The Flopperating System (tfos) - x86 32 bit Hobby OS | Operating System Development | Page 3

crimson plover May 23, 2025, 7:53 PM

#

: )

bold grove May 23, 2025, 8:15 PM

#

crimson plover : )

I feel like I’m losing my mind

crimson plover May 23, 2025, 8:15 PM

#

rip

#

dont let it run too far

bold grove May 23, 2025, 8:17 PM

#

I’ll send you a paste bin in a bit

#

I have that assembly function

#

I’m just

#

Idk

#

Bro :(

#

Now it’s just stuck crashing

#

Like in a loop

#

Not even —no-shutdown does anything

#

and it keeps pointing to garbage in my kernel just random points

crimson plover May 23, 2025, 8:33 PM

#

lmk when u can vc lol

bold grove May 23, 2025, 8:37 PM

#

Yeah shouldn’t be too long

bold grove May 23, 2025, 8:56 PM

#

okay well i at least got that crashing issue fixed

#

@crimson plover look at this

#

thread_t* thread_create(int (*fn)(void*), void* arg, uint32_t* stack) {
    if (!stack) {
        stack = kmalloc(STACK_SIZE);
        stack = stack + (STACK_SIZE / sizeof(uint32_t));
    }

    thread_t* t = thread_alloc();

    *--stack = 0x200;
    *--stack = 0;
    *--stack = 0;
    *--stack = 0;
    *--stack = 0;
    *--stack = (uint32_t)arg;
    *--stack = (uint32_t)fn;
    *--stack = (uint32_t)_thread_entry;

    t->esp = (uint32_t)stack;

    thread_enqueue(t);
    return t;
}

crimson plover May 23, 2025, 9:06 PM

#

idk what to make of it

rigid root May 24, 2025, 12:17 AM

#

you should really check those allocations for null

bold grove May 24, 2025, 3:22 AM

#

rigid root you should really check those allocations for null

The allocations work but I’ll add some asserts

crimson plover May 24, 2025, 7:36 AM

#

ngl I just panic inside of my pmm if I run out of mem

#

lol

#

galaxybrain

#

future me will hate me

primal nymph May 24, 2025, 8:06 AM

#

btw @bold grove i had some similar crash, it's better if you do that in asm

#

the problem was setting the stack value in c++

#

this is how i ended up doing it

bold grove May 24, 2025, 11:25 PM

#

primal nymph the problem was setting the stack value in c++

Yeah my issue was setting the stack value in C. I did it in assembly and it works now. This was my issue and you just saved my life

primal nymph May 24, 2025, 11:26 PM

#

bold grove Yeah my issue was setting the stack value in C. I did it in assembly and it work...

i just fixed my scheduler too XD

#

i was setting the next entry of a linked list to itself

#

and it took me 2 days to find why

bold grove May 24, 2025, 11:30 PM

#

@primal nymph this is the issue I was having for over a month

#

Bexause I’m just

#

galaxybrain

primal nymph May 24, 2025, 11:30 PM

#

lol

#

btw, how does a thread know when it should be stopped?

#

like, do you have to manually set it to terminated state?

bold grove May 24, 2025, 11:32 PM

#

primal nymph like, do you have to manually set it to terminated state?

Yeah

primal nymph May 24, 2025, 11:32 PM

#

ah f

bold grove May 24, 2025, 11:32 PM

#

Wait no

#

thread_kill() does that itself

primal nymph May 24, 2025, 11:32 PM

#

syscall?

bold grove May 24, 2025, 11:33 PM

#

Nah not for now but I’ll make it one

primal nymph May 24, 2025, 11:33 PM

#

#

this is how i do it

#

i could easily not even set it to TERMINATED

#

bc i just call schedule again

bold grove May 24, 2025, 11:38 PM

#

primal nymph bc i just call schedule again

Ahhh I see

#

Well

#

I can get into the initialized state of the scheduler but

#

Idk

#

I’ll send a paste bin later

#

Maybe you could
Help me out

primal nymph May 24, 2025, 11:39 PM

#

k

bold grove May 30, 2025, 5:34 PM

#

Damn boys

#

I haven’t

#

Updated this in a while

#

I’ve been working on the sch*^#ler

#

Schedulers are cancer. Not really but , you know

#

My actual algorithm is fine

#

I use a doubly linked list

#

For thread queues

#

And I can queue threads

#

But

#

Context switching is just fucky

primal nymph May 30, 2025, 5:39 PM

#

hmm

#

Common Scheduler:
https://git.voxelgames.eu/diamantino/horizonos/src/kernel/kernel-common/src/threading/Scheduler.cpp
https://git.voxelgames.eu/diamantino/horizonos/src/kernel/kernel-common/src/threading/Scheduler.hpp

X86_64 Arch Scheduler:
https://git.voxelgames.eu/diamantino/horizonos/src/kernel/kernel-x86_64/src/threading/SchedulerX86.cpp
https://git.voxelgames.eu/diamantino/horizonos/src/kernel/kernel-x86_64/src/threading/SchedulerX86.hpp

X86_64 Ctx Switch and stack init: https://git.voxelgames.eu/diamantino/horizonos/src/kernel/kernel-x86_64/src/threading/SchedulerX86.s

#

if you wanna look at mine

primal nymph May 30, 2025, 6:06 PM

#

@bold grove

#

and if you need help send the code

bold grove May 30, 2025, 6:06 PM

#

yeah im looking sorry

#

i will soon

#

take alook

primal nymph May 30, 2025, 6:06 PM

#

wait, i see you have vscode open

#

i can join you if you want

bold grove May 30, 2025, 8:09 PM

#

primal nymph i can join you if you want

Yeah let’s talk later but

#

I’m like pretty close to get Rid of the issue

primal nymph May 30, 2025, 8:09 PM

#

gg

bold grove May 30, 2025, 8:10 PM

#

Bro

#

It was bexause

#

I was calling a c function

#

Which fucked up the value of esp

primal nymph May 30, 2025, 8:10 PM

#

XD

#

how

bold grove May 30, 2025, 8:10 PM

#

From assembly

primal nymph May 30, 2025, 8:11 PM

#

i do that too

#

oh wait no

#

i removed that bc it gave me problems too

bold grove May 30, 2025, 8:11 PM

#

it was _thread_confirm_esp (uint32_t *esp)

#

And when I passed eax

primal nymph May 30, 2025, 8:11 PM

#

lol

bold grove May 30, 2025, 8:11 PM

#

It like

#

Fucked up somefhing

#

Giving me a bullshit ESP

#

and eventually my instruction pointer was literally nowehere

#

The instruction pointer was 00FF0077 then a bunch of weird characters

#

Qemu info registers was

#

Just

#

00000000

primal nymph May 30, 2025, 8:12 PM

#

XD

#

well, at least you figured it out

bold grove May 30, 2025, 8:13 PM

#

Yeah

#

Now

#

I can call _thread_entry

#

And yield back to sched

#

But

#

Once I switch tasks like

#

Eventually the esp gets fucked yo again

#

Like the offset is fucked up like

#

I’m pushing the correct offset then I check it in C and it’s wrong

primal nymph May 30, 2025, 8:14 PM

#

are you saving it correctly tho?

bold grove May 30, 2025, 8:14 PM

#

Yes

#

I followed literally

#

The intel manual

#

And I’m saving it correctly

#

I think it’s coming down to an issue with my allocator which is funny

#

Okay not my allocator but

#

The slab is just

#

Being weird

primal nymph May 30, 2025, 8:15 PM

#

why slab

bold grove May 30, 2025, 8:15 PM

#

I lowkey might just

bold grove May 30, 2025, 8:15 PM

#

primal nymph why slab

I fucked off offset somewhere and it’s causing issues

primal nymph May 30, 2025, 8:15 PM

#

slab might be one of the worst ones for threads

bold grove May 30, 2025, 8:15 PM

#

Yeah I kinda like realized that

#

Now I’m just doing

#

Vmm_malloc

primal nymph May 30, 2025, 8:16 PM

#

are you also sharing kernel pages?

bold grove May 30, 2025, 8:19 PM

#

primal nymph are you also sharing kernel pages?

No

primal nymph May 30, 2025, 8:19 PM

#

you should

bold grove May 30, 2025, 8:19 PM

#

I’m making a region for threads

#

Is that bad

primal nymph May 30, 2025, 8:19 PM

#

yeah prob

#

i made my allocator using contexts

bold grove May 30, 2025, 8:19 PM

#

Okay I’ll just map to kernel pages instead

primal nymph May 30, 2025, 8:19 PM

#

so every process gets it's own context

bold grove May 30, 2025, 8:19 PM

#

Yeah

primal nymph May 30, 2025, 8:19 PM

#

process, not thread btw

#

doing one per thread might be bad

bold grove May 30, 2025, 8:22 PM

#

primal nymph doing one per thread might be bad

Yeah no I’m not doing that

#

Multiple threads can make up a process and they share the same address space

bold grove Jun 1, 2025, 9:13 PM

#

I gotta update this damn repo fuck

#

Sorry guys I’ve just been working on memorizing this stupid page of tchaikovsky

primal nymph Jun 1, 2025, 9:43 PM

#

bold grove Sorry guys I’ve just been working on memorizing this stupid page of tchaikovsky

wtf is this mess

bold grove Jun 1, 2025, 9:44 PM

#

primal nymph wtf is this mess

Classical music.

primal nymph Jun 1, 2025, 9:44 PM

#

pain

bold grove Jun 1, 2025, 9:44 PM

#

It’s what I’m doing for my senior violin recital

primal nymph Jun 1, 2025, 9:45 PM

#

lol, there is italian in there XD

#

that is so complicated tho

#

i only ever seen the simple ones XD

bold grove Jun 1, 2025, 10:04 PM

#

It’s all Italian

#

Nah it’s pretty bad

bold grove Jun 1, 2025, 10:04 PM

#

primal nymph that is so complicated tho

Yeah it’s a common “hard” violin piece that’s played with an orchestra

#

I would send a recording but I’m a butt ugly Bosnian so no thank you

primal nymph Jun 1, 2025, 10:04 PM

#

lol, dw

noble remnant Jun 2, 2025, 12:38 AM

#

bold grove Sorry guys I’ve just been working on memorizing this stupid page of tchaikovsky

