Nullium

[wild glen]

or not idk

move multiboot2 entry to userland

[wild glen]

im done with the powerpc bullshit ima go back to actually writing code

[hushed briar]

Fire

[wild glen]

all hail somewhat working linux

(it breaks sometimes)

[wild glen]

ok so i moved grub to userspace for efficiency and speed

the gdt is now on a separate computer and has to be loaded through a network protocol at 1 byte per hour

if an error occurs a robot spawns in and lights your house on fire and shoots you in the head afterwards

also trying to add paging now

[wild glen]

be me
do thing correctly
triple fault

[wild glen]

ridiculously small font

[fluid moon]

A

[gloomy musk]

very smol font lol

[wild glen]

its for a text mode

you'd see the whole universe on a single screen

C

should i make it larger though

[gloomy musk]

it is a bit too tiny ye 😅

[wild glen]

im gonna use the default system font for the terminal actually

its just big enough to be good

probably

[wild glen]

and/or small enough

[wild glen]

boot options, probably (system extensions are gonna be drivers, idea stolen from mac os obviously)

ok anyways time to actually make something

[wild glen]

text mode option removed im gonna add another way to enter it later

also progress will slow down

sadly

[wild glen]

nvm menu gone

instead

like apple

"HOLD [insert key] DURING BOOT HOPE IT DOES SOMETHING "

also theres visual output if your goofy keypress worked

i didnt steal anything i swear on my mom trust me

ok anyways im gonna sleep actual progress tomorrow hopefully

[lean relic]

poor mom

[fluid moon]

the fact that it is off-center hurts me

[wild glen]

fixed later because im stupid

[wild glen]

poor mom indeed (i apologized to her afterwards)

im gonna consider stealing less from apple

before they shoot me to death with their big epic patents

[wild glen]

my best idea is to remove the text because its said that extensions are off way before this

garphibc desgn

[wild glen]

im gonna remove some of the mac os ideas

its not really gonna feel like mac os but might be similar to use

[wild glen]

same what in the fuck was i thinking

[fluid moon]

https://tenor.com/view/skillissue-skill-issue-gif-22125481

[wild glen]

my brain works periodically

by that i mean it doesnt

im stupid

[wild glen]

ok im back to doing things except progress is actually gonna be slow

almost done with paging

[wild glen]

i fucked up paging

truly skill issue

multiboot2 so functional amirite

[wild glen]

ok so apparently i am not making a unix im probably making a mac os and dos hybrid in some form

because no posix compliance

i mean i want to put bash on it eventually as an app

and also a terminal text mode for nerds and shit

maybe thats the unix part

but idk

classic mac os and unix is literally a/ux so its still a/ux like

probably

[gloomy musk]

I mean... you could probably force bash to work on a non-UNIX system with enough effort... probably..... okay maybe not......... okay yeah this sounds sillier the more I think about it

(was classic macOS not a unix-based system?)

[wild glen]

(no its a shitty ass cooperative multitasking thing)

[gloomy musk]

ahhh, I see

[lethal summit]

was it paging or Paging that you messed up

[wild glen]

partially firmware

[lethal summit]

Paging is easy to mess up but paging is pretty simple

[wild glen]

whats the difference 😭

[wild glen]

what do i need to run bash other than like

stdlibs

[lethal summit]

paging is just mapping and umapping pages (the act of), whereas Paging is the process of managing pages, such as performing page aging, page-out, demand paging, etc.

[wild glen]

oh

loading in my os afterwards was the problem

[gloomy musk]

...no clue, never tried to port it, never will 😅 I would imagine it's pretty dependent on the system being UNIX though

[wild glen]

idk

[lethal summit]

so it was paging

ok 👍

[wild glen]

https://tenor.com/view/osdev-programming-qemu-kvm-code-gif-13251797

found this on wiki

for bash

[lethal summit]

the wiki literally suggests triple faulting as a method for reboot

i would not trust it

no joke check the reboot page

[wild glen]

already did

[lethal summit]

