EvalynOS

[hazy saddle]

https://github.com/EvalynGoemer/blahajEngine/blob/main/src/engine/lua.h

This is the helper functions

[devout path]

xd the outcome of chaos is indeed stability

im impressed

[hazy saddle]

Just don’t fuck up and your ok

[devout path]

LMAO

INDEED

[hazy saddle]

If you do the same in C++ code it would crash the same

I’m letting Lua devs feel my pain

The engine is pretty complete for the most part in terms of adding new features for games

The rendering pipeline is what needs the most work

[devout path]

XD

gimme a moment my ethernet cable that's kidna fucked up aint' working properly, i broke the thing where it connects to LOL

[hazy saddle]

I got it to play a music video by loading every frame as a texture and swapping between them and loading them in and out of vram and then calling a system media player from Luas shell execution

[devout path]

ooh

[hazy saddle]

It was like a 2gb somthing game tho

So I’m not sharing it XD

https://github.com/EvalynGoemer/uservice

I also have this little thing that you give a sub set of systemd style commands to

And it will figure out your init ststem

And then execute the proper command for your system

So you don’t have to remember many different syntaces

[devout path]

oooh

[hazy saddle]

It has auto detection and a config file to override it

[devout path]

oh nice nice

[hazy saddle]

It can do all of this

[devout path]

daamnnn

[hazy saddle]

And it’s all a single C file of like 200 lines

But this stuff isn’t that impressive

Just some code I threw together and stuffs

https://tenor.com/view/noelle-sad-spongebob-sad-spongebob-crying-noelle-crying-gif-23355068

I just sit in my bedroom all day on my computer and make code and play video games

[devout path]

XD

eh i think it's just u being built diff

possibly u having fun making these

i alr got burned out

[hazy saddle]

I don’t constantly code and I get un burnt fast

So I am able to pace myself to keep going

And yeah it’s 8am with no sleep

https://tenor.com/view/jgmm-cat-meme-ragebait-cat-think-cat-ponder-gif-7034802297606826390

[devout path]

d o n o t

copy me LMAO

go sleep

i swear

u will get burned out like me even though u don't code

i mean even if you don't code

[hazy saddle]

I’ve been doing this for the last few weeks now

[devout path]

XD

[hazy saddle]

And I keep trying to fix it

[devout path]

nah just buy monster

[hazy saddle]

And then it returns

[hazy saddle]

I drank two two liters of Dr Pepper once

[devout path]

that is kinda not optimal XD

[hazy saddle]

I mean is any of my code “optimal”

A lot of my code is made at 3am working with no sleep

That’s why I also chose to live dangerously

[devout path]

i can now see a pattern LMAO

[hazy saddle]

Dangerously cheesy

[devout path]

your code says a lot about who you are

[hazy saddle]

DW I know how to make “safer” code and other stuff

Add the checks and other stuff etc

Don’t give Lua raw pointers but give handles

Etc etc

[devout path]

not safer code but like

the pattern of how it goes XD

[hazy saddle]

Tell me then

What kind of pattern / who am I

[devout path]

i mean like u can write safer code but chooses not to LMAO

that pattern

im operating on sub optimal braincells right now pardon if i say something out of the question

[hazy saddle]

It’s more fun on the edge and it’s more performant

And if stuff happens where it crashes something else already fucked up bad for the most part

[devout path]

then that tells a lot about u XD

[hazy saddle]

(For the kernel side)

I have no excuse for the LUA shit I’m pulling 😭

I should fix that

[hazy saddle]

I don’t get to do a lot irl though

I’m not allowed to leave my house much and I’m home schooled

[devout path]

oh zamn ;-;

[hazy saddle]

So this might just be how I get out my dangerous things since I can’t do anything fun irl

But like subconsciously

[devout path]

don't worry the outside world is just fucked as it is the same as the world we live in on the internet

[hazy saddle]

I’m gonna have to get a GED and shit

[devout path]

wat's GED?

as a malware analyst for quite a while now i can attest that the world we live in ain't kind as it seems

[hazy saddle]

High school diploma but not through high school

[devout path]

ooh

[hazy saddle]

It’s an “equivalent”

[devout path]

should be the same XD

[hazy saddle]

Also I have no ID 😭

My parents haven’t gotten it yet

So no job for me

[devout path]

i tihnk you received better education than the current ediucation that we live in to be honest

a lot of factors regarding how the external factors such as classmates, environment tells a lot that it's failing

and hwo they teach u

[hazy saddle]

I’ve been self learning everything

Because my parents do nothing XD

[devout path]

i resent the current education system but yea XD

your work would be the US equivalent of uuh

what's that called again

AP subjects

yes yes

[hazy saddle]

Honestly if Covid didn’t happen I would be still in school

[devout path]

oh damn

[hazy saddle]

My life would of ended up much different and I might not of discovered my self

[devout path]

yeaa ;-;

[hazy saddle]

But I should sleep because the keys on my phone keyboard look like they are rotating

These are uhhh not rated right

Rotated

[devout path]

wat da XD

i think i foudn the bug

[hazy saddle]

Oooo

[devout path]

we needa keep the interrupts disabled until the idle thread is fully done and at least one real kernel thread is linked

[hazy saddle]

Even though I am tired and seeing things my brain is still mostly at full capacity for programming

[devout path]

or don't shouldSchedule = 1 until inserted an actual thread

should i explain how it works XD

[devout path]

how to be like u

i envy u a little XD

[hazy saddle]

Ohhhhh shit right. I need to check if it was enabled first before doing the swapping logic

[devout path]

teach me ur ways

[devout path]

and it only took us a collaborative effort of an hour to debug XD

[hazy saddle]

Partial Mental insanity?