I feel your pain

#

memorising music sucks

bold grove Jun 2, 2025, 1:23 AM

#

noble remnant memorising music sucks

Yeah I have 2 weeks to memorize 11 pages

#

Not even 3 in

bold grove Jun 2, 2025, 2:00 AM

#

average git commit

#

bold grove Jun 2, 2025, 5:36 AM

#

So the scheduler is done

bold grove Jun 2, 2025, 5:53 AM

#

https://github.com/amar454/floppaos/blob/main/task/sched.c

GitHub

floppaos/task/sched.c at main · amar454/floppaos

floppaOS is a free and open source 32 bit operating system made in C. It is themed around the famous internet Caracal "Big Floppa" - amar454/floppaos

bold grove Jun 2, 2025, 7:02 PM

#

#

yessss

bold grove Jun 2, 2025, 7:27 PM

#

well

#

now that this scheduler is done

#

i think i can restrcuture the source code

#

i might just name it

#

something else

#

idk

pliant oriole Jun 2, 2025, 9:22 PM

#

ayyy good job

bold grove Jun 2, 2025, 10:48 PM

#

okay im fixing some stuff

#

in the scheduler

#

some issies

#

arose

#

the function pointer inside of my thread_t was being pushed with the wrong offset

#

and i couldnt figure out where

#

so

#

#

but im fixing it

bold grove Jun 2, 2025, 11:09 PM

#

fucking hell

#

bold grove Jun 3, 2025, 2:46 AM

#

Aight

#

We are good

#

Time to restructure the source code

#

It’s gonna be

#

arch/
drivers/
fs/
mm/
sched/
userspace/
kernel/
flanterm/
multiboot/
limine/

#

Something actually reasonably

#

So imma have ia32

#

Which is what I’m doing now

#

But I’ll do arch separation

#

So I’ll get that limine c template

#

and make it use that for booting

#

So I can get long mode

#

And limine supports multiboot protocol so

#

Yeah imma do that tonight okay ty for reading :)

small hound Jun 3, 2025, 4:50 AM

#

bold grove in the scheduler

is said scheduler multicore

bold grove Jun 3, 2025, 4:51 AM

#

small hound is said scheduler multicore

not yet

#

next commit hopefully yes

small hound Jun 3, 2025, 4:51 AM

#

bold grove next commit hopefully yes

how will you do multicore scheduling

bold grove Jun 3, 2025, 4:52 AM

#

small hound how will you do multicore scheduling

ill just create a queue for each cpu

small hound Jun 3, 2025, 4:52 AM

#

how will you balance and sync it globally

bold grove Jun 3, 2025, 4:53 AM

#

small hound how will you balance and sync it globally

round robin

small hound Jun 3, 2025, 4:53 AM

#

that doesnt mean anything

bold grove Jun 3, 2025, 4:53 AM

#

oh wait

#

my bad

small hound Jun 3, 2025, 4:53 AM

#

you only have one global queue of tasks how does each core know what to schedule lol

bold grove Jun 3, 2025, 4:54 AM

#

no im gonna implement smp

small hound Jun 3, 2025, 4:54 AM

#

wat

#

ok i think u might not be getting what I mean lol

#

so you have some tasks you wanna run

#

and you need to figure out how to assign each core to a given set of tasks

#

how will you do that

#

i'm gonna make core-local queues with periodic global resyncs and rebalancing

#

so i have a global circular doubly linked list and i take portions of that list for each task

bold grove Jun 3, 2025, 4:55 AM

#

oh wait cant i do priority scheduling

small hound Jun 3, 2025, 4:55 AM

#

💥

#

wdym

bold grove Jun 3, 2025, 4:56 AM

#

wait

#

cant i have each cpu core check for the highest priority scheduled task like

small hound Jun 3, 2025, 4:59 AM

#

what does this mean

bold grove Jun 3, 2025, 4:59 AM

#

in like a round robin

small hound Jun 3, 2025, 4:59 AM

#

why would this change anything tho

#

you cant run the same task on multiple cores at the same time

bold grove Jun 3, 2025, 4:59 AM

#

i know

small hound Jun 3, 2025, 4:59 AM

#

💥

#

so why does it matter who is running the high priority tasks

#

just give them longer timeslices

#

💥

#

if (task_is_high_priority && timeslice_isnt_up) dont_schedule_next();

#

💥

bold grove Jun 3, 2025, 5:00 AM

#

i see what you mean

torpid viper Jun 3, 2025, 5:36 AM

#

bold grove So the scheduler is done

cool

bold grove Jun 5, 2025, 4:39 PM

#

hi guysss

#

im redoing the thread structsss

#

#include <stdint.h>
#include <stddef.h>
#include <stdbool.h>
#include "sched.h"
#include "thread.h"
#include "../mem/alloc.h"
#include "../apps/echo.h"
#include "../mem/pmm.h"
#include "../mem/vmm.h"
#include "../lib/logging.h"
#include "../lib/str.h"
#include "../lib/assert.h"
#include "../mem/utils.h"


typedef void (*thread_fn) (void *arg);
typedef enum {
    THREAD_READY,
    THREAD_RUNNING,
    THREAD_EXITED,
    THREAD_DEAD
} thread_state_t;

typedef struct thread_list_node {
    struct thread_list_node* next;
    struct list_node* prev;
} thread_list_node_t;

typedef struct {
    thread_list_node_t* head;
    thread_list_node_t* tail;
    size_t count;
    thread_list_node_t node;
} thread_list_t;

typedef enum {
    USER,
    KERNEL
} thread_mode_t;

typedef struct thread {
    thread_fn fn;
    uint32_t thread_id;
    thread_state_t state;
    thread_list_node_t node;
    thread_mode_t mode;
    uintptr_t esp;
    uintptr_t base_esp;
} thread_t;

what do you think
and im just having threads point to C functions
which im assuming is fine
because i can enter whatever from a c function
once i get elf working

small hound Jun 6, 2025, 12:24 AM

#

still no x64 💥

#

lgtm

bold grove Jun 6, 2025, 3:47 AM

#

small hound still no x64 💥

There’s like no difference between threads in ia32 and x64

small hound Jun 6, 2025, 3:47 AM

#

bold grove There’s like no difference between threads in ia32 and x64

esp

bold grove Jun 6, 2025, 3:47 AM

#

Yeah

#

Fine

#

Just a different struct

small hound Jun 6, 2025, 3:48 AM

#

when are we getting x64 support and testing on real hw ⁉️

bold grove Jun 6, 2025, 3:48 AM

#

Runs on real hw

#

X64 support will come when it does I’m not sure

small hound Jun 6, 2025, 3:48 AM

#

what does "run" mean

#

does that mean it booted faulted and gave up

bold grove Jun 6, 2025, 3:49 AM

#

small hound what does "run" mean

Boots and every single feature works

small hound Jun 6, 2025, 3:49 AM

#

oh wowa

#

what about drivers

bold grove Jun 6, 2025, 3:49 AM

#

Yup. keyboard works just fine

#

I can acpi power off and on

small hound Jun 6, 2025, 3:49 AM

#

bold grove Yup. keyboard works just fine

is this the usb keyboard 😲

bold grove Jun 6, 2025, 3:50 AM

#

Yeah but most computers have ps2 emulation

small hound Jun 6, 2025, 3:50 AM

#

i wonder how lenient qemu actually is with device drivers

bold grove Jun 6, 2025, 3:50 AM

#

small hound i wonder how lenient qemu actually is with device drivers

Extremely

small hound Jun 6, 2025, 3:50 AM

#

all 3 of my disk drivers might suck on real hw if disks love to fail

#

i have no fail state checks

bold grove Jun 6, 2025, 3:50 AM

#

The only one that would transfer well is the NVME

small hound Jun 6, 2025, 3:51 AM

#

what about ahci and ide

bold grove Jun 6, 2025, 3:51 AM

#

Because it’s specification is solid and it’s not unreliable

small hound Jun 6, 2025, 3:51 AM

#

small hound what about ahci and ide

do u know

bold grove Jun 6, 2025, 3:51 AM

#

small hound what about ahci and ide

IDE is a fucking mess

small hound Jun 6, 2025, 3:51 AM

#

bold grove Because it’s specification is solid and it’s not unreliable

all the specifications are solid wdym

bold grove Jun 6, 2025, 3:51 AM

#

On my old kernel

small hound Jun 6, 2025, 3:51 AM

#

bold grove On my old kernel

is this your fault or ide's fault

bold grove Jun 6, 2025, 3:51 AM

#

small hound all the specifications are solid wdym

Well yes, but Ide drives are more susceptible to failure

bold grove Jun 6, 2025, 3:51 AM

#

small hound is this your fault or ide's fault

Idk but it worked completely fine in every emulator

small hound Jun 6, 2025, 3:51 AM

#

so does that mean i need to re-attempt the command like a floppy if it fails

small hound Jun 6, 2025, 3:52 AM

#

bold grove Idk but it worked completely fine in every emulator

are you sure the disk was in ide emulation mode or whatever

bold grove Jun 6, 2025, 3:52 AM

#

small hound are you sure the disk was in ide emulation mode or whatever

Yes . I was able to get the signature but not read or write without corruption

small hound Jun 6, 2025, 3:52 AM

#

was the disk in question highly problematic

bold grove Jun 6, 2025, 3:53 AM

#

512 Samsung evo ssd

#

Very new

#

And I know it works

#

I think the issue

#

Is that most hobbyist drivers

#

Don’t really account for

#

Cancer

#

That happens on real hardware

small hound Jun 6, 2025, 3:53 AM

#

are you sure that these fancy new sata drives work fine with ide emulation

bold grove Jun 6, 2025, 3:54 AM

#

Well it was one of these

#

#

So

small hound Jun 6, 2025, 3:55 AM

#

i dont think these new guys play well with IDE controller support

bold grove Jun 6, 2025, 3:55 AM

#

Oh well

small hound Jun 6, 2025, 3:55 AM

#

they would probably like ahci better (perhaps)

bold grove Jun 6, 2025, 3:55 AM

#

Yeah never got around to that

#

Idk tho

#

Because I was able to

#

Like

#

Read from sectors

small hound Jun 6, 2025, 3:55 AM

#

ahci easy peasy you should get to it

bold grove Jun 6, 2025, 3:55 AM

#

And detect them fine

bold grove Jun 6, 2025, 3:56 AM

#

small hound ahci easy peasy you should get to it

Yeah I will after I do arch separation very soon

#

And we will finally have

#

64 bit

#

finally not just ia32

small hound Jun 6, 2025, 3:56 AM

#

what was the logic behind just doing ia32 at first

#

are you rolling ur own bootloader

bold grove Jun 6, 2025, 3:57 AM

#

Nah multiboot 1

bold grove Jun 6, 2025, 3:58 AM

#

small hound what was the logic behind just doing ia32 at first

My dad made a hobby kernel in the 90s rhat was ia32 and I wanted to do what he did at first

small hound Jun 6, 2025, 3:58 AM

#

is that not your own bootloader

small hound Jun 6, 2025, 3:58 AM

#

small hound is that not your own bootloader

huh what

bold grove Jun 6, 2025, 3:58 AM

#

small hound is that not your own bootloader

No…

#

Multiboot is

small hound Jun 6, 2025, 3:58 AM

#

bold grove My dad made a hobby kernel in the 90s rhat was ia32 and I wanted to do what he d...