💥

[wild glen]

idk why the fuck would you triple fault to reboot

[lethal summit]

lazy people

[wild glen]

i mean its easy but like quadruple stupid

the only thing i use that for is some forms of debugging by "HMMM THIS DOESNT WORK MAYBE IF IT GETS TRIGGERED I SHIT COMPUTER"

[gloomy musk]

that's certainly one of the more egregious things the wiki has said XD

[wild glen]

the wiki says a ton of random bs other than this

that you need like 3000 years of experience

(i only programmed for directx 9 and a ti84+ ce calculator before osdev)

smells like total bullshit

[gloomy musk]

I mean, you certainly need to be comfortable with a lang like C, but asking for 12+ years of experience is a bit much lol

[wild glen]

true

i know enough c to understand everything well enough

12 years is insane

i feel like people gatekeep a lot to look smart

"you can only program if you know 3000 types of math and have 30 phd's"

[gloomy musk]

I mean, personally I don't think that was the intention of the paragraph, I think it was just worded a little poorly

[wild glen]

true i guess

it probably intents to warn dumbasses that "you cant make an os without knowing shit in a low level language" (RIS )

except ris

it should be worshipped

[fluid moon]

reasonable, who needs ACPI

[wild glen]

who needs to pulse the ps/2 keyboard

you can just make your processor lose its shit

[wild glen]

making a physical memory allocator

paging later

(i know its fucking required or yall are gonna kill me)

[gloomy musk]

I mean.... you don't have to do paging if you don't want to

[wild glen]

let's overwrite the kernel!!!

ironically classic mac os allowed itself to be overwriten

[gloomy musk]

Segmentation

[wild glen]

there were no defenses

like actually 😭

[gloomy musk]

wait really

[wild glen]

yeah lmfao

BeOS started like this

it literally overwrote mac os

[gloomy musk]

that's funny XD

[wild glen]

apple only unfucked it when mac os 10 came out

it was a unix now

[wild glen]

i spent like 400 years figuring out how to deal with the kernel in the memory alloc thing

i figured it out just allocate pages for the kernel (damn im stupid)

divide end addr of kernel by page size and then take that many pages

not actual paging its currently just physical memory allocation

not a lot of images because no visual changes

[wild glen]

did someone say something

i got pinged

[fluid moon]

me :3
but I had just woken up, so I was probably mistaken and deleted the msg

[wild glen]

https://cdn.discordapp.com/attachments/1260300556950442095/1401593482640097391/pyro-tf2.png-2.gif

what was the message though im interested

[fluid moon]

it was something like "you should map some memory instead"

[wild glen]

fair enough

[fluid moon]

https://tenor.com/view/sonic-wave-gd-geometry-dash-extreme-demon-meme-gif-4187432703063583914

[wild glen]

i has have int to ascii (binary???) converter

how do i do a regdump though

nevermind i doubt i need a register dump

i still have an itoa

[wild glen]

is this deadlocked

guys everyone who wants to actually read the github code i am sorry for my war crimes

jk i got the best code ever

i worked at blizzard for 90 years

[fluid moon]

for interrupts?

[wild glen]

uhhh idk i think i recall some people actually dumping processor registers in a kernel panic

[fluid moon]

you want an assembly stub to save registers to the stack and pass a pointer of it

[wild glen]

mk

[wild glen]

FUCK

[hushed briar]

W

[wild glen]

FUCK

[wild glen]

itoa works now

why does it end at 0x0

wtf is my linker on

crash screen

unlike the regular one (this one has to be toggled on by user) it has more info

grammatical error is intentional i think

anyways time to do the funny memory alloc

i needed to know at what addr the kernel ends

made an itoa and this crash screen

time to stop using that fucking crash screen for tests

thought this is like ultrabullshit for a moment

apparently its not

convert this to decimal then to megabytes

terminal on the left to prove that its 2 gigs

what if i give it 4

new number

damn

LETS GOOOOOOOOOOOO

[wild glen]

i forgot i have base10

anyways if anyone wants to know what the numbers mean look to the code screenshot

shit the numbers are cut off

now not

the left numbers are in kilobytes

the right ones are in bytes

top is total amount of ram (no shit)

[fluid moon]

so skibidi

[wild glen]

memory management isnt as painful when you know what in the fuck you are working with

[fluid moon]

do you output to COM1 or is the UI your only way to show text?

