uACPI - a portable and easy-to-integrate ACPI implementation | Operating System Development | Page 71

hallow marten Jun 21, 2026, 7:16 PM

#

then linux

slim panther Jun 21, 2026, 7:16 PM

#

huge

hallow marten Jun 21, 2026, 7:25 PM

#

what IRC client do you use btw?

#

is it konversation?

full nimbus Jun 21, 2026, 7:25 PM

#

This looks like the lounge?

hallow marten Jun 21, 2026, 7:25 PM

#

what

full nimbus Jun 21, 2026, 7:25 PM

#

Atleast the color scheme and layout

full nimbus Jun 21, 2026, 7:25 PM

#

hallow marten what

Self hostable web client

slim panther Jun 21, 2026, 7:26 PM

#

hallow marten what IRC client do you use btw?

the lounge

full nimbus Jun 21, 2026, 7:26 PM

#

Called it

#

slim panther Jun 21, 2026, 7:26 PM

#

it's a web client

hallow marten Jun 21, 2026, 7:26 PM

#

#

I like weechat

slim panther Jun 21, 2026, 7:26 PM

#

on zinnia i use irssi

#

or sometimes thunderbird but it kinda sucks

jaunty fox Jun 21, 2026, 7:27 PM

#

hallow marten then linux

we just need to hire one of those hardcore PR firms

#

they can mount a smear campaign against ACPICA

#

as OGs of this thread know there is more than enough fuel for that fire

full nimbus Jun 21, 2026, 7:28 PM

#

jaunty fox they can mount a smear campaign against ACPICA

I mean the fact they have been slopping a ton of shit apparently and breaking things

#

We don't even need to smear them

#

Just tell the truth

jaunty fox Jun 21, 2026, 7:29 PM

#

full nimbus I mean the fact they have been slopping a ton of shit apparently and breaking th...

they broke things real good

fiery turtle Jun 21, 2026, 7:29 PM

#

Who is that?

jaunty fox Jun 21, 2026, 7:29 PM

#

just ask the guy with the corvette

full nimbus Jun 21, 2026, 7:30 PM

#

ACPICA also killed my parents

fiery turtle Jun 21, 2026, 7:30 PM

#

the acpica maintainer corvette incident

slim panther Jun 21, 2026, 7:30 PM

#

fiery turtle Who is that?

zircon guy

fiery turtle Jun 21, 2026, 7:31 PM

#

zircon?

jaunty fox Jun 21, 2026, 7:32 PM

#

fiery turtle zircon?

fuschia/magenta

hallow marten Jun 21, 2026, 7:33 PM

#

fiery turtle Who is that?

little kernel guy

pallid lava Jun 21, 2026, 7:34 PM

#

uacpi first got into react

vocal geyser Jun 21, 2026, 7:34 PM

#

full nimbus ACPICA also killed my parents

🥀

pallid lava Jun 21, 2026, 7:34 PM

#

then it has to get into fuchsia, linux

#

finally windows

fiery turtle Jun 21, 2026, 7:34 PM

#

jaunty fox fuschia/magenta

ah cool

pallid lava Jun 21, 2026, 7:34 PM

#

uacpi becomes the new reference

full nimbus Jun 21, 2026, 7:34 PM

#

vocal geyser 🥀

Epstein also used ACPICA trl

pallid lava Jun 21, 2026, 7:34 PM

#

sunflush

vocal geyser Jun 21, 2026, 7:34 PM

#

full nimbus Epstein also used ACPICA <:trl:1477775551505633320>

interesting

hallow marten Jun 21, 2026, 7:35 PM

#

we should tell israel ACPICA is anti-semitic

#

they'll handle it in no time

full nimbus Jun 21, 2026, 7:35 PM

#

True

#

😭

vocal geyser Jun 21, 2026, 7:36 PM

#

🙁

Screenshot_2026-06-21-22-36-25-266_com.android.chrome-edit.jpg

pallid lava Jun 21, 2026, 7:36 PM

#

wait

#

geist is in this server

#

now I remember

fiery turtle Jun 21, 2026, 7:49 PM

#

vocal geyser 🙁

It's funny how it's saying "please try a different search", as if it really wants you to dig something up

lethal trench Jun 21, 2026, 8:32 PM

#

vocal geyser 🙁

Would of been funny if some random email from Epstein like “this acpica thing is really cool” came up

kind mantle Jun 21, 2026, 8:32 PM

#

doubt

#

it was basically never happening

#

it's just too niche

full nimbus Jun 21, 2026, 8:34 PM

#

kind mantle it was basically never happening

I mean If Epstein ever used Linux

#

He used ACPICA by proxy

#

Which was the joke I was going for

hallow marten Jun 21, 2026, 8:37 PM

#

#

this is in epstein's mail

full nimbus Jun 21, 2026, 8:40 PM

#

hallow marten

everyone0 everyone1 everyone2 omg look

steel inlet Jun 21, 2026, 8:40 PM

#

full nimbus I mean If Epstein ever used Linux

there is a bash manual in the files

fiery turtle Jun 21, 2026, 8:40 PM

#

lol

steel inlet Jun 21, 2026, 8:41 PM

#

and i saw some emails of orders for hp linux servers or smth iirc

#

not sure what that was about

hallow marten Jun 21, 2026, 8:43 PM

#

#

bunch of linux mentions from that

full nimbus Jun 21, 2026, 8:46 PM

#

steel inlet and i saw some emails of orders for hp linux servers or smth iirc

SEE! SEE!

lethal trench Jun 21, 2026, 8:47 PM

#

hallow marten

omg real?

#

(:P)

pallid lava Jun 23, 2026, 11:22 AM

#

got back in my hands an old laptop with a touchscreen

#

the madt is fucked

#

#

non-sequential APIC ids

#

#

this one is subtler, but it specifies NMIs as level triggered

#

which is prohibited by the intel sdm

#

and i cant get my os to proceed because i get an apic error

#

someone is sending me an IPI with vector 0-15

#

it is not me

kindred beacon Jun 23, 2026, 11:24 AM

#

pallid lava got back in my hands an old laptop with a touchscreen

do you know how the touch is wired? usb/i2c/something else?

pallid lava Jun 23, 2026, 11:24 AM

#

no idea

#

the laptop is super fucked overall

#

the quality is bad

#

the keyboard's up arrow is spammed constantly

kindred beacon Jun 23, 2026, 11:25 AM

#

you can determine if it's usb by looking at lsusb output

pallid lava Jun 23, 2026, 11:25 AM

#

typing commands is a pain when im in grub

#

but i have ssh so

#

i can check it now

kindred beacon Jun 23, 2026, 11:25 AM

#

alternatively you can get a nice listing of all input devices with evtest

fiery turtle Jun 23, 2026, 11:26 AM

#

pallid lava this one is subtler, but it specifies NMIs as level triggered

Thats why linux never even looks at nmi entries

fiery turtle Jun 23, 2026, 11:26 AM

#

pallid lava non-sequential APIC ids

Half of my laptops are this way lol, they're rarely sequential

pallid lava Jun 23, 2026, 11:26 AM

#

fiery turtle Thats why linux never even looks at nmi entries

so how does it wire the nmi handler?

