uACPI - a portable and easy-to-integrate ACPI implementation | Operating System Development | Page 31

winter orbit Jan 7, 2025, 11:16 AM

#

Bootmgr has an error code for ACPI failure

#

That can be cause by well it’s initialization failing

#

Go figure

#

meme

fiery turtle Jan 7, 2025, 11:16 AM

#

yeah

#

makes sense i guess

#

i wanted to make a separate uacpi barebones mode for this

#

with just the table api

#

so e.g. microkernels can have two instances of uacpi

#

one in the kernel for just table parsing

#

and the full one in userspace

winter orbit Jan 7, 2025, 11:18 AM

#

If you did that I’d probably replace what we’re doing in Hal with it too

fiery turtle Jan 7, 2025, 11:18 AM

#

nice

winter orbit Jan 7, 2025, 11:18 AM

#

User mode uACPI would be cool

fiery turtle Jan 7, 2025, 11:19 AM

#

well that exists already meme

winter orbit Jan 7, 2025, 11:19 AM

#

I wanted to design a micro kernel but ROS kinda took over life

fiery turtle Jan 7, 2025, 11:19 AM

#

e.g. pmos runs it in userspace

fiery turtle Jan 7, 2025, 11:19 AM

#

winter orbit I wanted to design a micro kernel but ROS kinda took over life

(personally im not a fan of microkernels lol but yeah)

winter orbit Jan 7, 2025, 11:20 AM

#

Well there’s no main stream desktop os that uses that design for a reason

#

spurdo

loud ice Jan 7, 2025, 11:20 AM

#

lol

#

yeah

fiery turtle Jan 7, 2025, 11:20 AM

#

yeah

#

theres no main stream microkernel os period

#

other than niche stuff

#

(switch etc)

loud ice Jan 7, 2025, 11:21 AM

#

the switch kernel is the most general purpose it gets i think

fiery turtle Jan 7, 2025, 11:22 AM

#

yeah turns out 10 ipc calls per disk read request is not that great

winter orbit Jan 7, 2025, 11:22 AM

#

psh

#

it’s fineeeee

loud ice Jan 7, 2025, 11:22 AM

#

you only need one

#

or two

#

if you are extremely careful about implementation

fiery turtle Jan 7, 2025, 11:22 AM

#

u have to have insanely clean design to allow one

#

two is more realistic

loud ice Jan 7, 2025, 11:23 AM

#

to allow one you need to bake disk drivers with filesystem drivers

winter orbit Jan 7, 2025, 11:23 AM

#

I’m more likely to complete this terrible thing then ever make a micro kernel now

loud ice Jan 7, 2025, 11:23 AM

#

or mess around with ebpf or something like that

fiery turtle Jan 7, 2025, 11:23 AM

#

or have some sort of transparent handle that can dispatch directly to the nvme server

loud ice Jan 7, 2025, 11:23 AM

#

winter orbit I’m more likely to complete this terrible thing then ever make a micro kernel n...

do you have a posix subsystem?

#

damn

winter orbit Jan 7, 2025, 11:24 AM

#

no

#

I made one

fiery turtle Jan 7, 2025, 11:24 AM

#

winter orbit I’m more likely to complete this terrible thing then ever make a micro kernel n...

linux subsystem for reactos?

loud ice Jan 7, 2025, 11:24 AM

#

ah

winter orbit Jan 7, 2025, 11:24 AM

#

no not Linux subsystem

#

A NT style posix subsystem

loud ice Jan 7, 2025, 11:24 AM

#

posix subsystem

#

it was a thing in nt a loooong time ago

winter orbit Jan 7, 2025, 11:24 AM

#

Binaries are PE,
It’s an NT personality that’s posix 1.1 compliant

loud ice Jan 7, 2025, 11:24 AM

#

before the NT people got mad at the open group for being annoying

#

and didnt implement the rest of posix

winter orbit Jan 7, 2025, 11:25 AM

#

I’m not gonna implement the rest of posix either /s

fiery turtle Jan 7, 2025, 11:25 AM

#

whats more annoying is they expose some posix api but with underscores for some reason

winter orbit Jan 7, 2025, 11:25 AM

#

This had a GDI TO X server wrapper thing

#

That I could compile xeyes

#

Without modifying the source really at all into an exe

#

That would work

#

Because this was based off another guys idea from 2002

#

And I just fixed it

#

KEKW

loud ice Jan 7, 2025, 11:26 AM

#

winter orbit This had a GDI TO X server wrapper thing

wha

#

wtf

#

thats based though

winter orbit Jan 7, 2025, 11:26 AM

#

Only ever ran xeyes

#

but I could port most Linux console apps

loud ice Jan 7, 2025, 11:27 AM

#

still incredibly based

fiery turtle Jan 7, 2025, 11:27 AM

#

i mean isnt the posix layer for reactos fundamentally going to suck because e.g. fork takes 20 syscalls and still cant compete with native fork in terms of accuracy

winter orbit Jan 7, 2025, 11:27 AM

#

nah windows has a native fork

#

It’s just hidden

fiery turtle Jan 7, 2025, 11:27 AM

#

they made one for wsl1 or something?

winter orbit Jan 7, 2025, 11:28 AM

#

no they made it for the original posix subsystem

fiery turtle Jan 7, 2025, 11:28 AM

#

how well does it work

winter orbit Jan 7, 2025, 11:28 AM

#

decent

Reactos’s terrible

fiery turtle Jan 7, 2025, 11:28 AM

#

lmfao

winter orbit Jan 7, 2025, 11:28 AM

#

I’ve ported my code to windows 2000 before

#

And it was just Linux but better

#

(I’m bias)

#

but you know what I mean,
MS could have made the posix subsystem great

fiery turtle Jan 7, 2025, 11:29 AM

#

embrace extinguish.. whatever

winter orbit Jan 7, 2025, 11:29 AM

#

Reactos is more likely to implement a WSL replacement

calm latch Jan 7, 2025, 11:30 AM

#

I init uACPI very late in pmOS
Like it's running in userspace so the kernel is fully inited

fiery turtle Jan 7, 2025, 11:30 AM

#

i mean the way wsl2 is implemented is the most optimal way to do it

#

just a vm

winter orbit Jan 7, 2025, 11:30 AM

#

Pretty much

fiery turtle Jan 7, 2025, 11:30 AM

#

hyper-v for ros when?

winter orbit Jan 7, 2025, 11:30 AM

#

I spent a lot of
Time on AMD-SVM

#

someone was working on it

#

The funny thing is

#

I want to see us make an android app box

#

Like how wsl works

fiery turtle Jan 7, 2025, 11:31 AM

#

the only downside of wsl2 is it uses virtio-fs which is slow as shit

winter orbit Jan 7, 2025, 11:31 AM

#

So our mobile OS can take over the world

fiery turtle Jan 7, 2025, 11:32 AM

#

winter orbit So our mobile OS can take over the world

lol

#

yeah phones with acpi and uefi would be nice

#

instead of proprietary crap linux loader

winter orbit Jan 7, 2025, 11:32 AM

#

They exist now

calm latch Jan 7, 2025, 11:32 AM

#

fiery turtle theres no main stream microkernel os period

MacOS has a kernel which was a microkernel at some point KEKW

fiery turtle Jan 7, 2025, 11:32 AM

#

i know but barely

winter orbit Jan 7, 2025, 11:33 AM

#

Also I had a design for a DT.sys

#

and a Linux boot stub to start NT

#

Chaos

calm latch Jan 7, 2025, 11:33 AM

#

loud ice if you are extremely careful about implementation

You just cache disks to kernel (?)

fiery turtle Jan 7, 2025, 11:33 AM

#

winter orbit Also I had a design for a DT.sys

nooooo

winter orbit Jan 7, 2025, 11:34 AM

#

you’re right

We willl see if qcom makes good on their promises

#

They have some what projects popping up now taking the uefi blobs and porting mainline EDK

#

With the blobs acting under the driver architecture

fiery turtle Jan 7, 2025, 11:34 AM

#

one thing u can respect ms for is them pushing actual standards

#

instead of insanity we have now

winter orbit Jan 7, 2025, 11:34 AM

#

That’s why I choose ROS tbh

#

I was either going to work on haiku or ROS

#

I wanted to do an nvidia gpu driver

#

Oh how plans changed

fiery turtle Jan 7, 2025, 11:35 AM

#

lol

#

yeah

winter orbit Jan 7, 2025, 11:36 AM

#

Now I’m just running official one like a sane person

#

Well

More sane

fiery turtle Jan 7, 2025, 11:36 AM

#

indeed

calm latch Jan 7, 2025, 11:41 AM

#

uDRM wanted to have an Nvidia driver (?)

fiery turtle Jan 7, 2025, 11:41 AM

#

if marvin doesnt lose motivation

winter orbit Jan 7, 2025, 11:42 AM

#

I think reversing nvidia graphics will be easier with the WDDM stuff I’m doing so maybe that’ll be possible one day

fiery turtle Jan 7, 2025, 11:44 AM

#

also just found a new ACPICA bug

#

should probably fix it

#

they forgor an if statement

calm latch Jan 7, 2025, 11:45 AM

#

I wanted to look at AMD GPUs, but I have no idea how GPUs work at all so idk

fiery turtle Jan 7, 2025, 11:46 AM

#

amd gpus are pretty well documented

winter orbit Jan 7, 2025, 11:46 AM

#

And Intel.

I started with messing with Intel HD graphics

calm latch Jan 7, 2025, 11:46 AM

#

fiery turtle amd gpus are pretty well documented

They are not

#

The interesting parts are proprietary

fiery turtle Jan 7, 2025, 11:47 AM

#

winter orbit And Intel. I started with messing with Intel HD graphics

i did at one point as well, did u see my baremetal hello world triangle?

winter orbit Jan 7, 2025, 11:47 AM

#

no I didn’t show me

fiery turtle Jan 7, 2025, 11:48 AM

#

#

this is full 3d pipeline igpu triangle

winter orbit Jan 7, 2025, 11:48 AM

#

That’s most of the way there to be honest

fiery turtle Jan 7, 2025, 11:48 AM

#

i dumped the shaders from mesa and stuff

winter orbit Jan 7, 2025, 11:48 AM

#

Super cool

fiery turtle Jan 7, 2025, 11:49 AM

#

then assembled the batch buffer and submitted it to the command streamer ring

#

a lot of trial and error and black screen or even cpu resets

winter orbit Jan 7, 2025, 11:49 AM

#

What OS is this on?

#

Or just like directly snapped on blob

fiery turtle Jan 7, 2025, 11:49 AM

#

its just a bare kernel with literally a main and nothing else

winter orbit Jan 7, 2025, 11:49 AM

#

did you post it anywhere?

fiery turtle Jan 7, 2025, 11:50 AM

#

to leos dms meme

#

i have it available

#

its a mess

#

i can upload to pastebin if u want

winter orbit Jan 7, 2025, 11:50 AM

#

I’m curious for sure

#

I need to find my first projects with it

#

I promise it’s more of a mess

#

chad

fiery turtle Jan 7, 2025, 11:51 AM

#

https://pastebin.com/8jDjB04B

Pastebin

#include "common/log.h"#include "common/types.h"#include "common/al...

