uACPI - a portable and easy-to-integrate ACPI implementation | Operating System Development | Page 52

fiery turtle May 1, 2025, 3:15 PM

#

Still

fiery ice May 1, 2025, 3:15 PM

#

just added a return

fiery turtle May 1, 2025, 3:15 PM

#

You cant use aml on real hw if u have half of the api stubbed

fiery ice May 1, 2025, 3:16 PM

#

i'll add the PCI driver then

fiery turtle May 1, 2025, 3:17 PM

#

I swear I need to add api smart checking and prefix every log message with [BAD KAPI] or something lol

#

(If it detects stubbed stuff)

dense steppe May 1, 2025, 3:35 PM

#

have a -DUACPI_SANITY_CHECK_KAPI and require any bug reports to run with it enabled

calm latch May 1, 2025, 4:23 PM

#

fiery turtle I swear I need to add api smart checking and prefix every log message with [BAD ...

Maybe you can just log when kernel API fails, and add a define to silence it?

#

Like not only if it's just stubbed

left orbit May 1, 2025, 4:24 PM

#

no

#

kernel api is allowed to fail lol

#

it's your fault for not implementing it

#

the kernel api is there literally so you implement it

calm latch May 1, 2025, 4:25 PM

#

that's why you make it a compile time option

left orbit May 1, 2025, 4:26 PM

#

yeah but it's stupid lol

calm latch May 1, 2025, 4:26 PM

#

The failures are usually because of OOM

left orbit May 1, 2025, 4:26 PM

#

exactly

calm latch May 1, 2025, 4:26 PM

#

A lot of kernels panic because of it anyway

left orbit May 1, 2025, 4:26 PM

#

which the kernel should be able to either on its own or know that it had happened

calm latch May 1, 2025, 4:26 PM

#

And if something returns NULL, it's usually because it's stubbed

#

So if you know that the kernel returns NULL when it's out of memory, you can silence it

#

And you can make it a debug/notice log level anyway

#

so a log message is harmless

left orbit May 1, 2025, 4:28 PM

#

making it a debug log makes sense

calm latch May 1, 2025, 4:28 PM

#

But I agree it's bad

left orbit May 1, 2025, 4:28 PM

#

i agree

calm latch May 1, 2025, 4:28 PM

#

Like the fault is on the kernel

left orbit May 1, 2025, 4:28 PM

#

i take messages a bit literally sometimes

calm latch May 1, 2025, 4:29 PM

#

and uACPI is a serious library and perhaps should be doing this stuff because some kernel doesn't follow the rules

#

out of 20 now?

#

You can have kernel api fail log level define or something, and make it a warning by default so it's obvious

#

And people could redefine it to debug/disable?

#

Idk

fiery turtle May 1, 2025, 4:35 PM

#

Its true that its super unlikely that any of the uacpi allocation APIs fail in a serious kernel because those usually have tons of mechanisms to stall a kernel allocation by all means possible to satisfy it

#

Like call into page reclaim, dump page cache, invoke oom killer, use emergency pool, compact existing memory etc

#

But its still not an error to oom and uacpi should handle it gracefully

calm latch May 1, 2025, 4:42 PM

#

i'm saying you could make it an explicit warning/log message

left orbit May 1, 2025, 4:43 PM

#

i still think it's dumb to make a dedicated "feature" for this

#

i would say it's okay if uacpi logged a debug message every time a kernel api failed

#

like, the first line of debugging something is enabling debug logging

#

and that will tell you which kernel api failed and why

fiery turtle May 1, 2025, 4:45 PM

#

Yeah, just too many places using kernel APIs and too annoying to maintain this

calm latch May 1, 2025, 4:50 PM

#

You could perhaps make kernel API defines use different error codes?

#

Like set msb bit

left orbit May 1, 2025, 4:51 PM

#

that's even more work than just logging failures lol

#

oh

#

like that

calm latch May 1, 2025, 4:52 PM

#

Then you just go over the codebase with a giant replace

fiery turtle May 1, 2025, 4:57 PM

#

calm latch Like set msb bit

Hmm, why is this helpful

#

U can do this right now, set the msb when returning errors from your api

left orbit May 1, 2025, 4:57 PM

#

so users can differentiate between uacpi and kernel api errors i guess

fiery turtle May 1, 2025, 4:58 PM

#

Sure it's not a valid enum but it is for you

#

The only difference is no error to string for uacpi

left orbit May 1, 2025, 4:58 PM

#

uacpi will just say invalid error though

#

yup

fiery turtle May 1, 2025, 4:58 PM

#

left orbit uacpi will just say invalid error though

It doesn't really check other than to stringify it

left orbit May 1, 2025, 4:58 PM

#

that's the problem

calm latch May 1, 2025, 4:58 PM

#

Error string can say returned by kernel in parenthesis

left orbit May 1, 2025, 4:58 PM

#

i think the idea is that uacpi would treat it as a regular error without the msb

fiery turtle May 1, 2025, 4:58 PM

#

And u can easily hack enum to string to pop the top bit

left orbit May 1, 2025, 4:58 PM

#

but you can check that bit

fiery turtle May 1, 2025, 4:58 PM

#

That's possible yeah

left orbit May 1, 2025, 4:59 PM

#

calm latch Error string can say returned by kernel in parenthesis

that's pain

calm latch May 1, 2025, 4:59 PM

#

Though what if it's null

left orbit May 1, 2025, 4:59 PM

#

you'd need to either allocate the string then append based on the bit or duplicate all the cases

calm latch May 1, 2025, 4:59 PM

#

Anyway

fiery turtle May 1, 2025, 4:59 PM

#

left orbit you'd need to either allocate the string then append based on the bit or duplica...

Yep

calm latch May 1, 2025, 4:59 PM

#

These are solutions against dumb

left orbit May 1, 2025, 5:00 PM

#

but i still think it's dumb for uacpi to add more code to catch someone returning not implemented from a kernel api lol

calm latch May 1, 2025, 5:00 PM

#

Which shouldn't be needed in context of osdev

fiery turtle May 1, 2025, 5:01 PM

#

Yeah like

#

Either use barebones or implement everything

#

The other use cases aren't supported officially

vast kestrel May 1, 2025, 5:08 PM

#

tbh it just feels like having kapi checks and self test is a bit too much of hand holding

#

Do you have a uacpi test image btw?

left orbit May 1, 2025, 5:09 PM

#

there's a test runner

vast kestrel May 1, 2025, 5:09 PM

#

vast kestrel tbh it just feels like having kapi checks and self test is a bit too much of han...

I mean something to test uacpi on a real hardware

#

Fuck I replied to myself

left orbit May 1, 2025, 5:10 PM

#

there's nothing iirc

#

you can grab a random kernel that uses uacpi

fiery turtle May 1, 2025, 5:10 PM

#

nah, it's just too much work

left orbit May 1, 2025, 5:10 PM

#

but ymmv

fiery turtle May 1, 2025, 5:10 PM

#

but it will be the ultra kernel trl

vast kestrel May 1, 2025, 5:10 PM

#

Fuck it, I will turn uacpi-os to a test image

left orbit May 1, 2025, 5:10 PM

#

still waiting for the report thread™️

#

and the fancy logger

#

i need to star the repo actually

fiery turtle May 1, 2025, 5:10 PM

#

left orbit and the fancy logger

same, im waiting for it to self materialize trl

left orbit May 1, 2025, 5:10 PM

#

starred.

fiery turtle May 1, 2025, 5:11 PM

#

thanks

fiery turtle May 1, 2025, 5:11 PM

#

fiery turtle same, im waiting for it to self materialize <:trl:829316641191886888>

(while i waste time on minecraft)

fiery turtle May 1, 2025, 5:12 PM

#

vast kestrel Fuck it, I will turn uacpi-os to a test image

is that your python-to-aml project?

left orbit May 1, 2025, 5:12 PM

#

i wish my kernel would just appear out of thin air like that too

vast kestrel May 1, 2025, 5:12 PM

#

yeah

#

I will just remove that part and turn it into a simple uacpi runner instead

#

it already has some basic threadig and stuff iirc

fiery turtle May 1, 2025, 5:12 PM

#

nice

fiery turtle May 1, 2025, 5:12 PM

#

left orbit i wish my kernel would just appear out of thin air like that too

lol

slim panther May 1, 2025, 9:26 PM

#

@fiery turtle are uacpi_kernel_io_{read,write}* functions ever going to get called on !x86?

frank canopy May 1, 2025, 9:28 PM

#

i think technically uncertain but almost certainly not

slim panther May 1, 2025, 9:30 PM

#

i'm doing some refactoring to my abstraction layers and the question is if i do conditional compilation inside of those functions or only include a definition when compiling for x86_64

hollow crescent May 1, 2025, 9:32 PM

#

left orbit so users can differentiate between uacpi and kernel api errors i guess

why should users even see uacpi error codes

left orbit May 1, 2025, 9:33 PM

#

slim panther i'm doing some refactoring to my abstraction layers and the question is if i do ...

just do if !x86_64 return not implemented inside them

left orbit May 1, 2025, 9:33 PM

#

hollow crescent why should users even see uacpi error codes

wha

#

wdym

#

would you rather have a boolean return value from functions that return status codes right now? lol

#

status codes have a great use case: they let you figure out what went wrong

hollow crescent May 1, 2025, 9:35 PM

#

you said users

#

why would users care about who issued the error

#

and why would users have access to uACPI anyways

#

isnt that a kernel thing

frank canopy May 1, 2025, 9:35 PM

#

i think what was meant was users of the library

#

not users as in usermode

hollow crescent May 1, 2025, 9:36 PM

#

in that case, yes of course you should return your own error codes

#

and the kernel designer had better not confuse them with their own error codes

left orbit May 1, 2025, 9:36 PM

#

💀

#

i mean the return of the kernel api functions is a uacpi_status

hollow crescent May 1, 2025, 9:36 PM

#

fine

left orbit May 1, 2025, 9:37 PM

#

and, while it's an arbitrary integer, you should probably return one of the uacpi status codes

hollow crescent May 1, 2025, 9:37 PM

#

yeah

#

so that uacpi knows what's wrong

frank canopy May 1, 2025, 9:37 PM

#

technically not all arbitrary, the success value is set

hollow crescent May 1, 2025, 9:38 PM

#

off the topic but i was definitely thinking how i could integrate uacpi into boron and i was thinking. if i'm going to make it a dll i should probably allow drivers to link against each other

fiery turtle May 1, 2025, 9:39 PM

#

slim panther <@353593948523986945> are `uacpi_kernel_io_{read,write}*` functions ever going t...

yes

#

nothing prevents it

slim panther May 1, 2025, 9:39 PM

#

hm

#

what should it do on e.g. riscv?

