027870d0
[    1.885632] i2c i2c-11: installing i2c space handler, parent=0xffff8aff1cc87000
[    1.885819] i2c i2c-12: installing i2c space handler, parent=0xffff8aff1cc8a000
[    1.885888] i2c i2c-13: installing i2c space handler, parent=0xffff8aff1cc8a800
[    1.885948] i2c i2c-14: installing i2c space handler, parent=0xffff8aff1cc8e800
[    4.358038] i2c i2c-15: installing i2c space handler, parent=0xffff8aff0312e0d0```

#

and if you look at the uacpi logs its i2c 3-10 that fail

fiery turtle Jun 18, 2026, 8:44 AM

#

maybe the way it checks for _STA on those is somehow returning the wrong value? Although that still doesnt explain the same parent handle

#

you should be able to generate the absolute path to the device object btw i think

gentle peak Jun 18, 2026, 8:49 AM

#

when logging what's causing the early exit: Jun 18 10:48:35 archlinux kernel: ACPI Error: Unknown exception code: 0xFFFFFFED (20251212/utexcep-41) Jun 18 10:48:35 archlinux kernel: i2c i2c-3: i2c space handler: acpi_bus_attach_private_data failed: UNKNOWN_STATUS_CODE Jun 18 10:48:35 archlinux kernel: ACPI Error: Unknown exception code: 0xFFFFFFED (20251212/utexcep-41) Jun 18 10:48:35 archlinux kernel: i2c i2c-4: i2c space handler: acpi_bus_attach_private_data failed: UNKNOWN_STATUS_CODE Jun 18 10:48:35 archlinux kernel: ACPI Error: Unknown exception code: 0xFFFFFFED (20251212/utexcep-41) Jun 18 10:48:35 archlinux kernel: i2c i2c-5: i2c space handler: acpi_bus_attach_private_data failed: UNKNOWN_STATUS_CODE Jun 18 10:48:35 archlinux kernel: ACPI Error: Unknown exception code: 0xFFFFFFED (20251212/utexcep-41) Jun 18 10:48:35 archlinux kernel: i2c i2c-6: i2c space handler: acpi_bus_attach_private_data failed: UNKNOWN_STATUS_CODE Jun 18 10:48:35 archlinux kernel: ACPI Error: Unknown exception code: 0xFFFFFFED (20251212/utexcep-41) Jun 18 10:48:35 archlinux kernel: i2c i2c-7: i2c space handler: acpi_bus_attach_private_data failed: UNKNOWN_STATUS_CODE Jun 18 10:48:35 archlinux kernel: ACPI Error: Unknown exception code: 0xFFFFFFED (20251212/utexcep-41) Jun 18 10:48:35 archlinux kernel: i2c i2c-8: i2c space handler: acpi_bus_attach_private_data failed: UNKNOWN_STATUS_CODE Jun 18 10:48:35 archlinux kernel: ACPI Error: Unknown exception code: 0xFFFFFFED (20251212/utexcep-41) Jun 18 10:48:35 archlinux kernel: i2c i2c-9: i2c space handler: acpi_bus_attach_private_data failed: UNKNOWN_STATUS_CODE Jun 18 10:48:35 archlinux kernel: ACPI Error: Unknown exception code: 0xFFFFFFED (20251212/utexcep-41) Jun 18 10:48:35 archlinux kernel: i2c i2c-10: i2c space handler: acpi_bus_attach_private_data failed: UNKNOWN_STATUS_CODE

fiery turtle Jun 18, 2026, 8:51 AM

#

huh

fiery turtle Jun 18, 2026, 8:52 AM

#

gentle peak when logging what's causing the early exit: ``` Jun 18 10:48:35 archlinux kernel...

thats so strange

#

thats on acpica right?

gentle peak Jun 18, 2026, 8:53 AM

#

yeah

fiery turtle Jun 18, 2026, 8:53 AM

#

acpi_bus_attach_private_data calls acpi_attach_data, which is supposed to attach a user pointer to an aml namespace node, but uacpi doesnt even have a mechanism like that so it uses a clanker designed hash table

#

but why it would fail on acpica is beyond me

#

and with an UNKNOWN_STATUS_CODE

#

maybe log it as an int

gentle peak Jun 18, 2026, 8:53 AM

#

oh the unknown_status_code is because acpi_bus_attach_private_data blindly returns -ENODEV instead of the actual acpi status

#

i'm adding a printf to acpi_bus_attach_private_data now

fiery turtle Jun 18, 2026, 8:54 AM

#

gentle peak oh the unknown_status_code is because acpi_bus_attach_private_data blindly retur...

brilliant KEKW

gentle peak Jun 18, 2026, 8:56 AM

#

AE_ALREADY_EXISTS

fiery turtle Jun 18, 2026, 8:56 AM

#

💀

#

how the hell does it already exist for a unique i2c object

#

Does the clanker rolled hash table attachment check for duplicates also?

gentle peak Jun 18, 2026, 8:57 AM

#

nope

fiery turtle Jun 18, 2026, 8:58 AM

#

Ah, that must be it

#

Ill add the native uacpi_namespace_node_attach_user_data in the next version anyway

gentle peak Jun 18, 2026, 8:59 AM

#

these are the specific semantics acpica has

#

so you can have multiple data things on the same node if they have different handlers

fiery turtle Jun 18, 2026, 9:00 AM

#

gentle peak so you can have multiple data things on the same node if they have different han...