#

when i type nmi in the qemu monitor it is received by linux

#

how does it know what parameters and what lints

fiery turtle Jun 23, 2026, 11:26 AM

#

It hardcodes the classic config

#

Its the only one that works

pallid lava Jun 23, 2026, 11:28 AM

#

i see

#

@kindred beacon ```
Available devices:
/dev/input/event0: Lid Switch
/dev/input/event1: Power Button
/dev/input/event10: HDA Intel PCH Mic
/dev/input/event11: HDA Intel PCH Speaker Front
/dev/input/event12: HDA Intel PCH HDMI/DP,pcm=3
/dev/input/event13: HDA Intel PCH HDMI/DP,pcm=7
/dev/input/event14: HDA Intel PCH HDMI/DP,pcm=8
/dev/input/event2: AT Translated Set 2 keyboard
/dev/input/event3: Video Bus
/dev/input/event4: Intel HID events
/dev/input/event5: Intel HID 5 button array
/dev/input/event6: PC Speaker
/dev/input/event7: FTSC1000:00 2808:50C5
/dev/input/event8: FTSC1000:00 2808:50C5 UNKNOWN
/dev/input/event9: SYNA3602:00 0911:5288 Touchpad

#

the touchscreen is event7

kindred beacon Jun 23, 2026, 11:31 AM

#

what the fuck

fiery turtle Jun 23, 2026, 11:31 AM

#

synaptics touchpad letsgo

pallid lava Jun 23, 2026, 11:32 AM

#

so yes, over i2c

kindred beacon Jun 23, 2026, 11:32 AM

#

jeez

#

good luck

pallid lava Jun 23, 2026, 11:32 AM

#

oh by the way

#

this laptop is from 2017

#

and has no csm

#

quite visionary

kindred beacon Jun 23, 2026, 11:32 AM

#

that's based actually

fiery turtle Jun 23, 2026, 11:32 AM

#

who needs csm tbh

kindred beacon Jun 23, 2026, 11:32 AM

#

fuck that old cruft

pallid lava Jun 23, 2026, 11:33 AM

#

back in 2020 i was forced to use tomatboot 😔

fiery turtle Jun 23, 2026, 11:33 AM

#

i was surprised to learn none of my laptops support the PXE GET_FILE_SIZE request, which is a fucking pain

#

havent tried the bios PXE yet, maybe that one is better

pallid lava Jun 23, 2026, 11:33 AM

#

let me now debug harder what is the apic error issue and i'll try to see how uacpi goes

kindred beacon Jun 23, 2026, 11:33 AM

#

only use TFTP for limine + kernel binary + initrd, NVMe-oF for the rest meme

pallid lava Jun 23, 2026, 11:34 AM

#

fiery turtle havent tried the bios PXE yet, maybe that one is better

i have written a pxe rom chad

fiery turtle Jun 23, 2026, 11:34 AM

#

pallid lava let me now debug harder what is the apic error issue and i'll try to see how uac...

the apic error in your kernel u mean?

pallid lava Jun 23, 2026, 11:34 AM

#

yes

fiery turtle Jun 23, 2026, 11:34 AM

#

ah

fiery turtle Jun 23, 2026, 11:34 AM

#

pallid lava i have written a pxe rom <:chad:815144781842743334>

for what card?

pallid lava Jun 23, 2026, 11:34 AM

#

virtio net

#

trl

kindred beacon Jun 23, 2026, 11:34 AM

#

actual troll

fiery turtle Jun 23, 2026, 11:34 AM

#

doesnt that already have one?

pallid lava Jun 23, 2026, 11:35 AM

#

yeah ofc

fiery turtle Jun 23, 2026, 11:35 AM

#

ah

pallid lava Jun 23, 2026, 11:40 AM

#

the storage is also unironically an emmc card

mortal yoke Jun 23, 2026, 11:42 AM

#

I have always wondered why do the cheap laptops/chromebooks use emmc instead of ufs

#

its literally worse than phones lol

calm latch Jun 23, 2026, 11:45 AM

#

It's cheaper?

#

UFS is like in the flagship smartphones, and only became a thing recently

#

I mean everyone uses NVMe now?

#

Smartphones, SD cards

mortal yoke Jun 23, 2026, 11:46 AM

#

calm latch UFS is like in the flagship smartphones, and only became a thing recently

its not flagships only, mid-range phones in 2020 already had it for an example

#

like oneplus nord n10 that cost like 250€ back then

vast kestrel Jun 23, 2026, 11:48 AM

#

I think it's because emmc works over SD bus that chipsets already have

#

Meanwhile ufs is a different bus

#

And nvme is just pcie

full nimbus Jun 23, 2026, 11:48 AM

#

how much cheaper even is emmc to skip nvmes? iirc you can get 256gb ones for like 30 bucks wholesale

mortal yoke Jun 23, 2026, 11:49 AM

#

the nice thing about UFS over emmc is that the controller interface is standardized (though ofc it can also come with quirks like all hardware)

pallid lava Jun 23, 2026, 12:12 PM

#

rip, i have to actualy implement the ec region handlers nooo

#

no idea why the apic error happens

#

it happens only once after i sti

#

the error status register has bit 6 set, which means that someone is sending me an interrupt with a vector from 0 to 15, which is illegal

#

it happens with the other cpu too when i bring it up

pallid lava Jun 23, 2026, 12:15 PM

#

pallid lava rip, i have to actualy implement the ec region handlers <:nooo:75046687099322375...

okay, this took me a grand total of 5 seconds

#

also, during initialization, i get a gpe before they're enabled

#

okay, there is something worse happening

#

when uacpi calls the battery's _STA, it accesses a register inside the embedded controller space

#

and i have no handler for it, because i have to first initialize the namespace, and then iterate over the devices looking for the embedded controller, and after that attach a handler?

calm latch Jun 23, 2026, 12:23 PM

#

mortal yoke the nice thing about UFS over emmc is that the controller interface is standardi...

Isn't emmc also standardized?

#

Like it's just the SD card

fiery turtle Jun 23, 2026, 12:28 PM

#

pallid lava and i have no handler for it, because i have to first initialize the namespace, ...

ECDT table

pallid lava Jun 23, 2026, 12:28 PM

#

i dont have that

fiery turtle Jun 23, 2026, 12:28 PM

#

Lol

#

You can find the ec device before initializing the namespace, after loading it

pallid lava Jun 23, 2026, 12:29 PM

#

oh okay

fiery turtle Jun 23, 2026, 12:29 PM

#

Idr how linux handles it exactly, I did look into it a while ago

mortal yoke Jun 23, 2026, 12:35 PM

#

calm latch Isn't emmc also standardized?

the thing itself yes, I guess it depends on the device whether it uses the standardized SD host controller interface or something proprietary

#

though even on devices that implement the sd host controller interface sometimes they are pretty broken and you have to write values to fields that are marked as RO in the spec (e.g. a register where you are supposed to discover supported voltages from so then the generic driver can properly read them from there) meme

slim panther Jun 23, 2026, 6:59 PM

#

pallid lava non-sequential APIC ids

is that not legal?

pallid lava Jun 23, 2026, 7:00 PM

#

why would it not be legal

#

it's just quirky lol

#

mfs want to be unique

vocal geyser Jun 23, 2026, 7:03 PM

#

slim panther is that not legal?

my laptop have 0 and 2 lapic ids

slim panther Jun 23, 2026, 7:03 PM

#

pallid lava why would it not be legal

idk otherwise who cares

#

if it works anyways

#

we know firmware devs are insane

vast kestrel Jun 23, 2026, 8:56 PM

#

It is very common, because on some CPUs the IDs encode the hierarchy iirc

#

frank canopy Jun 23, 2026, 9:13 PM

#

theres even a cpuid to discover how many bits are used for each field

pallid lava Jun 23, 2026, 9:13 PM

#

yes

calm latch Jun 23, 2026, 9:47 PM

#

My old PC starts at 16 and doesn't match CPUID trl

fiery turtle Jun 24, 2026, 12:41 AM

#

https://www.phoronix.com/news/DT-ACPI-Hybrid-Mode-Linux

DeviceTree-ACPI Hybrid Mode Proposed For Improving Linux Support On...