#

or aarch

#

just mmio?

fiery turtle May 1, 2025, 9:39 PM

#

u can return UACPI_STATUS_UNIMPLEMENTED

hollow crescent May 1, 2025, 9:39 PM

#

hollow crescent off the topic but i was definitely thinking how i could integrate uacpi into bor...

so that for example stornvme.sys would link against the kernel and also against uacpi.sys for bus enumeration (and power management probably)

fiery turtle May 1, 2025, 9:40 PM

#

e.g. on linux it would read from pci io ports

frank canopy May 1, 2025, 9:40 PM

#

yeah id return unimplemented for non-x86 io port things and if it ever trips that then figure out for that arch

slim panther May 1, 2025, 9:40 PM

#

the problem is my abstraction layer rn 😓

#

i can't call asm wrapper functions outside of the arch/ module

#

maybe a skill issue on my end, but i really want to discourage using inline asm in generic code

frank canopy May 1, 2025, 9:41 PM

#

you can always have per-arch impls in each arch/ module

#

and have the non-x86 ones return unimplemented for now

fiery turtle May 1, 2025, 9:42 PM

#

hollow crescent off the topic but i was definitely thinking how i could integrate uacpi into bor...

you can make it a separate dll, but its usually a very core component of any os especially for further device discovery and pci initialization etc so usually its not even a module

frank canopy May 1, 2025, 9:43 PM

#

can def have a driver module on top of it though, ala windows acpi.sys

#

which idk if that does all acpi or just the device stuff

fiery turtle May 1, 2025, 9:44 PM

#

yeah but windows also has multiple smaller acpi drivers that are built into the kernel

frank canopy May 1, 2025, 9:44 PM

#

yeah

fiery turtle May 1, 2025, 9:44 PM

#

for earlier code

frank canopy May 1, 2025, 9:44 PM

#

hence the driver module on top of it part, the driver running on top of the core kernel's uacpi stuff

fiery turtle May 1, 2025, 9:44 PM

#

yeah thats possible

frank canopy May 1, 2025, 9:45 PM

#

thats technically what i do except my drivers are just their own files that are part of the same compilation unit

#

my api can technically do separate modules for drivers except i dont support dynamic linking or loading rn lol

fiery turtle May 1, 2025, 9:55 PM

#

i have yet to think about how im gonna resolve cross module dependencies

frank canopy May 1, 2025, 9:56 PM

#

just have everything being compiled together as a single unit like i do and like is the norm for zig its easy

vast kestrel May 1, 2025, 9:56 PM

#

ok, I have a "bare minimum" kernel that should be able to run uacpi properly

frank canopy May 1, 2025, 9:57 PM

#

nice

vast kestrel May 1, 2025, 9:57 PM

#

now to enable random stuff so you can test more than just the init

fiery turtle May 1, 2025, 9:57 PM

#

Damn haiku PR got revived

#

After a few months

slim panther May 1, 2025, 9:57 PM

#

and?

#

declined?

fiery turtle May 1, 2025, 9:58 PM

#

vast kestrel ok, I have a "bare minimum" kernel that should be able to run uacpi properly

Nice

fiery turtle May 1, 2025, 9:58 PM

#

slim panther and?

I said revived not reviewed

slim panther May 1, 2025, 9:58 PM

#

i'm blind

frank canopy May 1, 2025, 9:58 PM

#

vast kestrel now to enable random stuff so you can test more than just the init

what sort of random stuff, of curiosity?

fiery turtle May 1, 2025, 9:58 PM

#

So far the feedback is positive

vast kestrel May 1, 2025, 9:58 PM

#

frank canopy what sort of random stuff, of curiosity?

idk, I guess power button and wakeup sources

hollow elm May 1, 2025, 9:58 PM

#

slim panther and?

bro does not care 🥀

vast kestrel May 1, 2025, 9:58 PM

#

so I can also implement a sleep function or something alike

#

to test sleep and wakeup from sleep

slim panther May 1, 2025, 9:59 PM

#

hollow elm bro does not care 🥀

opposite

#

i was actually hyped

fiery turtle May 1, 2025, 9:59 PM

#

vast kestrel ok, I have a "bare minimum" kernel that should be able to run uacpi properly

Test on real hw chad

vast kestrel May 1, 2025, 9:59 PM

#

tbh I should before saying that my code works :^)

#

lemme try on my laptop

#

oh wait first I need to add a framebuffer thing

fiery ice May 1, 2025, 9:59 PM

#

i still need to add PCI lol

vast kestrel May 1, 2025, 9:59 PM

#

I am using 0xE9

frank canopy May 1, 2025, 10:00 PM

#

at some point ill test my stuff on hardware lol

#

someday

#

im also using 0xE9 though lol

fiery ice May 1, 2025, 10:00 PM

#

i'm working on threads now so i can run uACPI properly

fiery turtle May 1, 2025, 10:00 PM

#

Very few people know that u can detect 0xE9 prescence by checking that in(0xE9) == 0xE9

#

Just a random fact

vast kestrel May 1, 2025, 10:01 PM

#

yeah, its a cool thing

frank canopy May 1, 2025, 10:01 PM

#

WAIT REALLY?

#

dang

vast kestrel May 1, 2025, 10:01 PM

#

the sad thing tho

#

the sad thing is that you don't have input from e9

fiery turtle May 1, 2025, 10:01 PM

#

Lol

vast kestrel May 1, 2025, 10:01 PM

#

which is why I quickly switch to serial

#

I just realized

fiery turtle May 1, 2025, 10:01 PM

#

frank canopy dang

Yeah it works on both qemu and bochs

vast kestrel May 1, 2025, 10:01 PM

#

once you initialize acpi, the kernel won't have ps2 emulation

#

problems for later ig

fiery turtle May 1, 2025, 10:02 PM

#

vast kestrel which is why I quickly switch to serial

Would be nice if you added repl to run random uacpi functions

frank canopy May 1, 2025, 10:02 PM

#

i kinda want to get an x86 SBC with a hardware serial port just to test osdev things on tbh

fiery turtle May 1, 2025, 10:02 PM

#

Like eval methods, dump the namespace

vast kestrel May 1, 2025, 10:02 PM

#

fiery turtle Would be nice if you added repl to run random uacpi functions

that is exactly my idea

#

I want to have a c like repl

fiery turtle May 1, 2025, 10:02 PM

#

Damn

vast kestrel May 1, 2025, 10:02 PM

#

and have all the uacpi functions as symbols

fiery turtle May 1, 2025, 10:02 PM

#

Nice

fiery ice May 1, 2025, 10:02 PM

#

frank canopy i kinda want to get an x86 SBC with a hardware serial port just to test osdev th...

technically most desktop computers have hw serial

#

i just don't know if the OS can use it

frank canopy May 1, 2025, 10:02 PM

#

fiery ice technically most desktop computers have hw serial

well yeah true ig

fiery ice May 1, 2025, 10:03 PM

#

for example, my motherboard exposes the pins for it

frank canopy May 1, 2025, 10:03 PM

#

but either way i want an SBC that i can use normal basic serial to get info out of to then do osdev with

hollow crescent May 1, 2025, 10:03 PM

#

fiery ice i just don't know if the OS can use it

should

hollow crescent May 1, 2025, 10:03 PM

#

fiery turtle i have yet to think about how im gonna resolve cross module dependencies

for each module
if module contains function then
thats the module

#

I see myself designing it like this at least

#

Kernel modules are compiled with the option that prevents the linker from throwing errors if functions are undefined

#

So I don't need to link one kernel driver to have it import another's functions

#

Which means that kernel drivers will be able to have circular dependencies and it would be allowed

#

The only thing that won't, is the kernel

calm latch May 1, 2025, 10:19 PM

#

slim panther what should it do on e.g. riscv?

I wonder if it could be used on loongarch...

slim panther May 1, 2025, 10:19 PM

#

does it have portio?

#

or working acpi meme

calm latch May 1, 2025, 10:22 PM

#

fiery turtle Test on real hw <:chad:815144781842743334>

My kernel kinda maybe had S3 sleep working with uACPI on physical hardware coolflushed

calm latch May 1, 2025, 10:22 PM

#

slim panther or working acpi <:meme:575445094589661215>

uACPI has managed to shut it down (S5) after recieving power button event in QEMU

#

So I guess

fiery ice May 1, 2025, 10:22 PM

#

mine doesn't even recognise S5 is even an option (i hope it's bc of the PCI api)

slim panther May 1, 2025, 10:22 PM

#

real

calm latch May 1, 2025, 10:23 PM

#

slim panther does it have portio?

It has iocsr, which is kinda like io on x86?

fiery ice May 1, 2025, 10:23 PM

#

fiery ice mine doesn't even recognise S5 is even an option (i hope it's bc of the PCI api)

this only on my laptop tho

calm latch May 1, 2025, 10:23 PM

#

calm latch It has iocsr, which is kinda like io on x86?

Like what if ACPI wants to access it, what does it do?

tired temple May 1, 2025, 10:23 PM

#

hollow crescent off the topic but i was definitely thinking how i could integrate uacpi into bor...

__declspec(dllimport) for real dlls.

calm latch May 1, 2025, 10:23 PM

#

calm latch uACPI has managed to shut it down (S5) after recieving power button event in QEM...

From GED

hollow crescent May 1, 2025, 10:24 PM

#

tired temple `__declspec(dllimport)` for real dlls.

i do not use msvc

#

i was talking about dlls as a concept rather than dlls as strictly PE format dynamic library objects

slim panther May 1, 2025, 10:24 PM

#

bro is pushing msvc again 😭

#

get away with this slop

calm latch May 1, 2025, 10:24 PM

#

calm latch My kernel kinda maybe had S3 sleep working with uACPI on physical hardware <:coo...