[devout path]

yes

XD

[hazy saddle]

The TTY should check if it is one before changing to zero and then one then

[devout path]

yea

[hazy saddle]

What about the cli and sti bug then

The qemu bug is fixed but not the physical one

[devout path]

the qemu?

wait that's the phyiscal bug fix XD

it's both a fix for it

try it

[hazy saddle]

Well true true

But the physical bug was from cli and sti

But I guess this has the same effect

[devout path]

ye so basiaclly in real hardware uuh

setup_idt turns interupts on which is sti while we're still in early boot

as pointed in the kernel.c

uuuuuuuh

aparently

wait where am i

ok so

setup_threading() immediately calls create_thread(null)

rihgt?

and it pushes null as the return address for the new thread and at the end sets shouldSchedule = 1

[hazy saddle]

But the tty code shouldn’t get executed until after the IDT no?

[hazy saddle]

Yeah that’s a work around for a big

If I only have one thread it executes

But if I have two

The first one never executes

[devout path]

let me confirm if that's tthe actual thing htats happening here

maybe im just drunk off monster energy

[hazy saddle]

I wonder if yall have a fix for this because I hate that hacky work arround ngl 😭

[devout path]

hmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmm

we shall see XD

[hazy saddle]

I’m gonna sleep

[devout path]

wait wait

before u do

[hazy saddle]

Il wake up and test the new TTY fix file the optimization

[devout path]

let me answer this

[hazy saddle]

Alright alright

[devout path]

oke oke

oke so

[hazy saddle]

And maybe yall can find the issue with the scheduler because that’s interesting too with the linked list

[devout path]

tty doesn't run until later but the kernel still hits the race long before that

[devout path]

oh yeah i found a fix for that while i was lookig at it

i was cross referencing to le amd64 funny book

[devout path]

uuh

the order inside kmain is

setup_idt() installs the idt and executes sti

so irqs become live!

[hazy saddle]

Which is what I would want right

[devout path]

uuuuuuuh

wait lemme read this again

im going to triple read this fuckin

uuuuh

timer interrupts start running yeh?

while create_thread(NULL) is still building the dummy thread

so the scheduler

switches to that thread and tries to execute a NULL poilt entry

point

would u like a comprehensive more non dumbed down explanation?

[hazy saddle]

And when the TTY renders that causes the stupid thingy issue because it sets it should shceduls

[devout path]

yes

[hazy saddle]

Is there a way to patch it to not need a null dummy thread

[devout path]

that's what im developing now

[hazy saddle]

Because that’s annoying and a waste of memory

Good good

[devout path]

since the 2 ppl that was helping me gave up MAO

i'll make a patch dw

this is fun learning x64 this way so

[hazy saddle]

And then once that gets PRd in to did the scheduler I can add my other patch to speed up the TTY

[devout path]

u see how this triggers the gpf right?

[hazy saddle]

Yup and the other PF

[devout path]

yeye

yuuuuuuup

my brain is fried

[hazy saddle]

But why the GPF when I’m not fucking with the scheduler stuff but just interupts being on or off?

[devout path]

i needa break LMAO

[hazy saddle]

I dunno

You should sleep and then do thissss

Because I lowkey got no idea how to fix that myself 💀

Otherwise I would not of bundled that hack

[devout path]

becauseee

[devout path]

i think it's because flipping interupts on letsthe PIT fire while the dummy thread still has NULL entry!

scheduler sees le shouldSchedue

it's set

and it ocntext switches to taht half built thread

[hazy saddle]

So it ends up falling through to the null entry because a race condition with some logic

[devout path]

iyes

the cpu raises the #GP

[hazy saddle]

You know I am gonna have to learn 32 bit at some point for a port on my kernel

[devout path]

XD DO NOT

[hazy saddle]

Maybe I can help with your kernel a bit XD

[devout path]

well u can look at my kernel

yippie !

tthank uou

[hazy saddle]

Maybe I can do a refactor to reorganize the files and shit

[devout path]

i'll use ur for my NUMA-aware research :DD

[hazy saddle]

A “Small” Refactor

[devout path]

https://github.com/Kawattenai/jimirOS

[devout path]

XDDDD

[devout path]

here :DD

i'll refactor the code once i'm ok

i've had too much caffeine to drink

[hazy saddle]

Yeah lmao

[devout path]

and i think it's my breaking point lMAO

[hazy saddle]

Is that libc for kernel or user

[devout path]

i fored myself to understand this

wai brb 2 mins

[hazy saddle]

This isn’t a bad structure for the kernel code

No git ignore

https://tenor.com/view/noelle-sad-spongebob-sad-spongebob-crying-noelle-crying-gif-23355068

[devout path]

oh it took me 2 hours to refactor

[hazy saddle]

Well it exists but not for binaries

[devout path]

XD my bad

[devout path]

yea

i forgot gitignore

[hazy saddle]

No kernel free

You can probably adapt the code I stole for the heap part atleast

As long as your PMM can free pages

[hazy saddle]

Just keep that licence header

[devout path]

yes yes ofc ofc

well ths is x32 so i might redesign it XD

[hazy saddle]

The PS2 code I took from there dosnt have that header because it was a smaller set and at that point I was in contact with the dev and they said to just put a small comment there

[devout path]

oooh

my memory management code is

suboptimal

LMAO

[hazy saddle]

The other files were a full port of the entire file lmao

I mean it’s a bitmap pmm so I could make a buddy allocator for ir at some point maybe

[devout path]

oh wait wat

oooh

[hazy saddle]

So that’s why I used the full header with all the fluff

[devout path]

yeah i wsas wondering about that XD

aight

go sleep XD

imma eat dinner

and do a little bit of planning on how to refactor your scheduler

and plus on point XD

im researching on schedulers

specifically on hint based topology aware schedulers

[viral bison]

huh?