Over the years while working at Red Hat, Hans de Goede was known for driving many wonderful Linux laptop improvements benefiting AMD/Intel x86_64 hardware

#

Looks like future snapdragon laptops will be acpi only

#

Thank god

pine leaf Jun 24, 2026, 5:31 AM

#

lame, device trees on top 🗣️ 🔥

#

trl

slim panther Jun 24, 2026, 8:06 AM

#

fiery turtle Looks like future snapdragon laptops will be acpi only

huh

kindred beacon Jun 24, 2026, 10:02 AM

#

fiery turtle https://www.phoronix.com/news/DT-ACPI-Hybrid-Mode-Linux

cc @wind fiber

kind mantle Jun 25, 2026, 12:53 AM

#

If someone has that Linux kernel with uACPI, I'd be willing to test

rugged silo Jun 25, 2026, 4:59 AM

#

@fiery turtle i have some problems with implementing uacpi, somehow the compiler wont see the uacpi functions but the ide sees them perfectly
https://codeberg.org/aiqe/forward/src/branch/main/kernel/src/acpi/acpi_init.c
compiler errors:

src/acpi/acpi_init.c: In function ‘acpi_init’:
src/acpi/acpi_init.c:9:24: error: implicit declaration of function ‘uacpi_initialize’ [-Wimplicit-function-declaration]
    9 |     uacpi_status ret = uacpi_initialize(0);
      |                        ^~~~~~~~~~~~~~~~
src/acpi/acpi_init.c:14:11: error: implicit declaration of function ‘uacpi_namespace_load’ [-Wimplicit-function-declaration]
   14 |     ret = uacpi_namespace_load();
      |           ^~~~~~~~~~~~~~~~~~~~
src/acpi/acpi_init.c:19:11: error: implicit declaration of function ‘uacpi_namespace_initialize’ [-Wimplicit-function-declaration]
   19 |     ret = uacpi_namespace_initialize();
      |           ^~~~~~~~~~~~~~~~~~~~~~~~~~
src/acpi/acpi_init.c:24:5: error: implicit declaration of function ‘uacpi_set_interrupt_model’ [-Wimplicit-function-declaration]
   24 |     uacpi_set_interrupt_model(1);
      |     ^~~~~~~~~~~~~~~~~~~~~~~~~
src/acpi/acpi_init.c:25:11: error: implicit declaration of function ‘uacpi_finalize_gpe_initialization’ [-Wimplicit-function-declaration]
   25 |     ret = uacpi_finalize_gpe_initialization();
      |           ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Codeberg.org

forward/kernel/src/acpi/acpi_init.c at main

fiery turtle Jun 25, 2026, 8:01 AM

#

rugged silo <@353593948523986945> i have some problems with implementing uacpi, somehow the...

What compile options for uacpi

rugged silo Jun 25, 2026, 2:15 PM

#

fiery turtle What compile options for uacpi

wdym

#

config.h defines?

fiery turtle Jun 25, 2026, 2:16 PM

#

Anyway the most likely reason is you have barebones mode enabled

rugged silo Jun 25, 2026, 2:19 PM

#

fiery turtle Anyway the most likely reason is you have barebones mode enabled

yes

#

wait

#

didnt i say that

#

oh yeah im stupid

#

yeah, i use barebones mode

#

is there smth wrong with it

fiery turtle Jun 25, 2026, 2:21 PM

#

well those functions dont exist in barebones mode

rugged silo Jun 25, 2026, 2:22 PM

#

oh

#

ig its not very possible to use uacpi this early

#

okay i will delay it then

fiery turtle Jun 25, 2026, 2:24 PM

#

barebones mode only includes tables.h

rugged silo Jun 25, 2026, 2:25 PM

#

uacpi needs pci, pci needs interrupt controller, interrupt controller (apic) needs uacpi

#

well ig i could bootstrap pci with pic

fiery turtle Jun 25, 2026, 2:26 PM

#

thats why theres the early_table_access helper

#

so u can parse static tables before doing aml or allocations or anything

#

and uacpi doesnt need pci until you start running aml

rugged silo Jun 25, 2026, 2:27 PM

#

oh okay thanks

rugged silo Jun 25, 2026, 2:50 PM

#

gpfs with error code 0 in at if (uacpi_signatures_match(hdr, ACPI_FACS_SIGNATURE)) somehow, what could this relate to

static void dump_table_header(
    uacpi_phys_addr phys_addr, void *hdr
)
{
    struct acpi_sdt_hdr *sdt = hdr;

    if (uacpi_signatures_match(hdr, ACPI_FACS_SIGNATURE)) {
        uacpi_info(
            "FACS 0x%016"UACPI_PRIX64" %08X", UACPI_FMT64(phys_addr),
            sdt->length
        );
        return;
    }

    if (!uacpi_memcmp(hdr, ACPI_RSDP_SIGNATURE, sizeof(ACPI_RSDP_SIGNATURE) - 1)) {
        struct acpi_rsdp *rsdp = hdr;

        uacpi_info(
            "RSDP 0x%016"UACPI_PRIX64" %08X v%02X (%6.6s)",
            UACPI_FMT64(phys_addr), rsdp->revision >= 2 ? rsdp->length : 20,
            rsdp->revision, rsdp->oemid
        );
        return;
    }

    uacpi_info(
        "%.4s 0x%016"UACPI_PRIX64" %08X v%02X (%6.6s %8.8s)",
        sdt->signature, UACPI_FMT64(phys_addr), sdt->length, sdt->revision,
        sdt->oemid, sdt->oem_table_id
    );
}

fiery turtle Jun 25, 2026, 2:51 PM

#

Bad implementation of uacpi_kernel_map

rugged silo Jun 25, 2026, 2:55 PM

#

that workaround should work i suppose

void *uacpi_kernel_map(uacpi_phys_addr addr, uacpi_size len)
{
    void *AddrVirt = (uint64_t)addr + offset;
    return (void *)AddrVirt;
}

#

i feel like its incredibly stupid but cant prove it because im incredibly stupid bruh

rugged silo Jun 25, 2026, 2:57 PM

#

rugged silo that workaround should work i suppose ```C void *uacpi_kernel_map(uacpi_phys_add...

though it didnt work

fiery ice Jun 25, 2026, 3:08 PM

#

rugged silo that workaround should work i suppose ```C void *uacpi_kernel_map(uacpi_phys_add...

💀

#

at least use the hhdm

#

i mean, if you mapped all of it it should work then

full nimbus Jun 25, 2026, 3:16 PM

#

fiery ice i mean, if you mapped all of it it should work then

You probably don't want to map everything in the HHDM

fiery ice Jun 25, 2026, 3:16 PM

#

full nimbus You probably don't want to map everything in the HHDM

i mean, for uacpi you could do that

#

especially if running in the kernel

full nimbus Jun 25, 2026, 3:16 PM

#

Sure, but you don't want to do that

#

The physical memory could have anything, things that may need different caching modes etc.

#

You can even get machine checks due to mapping shit improperly

#

Eg MTRRs not liking huge pages in some cases

kind mantle Jun 25, 2026, 3:50 PM

#

rugged silo that workaround should work i suppose ```C void *uacpi_kernel_map(uacpi_phys_add...

Can't work

#

Not everything that ACPI will want is gonna be in the HHDM

rustic compass Jun 25, 2026, 3:55 PM

#

kind mantle Not everything that ACPI will want is gonna be in the HHDM

unless you map everything in the HHDM but that will require you to map most stuff UC + 4K or things like Evalyn described will happen

north hull Jun 25, 2026, 3:56 PM

#

if your kernel cant easily impl this api then u are not ready to impl uacpi

#

u have other things to worry about

fiery turtle Jun 25, 2026, 3:57 PM

#

rugged silo that workaround should work i suppose ```C void *uacpi_kernel_map(uacpi_phys_add...

