uACPI - a portable and easy-to-integrate ACPI implementation | Operating System Development | Page 43

frank canopy Feb 21, 2025, 7:12 PM

#

and the concepts should translate fine

calm latch Feb 21, 2025, 7:12 PM

#

It's fine

fiery turtle Feb 21, 2025, 7:15 PM

#

yeah

#

Nice

frank canopy Feb 21, 2025, 7:17 PM

#

ill be swapping this to iterator api soon so sent a permalink lol

#

the descending/ascending is clearer and more like what most projects will use i think

fiery turtle Feb 21, 2025, 7:17 PM

#

Ill probably merge iterator api today

frank canopy Feb 21, 2025, 7:17 PM

#

cool

calm latch Feb 21, 2025, 7:18 PM

#

I can try the iterator thing, I have no preference what to use

fiery turtle Feb 21, 2025, 7:18 PM

#

Especially if @north holly tests the new pr meme

frank canopy Feb 21, 2025, 7:18 PM

#

what i also might do is create a device object for specifically the _SB namespace and iterate under that

fiery turtle Feb 21, 2025, 7:18 PM

#

calm latch I can try the iterator thing, I have no preference what to use

Nah its not needed in C

frank canopy Feb 21, 2025, 7:18 PM

#

also not merged yet

#

its very good for working in zig and absolutely unnecessary in C

#

because callbacks through C cant propogate zig errors and error tracing

fiery turtle Feb 21, 2025, 7:19 PM

#

Haiku uses it for c++ for whatever reason

north holly Feb 21, 2025, 7:19 PM

#

fiery turtle Especially if <@841762704766140486> tests the new pr <:meme:575445094589661215>

that's not happening, got my report card

fiery turtle Feb 21, 2025, 7:19 PM

#

The what

north holly Feb 21, 2025, 7:19 PM

#

Yes.

fiery turtle Feb 21, 2025, 7:19 PM

#

Like school stuff?

north holly Feb 21, 2025, 7:19 PM

#

yes

fiery turtle Feb 21, 2025, 7:19 PM

#

Ah

#

Ig ill test myself with astral then

frank canopy Feb 21, 2025, 7:21 PM

#

if i wasnt busy today id pull in the PR and test that

#

its one cli command to swap over and after that i just need to write binding externs

calm latch Feb 21, 2025, 7:21 PM

#

What do you need to test?

frank canopy Feb 21, 2025, 7:21 PM

#

the new iterator api

fiery turtle Feb 21, 2025, 7:21 PM

#

Just normal operation

frank canopy Feb 21, 2025, 7:21 PM

#

where you call next instead of giving a callback

fiery turtle Feb 21, 2025, 7:21 PM

#

That nothing regressrd

#

And yeah the iterator api

frank canopy Feb 21, 2025, 7:22 PM

#

but im stuck doing (thankfully paying) web dev work today 😭 so no testing new uacpi things

fiery turtle Feb 21, 2025, 7:23 PM

#

Good for you anyway

north holly Feb 21, 2025, 7:23 PM

#

if I get the chance I will be testing the iterator api

#

and will be migrating my acpi driver detection to use that

fiery turtle Feb 21, 2025, 7:23 PM

#

If you just check that nothing in obos regressed that'd be great

fiery turtle Feb 21, 2025, 7:23 PM

#

north holly and will be migrating my acpi driver detection to use that

Its worse than the old api

#

So dont do that

north holly Feb 21, 2025, 7:24 PM

#

worse how

fiery turtle Feb 21, 2025, 7:24 PM

#

Requires dynamic memory or recursion

north holly Feb 21, 2025, 7:24 PM

#

nvm

#

I will temporarily change it

#

but revert it after finding it works

#

or doesn't work

fiery turtle Feb 21, 2025, 7:24 PM

#

Im implementing it because of haiku basically

#

Btw when are you gonna be back

north holly Feb 21, 2025, 7:25 PM

#

depends

#

I will not elaborate.

fiery turtle Feb 21, 2025, 7:25 PM

#

Fair

north holly Feb 21, 2025, 7:25 PM

#

anywhere from a few days to 5 months

#

meme

fiery turtle Feb 21, 2025, 7:26 PM

#

50/50 either today or not

north holly Feb 21, 2025, 7:26 PM

#

yes

frank canopy Feb 21, 2025, 7:26 PM

#

fiery turtle Im implementing it because of haiku basically

im literally maintaining a hack that breaks error stack traces rn because of not having an iterator style api so i appreciate this too lol

fiery turtle Feb 21, 2025, 7:27 PM

#

Yeah I didn't realize other languages needed it

north holly Feb 21, 2025, 7:27 PM

#

zig L

left orbit Feb 21, 2025, 7:27 PM

#

yeah sadly no closures in zig

#

or anonymous functions (without weird syntax at least)

#

you can technically &struct { fn x() void {} }.x but 🤢

frank canopy Feb 21, 2025, 7:33 PM

#

yeah no

#

the closures is intended because of implicit allocation potential there

#

which any memory management other than declaring a variable (local or global) has to be explicit

#

the lack of anon functions is kinda annoying but i was doing work with java before that lang got them so i easily get used to that

left orbit Feb 21, 2025, 7:35 PM

#

i mean C doesn't have them either soooo

#

¯_(ツ)_/¯

frank canopy Feb 21, 2025, 7:53 PM

#

yeah

#

i dont find it too annoying tbh

#

its just that i cant return an error union from a function that returns uacpi_iteration_result or whatever the enum is called

calm latch Feb 21, 2025, 8:19 PM

#

So do I just run eval on _ADR, _CID, _HID?

#

uacpi_eval_typed

frank canopy Feb 21, 2025, 8:19 PM

#

calm latch So do I just run eval on _ADR, _CID, _HID?

yes but if you get namespace_node_info that function does that for you

calm latch Feb 21, 2025, 8:20 PM

#

oh, fun

frank canopy Feb 21, 2025, 8:29 PM

#

the struct has bools for if things are present and then a field with the thing if it exists its real nice

#

its what i use

calm latch Feb 21, 2025, 8:30 PM

#

Yeah I guess I'll also do that...

#

Wth

6.1.3 _CLS (Class Code)
This object is used to supply OSPM with the PCI-defined base-class, sub-class and programming interface for a device.
This object is optional. However, it may be useful for loading generic drivers on hardware that is compatible with PCI
-defined device classes, but that is not implemented on the PCI bus (and is therefore enumerated by ACPI.)

frank canopy Feb 21, 2025, 8:32 PM

#

yeah

#

i think my code might even handle that i dont remember

#

nope apparently i dont

fiery turtle Feb 22, 2025, 10:45 AM

#

#

there goes Astral new uACPI integration LULW

slim panther Feb 22, 2025, 10:46 AM

#

astral has no memmove?

fiery turtle Feb 22, 2025, 10:48 AM

#

yup

gentle peak Feb 22, 2025, 10:51 AM

#

pretty sure uacpi-linux doesn't have any of the mem* functions

fiery turtle Feb 22, 2025, 10:51 AM

#

anyway i just added UACPI_USE_BUILTIN_STRING

fiery turtle Feb 22, 2025, 10:51 AM

#

gentle peak pretty sure uacpi-linux doesn't have any of the mem* functions

yeah we'll have to pass that define as well

median crest Feb 22, 2025, 10:53 AM

#

fiery turtle there goes Astral new uACPI integration <:LULW:543420553021882371>

Astral has fallen billions must die

fiery turtle Feb 22, 2025, 10:53 AM

#

it's saved

#

dw

left orbit Feb 22, 2025, 10:53 AM

#

uacpi-linux being uacpi in the linux kernel or uacpi running in userspace?

fiery turtle Feb 22, 2025, 10:53 AM

#

the latter ofc

left orbit Feb 22, 2025, 10:53 AM

#

i should get back to the first one :^)

#

if haiku can do it then why cant we

fiery turtle Feb 22, 2025, 10:54 AM

#

this makes me wonder, shoudl i make it possible for uacpi to export its own impls of vsnprintf and even string as globally available without the uacpi_ prefix

fiery turtle Feb 22, 2025, 10:54 AM

#

left orbit i should get back to the first one :^)

LETSFUCKINGGOOOOOO

fiery turtle Feb 22, 2025, 10:54 AM

#

left orbit if haiku can do it then why cant we

true ture

gentle peak Feb 22, 2025, 10:54 AM

#

left orbit if haiku can do it then why cant we

because unlike haiku (apparently) linux uses acpica apis directly kernel wide so migration is a paaaain

median crest Feb 22, 2025, 10:54 AM

#

fiery turtle this makes me wonder, shoudl i make it possible for uacpi to export its own impl...

uprintf when

fiery turtle Feb 22, 2025, 10:55 AM

#

just call uacpi_vsnprintf

#

u already have it

left orbit Feb 22, 2025, 10:55 AM

#

gentle peak because unlike haiku (apparently) linux uses acpica apis directly kernel wide so...

i mean, most apis seem to translate pretty much 1:1

#

it must be doable and not that bad

median crest Feb 22, 2025, 10:55 AM

#

fiery turtle just call uacpi_vsnprintf

"time to switch acpi libraries!!!"
"wait why is my printf not working"

fiery turtle Feb 22, 2025, 10:56 AM

#

most code could probably be converted with sed

fiery turtle Feb 22, 2025, 10:56 AM

#

median crest "time to switch acpi libraries!!!" "wait why is my printf not working"

lol

median crest Feb 22, 2025, 10:56 AM

#

Load bearing uacpi

gentle peak Feb 22, 2025, 10:56 AM

#

it's still like tens of thousands of changes to call sites, types, method names, parameter ordering, etc

left orbit Feb 22, 2025, 10:56 AM

#

true true, but it can be done with enough time and will power lmao

gentle peak Feb 22, 2025, 10:57 AM

#

yeah

#

my main objection is that such a huge effort isn't really worth it unless it has a chance of getting mainlined, and that's not happening unless you add a config option to choose between acpica and uapci, so you essentially have to write your own independent abstraction layer that can be backed by either

fiery turtle Feb 22, 2025, 10:58 AM

#

basically yeah

#

maybe an abstraction later isnt that expensive

#

since so many osses have it

left orbit Feb 22, 2025, 10:59 AM

#

i mean i wanted to have a config switch for uacpi from the start tbh

gentle peak Feb 22, 2025, 10:59 AM

#

expensive almost certainly not, I doubt acpi is in any kind of performance critical path anyway, but it is even more effort

left orbit Feb 22, 2025, 10:59 AM

#

but it might be a big project lol thats true

fiery turtle Feb 22, 2025, 10:59 AM

#

gentle peak expensive almost certainly not, I doubt acpi is in any kind of performance criti...

i mean coding time wise

gentle peak Feb 22, 2025, 10:59 AM

#

ah

#

yeah idk

fiery turtle Feb 22, 2025, 11:00 AM

#

u could just draw inspiration from NTs interface ig

gentle peak Feb 22, 2025, 11:00 AM

#

I feel like no matter what you do the migration is gonna be the most time consuming part of it

fiery turtle Feb 22, 2025, 11:00 AM

#

like changing code to call into it?

gentle peak Feb 22, 2025, 11:00 AM

#

yeah

fiery turtle Feb 22, 2025, 11:00 AM

#

yeah

left orbit Feb 22, 2025, 11:01 AM

#

yeah

flat badge Feb 22, 2025, 11:01 AM

#

Linux maintainers will never merge a compat layer

gentle peak Feb 22, 2025, 11:01 AM

#

fiery turtle u could just draw inspiration from NTs interface ig

what is nt's interface like anyway? I'm curious

flat badge Feb 22, 2025, 11:01 AM

#

If you want to get uacpi into linux, the only reasonable path is probably reimplementing the acpica API on top of uacpi