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.

winter orbit Jan 7, 2025, 11:52 AM

#

Still less code than a Vulkan triangle

#

KEKW KEKW KEKW KEKW KEKW KEKW KEKW KEKW

fiery turtle Jan 7, 2025, 11:52 AM

#

the most interesting thing is build_command_buffer

#

and ig render_ring_build

fiery turtle Jan 7, 2025, 11:52 AM

#

winter orbit Still less code than a Vulkan triangle

lmfao

winter orbit Jan 7, 2025, 11:54 AM

#

That’s super neat

#

Is this 4th gen?

#

Like hd4000

fiery turtle Jan 7, 2025, 11:54 AM

#

somewhere around that

#

actually no

#

its ironlake aka gen5

winter orbit Jan 7, 2025, 11:55 AM

#

Ohh

fiery turtle Jan 7, 2025, 11:55 AM

#

i even patched the mesa disassembler while working on this

#

upstreamed some stuff

#

a fix for the igpu page table dump tool as well

#

fun times

#

ill get back to this after uacpi 1.0

winter orbit Jan 7, 2025, 11:57 AM

#

I really need to get back to this too but :

calm latch Jan 7, 2025, 11:57 AM

#

fiery turtle ill get back to this after uacpi 1.0

As uDRM?

winter orbit Jan 7, 2025, 11:57 AM

#

This is a DX 10 surface functioning on a Mali g610

#

In arm64 windows 11

fiery turtle Jan 7, 2025, 11:57 AM

#

calm latch As uDRM?

nah, my own thing

fiery turtle Jan 7, 2025, 11:57 AM

#

winter orbit This is a DX 10 surface functioning on a Mali g610

your own driver?

winter orbit Jan 7, 2025, 11:57 AM

#

Yup

fiery turtle Jan 7, 2025, 11:57 AM

#

damn

#

how did u reverse it?

winter orbit Jan 7, 2025, 11:58 AM

#

I didn’t really have to

Pan frost/panthor did a lot of
The reversing work

fiery turtle Jan 7, 2025, 11:58 AM

#

ah

winter orbit Jan 7, 2025, 11:58 AM

#

all I had to do was take their driver and mesa implementation and port some stuff over to windows

#

This is actually the Mali mesa driver working on windows

#

As a dx10 UMD

fiery turtle Jan 7, 2025, 11:59 AM

#

calm latch As uDRM?

I've had this way before uDRM

#

although its private now lol

#

until its anything interesting

fiery turtle Jan 7, 2025, 12:00 PM

#

winter orbit all I had to do was take their driver and mesa implementation and port some stuf...

ah cool

winter orbit Jan 7, 2025, 12:00 PM

#

LHobbies OSdev projects can’t have gpu drivers”

Lies trl

fiery turtle Jan 7, 2025, 12:00 PM

#

Lol we'll see

#

i think my strength is i stick to one project even if i get tired of it

#

like ive been working on uacpi for two years

winter orbit Jan 7, 2025, 12:00 PM

#

Poor haiku

fiery turtle Jan 7, 2025, 12:00 PM

#

without distractions

#

even though i did want to work on other stuff

winter orbit Jan 7, 2025, 12:01 PM

#

you’ve almost got a completely usable product too

fiery turtle Jan 7, 2025, 12:01 PM

#

basically yeah

#

at this point i can vouch for its stability

#

its been tested on so much hw thanks to this server

fiery turtle Jan 7, 2025, 12:02 PM

#

winter orbit Poor haiku

iirc they port linux drivers or something like that?

winter orbit Jan 7, 2025, 12:02 PM

#

That was a proposed plan

fiery turtle Jan 7, 2025, 12:02 PM

#

L

winter orbit Jan 7, 2025, 12:03 PM

#

is haiku considered hobbiest

#

I think so

fiery turtle Jan 7, 2025, 12:03 PM

#

nah ofc not

#

since it has paid developers no?

winter orbit Jan 7, 2025, 12:03 PM

#

what

#

it has contracts

#

Okay well if that is the line then ROS isn’t hobbiest either

#

KEKW

fiery turtle Jan 7, 2025, 12:03 PM

#

yeah

#

i dont think ros is a hobby os lol

#

maybe our definitions of hobby are different

winter orbit Jan 7, 2025, 12:04 PM

#

It’s hard for me to consider it one when it’s rapidly getting more attention and funding

fiery turtle Jan 7, 2025, 12:04 PM

#

ROS?

winter orbit Jan 7, 2025, 12:04 PM

#

and haiku

#

Both of them

fiery turtle Jan 7, 2025, 12:05 PM

#

haiku is getting more attention?

#

thats a new one for me

winter orbit Jan 7, 2025, 12:05 PM

#

Both ROS and haiku are getting exponentiall more people using it

fiery turtle Jan 7, 2025, 12:05 PM

#

Why is that btw?

#

ROS i understand since its doing cool things

winter orbit Jan 7, 2025, 12:05 PM

#

Haiku is pretty solid these days

#

Firefox with discord

#

And some usb audio

#

And I use it in my framework 13

fiery turtle Jan 7, 2025, 12:06 PM

#

ig

#

i just dont find it super interesting because i know nothing about beos

winter orbit Jan 7, 2025, 12:06 PM

#

I guess it’s also because of people like action retro making more videos about its improvements

#

there’s not a lot of beos left in haiku

fiery turtle Jan 7, 2025, 12:07 PM

#

MJD reactos videos 💀

winter orbit Jan 7, 2025, 12:07 PM

#

hey his last one he talked with me a ton first

#

In his defense

fiery turtle Jan 7, 2025, 12:07 PM

#

yeah good thing u contacted him

winter orbit Jan 7, 2025, 12:07 PM

#

He spent nearly 36 hours with me

wind fiber Jan 7, 2025, 12:08 PM

#

For me the distinction of hobby is that you can’t have (almost) full time paid developers, paid by the project. If you manage to hire that at least once you’re basically out for me (of the hobby sphere, project is still awesome!)

winter orbit Jan 7, 2025, 12:08 PM

#

okay ROS isn’t hobby then and should get its shit together

#

KEKW KEKW KEKW

wind fiber Jan 7, 2025, 12:08 PM

#

meme

fiery turtle Jan 7, 2025, 12:08 PM

#

winter orbit okay ROS isn’t hobby then and should get its shit together

fix those damn Fast mutexes

winter orbit Jan 7, 2025, 12:08 PM

#

“Fast”

wind fiber Jan 7, 2025, 12:09 PM

#

That puts me in a predicament tho, I’d love to be paid by Managarm GmbH but then it’s no longer hobby death

winter orbit Jan 7, 2025, 12:09 PM

#

Slowtexes

fiery turtle Jan 7, 2025, 12:09 PM

#

monkuous would vomit if you tell them your ops/s

winter orbit Jan 7, 2025, 12:09 PM

#

It’s pretty funny

wind fiber Jan 7, 2025, 12:09 PM

#

wind fiber That puts me in a predicament tho, I’d love to be paid by Managarm GmbH but then...

So I pay Managarm now meme

fiery turtle Jan 7, 2025, 12:09 PM

#

winter orbit Slowtexes

lmao

winter orbit Jan 7, 2025, 12:10 PM

#

tbh though I think more OSes paying for devs is cool

#

The top few have dominated too long

fiery turtle Jan 7, 2025, 12:10 PM

#

its not like a bad thing to not be a hobby os lol

#

its the opposite

winter orbit Jan 7, 2025, 12:10 PM

#

For sure

wind fiber Jan 7, 2025, 12:10 PM

#

Oh I agree

#

The distinction is only useful for comparison work imho

fiery turtle Jan 7, 2025, 12:13 PM

#

Any C++ experts? Is this necessary to get a zeroed buffer?

calm latch Jan 7, 2025, 12:17 PM

#

http://jkhvatov.blogspot.com/2009/11/c-what-is-difference-between-new-t-from.html ?

C++: what is difference between "new T" from "new T()"

Recently I received an interesting question: what difference between " new T " from " new T() " in C++? Let's look at example: #includ...

fiery turtle Jan 7, 2025, 12:18 PM

#

looking at this otherwise, the object is zero-initialized.

#

so i guess not?

calm latch Jan 7, 2025, 12:19 PM

#

If the new-initializer is omitted:

If T is a (possibly cv-qualified) non-POD class type (or array thereof), the object is default-initialized (8.5). If T is a const-qualified type, the underlying class type shall have a user-declared default constructor.

Otherwise, the object created has indeterminate value. If T is a const-qualified type, or a (possibly cv-qualified) POD class type (or array thereof) containing (directly or indirectly) a member of const-qualified type, the program is ill-formed;

fiery turtle Jan 7, 2025, 12:19 PM

#

FlushyBaka

calm latch Jan 7, 2025, 12:20 PM

#

POD -> plain old data, e.g. int

fiery turtle Jan 7, 2025, 12:20 PM

#

yeah i know what a pod is

#

i spent like 3 years writing c++

calm latch Jan 7, 2025, 12:21 PM

#

fiery turtle Any C++ experts? Is this necessary to get a zeroed buffer?

So like you do need that I think

fiery turtle Jan 7, 2025, 12:21 PM

#

back when SFINAE was cool

fiery turtle Jan 7, 2025, 12:21 PM

#

calm latch So like you do need that I think

ah cool

calm latch Jan 7, 2025, 12:22 PM

#

Why is it not cool now? ultrameme

fiery turtle Jan 7, 2025, 12:22 PM

#

concepts

calm latch Jan 7, 2025, 12:22 PM

#

Tbh I've spent 4 years so idk if I'm much of expert

fiery turtle Jan 7, 2025, 12:22 PM

#

depends on what u did

#

i wrote like heavily templated shit all the time

calm latch Jan 7, 2025, 12:23 PM

#

calm latch http://jkhvatov.blogspot.com/2009/11/c-what-is-difference-between-new-t-from.htm...

https://en.cppreference.com/w/cpp/language/value_initialization ?

cpp/language/value initialization

#

Idk I wrote pmOS as my flagship hobby project

fiery turtle Jan 7, 2025, 12:31 PM

#

I wonder what the fuck they meant by this

calm latch Jan 7, 2025, 12:38 PM

#

Like ACPI being able to read and write MSRs?

fiery turtle Jan 7, 2025, 12:39 PM

#

#

yeah i found some intel docs

#

and riscv https://github.com/riscv-non-isa/riscv-acpi-ffh

GitHub

GitHub - riscv-non-isa/riscv-acpi-ffh: The repo will be used to hol...

