uACPI - a portable and easy-to-integrate ACPI implementation | Operating System Development | Page 17

frank canopy Dec 23, 2024, 8:59 PM

#

but it should work

mortal yoke Dec 23, 2024, 8:59 PM

#

maybe there could be an environment variable that would allow overriding the flags (that the kernel would then set in eg. .cargo/config.toml if it wants to override them)

#

then if you want to use msvc for some reason you would set that to whatever flags you use for your kernel

rustic compass Dec 23, 2024, 9:00 PM

#

mortal yoke maybe there could be an environment variable that would allow overriding the fla...

thats the next step, i just want it to compile for now

frank canopy Dec 23, 2024, 9:02 PM

#

so there is a /kernel flag for msvc

#

but thats for windows kernel modules so like idk

#

my recommendation is to just install mingw clang or llvm in general

#

and use that

rustic compass Dec 23, 2024, 9:03 PM

#

depending on how fast i can find the flags msvc wants compared to the installation of clang for win11, it will decide if uacpi-rs will support msvc as a compiler

frank canopy Dec 23, 2024, 9:05 PM

#

/link /NODEFAULTLIB outta be equiv-ish for nostdlib

fiery turtle Dec 23, 2024, 9:06 PM

#

khitiara solo saving uacpi-rs on MSVC LULW

#

the MVP

frank canopy Dec 23, 2024, 9:07 PM

#

general regs only is included in /kernel

rustic compass Dec 23, 2024, 9:07 PM

#

3 flags to go

frank canopy Dec 23, 2024, 9:07 PM

#

and i dont think the stack protector exists on the ms callconvs but im not 100% sure so still looking

#

also the /link stuff has to be at the end

#

/link means everything afterwards gets appended to the linker command line

#

ok yeah afaict the stack protector/red zone/whatever you call it isnt a thing on the windows calling convention so that flag isnt relevant

fiery turtle Dec 23, 2024, 9:11 PM

#

yea

frank canopy Dec 23, 2024, 9:11 PM

#

so i think /kernel /link /NODEFAULTLIB oughtta work?

#

try it and let us know ig

#

as long as the link nodefaultlib part is at the end of the command line

#

there can be nothing after it

fiery turtle Dec 23, 2024, 9:12 PM

#

some things people recommend

frank canopy Dec 23, 2024, 9:12 PM

#

unless you have a separate linker command line thing, in which case put /NODEFAULTLIB on the linker command line

fiery turtle Dec 23, 2024, 9:13 PM

#

GS- disables stack probing i think

frank canopy Dec 23, 2024, 9:13 PM

#

Zi is debugging info

#

Oi generates intrinsics equiv to the __builtin stuff

#

the s after EH enables c++ throw/catch and the c after EH puts nothrow on extern "C" by default

#

yep /GS- is your -fno-stack-protector ig

#

and `/c compiles without linking

#

base is your base virt addr, entry is your entry point, and debug full is obvious on the linker command line there

fiery turtle Dec 23, 2024, 9:17 PM

#

so i geuss we need /GS- and /EHsh

frank canopy Dec 23, 2024, 9:17 PM

#

and subsystem:native is "Kernel mode drivers for Windows NT" so yeah that one is right

#

idk about the EHsh tbh depends on if youve got c++ exceptions and stuff

#

also

#

the docs say that any EH option other than /EH- is incompatible with /kernel

fiery turtle Dec 23, 2024, 9:18 PM

#

ah ok

#

then yeah we can drop it

frank canopy Dec 23, 2024, 9:19 PM

#

(for reference the other SUBSYSTEM options are windows boot environment, four different efi variants, posix subsystem for windows nt, console where windows gives it a console when you run it, and windows where you dont get a console because you are supposed to be opening win32 windows)

#

also for reference, /kernel disables c++ exceptions, RTTI, and requires the new() and delete() operators to be defined explicitly, does #define _KERNEL_MODE 1, and prevents use of sse avx etc, and passes /KERNEL to the linker which prevents incremental compilation and issues a warning if you link it with an object that wasnt compiled with /kernel

#

interestingly on x86_64 youre allowed to use SSE and SSE2 but still not allowed to use AVX etc, but on 32-bit x86 you cant use any vectorized things

fiery turtle Dec 23, 2024, 9:24 PM

#

iirc nt has kernel api to use SSE

frank canopy Dec 23, 2024, 9:24 PM

#

on 64-bit anyway

#

whether you use vectorization in the kernel really comes down to what registers youre willing to deal with saving on interrupts lmao

fiery turtle Dec 23, 2024, 9:26 PM

#

no what they do is like

xsave()
sse_stuff()
xrstor()

frank canopy Dec 23, 2024, 9:26 PM

#

ah ok

rustic compass Dec 23, 2024, 9:27 PM

#

comand line error: name of source file not found

frank canopy Dec 23, 2024, 9:29 PM

#

now that i got no idea

rustic compass Dec 23, 2024, 9:41 PM

#

I will try it with clang, if it works then msvc is not supported for now

hasty plinth Dec 24, 2024, 3:53 AM

#

no chad

#

I'm not strongly tied to my vote there btw, and tbh I dont see myself having any use for the io port versions of thse functions. They're handy hooks to have for access to any resource though (I found the lack of them for pci accesses to be annoying, but possibly thats a design issue on my end) so I'm hesitant to vote for removing just because I dont see a use for them right now.

#

sorry for the late reply btw, crazy time of year nooo

fiery turtle Dec 24, 2024, 10:17 AM

#

No rush lol

#

Thanks for sharing

rustic compass Dec 24, 2024, 4:26 PM

#

im getting a shit ton of warnings from uacpi when compiling with clang and it aborts with errors even tho there are none in the log:

📎 message.txt

#

flags:
.no_default_flags(true)
.remove_flag("-Wall")
.remove_flag("-Wextra")
.flag("-fno-stack-protector")
.flag("-mgeneral-regs-only")
.flag("-nostdlib")
.flag("-ffreestanding");
cc.flag("-mno-red-zone");

north holly Dec 24, 2024, 4:27 PM

#

#

to me those look like errors

rustic compass Dec 24, 2024, 4:27 PM

#

missed those nooo

north holly Dec 24, 2024, 4:27 PM

#

125:warning: [email protected]: C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:15:5: error: expected '}' 
149:warning: [email protected]: C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:15:5: error: expected '}' 
173:warning: [email protected]: C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:15:5: error: expected '}' 
197:warning: [email protected]: C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:15:5: error: expected '}' 
221:warning: [email protected]: C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:15:5: error: expected '}' 
245:warning: [email protected]: C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:19:5: error: expected '}' 
263:warning: [email protected]: C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:19:5: error: expected '}' 
281:warning: [email protected]: 7 errors generated.```

#

that's the log with numbers

#

line numbers

#

of where in the log it happened

#

I need to relearn english

rustic compass Dec 24, 2024, 4:29 PM

#

now the question is are these errors caused by cc or did i use a broken commit?

left orbit Dec 24, 2024, 4:32 PM

#

did the CI pass on this commit?

#

also show the actual file it errors out on

#

maybe it's partially written or corrupted

rustic compass Dec 24, 2024, 4:41 PM

#

Commit 55478816db3a4242d714538891a1842634c8ca33
https://github.com/UltraOS/uACPI/commit/55478816db3a4242d714538891a1842634c8ca33
It's the latest can't see if ci pass

GitHub

io: make sure the global lock is released for locked IndexField · U...

Signed-off-by: Daniil Tatianin [email protected]

north holly Dec 24, 2024, 4:41 PM

#

compiled for me

rustic compass Dec 24, 2024, 4:41 PM

#

I will reclone the repository and try again once I'm back home

north holly Dec 24, 2024, 4:43 PM

#

using clang-19, uacpi compiled

rustic compass Dec 24, 2024, 4:43 PM

#

Well what clang do I get when installing it via vs22?

north holly Dec 24, 2024, 4:43 PM

#

idk check

#

use the vs command prompt

#

and do clang --version

#

clang-19 is latest afaik

rustic compass Dec 24, 2024, 4:44 PM

#

Are you aware of a clang version that will definitely not work with uacpi?

north holly Dec 24, 2024, 4:44 PM

#

clang-20 hasn't been released afaict

north holly Dec 24, 2024, 4:44 PM

#

rustic compass Are you aware of a clang version that will definitely not work with uacpi?

nope

left orbit Dec 24, 2024, 4:45 PM

#

i don't think clang-cl does llvm 19 yet

#

but i might be wrong

north holly Dec 24, 2024, 4:45 PM

#

north holly clang-20 hasn't been released afaict

at least debian sid repos don't have it

left orbit Dec 24, 2024, 4:46 PM

#

llvm 20 isn't released yet

rustic compass Dec 24, 2024, 4:46 PM

#

I used clang not clang-cl as clang-cl showed the same behavior to the c flags as msvc

north holly Dec 24, 2024, 4:46 PM

#

it hasn't been released

#

yeah

fiery turtle Dec 24, 2024, 5:06 PM

#

left orbit did the CI pass on this commit?

Yup, on all platforms

left orbit Dec 24, 2024, 5:06 PM

#

how do you know 💀

#

ah

#

i see

#

i blind ☠️

fiery turtle Dec 24, 2024, 5:38 PM

#

basically clicking on green checkmark

winter orbit Dec 24, 2024, 6:54 PM

#

lol I should add windows 10 to the uACPI leader board

#

Would that mean

#

KEKW KEKW

fiery turtle Dec 24, 2024, 6:54 PM

#

Lmao sure that would be hilarious

winter orbit Dec 24, 2024, 6:55 PM

#

goal for the day established

#

meme

fiery turtle Dec 24, 2024, 6:55 PM

#

Do you somehow route all gpes and stuff to windows 10 code as well or how do these bindings even work

#

Since its acpi driver probably has very specific APIs

#

Or e.g. _CRS

winter orbit Dec 24, 2024, 6:57 PM

#

There’s a massive private dispatch table between HAL,
The PPM part of Hal

And pci essentially

do you know how we describe windows as a hybrid kernel usually

#

This isn’t FULLY true

#

there’s a thing called NTOS CORE,

ACPI, PCI,
HAL, ntoskrnl are a “kernel”

#

It being separate files is only spelt of the story since a bunch of ACPI logic is shared between all 4 components, and vice Versa etc etc

fiery turtle Dec 24, 2024, 6:58 PM

#

So how does it interface with acpi.sys?

winter orbit Dec 24, 2024, 6:59 PM

#

You have 3 parts to it I’ll see if I can get all of
Them into some links

#

Hang on..

fiery turtle Dec 24, 2024, 6:59 PM

#

Im curious how u can even convert uacpi specific structs to stuff it expects to get

winter orbit Dec 24, 2024, 7:00 PM

#

it’s not a super easy plug in

fiery turtle Dec 24, 2024, 7:01 PM

#

I suspected as much lol

winter orbit Dec 24, 2024, 7:02 PM

#

https://www.vergiliusproject.com/kernels/x86/windows-10/22h2/HAL_PRIVATE_DISPATCH

https://www.vergiliusproject.com/kernels/x86/windows-10/22h2

Vergilius Project

Take a look into the depths of Windows kernels and reveal more than 60000 undocumented structures.

Vergilius Project

Take a look into the depths of Windows kernels and reveal more than 60000 undocumented structures.