[devout path]

ah yes don't worry too much about it

[hazy saddle]

We got a little off track on optimization and patching bugs lol

Talking about other kernels etc

[viral bison]

oh, lol

if you need help with schedulers, i might be able to help

[hazy saddle]

Btw @devout path thanks for all the help with this and il finnaly have a scheduler that dosnt need a stupid hack XD

[devout path]

you're welcome !

[viral bison]

i finally fixed mine a few days ago XD

[devout path]

i'm happy to help !

[hazy saddle]

Mine mostly works but just needs a null entry for some reason

So I added that but it’s causing other issues that @devout path found so they are gonna make a PR

[viral bison]

wdym null entry?

[devout path]

thank you too for enlightening me with x64 knowledge that i wouldn't be able to have the motivation to learn

[hazy saddle]

A task that has a super small stack and an IP of 0

[viral bison]

strange

[hazy saddle]

Because my schedular with more than one task refuses to schedular the first one on the linked list

For some reason

[devout path]

here

[hazy saddle]

[viral bison]

https://github.com/Diamantino-Op/HorizonOS/blob/main/src/kernel/kernel-common/src/threading/Scheduler.cpp
https://github.com/Diamantino-Op/HorizonOS/blob/main/src/kernel/kernel-x86_64/src/threading/SchedulerX86.cpp
https://github.com/Diamantino-Op/HorizonOS/blob/main/src/kernel/kernel-x86_64/src/threading/SchedulerX86.s

here if you wanna look at mine

[devout path]

nice

[viral bison]

also, this might help: https://github.com/Diamantino-Op/HorizonOS/blob/main/src/libs/hos-base/src/LinkedList.hpp

linked list that should work very well

using templates to make it easier to use

[devout path]

ooh thakns thanks

[hazy saddle]

Ooo a less common c++ kernel

[devout path]

@hazy saddle if you ever run into somethig that needs debugging come to me i will 100% debug it

[hazy saddle]

I use C

[devout path]

just refactor it to c :))

XD im joking

[viral bison]

i prefer c++ bc of templates, namespaces and classes

[hazy saddle]

I use C because of a meme song

[devout path]

i use c because it's apparently the industry standard (i don't believe it)

[hazy saddle]

And because for lower level stuff its “better”

[viral bison]

also, i have performance tracking on my OS too

[hazy saddle]

Ooo nice

[devout path]

    bool had_threads = (threads != NULL);

    struct thread* new_thread = malloc(sizeof(struct thread));
    memset(new_thread, 0, sizeof(struct thread));

    new_thread->stack = malloc(STACK_SIZE);
    memset(new_thread->stack, 0, STACK_SIZE);
    new_thread->stack_top = (void *)(((uintptr_t)new_thread->stack + STACK_SIZE) & ~0xFULL);

    new_thread->is_user_task = 0;

    uint64_t *stack = (uint64_t *)new_thread->stack_top;

    *--stack = (uint64_t)task_quit;
    if (entry_point == NULL) {
        entry_point = task_quit;
    }
    *--stack = (uint64_t)entry_point;
    *--stack = 0x202;
    for (int i = 0; i < 15; i++) {
        *--stack = 0x0 + i;
    }

    new_thread->rsp = (uint64_t)stack;

    struct thread_node* new_node = malloc(sizeof(struct thread_node));
    new_node->thread = new_thread;

    if (!had_threads) {
        new_node->next_thread = new_node;
        threads = new_node;
    } else {
        new_node->next_thread = threads->next_thread;
        threads->next_thread = new_node;
    }

    if (had_threads) {
        shouldSchedule = 1;
    }
}```

[viral bison]

so i can do flamegraphs

[devout path]

oh sorry didn't send in code block

[devout path]

@hazy saddle wouldn't this suffice?

[hazy saddle]

I assume this patches it to work without the null block and stuff?

[devout path]

yes

[viral bison]

i would suggest something btw: you create a kernel thread that is just idling, and set it as the currently executing thread

so you never have a null current thread

i create an idle thread per core

[hazy saddle]

I mean that null thread for some reason in most cases NEVER gets executed

[viral bison]

so each core will always have a thread

[devout path]

it stops the PIT from doing funny shiz while the dummy still has a NULL :)

wellllllllllll

heh

lemme see im kinda stupid rn

[viral bison]

are you disabling ints while switching?

[hazy saddle]

I dunno

[viral bison]

and also stopping the isr handler

[hazy saddle]

the other threads execute perfectly

as many as i add

just the first one i add breaks

[viral bison]

are you creating the idle thread the same as the other?

[hazy saddle]

this is a NULL thread not an idle thread

i dont have idle threads

[viral bison]

make an idle thread

that just spams pause

done

[devout path]

wouldn't that be

not optimal

[viral bison]

it is optimal

[hazy saddle]

i can make the null, thread be a idle thread

[viral bison]

or well, you can make it request to schedule again after a few ms

[viral bison]

yeah, that might cause problems

[hazy saddle]

code highlighting

void create_thread(void (*entry_point)(void*)) {
    bool had_threads = (threads != NULL);

    struct thread* new_thread = malloc(sizeof(struct thread));
    memset(new_thread, 0, sizeof(struct thread));

    new_thread->stack = malloc(STACK_SIZE);
    memset(new_thread->stack, 0, STACK_SIZE);
    new_thread->stack_top = (void *)(((uintptr_t)new_thread->stack + STACK_SIZE) & ~0xFULL);

    new_thread->is_user_task = 0;

    uint64_t *stack = (uint64_t *)new_thread->stack_top;

    *--stack = (uint64_t)task_quit;
    if (entry_point == NULL) {
        entry_point = task_quit;
    }
    *--stack = (uint64_t)entry_point;
    *--stack = 0x202;
    for (int i = 0; i < 15; i++) {
        *--stack = 0x0 + i;
    }

    new_thread->rsp = (uint64_t)stack;

    struct thread_node* new_node = malloc(sizeof(struct thread_node));
    new_node->thread = new_thread;

    if (!had_threads) {
        new_node->next_thread = new_node;
        threads = new_node;
    } else {
        new_node->next_thread = threads->next_thread;
        threads->next_thread = new_node;
    }

    if (had_threads) {
        shouldSchedule = 1;
    }
}

[hazy saddle]

diddnt until now

[viral bison]

and also is harder to do than just an idle thread

[devout path]

can u try this in ur scheduler.c :)

if u are still in ur

[hazy saddle]

i am rn

[devout path]

pc

oke did u try?

wat happend

[hazy saddle]

im looking at the diff rn

this might make thread exit when it shouldnt?

leme test

[devout path]

hmmmmmmmmmmmmmmmmmmmmmmmm

test it

in theory it should work

[hazy saddle]

we arent crashin

[devout path]

YOOOOOOOOO

[hazy saddle]

can i remove dummy thread maybe?

