OBOS (not vibecoded) | Operating System Development | Page 28

real pecan Dec 20, 2024, 1:24 AM

#

iasl

flint idol Dec 20, 2024, 1:24 AM

#

thx

flint idol Dec 20, 2024, 1:24 AM

#

real pecan iasl

how do I get it to work with the acpidump output

#

like if I did acpidump > acpi.txt

#

how do I decompile acpi.txt

real pecan Dec 20, 2024, 1:25 AM

#

acpidump > dump.txt && acpixtract -a dump.txt && iasl *.dat && code dsdt.dsl

flint idol Dec 20, 2024, 1:25 AM

#

thanks

#

wait my laptop actually has "sane" aml for wake from suspend

#

well as sane as it gets

#

actually let me make sure

#

nvm

#

it delulu

real pecan Dec 20, 2024, 1:29 AM

#

lmao

#

whats wrong

flint idol Dec 20, 2024, 1:29 AM

#

it tells me to put devices into D2 for suspend if I want them to wake me
D2 doesn't exist

real pecan Dec 20, 2024, 1:30 AM

#

lmao

flint idol Dec 20, 2024, 1:30 AM

#

same thing on my dev pc

real pecan Dec 20, 2024, 1:31 AM

#

perhaps ur misunderstanding what it does

#

have u looked at how linux handles these values

flint idol Dec 20, 2024, 1:31 AM

#

set _S3D to 2 to tell the kernel that this device can only be at a deepest state of 2 to wake the kernel up
no D2 state

flint idol Dec 20, 2024, 1:31 AM

#

real pecan have u looked at how linux handles these values

linux was my reference

#

iirc

real pecan Dec 20, 2024, 1:31 AM

#

how does it react in this case

flint idol Dec 20, 2024, 1:32 AM

#

well it just hangs after a bit

#

idk why

real pecan Dec 20, 2024, 1:32 AM

#

i mean linux

flint idol Dec 20, 2024, 1:32 AM

#

while setting the wake devices

flint idol Dec 20, 2024, 1:32 AM

#

real pecan how does it react in this case

let me see

real pecan Dec 20, 2024, 1:32 AM

#

u can add printfs there like mishakov did to figure out where the power button press comes from

flint idol Dec 20, 2024, 1:32 AM

#

suspends

#

also wakes up

real pecan Dec 20, 2024, 1:32 AM

#

to see which state it puts the devices in and why

flint idol Dec 20, 2024, 1:32 AM

#

real pecan u can add printfs there like mishakov did to figure out where the power button p...

I'm sorry but I don't hate myself

#

(reading linux code has traumatized me)

real pecan Dec 20, 2024, 1:32 AM

#

its not that hard to compile linux lol

#

also code is very easy to read

flint idol Dec 20, 2024, 1:33 AM

#

real pecan its not that hard to compile linux lol

it takes time
time is something I don't have much of rn

#

anyway, what obos does in this case is not put it in any D state

real pecan Dec 20, 2024, 1:34 AM

#

what if u force put it in d3

#

does it have that

flint idol Dec 20, 2024, 1:34 AM

#

let me check

#

nope

real pecan Dec 20, 2024, 1:35 AM

#

Which does it have

flint idol Dec 20, 2024, 1:35 AM

#

none

#

no D state objects

#

no wake power resources either

real pecan Dec 20, 2024, 1:35 AM

#

So then it doesnt matter what it says in _S right

flint idol Dec 20, 2024, 1:36 AM

#

yeah

real pecan Dec 20, 2024, 1:36 AM

#

So what happens on suspend

flint idol Dec 20, 2024, 1:37 AM

#

I think because _S3D is 0x2, which means the deepest state the device can be to wake us is D2, it's fine if I keep it in D0

flint idol Dec 20, 2024, 1:37 AM

#

real pecan So what happens on suspend

it doesn't, it hangs while registering wake gpes

#

before suspend

real pecan Dec 20, 2024, 1:37 AM

#

Did u figure out where

flint idol Dec 20, 2024, 1:37 AM

#

flint idol it doesn't, it hangs while registering wake gpes

*configuring wake devices

flint idol Dec 20, 2024, 1:37 AM

#

real pecan Did u figure out where

what I'm tryna do

real pecan Dec 20, 2024, 1:37 AM

#

flint idol I think because _S3D is 0x2, which means the deepest state the device can be to ...

Yeah

flint idol Dec 20, 2024, 1:38 AM

#

just an acpi spec skill issue

flint idol Dec 20, 2024, 1:20 PM

#

my obos clone on my laptop

#

is 170 commits behind

#

and one commit ahead

#

on commitf989a176ebaaf320ab3e81929508df331922ca83

#

on the repo

flint idol Dec 20, 2024, 1:21 PM

#

flint idol and one commit ahead

of master

#

also why is cmake+make so slow compared to cmake+ninja

thick jolt Dec 20, 2024, 1:26 PM

#

ninja

haughty abyss Dec 20, 2024, 1:31 PM

#

flint idol also why is cmake+make so slow compared to cmake+ninja

i think its because cmake's make target is really bad

#

but also make is just slower

thick jolt Dec 20, 2024, 4:39 PM

#

haughty abyss i think its because cmake's make target is really bad

https://tenor.com/view/mods-give-him-a-low-taper-fade-meme-low-taper-fade-low-tier-god-gif-5578247064056451203

Tenor

torpid wigeon Dec 20, 2024, 9:26 PM

#

flint idol also why is cmake+make so slow compared to cmake+ninja

don't use make

flint idol Dec 21, 2024, 1:53 PM

#

time for more debugging

#

wake up
shit
get out of bed
do osdev
sleep
repeat

weak kestrel Dec 21, 2024, 2:17 PM

#

flint idol >wake up >shit >get out of bed >do osdev >sleep >repeat

you shite in bed?

flint idol Dec 21, 2024, 2:17 PM

#

typo

#

I meant to put shit after get out of bed

#

wake up
get out of bed
shit
do osdev
sleep
repeat

weak kestrel Dec 21, 2024, 2:17 PM

#

https://tenor.com/view/sus-cat-cat-stare-stare-bruh-cat-really-n-gif-25597234

Tenor

flint idol Dec 21, 2024, 2:35 PM

#

/home/oberrow/Code/obos/src/oboskrnl/power/device.c:244:12: warning: 'useDSW' may be used uninitialized [-Wmaybe-uninitialized]
  244 |         if (useDSW)
      |            ^
/home/oberrow/Code/obos/src/oboskrnl/power/device.c:198:10: note: 'useDSW' was declared here
  198 |     bool useDSW = false;```
we just love this warning

thick jolt Dec 21, 2024, 2:37 PM

#

flint idol ``` /home/oberrow/Code/obos/src/oboskrnl/power/device.c:244:12: warning: 'useDSW...

me after turning on all warnings

flint idol Dec 21, 2024, 2:37 PM

#

(that warning is not only a false-positive, a stupid one)

#

((the line it shows me shows the variable being initialized))

thick jolt Dec 21, 2024, 2:38 PM

#

truly a

#

gcc skill issue

#

richard stallman

#

😡

flint idol Dec 21, 2024, 2:38 PM

#

soon I port obos to clang

thick jolt Dec 21, 2024, 2:38 PM

#

llvm glazer alert

#

😡 😡

#

all should write their own compiler

#

trl

flint idol Dec 21, 2024, 2:39 PM

#

u confoosing

thick jolt Dec 21, 2024, 2:39 PM

#

listen

#

just write ur own tooling

#

gcc? clang? msvc?

#

no

flint idol Dec 21, 2024, 2:39 PM

#

you do it first

thick jolt Dec 21, 2024, 2:39 PM

#

nyauxcompiler

#

trl

#

one day maybe

flint idol Dec 21, 2024, 2:39 PM

#

no thanks

thick jolt Dec 21, 2024, 2:39 PM

#

i might give it a go

flint idol Dec 21, 2024, 2:39 PM

#

I don't want my compiler to page fault when I throw at it a hello world program

thick jolt Dec 21, 2024, 2:40 PM

#

flint idol no thanks

yes nyauxcompiler and it gives u even more UB!!!!

thick jolt Dec 21, 2024, 2:40 PM

#

flint idol I don't want my compiler to page fault when I throw at it a hello world program

😭

#

halfmemeright

#

itll do better and somehow kill ur entire kernel

#

and brick ur system

#

trl

flint idol Dec 21, 2024, 2:41 PM

#

#include <cstdio>
#include <cstring>
#include <type_traits>

int main() {
}

struct A {
    static void (*&_)(int);
    A(float b) { int i; _(*(int*)memcpy(&i, &b, 4)); }
};
std::remove_cvref_t<decltype(A::_)> **O, *P, Q;
decltype(A::_) A::_ = (O = &P, P = &Q, **O);

template<float B, int = 0>
struct _ {
    static const inline A a = (Q = [](int _) { putchar(_); }, A(B));
};

template struct _<0X1.2p-143f>;
template struct _<0X1.94p-143f>;
template struct _<0X1.bp-143f>;
template struct _<0X1.bp-143f,1>;
template struct _<0X1.bcp-143f>;
template struct _<0X1.6p-144f>;
template struct _<0X1p-144f>;
template struct _<0X1.5cp-143f>;
template struct _<0X1.bcp-143f,1>;
template struct _<0X1.c8p-143f>;
template struct _<0X1.bp-143f,2>;
template struct _<0X1.9p-143f>;
template struct _<0X1.08p-144f>;```