And kinda not even kernel

lofty dragon May 1, 2025, 10:26 PM

#

💀

tired temple May 1, 2025, 10:26 PM

#

hollow crescent i do not use msvc

I meant mechanism behind it. It's not compiler specific. I'm not pushing anything. I am just also thinking on integrating many things and shared my thoughts on the close subject. I'd make it a driver, boot time one, some kind of root enumerator. I'm actually close to start with it.

hollow crescent May 1, 2025, 10:27 PM

#

tired temple I meant mechanism behind it. It's not compiler specific. I'm not pushing anythin...

the way gcc works is you have to explicitly link against a shared object in order for the linker to say "fine, we will leave this as an import"

#

so like, -lboron if you want to link against libboron.so

#

this works fine for user space but i kinda do not want the kernel to be a "shared object" (because idk if limine supports loading those and im not taking chances)

#

(although i could just create a libboronkernel.so which is literally just the kernel objects in library form and then link against that and serve it relocations from the boron kernel proper actually)

#

(because the modules only remember the names of imported functions and nothing else)

calm latch May 1, 2025, 10:30 PM

#

just have a prekernel trl

hollow crescent May 1, 2025, 10:30 PM

#

calm latch just have a prekernel <:trl:829316641191886888>

no

#

i wanted that for a while but just decided not to bother

tired temple May 1, 2025, 10:31 PM

#

slim panther get away with this slop

You know what "slop" really is? that pathetic overhyped rust the only outcome of using which is dividing by 20 the number in a phallometric contest.

slim panther May 1, 2025, 10:31 PM

#

lmao

hollow crescent May 1, 2025, 10:31 PM

#

okay "phallometric contest" is funny

tired temple May 1, 2025, 10:41 PM

#

hollow crescent the way gcc works is you have to explicitly link against a shared object in orde...

yes, elf can't provide real dlls, so if you stick with it, you can't use this useful mechanism for driver intercomuniction. however clang can make PEs. For x86 and arm. And since they added mips pe support as I heard, maybe dllimport is available for it as well. And maybe even for loongarch too, since it's a mips clone. Fortunately, having picked what I really need for the wanted design, I don't have such issues. My issue is the extremely poor ACPI support for arm targets. So I need some kind of dtb engine equivalent to the acpi enumerator for machine topogy discovery. And that fdt is so fkn sucky, so demotivating.

slim panther May 1, 2025, 10:41 PM

#

dtbs are easy

hollow crescent May 1, 2025, 10:41 PM

#

tired temple yes, elf can't provide real dlls, so if you stick with it, you can't use this us...

elf shared objects are a type of DLL and there is no "real DLL". DLL, shared object or dynamic library are names for the same concept

#

PE DLLs have one feature that ELF shared objects don't have though which is that function imports are tied to a DLL within an application/DLL

#

while ELF just has a list of SOs that need to be loaded

tired temple May 1, 2025, 10:43 PM

#

slim panther dtbs are easy

Because they don't do anything. Without "bindings" they are just nothing. And those bindings in fact mean "just write a specific driver everytime for eachndevice". Still you won't have such nice things as s states transitions anything that aml does. Dtb are crap.

slim panther May 1, 2025, 10:44 PM

#

dtbs are a solution for a different problem

#

they just describe a device

tired temple May 1, 2025, 10:46 PM

#

hollow crescent PE DLLs have one feature that ELF shared objects don't have though which is that...

That very feature, you'd benefit from for what you wete talking about and to what I commented. Easily provide "helper" functions from a driver to other drivers. The NT way.

tired temple May 1, 2025, 10:51 PM

#

slim panther dtbs are a solution for a different problem

Yet you don't get an equivalent for "everything else" dtbs are incapable of on dtb supplied machines. You just get this garbage dtb full of linuxisms and nothing working even in the stock linux. try to get an arm sbc with suspend to ram or hibernation. On freaking "power saving arm" platforms. You can fry eggs with them and linux. This is how it works without having acpi they are so arrogant towards. With the fdt pulled off cluelessly from openfirmware without much thinking.

slim panther May 1, 2025, 10:52 PM

#

that's why x86 is the only real platform

tired temple May 1, 2025, 10:52 PM

#

slim panther that's why x86 is the only real platform

Because red hat are lamers?

slim panther May 1, 2025, 10:53 PM

#

what?

#

where does rh come in here

tired temple May 1, 2025, 10:53 PM

#

slim panther what?

They introduced this fdt to use instead of acpi.

jaunty fox May 1, 2025, 10:59 PM

#

hollow crescent this works fine for user space but i kinda do not want the kernel to be a "share...

i was experimenting with the automatic creation of a symbols file & a pseudo kernel library (linked against by kernel modules, but the kernel dynamic linker knows it's really a placeholder for the kernel) to this end

#

the alternative is to use that linker option that ignores undefined symbols

#

but i don't like that one

vast kestrel May 2, 2025, 1:49 AM

#

the power button works on my laptop without needing an EC driver Smuggy

#

and its a really new laptop, so its quite surprising

#

there are some aml errors, not sure if its related to the ec or not, but I am gonna write an ec driver regardless

fiery turtle May 2, 2025, 8:44 AM

#

vast kestrel the power button works on my laptop without needing an EC driver <:Smuggy:567581...

Nice

fiery turtle May 2, 2025, 8:44 AM

#

vast kestrel there are some aml errors, not sure if its related to the ec or not, but I am go...

Can you show them

trim frost May 2, 2025, 8:52 AM

#

vast kestrel the power button works on my laptop without needing an EC driver <:Smuggy:567581...

can't relate, my kernel doesn't even run on my laptop :/

fiery turtle May 2, 2025, 8:53 AM

#

vast kestrel and its a really new laptop, so its quite surprising

Probably they just didn't cheap out on circuitry to route it properly

vast kestrel May 2, 2025, 9:12 AM

#

fiery turtle Can you show them

So some of them are object already exists that also happen on Linux

#

But one of them is the bat0 complaining about undefined in op

fiery turtle May 2, 2025, 9:12 AM

#

ah ive seen this already

vast kestrel May 2, 2025, 9:13 AM

#

Now I checked the aml (I can send you later) but it does actually properly check for _REG in EC being called

fiery turtle May 2, 2025, 9:13 AM

#

this is basically untested code brecause no ec driver

#

so it hits a path with undefined local

vast kestrel May 2, 2025, 9:13 AM

#

Yeah but it's weirder

#

There is no ECDT

#

And this doesn't happen in Linux

fiery turtle May 2, 2025, 9:13 AM

#

yeah but u run _REG before initialize usually

#

this is why it works fine

#

anyway yeah send the dump

#

but i think one of mishakov's laptop is like this

vast kestrel May 2, 2025, 9:14 AM

#

Is there a chance you run _REG by accident with default handler?

fiery turtle May 2, 2025, 9:14 AM

#

nope

#

there isnt a default handler for ec

vast kestrel May 2, 2025, 9:14 AM

#

I'll check it again later then, I looked at it quickly and I didn't see it make a mistake, but I haven't looked too deep

fiery turtle May 2, 2025, 9:14 AM

#

u say it checks for ECAV

vast kestrel May 2, 2025, 9:15 AM

#

Yeah

fiery turtle May 2, 2025, 9:15 AM

#

this is what ive seen on mishakovs laptop as well

#

but in the else

#

it forgets to initialize a local

vast kestrel May 2, 2025, 9:15 AM

#

And it sets it in _REG

fiery turtle May 2, 2025, 9:15 AM

#

so its undefined

vast kestrel May 2, 2025, 9:15 AM

#

Could be then

fiery turtle May 2, 2025, 9:15 AM

#

and then it uses it

vast kestrel May 2, 2025, 9:15 AM

#

I really haven't looked too deep

#

But wait, if there is no ECDT, how does that work

fiery turtle May 2, 2025, 9:15 AM

#

its a common untested thing they have in BAT0._STA

vast kestrel May 2, 2025, 9:15 AM

#

Yeah it might be the same

#

Was it a Lenovo?

fiery turtle May 2, 2025, 9:16 AM

#

linux doesnt call BAT0._STA because it knows it has no _INI since ACPICA does subtree prescan to speed up namespace_initialize()

#

basically its a cosmetic issue

fiery turtle May 2, 2025, 9:16 AM

#

vast kestrel Was it a Lenovo?

yeah

vast kestrel May 2, 2025, 9:16 AM

#

I see

fiery turtle May 2, 2025, 9:16 AM

#

but yeah u can just paste the code here later

#

to verify

vast kestrel May 2, 2025, 9:17 AM

#

It is really weird to me that there is no ECDT tho

#

Like, when am I expected to install my own drivers?

fiery turtle May 2, 2025, 9:17 AM

#

i mean not all laptops need early EC

vast kestrel May 2, 2025, 9:17 AM

#

Yeah but the _STA needs it technically

fiery turtle May 2, 2025, 9:17 AM

#

vast kestrel Like, when am I expected to install my own drivers?

after initialize() iirc

fiery turtle May 2, 2025, 9:17 AM

#

vast kestrel Yeah but the _STA needs it technically

_STA would be called way later

vast kestrel May 2, 2025, 9:17 AM

#

Ok so I am doing it in the right place

fiery turtle May 2, 2025, 9:17 AM

#

when BAT0 driver is being probed

#

at that point EC is already live

vast kestrel May 2, 2025, 9:18 AM

#

Wait so why do you call it?

fiery turtle May 2, 2025, 9:18 AM

#

during namespace_initialize it expects all _INI to be called

#

_INI can only be called if _STA returns 0xF

#

and u cant go into child namespaces if _STA failed

#

so u kinda have to

#

or u can pre-scan to know what subtrees u can skip

#

but i just dont do it because its fast anyway

vast kestrel May 2, 2025, 9:19 AM

#

I see

#

Is that what windows does as well?