[viral bison]

also, you will need to add xsave and xload in the ctx switch too

[devout path]

maybe

i think

[viral bison]

if you are doing SIMD stuff

[hazy saddle]

only for user sitches

[viral bison]

no

[hazy saddle]

kernel should not use SIMD/MMX

[viral bison]

for all

[devout path]

theoretically.. it stops PIT from preempting while the dummy still has a null.. it only enables shouldSchedule once the run queue already has a thread

[viral bison]

you might have kernel threads too

and they do benefit a lot from simd

[devout path]

im only good at theory @hazy saddle pardon ;-;

[hazy saddle]

and then there is more context switch overhead

[viral bison]

ehm

that is very bad

[devout path]

XD

[hazy saddle]

if i remove the dummy thread it breaks

[devout path]

well

[viral bison]

also, why the PIT...

[hazy saddle]

lazy

[devout path]

^^ same reason

[viral bison]

it's so bad

[devout path]

if it works it works

[viral bison]

it works very bad

[devout path]

wat should a scheduler be like XD

[viral bison]

i use APIC to for scheduler

[hazy saddle]

this pit shouldnt make me need a dummy thread

[devout path]

hmm yeah makes sense

[hazy saddle]

deleting this makes me not need dummy thread it seewms?

(i mean adding comments)

[devout path]

most likely

[viral bison]

also, why it crashes?

what exception it throws at you?

[hazy saddle]

dosnt crash anymore

it was GPF and or PF

[devout path]

here

[hazy saddle]

wait i still need to test real hardware

[devout path]

okeoke

[viral bison]

that's why null threads are bad + you should do a nullcheck at the top

[hazy saddle]

nah we live on the edge

this is the diff

[devout path]

Test

Hi

My discord ain't working on my pc

What the sigma

[devout path]

XD

Yra

Yea*

[hazy saddle]

ok let me do a real hardware test\

[devout path]

Okeoke

Pray it doesn't gpf

I would cry

[hazy saddle]

even smaller

[viral bison]

9 AM

wow

6h behind me XD

[hazy saddle]

Uhhhh sooooo errrmmm @devout path

[devout path]

XD

[hazy saddle]

Your not gonna like this

[devout path]

IT GPF?!?!?

OMAGHGAH

WELP BACK TO DEBUGGING

[devout path]

HUUUUUUUUUH

[hazy saddle]

This is with the inline assembly

[devout path]

i think so

[viral bison]

have you tried kvm?

send me the iso a sec

[hazy saddle]

Let me try it with the should schedule

[viral bison]

i'll try to load it with kvm

[devout path]

When it runs on qemu it's fine

[hazy saddle]

This is on real hardware. It works on KVM. And there is no ISO the QEMU uses the project folder as the mount

[devout path]

The script is built for kvm

THAT'S CONFUSINGG

Maybe it is a cpu bug...

[hazy saddle]

Let’s try it without the interupt shit

[devout path]

Oke

[hazy saddle]

And shit disabling and setting should schedule

[viral bison]

maybe it's the cpu?

[devout path]

Absolute computer science work here

[viral bison]

lets' try on multiple kvm devices

and see

[hazy saddle]

Download my repo and il send a patch file

Just a moment

[viral bison]

can't you just send the iso?

[hazy saddle]

There is no ISO

[viral bison]

...

[devout path]

Build -> fail-> refactor until we win

[viral bison]

how do you launch it with kvm then?

[hazy saddle]

I tell QEMU to use the project folder as a file system to mount and the bootloader can read from
There

[viral bison]

which bootloader?

[hazy saddle]

limine

[devout path]

limine

[hazy saddle]

    qemu-system-x86_64 \
        -machine q35,accel=kvm \
        -cpu host \
        -m 512M \
        -drive if=pflash,format=raw,readonly=on,file=./OVMF_CODE.4m.fd \
        -drive if=pflash,format=raw,readonly=on,file=./OVMF_VARS.4m.fd \
        -drive format=raw,file=fat:rw:. \
        -boot d \
        -audiodev pa,id=speaker -machine pcspk-audiodev=speaker

[viral bison]

how does it work without iso...

[hazy saddle]

like that

[devout path]

so it's simulating a fat file system yeah?

[hazy saddle]

yup

[viral bison]

i use these flags, lol

[hazy saddle]

thats the patch

[viral bison]

how do i even use patch files?

[hazy saddle]

its a git thing

[devout path]

yeah

[hazy saddle]

git should let you import it

[devout path]

git apply

[hazy saddle]

cat patch.patch | git apply i think

[devout path]

oh i only run git apply <patchfile>

and it works

[viral bison]

it works

the os i mean

[devout path]

@hazy saddle i will be back in an hour tryna figure out hwo to make this scheduler

[hazy saddle]

in qemu yes

[devout path]

expect me to fail so correct me XD

[hazy saddle]

not on real hardware with that patch

[devout path]

but i will be here chatting XD

while i code

[viral bison]

that's so strange

[hazy saddle]

jank

[viral bison]

use flanterm...