The repo will be used to hold the draft non-ISA RISC-V ACPI Functional Fixed Hardware (FFH) specification - riscv-non-isa/riscv-acpi-ffh

calm latch Jan 7, 2025, 12:39 PM

#

I'll implement that in pmOS

#

Are there other operating systems, using uACPI and working on RISC-V?

fiery turtle Jan 7, 2025, 12:40 PM

#

keyronex

#

although idk the state of keyronex-on-riscv

calm latch Jan 7, 2025, 12:40 PM

#

My implementation is (should be) on par with x86

#

The only issue is that I don't have any physical hardware

fiery turtle Jan 7, 2025, 12:41 PM

#

physical riscv hw is mostly trash

calm latch Jan 7, 2025, 12:41 PM

#

But QEMU is boring

#

trash is fun sometimes

winter orbit Jan 7, 2025, 12:41 PM

#

does your riscv build boot on uboot

#

KEKW

fiery turtle Jan 7, 2025, 12:42 PM

#

well theres riscv framework i think?

calm latch Jan 7, 2025, 12:42 PM

#

uboot and ovmf

fiery turtle Jan 7, 2025, 12:42 PM

#

idk if it does acpi

calm latch Jan 7, 2025, 12:42 PM

#

through UEFI

#

But uboot doesn't (didn't last time I checked) give you ACPI

winter orbit Jan 7, 2025, 12:42 PM

#

I don’t think there’s a full edk2 port for this yet

calm latch Jan 7, 2025, 12:42 PM

#

?

winter orbit Jan 7, 2025, 12:43 PM

#

https://milkv.io/jupiter

Milk-V

calm latch Jan 7, 2025, 12:43 PM

#

It works ok on QEMU

winter orbit Jan 7, 2025, 12:43 PM

#

Is what I have on my desk

calm latch Jan 7, 2025, 12:43 PM

#

calm latch Jan 7, 2025, 12:43 PM

#

winter orbit Is what I have on my desk

You could try pmOS ultrameme

winter orbit Jan 7, 2025, 12:43 PM

#

It has working GPU in Linux though with AMD GPUs

calm latch Jan 7, 2025, 12:43 PM

#

I haven never ran it on physical hardware

winter orbit Jan 7, 2025, 12:44 PM

#

Yeah I want more things to try

#

Would be cool if it had an edk2 port though… let me see

calm latch Jan 7, 2025, 12:44 PM

#

efi uboot was working in qemu

winter orbit Jan 7, 2025, 12:45 PM

#

EFI uboot works on this I think

#

I haven’t gotten haiku to boot

#

Only ROS bootloader

calm latch Jan 7, 2025, 12:45 PM

#

Oh cool it's not that expensive

winter orbit Jan 7, 2025, 12:46 PM

#

No it’s a great board too

calm latch Jan 7, 2025, 12:46 PM

#

There was another board with Spacemit M1 though

#

But it says it's sold out everywhere

#

Do you have uart?

slim panther Jan 7, 2025, 12:46 PM

#

fiery turtle idk if it does acpi

it does not

calm latch Jan 7, 2025, 12:47 PM

#

calm latch Do you have uart?

(if you're willing to try)

slim panther Jan 7, 2025, 12:47 PM

#

fiery turtle well theres riscv framework i think?

that's just a visionfive2 in laptop format

winter orbit Jan 7, 2025, 12:47 PM

#

calm latch (if you're willing to try)

yup

#

And it has UART

calm latch Jan 7, 2025, 12:47 PM

#

The kernel logs to it

winter orbit Jan 7, 2025, 12:48 PM

#

Let’s see what chaos we can unleash

#

Gimme a moment

calm latch Jan 7, 2025, 12:48 PM

#

I'm using limine

fiery turtle Jan 7, 2025, 12:48 PM

#

slim panther it does not

largest L ive seen

slim panther Jan 7, 2025, 12:48 PM

#

why should it

fiery turtle Jan 7, 2025, 12:48 PM

#

windows

calm latch Jan 7, 2025, 12:48 PM

#

Wait a sec I need to recompile everything

hasty plinth Jan 7, 2025, 12:48 PM

#

fiery turtle well theres riscv framework i think?

Still WIP and afaik it's just a vf2 in disguise

slim panther Jan 7, 2025, 12:48 PM

#

there is no windows

winter orbit Jan 7, 2025, 12:48 PM

#

Well

calm latch Jan 7, 2025, 12:48 PM

#

slim panther why should it

There was a spec mandating it

fiery turtle Jan 7, 2025, 12:48 PM

#

hence the L

winter orbit Jan 7, 2025, 12:48 PM

#

There is windows

slim panther Jan 7, 2025, 12:48 PM

#

these are basically exclusively embedded cpus

calm latch Jan 7, 2025, 12:48 PM

#

Which got archived

slim panther Jan 7, 2025, 12:49 PM

#

winter orbit There is windows

?

hasty plinth Jan 7, 2025, 12:49 PM

#

Ah lol, I remember that spec

winter orbit Jan 7, 2025, 12:49 PM

#

Sorry to burst your bubble

#

KEKW

#

It’s just not public

fiery turtle Jan 7, 2025, 12:49 PM

#

unarchive the spec and mandate acpi for all riscv chips 😡

#

otherwise the impl is considered invalid

slim panther Jan 7, 2025, 12:49 PM

#

winter orbit It’s just not public

source?

fiery turtle Jan 7, 2025, 12:49 PM

#

NDA

slim panther Jan 7, 2025, 12:50 PM

#

they should have something like systemsready

#

something that mandates uefi

#

fuck device tree fw config

winter orbit Jan 7, 2025, 12:50 PM

#

slim panther source?

I’ve just seen some signs of it really nothing officially

calm latch Jan 7, 2025, 12:51 PM

#

slim panther they should have something like systemsready

there was something like it

winter orbit Jan 7, 2025, 12:51 PM

#

I don’t think risc-v has a future if it keeps heading down the path it is

slim panther Jan 7, 2025, 12:51 PM

#

yea

#

i would love for it to succeed but wtf are they doing

calm latch Jan 7, 2025, 12:51 PM

#

MIPS routers are still being made

slim panther Jan 7, 2025, 12:51 PM

#

they're fucking up everything

fiery turtle Jan 7, 2025, 12:51 PM

#

winter orbit I don’t think risc-v has a future if it keeps heading down the path it is

yes

#

did u see the format they wanted to use for the cpuid thing

slim panther Jan 7, 2025, 12:52 PM

#

💀 yea

fiery turtle Jan 7, 2025, 12:52 PM

#

idk which ass they pulled that out of

slim panther Jan 7, 2025, 12:52 PM

#

asn.1

loud ice Jan 7, 2025, 12:52 PM

#

fiery turtle Any C++ experts? Is this necessary to get a zeroed buffer?

yes

hasty plinth Jan 7, 2025, 12:52 PM

#

fiery turtle did u see the format they wanted to use for the cpuid thing

The cpuid thing that was only meant for m mode

slim panther Jan 7, 2025, 12:53 PM

#

that's even more stupid

hasty plinth Jan 7, 2025, 12:53 PM

#

I.e totally useless

#

Because firmware already knows that info

slim panther Jan 7, 2025, 12:53 PM

#

what's so hard in making a user-RO CSR

#

imagine you have a highly flexible and extensible ISA, only to not have a way to detect which extensions are actually available

#

like

#

who the fuck thought that would be a good idea

loud ice Jan 7, 2025, 12:54 PM

#

slim panther imagine you have a highly flexible and extensible ISA, only to not have a way to...

yeah lmao

#

dont worry they are planning to put some shit ass asn.1 in there

#

to make it so that you can detect them (in kernel mode of course)

slim panther Jan 7, 2025, 12:54 PM

#

especially not even the fucking supervisor has access to that info

winter orbit Jan 7, 2025, 12:54 PM

#

RIP risc-v

#

Then

slim panther Jan 7, 2025, 12:54 PM

#

how the hell is it supposed to know that

loud ice Jan 7, 2025, 12:54 PM

#

slim panther especially not even the fucking supervisor has access to that info

you have the DTB

hasty plinth Jan 7, 2025, 12:54 PM

#

Well they did (misa), except that's m mode only and only supports ~20 extensions

slim panther Jan 7, 2025, 12:54 PM

#

the dtb doesn't count

loud ice Jan 7, 2025, 12:55 PM

#

thats basically complete

#

right?

hasty plinth Jan 7, 2025, 12:55 PM

#

Yeah but how do you get the dtb

loud ice Jan 7, 2025, 12:55 PM

#

that is definitely how you do good device discovery

hasty plinth Jan 7, 2025, 12:55 PM

#

Its unspecified

loud ice Jan 7, 2025, 12:55 PM

#

hasty plinth Yeah but how do you get the dtb

m mode gives you

slim panther Jan 7, 2025, 12:55 PM

#

hasty plinth Yeah but how do you get the dtb

bootloader meme

loud ice Jan 7, 2025, 12:55 PM

#

im not saying its a good solution btw

hasty plinth Jan 7, 2025, 12:55 PM

#

Show me the spec meme

#

SBI is optional

loud ice Jan 7, 2025, 12:56 PM

#

hasty plinth Show me the spec <:meme:575445094589661215>

https://docs.kernel.org/arch/riscv/boot.html "$a1 to contain the address of the devicetree in memory."

slim panther Jan 7, 2025, 12:56 PM

#

the kernel expects

#

that's not a risc-v spec

hasty plinth Jan 7, 2025, 12:56 PM

#

Lmao

loud ice Jan 7, 2025, 12:56 PM

#

its not

#

it is however

#

the spec

hasty plinth Jan 7, 2025, 12:56 PM

#

I guess that's as good as we get

loud ice Jan 7, 2025, 12:56 PM

#

the only one, as far as i know

#

lmao

#

it also is a spec

slim panther Jan 7, 2025, 12:56 PM

#

i fucking hate that so much risc-v stuff is 100% linux-centric

loud ice Jan 7, 2025, 12:56 PM

#

it specifies the requirements from the kernel on the firmware

slim panther Jan 7, 2025, 12:56 PM

#

like

#

what a fucking good idea to focus everything around a single kernel

loud ice Jan 7, 2025, 12:57 PM

#

uefi specifies this shit too in a way that is 10 times more convoluted btw

slim panther Jan 7, 2025, 12:57 PM

#

i'd rather have a uefi spec than linux

loud ice Jan 7, 2025, 12:57 PM

#

tbh the linux boot proto on arm and riscv is like

#

actually not that bad

#

its a couple registers

slim panther Jan 7, 2025, 12:57 PM

#

multiboot3

loud ice Jan 7, 2025, 12:57 PM

#

if you want to read the ISA string from userspace, you just open("/proc/device-tree/cpus/cpu@0/riscv,isa", O_RDONLY)

#

problem solved

#

no flaws

#

do not think about this btw

#

or if you want to find out what kind of cpu that is you have /proc/device-tree/cpus/cpu@0/compatible

#

so for example my cpu is "apple,icestorm"

#

hmm i see one flaw with this approach meme

hasty plinth Jan 7, 2025, 12:58 PM

#

Seems portable enough meme

loud ice Jan 7, 2025, 12:59 PM

#

portable
riscv

#

lmfao

hasty plinth Jan 7, 2025, 12:59 PM

#

Lol

loud ice Jan 7, 2025, 12:59 PM

#

theres a lot of stuff under /proc/device-tree that is super useful too!

#

like /proc/device-tree/model

slim panther Jan 7, 2025, 12:59 PM

#

loud ice if you want to read the ISA string from userspace, you just `open("/proc/device-...

fucking stupid

#

i hate device trees

calm latch Jan 7, 2025, 12:59 PM

#

pmOS RISC-V build ultrameme

📎 pmOS.iso

loud ice Jan 7, 2025, 12:59 PM

#

slim panther fucking stupid

i dont think that is the official guidance btw

slim panther Jan 7, 2025, 12:59 PM

#

i don't want to use them

loud ice Jan 7, 2025, 12:59 PM

#

slim panther i hate device trees

FDTs?