#

Like, running untested acpi code paths is a bit scary 3sGudaEhehe

fiery turtle May 2, 2025, 9:19 AM

#

u mean would windows call _STA there?

vast kestrel May 2, 2025, 9:19 AM

#

Yeah

#

And it's not like it's a missing ec driver, it doesn't provide ECDT, so it would happen on windows as well

fiery turtle May 2, 2025, 9:20 AM

#

it would probably have an EC driver at that point, but I doubt it has any sort of advanced tree pre-scan

#

windows just doesnt do any advanced things with aml

#

would be interesting to check I guess

vast kestrel May 2, 2025, 9:20 AM

#

But it won't since the firmware doesn't provide ECDT you can't install the EC driver before initialize

fiery turtle May 2, 2025, 9:21 AM

#

technically

#

the namespace is already loaded at that point

#

so u can find the EC device

vast kestrel May 2, 2025, 9:21 AM

#

Yeah you could install the driver

fiery turtle May 2, 2025, 9:21 AM

#

no one knows what order the windows driver does it

vast kestrel May 2, 2025, 9:21 AM

#

But is that what I should do

#

Tru

fiery turtle May 2, 2025, 9:21 AM

#

vast kestrel But is that what I should do

well thats not what linux does at least i think

#

id have to re-check the code

#

but i should experiment to see what widnows does in this case

vast kestrel May 2, 2025, 9:22 AM

#

Btw for the love of god can you add to the osdev docs the set interrupt controller function? I forgot it's name and had to search all the header files for it 3sGudaEhehe

fiery turtle May 2, 2025, 9:22 AM

#

lol

#

yeah

vast kestrel May 2, 2025, 9:22 AM

#

And given it's a pretty important part of the acpi init it could be helpful

fiery turtle May 2, 2025, 9:22 AM

#

its in utilities.h

vast kestrel May 2, 2025, 9:23 AM

#

Also I feel like the namespace iterations are missing a few docs, like I couldn't figure out of find_node gives you an existing reference or if you need to unref it (from the code it seems like an existing reference)

#

But if it gives an existing reference can't it be a bit dangerous?

fiery turtle May 2, 2025, 9:23 AM

#

namespace nodes are permanent

#

u cant ref or unref them with public api

vast kestrel May 2, 2025, 9:24 AM

#

I see

fiery turtle May 2, 2025, 9:24 AM

#

but yeah a comment would be nice

vast kestrel May 2, 2025, 9:24 AM

#

Right now the problem I had is that the second ec resource had a type of zero, so I might have a bug somewhere in there

fiery turtle May 2, 2025, 9:24 AM

#

vast kestrel But if it gives an existing reference can't it be a bit dangerous?

it explicitly denies access to namespace nodes that are temporary

vast kestrel May 2, 2025, 9:24 AM

#

Yeah I saw that

fiery turtle May 2, 2025, 9:24 AM

#

aka only exist if theres currently aml being executed

vast kestrel May 2, 2025, 9:25 AM

#

All the temporary node stuff

fiery turtle May 2, 2025, 9:25 AM

#

vast kestrel Right now the problem I had is that the second ec resource had a type of zero, s...

isnt that just the resource terminator?

vast kestrel May 2, 2025, 9:26 AM

#

I am using the get_current_resources, which returns length + entries

#

I check the length >= 2, the first resource is IO as expected, the second has a type of zero, which idr what it is

fiery turtle May 2, 2025, 9:26 AM

#

0 is IRQ

#

from the looks of it

fiery turtle May 2, 2025, 9:26 AM

#

vast kestrel I check the length >= 2, the first resource is IO as expected, the second has a ...

wait

vast kestrel May 2, 2025, 9:26 AM

#

Yeah, and I looked at the aml and it doesn't have that, so probably did something funny in my code

fiery turtle May 2, 2025, 9:26 AM

#

resources are variable length

#

2 means 2 bytes

#

not two resources

vast kestrel May 2, 2025, 9:27 AM

#

Wait

#

But it returns an array of resource pointers?

#

it's uacpi_resources

fiery turtle May 2, 2025, 9:27 AM

#

yes, but see https://github.com/uACPI/uACPI/blob/master/include/uacpi/resources.h#L647

vast kestrel May 2, 2025, 9:27 AM

#

Ok fuck

#

Ok ok I see how the api is supposed to be used

#

Lmao

#

more docs plz

fiery turtle May 2, 2025, 9:28 AM

#

i tried to hint it by doing resources[]

#

or maybe not

#

yeah it needs a comment

vast kestrel May 2, 2025, 9:28 AM

#

Nah, I would probably do the same

#

Might be worth a comment, but if I looked more than two seconds at the struct it would also tell me what I did wrong :^)

#

You just saved me from adding heap guards and expecting a bug with the allocator

fiery turtle May 2, 2025, 9:29 AM

#

lol

#

to be fair that macro is also 1 line above the struct

#

btw it might be easier to just uacpi_for_each_device_resource

#

that does free on its own

#

u dont even have to keep a pointer

vast kestrel May 2, 2025, 9:30 AM

#

It probably is

#

I haven't seen that, will def use it

fiery turtle May 2, 2025, 9:31 AM

#

uacpi_for_each_device_resource(dev, "_CRS", lambda, ctx)

#

there's also uacpi_for_each_resource for an already existing resource buffer

vast kestrel May 2, 2025, 9:37 AM

#

Yeah uacpi had a lot of helpers, really nice to use it

#

Even better when using it correctly

#

:^)

fiery turtle May 2, 2025, 9:38 AM

#

thx

dense steppe May 2, 2025, 9:51 AM

#

hollow crescent Which means that kernel drivers will be able to have circular dependencies and i...

why are circular dependencies useful?

because consider a strongly connected component in the dependency graph. if any of the modules in it are loaded, all of the modules in it have to be loaded. therefore, we might as well consider the strongly connected component a single module.

hollow crescent May 2, 2025, 9:54 AM

#

dense steppe why are circular dependencies useful? because consider a strongly connected com...

So that I don't have to link uacpi into my HAL

#

(Although it's possible (but slim chances) that there is no need for UACPI to use the HAL or the HAL to use UACPI)

#

Also why would I not want to link uACPI with the HAL? LARPing purposes

dense steppe May 2, 2025, 9:55 AM

#

hollow crescent (Although it's possible (but slim chances) that there is no need for UACPI to us...

yeah so this won't happen

#

because the kernel depends on the HAL

hollow crescent May 2, 2025, 9:55 AM

#

yes

dense steppe May 2, 2025, 9:55 AM

#

the HAL needs ACPI tables for initialization (MADT, etc)

#

and ACPI needs the kernel for memory allocations, IRQ handling, etc.

hollow crescent May 2, 2025, 9:56 AM

#

if I really didn't want to have circular dependencies then I could ship my own implementation of ACPI tables for the HAL

#

in fact, I already do that

#

but this is a very good point. if uacpi and hal depend on each other what's the point of them being separated?

fiery turtle May 2, 2025, 9:57 AM

#

btw uacpi has a mode that strips all code but table managment

#

u could reuse that

dense steppe May 2, 2025, 9:58 AM

#

two uACPIs in one kernel? 🤯

fiery turtle May 2, 2025, 9:58 AM

#

lol

hollow crescent May 2, 2025, 9:59 AM

#

fiery turtle btw uacpi has a mode that strips all code but table managment

thats fun

fiery turtle May 2, 2025, 9:59 AM

#

https://review.haiku-os.org/c/haiku/+/8937 haiku pr is finally doing something

hollow crescent May 2, 2025, 9:59 AM

#

but nah

fiery turtle May 2, 2025, 9:59 AM

#

we'll see what happens

hollow crescent May 2, 2025, 10:00 AM

#

so a couple options

build in uACPI with my HAL (no separating the two)
the HAL and uACPI are circularly dependent (de facto no separating the two and only for larp purposes)
the HAL comes with its own ACPI table parser and uACPI doesn't directly depend on the HAL (the two can be separated but there is code redundancy and my ACPI table parser is probably inferior)

fiery turtle May 2, 2025, 10:03 AM

#

What is LARP

fiery turtle May 2, 2025, 10:05 AM

#

hollow crescent so a couple options - build in uACPI with my HAL (no separating the two) - the ...

Yeah I'm not sure, what's the advantage of decoupling uacpi from Hal on x86?

hollow crescent May 2, 2025, 10:05 AM

#

fiery turtle Yeah I'm not sure, what's the advantage of decoupling uacpi from Hal on x86?

being able to boot without ACPI support

hollow crescent May 2, 2025, 10:05 AM

#

fiery turtle What is LARP

Live action role play

#

Pretending to be something you're not, in this case

fiery turtle May 2, 2025, 10:08 AM

#

hollow crescent being able to boot without ACPI support

Yeah but why would you want this

#

Does windows have a mode like this

hollow crescent May 2, 2025, 10:08 AM

#

Just in case

hollow crescent May 2, 2025, 10:09 AM

#

fiery turtle Does windows have a mode like this

I think so

fiery turtle May 2, 2025, 10:09 AM

#

hollow crescent Just in case

Is it not possible to make it skip acpi init behind a flag for exceptional cases like that?

hollow crescent May 2, 2025, 10:09 AM

#

I know you can disable the ACPI shutdown and have it show the "It's now safe to turn off your computer" screen in the latest Windows 11 but I'm not sure that disables the entirety of ACPI

hollow crescent May 2, 2025, 10:09 AM

#

fiery turtle Is it not possible to make it skip acpi init behind a flag for exceptional cases...

Yes you can

#

Or you can just omit uacpi.sys from loading

fiery turtle May 2, 2025, 10:10 AM

#

hollow crescent I know you can disable the ACPI shutdown and have it show the "It's now safe to ...

Probably not, because it need acpi for device discovery and PCI brdige allocation and discovery as well

#

Like that would probably disable all of PCI and i2c as well

#

Potentially USB

hollow crescent May 2, 2025, 10:11 AM

#

Why would it disable all of PCI? You can scan PCI devices the brute force way

#

By scanning every bus, slot, and function

fiery turtle May 2, 2025, 10:11 AM

#

You can ofc lol, but I doubt windows has a brute force scan fallback

#

Its not safe either

#

Probing bogus busses might hang some controllers

hollow crescent May 2, 2025, 10:12 AM

#

ah

#

I see

fiery turtle May 2, 2025, 10:12 AM

#

In this case it also has to trust bios for bar window allocation

#

Which is probably fine but yeah idk

hollow crescent May 2, 2025, 10:13 AM

#

It has to trust the firmware for a lot of things anyway

fiery turtle May 2, 2025, 10:13 AM

#

Yeah basically it cant check any resource collisions

vast kestrel May 2, 2025, 10:21 AM

#

what would be better, micropython + ipython shell, or a C like repl shell

#

both would have full access to uacpi functions and what not

fiery turtle May 2, 2025, 10:26 AM

#

Python is probably way more powerful

vast kestrel May 2, 2025, 10:27 AM

#

it would require some ffi glue but yeah

loud ice May 2, 2025, 10:43 AM

#

vast kestrel what would be better, micropython + ipython shell, or a C like repl shell

does micropython have enough stuff to run ipython?