[devout path]

heh

[hazy saddle]

no

[devout path]

heh

LMAO

indeed no

[hazy saddle]

i diddnt know flanterm existed

[viral bison]

...

[hazy saddle]

then i made my own think

[viral bison]

lol

[hazy saddle]

and i dont wana admit defeat

[devout path]

well i mean it's quite fun building an os and having stuff completed

but i find more fun tryna find bugs like these

[viral bison]

gl testing on your pc without printing all the regs

[devout path]

XD

[hazy saddle]

i coukld if the panic hadnt paniced

[devout path]

u feel me @hazy saddle

xd

[viral bison]

bruh

how can the panic panic

[hazy saddle]

it uses the TTY

which is whats causing panics

[viral bison]

WHY

[hazy saddle]

because when i did it without it my code was a mess

[viral bison]

...

[hazy saddle]

its fineeeeeeee

It could still panic even without using the TTY

Some fucking how

@devout path we aren’t panicking yet

[devout path]

AYOOOOOOOO

NICE NICE

WE'RE GETTING SOMEWHERE

[hazy saddle]

Bad Apple is running

[devout path]

AYOOOOOOOOOOOOOOOO

[hazy saddle]

It’s faster

[devout path]

the architecture is indeed questionable but it works :D

[hazy saddle]

But it also fucks up the bad apple demo

[devout path]

oh damn

[hazy saddle]

Fuck it what happens if I grab flanterm

[devout path]

XD

[hazy saddle]

Will it be faster

[viral bison]

yes

• support all bash chars

and has it's own allocator

i use flanterm + nanoprintf

[hazy saddle]

Damn nice

[viral bison]

work very good

[hazy saddle]

What font is it gonna use lmao

[viral bison]

https://github.com/Diamantino-Op/HorizonOS/blob/main/src/kernel/kernel-common/src/Terminal.cpp

[viral bison]

i think you can choose

but it should have a default one

[devout path]

welp

she gave up XD

[hazy saddle]

Add flanterm as a git sub module maybe

[devout path]

ngl flanterm is good

[viral bison]

also, i highly recommend nanoprintf

very easy to implement

[hazy saddle]

Hey atleast I can say I did make a working terminal myself

And then I take the better one for realisies

@viral bison @devout path should I have an external folder where I put all the sub modules in maybe?

Like it will be /src/kernel/external/flanterm

[viral bison]

maybe

[hazy saddle]

And then if I need to I can make symlinks that go there from other places

[viral bison]

i just make CMake get them

and also update

[hazy saddle]

I use regular make files

[viral bison]

and use github to autobuild llvm for me

so i can download the binaries easily

[devout path]

yeah

u can always just add that to ur make file

[hazy saddle]

It should auto get globbed by the make file when it’s there

[devout path]

oh yeah

yes yes

[hazy saddle]

added to credits and its sub moduled in

uacpi can also go in that folder too

and nanoprintf

[devout path]

ooh

[hazy saddle]

everyone deserves credit :3

even stackoverflow

[devout path]

XD

which tells me

i forgot to credit limine on my os

oh wait nvm i haven't published that

im using grub

ok im stupid

:D

[hazy saddle]

oh wait both of you if you want @devout path and @viral bison can be in my credits under the good friends section XD

[devout path]

oke oke :DD

[hazy saddle]

I only credit the boiler plate and not the library usage

[devout path]

will u keep ur current scheduler or would u like me to make an optimal scheduler?

[hazy saddle]

I can probably add the full limine to the credits which u could

[devout path]

i think i already have a good idea

[devout path]

ooh

[hazy saddle]

I am probsly gonna fix it and improve it later

It dosnt even support multiple user threads yet

[devout path]

i'll put suggestions in pull requests in git :D

[hazy saddle]

maybe mark as an issue if it dosnt have code to merge but just ideas?

[devout path]

oke oke

i'll put code too dw

so i can practice

[viral bison]

user threads are the same as kernel ones

you just need to trampoline them into usermode first

[hazy saddle]

i know but they need extra context swithcing logic for see registerd

[viral bison]

no

only if you enable simd

[hazy saddle]

well if a thread has simd then i have to fo thoes

and user threads will have simd enabled

and it is a kernel thread that tromplines

[viral bison]

you can't enable simd just for 1 thread tho

..

why

[hazy saddle]

the context saving and switching

other threads arent meant to use them

[viral bison]

you just need to trampoline once i think

per thread

[hazy saddle]

once per new thread

[viral bison]

yeah

i have a function is asm specifically for that

instead of setting the RIP of the new thread to the program code, i set the RIP to the trampoline, and pass the program RIP as a second parameter

[hazy saddle]

it seems nanoprintf is header only so i wont submoudle it

but it still goes into the external folder because its not being modified

[viral bison]

you can look at my Terminal.cpp for how to use it

[hazy saddle]

hopefuly symlinks work with git

@devout path this looks good

[devout path]

oooh nice nice

[hazy saddle]

shit im gonna have to rethink this

https://tenor.com/view/technical-difficulties-tenna-deltarune-meme-undertale-gif-4437984343073759770

[devout path]

XD

take ur time

[hazy saddle]

i dont think sub modules are right for this exactly

[devout path]

oh damn

welp

back to the drawing board it seems

[hazy saddle]

nah ive already got it

and im adding eveyrthing to creddits

i can use this

[devout path]

ooh nice nice

wat os are u using

[hazy saddle]

cachyos

[devout path]

wait wa

no that was not the question

wait what

[hazy saddle]

arch derivitive

linux

[devout path]

ohh

ohhhhhhhhhhhhhhhhhhhhhhh

ok nvm i thought i asked wrongly

nvm i thought i asked what os are u making

my eyes aer deceiving me now

oke oke

[hazy saddle]