How did you manage to gpf with this tho

#

At least add logs

#

To trace what you return

full nimbus Jun 25, 2026, 4:00 PM

#

rustic compass unless you map everything in the HHDM but that will require you to map most stuf...

in theory you could map it to the HHDM region on demand if you are very careful when you construt it but in many cases its not worth it

#

eg mapping the memory map exactly and when using large pages make sure they dont spill over into unknown memory

#

Loongarch64 would also let you cheat this with DMWs since it's a safe way to have an HHDM of all physical memory and with different caching modes

#

But again it's much easier to just have a proper virtual memory allocator

#

You won't be starving for virtual address space especially if you support 5 level paging

full nimbus Jun 25, 2026, 4:03 PM

#

full nimbus But again it's much easier to just have a proper virtual memory allocator

And you will need this later regardless

full nimbus Jun 25, 2026, 4:03 PM

#

fiery turtle At least add logs

Or atleast grab the QEMU log and addr2line it

rugged silo Jun 26, 2026, 4:46 AM

#

yeah i will use the vmm then

#

find free unmapped address and use it

slim panther Jun 28, 2026, 12:29 PM

#

ayo?

#

uacpi 6?

fiery turtle Jun 28, 2026, 12:32 PM

#

Yep

#

https://github.com/uACPI/uACPI/releases/tag/6.0.0
I've just released uACPI 6.0.0, this one is pretty important as it includes a lot of bug fixes

steel inlet Jun 28, 2026, 12:34 PM

#

me (again), still on 3.2.0:

lethal trench Jun 28, 2026, 12:34 PM

#

fiery turtle https://github.com/uACPI/uACPI/releases/tag/6.0.0 I've just released uACPI 6.0.0...

woaw based

slim panther Jun 28, 2026, 12:34 PM

#

steel inlet me (again), still on 3.2.0:

rugged silo Jun 28, 2026, 12:51 PM

#

fiery turtle https://github.com/uACPI/uACPI/releases/tag/6.0.0 I've just released uACPI 6.0.0...

thanks for the update, i have blessed my os with version 6.0.0

slim panther Jun 28, 2026, 1:06 PM

#

let's see

fiery turtle Jun 28, 2026, 1:30 PM

#

Next up I want to add an ability to attach arbitrary user data to a uacpi namespace node (to simplify uacpi-linux), and an optional 64-bit division platform helper

kind mantle Jun 28, 2026, 2:41 PM

#

I wonder if uACPI for Linux is ever going to be a real thing

#

As opposed to a tech demo

calm latch Jun 28, 2026, 2:45 PM

#

Considering it fixed a bunch of stuff for reactos...

fiery turtle Jun 28, 2026, 3:30 PM

#

kind mantle I wonder if uACPI for Linux is ever going to be a real thing

That just needs a bunch of work put in

rugged silo Jun 28, 2026, 3:32 PM

#

fiery turtle That just needs a bunch of work put in

uacpi 33.0.0

fiery turtle Jun 28, 2026, 3:45 PM

#

Nah the work is mostly on the uacpi linux side

calm latch Jun 28, 2026, 8:29 PM

#

Linux 33.0.0

pallid lava Jun 28, 2026, 8:33 PM

#

just updated to latest

#

have there been any changes to the gpe code, or my kernel is broken?

#

the power off button does not send any ec gpe

#

ok nvm, you removed the got gpe message flesh

fiery turtle Jun 28, 2026, 9:25 PM

#

There was such a message?

#

What version were you on before lol

pallid lava Jun 28, 2026, 9:33 PM

#

february/march i think

fiery turtle Jun 28, 2026, 9:40 PM

#

Maybe that was your kernel

fiery turtle Jun 29, 2026, 6:16 PM

#

nice! new NT regression after recent update

#

#

am i supposed to migrate my pc to uacpi now KEKW

fiery ice Jun 29, 2026, 6:17 PM

#

XD

full nimbus Jun 29, 2026, 6:18 PM

#

Windows to switch to uACPI

#

uACPI to take over the world

#

uACPI for president

pallid lava Jun 29, 2026, 6:50 PM

#

thought i was going crazy... but actually, uacpi_find_devices ignores devices whose _STA reports they're not active (skrill issue)

pallid lava Jun 29, 2026, 6:51 PM

#

fiery turtle am i supposed to migrate my pc to uacpi now <:KEKW:913708372329644042>

yes

fiery turtle Jun 29, 2026, 7:29 PM

#

pallid lava thought i was going crazy... but actually, uacpi_find_devices ignores devices wh...

Did you run into some issue because of that

pallid lava Jun 29, 2026, 7:29 PM

#

nah, it's not for real hw

#

i am implementing memory hotplugging, and obviously unplugged memory is reported as not present

fiery turtle Jun 29, 2026, 7:29 PM

#

Ah

#

Oh cool

pallid lava Jun 29, 2026, 7:30 PM

#

yes coolmeme

hallow marten Jun 29, 2026, 7:41 PM

#

pallid lava i am implementing memory hotplugging, and obviously unplugged memory is reported...

are you planning to support CPU hotplugging as well?

#

that sounds like hell

pallid lava Jun 29, 2026, 7:55 PM

#

yes

#

it's the 2 things i want to do before continuing

#

i should create a thread tbh once i clean some things up

hallow marten Jun 29, 2026, 8:01 PM

#

Actually cpu hotplug shouldn't be that bad now that I think of it

#

You "just" try migrating everything

pallid lava Jun 29, 2026, 8:03 PM

#

yeah

chrome crow Jun 29, 2026, 8:30 PM

#

fiery turtle nice! new NT regression after recent update

what did you do

fiery turtle Jun 29, 2026, 8:30 PM

#

Booted my computer

calm latch Jun 29, 2026, 9:25 PM

#

hallow marten You "just" try migrating everything

Isn't it basically the same as CPU parking?

pallid lava Jun 29, 2026, 9:26 PM

#

yes

hallow marten Jun 29, 2026, 9:27 PM

#

actually I'll look into this eventually

pallid lava Jun 29, 2026, 9:27 PM

#

implementing hotplug support is not necessary for having in your os support for software disabling cpus during runtime, making them sleep for a long time and whatever

hallow marten Jun 29, 2026, 9:27 PM

#

doesn't really seem that hard

pallid lava Jun 29, 2026, 9:27 PM

#

i just want to get my hands dirty with acpi

#

(aka match processors to processor objects and call _ej0)

hallow marten Jun 29, 2026, 9:27 PM

#

implement it in zag

#

trl

#

you will get mentorship from the zag foundation

#

(™)

pallid lava Jun 29, 2026, 9:28 PM

#

based?

fiery ice Jun 29, 2026, 11:52 PM

#

hmm @fiery turtle what could cause uacpi to hang before saying it loaded all the tables and did tot ops in tot ms?

fiery turtle Jun 29, 2026, 11:53 PM

#

gdb

fiery ice Jun 29, 2026, 11:54 PM

#

i can't 💀

#

it's on real hw

#

it only happens there

fiery turtle Jun 29, 2026, 11:54 PM

#

debug prints

#

uacpi_context_set_log_level

fiery ice Jun 30, 2026, 12:01 AM

#

fiery turtle uacpi_context_set_log_level

#

gets stuck there

fiery turtle Jun 30, 2026, 12:05 AM

#

if i had to guess it hangs in your allocator

fiery ice Jun 30, 2026, 12:05 AM

#

hmm

#

i'll check it then thx

fiery turtle Jun 30, 2026, 12:06 AM

#

yeah just add a bunch of prints

fiery ice Jun 30, 2026, 12:07 AM

#

another strange thing would be the messages for port 0

#

tho maybe that's another process

#

and i'm gonna check it now

#

hmm, no, it's uacpi

#

i think it's hanging on the pci read or something? since that's where i'm using ports

fiery turtle Jun 30, 2026, 12:32 AM

#

Create namestring just allocates

fiery ice Jun 30, 2026, 12:34 AM

#

fiery turtle Create namestring just allocates

ok, i think it 100% is pci stuff

#

for some reason on kvm it doesn't request pci reads / writes before i call send mcfg