so we dont exactly live in the 90s i think

bold grove Jun 6, 2025, 3:58 AM

#

A gnu boot specification

small hound Jun 6, 2025, 3:58 AM

#

bold grove Multiboot is

oh the magic number thing

bold grove Jun 6, 2025, 3:58 AM

#

small hound so we dont exactly live in the 90s i think

I know but I wanted to make him proud and then further do 64 bit

small hound Jun 6, 2025, 3:59 AM

#

interesting lore

bold grove Jun 6, 2025, 3:59 AM

#

Yeah it’s also why I did only vga text mode for a long time and vga graphics mode

#

Because my dad made games for his own OS

small hound Jun 6, 2025, 4:00 AM

#

😲

#

awesome sauce

bold grove Jun 6, 2025, 4:00 AM

#

And he made this super sick spreadsheet application in vga text mode that is lowkey still really good

small hound Jun 6, 2025, 4:01 AM

#

most sane 90's c programer

bold grove Jun 6, 2025, 4:01 AM

#

C programmers were on one 20 years ago

timber terrace Jun 6, 2025, 4:01 AM

#

i wish my dad did shit like this he just made hard drives spin to sound like the doom soundtrack

bold grove Jun 6, 2025, 4:01 AM

#

timber terrace i wish my dad did shit like this he just made hard drives spin to sound like the...

Hiii welcome but yeah

#

My dad now does embedded systems at General Electric

small hound Jun 6, 2025, 4:02 AM

#

does he make their engines too

bold grove Jun 6, 2025, 4:03 AM

#

No he makes the software that allows them to work

small hound Jun 6, 2025, 4:03 AM

#

when will we see an ada kernel developed here

#

has there bee one

small hound Jun 6, 2025, 4:03 AM

#

bold grove No he makes the software that allows them to work

does flopper operating system have a thingy mabob like a roadmap

bold grove Jun 6, 2025, 4:04 AM

#

small hound does flopper operating system have a thingy mabob like a roadmap

No. I’ve honestly been slacking and the source code is really messy

#

Honestly like

#

I’m very close to a new commit with a better structure

small hound Jun 6, 2025, 4:05 AM

#

huge commits with 1 line messages

bold grove Jun 6, 2025, 4:05 AM

#

Absolutely

#

So

#

This next commit will come tonight but

#

The main deal right now is that I want to make syscalls

#

After I tidy up my filesystem

small hound Jun 6, 2025, 4:07 AM

#

do you think my kernel is okayish in organization (shameless plug)

bold grove Jun 6, 2025, 4:07 AM

#

small hound do you think my kernel is okayish in organization (shameless plug)

Not bad at all. Pretty good

small hound Jun 6, 2025, 4:08 AM

#

neat

#

i was making things up while organizing it

bold grove Jun 6, 2025, 4:08 AM

#

Sam

#

Same*

small hound Jun 6, 2025, 4:08 AM

#

menix is well organized (the legacy branch)

bold grove Jun 6, 2025, 4:08 AM

#

It’s aight

#

Managarms source code hurts my head

small hound Jun 6, 2025, 4:08 AM

#

it is strangely organized

#

me when i kernel/thor/generic/thor-internal/arch-generic/cursor.hpp

bold grove Jun 6, 2025, 4:09 AM

#

Its structure is strange but

#

Mainly like

#

Lemme just show an example

small hound Jun 6, 2025, 4:10 AM

#

u get used to it

bold grove Jun 11, 2025, 5:23 AM

#

dead os

#

sad days

#

im workig on some stuff though

#

let me cook

#

maybe

sleek mulch Jun 17, 2025, 11:16 AM

#

We wait

crimson plover Jun 17, 2025, 12:26 PM

#

https://tenor.com/view/catjam-cat-disco-catjamming-gif-16989864924126703156

Tenor

bold grove Jun 17, 2025, 10:05 PM

#

sleek mulch We wait

nooo nooo

#

Don’t worry guys im just depressed

#

Work on the os hasn’t started again but

#

It will soon I hope

small hound Jun 17, 2025, 10:18 PM

#

bold grove Don’t worry guys im just depressed

implement the XHCI USB controller to be even more depressed

bold grove Jun 17, 2025, 10:30 PM

#

small hound implement the XHCI USB controller to be even more depressed

I’m still working on scheduling. I’ve gotten threads but now I need to actually get them in userspace

#

Although it wasn’t my goal

#

I’ll probably follow the posix api

#

For that

#

I plan to make all the software for it though so

#

Maybe I can do my own thing

small hound Jun 17, 2025, 10:30 PM

#

i thought this was not a unix-like os

bold grove Jun 17, 2025, 10:30 PM

#

Yeah

#

It isn’t

#

So forget what I said lol

#

I’m just gonna cope

#

Currently

#

I can

small hound Jun 17, 2025, 10:31 PM

#

what 😭

bold grove Jun 17, 2025, 10:31 PM

#

Sorry im bad at typing

#

Give me a bit and I’ll explain what I’m up to

bold grove Jun 17, 2025, 10:34 PM

#

small hound what 😭

void thread_enqueue(thread_t* thr)
Works and now I can create processes. Many threads can be attached to a process and share the same address space. I’m using a round robin and im going to implement SMP eventually

small hound Jun 17, 2025, 10:34 PM

#

neat

bold grove Jun 17, 2025, 10:35 PM

#

My first process is going to be the reaper

#

Then im gonna implement syscalls

#

And then make an actual shell

#

The worst part right now is

#

fork() FOR FUCKS SAKE

small hound Jun 17, 2025, 10:41 PM

#

bold grove `fork()` FOR FUCKS SAKE

why not take a break and work on something else

#

on-disk filesystems

#

block device drivers

#

usb

#

networking

#

audio cards

bold grove Jun 17, 2025, 10:41 PM

#

You’re right

#

I already got a lot done

#

I’ll push later

#

And then I’ll make a proper VFS

#

and then I’ll implement NVME

small hound Jun 17, 2025, 10:45 PM

#

why not do those now not later

crimson plover Jun 17, 2025, 10:55 PM

#

I like working on core stuff just not worrying about getting to userspace or anything

#

hell I am like 200 todo items away from starting to port anything

bold grove Jun 17, 2025, 11:02 PM

#

small hound why not do those now not later

Bc I’m depressed and not next to my computer lol

#

When I get the urge to leave this bed after 45 hours I’ll do it

small hound Jun 17, 2025, 11:03 PM

#

crimson plover I like working on core stuff just not worrying about getting to userspace or any...

relatable

small hound Jun 17, 2025, 11:04 PM

#

bold grove Bc I’m depressed and not next to my computer lol

eat ur happy pills

bold grove Jun 17, 2025, 11:06 PM

#

small hound eat ur happy pills

Already did

#

I’ve ran out of like weed and coffee so

#

I’m kinda just on pause rn

bold grove Jun 17, 2025, 11:07 PM

#

crimson plover hell I am like 200 todo items away from starting to port anything

I’ll never port anything ever because I’m not conforming to Unix stuff and im not gonna implement winapi lol

bold grove Jun 18, 2025, 3:42 AM

#

crimson plover hell I am like 200 todo items away from starting to port anything

I parted ways with my todo list because it’s fucking massive

#

Todo: not make dogshit OS

#

okay

#

So

#

Working on VFS rn

small hound Jun 18, 2025, 3:44 AM

#

bold grove Todo: not make dogshit OS

difficulty: impossible

bold grove Jun 18, 2025, 3:44 AM

#

I’ve gotten basic operations

#

And proper vfs nodes and stuff

#

Now I can mount my temporary file system to /tmp

#

Now

#

I’m thinking

#

Either fat or ext2

#

I’ve implemented EXT2 in userspace before

#

So

#

Maybe I should do that but FAT isn’t that hard from what I’ve heard

#

My old ext2 thing for userspace was like 2000loc from 2018

#

So

#

Time to die looking through that

small hound Jun 18, 2025, 3:46 AM

#

bold grove Now I can mount my temporary file system to /tmp

doesn't this mean you need tmpfs

bold grove Jun 18, 2025, 3:46 AM

#

small hound doesn't this mean you need tmpfs

Which I have

small hound Jun 18, 2025, 3:46 AM

#

bold grove My old ext2 thing for userspace was like 2000loc from 2018

that's surprisingly small did you have everything

small hound Jun 18, 2025, 3:46 AM

#

bold grove Which I have

neat

bold grove Jun 18, 2025, 3:46 AM

#

small hound that's surprisingly small did you have everything

All file operations and nodes and shit yeah

#

I made a JVM like environment

small hound Jun 18, 2025, 3:47 AM

#

truncate, allocate, free, mkdir, rmdir, some basic integrity checks, all of it?

bold grove Jun 18, 2025, 3:47 AM

#

small hound truncate, allocate, free, mkdir, rmdir, some basic integrity checks, all of it?

