fiery turtle Jan 24, 2025, 1:14 PM

#

Since its an old obos iso it can be an obos regression as well lol

north holly Jan 24, 2025, 1:15 PM

#

not old

#

it's on master, but behind a commit for RTC support, which is used in zero places in obos

fiery turtle Jan 24, 2025, 1:15 PM

#

Can u do more debugging pls?

#

The address spaces support one

north holly Jan 24, 2025, 1:16 PM

#

I've debugged to my full extent

keen marsh Jan 24, 2025, 1:17 PM

#

@ me if you need any more testing

north holly Jan 24, 2025, 1:17 PM

#

and have found it's in uacpi_finalize_gpe_initialization

fiery turtle Jan 24, 2025, 1:17 PM

#

north holly I've debugged to my full extent

No like put prints in uacpi code

north holly Jan 24, 2025, 1:17 PM

#

sure

fiery turtle Jan 24, 2025, 1:17 PM

#

This doesnt tell me much because this pr didn't change any event code

#

So it must be something else

#

Like an obos bug chad

north holly Jan 24, 2025, 1:18 PM

#

obos on m68k-port-fixes doesn't hang

#

obos on master+updated uacpi does

fiery turtle Jan 24, 2025, 1:18 PM

#

north holly obos on m68k-port-fixes doesn't hang

What about this plus new uacpi

north holly Jan 24, 2025, 1:18 PM

#

I'll do that

fiery turtle Jan 24, 2025, 1:19 PM

#

If that still hangs pls put prints in uacpi, finalize gpe initialization is really small

north holly Jan 24, 2025, 1:23 PM

#

bruh wtf is cmake smoking

#

fatal: unable to read tree (63ab369748db5764b59a728f54da0d7e5390f407)
CMake Error at /home/oberrow/Code/obos/build/x86_64-release/_deps/uacpi-subbuild/uacpi-populate-prefix/tmp/uacpi-populate-gitclone.cmake:61 (message):
  Failed to checkout tag: '63ab369748db5764b59a728f54da0d7e5390f407'

#

ah nvm

#

infy rewriting history

calm latch Jan 24, 2025, 1:25 PM

#

📎 pmOS.iso

#

I have to go in like 10 minutes...

fiery turtle Jan 24, 2025, 1:26 PM

#

north holly infy rewriting history

Yes

north holly Jan 24, 2025, 1:26 PM

#

@keen marsh

📎 obos.iso

keen marsh Jan 24, 2025, 1:29 PM

#

north holly <@432950924462522368>

Same as last time

north holly Jan 24, 2025, 1:29 PM

#

guess I'm adding prints to uacpi code

#

I can probably test on my own from now, then when I've actually found a bug*

#

I can ask you

calm latch Jan 24, 2025, 1:29 PM

#

calm latch I have to go in like 10 minutes...

and framebuffer's broken on sony laptop nooo

fiery turtle Jan 24, 2025, 1:30 PM

#

north holly guess I'm adding prints to uacpi code

Lets go

north holly Jan 24, 2025, 1:30 PM

#

north holly I can probably test on my own from now, then when I've actually found a bug*

*in uacpi

calm latch Jan 24, 2025, 1:31 PM

#

calm latch and framebuffer's broken on sony laptop <:nooo:750466870993223750>

and on the macbook

#

I can't test it nooo

calm latch Jan 24, 2025, 1:32 PM

#

calm latch I have to go in like 10 minutes...

I'll return in 3 days

#

In the meantime I'll only have my Lenovo laptop

#

At least that's one which has GPIO meme

fiery turtle Jan 24, 2025, 1:33 PM

#

Good test bed

north holly Jan 24, 2025, 1:34 PM

#

        ret = set_gpe_state(event, GPE_STATE_ENABLED);
        if (uacpi_unlikely_error(ret)) {
            gp_event_toggle_masks(event, UACPI_FALSE);
            event->num_users--;
        }```

#

uacpi is last seen here

#

    ret = uacpi_gas_read(&reg->enable, &enable_mask);
    if (uacpi_unlikely_error(ret))
        goto out;

    switch (state) {
    case GPE_STATE_ENABLED:
        enable_mask |= event_bit;
        break;
    case GPE_STATE_DISABLED:
        enable_mask &= ~event_bit;
        break;
    default:
        ret = UACPI_STATUS_INVALID_ARGUMENT;
        goto out;
    }

    ret = uacpi_gas_write(&reg->enable, enable_mask);```

#

then hangs somewhere here

fiery turtle Jan 24, 2025, 1:35 PM

#

Gas write dispatches into your code btw

north holly Jan 24, 2025, 1:35 PM

#

in between the two locks

north holly Jan 24, 2025, 1:36 PM

#

fiery turtle Gas write dispatches into your code btw

which apis

fiery turtle Jan 24, 2025, 1:36 PM

#

Probably io_map and then io_write

#

But add more prints

north holly Jan 24, 2025, 1:47 PM

#

uacpi is last seen during or after the gas write

#

I think it hangs in outb

#

writing 0x22 to I/O port 0x182a is the last thing I see

fiery turtle Jan 24, 2025, 1:50 PM

#

Wtf

north holly Jan 24, 2025, 1:50 PM

#

byte width is 1

fiery turtle Jan 24, 2025, 1:50 PM

#

And that never returns?

north holly Jan 24, 2025, 1:50 PM

#

so it's an outb

north holly Jan 24, 2025, 1:50 PM

#

fiery turtle And that never returns?

nope

#

there was a previous write to that port:

writing 0x02 to I/O port 0x182a```

#

a read made right before that indeed returned to the caller

#

with 0x02 as the value read

fiery turtle Jan 24, 2025, 1:52 PM

#

Huh

north holly Jan 24, 2025, 1:52 PM

#

basically:

acquired [some random spinlock]
read 0x02 from 0x182a
writing 0x22 to I/O port 0x182a
HANGS

fiery turtle Jan 24, 2025, 1:52 PM

#

Can u roll back to previous uacpi but keep these io traces? To see if the values written have changed or something

north holly Jan 24, 2025, 1:53 PM

#

sure

fiery turtle Jan 24, 2025, 1:53 PM

#

Thanks

#

Although I dont understand how outb can hang

#

Any chance u get an irq during that?

north holly Jan 24, 2025, 1:54 PM

#

IRQL is at dispatch

#

so no timer/scheduler IRQs

fiery turtle Jan 24, 2025, 1:54 PM

#

What about acpi

north holly Jan 24, 2025, 1:54 PM

#

GPEs are blocked at IRQL 8 (some random IRQL, nothing special about it)

#

and IRQL_GPE > IRQL_DISPATCH

#

so it might get a GPE which throws stuff off, but it seems unlikely that it would happen every time

fiery turtle Jan 24, 2025, 1:56 PM

#

Because that thing might enable a gpe with a spinlock held, if you dont actually disable irqs and call into uacpi again that will hang

#

Do u trace spinlock acquires?

north holly Jan 24, 2025, 1:56 PM

#

spinlock acquires are being traced

#

uacpi_cpu_flags uacpi_kernel_lock_spinlock(uacpi_handle hnd)
{
    spinlock* lock = (spinlock*)hnd;
    printf("acquiring %p\n", hnd);
    irql r = Core_SpinlockAcquire(lock);
    printf("acquired %p\n", hnd);
    return r;
}```

#

and there are no logs related to spinlocks after the write

fiery turtle Jan 24, 2025, 1:58 PM

#

I see

#

But yeah lets see the io traces with older uacpi

north holly Jan 24, 2025, 1:58 PM

#

is master good?

#

or is that too old?

#

actually I'd ideally be on the commit that I knew to work before

#

1d636a34152dc82833c89175b702f2c0671f04e3

fiery turtle Jan 24, 2025, 1:59 PM

#

north holly is master good?

Yeah

fiery turtle Jan 24, 2025, 1:59 PM

#

north holly `1d636a34152dc82833c89175b702f2c0671f04e3`

How old is that

#

Compared to master

north holly Jan 24, 2025, 1:59 PM

#

https://github.com/uACPI/uACPI/commit/1d636a34152dc82833c89175b702f2c0671f04e3

GitHub

Fix a bunch of warnings for newer GCC versions · uACPI/uACPI@1d636a3

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

fiery turtle Jan 24, 2025, 1:59 PM

#

Thats basically master

north holly Jan 24, 2025, 2:00 PM

#

ye

#

HEAD~1

fiery turtle Jan 24, 2025, 2:01 PM

#

Ye

#

And the last commit was cosmetic

north holly Jan 24, 2025, 2:02 PM

#

cmake is being retarted

#

*restarted

#

huh what

#

it still hangs on old uacpi

#

same value written

fiery turtle Jan 24, 2025, 2:05 PM

#

But u said it just worked on the old uacoi 💀💀💀

north holly Jan 24, 2025, 2:05 PM

#

it worked on old uacpi

#

on his PC

#

guess I never bothered to check my pc

fiery turtle Jan 24, 2025, 2:05 PM

#

Bruh

#

Maybe try older uacpi or obos something?

#

And make sure u did overwrite the usb

north holly Jan 24, 2025, 2:06 PM

#

yeah I did

fiery turtle Jan 24, 2025, 2:06 PM

#

Oh also

north holly Jan 24, 2025, 2:06 PM

#

overwrite the usb

fiery turtle Jan 24, 2025, 2:06 PM

#

What if u dont do gpe set for wake stuff

#

And just do finalize

north holly Jan 24, 2025, 2:07 PM

#

soo skip the call to OBOS_InitWakeGPEs

fiery turtle Jan 24, 2025, 2:07 PM

#

Basically

#

And if even that doesnt work we can try pmos or astral

fiery turtle Jan 24, 2025, 2:08 PM

#

north holly it worked on old uacpi

Also maybe it just doesnt repro every time?

north holly Jan 24, 2025, 2:09 PM

#

north holly soo skip the call to `OBOS_InitWakeGPEs`

still hangs

#

without that

#

and I can confirm the usb was overwritten

#

I'm going to download pmOS

fiery turtle Jan 24, 2025, 2:09 PM

#

Try that ig

#

That might have a fucked framebuffer tho

#

If that doesnt work will try proxima

north holly Jan 24, 2025, 2:10 PM

#

fiery turtle That might have a fucked framebuffer tho

bah whatever

#

I'll just see if the power button shuts down pmOS

fiery turtle Jan 24, 2025, 2:10 PM

#

Yeah

north holly Jan 24, 2025, 2:10 PM

#

which should be enough to see if it works

north holly Jan 24, 2025, 2:11 PM

#

calm latch

this iso worked...