#

but on real hw yes

#

the hanging is caused by the pci service not responding bc it's not active yet

fiery ice Jun 30, 2026, 1:28 AM

#

fiery turtle Create namestring just allocates

ok, it was 100% the pci stuff

#

it is doing pci operations during namespace load

#

fiery turtle Jun 30, 2026, 8:02 AM

#

fiery ice

Why is every message duplicated

slim panther Jun 30, 2026, 8:40 AM

#

fiery turtle Why is every message duplicated

once debug and once information KEKW

#

inb4 switch on log type but no break;

pallid lava Jun 30, 2026, 9:42 AM

#

small typo in include/uacpi/types.h:37

pallid lava Jun 30, 2026, 10:41 AM

#

okay, so, i have written a wrapper for _OST, but i am not sure if it's entirely correct

#

int
acpi_device_ost(uacpi_namespace_node *node, unsigned int source_event,
                unsigned int status_code)
{
        uacpi_status status;
        uacpi_namespace_node *ost;
        uacpi_object_array args;
        uacpi_object *arg0, *arg1, *args_ptr[3];

        status = uacpi_namespace_node_find(node, "_OST", &ost);
        if (status == UACPI_STATUS_NOT_FOUND) {
                return 0;
        } else if (status != UACPI_STATUS_OK) {
                return -1;
        }

        arg0 = uacpi_object_create_integer(source_event);
        if (!arg0) {
                return -ERROR_OUT_OF_MEMORY;
        }

        arg1 = uacpi_object_create_integer(status_code);
        if (!arg1) {
                free(arg0);
                return -ERROR_OUT_OF_MEMORY;
        }

        args_ptr[0] = arg0;
        args_ptr[1] = arg1;
        args_ptr[2] = UACPI_NULL;
        args.count = 3;
        args.objects = args_ptr;

        status = uacpi_execute(ost, nullptr, &args);
        free(arg0);
        free(arg1);
        if (uacpi_likely_success(status)) {
                return 0;
        } else {
                return -1;
        }
}

#

is this the right way of passing optional arguments?

#

specifically, _OST takes three arguments: arg0 (integer), arg1(integer), arg2(optional buffer)

fiery ice Jun 30, 2026, 10:42 AM

#

fiery turtle Why is every message duplicated

I redirected dbg to screen

fiery turtle Jun 30, 2026, 10:49 AM

#

pallid lava specifically, _OST takes three arguments: arg0 (integer), arg1(integer), arg2(op...

when the spec says "may be null" it probably means the buffer may be empty

#

you cant pass 2 arguments to methods that accept 3

#

you also cant just free the arguments, you must use uacpi_object_unref

#

since aml can just arbitrarily grab a reference to it and you wont know

pallid lava Jun 30, 2026, 10:52 AM

#

oh okay

#

so, how do i create an empty buffer? do i use uacpi_object_create_buffer, with view.length = 0, or how?

fiery turtle Jun 30, 2026, 10:53 AM

#

yeah

pallid lava Jun 30, 2026, 10:53 AM

#

i see

#

thanks letsgo

fiery turtle Jun 30, 2026, 10:53 AM

#

npnp

#

https://elixir.bootlin.com/linux/v7.1.2/source/drivers/acpi/utils.c#L541

#

i think linux does that also

pallid lava Jun 30, 2026, 10:56 AM

#

perfect then

#

lmao

#

_OST for memory hotplug in qemu is just this

#

#

tbf, there's not much useful stuff that qemu can do

#

at most, print a message or trace _OST calls, which is what it does

fiery turtle Jun 30, 2026, 11:09 AM

#

makes sense lol

quaint lotus Jun 30, 2026, 11:22 AM

#

I know this is mostly NIH but using a whole ACPI system feels kinda like cheating

#

Is there a way to trim uACPI down to just an AML interpreter

pallid lava Jun 30, 2026, 11:25 AM

#

but... it's an aml interpreter?

#

just don't use the helpers

quaint lotus Jun 30, 2026, 11:30 AM

#

pallid lava but... it's an aml interpreter?

From the git it seems to also do those:

A fairly advanced event subsystem (GPE/fixed, wake, implicit notify, AML handlers)
Table management API (search, dynamic installation/loading, overrides, etc.)
Operation region subsystem (user handlers, support for BufferAcc opregions, builtins for common types)
Sleep state management (transition to any S state, wake vector programming)
PCI routing table retrieval & interrupt model API
Device search API
Interface & feature management exposed via _OSI
Client-defined Notify() handlers
Firmware global lock management (_GL, locked fields, public API)
GAS read/write API

fiery ice Jun 30, 2026, 11:30 AM

#

fiery turtle Why is every message duplicated

btw infy, how can i fix those warning during namespace load?

pallid lava Jun 30, 2026, 11:31 AM

#

you can use or not use all that stuff

fiery ice Jun 30, 2026, 11:31 AM

#

i didn't think it wanted to interact with pci things during namespace load

fiery turtle Jun 30, 2026, 11:31 AM

#

fiery ice btw infy, how can i fix those warning during namespace load?

you cant, its a genuine bug in your firmware

fiery ice Jun 30, 2026, 11:31 AM

#

💀

#

damn you samsung

lethal trench Jun 30, 2026, 11:31 AM

#

fiery turtle you cant, its a genuine bug in your firmware

well well well...

quaint lotus Jun 30, 2026, 11:32 AM

#

fiery turtle you cant, its a genuine bug in your firmware

least buggy firmware:

fiery turtle Jun 30, 2026, 11:32 AM

#

this one is relatively harmless at least

lethal trench Jun 30, 2026, 11:32 AM

#

at this point I'm convinced firmware devs are paid by microsoft to add bugs to firmware

fiery ice Jun 30, 2026, 11:32 AM

#

well, i have a pci write / read before loading now

#

it is legacy only

#

but should help a bit XD

fiery turtle Jun 30, 2026, 11:32 AM

#

u can parse mcfg before namespace load no? lol

fiery ice Jun 30, 2026, 11:33 AM

#

no

#

uacpi wants namespace loaded at least

#

with my current method i mean

#

i use uacpi_find_devices_at

fiery turtle Jun 30, 2026, 11:33 AM

#

table_find("MCFG")?

fiery ice Jun 30, 2026, 11:34 AM

#

hmm, i do it on the callback of uacpi_find_devices_at

fiery turtle Jun 30, 2026, 11:36 AM

#

anyway legacy access works fine

fiery turtle Jun 30, 2026, 11:37 AM

#

fiery ice damn you samsung

u should see the same warnings under linux btw

#

with acpica

fiery ice Jun 30, 2026, 11:37 AM

#

oh

fiery turtle Jun 30, 2026, 11:37 AM

#

check your dmsg

steel inlet Jun 30, 2026, 8:49 PM

#

https://uacpi.com/

uacpi.com

DISCRETE INVESTIGATIONS, TRUSTED RESULTS.

UAC Agency Licenses: Utah P112781, Arizona 1819040, California PI190144

fiery turtle Jun 30, 2026, 8:52 PM

#

💀

pallid lava Jun 30, 2026, 9:20 PM

#

dw, it's not over for you

#

https://nasm.org and https://nasm.us -- guess which one is for the assembler we all love

full nimbus Jun 30, 2026, 9:24 PM

#

pallid lava <https://nasm.org> and <https://nasm.us> -- guess which one is for the assembler...