#

also micropython is quite annoying

vast kestrel May 2, 2025, 10:43 AM

#

no idea

loud ice May 2, 2025, 10:43 AM

#

what about kuroko :^)

#

its also a better language

vast kestrel May 2, 2025, 10:44 AM

#

"Kuroko's Basketball is a Japanese sports manga series written and illustrated by Tadatoshi Fujimaki"

loud ice May 2, 2025, 10:44 AM

#

lol

#

https://github.com/kuroko-lang/kuroko

#

its klange's not quite python

#

with sane variable binding rules

vast kestrel May 2, 2025, 10:44 AM

#

its something someone on the server made?

#

is there a way to tell limine to use the highest resolution available? it annoys me that it doesn't use full screen on my laptop

fiery turtle May 2, 2025, 11:04 AM

#

vast kestrel is there a way to tell limine to use the highest resolution available? it annoys...

it probably tries by default

#

its likely that your laptop doesnt provide an edid via uefi

#

so limine just picks whatever is safe

#

although laptops usually do provide it so idk

vast kestrel May 2, 2025, 11:04 AM

#

in the limine screen it is full, in the os screen it is not

#

or is this a flanterm thing

fiery turtle May 2, 2025, 11:04 AM

#

limine probably doesnt do modesetting

#

and default gop is native res usually

#

but idk ask mint

vast kestrel May 2, 2025, 11:05 AM

#

~~ever since the discord was killed its a bit harder to ask limine questions~~

trim frost May 2, 2025, 11:41 AM

#

updated score on i5 12400f

lofty dragon May 2, 2025, 11:49 AM

#

vast kestrel ~~ever since the discord was killed its a bit harder to ask limine questions~~

you can just ping me

#

tbh

vast kestrel May 2, 2025, 11:49 AM

#

tru

lofty dragon May 2, 2025, 11:49 AM

#

and no Limine itself doesn't do modesetting, it uses GOP or VBE

vast kestrel May 2, 2025, 11:50 AM

#

tbf I rarely need limine help because it just works Smuggy

lofty dragon May 2, 2025, 11:50 AM

#

vast kestrel in the limine screen it is full, in the os screen it is not

if you mean that Flanterm just uses a smaller rectangle in the middle of the screen, that is limited due to using the Flanterm internal bump allocator

#

if you use a proper allocator you can get full screen

vast kestrel May 2, 2025, 11:51 AM

#

ooooooo

#

ok

#

then ill do that

#

thankss

lofty dragon May 2, 2025, 11:51 AM

#

np

vast kestrel May 2, 2025, 12:46 PM

#

@fiery turtle yeah its the exact same thing with the local not being initialized

fiery ice May 2, 2025, 12:49 PM

#

lofty dragon if you mean that Flanterm just uses a smaller rectangle in the middle of the scr...

can i use the bump alloc at the start, and after i init my vma switch to that?

vast kestrel May 2, 2025, 12:50 PM

#

you will need to destroy the context and create a new one

torpid root May 2, 2025, 12:50 PM

#

trim frost updated score on i5 12400f

any stubbed api functions?

fiery ice May 2, 2025, 12:50 PM

#

vast kestrel you will need to destroy the context and create a new one

there is fb_reinit i saw

fiery turtle May 2, 2025, 12:50 PM

#

vast kestrel <@353593948523986945> yeah its the exact same thing with the local not being ini...

makes sense

trim frost May 2, 2025, 12:51 PM

#

torpid root any stubbed api functions?

thread id is 1*

#

and schedule stuff is unimplemented

lofty dragon May 2, 2025, 12:57 PM

#

fiery ice can i use the bump alloc at the start, and after i init my vma switch to that?

yeah, what itay said

#

you have to create a new context

torpid root May 2, 2025, 1:31 PM

#

@fiery turtle will anything break if I noop in spinlock/mutex acquire/release until the scheduler is initialised?

#

I assume it would not

fiery turtle May 2, 2025, 2:31 PM

#

should be fine

frank canopy May 2, 2025, 6:27 PM

#

further resource type support pushed to zuacpi

#

as well as actually accepting the existence of alignment in resources

fiery turtle May 2, 2025, 7:11 PM

#

nice!

frank canopy May 2, 2025, 7:17 PM

#

it now has enough resource types to handle everything that q35's acpi tables gives me

hollow scaffold May 2, 2025, 7:18 PM

#

I love this project

frank canopy May 2, 2025, 7:18 PM

#

(plus an nvme controller on top)

vast kestrel May 2, 2025, 8:18 PM

#

I spent the entire day rewriting the threading and sync code because I thought I had a bug in it, only to release I got the order of the data and command port wrong in the CRS AMfacepalm

trim frost May 2, 2025, 8:19 PM

#

vast kestrel I spent the entire day rewriting the threading and sync code because I thought I...

skill issue

vast kestrel May 2, 2025, 8:20 PM

#

I still have a bug in the scheduler tho, waking a thread from interrupt doesn't wakeup the core to actually run the thread I woke up Smuggy

fiery turtle May 2, 2025, 8:20 PM

#

vast kestrel I spent the entire day rewriting the threading and sync code because I thought I...

Lol

#

At least they're better now I hope

vast kestrel May 2, 2025, 8:20 PM

#

they are simpler

fiery turtle May 2, 2025, 8:21 PM

#

vast kestrel I spent the entire day rewriting the threading and sync code because I thought I...

Use mwait chad

vast kestrel May 2, 2025, 8:21 PM

#

before I had a futex like primitive, and I switched it to just having a semaphore+cond as the primitives instead

#

smh

#

I actually removed a bunch of random from the kernel I based off since I want this to run on older hardware as well lol

#

I still depend on constant tsc, since that is the simplest time keeping, but I am not using tsc deadline for example

#

I use the lapic

fiery turtle May 2, 2025, 8:22 PM

#

nice

frank canopy May 2, 2025, 8:23 PM

#

reminds me i need a futex primitive

frank canopy May 2, 2025, 8:23 PM

#

vast kestrel I spent the entire day rewriting the threading and sync code because I thought I...

is your think open source? curious to take a look at scheduler code

vast kestrel May 2, 2025, 8:23 PM

#

yeah, its uacpi-os

#

https://github.com/Itay2805/uACPI-OS/tree/master

#

I will need to implement the workqueue stuff eventually, I am starting to get TODO prints for wait_for_completion nervous_sweat_FB

#

also reminds me I haven't implemented timers at all yet, so there are no timeouts

fiery turtle May 2, 2025, 8:26 PM

#

i wonder where thats called from for u

vast kestrel May 2, 2025, 8:26 PM

#

I added a lid driver

#

so I register a notify

#

I think once I added that it started calling it

fiery turtle May 2, 2025, 8:27 PM

#

ah yeah

vast kestrel May 2, 2025, 8:28 PM

#

would it be more correct to clear the gpe inside of the gpe handler, or inside of the worker thread that handles it ASthink

#

wait I guess I need to clear it at the handler, in case it wasn't for SCI_EVT

fiery turtle May 2, 2025, 8:29 PM

#

depends on the gpe

#

take a look when managarm clears it idr

#

or haiku is probably a better EC driver

trim frost May 2, 2025, 8:31 PM

#

what's fixed event 3?

fiery turtle May 2, 2025, 8:31 PM

#

see the enum in event.h

vast kestrel May 2, 2025, 8:31 PM

#

sleep button I think

#

trim frost May 2, 2025, 8:32 PM

#

yeah i get an error that that failed to enable when booting from my laptop and it makes sense

#

thanks

fiery turtle May 2, 2025, 8:33 PM

#

u get a sleep button event?

vast kestrel May 2, 2025, 8:34 PM

#

he tried to install a sleep button event

#

but it failed

fiery turtle May 2, 2025, 8:34 PM

#

sleep button is routed via notify probably

vast kestrel May 2, 2025, 8:34 PM

#

I think uacpi has some method to check if a specific fixed event is present in hardware

fiery turtle May 2, 2025, 8:35 PM

#

its a bit in fadt

vast kestrel May 2, 2025, 8:35 PM

#

oh wait no I confused it with uacpi_fixed_event_info

#

it has different stuff

trim frost May 2, 2025, 8:36 PM

#

10883 ops lol

#

vast kestrel May 2, 2025, 8:37 PM

#

ok I finally got to getting the query value from the EC

#

I now need to call QXX or something right

fiery turtle May 2, 2025, 8:38 PM

#

_QXX

vast kestrel May 2, 2025, 8:52 PM

#

It's doing stuff AmeliaEyes

#

I need to add the schedule work tho now

#

So sadge lif

fiery turtle May 2, 2025, 8:53 PM

#

damn very cool

#

also first time seeing hardware with multiple GPE blocks

vast kestrel May 2, 2025, 8:54 PM

#

Nice

#

It's called Intel ultra for a reason :^)

fiery turtle May 2, 2025, 8:54 PM

#

tru

#

does it also not have S3?

vast kestrel May 2, 2025, 8:55 PM

#

#

SS3 is set to zero it seems

fiery turtle May 2, 2025, 8:55 PM

#

modern standby crap

#

sad

vast kestrel May 2, 2025, 8:56 PM

#

it has s4 :^)

#

but yeah pretty cringe

#

when ACPI rejects you

fiery turtle May 2, 2025, 8:57 PM

#

oh yeah _OSC

#

no hobby os using it so far

vast kestrel May 2, 2025, 8:58 PM

#

there are a ton of debug prints in this aml ngl

#

I think most of it is from a template being replicated for each root port tho

fiery turtle May 2, 2025, 8:59 PM

#

fair

vast kestrel May 2, 2025, 9:00 PM

#

ok time to implement the work queue thing

leaden fox May 2, 2025, 9:06 PM

#

fiery turtle oh yeah _OSC

what's _OSC?

fiery turtle May 2, 2025, 9:09 PM

#

https://uefi.org/specs/ACPI/6.5/06_Device_Configuration.html#osc-operating-system-capabilities

leaden fox May 2, 2025, 9:10 PM

#

thx

hollow elm May 2, 2025, 9:17 PM

#

did a bios update on my laptop and unfortunately there are no acpi table changes

#

or well, two tpm-related addresses changed, but no logic changes unlike the previous bios update

fiery turtle May 2, 2025, 9:18 PM

#

because its already flawless clearly meme

hollow elm May 2, 2025, 9:19 PM

#

yeah meme ```
[ 1.046741] ACPI Error: AE_NOT_FOUND, While resolving a named reference package element - _PR_.P000 (20240827/dspkginit-438)
[ 1.046762] ACPI: _TZ_.THRM: _PSL evaluation failure
[ 1.046792] ACPI: thermal: [Firmware Bug]: Invalid critical threshold (-274000)

fiery turtle May 2, 2025, 9:20 PM

#

oh well

hollow elm May 2, 2025, 9:20 PM

#

i assume the philosophy is the classic "if it boots windows it's good enough"

fiery turtle May 2, 2025, 9:21 PM

#

windows also doesnt resolve named references in the interpreter

#

so maybe it works in the windows workflow somehow

vast kestrel May 2, 2025, 9:41 PM

#

#

It worksssssss

fiery turtle May 2, 2025, 9:42 PM

#

niice

fiery ice May 2, 2025, 9:42 PM

#

wait, how did you detect the lid change?

vast kestrel May 2, 2025, 9:42 PM

#

Also bonus thing, clicking the power button prints a process list

vast kestrel May 2, 2025, 9:42 PM

#

fiery ice wait, how did you detect the lid change?