#

which would probably be ugly af

fiery turtle Feb 22, 2025, 11:02 AM

#

gentle peak what is nt's interface like anyway? I'm curious

https://learn.microsoft.com/en-us/windows-hardware/drivers/ddi/acpiioct/ni-acpiioct-ioctl_acpi_eval_method_ex

IOCTL_ACPI_EVAL_METHOD_EX (acpiioct.h) - Windows drivers

A driver for a device can use the IOCTL_ACPI_EVAL_METHOD_EX device control request to synchronously evaluate an ACPI control method that is supported by a child device in the namespace of the device.

fiery turtle Feb 22, 2025, 11:02 AM

#

flat badge If you want to get uacpi into linux, the only reasonable path is probably reimpl...

thats probably the cheapest way

fiery turtle Feb 22, 2025, 11:02 AM

#

flat badge Linux maintainers will never merge a compat layer

im not sure why you're so certain about that

strong heath Feb 22, 2025, 11:02 AM

#

you would be if you were up to date on linux happenings

fiery turtle Feb 22, 2025, 11:03 AM

#

i am

#

this aint rust

#

and different subsystems have different maintainers

fiery turtle Feb 22, 2025, 11:03 AM

#

fiery turtle https://learn.microsoft.com/en-us/windows-hardware/drivers/ddi/acpiioct/ni-acpii...

nvm wrong link

flat badge Feb 22, 2025, 11:03 AM

#

Hardware abstraction layers regularly get rejected in linux

fiery turtle Feb 22, 2025, 11:04 AM

#

fiery turtle nvm wrong link

or no its correct i think(?)

torpid root Feb 22, 2025, 11:04 AM

#

uACPICA layer over uACPI :p

flat badge Feb 22, 2025, 11:04 AM

#

It's not too uncommon that some company wants to merge a driver that's portable across, say, freebsd and linux and it usually results in heated discussion

strong heath Feb 22, 2025, 11:04 AM

#

yeah im not necessarily talking about rust anything

fiery turtle Feb 22, 2025, 11:04 AM

#

flat badge It's not too uncommon that some company wants to merge a driver that's portable ...

hm yeah

#

i mean an acpica interface layer is also possible, most apis are translatable 1:1

strong heath Feb 22, 2025, 11:05 AM

#

I am talking about the amdgpus of the world being refused to be portable between the OSes so freebsd needs linux compat

flat badge Feb 22, 2025, 11:05 AM

#

flat badge It's not too uncommon that some company wants to merge a driver that's portable ...

For example drivers that wrap mem allocation or PCI into a compat layer regularly get NAKd

slim panther Feb 22, 2025, 11:05 AM

#

stupid

fiery turtle Feb 22, 2025, 11:05 AM

#

yeah its kinda stupid ngl

#

they dont want anyone other than linux to be viable

left orbit Feb 22, 2025, 11:05 AM

#

maybe it would be smart to reach out to a maintainer and see if it has a chance to pass?

#

idk

strong heath Feb 22, 2025, 11:06 AM

#

whatever you do, if you are serious about upstreaming it, you will need to talk it out anyways

torpid root Feb 22, 2025, 11:06 AM

#

@LinusTorvalds

strong heath Feb 22, 2025, 11:06 AM

#

so better start at that

fiery turtle Feb 22, 2025, 11:06 AM

#

left orbit maybe it would be smart to reach out to a maintainer and see if it has a chance ...

yeah

#

looks like these arent intel employees at least

left orbit Feb 22, 2025, 11:06 AM

#

hey hes polish

#

:^)

strong heath Feb 22, 2025, 11:07 AM

#

if you are also russian, so you want to clear that up first if you are employed by a russian company

#

unless you want to be in a phoronix article

fiery turtle Feb 22, 2025, 11:07 AM

#

left orbit hey hes polish

LETSFUCKINGGOOOOOO lol

fiery turtle Feb 22, 2025, 11:07 AM

#

strong heath if you are also russian, so you want to clear that up first if you are employed ...

unrelated, this is my personal project lol

left orbit Feb 22, 2025, 11:07 AM

#

or at least a very polish sounding last name lol

strong heath Feb 22, 2025, 11:07 AM

#

im just clearing items of the list

#

better safe than sorry

fiery turtle Feb 22, 2025, 11:07 AM

#

i think i've talked to him before as well

#

when sending some patches for acpica fixes

#

to the linux mailing list

left orbit Feb 22, 2025, 11:08 AM

#

so how do you usually reach these people?

strong heath Feb 22, 2025, 11:08 AM

#

mail

fiery turtle Feb 22, 2025, 11:08 AM

#

mail, yeah

left orbit Feb 22, 2025, 11:08 AM

#

like, literally send an email to one from above?

fiery turtle Feb 22, 2025, 11:09 AM

#

yup

left orbit Feb 22, 2025, 11:09 AM

#

one of the addresses

#

ah ok

fiery turtle Feb 22, 2025, 11:09 AM

#

thats how mailing lists work

left orbit Feb 22, 2025, 11:09 AM

#

fair lol

fiery turtle Feb 22, 2025, 11:09 AM

#

usually u Cc the mailing list as well, but for inqueries u dont have to

left orbit Feb 22, 2025, 11:10 AM

#

apparently the guy works at intel and lives in poland

#

from a quick google search lol

fiery turtle Feb 22, 2025, 11:10 AM

#

ah

strong heath Feb 22, 2025, 11:10 AM

#

full doxx in preparation of the talks

left orbit Feb 22, 2025, 11:10 AM

#

:^)

fiery turtle Feb 22, 2025, 11:10 AM

#

lol

strong heath Feb 22, 2025, 11:10 AM

#

bring pictures of him cheating on his wife

#

as blackmail

left orbit Feb 22, 2025, 11:11 AM

#

gotta have the upper hand /s

fiery turtle Feb 22, 2025, 11:12 AM

#

true

calm latch Feb 22, 2025, 11:18 AM

#

fiery turtle just call uacpi_vsnprintf

I like to live dangerously and use sprintf which I wrote myself with uACPI KEKW

fiery turtle Feb 22, 2025, 11:19 AM

#

more fake bug reports for me to figure out thanks trl

slim panther Feb 22, 2025, 11:20 AM

#

im guessing then that linux pulls in uacpi sources directly, right?

#

*would pull

fiery turtle Feb 22, 2025, 11:21 AM

#

yeah

calm latch Feb 22, 2025, 11:22 AM

#

just let everyone else (haiku, bsds, ros) switch to uacpi

#

then it's gonna be linux's problem that they're the only ones using acpica

slim panther Feb 22, 2025, 11:23 AM

#

calm latch just let everyone else (haiku, bsds, ros) switch to uacpi

(windows)

#

meme

calm latch Feb 22, 2025, 11:23 AM

#

trl

fiery turtle Feb 22, 2025, 11:23 AM

#

freebsd would be cool as well

strong heath Feb 22, 2025, 11:23 AM

#

what is a ros

fiery turtle Feb 22, 2025, 11:23 AM

#

maybe after haiku(?)

calm latch Feb 22, 2025, 11:23 AM

#

ReactOS

slim panther Feb 22, 2025, 11:23 AM

#

freebsd mlibc would be cool

strong heath Feb 22, 2025, 11:23 AM

#

reactos exists? damn

strong heath Feb 22, 2025, 11:23 AM

#

slim panther freebsd mlibc would be cool

no chance

#

dont push your luck

fiery turtle Feb 22, 2025, 11:23 AM

#

reactos is getting better

left orbit Feb 22, 2025, 11:23 AM

#

freebsd uacpi would be cool

slim panther Feb 22, 2025, 11:23 AM

#

why

strong heath Feb 22, 2025, 11:24 AM

#

because freebsd is all into using their own stuff

fiery turtle Feb 22, 2025, 11:24 AM

#

i wonder if freebsd also calls into acpica directly

calm latch Feb 22, 2025, 11:24 AM

#

fiery turtle reactos is getting better

I think they've said that uACPI has fixed a bunch of issues for them

fiery turtle Feb 22, 2025, 11:24 AM

#

yeah

strong heath Feb 22, 2025, 11:24 AM

#

they only dont use their own stuff when they have absolutely no other chance, like, using the AMD drivers

fiery turtle Feb 22, 2025, 11:24 AM

#

(or acpi)

slim panther Feb 22, 2025, 11:24 AM

#

meme

#

turns out writing gpu drivers is fucking painful

#

(who would have guessed)

fiery turtle Feb 22, 2025, 11:25 AM

#

although there is a BSD that rolled their own acpi

left orbit Feb 22, 2025, 11:25 AM

#

who would have thought

fiery turtle Feb 22, 2025, 11:25 AM

#

idk how well that went

calm latch Feb 22, 2025, 11:25 AM

#

slim panther <:meme:575445094589661215>

how's uDRM? ultrameme

slim panther Feb 22, 2025, 11:25 AM

#

slim panther turns out writing gpu drivers is fucking painful

.

left orbit Feb 22, 2025, 11:25 AM

#

fiery turtle although there is a BSD that rolled their own acpi

it’s apparently not that big of a task

slim panther Feb 22, 2025, 11:25 AM

#

meme

left orbit Feb 22, 2025, 11:25 AM

#

considering you did it solo and it’s better than acpica

fiery turtle Feb 22, 2025, 11:25 AM

#

true

#

qwinci did as well

left orbit Feb 22, 2025, 11:25 AM

#

but is it really that high of a bar :^)

slim panther Feb 22, 2025, 11:26 AM

#

and there's a lot of time and dedication that went into it

left orbit Feb 22, 2025, 11:26 AM

#

yeah ofc

fiery turtle Feb 22, 2025, 11:26 AM

#

it only took me 2 years trl

calm latch Feb 22, 2025, 11:26 AM

#

left orbit considering you did it solo and it’s better than acpica

but there are at least 18 operating systems that've been using/testing it

left orbit Feb 22, 2025, 11:26 AM

#

slim panther and there's a lot of time and dedication that went into it

not trying to discredit infy or qwinci here, dont get me wrong

strong heath Feb 22, 2025, 11:26 AM

#

you could concievably have the acpi guy on a 10-15 people project

#

just working on that

#

if I had those resources in Ironclad, I would maybe do that

#

because C vs Ada

fiery turtle Feb 22, 2025, 11:27 AM

#

strong heath you could concievably have the acpi guy on a 10-15 people project

wdym by that

left orbit Feb 22, 2025, 11:27 AM

#

formally verified aml interpreter

calm latch Feb 22, 2025, 11:27 AM

#

and then aml just does while (1) ;

strong heath Feb 22, 2025, 11:28 AM

#

fiery turtle wdym by that

its not a thing id have a 1 person out of a 5 person team do

#

but 1 out of 15 seems fine use of resources

fiery turtle Feb 22, 2025, 11:28 AM

#

ah

#

u should advertise how much u pay for ironclad features

#

people assume its probably like 10 euro

strong heath Feb 22, 2025, 11:28 AM

#

I plan on doing so

#

put some job posting, some bug bounties

fiery turtle Feb 22, 2025, 11:29 AM

#

yeah

strong heath Feb 22, 2025, 11:29 AM

#

I also got an announcement to do regarding that on some Ada events

#

to get some eyes on it

fiery turtle Feb 22, 2025, 11:29 AM

#

nice

fiery turtle Feb 22, 2025, 11:30 AM

#

calm latch but there are at least 18 operating systems that've been using/testing it

thats also important tbh, i dont have that amount of resources or computes to constantly test it

#

and find rare bugs etc

median crest Feb 22, 2025, 11:33 AM

#

Time to learn ada and put even more bugs in ironclad

#

So you can get even more payouts

#

Business grindset $$$$

strong heath Feb 22, 2025, 11:33 AM

#

C++ mindset

fiery turtle Feb 22, 2025, 11:33 AM

#

strong heath to get some eyes on it

btw how did u discover ada

strong heath Feb 22, 2025, 11:34 AM

#

I was one day bored and say in quora that american boats use Ada

#

I like boats

#

so I learned Ada

fiery turtle Feb 22, 2025, 11:34 AM

#

i had a similar experience but with planes

strong heath Feb 22, 2025, 11:35 AM

#

where is your Ada uACPI then

fiery turtle Feb 22, 2025, 11:35 AM

#

apparently all autopilot/fmc software is ada

strong heath Feb 22, 2025, 11:35 AM

#

yeah pretty much

fiery turtle Feb 22, 2025, 11:35 AM

#

naah im a c guy

left orbit Feb 22, 2025, 11:35 AM

#

based

strong heath Feb 22, 2025, 11:35 AM

#

one of us.... one of us.... one of us....