What the fuck 😭

steel inlet Jun 30, 2026, 9:37 PM

#

pallid lava <https://nasm.org> and <https://nasm.us> -- guess which one is for the assembler...

nasm.us

#

(because i already knew)

#

never heard of nasm.org tho, wtf is that

quaint lotus Jun 30, 2026, 9:42 PM

#

pallid lava <https://nasm.org> and <https://nasm.us> -- guess which one is for the assembler...

nasm.us for sure

#

I went to that site way too many times

gloomy helm Jul 1, 2026, 8:16 AM

#

me writing my own aml parser and shi cuz uacpi is too complex: ._.

torpid root Jul 1, 2026, 8:17 AM

#

stare

gloomy helm Jul 1, 2026, 8:18 AM

#

(also 70000th message!! trl )

#

nuke3

lethal trench Jul 1, 2026, 8:18 AM

#

gloomy helm me writing my own aml parser and shi cuz uacpi is too complex: ._.

im sure there isn't a reason it's so complex and it's just like that because infy likes writing complex code :p

calm latch Jul 1, 2026, 8:21 AM

#

gloomy helm (also 70000th message!! <:trl:829316641191886888>)

lounge-2

fiery turtle Jul 1, 2026, 8:23 AM

#

yeah its 30k lines mostly because i wanted to troll

#

it actually requires 10 loc

lethal trench Jul 1, 2026, 8:23 AM

#

not even

#

if(firmware_is_bugged()) {
  panic("buggy firmware I aint even gonna bother");
}

#

this always panics

fiery turtle Jul 1, 2026, 8:24 AM

#

does this check just return true always?

torpid root Jul 1, 2026, 8:25 AM

#

it calls is_computer_on()

lethal trench Jul 1, 2026, 8:26 AM

#

fiery turtle does this check just return true always?

yeah it's 100% right about 90% of the time

rugged silo Jul 1, 2026, 8:27 AM

#

lethal trench ```c if(firmware_is_bugged()) { panic("buggy firmware I aint even gonna bother...

gcc gonna optimize that into if(1) trl

gloomy helm Jul 1, 2026, 8:27 AM

#

fiery turtle yeah its 30k lines mostly because i wanted to troll

mine is 1000 lines nooo

fiery turtle Jul 1, 2026, 9:05 AM

#

if you strip references from AML it could be much simpler

#

most of the complexity i would say comes from them

#

and no one really knows how they are supposed to work, especially not the spec

torpid root Jul 1, 2026, 9:09 AM

#

lol

lethal trench Jul 1, 2026, 9:10 AM

#

fiery turtle and no one really knows how they are supposed to work, especially not the spec

how they are supposed to work
bold of you to assume they're supposed to work in the first place

fiery turtle Jul 1, 2026, 9:12 AM

#

the entire aml/asl spec is designed around ignoring the existence of edge cases

#

you would never do multilevel references anyway right trl

lethal trench Jul 1, 2026, 9:13 AM

#

sureeeee

fiery turtle Jul 1, 2026, 9:13 AM

#

you would never try to make local references escape the method right trl

#

and if you're the implementor, go figure it out lol

#

gl

lethal trench Jul 1, 2026, 9:14 AM

#

"it worked when we tested it on nt"

#

"acpica threw errors but it booted so who cares"

fiery turtle Jul 1, 2026, 9:14 AM

#

fiery turtle you would never try to make local references escape the method right <:trl:82931...

the way acpica solves it is by segfaulting

#

so this whole thing works only becaused firmware engineers made sure to only use the bare minimum of features that never trigger such edge cases

gloomy helm Jul 1, 2026, 9:15 AM

#

fiery turtle the way acpica solves it is by segfaulting

acpica my despised

fiery turtle Jul 1, 2026, 9:15 AM

#

like its not even anything crazy, e.g.

Method (TEST) {
    Return (RefOf(Local0))
}

#

no one really knows how this is supposed to work

lethal trench Jul 1, 2026, 9:16 AM

#

replace aml with jvm bytecode atp 🙏

fiery turtle Jul 1, 2026, 9:17 AM

#

unironically it would be simpler

#

or at least have defined behavior

fiery turtle Jul 1, 2026, 9:19 AM

#

fiery turtle unironically it would be simpler

Like my interpreter.c is 6k loc not counting utility functions, you can probably implement a simple java vm in that many loc

lethal trench Jul 1, 2026, 9:20 AM

#

i am now tempted but I must resist

fiery turtle Jul 1, 2026, 9:20 AM

#

yeah imo bytecode vms are super fun

#

or any emulators or interpreters

lethal trench Jul 1, 2026, 9:21 AM

#

and that's why I'm so tempted 😭

fiery turtle Jul 1, 2026, 9:21 AM

#

the worst one jvm has is like invokedynamic or whatever

full nimbus Jul 1, 2026, 1:18 PM

#

lethal trench im sure there isn't a reason it's so complex and it's just like that because inf...

Also there are better reasons to want to make an AML thing and not use uACPI

pallid lava Jul 1, 2026, 3:45 PM

#

even our beloved emulator (qemu) does not implement properly the acpi spec nooo

#

for memory devices, The physical address space described by _CRS object must be described using the Extended Address Space Resource Descriptor macro.

what i actually get: a qword memory descriptor

fiery turtle Jul 1, 2026, 4:17 PM

#

Gg

frank canopy Jul 1, 2026, 4:28 PM

#

the q35 machine in qemu is missing an aml field that the spec defines as required for the pci root

#

there is no consistency or correctness. only what will work under windows

pallid lava Jul 1, 2026, 4:30 PM

#

which field?

frank canopy Jul 1, 2026, 4:30 PM

#

BBN iirc

pallid lava Jul 1, 2026, 4:30 PM

#

i think it can be assumed to 0?

frank canopy Jul 1, 2026, 4:31 PM

#

_SEG is defined 0 if missing but _BBN is officially supposed to be required despite being assume 0 if missing in practice

pallid lava Jul 1, 2026, 4:31 PM

#

oh yes, you're right

frank canopy Jul 1, 2026, 4:32 PM

#

so like it's fine it's just more evidence that the acpi spec is just a suggestion

pallid lava Jul 1, 2026, 4:40 PM

#

pallid lava for memory devices, ```The physical address space described by _CRS object must ...

this is because linux itself only supports this type of descriptor chad

fiery turtle Jul 1, 2026, 4:42 PM

#

Lmao

#

Finally, linux support in aml

pallid lava Jul 1, 2026, 4:43 PM

#

pallid lava this is because linux itself only supports this type of descriptor <:chad:815144...

im not sure

#

#

it uses this macro/function (which is not defined anywhere)

#

and according to gemini, it converts any memory resource to that

#

and the qemu aml is capable of creating a dword memory descriptor, if the hp region is in low memory

pallid lava Jul 1, 2026, 4:45 PM

#

pallid lava and according to gemini, it converts any memory resource to that

okay yes, indeed it's happening this

#

elixir is tweaking for some reason

gentle peak Jul 1, 2026, 4:48 PM

#

pallid lava it uses this macro/function (which is not defined anywhere)

it's defined in acpica

pallid lava Jul 1, 2026, 4:48 PM

#

acpi_resource_to_address64 takes a word, dword or qword descriptor, and converts it to a qword descriptor

gentle peak Jul 1, 2026, 4:48 PM

#

drivers/acpi/acpica is excluded from elixir for some reason

pallid lava Jul 1, 2026, 4:48 PM

#

and if there is an extended one (the required one by the spec), it fails chad

pallid lava Jul 1, 2026, 4:49 PM

#

gentle peak drivers/acpi/acpica is excluded from elixir for some reason

i see

fiery turtle Jul 1, 2026, 4:51 PM

#

gentle peak it's defined in acpica

I should add a helper like that

fiery turtle Jul 1, 2026, 4:51 PM

#

pallid lava and if there is an extended one (the required one by the spec), it fails <:chad:...