[wild glen]

the ui

[fluid moon]

bc you should consider having serial communication to the host

[wild glen]

idk how to output to com1

[fluid moon]

it’s just I/O, you outb an initialization sequence that sets the frequency to COM1 and then you’ll be able to use the port as a bidirectional text buffer

[wild glen]

fair

can it be used for like communication between multiple pc's

[fluid moon]

don’t rely on it as a user input source, use it to dump info

[wild glen]

mk

[fluid moon]

I think arduino heavily relies on it for software flashing

[wild glen]

idk

[fluid moon]

.

[wild glen]

ok for actual communication id probably just add ethernet drivers later

fair enough

[fluid moon]

my logging functions print both on the “UI terminal” and on serial

[wild glen]

im gonna make an os log program later

because theres no terminal text mode

[fluid moon]

yeah, just setup a BUS ethernet topology

[fluid moon]

you should still be able to log info on serial

it can save you a lot of debugging time

[wild glen]

true

[wild glen]

is ethernet bidirectional? (googled it, yeah it is)

[fluid moon]

10base2

[fluid moon]

depends though

I’m not a networking expert

but

[wild glen]

me neither

i really just want to send bytes between computers

(a protocol™ :0)

[fluid moon]

you should be able to send data to a beacon (likely your router) or another another device if you’re using a bus topology

[fluid moon]

bus topology is cool and mainly relies on protocols rather than hardware configuration

[fluid moon]

I studied it in school :)

[fluid moon]

there are lots of layers when treating networks, unfortunately

[wild glen]

https://tenor.com/view/tense-table-smash-mad-gif-13656077

a

[fluid moon]

LAN communication it’s already complex by itself, on top of that you have addressing protocols like IP, transmission protocols like TCP and application layer protocols like HTTP

[wild glen]

oh god

[fluid moon]

at the end you’ll end up with a cascade structure that becomes more abstract layer by layer:

LAN transmissions (a lot of theory, lots of hardware stuff)
Addressing (theory and implementation)
Transport (mainly algorithms implementation and some security related development)
Applications (relatively the easiest part, you just follow protocol specifications and implement them)

[wild glen]

basically app wants to send thing, it goes into a queue, then it makes it into like a valid format, then it sends it through the driver? idk this is what it looks like to me

getting stuff is just that but in reverse

[fluid moon]

that’s the TX queue in ethernet comms

[wild glen]

a

damn

[fluid moon]

you want to send a packet to another host (likely the router) so it’s wrapped into a {data,length,flags} structure and put in the transmission queue

[wild glen]

sounds ok enough

[fluid moon]

that’s all I know about detailed packet transmission

I’m still learning as I write an ethernet driver

[wild glen]

idk it sounds understandable in theory

[fluid moon]

I might have got it all wrong so don’t quote me on it, if you really want to learn more read some specification or ask someone more fond :)

[wild glen]

https://cdn.discordapp.com/attachments/1260300556950442095/1401593482640097391/pyro-tf2.png-2.gif

more memory management:

top number is pages required for kernel
bottom number is how much memory the kernel uses

is it me or is that a shit ton

4kb pages btw

c does floor divisions i might have to round it up later

that is um

mildly different

jesus christ 😭

that is even more of a shit ton

wait i divided it by a regular 1000

not the hex 1000

me dum dum

[wild glen]

ok its 1066 instead of 1065

yay

i need 1066 pages (damn)

is that a lot or did i do everything right

it includes like literally everything

like the stack and all that

[fluid moon]

you could ceil divide: #define CEIL_DIV(x,y) (x + y - 1) / y

[wild glen]

alr done

its rounded up now

[fluid moon]

coolest C macro

[wild glen]

real

[wild glen]

more memory progress tmrw

you people are smart and know your shit but like i have a bitmap of 4 kb pages and like it can allocate groups of these pages and they haved to be freed otherwise memory leak