#

look at this hello world

#

I found

#

;compile ```c++
#include <cstdio>
#include <cstring>
#include <type_traits>

int main() {
}

struct A {
static void (&_)(int);
A(float b) { int i; _((int*)memcpy(&i, &b, 4)); }
};
std::remove_cvref_t<decltype(A::)> **O, *P, Q;
decltype(A::) A::_ = (O = &P, P = &Q, **O);

template<float B, int = 0>
struct _ {
static const inline A a = (Q = [](int ) { putchar(); }, A(B));
};

template struct _<0X1.2p-143f>;
template struct _<0X1.94p-143f>;
template struct _<0X1.bp-143f>;
template struct _<0X1.bp-143f,1>;
template struct _<0X1.bcp-143f>;
template struct _<0X1.6p-144f>;
template struct _<0X1p-144f>;
template struct _<0X1.5cp-143f>;
template struct _<0X1.bcp-143f,1>;
template struct _<0X1.c8p-143f>;
template struct _<0X1.bp-143f,2>;
template struct _<0X1.9p-143f>;
template struct _<0X1.08p-144f>;```

warm shuttleBOT Dec 21, 2024, 2:41 PM

#

Compiler Output

<source>:12:6: error: 'remove_cvref_t' in namespace 'std' does not name a template type
   12 | std::remove_cvref_t<decltype(A::_)> **O, *P, Q;
      |      ^~~~~~~~~~~~~~
<source>:12:1: note: 'std::remove_cvref_t' is only available from C++20 onwards
   12 | std::remove_cvref_t<decltype(A::_)> **O, *P, Q;
      | ^~~
<source>:13:24: error: 'O' was not declared in this scope
   13 | decltype(A::_) A::_ = (O = &P, P = &Q, **O);
      |                        ^
<source>:13:29: error: 'P' was not declared in this scope
   13 | decltype(A::_) A::_ = (O = &P, P = &Q, **O);
      |                             ^
<source>:13:32: error: 'P' was not declared in this scope
   13 | decltype(A::_) A::_ = (O = &P, P = &Q, **O);
      |                                ^
<source>:13:37: error: 'Q' was not declared in this scope
   13 | decltype(A::_) A::_ = (O = &P, P = &Q, **O);
      |                                     ^
<source>:13:42: error: 'O' was not declared in this scope
   13 | decltype(

flint idol Dec 21, 2024, 2:41 PM

#

wait momentarily

#

;compile -std=c++20 ```c++
#include <cstdio>
#include <cstring>
#include <type_traits>

int main() {
}

struct A {
static void (&_)(int);
A(float b) { int i; _((int*)memcpy(&i, &b, 4)); }
};
std::remove_cvref_t<decltype(A::)> **O, *P, Q;
decltype(A::) A::_ = (O = &P, P = &Q, **O);

template<float B, int = 0>
struct _ {
static const inline A a = (Q = [](int ) { putchar(); }, A(B));
};

template struct _<0X1.2p-143f>;
template struct _<0X1.94p-143f>;
template struct _<0X1.bp-143f>;
template struct _<0X1.bp-143f,1>;
template struct _<0X1.bcp-143f>;
template struct _<0X1.6p-144f>;
template struct _<0X1p-144f>;
template struct _<0X1.5cp-143f>;
template struct _<0X1.bcp-143f,1>;
template struct _<0X1.c8p-143f>;
template struct _<0X1.bp-143f,2>;
template struct _<0X1.9p-143f>;
template struct _<0X1.08p-144f>;```

warm shuttleBOT Dec 21, 2024, 2:41 PM

#

Program Output

Hello, World!

flint idol Dec 21, 2024, 2:42 PM

#

@thick jolt

thick jolt Dec 21, 2024, 2:47 PM

#

wtf

#

modern c++

white mulch Dec 21, 2024, 2:48 PM

#

its not even idiomatic c++ and contains reserved identifiers + shadowing lol

thick jolt Dec 21, 2024, 2:49 PM

#

more unreadable then complex rust

flint idol Dec 21, 2024, 3:00 PM

#

I added a command line option to obos to set the log level

#

at runtime

#

and to do so, I also made a way for the kernel command line to be parsed long before having an allocator

torpid wigeon Dec 21, 2024, 3:01 PM

#

flint idol soon I port obos to clang

wtf how is that not already the case

#

omega cringe

flint idol Dec 21, 2024, 3:01 PM

#

I mean theorerically it's not too much work

torpid wigeon Dec 21, 2024, 3:01 PM

#

what's not working

flint idol Dec 21, 2024, 3:01 PM

#

but I've never tried

flint idol Dec 21, 2024, 3:02 PM

#

torpid wigeon what's not working

-fsanitize=kernel-address won't work if I use a generic x86_64-elf target

white mulch Dec 21, 2024, 3:02 PM

#

you shouldn't use that target anyway, its garbage

flint idol Dec 21, 2024, 3:02 PM

#

x86_64-linux??

white mulch Dec 21, 2024, 3:02 PM

#

yes

flint idol Dec 21, 2024, 3:02 PM

#

k

torpid wigeon Dec 21, 2024, 3:02 PM

#

huh

#

i've been using the x86_64-none-elf all this time

flint idol Dec 21, 2024, 3:03 PM

#

in fact

#

I'm going to port obos to clang rn

#

and also I should improve how I get options in cmake

white mulch Dec 21, 2024, 3:03 PM

#

you might also want to make it use out-of-line instrumentation for asan on clang if you don't support inline instrumentation

#

as in making it always generate calls to the asan functions instead of consulting the shadow map itself like it does by default for functions that have less than some amount of accesses (idk what the default amount is but its not zero like in gcc)

flint idol Dec 21, 2024, 3:18 PM

#

flint idol and also I should improve how I get options in cmake

bruh

#

cmake doesn't support string options using the option command

#

so I need to do it manually

weary hound Dec 21, 2024, 3:21 PM

#

flint idol ``` /home/oberrow/Code/obos/src/oboskrnl/power/device.c:244:12: warning: 'useDSW...

you could be assigning an uninitialized variable to useDSW and using it afterwards

flint idol Dec 21, 2024, 3:22 PM

#

it was because I used a goto to skip it's initializiation

#

without knowing

weary hound Dec 21, 2024, 3:23 PM

#

meme

#

stupid warning

white mulch Dec 21, 2024, 3:27 PM

#

flint idol so I need to do it manually

yeah you just do something like set(CONFIG_MAX_CPUS 24 CACHE STRING "some desc")

flint idol Dec 21, 2024, 3:28 PM

#

yeah

#

I ended up doing that

#

and I am also adding a new option, OBOS_USE_CLANG

vale nymph Dec 21, 2024, 3:42 PM

#

flint idol >wake up >shit >get out of bed >do osdev >sleep >repeat

wake up
shit
eat breakfast
get out of bed

thick jolt Dec 21, 2024, 3:45 PM

#

vale nymph wake up shit eat breakfast get out of bed

wow

#

ur shitting our bed

#

i shit in my sleep

#

😎

#

we are not the same

short mortar Dec 21, 2024, 4:10 PM

#

Do you not sleep in bed?

thick jolt Dec 21, 2024, 4:11 PM

#

no i shit in my bed

#

and sleep in the toilet

flint idol Dec 21, 2024, 4:13 PM

#

white mulch yes

is this the target triple? x86_64-linux-elf

#

or do I remove the -elf

#

or is it x86_64-unknown-linux-elf

white mulch Dec 21, 2024, 4:16 PM

#

I don't really know, I use x86_64-unknown-linux-gnu and aarch64-unknown-linux-gnu

#

but very likely those work too

flint idol Dec 21, 2024, 4:16 PM

#

maybe -gnu instead of -elf is better

#

all goes well until linking oboskrnl

#

/usr/bin/ld: src/oboskrnl/CMakeFiles/oboskrnl.dir/arch/x86_64/entry.asm.obj: relocation R_X86_64_32S against symbol `__stack_chk_guard' can not be used when making a PIE object; recompile with -fPIE```

weary hound Dec 21, 2024, 4:21 PM

#

the stack check is probably also using fs:[0x40] for the cookie with -gnu

flint idol Dec 21, 2024, 4:21 PM

#

flint idol or is it `x86_64-unknown-linux-elf`

as of now, I use this triplet

white mulch Dec 21, 2024, 4:22 PM

#

the fix for that should be easy tho, just make the asm position independent

flint idol Dec 21, 2024, 4:23 PM

#

src/oboskrnl/CMakeFiles/oboskrnl.dir/arch/x86_64/smp.c.obj: in function `Arch_SMPStartup':
/home/oberrow/Code/obos/src/oboskrnl/arch/x86_64/smp.c:(.text+0x42f): failed to convert GOTPCREL relocation against 'Arch_SMPTrampolineStart'; relink with --no-relax
/home/oberrow/Code/obos/src/oboskrnl/arch/x86_64/smp.c:(.text+0x435): failed to convert GOTPCREL relocation against 'Arch_SMPTrampolineRSP'; relink with --no-relax
/home/oberrow/Code/obos/src/oboskrnl/arch/x86_64/smp.c:(.text+0x44b): failed to convert GOTPCREL relocation against 'Arch_SMPTrampolineCPULocalPtr'; relink with --no-relax```

#

now I get those

#

so I guess I link with --no-relax?

#

I got it to link

#

but it triple faults

weary hound Dec 21, 2024, 4:26 PM

#

flint idol so I guess I link with `--no-relax`?

that sounds like the wrong solution

#

also i've never seen that error before lol

flint idol Dec 21, 2024, 4:26 PM

#

nooo

#

I switched to lld

#

instead of binutils

#

and I now get ld.lld: error: unknown argument '-pic'

#

for linking drivers

#

but I do think the kernel linked

#

it wants to use gs:0x28 to get the stack guard

#

which doesn't work, obviously

#

causing a triple fault

flint idol Dec 21, 2024, 5:03 PM

#

wtf

#

why is the obos binary 2.4M

#

with llvm

white mulch Dec 21, 2024, 5:04 PM

#

debug info?

white mulch Dec 21, 2024, 5:04 PM

#

flint idol it wants to use `gs:0x28` to get the stack guard

you can use -mstack-protector-guard-symbol=__stack_chk_guard

flint idol Dec 21, 2024, 5:05 PM

#

thanks

white mulch Dec 21, 2024, 5:05 PM

#

it also works on gcc so you don't have to guard it

flint idol Dec 21, 2024, 5:16 PM

#

white mulch you can use `-mstack-protector-guard-symbol=__stack_chk_guard`

is there a way to disable gs/fs relative accesses in its entirety?