#

idk how new of uacpi it is

#

unless

#

it doesn't

#

idk if it's hanging or what

#

but the power button isn't shutting it down

#

it says that it enabled 11 GPEs

fiery turtle Jan 24, 2025, 2:14 PM

#

north holly idk how new of uacpi it is

Its pr head

fiery turtle Jan 24, 2025, 2:15 PM

#

north holly but the power button isn't shutting it down

@calm latch is it supposed to?

fiery turtle Jan 24, 2025, 2:15 PM

#

north holly it says that it enabled 11 GPEs

Btw can u give a dump of this pc

north holly Jan 24, 2025, 2:16 PM

#

you already have it

calm latch Jan 24, 2025, 2:16 PM

#

fiery turtle <@350748255929499649> is it supposed to?

It is

fiery turtle Jan 24, 2025, 2:16 PM

#

Which one is it

calm latch Jan 24, 2025, 2:16 PM

#

Idk it only works in QEMU

north holly Jan 24, 2025, 2:16 PM

#

fiery turtle Which one is it

it's an acer pc if that helps

north holly Jan 24, 2025, 2:16 PM

#

calm latch Idk it only works in QEMU

does it use fixed events and EC?

north holly Jan 24, 2025, 2:17 PM

#

north holly it's an acer pc if that helps

but it's probably the first one

calm latch Jan 24, 2025, 2:17 PM

#

The only nice thing is that now it's consistently broken

calm latch Jan 24, 2025, 2:17 PM

#

north holly does it use fixed events and EC?

Yes

north holly Jan 24, 2025, 2:17 PM

#

it isn't shutting down

calm latch Jan 24, 2025, 2:17 PM

#

calm latch It is

I would imagine the kernel is panicking

north holly Jan 24, 2025, 2:17 PM

#

on power button

#

last thing I see if is a test exception log

calm latch Jan 24, 2025, 2:17 PM

#

And nothing?

north holly Jan 24, 2025, 2:17 PM

#

then nothing

calm latch Jan 24, 2025, 2:17 PM

#

So the framebuffer is working

#

Does keyboard work?

north holly Jan 24, 2025, 2:17 PM

#

also it seems to test the AHCI controller

north holly Jan 24, 2025, 2:18 PM

#

calm latch Does keyboard work?

spamming the keys on the keyboard does nothing

calm latch Jan 24, 2025, 2:18 PM

#

north holly also it seems to test the AHCI controller

Yeah it's the thing I'm working on

#

bruh

#

I'm not at home but will debug this evening

#

Maybe the AHCI driver is messing everything up

fiery turtle Jan 24, 2025, 2:20 PM

#

@north holly can u try this?

north holly Jan 24, 2025, 2:20 PM

#

k

fiery turtle Jan 24, 2025, 2:21 PM

#

alothugh fuck i dont think it hooks up the power button

#

@gentle peak (sry for ping) where can we get the latest proxima build that shuts down on power button?

north holly Jan 24, 2025, 2:21 PM

#

yeah it also doesn't boot

fiery turtle Jan 24, 2025, 2:21 PM

#

wdym

#

it boots here

north holly Jan 24, 2025, 2:21 PM

#

yeah it triple faults

fiery turtle Jan 24, 2025, 2:21 PM

#

fuck

#

lmfao

gentle peak Jan 24, 2025, 2:21 PM

#

fiery turtle alothugh fuck i dont think it hooks up the power button

iirc that one does

fiery turtle Jan 24, 2025, 2:22 PM

#

like in qemu?

fiery turtle Jan 24, 2025, 2:22 PM

#

gentle peak iirc that one does

doesnt work in my qemu

north holly Jan 24, 2025, 2:22 PM

#

no, it triple faults in real hardware

fiery turtle Jan 24, 2025, 2:22 PM

#

and doesnt shut down in qemu

gentle peak Jan 24, 2025, 2:22 PM

#

anything built after dec 31 should have the power button stuff

#

2s

north holly Jan 24, 2025, 2:23 PM

#

last thing I see is limine messages

fiery turtle Jan 24, 2025, 2:23 PM

#

i tried looking for your ci image but i dont se the links in the repo

north holly Jan 24, 2025, 2:23 PM

#

saying it's loading the kernel

#

then a triple fault instantly

gentle peak Jan 24, 2025, 2:24 PM

#

apparently i lied in that message and the build was not for x86-64-v0

north holly Jan 24, 2025, 2:24 PM

#

bruh

gentle peak Jan 24, 2025, 2:24 PM

#

i'll get the latest from ci

north holly Jan 24, 2025, 2:24 PM

#

my pc is x86_64-v3 iirc

#

either that or x86_64-v2

#

it's haswell

fiery turtle Jan 24, 2025, 2:24 PM

#

maybe that was the SSE build

gentle peak Jan 24, 2025, 2:24 PM

#

i believe that one was where i built with -march=raptorlake -mtune=raptorlake accidentally because i forgot you have to fully reconfigure with meson for those to apply

fiery turtle Jan 24, 2025, 2:25 PM

#

makes sense

north holly Jan 24, 2025, 2:25 PM

#

yeah I think haswell is a bit older than raptorlake

gentle peak Jan 24, 2025, 2:25 PM

#

here's the latest from CI

📎 proxima.iso

fiery turtle Jan 24, 2025, 2:26 PM

#

hmm still no shutdown

#

#

is this the expected output?

gentle peak Jan 24, 2025, 2:26 PM

#

that's the expected output yeah

north holly Jan 24, 2025, 2:27 PM

#

still triple faults

gentle peak Jan 24, 2025, 2:27 PM

#

i forgot currently it panics on init and therefore irqs are disabled

fiery turtle Jan 24, 2025, 2:27 PM

#

bruh we dont have a single non broken kernel to test uacpi with 💀

gentle peak Jan 24, 2025, 2:27 PM

#

it's not really broken

north holly Jan 24, 2025, 2:27 PM

#

astral

gentle peak Jan 24, 2025, 2:27 PM

#

everything works as intended

#

that image just doesn't have a /bin/init

fiery turtle Jan 24, 2025, 2:27 PM

#

gentle peak i forgot currently it panics on init and therefore irqs are disabled

ah

gentle peak Jan 24, 2025, 2:27 PM

#

so it panics on init and disables irqs

fiery turtle Jan 24, 2025, 2:27 PM

#

can u not do that?

#

well it also triple faults on oberrows pc

north holly Jan 24, 2025, 2:27 PM

#

gentle peak it's not really broken

it triple faults on my PC where the bug I'm trying to debug is on

gentle peak Jan 24, 2025, 2:27 PM

#

yeah give me like 2s

gentle peak Jan 24, 2025, 2:28 PM

#

fiery turtle i tried looking for your ci image but i dont se the links in the repo

it's github actions, you can't download the artifact if you're not logged in and idk if you need to be a member of the org or anything like that

fiery turtle Jan 24, 2025, 2:28 PM

#

north holly >astral

astral doesnt enable events

north holly Jan 24, 2025, 2:28 PM

#

not even finalize gpe init?

fiery turtle Jan 24, 2025, 2:28 PM

#

there's managarm but its a 5gb iso

#

if u want to dd that

fiery turtle Jan 24, 2025, 2:28 PM

#

north holly not even finalize gpe init?

ye

gentle peak Jan 24, 2025, 2:28 PM

#

latest local build, which actually has a /bin/init

📎 proxima.iso

gentle peak Jan 24, 2025, 2:29 PM

#

north holly it triple faults on my PC where the bug I'm trying to debug is on

that's very frustrating to debug fuck

fiery turtle Jan 24, 2025, 2:29 PM

#

gentle peak latest local build, which actually has a /bin/init

hmm it says shutting down in 3 seconds but actually shuts down in like 30

north holly Jan 24, 2025, 2:29 PM

#

fiery turtle ye

btw pmOS does boot past finalize_gpe_init

gentle peak Jan 24, 2025, 2:29 PM

#

it counts down to 0 for me and then immediately shuts down

#

this is in qemu yeah?

fiery turtle Jan 24, 2025, 2:30 PM

#

yeah

#

no invariant tsc here tho

fiery turtle Jan 24, 2025, 2:30 PM

#

north holly btw pmOS does boot past `finalize_gpe_init`

yes but it doesnt shut down

#

which ur pc should because its a fixed event i think

gentle peak Jan 24, 2025, 2:30 PM

#

fiery turtle no invariant tsc here tho

as in qemu doesn't report it or your host doesn't have it

north holly Jan 24, 2025, 2:30 PM

#

yeah it's fixed

gentle peak Jan 24, 2025, 2:30 PM

#

because proxima ignores what cpuid says about invtsc presence

fiery turtle Jan 24, 2025, 2:30 PM

#

gentle peak as in qemu doesn't report it or your host doesn't have it

the former

median crest Jan 24, 2025, 2:30 PM

#

fiery turtle astral doesnt enable events

I thought you did that since the power button works

fiery turtle Jan 24, 2025, 2:30 PM

#

nah thats a separate thing

median crest Jan 24, 2025, 2:31 PM

#

Oh

fiery turtle Jan 24, 2025, 2:31 PM

#

the thign that hangs on obos is not called on astral

#

astral doesnt call it because i forgor

median crest Jan 24, 2025, 2:32 PM

#

:skull

gentle peak Jan 24, 2025, 2:32 PM

#

oh the build i sent also triple faults on my own haswell test machine

#

that makes it easier to debug at least

fiery turtle Jan 24, 2025, 2:32 PM

#

💀

#

pls fix it so i can prove its an obos bug

gentle peak Jan 24, 2025, 2:32 PM

#

give me like an hour

#

i should really set up pxe

fiery turtle Jan 24, 2025, 2:33 PM

#

letsgo

#

although i mean pmos doesnt hang so

#

i kinda did prove its an obos bug

north holly Jan 24, 2025, 2:33 PM

#

it is a outb

gentle peak Jan 24, 2025, 2:33 PM

#

oh also this uses uacpi master not the 1.0 rc

fiery turtle Jan 24, 2025, 2:33 PM

#

@north holly are u brave enough to try latest managarm?

north holly Jan 24, 2025, 2:33 PM

#

idk how that would be bugged

north holly Jan 24, 2025, 2:33 PM

#

fiery turtle <@841762704766140486> are u brave enough to try latest managarm?

no

fiery turtle Jan 24, 2025, 2:34 PM

#

gentle peak oh also this uses uacpi master not the 1.0 rc

thats fine, obos hangs on master too

fiery turtle Jan 24, 2025, 2:34 PM

#

north holly idk how that would be bugged

because that enables irq generation for those gpes

north holly Jan 24, 2025, 2:34 PM

#

north holly no

I don't feel like finding an iso

fiery turtle Jan 24, 2025, 2:34 PM

#

i suspect u get an irq and hang there

#

i can send u the link

north holly Jan 24, 2025, 2:34 PM

#

I mask IRQs

#

now

#

still hung

fiery turtle Jan 24, 2025, 2:35 PM

#

probably didnt mask well enough

#

because outb cant hang

#

unless it triggers and smi, which this one shouldnt i hope

north holly Jan 24, 2025, 2:35 PM

#

it's at irql 15

north holly Jan 24, 2025, 2:35 PM

#

north holly it's at irql 15

the only thing that can happen is NMI

#

and SMI

#

probably

fiery turtle Jan 24, 2025, 2:36 PM

#

@north holly can u try this one pls https://builds.managarm.org/project/managarm/files/2025-01-22T21:56:41

north holly Jan 24, 2025, 2:37 PM

#

993.0M

fiery turtle Jan 24, 2025, 2:37 PM

#

yes

north holly Jan 24, 2025, 2:37 PM

#

too much time to download

fiery turtle Jan 24, 2025, 2:37 PM

#

fuck

#

@kindred beacon do u have a small bootable managarm image perchance?

gentle peak Jan 24, 2025, 2:38 PM

#

ok what the fuck?