Yeqh and ref counting

small hound Jun 18, 2025, 3:47 AM

#

bold grove Yeqh and ref counting

do you honor ext2 flags like secure delete where you must zero all child blocks of a file before deleting it

bold grove Jun 18, 2025, 3:47 AM

#

small hound do you honor ext2 flags like secure delete where you must zero all child blocks ...

Yeah it calloc() all of the blocks of the parent

small hound Jun 18, 2025, 3:48 AM

#

bold grove Yeqh and ref counting

wait so how did you keep it that small, did you only handle the first 11 direct blocks?

#

I have 2 different walk functions for blocks and dirents

bold grove Jun 18, 2025, 3:48 AM

#

Remember. Making it in userspace requires less low level stuff

#

So

small hound Jun 18, 2025, 3:48 AM

#

bold grove Yeah it calloc() all of the blocks of the parent

what lol huh

bold grove Jun 18, 2025, 3:49 AM

#

When you delete a file

#

It zeros

small hound Jun 18, 2025, 3:49 AM

#

bold grove Yeah it calloc() all of the blocks of the parent

that's not what secure delete does tho? calloc all blocks of a parent?

bold grove Jun 18, 2025, 3:49 AM

#

All of its child blocks

small hound Jun 18, 2025, 3:49 AM

#

bold grove All of its child blocks

including the indirect ones too?

#

neat

bold grove Jun 18, 2025, 3:49 AM

#

Yes. I’ll post it sometime

#

Or

small hound Jun 18, 2025, 3:49 AM

#

you probably shouldn't do that unless secure delete is specified tho

#

since like

#

slow

bold grove Jun 18, 2025, 3:49 AM

#

small hound you probably shouldn't do that unless secure delete is specified tho

Yeah the interface was a command line and it had a flag

small hound Jun 18, 2025, 3:49 AM

#

but cool that it works

bold grove Jun 18, 2025, 3:50 AM

#

—secure-del

small hound Jun 18, 2025, 3:50 AM

#

bold grove Yeah the interface was a command line and it had a flag

oh wait I get it so you can just treat the /dev/device as your disk

bold grove Jun 18, 2025, 3:50 AM

#

Yes

small hound Jun 18, 2025, 3:50 AM

#

but like a file or large stream

bold grove Jun 18, 2025, 3:50 AM

#

Exactly

small hound Jun 18, 2025, 3:50 AM

#

oh that makes life easier

bold grove Jun 18, 2025, 3:50 AM

#

Yeah so

#

It was simpler than implementing it for an operating system

small hound Jun 18, 2025, 3:50 AM

#

very cool that you got it all to work

bold grove Jun 18, 2025, 3:50 AM

#

It was painful

small hound Jun 18, 2025, 3:50 AM

#

did you also fsck to see if fsck is happy about your impl

#

that was a little annoying because there were linux specific fields in the superblock for me

#

I just ignored em

#

fsck can yap all it wants

bold grove Jun 18, 2025, 3:51 AM

#

Yeah no I didn’t do the Linux field

#

Fuck that shit

small hound Jun 18, 2025, 3:51 AM

#

can u send code soon

#

me curious to see how this is done in userspace

bold grove Jun 18, 2025, 3:51 AM

#

I mean honestly

#

Just check out

#

@west scarab

#

Look at his

small hound Jun 18, 2025, 3:51 AM

#

no but where is yours

bold grove Jun 18, 2025, 3:52 AM

#

Not posted because it’s ass and I hate the way it looks

small hound Jun 18, 2025, 3:52 AM

#

it's not that bad i won't hate lol

#

I wrote some diabolical XHCI stuff earlier

#

would be hypocritical

bold grove Jun 18, 2025, 3:52 AM

#

It will be posted in about a week or so. The only issue is that I leak some memory when removing directories

#

Idk where

small hound Jun 18, 2025, 3:52 AM

#

dang wow we got a shipping delay on an internet repository

bold grove Jun 18, 2025, 3:53 AM

#

Yes sir

small hound Jun 18, 2025, 3:53 AM

#

😢

#

how come

bold grove Jun 18, 2025, 3:53 AM

#

Because I just don’t like it right now and it’s scattered into a bullshit api

#

And written in c++ which is its own issue

#

And it won’t even compile now after gcc updates

small hound Jun 18, 2025, 3:54 AM

#

what did bro do

#

compiler behavior hacks is crazy

#

idc if it compiles lol I'm just curious to see how these are done purely in userspace

bold grove Jun 18, 2025, 3:54 AM

#

Well I’ll upload it soon but for now look at mishakovs

#

I’m also just not at my old computer so

#

And im too fucked up rn to go up the stairs

#

Also my girlfriend is clingy to me rn lol

#

She isn’t letting me get up bc she’s cuddling me

#

Sleepy girlfriend = terrifying

#

@small hound also for ext2 it’s extremely unnecessary to implement the Linux specific fields

#

It doesn’t affect posix compatibility

#

However , if you want to use Linux software eventually you might need to but idk

small hound Jun 18, 2025, 4:07 AM

#

bold grove <@710548131992961074> also for ext2 it’s extremely unnecessary to implement the ...

fsck was yelling at me

bold grove Jun 18, 2025, 4:07 AM

#

Like in an os dev context those are useless

small hound Jun 18, 2025, 4:07 AM

#

i never implemented them

bold grove Jun 18, 2025, 4:08 AM

#

small hound fsck was yelling at me

fsck is horrible outdated software

small hound Jun 18, 2025, 4:08 AM

#

did you have sanity checks in your filesystem implementation

small hound Jun 18, 2025, 4:08 AM

#

bold grove fsck is horrible outdated software

what lol

bold grove Jun 18, 2025, 4:08 AM

#

small hound did you have sanity checks in your filesystem implementation

I had asserts everywhere

small hound Jun 18, 2025, 4:08 AM

#

how would i verify filesystem integrity after modification

small hound Jun 18, 2025, 4:08 AM

#

bold grove I had asserts everywhere

what does this mean?

bold grove Jun 18, 2025, 4:08 AM

#

Sanity checks

small hound Jun 18, 2025, 4:08 AM

#

do you validate an inode's fields as being "sane enough" for the operation to continue

#

give an example

#

show like a code snippet

#

or something

bold grove Jun 18, 2025, 4:09 AM

#

Bet I don’t have it but I’ll just type it up real quick

small hound Jun 18, 2025, 4:09 AM

#

no you can just say it

#

tell me like what kind of checks you had in place before operating on a possibly corrupted inode

#

cos i will probably add this soon

bold grove Jun 18, 2025, 4:10 AM

#

if (inode->size > (inode->blocks * BLOCK_SIZE)) {
        return false; 
    }

#

Throw some sort of error

#

Instead of false

#

And then

#

This is useful

#

if (inode->nlinks == 0 || inode->nlinks > MAX_LINKS) {
        return false;
    }

small hound Jun 18, 2025, 4:11 AM

#

bold grove ``` if (inode->size > (inode->blocks * BLOCK_SIZE)) { return false; ...

huh what

#

oh intersting ok

bold grove Jun 18, 2025, 4:11 AM

#

Also

small hound Jun 18, 2025, 4:12 AM

#

bold grove ``` if (inode->nlinks == 0 || inode->nlinks > MAX_LINKS) { return false;...

this... no?

bold grove Jun 18, 2025, 4:12 AM

#

if (inode->blocks > 0) {
        bool all_zero = true;
        for (int i = 0; i < 15; i++) {
            if (inode->block[i] != 0) {
                all_zero = false;
                break;
            }
        }
        if (all_zero) {
           // give error maybe idk 
        }
    }

small hound Jun 18, 2025, 4:12 AM

#

what huh

small hound Jun 18, 2025, 4:12 AM

#

bold grove ```c if (inode->blocks > 0) { bool all_zero = true; for (int i =...

what??? these are valid scenarios??

#

huhhuhuh

bold grove Jun 18, 2025, 4:12 AM

#

Yes

small hound Jun 18, 2025, 4:12 AM

#

no no i'm saying

#

this is not corruption

#

you can have an inode with no blocks allocated to it

#

why would you produce an error here

#

💥

bold grove Jun 18, 2025, 4:13 AM

#

No no, you would check this BEFORE writing to an inode’s block for the first time

small hound Jun 18, 2025, 4:13 AM

#

bold grove ``` if (inode->nlinks == 0 || inode->nlinks > MAX_LINKS) { return false;...

you can have an inode that has no links if it's still open somewhere but deleted, you'd access it by FD but it would be nonexistent if u access it by path

small hound Jun 18, 2025, 4:14 AM

#

bold grove No no, you would check this BEFORE writing to an inode’s block for the first tim...

what ?!

#

an file that you just created with no data would have no blocks assigned to it

#

this is throwing an error for a perfectly valid scenario

#

crazy lol

bold grove Jun 18, 2025, 4:15 AM

#

There was some point where I checked something like that but yet again

small hound Jun 18, 2025, 4:15 AM

#

perfectly valid filesystem throwing an FS_CORRUPT would be wild

small hound Jun 18, 2025, 4:15 AM

#

bold grove There was some point where I checked something like that but yet again

so then your implementation is just incorrect here then

#

💥

bold grove Jun 18, 2025, 4:19 AM

#

small hound so then your implementation is just incorrect here then

So sorry

#

I explained this wrong

#

before assigning blocks to an inode for the first time, do a sanity check to make sure the inode does not already reference any blocks

#

If the inode has NOT been referenced

#

This doesn’t apply to like device nodes or empty files

small hound Jun 18, 2025, 4:21 AM

#

bold grove before assigning blocks to an inode for the first time, do a sanity check to mak...

why would you throw an error ??

#

just zero the blocks

#

if you have a lazily deleted inode that just removes the entry, of course it would reference blocks

bold grove Jun 18, 2025, 4:21 AM

#

Well because if that was happening im the first place something somewhere else had to have been fucked up

small hound Jun 18, 2025, 4:22 AM

#

no?