(i like talking im sorry)

i want a separate bitmap that instead tells which task took which pages

also its physical memory only currently

virtual memory is later

[wild glen]

so if you decide to kill the task it will de allocate everything

hopefully

if something takes too much memory my only suggestion is to kernel panic the system

or just say "you took too much memory go end your apps or some shit"

windows does that

i think

triple fault it

[lethal summit]

bitmap allocator 💔

[wild glen]

give me a better one then

then why

[lethal summit]

buddy allocator

[wild glen]

:0

[lethal summit]

bitmaps are egregiously slow once you have more than a few megabytes of memory

[wild glen]

i guess

[lethal summit]

no joke it can take multiple seconds to scan through an 8gb address space

they also dont provide any alignment guarantees

buddy allocators by default align to the size specified

[wild glen]

fair

how does it work

[lethal summit]

u should read the dedicated articles and implementations of it since those provide much more concise and thought out info whereas me yapping would sound like an unhelpful yapfest lol

[wild glen]

fair

is it multiple bitmaps layered on each other

[lethal summit]

but the idea is you have blocks that you allocate from of different sizes and when two adjacent same size blocks become free you merge them together into a larger block

reduces fragmentation

[wild glen]

oh so a list of blocks

[lethal summit]

basically yea

[wild glen]

im gonna try making that in a simpler programming language first

because im stupid

[lethal summit]

faster + alignment guarantees + reduced fragmentation + all the cool kids use it

it's a win win

[wild glen]

linux etc

[wild glen]

true

wait im allocating pages though

ima still use it idk

oh

WAIT HOLY SHIT

i wanted to make that at 3 am

or something

idk its partitions

[lethal summit]

💥

[rocky elbow]

@wild glen checkout my gui

[lethal summit]

kernel window manager

[wild glen]

most informative kernel panic

the color palette is a bit goofy but its cool

[wild glen]

guys rip the 4 kb of ram that is never going to be used in the os

and the first two megabytes of ram as well

hlep i need this number divided

what can it be divided by

actually fuck it more than 4 kb is lost

at least 8

[lean relic]

https://tenor.com/view/budanov-gif-3507620829506990597

[wild glen]

Siegma

[wild glen]

idc about anything until this number is divided

or not all memory would able to be used :(

fuck it im fine with losing 8 kb

its not that much

instead of 1047920 pages it now can scan 261980 instead (buddy allocator, i think, it's like a directory tree)

do i need to reduce that number more

[lean relic]

1

[wild glen]

no shit

[lean relic]

523961

[wild glen]

yes shit

[lean relic]

https://tenor.com/view/black-guy-suit-black-guy-funny-guy-in-suit-skuffed-notskuf-gif-25390799

[wild glen]

he smort

math

anyways

im doing more directories

directories™

the buddy allocation of all time

(i know this is 4 gb in pages, some of them are going to be blocked out depending on how much memory is in your pc)

(this should work fine for 32 bit)

( i think idk you people reading this know way more than i do )

[wild glen]

i suck at thinking so i just booted up the fucking scratch editor

idk is this how buddy allocators work im braindead

shouldve used a spreadsheet editor actually

i might be making something else

whatever im just gonna write a fucking memory allocator on scratch first

then port it

im geniunely too stupid to write it in c from the start

[viral leaf]

Just write in pseudo language

[wild glen]

na

[lean relic]

^

if i cant get figure something out i write it in psuedocode, sometimes port parts of it to a scripting language

like python

[fluid moon]

LMFAO

gold message

[fluid moon]

are you writing a PMM or a heap allocator?