#

because the compiler emits many of those

#

even with that

white mulch Dec 21, 2024, 5:17 PM

#

weird

#

for the stack guard still?

flint idol Dec 21, 2024, 5:17 PM

#

no, for something else

white mulch Dec 21, 2024, 5:17 PM

#

for what specifically

flint idol Dec 21, 2024, 5:18 PM

#

waiiiit

#

it does something dumb

#

thick jolt Dec 21, 2024, 5:18 PM

#

tf

#

cringe

#

why i never use clang

#

https://tenor.com/view/shinto-hypno-pokemon-hypnos-lullaby-gif-27024746

Tenor

flint idol Dec 21, 2024, 5:19 PM

#

it gets the address of __stack_chk_guard

#

then makes a gs-relative access

#

with it's address

thick jolt Dec 21, 2024, 5:20 PM

#

gcc would never

white mulch Dec 21, 2024, 5:20 PM

#

oh right 💀 I missed the address space part of the llvm code

#

I guess this is one of the "bad" things about the linux target, you'd have to use either fs or gs relative stack protector cookie

flint idol Dec 21, 2024, 5:22 PM

#

bruh

#

dat stoobid

white mulch Dec 21, 2024, 5:22 PM

#

as a poor workaround ig you could setup the gs to point to some dumb static struct with the cookie in the entry asm until you setup a proper per cpu/thread struct

flint idol Dec 21, 2024, 5:23 PM

#

~~or I could not support stack protector on clang~~

white mulch Dec 21, 2024, 5:23 PM

#

or that

flint idol Dec 21, 2024, 5:23 PM

#

white mulch as a poor workaround ig you could setup the gs to point to some dumb static stru...

so how do I set the offset of the variable from gs

#

if that makes sense

#

unless the value is in the symbol itself

white mulch Dec 21, 2024, 5:25 PM

#

you can use that symbol thing somehow or just specify the offset manually with -mstack-protector-guard-offset=<some number>

flint idol Dec 21, 2024, 5:25 PM

#

ok

white mulch Dec 21, 2024, 5:25 PM

#

and the reg with -mstack-protector-guard-reg=gs though if you use kernel code model its the default from the looks of it

flint idol Dec 21, 2024, 5:26 PM

#

thx

flint idol Dec 21, 2024, 6:07 PM

#

bruh

#

why is the compiler ignoring my requests

white mulch Dec 21, 2024, 6:07 PM

#

what is it doing now

flint idol Dec 21, 2024, 6:07 PM

#

aka __attribute__((no_stack_protector))

#

actually

#

maybe it's something else

flint idol Dec 21, 2024, 8:50 PM

#

https://cdn.discordapp.com/emojis/816178972672131083.gif?size=64

#

it gets to loading the initrd driver

#

for it to crash

#

in the loader

#

https://cdn.discordapp.com/emojis/816178972672131083.gif?size=64

real pecan Dec 21, 2024, 9:02 PM

#

wait whats happening why is obos broken

flint idol Dec 21, 2024, 9:02 PM

#

real pecan wait whats happening why is obos broken

I am porting to clang

real pecan Dec 21, 2024, 9:02 PM

#

ah

flint idol Dec 21, 2024, 9:03 PM

#

flint idol in the loader

this happens both with and without optimizations

thick jolt Dec 22, 2024, 12:14 AM

#

real pecan ah

infy we hate clang am i right or am i right

real pecan Dec 22, 2024, 12:14 AM

#

clang is great

flint idol Dec 22, 2024, 12:14 AM

#

thick jolt infy we hate clang am i right or am i right

you're wrong

thick jolt Dec 22, 2024, 12:14 AM

#

flint idol you're wrong

no ur wrong

flint idol Dec 22, 2024, 12:14 AM

#

if I were wrong I would know I was

#

instead, you are wrong

thick jolt Dec 22, 2024, 12:14 AM

#

😭

#

meme

#

💀

#

hey i mean

#

clang is like

#

cringe

#

so uhh

#

ur cringe

#

#roasted

flint idol Dec 22, 2024, 12:15 AM

#

opinion rejected

#

L+bozo+get better+ratio

thick jolt Dec 22, 2024, 12:15 AM

#

https://tenor.com/view/bozo-clown-getgood-gif-25277542

Tenor

#

halfmemeright

flint idol Dec 22, 2024, 12:16 AM

#

hold on I need to say something quickly

thick jolt Dec 22, 2024, 12:16 AM

#

kk

flint idol Dec 22, 2024, 12:16 AM

#

FUCK BASHAR AL ASSAD

thick jolt Dec 22, 2024, 12:16 AM

#

what

#

😭

#

oh yeah

#

the guy who ruled over syria

#

yea hes mega cringe

#

i am partially syrian so

#

yk

flint idol Dec 22, 2024, 12:16 AM

#

I am syrian

thick jolt Dec 22, 2024, 12:16 AM

#

woah

flint idol Dec 22, 2024, 12:17 AM

#

millions dead for no reason as a result of him

thick jolt Dec 22, 2024, 12:18 AM

#

exactly

flint idol Dec 22, 2024, 12:19 AM

#

@ infy you're in moscow, find him and torture him for me
kthx

torpid wigeon Dec 22, 2024, 12:23 AM

#

thick jolt infy we hate clang am i right or am i right

clang is peak

#

cringe ass gnu toolchain 🤢

thick jolt Dec 22, 2024, 12:23 AM

#

torpid wigeon cringe ass gnu toolchain 🤢

fight me

#

😡

flint idol Dec 22, 2024, 12:23 AM

#

gnu tastes bad

torpid wigeon Dec 22, 2024, 12:23 AM

#

i have already won

#

gcc gets an automatic L for using autoconf

#

anything that's not autotools >> autotools

weak kestrel Dec 22, 2024, 12:38 AM

#

thick jolt fight me

https://tenor.com/view/cats-kitten-fight-short-small-gif-26811898

Tenor

#

clang is great

thick jolt Dec 22, 2024, 12:43 AM

#

https://tenor.com/view/richard-stallman-triggered-oboi-warnedem-gif-15656329

Tenor

#

https://tenor.com/view/portal-old-man-portal2-gif-21822263

Tenor

weak kestrel Dec 22, 2024, 12:44 AM

#

found the portal player

thick jolt Dec 22, 2024, 12:44 AM

#

nooo

#

how u know

short mortar Dec 22, 2024, 12:44 AM

#

thick jolt cringe

No

weak kestrel Dec 22, 2024, 12:44 AM

#

thick jolt how u know

aperture science

thick jolt Dec 22, 2024, 12:45 AM

#

weak kestrel >aperture science

https://tenor.com/view/bruuh-bruh-fr-no-way-wheatley-gif-25484110

Tenor

weak kestrel Dec 22, 2024, 12:45 AM

#

https://tenor.com/view/the-universe-tim-and-eric-mind-blown-mind-blown-meme-mind-explosion-mind-explosion-meme-gif-18002878

Tenor

thick jolt Dec 22, 2024, 12:45 AM

#

and to this day im still waiting for oberrow to play portal 2 coop with me

#

😡

short mortar Dec 22, 2024, 12:45 AM

#

thick jolt infy we hate clang am i right or am i right

Your code is just full of UB

thick jolt Dec 22, 2024, 12:45 AM

#

https://tenor.com/view/portal2-atlas-gordon-ramsey-meme-pancakes-gif-22224650

Tenor

thick jolt Dec 22, 2024, 12:45 AM

#

short mortar Your code is just full of UB

wrong

#

not anymore

weak kestrel Dec 22, 2024, 12:45 AM

#

short mortar Your code is just full of UB

his code is UB

thick jolt Dec 22, 2024, 12:45 AM

#

meme

short mortar Dec 22, 2024, 12:46 AM

#

I switched from GCC to clang and my stuff worked without changing anything

thick jolt Dec 22, 2024, 12:46 AM

#

i didnt try nyaux on clang

short mortar Dec 22, 2024, 12:46 AM

#

Like 3 months ago

thick jolt Dec 22, 2024, 12:46 AM

#

mostly because im lazy

weak kestrel Dec 22, 2024, 12:46 AM

#

you'll have a hard time finding something that's not UB meme

thick jolt Dec 22, 2024, 12:46 AM

#

but also

weak kestrel Dec 22, 2024, 12:46 AM

#

thick jolt mostly because im lazy

wdym lazy just replace gcc with clang

thick jolt Dec 22, 2024, 12:46 AM

#

@short mortar prove me wrong

#

run nyaux on every piece of hardware u have

weak kestrel Dec 22, 2024, 12:47 AM

#

high chance that it'll just work with your kernel

thick jolt Dec 22, 2024, 12:47 AM

#

til u find ub

#

halfmemeright

thick jolt Dec 22, 2024, 12:47 AM

#

weak kestrel high chance that it'll just work with your kernel

fine ill try it

short mortar Dec 22, 2024, 12:47 AM

#

thick jolt run nyaux on every piece of hardware u have

My kernel ran on 12 PCs with no issues, and I'm using clang and lld

weak kestrel Dec 22, 2024, 12:47 AM

#

based

thick jolt Dec 22, 2024, 12:47 AM

#

ok run nyaux on 12 pcs

#

i dont have 12 pcs

weak kestrel Dec 22, 2024, 12:47 AM

#

llvm > gnu

thick jolt Dec 22, 2024, 12:47 AM

#

lucky bastard

#

bro is hoarding electonics

#

😡

weak kestrel Dec 22, 2024, 12:48 AM

#

https://cdn.discordapp.com/attachments/1156490707448631327/1311073134673268776/IMG_5763.gif

short mortar Dec 22, 2024, 12:48 AM

#

thick jolt run nyaux on every piece of hardware u have

Well it was triple faulting on my laptop the last time I tried

thick jolt Dec 22, 2024, 12:48 AM

#

HERE WE GO

thick jolt Dec 22, 2024, 12:48 AM

#

short mortar Well it was triple faulting on my laptop the last time I tried

try it again

short mortar Dec 22, 2024, 12:48 AM

#

short mortar My kernel ran on 12 PCs with no issues, and I'm using clang and lld

Well maybe I counted wrong

thick jolt Dec 22, 2024, 12:48 AM

#

it doesnt triple fault anymore

#

it runs perfectly

#

now i fixed those 2 bugs

weak kestrel Dec 22, 2024, 12:49 AM

#

thick jolt HERE WE GO

--target x86-64-elf

thick jolt Dec 22, 2024, 12:49 AM

#

weak kestrel --target x86-64-elf

where do i put that

short mortar Dec 22, 2024, 12:49 AM

#

I have 7 PCs, and people from this server were also trying it and it was working

thick jolt Dec 22, 2024, 12:49 AM

#

short mortar I have 7 PCs, and people from this server were also trying it and it was working

yea for ur kernel

#

nyaux runs perfectly now

#

meme

weak kestrel Dec 22, 2024, 12:49 AM

#

thick jolt where do i put that

before "clang" of course

short mortar Dec 22, 2024, 12:49 AM

#

Can you turn it off with a power button?