#

Search ACPI on this site

fiery turtle Dec 24, 2024, 7:03 PM

#

These are like structs and exports it expects?

winter orbit Dec 24, 2024, 7:04 PM

#

#

And a lot more

#

The two above are windows 2003

#

NOT what we’re doing

fiery turtle Dec 24, 2024, 7:05 PM

#

Interesting

#

So you implemented literally all of those to make uacpi work with windows 10?

winter orbit Dec 24, 2024, 7:05 PM

#

A lot more than this

#

I’m just at fire stone

#

So I can’t show you my versions of the dispatch tables

#

But it’s 10
Times bigger

fiery turtle Dec 24, 2024, 7:06 PM

#

Can it run with some of them unimplemeted?

winter orbit Dec 24, 2024, 7:06 PM

#

yup it can

fiery turtle Dec 24, 2024, 7:06 PM

#

How does that work?

fiery turtle Dec 24, 2024, 7:07 PM

#

winter orbit I’m just at fire stone

The what?

winter orbit Dec 24, 2024, 7:07 PM

#

it’s a tire store im getting an alignment done on my car

fiery turtle Dec 24, 2024, 7:07 PM

#

Ah

winter orbit Dec 24, 2024, 7:08 PM

#

fiery turtle How does that work?

It depends on what it is

PPM -> okay if you ignore most of this you won’t have cpu power states and Po will just ignore it

fiery turtle Dec 24, 2024, 7:09 PM

#

Does it need a way to evaluate arbitrary aml?

winter orbit Dec 24, 2024, 7:11 PM

#

ACPI.sys is supposed to handled it all internally

fiery turtle Dec 24, 2024, 7:11 PM

#

Ohh I see

#

So all u get is abstracted away helpers?

#

Im curious about how it abstracts all resources that devices consume

winter orbit Dec 24, 2024, 7:12 PM

#

yeah that’s how modern windows is architected atm

now this hasn’t ALWAYS been the case

#

Windows ACPI/Pci/hal has gone through a lot of changes

#

I basically am just choosing 2 versions for now and conforming to that

#

Vista and 10 1607

fiery turtle Dec 24, 2024, 7:13 PM

#

What are the differences between them?

#

And which will reactos ship with

#

And which one does it work with now

winter orbit Dec 24, 2024, 7:14 PM

#

my plan currently is to make it act like vista OOTB and add some preprocessor options to enable the 1607 capabilitiy

#

there’s a few reasons for this

#

But mostly it comes down to us trying to move our kernel to act like vista, and I want to be able to inject our pci and ACPI into vista when needed

#

And then windows 10 is just better for.. many reasons

#

Just to be clear ACPI.sys and Hal have been rewritten like 16 times I think

#

May be more.. I counted at least 16

#

If I had to guess ACPI is written when engineers from bios vendors get contracted out to MS

#

Which happens every few years

#

anyway the other reason is vista ACPI is primitive as hell

fiery turtle Dec 24, 2024, 7:29 PM

#

Interesting

#

I wonder how it implements the ec driver for example

#

And device notifications

winter orbit Dec 24, 2024, 7:30 PM

#

Po is what handles device notifications.

fiery turtle Dec 24, 2024, 7:30 PM

#

Given it has to be this abstraction

winter orbit Dec 24, 2024, 7:30 PM

#

I can show you this actually

fiery turtle Dec 24, 2024, 7:31 PM

#

winter orbit Po is what handles device notifications.

Are they propagated to the respective device driver?

winter orbit Dec 24, 2024, 7:31 PM

#

https://github.com/reactos/reactos/pull/7559

https://github.com/reactos/reactos/pull/5719

GitHub

[COMPBATT] Resuscitate the Composite Battery driver by GeoB99 · Pul...

Purpose
This PR is an attempt to resuscitate the Composite Battery driver which is broken, barely usable and it even asserts the kernel because it improperly uses a function with incorrect paramete...

GitHub

[NTOSKRNL] Implement the power manager by GeoB99 · Pull Request #57...

ReactOS lacks on any fundamental power management mechanisms in the kernel. We do not have power policies support, no power action functions, battery support is inexistent, we do not implement hibe...

#

Depends on the device but most of the time yeah

fiery turtle Dec 24, 2024, 7:31 PM

#

12k lines lol

winter orbit Dec 24, 2024, 7:32 PM

#

You register a notification with Po and then you register a handler with either the device driver or generic

#

and yup that’s just the start

fiery turtle Dec 24, 2024, 7:32 PM

#

Thanks ill take a look

winter orbit Dec 24, 2024, 7:32 PM

#

Our Po is being designed off modern windows so you can guarantee it’s close to what’s happening now in win11 :)

fiery turtle Dec 24, 2024, 7:32 PM

#

Would be cool to have a full win10 compatible implementation for regression testing etc

loud ice Dec 24, 2024, 7:32 PM

#

what is Po?

#

power?

winter orbit Dec 24, 2024, 7:33 PM

#

Power management subsystem

loud ice Dec 24, 2024, 7:33 PM

#

shouldnt it be like Pwr

winter orbit Dec 24, 2024, 7:33 PM

#

fiery turtle Would be cool to have a full win10 compatible implementation for regression test...

Portable headers and a cmake target can make that happen I think

fiery turtle Dec 24, 2024, 7:33 PM

#

winter orbit Portable headers and a cmake target can make that happen I think

I can live with that

winter orbit Dec 24, 2024, 7:33 PM

#

loud ice shouldnt it be like Pwr

nah this is back when all the subsystems were named objects

#

Power Object

fiery turtle Dec 24, 2024, 7:34 PM

#

Then I guess id have to load it as a driver somehow

winter orbit Dec 24, 2024, 7:34 PM

#

ObO

loud ice Dec 24, 2024, 7:34 PM

#

ObO?

winter orbit Dec 24, 2024, 7:34 PM

#

object management Object

#

this is shitty inconsistent naming schemes we shouldn’t be mimicking

#

but all of the functions are PoDoSomeShit

fiery turtle Dec 24, 2024, 7:35 PM

#

winter orbit ObO

So this is what obos stands for @north holly

winter orbit Dec 24, 2024, 7:35 PM

#

stdcall PoRegisterDeviceForIdleDetection(ptr long long long)
@ stdcall PoRegisterDeviceNotify(ptr long long long ptr ptr)
@ stdcall PoRegisterSystemState(ptr long)
@ stdcall PoRequestPowerIrp(ptr long long ptr ptr ptr)
@ stdcall PoRequestShutdownEvent(ptr)

loud ice Dec 24, 2024, 7:35 PM

#

is any of this stable?

#

ah wonderful

#

yes

winter orbit Dec 24, 2024, 7:35 PM

#

not yet this is what I’ve been doing

north holly Dec 24, 2024, 7:35 PM

#

fiery turtle So this is what obos stands for <@841762704766140486>

Yes definitely

winter orbit Dec 24, 2024, 7:36 PM

#

I’m rewriting a bunch of PCI HAL and all of ACPI

Po is getting stable

loud ice Dec 24, 2024, 7:36 PM

#

winter orbit > stdcall PoRegisterDeviceForIdleDetection(ptr long long long) @ stdcall PoRegis...

what syntax is that lol

winter orbit Dec 24, 2024, 7:36 PM

#

Spec files

#

It’s a wine thing

loud ice Dec 24, 2024, 7:36 PM

#

ah

#

huh

fiery turtle Dec 24, 2024, 7:36 PM

#

north holly Yes definitely

Object management object system

winter orbit Dec 24, 2024, 7:36 PM

#

by the way

#

All of this is the root cause of our driver instability

loud ice Dec 24, 2024, 7:36 PM

#

winter orbit Power Object

you sure it stands for that?

winter orbit Dec 24, 2024, 7:36 PM

#

no

loud ice Dec 24, 2024, 7:36 PM

#

it sounds like it should be POwer

winter orbit Dec 24, 2024, 7:37 PM

#

I believe it

loud ice Dec 24, 2024, 7:37 PM

#

which is why pwr would make more sense

#

but otoh

#

they also have Zw

#

which doesnt stand for anything clearly

#

right?

winter orbit Dec 24, 2024, 7:37 PM

#

no clue I’m not actually a big fan of Mimicking MS as much as the other devs

#

If Zw stands for something I’ll eat my shoe

loud ice Dec 24, 2024, 7:38 PM

#

yeah no meaning

#

its to avoid conflict with other symbols or something

north holly Dec 24, 2024, 7:39 PM

#

winter orbit If Zw stands for something I’ll eat my shoe

Zebra Win

winter orbit Dec 24, 2024, 7:39 PM

#

oh snap

loud ice Dec 24, 2024, 7:39 PM

#

ZavoidingconflictWithothercode

#

:^)

north holly Dec 24, 2024, 7:39 PM

#

winter orbit oh snap

Eat ur shoe now

winter orbit Dec 24, 2024, 7:39 PM

#

ZwINeedANap

winter orbit Dec 24, 2024, 7:39 PM

#

north holly Eat ur shoe now

Yes sir

north holly Dec 24, 2024, 7:39 PM

#

winter orbit ZwINeedANap

ZwFr

loud ice Dec 24, 2024, 7:39 PM

#

winter orbit ZwINeedANap

no NtINeedANap to match it

#

so its a fake api clearly

#

you lied

winter orbit Dec 24, 2024, 7:40 PM

#

Exposed

loud ice Dec 24, 2024, 7:40 PM

#

yeah

#

busted

north holly Dec 24, 2024, 7:40 PM

#

ZwIAmInMyBedRn

#

ZwItIsOnly3PM

winter orbit Dec 24, 2024, 7:40 PM

#

Wait till the ROS prefix comes into play

#

I have tons of those in my modern video driver work

north holly Dec 24, 2024, 7:40 PM

#

ZwIShouldNotBeThisTired

#

@fiery turtle is there a way to integrate uacpi in my brain so I can call _WAK and wake up

fiery turtle Dec 24, 2024, 7:42 PM

#

yes

#

implement kernel api.h

north holly Dec 24, 2024, 7:42 PM

#

Ok

fiery turtle Dec 24, 2024, 7:42 PM

#

how good is your brains aml

north holly Dec 24, 2024, 7:43 PM

#

Like server hardware

#

Aml

#

Well tested

winter orbit Dec 24, 2024, 7:43 PM

#

lmao

north holly Dec 24, 2024, 7:43 PM

#

Except uacpi just pfed

#

In my brain

#

Damn it memory corruption

winter orbit Dec 24, 2024, 7:43 PM

#

I should see if uACPI blows up in my dell server

#

and get a dump from it and PR it to that repo

north holly Dec 24, 2024, 7:44 PM

#

north holly Damn it memory corruption

I'll reboot

#

And see if that fixes the buf

#

*bug

fiery turtle Dec 24, 2024, 7:44 PM

#

winter orbit I should see if uACPI blows up in my dell server

it runs quite a lot of server blobs during CI so probably not

#

but would be nice if u tried

winter orbit Dec 24, 2024, 7:44 PM

#

very cool

fiery turtle Dec 24, 2024, 7:44 PM

#

winter orbit and get a dump from it and PR it to that repo

big

winter orbit Dec 24, 2024, 7:44 PM

#

I have 400 devices

#

I wonder how many I can add

#

Want to see my hardware testing cart

fiery turtle Dec 24, 2024, 7:45 PM

#

wait what

#

how

north holly Dec 24, 2024, 7:45 PM

#

Test obos on them

#

Inb4 crashes on 399/400

fiery turtle Dec 24, 2024, 7:45 PM

#

403/400

winter orbit Dec 24, 2024, 7:45 PM

#

fiery turtle Dec 24, 2024, 7:45 PM

#

daamn

#

is this just hobby or why do u have so many

dusky glade Dec 24, 2024, 7:46 PM

#

winter orbit

test nyaux on them

#

please

north holly Dec 24, 2024, 7:46 PM

#

It crashes on all

fiery turtle Dec 24, 2024, 7:46 PM

#

inb4 it kills them all

north holly Dec 24, 2024, 7:46 PM

#

We already know the answer

winter orbit Dec 24, 2024, 7:46 PM

#

fiery turtle is this just hobby or why do u have so many

Well

#

I’m trying to fix reactos

north holly Dec 24, 2024, 7:46 PM

#

Obos crashing in the EC driver on his hardware

winter orbit Dec 24, 2024, 7:46 PM

#

And I work as a bios dev

north holly Dec 24, 2024, 7:47 PM

#

Infy has said some stuff about firmware devs

winter orbit Dec 24, 2024, 7:47 PM

#

He’s right

fiery turtle Dec 24, 2024, 7:47 PM

#

does it boot on all?