#

if i set up an hdd image on the usb with the exact same kernel and initrd images it boots fine

#

but if i dd an iso onto it it's broken

fiery turtle Jan 24, 2025, 2:38 PM

#

maybe its a limine bug with some memory map edge case again

#

or similar

gentle peak Jan 24, 2025, 2:39 PM

#

i mean i saw it get into the kernel on the iso boot

#

it triple faulted after time: tsc is ... MHz

fiery turtle Jan 24, 2025, 2:39 PM

#

ah

north holly Jan 24, 2025, 2:39 PM

#

gentle peak it triple faulted after `time: tsc is ... MHz`

I didn't

kindred beacon Jan 24, 2025, 2:39 PM

#

fiery turtle <@182819561203040256> do u have a small bootable managarm image perchance?

Small is 8 gigs for me

gentle peak Jan 24, 2025, 2:39 PM

#

i can confirm the shutdown works on my machine though

kindred beacon Jan 24, 2025, 2:39 PM

#

But we have a nightly image linked in the readme

fiery turtle Jan 24, 2025, 2:39 PM

#

its 4 gigs tho

#

and oberrow cant download that

north holly Jan 24, 2025, 2:40 PM

#

I mean I can but it takes many long time

kindred beacon Jan 24, 2025, 2:40 PM

#

We default to 4 gigs lol

fiery turtle Jan 24, 2025, 2:40 PM

#

gentle peak i can confirm the shutdown works on my machine though

haswell fix pleading

gentle peak Jan 24, 2025, 2:40 PM

#

i did not make any changes

#

that's the thing

#

i literally just switched from dd'ing an iso to setting up a proper hddboot on the usb for ease of debugging

fiery turtle Jan 24, 2025, 2:41 PM

#

can u send over that raw hdd image?

slim panther Jan 24, 2025, 2:42 PM

#

fiery turtle <@182819561203040256> do u have a small bootable managarm image perchance?

arent all managarm images at least 4gb

flat badge Jan 24, 2025, 2:42 PM

#

It's not 4gigs compressed

fiery turtle Jan 24, 2025, 2:42 PM

#

nah qookie was able to make a small one

#

of a few megs

slim panther Jan 24, 2025, 2:42 PM

#

well yea compressed

#

it's mostly 0

#

but image-size is 4G

north holly Jan 24, 2025, 2:42 PM

#

also, can't outb technically hang if you do a bad read/write

#

to an IO port

fiery turtle Jan 24, 2025, 2:43 PM

#

nop

#

gimme a sec ill take a look at your dump

north holly Jan 24, 2025, 2:43 PM

#

port in question is 0x182a

fiery turtle Jan 24, 2025, 2:44 PM

#

is it ASUS?

north holly Jan 24, 2025, 2:44 PM

#

ACER

fiery turtle Jan 24, 2025, 2:44 PM

#

I have LENOVO, DELL, and ASUS

#

3 of your dumps

north holly Jan 24, 2025, 2:44 PM

#

I don't own anything ASUS

fiery turtle Jan 24, 2025, 2:45 PM

#

the motherboard might be asus

flat badge Jan 24, 2025, 2:45 PM

#

Somebody with a managarm dev env could just do xbstrap install base && xbstrap run initialize-empty-image remake-image

#

to get a minimal image

north holly Jan 24, 2025, 2:45 PM

#

fiery turtle the motherboard might be asus

what's the OEM ID

#

the one I have is ACRSYS

#

in the dsdt

fiery turtle Jan 24, 2025, 2:46 PM

#

north holly port in question is 0x182a

[050h 0080 4] GPE0 Block Address : 00001820

#

with lenght 16

#

so yeah its correct

north holly Jan 24, 2025, 2:46 PM

#

what's offset 0xa supposed to mean?

frank canopy Jan 24, 2025, 2:46 PM

#

now I'm getting tempted to make a kernel that just dumps out the entire acpi tables to com1 or something just to make making the dumps easier

fiery turtle Jan 24, 2025, 2:46 PM

#

half is status half is enable bits

gentle peak Jan 24, 2025, 2:47 PM

#

gentle peak i literally just switched from dd'ing an iso to setting up a proper hddboot on t...

so i guess the fix is ```sh
DEVICE=/dev/...
LIMINE=$(limine --print-datadir)

sfdisk "$DEVICE" << EOF
,+,0c,-
EOF
limine bios-install "$DEVICE"
mkfs.fat "${DEVICE}1"
mount "${DEVICE}1" /mnt
mkdir -p /mnt/EFI/BOOT
cp "$LIMINE/BOOTX64.EFI" /mnt/EFI/BOOT
cp "$LIMINE/limine-bios.sys" /mnt
cat > /mnt/limine.conf << EOF
timeout: 5

/Proxima
protocol: limine
kernel_path: boot():/boot/hydrogen
module_path: boot():/boot/proxima.tar
EOF
umount /mnt

echo "Now copy /boot/hydrogen and /boot/proxima.tar from the ISO to the USB"

fiery turtle Jan 24, 2025, 2:47 PM

#

offset A is GPE 0xA - (16 / 2)

north holly Jan 24, 2025, 2:47 PM

#

so 0x182a is GPE 10 of GPE block zero?

fiery turtle Jan 24, 2025, 2:48 PM

#

no, its at least GPE 16

#

each bit is one gpe

#

0xA - (16 / 2) is 2

#

so 16 bits

#

and then we would check the bit that it sets in that reg

north holly Jan 24, 2025, 2:49 PM

#

I see

fiery turtle Jan 24, 2025, 2:51 PM

#

btw the oem id is dell @north holly

north holly Jan 24, 2025, 2:51 PM

#

wrong ACPI tables

fiery turtle Jan 24, 2025, 2:51 PM

#

bruh

north holly Jan 24, 2025, 2:51 PM

#

[000h 0000   4]                    Signature : "FACP"    [Fixed ACPI Description Table (FADT)]
[004h 0004   4]                 Table Length : 0000010C
[008h 0008   1]                     Revision : 05
[009h 0009   1]                     Checksum : F5
[00Ah 0010   6]                       Oem ID : "ACRSYS"
[010h 0016   8]                 Oem Table ID : "ACRPRDCT"
[018h 0024   4]                 Oem Revision : 01072009
[01Ch 0028   4]              Asl Compiler ID : "AMI "
[020h 0032   4]        Asl Compiler Revision : 00010013

fiery turtle Jan 24, 2025, 2:51 PM

#

dont have this dump

#

is it same gpe block base?

north holly Jan 24, 2025, 2:52 PM

#

yes

fiery turtle Jan 24, 2025, 2:52 PM

#

and length?

north holly Jan 24, 2025, 2:52 PM

#

[050h 0080   4]           GPE0 Block Address : 00001820

#

[05Ch 0092   1]            GPE0 Block Length : 10

fiery turtle Jan 24, 2025, 2:52 PM

#

ye

north holly Jan 24, 2025, 2:53 PM

#

📎 dump.txt

fiery turtle Jan 24, 2025, 2:53 PM

#

64 gpe regs, 128 bits total, 64 status bits, 64 enable bits

north holly Jan 24, 2025, 2:53 PM

#

north holly

that's the ACPI dump of this PC

#

for future reference

fiery turtle Jan 24, 2025, 2:53 PM

#

thanks

north holly Jan 24, 2025, 2:56 PM

#

bro wtf is this

#

nooo

fiery turtle Jan 24, 2025, 2:56 PM

#

wdym

north holly Jan 24, 2025, 2:56 PM

#

it just wants to hang

#

on outb

gentle peak Jan 24, 2025, 2:56 PM

#

for reference: this hddboot image works on my haswell machine

📎 proxima.img

fiery turtle Jan 24, 2025, 2:57 PM

#

@north holly ^

#

inb4 also hangs

north holly Jan 24, 2025, 2:58 PM

#

gentle peak for reference: this hddboot image works on my haswell machine

and it doesn't on mine

#

at least when I dd the image

#

I'm gonna go get something to eat now

fiery turtle Jan 24, 2025, 2:58 PM

#

bruh

#

so unlucky

#

@gentle peak btw if you have time, could you test out the 1.0rc pr?

gentle peak Jan 24, 2025, 3:06 PM

#

fiery turtle <@938223886439436320> btw if you have time, could you test out the 1.0rc pr?

origin/address-spaces-support?

fiery turtle Jan 24, 2025, 3:07 PM

#

yup

#

I wonder why proxima triple faults on obos tho

#

oberrows pc i mean

gentle peak Jan 24, 2025, 3:11 PM

#

if i had a bit more patience i'd investigate it

#

but i have literally nothing to go off of given that (presumably) the triple fault occurs before the display "recovers" from limine's modeset

#

so it'd take a long time to narrow down

fiery turtle Jan 24, 2025, 3:12 PM

#

true

gentle peak Jan 24, 2025, 3:14 PM

#

fiery turtle <@938223886439436320> btw if you have time, could you test out the 1.0rc pr?

uacpi regression!

#

not receiving the power button event

#

i'm getting the VALZ no listeners warning

#

but the actual handler isn't getting invoked

fiery turtle Jan 24, 2025, 3:14 PM

#

ok this is actually good

north holly Jan 24, 2025, 3:15 PM

#

I have found out it's tryna do something to GPE 21

#

it's trying to enable it

#

now time to see what GPE 21 is

gentle peak Jan 24, 2025, 3:15 PM

#

gentle peak not receiving the power button event

this happens on real hardware too

#

on qemu i mean

#

sorry

#

i'm not getting the valz no listeners thing on qemu obv because that's a toshiba thing

#

but the power button handler isn't getting invoked either

fiery turtle Jan 24, 2025, 3:19 PM

#

nice because i can probably repro it in astral locally and then debug from there

gentle peak Jan 24, 2025, 3:19 PM

#

yeah it happens on debug builds running in tcg and everything

fiery turtle Jan 24, 2025, 3:20 PM

#

as a quick guess, perhaps reverting dfd92d5d87acadc574619b1b0401dc59a92da422 could fix it

#

potentially followed by e7ffcaa016515e4857630af6ae342fb0eb97ceb6

calm latch Jan 24, 2025, 3:21 PM

#

calm latch

Power button's also broken in QEMU for some reason...

fiery turtle Jan 24, 2025, 3:21 PM

#

yeye same in proxima

calm latch Jan 24, 2025, 3:21 PM

#

calm latch Power button's also broken in QEMU for some reason...