#

you will one day..... be one of us....

left orbit Feb 22, 2025, 11:35 AM

#

~~zig when~~

#

meme

strong heath Feb 22, 2025, 11:36 AM

#

never

fiery turtle Feb 22, 2025, 11:36 AM

#

ill give it a try i swear lol, ive been wanting to

left orbit Feb 22, 2025, 11:36 AM

#

been waiting so long nooo

#

i believe 🙏

median crest Feb 22, 2025, 11:39 AM

#

Write an os in lua because lua is brazilian tech so its the best

fiery turtle Feb 22, 2025, 11:39 AM

#

maybe ill contribute sometihng to your zig os eventually meme

strong heath Feb 22, 2025, 11:39 AM

#

median crest Write an os in lua because lua is brazilian tech so its the best

https://tenor.com/3OVz.gif

Tenor

left orbit Feb 22, 2025, 11:39 AM

#

yooo that’d be sick

fiery turtle Feb 22, 2025, 11:40 AM

#

letsgo

slim panther Feb 22, 2025, 11:40 AM

#

someone needs to pull through with a c# os

strong heath Feb 22, 2025, 11:40 AM

#

I had one

#

it was aight

left orbit Feb 22, 2025, 11:40 AM

#

i got a basic scheduler and dispatcher done yesterday :^) so maybe i will work on the io stack today

fiery turtle Feb 22, 2025, 11:40 AM

#

when new thread, when open source ultrameme

left orbit Feb 22, 2025, 11:40 AM

#

then a simple acpi and pci drivers for discovering devices

#

soon™️

#

i need a little reorganization of code

#

currently i call into hal for 90% of things that should be in ke lol

#

so it needs a little clean up

fiery turtle Feb 22, 2025, 11:42 AM

#

ill be waiting

median crest Feb 22, 2025, 11:43 AM

#

https://tenor.com/view/keaquireskinlock-gif-13675156356974872901

Tenor

left orbit Feb 22, 2025, 11:48 AM

#

oh thanks for reminding me i have to write the C api too

#

i want to be able to write drivers in C and C++

slim panther Feb 22, 2025, 11:49 AM

#

rust api when

#

actually

#

i should make a bindgen for my modules

fiery turtle Feb 22, 2025, 11:50 AM

#

left orbit i want to be able to write drivers in C and C++

daamn

#

thats cool

median crest Feb 22, 2025, 11:50 AM

#

left orbit i want to be able to write drivers in C and C++

C++ :(

slim panther Feb 22, 2025, 11:50 AM

#

C :)

left orbit Feb 22, 2025, 11:50 AM

#

i mean i get C++ for free if i have a C api

median crest Feb 22, 2025, 11:50 AM

#

The faces cancel each other so its just :

left orbit Feb 22, 2025, 11:50 AM

#

so like

#

i’m not gonna make a C++ api but you can still write C++ drivers

slim panther Feb 22, 2025, 11:52 AM

#

median crest The faces cancel each other so its just :

the c too

#

++ :

#

i should switch to meson so i can have native rust build support

mortal yoke Feb 22, 2025, 11:56 AM

#

I could probably write drivers in rust too, I bet there is some kind of wdk crate out there meme though one bad thing is that there is no native seh support in rust so gl if you pass some garbage pointer to some kernel function and it raises an exception

fiery turtle Feb 22, 2025, 11:56 AM

#

truue

slim panther Feb 22, 2025, 11:57 AM

#

mortal yoke I could probably write drivers in rust too, I bet there is some kind of wdk crat...

wdym

mortal yoke Feb 22, 2025, 12:18 PM

#

slim panther wdym

well actually most of the kernel functions don't generate any exceptions visible to the outside but you could have smth like ```rust
unsafe fn handle_device_control(device: *mut DEVICE_OBJECT, irp: *mut IRP) {
let slot = IoGetCurrentIrpStackLocation(irp);
if (*slot).RequestorMode == UserMode {
let addr = (*irp).AssociatedIrp.Systembuffer;
// generates STATUS_ACCESS_VIOLATION exception if a kernel mode address/unaligned address is passed
ProbeForRead(addr, 1, 0);
// can generate an exception too if the user address was unmapped/mprotected
(addr as *mut u8).write_volatile(1);
}
}

slim panther Feb 22, 2025, 12:19 PM

#

oh you mean in NT context

mortal yoke Feb 22, 2025, 12:19 PM

#

yes lol

slim panther Feb 22, 2025, 12:20 PM

#

i skipped over the wdk part

#

xd

fiery turtle Feb 22, 2025, 1:13 PM

#

just released uacpi 2.0

lofty dragon Feb 22, 2025, 1:16 PM

#

🎉

strong heath Feb 22, 2025, 1:17 PM

#

okay okay

vale isle Feb 22, 2025, 1:23 PM

#

we getting to chromium version numbers with this one

lofty dragon Feb 22, 2025, 1:23 PM

#

<3 semver

median crest Feb 22, 2025, 1:24 PM

#

The year is 2057. It is the year of the managarm desktop. Uacpi 327492.0.2 is out

fiery turtle Feb 22, 2025, 1:24 PM

#

lol

slim panther Feb 22, 2025, 1:24 PM

#

fiery turtle just released uacpi 2.0

that was fast

fiery turtle Feb 22, 2025, 1:25 PM

#

it's a breaking change technically so i have to bump the major number sad

vale isle Feb 22, 2025, 1:25 PM

#

y'all know how some projects go like 1.6, 1.7 and then drop the 1 and release 8.0

slim panther Feb 22, 2025, 1:25 PM

#

imagine mlibc did that 🤯

vale isle Feb 22, 2025, 1:25 PM

#

vale isle y'all know how some projects go like 1.6, 1.7 and then drop the 1 and release 8....

is it possible to do the reverse

#

like add a new major-major

strong heath Feb 22, 2025, 1:26 PM

#

fiery turtle it's a breaking change technically so i have to bump the major number <:sad:1291...

you do well

slim panther Feb 22, 2025, 1:26 PM

#

at least you do semver (based)

strong heath Feb 22, 2025, 1:26 PM

#

its better to have a 13.0.1 than having to change half of your code with a major release every 6 months

vale isle Feb 22, 2025, 1:26 PM

#

slim panther imagine mlibc did that 🤯

i thought mlibc followed semver but rarely made a release?

strong heath Feb 22, 2025, 1:26 PM

#

these people are haters

strong heath Feb 22, 2025, 1:26 PM

#

vale isle i thought mlibc followed semver but rarely made a release?

mlibc doesnt do releases, thats their brain worms

median crest Feb 22, 2025, 1:26 PM

#

I wouldnt know how to do semver properly for a project like an os

vale isle Feb 22, 2025, 1:27 PM

#

median crest I wouldnt know how to do semver properly for a project like an os

SCI and stuff shouldn't be semver, just don't make any breaking changes

strong heath Feb 22, 2025, 1:27 PM

#

I use semver for Ironclad, I am pre 1.0 though so it didnt really kick in

vale isle Feb 22, 2025, 1:27 PM

#

:clueless:

fiery turtle Feb 22, 2025, 1:28 PM

#

when ironclad 1.0

median crest Feb 22, 2025, 1:28 PM

#

Just call it alpha for all of infnity

vale isle Feb 22, 2025, 1:28 PM

#

what I plan on doing for my OS is pre-1.0 cargo-flavoured-semver

slim panther Feb 22, 2025, 1:28 PM

#

i mean at what point is a kernel really 1.0

vale isle Feb 22, 2025, 1:29 PM

#

vale isle what I plan on doing for my OS is pre-1.0 cargo-flavoured-semver

until I can call a release 1.0 and then "ratified" interfaces are frozen but new interfaces will have an experiment stage or so

median crest Feb 22, 2025, 1:30 PM

#

slim panther i mean at what point is a kernel really 1.0

When you can play a full 100 hour run of factorio in it without it crashing once

strong heath Feb 22, 2025, 1:30 PM

#

fiery turtle when ironclad 1.0

whenever someone pays me 13 quintillion gazzidollars

slim panther Feb 22, 2025, 1:30 PM

#

is factorio portable

strong heath Feb 22, 2025, 1:30 PM

#

I honestly dont make releases because I am lazy

slim panther Feb 22, 2025, 1:30 PM

#

i mean the wine version should be

vale isle Feb 22, 2025, 1:30 PM

#

slim panther is factorio portable

wine is

median crest Feb 22, 2025, 1:30 PM

#

slim panther is factorio portable

You can wine or do glibc shim like qwinci did

#

Managarm ran factorio

slim panther Feb 22, 2025, 1:30 PM

#

omega zased

median crest Feb 22, 2025, 1:31 PM

#

I tried running factorio in astral but it segfaults in wine and the glibc shim

#

💔

fiery turtle Feb 22, 2025, 1:31 PM

#

fix it 😡

#

i didnt compile the entire astral for no reason

median crest Feb 22, 2025, 1:31 PM

#

I give you the factorio zip, instructions to enable syscall logging and you do it meme

fiery turtle Feb 22, 2025, 1:32 PM

#

damn

median crest Feb 22, 2025, 1:32 PM

#

I like that they give drm free versions if you own the game

#

I was trying to get wow64 working on astral but wine seems to not look in /lib/wine for the dlls for some fuckin reason

fiery turtle Feb 22, 2025, 1:33 PM

#

whats next on the list for astral atm

hollow elm Feb 22, 2025, 1:35 PM

#

astral automated teller machine

slim panther Feb 22, 2025, 1:35 PM

#

steam

vale isle Feb 22, 2025, 1:35 PM

#

Versions are considered compatible if their left-most non-zero major/minor/patch component is the same. This is different from SemVer which considers all pre-1.0.0 packages to be incompatible.

cargo-flavoured semver is the best

median crest Feb 22, 2025, 1:35 PM

#

fiery turtle whats next on the list for astral atm

Rework scheduler, rework irq system, rework drivers, fix wine, idk

lofty dragon Feb 22, 2025, 1:36 PM

#

slim panther imagine mlibc did that 🤯

imagine if i didn't have to fight tooth and nail to have every single reasonable change merged...

fiery turtle Feb 22, 2025, 1:36 PM

#

median crest Rework scheduler, rework irq system, rework drivers, fix wine, idk

i personally like the last point

slim panther Feb 22, 2025, 1:36 PM

#

idk

vale isle Feb 22, 2025, 1:37 PM

#

hollow elm astral automated teller machine

ASTRAL stands for automated static teller real-mode advanced libre

#

ok it needs a noun

median crest Feb 22, 2025, 1:37 PM

#

Time to shill astral for the higher ups and have it run on the atms

vale isle Feb 22, 2025, 1:37 PM

#

what's a noun that starts with l

#

related to finance

fiery turtle Feb 22, 2025, 1:38 PM

#

median crest Time to shill astral for the higher ups and have it run on the atms

imagine when they see "address is not kernec"

#

this is how u know its top quality software

lofty dragon Feb 22, 2025, 1:39 PM

#

lol

median crest Feb 22, 2025, 1:42 PM

#

fiery turtle imagine when they see "address is not kernec"

ç

dire owl Feb 22, 2025, 2:24 PM

#

vale isle what's a noun that starts with l

loss

#

liability

vale isle Feb 22, 2025, 2:37 PM

#

lol

dense steppe Feb 22, 2025, 2:43 PM

#

leverage

severe swift Feb 22, 2025, 3:44 PM

#

leveraged to my personal risk tolerance

#

(x125 lever on bitcoin shorts on binance)

flat badge Feb 22, 2025, 4:13 PM

#

median crest I tried running factorio in astral but it segfaults in wine and the glibc shim

What's the glibc shim?

mortal yoke Feb 22, 2025, 4:22 PM

#

flat badge What's the glibc shim?

the preloadable lib that I made that contains aliases for some functions and overwrites some others like a lot of the pthread functions to make them usable from both libraries compiled against mlibc and from an executable linked against glibc

#

https://github.com/Qwinci/mlibc-glibc-compat

flat badge Feb 22, 2025, 4:29 PM

#

Why is that needed?