@devout path jusr a few more patches and we are good

Holy shit it’s so much faster

I do not regret this

Making a printf and terminal shit is kind of an ass

[devout path]

ooh niceee

:000

[hazy saddle]

im gonna end up refactoring alot

another "Small Refactor"

and usinf flanterm i can restore the other print messages

[viral bison]

and also move the cursor btw

[hazy saddle]

it dose

when you press enter

it goes to the right place when the shell starts

[viral bison]

no i mean backwards and left/right

[hazy saddle]

oh

im too lazy for that right now

[devout path]

XD

[hazy saddle]

it can print this properly now

with nanoprintf

[viral bison]

yep, so good

[hazy saddle]

I’m just moving around every single header file and re structuring it

I might also change the make file up a bit

“Small” things

Like also completely changing the tty and printf backend

Well il call it “Small” Refactor 2: Electric Boogaloo

[devout path]

xdddd

totally small

[hazy saddle]

And then the next will be “Small” Refactor 3: Tokyo Drift

Holy this shit is instant on physical hardware

HOW

[viral bison]

actual physical gpu

[hazy saddle]

Wait what

Isn’t it still going through the same limine frame buffer

[viral bison]

and you have the full speed of the cpu

instead of it being virtualised

[hazy saddle]

But my implementation was slow as shit

[viral bison]

poor code

[hazy saddle]

I mean i would write every character to the screen weather null or not etc

But how dose this reach this kind of insane speeds

Jesus Christ

[devout path]

XDD

niceee

ooh

[hazy saddle]

wayyyy better

now to di this everywehrr

[devout path]

ayeee

[devout path]

go sleep ;-;

imma sleep now

[hazy saddle]

no

[devout path]

XD

i nneda sleep XD

[hazy saddle]

im gonna finisht he headers

[devout path]

i haven't worked on the os yet cuz i have band rehearsals tmr

[hazy saddle]

then tomorrow reorganise

[devout path]

welp good luck with that XD

[hazy saddle]

and then redo VFS

and then commit

the big beautiful commit /s

[devout path]

xd

[hazy saddle]

im also gonna add an IDE setup readme for KATE

[hazy saddle]

Alright @devout path

And this isn’t even half way done

This is what I have left to do

I am gonna eat and then sleep

[viral bison]

also a tip: you can try CLion

it's free for community use

[devout path]

Nice

[hazy saddle]

a much much MUCH better panic is almost done

[hazy saddle]

back to the old 8x8 font and trying to find the best ascii art fot this

[gentle quest]

me when i want to read some info about the crash but half the screen is occupied with drawings

[hazy saddle]

there is enough room for it even on the smaller screen so it should be fineeee

this one might be a little too bgi

[marble ridge]

lol

[hazy saddle]

i think this should do

this is the new finished kernel panic using flanterm + nanoprintf

[viral bison]

for error you can just use a number and not an hex btw

[hazy saddle]

I’m printing it as binary so I can get the bit flags and stuff

[hazy saddle]

added some colors to make it easier to read the seperate sections

may tweak them later

[hazy saddle]

@marble ridge I might of found an issue with flanterm

I’m trying to use my 8x8 font with it and it works in emulator but on real hardware I get a TF and by changing the font to NULL it will then work on real hardware

I have tried

• hard coding the font instead of generating it from my old font (endianness needed to be changed, it might be a nice option to be able to have flanterm flip the font too)
• adding padding to the font so it takes up 256 glyphs which is what flanterm seems to want
• diffrent ways of giving the pointer to flanterm

Sending code in a sec once my computer is back in linux

Also sorry about the ping

    ft_ctx = flanterm_fb_init(
        NULL, NULL,
        framebuffer_request.response->framebuffers[0]->address,
        framebuffer_request.response->framebuffers[0]->width,
        framebuffer_request.response->framebuffers[0]->height,
        framebuffer_request.response->framebuffers[0]->pitch,
        framebuffer_request.response->framebuffers[0]->red_mask_size,
        framebuffer_request.response->framebuffers[0]->red_mask_shift,
        framebuffer_request.response->framebuffers[0]->green_mask_size,
        framebuffer_request.response->framebuffers[0]->green_mask_shift,
        framebuffer_request.response->framebuffers[0]->blue_mask_size,
        framebuffer_request.response->framebuffers[0]->blue_mask_shift,
        NULL, NULL, NULL, NULL, NULL, NULL, NULL, (void*)font8x8_basic_ft,
        8, 8, 1, 0, 0, 0
    );

static const uint8_t  __attribute__((unused)) font8x8_basic_ft[] = {
    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00 ,
    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00 ,
    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00 ,
    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00 ,
    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00 ,
    0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00 ,

[gentle quest]

i think oberrow ran into that

the fix was increasing some buffer or something idk

[hazy saddle]

What buffer should I increase? Also my font data is smaller than the default font and should use less data?

[hazy saddle]

Aha increasing the bump allocation limit seems to of fixed it! I will do some tweaking to find an optimal value for my kernel

Thank you

(Stack trace failed because this was built without symbols)

[hazy saddle]

cleaned up the dependancies section and got patching support to patch dependancoes

[marble ridge]

what's a TF?

[hazy saddle]

Triple fault

[marble ridge]

ah

[hazy saddle]

I fixed it by increasing the size of the bump allocator

[marble ridge]

i am fairly sure that was fixed? or was it not?

[hazy saddle]

Is a bit annoying to have to do that with the patch file and stuffs

[hazy saddle]

Nope

[marble ridge]

interesting

thanks for letting me know

[hazy saddle]

Np, could just increase the default size to be a bit bigger just incase

Or let you specify it with a #define before including it

Because afaik you have to edit the .C file as it’s not defined in header file

-#define FLANTERM_FB_BUMP_ALLOC_POOL_SIZE 873000
+#define FLANTERM_FB_BUMP_ALLOC_POOL_SIZE 0x180000

this is the patch i made

[hazy saddle]

got sub modules working properly now instead of the get deps script

[marble ridge]

thing is: it shouldn't really crash

have you actually checked that the init function didn't return NULL because it failed to allocate?