handler is always the same tho, acpi_bus_private_data_handler

#

void acpi_bus_private_data_handler(acpi_handle handle,
                   void *context)
{
    return;
}
EXPORT_SYMBOL(acpi_bus_private_data_handler);

#

very important

gentle peak Jun 18, 2026, 9:00 AM

#

in this case yes but i'd like to actually be correct

fiery turtle Jun 18, 2026, 9:00 AM

#

yeah ofc

#

looks like its also called here https://elixir.bootlin.com/linux/v7.1/source/drivers/acpi/scan.c#L718, but with a different handler

gentle peak Jun 18, 2026, 9:01 AM

#

this should work i hope

fiery turtle Jun 18, 2026, 9:02 AM

#

yeah looks correct

gentle peak Jun 18, 2026, 9:02 AM

#

oh wait no i think i'm overriding e there

fiery turtle Jun 18, 2026, 9:02 AM

#

ah lol

#

yes

gentle peak Jun 18, 2026, 9:09 AM

#

also I just realized: those functions errored on acpica anyway for those objects, they just did so silently, so this can't be the underlying issue

fiery turtle Jun 18, 2026, 9:18 AM

#

maybe something attached some data to them earlier?

gentle peak Jun 18, 2026, 9:19 AM

#

no I mean that the whole i2c failed to attach thing can't be causing all the other issues I experienced

#

the timeouts, the lack of networking, touchpad, etc

fiery turtle Jun 18, 2026, 9:19 AM

#

ah, right

fiery turtle Jun 18, 2026, 9:23 AM

#

gentle peak the timeouts, the lack of networking, touchpad, etc

so does this same kernel boot with all of those working under acpica?

gentle peak Jun 18, 2026, 9:23 AM

#

yes

fiery turtle Jun 18, 2026, 9:24 AM

#

maybe its not related to i2c afterall

gentle peak Jun 18, 2026, 9:24 AM

#

also backlight brightness modification isn't working

fiery turtle Jun 18, 2026, 9:24 AM

#

yeah something really cursed is going on

full nimbus Jun 18, 2026, 10:36 AM

#

flat badge Afaict the main use cases for NPUs etc are for example doing speed recognition l...

Yeah they are good for background image bluring, speech recognition, maybe OCR things too.
This is why iphones have been shipping with them for ages

fiery turtle Jun 18, 2026, 11:12 PM

#

@gentle peak btw are you going to try to debug uacpi-linux hangs & other weird behavior on that laptop further? Id be happy to help where I can, would be cool to get that thing working

gentle peak Jun 18, 2026, 11:13 PM

#

maybe in the future but not any time soon

fiery turtle Jun 18, 2026, 11:13 PM

#

no worries

#

got some other stuff going atm?

fiery turtle Jun 20, 2026, 11:28 AM

#

Wake up babe, new uACPI feature dropped. AML backtraces are now more detailed, they include the offset within the method as well as the opcode that caused the failure. This should make pinpointing real hardware problems easier.

Some examples:

[uACPI][ERROR] aborting method invocation due to previous error: hanging AML while loop
[uACPI][ERROR]     #0 in \HANG()+14/14 at WhileOp

[uACPI][ERROR] attempted access to deleted (<unknown>) operation region \CAS4.MYRE: invalid argument
[uACPI][ERROR] aborting method invocation due to previous error: invalid argument
[uACPI][ERROR]     #0 in \CAS4()+47/47 at StoreOp

[uACPI][ERROR] unable to perform a read from field 00007385aade1d50: parent opregion gone
[uACPI][ERROR] aborting table load due to previous error: invalid argument
[uACPI][ERROR]     #0 in \()+126/137 at DerefOfOp

[uACPI][ERROR] aborting method invocation due to previous error: reached maximum AML call stack depth
[uACPI][ERROR]     #0 in \HANG()+9/9 at MethodCallOp
[uACPI][ERROR]     #1 in \HANG()+9/9 at MethodCallOp
[uACPI][ERROR]     #2 in \HANG()+9/9 at MethodCallOp
[uACPI][ERROR]     #3 in \HANG()+9/9 at MethodCallOp

[uACPI][ERROR] attempted to perform an invalid implicit cast (String -> Package)
[uACPI][ERROR] aborting table load due to previous error: incompatible AML object type
[uACPI][ERROR]     #0 in \()+372/775 at StoreOp

[uACPI][ERROR] Load: table size 17 is too small
[uACPI][ERROR] aborting table load due to previous error: invalid table length
[uACPI][ERROR]     #0 in \()+56/370 at LoadOp
[uACPI][INFO] dynamic load of 'DSDT' (OEM ID 'uTEST' OEM Table ID 'TESTTABL')
[uACPI][ERROR] aborting table load due to previous error: invalid table length
[uACPI][ERROR]     #0 in \()+145/370 at LoadTableOp

#

(not yet merged but going to be there for 6.0)

hollow scaffold Jun 20, 2026, 11:30 AM

#

pretty!

slim panther Jun 21, 2026, 7:14 PM

#

@fiery turtle uacpi has been shilled to geist

#

hallow marten Jun 21, 2026, 7:15 PM

#

imagine fuchsia gets uACPI

#uACPI - a portable and easy-to-integrate ACPI implementation