#

i mean reactos

winter orbit Dec 24, 2024, 7:47 PM

#

yeah as of a couple months ago

dusky glade Dec 24, 2024, 7:47 PM

#

fiery turtle inb4 it kills them all

😭

#

NUH UH

fiery turtle Dec 24, 2024, 7:47 PM

#

winter orbit yeah as of a couple months ago

nice

north holly Dec 24, 2024, 7:47 PM

#

#1217009725711847465 message

fiery turtle Dec 24, 2024, 7:47 PM

#

do u write aml at work meme

winter orbit Dec 24, 2024, 7:47 PM

#

I have like 6 of these carts

winter orbit Dec 24, 2024, 7:47 PM

#

fiery turtle do u write aml at work <:meme:575445094589661215>

Only a tiny bit

north holly Dec 24, 2024, 7:47 PM

#

#schedulers message

winter orbit Dec 24, 2024, 7:48 PM

#

My position is a UEFI Innovation Engineer

fiery turtle Dec 24, 2024, 7:48 PM

#

lmao

winter orbit Dec 24, 2024, 7:48 PM

#

I do a lot of things

fiery turtle Dec 24, 2024, 7:48 PM

#

winter orbit My position is a UEFI Innovation Engineer

yeah i remember that

north holly Dec 24, 2024, 7:48 PM

#

fiery turtle Dec 24, 2024, 7:48 PM

#

so mostly uefi-related work?

winter orbit Dec 24, 2024, 7:48 PM

#

north holly

This is true

winter orbit Dec 24, 2024, 7:48 PM

#

fiery turtle so mostly uefi-related work?

yes

#

I have never seen anything conform to UEFI and ACPI spec perfectly

fiery turtle Dec 24, 2024, 7:48 PM

#

very true

winter orbit Dec 24, 2024, 7:48 PM

#

driver#### is the most broken thing ever

fiery turtle Dec 24, 2024, 7:49 PM

#

what is that?

north holly Dec 24, 2024, 7:49 PM

#

#1141057599584878645 message

winter orbit Dec 24, 2024, 7:49 PM

#

do you know what boot#### is

fiery turtle Dec 24, 2024, 7:49 PM

#

nope

north holly Dec 24, 2024, 7:49 PM

#

retarbios

winter orbit Dec 24, 2024, 7:49 PM

#

Uh think of it like a list of allowed, and ordered boot options

#

Driver#### is that for driver options

fiery turtle Dec 24, 2024, 7:50 PM

#

is this like a spec, an app, or what?

winter orbit Dec 24, 2024, 7:50 PM

#

hang on

#

https://uefi.org/specs/UEFI/2.10/03_Boot_Manager.html

#

This one

fiery turtle Dec 24, 2024, 7:51 PM

#

ah that one

winter orbit Dec 24, 2024, 7:51 PM

#

Each load option entry resides in a Boot####, Driver####, SysPrep####, OsRecovery#### or PlatformRecovery#### variable where #### is replaced by a unique option number in printable hexadecimal representation using the digits 0-9, and the upper case versions of the characters A-F (0000-FFFF).

fiery turtle Dec 24, 2024, 7:51 PM

#

why is it broken?

winter orbit Dec 24, 2024, 7:51 PM

#

Well on Intel AMI bioses you can replace one of them with zeros and brick your system forever

loud ice Dec 24, 2024, 7:51 PM

#

im guessing people dont implement it correctly

fiery turtle Dec 24, 2024, 7:52 PM

#

winter orbit Well on Intel AMI bioses you can replace one of them with zeros and brick your s...

💀

winter orbit Dec 24, 2024, 7:52 PM

#

most systems have it so it kills itself after the first boot of it existing

loud ice Dec 24, 2024, 7:52 PM

#

lmfao

winter orbit Dec 24, 2024, 7:52 PM

#

I got a lot of horror stories

fiery turtle Dec 24, 2024, 7:52 PM

#

why does that happen

loud ice Dec 24, 2024, 7:52 PM

#

does a cmos reset not fix it?

winter orbit Dec 24, 2024, 7:52 PM

#

nah

#

KEKW

loud ice Dec 24, 2024, 7:52 PM

#

holy shit

winter orbit Dec 24, 2024, 7:52 PM

#

AMI thinks cmos is old people

#

let’s store my uefi boot### into permissible storage

#

or wait here’s a good one

north holly Dec 24, 2024, 7:53 PM

#

you mean this stuff?

loud ice Dec 24, 2024, 7:53 PM

#

yeah but you can still reset efivars i thought?

fiery turtle Dec 24, 2024, 7:53 PM

#

why does removing it brick the system?

winter orbit Dec 24, 2024, 7:54 PM

#

dunno I work at phoenix not AMI I won’t investigate it for them for free

winter orbit Dec 24, 2024, 7:54 PM

#

loud ice yeah but you can still reset efivars i thought?

spec claims yes

fiery turtle Dec 24, 2024, 7:54 PM

#

lol

winter orbit Dec 24, 2024, 7:54 PM

#

Firmware vendors wipe ass with spec

north holly Dec 24, 2024, 7:54 PM

#

wow even firmware devs hate firmware devs

winter orbit Dec 24, 2024, 7:54 PM

#

I do hate firmware devs

#

They’re evil

#

Anyway insyde is the best bios

#

Phoenix is okay most of the time but the latest thinkpads suck

north holly Dec 24, 2024, 7:55 PM

#

winter orbit Anyway insyde is the best bios

were you paid to say that or...

winter orbit Dec 24, 2024, 7:55 PM

#

no I work at phoenix

#

we admit our flaws

#

Insyde rocks

north holly Dec 24, 2024, 7:56 PM

#

you guys admit your flaws inside of rocks /j

winter orbit Dec 24, 2024, 7:56 PM

#

Only then will a true bios be made

north holly Dec 24, 2024, 7:57 PM

#

once infy qwinci and I found a bug with seabios where they would restore ECAM with ECAM after resume from S3

winter orbit Dec 24, 2024, 7:57 PM

#

lol

torpid ferry Dec 24, 2024, 7:57 PM

#

winter orbit Well on Intel AMI bioses you can replace one of them with zeros and brick your s...

i got a samsung laptop (and qookie too) with the uefi brick bug

winter orbit Dec 24, 2024, 7:57 PM

#

You should dump the HII database from any AMD laptop

#

It’s a hilarious

winter orbit Dec 24, 2024, 7:58 PM

#

torpid ferry i got a samsung laptop (and qookie too) with the uefi brick bug

That’s cute

#

Is it a book 3?

north holly Dec 24, 2024, 7:58 PM

#

winter orbit You should dump the HII database from any AMD laptop

how

winter orbit Dec 24, 2024, 7:58 PM

#

Or any of that stupid line

winter orbit Dec 24, 2024, 7:58 PM

#

north holly how

By hooking into the set up screen

north holly Dec 24, 2024, 7:58 PM

#

not that I have an amd laptop

north holly Dec 24, 2024, 7:58 PM

#

winter orbit By hooking into the set up screen

no fnuy command to do it?

torpid ferry Dec 24, 2024, 7:58 PM

#

winter orbit Is it a book 3?

series 5 from 2013

winter orbit Dec 24, 2024, 7:59 PM

#

north holly no fnuy command to do it?

I can’t really tell you that

winter orbit Dec 24, 2024, 7:59 PM

#

torpid ferry series 5 from 2013

I’m sorry

north holly Dec 24, 2024, 7:59 PM

#

winter orbit I can’t really tell you that

🕵️

winter orbit Dec 24, 2024, 7:59 PM

#

You should put it to rest

#

My contract super loose about what I’m allowed to talk about. That’s why I can say these funny things.

#

But there are a few things I can’t say

torpid ferry Dec 24, 2024, 8:00 PM

#

winter orbit You should put it to rest

use it as my server now, it did 10 years of honorable daily driving

winter orbit Dec 24, 2024, 8:00 PM

#

That’s insanely long for a bomb

torpid ferry Dec 24, 2024, 8:00 PM

#

had a tiny ass battery, like 46wh, for a 17.6 inch monster

winter orbit Dec 24, 2024, 8:01 PM

#

I have a Samsung book 3

north holly Dec 24, 2024, 8:01 PM

#

winter orbit That’s insanely long for a bomb

are you the guys who give lenovo their firmware?

#

phoneix

#

or wait

#

brain fart

winter orbit Dec 24, 2024, 8:01 PM

#

Only for a couple lines

#

I mean, pick up a ThinkPad and look at who made the bios

north holly Dec 24, 2024, 8:02 PM

#

I've an ideapad

winter orbit Dec 24, 2024, 8:02 PM

#

That’s not us

#

That has a secret menu by the way

#

With funny options

#

You should unlock it

torpid ferry Dec 24, 2024, 8:03 PM

#

acer has one too

north holly Dec 24, 2024, 8:03 PM

#

winter orbit You should unlock it

how do I find it

torpid ferry Dec 24, 2024, 8:03 PM

#

quite hilarious as you have to press a konami code ass sequence to unlock it

#

and you need it to even enable svm on cpus that support it

winter orbit Dec 24, 2024, 8:03 PM

#

north holly how do I find it

No idea offhand

north holly Dec 24, 2024, 8:10 PM

#

my laptop's bios is InsydeH20

winter orbit Dec 24, 2024, 8:11 PM

#

yes insyde let’s goooo

kindred beacon Dec 24, 2024, 8:11 PM

#

winter orbit https://github.com/reactos/reactos/pull/7559 https://github.com/reactos/reactos...

Managarm more advanced than reactos???????

north holly Dec 24, 2024, 8:12 PM

#

but I can't for the life of me find the secret bios menu

winter orbit Dec 24, 2024, 8:12 PM

#

Managarm crashes less

loud ice Dec 24, 2024, 8:12 PM

#

you sure

winter orbit Dec 24, 2024, 8:12 PM

#

yeah

#

KEKW

loud ice Dec 24, 2024, 8:12 PM

#

doubtful

#

assertos crashes a lot too afaik

winter orbit Dec 24, 2024, 8:13 PM

#

I mean oh no we’re getting so much better dw we don’t crash as much

#

meme

north holly Dec 24, 2024, 8:13 PM

#

damn it

#

https://gist.github.com/Minionguyjpro/444c565eb37385af61e672e40ec68ea6

Gist

How to Unlock Advanced BIOS Settings in InsydeH2O BIOS for 320-14IK...

How to Unlock Advanced BIOS Settings in InsydeH2O BIOS for 320-14IKB, 320-14ISK, 320-15IKB (Type 80XL, Type 80YE), 320 Touch-15IKB (Type 80XN), 320-15ISK, 320-17IKB (Type 80XM), 320-17ISK, 520-15IK...