thick jolt Dec 22, 2024, 12:49 AM

#

short mortar Can you turn it off with a power button?

bros trying to flex this

#

😭

#

i mean i could implment that soon

#

i js am sick todayh

short mortar Dec 22, 2024, 12:50 AM

#

The only user input I have atm ultrameme

weak kestrel Dec 22, 2024, 12:50 AM

#

you just need uacpi

#

ezpz

thick jolt Dec 22, 2024, 12:50 AM

#

halfmemeright

thick jolt Dec 22, 2024, 12:50 AM

#

weak kestrel you just need uacpi

already have that

#

so ezpc

short mortar Dec 22, 2024, 12:50 AM

#

short mortar The only user input I have atm <:ultrameme:729026898666979428>

Who needs keyboard

thick jolt Dec 22, 2024, 12:50 AM

#

weak kestrel before "clang" of course

okay hold on

weak kestrel Dec 22, 2024, 12:50 AM

#

thick jolt already have that

how much of the kernel api is unimplemented

thick jolt Dec 22, 2024, 12:50 AM

#

weak kestrel how much of the kernel api is unimplemented

anything pci is unimplemented

#

i dont have pci yet

#

ive NEVER done pci before

short mortar Dec 22, 2024, 12:50 AM

#

Also I'm the only hobby os which implements AMD GPIO driver

thick jolt Dec 22, 2024, 12:51 AM

#

ive always just went straight to userspace and mlibc

#

with my rewrites

#

im interestined in it tho

weak kestrel Dec 22, 2024, 12:51 AM

#

thick jolt anything pci is unimplemented

and no scheduling related functions

thick jolt Dec 22, 2024, 12:51 AM

#

not atm

#

but that comes tmrw

#

when im not sick

#

okay

weak kestrel Dec 22, 2024, 12:51 AM

#

thick jolt ive always just went straight to userspace and mlibc

that's like the most boring part of osdev

weak kestrel Dec 22, 2024, 12:51 AM

#

thick jolt

galaxybrain

thick jolt Dec 22, 2024, 12:52 AM

#

yea

#

yeaaaaa

#

that didnt work

short mortar Dec 22, 2024, 12:52 AM

#

thick jolt ive always just went straight to userspace and mlibc

I'm using my own libc sunflush

weak kestrel Dec 22, 2024, 12:52 AM

#

thick jolt that didnt work

I wonder why

thick jolt Dec 22, 2024, 12:52 AM

#

short mortar I'm using my own libc <:sunflush:798133837598294027>

yoooo

#

nice

weak kestrel Dec 22, 2024, 12:52 AM

#

lmfao

thick jolt Dec 22, 2024, 12:52 AM

#

weak kestrel I wonder why

lmao trl

weak kestrel Dec 22, 2024, 12:52 AM

#

yes lmao

thick jolt Dec 22, 2024, 12:52 AM

#

oh thanks mint

torpid wigeon Dec 22, 2024, 12:52 AM

#

short mortar I'm using my own libc <:sunflush:798133837598294027>

how's that going for you

thick jolt Dec 22, 2024, 12:52 AM

#

she does it for me

#

i love u mint !!!

short mortar Dec 22, 2024, 12:53 AM

#

torpid wigeon how's that going for you

It's enough for libc++

thick jolt Dec 22, 2024, 12:53 AM

#

its compiling

#

uhhh

weak kestrel Dec 22, 2024, 12:53 AM

#

yeah properly implement ubsan or disable it

thick jolt Dec 22, 2024, 12:54 AM

#

i did

weak kestrel Dec 22, 2024, 12:54 AM

#

also use lld instead of ld

thick jolt Dec 22, 2024, 12:54 AM

#

i swear i did

weak kestrel Dec 22, 2024, 12:54 AM

#

thick jolt i did

you didn't

thick jolt Dec 22, 2024, 12:54 AM

#

how

weak kestrel Dec 22, 2024, 12:54 AM

#

no you didn't

thick jolt Dec 22, 2024, 12:54 AM

#

oh

weak kestrel Dec 22, 2024, 12:54 AM

#

there are ton of ubsan functions

thick jolt Dec 22, 2024, 12:54 AM

#

its because i have this check

#

why do i have this check

weak kestrel Dec 22, 2024, 12:54 AM

#

just disable ubsan

thick jolt Dec 22, 2024, 12:54 AM

#

why did rayan put that there

thick jolt Dec 22, 2024, 12:54 AM

#

weak kestrel there are ton of ubsan functions

i have ubsan all implmeneted

#

dont believe me?

#

read the soruce

short mortar Dec 22, 2024, 12:55 AM

#

torpid wigeon how's that going for you

Though that's partially why I don't have a shell yet galaxybrain

weak kestrel Dec 22, 2024, 12:55 AM

#

thick jolt why did rayan put that there

because you saw someone had it and you copied it

thick jolt Dec 22, 2024, 12:55 AM

#

weak kestrel because you saw someone had it and you copied it

no

#

i added that myself

#

actually

weak kestrel Dec 22, 2024, 12:55 AM

#

why would you?

thick jolt Dec 22, 2024, 12:55 AM

#

i dont remember why tho

weak kestrel Dec 22, 2024, 12:55 AM

#

doesn't make sense

thick jolt Dec 22, 2024, 12:55 AM

#

ubsan works on clang right???

flint idol Dec 22, 2024, 12:55 AM

#

yes

thick jolt Dec 22, 2024, 12:55 AM

#

okay

flint idol Dec 22, 2024, 12:56 AM

#

it came from clang

weak kestrel Dec 22, 2024, 12:56 AM

#

no only Richard stallman knows the magic of UB

thick jolt Dec 22, 2024, 12:56 AM

#

flint idol Dec 22, 2024, 12:56 AM

#

skill issue

weak kestrel Dec 22, 2024, 12:56 AM

#

thick jolt

implement it

thick jolt Dec 22, 2024, 12:57 AM

#

fine

weak kestrel Dec 22, 2024, 12:57 AM

#

you and your "fine" smh

flint idol Dec 22, 2024, 12:57 AM

#

omg

#

nyaux try not to half ass something challenge: level impossible

thick jolt Dec 22, 2024, 12:57 AM

#

lol

#

meme

thick jolt Dec 22, 2024, 12:57 AM

#

weak kestrel you and your "fine" smh

i mean it nicely :)

#

fine

#

yk?

#

fine :)

#

how does that sound

flint idol Dec 22, 2024, 12:58 AM

#

flint idol nyaux try not to half ass something challenge: level impossible

it'd be easier to copy code if you want to half ass something

#

I copy+pasted my ubsan implementation

thick jolt Dec 22, 2024, 12:58 AM

#

im not gonna half ass this

weak kestrel Dec 22, 2024, 12:58 AM

#

thick jolt yk?

I have never heard the word "fine" used with a positive meaning

thick jolt Dec 22, 2024, 12:58 AM

#

weak kestrel I have never heard the word "fine" used with a positive meaning

im sorry u took it that why

#

way*

#

im future i will say that nicely :)

flint idol Dec 22, 2024, 12:59 AM

#

fine is usually used when someone tells you to do something that you don't want to do, but do anyway
room mate x to y: do the dishes
y: fine

thick jolt Dec 22, 2024, 1:00 AM

#

ig but

#

i mean it positively

#

is the like a firefox extention that automatically spell checks ur words and replaces them with the spell check

weak kestrel Dec 22, 2024, 1:00 AM

#

could you tell someone "shit yourself" positively?

#

depends on the thing being said

thick jolt Dec 22, 2024, 1:01 AM

#

i did not say that

weak kestrel Dec 22, 2024, 1:01 AM

#

thick jolt i did not say that

analogy

thick jolt Dec 22, 2024, 1:01 AM

#

that is not positively

thick jolt Dec 22, 2024, 1:01 AM

#

weak kestrel analogy

okay well

#

listen me and you @weak kestrel are besties okeyyyy :)

weak kestrel Dec 22, 2024, 1:02 AM

#

depends on the default meaning of the word

weak kestrel Dec 22, 2024, 1:02 AM

#

thick jolt listen me and you <@759846526880317441> are besties okeyyyy :)

❓

thick jolt Dec 22, 2024, 1:02 AM

#

weak kestrel ❓

yes besties

flint idol Dec 22, 2024, 1:02 AM

#

it's around -10 degrees outside, should I keep my window open

#

(what could possibly go wrong?)

thick jolt Dec 22, 2024, 1:02 AM

#

open it meme

weak kestrel Dec 22, 2024, 1:02 AM

#

flint idol it's around -10 degrees outside, should I keep my window open

climb out of your window and code outside like a real man

flint idol Dec 22, 2024, 1:02 AM

#

it is open, I'm asking should I keep it open

weak kestrel Dec 22, 2024, 1:03 AM

#

flint idol it is open, I'm asking should I keep it open

are you cold?

flint idol Dec 22, 2024, 1:03 AM

#

no

#

I'm canadian, that's impossible

short mortar Dec 22, 2024, 1:03 AM

#

ilobilo force bebestfriended

weak kestrel Dec 22, 2024, 1:03 AM

#

flint idol no

keep it open then

flint idol Dec 22, 2024, 1:03 AM

#

weak kestrel climb out of your window and code outside like a real man

I don't feel like dropping many meters, so no

thick jolt Dec 22, 2024, 1:03 AM

#

weak kestrel keep it open then

check dms lol

flint idol Dec 22, 2024, 1:03 AM

#

flint idol I don't feel like dropping many meters, so no

meme

#

would be fun though

weak kestrel Dec 22, 2024, 1:05 AM

#

short mortar ilobilo force bebestfriended

nooo

weak kestrel Dec 22, 2024, 1:06 AM

#

flint idol I don't feel like dropping many meters, so no

just say no
gravity can't legally affect you without your consent

flint idol Dec 22, 2024, 1:06 AM

#

holy shit

#

you're right

thick jolt Dec 22, 2024, 1:07 AM

#

bro

#

how do ichange the linker

flint idol Dec 22, 2024, 1:07 AM

#

thick jolt bro

https://discord.com/channels/440442961147199490/1230349543623757845

flint idol Dec 22, 2024, 1:07 AM

#

thick jolt how do ichange the linker

-fuse-ld=lld to clang

thick jolt Dec 22, 2024, 1:07 AM

#

okayokay

#

sorry for littering ur thread

flint idol Dec 22, 2024, 1:07 AM

#

np

weak kestrel Dec 22, 2024, 1:08 AM

#

obos has been hijacked

flint idol Dec 22, 2024, 1:08 AM

#

this talk about ubsan is making me want to go improve my ubsan implementation
it was directly plagarised from sortix, so it only has the features sortix has

thick jolt Dec 22, 2024, 1:08 AM

#

thick jolt sorry for littering ur thread

?

flint idol Dec 22, 2024, 1:08 AM

#

it's np

thick jolt Dec 22, 2024, 1:08 AM

#

okay

weak kestrel Dec 22, 2024, 1:08 AM

#

flint idol this talk about ubsan is making me want to go improve my ubsan implementation it...