#

or all of them

calm latch Jan 7, 2025, 12:59 PM

#

calm latch pmOS RISC-V build <:ultrameme:729026898666979428>

For some reason it still calls exit after fork but whatever

slim panther Jan 7, 2025, 12:59 PM

#

loud ice FDTs?

fdts

#

OF dts

loud ice Jan 7, 2025, 12:59 PM

#

acpi is also a device tree

loud ice Jan 7, 2025, 1:00 PM

#

slim panther OF dts

OF dts are different and good

#

OF dts have AML except more sane

slim panther Jan 7, 2025, 1:00 PM

#

i have uacpi for that

loud ice Jan 7, 2025, 1:00 PM

#

and are native endian too

loud ice Jan 7, 2025, 1:00 PM

#

slim panther i have uacpi for that

they also have names that arent four characters

#

god i hate acpi names so much

slim panther Jan 7, 2025, 1:00 PM

#

loud ice and are native endian too

i thought fdts are big endian

loud ice Jan 7, 2025, 1:00 PM

#

slim panther i thought fdts are big endian

i said native endian

#

i never specified the platform endianness

slim panther Jan 7, 2025, 1:01 PM

#

what does that mean

winter orbit Jan 7, 2025, 1:01 PM

#

calm latch pmOS RISC-V build <:ultrameme:729026898666979428>

Here let’s move this to your channel

slim panther Jan 7, 2025, 1:01 PM

#

fdts are always big endian

#

smh

loud ice Jan 7, 2025, 1:01 PM

#

they are native endian

#

OF dts are a PowerPC thing

#

powerpc is big endian

#

FDTs arent OF dts

#

OF dts are basically ACPI/AML except with names that arent four or eight letters

slim panther Jan 7, 2025, 1:03 PM

#

loud ice FDTs arent OF dts

then whatever the device tree spec defines

#

those are big-endian

fiery turtle Jan 7, 2025, 1:03 PM

#

loud ice OF dts are basically ACPI/AML except with names that arent four or eight letters

aml doesnt have 8 letter names meme

loud ice Jan 7, 2025, 1:04 PM

#

fiery turtle aml doesnt have 8 letter names <:meme:575445094589661215>

PNP IDs are

loud ice Jan 7, 2025, 1:04 PM

#

slim panther then whatever the device tree spec defines

FDTs

fiery turtle Jan 7, 2025, 1:04 PM

#

well

loud ice Jan 7, 2025, 1:04 PM

#

and yes

#

and they are rather stupid in that way

#

but its also kinda not that big of an issue

fiery turtle Jan 7, 2025, 1:04 PM

#

idk why they did such stupid pnp id names

#

technically _HID can store any arbitrary string

loud ice Jan 7, 2025, 1:05 PM

#

wait is there constraint

#

what the fuck

fiery turtle Jan 7, 2025, 1:05 PM

#

which some devices do

loud ice Jan 7, 2025, 1:05 PM

#

fuck acpi

#

this shit is stupid

#

my devices are gonna have a PNP string that is pitust,...

#

and all the properties in the linux fwnode format

fiery turtle Jan 7, 2025, 1:05 PM

#

small pnp ids can be encoded in a dword

loud ice Jan 7, 2025, 1:06 PM

#

loud ice and all the properties in the linux fwnode format

not that i have any i guess

loud ice Jan 7, 2025, 1:06 PM

#

fiery turtle small pnp ids can be encoded in a dword

memory is free

fiery turtle Jan 7, 2025, 1:06 PM

#

maybe thats why they did it this way

loud ice Jan 7, 2025, 1:06 PM

#

none of this is a concern

fiery turtle Jan 7, 2025, 1:06 PM

#

Theres an EisaId macro in iasl

#

to do that

loud ice Jan 7, 2025, 1:06 PM

#

memory is literally free

#

or close enough

fiery turtle Jan 7, 2025, 1:06 PM

#

yes

loud ice Jan 7, 2025, 1:06 PM

#

this is like 20 bytes of overhead

#

max

#

this is why i dont like acpi

#

not that FDTs are better

fiery turtle Jan 7, 2025, 1:06 PM

#

if _HID was just normal names it would be way better i agree

loud ice Jan 7, 2025, 1:07 PM

#

tbh the real issue with FDTs is that they are not a thing on all platforms, and they don't have code so have fun doing your own drivers

fiery turtle Jan 7, 2025, 1:07 PM

#

basically yeah

loud ice Jan 7, 2025, 1:07 PM

#

fiery turtle if _HID was just normal names it would be way better i agree

the name "_HID" is in itself the ACPI name brain damage

fiery turtle Jan 7, 2025, 1:07 PM

#

if fdts had some sort of a sane bytecode to do the bare minimum platform config and stuff

loud ice Jan 7, 2025, 1:07 PM

#

yeah

fiery turtle Jan 7, 2025, 1:07 PM

#

but they dont care because linux is the only kernel

#

and it has every driver

loud ice Jan 7, 2025, 1:08 PM

#

honestly that is not a bad approach

fiery turtle Jan 7, 2025, 1:08 PM

#

i kinda hate it

loud ice Jan 7, 2025, 1:08 PM

#

linux is the only (important) kernel

#

i know it sucks for everyone except linux

fiery turtle Jan 7, 2025, 1:08 PM

#

ms is better than linux because they dont consider themselves the only os

loud ice Jan 7, 2025, 1:08 PM

#

but uh

loud ice Jan 7, 2025, 1:08 PM

#

fiery turtle ms is better than linux because they dont consider themselves the only os

acpi.sys is so full of bugs they may as well

fiery turtle Jan 7, 2025, 1:09 PM

#

fault of whoever implemented that crap meme

loud ice Jan 7, 2025, 1:09 PM

#

yeah

#

oh right its the firmware people

#

which did not care about other people existing

fiery turtle Jan 7, 2025, 1:14 PM

#

i see a pattern

#

meme

calm latch Jan 7, 2025, 1:15 PM

#

risc-v needs microsoft

vast kestrel Jan 7, 2025, 3:49 PM

#

riscv needs Jesus

flat badge Jan 7, 2025, 4:04 PM

#

fiery turtle if fdts had some sort of a sane bytecode to do the bare minimum platform config ...

FDTs just fill a different design space compared to ACPI

#

FDTs are useful for small socs where you want to run a customized kernel on

#

No bytecode is a feature in this regard

calm latch Jan 7, 2025, 4:05 PM

#

The issue with that is that it's very Linux specific

flat badge Jan 7, 2025, 4:06 PM

#

How are they Linux specific?

vast kestrel Jan 7, 2025, 4:07 PM

#

I seen drivers breaking the meaning of fields (or just changing the content in general without replacing the compatibility) without any regard to others, so if you are not Linux and use a dtb, your driver might just break one day

#

Or even if you are a Linux of a different version

calm latch Jan 7, 2025, 4:08 PM

#

And when the "small" Arm/RISC-V SoCs appear in PC class computers or servers

flat badge Jan 7, 2025, 4:08 PM

#

That sounds unlikely given the backwards compat philosophy of Linux

#

Also, you're bundling the DT with the kernel anyway, so that's a non issue

flat badge Jan 7, 2025, 4:09 PM

#

calm latch And when the "small" Arm/RISC-V SoCs appear in PC class computers or servers

Arm servers use ACPI

calm latch Jan 7, 2025, 4:09 PM

#

flat badge How are they Linux specific?

Somehow when implementing FDT support in my OS, the only way to understand the PLIC controller was to look at the Linux source, as the spec was clearly written after Linux

calm latch Jan 7, 2025, 4:10 PM

#

flat badge Arm servers use ACPI

Yes, because of the server ready or whatever initiative

flat badge Jan 7, 2025, 4:10 PM

#

Well that's true

calm latch Jan 7, 2025, 4:10 PM

#

A lot of smaller boards, like raspberry pi do not use ACPI

flat badge Jan 7, 2025, 4:10 PM

#

FDTs are a linux invention

calm latch Jan 7, 2025, 4:10 PM

#

calm latch Yes, because of the server ready or whatever initiative

And because they have to support other operating systems

flat badge Jan 7, 2025, 4:11 PM

#

But the data in them is pretty much OS agnostic

calm latch Jan 7, 2025, 4:11 PM

#

But very little is standardized, besides how the data is stored

#

Like a field could mean anything and it depends on driver

flat badge Jan 7, 2025, 4:12 PM

#

calm latch Yes, because of the server ready or whatever initiative

I don't think it's about pressure from ARM, but acpi just makes more sense for servers that need to run a large variety of kernels

calm latch Jan 7, 2025, 4:12 PM

#

And the definition a lot of times is "what Linux does"

calm latch Jan 7, 2025, 4:12 PM

#

flat badge I don't think it's about pressure from ARM, but acpi just makes more sense for s...

Yeah, it's the demands of potential enterprise clients I'd suppose

flat badge Jan 7, 2025, 4:12 PM

#

I doubt that e.g. Amazon uses acpi because server ready exists, it's rather just the best option for their customers

calm latch Jan 7, 2025, 4:13 PM

#

Yeah

flat badge Jan 7, 2025, 4:13 PM

#

And AWS servers probably account for the vast majority of all ARM servers in use

#

People want as little friction as possible when transferring their workloads from x86 servers to ARM servers

calm latch Jan 7, 2025, 4:14 PM

#

But if you buy a random Arm SBC/computer, a lot of times Linux is the only thing it supports and it does not use ACPI

calm latch Jan 7, 2025, 4:14 PM

#

calm latch But if you buy a random Arm SBC/computer, a lot of times Linux is the only thing...

Whereas unless you're talking about something like game consoles, all x86 have ACPI

flat badge Jan 7, 2025, 4:14 PM

#

But that's not true for, say, Android phones where the vendor just doesn't care about compat with a wide range of kernels

calm latch Jan 7, 2025, 4:15 PM

#

Like it's fine when a router or something uses FDT, but it bleeds over to things which have removable storage and allow you to change the operating system

flat badge Jan 7, 2025, 4:16 PM

#

There's no point in shipping a firmware that supports arbitrary kernels because the vendor supplied kernel simply compiles in exactly the drivers that are needed for that particular device

jaunty fox Jan 7, 2025, 6:01 PM

#

flat badge FDTs are a linux invention

Apple had them before in an alternative form

#

And they ultimately are a bastardised subset of an open firmware

deft canopy Jan 7, 2025, 6:32 PM

#

jaunty fox Apple had them before in an alternative form

Did apple provide the full openfirmware experience?

#

Or are you saying apple made the initial bastardized subset?

mortal yoke Jan 7, 2025, 8:48 PM

#

@fiery turtle letsgo now the two remaining ones are the osi one (because of it being a hardcoded aml blob) and the IO from deleted fields doesn't crash test because it doesn't crash (which I think is because of the way I reference op regions within fields)

fiery turtle Jan 7, 2025, 8:56 PM

#

mortal yoke <@353593948523986945> <:letsgo:1282155513156468747> now the two remaining ones a...

LETSFUCKINGGOOOOOO

#

feel free to add osi detection for qacpi or make it more agnostic idk

fiery turtle Jan 7, 2025, 8:59 PM

#

mortal yoke <@353593948523986945> <:letsgo:1282155513156468747> now the two remaining ones a...

which ones are expected btw

mortal yoke Jan 7, 2025, 9:30 PM

#

fiery turtle which ones are expected btw

its actually the expected fail count and its there to make it return 0 for ci purposes with that many failed tests

fiery turtle Jan 7, 2025, 9:31 PM

#

mortal yoke its actually the expected fail count and its there to make it return 0 for ci pu...

ah

winter orbit Jan 7, 2025, 9:48 PM

#

i am having a moment

#

okay so trying to get just uACPI built alone so we can have something to start with as a drafy

#

github actions are happy 👍

#

Then just move to my laptop to test something

#

seemingly angry with this?? like 1000+ spam of this same error but in different spots

#

Mind you i can push from this laptop, go to my desktop or github actions and it builds fine lol???