#

there is something I can run

#

but it's windows only

#

not that I trust it anyway

winter orbit Dec 24, 2024, 8:13 PM

#

damn

loud ice Dec 24, 2024, 8:14 PM

#

the binaries are gone

#

so

#

🤷

left orbit Dec 24, 2024, 8:16 PM

#

https://web.archive.org/web/20240428080531/https://github.com/SmokelessCPUv2/LenovoH2O-Unlocker/releases/latest/download/Unlocker.exe

#

it's just setting an efi variable {6ACCE65D-DA35-4B39-B64B-5ED927A7DC7E}

winter orbit Dec 24, 2024, 8:17 PM

#

we love that

north holly Dec 24, 2024, 8:17 PM

#

hmm setting random EFI variables

#

how safe

winter orbit Dec 24, 2024, 8:17 PM

#

lol

north holly Dec 24, 2024, 8:18 PM

#

(I'll pass)

winter orbit Dec 24, 2024, 8:18 PM

#

It’ll be fineeee

left orbit Dec 24, 2024, 8:18 PM

#

pBuffer = 1;
strcpy(Guid, "{6ACCE65D-DA35-4B39-B64B-5ED927A7DC7E}");
strcpy(Name, "cE!");
if ( GetFirmwareEnvironmentVariableA(Name, Guid, &pBuffer, 1u) )
{
  sub_140001020("Locking\n");
  v4 = 0;
}
else
{
  pBuffer = 1;
  sub_140001020("Unlocking\n");
  v4 = 1;
}
if ( SetFirmwareEnvironmentVariableA(Name, Guid, &pBuffer, v4) )
{
  sub_140001020("Success\n");
}```

#

decompiled

#

not sure how that works

winter orbit Dec 24, 2024, 8:18 PM

#

If you can do it from windows it’s probs pretty safe

#

but then again

#

Fw vendors are dumb

left orbit Dec 24, 2024, 8:19 PM

#

poking random efi variables is probably never safe

#

lmao

#

knowing vendors yeah

winter orbit Dec 24, 2024, 8:19 PM

#

Don’t trust firmware devs

#

We’re all evil

left orbit Dec 24, 2024, 8:19 PM

#

rm -rf /sys/firmware/efi/efivars

winter orbit Dec 24, 2024, 8:19 PM

#

We take an ancient oath to never conform to the spec properly

left orbit Dec 24, 2024, 8:19 PM

#

go see if ur hw vendor is retarded

winter orbit Dec 24, 2024, 8:19 PM

#

no better

#

Replace all data portions with zero

left orbit Dec 24, 2024, 8:19 PM

#

(dont, or do, but i am not responsible for ur hardware meme )

north holly Dec 24, 2024, 8:19 PM

#

left orbit go see if ur hw vendor is retarded

I think doing that is retarted

winter orbit Dec 24, 2024, 8:20 PM

#

it is

mortal yoke Dec 24, 2024, 8:20 PM

#

just modify the bios

left orbit Dec 24, 2024, 8:20 PM

#

oh yeah it absolutely is

winter orbit Dec 24, 2024, 8:20 PM

#

uACPI become “fw-retard” channel

left orbit Dec 24, 2024, 8:20 PM

#

what's even more retarded is that there is a 50% chance that this will brick ur motherboard

winter orbit Dec 24, 2024, 8:20 PM

#

I bet you it’s more than 50%

mortal yoke Dec 24, 2024, 8:21 PM

#

and garbo laptops likely have no good way of emergency flashing a bios lol

winter orbit Dec 24, 2024, 8:21 PM

#

Nothing made in the last 10 years will survive that

north holly Dec 24, 2024, 8:21 PM

#

mortal yoke and garbo laptops likely have no good way of emergency flashing a bios lol

garbo=lenovo?

north holly Dec 24, 2024, 8:21 PM

#

winter orbit Nothing made in the last 10 years will survive that

well jokes on you

#

this was made 11 years ago

#

meme

winter orbit Dec 24, 2024, 8:21 PM

#

listen buddy

#

nooo

left orbit Dec 24, 2024, 8:21 PM

#

mortal yoke and garbo laptops likely have no good way of emergency flashing a bios lol

spi flash that bitch

#

or whatever interface modern laptops use anyway

north holly Dec 24, 2024, 8:21 PM

#

EC

winter orbit Dec 24, 2024, 8:22 PM

#

north holly garbo=lenovo?

I am financially obligated to question this statement

mortal yoke Dec 24, 2024, 8:22 PM

#

thankfully on desktop mobos its pretty common for them to have a dedicated flasher chip on-board that can flash the bios from an usb

#

(on decent ones at least)

left orbit Dec 24, 2024, 8:22 PM

#

yeah but the firmware chip

#

is probably not the ec

#

or part of it

loud ice Dec 24, 2024, 8:22 PM

#

imagine not having CCD meme

#

(cros_ec lets you reflash the bios via special usb afaik)

north holly Dec 24, 2024, 8:23 PM

#

I wonder what "reboot to setup mode" does

#

in my bios

#

it says it clears pk

#

disables secure boot

left orbit Dec 24, 2024, 8:23 PM

#

enters tpm setup mode?

north holly Dec 24, 2024, 8:23 PM

#

and enters setup boot

north holly Dec 24, 2024, 8:23 PM

#

left orbit enters tpm setup mode?

no tpm here iirc

winter orbit Dec 24, 2024, 8:23 PM

#

Probably bricks it too

north holly Dec 24, 2024, 8:24 PM

#

only one way to find out /j

#

it's tempting

winter orbit Dec 24, 2024, 8:25 PM

#

randomly resetting shit has never gone wrong

mortal yoke Dec 24, 2024, 8:26 PM

#

also one thing I don't like is that fw vendors are always hiding the fun options

loud ice Dec 24, 2024, 8:26 PM

#

yeah

winter orbit Dec 24, 2024, 8:26 PM

#

they never hide it very well

#

More people need to read the HII spec

#

https://uefi.org/specs/UEFI/2.10/33_Human_Interface_Infrastructure.html

mortal yoke Dec 24, 2024, 8:27 PM

#

mortal yoke also one thing I don't like is that fw vendors are always hiding the fun options

like the device mode usb controller toggle, random debug settings or some pci hw level things troll

winter orbit Dec 24, 2024, 8:28 PM

#

Steam deck definitely isn’t hiding funny things beneath surface

#

It is definitely not hiding the funniest thing ever

#

Zero funny things in steam deck

north holly Dec 24, 2024, 8:30 PM

#

https://tenor.com/view/i-saw-what-you-deleted-cat-gif-25407007

Tenor

winter orbit Dec 24, 2024, 8:31 PM

#

those who know

north holly Dec 24, 2024, 8:31 PM

#

mango mango?

mortal yoke Dec 24, 2024, 8:32 PM

#

just throw your bios to uefi editor and look what menus they hide and make them accessible

winter orbit Dec 24, 2024, 8:32 PM

#

Uefi easy patcher . Exe

mortal yoke Dec 24, 2024, 8:32 PM

#

thankfully its not exe lol

slim panther Dec 24, 2024, 8:33 PM

#

.EFI

mortal yoke Dec 24, 2024, 8:33 PM

#

(instead its a website)

#

anyway this isn't related to uacpi anymore

north holly Dec 24, 2024, 8:35 PM

#

winter orbit Uefi easy patcher . Exe

Does Phoenix make firmware for acer

winter orbit Dec 24, 2024, 8:35 PM

#

no

#

wait maybe in the past

north holly Dec 24, 2024, 8:35 PM

#

2013

winter orbit Dec 24, 2024, 8:35 PM

#

I doubt it

mortal yoke Dec 24, 2024, 8:35 PM

#

my acer has insyde and its garbage

winter orbit Dec 24, 2024, 8:36 PM

#

nooo insyde my poster child

#

:(

torpid ferry Dec 24, 2024, 8:44 PM

#

somehow the most uefi compliant machine i have is a lumia

hollow elm Dec 24, 2024, 9:24 PM

#

torpid ferry use it as my server now, it did 10 years of honorable daily driving

same it serves nfs and acts as a wifi ap :^)

torpid ferry Dec 24, 2024, 9:34 PM

#

i use it to ssh into it since i only bring my tablet to uni

flat badge Dec 24, 2024, 9:37 PM

#

I'm surprised that you can set a laptop's wifi to ap mode

#

Iirc i tried it on the raspi4 and it didn't work

#

Although it works on some intel chipset that i have

torpid ferry Dec 24, 2024, 9:41 PM

#

works on all of my machines

flat badge Dec 24, 2024, 9:44 PM

#

Ah nvm the issue was not ap mode but attaching the wifi to a bridge

#

Some wifi firmware / hardware refuses to send packets with mac address != the wifi adapter's mac address

pine leaf Dec 24, 2024, 9:46 PM

#

winter orbit no I work at phoenix

wait they're still around? I thought they vanished after fading into obscurity in the 2010s KEKW

#

everything I've seen from the past 10 years is either AMI or Insyde

calm latch Dec 24, 2024, 10:23 PM

#

My laptop has Phoenix bios

#

Made in 2023

vast kestrel Dec 24, 2024, 10:24 PM

#

AMI master race

#

All my hardware has AMI

calm latch Dec 24, 2024, 10:25 PM

#

I think my parents' old laptop also has Phoenix

rustic compass Dec 24, 2024, 10:27 PM

#

left orbit also show the actual file it errors out on

opcodes.c
14| static const struct uacpi_op_spec opcode_table[0x100] = {
15|    UACPI_ENUMERATE_OPCODES  //<<Erros on first char
16| };

18|static const struct uacpi_op_spec ext_opcode_table[] = {
19|    UACPI_ENUMERATE_EXT_OPCODES  //<<Erros on first char
20|};

#

lines 14-20

#

clang version is 18.1.8

#

@fiery turtle clang complains about this code with these error messages:

125:warning: [email protected]: C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:15:5: error: expected '}' 
149:warning: [email protected]: C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:15:5: error: expected '}' 
173:warning: [email protected]: C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:15:5: error: expected '}' 
197:warning: [email protected]: C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:15:5: error: expected '}' 
221:warning: [email protected]: C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:15:5: error: expected '}' 
245:warning: [email protected]: C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:19:5: error: expected '}' 
263:warning: [email protected]: C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:19:5: error: expected '}' 
281:warning: [email protected]: 7 errors generated.

hollow elm Dec 24, 2024, 10:31 PM

#

can you post the preprocessed source code?

#

copy the clang command, add -E and change the -o ... to point to a different file

rustic compass Dec 24, 2024, 10:33 PM

#

where do i have to add the -E?

hollow elm Dec 24, 2024, 10:34 PM

#

to the clang command used to compile the opcodes.o file

rustic compass Dec 24, 2024, 10:36 PM

#

Command "C:\\Program Files\\Microsoft Visual Studio\\2022\\Community\\VC\\Tools\\Llvm\\bin\\clang.exe" "-I" "C:\\Users\\juven\\SynologyDrive\\uacpi-rs\\uacpi-sys\\uacpi-src/include" "-Wall" "-Wextra" "-fno-stack-protector" "-mgeneral-regs-only" "-nostdlib" "-ffreestanding" "-mno-red-zone" "-DUACPI_SIZED_FREES=1" "-o" "C:\\Users\\juven\\Desktop\\9b5d1a837e7b6553-opcodes.o" "-c" "C:\\Users\\juven\\SynologyDrive\\uacpi-rs\\uacpi-sys\\uacpi-src/source/opcodes.c

where exactly do i need to add -E, or can i just append it?

hollow elm Dec 24, 2024, 10:37 PM

#

it can go wherever, just change the path after -o to something else as well

rustic compass Dec 24, 2024, 10:43 PM

#

i hate windows so much

#

cc can execute the compiler invocation command just fine but when i do it in powershell it breaks and complains about too many arguments

north holly Dec 24, 2024, 10:46 PM

#

rustic compass i hate windows so much

Fr

#

https://tenor.com/view/fuck-microsoft-fuck-microsoft-john-malkovich-malkovich-gif-24980256

Tenor

fiery turtle Dec 24, 2024, 10:48 PM

#

rustic compass <@353593948523986945> clang complains about this code with these error messages:...