#

To run closed source linux programs?

mortal yoke Feb 22, 2025, 4:30 PM

#

yes

slim panther Feb 22, 2025, 4:30 PM

#

games probably

mortal yoke Feb 22, 2025, 4:36 PM

#

the thing that I like the most about that thing is the logic used to determine whether the caller was linked against glibc or not troll it uses dladdr1 to get the dynamic section of the caller, gets the symbol versioning information from there and looks if any symbol begins with GLIBC_ (and then if it finds symbol like that and the caller uses pthread_mutex_init etc a special flag in the mutex is set to indicate that the mutex was created from a glibc expecting executable so the other code like pthread_mutex_lock knows to use a different struct layout)

frank canopy Feb 22, 2025, 4:38 PM

#

fiery turtle maybe ill contribute sometihng to your zig os eventually <:meme:5754450945896612...

you could contribute to my zig os instead KEKW

fiery turtle Feb 22, 2025, 4:38 PM

#

true

frank canopy Feb 22, 2025, 4:39 PM

#

ive not been doing irp functions until i need them so rn the only thing drivers can do is query properties but i do have a fully working driver stack complete with enumeration and even async completion support

flat badge Feb 22, 2025, 4:40 PM

#

mortal yoke the thing that I like the most about that thing is the logic used to determine w...

How does it work? If it detects glibc, it provides alternative implementations for some functions?

mortal yoke Feb 22, 2025, 4:43 PM

#

well right now it just unconditionally overrides a lot of the pthread functions and the impl for them is basically copied from mlibc with some small differences to make it handle eg. the mutex created by glibc app but locked by some library case

frank canopy Feb 22, 2025, 4:44 PM

#

anyway time to look at the uacpi 2.0 stuff and see what i can yeet and what changes i can make

#

so defining uacpi_use_builtin_string tells it to not assume builtins?

#

if my reading this is right?

fiery turtle Feb 22, 2025, 4:46 PM

#

frank canopy so defining uacpi_use_builtin_string tells it to not assume builtins?

basically tells it to use own impl for memcpy etc if you dont have your overrides

frank canopy Feb 22, 2025, 4:46 PM

#

ok

#

weird name for the define but sure

#

the naming confusion between builtin (a thing what starts with __builtin and comes from compiler) and builtin (a thing what is built into uacpi) is confusing

calm latch Feb 22, 2025, 5:04 PM

#

I think I kinda have an ACPI bus discovery mechanism rust

frank canopy Feb 22, 2025, 5:04 PM

#

nice

calm latch Feb 22, 2025, 5:05 PM

#

(now I just need to implement deserialization in C++)

calm latch Feb 22, 2025, 5:08 PM

#

calm latch I think I kinda have an ACPI bus discovery mechanism <:rust:740835909964136479>

Also, why does _UID in info suspiciously looks like _MLS?

calm latch Feb 22, 2025, 5:09 PM

#

calm latch I think I kinda have an ACPI bus discovery mechanism <:rust:740835909964136479>

(This is C server publishing to Rust server trl )

frank canopy Feb 22, 2025, 5:09 PM

#

uid is context dependent i think?

#

but yeah tbh i can see why uid would be like the english version of mls for some things

#

only requirement for uid is it must be unique across the same HID/CIDs and must persist across reboots

calm latch Feb 22, 2025, 5:11 PM

#

But it's the same between multiple objects?

frank canopy Feb 22, 2025, 5:12 PM

#

yeah thats weird

#

" The _UID must be unique across all devices with either a common _HID or _CID. This is because a device needs to be uniquely identified to the OSPM, which may match on either a _HID or a _CID to identify the device."

#

so idk

#

the pci interrupt link devices are incomprehensible to me and those are the ones im seeing have repeats though

#

so idk on there

#

(also MLS has multiple languages to it so it should be localizable. idk how to access that info though and my kernel doesnt do localization at this point anyway)

fiery turtle Feb 22, 2025, 5:17 PM

#

frank canopy the naming confusion between builtin (a thing what starts with `__builtin` and c...

Yeah I know, dunno how to make it better and it doesnt matter because there's config.h that explains all options

fiery turtle Feb 22, 2025, 5:17 PM

#

calm latch I think I kinda have an ACPI bus discovery mechanism <:rust:740835909964136479>

LETSFUCKINGGOOOOOO

calm latch Feb 22, 2025, 5:17 PM

#

(It only publishes devices for now)

#

Like now I just gotta write the init server

frank canopy Feb 22, 2025, 5:17 PM

#

fiery turtle Yeah I know, dunno how to make it better and it doesnt matter because there's co...

the config.h had the same exact confusion to it and it took me multiple minutes to realise the docs werent wrong because of that

fiery turtle Feb 22, 2025, 5:18 PM

#

What part specifically?

frank canopy Feb 22, 2025, 5:18 PM

#

"Makes uACPI use the builtin versions of mem{cpy,move,set,cmp} instead of relying on the host to provide them." implies it should be using the __builtin if the flag is set to me

#

because thats what the phrase "builtin versions" would be expected to mean, no?

#

and then the end of the docstring on that define says that the compiler might depend on them being available even if the thing is defined which feels like a contradiction

#

how that doc comment reads to me is saying if its defined then uacpi will use the __builtin_memxxx but be aware the compiler might generate dependencies on them even if you define it (which is implied to make uacpi depend on them explicitly so of course theres a dependency on the __builtin_memxxx)

fiery turtle Feb 22, 2025, 5:21 PM

#

frank canopy "Makes uACPI use the builtin versions of mem{cpy,move,set,cmp} instead of relyin...

That would make no sense, since it should really be detecting if those exist itself

#

Uacpi doesnt have any config options that are compiler dependent

frank canopy Feb 22, 2025, 5:22 PM

#

i just dont see how that first sentence implies anything other than forcing the use of __builtin_memxxx

#

which is the opposite of what the define does

fiery turtle Feb 22, 2025, 5:26 PM

#

Built-in to uacpi, it doesnt say “compiler intrinsic” anywhere tho

#

Im open to suggestions for better wording tho

frank canopy Feb 22, 2025, 5:27 PM

#

tbh if the intrinsic didnt have builtin in the name itd be less confusing

#

id call it "default implementation" or "uacpi-provided" or something like that personally

#

uacpi-provided is my instinct

#

at least for in the doc string

#

can keep the define name for version compat if the docs make it clear

fiery turtle Feb 22, 2025, 5:30 PM

#

Yeah thx

vale isle Feb 22, 2025, 5:40 PM

#

calm latch I think I kinda have an ACPI bus discovery mechanism <:rust:740835909964136479>

yay!!

#

is this rust?

vale isle Feb 22, 2025, 5:40 PM

#

calm latch (This is C server publishing to Rust server <:trl:829316641191886888>)

ah, so that explains why the formatting looks like rust's Debug

#

it's deserialized in rust?

calm latch Feb 22, 2025, 5:59 PM

#

yes

north holly Feb 22, 2025, 8:13 PM

#

I will update my uacpi one day

fiery turtle Feb 22, 2025, 8:17 PM

#

do it two day

keen shard Feb 22, 2025, 9:05 PM

#

fiery turtle do it two day

How do you test uACPI? Do you have your own operating system kernel you use it with or do you use some common unix distribution

fiery turtle Feb 22, 2025, 9:05 PM

#

see tests/

#

i have a userspace port of uacpi

#

that i run against the test suite as well as real hardware blobs

keen shard Feb 22, 2025, 9:10 PM

#

fiery turtle that i run against the test suite as well as real hardware blobs

I see, That’s dope

#

Does the user space port do all the same stuff just using system calls or how does that work

fiery turtle Feb 22, 2025, 9:14 PM

#

keen shard Does the user space port do all the same stuff just using system calls or how do...

see for yourself: https://github.com/uACPI/uACPI/blob/master/tests/runner/interface_impl.cpp

keen shard Feb 22, 2025, 9:28 PM

#

fiery turtle see for yourself: https://github.com/uACPI/uACPI/blob/master/tests/runner/interf...

Dang bro I’ve been looking around the repo and I wanna say this is one of the most robust open source software projects I’ve seen in years, congratulations

fiery turtle Feb 22, 2025, 9:31 PM

#

thank you

keen shard Feb 22, 2025, 9:32 PM

#

fiery turtle thank you

What was your main motivation? Lack of non bullshit ACPI implementations?

fiery turtle Feb 22, 2025, 9:32 PM

#

i wanted a fully NIH OS, and ACPI was going to be a big part of that

#

when i started out i didnt know anything about ACPICA's quality

#

i thought it was good since everyone used it

keen shard Feb 22, 2025, 9:40 PM

#

fiery turtle i thought it was good since everyone used it

ACPICA is actually a joke

#

But that’s fire bro keep it up

#

I’m gonna port it to mine soon

#

I’ve gotten almost all of the deps done

#

I’ll let you know when I do 💯💯💯

fiery turtle Feb 22, 2025, 9:42 PM

#

letsgo!

keen shard Feb 22, 2025, 9:57 PM

#

fiery turtle letsgo!

meme meme

north holly Feb 22, 2025, 10:54 PM

#

fiery turtle that i run against the test suite as well as real hardware blobs

then you ping oberrow meme

keen shard Feb 22, 2025, 10:57 PM

#

north holly then you ping oberrow <:meme:575445094589661215>

Then you die because you forgot to put a lock ultrameme

north holly Feb 22, 2025, 10:58 PM

#

wrong kernel

#

it dies because of obos regression #844868385737

#

(of the day)

keen shard Feb 22, 2025, 10:59 PM

#

Yeah I just get the

Wel%co[*{%]*]e to FlqodidococcOS 1.02948293738;8.29382Segmentatjon fault at 0x02938382912728sjskxjdnxiekenjabcdefihatemlifeeucidnejxidodoeen

fiery turtle Feb 22, 2025, 11:12 PM

#

north holly then you ping oberrow <:meme:575445094589661215>

true

dense steppe Feb 23, 2025, 4:14 PM

#

How is uacpi_kernel_reset_event() used, and is it guaranteed that there is no contention on the semaphore when it is called?

fiery turtle Feb 23, 2025, 4:15 PM

#

ResetOp from AML

#

no guarantees

dense steppe Feb 23, 2025, 4:15 PM

#

shit

#

that means I need to actually do stuff, then?