#

worst doing this fixes it

#define UACPI_BUILD_BUG_ON_WITH_MSG(expr, msg) //UACPI_STATIC_ASSERT((!expr), msg)

#

Like what???

loud ice Jan 7, 2025, 10:14 PM

#

i guess msvc is a shitty compiler?

loud ice Jan 7, 2025, 10:14 PM

#

winter orbit worst doing this fixes it ```c #define UACPI_BUILD_BUG_ON_WITH_MSG(expr, msg) //...

also btw that macro is broken

gentle peak Jan 7, 2025, 10:15 PM

#

yeah the excl should be before the 2nd paren

loud ice Jan 7, 2025, 10:15 PM

#

my guess is that your version of msvc just doesnt have static assert

winter orbit Jan 7, 2025, 10:15 PM

#

oh this is just in uacpi repo

#

this is vs2022

#

so id assume it would

loud ice Jan 7, 2025, 10:15 PM

#

lol

#

the msvc people are the firmware devs of the compiler world

winter orbit Jan 7, 2025, 10:16 PM

#

loud ice also btw that macro is broken

my first pr to uacpi time /s?

#

hmm no static assert works fine

#

#

oof

fiery turtle Jan 7, 2025, 10:28 PM

#

loud ice also btw that macro is broken

wdym

hollow elm Jan 7, 2025, 10:28 PM

#

i guess they dont static assert with messages

#

static assert with message is a C23 feature

winter orbit Jan 7, 2025, 10:28 PM

#

Microsoft 👍

loud ice Jan 7, 2025, 10:29 PM

#

you need to do UACPI_STATIC_ASSERT(!(expr)) and not UACPI_STATIC_ASSERT((!expr))

fiery turtle Jan 7, 2025, 10:29 PM

#

hollow elm static assert with message is a C23 feature

💀

loud ice Jan 7, 2025, 10:29 PM

#

fiery turtle 💀

lol yeah

#

its a C moment

winter orbit Jan 7, 2025, 10:29 PM

#

welp

#

it is what it is

#

trollface

frank canopy Jan 7, 2025, 10:30 PM

#

there might be a specific define for the c23 thing you can use idk

fiery turtle Jan 7, 2025, 10:30 PM

#

loud ice you need to do `UACPI_STATIC_ASSERT(!(expr))` and not `UACPI_STATIC_ASSERT((!exp...

it is in uacpi's repo #define UACPI_BUILD_BUG_ON_WITH_MSG(expr, msg) UACPI_STATIC_ASSERT(!(expr), msg)

hollow elm Jan 7, 2025, 10:30 PM

#

ig it makes sense that vs2022 is missing a feature from a standard that was officially published in 2024

loud ice Jan 7, 2025, 10:30 PM

#

fiery turtle it is in uacpi's repo `#define UACPI_BUILD_BUG_ON_WITH_MSG(expr, msg) UACPI_STAT...

yeah that makes sense

loud ice Jan 7, 2025, 10:30 PM

#

hollow elm ig it makes sense that vs2022 is missing a feature from a standard that was offi...

fair

hollow elm Jan 7, 2025, 10:30 PM

#

solution: install vs2024

loud ice Jan 7, 2025, 10:30 PM

#

although

hollow elm Jan 7, 2025, 10:30 PM

#

meme

loud ice Jan 7, 2025, 10:30 PM

#

its a C++ thing for a very long time

hollow elm Jan 7, 2025, 10:30 PM

#

true

loud ice Jan 7, 2025, 10:30 PM

#

and its an obvious extension

hollow elm Jan 7, 2025, 10:30 PM

#

but it's msvc

winter orbit Jan 7, 2025, 10:31 PM

#

fiery turtle it is in uacpi's repo `#define UACPI_BUILD_BUG_ON_WITH_MSG(expr, msg) UACPI_STAT...

can't believe uacpi requires C23

fiery turtle Jan 7, 2025, 10:31 PM

#

@winter orbit do u want to pr msvc compat for that header or should i fix it

#

trl

winter orbit Jan 7, 2025, 10:31 PM

#

it's fine i'd leave it as is

#

the _STATIC_ASSERT im using here is NT_ASSERT wrapped

fiery turtle Jan 7, 2025, 10:32 PM

#

winter orbit my first pr to uacpi time /s?

but its different in uacpis repo lol, did u move the ! inside parens yourself?

gentle peak Jan 7, 2025, 10:32 PM

#

hollow elm static assert with message is a C23 feature

no? static assert with messages has been in there from c11, aka from the same moment as static assert itself

#

the message was required until c23 actually

winter orbit Jan 7, 2025, 10:32 PM

#

oh no i didn't

fiery turtle Jan 7, 2025, 10:32 PM

#

winter orbit it's fine i'd leave it as is

it should be in compiler.h anyway

#

because helpers.h is not overridable

fiery turtle Jan 7, 2025, 10:33 PM

#

gentle peak no? static assert with messages has been in there from c11, aka from the same mo...

oh rly?

#

is msvc broken then or what

gentle peak Jan 7, 2025, 10:34 PM

#

maybe static assert without messages was a C++ thing from before C's static asserts? that's the only thing i can think of

loud ice Jan 7, 2025, 10:34 PM

#

doubtful

#

i bet its just average MSVC jank

fiery turtle Jan 7, 2025, 10:34 PM

#

this is like semi latest msvc

loud ice Jan 7, 2025, 10:34 PM

#

yeah but its msvc

fiery turtle Jan 7, 2025, 10:34 PM

#

i think?

winter orbit Jan 7, 2025, 10:34 PM

#

this is latest vs2022

hollow elm Jan 7, 2025, 10:35 PM

#

gentle peak no? static assert with messages has been in there from c11, aka from the same mo...

oh yeah you're right, my bad

#

i completely misread the cppref page

fiery turtle Jan 7, 2025, 10:35 PM

#

anyway since it's possible to have a broken static assert ill move UACPI_STATIC_ASSERT to compiler.h and make it overridable

fiery turtle Jan 7, 2025, 10:37 PM

#

winter orbit this is latest vs2022

is it possible you're compiling that header in c++ mode?

winter orbit Jan 7, 2025, 10:37 PM

#

yup

#

wait

#

nope

fiery turtle Jan 7, 2025, 10:38 PM

#

it's part of the public headers so if u compile anything in ros as c++ it might be part of that

fiery turtle Jan 8, 2025, 1:31 AM

#

Yay

#

With this, uACPI should be at pretty much full feature parity with ACPICA barring misc things

#

Now ill do some extensive testing and reviewing and hopefully merge this in a few days

#

its 19 commits atm

#

GenericSerialBus was the final boss

#

since it combines all annoying things

#

typedef struct uacpi_region_serial_rw_data
{
    void *handler_context;
    void *region_context;
    uacpi_u64 command;
    uacpi_data_view connection;
    uacpi_data_view in_out_buffer;
    uacpi_access_attribute access_attribute;

    /*
     * Applicable if access_attribute is one of:
     * - UACPI_ACCESS_ATTRIBUTE_BYTES
     * - UACPI_ACCESS_ATTRIBUTE_RAW_BYTES
     * - UACPI_ACCESS_ATTRIBUTE_RAW_PROCESS_BYTES
     */
    uacpi_u8 access_length;
} uacpi_region_serial_rw_data;

This is how many args u get per one AML access to a GSB field

#

gone are the simple days

calm latch Jan 8, 2025, 1:40 AM

#

Does this add new kernel APIs?

fiery turtle Jan 8, 2025, 1:40 AM

#

nope

calm latch Jan 8, 2025, 1:40 AM

#

It's not the msr stuff then¿

fiery turtle Jan 8, 2025, 1:40 AM

#

same api as e.g. EC

#

u handle whatever is needed yourself

#

u register a ffixedhw address space handler, then handle a UACPI_REGION_OP_FFIXEDHW_COMMAND

#

if that asks u do read an msr u do that

calm latch Jan 8, 2025, 1:41 AM

#

fun

fiery turtle Jan 8, 2025, 1:42 AM

#

tbf linux currently doesnt implement that

#

ffixedhw was added in 2022

#

https://elixir.bootlin.com/linux/v6.13-rc3/source/drivers/acpi/acpi_ffh.c

acpi_ffh.c - drivers/acpi/acpi_ffh.c - Linux source ...

Elixir Cross Referencer - source code of Linux v6.13-rc3: drivers/acpi/acpi_ffh.c

calm latch Jan 8, 2025, 1:43 AM

#

return -EOPNOTSUPP;
?

#

boring of linux

fiery turtle Jan 8, 2025, 1:43 AM

#

actually no

#

https://elixir.bootlin.com/linux/v6.13-rc3/source/drivers/acpi/arm64/ffh.c#L65

ffh.c - drivers/acpi/arm64/ffh.c - Linux source code...

Elixir Cross Referencer - source code of Linux v6.13-rc3: drivers/acpi/arm64/ffh.c

#

there's an impl for arm64

calm latch Jan 8, 2025, 1:45 AM

#

calm latch It's not the msr stuff then¿

I'm aswking because that will probably be a new syscall in pmOS

fiery turtle Jan 8, 2025, 1:46 AM

#

lucky for you this is very new stuff so no hw using it yet

#

besides some esoteric arm systems

calm latch Jan 8, 2025, 1:46 AM

#

there was a spec for risc-v though

#

ultrameme

fiery turtle Jan 8, 2025, 1:48 AM

#

hopefully i dont forget to fix an acpica bug i found tomorrow

fiery turtle Jan 8, 2025, 1:48 AM

#

calm latch there was a spec for risc-v though

true

#

this is what it takes to figure out whatever you're supposed to do with these esoteric regions

frank canopy Jan 8, 2025, 1:50 AM

#

lmao nice

fiery turtle Jan 8, 2025, 1:50 AM

#

having the entirety of acpica open LULW

#

but now that i understand how they work its more or less simple actually

frank canopy Jan 8, 2025, 1:51 AM

#

oh?

fiery turtle Jan 8, 2025, 1:52 AM

#

which part? lol

frank canopy Jan 8, 2025, 1:52 AM

#

fair enough lol

#

im mostly curious how the address spaces all work because i found like zero docs for any of them

fiery turtle Jan 8, 2025, 1:53 AM

#

which one are u curious about

frank canopy Jan 8, 2025, 1:53 AM

#

generic serial since it sounds like the worst of them

fiery turtle Jan 8, 2025, 1:53 AM

#

okay so

#

I can give u a real world example i guess

calm latch Jan 8, 2025, 1:54 AM

#

at least you can look at the source KEKW

fiery turtle Jan 8, 2025, 1:54 AM

#

Name (XPMC, ResourceTemplate ()
{
    I2cSerialBusV2 (0x0034, ControllerInitiated, 0x00061A80,
        AddressingMode7Bit, "\\_SB.I2C5",
        0x00, ResourceConsumer, , Exclusive,
        )
})
OperationRegion (XSCG, GenericSerialBus, Zero, 0x0100)
Field (XSCG, BufferAcc, NoLock, Preserve)
{
    Connection (XPMC), 
    AccessAs (BufferAcc, AttribByte), 
    XP00,   8, 
    XP01,   8
}

calm latch Jan 8, 2025, 1:54 AM

#

(and not have to reverse NT)

fiery turtle Jan 8, 2025, 1:54 AM

#

This declares a GenericSerialBus at [0, 100] (these numbers are meaningless)

calm latch Jan 8, 2025, 1:54 AM

#

fiery turtle ```py Name (XPMC, ResourceTemplate () { I2cSerialBusV2 (0x0034, ControllerIn...