#

you dont have to fully delete/zero all the contents of an inode when you remove its link

#

in ext2 you only have to remove it from the bitmap and remove all the child links from the bitmap (deallocate them)

#

so on-disk there will still be blocks on that inode when you re-allocate it

#

but it's fine, normal, and expected behavior

#

💥 did bro actually write this driver

bold grove Jun 18, 2025, 4:23 AM

#

small hound you dont have to fully delete/zero all the contents of an inode when you remove ...

But at least back then I thought it would be best to avoid this scenario

bold grove Jun 18, 2025, 4:23 AM

#

small hound 💥 did bro actually write this driver

Yeah but remember this was 6 years ago lmao

small hound Jun 18, 2025, 4:23 AM

#

you wrote an ext2 driver at age 12?

#

wow

small hound Jun 18, 2025, 4:24 AM

#

bold grove But at least back then I thought it would be best to avoid this scenario

i mean you were 12 so i'll give u credit for doing this at all or even getting remotely started

#

yea i'll give u some slack bro wasn't even a teenager

#

🔥

bold grove Jun 18, 2025, 4:25 AM

#

Apparently I wasn’t even 12

#

Because it’s dated to feb12 2019 but my birthday isn’t until March 29

small hound Jun 18, 2025, 4:25 AM

#

ok now i'm starting to doubt you wrote this at all

bold grove Jun 18, 2025, 4:26 AM

#

I did, at that point I had already made a bootloader and snake game on gameboy in assembly

#

I’d been into low level stuff for a while

#

And stuff like that

small hound Jun 18, 2025, 4:26 AM

#

age 11 ext2 driver?? wowzas

bold grove Jun 18, 2025, 4:26 AM

#

And my dad got me a book about file systems

#

And the C programming language

small hound Jun 18, 2025, 4:26 AM

#

and yet bro is still crashing out about schedulers 6 years later

#

reasonable

bold grove Jun 18, 2025, 4:26 AM

#

Ofc

#

I’m not crashing out over schedulers atm tho

#

Because it actually is fine now

small hound Jun 18, 2025, 4:27 AM

#

when will we drop the 11yo amar ext2 code 🙏

bold grove Jun 18, 2025, 4:27 AM

#

Fork() works

bold grove Jun 18, 2025, 4:27 AM

#

small hound when will we drop the 11yo amar ext2 code 🙏

My funeral

small hound Jun 18, 2025, 4:27 AM

#

bold grove I did, at that point I had already made a bootloader and snake game on gameboy i...

can i see ur bootloader

#

i might add support for my OS to boot on it for no reason

bold grove Jun 18, 2025, 4:28 AM

#

It’s not necessarily a boot loader

small hound Jun 18, 2025, 4:28 AM

#

why did u call it a bootloader

bold grove Jun 18, 2025, 4:28 AM

#

The game boy basically boots into a point of memory

#

Well

#

Because technically it is one

#

But the gameboy doesn’t have the facilities for me to be able to call it a true bootloader

small hound Jun 18, 2025, 4:28 AM

#

bold grove The game boy basically boots into a point of memory

shocker 💥

small hound Jun 18, 2025, 4:29 AM

#

bold grove But the gameboy doesn’t have the facilities for me to be able to call it a true ...

do u have these gameboy repos

bold grove Jun 18, 2025, 4:29 AM

#

Yeah I do

#

My girl is asleep so lemme get my old laptop rq

small hound Jun 18, 2025, 4:29 AM

#

awesome sauces

#

what's with ur internet harassment bot you made 😭

#

is this an april fools joke

bold grove Jun 18, 2025, 4:31 AM

#

Don’t worry about it

#

Plus insta

#

Nuked it

#

So It doesn’t work

small hound Jun 18, 2025, 4:31 AM

#

no your functions dont exist

#

lmao crazy stuff

bold grove Jun 18, 2025, 4:32 AM

#

Okay here’s the bootloader

#

https://pastecode.dev/s/bl1be6cg

PasteCode

Untitled - PasteCode.dev

PasteCode.dev is a website where you can store any code online for quick sharing. You can paste code and share code online for free.

#

This syntax is so cancer btw

#

Have fun reading any of that to be honest

small hound Jun 18, 2025, 4:36 AM

#

neat

bold grove Jun 18, 2025, 4:36 AM

#

The game boy register set is

#

A B C D E H L

#

Also combined

#

Pairs

#

16 bit ones

#

AF
BC
DE
HL

small hound Jun 18, 2025, 4:37 AM

#

bold grove https://pastecode.dev/s/bl1be6cg

isnt this literally a disassembly of the gameboy boot ROM with a few symbol names added

#

https://gist.github.com/drhelius/6063288

Gist

Game Boy Boot ROM Disassembly

Game Boy Boot ROM Disassembly. GitHub Gist: instantly share code, notes, and snippets.

bold grove Jun 18, 2025, 4:38 AM

#

It’s based off of it but yeah

small hound Jun 18, 2025, 4:38 AM

#

they look identical lol

#

typical reverse engineering experience

bold grove Jun 18, 2025, 4:38 AM

#

I didn’t actually use anything online like

#

I actually got the rom itself

#

And objdump the fuck out of it

small hound Jun 18, 2025, 4:38 AM

#

and you.... changed symbol names and called it ur own ?

#

like these are bar-for-bar identical lol

bold grove Jun 18, 2025, 4:39 AM

#

Because it’s the only way you can boot anything on a gameboy it has to be that way

#

I’ll show you the guide I followed

small hound Jun 18, 2025, 4:40 AM

#

bold grove Because it’s the only way you can boot anything on a gameboy it has to be that w...

cant you add your own custom little easter eggs

#

wow the instructions are all identical lol

bold grove Jun 18, 2025, 4:40 AM

#

small hound cant you add your own custom little easter eggs

Literally no

#

It has to be

#

256 bytes

#

In that exact order

#

No exceptions

small hound Jun 18, 2025, 4:41 AM

#

so this wasnt much of a bootloader then

#

just objdump -> change symbols -> recompile

#

darn

#

did u ever do any other niche hardware

bold grove Jun 18, 2025, 4:42 AM

#

I own an Apple 2

#

And I made many basic programs for it

#

That’s the extent

small hound Jun 18, 2025, 4:42 AM

#

neat

#

very nice pc

#

that's all pretty cool 💥

#

lucky kid lol

bold grove Jun 18, 2025, 4:43 AM

#

small hound cant you add your own custom little easter eggs

Wait you’re right

#

Let me see

#

What I can do for that

#

Fuck a VFS

small hound Jun 18, 2025, 4:44 AM

#

💥

bold grove Jun 18, 2025, 4:45 AM

#

Now I’m purposely gonna add Easter eggs

#

Okay I figured out how to actually flow the code differently

small hound Jun 18, 2025, 5:03 AM

#

'flow the code'

bold grove Jun 19, 2025, 5:58 PM

#

Alright guys

#

Enough bullshit

#

Enough stupid source structure

#

Rewriting

#

Using limine

#

I already got Flanterm and pmm working :)

#

And gdt idt

#

I’m gonna rename it Zarugan (Za - roo - gun) which is the name of my fictional language I made

torpid viper Jun 19, 2025, 6:44 PM

#

bold grove I already got Flanterm and pmm working :)

cool

bold grove Jun 19, 2025, 7:39 PM

#

long mode

#

im making a build system

#

for shits and giggles

blissful grove Jun 20, 2025, 10:03 AM

#

bold grove Rewriting

Wow its happening, good luck

bold grove Jun 20, 2025, 7:09 PM

#

blissful grove Wow its happening, good luck

Yessir

#

Wanna look at my new build system? Just started making it

#

Here’s the syntax, I already made the ast and parser and stuff so

#

decl cc = gcc;
decl src csource= main.c; // src is a built in
decl bin dest= build/main; // bin is a built in
decl args cargs = -foo;

method all (main.c) {
   @cc @cargs -o @dest;
}

small hound Jun 20, 2025, 7:44 PM

#

"We have make at home"

bold grove Jun 20, 2025, 8:06 PM

#

Idk it compiles my kernel quicker

#

And also compiles qemu decently quick with some problem spots lol

small hound Jun 26, 2025, 3:26 AM

#

are you making another 50,000 line commit called "fixes."

bold grove Jun 26, 2025, 5:33 AM

#

small hound are you making another 50,000 line commit called "fixes."

No it’s a complete rewrite

#

In 64 bit

#

Using limine

#

And I have some stuff

small hound Jun 26, 2025, 5:34 AM

#

"fixes."

+50000
-0

crimson plover Jun 26, 2025, 5:56 AM

#

lol

bold grove Jun 26, 2025, 7:55 PM

#

“Fuck this shit. Fixes”

+10000000
-0

bold grove Jun 30, 2025, 10:20 PM

#

Dead os

#

I’m not as good at this shit as I need to be

small hound Jun 30, 2025, 10:53 PM

#

🥀💔

bold grove Jul 8, 2025, 3:57 AM

#

I rewrote paging.c 🙏

small hound Jul 8, 2025, 4:02 AM

#

is this os still dead

small hound Jul 8, 2025, 4:02 AM

#

bold grove I’m not as good at this shit as I need to be

unc fell off 💔

bold grove Jul 8, 2025, 4:07 AM

#

I’m actually working on it a lot rn

bold grove Jul 8, 2025, 9:50 PM

#

small hound unc fell off 💔

Unc is cooking a little

small hound Jul 8, 2025, 9:56 PM

#

bold grove Unc is cooking a little

is unc making an asynchronous io stack ⁉️

#

oooooooooooooo

bold grove Jul 8, 2025, 10:03 PM

#

Nah nah nah I just rewrote my vmm and paging shit

#

And now cleaning up scheduler

#

I made all threads kernel threads

#

It’s just user mode threads are just kernel threads that jump to my user mode entry routine in C

blissful grove Jul 9, 2025, 9:14 AM

#

Nice, good to see you're still going with this

bold grove Jul 9, 2025, 5:21 PM

#

I’m trying lol

bold grove Jul 10, 2025, 5:35 AM

#

I’ve rewritten many parts

#

Of mem/ and sched/

bold grove Jul 10, 2025, 9:17 PM

#


void ctx_switch(cpu_ctx_t* _old, cpu_ctx_t* _new) {
    asm volatile (
        "pushf\n"
        "pusha\n"
        "movl %%esp, %[old]\n"
        "movl %[new], %%esp\n"
        "popa\n"
        "popf\n"
        "ret\n"
        : [old] "=m"(*_old)
        : [new] "m"(*_new)
        : "memory"
    );
}

#

yee

small hound Jul 10, 2025, 10:07 PM

#

bold grove ```c void ctx_switch(cpu_ctx_t* _old, cpu_ctx_t* _new) { asm volatile ( ...