#

should I wake up waiters and tell them the acquire failed?

fiery turtle Feb 23, 2025, 4:16 PM

#

i wouldnt

#

but tbh id just see what linux does

dense steppe Feb 23, 2025, 4:16 PM

#

yeah, and it'd overcomplicate stuff

dense steppe Feb 23, 2025, 4:16 PM

#

fiery turtle but tbh id just see what linux does

just looked at uacpi-linux

#

and what it does is much easier

fiery turtle Feb 23, 2025, 4:17 PM

#

what does it do?

dense steppe Feb 23, 2025, 4:17 PM

#

basically this:

while (trylock (semaphore))
        ;

fiery turtle Feb 23, 2025, 4:17 PM

#

thats smart

dense steppe Feb 23, 2025, 4:18 PM

#

for me, sema_lock_timeout (sema, 0) is equivalent to trylock

#

so I can just do that instead

fiery turtle Feb 23, 2025, 4:18 PM

#

yeah

dense steppe Feb 23, 2025, 4:21 PM

#

I just discovered a bug in my acpi mutex implementation

fiery turtle Feb 23, 2025, 4:21 PM

#

almost makes me think reset should be implemented internally be uacpi

dense steppe Feb 23, 2025, 4:21 PM

#

do you see the bug?

fiery turtle Feb 23, 2025, 4:22 PM

#

overflow?

dense steppe Feb 23, 2025, 4:22 PM

#

no

#

it doesn't handle 0xffff

fiery turtle Feb 23, 2025, 4:22 PM

#

ah right

#

ill get rid of uacpi_kernel_reset_event and just have a loop of uacpi_kernel_wait_for_event(event, 0)

#

how did i not think of that implementation

#

well i didnt because i basically stole acpica's api for events sort of

dense steppe Feb 23, 2025, 4:24 PM

#

fiery turtle ill get rid of `uacpi_kernel_reset_event` and just have a loop of `uacpi_kernel_...

why? someone might have a faster sema_slurp or something which is basically

void
sema_slurp (struct semaphore *sema)
{
        spin_lock (&sema->lock);
        sema->count = 0;
        spin_unlock (&sema->lock);
}

fiery turtle Feb 23, 2025, 4:24 PM

#

not sure it matters

dense steppe Feb 23, 2025, 4:24 PM

#

yeah, it probably doesn't

#

anyway

#

why are the return types of uacpi_kernel_acquire_mutex and uacpi_kernel_wait_for_event inconsistent?

#

the former returns a uacpi_status and the latter returns a uacpi_bool indicating success

#

it just seems strange

fiery turtle Feb 23, 2025, 4:26 PM

#

good question

#

but tldr is there's nothing uacpi can currently do with an error for wait

#

since events are mostly used for aml

dense steppe Feb 23, 2025, 4:27 PM

#

makes sense

#

yeah

fiery turtle Feb 23, 2025, 4:27 PM

#

aborting is probably worse than just not entering the branch that does Wait(...) in aml

dense steppe Feb 23, 2025, 4:28 PM

#

there's going to be like zero use of semaphores in my kernel besides ACPI that I can think of now

#

lol

fiery turtle Feb 23, 2025, 4:28 PM

#

but yeah it was a mistake not to convert events alongside mutexes

#

i might do it later for 3.0

fiery turtle Feb 23, 2025, 4:28 PM

#

dense steppe there's going to be like zero use of semaphores in my kernel besides ACPI that I...

i think same goes for linux

#

not many users of it

dense steppe Feb 23, 2025, 4:31 PM

#

like everything is either plain mutex or rwmutex

mortal yoke Feb 23, 2025, 4:31 PM

#

I mean you can use them to implement mutexes

#

uacpi_status uacpi_kernel_acquire_mutex(uacpi_handle mutex, uacpi_u16 timeout) {
    i64 timeout_value = static_cast<i64>(timeout * (NS_IN_MS / 100));
    auto status = KeWaitForSingleObject(
        mutex,
        KWAIT_REASON::Executive,
        KernelMode,
        false,
        timeout != 0xFFFF ? &timeout_value : nullptr);
    return status == STATUS_TIMEOUT ? UACPI_STATUS_TIMEOUT : UACPI_STATUS_OK;
}
void uacpi_kernel_release_mutex(uacpi_handle mutex) {
    auto semaphore = static_cast<KSEMAPHORE*>(mutex);
    KeReleaseSemaphore(semaphore, 0, 1, false);
}
uacpi_bool uacpi_kernel_wait_for_event(uacpi_handle event, uacpi_u16 timeout) {
    return uacpi_kernel_acquire_mutex(event, timeout) == UACPI_STATUS_OK;
}
void uacpi_kernel_signal_event(uacpi_handle event) {
    auto semaphore = static_cast<KSEMAPHORE*>(event);
    KeReleaseSemaphore(semaphore, 0, 1, false);
}
void uacpi_kernel_reset_event(uacpi_handle event) {
    auto semaphore = static_cast<KSEMAPHORE*>(event);
    KeInitializeSemaphore(semaphore, 0, INT32_MAX);
}

dense steppe Feb 23, 2025, 4:32 PM

#

mortal yoke I mean you can use them to implement mutexes

sure, but then you lose out on this fastpath:

torpid root Feb 23, 2025, 4:32 PM

#

initialising uacpi on a thread instead of before the scheduler results in a few hundred thousand points more on average

#

only on uefi though, on bios the performance is affected negatively

#

well

#

I'm getting 4.5 million consistently on my cpu

dense steppe Feb 23, 2025, 4:34 PM

#

and your CPU is?

fiery turtle Feb 23, 2025, 4:35 PM

#

torpid root only on uefi though, on bios the performance is affected negatively

that doesnt make much sense ngl

#

since aml is the same

torpid root Feb 23, 2025, 4:35 PM

#

dense steppe and your CPU is?

ryzen 5 pro 4650g

torpid root Feb 23, 2025, 4:35 PM

#

fiery turtle that doesnt make much sense ngl

yeah I don't know what's happening

mortal yoke Feb 23, 2025, 4:37 PM

#

dense steppe sure, but then you lose out on this fastpath:

yeah maybe, though a semaphore acquire isn't that costly either as it can also just be a single cmpxchg

torpid root Feb 23, 2025, 4:37 PM

#

torpid root initialising uacpi on a thread instead of before the scheduler results in a few ...

oh now it's almost the same

#

I'm really cursed

fiery turtle Feb 23, 2025, 4:38 PM

#

we really need a good automated test runner trl

north holly Feb 23, 2025, 4:38 PM

#

obos is basically an automated test runner

torpid root Feb 23, 2025, 4:38 PM

#

run - triple fault - run - triple fault - ...

#

meme

north holly Feb 23, 2025, 4:39 PM

#

what does iloblix do again?

torpid root Feb 23, 2025, 4:39 PM

#

nooo

#

so true

north holly Feb 23, 2025, 4:39 PM

#

you'll get there one day

#

right now I forgor how file reading works

median crest Feb 23, 2025, 4:45 PM

#

fiery turtle we really need a good automated test runner <:trl:829316641191886888>

ukernel time

fiery turtle Feb 23, 2025, 4:46 PM

#

i mean a piece of hardware

strong heath Feb 23, 2025, 4:47 PM

#

Use a watchdog and any existing hardware

hollow elm Feb 23, 2025, 4:51 PM

#

picokvm with it wired up to press the power and reset buttons

flat badge Feb 23, 2025, 4:56 PM

#

A test runner is not very hard to wire up

#

I have a bpi-f3 wired up as a test runner for RV

#

I just use a smart wall plug to turn power on/off

#

tftp boot + uart wired to a different machine

calm latch Feb 23, 2025, 5:17 PM

#

We need a leaderboard for RISC-V trl

fiery turtle Feb 23, 2025, 5:20 PM

#

risc-v qemu acpi is like 10 opcodes

kind mantle Feb 23, 2025, 5:22 PM

#

lmao

flat badge Feb 23, 2025, 5:35 PM

#

The board doesn't have acpi

#

I use it to test managarm, not uacpi

#

But the same approach would also work for uacpi on x86

calm latch Feb 23, 2025, 8:26 PM

#

Do ARM and RISC-V have sleep states as well?

fiery turtle Feb 23, 2025, 8:26 PM

#

yes

flat badge Feb 23, 2025, 8:29 PM

#

RISC-V can sleep using SBI

#

I'm not sure how this is wired up to ACPI

#

it'd help if they released a proper acpi spec with risc-v support in it meme

calm latch Feb 23, 2025, 8:32 PM

#

there are ECRs in github issues pointing to google drive links trl

kindred beacon Feb 23, 2025, 8:33 PM

#

last time I checked the links were dead

#

this does not include RISC-V shit tho??????????????????????

#

bruh

#

waiting for 6.6 then

#

https://tenor.com/view/caveira-skeleton-ive-been-waiting-gif-13523984

Tenor

calm latch Feb 23, 2025, 8:37 PM

#

I have these

📎 ACPI_RHCT_RISC-V_ECR_B2.pdf 📎 ACPI_MADT_RISC-V_PLIC_AIA_ECR_B2.pdf 📎 ACPI_RHCT_RISC-V_ECR.pdf

#

Should've saved the rest

#

Is it even worth supporting 64 bit wakeup vector memes?

kindred beacon Feb 23, 2025, 8:41 PM

#

calm latch I have these

based

calm latch Feb 23, 2025, 8:41 PM

#

Like they were not dead a year ago

fiery turtle Feb 23, 2025, 8:42 PM

#

calm latch Is it even worth supporting 64 bit wakeup vector memes?

nah

calm latch Feb 23, 2025, 8:44 PM

#

So like just having the 16 bit mode one is fine?

fiery turtle Feb 23, 2025, 8:45 PM

#

the 64-bit one is protected mode

#

and idk if anyhing supports it

calm latch Feb 23, 2025, 8:45 PM

#

Does everyone folow this?

Platform runtime firmware restores the pre-sleep configuration or initial boot configuration of
each CPU (MSR, MTRR, firmware update, SMBase, and so on). Interrupts must be disabled (for IA-32 processors, disabled by CLI instruction).

#

Wait, initial boot

#

Still

#

I think someone was saving MTRRs here?

fiery turtle Feb 23, 2025, 8:46 PM

#

i wouldnt rely on it

calm latch Feb 23, 2025, 8:48 PM

#

Ok the spec is very short and looks very easy

#

(famous last words)

fiery turtle Feb 23, 2025, 8:48 PM

#

did u read the stuff about _PRW etc

#

oberrow will know more

calm latch Feb 23, 2025, 8:49 PM

#

I've read chapter 16

#

Which is about _TTS, _PRW, _PTS and _WAK

north holly Feb 23, 2025, 9:07 PM

#

idk what _TTS does

#

_WAK is handled by uacpi_prepare_for_wake_from_sleep_state

#

_PTS is from uacpi_prepare_for_sleep_state

#

_PRW is fnuy

#

let me go back to my suspend code to see what it did again

#

ok yeah so PRW basically tells you the power resources needed to be enabled for the device to wake u

#

and also tells the deepest state the device can wake you frame

#

as well as the gpe index that you should use uacpi_setup_gpe_for_wake on

#

btw it mentions something about gpe block devices, ignore that

#

infy said a while back they don't exist

north holly Feb 23, 2025, 9:10 PM

#

north holly as well as the gpe index that you should use `uacpi_setup_gpe_for_wake` on

btw before doing that you should evaluate _DSW if it exists, or _PSW if _DSW does not exist

#

if _PSW doesn't exist, obos ignores the error and continues

#

after doing that you can simply put the device into the deepest state supported for wake from _Sn

#

(there was a chart for that in the acpi spec)

calm latch Feb 23, 2025, 9:11 PM

#

north holly ok yeah so PRW basically tells you the power resources needed to be enabled for ...

Do I need them?

north holly Feb 23, 2025, 9:12 PM

#

I just call _ON on each one

#

but technically they should all be enabled before suspend, unless manually disabled

#

but I enable them manually anyway

calm latch Feb 23, 2025, 9:12 PM

#

What do I do with TTS?

north holly Feb 23, 2025, 9:13 PM

#

never heard of it

calm latch Feb 23, 2025, 9:13 PM

#

16.1.6 Transitioning from the Working to the Sleeping State
On a transition of the system from the working to the sleeping state, the following occurs:

OSPM decides (through a policy scheme) to place the system into the sleeping state.