Lmao

lethal trench Jul 1, 2026, 5:28 PM

#

pallid lava and if there is an extended one (the required one by the spec), it fails <:chad:...

I fucking love acpi <3

gloomy helm Jul 1, 2026, 5:29 PM

#

lethal trench I fucking love acpi <3

i hate acpi >:(

lethal trench Jul 1, 2026, 5:29 PM

#

gloomy helm i hate acpi >:(

don't worry acpi hates you

#

acpi hates all of us

pallid lava Jul 1, 2026, 5:29 PM

#

it isn't acpi's fault, tbf

#

it's linux being stupid now this time for real

lethal trench Jul 1, 2026, 5:31 PM

#

fair enough but it works 😭

full nimbus Jul 1, 2026, 6:12 PM

#

frank canopy there is no consistency or correctness. only what will work under windows

It feels like every single company takes spec for almost everything takes it and uses it to wipe their ass after shitting and then proceeding to take a piss on It

frank canopy Jul 1, 2026, 7:23 PM

#

full nimbus It feels like every single company takes spec for almost everything takes it and...

yeah pretty much tbh

#

and then they get mad when other people do that to their specs

quaint lotus Jul 3, 2026, 12:19 PM

#

Out of curiosity what does AML typically do?

#

Like does it only contain data, or can it write to ports or memory addresses and crazy stuff like that

torpid root Jul 3, 2026, 12:19 PM

#

quaint lotus Out of curiosity what does AML typically do?

find unhandled edge cases in aml interpreters

mortal yoke Jul 3, 2026, 12:21 PM

#

quaint lotus Like does it only contain data, or can it write to ports or memory addresses and...

it does all of them

quaint lotus Jul 3, 2026, 12:23 PM

#

fiery turtle the entire aml/asl spec is designed around ignoring the existence of edge cases

this is rather annoying

gloomy helm Jul 3, 2026, 12:23 PM

#

aml allows for some kewl shit like acpi shutdown

#

and more

#

so in a sense it does allow memory writes and shit

pallid lava Jul 3, 2026, 12:26 PM

#

shutdown would have been good if the s5 slp_type was in the fadt and there were no need of doing whatever _pts garbage

fiery turtle Jul 3, 2026, 12:34 PM

#

quaint lotus Like does it only contain data, or can it write to ports or memory addresses and...

Its a turing complete programming language/byte code

#

It does everything and anything it wants

#

Typically programming the chipset, talking to the firmware over SMM and many other things

quaint lotus Jul 3, 2026, 12:35 PM

#

At that point why don’t hardware developers just write their drivers in AML trl

fiery turtle Jul 3, 2026, 12:35 PM

#

They do ship a lot of power management drivers in aml

#

Thats how you retrieve the battery percentage for example

#

Some roll entire i2c drivers

#

And the nice thing for the kernel is it just runs a method which does this black magic under the hood

quaint lotus Jul 3, 2026, 12:37 PM

#

Why can’t this apply to more drivers? Like GPU drivers for instance

#

Imagine AML mode setting

pallid lava Jul 3, 2026, 12:37 PM

#

that was called running the rom in v8086 nooo

#

aml is too much trouble and is not very performant

fiery turtle Jul 3, 2026, 12:38 PM

#

quaint lotus Why can’t this apply to more drivers? Like GPU drivers for instance

You get backlight control and even mux control in aml for hybrid laptops

#

But not modeset yeah

#

Idk why they never did that

#

Probably because that wasn't needed and was covered by generic gpu drivers

#

Whereas backlight and mux is always ad hoc chipset specific stuff

full nimbus Jul 3, 2026, 4:06 PM

#

fiery turtle Idk why they never did that

Could also have been too funky across GPU versions to be shipping laptops that may have diff cards?

fiery turtle Jul 3, 2026, 4:27 PM

#

full nimbus Could also have been too funky across GPU versions to be shipping laptops that m...

Well gpus provide the tables themselves

full nimbus Jul 3, 2026, 4:28 PM

#

ah true

flat badge Jul 3, 2026, 8:49 PM

#

quaint lotus Like does it only contain data, or can it write to ports or memory addresses and...

see, the purpose of AML is pinging an EC that tells a proper microcontroller that runs software written in an actual language to do the actual work

#

meme

kind mantle Jul 4, 2026, 12:22 AM

#

fiery turtle Well gpus provide the tables themselves

... Do GPUs have AML in them?

fiery turtle Jul 4, 2026, 12:23 AM

#

yes

kind mantle Jul 4, 2026, 12:58 AM

#

Damn

flat badge Jul 4, 2026, 9:57 AM

#

Hm?

#

which GPUs do that?

#

for integrated GPUs, the AML is typically embedded in the uefi firmware image

#

and not in the gpu

fiery turtle Jul 4, 2026, 10:28 AM

#

flat badge which GPUs do that?

Nvidia afaik

fiery turtle Jul 4, 2026, 10:29 AM

#

flat badge for integrated GPUs, the AML is typically embedded in the uefi firmware image

Firmware image of what

#

Where does said image live

flat badge Jul 4, 2026, 10:29 AM

#

rom on the mainboard

#

spi flash or whatever

#

with the rest of the uefi firmware

fiery turtle Jul 4, 2026, 10:30 AM

#

Mainboard of the gpu?

flat badge Jul 4, 2026, 10:30 AM

#

no, mainboard

fiery turtle Jul 4, 2026, 10:30 AM

#

Oh ok no thats not how that works

flat badge Jul 4, 2026, 10:30 AM

#

for integrated gpus? yes, it is

fiery turtle Jul 4, 2026, 10:30 AM

#

I said Nvidia

flat badge Jul 4, 2026, 10:31 AM

#

for integrated GPUs, the AML is typically embedded in the uefi firmware image

#

that's what i said

fiery turtle Jul 4, 2026, 10:31 AM

#

I mean where else would it live for an integrated gpu lol

flat badge Jul 4, 2026, 10:31 AM

#

gpu option rom

#

but that's not how it's done for integrated

#

I don't think nvidia has any AML (?)

fiery turtle Jul 4, 2026, 10:32 AM

#

Definitely does

flat badge Jul 4, 2026, 10:33 AM

#

discrete nvidia gpus?

fiery turtle Jul 4, 2026, 10:34 AM

#

Yes, ill take a look at a few dumps from desktops later to check

flat badge Jul 4, 2026, 10:35 AM

#

I don't have any on my machine

#

i never saw a SSDT with nvidia vendor

#

on a desktop

fiery turtle Jul 4, 2026, 10:35 AM

#

They dont have the Nvidia vendor

#

But if you grep for a gpu object you will see multiple

#

One for integrated and one for discrete

flat badge Jul 4, 2026, 10:36 AM

#

which vendor do they have?

fiery turtle Jul 4, 2026, 10:36 AM

#

Ill get back to you

flat badge Jul 4, 2026, 10:43 AM

#

$ grep -rio 'GFX.' | uniq
DSDT.dsl:GFX0
SSDT2.dsl:GFX0
SSDT2.dsl:Gfx 
SSDT2.dsl:GFX0

#

only has integrated

slim panther Jul 4, 2026, 10:43 AM

#

what gpu do you have

flat badge Jul 4, 2026, 10:44 AM

#

this machine has a 1660

#

but i also never saw it on other desktops

#

the bus that the card is on is not even in ACPI

fiery turtle Jul 4, 2026, 10:48 AM

#

you mean as _BBN? That only exposes root ports

flat badge Jul 4, 2026, 10:48 AM

#

no, i mean as _ADR

fiery turtle Jul 4, 2026, 10:48 AM

#

_ADR doesnt include the bus

flat badge Jul 4, 2026, 10:49 AM

#

how do you address a device on a non root bus then?