[wild glen]

uhhhhhhhhh

currently a pmm

i wasnt at home for like the last 2 hours

[wild glen]

oh wait im making something else

yeah im reconsidering life choices 🔥

buddy allocator time

[lethal summit]

goated

[wild glen]

i made a buddy allocator on scratch

context: lol no

ima post the project file once im done with this fuckass

its gonna be ported into my os

there's no structs in scratch i cant make an array of structs tho

its gonna have structs in c dont shoot me to death

to have it make more sense i have to use a sorting algorithm

wait i found a bug

j

[wild glen]

i have to do a sorting algorithm

F

its for easier merging of free memory blocks

oh wait

ok its not a traditional sorting algorithm

its a filter

it finds free blocks in the list

puts them on a list and then merges the ones close by

[lethal summit]

you typically just check if you can merge upon free instead of having this as a separate thing

let's say

allocate 1 page, call this LEFT

allocate 1 page, call this RIGHT

free LEFT, check if we can merge -> we can't merge

then

free RIGHT, check if we can merge -> we can -> merge it with LEFT

[wild glen]

true

what if free segments are fragmented tho

how the hell are you going to fix something like this

(scratch because im still not done with the testing shit)

[lethal summit]

you always allocate with sizes that are powers of two

that's the caveat of this all

[wild glen]

fair

[lethal summit]

you merge 1 and 3 to make 0..100 of size 100 and type free

[wild glen]

true

ok i think i figured out how to do that

thats what i was going to do either way but i think i figured out how to program that

[lethal summit]

have a blast 😃

[wild glen]

aight

[fluid moon]

why don't you use a bitmap

it's very unlikely that you'll need a buddy allocator for pmm

[wild glen]

he said the reason why afterwards

[fluid moon]

having a buddy allocator as PMM it's a nice flex but also a pain inthe ass to write

[wild glen]

maybe

idk

[fluid moon]

this could be considered incitement to suicide

[lethal summit]

false propaganda from Big Bitmap

[wild glen]

memory allocation wars

[lethal summit]

it took like 50 minutes to write gng

ts not hard

[fluid moon]

overkill

[lethal summit]

now my buddy allocator is not good and it has a Big Function Lock, I'm planning on parallelizing it soon

[wild glen]

same

my only problem is merging really and i think i solved it

[lethal summit]

but it is a buddy allocator

[lethal summit]

I'll also be doing processor topology representation in software soon (getting it figured out) so I can make it NUMA aware for fun

[fluid moon]

I like segmentation when allocating heap instead

[lethal summit]

why is it overkill

it genuinely isn't bad

"Having performant code... overkil...."

💔

[fluid moon]

okay it's better, I'm lazy

https://cdn.discordapp.com/attachments/1307394519410540575/1403732634328236112/togif.gif

[wild glen]

booting from rom

[lethal summit]

there we go

yep buddy allocators are Very Fun

have a blast pal

[wild glen]

propaganda solved 👍

[fluid moon]

Booting from ROM...

why isn't kernel.c booting

plus who tf writes "file.C" (with capital C)

[wild glen]

braindead people

[lethal summit]

michael soft

[wild glen]

MICHAEL SOFT?!?!?!!?

theyre braindead anyways

[lethal summit]

i said michael soft not microsoft

microsoft doesnt code like that

michael soft might

[fluid moon]

his brother, michael hard, works on electronics

[lethal summit]

facts

[wild glen]

john michael is the leader of the board of directors at michael soft

and/or the ceo

[fluid moon]

binbows

[sour pollen]

literally no.

[fluid moon]

.

IT IS better than bitmap

[sour pollen]

even then, a simple freelist allocator is much easier to get right than a bitmap allocator anyway

[gloomy musk]

Freelist my beloved ❤️

[wild glen]

memory allocation wars

[sour pollen]

busting the bitmap propaganda

[wild glen]

protesting against the bitmap kingdom

absolute revolution

world war 3

[fluid moon]

what complexity do freelist and buddy allocator have per allocation?

[wild glen]

buddy allocator starts with one big ass free space, then it just gets cut and the new space is treated as "used"

[wild glen]

see this image

merging two free entries is a pain in the ass though if you dont know how to do it

[hushed briar]

Buddy is just division by 2

Free list is just… a free list

[fluid moon]

so O(log N) worst case

[hushed briar]

Yes

I have an impl

If you’re interested

[fluid moon]

show me, I'm falling for buddy propaganda

[wild glen]

i also do except its on fucking scratch

(its gonna get ported i have no other option)

[hushed briar]

https://github.com/amar454/floppaos/blob/main/mem/pmm.c

[sour pollen]

freelist is always O(1) for alloc/free, atleast if you only do page sized allocations. You can get away with scanning the list though, for higher alloc sizes