OSPM invokes the _TTS method to indicate the deepest possible system state the system will transition to (1, 2,
3, or 4 representing S1, S2, S3, and S4)

north holly Feb 23, 2025, 9:13 PM

#

ignore

#

if you want to go to _S3

#

you check if S3 exists

calm latch Feb 23, 2025, 9:13 PM

#

Control method used to prepare to sleep and run once awakened

north holly Feb 23, 2025, 9:14 PM

#

uacpi probably does that for you

#

in prepare for sleep state

#

cc @fiery turtle

calm latch Feb 23, 2025, 9:15 PM

#

https://uefi.org/htmlspecs/ACPI_Spec_6_4_html/07_Power_and_Performance_Mgmt/oem-supplied-system-level-control-methods.html

north holly Feb 23, 2025, 9:15 PM

#

yes I see it

fiery turtle Feb 23, 2025, 9:15 PM

#

calm latch > 16.1.6 Transitioning from the Working to the Sleeping State > On a transition ...

it probably means _SST, and yeah uacpi calls it

north holly Feb 23, 2025, 9:16 PM

#

hmm it says _TTS

#

#

bruh

calm latch Feb 23, 2025, 9:17 PM

#

9.2.1 _SST (System Status)
This optional object is a control method that OSPM invokes to set the system status indicator as desired

calm latch Feb 23, 2025, 9:17 PM

#

north holly

bruh

fiery turtle Feb 23, 2025, 9:17 PM

#

https://elixir.bootlin.com/linux/v6.13.4/source/drivers/acpi/sleep.c#L36

sleep.c - drivers/acpi/sleep.c - Linux source code v...

Elixir Cross Referencer - source code of Linux v6.13.4: drivers/acpi/sleep.c

#

linux calls it on its own

north holly Feb 23, 2025, 9:18 PM

#

oops

calm latch Feb 23, 2025, 9:18 PM

#

inb4 seabios bug was obos bug

north holly Feb 23, 2025, 9:18 PM

#

no it is a seabios bug

#

crescent also works with my patch

fiery turtle Feb 23, 2025, 9:19 PM

#

that said obos did die on some real hw

north holly Feb 23, 2025, 9:19 PM

#

said hardware doesn't have _TTS

fiery turtle Feb 23, 2025, 9:20 PM

#

out of 500 blobs only macbook pro 12/14 have it

calm latch Feb 23, 2025, 9:21 PM

#

So basically,

I have to implement wakeup vector
I have to call a bunch of ACPI methods
Profit?

fiery turtle Feb 23, 2025, 9:21 PM

#

❯ grep -rn '_TTS'
Notebook/Apple/MacBookPro12/MacBookPro12,1/A236F4DD9072:9015:    Method (_TTS, 1, NotSerialized)  // _TTS: Transition To State
Notebook/Apple/MacBookPro12/MacBookPro12,1/A236F4DD9072.bin:2676:    79A0: 5F 12 08 03 0A 07 0A 07 0A 00 14 0C 5F 54 54 53  _..........._TTS
Notebook/Apple/MacBookPro14/MacBookPro14,3/91077C25D705:10177:    Method (_TTS, 1, NotSerialized)  // _TTS: Transition To State

north holly Feb 23, 2025, 9:21 PM

#

north holly (there was a chart for that in the acpi spec)

oh btw you do all that for each wake capable device you wish to enable

north holly Feb 23, 2025, 9:21 PM

#

calm latch So basically, 1. I have to implement wakeup vector 2. I have to call a bunch of ...

wkaeup vector is literally just

#

smp trampoline

calm latch Feb 23, 2025, 9:21 PM

#

fiery turtle ``` ❯ grep -rn '_TTS' Notebook/Apple/MacBookPro12/MacBookPro12,1/A236F4DD9072:90...

I think 12,1 is a model number

calm latch Feb 23, 2025, 9:21 PM

#

north holly smp trampoline

I don't have smp trampoline

north holly Feb 23, 2025, 9:22 PM

#

ah well

calm latch Feb 23, 2025, 9:22 PM

#

And I need some system to park cores?

fiery turtle Feb 23, 2025, 9:22 PM

#

people raised by limine

north holly Feb 23, 2025, 9:22 PM

#

north holly oh btw you do all that for each wake capable device you wish to enable

and you do it before sleep, of course

fiery turtle Feb 23, 2025, 9:22 PM

#

dont have smp startup code

#

L

north holly Feb 23, 2025, 9:22 PM

#

L

calm latch Feb 23, 2025, 9:22 PM

#

L

north holly Feb 23, 2025, 9:22 PM

#

smp trampolines are always fun to write

calm latch Feb 23, 2025, 9:22 PM

#

My pmm has memory zones though

north holly Feb 23, 2025, 9:22 PM

#

idk what that does

calm latch Feb 23, 2025, 9:22 PM

#

So it shouldn't be hard to allocate memory below 1MB

north holly Feb 23, 2025, 9:23 PM

#

I just hardcode the smp trampoline at 0x1000 meme

#

(btw doing it at phys. 0 will not work in any case)

calm latch Feb 23, 2025, 9:23 PM

#

No, this is too boring

#

I need to dynamically allocate it

north holly Feb 23, 2025, 9:23 PM

#

north holly (btw doing it at phys. 0 will not work in any case)

((if you workaround the bios not letting you, it'll just cause you to triple fault because you'd have overwritten the IVT))

calm latch Feb 23, 2025, 9:24 PM

#

And patch the code to load the right page table top level pointer

#

Into cr3

#

Anyway

north holly Feb 23, 2025, 9:25 PM

#

other than acpi stuff, you should probably get a way to tell your drivers that you are going into suspend

calm latch Feb 23, 2025, 9:25 PM

#

I can then just pagefault to success

fiery turtle Feb 23, 2025, 9:25 PM

#

calm latch So basically, 1. I have to implement wakeup vector 2. I have to call a bunch of ...

also park other cores under cli;hlt

calm latch Feb 23, 2025, 9:25 PM

#

fiery turtle also park other cores under cli;hlt

No, but then how do I wake them up

north holly Feb 23, 2025, 9:25 PM

#

north holly other than acpi stuff, you should probably get a way to tell your drivers that y...

and vice versa

calm latch Feb 23, 2025, 9:25 PM

#

Like this needs to be a syscall

north holly Feb 23, 2025, 9:25 PM

#

make a worker thread, bind it to one core

#

syscall to park other cores

calm latch Feb 23, 2025, 9:26 PM

#

Because I will not do @ monkous memes and I will let uACPI enter S3 from userspace instead

north holly Feb 23, 2025, 9:26 PM

#

north holly syscall to park other cores

and continue doing shit

calm latch Feb 23, 2025, 9:27 PM

#

calm latch Because I will not do @ monkous memes and I will let uACPI enter S3 from userspa...

1 stack per CPU should help with wakeup

#

So like the only thing I need from kernel is a syscall to flush caches and a wakeup vector?

north holly Feb 23, 2025, 9:30 PM

#

btw here is a list of what you need to save before S3 (that you might forget about) on x86-64:

HPET status
LAPIC status
IOAPIC status
PCI status
ACPI NVS ram regions (some devices don't need to / cannot do it, so add a kernel command line option to enable/disable this)
and that should be it

north holly Feb 23, 2025, 9:30 PM

#

calm latch So like the only thing I need from kernel is a syscall to flush caches and a wak...

pretty much

calm latch Feb 23, 2025, 9:30 PM

#

north holly btw here is a list of what you need to save before S3 (that you might forget abo...

What's PCI status?

north holly Feb 23, 2025, 9:30 PM

#

and also you will need to wake other CPUs again

north holly Feb 23, 2025, 9:30 PM

#

calm latch What's PCI status?

literally anything in the PCI config space

calm latch Feb 23, 2025, 9:31 PM

#

bruh

#

Shouldn't be hard though

north holly Feb 23, 2025, 9:31 PM

#

nah

#

(it only resulted in a full PCI interface rewrite for me)

#

((nothing big))

#

btw unless you have my patched seabios, if you use PCIe, it will be broken

#

after wake from suspend

calm latch Feb 23, 2025, 9:32 PM

#

I will see

#

My ~~kernel~~ devicesd server already keeps the list of PCI devices

#

Apparently

north holly Feb 23, 2025, 9:33 PM

#

what about BARs

#

command register

#

MSI

#

any IRQ pin status

calm latch Feb 23, 2025, 9:33 PM

#

I just add a few fields and then memcpy?

#

Also I can just let device drivers take care of it
(except that it would be a race condition)

north holly Feb 23, 2025, 9:34 PM

#

ok...

#

iirc the PCI cap list is intact after wake from suspend

calm latch Feb 23, 2025, 9:35 PM

#

Ok, I need to look how to implement SMP wakeup vector

#

/ACPI

#

(My hyper builds have no SMP for now)

#

(so that also needs to be taken care of (eventually))

north holly Feb 23, 2025, 9:35 PM

#

https://github.com/OBOS-dev/obos/blob/master/src/oboskrnl/arch/x86_64/smp.asm

GitHub

obos/src/oboskrnl/arch/x86_64/smp.asm at master · OBOS-dev/obos

I was bored... Contribute to OBOS-dev/obos development by creating an account on GitHub.

calm latch Feb 23, 2025, 9:35 PM

#

calm latch Ok, I need to look how to implement SMP wakeup vector

And my grub memes were extremely cursed

calm latch Feb 23, 2025, 9:36 PM

#

north holly https://github.com/OBOS-dev/obos/blob/master/src/oboskrnl/arch/x86_64/smp.asm

🤮 intel syntax

north holly Feb 23, 2025, 9:36 PM

#

kys (keep yourself safe)

calm latch Feb 23, 2025, 9:37 PM

#

(/s)

#

(although I prefer AT&T trl )

#

Do you keep cr3 in < 4GB?

north holly Feb 23, 2025, 9:38 PM

#

yes

calm latch Feb 23, 2025, 9:39 PM

#

unfortunate...

north holly Feb 23, 2025, 9:39 PM

#

wha

#

it is quite fortunate

#

because my pmm has zones

#

so I just tell it to allocate a 32-bit cr3 for the kernel

#

obos_status Arch_InitializeKernelPageTable()
{
    obos_status status = OBOS_STATUS_SUCCESS;
    uintptr_t newCR3 = Mm_AllocatePhysicalPages32(1,1, &status);
    uintptr_t oldCR3 = getCR3();
    // ...```

frank canopy Feb 23, 2025, 9:41 PM

#

i should work on smp at some point tbh

calm latch Feb 23, 2025, 9:41 PM

#

I can't load 64 bit idt (except...) and I can't jump to 64 bit address (also except...)

#

Does jmp extend sign?

#

If the kernel is in -2GB

north holly Feb 23, 2025, 9:42 PM

#

huh

#

why would you load the idt before being in long mode

#

in this case

calm latch Feb 23, 2025, 9:42 PM

#

So I can pagefault to success

north holly Feb 23, 2025, 9:42 PM

#

what

#

go to long mode

calm latch Feb 23, 2025, 9:42 PM

#

Or I need to jump to somewhere in -2GB

north holly Feb 23, 2025, 9:42 PM

#

load idt

#

then do whatever shit you want

calm latch Feb 23, 2025, 9:43 PM

#

Also kaslr says trl

#

I'll just patch assembly trl

mortal yoke Feb 23, 2025, 9:43 PM

#

?

calm latch Feb 23, 2025, 9:44 PM

#

not assembly, but assembled code

#

idk I'm tired today

mortal yoke Feb 23, 2025, 9:44 PM

#

you can just have a structure at the end of the trampoline and fill in the jump address to there

north holly Feb 23, 2025, 9:44 PM

#

you are not making sense go to sleep and come back

north holly Feb 23, 2025, 9:44 PM

#

mortal yoke you can just have a structure at the end of the trampoline and fill in the jump ...