I see I2C

#

So if I write SMBus driver, it will somehow need to hook into ACPI?

fiery turtle Jan 8, 2025, 1:55 AM

#

Connection (XPMC), connects all following fields with the specified i2c controller

#

in our case I2C5

#

AccessAs (BufferAcc, AttribByte), specifies the SMBus protocol to use for access

#

in this case AttribByte aka read or write one byte

#

Then reading e.g.XP00 would return a 3 byte buffer, status, length, and the response byte

frank canopy Jan 8, 2025, 1:57 AM

#

whats the 8 do then?

fiery turtle Jan 8, 2025, 1:57 AM

#

Since XP00 is at offset 0 it would read from this i2c with command "0"

fiery turtle Jan 8, 2025, 1:57 AM

#

frank canopy whats the 8 do then?

meaningless

#

but it means 8 bits

frank canopy Jan 8, 2025, 1:57 AM

#

ah ok

fiery turtle Jan 8, 2025, 1:57 AM

#

in this case its ignored

fiery turtle Jan 8, 2025, 1:57 AM

#

calm latch So if I write SMBus driver, it will somehow need to hook into ACPI?

yeah

calm latch Jan 8, 2025, 1:58 AM

#

nooo

#

Can ACPI write to PS/2? Thonk

frank canopy Jan 8, 2025, 1:58 AM

#

so in this case its saying use smbus to access the resource from XPMC?

fiery turtle Jan 8, 2025, 1:58 AM

#

calm latch Can ACPI write to PS/2? <:Thonk:1271231807945441402>

via normal SystemIO, sure

fiery turtle Jan 8, 2025, 1:58 AM

#

frank canopy so in this case its saying use smbus to access the resource from XPMC?

its saying in order to read the following fields u must use the aforementioned i2c device

calm latch Jan 8, 2025, 1:59 AM

#

No, but not for normal "your PS/2 controller is actually superio constoller trl "

#

But via this

frank canopy Jan 8, 2025, 1:59 AM

#

fiery turtle its saying in order to read the following fields u must use the aforementioned i...

so how does this not require more kernel apis to implement stuff like i2c?

left orbit Jan 8, 2025, 2:00 AM

#

everything is done over mmio or system io

#

uacpi implements it all for you

calm latch Jan 8, 2025, 2:00 AM

#

not everything

fiery turtle Jan 8, 2025, 2:00 AM

#

frank canopy so how does this not require more kernel apis to implement stuff like i2c?

uacpi_install_address_space_handler(GENERIC_SERIAL_BUS, handle_i2c_io, ...

calm latch Jan 8, 2025, 2:00 AM

#

there are gpio memes for example

fiery turtle Jan 8, 2025, 2:00 AM

#

left orbit everything is done over mmio or system io

not on all systems

frank canopy Jan 8, 2025, 2:01 AM

#

fiery turtle `uacpi_install_address_space_handler(GENERIC_SERIAL_BUS, handle_i2c_io, ...`

ahhhh ok so for that case id need to add handlers for stuff like that then, got it

fiery turtle Jan 8, 2025, 2:01 AM

#

but most blobs have a handrolled smbus driver in aml

left orbit Jan 8, 2025, 2:01 AM

#

what else could it be, msr on x86?

#

or like a proper driver?

fiery turtle Jan 8, 2025, 2:01 AM

#

frank canopy ahhhh ok so for that case id need to add handlers for stuff like that then, got ...

yeah

fiery turtle Jan 8, 2025, 2:01 AM

#

left orbit what else could it be, msr on x86?

any of the following:

UACPI_ADDRESS_SPACE_SYSTEM_MEMORY = 0,
UACPI_ADDRESS_SPACE_SYSTEM_IO = 1,
UACPI_ADDRESS_SPACE_PCI_CONFIG = 2,
UACPI_ADDRESS_SPACE_EMBEDDED_CONTROLLER = 3,
UACPI_ADDRESS_SPACE_SMBUS = 4,
UACPI_ADDRESS_SPACE_SYSTEM_CMOS = 5,
UACPI_ADDRESS_SPACE_PCI_BAR_TARGET = 6,
UACPI_ADDRESS_SPACE_IPMI = 7,
UACPI_ADDRESS_SPACE_GENERAL_PURPOSE_IO = 8,
UACPI_ADDRESS_SPACE_GENERIC_SERIAL_BUS = 9,
UACPI_ADDRESS_SPACE_PCC = 0x0A,
UACPI_ADDRESS_SPACE_PRM = 0x0B,
UACPI_ADDRESS_SPACE_FFIXEDHW = 0x7F,

left orbit Jan 8, 2025, 2:02 AM

#

hmm okay

#

and how do you usually handle those?

#

is it like platform specific drivers?

fiery turtle Jan 8, 2025, 2:03 AM

#

left orbit is it like platform specific drivers?

Yup

#

Usually none of these are mandatory besides the base ones that uacpi handles for you

#

But some platforms may need them e.g. to retrieve battery stats

#

Or other info

frank canopy Jan 8, 2025, 2:04 AM

#

fiery turtle Usually none of these are mandatory besides the base ones that uacpi handles for...

that being system memory, system io, and pci config?

fiery turtle Jan 8, 2025, 2:04 AM

#

frank canopy that being system memory, system io, and pci config?

Yeah

winter orbit Jan 8, 2025, 3:15 AM

#

fiery turtle Yay

Nice

#

Can re try installing the official Ryzen AMD drivers for GPIO and friends

#

oof good times

vast kestrel Jan 8, 2025, 7:54 AM

#

hollow elm static assert with message is a C23 feature

Isn't it the opposite? C23 has the ability to not provide a message iirc

#

https://en.cppreference.com/w/c/language/_Static_assert

c/language/ Static assert

hollow elm Jan 8, 2025, 8:01 AM

#

vast kestrel Isn't it the opposite? C23 has the ability to not provide a message iirc

yes someone mentioned it earlier, my mistake

#

i misread the cppref page

vast kestrel Jan 8, 2025, 8:02 AM

#

Ah fuck, I started scrolling and saw there were too many messages so just skipped at one point :^)

fiery turtle Jan 8, 2025, 9:22 AM

#

winter orbit Can re try installing the official Ryzen AMD drivers for GPIO and friends

Lol @calm latch actually wrote one himself for power button notifications

fiery turtle Jan 8, 2025, 11:23 AM

#

@winter orbit https://learn.microsoft.com/en-us/cpp/c-language/static-assert-c?view=msvc-170 this page says it was introduced in C11

_Static_assert keyword and static_assert macro (C11)

Describes the C11 _Static_assert keyword and the C11 static_assert macro.

#

which C mode were u compiling in?

fiery turtle Jan 8, 2025, 12:43 PM

#

working on address space tests... kinda boring but the peace of mind is worth it

hallow marten Jan 8, 2025, 4:32 PM

#

fiery turtle working on address space tests... kinda boring but the peace of mind is worth it

is aml turing complete

#

can I write a raytracer in it

fiery turtle Jan 8, 2025, 4:50 PM

#

hallow marten is aml turing complete

Absolutely

fiery turtle Jan 8, 2025, 4:51 PM

#

hallow marten can I write a raytracer in it

Pretty much anything

hallow marten Jan 8, 2025, 4:51 PM

#

are there floats

#

I doubt it but would be fun meme

mortal yoke Jan 8, 2025, 4:52 PM

#

no

#

but you could do fixed point

fiery turtle Jan 8, 2025, 4:52 PM

#

^

#

You could write a soft float library

mortal yoke Jan 8, 2025, 4:53 PM

#

or I guess you could also literally use the ieee float format and write a software impl lol

fiery turtle Jan 8, 2025, 4:53 PM

#

Yep

mortal yoke Jan 8, 2025, 4:53 PM

#

but that's more work than fixed point

gentle peak Jan 8, 2025, 4:55 PM

#

is it possible to write a jit compiler in aml

hallow marten Jan 8, 2025, 4:55 PM

#

mortal yoke but you could do fixed point

yeah

#

just use bigger numbers to represent decimals

mortal yoke Jan 8, 2025, 4:56 PM

#

gentle peak is it possible to write a jit compiler in aml

I don't see why not, though ofc you can't execute it if you mean like native code

#

actually that could be fun, compiling some kind of textual language to aml in aml and then Load'ing it KEKW

hallow marten Jan 8, 2025, 4:57 PM

#

if there are functions and stuff then it's pretty much at the same level as C

fiery turtle Jan 8, 2025, 4:57 PM

#

hallow marten if there are functions and stuff then it's pretty much at the same level as C

Yeah

gentle peak Jan 8, 2025, 5:01 PM

#

mortal yoke I don't see why not, though ofc you can't execute it if you mean like native cod...

yeah I mean like compiling to aml

#

if you pair that with a JIT-based AML interpreter I wonder what the performance of e.g. a JVM implementation would be

mortal yoke Jan 8, 2025, 5:05 PM

#

kinda unrelated to this I got a cursed idea for the qacpi _OSI method from this, what if I make the statically compiled aml generate the _OSI aml based on a list of strings you pass to it and then Load'ing that KEKW then the user could remove/add osi strings and call a recompile method that compiles the _OSI method (assuming you don't need to change them during runtime which I think is a reasonable assumption)

#

and tbh changing it during runtime isn't really a problem either, I can just remove the _OSI node before recompiling to update it

fiery turtle Jan 8, 2025, 5:07 PM

#

mortal yoke and tbh changing it during runtime isn't really a problem either, I can just rem...

You can just copy object to it

#

Cursed af yeah

#

But should work

mortal yoke Jan 8, 2025, 5:09 PM

#

wouldn't the method would have to have a name that's not already in the namespace tho

#

the generated one that is, else its just going to get ignored

fiery turtle Jan 8, 2025, 5:10 PM

#

mortal yoke the generated one that is, else its just going to get ignored

Ah right

#

Can always special case it 💀

mortal yoke Jan 8, 2025, 5:10 PM

#

also one minor problem I can see with this is that the data buffer that was previously loaded is not going to get freed if you recompile it multiple times

fiery turtle Jan 8, 2025, 5:11 PM

#

Why not just have a package object of allowed strings instead

#

Then u can just copy object to that

#

Then your osi would just iterate it

mortal yoke Jan 8, 2025, 5:11 PM

#

that could actually be good solution tbh

fiery turtle Jan 8, 2025, 5:12 PM

#

The only problem is you could run into a name thats already taken

#

So u have to be careful

mortal yoke Jan 8, 2025, 5:16 PM

#

then imagine some firmware dev starts using that to check for this specific interpreter troll

fiery turtle Jan 8, 2025, 5:19 PM

#

Lol

winter orbit Jan 8, 2025, 5:23 PM

#

mortal yoke then imagine some firmware dev starts using that to check for this specific inte...