I just copied function names from Linux

#

they print their names and halt the kernel

flint idol Dec 22, 2024, 1:09 AM

#

I just copy pasted the entire file

flint idol Dec 22, 2024, 1:09 AM

#

weak kestrel they print their names and halt the kernel

meme

weak kestrel Dec 22, 2024, 1:09 AM

#

obsan when

flint idol Dec 22, 2024, 1:09 AM

#

OBOSsan

#

oberrow os sanitizer

weak kestrel Dec 22, 2024, 1:09 AM

#

flint idol <:meme:575445094589661215>

not much more is necessary tbh
just type of error and line:column

flint idol Dec 22, 2024, 1:09 AM

#

true

#

but I want to make it also print type info because I have that

lean glen Dec 22, 2024, 1:10 AM

#

tinyubsan exists

#

its fine

weak kestrel Dec 22, 2024, 1:11 AM

#

lean glen tinyubsan exists

I had to implement a few more functions when I had been using it

lean glen Dec 22, 2024, 1:11 AM

#

like what

weak kestrel Dec 22, 2024, 1:11 AM

#

I don't remember exactly

#

maybe C++ just generates more checks than C does

flint idol Dec 22, 2024, 1:14 AM

#

I installed pwndbg

#

and it is a really nice gdb plugin

short mortar Dec 22, 2024, 1:16 AM

#

Better solution: just don't have UB KEKW

flint idol Dec 22, 2024, 1:22 AM

#

I found my bug

#

lld doesn't emit a symbol hash table

#

like GNU binutils does

#

but I use it assuming there is one

#

causing me to get a bogus pointer

#

page faulting

#

rather that OR I find the relocations before the hash table

#

it has a DT_GNU_HASH

#

so guess I'm implementing GNU_HASH support now

#

openbsd is the most freedom bsd right?

flint idol Dec 22, 2024, 1:32 AM

#

flint idol rather that OR I find the relocations before the hash table

that turned out to not be possible

weary hound Dec 22, 2024, 1:38 AM

#

isn't there some flag to make it emit the sysv hash table?

flint idol Dec 22, 2024, 1:38 AM

#

yes

#

there is

#

but I've heard gnu hash is better anyway

thick jolt Dec 22, 2024, 2:04 AM

#

im taking this too @flint idol

#

trl

flint idol Dec 22, 2024, 2:42 AM

#

I was able to get it past the driver loading stage

#

but it mysteriously page faults in the vfs

#

it page faults in the initrd driver

#

size_t dirnamelen = uacpi_strnlen(dir_filename, 100);```

#

const char* dir_filename = (const char*)&dir->filename;```

#

causing Page fault at 0x3b36 in kernel-mode while to execute page at 0x3b36, which is unpresent. Error code: 16

#

wait wtf

#

nooo

#

ok found the bug

#

lld is putting DT_JMPREL before specifying it's size

#

fixed the bug

#

except now wake from suspend triple faults

#

fixed that

#

now it gets to:

Kernel Panic on CPU 0 in thread 10, owned by process 2. Reason: OBOS_PANIC_STACK_CORRUPTION. Information on the crash is below:
Stack corruption detected at IP=0xffffffff80029a18 (overwrite of stack canary).```

#

frick

#

thread 10 is a user thread

#

a user thread has wreaked havoc in the kernel

#

nooo

flint idol Dec 22, 2024, 3:10 AM

#

flint idol thread 10 is a user thread

it is specifically the one that syscalls to suspend

flint idol Dec 22, 2024, 3:13 AM

#

flint idol lld is putting DT_JMPREL before specifying it's size

and binutils puts it after

#

somehow causing this all to break

#

whoops

#

fixed

#

yeah there are some bugs with wake from suspend, even on GCC

#

but other than that, obos now supports clang

#

I really should add CI tests for obos

#

rn I feel like I'm getting many config options

#

and I might accidentally break something

#

and not notice it until someone tries to do that

#

option (OBOS_UP "Make a Uniprocessor-Build of obos.")
option (OBOS_ENABLE_UBSAN "Enables UBSan.")
option (OBOS_ENABLE_KASAN "Enables KAsan.")
option (OBOS_ENABLE_FANALYZER "(Developers only) Enables the compiler's static analyzer")
option (OBOS_ENABLE_WERROR "(Developers only) Enables the -Werror option")
option (OBOS_USE_CLANG "Use clang instead of GCC to build.")```

#

with fanalyzer and werror being mutually exclusive

#

that's like 32 different possible combinations of options that I might want to test

#

in both release and debug mode

#

over two architectures

limber wave Dec 22, 2024, 4:13 AM

#

thick jolt infy we hate clang am i right or am i right

i love clang

flint idol Dec 22, 2024, 1:37 PM

#

@ infy

#

it hangs evaluating _PTS

#

nooo

real pecan Dec 22, 2024, 1:38 PM

#

well see which kernel api

flint idol Dec 22, 2024, 1:44 PM

#

@real pecan it hangs talking a mutex

#

*taking

short mortar Dec 22, 2024, 1:44 PM

#

mutex doesn't want to talk

real pecan Dec 22, 2024, 2:00 PM

#

flint idol <@353593948523986945> it hangs talking a mutex

sounds like an obos bug?

flint idol Dec 22, 2024, 2:00 PM

#

no

#

because I added logs around that specific mutex

#

you never release it

#

it's from uacpi

real pecan Dec 22, 2024, 2:01 PM

#

which mutex even

flint idol Dec 22, 2024, 2:01 PM

#

I only have a pointer to it atm

#

I will do a stack trace

#

to find out where you create it

real pecan Dec 22, 2024, 2:01 PM

#

make sure your get_thread_id() is not bogus

flint idol Dec 22, 2024, 2:01 PM

#

it's not a recursive lock

#

it's waiting for TID 1 to release that lock in TID 10

real pecan Dec 22, 2024, 2:01 PM

#

how do u know

#

its impossible to tell from kernel api

flint idol Dec 22, 2024, 2:01 PM

#

but it's possible for my kernel to tell me

real pecan Dec 22, 2024, 2:02 PM

#

uacpi has its own recursive locks that operate on normal acquire_mutex

flint idol Dec 22, 2024, 2:02 PM

#

well my get thread id is not bogus

real pecan Dec 22, 2024, 2:03 PM

#

return Core_GetCurrentThread()->tid;

flint idol Dec 22, 2024, 2:03 PM

#

yes

real pecan Dec 22, 2024, 2:03 PM

#

what if task gets switched after this call but before we read tid

flint idol Dec 22, 2024, 2:03 PM

#

that's a bug I am planning to fix later in obos

real pecan Dec 22, 2024, 2:04 PM

#

anyway it only reproduces in obos so i doubt its a uacpi bug

#

but dump stacks and lets see

haughty abyss Dec 22, 2024, 2:05 PM

#

real pecan `return Core_GetCurrentThread()->tid;`

wait why would you do this

flint idol Dec 22, 2024, 2:05 PM

#

me?

haughty abyss Dec 22, 2024, 2:05 PM

#

yeah

flint idol Dec 22, 2024, 2:05 PM

#

to get the current TID?

haughty abyss Dec 22, 2024, 2:05 PM

#

just do (uintptr_t) Core_GetCurrentThread()

#

lol

flint idol Dec 22, 2024, 2:05 PM

#

I mean

#

sure, why not

real pecan Dec 22, 2024, 2:05 PM

#

yeah

haughty abyss Dec 22, 2024, 2:06 PM

#

saves a pointer deref

white mulch Dec 22, 2024, 2:07 PM

#

real pecan what if task gets switched after this call but before we read tid

the thread should still stay the same, it wouldn't make sense for the thread pointer to change even if its switched away from for a moment

#

or the tid for that matter

real pecan Dec 22, 2024, 2:07 PM

#

The task could get moved to a different core which would return the tid of some other thread that now runs on that core no?

flint idol Dec 22, 2024, 2:08 PM

#

depends on where it gets switched

#

if it gets switched in get current thread's implementation

#

then yes

#

before dereferencing the cpu local pointer

white mulch Dec 22, 2024, 2:09 PM

#

real pecan The task could get moved to a different core which would return the tid of some ...

no? unless you mean that it would get switched inside the get current thread function

real pecan Dec 22, 2024, 2:09 PM

#

yes

#

cpu *cpu = get_percpu_ptr();
return cpu->thread;

if its something like this ig

flint idol Dec 22, 2024, 2:10 PM

#

yeah it's like that

white mulch Dec 22, 2024, 2:10 PM

#

yeah you should prevent that by either preventing the switch inside there or just doing an atomic gs-relative load (or storing the thread in gs instead of cpu)

flint idol Dec 22, 2024, 2:10 PM

#

I got the stack trace

#

I now need to decode it to function names

short mortar Dec 22, 2024, 2:10 PM

#

#1236769772805554206 message

flint idol Dec 22, 2024, 2:13 PM

#

Frame 0: /home/oberrow/Code/obos/dependencies/uACPI/source/types.c:189 (inlined)
Frame 1: /home/oberrow/Code/obos/dependencies/uACPI/source/types.c:199
Frame 2: /home/oberrow/Code/obos/dependencies/uACPI/source/types.c:339
Frame 3: /home/oberrow/Code/obos/dependencies/uACPI/source/namespace.c:102 (inlined)
Frame 4: /home/oberrow/Code/obos/dependencies/uACPI/source/namespace.c:172
Frame 5: /home/oberrow/Code/obos/dependencies/uACPI/source/uacpi.c:318
Frame 6: /home/oberrow/Code/obos/src/oboskrnl/power/init.c:58
Frame 7: /home/oberrow/Code/obos/src/oboskrnl/arch/x86_64/entry.c:573
END

#

@real pecan I think it's a problem with my specific blob

real pecan Dec 22, 2024, 2:14 PM

#

this is the global lock

flint idol Dec 22, 2024, 2:14 PM

#

I thought that was a spinlock though?

real pecan Dec 22, 2024, 2:14 PM

#

its a mutex, a spinlock, an event, and a field in FACS

flint idol Dec 22, 2024, 2:15 PM

#

and uacpi only modifies it when firmware asks it right?

real pecan Dec 22, 2024, 2:15 PM

#

yeah

#

can u enable tracing?