[wild glen]

this is the most stupid code ever

anyways my buddy allocator works i can port it now 💥

(i should port my os to scratch )

this is the scratch project file for the buddy allocator test thing

yall can do whatever you want with this (its a theoretical simulation of memory)

[fluid moon]

https://cdn.discordapp.com/attachments/1406744559349796876/1406744568975855776/buddy_allocator.gif

[wild glen]

golden

im not sure if this is public shaming or is this the most random shit ever

[fluid moon]

nah it's just a meme

[wild glen]

fair

[wild glen]

Dam

[wild glen]

there's also the "Do not" error

[fluid moon]

https://cdn.discordapp.com/attachments/1151529419589025892/1406449523030491176/makefile.gif

[wild glen]

https://cdn.discordapp.com/attachments/1159299233141248011/1405269870160908429/togif.gif

[fluid moon]

https://cdn.discordapp.com/attachments/1307394519410540575/1403732634328236112/togif.gif

[wild glen]

booting an uncompiled c file is crazy

even though this was already discussed

its still insane

[fluid moon]

going through a rewrite might be the best thing I've ever done

[wild glen]

of your memory manager?

[fluid moon]

of the whole kernel

[wild glen]

oh

[fluid moon]

I've dumped the old one

the codebase became spaghetti code

[wild glen]

the best thing i did is to scrap my shitty vga text mode os because half the code was stolen

this one is my actual try

[fluid moon]

on top of that uacpi was crashing, something was breaking somewhere and I couldn't find out where

[wild glen]

rip

[wild glen]

75%

i was also following the shittiest tutorial on earth

[fluid moon]

I can finally define a virtual memory layout

your progress report thread became misc0

[wild glen]

i permit it lmfao

this went from buddy memory allocation to kernel rewrites

if you want we can go to misc 0 idk

[fluid moon]

I think I'll go to sleep, keep reporting progress and I'll keep posting in the thread :)

[wild glen]

aight

gn

do i still do buddy allocation with pages

fuck it ima do it with pages

The

ok anyways

[wild glen]

ported struct

I LOVE THE ITOA FUNCTION AISKJDHAFHQUJKRHQWUJQWMNDHGIJSEKFQWIL>WQKHJDQWHJKD

anyways i can add malloc now

i made it horizontal

i want to print all the values in the list thing out

debugging 👍

it allocated yay

function for it

it shouldnt return 0 but whatever

lemme make it not return 0

this is page allocation (palloc), malloc will return the actual memory address

idk if its called palloc usually

i just named it palloc because page allocation

[lethal summit]

windows nt code style

windows nt function call style

[wild glen]

mac os clone
looks inside
windows nt

it isnt intentional

how does winnt code even look like

[lethal summit]

https://github.com/microsoft/Windows-driver-samples/blob/e1bec22a08c3d4f176bef50f79dc76afc34ebbe6/filesys/fastfat/workque.c#L157

windows nt code style

[wild glen]

damn

i am now michael soft

[lethal summit]

true

[wild glen]

idk i remember seeing directx instructions like that

with these function calls

i think microsoft made them

these functions were long as actual fuck

idk i just code functions with a lot of parameters like that now

[viral leaf]

No nt style would be

NT_RESERVE_PAGE_ALLOCATOR_HELPER_FUNCTION(...)

[lean relic]

NT_RESERVE_PAGE_ALLOCATOR_HELPER_FUNCTION(
a,
b,
c,
d,
e,
f
)

[wild glen]

yo guys so like heres my plan (in order)

• memory management
• more memory stuff (paging)
• scheduler
• hard disk access
• window management
• file systems
• programs

maybe swap fs and window management

[lethal summit]

did you already do the scheduler?