I can make it happen

frank canopy Jan 8, 2025, 5:37 PM

#

lmao

mortal yoke Jan 8, 2025, 6:03 PM

#

If (CondRefOf (\QACP)) {
    // refuse run on qacpi, we don't support that.
    DataTableRegion (FACP, "FACP", "", "")

    Field (FACP, DWordAcc, NoLock) {
        Offset (64),
        PM1A, 32,
        PM1B, 32
    }

    If (CondRefOf (\_PTS)) {
        \_PTS (5)
    }

    Local0 = \_S5
    If (SizeOf (Local0) == 1) {
        Local0 = DerefOf (Local0[0])
        Local1 = (Local0 >> 8) & 0xFF
        Local0 = Local0 & 0xFF
    }
    Else {
        Local1 = DerefOf (Local0[1])
        Local0 = DerefOf (Local0[0])
    }

    OperationRegion (PMA, SystemIO, PM1A, 2)
    OperationRegion (PMB, SystemIO, PM1B, 2)
    Field (PMA, WordAcc, NoLock) {
        PMA1, 16
    }
    Field (PMB, WordAcc, NoLock) {
        PMB1, 16
    }

    Local2 = PMA1
    If (PM1B != 0) {
        Local2 = Local2 | PMB1
    }
    Local2 = Local2 & ~0x3C00
    Local3 = Local2
    Local2 = Local2 | (Local0 << 10)
    Local3 = Local3 | (Local1 << 10)
    PMA1 = Local2
    If (PM1B != 0) {
        PMB1 = Local3
    }

    Local2 = Local2 | (1 << 13)
    Local3 = Local3 | (1 << 13)
    PMA1 = Local2
    If (PM1B != 0) {
        PMB1 = Local3
    }
    Sleep (1000 * 60 * 60 * 24)
}

fiery turtle Jan 8, 2025, 6:03 PM

#

bruh

loud ice Jan 8, 2025, 6:04 PM

#

mortal yoke ``` If (CondRefOf (\QACP)) { // refuse run on qacpi, we don't support that. ...

genius

fiery turtle Jan 8, 2025, 6:04 PM

#

cant believe u actually spent time writing this LULW

loud ice Jan 8, 2025, 6:04 PM

#

i have a solution to this issue

#

implement an aml flag which makes it so that all identifiers that start with "QA" are namespaced differently

#

and separately

fiery turtle Jan 8, 2025, 6:05 PM

#

just scan the entire blob for references to QACP and rename to QACB

loud ice Jan 8, 2025, 6:05 PM

#

nah you can just rewrite your aml

mortal yoke Jan 8, 2025, 6:06 PM

#

fiery turtle cant believe u actually spent time writing this <:LULW:543420553021882371>

I didn't have anything else to do lol (or well I could always work on my kernel but I though it was kinda fun doing shutdown using the fadt from aml)

loud ice Jan 8, 2025, 6:06 PM

#

idea

#

FDT parser in AML

#

so you can use uACPI for FDT

fiery turtle Jan 8, 2025, 6:07 PM

#

💀

#

write an aml program that takes an FDT buffer as input and generates an aml namespace from it

loud ice Jan 8, 2025, 6:08 PM

#

yeah

#

thats what i meant obviously

fiery turtle Jan 8, 2025, 6:09 PM

#

thats surprisingly doable

#

u can even load that program as an SSDT dynamically

mortal yoke Jan 8, 2025, 6:09 PM

#

and the dsdt data can be put inside a table

loud ice Jan 8, 2025, 6:09 PM

#

yeah

#

its surprisingly doable

#

and that is the scary part

fiery turtle Jan 8, 2025, 6:13 PM

#

lmfao

winter orbit Jan 8, 2025, 6:15 PM

#

As someone whose daily driving uACPI this software is production ready

#

Okay now pin that out context trollface

mortal yoke Jan 8, 2025, 6:16 PM

#

did your windows uacpi experiment succeed btw?

winter orbit Jan 8, 2025, 6:16 PM

#

its still going

#

im using it rn

mortal yoke Jan 8, 2025, 6:16 PM

#

nice

loud ice Jan 8, 2025, 6:16 PM

#

damn

#

daily driving uacpi

winter orbit Jan 8, 2025, 6:16 PM

#

I put it on my main machine yeah

#

Only had one issue so far and it’s my code not uACPI I think

fiery turtle Jan 8, 2025, 6:22 PM

#

winter orbit As someone whose daily driving uACPI this software is production ready

fiery turtle Jan 8, 2025, 6:22 PM

#

winter orbit As someone whose daily driving uACPI this software is production ready

done

#

Framework's EC device is kinda fucked

#

it lists like 8 ports in _CRS

#

you're only supposed to have 2 💀

#

and a fucking GpioInt for some reason 💀 x2

winter orbit Jan 8, 2025, 6:23 PM

#

oof

#

This is my daily driver btw

#

trl

fiery turtle Jan 8, 2025, 6:24 PM

#

i think it expects the OS to only use the first two (since they're located at classic 0x62,0x66)

#

and the rest is for os bookkeeping purposes

#

classic unconditionally assuming EC presence if semi recent OS

#

firmware devs dont read the spec

#

nowhere in the spec does it say "treat _INI as _REG lol"

winter orbit Jan 8, 2025, 6:26 PM

#

I think I know one of the people that worked on this

#

chad

#

Let me go punch him

fiery turtle Jan 8, 2025, 6:27 PM

#

lol

leaden fox Jan 8, 2025, 6:36 PM

#

the og framework has had a huge battery drain firmware issue during suspend

#

no clue if they fixed it

winter orbit Jan 8, 2025, 6:37 PM

#

they did but this is the 16

leaden fox Jan 8, 2025, 6:37 PM

#

I just know the hotfix was using hibernate

winter orbit Jan 8, 2025, 6:37 PM

#

it’s a very different firmware

#

This is far closer to the steam deck

leaden fox Jan 8, 2025, 6:37 PM

#

how's the big one

#

any good?

winter orbit Jan 8, 2025, 6:37 PM

#

I like it I’ve been using it for awhile

#

It’s my daily driver

#

well right now obviously I’m running uACPI in it in win11

#

But yeah

#

The things suspend works amazing too I put it in my bag and all is well

vale isle Jan 8, 2025, 6:39 PM

#

winter orbit This is far closer to the steam deck

steam deck uacpi when

#

ig that's just linux uacpi

leaden fox Jan 8, 2025, 6:39 PM

#

winter orbit The things suspend works amazing too I put it in my bag and all is well

that's sick

#

my framework drains so quick

#

the fan is also dirty I think which makes it pretty laggy sometimes

winter orbit Jan 8, 2025, 6:40 PM

#

I like how my fans are interchangeable with the gpu module basically

winter orbit Jan 8, 2025, 6:40 PM

#

vale isle ig that's just linux uacpi

I could just boot win11 on it

#

Or reactos

#

Either way I’ll get a dump of stream deck for the repo

vale isle Jan 8, 2025, 6:41 PM

#

the elgato stream deck meme

#

uacpi on the stream deck

winter orbit Jan 8, 2025, 6:42 PM

#

KEKW

calm latch Jan 8, 2025, 6:47 PM

#

vale isle steam deck uacpi when

My friend has a steam deck, maybe I could ask him to test pmOS trl

loud ice Jan 8, 2025, 7:00 PM

#

fiery turtle Framework's EC device is kinda fucked

its cros_ec part 2

#

i think

fiery turtle Jan 8, 2025, 7:02 PM

#

loud ice its cros_ec part 2

True

loud ice Jan 8, 2025, 7:03 PM

#

see if those ports are used by cros_ec

vagrant hull Jan 8, 2025, 10:33 PM

#

sorry for the stupid question but how are you even running uACPI in Windows

#

does Windows like, allow you to swap ACPI drivers willy nilly or what?

#

or is the ROS acpi.sys intercompatible with the Windows one?

#

and you just made a uACPI acpi.sys and Windows just swallows it?

winter orbit Jan 8, 2025, 10:39 PM

#

vagrant hull does Windows like, allow you to swap ACPI drivers willy nilly or what?

No nope

this is disabling all signatures, enabled boot KD debugging and replacing the acpi.sys driver with my new one

winter orbit Jan 8, 2025, 10:39 PM

#

vagrant hull or is the ROS acpi.sys intercompatible with the Windows one?

it's never been

This new one is going to be :)

vagrant hull Jan 8, 2025, 10:40 PM

#

i see

#

ty

loud ice Jan 8, 2025, 10:51 PM

#

winter orbit No nope this is disabling all signatures, enabled boot KD debugging and replaci...

that sounds like a great configuration for your main system to be in trl

#

in unrelated news i think you should run my new game, the executable is called BiosDevSecretStealerBootkitInstaller.exe meme

winter orbit Jan 8, 2025, 11:02 PM

#

loud ice that sounds like a great configuration for your main system to be in <:trl:82931...

I never said it was a GOOD idea

#

Just that I’m doing it

#

And it’s working

#

drink

vagrant hull Jan 8, 2025, 11:24 PM

#

@fiery turtle i have a new BIOS :^)

fiery turtle Jan 8, 2025, 11:25 PM

#

Give meme

vagrant hull Jan 8, 2025, 11:26 PM

#

sent

#

they seem to have fixed whatever was triggering that ACPICA error

#

from the last revision

fiery turtle Jan 8, 2025, 11:43 PM

#

vagrant hull they seem to have fixed whatever was triggering that ACPICA error

lol

#

thanks ill take a look at the diff

left orbit Jan 8, 2025, 11:43 PM

#

my bios update today added 80KB of shit related to audio???

#

to the DSDT

fiery turtle Jan 8, 2025, 11:44 PM

#

lol

dense steppe Jan 8, 2025, 11:44 PM

#

you folks get bios updates?

frank canopy Jan 8, 2025, 11:46 PM

#

ikr cant remember the last time ive noticed a bios update

#

maybe theyre sneaking past me but idk

dense steppe Jan 8, 2025, 11:46 PM

#

(obviously there are bios updates for my computer but I am lazy and won't do anything about that unless something breaks or a security issue arises meme )

fiery turtle Jan 8, 2025, 11:47 PM

#

dense steppe (obviously there are bios updates for my computer but I am lazy and won't do any...

Same

#

My Mobo is boasting that 2019 built-in bios

dense steppe Jan 8, 2025, 11:48 PM

#

frank canopy maybe theyre sneaking past me but idk

some BIOS configurations allow Windows Update to install a "capsule" that gets installed onto the motherboard at next boot.

#

though I have that turned off

#

(and very seldom do I boot into Windows)

#

actually semi-recently (like mid 2024 or something) there was a bug for recent AMD processors that iirc had weak but not non-existent security implications

#

this issue was supposedly fixed in microcode

#

iirc BIOS updates were recommended, so that the fix gets applied even before the kernel boots

#

IMO it is entirely fine to patch the microcode when the kernel boots instead. I don't know why it wouldn't be possible to do so.

left orbit Jan 8, 2025, 11:51 PM

#

it is though

#

kernels do load microcode on boot

dense steppe Jan 8, 2025, 11:52 PM

#

yeah exactly.

winter orbit Jan 9, 2025, 12:10 AM

#

dense steppe some BIOS configurations allow Windows Update to install a "capsule" that gets i...