[hazy saddle]

i have not but i am not using malloc or free with it? it should be using the internal bump allocator no?

[marble ridge]

yes but as you figured, the internal bump allocator can run out of space :p

[hazy saddle]

oh right but what would i do when it runs out of space?

i can catch it from TFing but then there is no tty? or i can deley it to where malloc works

[marble ridge]

you can't do anything

well, there is something Flanterm can do

and it does do already, but only for big screens, not "the font is so small that it takes too much space"

that being limiting how big the Flanterm box will be

so that is something we can do, that, or obviously using your own allocator

[hazy saddle]

I’ve just implemented my patching system so it can automatically increase the bump allocated space l. I can use my own allocator but I don’t wana setup a ring buffer to store a few prints before that gets setup

[hazy saddle]

I did not realize that a smaller font would need more memory XD I’m dumb

Once nice thing would be to be able to define that define from my own code and not having to patch it myself

[gentle quest]

well oberrow fixed it by increasing the size, other than that not sure

[hazy saddle]

The bump allocator is nice for being able to print before that’s out or if my allocator dies some how

[hazy saddle]

That’s the same way I fixed it too

[devout path]

ooh

[hazy saddle]

This is the complete version

[devout path]

XD nice

[hazy saddle]

Wana know what I based the ascii art off of

Oh and I got sub modules working

And patching of sub modules is supported

[devout path]

yes'

[hazy saddle]

That

[devout path]

XDD

[devout path]

noice

fell asleep last night so i couldn't reply

took a short nap of 14 hours

[hazy saddle]

Sameeee

I’m gonna work on some more and sleep again

I think all I have left is the VFS rework

Having it all in one list will make ls be way easier to implement too

[devout path]

yeaaah

[hazy saddle]

I mean even if it’s linear it’s still a VFS with the read write function indirection

And caching can be done easily

If a file is searched for a lot put it to the cache that gets searched first

And I can cache ls operations

I could even use a hash map for the cache so that would be really fucking fast?

It gives me a lot of flexibility to make things be super simple but still be possible to be fast

[devout path]

yup yup

[devout path]

hmmm

wat if a balanced binary search tree?

oh wait nvm

[hazy saddle]

also makes freeing things eassier because its one list

And the cache is just pointers

[devout path]

yeah true

less complex = less of a headache

[hazy saddle]

@devout path so much simpler

now to make things register properly

[hazy saddle]

i think mega refactor is complete

except the makefile ugh

[hazy saddle]

https://github.com/EvalynGoemer/evalynOS/commit/98270298d97b707102a849b76f51c705a761b0a5

and we are LIVEEE

i forgot to edit the readme about the get deps script 💀

it dose it automaticaly on make with the sub modules

well next commit with page table switching can fix that

[devout path]

nicee ;000

[hazy saddle]

Your impressed

[devout path]

yes

i ahven't done any os dev today so xd

[hazy saddle]

I’m gonna take a break and then work on the schedular

My friend can’t get theirs to work so they are gonna steal mine once it’s done

[devout path]

XD

how are u oging to design the scheduler @hazy saddle

a simple roundrobin scheduler?

[hazy saddle]

It’s already round robin

[devout path]

oh im stupid

i completely forgot

[hazy saddle]

But I need to extend support for user tasks better

[devout path]

o oke oke

[hazy saddle]

And there is no such thing as sleep

So I might extend it to do that too have have awake timers

[devout path]

im tryna make a scheduler that implements redblack trees

[devout path]

XD

[devout path]

o nice nice

[hazy saddle]

There is also a sys call to change the PIT speed

And to reset the PIT timer

Because the bad apple program expects that functions

[devout path]

h uh

oh well fair XD

[hazy saddle]

I could maybe implement that safer and changing the speed and then putting it back after coming to another thread

Or just add sleep

[devout path]

just

add sleep XD

cuz

[hazy saddle]

Also my kernel has a sleep thread

It’s the kernel main

That has a halt loop

[devout path]

oh

[hazy saddle]

So when there is nothing to schedule it should? Drop back to there and sleep

[devout path]

ooh

that's a hacky solution for it

[hazy saddle]

A hacky solution I did not even come up with

It just happened naturally XD

[devout path]

nice :D

[hazy saddle]

At that point is it even hacky?

[devout path]

oh nvm XD

fair point

[devout path]

ima take a good read at his

this*

[hazy saddle]

Look at the regular source tree tbh

So much as changed

[devout path]

wat da sigma

and u did this in a day ? crazt

crazy

[hazy saddle]

Well with sleep depravation yes

[devout path]

XD

daamn

go slep or somethin

[hazy saddle]

I’m staying up to 8pm

To get regular schedule

[devout path]

i slept a good 14 hours earlier

xd

[devout path]

fair point

[hazy saddle]

Was this really a lot of work tho XD

Just some refactoring of stuff

[devout path]

im just lazy iz all XD

my brain wouldn't comprehend a shit ton of refactoring

that's why i plan in advance

[hazy saddle]

It’s just shuffling some stuff around and fixing up systems

[hazy saddle]

I planned alot of this in advance

[devout path]

ooh

[devout path]

dw about me im that lazy XD

or well i just compliment alot

[hazy saddle]

@winged basalt moving over here

[winged basalt]

👍🏻👍🏻👍🏻

I've started building System level tools and utilities in C to get started for my own OS project

It'll take a few months but definitely worth it 😊

[hazy saddle]

Small update: the day after tomorrow I should have some more time to start working on things

@warped ermine because you wanted the link too il also ping you here

Just incase

[warped ermine]

hai

:3

[hazy saddle]

:3

[hazy saddle]

I do need to put stack trace symbols into the initramfs so I don’t have to compile twice for builds