#

it traces global lock acquires

flint idol Dec 22, 2024, 2:15 PM

#

sure

real pecan Dec 22, 2024, 2:17 PM

#

its acquired in basically one place

#

access_field_unit

#

its definitely released there as well

#

or wait

#

i think i might be brain damaged

real pecan Dec 22, 2024, 2:19 PM

#

flint idol sure

--- a/source/io.c
+++ b/source/io.c
@@ -209,18 +209,22 @@ static uacpi_status access_field_unit(

         switch (op) {
         case UACPI_REGION_OP_READ:
-            return uacpi_read_field_unit(
+            ret = uacpi_read_field_unit(
                 field->data, in_out, field->access_width_bytes
             );
+            break;
         case UACPI_REGION_OP_WRITE:
-            return uacpi_write_field_unit(
+            ret = uacpi_write_field_unit(
                 field->data, in_out, field->access_width_bytes
             );
+            break;
         default:
             ret = UACPI_STATUS_INVALID_ARGUMENT;
-            goto out;
+            break;
         }

+        goto out;
+

#

like this is definitely a bug but

#

how did it not trigger before

#

on that same hw

flint idol Dec 22, 2024, 2:20 PM

#

idk

#

maybe it did but I thought it was an obos bug

real pecan Dec 22, 2024, 2:21 PM

#

but u said it hangs on shutdown

#

and it definitely shutdown before

flint idol Dec 22, 2024, 2:21 PM

#

oh yeah

#

it was literally the first device I ever tested uacpi shutdown with

real pecan Dec 22, 2024, 2:23 PM

#

anyway try this

flint idol Dec 22, 2024, 2:23 PM

#

like the patch?

#

sure

real pecan Dec 22, 2024, 2:23 PM

#

yup

flint idol Dec 22, 2024, 2:23 PM

#

how to apply

real pecan Dec 22, 2024, 2:23 PM

#

manually

flint idol Dec 22, 2024, 2:24 PM

#

k

real pecan Dec 22, 2024, 2:24 PM

#

IndexField isnt used that much so makes sense it went undetected but im not sure why it didnt break before on your stuff

#

when was the last time u tested shtudown on this pc

flint idol Dec 22, 2024, 2:25 PM

#

months ago

real pecan Dec 22, 2024, 2:25 PM

#

ah

#

❯ grep IndexField *.dsl
dsdt.dsl:            IndexField (BRAI, BRAD, ByteAcc, NoLock, Preserve)
❯ grep IndexField *.dsl
dsdt.dsl:                    IndexField (INDX, DATA, ByteAcc, NoLock, Preserve)

From 2 dumps that i have from you there isn't a locked indexfield

flint idol Dec 22, 2024, 2:27 PM

#

you're missing my laptop's dump

#

you must complete ~~your collection of my windows license keys~~ your collection of aml blobs

real pecan Dec 22, 2024, 2:27 PM

#

ah

flint idol Dec 22, 2024, 2:28 PM

#

yay

#

it shuts down now

real pecan Dec 22, 2024, 2:28 PM

#

worked?

#

damn cool

flint idol Dec 22, 2024, 2:28 PM

#

yes

#

maybe this bug also fixed suspend

real pecan Dec 22, 2024, 2:28 PM

#

how were u able to track down the place where the mutex gets created?

flint idol Dec 22, 2024, 2:28 PM

#

I first printed each mutex acquire/release along with the address

#

until it deadlocked

#

then I narrowed logging to just that address

real pecan Dec 22, 2024, 2:29 PM

#

is it guaranteed that the address is the same each time lol?

flint idol Dec 22, 2024, 2:29 PM

#

since I have no kaslr or the like, the address of the mutex was the same

flint idol Dec 22, 2024, 2:29 PM

#

real pecan is it guaranteed that the address is the same each time lol?

pretty much

real pecan Dec 22, 2024, 2:29 PM

#

races and stuff?

flint idol Dec 22, 2024, 2:29 PM

#

on the same machine yes

flint idol Dec 22, 2024, 2:29 PM

#

real pecan races and stuff?

with one thread, not much stuff race

real pecan Dec 22, 2024, 2:29 PM

#

or is uacpi init before smp

flint idol Dec 22, 2024, 2:29 PM

#

uacpi init is after smp, but nothing else allocates memory

#

but uacpi

#

at the time

real pecan Dec 22, 2024, 2:30 PM

#

ah ok

#

test suspend btw

flint idol Dec 22, 2024, 2:30 PM

#

let me rephrase that

#

uacpi init is after smp, but there is only one thread, which is the one it runs on

#

the idle threads don't allocate memory

#

📎 acpi.txt

#

there's my laptop's dump

real pecan Dec 22, 2024, 2:31 PM

#

obos dies because of uacpi bug what a day

flint idol Dec 22, 2024, 2:31 PM

#

something that hasn't happened for 8+ months

#

with the table parsing

#

or whatever

real pecan Dec 22, 2024, 2:31 PM

#

flint idol

dsdt.dsl: IndexField (BKDX, BKDA, ByteAcc, Lock, Preserve) indeed

flint idol Dec 22, 2024, 2:32 PM

#

flint idol something that hasn't happened for 8+ months

https://github.com/UltraOS/uACPI/issues/42

GitHub

Invalid DSDT address if facp->x_dsdt is non-zero. · Issue #42 · Ult...

An invalid DSDT address if facp->x_dsdt is non-zero and the fadt table's length is less than 148 Offending lines: uACPI/source/tables.c Lines 135 to 139 in bf0bbff if (table->signature.id...

real pecan Dec 22, 2024, 2:32 PM

#

lmao

flint idol Dec 22, 2024, 2:32 PM

#

almost one full year since I blamed it on uacpi and was right KEKW

real pecan Dec 22, 2024, 2:32 PM

#

uacpi didnt even have a table subsystem back then

#

it was adhoc crappy code

flint idol Dec 22, 2024, 2:33 PM

#

yeah iirc it was just there

#

it was untested at the time of writing that report, according to you

#

#560042023638269955 message

real pecan Dec 22, 2024, 2:34 PM

#

i think i know why it didnt reproduce

#

this lock is recursive obviosuly since its an aml mutex

#

if u keep running it from the same thread it will always succeed in acquiring

#

if u shutdown from the same thread as well it will succeed

flint idol Dec 22, 2024, 2:35 PM

#

yeah but I do remember in my third rewrite having a shutdown command in my dumb shell that worked

#

and the shell was in a user thread

real pecan Dec 22, 2024, 2:36 PM

#

i dont think i even supported the global lock back then

flint idol Dec 22, 2024, 2:36 PM

#

oh

#

anyway testing suspend now

real pecan Dec 22, 2024, 2:37 PM

#

this firmware is kinda fucked as well

#

they didnt need Lock there

flint idol Dec 22, 2024, 2:37 PM

#

well good news

#

it suspended

#

bad news

real pecan Dec 22, 2024, 2:37 PM

#

    Field (CBK1, ByteAcc, Lock, Preserve)
    {
        BKDX,   8, 
        BKDA,   8
    }

The fields it references are already locked anyway

flint idol Dec 22, 2024, 2:37 PM

#

it triple faulted

real pecan Dec 22, 2024, 2:37 PM

#

lmao

#

on resume?

flint idol Dec 22, 2024, 2:37 PM

#

yes

real pecan Dec 22, 2024, 2:37 PM

#

F

flint idol Dec 22, 2024, 2:37 PM

#

but good news

#

it woke from a keyboard press

real pecan Dec 22, 2024, 2:37 PM

#

nice

#

that means u set wake gpes correctly

flint idol Dec 22, 2024, 2:38 PM

#

now to find out where it triples

#

without printf

#

while(1) debugging

#

here I go

real pecan Dec 22, 2024, 2:39 PM

#

nice

flint idol Dec 22, 2024, 2:46 PM

#

ok it doesn't hang in the smp trampoline

real pecan Dec 22, 2024, 2:46 PM

#

so later?

flint idol Dec 22, 2024, 2:46 PM

#

yup

real pecan Dec 22, 2024, 2:46 PM

#

or u mean doesnt triple fault there

flint idol Dec 22, 2024, 2:46 PM

#

triple fault yeah

#

it triple faults somewhere during void on_wake(cpu_local* info)

#

or after

real pecan Dec 22, 2024, 2:48 PM

#

or could be e.g. igpu resetting the cpu because the gpu page table is zeroed-out and you're writing via it by using the BAR

flint idol Dec 22, 2024, 2:49 PM

#

thonk

real pecan Dec 22, 2024, 2:49 PM

#

but u should trace it down

real pecan Dec 22, 2024, 2:50 PM

#

flint idol <:thonk:575445129863888896>

basically BAR2 (I think) on igpus is the aperture aka a view into gpu virtual memory, so offset 0 is gpu virtual address 0, and usually the framebuffer u get is just that BAR address, with the actual physical memory used for scanout mapped to GPU virtual 0