32 bit 💥

small hound Jul 11, 2025, 1:52 AM

#

why did you not call this the flopperating system

bold grove Jul 11, 2025, 2:07 AM

#

small hound why did you not call this the flopperating system

Phenomenal idea

#

I’m gonna do that

pliant oriole Jul 12, 2025, 10:25 PM

#

bold grove ```c void ctx_switch(cpu_ctx_t* _old, cpu_ctx_t* _new) { asm volatile ( ...

use an asm function ❌ use a C function that only has inline asm ✅

small hound Jul 12, 2025, 10:39 PM

#

pliant oriole use an asm function ❌ use a C function that only has inline asm ✅

is your scheduler working

#

did u fix it

#

😃

pliant oriole Jul 13, 2025, 12:03 AM

#

sorta

#

somehow mov rsp, 4096 right before i jump to userspace fixed it

small hound Jul 13, 2025, 12:09 AM

#

pliant oriole somehow mov rsp, 4096 right before i jump to userspace fixed it

was there a rewrite 😃

pliant oriole Jul 13, 2025, 12:28 AM

#

nah but i just finished cr3 switching so now the scheduler is basically done and i'm happy to do a rewrite

bold grove Jul 13, 2025, 1:24 AM

#

virtual memory manager done

#

the rewrite

#

https://paste.ofcode.org/uPuba8eNsTaL66ByWZgcaQ

bold grove Jul 13, 2025, 1:25 AM

#

blissful grove Nice, good to see you're still going with this

look!!

#

@small hound i wrote a goated vmm

small hound Jul 13, 2025, 1:44 AM

#

bold grove <@710548131992961074> i wrote a goated vmm

that's it?

#

oh ok

#

cool beans

#

I thought there would be a bit more when you said it was goated

bold grove Jul 13, 2025, 1:50 AM

#

im making it a bit better

#

addibng comments and also todos

#

and some other fun methods

small hound Jul 13, 2025, 1:52 AM

#

bold grove im making it a bit better

will we be getting demand paging

#

generic architecture agnostic paging system when

#

use a for loop to map pages and have each architecture implement an operation to do one layer of paging

small hound Jul 13, 2025, 1:54 AM

#

bold grove https://paste.ofcode.org/uPuba8eNsTaL66ByWZgcaQ

wait where is your tlb shootdown

#

also doing a full tlb flush after any paging action is kinda wild

#

the invlpg in question:

#

why does your region unmapping also deallocate the pages in your physical memory allocator 🤨

#

int unmap_range(uintptr_t vaddr, size_t size) {
    size_t pages = _align_up(size) / PAGE_SIZE;

    vm_region_t *cur = vm_regions, *prev = NULL;