That should be part of this next update

[hazy saddle]

i have now gotten a basic serial port driver working finnaly

it will not take in input as thats still through PS/2 but il be making a /dev/kbd to glob all the keyboard inputs together along with a /dev/term/serial for outputting to the serial device directly and for now printf and /dev/term/tty will goto serial

Reading from /dev/term/tty will give keyboard input from there and writing will output there of course

I should maybe reroute reads to /dev/term/tty to that new /dev/kbd for inputs

I know this isn’t the new userspace stuff but I’m working on another secret project that needs me to document the serial port and I need quick serotonin by making something work

oh wait i diddnt even call my init function and it still works XD

there we go. it now properly calls it and code respects checking if it properly was setup before using it just incase

[hazy saddle]

serial port now acts as a keybaord!

and it works in the shell thing too with qemu

[devout path]

ooh noice noice

[hazy saddle]

aswell serial and ps2m input get globbed into /dev/kbd

it will cause desyncs with other buffers but if they all use one its all good

and when i have open and close il flush the buffer when opening to fix that

also wait @devout path im commenting my code

why would i ever do this

[hazy saddle]

@devout path guess what. Tomorrow when I wakeup if I have the energy I am going to get multi app userspace working and then start being able to port a ton of fun stuff

And then it’s gonna be like

And then il probably add signals pipes forks other stuff after doom and basic other stuff perhaps

[devout path]

wai wha

[devout path]

LMAOOO

nice nice

[hazy saddle]

I try to avoid magic numbers and funny stuff in code so I don’t have to comment it a lot

But this is just a ton of magic numbers go brrrr

[devout path]

ooh

How's the progress been?

[hazy saddle]

None

I’ve been a bit busy with other projects

I am starting to get the scratch to work on it more

And the code base has been for the most part fully cleaned up so should go smoothly

And then I’m gonna be porting a libc + doom and core utils

Maybe il get multi frame buffer (monitor) support and SMP later?

So I could have two instances of doom at once

And once I get USB I could have one computer two people playing doom multiplayer maybe with a crude TCP/IP that’s loopback only?

Could be neat

Hey loop back networking could be something my kernel has and windows dosnt

(They broke it in one of the updates XD)

[devout path]

aww

i've been tryna learn zig xd

sorry late reply i slept the whole day

ooh nice nice

[devout path]

xddd

[hazy saddle]

@devout path I think I’ve got page cloning half way done

This code dosnt crash the system

So really fucking good

And it should be creating the top level dummy pages

[devout path]

uuuuuuuuuuuuuuuuuuuuuh

[hazy saddle]

So when I go and clone them it should all work

That get next level function also allocates with the true flag set

So that’s why it looks kinda stoopid

But this should mean the top 256 parts of the PML4 should be allocated

And they won’t get deallocated

[devout path]

oooh fair fair

should work in theory yes yes

[hazy saddle]

Next should just be a memcpy of the PML4 I think

[devout path]

yes yes

[hazy saddle]

@devout path hows this look

Wait I did malloc the wrong size

thats better

This is using services from the VMM but they shouldn’t be used until after init so it’s fine

The service being malloc

[sudden mirage]

Let me tell you that I've did this and lost a lot of precious code I wasn't able to find again :<

[hazy saddle]

my old kernel was lost this way

my kernel was originaly remaking that one

[sudden mirage]

I also lost my old kernel like this

Since then I started making a temporary directory for performing any git operation besides add/commit/push

[hazy saddle]

could prolly automate that

[devout path]

i uuh learned my lesson. before i put my code in a repo, i deleted my wsl ubuntu in exchange for actually booting ubuntu yea? then what happened is that i forgot i had progress prior to pushing the repo cuz i had a stable release at that time and then ... heh...

very sad moments in life

[devout path]

yes yes yes

this looks perfect

[hazy saddle]

im reallyt nervous here haha

[devout path]

wait why xd

[hazy saddle]

because its looking like its going good

[devout path]

hmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmm

[hazy saddle]

and im not gona have to bang my head against the wall

and thats

a lil sus

[devout path]

try checking if it returns null first, new_pagemap == NULL

for redundancy

[devout path]

heh that's programming in a nutshell

[hazy saddle]

my create thread is gonna take in a page map

and if its null it will be the kernel page map

[devout path]

yeaaa

uuh do u have a lower half

[hazy saddle]

yeh yeh thats where my userspace is

i have code runnign in ring3 already

just no proper mapping switches etc

[devout path]

so the line copies the entire pm4 table right?

im lazy to check so lemme ask u

[hazy saddle]

which line

[devout path]

it won't completely start a new process with a completely empty userspace no?

u didn't map it to userspace, the kernel pagemap i mean

[hazy saddle]

cloning happens from the root kernel mappings

which will have nothin down there

eg this

[devout path]

dya use recursive paging?

[devout path]

oh

oke oke

[hazy saddle]

the kernel page map is the clean one made by the kernel pre userspace

so i can use that easi;ly

[hazy saddle]

whats that

[devout path]

where the last entry points to pml4 itself

oke i guess u don't

anyways

[hazy saddle]

what would that do?

i never understood it

[devout path]

i think u zero the page first for redundancy just to be sure and ur fine

i dont mind

i mean i think it's all good

what the hell came out of my keyboard im sleepy as hell

[devout path]

ah

kinda avoids the which comes first problem

of needing page table entries

[hazy saddle]

?????

it dosnt have the type

its in the header file

and its included

[devout path]

heh???

[hazy saddle]

oh i forgor the name

[devout path]

oh

[hazy saddle]

XD

[devout path]

OH

YEAH that's kinda funnny XD

[hazy saddle]

WE ARENT CRASHING

[devout path]

LES FUCKIN GOOOOOOOOOOO

[hazy saddle]

Now I would want to make some test programs that run in the same virt but diff phys