🥚sactly

calm latch Feb 23, 2025, 9:45 PM

#

mortal yoke you can just have a structure at the end of the trampoline and fill in the jump ...

yeah that's what I was saying

#

but I want my trampoline to be relocatable

#

(Which does not change anything)

mortal yoke Feb 23, 2025, 9:45 PM

#

you can do cs relative accesses

calm latch Feb 23, 2025, 9:46 PM

#

yeah

mortal yoke Feb 23, 2025, 9:46 PM

#

if you don't want to patch instructions that is

calm latch Feb 23, 2025, 9:46 PM

#

But still

north holly Feb 23, 2025, 9:46 PM

#

jump to the jump address in long mode
so then you can use rip-relative accesses

calm latch Feb 23, 2025, 9:46 PM

#

I'm doing far jump

#

Setting cs to 64 bit one and all the fancy stuff

fiery turtle Feb 23, 2025, 9:46 PM

#

just mov rax, addr jump rax

calm latch Feb 23, 2025, 9:47 PM

#

fiery turtle just `mov rax, addr jump rax`

From real mode

#

Does it do sign extension?

mortal yoke Feb 23, 2025, 9:47 PM

#

you first enter long mode to a label within your tramp

fiery turtle Feb 23, 2025, 9:47 PM

#

^

#

offset can be relative

mortal yoke Feb 23, 2025, 9:47 PM

#

and after you are in long mode you do a normal jump to the actual dest

calm latch Feb 23, 2025, 9:47 PM

#

I want that label to be in higher half directly

north holly Feb 23, 2025, 9:47 PM

#

why

calm latch Feb 23, 2025, 9:47 PM

#

Because otherwise I would need id mapping

#

(which is a single PDE)

#

(but anyway)

mortal yoke Feb 23, 2025, 9:48 PM

#

its very likely not possible

north holly Feb 23, 2025, 9:48 PM

#

exception thrown in my brain: e.what(): "what are you smoking?"

mortal yoke Feb 23, 2025, 9:49 PM

#

you need an identity mapping for enabling paging anyway

#

like you can't just be in a random lower half page that is not mapped when you enable paging (and you can't be in higher half either because paging is not active yet)

calm latch Feb 23, 2025, 9:51 PM

#

ChatGPT says this is possible trl

left orbit Feb 23, 2025, 9:51 PM

#

chatgpt says a lot of things

#

gg

calm latch Feb 23, 2025, 9:51 PM

#

Real mode doesn't have paging

#

So I enter long mode from real mode, the far jump would enable paging

#

If jmp does sign extension, I can jump directly to higher half address

#

Or am I too tired

left orbit Feb 23, 2025, 9:52 PM

#

i dont think so lol

mortal yoke Feb 23, 2025, 9:53 PM

#

afaik paging is activated immediately when you enable it in cr0 and not delayed like that

calm latch Feb 23, 2025, 9:53 PM

#

Ok I forgot how to x86

calm latch Feb 23, 2025, 9:53 PM

#

mortal yoke afaik paging is activated immediately when you enable it in cr0 and not delayed ...

yeah

left orbit Feb 23, 2025, 9:53 PM

#

so you definitely do need identity mapping

#

at least a single page

calm latch Feb 23, 2025, 9:54 PM

#

It enters compat mode

#

then 2, if idtr does sign extention, I could pagefault to the right address

mortal yoke Feb 23, 2025, 9:55 PM

#

bru

calm latch Feb 23, 2025, 9:56 PM

#

Ok fine I'll just map that page

#

But then I won't be able to detect kernel nullptr derefs

mortal yoke Feb 23, 2025, 9:56 PM

#

it doesn't have to be page 0

calm latch Feb 23, 2025, 9:56 PM

#

It can be page 0

north holly Feb 23, 2025, 9:56 PM

#

wrong

calm latch Feb 23, 2025, 9:57 PM

#

My pmm can suddenly decide it's page 0

north holly Feb 23, 2025, 9:57 PM

#

the acpi spec does not allow that for wake from suspend's

north holly Feb 23, 2025, 9:57 PM

#

north holly the acpi spec does not allow that for wake from suspend's

wake vector

mortal yoke Feb 23, 2025, 9:57 PM

#

yeah that too

calm latch Feb 23, 2025, 9:57 PM

#

That's anoying

mortal yoke Feb 23, 2025, 9:57 PM

#

its not

north holly Feb 23, 2025, 9:58 PM

#

no not really

#

also the IVT

#

is at address zero

#

so you can't even put it at phys. zero if you tried

#

otherwise you'd overwrite the IVT

#

and cause the firmware to triple fault (which I have experienced on qemu!!)

mortal yoke Feb 23, 2025, 9:59 PM

#

also you could always only map the trampoline before needing it and unmap after you are done using it

north holly Feb 23, 2025, 9:59 PM

#

this

calm latch Feb 23, 2025, 9:59 PM

#

Anyway, before the smp trampoline
I have to make sure that entering S3 from userspace (by which I mean writing whatever to SLP_TYPx in Ring 3) doesn't destroy kernel state

north holly Feb 23, 2025, 10:00 PM

#

park all other CPUs

#

and you're fine

#

*if you also make sure the current CPU doesn't get rescheduled until the wake

calm latch Feb 23, 2025, 10:00 PM

#

I need to make sure I don't get a timer interrupt or whatever and uACPI doesn't get preempted unexpectedly

#

and also that the kernel is able to schedule uACPI server after wakeup

#

the hardest part is timers (?)

#

Can I have problems if I don't enter S3 from bsp?

north holly Feb 23, 2025, 10:05 PM

#

hopefully not

#

although I don't depend on that

#

and make sure the suspend helper frame is pinned to the BSP

calm latch Feb 23, 2025, 10:11 PM

#

Do I get NVS regions from limine?

fiery turtle Feb 23, 2025, 10:12 PM

#

yes

calm latch Feb 23, 2025, 10:54 PM

#

Ok, entering S3 was trivial

#

Now I just need to wake up trl

#

(my testing PC is very confused trl )

#

It showed 03 POST code, then a bunch of stuff, then went to sleep

frank canopy Feb 23, 2025, 11:56 PM

#

i wish i could make limine give me the mp info without bootstrapping the APs tbh, im gonna need a trampoline anyway for possible suspend/wake so like what would be the point in having limine do it

#

uacpi tables only mode when lol

#

(so i can make a bootloader that does what i want)

left orbit Feb 23, 2025, 11:58 PM

#

how about making a pr to limine to do exactly that? :^)

#

adding a simple flag to the mp request would be simple enough

#

limine source code is very approachable

slim panther Feb 23, 2025, 11:59 PM

#

frank canopy i wish i could make limine give me the mp info without bootstrapping the APs tbh...

atp just diy

kindred beacon Feb 23, 2025, 11:59 PM

#

frank canopy i wish i could make limine give me the mp info without bootstrapping the APs tbh...

but why

left orbit Feb 23, 2025, 11:59 PM

#

tbh it doesnt matter that limine starts them up for you

kindred beacon Feb 23, 2025, 11:59 PM

#

I honestly don't get the point

left orbit Feb 23, 2025, 11:59 PM

#

you can just ipi them and start them up again yourself

frank canopy Feb 24, 2025, 12:01 AM

#

left orbit you can just ipi them and start them up again yourself

wait you can ipi an already running processor? didnt know that, assumed it wasnt allowed

left orbit Feb 24, 2025, 12:01 AM

#

of course you can

north holly Feb 24, 2025, 12:01 AM

#

ye ofc

frank canopy Feb 24, 2025, 12:01 AM

#

left orbit limine source code is very approachable

ok but its C and i wont want to write C lmao

#

ok in that case im good

left orbit Feb 24, 2025, 12:02 AM

#

C is just like zig

#

but less ziggy

#

and more like C

frank canopy Feb 24, 2025, 12:02 AM

#

the ziggyness is why i like zig over c lol

rustic compass Feb 24, 2025, 8:29 PM

#

lenovo x270 aquired letsgo

rustic compass Feb 24, 2025, 9:01 PM

#

now i need a os to test uacpi on it and dump the tables

kindred beacon Feb 24, 2025, 9:06 PM

#

how is the policy for adding table structs to uACPI?

#

esp for non-standard (as in, not part of the ACPI spec) tables like DMAR

flat badge Feb 24, 2025, 9:13 PM

#

@fiery turtle

fiery turtle Feb 24, 2025, 9:16 PM

#

any new tables are welcome

kindred beacon Feb 24, 2025, 9:17 PM

#

surprising but nice

fiery turtle Feb 24, 2025, 9:18 PM

#

i mean it already has hpet and mcfg

kindred beacon Feb 24, 2025, 9:57 PM

#

fair

kind mantle Feb 24, 2025, 10:03 PM

#

fiery turtle i mean it already has hpet and mcfg

lmao things that essential are not in the ACPI spec?

calm latch Feb 24, 2025, 10:39 PM

#

Is fadt preserved between sleeps?

#

Or do I need to be setting wakeup vector every time?

fiery turtle Feb 24, 2025, 10:39 PM

#

The wake-up vector is stored in facs

#

Its just ram so I don't see why it wouldn't

calm latch Feb 24, 2025, 10:40 PM

#

what if there's some insane firmware which overwrites it, idk

fiery turtle Feb 24, 2025, 10:41 PM

#

Perhaps

left orbit Feb 24, 2025, 10:41 PM

#

then that isn't your problem, perhaps

#

print a big ass warning with bios interrupts

#

YOUR FIRMWARE IS STUPID

#

PLEASE BUY A NEW COMPUTER

#

and halt in a loop

#

easy

fiery turtle Feb 24, 2025, 10:42 PM

#

kind mantle lmao things *that* essential are *not* in the ACPI spec?

Well technically one is an unrelated device, and the other is part of the pcie specification which costs 2 thousand dollars

left orbit Feb 24, 2025, 10:42 PM

#

but realistically just see what linux does

lofty dragon Feb 24, 2025, 10:43 PM

#

i will now port uACPI to Vinix

fiery turtle Feb 24, 2025, 10:43 PM

#

Letsgo

left orbit Feb 24, 2025, 10:43 PM

#

finally 🙏

lofty dragon Feb 24, 2025, 10:43 PM

#

i will do that until i am eepy to go to sleep

fiery turtle Feb 24, 2025, 10:43 PM

#

Good way to fall asleep

lofty dragon Feb 24, 2025, 10:43 PM

#

fr

#

it's building llvm

#

i need to wait for it to finish now

#

i don't wanna interrupt it

fiery turtle Feb 24, 2025, 10:44 PM

#

Might as well sleep

lofty dragon Feb 24, 2025, 10:44 PM

#

ok i'll interrupt it

fiery turtle Feb 24, 2025, 10:44 PM

#

Lol

fiery turtle Feb 24, 2025, 10:45 PM

#

lofty dragon i will now port uACPI to Vinix

How good is v to c integration

lofty dragon Feb 24, 2025, 10:45 PM

#

acceptable

#

but i still have to write the bindings

fiery turtle Feb 24, 2025, 10:46 PM

#

Can you just import c headers

#

Oh

left orbit Feb 24, 2025, 10:46 PM

#

oh god lmao have fun

#

isn't there v2c?

fiery turtle Feb 24, 2025, 10:46 PM

#

left orbit oh god lmao have fun

True

#

Convert the entire types.h and resources.h trl

#

Also me when there's like 3 or 4 kernels written in zig that use uACPI but no official bindings sad