Maybe I messed something up

fiery turtle Jan 24, 2025, 3:21 PM

#

its a uacpi bug in the pr

#

some caching logic might be incorrect i think

#

ill debug it soon

north holly Jan 24, 2025, 3:22 PM

#

I'm gonna test a really old obos commit

#

before I even had an EC driver

#

to see what happens there

#

seriously infy

#

you gotta stop rewriting git history

#

Cloning into 'uACPI'...
fatal: unable to read tree (cc2ac5f788d81874b9c677c763b803aa375e9395)
CMake Error at /home/oberrow/Code/obos-old/build/_deps/uacpi-subbuild/uacpi-populate-prefix/tmp/uacpi-populate-gitclone.cmake:61 (message):
  Failed to checkout tag: 'cc2ac5f788d81874b9c677c763b803aa375e9395'

fiery turtle Jan 24, 2025, 3:24 PM

#

i think i know whats going on

#

but for that i have to fix astral builds

fiery turtle Jan 24, 2025, 3:25 PM

#

fiery turtle as a quick guess, perhaps reverting dfd92d5d87acadc574619b1b0401dc59a92da422 cou...

@gentle peak does reverting both of these on top of uacpi's pr fix it for u in qemu?

gentle peak Jan 24, 2025, 3:28 PM

#

i'll try that in a sec but i'm stepping through the code a bit and i noticed that when do_read_one(mapping, 0, &value0) calls uacpi_gas_read_mapped, gas->total_bit_width is 0 so it doesn't read anything, and do_read_one(mapping, 1, &value1) does an early exit because it isn't in REGISTER_MAPPING_STATE_MAPPED

#

this is while reading PM1_STS

#

which seems wrong

north holly Jan 24, 2025, 3:28 PM

#

obos 3783b02138985e269bceb137233328b658411427 works

fiery turtle Jan 24, 2025, 3:28 PM

#

yeah because PM1_STS can have two separate registers

#

and pm1b is not implemented on qemu

north holly Jan 24, 2025, 3:29 PM

#

which is the commit right before any sort of EC driver was added

fiery turtle Jan 24, 2025, 3:29 PM

#

so it just skips reading it

fiery turtle Jan 24, 2025, 3:29 PM

#

gentle peak i'll try that in a sec but i'm stepping through the code a bit and i noticed tha...

oh wait, bit width is 0 even for the first one? yeah tahts wrong

gentle peak Jan 24, 2025, 3:29 PM

#

fiery turtle yeah because PM1_STS can have two separate registers

i'm more so worried about the first one

#

oh great the revert has a merge conflict

fiery turtle Jan 24, 2025, 3:30 PM

#

yeah nvm dont even try to revet it, obv thats the bug

#

im not sure why it gets a zero there tho

gentle peak Jan 24, 2025, 3:31 PM

#

in map_one for PM1a, the temp_gas has 0 for register_bit_{width,offset}

#

spec->access_width is 0

fiery turtle Jan 24, 2025, 3:32 PM

#

thats correct

#

its only specified for non-gas

#

gas_validate somehow returns 0 for total_width i guess?

gentle peak Jan 24, 2025, 3:32 PM

#

yeah but that info is then copied into the mapping by uacpi_map_gas_noalloc

fiery turtle Jan 24, 2025, 3:33 PM

#

gentle peak yeah but that info is then copied into the `mapping` by `uacpi_map_gas_noalloc`

wdym? total_width is returned from gas_validate right?

gentle peak Jan 24, 2025, 3:33 PM

#

the total_width calculated by gas_validate is align_up(bit_offset + bit_width, uacpi_size)

#

both of which are 0

fiery turtle Jan 24, 2025, 3:34 PM

#

yeah, but it aligns up to computed access width

#

or i guess it aligns the zero up to zero?

gentle peak Jan 24, 2025, 3:34 PM

#

this is in gas_get_access_bit_width

#

and again register_bit_width is 0

fiery turtle Jan 24, 2025, 3:34 PM

#

wait a sec

#

why is bit_width zeroed out even

#

it shouldnt be i think

gentle peak Jan 24, 2025, 3:36 PM

#

spec->access_width is 0

#

(the one passed to map_one)

fiery turtle Jan 24, 2025, 3:36 PM

#

but pm1 is REGISTER_KIND_GAS no?

gentle peak Jan 24, 2025, 3:37 PM

#

access_kind is 1

#

REGISTER_KIND_IO

fiery turtle Jan 24, 2025, 3:37 PM

#

whats the backtrace?

gentle peak Jan 24, 2025, 3:37 PM

#

#

#0  gas_validate (gas=0xffffffff9034ee84, access_bit_width=0xffffffff9034ee4b "", bit_width=0xffffffff9034ee4a "") at ../../../../../packages/hydrogen/src/subprojects/uacpi/source/io.c:727
#1  0xffffffff800437e0 in uacpi_map_gas_noalloc (gas=0xffffffff9034ee84, out_mapped=0xffffffff8004f500 <g_register_mappings>) at ../../../../../packages/hydrogen/src/subprojects/uacpi/source/io.c:837
#2  0xffffffff80044d76 in map_one (spec=0xffffffff80003da0 <g_registers>, mapping=0xffffffff8004f500 <g_register_mappings>, idx=0 '\000') at ../../../../../packages/hydrogen/src/subprojects/uacpi/source/registers.c:139
#3  0xffffffff80044e39 in ensure_register_mapped (spec=0xffffffff80003da0 <g_registers>, mapping=0xffffffff8004f500 <g_register_mappings>) at ../../../../../packages/hydrogen/src/subprojects/uacpi/source/registers.c:168
#4  0xffffffff80045064 in uacpi_read_register (reg_enum=UACPI_REGISTER_PM1_STS, out_value=0xffffffff9034ef20) at ../../../../../packages/hydrogen/src/subprojects/uacpi/source/registers.c:236
#5  0xffffffff80046773 in handle_fixed_events () at ../../../../../packages/hydrogen/src/subprojects/uacpi/source/event.c:205
#6  0xffffffff8004877a in handle_sci (ctx=0xffff800007f37440) at ../../../../../packages/hydrogen/src/subprojects/uacpi/source/event.c:2099
#7  0xffffffff80011835 in acpi_irq_dispatcher (ptr=0xffff800007d53980) at ../../../../../packages/hydrogen/src/kernel/src/drv/acpi.c:374
#8  0xffffffff80023665 in sched_init_task (func=0xffffffff800117de <acpi_irq_dispatcher>, ctx=0xffff800007d53980, task=0xffff800007d3d800) at ../../../../../packages/hydrogen/src/kernel/src/sched/sched.c:439
#9  0xffffffff8000eb7e in task_init_stub () at ../../../../../packages/hydrogen/src/kernel/src/sched/sched.S:42```

#

here's the raw one from gdb

fiery turtle Jan 24, 2025, 3:38 PM

#

    [UACPI_REGISTER_PM1_STS] = {
        .kind = REGISTER_KIND_GAS,

💀

gentle peak Jan 24, 2025, 3:38 PM

#

yeah it gets changed somewhere

fiery turtle Jan 24, 2025, 3:39 PM

#

its const too

#

so it should be mapped RO

gentle peak Jan 24, 2025, 3:39 PM

#

oh wait i was looking at access_kind not kind

#

you should really make the fields in the struct themselves enums instead of chars so the debugger prints them nicer

#

yeah kind is 0

fiery turtle Jan 24, 2025, 3:39 PM

#

the struct would be huge then

#

like u cant have typed enums in c

#

unless its c23

gentle peak Jan 24, 2025, 3:40 PM

#

i mean

#

not really

#

right now it's 3*u8 followed by a pointer

#

if you changed them to enums you'd only be wasting 4 more bytes

fiery turtle Jan 24, 2025, 3:40 PM

#

on 64-bit platforms

#

but yeah

gentle peak Jan 24, 2025, 3:41 PM

#

ah i see you made the same mistake i did

fiery turtle Jan 24, 2025, 3:41 PM

#

bruh

#

brain damage

gentle peak Jan 24, 2025, 3:42 PM

#

if i change that to spec->kind it works perfectly

fiery turtle Jan 24, 2025, 3:42 PM

#

nice

#

lemme make the fix

#

alright fixed

#

thanks for your help

gentle peak Jan 24, 2025, 3:43 PM

#

no problem

fiery turtle Jan 24, 2025, 3:43 PM

#

does real hw work as well now?

gentle peak Jan 24, 2025, 3:43 PM

#

testing now

fiery turtle Jan 24, 2025, 3:43 PM

#

(ideally on current pr head)

gentle peak Jan 24, 2025, 3:45 PM

#

did you rewrite git history or something? git's saying the local and origin branches have diverged

#

it's not an issue but still

fiery turtle Jan 24, 2025, 3:46 PM

#

yeah ofc, i just pushed the fix into the commit which introduced the bug

gentle peak Jan 24, 2025, 3:46 PM

#

works on real hw

fiery turtle Jan 24, 2025, 3:46 PM

#

nice

#

there's another behavior change which i dont like

#

    total_width = UACPI_ALIGN_UP(
        gas->register_bit_offset + gas->register_bit_width,
        *access_bit_width, uacpi_size
    );

#

previously this wasnt getting aligned up when doing io

#

not that it should matter in this case

#

@north holly ive just fixed a relatively important bug, can u retest

#

probably not the one that caused your hang but

frank canopy Jan 24, 2025, 3:50 PM

#

it might be

#

who knows

gentle peak Jan 24, 2025, 3:50 PM

#

i mean it could've been the cause, since the bug had the potential to access mmio as pio

fiery turtle Jan 24, 2025, 3:50 PM

#

the other way around

#

but yeah

frank canopy Jan 24, 2025, 3:54 PM

#

with how interconnected everything is i still wouldnt be shocked if it was related to the hang in obos

fiery turtle Jan 24, 2025, 3:55 PM

#

thing is, obos hangs even on master

#

before any of these changes

gentle peak Jan 24, 2025, 3:56 PM

#

fair

frank canopy Jan 24, 2025, 3:56 PM

#

oh true

fiery turtle Jan 24, 2025, 3:56 PM

#

and e.g. latest pmos did make it past the spot where obos hung

#

but didnt shutdown because of this bug that monkuous found

#

@calm latch btw can u git fetch && git reset --hard origin/address-spaces-support and rebuild the image?