    // iterate through the region list to find pages to unmap
    while (cur) {
        // check if the current region overlaps with the range to be unmapped
        if (cur->start == vaddr && cur->end == vaddr + size) {
            // if the current region is the first in the list, update the head of the list
            if (prev) {
                prev->next = cur->next;
            }
            // if the current region is the last in the list, update the tail of the list
            else {
                vm_regions = cur->next;
            }
            pmm_free_page(cur);
            break;
        }
        prev = cur;
    }
    // iterate through the pages to be unmapped
    for (size_t i = 0; i < pages; i++) {
        if (unmap_page(vaddr + i * PAGE_SIZE) < 0)
            return -1;
    }
    return 0;
}
``` sus

#

dunno why you'd wanna deallocate pages in page unmapping logic

bold grove Jul 13, 2025, 2:11 AM

#

im pretty sure i pressed enter too many times in copilot so im ficing that now

#

@small hound this should be better

#


int unmap_range(uintptr_t vaddr, size_t size) {
    size_t pages = _align_up(size) / PAGE_SIZE;

    vm_region_t *cur = vm_regions, *prev = NULL;

    while (cur) {
        // check if the range is within the current region
        if (cur->start == vaddr && cur->end == vaddr + size) {

            if (prev) {
                prev->next = cur->next;

            } else {
                vm_regions = cur->next;
            }
            pmm_free_page(cur);
            break;
        }

        // move to next region
        prev = cur;
        cur = cur->next;
    }

    for (size_t i = 0; i < pages; i++) {
        uintptr_t va = vaddr + i * PAGE_SIZE;
        if (unmap_page(va) < 0) {
            // todo: handle error
            return -1;
        }
    }
    return 0;
}

bold grove Jul 13, 2025, 2:12 AM

#

small hound wait where is your tlb shootdown

its in my paging.c

#

here:

#

void _flush_tlb(void) {
    uintptr_t cr3;
    __asm__ volatile("mov %%cr3, %0" : "=r"(cr3));
    __asm__ volatile("mov %0, %%cr3" :: "r"(cr3));
}

small hound Jul 13, 2025, 2:13 AM

#

bold grove ```c void _flush_tlb(void) { uintptr_t cr3; __asm__ volatile("mov %%cr3,...

this is NOT tlb shootdown

bold grove Jul 13, 2025, 2:14 AM

#

small hound this is NOT tlb shootdown

what would it look like then

small hound Jul 13, 2025, 2:15 AM

#

bold grove what would it look like then

it is when you tell other processors to invalidate their TLB entries

bold grove Jul 13, 2025, 2:15 AM

#

oh

#

isnt taht

small hound Jul 13, 2025, 2:15 AM

#

on x86 this would be via an IPI

bold grove Jul 13, 2025, 2:15 AM

#

yeah

small hound Jul 13, 2025, 2:15 AM

#

do you have that

bold grove Jul 13, 2025, 2:15 AM

#

no but

#

i had in my old src file

#

gonna go get it

#

wait

#

isnt it literally

#

just

#

asm volatile ("invlpg (%0)" : : "a" (vaddr));

#

in my case

small hound Jul 13, 2025, 2:17 AM

#

yes and you need a way to send that address to other processors and wait for them to all invalidate that entry in their TLBs

#

flushing the entire TLB after changing one page mapping is very inefficient

bold grove Jul 13, 2025, 2:18 AM

#

how would i be able to sen that addr to other processors

small hound Jul 13, 2025, 2:18 AM

#

static void do_tlb_shootdown(uint64_t addr) {
    uint64_t this_core = get_sch_core_id();
    uint64_t cores = scheduler_get_core_count();
    for (uint64_t i = 0; i < cores; i++) {
        if (i == this_core)
            continue;

        struct core *target = global_cores[i];
        atomic_store_explicit(&target->tlb_shootdown_page, addr,
                              memory_order_release);
        lapic_send_ipi(i, TLB_SHOOTDOWN_ID);
        while (atomic_load_explicit(&target->tlb_shootdown_page,
                                    memory_order_acquire) != 0)
            cpu_relax();
    }
}``` I cheesed this a bit with an atomic ptr

bold grove Jul 13, 2025, 2:19 AM

#

so if i dont have many cores rn

#

can i just

#

asm volatile ("invlpg (%0)" : : "a" (va));

#

for now

small hound Jul 13, 2025, 2:19 AM

#

bold grove for now

do you plan to support SMP

bold grove Jul 13, 2025, 2:19 AM

#

and i dont have lapic just pit

#

yeah eventually but

#

not at the moment

small hound Jul 13, 2025, 2:20 AM

#

ok then just remember to do the tlb shootdown

bold grove Jul 13, 2025, 2:20 AM

#

okay

#

but is asm volatile ("invlpg (%0)" : : "a" (va)); good enough for now after unmapping?

small hound Jul 13, 2025, 2:20 AM

#

yep 👍

#

do you support 2MB pages

#

not sure if 32 bit supports that

#

pretty sure it does

bold grove Jul 13, 2025, 2:21 AM

#

yes

#

its not that off from 64'

#

just remember 4gb limited mem

#

besides that its the same shit

small hound Jul 13, 2025, 2:21 AM

#

also where is your actual page table walking logic

#

int map_page(uintptr_t vaddr, uintptr_t paddr, PageAttributes attrs) {
    vaddr = _align_down(vaddr);
    paddr = _align_down(paddr);

    // get/create page table
    PTE* pt = _get_or_make_pt(vaddr);
    if (!pt) return -1;

    // map physical frame to virtual address
    uint32_t pt_idx = _pt_index(vaddr);
    attrs.frame_addr = paddr >> 12;
    SET_PF(&pt[pt_idx], attrs);

    _flush_tlb();
    return 0;
}```

#

where are you walking the page tables

#

void vmm_map_page(uintptr_t virt, uintptr_t phys, uint64_t flags) {
    if (virt == 0) {
        k_panic("CANNOT MAP PAGE 0x0!!!\n");
    }

    bool interrupts = spin_lock(&vmm_lock);
    struct page_table *current_table = kernel_pml4;

    for (uint64_t i = 0; i < 3; i++) {
        uint64_t level = virt >> (39 - (i * 9)) & 0x1FF;
        pte_t *entry = &current_table->entries[level];
        if (!ENTRY_PRESENT(*entry))
            pte_init(entry, 0);

        current_table =
            (struct page_table *) ((*entry & PAGING_PHYS_MASK) + hhdm_offset);
    }

    uint64_t L1 = (virt >> 12) & 0x1FF;
    pte_t *entry = &current_table->entries[L1];
    *entry = (phys & PAGING_PHYS_MASK) | flags | PAGING_PRESENT;

    invlpg(virt);
    do_tlb_shootdown(virt);
    spin_unlock(&vmm_lock, interrupts);
}``` like I have this here

#

the loop

#

but where do you walk your page tables

bold grove Jul 13, 2025, 2:22 AM

#

ohhh i see

#

i hadthat implmented '

#

in paging.c

#

but forgot to add that subroutine

#

in map page

#

so let me

#

fix that

small hound Jul 13, 2025, 2:23 AM

#

💀 this is like the most important part of your virtual memory management

#

is this blob of 300 lines of code you sent tested at all

#

does this even run

bold grove Jul 13, 2025, 2:23 AM

#

i literally wrote it all today

#

and yes it runs just having issues but that was the prob lem

small hound Jul 13, 2025, 2:23 AM

#

and copilot too

bold grove Jul 13, 2025, 2:23 AM

#

im tired

#

nah

#

copilot i just use for cmments sometimes

small hound Jul 13, 2025, 2:23 AM

#

bold grove and yes it runs just having issues but that was the prob lem

how does it run if you aren't walking the page tables

bold grove Jul 13, 2025, 2:25 AM

#

it initializes fine

#

but yeah

#

that was the issue

small hound Jul 13, 2025, 2:25 AM

#

how is it initializing if your mapping logic doesn't walk the page tables...

#

💥

#

if vmm_init(void) calls _map_kernel(pstart, ksize), and

_map_kernel(uintptr_t pstart, size_t size)
calls map_page(vaddr, paddr, k_pg_attrs), and
map_page(uintptr_t vaddr, uintptr_t paddr, PageAttributes attrs) does this

int map_page(uintptr_t vaddr, uintptr_t paddr, PageAttributes attrs) {
    vaddr = _align_down(vaddr);
    paddr = _align_down(paddr);

    // get/create page table
    PTE* pt = _get_or_make_pt(vaddr);
    if (!pt) return -1;

    // map physical frame to virtual address
    uint32_t pt_idx = _pt_index(vaddr);
    attrs.frame_addr = paddr >> 12;
    SET_PF(&pt[pt_idx], attrs);

    _flush_tlb();
    return 0;
}```

#

then how is it initializing

#

👻

#

page table poltergeist

bold grove Jul 13, 2025, 2:34 AM

#

Well

#

I’m going out rn

#

So I will message later @small hound

bold grove Jul 16, 2025, 6:10 AM

#

@small hound

small hound Jul 16, 2025, 6:10 AM

#

yes

#

it has been a while

bold grove Jul 16, 2025, 6:11 AM

#

im walking page tables like you said

#

let me show

small hound Jul 16, 2025, 6:13 AM

#

👍

bold grove Jul 16, 2025, 6:14 AM

#

https://pastecode.io/s/7hshippc

#

what do you think

#

theres only 2 levels of paging

#

in 32 bit

#

mind ou

#

you

small hound Jul 16, 2025, 6:15 AM

#

bold grove theres only 2 levels of paging

the goto is cursed

#

separate those out into functions

bold grove Jul 16, 2025, 6:16 AM

#

okay bet

bold grove Jul 16, 2025, 6:24 AM

#

small hound separate those out into functions

Look fine besides that though?

small hound Jul 16, 2025, 6:25 AM

#

bold grove Look fine besides that though?

you can also check if the higher level pages are also empty rather than just looking at the lowest one

#

for freeing

#

looks ok

#

👍

bold grove Jul 16, 2025, 6:31 AM

#

@small hound i modularized this nicer

#

int map_page(uintptr_t vaddr, uintptr_t paddr, PageAttributes attrs) {
    vaddr = _align_down(vaddr);
    paddr = _align_down(paddr);
    if (vaddr == 0) return -1;

    uint32_t indices[2] = {_pd_index(vaddr), _pt_index(vaddr)};
    PTE* tables[2] = {RECURSIVE_PD, NULL};
    for (int i = 0; i < 2; i++) { // walk page table levels
        switch (i) {
            case 0:
                _pd_lvl(tables, indices, attrs);
            case 1:
                _pt_lvl(tables, indices, paddr, attrs);
        }
    }

    asm volatile ("invlpg (%0)" : : "a" (vaddr));
    return 0;
}

small hound Jul 16, 2025, 6:32 AM

#

bold grove ```c int map_page(uintptr_t vaddr, uintptr_t paddr, PageAttributes attrs) { ...

lol why is there a loop at all

bold grove Jul 16, 2025, 6:32 AM

#

sorry

#

bro

#

because you said

#

there should be a loop

bold grove Jul 16, 2025, 6:32 AM

#

small hound the loop

^

small hound Jul 16, 2025, 6:32 AM

#

I forgot you're on 32 bit 💀

bold grove Jul 16, 2025, 6:32 AM

#

so ive been losing my shit

small hound Jul 16, 2025, 6:32 AM

#

it's usually a bit better on 64 bit

#

I mean if it works it works

bold grove Jul 16, 2025, 6:33 AM

#

well

#

ive spent hours

#

so

#

idc

small hound Jul 16, 2025, 6:33 AM

#

✅ great job

bold grove Jul 16, 2025, 6:33 AM

#

heres how im doing ranges

#

int unmap_range(uintptr_t vaddr, size_t size) {
    size_t pages = _align_up(size) / PAGE_SIZE;

    vm_region_t *cur = vm_regions, *prev = NULL;

    while (cur) {
        // check if the range is within the current region
        if (cur->start == vaddr && cur->end == vaddr + size) {

            if (prev) {
                prev->next = cur->next;

            } else {
                vm_regions = cur->next;
            }
            break;
        }

        // move to next region
        prev = cur;
        cur = cur->next;
    }

    for (size_t i = 0; i < pages; i++) {
        uintptr_t va = vaddr + i * PAGE_SIZE;
        if (unmap_page(va) < 0) {
            // todo: handle error
            return -1;
        }
    }
    return 0;
}

int map_range(uintptr_t vaddr, uintptr_t paddr, size_t size, PageAttributes attrs) {
    size_t pages = _align_up(size) / PAGE_SIZE;

    vm_region_t* r = _make_region(vaddr, vaddr + size, attrs);
    if (!r) {
        return -1;
    }
    r->next = vm_regions;
    vm_regions = r;

    for (size_t i = 0; i < pages; i++) {
        uintptr_t va = vaddr + i * PAGE_SIZE;
        uintptr_t pa = paddr + i * PAGE_SIZE;
        if (map_page(va, pa, attrs) < 0) {
            // rollback if failed 
            // prevent mem leaks
            // todo: actual error handling
            for (size_t j = 0; j < i; j++) {
                uintptr_t undo_va = vaddr + j * PAGE_SIZE;
                unmap_page(undo_va);
            }
            vm_regions = r->next;
            return -1;
        }
    }
    return 0;
}

small hound Jul 16, 2025, 6:34 AM

#

sounds great 😃

bold grove Jul 16, 2025, 6:40 AM

#

heres the whole new thing

#

https://pastecode.io/s/gghv0in5

small hound Jul 17, 2025, 4:22 PM

#

by the way mister flopmaster, my TLB eviction is kind of not the best way to do it. it's not exactly wrong but it also isnt the most efficient way

#

the better way would be to have a list (global or per core) of pages to invalidate, so that when you map/unmap a region, rather than waiting after each page, you add all pages to the list and send the IPI to start doing TLB shootdowns