I mean I compiled it with so many compilers and never seen this error so no idea

#

If it works on MSVC no way it breaks on clang

slim panther Dec 24, 2024, 10:49 PM

#

rustic compass cc can execute the compiler invocation command just fine but when i do it in pow...

try a batch file?

rustic compass Dec 24, 2024, 10:50 PM

#

slim panther try a batch file?

that worked even less

#

i now got it

#

📎 9b5d1a837e7b6553-opcodes.o

#

.\clang.exe -I "C:\\Users\\juven\\SynologyDrive\\uacpi-rs\\uacpi-sys\\uacpi-src/include" -Wall -Wextra -fno-stack-protector -mgeneral-regs-only -E -nostdlib -ffreestanding -mno-red-zone -DUACPI_SIZED_FREES=1 -o "C:\\Users\\juven\\Desktop\\9b5d1a837e7b6553-opcodes.o" -c "C:\\Users\\juven\\SynologyDrive\\uacpi-rs\\uacpi-sys\\uacpi-src/source/opcodes.c"

rustic compass Dec 24, 2024, 10:53 PM

#

fiery turtle If it works on MSVC no way it breaks on clang

msvc broke so bady that i resorted to giving cc clang manually

#

which didnt help much

fiery turtle Dec 24, 2024, 10:54 PM

#

Iirc it compiles on apple clang for Mac os ci

#

So that works at least

hollow elm Dec 24, 2024, 10:55 PM

#

well managarm's kernel is compiled with clang so

rustic compass Dec 24, 2024, 10:56 PM

#

https://github.com/Shinribo/uacpi-rs/tree/1.0-dev this is the branch i am working on and where these errors happen

GitHub

GitHub - Shinribo/uacpi-rs at 1.0-dev

UACPI bindings for Rust. Contribute to Shinribo/uacpi-rs development by creating an account on GitHub.

north holly Dec 24, 2024, 10:57 PM

#

fiery turtle Iirc it compiles on apple clang for Mac os ci

Uacpi compiles for me with clang

rustic compass Dec 24, 2024, 10:58 PM

#

i hope you guys find the error, i will go to bed now and install clang standalone tomorrow if the error cant be found

#

and if that doesnt work i will compile uacpi by hand because apparently clang doesnt have a problem then

hollow elm Dec 24, 2024, 10:59 PM

#

hmm i can reproduce this locally meme ```
qookie@selenium /tmp λ clang 9b5d1a837e7b6553-opcodes.c -c -o foo.o -m32 -std=c23 -target x86_64-win32
C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:15:28656: error: expected '}'
C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:15:28543: note: to match this '{'
C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:15:29052: error: expected '}'
C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:15:28939: note: to match this '{'
C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:15:29448: error: expected '}'
C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:15:29335: note: to match this '{'
C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:15:29843: error: expected '}'
C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:15:29730: note: to match this '{'
C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:15:30490: error: expected '}'
C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:15:30377: note: to match this '{'
C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:19:5193: error: expected '}'
C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:19:5111: note: to match this '{'
7 errors generated.

fiery turtle Dec 24, 2024, 11:00 PM

#

hollow elm hmm i can reproduce this locally <:meme:575445094589661215> ``` qookie@selenium ...

Where -I?

hollow elm Dec 24, 2024, 11:01 PM

#

the file is preprocessed

fiery turtle Dec 24, 2024, 11:01 PM

#

Ah

#

Any ideas what it doesn't like and why it works everywhere else?

slim panther Dec 24, 2024, 11:01 PM

#

hollow elm hmm i can reproduce this locally <:meme:575445094589661215> ``` qookie@selenium ...

can you paste that file pls

hollow elm Dec 24, 2024, 11:02 PM

#

rustic compass

.

#

just rename .o to .c

slim panther Dec 24, 2024, 11:02 PM

#

same line nums?

hollow elm Dec 24, 2024, 11:02 PM

#

yeah i didn't edit the file

hollow elm Dec 24, 2024, 11:03 PM

#

fiery turtle Any ideas what it doesn't like and why it works everywhere else?

a bunch of defines weren't replaced

slim panther Dec 24, 2024, 11:03 PM

#

these line nubers make 0 sense

hollow elm Dec 24, 2024, 11:03 PM

#


static const struct uacpi_op_spec ext_opcode_table[] = {
    { "ReservedExtOp", .decode_ops = { UACPI_PARSE_OP_BAD_OPCODE, }

fiery turtle Dec 24, 2024, 11:03 PM

#

Thats one long line

hollow elm Dec 24, 2024, 11:03 PM

#

UACPI_PARSE_OP_BAD_OPCODE should've been replaced by the define's value

slim panther Dec 24, 2024, 11:03 PM

#

is that not an enum

fiery turtle Dec 24, 2024, 11:03 PM

#

hollow elm `UACPI_PARSE_OP_BAD_OPCODE` should've been replaced by the define's value

Why, it’s an enum no?

hollow elm Dec 24, 2024, 11:03 PM

#

oh is it?

slim panther Dec 24, 2024, 11:03 PM

#

uacpi_parse_op

hollow elm Dec 24, 2024, 11:04 PM

#

so it is

#

why is it all uppercase then meme

slim panther Dec 24, 2024, 11:04 PM

#

^

#

coding style L

fiery turtle Dec 24, 2024, 11:04 PM

#

Like literally every uacpi enum

#

Id make it lower if it was scoped like c++

hollow elm Dec 24, 2024, 11:05 PM

#

fair enough

vast kestrel Dec 24, 2024, 11:05 PM

#

It's pretty common to use uppercase for enums as well

hollow elm Dec 24, 2024, 11:05 PM

#

slim panther these line nubers make 0 sense

because the preprocessor directives change the file and line numbers

slim panther Dec 24, 2024, 11:06 PM

#

💀

#

hell naw

hollow elm Dec 24, 2024, 11:06 PM

#

opcodes.c line 15 is at line 1366 in the preprocessed file

fiery turtle Dec 24, 2024, 11:06 PM

#

slim panther 💀

💀

hollow elm Dec 24, 2024, 11:07 PM

#

https://uwu.foundation/7p42VZs.png what line are we on? i dunno

slim panther Dec 24, 2024, 11:07 PM

#

forgor

hollow elm Dec 24, 2024, 11:08 PM

#

there's a comma missing here

#

!!!

slim panther Dec 24, 2024, 11:08 PM

#

!!!!!!!!!!!!!!!

hollow elm Dec 24, 2024, 11:08 PM

#

https://uwu.foundation/6x3EQNh.png

#

!!!

fiery turtle Dec 24, 2024, 11:09 PM

#

Wait but how does it still compile

#

I think it got preprocessed incorrectly

#

Any chance its a FieldOp?

hollow elm Dec 24, 2024, 11:11 PM

#

CreateDWordFieldOp

fiery turtle Dec 24, 2024, 11:11 PM

#

Yup

#

#define UACPI_DO_BUILD_BUFFER_FIELD_OP(type, code, node_idx, ...)     \
UACPI_OP(                                                             \
    type##FieldOp, code,                                              \
    {                                                                 \
        UACPI_PARSE_OP_TERM_ARG_UNWRAP_INTERNAL,                      \
        UACPI_PARSE_OP_TYPECHECK, UACPI_OBJECT_BUFFER,                \
        UACPI_PARSE_OP_OPERAND,                                       \
        __VA_ARGS__                                                   \
        UACPI_PARSE_OP_CREATE_NAMESTRING_OR_NULL_IF_LOAD,             \
        UACPI_PARSE_OP_SKIP_WITH_WARN_IF_NULL, node_idx,              \
        UACPI_PARSE_OP_OBJECT_ALLOC_TYPED, UACPI_OBJECT_BUFFER_FIELD, \
        UACPI_PARSE_OP_INVOKE_HANDLER,                                \
        UACPI_PARSE_OP_INSTALL_NAMESPACE_NODE, node_idx,              \
    }                                                                 \
)

#

va args removes the comma

hollow elm Dec 24, 2024, 11:11 PM

#

trolled

slim panther Dec 24, 2024, 11:12 PM

#

##__VA_ARGS__ meme

hollow elm Dec 24, 2024, 11:12 PM

#

__VA_OPT__(,) :^)

fiery turtle Dec 24, 2024, 11:12 PM

#

but its not optional in this case right

hollow elm Dec 24, 2024, 11:12 PM

#

ah true

fiery turtle Dec 24, 2024, 11:13 PM

#

UACPI_DO_BUILD_BUFFER_FIELD_OP(                             \
    Create, UACPI_EXT_OP(0x13), 3,                          \
    UACPI_PARSE_OP_OPERAND,                                 \
)                                                           \

#

one example

#

i guess that last comma gets stripped

#

is it possible c23 has something to do with that

hollow elm Dec 24, 2024, 11:14 PM

#

add a comma after __VA_ARGS__?

fiery turtle Dec 24, 2024, 11:14 PM

#

that will break every other compiler no?

hollow elm Dec 24, 2024, 11:14 PM

#

fiery turtle is it possible c23 has something to do with that

the file was preprocessed without any -std

#

and i only have the preprocessed one so

fiery turtle Dec 24, 2024, 11:15 PM

#

like that would fix your version of clang

#

but break other versions i guess

#

because it would be UACPI_PARSE_OP_OPERAND,,

slim panther Dec 24, 2024, 11:16 PM

#

how would you even use this?

fiery turtle Dec 24, 2024, 11:16 PM

#

is there something like negated __VA_OPT__? meme

fiery turtle Dec 24, 2024, 11:16 PM

#

slim panther how would you even use this?

what specifically?

slim panther Dec 24, 2024, 11:17 PM

#

the macro

#

why is it a vararg?

fiery turtle Dec 24, 2024, 11:17 PM

#

fiery turtle ``` UACPI_DO_BUILD_BUFFER_FIELD_OP( \ Create, UA...

.

#

most field ops dont need anything in there

slim panther Dec 24, 2024, 11:18 PM

#

ah

fiery turtle Dec 24, 2024, 11:18 PM

#

but this one has an extra opcode inserted in the middle

slim panther Dec 24, 2024, 11:19 PM

#

slim panther `##__VA_ARGS__` <:meme:575445094589661215>

just tested, this works

fiery turtle Dec 24, 2024, 11:19 PM

#

what does this do?

slim panther Dec 24, 2024, 11:20 PM

#

removes the comma if no vaarg is given

#

you can add a , at the end

fiery turtle Dec 24, 2024, 11:21 PM

#

--- a/include/uacpi/internal/opcodes.h
+++ b/include/uacpi/internal/opcodes.h
@@ -396,7 +396,7 @@ UACPI_OP(                                                             \
         UACPI_PARSE_OP_TERM_ARG_UNWRAP_INTERNAL,                      \
         UACPI_PARSE_OP_TYPECHECK, UACPI_OBJECT_BUFFER,                \
         UACPI_PARSE_OP_OPERAND,                                       \
-        __VA_ARGS__                                                   \
+        ##__VA_ARGS__,                                                \
         UACPI_PARSE_OP_CREATE_NAMESTRING_OR_NULL_IF_LOAD,             \
         UACPI_PARSE_OP_SKIP_WITH_WARN_IF_NULL, node_idx,              \
         UACPI_PARSE_OP_OBJECT_ALLOC_TYPED, UACPI_OBJECT_BUFFER_FIELD, \
@@ -1146,7 +1146,7 @@ UACPI_OP(                                                   \
 )                                                           \
 UACPI_DO_BUILD_BUFFER_FIELD_OP(                             \
     Create, UACPI_EXT_OP(0x13), 3,                          \
-    UACPI_PARSE_OP_OPERAND,                                 \
+    UACPI_PARSE_OP_OPERAND                                  \
 )                                                           \
 UACPI_OUT_OF_LINE_OP(                                       \
     LoadTableOp, UACPI_EXT_OP(0x1F),                        \

#

this seems to work on MSVC at least

hollow elm Dec 24, 2024, 11:22 PM

#

god bless the c preprocessor

fiery turtle Dec 24, 2024, 11:22 PM

#

does it work for u as well qookie?

hollow elm Dec 24, 2024, 11:23 PM

#

i mean i didn't have any issues either way

#

as i said, i'm only testing the preprocessed file Shinribo posted

fiery turtle Dec 24, 2024, 11:23 PM

#

ah

fiery turtle Dec 24, 2024, 11:23 PM

#

fiery turtle ```diff --- a/include/uacpi/internal/opcodes.h +++ b/include/uacpi/internal/opco...