#

so oberrow can retest on his pc

#

because proxima triple faults on that

frank canopy Jan 24, 2025, 3:59 PM

#

upsides of using zig fetch & co for dep management:
- zig fetch --save=uacpi git+https://github.com/uACPI/uACPI#address-spaces-support and its done without having to deal with git history management on a submodule
downsides:
- no local copy to edit for adding printfs

fiery turtle Jan 24, 2025, 3:59 PM

#

no local copy kinda sucks yeah

frank canopy Jan 24, 2025, 4:00 PM

#

i think it stores one in the build cache but thats checksummed so cant edit it and hard to find

#

but if i dont need to edit uacpi then dang is it nice

#

can also put in a path instead of a url to do it with submodules or whatever in case of emergency

#

thats what i do for other deps i have needed to edit

fiery turtle Jan 24, 2025, 4:04 PM

#

True

north holly Jan 24, 2025, 4:13 PM

#

fiery turtle <@841762704766140486> ive just fixed a relatively important bug, can u retest

I'm going to do that after playing Minecraft

fiery turtle Jan 24, 2025, 4:14 PM

#

nice

#

vanilla or modded

north holly Jan 24, 2025, 4:15 PM

#

Hypixel

#

My computer is too shit for mods

#

And I don't want to make a new world

#

bruh 1.13 still doesn't work on debian sid

#

I'll just send a crash report

fiery turtle Jan 24, 2025, 4:17 PM

#

ah

north holly Jan 24, 2025, 4:17 PM

#

hopefully they fix that (they won't)

#

but anyway, I guess I'll play hypixel in the meantime

north holly Jan 24, 2025, 4:18 PM

#

north holly hopefully they fix that (they won't)

last time I looked at it, some library was failing

left orbit Jan 24, 2025, 4:20 PM

#

north holly bruh 1.13 still doesn't work on debian sid

how

#

it's java

#

how can it not work on debian

median crest Jan 24, 2025, 4:21 PM

#

north holly My computer is too shit for mods

When I was a kid I played mc with mods on a celeron with intel hd graphics 3000, you just gotta optifine it hard enough

#

And use like 2 render distance

north holly Jan 24, 2025, 4:22 PM

#

nvm I'm not playing hypixel anymore

north holly Jan 24, 2025, 4:26 PM

#

left orbit how can it not work on debian

yeah idk

#

library initialization failed - unable to allocate file descriptor table - out of memoryAborted (core dumped)

#

hm

#

huh changing the command to use java instead of whatever the mc launcher packages

#

gets it to work

#

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

Tenor

calm latch Jan 24, 2025, 4:41 PM

#

fiery turtle <@350748255929499649> btw can u `git fetch && git reset --hard origin/address-sp...

I'll be home in 2 hours, but I can explain how to build it ultrameme

#

(I think I need a better build system, but whatever)

calm latch Jan 24, 2025, 4:45 PM

#

fiery turtle and e.g. latest pmos did make it past the spot where obos hung

I can't test now, but it was also not booting on my PCs

#

Though it might have been unrelated

flat badge Jan 24, 2025, 4:57 PM

#

calm latch (I think I need a better build system, but whatever)

shameless xbstrap suggestion

calm latch Jan 24, 2025, 4:58 PM

#

I mean it's just make all to get an iso

#

After you build clang trl

north holly Jan 24, 2025, 5:01 PM

#

minecraft break done

north holly Jan 24, 2025, 5:18 PM

#

bro why would an outb hang

#

nooo

#

uacpi_status uacpi_kernel_io_write8(
    uacpi_handle hnd, uacpi_size offset, uacpi_u8 in_value
)
{
    uintptr_t port = (uintptr_t)hnd;
#if defined(__x86_64__) || defined(__i386__)
    outb(port+offset, in_value);
#endif
    return UACPI_STATUS_OK;
}```

frank canopy Jan 24, 2025, 5:24 PM

#

did you put a printf after the outb?

#

and is it just triggering an interrupt the instant the outb goes out and then thats hanging?

fiery turtle Jan 24, 2025, 5:28 PM

#

That was my assumption

fiery turtle Jan 24, 2025, 5:28 PM

#

north holly ```c uacpi_status uacpi_kernel_io_write8( uacpi_handle hnd, uacpi_size offse...

Try no-oping uacpi irq installation

north holly Jan 24, 2025, 5:29 PM

#

UACPI_STATUS_OK

#

my IOAPIC code has struck again

#

this is the second time this has happened

#

on that same computer

fiery turtle Jan 24, 2025, 5:35 PM

#

Wdym

#

Like doesnt hang anymore?

#

I guess it was hanging because your polarity was bogus

#

And u did get an irq right after the outb

north holly Jan 24, 2025, 5:59 PM

#

fiery turtle Like doesnt hang anymore?

No

#

It doesn't hang without irqs

fiery turtle Jan 24, 2025, 6:00 PM

#

fix your goddamn gsi routing code

north holly Jan 24, 2025, 6:00 PM

#

north holly It doesn't hang without irqs

Being registered

north holly Jan 24, 2025, 6:00 PM

#

fiery turtle fix your goddamn gsi routing code

Yes

#

In a few hours tho

#

Gtg now

fiery turtle Jan 24, 2025, 6:02 PM

#

sure

north holly Jan 24, 2025, 6:02 PM

#

Weird that it worked in a very old obos commit, even though I hadn't changed any ioapic code

#

Since then

fiery turtle Jan 24, 2025, 6:03 PM

#

u probably did

#

by accident maybe

#

or maybe ur corrupting your redirect table

north holly Jan 24, 2025, 6:11 PM

#

fiery turtle u probably did

git diff never lies

fiery turtle Jan 24, 2025, 6:11 PM

#

u mean git log

north holly Jan 24, 2025, 6:11 PM

#

I used git diff

#

To compare whatever the commit that I tried was

#

To master

#

And the ioapic code wasn't changed except for one small detail which I am sure doesn't matter

fiery turtle Jan 24, 2025, 6:13 PM

#

inb4 memory corruption

frank canopy Jan 24, 2025, 6:23 PM

#

one small detail which i am sure doesn't matter
🤔

strong heath Jan 24, 2025, 6:36 PM

#

does uacpi support modern standby

fiery turtle Jan 24, 2025, 6:36 PM

#

its up to the os to support it

#

but uacpi has all the primitives to implement it

strong heath Jan 24, 2025, 6:38 PM

#

what does it entail

fiery turtle Jan 24, 2025, 6:38 PM

#

https://elixir.bootlin.com/linux/v6.12.6/source/drivers/acpi/x86/s2idle.c

s2idle.c - drivers/acpi/x86/s2idle.c - Linux source ...

Elixir Cross Referencer - source code of Linux v6.12.6: drivers/acpi/x86/s2idle.c

strong heath Jan 24, 2025, 6:39 PM

#

thanks ill give it a look

fiery turtle Jan 24, 2025, 6:40 PM

#

npnp

strong heath Jan 24, 2025, 7:07 PM

#

does uacpi need to be initialized to return tables with uacpi_table_find_by_signature

gentle peak Jan 24, 2025, 7:08 PM

#

you can use early table access

strong heath Jan 24, 2025, 7:09 PM

#

what do you mean by that

fiery turtle Jan 24, 2025, 7:09 PM

#

strong heath does uacpi need to be initialized to return tables with `uacpi_table_find_by_sig...

uacpi_setup_early_table_access

strong heath Jan 24, 2025, 7:11 PM

#

oh thats neat

#

thanks

fiery turtle Jan 24, 2025, 7:11 PM

#

np!

north holly Jan 24, 2025, 7:19 PM

#

frank canopy > one small detail which i am sure doesn't matter 🤔

It's basically just switching the order a 64 bit register gets written in

#

From low first then high to high first then low

frank canopy Jan 24, 2025, 7:20 PM

#

ok yeah that shouldnt be an issue lol

strong heath Jan 24, 2025, 7:20 PM

#

(0000000000) successfully loaded 1 AML blob, 1712 ops in 10ms (avg 171200/s)

fiery turtle Jan 24, 2025, 7:20 PM

#

q35 + kvm?

strong heath Jan 24, 2025, 7:21 PM

#

for the sake of my ego ill just not say

fiery turtle Jan 24, 2025, 7:21 PM

#

btw do u want me to add ironclad to the mentioned projects list

strong heath Jan 24, 2025, 7:21 PM

#

ill tell you when to do it since its not upstreamed yet

#

if you care about that

fiery turtle Jan 24, 2025, 7:21 PM

#

sure

#

yeah

calm latch Jan 24, 2025, 7:26 PM

#

It gets power button notification, but doesn't shut down untill I press the keyboard button for some reason

fiery turtle Jan 24, 2025, 7:26 PM

#

did u update to latest pr head

#

wait what does that even mean

calm latch Jan 24, 2025, 7:26 PM

#

I have very fast internet

#

fiery turtle Jan 24, 2025, 7:27 PM

#

no like

calm latch Jan 24, 2025, 7:27 PM

#

It sent messages in wrong order

fiery turtle Jan 24, 2025, 7:27 PM

#

u get the notification

calm latch Jan 24, 2025, 7:27 PM

#

fiery turtle <@350748255929499649> btw can u `git fetch && git reset --hard origin/address-sp...

📎 pmOS.iso

fiery turtle Jan 24, 2025, 7:27 PM

#

not when u press the power button?

calm latch Jan 24, 2025, 7:27 PM

#

*keyboard button

#

not power button

#

maybe it's qemu thing

fiery turtle Jan 24, 2025, 7:27 PM

#

i dont think its a qemu thing

calm latch Jan 24, 2025, 7:27 PM

#

I press power button, it gets notification and doesn't do anything

#

untill I press enter

#

weird

fiery turtle Jan 24, 2025, 7:28 PM

#

pmos regression?