#

I also saw one on github recently

left orbit Feb 24, 2025, 10:50 PM

#

imaginarium has some bindings in tree

#

but i havent looked at them

#

~~learn zig and write official bindings~~

fiery turtle Feb 24, 2025, 10:50 PM

#

Does zig have like a package manager where you would put them or how does that work

left orbit Feb 24, 2025, 10:51 PM

#

sort of

#

how it works is you just fetch tarball or other form of sources

#

like uhhh

#

one sec

fiery turtle Feb 24, 2025, 10:51 PM

#

left orbit how it works is you just fetch tarball or other form of sources

Lol

left orbit Feb 24, 2025, 10:51 PM

#

there's thie build.zig.zon file you can put your dependencies in

#

and build.zig fetches them for you

#

.{
    .name = "zigux",
    .version = "0.0.0",
    .dependencies = .{
        .@"freestnd-c-hdrs" = .{
            .url = "https://github.com/osdev0/freestnd-c-hdrs/archive/21b59ecd6ef67bb32f893da8288ce08a324d1986.tar.gz",
            .hash = "1220a5942e368a8780187b9138b5db8c24f098533d6084c58d1122378bb4a31dcfd5",
        },
        .limine = .{
            .url = "https://github.com/48cf/limine-zig/archive/7b29b6e6f6d35052f01ed3831085a39aae131705.tar.gz",
            .hash = "1220f946f839eab2ec49dca1c805ce72ac3e3ef9c47b3afcdecd1c05a7b35f66d277",
        },
        .uacpi = .{
            .url = "https://github.com/uACPI/uACPI/archive/757dcece1a9f1c069355b23f7feaf21656728f94.tar.gz",
            .hash = "12200468fb671c04225f04d87e0fe1131e49ccd153e870be6e858dc80e39cb4e18df",
        },
    },
    .paths = .{"."},
}

fiery turtle Feb 24, 2025, 10:52 PM

#

Interesting

left orbit Feb 24, 2025, 10:52 PM

#

you can grab them in build.zig, either treat them as zig modules (if they have a build.zig) or just use the include directories or source files directly

fiery turtle Feb 24, 2025, 10:52 PM

#

Also os name leak

left orbit Feb 24, 2025, 10:52 PM

#

const freestnd_c_hdrs = b.dependency("freestnd-c-hdrs", .{});

#

this is how you refer to them in build.zig

#

kernel.addIncludePath(
    freestnd_c_hdrs.path(b.fmt("{s}/include", .{@tagName(arch)})),
);```

fiery turtle Feb 24, 2025, 10:53 PM

#

Makes sense

left orbit Feb 24, 2025, 10:53 PM

#

and here i add the include path for the freestanding headers to the kernel executable

#

so it ends up being ${source_dir}/${arch}/include

#

the source dir for the dependencies usually points somewhere inside the zig cache

#

there's 2 zig caches though, one in your home directory and one in the project root

#

~/.cache/zig and ./.zig-cache

fiery turtle Feb 24, 2025, 10:54 PM

#

Whats the global cache for?

left orbit Feb 24, 2025, 10:55 PM

#

honestly idk what its used for but i think caching dependencies

#

so the same tarballs can be reused across multiple projects

#

they are identified by the hash so

#

if it sees the same hash in the global cache it can just get it instead of downloading it again

#

thats why its important to change the hash when updating the url in the build.zig.zon file

#

or dont do it manually and just use zig fetch :^)

#

hacker@raptor:~$ zig fetch --help
Usage: zig fetch [options] <url>
Usage: zig fetch [options] <path>

    Copy a package into the global cache and print its hash.

Options:
  -h, --help                    Print this help and exit
  --global-cache-dir [path]     Override path to global Zig cache directory
  --debug-hash                  Print verbose hash information to stdout
  --save                        Add the fetched package to build.zig.zon
  --save=[name]                 Add the fetched package to build.zig.zon as name
  --save-exact                  Add the fetched package to build.zig.zon, storing the URL verbatim
  --save-exact=[name]           Add the fetched package to build.zig.zon as name, storing the URL verbatim```

calm latch Feb 24, 2025, 10:57 PM

#

how do I wake qemu up? meme

fiery turtle Feb 24, 2025, 10:57 PM

#

left orbit if it sees the same hash in the global cache it can just get it instead of downl...

Ah

left orbit Feb 24, 2025, 10:57 PM

#

system_wakeup

#

in monitor

calm latch Feb 24, 2025, 10:58 PM

#

I've put a while 1 loop in assembly at my wakeup vector, and gdb doesn't reach it

fiery turtle Feb 24, 2025, 10:58 PM

#

calm latch how do I wake qemu up? <:meme:575445094589661215>

Press key

left orbit Feb 24, 2025, 10:58 PM

#

ah even better, didnt know

fiery turtle Feb 24, 2025, 10:58 PM

#

It won't wake up if u didnt do any wake gpe configuration

gentle peak Feb 24, 2025, 10:58 PM

#

calm latch I've put a while 1 loop in assembly at my wakeup vector, and gdb doesn't reach i...

are you using break or hbreak

#

break can be a bit finicky during cpu init

calm latch Feb 24, 2025, 10:58 PM

#

just ctrl+c whatever?

gentle peak Feb 24, 2025, 10:58 PM

#

ah

calm latch Feb 24, 2025, 10:58 PM

#

gentle peak Feb 24, 2025, 10:58 PM

#

nevermind then

calm latch Feb 24, 2025, 10:59 PM

#

fiery turtle Feb 24, 2025, 10:59 PM

#

calm latch

U sure u suspended?

calm latch Feb 24, 2025, 10:59 PM

#

It said "QEMU paused"

#

Also that same code (without complications) did suspend my test PC

frank canopy Feb 24, 2025, 10:59 PM

#

left orbit ``` .{ .name = "zigux", .version = "0.0.0", .dependencies = .{ ...

you can also git+https://github.com/osdev0/freestnd-c-hdrs#21b59ecd6ef67bb32f893da8288ce08a324d1986

#

instead of the archive url

left orbit Feb 24, 2025, 10:59 PM

#

damn that's nice

#

that is not documented anywhere lol

gentle peak Feb 24, 2025, 11:00 PM

#

why would you choose to bypass the CDN like that

frank canopy Feb 24, 2025, 11:00 PM

#

and if you give it a tag/ref on the command line itll figure out what commit to use

gentle peak Feb 24, 2025, 11:00 PM

#

or well

#

not bypass

#

but still git clones are significantly slower than regular tarball downloads in my experience

calm latch Feb 24, 2025, 11:00 PM

#

calm latch

(it's supposed to be 0x9E000)

frank canopy Feb 24, 2025, 11:00 PM

#

its more about being able to tell it a ref and have it resolve it

fiery turtle Feb 24, 2025, 11:00 PM

#

calm latch It said "QEMU paused"

Yeah idk only oberrow knows how any of this works

#

And he won't write a wiki article

gentle peak Feb 24, 2025, 11:00 PM

#

calm latch

This seems like a triple fault with -no-reboot -no-shutdown

left orbit Feb 24, 2025, 11:00 PM

#

gatekeeping the suspend to ram knowledge smh

calm latch Feb 24, 2025, 11:00 PM

#

I could use bochs, but I need to fix my PCI functions

fiery turtle Feb 24, 2025, 11:01 PM

#

left orbit gatekeeping the suspend to ram knowledge smh

Ikr

calm latch Feb 24, 2025, 11:01 PM

#

gentle peak This seems like a triple fault with -no-reboot -no-shutdown

I don't pass those

frank canopy Feb 24, 2025, 11:01 PM

#

also worth noting on the zig dependency things, the dependency you fetch doesnt have to have any zig build support - the dependency object you get has a path method to get a dependency-relative path you can give to things in your main build. i use that to get both uacpi and limine binaries

gentle peak Feb 24, 2025, 11:01 PM

#

Oh, hmm

calm latch Feb 24, 2025, 11:01 PM

#

Like before I was setting wakeup vector, it was properly tripple faulting

fiery turtle Feb 24, 2025, 11:02 PM

#

So does it wake or not

left orbit Feb 24, 2025, 11:02 PM

#

frank canopy also worth noting on the zig dependency things, the dependency you fetch doesnt ...

yeah, no, you don't even need zig code to use build.zig lol, that's the great part about it

calm latch Feb 24, 2025, 11:02 PM

#

calm latch I could use bochs, but I need to fix my PCI functions

Since I only support ecam

calm latch Feb 24, 2025, 11:02 PM

#

fiery turtle So does it wake or not

Idk I think it does? But it doesn't reach my assembly

fiery turtle Feb 24, 2025, 11:02 PM

#

Also make sure you're not trashing bios state with your low mem allocator

calm latch Feb 24, 2025, 11:02 PM

#

it shouldn't be

fiery turtle Feb 24, 2025, 11:03 PM

#

calm latch Since I only support ecam

Yeah ecam will be broken on qemu after resume as well

#

Because of a bios bug

calm latch Feb 24, 2025, 11:03 PM

#

anyway, I don't care about PCI for now

fiery turtle Feb 24, 2025, 11:03 PM

#

Yeah idk why your wake vector is bogus

#

Check bios logs

#

Isa-debugcon or whatever

frank canopy Feb 24, 2025, 11:05 PM

#

oh and last zig uacpi bindings note: you dont technically need them at all

calm latch Feb 24, 2025, 11:05 PM

#

bruh

left orbit Feb 24, 2025, 11:06 PM

#

frank canopy oh and last zig uacpi bindings note: you dont technically need them at all

but native zig bindings would be much nicer

frank canopy Feb 24, 2025, 11:06 PM

#

can just

const uacpi = @cImport({
    @cInclude("uacpi.h");
    // repeat for other headers
});

#

but yeah native bindings are better

calm latch Feb 24, 2025, 11:06 PM

#

I need to printf

loud ice Feb 24, 2025, 11:06 PM

#

~~i thought you dont need one?~~ ah nvm

gentle peak Feb 24, 2025, 11:06 PM

#

calm latch <:bruh:761542241381908491>

seems like a wrong resume vector

frank canopy Feb 24, 2025, 11:06 PM

#

ive got a whole bunch of stuff already bound and ill probably throw a license exception thing for the bindings in imaginarium allowing uacpi to adopt them officially

fiery turtle Feb 24, 2025, 11:07 PM

#

Is imaginarium gpl?

left orbit Feb 24, 2025, 11:07 PM

#

mpl-2.0

#

according to github

frank canopy Feb 24, 2025, 11:07 PM

#

mpl-2 yeah

fiery turtle Feb 24, 2025, 11:07 PM

#

Ohh

#

Would be great ofc

frank canopy Feb 24, 2025, 11:08 PM

#

i dont use the incompatible with secondary licenses thing but ill still probably add a note for uacpi adoption for clarity anyway

fiery turtle Feb 24, 2025, 11:08 PM

#

Letsgo

calm latch Feb 24, 2025, 11:09 PM

#

calm latch <:bruh:761542241381908491>

Wtf

kind mantle Feb 24, 2025, 11:09 PM

#

calm latch Wtf

Hm yes lettuce wake up in real mode

north holly Feb 24, 2025, 11:09 PM

#

bruh

calm latch Feb 24, 2025, 11:09 PM

#

I need more printfs

fiery turtle Feb 24, 2025, 11:09 PM

#

calm latch Wtf

Looks correct?

north holly Feb 24, 2025, 11:09 PM

#

kind mantle Hm yes lettuce wake up in real mode

well... yes?

gentle peak Feb 24, 2025, 11:10 PM

#

fiery turtle Looks correct?

he said it was supposed to be 9e000 earlier

fiery turtle Feb 24, 2025, 11:10 PM

#

Ohh

kind mantle Feb 24, 2025, 11:10 PM

#

north holly well... yes?

lmao I learn more about how little I know every day

#

I thought it woke up in long mode

north holly Feb 24, 2025, 11:10 PM

#

_S3 throws you in real mode on wake

fiery turtle Feb 24, 2025, 11:11 PM

#

I mean yeah CPU is reset

north holly Feb 24, 2025, 11:11 PM

#

kind mantle I thought it woke up in long mode

lmaoooo

#

poor thing

fiery turtle Feb 24, 2025, 11:11 PM

#

So it starts in real mode

north holly Feb 24, 2025, 11:11 PM

#

that's only _S2/_S1

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