@rustic compass let me know if this fixes it

fiery turtle Dec 24, 2024, 11:24 PM

#

hollow elm god bless the c preprocessor

so stupid lmao

#

since it works in clang 19 i guess its a temporary regression they had?

hollow elm Dec 24, 2024, 11:25 PM

#

seems so

fiery turtle Dec 24, 2024, 11:25 PM

#

it repros for u on clang 18 right

hollow elm Dec 24, 2024, 11:25 PM

#

could probably dig through github issues to find out

#

managarm has clang 18.1.2 and that works

fiery turtle Dec 24, 2024, 11:26 PM

#

hmm

hollow elm Dec 24, 2024, 11:26 PM

#

the failure was on 18.1.8

fiery turtle Dec 24, 2024, 11:26 PM

#

i guess it was very temporary then

hollow elm Dec 24, 2024, 11:26 PM

#

so still chance that it's a regression

#

but yeah it was likely fixed

fiery turtle Dec 24, 2024, 11:26 PM

#

hollow elm managarm has clang 18.1.2 and that works

i keep forgetting u ran an already preprocessed file

hollow elm Dec 24, 2024, 11:27 PM

#

lol

#

i can test locally with clang 19.1.4

fiery turtle Dec 24, 2024, 11:27 PM

#

yeah u can try compiling the test runner with clang

hollow elm Dec 24, 2024, 11:28 PM

#

doing that rn

fiery turtle Dec 24, 2024, 11:28 PM

#

thx

hollow elm Dec 24, 2024, 11:28 PM

#

that compiles

fiery turtle Dec 24, 2024, 11:28 PM

#

i guess it was then

hollow elm Dec 24, 2024, 11:28 PM

#

btw in other news, gcc 15 adds a new warning that triggers for uacpi: ```
/tmp/uACPI/source/namespace.c:24:56: error: initializer-string for array of ‘char’ is too long [-Werror=unterminated-string-initialization]
24 | [UACPI_PREDEFINED_NAMESPACE_GPE] = MAKE_PREDEFINED("_GPE"),
| ^~~~~~
/tmp/uACPI/source/namespace.c:17:22: note: in definition of macro ‘MAKE_PREDEFINED’
17 | .name.text = name_str,
| ^~~~~~~~

fiery turtle Dec 24, 2024, 11:29 PM

#

PLS

hollow elm Dec 24, 2024, 11:29 PM

#

it's crying because there is no space for the null terminator

slim panther Dec 24, 2024, 11:29 PM

#

🤓

fiery turtle Dec 24, 2024, 11:29 PM

#

leave me alone compilers crywojak

hollow elm Dec 24, 2024, 11:29 PM

#

i mean it's a good warning in general, uncovered some bugs in leenocks iirc

fiery turtle Dec 24, 2024, 11:29 PM

#

yes

#

ig

slim panther Dec 24, 2024, 11:29 PM

#

is this because an int cast of a string?

fiery turtle Dec 24, 2024, 11:30 PM

#

its a char[4]

slim panther Dec 24, 2024, 11:30 PM

#

🗿

fiery turtle Dec 24, 2024, 11:30 PM

#

assigned a char[5]

hollow elm Dec 24, 2024, 11:30 PM

#

no, it's because text is char[4] but "_GPE" is char[5]

slim panther Dec 24, 2024, 11:30 PM

#

{'_', 'G', 'P', 'E'}

fiery turtle Dec 24, 2024, 11:30 PM

#

\0

mortal yoke Dec 24, 2024, 11:30 PM

#

or just make it a char[5]

fiery turtle Dec 24, 2024, 11:31 PM

#

it's a uacpi_object_name struct so i can quickly type pun to ints

#

aligned on a 32-bit boundary

hollow elm Dec 24, 2024, 11:31 PM

#

#pragma GCC diagnostic ignored "-Wunterminated-string-initialization" meme

mortal yoke Dec 24, 2024, 11:31 PM

#

fiery turtle it's a uacpi_object_name struct so i can quickly type pun to ints

why do you do that tho?

fiery turtle Dec 24, 2024, 11:31 PM

#

mortal yoke why do you do that tho?

name.id == other_name.id

#

so u32 vs u32 cmp

#

instead of memcmp etc

slim panther Dec 24, 2024, 11:32 PM

#

union trolling {
    char fuck[4];
    uint32_t you;
};

mortal yoke Dec 24, 2024, 11:32 PM

#

pretty sure the compiler would optimize that out anyway

fiery turtle Dec 24, 2024, 11:32 PM

#

it cant unless there's an alignment guarantee

#

also not unless people define it to __builtin_memcmp

#

also memcmp is like harder to type out lol

mortal yoke Dec 24, 2024, 11:33 PM

#

bruh

fiery turtle Dec 24, 2024, 11:33 PM

#

if (uacpi_memcmp(node0.name, node1.name, sizeof(node0.name)) == 0)

#

repeat this every time

#

id rather do node0.name.id == node1.name.id

#

and i can quickly check agaisnt constants etc

#

anyway not that important tbh

hollow elm Dec 24, 2024, 11:37 PM

#

that's not that bad imo

mortal yoke Dec 24, 2024, 11:37 PM

#

fiery turtle it cant unless there's an alignment guarantee

it does optimize it on arches that support unaligned loads tho

fiery turtle Dec 24, 2024, 11:38 PM

#

With built-in or without

kindred beacon Dec 24, 2024, 11:38 PM

#

mortal yoke it does optimize it on arches that support unaligned loads tho

I imagine that would depend on the march

mortal yoke Dec 24, 2024, 11:39 PM

#

well on arches like x86_64 that unconditionally support it it doesn't depend on march

#

and if the arch doesn't support it and you do it manually its going to be screwed that way too

fiery turtle Dec 24, 2024, 11:40 PM

#

mortal yoke and if the arch doesn't support it and you do it manually its going to be screwe...

Union forces the alignment

mortal yoke Dec 24, 2024, 11:40 PM

#

ah yeah

kindred beacon Dec 24, 2024, 11:40 PM

#

You sure? Last time I checked march has a bunch of cost factors for different operations, I'd imagine unaligned access being one

mortal yoke Dec 24, 2024, 11:41 PM

#

well at least when you don't pass any march it optimizes it to just a normal u32 load

#

or with march sandybridge/alderlake

fiery turtle Dec 24, 2024, 11:42 PM

#

And if u use a custom memcmp

mortal yoke Dec 24, 2024, 11:42 PM

#

well yeah if you'd not define uacpi memcmp to the builtin one and compile with ffreestanding then it would be slower

#

but there aren't really a lot of reasons to not define them to the builtin ones to let the compiler optimize calls to them

#

anyway I just wanted to see if the compiler would optimize it, you can ofc keep it how it is KEKW

fiery turtle Dec 24, 2024, 11:44 PM

#

Not a lot of people are well versed in compiler internals

#

And know to use builtins unfortunately

#

Nyaux probably doesn't

#

LULW

mortal yoke Dec 24, 2024, 11:45 PM

#

lol

kindred beacon Dec 24, 2024, 11:45 PM

#

The shade

fiery turtle Dec 24, 2024, 11:46 PM

#

Lmao

#

Nothing personnel kid

left orbit Dec 24, 2024, 11:46 PM

#

when you call the builtin it emits a call to mem* anyway 90% of the time

#

so it's not even like a "magic" fix, you still need to implement the function

fiery turtle Dec 24, 2024, 11:47 PM

#

Yeah but its IR so it can optimize cases like we've just discussed

#

And not actually emit a call

left orbit Dec 24, 2024, 11:47 PM

#

yea i didn't really read it

#

but yeah that's probably the 10% i was talking about

fiery turtle Dec 24, 2024, 11:48 PM

#

Yeah

#

I mean probably most small compares can be optimized to integer loads

#

So its about the size of the objects you're comparing

calm latch Dec 24, 2024, 11:56 PM

#

Don't compilers replace calls to those functions with builtins automatically?

slim panther Dec 24, 2024, 11:56 PM

#

not guaranteed

calm latch Dec 24, 2024, 11:56 PM

#

I had some fun where my libc would break because GCC would replace calloc with call to itself

#

Which would stack overflow

slim panther Dec 24, 2024, 11:57 PM

#

fun

fiery turtle Dec 24, 2024, 11:57 PM

#

Thats a different optimization

#

Recognizing a memcmp and just replacing it with that call

#

Builtins are the other way around, they try to avoid memcmp work entirely and do an ad hoc version

calm latch Dec 24, 2024, 11:57 PM

#

I think what GCC does is replace memcpy with builtin

#

And functions like that

left orbit Dec 24, 2024, 11:58 PM

#

if they match the right signature, it probably does

mortal yoke Dec 24, 2024, 11:58 PM

#

calm latch Don't compilers replace calls to those functions with builtins automatically?

yes but if you use ffreestanding then no

calm latch Dec 24, 2024, 11:58 PM

#

I don't

left orbit Dec 24, 2024, 11:58 PM

#

i would still expect that to happen even in freestanding given the signature is correct

mortal yoke Dec 24, 2024, 11:58 PM

#

it doesn't happen

#

ffreestanding implies fno-builtin

calm latch Dec 24, 2024, 11:59 PM

#

And in freestanding you can tell compilers that the functions should be treated as builtins

left orbit Dec 24, 2024, 11:59 PM

#

ah

calm latch Dec 24, 2024, 11:59 PM

#

Like the printf attribute

fiery turtle Dec 24, 2024, 11:59 PM

#

calm latch And in freestanding you can tell compilers that the functions should be treated ...

Then watch your memcpy get optimized into itself KEKW

calm latch Dec 24, 2024, 11:59 PM

#

I pass -fno-builtins to select c files in my libc

mortal yoke Dec 25, 2024, 12:00 AM

#

calm latch Like the printf attribute

that only affects the format string part tho

#

it doesn't make it an actual builtin the compiler can optimize calls to

calm latch Dec 25, 2024, 12:00 AM

#

Or you can also do macros shit

mortal yoke Dec 25, 2024, 12:00 AM

#

yeah that works

fiery turtle Dec 25, 2024, 12:00 AM

#

calm latch Or you can also do macros shit

You should yeah

calm latch Dec 25, 2024, 12:01 AM

#

uACPI doesn't call memcpy

#

?

fiery turtle Dec 25, 2024, 12:01 AM

#

It calls uacpi_memcpy

#

Which is whatever u tell it

calm latch Dec 25, 2024, 12:02 AM

#

Then there's also loop unrolling I would imagine

#

But idk how well that works

#

"I don't know what I'm doing"

frank canopy Dec 25, 2024, 12:26 AM

#

fiery turtle it's a uacpi_object_name struct so i can quickly type pun to ints

I actually did the exact same thing for the table signatures in zig being able to treat the four byte string as an int is so nice

fiery turtle Dec 25, 2024, 12:28 AM

#

Yeah

frank canopy Dec 25, 2024, 12:30 AM

#

plus i can make the cast an inline function and then itll evaluate at comptime to let me define actual enum constants

fiery turtle Dec 25, 2024, 12:31 AM

#

Nice

frank canopy Dec 25, 2024, 12:39 AM

#

still dont know why ACPI only lets names be four characters long but it sure is convenient for exactly this reason lol

sterile egret Dec 25, 2024, 3:49 AM

#

winter orbit

NEED

winter orbit Dec 25, 2024, 3:54 AM

#

Thank you

sterile egret Dec 25, 2024, 4:04 AM

#

Also...