fiery turtle Jul 4, 2026, 10:50 AM

#

_BBN of the parent device

#

flat badge Jul 4, 2026, 10:51 AM

#

ah

#

so it'd be nested under the bridge

#

under PCI0.PEG0 or whatever in this case

fiery turtle Jul 4, 2026, 10:52 AM

#

yeah it would have to be under a parent bridge dev, so like the pcie root port device ig

#

yeah nothing on my amd desktop, i have like 3 tables with gfx, but i think they're all talking about the embedded gpu

flat badge Jul 4, 2026, 10:55 AM

#

yeah okay it has a device there but that's supplied by the firmware

#

not by the gpu

#

and it only has generic stuff, nothing nvidia specific

fiery turtle Jul 4, 2026, 10:55 AM

#

i could've sworn i had a table on my rtx2080 that had nvidia stuff

#

but this one is 5xxx series

flat badge Jul 4, 2026, 10:56 AM

#

that'd mean that the option rom would need to install it

fiery turtle Jul 4, 2026, 10:56 AM

#

yeah

flat badge Jul 4, 2026, 10:56 AM

#

which is not impossible but kinda rare

#

proprietary devices do this sometimes (e.g., to boot from raid) but there's probably no point in doing it for a gpu

#

since you'll have a driver anyway once you're in the proper OS

fiery turtle Jul 4, 2026, 10:57 AM

#

so like, in a setup with a desktop motherboard, the cpu is not known at firmware compile time

#

so someone has to install it dynamically

#

i mean e.g. the amd integrated gpu tables

flat badge Jul 4, 2026, 10:58 AM

#

yeah the firmware hard codes a set of igpus

#

that work on the particular machine

fiery turtle Jul 4, 2026, 10:58 AM

#

you think it has a hardcoded set of supported igpus?

#

What if i plug in a zen6 cpu that didnt exist when it was made

flat badge Jul 4, 2026, 10:59 AM

#

you have to update your bios

#

at least that's how it works on intel

fiery turtle Jul 4, 2026, 11:00 AM

#

maybe

mortal yoke Jul 4, 2026, 11:00 AM

#

the gop driver is also in the bios image on intel

flat badge Jul 4, 2026, 11:00 AM

#

yeah

fiery turtle Jul 4, 2026, 11:00 AM

#

true

#

i saw the gop driver version hardcoded in bios

#

but yeah that was intel

flat badge Jul 4, 2026, 11:01 AM

#

idk how it works on AMD. AMD has a wider set of CPUs per socket, so maybe they're doing it in a less static way

fiery turtle Jul 4, 2026, 11:02 AM

#

perhaps

kindred beacon Jul 4, 2026, 2:21 PM

#

The UEFI ROMs can install tables at runtime

frank canopy Jul 4, 2026, 3:10 PM

#

yup

#

theres a uefi protocol for it

flat badge Jul 4, 2026, 3:32 PM

#

yes, it's clear that they can do it, the question is whether they do it in practice

kindred beacon Jul 4, 2026, 5:14 PM

#

I have seen that, yes

#

For AMD iirc

hearty pike Jul 12, 2026, 2:37 PM

#

Is there any way obtain the uacpi_generic_region_info associated with a UACPI_OBJECT_OPERATION_REGION?

fiery turtle Jul 12, 2026, 2:42 PM

#

What are you trying to do?

#

This info is passed to you on region attach

hearty pike Jul 12, 2026, 2:43 PM

#

I'm figuring out what all needs to be done for the Linux integration. acpi_check_address_range seems to require something like that.

fiery turtle Jul 12, 2026, 2:44 PM

#

But Linux integration already exists, or are you doing something else?

hearty pike Jul 12, 2026, 2:47 PM

#

From what I saw, a decent number of the functions were not implemented such that linux got the full use of them (including acpi_check_address_range and acpi_initialize_subsystem).

fiery turtle Jul 12, 2026, 2:48 PM

#

Ill have to get back to you because idk what check_address_range does off the top of my head, but iirc most unimplemented functions were either not super useful or something very misc

hearty pike Jul 12, 2026, 2:51 PM

#

If the most up to date version is the repository posted online, then it seems to lack at least some of the functionality that Linux uses to protect against buggy firmware.

fiery turtle Jul 12, 2026, 2:52 PM

#

Yeah its possible

#

Are you working on improving the port?

#

If anything, since linux is using uacpi source directly, it can just import the private api from internal/ and access any fields of an object

craggy epoch Jul 15, 2026, 6:28 PM

#

torpid root find unhandled edge cases in aml interpreters

Unfortunately

kindred beacon Jul 18, 2026, 9:43 PM

#

quick question, is IVRS something you'd want to have upstream

#

also, are you taking sloperated imhex patterns for tables meme

fiery turtle Jul 18, 2026, 9:44 PM

#

kindred beacon quick question, is IVRS something you'd want to have upstream

yeah ofc, any tables are welcome

fiery turtle Jul 18, 2026, 9:44 PM

#

kindred beacon also, are you taking sloperated imhex patterns for tables <:meme:575445094589661...

what is that?

kindred beacon Jul 18, 2026, 9:45 PM

#

#

hex editor that has a pattern lang for dissecting stuff

#

https://github.com/werwolv/imhex

GitHub

GitHub - WerWolv/ImHex: 🔍 A Hex Editor for Reverse Engineers, Pr...

🔍 A Hex Editor for Reverse Engineers, Programmers and people who value their retinas when working at 3 AM. - WerWolv/ImHex

fiery turtle Jul 18, 2026, 9:46 PM

#

we would probably need a separate "disassembler" project for that like iasl in acpica

#

have u tried running that table against iasl btw

#

it can disassemble stuff nicely

kindred beacon Jul 18, 2026, 9:46 PM

#

didn't know

#

sec

fiery turtle Jul 18, 2026, 9:46 PM

#

its just iasl ivrs.dat

kindred beacon Jul 18, 2026, 9:47 PM

#

with .dat just being the binary dump?

#

TIL

fiery turtle Jul 18, 2026, 9:48 PM

#

yep

#

i typically just do iasl *.dat to get the disassembled versions for all tables in a dump

kindred beacon Jul 18, 2026, 9:48 PM

#

yeah it decodes IVRS as well

frank canopy Jul 18, 2026, 9:49 PM

#

ive got a non-slop imhex pattern for table headers and also has a thing for pkglen decoding

#

its very nice

fiery turtle Jul 18, 2026, 9:50 PM

#

how is your aml project going btw

frank canopy Jul 18, 2026, 9:50 PM

#

fiery turtle how is your aml project going btw

i havent done any dev on anything in weeks and im sad about it. might work more tonight if i can get the motivation to cooperate tho

fiery turtle Jul 18, 2026, 9:52 PM

#

ah damn

frank canopy Jul 18, 2026, 9:56 PM

#

oh also theres a massive thunderstorm coming with a tornado watch so i might end up sheltering in the basement instead of doing dev work tonight lmao

fiery turtle Jul 18, 2026, 9:59 PM

#

holy

#

stay safe

frank canopy Jul 18, 2026, 10:01 PM

#

its not uncommon here this time of year, and ive never actually been caught in a tornado despite that

#

but still will play it safe if theres a warning on one

flat badge Jul 18, 2026, 10:01 PM

#

where do you live?

frank canopy Jul 18, 2026, 10:02 PM

#

pennsylvania

#

in the greater philadelphia area, which is big enough that i feel ok saying it publicly

#

we dont get big tornadoes

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