Apci has a lid device

fiery ice May 2, 2025, 9:43 PM

#

oh cool

vast kestrel May 2, 2025, 9:43 PM

#

It triggers a notification on a change and you can query it

hollow elm May 2, 2025, 9:43 PM

#

fiery ice wait, how did you detect the lid change?

fucking magnets, how do they work

vast kestrel May 2, 2025, 9:43 PM

#

as simple as this https://github.com/Itay2805/uACPI-OS/blob/master/kernel/drivers/acpi-lid/lid.c

fiery ice May 2, 2025, 9:43 PM

#

thx

fiery turtle May 2, 2025, 9:44 PM

#

vast kestrel Also bonus thing, clicking the power button prints a process list

how do u shutdown then lol

vast kestrel May 2, 2025, 9:44 PM

#

it does need the EC in my case, but yeah

#

the problem is

#

that I fucking hate PS2

fiery ice May 2, 2025, 9:44 PM

#

XD

#

PS2 is needed unfortunately

vast kestrel May 2, 2025, 9:44 PM

#

especially with how they also mixed in the EC to it

#

its just a bunch of cancer to deal with correctly

#

I have no idea how my keyboard is connected on this laptop tho

#

the touchpad is i2c

#

from what I seen

hollow elm May 2, 2025, 9:45 PM

#

ps/2 kbd and i2c touchpad seems like a common arrangement

vast kestrel May 2, 2025, 9:45 PM

#

oh also I didn't actually implement the ec address space operations lol, since they were not needed yet

fiery ice May 2, 2025, 9:45 PM

#

keyboard is prob ps2

hollow elm May 2, 2025, 9:45 PM

#

that's how it is on my laptop

fiery ice May 2, 2025, 9:46 PM

#

same

frank canopy May 2, 2025, 9:46 PM

#

yeah its ps2 on my laptop as well

vast kestrel May 2, 2025, 9:46 PM

#

its so freaking cancer

#

how am I supposed to know when stuff is keyboard and when stuff is ec

hollow elm May 2, 2025, 9:46 PM

#

vast kestrel Also bonus thing, clicking the power button prints a process list

at least on your laptop the power button doesn't require an amd gpio driver :^)

vast kestrel May 2, 2025, 9:47 PM

#

the fuck is this

hollow elm May 2, 2025, 9:47 PM

#

virutal

fiery ice May 2, 2025, 9:48 PM

#

XDD

frank canopy May 2, 2025, 9:49 PM

#

lmao

#

its an intel usbgpio device obviously

fiery ice May 2, 2025, 9:49 PM

#

wtf is an intel usbgpio lol

hollow elm May 2, 2025, 9:49 PM

#

drivers/usb/misc/usb-ljca.c

#

it looks like it's an internal usb device that has an i2c, spi and gpio controller?

#

good shit

#

i hope i can some day get the same crack that hw engineers get

vast kestrel May 2, 2025, 9:51 PM

#

dafuq

#

why

#

whats wrong with the pch

#

the lpc exists for a reason

#

anyways I guess to actually continue and add a proper shell I will need an actual ps2 driver

#

sad lif

fiery turtle May 2, 2025, 9:56 PM

#

make a uPS2 while you're at it meme

vast kestrel May 2, 2025, 9:56 PM

#

glhf

#

ps2 is just too cancer

vast kestrel May 2, 2025, 10:51 PM

#

fiery turtle make a uPS2 while you're at it <:meme:575445094589661215>

Tbh it is kinda tempting to do so

#

Tho is uAPI a thing now

#

Did you ever continue working on it

fiery turtle May 2, 2025, 10:53 PM

#

we collectively abandoned it

#

we decided its not worth it

vast kestrel May 2, 2025, 10:54 PM

#

Sadge

#

So each driver should just request what it wants

#

With its own api

fiery turtle May 2, 2025, 10:55 PM

#

basically

vast kestrel May 2, 2025, 10:57 PM

#

At least lto is a thing

fiery turtle May 2, 2025, 10:57 PM

#

attribute alias as well

vast kestrel May 2, 2025, 11:02 PM

#

The main annoyance with PS2 is the fact that there is a lot of weird synchronization between the keyboard and mouse

#

And I think the data port is shared between the EC and PS2

#

Command port is different tho

fiery turtle May 2, 2025, 11:04 PM

#

vast kestrel And I think the data port is shared between the EC and PS2

huh

#

how would they synchronize it

vast kestrel May 2, 2025, 11:05 PM

#

Like, verify me if I'm wrong, but from what I seen 0x64 is the data and 0x66 is the command for the EC on my PC, and PS2 is 0x64 data and 0x60 command iirc

fiery turtle May 2, 2025, 11:07 PM

#

i mean yeah i think so

#

i just dont understand how that works

#

without race conditions

vast kestrel May 2, 2025, 11:08 PM

#

Maybe if you are fully interrupt driven it won't fire multiple interrupts (?)

#

Or well, how does it know when you are done

fiery turtle May 2, 2025, 11:09 PM

#

technically EC is entirely event driven

#

and irq driven

#

so idk

vast kestrel May 2, 2025, 11:43 PM

#

It seems like Linux has multiple layers of abstractions

#

So you have the keyboard/mouse on top of libps2 which is on top of serio

fiery turtle May 2, 2025, 11:47 PM

#

ofc it would have 10 layers for ps2 lol

#

its linux afterall

#

@vast kestrel should we put this kernel in the uacpi org after its done?

vast kestrel May 2, 2025, 11:49 PM

#

Sure

fiery turtle May 2, 2025, 11:49 PM

#

maybe I could add some CI runs at least in QEMU on every commit

vast kestrel May 2, 2025, 11:49 PM

#

I mean define done :^)

fiery turtle May 2, 2025, 11:49 PM

#

whenever u feel like its what u envision

vast kestrel May 2, 2025, 11:49 PM

#

Yeah having some automated tests could be cool

fiery turtle May 2, 2025, 11:50 PM

#

like i have tons of tests but none of them test kernel mode

vast kestrel May 2, 2025, 11:50 PM

#

I want this to be enough to take a random computer and ensure that uacpi works on it with a known good kernel

fiery turtle May 2, 2025, 11:50 PM

#

that would be cool yeah

vast kestrel May 2, 2025, 11:51 PM

#

So if someone comes with some hardware error they can run it first and know if it's expected or not with uacpi

#

It's also why I want to keep the kernel pretty simple

#

Or well, as simple as something that can run acpi can be

#

Btw it seems that the acpi EC in Linux doesn't use any of these stuff, it just works directly against the controller

#

(the libps2/serio that is)

#

Maybe it's up to the controller to ensure it doesn't mix stuff, even tho it would be funny if as long as the EC is doing stuff keyboard doesn't work lol

fiery turtle May 2, 2025, 11:57 PM

#

yeah its probably synchronized at chipset level

#

i like that u use uacpi_snprintf in your kernel

#

i was gonna do that as well lmao

hollow elm May 2, 2025, 11:59 PM

#

if rip in range(windows_ps2_driver_start, windows_ps2_driver_size): dont_do_ec_stuff()

fiery turtle May 3, 2025, 12:00 AM

#

I think both ps2 and ec have a bit that says if there's anything in the data port

#

So that's how they synchronize probably

#

Although I wonder if that means the keyboard hangs if ec is not handled

#

And its most likely also routed via i2c so you don't even have to use ps2

#

So maybe they assume if you have an ec driver u also have i2c

vast kestrel May 3, 2025, 12:09 AM

#

But the keyboard stays as PS2 no?

vast kestrel May 3, 2025, 12:10 AM

#

fiery turtle i like that u use uacpi_snprintf in your kernel

Idk if you noticed but all of the error handling is based on uacpi_status lol

fiery turtle May 3, 2025, 12:15 AM

#

vast kestrel Idk if you noticed but all of the error handling is based on uacpi_status lol

Nice

fiery turtle May 3, 2025, 12:15 AM

#

vast kestrel But the keyboard stays as PS2 no?