#

what if u revert to old uacpi btw

calm latch Jan 24, 2025, 7:28 PM

#

I've removed the old api

#

wait a sec

fiery turtle Jan 24, 2025, 7:28 PM

#

i doubt it matters but

#

because it sounds like it hangs in userspace

#

until u press a key

calm latch Jan 24, 2025, 7:28 PM

#

which old

#

maybe my interrupts delivery is broken

fiery turtle Jan 24, 2025, 7:29 PM

#

master

calm latch Jan 24, 2025, 7:29 PM

#

or scheduler idk

fiery turtle Jan 24, 2025, 7:29 PM

#

but u get the notif

calm latch Jan 24, 2025, 7:29 PM

#

Maybe broken timers

#

Idk, it does sleep(3);

#

Lemmy check master

#

I added 3 new syscalls and changed my IPC stuff

#

maybe I broke something

#

in the process

fiery turtle Jan 24, 2025, 7:30 PM

#

That would explain it

calm latch Jan 24, 2025, 7:30 PM

#

(more messages to compete with Nyaux thread)

calm latch Jan 24, 2025, 7:31 PM

#

calm latch wait a sec

(for me to fix this)

#

bruh wtf is going on with my wifi

#

I can't even push a commit

fiery turtle Jan 24, 2025, 7:31 PM

#

You can have both so its easier to test

#

These APIs dont conflict with each other

calm latch Jan 24, 2025, 7:32 PM

#

I know

#

ugh

#

Oh I have a very old API

#

It's the same on master

#

It's weird

fiery turtle Jan 24, 2025, 7:37 PM

#

Remove the star

calm latch Jan 24, 2025, 7:38 PM

#

no, it's the old thing

#

I've fixed it

#

Idk what's going on

#

I'm blaming qemu

fiery turtle Jan 24, 2025, 7:38 PM

#

Doubt

#

Try bisecting your IPC changes

calm latch Jan 24, 2025, 7:39 PM

#

I think my timers code is broken

#

If I open QEMU monitor it does it as it should

#

sleep() that is

fiery turtle Jan 24, 2025, 7:39 PM

#

Yeah potentially

calm latch Jan 24, 2025, 7:40 PM

#

calm latch

@north holly ?

north holly Jan 24, 2025, 7:40 PM

#

At the mosque

#

Can't test

fiery turtle Jan 24, 2025, 7:40 PM

#

But why is pressing a button needed

calm latch Jan 24, 2025, 7:42 PM

#

It gets keyboard interrupt, which reschedules and restarts lapic timer, probably

#

https://gitlab.com/mishakov/pmos/-/tree/dev?ref_type=heads is on the new branch

GitLab

Files · dev · Mikhail Kovalev / pmOS · GitLab

GitLab.com

#

I don't remember, but I think I might be disabling the timer in idle thread Thonk

#

(I've also changed the timers' code)

calm latch Jan 24, 2025, 7:47 PM

#

calm latch It gets keyboard interrupt, which reschedules and restarts lapic timer, probably

Too defensive programming KEKW

fiery turtle Jan 24, 2025, 8:59 PM

#

https://en.wikipedia.org/wiki/Hamming_weight#Efficient_implementation

Hamming weight

The Hamming weight of a string is the number of symbols that are different from the zero-symbol of the alphabet used. It is thus equivalent to the Hamming distance from the all-zero string of the same length. For the most typical case, a string of bits, this is the number of 1's in the string, or the digit sum of the binary representation of a g...

#

i think im going to do this

#

instead of relying on __builtin_popcnt

strong heath Jan 24, 2025, 9:02 PM

#

What advantages does it bring

fiery turtle Jan 24, 2025, 9:02 PM

#

not relying on libgcc or x86_64-v1

#

// https://en.wikipedia.org/wiki/Hamming_weight#Efficient_implementation
uacpi_u8 uacpi_popcount(uacpi_u64 x)
{
    x -= (x >> 1) & 0x5555555555555555ull;
    x = (x & 0x3333333333333333ull) + ((x >> 2) & 0x3333333333333333);
    x = (x + (x >> 4)) & 0x0F0F0F0F0F0F0F0Full;

    return (x * 0x0101010101010101) >> 56;
}

#

this should work

strong heath Jan 24, 2025, 9:02 PM

#

fiery turtle not relying on libgcc or x86_64-v1

That would be very neat for me

#

Please do

fiery turtle Jan 24, 2025, 9:03 PM

#

yeah

#

everyone compiles for x86_64 base so it produces an undefined reference which is annoying

#

wait

#

this is used in like one place to count set bits of a 16 bit number

#

i can just get rid of this

#

for some reason i was under the impression that aml had a popcount instruction

frank canopy Jan 24, 2025, 9:12 PM

#

nice to get rid of things

fiery turtle Jan 24, 2025, 9:15 PM

#

yeah ill just do this ad-hoc loop, which is used on a very cold path anyway

frank canopy Jan 24, 2025, 9:24 PM

#

is has_builtin false if you compile without the instruction, for however popcout was implemented?

#

if so id use that just in case

fiery turtle Jan 24, 2025, 9:25 PM

#

the has_builtin is always true

#

because it falls back to the libgcc implementation

frank canopy Jan 24, 2025, 9:26 PM

#

ah ok

#

yeah if its a cold path then probably not worth it then

fiery turtle Jan 24, 2025, 9:26 PM

#

yeah this is resource conversion

#

so very very cold path

#

@north holly @calm latch I just made a few last minute optimizations (for GPEs) and popcount removal, so if you can pls retest

#

EC stuff might've regressed

north holly Jan 24, 2025, 9:40 PM

#

uh soon

#

in 5-6 hours probably

fiery turtle Jan 24, 2025, 9:42 PM

#

sure sure

#

im doing the final review pass and fixing small things atm

fiery turtle Jan 24, 2025, 9:59 PM

#

alright, I think that's it, if oberrow and mishakov dont find new regressions im merging it

north holly Jan 24, 2025, 10:00 PM

#

letsgo

fiery turtle Jan 24, 2025, 10:00 PM

#

and monkuous (if they're available to test)

north holly Jan 24, 2025, 10:00 PM

#

uacpi 1.0 tomorrow then 🥺

fiery turtle Jan 24, 2025, 10:00 PM

#

yup

#

as soon as i wake up

north holly Jan 24, 2025, 10:00 PM

#

letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo letsgo

#

letsgo letsgo letsgo letsgo letsgo

#

that's enough before banter mutes me

fiery turtle Jan 24, 2025, 10:00 PM

#

lol

north holly Jan 24, 2025, 10:01 PM

#

@stone reef I'm going to spam and there is nothing you can do about it!!21111!!!

fiery turtle Jan 24, 2025, 10:01 PM

#

this PR is one month old now

#

very good time to merge

pine leaf Jan 24, 2025, 10:05 PM

#

if there's some builds that can be tested on hw, I'll try them on a few different machines of mine

fiery turtle Jan 24, 2025, 10:05 PM

#

sure that would be cool

pine leaf Jan 24, 2025, 10:05 PM

#

I have desktops and laptops from a very wide range of eras and manufacturers

fiery turtle Jan 24, 2025, 10:05 PM

#

how old are these machines

pine leaf Jan 24, 2025, 10:06 PM

#

oldest is a Core 2 Duo from 2007
newest 2 are an AM4 board from ASRock and an LGA1700 board from ASUS

#

and several in between

fiery turtle Jan 24, 2025, 10:06 PM

#

core 2 duo could have pretty insane aml so thats nice

#

pre-vista aml was insane

pine leaf Jan 24, 2025, 10:09 PM

#

I can test on all of these

Dell Optiplex 755 - C2D E8400
Dell Precision T1650 - Xeon E3-1240v2
HP ProBook 4540s - i5-3210M (sadly has known but unfixed graphics issues in Limine - iirc only when booting via legacy BIOS?)
HP EliteDesk something or other - some 4th gen i7 (I haven't touched this one in a while, I forget what exactly it is)
Dell Latitude E5250 - i5-5200U
ThinkPad T14 Gen 1 AMD - Ryzen 7 PRO 4750U
Beelink SER5 - Ryzen 5 5500U
ASUS PRIME H610M-E D4 - i3-12100f
ASRock B450 Steel Legend - Ryzen 5 3500X

fiery turtle Jan 24, 2025, 10:09 PM

#

damn

#

obos uses hyper so

#

limine problems might not affect it

pine leaf Jan 24, 2025, 10:10 PM

#

well that's only on that one machine, and it's purely a graphical issue

#

it actually executes fine, but the fb pitch is wrong or some nonsense

fiery turtle Jan 24, 2025, 10:10 PM

#

but yeah would be great ofc

#

for obos, pmos and proxima

calm latch Jan 24, 2025, 10:11 PM

#

pine leaf oldest is a Core 2 Duo from 2007 newest 2 are an AM4 board from ASRock and an LG...

This should be fun

#

I can also make a build with 32 bit userspace (and uACPI)

fiery turtle Jan 24, 2025, 10:12 PM

#

pls do

#

from the new PR head tho

calm latch Jan 24, 2025, 10:13 PM

#

pine leaf I can test on all of these - Dell Optiplex 755 - C2D E8400 - Dell Precision T165...

pmOS doesn't need framebuffer KEKW

fiery turtle Jan 24, 2025, 10:13 PM

#

core2duo probably supports 64-bit stuf tho

calm latch Jan 24, 2025, 10:13 PM

#

(it doesn't work half the times)

calm latch Jan 24, 2025, 10:13 PM

#

fiery turtle core2duo probably supports 64-bit stuf tho

Yeah it's an ok 64 bit

#

There's stuff like athlon 64

#

which is like early 64 bit

calm latch Jan 24, 2025, 10:14 PM

#

pine leaf I can test on all of these - Dell Optiplex 755 - C2D E8400 - Dell Precision T165...

I had B450M steel legend lol

#

anyway

#

time to recompile llvm

#

git fetch && git reset --hard origin/address-spaces-support?

fiery turtle Jan 24, 2025, 10:15 PM

#

yup

#

should see thisstdlib: get rid of uacpi_popcount

calm latch Jan 24, 2025, 10:25 PM

#

/home/misha/llvm-pmos/libcxx/src/memory.cpp:102:43: error: size of array element of type '__libcpp_mutex_t' (aka 'pthread_mutex_t') (32 bytes) isn't a multiple of its alignment (64 bytes) bruh

fiery turtle Jan 24, 2025, 10:28 PM

#

what the fuck

calm latch Jan 24, 2025, 10:29 PM

#

Idk

#

llvm insanity

#

It's defined as ```C
typedef struct {
__pmos64u_t block_count;
__pmos64u_t blocking_thread_id;
struct __pthread_waiter *waiters_list_head;
struct __pthread_waiter *waiters_list_tail;
unsigned long recursive_lock_count;
int type;
#ifdef i386
} pthread_mutex_t attribute((aligned(64)));
#else
} pthread_mutex_t;

north holly Jan 24, 2025, 10:34 PM

#

@fiery turtle this new commit caused obos to shutdown on suspend

#

Either that or I forgot to do something

fiery turtle Jan 24, 2025, 10:34 PM

#

which commit lol

#

u did i think

north holly Jan 24, 2025, 10:34 PM

#

Ah nvm I forgot something

fiery turtle Jan 24, 2025, 10:34 PM

#

but that would be a funny bug

north holly Jan 24, 2025, 10:35 PM

#

I forgot to put the wake gpe registering code back

left orbit Jan 24, 2025, 10:35 PM

#

lmao

north holly Jan 24, 2025, 10:35 PM

#

Silly me

calm latch Jan 24, 2025, 10:36 PM

#

clang's -E is refusing to work bruh

north holly Jan 24, 2025, 10:37 PM

#

Ah nvm it still broke

#

@fiery turtle

fiery turtle Jan 24, 2025, 10:37 PM

#

what

north holly Jan 24, 2025, 10:37 PM

#

The HEAD of the PR broke suspend

fiery turtle Jan 24, 2025, 10:37 PM

#

can u please triple check

north holly Jan 24, 2025, 10:38 PM

#

Yes

fiery turtle Jan 24, 2025, 10:38 PM

#

that its not an obos bug

north holly Jan 24, 2025, 10:38 PM

#

I will be going back in time to make sure

calm latch Jan 24, 2025, 10:39 PM

#

I am going insane ```C

230 "/home/misha/pmos/sysroot/usr/include/__posix_types.h" 3

typedef struct {
__pmos64u_t block_count;
__pmos64u_t blocking_thread_id;
struct __pthread_waiter *waiters_list_head;
struct __pthread_waiter *waiters_list_tail;
unsigned long recursive_lock_count;
int type;

} pthread_mutex_t attribute((aligned(64)));

typedef pthread_mutex_t __libcpp_mutex_t;

#

what the fuck

#

cpp output by clang

#

I need clang wizzards

dense steppe Jan 24, 2025, 10:41 PM

#

calm latch I am going insane ```C # 230 "/home/misha/pmos/sysroot/usr/include/__posix_types...

no futex? trl

slim panther Jan 24, 2025, 10:41 PM

#

hello

dense steppe Jan 24, 2025, 10:41 PM

#

also what is strange about this CPP output?

calm latch Jan 24, 2025, 10:41 PM

#

dense steppe no `futex`? <:trl:829316641191886888>

My mutex is done with IPC bruh

#

Which needs to go because of priority inversion issues, and besides the point

calm latch Jan 24, 2025, 10:42 PM

#

calm latch `/home/misha/llvm-pmos/libcxx/src/memory.cpp:102:43: error: size of array elemen...