It’s not a windows thing it’s just called a windows thing for marketing

#

They’re actually tiny OSes that do it

#

And have signed permission to do it

#

The capsule format is public

#

Linux even has an updater for it

hollow elm Jan 9, 2025, 12:13 AM

#

on hp laptops (or at least the one i have) bios updates are done via an efi application

#

the windows program just drops some files in the ESP and reboots into it

#

it also doesn't clean up anything so i have the updater and the whole bios image on the ESP still

winter orbit Jan 9, 2025, 12:26 AM

#

hollow elm the windows program just drops some files in the ESP and reboots into it

yeah

winter orbit Jan 9, 2025, 12:27 AM

#

hollow elm on hp laptops (or at least the one i have) bios updates are done via an efi appl...

Even HPs utility used the same protocol though

vagrant hull Jan 9, 2025, 8:36 PM

#

dead project :^)

fiery turtle Jan 9, 2025, 8:39 PM

#

sad

#

writing tests for the 1.0 PR

#

its mostly done

#

also irl work soo

rustic compass Jan 10, 2025, 7:06 AM

#

@fiery turtle in type.h:

typedef enum uacpi_region_op {
    UACPI_REGION_OP_ATTACH = 1,
    UACPI_REGION_OP_READ = 2,
    UACPI_REGION_OP_WRITE = 3,
    UACPI_REGION_OP_DETACH = 4,
} uacpi_region_op;

typedef struct uacpi_region_attach_data {
    void *handler_context;
    uacpi_namespace_node *region_node;
    void *out_region_context;
} uacpi_region_attach_data;

typedef struct uacpi_region_rw_data {
    void *handler_context;
    void *region_context;
    union {
        uacpi_phys_addr address;
        uacpi_u64 offset;
    };
    uacpi_u64 value;
    uacpi_u8 byte_width;
} uacpi_region_rw_data;

typedef struct uacpi_region_detach_data {
    void *handler_context;
    void *region_context;
    uacpi_namespace_node *region_node;
} uacpi_region_detach_data;

typedef uacpi_status (*uacpi_region_handler)
    (uacpi_region_op op, uacpi_handle op_data);

typedef uacpi_status (*uacpi_notify_handler)
    (uacpi_handle context, uacpi_namespace_node *node, uacpi_u64 value);

is context user defineable content?

#

as in when registering an interrupt it takes the user supplied pointer and on handler invocation it passes it to the handler?

#

also can i rely on the heap when you need/can register handler?

#

as in it must be present to even get to the point of registering handlers?

fiery turtle Jan 10, 2025, 7:30 AM

#

rustic compass <@353593948523986945> in type.h: ```C typedef enum uacpi_region_op { UACPI_R...

Both are, one is global per handler, the other is per detected opregion with this type

fiery turtle Jan 10, 2025, 7:30 AM

#

rustic compass as in when registering an interrupt it takes the user supplied pointer and on ha...

Yes

fiery turtle Jan 10, 2025, 7:31 AM

#

rustic compass also can i rely on the heap when you need/can register handler?

Yes

fiery turtle Jan 10, 2025, 7:31 AM

#

rustic compass as in it must be present to even get to the point of registering handlers?

Registering handlers happens fairly late so yea

thorny orbit Jan 10, 2025, 10:58 AM

#

what tables used in uacpi_find_devices()

fiery turtle Jan 10, 2025, 10:59 AM

#

none

thorny orbit Jan 10, 2025, 11:00 AM

#

fiery turtle none

ok

torpid root Jan 10, 2025, 11:04 AM

#

did that "ok" have to be a ping

slim panther Jan 10, 2025, 11:04 AM

#

torpid root did that "ok" have to be a ping

ok

vagrant hull Jan 10, 2025, 12:56 PM

#

@fiery turtle did you check the diff of my BIOS yet?

fiery turtle Jan 10, 2025, 2:19 PM

#

Not yet but I also updated my bios from 2018 and my monitors firmware

#

Will do today

fiery turtle Jan 10, 2025, 6:08 PM

#

vagrant hull <@353593948523986945> did you check the diff of my BIOS yet?

mostly WMI memes

#

moved a lot of code around

#

some misc changes

vagrant hull Jan 10, 2025, 6:34 PM

#

still lots of debug stuff around?

fiery turtle Jan 10, 2025, 7:04 PM

#

even more added

fiery turtle Jan 11, 2025, 9:52 AM

#

hmm can I throw C++ exceptions from uacpi callbacks in my test runner

#

that would make it much easier to see what failed

#

SO seems to suggest i can

#

as long as I compile everything with the same toolchain

calm latch Jan 11, 2025, 10:27 AM

#

i think you can

#

I've read somewhere that malloc throws exceptions with some libc s trl

hasty plinth Jan 11, 2025, 10:29 AM

#

I guess its not marked nothrow

#

wtf though

fiery turtle Jan 11, 2025, 10:40 AM

#

calm latch I've read somewhere that malloc throws exceptions with some libc s <:trl:8293166...

wtf

calm latch Jan 11, 2025, 10:41 AM

#

Ok maybe not malloc

#

But I remember reading something about it

fiery turtle Jan 11, 2025, 10:49 AM

#

malloc has special rules in c++

#

so it throwing wouldnt surprise me tbh

left orbit Jan 11, 2025, 11:15 AM

#

fiery turtle hmm can I throw C++ exceptions from uacpi callbacks in my test runner

where do you want to catch them?

#

like i guess as long as you throw and catch within the callback its gonna be finee

#

but throwing across the uacpi boundary might me tough

fiery turtle Jan 11, 2025, 11:16 AM

#

left orbit like i guess as long as you throw and catch within the callback its gonna be fin...

no im literally throwing them via the entire uacpi stack

#

it works fine

left orbit Jan 11, 2025, 11:16 AM

#

wtf

fiery turtle Jan 11, 2025, 11:16 AM

#

although it leaks everything of course

left orbit Jan 11, 2025, 11:16 AM

#

that sounds sketchy as fuck

#

oh yeah

#

ofc it works

#

but thats waht i meant by "might be tough"

fiery turtle Jan 11, 2025, 11:16 AM

#

i dont care because its just for my tests and happens only on error anyway

#

yeah

#

address spaces tests progressing

#

but not feeling well atm

#

i also made custom tracing for GeneralPurposeIO etc

#

so it prints out pin numbers and stuff

#

💀

#

ah nvm

#

uacpi_bool uacpi_is_buffer_access_address_space(uacpi_address_space space)
{
    switch (space) {
    case UACPI_ADDRESS_SPACE_SMBUS:
    case UACPI_ADDRESS_SPACE_IPMI:
    case UACPI_ADDRESS_SPACE_GENERAL_PURPOSE_IO:
    case UACPI_ADDRESS_SPACE_GENERIC_SERIAL_BUS:
    case UACPI_ADDRESS_SPACE_PRM:
    case UACPI_ADDRESS_SPACE_FFIXEDHW:
        return UACPI_TRUE;
    default:
        return UACPI_FALSE;
    }
}

a bit of brain damage in this switch

#

💀 x2

#

LETSFUCKINGGOOOOOO

#

least overengineered test

calm latch Jan 11, 2025, 12:18 PM

#

C just needs defer

mortal yoke Jan 11, 2025, 12:19 PM

#

defer sucks

#

well its better than nothing but proper destructors are way better

calm latch Jan 11, 2025, 12:20 PM

#

https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2895.htm

calm latch Jan 11, 2025, 12:21 PM

#

mortal yoke well its better than nothing but proper destructors are way better

Then you need classes, RAII and the whole shebang

calm latch Jan 11, 2025, 12:21 PM

#

calm latch https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2895.htm

Someone proposed it

#

It's a simple and elegant solution

#

imo

#

Like I've used it in Go and quite liked it

torpid root Jan 11, 2025, 12:21 PM

#

calm latch Then you need classes, RAII and the whole shebang

that would be a whole another language
like better c

calm latch Jan 11, 2025, 12:22 PM

#

Like RAII of course is more powerful

torpid root Jan 11, 2025, 12:22 PM

#

maybe we could call it c+1

calm latch Jan 11, 2025, 12:22 PM

#

Thonk what if you call it c but incremented so c++

mortal yoke Jan 11, 2025, 12:23 PM

#

calm latch Then you need classes, RAII and the whole shebang

well yeah it would indeed need more things than it too which is why defer would fit c better even though its a non-ideal solution (though c itself is non-ideal for basically everything so it would fit lol)

calm latch Jan 11, 2025, 12:23 PM

#

Also there's D KEKW

torpid root Jan 11, 2025, 12:23 PM

#

calm latch <:Thonk:1271231807945441402> what if you call it c but incremented so c++

would c++ return c though?
what about ++c

calm latch Jan 11, 2025, 12:23 PM

#

D

torpid root Jan 11, 2025, 12:24 PM

#

basically 'C' incremented lol

calm latch Jan 11, 2025, 12:24 PM

#

calm latch https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2895.htm

Why

#

this is just labdas at this point

#

who proposed it

mortal yoke Jan 11, 2025, 12:25 PM

#

lambdas would be nice too

#

but I'd much more like generics

#

(though I barely use c anymore so eh, I don't really care)

calm latch Jan 11, 2025, 12:25 PM

#

lambdas would probably require a better type system

#

__attribute__((__cleanup__(callback))) Thonk

#

Like compilers already implement it just add it to the language nooo

gentle peak Jan 11, 2025, 1:02 PM

#

calm latch C just needs defer

there's this but it'd be better if it had an expression as argument instead of a function https://gcc.gnu.org/onlinedocs/gcc/Common-Variable-Attributes.html#index-cleanup-variable-attribute

Common Variable Attributes (Using the GNU Compiler Collection (GCC))

calm latch Jan 11, 2025, 1:03 PM

#

No but like I want to say that it would be a very useful feature, if they added it to the language

#

Like if you're a C++ compiler you already have infrastructure for it, and if you're an embedded compiler targeting something small and simple, it should still be easily implementable

gentle peak Jan 11, 2025, 1:04 PM

#

fair

dense steppe Jan 11, 2025, 1:06 PM

#

I believe it is possible to implement defer using macros and __attribute__((cleanup)). Maybe the actual code needs to be within parentheses like defer(<code here>);, but it is definitely possible

#

It is possible that this requires GCC lambdas

gentle peak Jan 11, 2025, 1:08 PM

#

if you use nested functions it should be possible yeah

Nested Functions (Using the GNU Compiler Collection (GCC))

#

because they can already access the parent scope's locals

dense steppe Jan 11, 2025, 1:09 PM

#

aren't nested functions that access parent scope locals sometimes implemented using self-modifying code? except in the case of defer, that would be optimized away

mortal yoke Jan 11, 2025, 1:10 PM

#

I am pretty sure nested functions suck in some ways yeah, I am not sure if its that but something there is that makes them not so good

#

something like clang blocks is more sane

#

but gcc doesn't support them

calm latch Jan 11, 2025, 1:11 PM

#

Why do you even need nested functions?

dense steppe Jan 11, 2025, 1:11 PM

#

to use as the cleanup function in __attribute__((cleanup))

calm latch Jan 11, 2025, 1:11 PM

#

like just create raii-style destructors for everything

gentle peak Jan 11, 2025, 1:11 PM

#

cleanup can only call functions, and you want to be able to use locals in your cleanup handler

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