No idea

#

But yeah the kernel is shaping up great so far

vast kestrel May 3, 2025, 12:16 AM

#

Yeah I like how it's built, it's relatively clean

#

I need to cleanup the scheduler a little, since right now everything stays on the first core

fiery turtle May 3, 2025, 12:24 AM

#

You could also add support for memory and CPU hotplug in qemu for fun

#

Well and PCI

vast kestrel May 3, 2025, 12:28 AM

#

That could be fun

#

And I can test it in qemu unlike EC stuff

#

Smuggy

fiery turtle May 3, 2025, 12:29 AM

#

Its a matter of installing notify

#

Yeah

vast kestrel May 3, 2025, 12:29 AM

#

I think implementing most of the standard acpi devices could be nice

#

As a reference

fiery turtle May 3, 2025, 12:30 AM

#

Yeah

#

And AMD GPIO chad

#

To support every way of routing a power button

vast kestrel May 3, 2025, 12:30 AM

#

For a second I read it as AMD GPU 3sCoconaSweat

fiery turtle May 3, 2025, 12:30 AM

#

Mishakov did it for his kernel

vast kestrel May 3, 2025, 12:30 AM

#

Nice

fiery turtle May 3, 2025, 12:30 AM

#

Its not that big of a driver

vast kestrel May 3, 2025, 12:31 AM

#

90% of GPIO drivers is a bit mask of config and data :^)

fiery turtle May 3, 2025, 12:31 AM

#

Yah

vast kestrel May 3, 2025, 12:31 AM

#

Intel chipsets have a ton of stuff to route GPIO as gpes and smis and scis

fiery turtle May 3, 2025, 12:31 AM

#

There's also GED on riscv

#

Which is like EC but simpler

fiery turtle May 3, 2025, 12:32 AM

#

vast kestrel Intel chipsets have a ton of stuff to route GPIO as gpes and smis and scis

Ye

vast kestrel May 3, 2025, 12:32 AM

#

Having the kernel be multi-arch would be interesting

fiery turtle May 3, 2025, 12:32 AM

#

Definitely

vast kestrel May 3, 2025, 12:32 AM

#

Hardware reduced acpi support :^)

#

Well, I am not adding support for other arches so PRs are welcomed AMtoiletspin

fiery turtle May 3, 2025, 12:33 AM

#

Lol

fiery turtle May 3, 2025, 12:33 AM

#

vast kestrel Hardware reduced acpi support :^)

More like lack of support for other features

vast kestrel May 3, 2025, 12:34 AM

#

There is some random, like gpe devices or whatever it was

#

Tho no idea if any hardware actually uses it

fiery turtle May 3, 2025, 12:34 AM

#

The ged yeah, but its 100 lines

#

Once you have all these helpers the drivers just fall into place

vast kestrel May 3, 2025, 12:37 AM

#

I just remembered I wanted to suggest some uacpi helper function but I completely forgot what it was

fiery turtle May 3, 2025, 12:38 AM

#

Lol

trim frost May 3, 2025, 4:27 AM

#

fiery turtle no hobby os using it so far

That's wrong, I use it but I don't restore the state (so it's just a shutdown) lol

calm latch May 3, 2025, 8:00 AM

#

fiery turtle So maybe they assume if you have an ec driver u also have i2c

How do you support i2c?

calm latch May 3, 2025, 8:02 AM

#

fiery turtle There's also GED on riscv

I think it didn't get merged to QEMU risc

leaden fox May 3, 2025, 8:48 AM

#

uUSB when

fiery turtle May 3, 2025, 9:19 AM

#

calm latch How do you support i2c?

Smbus drivers

vast kestrel May 3, 2025, 9:46 AM

#

Ok I think I realized something, because the command registers are different, it can know if what you are about to write is for keyboard or EC

fiery turtle May 3, 2025, 9:47 AM

#

probably

vast kestrel May 3, 2025, 9:47 AM

#

And assuming that the EC is not buggy, indeed the OBF can be set for either

#

So in the write case if you start a kbd/EC command it will wait for that one

#

And I guess it's the same for IBF

#

Wait no that's different

#

Even tho

#

I wonder if you write to the EC if the kbd will have IBF set as well

fiery turtle May 3, 2025, 9:49 AM

#

that would most definitely break lol

vast kestrel May 3, 2025, 9:49 AM

#

Also what if you write to one reg and then another

#

Because Linux doesn't lock it between them as we saw

fiery turtle May 3, 2025, 9:50 AM

#

i mean u cant write there randomly

#

only if it tells u that its empty

vast kestrel May 3, 2025, 9:50 AM

#

No I mean the command

#

Imagine writing command to EC and then KBD

fiery turtle May 3, 2025, 9:50 AM

#

command regs arent shared

#

so

vast kestrel May 3, 2025, 9:51 AM

#

It can work if that sets the IBF on the second command reg

fiery turtle May 3, 2025, 9:51 AM

#

the order in which the "data full" bit wil be set should be synchronzed

vast kestrel May 3, 2025, 9:51 AM

#

And it stays full until you are done with the other thing

fiery turtle May 3, 2025, 9:51 AM

#

probably

#

like EC is not polled so u just wait for it to tell u via irq

vast kestrel May 3, 2025, 9:51 AM

#

I wonder if there is any open source EC firmware

fiery turtle May 3, 2025, 9:51 AM

#

i remember reading an implementation

#

someone posted a link

#

dont remember where

#

maybe in this thread

vast kestrel May 3, 2025, 9:52 AM

#

I'll search when I get home

hasty plinth May 3, 2025, 10:32 AM

#

the framework laptop ec firmware is open source

#

if thats of any use

vast kestrel May 3, 2025, 10:38 AM

#

yeah that works

#

lets see

#

I sure do like it when the firmware asks a question

#

ok actually all the 3 cases are funny

#

ok so this is the keyboard handler https://github.com/system76/ec/blob/86dbc9ba3a6d0a77a54208037c99990188a3f660/src/board/system76/common/kbc.c#L453

#

it seems to run once every 1ms

#

which sounds a bit high ngl

#

#

the output will check if the state allows it to send back keyboard/touchpad/mouse data

#

so while you are doing a command it will not send keyboard data

#

but outside of a command it will

fiery turtle May 3, 2025, 10:47 AM

#

Is there a KBC STATE EC?

vast kestrel May 3, 2025, 10:47 AM

#

nope

#

do we have the acpi tables for System76 laptops?

fiery turtle May 3, 2025, 10:47 AM

#

Maybe its not even shared on the framework

fiery turtle May 3, 2025, 10:47 AM

#

vast kestrel do we have the acpi tables for System76 laptops?

I have one dump

vast kestrel May 3, 2025, 10:47 AM

#

gib

fiery turtle May 3, 2025, 10:48 AM

#

Scroll up

vast kestrel May 3, 2025, 10:48 AM

#

actually lemme find the EC stuff

fiery turtle May 3, 2025, 10:48 AM

#

📎 acpi.tar

vast kestrel May 3, 2025, 10:50 AM

#

WTF

fiery turtle May 3, 2025, 10:51 AM

#

linux would use the first two probably

#

the rest are just for bookkeeping

#

is 62/66 shared with kbd?

hollow elm May 3, 2025, 10:52 AM

#

ps/2 uses 0x60 and 0x64

vast kestrel May 3, 2025, 10:52 AM

#

nope those are not shared

#

maybe I was dreaming

#

hold on

#

yeah skill issue on my side AMfacepalm

#

its 0x62 as well

#

I remembered 0x60

#

ok so there is actually no problem

#

nice

#

does it actually use the GPIO instead of SCI on the system76 laptop?

hollow elm May 3, 2025, 10:54 AM

#

if it's an amd laptop that's not unlikely

#

me and @calm latch have laptops that wire the power button through the amd gpio gunk

vast kestrel May 3, 2025, 10:55 AM

#

ok so there is no problem with ps2 and ec, just me being completely blind

#

so time to add a ps2 driver for fun and profit

#

one thing I do wonder is how the ec handles the mouse stuff

hollow elm May 3, 2025, 10:57 AM

#

wdym? if the touchpad is i2c it doesn't go through the ec

vast kestrel May 3, 2025, 10:57 AM

#

wait fuck yeah it won't do the dual port mouse in there

hollow elm May 3, 2025, 10:57 AM

#

you need an i2c controller driver and it just uses hid over i2c

vast kestrel May 3, 2025, 10:57 AM

#

I was hoping to see how the controller handles it when ps2 has two devices Sad

hollow elm May 3, 2025, 10:58 AM

#

hollow elm you need an i2c controller driver and it just uses hid over i2c

usually a gpio for irqs

vast kestrel May 3, 2025, 10:59 AM

#

it seems on my laptop the i2c controllers are discoverable over pci/pcie

#

anyways I don't need mouse for this, keyboard is enough

hollow elm May 3, 2025, 11:00 AM

#

vast kestrel it seems on my laptop the i2c controllers are discoverable over pci/pcie

afaicu you still need to associate it with the aml node for it, since the aml for the touchpad refers to the controller that way

vast kestrel May 3, 2025, 11:01 AM

#

yeah it has a node in the AML

#

but that is standard PCI stuff

#

@fiery turtle is there some helper to lookup a pci node under a pci controller device?

fiery turtle May 3, 2025, 11:01 AM

#

Nop

#

Your i2c mouse probably has ps2 emulation anyway

vast kestrel May 3, 2025, 11:02 AM

#

ps2 emulation is turned off when you enable acpi

trim frost May 3, 2025, 11:04 AM

#

my i2c touchpad didnt have ps2 emulation even before i added acpi to my kernel

vast kestrel May 3, 2025, 11:24 AM

#

@fiery turtle do I need to call the _STA myself to ensure that I can deal with the device?

#

or well, that it exists

fiery turtle May 3, 2025, 11:28 AM

#

vast kestrel <@353593948523986945> do I need to call the _STA myself to ensure that I can dea...

what context

vast kestrel May 3, 2025, 11:28 AM

#

for example in the acpi driver dispatcher

fiery turtle May 3, 2025, 11:28 AM

#

uacpi_find_devices checks automatically

#

everything else u have to check yourself

vast kestrel May 3, 2025, 11:29 AM

#

so because I am using uacpi_namespace_for_each_child I need to call _STA myself

#

got it

fiery turtle May 3, 2025, 11:29 AM

#

in the main dispatcher u can just use uacpi_namespace_node_info

#

or actually, that doesnt have sta but yeah

#

sta u check yourself

#

uacpi_eval_sta

vast kestrel May 3, 2025, 11:30 AM

#

ok cool

#

should I first check _STA and then HID/CID or the opposite ASthink

#

I feel like first check _STA

fiery turtle May 3, 2025, 11:30 AM

#

_STA first, then namespace_node_info

#

which collects everything there is about a device

#

then u can just give that struct to probe for example

#

including uid, hid and cid

#

uacpi_get_namespace_node_info

vast kestrel May 3, 2025, 11:31 AM

#

oh I just now noticed that the node info has the cid and hid

#

I called the others manually until now lol

fiery turtle May 3, 2025, 11:31 AM

#

yeah u dont have to

vast kestrel May 3, 2025, 11:32 AM

#

I used it to check if it has an hid and then called the eval manually

#

lmao

fiery turtle May 3, 2025, 11:32 AM

#

theres also eval_{hid,cid} if u really want to do it manually

vast kestrel May 3, 2025, 11:32 AM

#

yeah that is what I did

fiery turtle May 3, 2025, 11:32 AM

#

vast kestrel I used it to check if it has an hid and then called the eval manually

bruh

#

u could also just unconditionally eval it and check against UACPI_STATUS_NOT_FOUND

vast kestrel May 3, 2025, 11:32 AM

#

meh

fiery turtle May 3, 2025, 11:33 AM

#

it cant collide with aml errors because aml errors have a separate enum value

vast kestrel May 3, 2025, 11:33 AM

#

I already use the info so why not

fiery turtle May 3, 2025, 11:33 AM

#

yea

trim frost May 3, 2025, 11:55 AM

#

how can i get battery info from uacpi?

vast kestrel May 3, 2025, 11:56 AM

#

there is acpi smart battery

fiery turtle May 3, 2025, 12:03 PM

#

trim frost how can i get battery info from uacpi?