This error

#

📎 memory.cpp.e

dense steppe Jan 24, 2025, 10:43 PM

#

are pointers 64-bit?

#

sanity question

calm latch Jan 24, 2025, 10:43 PM

#

32 bit

dense steppe Jan 24, 2025, 10:43 PM

#

ahhh

calm latch Jan 24, 2025, 10:43 PM

#

64 bytes Thonk

#

not 64 bits

#

I'm dumb...

dense steppe Jan 24, 2025, 10:44 PM

#

Ahh

#

you meant to do _Alignas (__pmos64u_t)?

calm latch Jan 24, 2025, 10:44 PM

#

yeah

dense steppe Jan 24, 2025, 10:45 PM

#

easy to get wrong :p

fiery turtle Jan 24, 2025, 10:50 PM

#

@north holly did u figure it out?

north holly Jan 24, 2025, 10:51 PM

#

I'm trying

fiery turtle Jan 24, 2025, 10:51 PM

#

is it an obos problem?

north holly Jan 24, 2025, 10:51 PM

#

Idk at this point

fiery turtle Jan 24, 2025, 10:51 PM

#

what happens exactly

north holly Jan 24, 2025, 10:51 PM

#

The signs do not point anywhere

calm latch Jan 24, 2025, 10:51 PM

#

Are there operating systems with i686 uACPI?

fiery turtle Jan 24, 2025, 10:52 PM

#

not besides reactos

calm latch Jan 24, 2025, 10:52 PM

#

fuck ld.lld: error: unable to find library -lclang_rt.builtins-i386

#

not this shit again

dense steppe Jan 24, 2025, 10:52 PM

#

calm latch Are there operating systems with i686 uACPI?

that's why you're doing 32-bit stuff?

calm latch Jan 24, 2025, 10:53 PM

#

I'm trying to build 32 bit userspace

fiery turtle Jan 24, 2025, 10:53 PM

#

can you make a normal 64-bit pmos build as well?

calm latch Jan 24, 2025, 10:53 PM

#

Yeah

#

It shouldn't matter

north holly Jan 24, 2025, 10:53 PM

#

Ok I think it works

fiery turtle Jan 24, 2025, 10:53 PM

#

north holly Ok I think it works

what was the issue

north holly Jan 24, 2025, 10:54 PM

#

Idk

#

I'm on an old commit for uacpi

fiery turtle Jan 24, 2025, 10:54 PM

#

which

north holly Jan 24, 2025, 10:54 PM

#

I forgot

#

Wait

fiery turtle Jan 24, 2025, 10:54 PM

#

can u do the pr head

north holly Jan 24, 2025, 10:54 PM

#

I'm on my phone

north holly Jan 24, 2025, 10:54 PM

#

fiery turtle can u do the pr head

Yes

calm latch Jan 24, 2025, 10:55 PM

#

-- Compiler-RT supported architectures: nooo

#

Why is it so difficult to compile compiler-rt?

#

clang makes me want to go back to gcc

#

which is not the vibe

north holly Jan 24, 2025, 10:58 PM

#

wat da shit

fiery turtle Jan 24, 2025, 10:58 PM

#

wtf is going on

north holly Jan 24, 2025, 10:59 PM

#

Page fault

fiery turtle Jan 24, 2025, 10:59 PM

#

in the allocator?

north holly Jan 24, 2025, 10:59 PM

#

Wait I think the IRQL is too high to handle page faults

#

Yup

calm latch Jan 24, 2025, 11:00 PM

#

north holly Wait I think the IRQL is too high to handle page faults

what

north holly Jan 24, 2025, 11:00 PM

#

I put a restriction on what IRQL you can page fault on

#

Because like swap n'shit

#

That might need to block

fiery turtle Jan 24, 2025, 11:01 PM

#

why do u get a pf there

north holly Jan 24, 2025, 11:03 PM

#

Demand paging

#

Using the non paged allocator should be the solution

fiery turtle Jan 24, 2025, 11:03 PM

#

how did this work previously

#

did u never get ec events

north holly Jan 24, 2025, 11:04 PM

#

I didn't get an irq during that ig

north holly Jan 24, 2025, 11:04 PM

#

fiery turtle did u never get ec events

Yeah probably

#

It's a timing issue

#

Or well

#

The irq is at the wrong time

fiery turtle Jan 24, 2025, 11:04 PM

#

maybe its a regression and u arent supposed to get an irq at that time?

#

whats even the event

north holly Jan 24, 2025, 11:04 PM

#

Beats me

#

Ok it works

fiery turtle Jan 24, 2025, 11:07 PM

#

do u get ec events?

north holly Jan 24, 2025, 11:07 PM

#

Whatever happened to the light on suspend

north holly Jan 24, 2025, 11:07 PM

#

fiery turtle do u get ec events?

I will check

fiery turtle Jan 24, 2025, 11:07 PM

#

north holly Whatever happened to the light on suspend

hm?

#

how is literally everything breaking LULW

north holly Jan 24, 2025, 11:08 PM

#

There is usually a light that's turned on at suspend

#

It doesn't turn on

dense steppe Jan 24, 2025, 11:08 PM

#

at this rate there will be davix suspend before obos suspend works trl

north holly Jan 24, 2025, 11:08 PM

#

But it's still in suspend

fiery turtle Jan 24, 2025, 11:08 PM

#

did it wake up at least

north holly Jan 24, 2025, 11:08 PM

#

dense steppe at this rate there will be davix suspend before obos suspend works <:trl:8293166...

Obos suspend has worked for a long time

#

Yes

fiery turtle Jan 24, 2025, 11:09 PM

#

try ec events and see if that light not turning on repros

dense steppe Jan 24, 2025, 11:09 PM

#

north holly Obos suspend has worked for a long time

(on the 1.0 PR. also I'm just trolling. I have no plans on doing suspend for a while)

north holly Jan 24, 2025, 11:09 PM

#

fiery turtle try ec events and see if that light not turning on repros

I don't think it worked on an older commit

#

So it's fine

fiery turtle Jan 24, 2025, 11:09 PM

#

north holly I don't think it worked on an older commit

the light?

north holly Jan 24, 2025, 11:09 PM

#

Yes

fiery turtle Jan 24, 2025, 11:09 PM

#

ah

#

im curious about ec as well

north holly Jan 24, 2025, 11:10 PM

#

kk

fiery turtle Jan 24, 2025, 11:11 PM

#

and also the ioapic fix so we can test on other pc

north holly Jan 24, 2025, 11:11 PM

#

EC events worked

north holly Jan 24, 2025, 11:11 PM

#

fiery turtle and also the ioapic fix so we can test on other pc

At my friend's house using my laptop

fiery turtle Jan 24, 2025, 11:11 PM

#

ah

#

well once you're back

north holly Jan 24, 2025, 11:11 PM

#

My laptop's EC events work

fiery turtle Jan 24, 2025, 11:11 PM

#

like lid and stuff?

north holly Jan 24, 2025, 11:11 PM

#

Yes

#

fiery turtle Jan 24, 2025, 11:11 PM

#

ok thats nice at least

#

charger too?

north holly Jan 24, 2025, 11:12 PM

#

Yes

fiery turtle Jan 24, 2025, 11:13 PM

#

letsgo

#

im assuming u shutdown after wake so that worked as well?

north holly Jan 24, 2025, 11:14 PM

#

Yes

#

Only thing untested is reboot

#

But whatever y'know

fiery turtle Jan 24, 2025, 11:15 PM

#

ok cool at least this laptop didnt regress

north holly Jan 24, 2025, 11:15 PM

#

Ye

hollow elm Jan 24, 2025, 11:15 PM

#

fiery turtle the light?

do not walk towards it

fiery turtle Jan 24, 2025, 11:15 PM

#

north holly Ye

but it worked with the broken pr as well earlier today right

fiery turtle Jan 24, 2025, 11:15 PM

#

hollow elm do not walk towards it

lol

north holly Jan 24, 2025, 11:20 PM

#

fiery turtle but it worked with the broken pr as well earlier today right

Yes

fiery turtle Jan 24, 2025, 11:20 PM

#

yeah this laptop doesnt care

calm latch Jan 24, 2025, 11:28 PM

#

I think I won

#

fun

north holly Jan 24, 2025, 11:32 PM

#

What even happened?

calm latch Jan 24, 2025, 11:32 PM

#

32 bit userspace broke kernel

north holly Jan 24, 2025, 11:33 PM

#

Damn

#

Anyway uacpi thread message count is above obos' thread

calm latch Jan 24, 2025, 11:34 PM

#

calm latch 32 bit userspace broke kernel

tldr is that i386 has too little registers, so I pass some of syscall arguments on stack

#

ok 32 bit userspace uACPI seems to be kinda working

#

I need to fix a pagefault

#

(fork is broken)

north holly Jan 24, 2025, 11:38 PM

#

hey mishakov I'm bored, can you throw at me some random thing I could implement in my kernel that only takes a few hours

#

@calm latch

calm latch Jan 24, 2025, 11:38 PM

#

idk

#

my estimations are always off

#

(is that the whole point?)

calm latch Jan 24, 2025, 11:39 PM

#

north holly hey mishakov I'm bored, can you throw at me some random thing I could implement ...

Do you support i386 userspace on 64 bit kernel? ultrameme

north holly Jan 24, 2025, 11:39 PM

#

no

#

(t yet)

#

ultrameme

#

but for that I'd need an i686-obos toolchain which I don't feel like doing

#

I recently added CMOS RTC support because I was bored

#

but no gettimeofday syscall to go with that

calm latch Jan 24, 2025, 11:45 PM

#

then do that

#

idk

north holly Jan 24, 2025, 11:46 PM

#

yeah

calm latch Jan 24, 2025, 11:53 PM

#

anyway 32 bit userspace pmos build

📎 pmOS.iso

#

@fiery turtle

#

@pine leaf ?