#

;compile ```c
#include <stdio.h>

int main(int argc, char **argv) {
int value = '\x01\x02\x03\x04';
printf("%x\n", value);
}```

autumn cairnBOT Dec 25, 2024, 4:05 AM

#

Program Output

Compiler Output

<source>: In function 'main':
<source>:4:15: warning: multi-character character constant [-Wmultichar]
    4 |   int value = '\x01\x02\x03\x04';
      |               ^~~~~~~~~~~~~~~~~~

sterile egret Dec 25, 2024, 4:06 AM

#

Multi-character literals are a thing meme

rustic compass Dec 25, 2024, 7:52 AM

#

fiery turtle <@257427487556894722> let me know if this fixes it

good and bad news, 5 error removed, 2 remaining

#

Will send the new error log in an hour

torpid root Dec 25, 2024, 9:25 AM

#

I pulled the latest uacpi and now it deadlocks after

[info] uacpi: enabled 2 GPEs in block _GPE@[0->64]

#

single thread

#

I don't believe it's an issue with my spinlock

#

previous clone was before the leaderboard

rustic compass Dec 25, 2024, 9:46 AM

#

warning: [email protected]: C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:19:5: error: expected '}' 
warning: [email protected]:    19 |     UACPI_ENUMERATE_EXT_OPCODES
warning: [email protected]:       |     ^

warning: [email protected]: C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\uacpi-src/source/opcodes.c:19:5: error: expected '}' 
warning: [email protected]:    19 |     UACPI_ENUMERATE_EXT_OPCODES
warning: [email protected]:       |     ^

📎 message.txt 📎 part2_opcodes.o

#

i rechecked the patch and its exactly what you send me

#

so i dont think i messed up

#

hopefully

fiery turtle Dec 25, 2024, 9:53 AM

#

torpid root I pulled the latest uacpi and now it deadlocks after ``` [info] uacpi: enabled 2...

Perhaps your get thread id is not good?

torpid root Dec 25, 2024, 9:53 AM

#

it's a stub

#

but that shouldn't matter for single thread?

fiery turtle Dec 25, 2024, 9:53 AM

#

You always run uacpi on one thread?

torpid root Dec 25, 2024, 9:54 AM

#

well I don't have a scheduler yet lol

fiery turtle Dec 25, 2024, 9:54 AM

#

Ah

#

Can you check where it deadlocks?

torpid root Dec 25, 2024, 9:54 AM

#

hmm can't currently

#

maybe in 8 ish hours

fiery turtle Dec 25, 2024, 9:55 AM

#

Sure

#

But I think it is an issue with your spinlocks or mutexes, currently I think managarm, pmos and obos all use the latest uacpi

torpid root Dec 25, 2024, 9:56 AM

#

my locks are basically frigg ticket lock

#

I even tried just doing bool is_locked

fiery turtle Dec 25, 2024, 9:56 AM

#

What are your mutexes

torpid root Dec 25, 2024, 9:56 AM

#

fiery turtle What are your mutexes

spinlocks without disabling interrupts halfmemeleft

fiery turtle Dec 25, 2024, 9:57 AM

#

Lol

#

Btw that message is like the last thing it ever prints in log during init

torpid root Dec 25, 2024, 9:57 AM

#

oh lol

fiery turtle Dec 25, 2024, 9:57 AM

#

How did u deduce that it deadlocks?

torpid root Dec 25, 2024, 9:58 AM

#

if I don't actually lock in my locks, it continues

fiery turtle Dec 25, 2024, 9:58 AM

#

It?

torpid root Dec 25, 2024, 9:59 AM

#

I mean the uacpi function returns and kernel continues fine

fiery turtle Dec 25, 2024, 9:59 AM

#

Ah so it never returns from that init?

torpid root Dec 25, 2024, 10:00 AM

#

I don't think so, I haven't checked which function it was, didn't have much time

fiery turtle Dec 25, 2024, 10:00 AM

#

Its namespace_initialize

#

Anyway lemme know once u do a bit more digging

fiery turtle Dec 25, 2024, 10:02 AM

#

rustic compass ``` warning: [email protected]: C:\Users\juven\SynologyDrive\uacpi-rs\uacpi-sys\ua...

BTW any way you can update your clang?

#

This version has some sort of regression in macros expansion

#

Or downgrade it works as well

rustic compass Dec 25, 2024, 10:09 AM

#

apparently vs22 forces clang 18.1.8

left orbit Dec 25, 2024, 10:09 AM

#

update

#

ur build tools

rustic compass Dec 25, 2024, 10:09 AM

#

how? i cant select the clang version in vs22

left orbit Dec 25, 2024, 10:10 AM

#

oh actually the latesat is 18.1.8, you are right

#

also, you can open the visual studio installer and click "update"

#

if there is a newer version it would be downloaded

fiery turtle Dec 25, 2024, 10:17 AM

#

Bruh did they put a shitty version of clang on purpose so people keep using MSVC

left orbit Dec 25, 2024, 10:17 AM

#

no

#

llvm has builds for windows

#

and they ship clang-cl

#

C:\Users\hacker>clang-cl --version
clang version 19.1.5
Target: x86_64-pc-windows-msvc
Thread model: posix
InstalledDir: C:\Program Files\LLVM\bin```

#

if you need this shit for whatever reason lol

fiery turtle Dec 25, 2024, 10:18 AM

#

rustic compass good and bad news, 5 error removed, 2 remaining

I’ll see if there's a similar pattern anywhere in these macros...

left orbit Dec 25, 2024, 10:23 AM

#

i mean realistically the only reason visual studio ships clang-cl is because that's the only way to use an llvm toolchain within visual studio (without cmake), but if you're using rust why are you using visual studio??? and even disregarding that you aren't tied to msbuild, so you can use a proper llvm installation

rustic compass Dec 25, 2024, 10:25 AM

#

In not using vs22 I'm just using clang from it

left orbit Dec 25, 2024, 10:25 AM

#

then don't lol

#

the only reason it's there so you can use it with msbuild projects

rustic compass Dec 25, 2024, 10:26 AM

#

Is there a installer for clang on windows or do I have to compile it from source?

left orbit Dec 25, 2024, 10:26 AM

#

yeah of course

#

https://github.com/llvm/llvm-project/releases

GitHub

Releases · llvm/llvm-project

The LLVM Project is a collection of modular and reusable compiler and toolchain technologies. - llvm/llvm-project

#

there should be an installer

#

llvm-19.1.6.exe

#

https://github.com/llvm/llvm-project/releases/download/llvmorg-19.1.6/LLVM-19.1.6-win64.exe

#

then just add <llvm install path>\bin to your path

gentle peak Dec 25, 2024, 10:54 AM

#

fiery turtle But I think it is an issue with your spinlocks or mutexes, currently I think man...

so does proxima now and it works fine

fiery turtle Dec 25, 2024, 11:24 AM

#

Big

frank canopy Dec 25, 2024, 11:49 AM

#

and imaginarium though my mutexes are still stubbed

rustic compass Dec 25, 2024, 12:48 PM

#

left orbit then just add `<llvm install path>\bin` to your path

fyi the installer asks if it should add the install to path

left orbit Dec 25, 2024, 12:49 PM

#

fair, that works too

rustic compass Dec 25, 2024, 12:55 PM

#

you gonna be shitting me, same errors

#

clang version 19.1.6

#

its using clang.exe from the llvm install and it still breaks

#

is it by any means possible that a certain compile order has to maintained?

fiery turtle Dec 25, 2024, 1:00 PM

#

rustic compass you gonna be shitting me, same errors

Hmm maybe its a windows specific clang bug then

#

Very interesting

#

I'll try to work around once I get to my pc

frank canopy Dec 25, 2024, 1:05 PM

#

I think I'm technically compiling uacpi via clang on windows but I'm doing it by way of zig cc not directly and I'm not sure if zig includes its own llvm here

fiery turtle Dec 25, 2024, 1:06 PM

#

Yeah no idea

#

Maybe shinribos pc is just cursed

rustic compass Dec 25, 2024, 1:07 PM

#

death

frank canopy Dec 25, 2024, 1:11 PM

#

well llvm isnt bundled with zig

#

so yeah im technically compiling uacpi with clang on windows

calm latch Dec 25, 2024, 1:14 PM

#

wls2?

frank canopy Dec 25, 2024, 1:14 PM

#

oh wait nvm again it might actually be linked into zig

frank canopy Dec 25, 2024, 1:14 PM

#

calm latch wls2?

nope, though that also works

#

ok yeah i was sorta wrong

#

zig does bundle its own clang/llvm

#

i didnt realise because there arent any big bundles of llvm files in zig's install dir

#

because it seems its statically linked

#

i think anyway, this is deeper into the zig source than ive gone before lol

#

anyway it is clang 19.1.0

#

idk if anyone sees anything i dont that might matter here but yeah

fiery turtle Dec 25, 2024, 2:51 PM

#

@rustic compass

📎 message.txt

#

give this a try

calm latch Dec 25, 2024, 3:46 PM

#

calm latch wls2?

Because I develop on windows, but I'm used to linux tooling

slow drift Dec 25, 2024, 4:19 PM

#

what font is that

north holly Dec 25, 2024, 4:19 PM

#

the one with english letters

slow drift Dec 25, 2024, 4:20 PM

#

huh

#

is it from uacpi itself

torpid root Dec 25, 2024, 4:28 PM

#

slow drift what font is that

I don't remember

fiery turtle Dec 25, 2024, 4:34 PM

#

the uacpi font

rustic compass Dec 25, 2024, 4:37 PM

#

fiery turtle <@257427487556894722>

works now

#

now i get a billion rust errors caused by the new uacpi bindings KEKW

#

but it fucking works now

#

o boy time to write wrappers around 7kloc of bindgen

#

but we have progress letsgo

slim panther Dec 25, 2024, 4:47 PM

#

@fiery turtle told you meme

#

i know my C preprocessor hacks

rustic compass Dec 25, 2024, 4:47 PM

#

this was your fault?

slim panther Dec 25, 2024, 4:47 PM

#

no, i suggested the fix

rustic compass Dec 25, 2024, 4:48 PM

#

may you have a long and healthy live

fiery turtle Dec 25, 2024, 4:50 PM

#

slim panther i know my C preprocessor hacks

Nice

fiery turtle Dec 25, 2024, 4:51 PM

#

rustic compass but we have progress<:letsgo:1282155513156468747>

LETSFUCKINGGOOOOOO

slim panther Dec 25, 2024, 4:51 PM

#

rustic compass o boy time to write wrappers around 7kloc of bindgen

don't forget this

#

#1217009725711847465 message

rustic compass Dec 25, 2024, 5:02 PM

#

@fiery turtle can you commit these fixes to github?

torpid root Dec 25, 2024, 5:21 PM

#

@fiery turtle this gets stuck, not namespace load

uacpi_install_fixed_event_handler(
    UACPI_FIXED_EVENT_TIMER_STATUS,
    [](uacpi_handle) -> uacpi_interrupt_ret
    {
        timers::acpipm::overflows++;
        return UACPI_INTERRUPT_HANDLED;
    }, nullptr
);

fiery turtle Dec 25, 2024, 5:32 PM

#

torpid root <@353593948523986945> this gets stuck, not namespace load ```c++ uacpi_install_f...

Look at the impl, its very very simple

#

Any chance u get an irq at the same time that tries to acquire the same lock

fiery turtle Dec 25, 2024, 5:33 PM

#

rustic compass <@353593948523986945> can you commit these fixes to github?

Yeah I will

torpid root Dec 25, 2024, 5:46 PM

#

fiery turtle Any chance u get an irq at the same time that tries to acquire the same lock

no uacpi function is called from interrupt handlers

torpid root Dec 25, 2024, 6:01 PM

#

slow drift what font is that

hack nerd font

#

https://github.com/ryanoasis/nerd-fonts/releases/download/v3.3.0/Hack.zip

fiery turtle Dec 25, 2024, 6:08 PM

#

where exactly

fiery turtle Dec 25, 2024, 6:08 PM

#

torpid root <@353593948523986945> this gets stuck, not namespace load ```c++ uacpi_install_f...

got a rip at least?