_BIF/_BIX

#

look at managarm

vast kestrel May 3, 2025, 12:34 PM

#

heyyy ps2 works

#

ngl

#

its really clean to have ACPI as a device bus

#

you just see devices and initialize them

fiery turtle May 3, 2025, 12:38 PM

#

nice

#

ultra progress this week: I made an atomic.h header

#

trl

#

uacpios progress: full scheduler, work queues, every acpi driver in existance, allocators

vast kestrel May 3, 2025, 1:11 PM

#

Smuggy

#

Tbf it helps to not care too much about performance or scalability

#

The sync primitives I made are based on go, so it's not too bad

slim panther May 3, 2025, 1:43 PM

#

wtf

#

is it normal for hw to time out or is that my fault?

left orbit May 3, 2025, 1:44 PM

#

bro flanterm looks so good

vast kestrel May 3, 2025, 1:44 PM

#

most likely your fault

#

especially for changing modes

fiery ice May 3, 2025, 1:45 PM

#

left orbit bro flanterm looks so good

how did you lose your introduced role?

left orbit May 3, 2025, 1:45 PM

#

i'm built different

slim panther May 3, 2025, 1:46 PM

#

vast kestrel especially for changing modes

what does that entail

vast kestrel May 3, 2025, 1:46 PM

#

lets see what uacpi does exactly

#

yeah it just writes an smi and waits for the mode to change

left orbit May 3, 2025, 1:47 PM

#

it just pokes fadt registers

north hull May 3, 2025, 1:47 PM

#

slim panther wtf

is that two webcams?

slim panther May 3, 2025, 1:47 PM

#

no

left orbit May 3, 2025, 1:47 PM

#

oh it writes an smi register

#

nvm

vast kestrel May 3, 2025, 1:47 PM

#

slim panther May 3, 2025, 1:47 PM

#

light sensor

north hull May 3, 2025, 1:47 PM

#

ah

left orbit May 3, 2025, 1:47 PM

#

makes sense

#

    [UACPI_REGISTER_SMI_CMD] = {
        .kind = REGISTER_KIND_IO,
        .access_kind = REGISTER_ACCESS_KIND_NORMAL,
        .access_width = 1,
        .accessors = { &g_uacpi_rt_ctx.fadt.smi_cmd, },
    },```

vast kestrel May 3, 2025, 1:47 PM

#

it waits for the SCI_EN to be set

left orbit May 3, 2025, 1:48 PM

#

so either your io kernel api is broken or fadt is garbage

#

i'm guessing the first one

#

https://github.com/menix-os/menix/blob/d763d49794267e9c2c7705607694a2c9d5926cdb/kernel/src/generic/platform/acpi/uacpi.rs#L126C1-L147C2

#

bro WHATWEASJITYGSEROJHSRDOIJFUK

vast kestrel May 3, 2025, 1:49 PM

#

lmao

left orbit May 3, 2025, 1:49 PM

#

OH

#

OMG

#

wait

slim panther May 3, 2025, 1:49 PM

#

?#

#

that's not there

#

anymore

left orbit May 3, 2025, 1:49 PM

#

which branch am i supposed ot be

#

looking at

#

😭😭

slim panther May 3, 2025, 1:49 PM

#

remove-talc

left orbit May 3, 2025, 1:49 PM

#

THANK god

#

where tf is ur io_read

slim panther May 3, 2025, 1:50 PM

#

arch

left orbit May 3, 2025, 1:50 PM

#

bruh

#

this guy💀

#

also ur not returning a handle

slim panther May 3, 2025, 1:50 PM

#

well what tf am i supposed to return

#

0?

left orbit May 3, 2025, 1:50 PM

#

u need to know

#

the base and length

#

well, the length might not be that important

slim panther May 3, 2025, 1:50 PM

#

idfk what this function does man

left orbit May 3, 2025, 1:50 PM

#

but the base yes

slim panther May 3, 2025, 1:50 PM

#

i keep forgetting

left orbit May 3, 2025, 1:50 PM

#

uacpi might map with base>0

#

and u need to keep track of that

fiery ice May 3, 2025, 1:51 PM

#

slim panther idfk what this function does man

which one

vast kestrel May 3, 2025, 1:51 PM

#

io_map takes the base, and io_read reads from the mapped relative to the mapped

slim panther May 3, 2025, 1:51 PM

#

the io_map functions

vast kestrel May 3, 2025, 1:51 PM

#

so if it maps 0x100, then read will read 0 to read the 0x100 port

left orbit May 3, 2025, 1:51 PM

#

it's base+offset for io_read

fiery ice May 3, 2025, 1:51 PM

#

for x86_64 basically nothing

vast kestrel May 3, 2025, 1:51 PM

#

or will read 1 to read 0x101

left orbit May 3, 2025, 1:51 PM

#

like ts so simple man

slim panther May 3, 2025, 1:51 PM

#

oh yea i forgor to do this

fiery ice May 3, 2025, 1:51 PM

#

slim panther May 3, 2025, 1:51 PM

#

smh

left orbit May 3, 2025, 1:52 PM

#

#[unsafe(no_mangle)]
extern "C" fn uacpi_kernel_io_read8(
    arg1: uacpi_handle,
    offset: uacpi_size,
    out_value: *mut uacpi_u8,
) -> uacpi_status {
    unsafe {
        (*out_value) = asm::read8((arg1 as usize + offset) as u16);
    }
    return UACPI_STATUS_OK;
}```

#

how did u implement it correctly

#

and then say

#

idk how it works...

#

also io_map/unmap should be arch specific too in that case

fiery ice May 3, 2025, 1:52 PM

#

that is io_read

left orbit May 3, 2025, 1:52 PM

#

no shit

#

i can read

#

but he uses the handle

fiery ice May 3, 2025, 1:52 PM

#

ah

left orbit May 3, 2025, 1:52 PM

#

and he never initializes it

#

he uses the handle as if it was the base

#

just like many kernels do

fiery ice May 3, 2025, 1:54 PM

#

left orbit May 3, 2025, 1:55 PM

#

if u wanna be cool then ```rs
#[unsafe(no_mangle)]
extern "C" fn uacpi_kernel_io_map(
base: uacpi_io_addr,
len: uacpi_size,
out_handle: *mut uacpi_handle,
) -> uacpi_status {
let io_window = Box::new((base, len));
unsafe {
out_handle.write(Box::into_raw(io_window));
}
return UACPI_STATUS_OK;
}

and then u can do bounds checking in io_read{8,16,32}

#

or you can just out_handle.write(base as uacpi_handle);

fiery turtle May 3, 2025, 2:02 PM

#

slim panther wtf

if i had a penny for every aml blob

slim panther May 3, 2025, 2:03 PM

#

anyways

#

yes works now

left orbit May 3, 2025, 2:04 PM

#

🗿

#

nice

#

i'm assuming that "now" means after implementing io_map properly?

#

slim panther May 3, 2025, 2:04 PM

#

true

left orbit May 3, 2025, 2:04 PM

#

epic

fiery turtle May 3, 2025, 2:05 PM

#

so basically previously u would write to random io registers

#

causing ur hw to do absolutely random shit

#

lol

slim panther May 3, 2025, 2:05 PM

#

wow on the fw13 there are 0 acpi warnings or errors

left orbit May 3, 2025, 2:05 PM

#

who would've thought that actually implementing kernel apis would make things work

#

clueless

vast kestrel May 3, 2025, 2:20 PM

#

I think my battery is low :^)

#

It started spamming these notifications

#

Also when connecting to power it gave a response to the query that aml doesn't have a handler for ASthink

fiery turtle May 3, 2025, 2:58 PM

#

Lol

fiery turtle May 3, 2025, 2:58 PM

#

vast kestrel It started spamming these notifications

How often

vast kestrel May 3, 2025, 2:58 PM

#

every few seconds

#

like 2 seconds or so

fiery turtle May 3, 2025, 2:58 PM

#

That's probably fine

#

Just discharge

vast kestrel May 3, 2025, 2:58 PM

#

yeah yeah

fiery turtle May 3, 2025, 2:58 PM

#

vast kestrel Also when connecting to power it gave a response to the query that aml doesn't h...

That's strange tho

#

So u don't get a notify when plugging in?

vast kestrel May 3, 2025, 2:59 PM

#

lemme quickly fix so it won't throw on it but warn instead

#

and see what it tried to call

#

#

Q37 and Q38

#

they were called

#

but they reference something that doesn't exist?

#

actually it doesn't have the same error as what I had before

#

oh wait

#

maybe because I threw in from this it caused the entire thread to abort

#

idk

#

anyways it seems to be working, and the PNOT not exists is true, its not defined anywhere

#

but its after the entire handler

#

#

or well, in Q38 its in the middle

#

I should add the smart battery device

#

but no first keyboard and shell

fiery turtle May 3, 2025, 3:08 PM

#

vast kestrel

is PNOT a thing?

vast kestrel May 3, 2025, 3:09 PM

#

nope

fiery turtle May 3, 2025, 3:09 PM

#

wh yis there a _Q4f lol

#

do u not upper case them

vast kestrel May 3, 2025, 3:09 PM

#

fiery turtle May 3, 2025, 3:09 PM

#

maybe SSDTs

vast kestrel May 3, 2025, 3:09 PM

#

yeah I do it lower case think

#

wait is that a problem

fiery turtle May 3, 2025, 3:10 PM

#

aml names are upper case so yes

vast kestrel May 3, 2025, 3:10 PM

#

ok that probably explains the not found I had later

#

where did you even see 4f

fiery turtle May 3, 2025, 3:10 PM

#

yep

vast kestrel May 3, 2025, 3:10 PM

#

OHOHOH ok now I see it

#

that is what that failed before

#

and because I exited the thread from the failure it didn't continue to the other handler

#

ok time to fix

#

Smuggy

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