bold grove Jul 17, 2025, 6:31 PM

#

Okay bet

#

Should u just create like a queue of pages to be invalidated?

small hound Jul 17, 2025, 7:31 PM

#

bold grove Should u just create like a queue of pages to be invalidated?

yah and you can either send an IPI where the interrupt handler does the shootdown or you can have a thread that does that and just wake up the thread on the other cores

#

depends on how you wanna do it

#

or if you have it, you can use a deferred procedure call

#

and somehow get the other cores to run that DPC

bold grove Jul 17, 2025, 7:55 PM

#

small hound yah and you can either send an IPI where the interrupt handler does the shootdow...

I’m prob just gonna make a kernel thread to do that for me

bold grove Jul 17, 2025, 10:53 PM

#

small hound yah and you can either send an IPI where the interrupt handler does the shootdow...

I made a kernel thread to do it for me

small hound Jul 17, 2025, 10:55 PM

#

bold grove I made a kernel thread to do it for me

does it work

#

you should make a unit and integration test suite to test these and such

bold grove Jul 17, 2025, 10:58 PM

#

Yeah right now my resting functions are just little static functions in my source files

#

So in vmm.c I have some static functions

small hound Jul 17, 2025, 10:59 PM

#

bold grove Yeah right now my resting functions are just little static functions in my sourc...

resting functions?

bold grove Jul 17, 2025, 10:59 PM

#

Testing

#

My bad

small hound Jul 17, 2025, 10:59 PM

#

oh ok

bold grove Jul 17, 2025, 10:59 PM

#

Like

#

Vmm_range_map_test

#

That bullshit

#

But yes I do plan to make a testing suite at some point

#

I haven’t pushed this yet only bc git is stupid on my laptop

small hound Jul 17, 2025, 11:01 PM

#

bold grove I haven’t pushed this yet only bc git is stupid on my laptop

how

bold grove Jul 17, 2025, 11:01 PM

#

It just gets stuck on “pushing”

small hound Jul 17, 2025, 11:02 PM

#

internet issue?

bold grove Jul 17, 2025, 11:02 PM

#

Nah

small hound Jul 17, 2025, 11:02 PM

#

did you sign in with gh-cli

bold grove Jul 17, 2025, 11:02 PM

#

Idk

#

Yes

#

I’m gonna figure this out rn

small hound Jul 17, 2025, 11:02 PM

#

reboot it

#

trl

#

everything fixes itself after reboot, trust

bold grove Jul 17, 2025, 11:02 PM

#

So I can actually update this god forsaken thing

#

Said every windows user ever

small hound Jul 17, 2025, 11:03 PM

#

for linux, be root, for windows, reboot

bold grove Jul 17, 2025, 11:04 PM

#

small hound for linux, be root, for windows, reboot

Holy fucking bars

small hound Jul 17, 2025, 11:10 PM

#

still no flopperating system 😢

high hawk Jul 17, 2025, 11:10 PM

#

small hound still no flopperating system 😢

but there's plenty of floppy operating systems

#

i.e. operating systems for floppy disks!

bold grove Jul 17, 2025, 11:15 PM

#

I wish i could run mine on one

#

Unfortunately multiboot is more bloated than a polish drunk dad

#

Why the fck do I need an EGA video mode

bold grove Jul 18, 2025, 12:21 AM

#

Flopperating System

#

Changed thread name 🗣️🗣️

#

Just saying

#

I fucking love the buddy system

#

Doesn’t get easier than this

#


void* pmm_alloc_pages(uint32_t order, uint32_t count) {
    if (order > MAX_ORDER || count == 0) return NULL;

    void* first_page = NULL;
    void* current_page = NULL;

    for (uint32_t i = 0; i < count; i++) {
        void* page = NULL;
        for (uint32_t j = order; j <= MAX_ORDER; j++) {
            if (pmm_buddy.free_list[j]) {
                struct Page* p = pmm_buddy.free_list[j];
                pmm_buddy.free_list[j] = p->next;
                p->is_free = 0;
                p->order = order;

                while (j > order) {
                    j--;
                    buddy_split(p->address, j);
                }

                page = (void*)p->address;
                break;
            }
        }

        if (!page) {
            if (first_page) {
                pmm_free_pages(first_page, order, i);
            }
            log("pmm: Out of memory!\n", RED);
            return NULL;
        }

        if (!first_page) {
            first_page = page;
        }
    }

    return first_page;
}

small hound Jul 18, 2025, 1:20 PM

#

bold grove ```c void* pmm_alloc_pages(uint32_t order, uint32_t count) { if (order > MA...

now make it NUMA aware trl

bold grove Jul 18, 2025, 2:02 PM

#

small hound now make it NUMA aware <:trl:829316641191886888>

In this economy? Only bitmap trl trl

small hound Jul 18, 2025, 2:02 PM

#

bold grove In this economy? Only bitmap <:trl:829316641191886888> <:trl:829316641191886888>

wut

#

you can make a NUMA aware bitmap allocator

#

it will just be bad

bold grove Jul 18, 2025, 2:13 PM

#

small hound you can make a NUMA aware bitmap allocator

That would be funny

#

Doesn’t that have nothing to do with bitmap / buddy

#

I thought it was where memory access time depends on the memory location relative to the processor

#

It’s an smp thing iirc

small hound Jul 18, 2025, 2:57 PM

#

bold grove I thought it was where memory access time depends on the memory location relativ...

yes

#

non uniform memory access

bold grove Jul 18, 2025, 2:58 PM

#

Yes

bold grove Jul 18, 2025, 3:01 PM

#

small hound non uniform memory access

Kinda a complicated concept but I get it after reading some more

bold grove Jul 18, 2025, 3:35 PM

#

Bro my parents are crazy fucks

#

They destroyed my fucking laptop which I just fucking bought for uni

#

For no fucking reason too besides being fucking assholes

#

It’s because I didn’t give them 200$ out my paycheck

#

When I literally

#

Don’t even make 500 after taxes a now each week

#

So consider the project on hold

#

All of my source code is gone

#

Computer is fucked

small hound Jul 18, 2025, 3:40 PM

#

bold grove So consider the project on hold

dang, I hope u can recover the disk off of the laptop and maybe get some stuff back

#

🙏

bold grove Jul 18, 2025, 3:40 PM

#

It’s just so fucking infuriating living with people like this

#

I got bread but god damn can I just have my laptop 😢

bold grove Jul 19, 2025, 7:44 PM

#

Update: I bought a new computer

#

Setting it up later

#

Flopperating system may be back

#

Hopefully I can fucking push git commits too

bold grove Jul 19, 2025, 8:02 PM

#

Okay we got it

bold grove Jul 19, 2025, 8:31 PM

#

dude

#

it took me so damn long

#

to write this shit

#

PDE* vmm_clone_pd(PDE* _src_pd) {
    // allocate new page directory (zero ts)
    PDE* new_pd = (PDE*)pmm_alloc_page();
    if (!new_pd) return NULL;

    for (uint32_t i = 0; i < PAGE_DIRECTORY_SIZE; i++) { 
        if ((i << 22) >= KERNEL_VADDR_BASE) {
            new_pd[i] = kernel_pd[i];
            continue;
        }
        if (!_src_pd[i].present)
            continue;
        
        // define the src pt and create a new one
        PTE* src_pt = (PTE*)(_src_pd[i].table_addr << 12);
        PTE* new_pt = (PTE*)pmm_alloc_page();
        if (!new_pt) return NULL;

        _zero_pt(new_pt);

        for (uint32_t j = 0; j < PAGE_TABLE_SIZE; j++) {
            if (!src_pt[j].present)
                continue;

            // allocate new frame 
            void* new_frame = pmm_alloc_page();
            if (!new_frame)
                return NULL;

            // memcpy old frame to new frame
            void* old_frame = (void*)(src_pt[j].frame_addr << 12);
            flop_memcpy(new_frame, old_frame, PAGE_SIZE);

            PageAttributes attrs = {
                .present = src_pt[j].present,
                .rw = src_pt[j].rw,
                .user = src_pt[j].user,
                .frame_addr = ((uintptr_t)new_frame) >> 12
            };

            SET_PF(&new_pt[j], attrs);
        }


        PageAttributes pt_attrs = {
            .present = 1,
            .rw = 1,
            .user = 1,
            .frame_addr = ((uintptr_t)new_pt) >> 12
        };

        SET_PF((PTE*)&new_pd[i], pt_attrs);
    }

    return new_pd;
}

small hound Jul 19, 2025, 9:51 PM

#

lmao what's with the constant code dumps

#

we have version management at home

bold grove Jul 19, 2025, 10:17 PM

#

Just too lazy

#

Plus I’d rather show it here if it’s just short blocks

blissful grove Jul 20, 2025, 7:28 AM

#

its alright, these threads are for sharing your progress

#

I reckon 50% of that is either kernel panics or code snippets lol

#

or screenshots of code 😠

noble remnant Jul 20, 2025, 12:36 PM

#

beautiful rename

#

how fast does it flopperate though

bold grove Jul 21, 2025, 1:49 AM

#

noble remnant how fast does it flopperate though

Never fast enough 😔

bold grove Jul 21, 2025, 1:49 AM

#

noble remnant beautiful rename

Courtesy of @small hound

bold grove Jul 22, 2025, 1:12 AM

#

GitHub repo update: 3-5 days from now

#

Maybe tonight if I don’t drink a fuck ton but

bold grove Jul 22, 2025, 1:14 AM

#

blissful grove its alright, these threads are for sharing your progress

Don’t worry my progress will be visible to you guys soon

#

I just fucking suck at version control

#

And it kinda makes me mad

pliant oriole Jul 22, 2025, 2:32 AM

#

noble remnant how fast does it flopperate though

10 exaflops/second

blissful grove Jul 22, 2025, 2:46 AM

#

Lol take your time, no one here is rushing you

bold grove Jul 22, 2025, 2:57 AM

#

blissful grove Lol take your time, no one here is rushing you

thanks man

#

kinda cooking atm so

#

we wil see

small hound Jul 22, 2025, 2:59 AM

#

oh yea a thing I found helpful was to encapsulate all globals that are used in a bunch of places in one struct like this

struct charmos_globals {
    char *root_partition;
    struct vfs_mount *mount_list_head;
    struct vfs_node *root_node;
    volatile enum bootstage current_bootstage;
    uint64_t core_count;
    struct scheduler **schedulers;
    struct core **cores;

    volatile bool panic_in_progress;
};

extern struct charmos_globals global;```

#

idk u might find it helpful

#

I found it very neat and useful

bold grove Jul 22, 2025, 3:06 AM

#

Oh shit yeah

#

That’s not a bad idea

#

I should probably reserve a bit of memory in the kernel to hold a struct like that at boot

small hound Jul 22, 2025, 3:17 AM

#

bold grove I should probably reserve a bit of memory in the kernel to hold a struct like th...

"reserve a bit of memory" wdym

#

just

struct charmos_globals global = {0};

bold grove Jul 22, 2025, 3:21 AM

#

small hound "reserve a bit of memory" wdym

yeah you right

#

anyways

#

i should invlpg only after unmapping, right?

small hound Jul 22, 2025, 3:22 AM

#

bold grove i should invlpg only after unmapping, right?

do you ever change PTEs in any situation besides unmapping

bold grove Jul 22, 2025, 3:22 AM

#

no

small hound Jul 22, 2025, 3:22 AM

#

are you sure

bold grove Jul 22, 2025, 3:23 AM

#

wait

small hound Jul 22, 2025, 3:23 AM

#

if 0xfff000 is mapped to 0x4000, and you go and map 0xfff000 to 0x1000, you are indeed changing page table entries

bold grove Jul 22, 2025, 3:23 AM

#

small hound do you ever change PTEs in any situation besides unmapping

actualy yeah

#

yeah so for mapping and unmapping

#

i need to invlpg

#

got it

small hound Jul 22, 2025, 3:23 AM

#

✅

bold grove Jul 22, 2025, 3:23 AM

#

anywhere i change pt

#

i need to invlpg

#

well great, no paging fault now

#

okay yessir

#

got it

#

lemme show

#

how im mapping my memory

pliant oriole Jul 22, 2025, 3:38 AM

#

small hound oh yea a thing I found helpful was to encapsulate all globals that are used in a...

this is good i should add this as well

bold grove Jul 22, 2025, 3:39 AM

#

int map_kernel_space(uintptr_t vaddr, size_t size) {
    if (vaddr < KERNEL_VADDR_BASE) {
        return -1;
    }

    page_attrs_t attrs = {
        .present = 1,
        .rw = 1,
        .user = 0
    };

    uintptr_t phys_addr = (uintptr_t)pmm_alloc_pages(0, size / PAGE_SIZE);
    return map_range(vaddr, phys_addr, size, attrs);
}

int map_user_space(uintptr_t vaddr, size_t size) {
    if (vaddr >= USER_SPACE_END) {
        return -1;
    }

    page_attrs_t attrs = {
        .present = 1,
        .rw = 1,
        .user = 1
    };

    uintptr_t phys_addr = (uintptr_t)pmm_alloc_pages(0, size / PAGE_SIZE);
    return map_range(vaddr, phys_addr, size, attrs);
}

#

and the user gets 3gb

#The Flopperating System (tfos) - x86 32 bit Hobby OS