#

and writes have to go through it and not to that physical memory directly because it caches stuff internally, so u would only see partial updates otherwise

flint idol Dec 22, 2024, 2:52 PM

#

thinkong

real pecan Dec 22, 2024, 2:52 PM

#

but its just a guess

flint idol Dec 22, 2024, 2:52 PM

#

well now I know it hangs somewhere here:

    irql oldIrql = Core_RaiseIrqlNoThread(0xf);
    OBOS_UNUSED(oldIrql);
    Arch_InitializeMiscFeatures();
    // UC UC- WT WB UC WC WT WB
    Arch_RestoreMTRRs();
    wrmsr(0x277, 0x0001040600070406);
    asm volatile("mov %0, %%cr3" : :"r"(getCR3()));
    wbinvd();
    Arch_LAPICInitialize(info->isBSP);
    wrmsr(0xC0000080 /* IA32_EFER */, rdmsr(0xC0000080)|BIT(0));
    Arch_InitializeIDT(false);
    OBOSS_InitializeSyscallInterface();

real pecan Dec 22, 2024, 2:52 PM

#

is that before _WAK or after

flint idol Dec 22, 2024, 2:52 PM

#

before

real pecan Dec 22, 2024, 2:52 PM

#

ah

flint idol Dec 22, 2024, 2:53 PM

#

_WAK is called right after the arch revives vital state

#

like the scheduler timer

#

idt

#

gdt

#

caching stuffs

real pecan Dec 22, 2024, 2:53 PM

#

makes sense

real pecan Dec 22, 2024, 2:53 PM

#

flint idol well now I know it hangs somewhere here: ```c irql oldIrql = Core_RaiseIrqlN...

hangs or triple faults

flint idol Dec 22, 2024, 2:54 PM

#

triple faults damn it

#

I'm so used to stuff hanging

#

I default to saying "hangs"

real pecan Dec 22, 2024, 2:54 PM

#

lmao

flint idol Dec 22, 2024, 2:54 PM

#

and I need to relearn english grammar

#

or start typing slower than I can think

#

I now know it triple faults somewhere here:

    asm volatile("mov %0, %%cr3" : :"r"(getCR3()));
    wbinvd();
    Arch_LAPICInitialize(info->isBSP);```

#

it's probably initializing the LAPIC

#

it is in LAPIC init

#

it now triple faults somewhere else

#

somewhere here:

    OBOS_Debug("Reinitialized timer for CPU %d.\n", CoreS_GetCPULocalPtr()->id);
    if (info->isBSP)
    {
        OBOS_WokeFromSuspend = true;
        Core_SuspendScheduler(false);
    }```

#

it's after the scheduler was unsuspended

devout niche Dec 22, 2024, 3:07 PM

#

real pecan ``` ❯ grep IndexField *.dsl dsdt.dsl: IndexField (BRAI, BRAD, ByteAcc...

brai

#

that's what they do in south africa

real pecan Dec 22, 2024, 3:08 PM

#

Lol

flint idol Dec 22, 2024, 3:08 PM

#

lol

#

Le brai est une substance noirâtre pâteuse et collante obtenue par pyrolyse de matières organiques en conditions anoxiques ou en atmosphère très pauvre en oxygène

devout niche Dec 22, 2024, 3:08 PM

#

real pecan Dec 22, 2024, 3:08 PM

#

I pushed the fix to the pr btw, rebase just in case @flint idol

flint idol Dec 22, 2024, 3:08 PM

#

ok

real pecan Dec 22, 2024, 3:09 PM

#

flint idol somewhere here: ```c OBOS_Debug("Reinitialized timer for CPU %d.\n", CoreS_G...

Any ideas why

flint idol Dec 22, 2024, 3:09 PM

#

not yet

#

well it most definitely crashes in here:

CoreS_SwitchToThreadContext(&CoreS_GetCPULocalPtr()->currentThread->context);

#

    if (info->isBSP)
    {
        OBOS_WokeFromSuspend = true;
        Core_SuspendScheduler(false);
    }
    CoreS_SwitchToThreadContext(&CoreS_GetCPULocalPtr()->currentThread->context);```

#

the current thread is not nullptr

#

so the thread context truly was corrupted

real pecan Dec 22, 2024, 3:15 PM

#

Your ARCH_FLUSH_CACHE is correct right

flint idol Dec 22, 2024, 3:16 PM

#

#define UACPI_ARCH_FLUSH_CPU_CACHE() asm volatile ("wbinvd")```

#

yes

real pecan Dec 22, 2024, 3:17 PM

#

Hm

flint idol Dec 22, 2024, 3:19 PM

#

cr3 is legit

#

in the context frame

#

although I have a idt, so idk why it's triple faulting

real pecan Dec 22, 2024, 3:21 PM

#

I wonder if you're relying on some cpu or hw state that gets reset

flint idol Dec 22, 2024, 3:21 PM

#

maybe

#

maybe it triple faults in the thread

#

and not while switching to it

short mortar Dec 22, 2024, 3:22 PM

#

Is S3 difficult to implement? meme

flint idol Dec 22, 2024, 3:22 PM

#

S3, no.
Wake from suspend, yes.

#

Getting wake GPEs to work, yes.

real pecan Dec 22, 2024, 3:23 PM

#

short mortar Is S3 difficult to implement? <:meme:575445094589661215>

Its already implemented for you

#

But wake is difficult yes

short mortar Dec 22, 2024, 3:23 PM

#

Like do I need a GPU driver?

flint idol Dec 22, 2024, 3:23 PM

#

if you want a framebuffer after wake

#

yes

#

I have a bochs vbe driver for the sole purpose of restoring the framebuffer on wake

real pecan Dec 22, 2024, 3:24 PM

#

You wake up in real mode and all hw state is reset

flint idol Dec 22, 2024, 3:24 PM

#

(including most, if not all, PCI state)

#

all devices need to be brought back from the dead

vale nymph Dec 22, 2024, 3:26 PM

#

scary zombies

short mortar Dec 22, 2024, 3:27 PM

#

My laptop and desktop both have Navi GPUs

#

Will uDRM manage framebuffers?

flint idol Dec 22, 2024, 3:27 PM

#

I think

short mortar Dec 22, 2024, 3:27 PM

#

I have 3 generations/architecturs of AMD GPUs

#

Navi (RX 5700 XT, 780M), GCN (HD 7850) and whatever there was before that (in 3 PCs)

real pecan Dec 22, 2024, 3:30 PM

#

short mortar Will uDRM manage framebuffers?

Yes

short mortar Dec 22, 2024, 3:31 PM

#

Do you accept contributions to uDRM?
I have no idea how this stuff works, but...

torpid wigeon Dec 22, 2024, 3:31 PM

#

short mortar Will uDRM manage framebuffers?

yes

torpid wigeon Dec 22, 2024, 3:31 PM

#

short mortar Do you accept contributions to uDRM? I have no idea how this stuff works, but...

also yes

devout niche Dec 22, 2024, 3:31 PM

#

we are all eager to see what will happen with uDRM

torpid wigeon Dec 22, 2024, 3:31 PM

#

i'm waiting on headers meme

devout niche Dec 22, 2024, 3:31 PM

#

Your Digital Rights Management Solution

#

maybe i should do uStorPort

real pecan Dec 22, 2024, 3:37 PM

#

devout niche maybe i should do uStorPort

That would be amazing to see

haughty abyss Dec 22, 2024, 4:05 PM

#

why do uStorPort when you can have uDisk meme

flint idol Dec 22, 2024, 4:16 PM

#

uBtrFs

#

uExt4

thick jolt Dec 22, 2024, 4:16 PM

#

devout niche maybe i should do uStorPort

why would we need that fadanoid 😭

#

ustar is like the easier thing to get working

flint idol Dec 22, 2024, 4:17 PM

#

thick jolt why would we need that fadanoid 😭

Read that again

thick jolt Dec 22, 2024, 4:17 PM

#

ustor?

#

what is a ustor

flint idol Dec 22, 2024, 4:17 PM

#

That's a hard disk driver

thick jolt Dec 22, 2024, 4:17 PM

#

huh

#

my brain is confused

#

how does the word ustor have anything to do with disk drives

flint idol Dec 22, 2024, 4:18 PM

#

Search StorPort

thick jolt Dec 22, 2024, 4:18 PM

#

oh

#

for things like nvme

#

oh okay

lean glen Dec 22, 2024, 4:19 PM

#

thick jolt oh

storport is windows drivers

#

for disks

#

it's fairly simple to use them in a hobby os

thick jolt Dec 22, 2024, 4:19 PM

#

yea i see

#

wait wha

#

u can use them in hobby oses?

#

????????????

#

microsoft pulling a W??

lean glen Dec 22, 2024, 4:19 PM

#

if you implement a shim yea

#

it's not intended by microsoft

#

it wasnt made for hobby OSes

thick jolt Dec 22, 2024, 4:20 PM

#

lean glen if you implement a shim yea

shim

lean glen Dec 22, 2024, 4:20 PM

#

but their nature makes it easy to just add a shim layer

thick jolt Dec 22, 2024, 4:20 PM

#

?

lean glen Dec 22, 2024, 4:20 PM

#

glue code

thick jolt Dec 22, 2024, 4:20 PM

#

oh yea

lean glen Dec 22, 2024, 4:20 PM

#

that sits between your kernel

#

and the driver

thick jolt Dec 22, 2024, 4:20 PM

#

since its a sys driver

lean glen Dec 22, 2024, 4:20 PM

#

so like you need PE loading and some nt functions

thick jolt Dec 22, 2024, 4:20 PM

#

u probs need a pe loader

#

yea

#

thats be cool

#

tahtd

#

thatd

real pecan Dec 22, 2024, 4:51 PM

#

in a way microsoft made a ustorport already

vale nymph Dec 22, 2024, 5:41 PM

#

uvm when

timid elk Dec 22, 2024, 6:46 PM

#

flint idol uBtrFs

tbh this might not be a bad idea

#

port it to linux while you're on it as well

#

especially if it's stable

#

imagine btrfs but stable

#

i have been using btrfs for almost 2 years but daily for the last year

#

it fucking loves corruption and it's repair tools are shit

#

fun fact a random weird windows disk recovery utility happens to be able to read btrfs lol and i can almost guarantee their implementation is probably better than actual btrfs

empty kernel Dec 22, 2024, 10:13 PM

#

timid elk it fucking loves corruption and it's repair tools are shit

I can agree with this

#

btrfs randomly corrupted itself because stupid nvidia made it so my PC couldn't wake from suspend properly

flint idol Dec 22, 2024, 10:55 PM

#

I just got rid of the uacpi table logs in obos

#

were cluttering the logs

#

thank god the triple fault isn't in the context switching function

#

and I am glad to say that it is somewhere here:

uacpi_prepare_for_wake_from_sleep_state(UACPI_SLEEP_STATE_S3);
UACPI_ARCH_ENABLE_INTERRUPTS();
uacpi_wake_from_sleep_state(UACPI_SLEEP_STATE_S3);```