#

it acquries a mutex, does two field writes, then releases it

torpid root Dec 25, 2024, 6:09 PM

#

yes in the spinlock while loop :(

fiery turtle Dec 25, 2024, 6:09 PM

#

which

torpid root Dec 25, 2024, 6:10 PM

#

which?

fiery turtle Dec 25, 2024, 6:10 PM

#

which function, which spinlock etc

#

give me at least something LULW

torpid root Dec 25, 2024, 6:10 PM

#

I guess this one: ret = uacpi_recursive_lock_acquire(&g_event_lock);

#

line 2227

fiery turtle Dec 25, 2024, 6:10 PM

#

thats a mutex

torpid root Dec 25, 2024, 6:10 PM

#

event.c

torpid root Dec 25, 2024, 6:11 PM

#

fiery turtle thats a mutex

yeah my mutex is a spinlock for now

fiery turtle Dec 25, 2024, 6:11 PM

#

can u use gdb at least

torpid root Dec 25, 2024, 6:16 PM

#

fiery turtle can u use gdb at least

an interrupt

fiery turtle Dec 25, 2024, 6:16 PM

#

do your uacpi spinlocks disable interrupts?

torpid root Dec 25, 2024, 6:16 PM

#

yes

fiery turtle Dec 25, 2024, 6:17 PM

#

apparently not

#

or well

#

that shouldnt matter

#

that spinlock is released right away in that same function

torpid root Dec 25, 2024, 6:19 PM

#

fiery turtle do your uacpi spinlocks disable interrupts?

but isn't a mutex locked in uacpi_install_fixed_event_handler

fiery turtle Dec 25, 2024, 6:19 PM

#

is your unlock_spinlock correct THINKAUTISTIC

torpid root Dec 25, 2024, 6:19 PM

#

torpid root but isn't a mutex locked in uacpi_install_fixed_event_handler

not spinlock

fiery turtle Dec 25, 2024, 6:19 PM

#

torpid root but isn't a mutex locked in uacpi_install_fixed_event_handler

this is a spinlock

#

unrelated to event.c

#

uacpi_status uacpi_write_register_field(
    enum uacpi_register_field field_enum, uacpi_u64 in_value
)
{
    uacpi_status ret;
    uacpi_u8 field_idx = field_enum;
    const struct register_field *field;
    const struct register_spec *reg;
    uacpi_u64 data;
    uacpi_cpu_flags flags;

    if (uacpi_unlikely(field_idx > UACPI_REGISTER_FIELD_MAX))
        return UACPI_STATUS_INVALID_ARGUMENT;

    field = &fields[field_idx];
    reg = &registers[field->reg];

    in_value = (in_value << field->offset) & field->mask;

    flags = uacpi_kernel_lock_spinlock(g_reg_lock);

    if (reg->kind == REGISTER_ACCESS_KIND_WRITE_TO_CLEAR) {
        if (in_value == 0) {
            ret = UACPI_STATUS_OK;
            goto out;
        }

        ret = do_write_register(reg, in_value);
        goto out;
    }

    ret = do_read_register(reg, &data);
    if (uacpi_unlikely_error(ret))
        goto out;

    data &= ~field->mask;
    data |= in_value;

    ret = do_write_register(reg, data);

out:
    uacpi_kernel_unlock_spinlock(g_reg_lock, flags);
    return ret;
}

#

this is where u hang, and this is the only place this spinlock is referenced

torpid root Dec 25, 2024, 6:22 PM

#

I don't understand this

#

beacause spinlock definitely disables interrupts

fiery turtle Dec 25, 2024, 6:22 PM

#

i think it doesnt matter whether it does or not

#

auto ticket = _next_ticket.fetch_add(1, std::memory_order_relaxed); wait why is this relaxed

#

relaxed won't produce lock add

#

relaxed just means it guarantees no tearing

#

but two acquires can produce the same ticket in your case

#

i doubt thats the problem here but i think its still a bug

fiery turtle Dec 25, 2024, 6:29 PM

#

torpid root an interrupt

dump the lock state i guess, which ticket is it expecting, which is the current

torpid root Dec 25, 2024, 6:29 PM

#

no need

#

I found the issue

#

as always, it's my skill problems

fiery turtle Dec 25, 2024, 6:29 PM

#

what was it?

torpid root Dec 25, 2024, 6:30 PM

#

auto current = _serving_ticket.load(std::memory_order_relaxed);
_serving_ticket.store(current + 1, std::memory_order_release);
if constexpr (ints)
    arch_unlock();

vs

if constexpr (ints)
    arch_unlock();
auto current = _serving_ticket.load(std::memory_order_relaxed);
_serving_ticket.store(current + 1, std::memory_order_release);

#

I'm sorry

fiery turtle Dec 25, 2024, 6:30 PM

#

lol

#

npnp

fiery turtle Dec 25, 2024, 6:31 PM

#

fiery turtle `auto ticket = _next_ticket.fetch_add(1, std::memory_order_relaxed);` wait why i...

should fix this as well probably

#

or well its fine as long as you dont have smp

torpid root Dec 25, 2024, 6:31 PM

#

fiery turtle relaxed won't produce `lock add`

that's a ticket counter

#

doesn't matter if it's acquite or not

fiery turtle Dec 25, 2024, 6:31 PM

#

in your case two threads will end up with the same ticket

#

if i understand relaxed semantics correctly

torpid root Dec 25, 2024, 6:32 PM

#

hmmm are you sure

fiery turtle Dec 25, 2024, 6:32 PM

#

relaxed just means a thread wont end up with garbage ticket as a result of other thread doing a partial write

#

which isnt possible on x86

#

so its a no-op

fiery turtle Dec 25, 2024, 6:33 PM

#

torpid root hmmm are you sure

objdump to be sure

torpid root Dec 25, 2024, 6:33 PM

#

because it's still atomic, thread one returns and increments while thread two waits for it to finish doing so

fiery turtle Dec 25, 2024, 6:33 PM

#

if it isnt a lock; inc then im right

torpid root Dec 25, 2024, 6:33 PM

#

sure

fiery turtle Dec 25, 2024, 6:33 PM

#

torpid root because it's still atomic, thread one returns and increments while thread two wa...

in this case i think its atomic in the sense that it will never be partially updated

hollow elm Dec 25, 2024, 6:34 PM

#

cc @flat badge because this is the frigg ticket spinlock impl :^)

fiery turtle Dec 25, 2024, 6:35 PM

#

only this operation

flat badge Dec 25, 2024, 6:35 PM

#

relaxed still generates lock xadd

fiery turtle Dec 25, 2024, 6:35 PM

#

hmm u sure?

torpid root Dec 25, 2024, 6:35 PM

#

yes

#

fiery turtle Dec 25, 2024, 6:35 PM

#

i wonder why

flat badge Dec 25, 2024, 6:36 PM

#

relaxed doesn't mean that it's not atomic

#

just that it's not ordered with other non-atomic accesses

slow drift Dec 25, 2024, 6:36 PM

#

torpid root hack nerd font

lol tysm

torpid root Dec 25, 2024, 6:36 PM

#

np

fiery turtle Dec 25, 2024, 6:36 PM

#

hmm well docs dont make it sound like that

flat badge Dec 25, 2024, 6:37 PM

#

only this operation's atomicity is guaranteed

fiery turtle Dec 25, 2024, 6:37 PM

#

yes, so u read 1 and write back 2, 2 is stored

flat badge Dec 25, 2024, 6:37 PM

#

no, that wouldn't be atomic

fiery turtle Dec 25, 2024, 6:37 PM

#

u dont read or write partial values

#

atomic means it's done as one singular op

hollow elm Dec 25, 2024, 6:38 PM

#

the whole fetch, modify, write is one atomic op

flat badge Dec 25, 2024, 6:38 PM

#

a relaxed atomic op is not the same as a non-atomic op

fiery turtle Dec 25, 2024, 6:38 PM

#

yes

#

non-atomic doesn't guarantee no tearing if an interrupt comes in the middle for example

#

a relaxed atomic does

flat badge Dec 25, 2024, 6:39 PM

#

fetch_add guarantees in any case:

#

Atomically replaces the current value with the result of arithmetic addition of the value and arg. That is, it performs atomic post-increment. The operation is a read-modify-write operation.

fiery turtle Dec 25, 2024, 6:39 PM

#

but apparently it also enforces lock in this case

flat badge Dec 25, 2024, 6:39 PM

#

the memory order is a guarantee on top of that

fiery turtle Dec 25, 2024, 6:39 PM

#

yeah i guess its just how fetch_add is to prevent foot shooting

flat badge Dec 25, 2024, 6:39 PM

#

if you want the op that you're describing, you need to do

fiery turtle Dec 25, 2024, 6:40 PM

#

just relaxed load and relaxed store

flat badge Dec 25, 2024, 6:40 PM

#

temp = x.load(std::memory_order_relaxed);
x.store(temp + 1, std::memory_order_relaxed);

fiery turtle Dec 25, 2024, 6:40 PM

#

yeah

flat badge Dec 25, 2024, 6:40 PM

#

which we actually do use sometimes in managarm

#

but in the context of a mutex, it's true that relaxed ordering would not be enough

#

the reason why it is enough here is that it's followed by a store-release

fiery turtle Dec 25, 2024, 6:41 PM

#

yea

flat badge Dec 25, 2024, 6:42 PM

#

torpid root ```c++ auto current = _serving_ticket.load(std::memory_order_relaxed); _serving_...

Also, here you need to consider whether this can be nested in a region that already has interrupts disabled

#

Usually you want to restore to the previous state instead of re-enabling them unconditionally in unlock()

fiery turtle Dec 25, 2024, 6:43 PM

#

wait u dont do that?

flat badge Dec 25, 2024, 6:43 PM

#

the naive implementation will break if you take two spinlocks at the same time

torpid root Dec 25, 2024, 6:43 PM

#

yes arch_unlock does that

flat badge Dec 25, 2024, 6:43 PM

#

ah i see, then that should work

fiery turtle Dec 25, 2024, 6:44 PM

#

torpid root yes arch_unlock does that

why not do the check locally instead of calling into arch-specific stuff

torpid root Dec 25, 2024, 6:44 PM

#

circular dependencies

fiery turtle Dec 25, 2024, 6:44 PM

#

like if (should_reenable) arch_really_unlock()

#

but u store previous state somewhere right

torpid root Dec 25, 2024, 6:44 PM

#

fiery turtle why not do the check locally instead of calling into arch-specific stuff

it's not arch specific, just in a separate source file

fiery turtle Dec 25, 2024, 6:45 PM

#

hmm

torpid root Dec 25, 2024, 6:45 PM

#

because arch.cppm -> lib.cppm -> lock.cppm
and arch specific functions are in arch.cppm
so lock.cpp imports it and calls those

fiery turtle Dec 25, 2024, 6:46 PM

#

I see

#

anyway I just realized uacpi has completely broken event initialization order

#

strange this never caused issues

#

or actually i guess its less bad than i thought but still

#

if there is a global firmware lock contention at namespace load time it will hang forever

#

because the SCI handler is not yet installed at that point

torpid root Dec 25, 2024, 6:49 PM

#

so this skill issue searching actually resulted in something good?

#

meme

fiery turtle Dec 25, 2024, 6:50 PM

#

torpid root so this skill issue searching actually resulted in something good?

yeah

#

always spot random stuff when people report issues

torpid root Dec 25, 2024, 6:50 PM

#

I shall do it more often then

fiery turtle Dec 25, 2024, 6:51 PM

#

lmao

torpid root Dec 25, 2024, 6:51 PM

#

jk

fiery turtle Dec 25, 2024, 6:51 PM

#

if i discover a uacpi bug for every false uacpi bug report its good enough i guess

calm latch Dec 25, 2024, 6:56 PM

#

fiery turtle anyway I just realized uacpi has completely broken event initialization order

Could it be the cause for EC trouble on my PC?

#

The cosmetic issue

fiery turtle Dec 25, 2024, 7:02 PM

#

nah

fiery turtle Dec 25, 2024, 7:02 PM

#

calm latch Could it be the cause for EC trouble on my PC?

what was the problem there exactly?

calm latch Dec 25, 2024, 7:05 PM

#

That it complains about some uninitialized variable

north holly Dec 25, 2024, 7:05 PM

#

test obos on it lol

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