#

fork is broken but it shouldn't matter

pine leaf Jan 24, 2025, 11:54 PM

#

yeah gimme a bit
I can test it later today

calm latch Jan 24, 2025, 11:55 PM

#

btw what's your timezone?

pine leaf Jan 24, 2025, 11:55 PM

#

GMT-7

calm latch Jan 24, 2025, 11:55 PM

#

since ugh

#

Bruh it triple faults on my laptop

#

Which is something new

#

I think the last time I saw pmOS triple fault was a year ago...

#

@fiery turtle it only shuts down when I press enter in QEMU KEKW

#

Anyway, normal build

📎 pmOS.iso

calm latch Jan 25, 2025, 12:07 AM

#

calm latch <@353593948523986945> it only shuts down when I press enter in QEMU <:KEKW:91370...

could it be WSL2 shenanigans?

calm latch Jan 25, 2025, 12:09 AM

#

calm latch Anyway, normal build

This works on my laptop

#

EC + AMD GPIO

#

(like power button + lid)

#

And it shuts down in 3 seconds, as it should

frank canopy Jan 25, 2025, 12:30 AM

#

took until literally now but i think i finally have an understanding in my brain of how to make my mm work so i can go back to acpi and drivers and actually fun things

fiery turtle Jan 25, 2025, 9:32 AM

#

calm latch <@353593948523986945> it only shuts down when I press enter in QEMU <:KEKW:91370...

Like on real hw it works fine?

fiery turtle Jan 25, 2025, 9:33 AM

#

calm latch Anyway, normal build

Is this from master or pr?

fiery turtle Jan 25, 2025, 9:38 AM

#

frank canopy took until literally now but i think i finally have an understanding in my brain...

Letsgo

calm latch Jan 25, 2025, 10:59 AM

#

fiery turtle Like on real hw it works fine?

At least on my laptop

fiery turtle Jan 25, 2025, 11:06 AM

#

calm latch At least on my laptop

the pr head?

calm latch Jan 25, 2025, 11:07 AM

#

Lemmy double check

fiery turtle Jan 25, 2025, 11:08 AM

#

how did u manage to have it regress on qemu but not real hw

calm latch Jan 25, 2025, 11:10 AM

#

misha@Yoga:~/pmos/devicesd/uACPI$ git log
commit 199b625a347e8f91efad3ec6c7a2200903a24f01 (HEAD -> master, origin/address-spaces-support)
Author: Daniil Tatianin <[email protected]>
Date:   Sat Jan 25 00:17:39 2025 +0300

    stdlib: get rid of uacpi_popcount

#

Is this the right one?

fiery turtle Jan 25, 2025, 11:12 AM

#

yup

calm latch Jan 25, 2025, 11:13 AM

#

I think I've managed to catch the framebuffer bug/kernel panick in virtualbox...

fiery turtle Jan 25, 2025, 11:13 AM

#

calm latch ``` misha@Yoga:~/pmos/devicesd/uACPI$ git log commit 199b625a347e8f91efad3ec6c7a...

so this is the one u tested with and no regressions on your laptop found?

calm latch Jan 25, 2025, 11:13 AM

#

fiery turtle so this is the one u tested with and no regressions on your laptop found?

It seems so

calm latch Jan 25, 2025, 11:13 AM

#

calm latch Anyway, normal build

This is the build

fiery turtle Jan 25, 2025, 11:13 AM

#

ok nice

#

why does 32-bit build die tho

calm latch Jan 25, 2025, 11:23 AM

#

it works in qemu

#

probably gdt is messed up or something

#

[ OK ] Loded GDT

fiery turtle Jan 25, 2025, 11:24 AM

#

lol

#

anyway looks like the pr should be good to merge then

#

@gentle peak if you're here maybe a final proxima test, since I've updated the PR since last time

#

especially curious about that laptop with legacy GPEs

#

because i added pre-mapping for GPE registers

calm latch Jan 25, 2025, 11:31 AM

#

calm latch I think I've managed to catch the framebuffer bug/kernel panick in virtualbox...

bruh

fiery turtle Jan 25, 2025, 11:32 AM

#

whats the opcode?

gentle peak Jan 25, 2025, 11:33 AM

#

fiery turtle <@938223886439436320> if you're here maybe a final proxima test, since I've upda...

sure, give me a minute since it seems like i managed to break my build system

fiery turtle Jan 25, 2025, 11:33 AM

#

npnp

#

not in a hurry anyway

calm latch Jan 25, 2025, 11:35 AM

#

216001: c5 f8 77 vzeroupper

fiery turtle Jan 25, 2025, 11:35 AM

#

gentle peak sure, give me a minute since it seems like i managed to break my build system

btw why did you decide to split proxima into multiple repos

fiery turtle Jan 25, 2025, 11:35 AM

#

calm latch `216001: c5 f8 77 vzeroupper`

should work on 32 bit

gentle peak Jan 25, 2025, 11:36 AM

#

fiery turtle btw why did you decide to split proxima into multiple repos

i don't like when things that are logically different packages are in the same repo

#

kernel and libc shouldn't belong in the same repo imo

calm latch Jan 25, 2025, 11:36 AM

#

Yeah why is it broken on everything but my Zen 4 laptop...

fiery turtle Jan 25, 2025, 11:36 AM

#

i guess

strong heath Jan 25, 2025, 11:37 AM

#

I would like to thank infy for making me able to remove all copied libgcc code from Ironclad, not used for anything else but uacpi, with the latest commit in the dev branch

#

a toast for you infy

#

https://tenor.com/YHMP.gif

Tenor

fiery turtle Jan 25, 2025, 11:37 AM

#

lol thanks

fiery turtle Jan 25, 2025, 11:38 AM

#

strong heath I would like to thank infy for making me able to remove all copied libgcc code f...

was there some other code besides popcount?

gentle peak Jan 25, 2025, 11:38 AM

#

fiery turtle especially curious about that laptop with legacy GPEs

works on this laptop

fiery turtle Jan 25, 2025, 11:38 AM

#

gentle peak works on this laptop

nice, and u get the notify for that proprietary device too right?

gentle peak Jan 25, 2025, 11:38 AM

#

yeah

fiery turtle Jan 25, 2025, 11:38 AM

#

cool cool

#

thx for testing

strong heath Jan 25, 2025, 11:39 AM

#

fiery turtle was there some other code besides popcount?

no

fiery turtle Jan 25, 2025, 11:39 AM

#

i was worried for a sec

calm latch Jan 25, 2025, 11:39 AM

#

pmOS build without AVX-512 KEKW

📎 pmOS.iso

gentle peak Jan 25, 2025, 11:40 AM

#

interestingly, while i do still get the valz notify when pressing the power button, i don't get it when opening/closing the lid anymore

#

i don't install handlers for those events but i used to get them anyway

calm latch Jan 25, 2025, 11:40 AM

#

fiery turtle how did u manage to have it regress on qemu but not real hw

It breaks in the exact same way in virtualbox 💀

fiery turtle Jan 25, 2025, 11:40 AM

#

gentle peak i don't install handlers for those events but i used to get them anyway

u should be getting them if you used to

#

does it happen if you revert?

calm latch Jan 25, 2025, 11:41 AM

#

only shuts down when I press enter

fiery turtle Jan 25, 2025, 11:41 AM

#

(or maybe try rebooting as well)

calm latch Jan 25, 2025, 11:42 AM

#

it was working fine before

loud ice Jan 25, 2025, 11:42 AM

#

strong heath I would like to thank infy for making me able to remove all copied libgcc code f...

and this entire time, you could have just used clang which never calls the builtin

fiery turtle Jan 25, 2025, 11:42 AM

#

calm latch it was working fine before

before what?

calm latch Jan 25, 2025, 11:42 AM

#

idk on old versions

#

like several months old

fiery turtle Jan 25, 2025, 11:43 AM

#

of pmos?

strong heath Jan 25, 2025, 11:43 AM

#

loud ice and this entire time, you could have just used clang which never calls the built...

I dont really have a choice with Ada, regardless of how much I shill GCC

calm latch Jan 25, 2025, 11:43 AM

#

maybe I did something in kernel

strong heath Jan 25, 2025, 11:43 AM

#

since GNAT is GCC only for the time being

calm latch Jan 25, 2025, 11:43 AM

#

fiery turtle of pmos?

yeah

fiery turtle Jan 25, 2025, 11:43 AM

#

ah

loud ice Jan 25, 2025, 11:43 AM

#

strong heath I dont really have a choice with Ada, regardless of how much I shill GCC

you can compile uacpi with clang though

fiery turtle Jan 25, 2025, 11:44 AM

#

what can it do other than calling the builtin tho?

strong heath Jan 25, 2025, 11:44 AM

#

loud ice you can compile uacpi with clang though

that is a mess (to require 2 compilers at once and link them statically and hope nothing breaks) and a bit hard to line up with the current buildsystem

#

I rather just slap the C popcnt

#

which I dont have to do anymore!

gentle peak Jan 25, 2025, 11:44 AM

#

fiery turtle does it happen if you revert?

doesn't happen on master either, maybe i'm misremembering and it never happened

fiery turtle Jan 25, 2025, 11:45 AM

#

gentle peak doesn't happen on master either, maybe i'm misremembering and it never happened

bruh no last minute regressions for me to debug nooo

gentle peak Jan 25, 2025, 11:45 AM

#

interestingly though, 1.0-rc seems to be a minor performance regression (very minor, 7.2-7.4M -> 7.0-7.2M on kvm)

loud ice Jan 25, 2025, 11:45 AM

#

strong heath that is a mess (to require 2 compilers at once and link them statically and hope...

fair enough

fiery turtle Jan 25, 2025, 11:45 AM

#

lemme take a look at your dump if u have that at all

gentle peak Jan 25, 2025, 11:45 AM

#

it's on the acpidumps repo

fiery turtle Jan 25, 2025, 11:45 AM

#

yeah i have it

vast kestrel Jan 25, 2025, 11:45 AM

#

gentle peak interestingly though, 1.0-rc seems to be a minor performance regression (very mi...

Def the popcount :^)

loud ice Jan 25, 2025, 11:46 AM

#

loud ice fair enough

#[unsafe(no_mangle)]
extern "C" fn __popcountdi2(a: i64) -> i32 {
    a.count_ones() as i32
}
``` my shitty test os had it implemented like this lul

fiery turtle Jan 25, 2025, 11:46 AM

#

gentle peak interestingly though, 1.0-rc seems to be a minor performance regression (very mi...

thats probably because of extra logic to handle special fields

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

230 "/home/misha/pmos/sysroot/usr/include/__posix_types.h" 3