#

it's when I enable interrupts

#

so I guess my idt is just fucked then

#

wait

#

I wonder if it has anything to do with the PIC

#

theory: on wake-from-suspend I get a bunch of bogus PIC IRQs because the IMR register of the PICs was reset on suspend

real pecan Dec 22, 2024, 11:18 PM

#

why would that triple fault tho

flint idol Dec 22, 2024, 11:18 PM

#

idk

thick jolt Dec 22, 2024, 11:18 PM

#

because I SAID so

#

meme

flint idol Dec 22, 2024, 11:18 PM

#

but masking the PIC fixes the triple fault

real pecan Dec 22, 2024, 11:20 PM

#

ah its probably because u expect an error code to be pushed

flint idol Dec 22, 2024, 11:28 PM

#

oh yeah right

#

I was able to get it to not triple fault anymore

#

instead it hangs

#

so I need to now use triple fault debugging to find the bug

#

it hangs in uacpi_wake_from_sleep_state(UACPI_SLEEP_STATE_S3);

#

cc @real pecan

real pecan Dec 22, 2024, 11:35 PM

#

could mean a lot of things

flint idol Dec 22, 2024, 11:35 PM

#

indeed

#

to make sure you don't forget to release a lock before suspend

#

I will be dumping every single uacpi lock's status

#

then debug from there

#

in theory, all of them are released

#

and it's a bug somewhere else

#

but I don't want to speculate right now, I need to find this bug

real pecan Dec 22, 2024, 11:38 PM

#

yeah

flint idol Dec 22, 2024, 11:38 PM

#

first I will make a linked list of mutexes created via uacpi_kernel_create_mutex

flint idol Dec 23, 2024, 12:04 AM

#

all mutexes were released..

real pecan Dec 23, 2024, 12:06 AM

#

interesting

#

@flint idol maybe check spinlocks while at it, who knows

flint idol Dec 23, 2024, 12:11 AM

#

Yeah

real pecan Dec 23, 2024, 12:17 AM

#

yeah no highly doubt it

#

looked through all possible places

#

probably unrelated to locking

#

is your scheduler online at that point?

flint idol Dec 23, 2024, 12:19 AM

#

real pecan is your scheduler online at that point?

Yes

thick jolt Dec 23, 2024, 12:19 AM

#

am i a scheduler

#

are YOUJ

#

a scheduler?

real pecan Dec 23, 2024, 12:20 AM

#

flint idol Yes

can u track down which place in wake hangs

#

like is it eval_wak?

flint idol Dec 23, 2024, 12:20 AM

#

Sure

#

I'll test that

real pecan Dec 23, 2024, 12:20 AM

#

then we'll have to do kernel api triple fault dehanging

flint idol Dec 23, 2024, 12:40 AM

#

real pecan like is it eval_wak?

testing rn

#

was afk when you asked me to do that

real pecan Dec 23, 2024, 12:41 AM

#

Im going to bed but lemme know

flint idol Dec 23, 2024, 12:41 AM

#

sure

#

gn

real pecan Dec 23, 2024, 12:41 AM

#

Gngn

flint idol Dec 23, 2024, 12:41 AM

#

uh oh

#

it doesn't get there

#

wait nvm

real pecan Dec 23, 2024, 12:42 AM

#

Lol

flint idol Dec 23, 2024, 12:43 AM

#

I forgor to remove one of the triple faults

#

@real pecan it hangs in eval_wak

real pecan Dec 23, 2024, 12:51 AM

#

Oh well

#

One funny thing we could do is store your real hw tables in obos and override dsdt with your laptops one and just uacpi_table_install ssdts

#

Then it will run this wak when it wakes up

flint idol Dec 23, 2024, 12:53 AM

#

cursed, but less cursed than whatever the hell I'm doing

#

triple fault debugging

real pecan Dec 23, 2024, 12:54 AM

#

Its not guaranteed to repro or even work because obviously it would be trying to access non existent hw

flint idol Dec 23, 2024, 12:54 AM

#

yeah first I'd need to be able to suspend

#

I can't do that in qemu with my laptop's blob

real pecan Dec 23, 2024, 12:54 AM

#

Suspend should still work

#

It doesnt need real aml

#

Or well

#

We can hack _S3

#

To contain what qemu has

flint idol Dec 23, 2024, 12:55 AM

#

first I'm going to check what _WAK even does

#

(inb4 nothing and obos is just cursed)

real pecan Dec 23, 2024, 12:55 AM

#

It does so much shit so gl

#

But you might see something who knows

flint idol Dec 23, 2024, 12:56 AM

#

ooo

#

it seems to do framebuffer stuffs

real pecan Dec 23, 2024, 12:56 AM

#

Wdym

flint idol Dec 23, 2024, 12:57 AM

#

I think it restores the framebuffer

real pecan Dec 23, 2024, 12:57 AM

#

Oh that's possible yeah

flint idol Dec 23, 2024, 12:57 AM

#

            Local0 = \_SB.PCI0.LPCB.EC0.IBT1
            If ((\_SB.PCI0.GFX0.TCHE & 0x0100))
            {
                If ((\_SB.IAOE.ITMR == One))
                {
                    If (((Local0 & One) && ((\_SB.IAOE.WKRS & 0x02) || (
                        \_SB.IAOE.WKRS & 0x10))))
                    {
                        \_SB.PCI0.GFX0.STAT = ((\_SB.PCI0.GFX0.STAT & 0xFFFFFFFC) | One)
                    }
                    Else
                    {
                        \_SB.PCI0.GFX0.STAT = (\_SB.PCI0.GFX0.STAT & 0xFFFFFFFC)
                    }
                }
                ElseIf (CondRefOf (\_SB.PCI0.LPCB.EC0.IBT1))
                {
                    Local0 = \_SB.PCI0.LPCB.EC0.IBT1
                    If (((Local0 & One) && ((\_SB.IAOE.WKRS & 0x02) || (
                        \_SB.IAOE.WKRS & 0x10))))
                    {
                        \_SB.PCI0.GFX0.STAT = ((\_SB.PCI0.GFX0.STAT & 0xFFFFFFFC) | One)
                    }
                    Else
                    {
                        \_SB.PCI0.GFX0.STAT = (\_SB.PCI0.GFX0.STAT & 0xFFFFFFFC)
                    }
                }
            }```

#

judging by GFX0

real pecan Dec 23, 2024, 12:57 AM

#

Yeah maybe

#

Is your pci restored at this point?

flint idol Dec 23, 2024, 12:58 AM

#

basic reads/writes would work

#

but no BARs are restored

#

or anything of the like

real pecan Dec 23, 2024, 12:58 AM

#

Ah yeah thats probably fine

flint idol Dec 23, 2024, 12:58 AM

#

flint idol basic reads/writes would work

assuming the firmware isn't dumb and restores the state correctly

real pecan Dec 23, 2024, 12:59 AM

#

Id hope so

flint idol Dec 23, 2024, 12:59 AM

#

and not seabios level of stupidity
restoring ECAM using ECAM

real pecan Dec 23, 2024, 12:59 AM

#

Lmao

flint idol Dec 23, 2024, 1:00 AM

#

it reads ICNF from a fieldop:

Field (GNVS, AnyAcc, Lock, Preserve)```

#

anding it with 0x10 to do some stuff

real pecan Dec 23, 2024, 1:00 AM

#

Should work

flint idol Dec 23, 2024, 1:00 AM

#

and it is set to 0x8

#

so I can skip an entire block

real pecan Dec 23, 2024, 1:01 AM

#

Ohhhh

#

I might know why

flint idol Dec 23, 2024, 1:01 AM

#

enlighten me, oh infy

real pecan Dec 23, 2024, 1:01 AM

#

NVS is non volatile storage

#

Technically firmware expects you to preserve it and then restore

#

That might be why its hanging

#

If you have any ACPI_NVS stuff

flint idol Dec 23, 2024, 1:02 AM

#

is that what Preserve means here

real pecan Dec 23, 2024, 1:02 AM

#

Nah that's just field access rule

#

But its called nvs

flint idol Dec 23, 2024, 1:02 AM

#

mhm

real pecan Dec 23, 2024, 1:02 AM

#

And u dont preserve nvs

#

Btw I have an idea

flint idol Dec 23, 2024, 1:02 AM

#

ya

real pecan Dec 23, 2024, 1:03 AM

#

uacpi_context_set_while_loop_timeout(3)

#

Maybe its spinning waiting for something that will never happen

flint idol Dec 23, 2024, 1:03 AM

#

wait what does TRAP() do

#

TrapOp

#

saw that in WAK and was curious

real pecan Dec 23, 2024, 1:03 AM

#

Thats just some method

#

TrapOp is not a thing

flint idol Dec 23, 2024, 1:04 AM

#

ah

#

it sounded like something to trap into the kernel or smth

real pecan Dec 23, 2024, 1:04 AM

#

Probably trap into SMM

#

But u can see implementation

#

But anyway I think its related to that if its not a bug because of something else

flint idol Dec 23, 2024, 1:05 AM

#

it's uacpi_context_set_loop_timeout(3)

real pecan Dec 23, 2024, 1:05 AM

#

Nvs is supposed to be preserved in s3, Linux suspend to ram impl might give u clues

#

I can take a look tomorrow as well

flint idol Dec 23, 2024, 1:05 AM

#

but I'll do that

real pecan Dec 23, 2024, 1:05 AM

#

flint idol it's `uacpi_context_set_loop_timeout(3)`

Yeah

#

Try that who knows

#

Or maybe your hpet stops counting

#

So it hangs

flint idol Dec 23, 2024, 1:06 AM

#

that wasn't it

flint idol Dec 23, 2024, 1:06 AM

#

real pecan Or maybe your hpet stops counting

wait

#

that could be it

real pecan Dec 23, 2024, 1:06 AM

#

Lmao

flint idol Dec 23, 2024, 1:06 AM

#

I completely forgor to restore hpet state

real pecan Dec 23, 2024, 1:07 AM

#

💀

flint idol Dec 23, 2024, 1:07 AM

#

if it isn't that, it's still a bug you just found for me

real pecan Dec 23, 2024, 1:07 AM

#

I highly doubt it keeps state after reset

thick jolt Dec 23, 2024, 1:10 AM

#

flint idol I completely forgor to restore hpet state

HOW COULD U FORGET THE HPET

#

😡

flint idol Dec 23, 2024, 1:10 AM

#

-# ignore

#

hmm should I save the entire HPET state

#OBOS (not vibecoded)