[wild glen]

no

[lethal summit]

so that is like a really really big deal

do you want multitasking in this project

[wild glen]

ok yeah add the scheduler to that list then

[lethal summit]

and you'll also want a side dish of synchronization primitives to go along with it

condvars, mutexes, semaphores, blah blah blah

[wild glen]

mk

oh

the concept is simple

[lethal summit]

also you didn't seem to know what a spinlock was before this, I suggest you definitely look into multithreading and write multithreaded userspace programs with pthreads in C

[wild glen]

i use interrupts for the mouse pointer already

[wild glen]

a

[lethal summit]

sure the concept is simple but once you add many layers on top of it things can get complex

like with mutexes

[wild glen]

fair

[lethal summit]

a simple mutex is, well, simple

but if you have thread prioirities, then you might introduce priority inheritance, and then if you want pointer sized locks, you need per-thread turnstiles and a global turnstile hash table

[wild glen]

fair

[fluid moon]

cringe, I atomically lock the spinlock and stop the whole scheduling system

[lethal summit]

do you not use locks anywhere else

wdym

[fluid moon]

not currently

[lethal summit]

how is your os functional

[fluid moon]

I lock the spinlock in critical functions like allocators

the scheduler notices and doesn't switch task

[lethal summit]

what do you mean "The Spinlock" is this a BKL type thing

https://cdn.discordapp.com/attachments/829403502349647952/1406807383006646333/mtx.gif

[fluid moon]

it is a BKL

[lethal summit]

crazy

[fluid moon]

I'm not even planning on adding SMP

[lethal summit]

your loss

[fluid moon]

I will fall for SMP propaganda someday

but not today

[wild glen]

what in the fuck happened

also

i am tragically informing you all

about the lack of the cd-rom, the box, the manual, and swiss cheese in the nullium cd set

no warranty included either

in the digital edition swiss cheese has to be removed as well

i couldnt afford the free version of chatgpt so i have to inform you people about this tragic loss not with an essay

but instead 8 discord messages

to protest this crime cheese react the thread

or something

[hushed briar]

Nice

[wild glen]

i wanted to add tetris except the company making it will not like it

at some point

in the os bundle

which is a major ass disappointment

[hushed briar]

Did you fall for my buddy propaganda

[hushed briar]

That’s not an os thing tho

Like

Just create a kernel

And a userspace

Then just make your own Tetris game in c in userspace

Running Tetris in the kernel is

Bound to cause maybe a few issues

[wild glen]

i know

thats what i wanted to do

the company will still not like it

the tetris company

i have to license it for all my money

times 5403002

idk i dont want to run tetris in the kernel no matter how goofy asf that would be

[wild glen]

he did i think

i remember a guy writing a tetris os

there was a youtube video or smth

[hushed briar]

yeah

there is thois popular programming youtuber who did it

[wild glen]

fair

[hushed briar]

although

its dogshit

like

its a cool concept but

the code is shit

[wild glen]

youtubers suck at programming always

except terry davis (brainfart)

he was a streamer though

[hushed briar]

terry davis wasnt bad

shit fart

[wild glen]

fart shit

idk youtubers absolutely suck ass at coding

idk how

idk why

[hushed briar]

you should help me with the ui of my os

at some point

when i reach userspace again

[wild glen]

why me lmao

[hushed briar]

because your shit looks great

[wild glen]

thank

lemme finish my homework rq i can actually finish my crappy mem allocator (PALLOC MAUAHAHAHAHAHAHAHAHAHAH how did people live with 6 letter labels)

[hushed briar]

no youre good, just at some point id be interested in that

if you need help with your allocator let me know

i have the buddy system implemented and i know a decent bit about it

[wild glen]

fair

i have a scratch project for it for some ungodly reason

you can scroll up and find it

[hushed briar]

yeah i saw

sometimes we need visual representations for some concepts

i used scratch to help me implement virtual memory

no shame

[wild glen]

easy to understand lego programming language that doesnt need to compile

shit has more debugging tools than visual studio

[hushed briar]

yeah its elite

[wild glen]

fr

[hushed briar]

i designed my clone pagemap

in that

because

it was fucking weird

[wild glen]

damn

[hushed briar]

int clone_pagemap(uintptr_t src_base, uintptr_t dst_base, size_t size) {
    size_t pages = _align_up(size) / PAGE_SIZE;

    for (size_t i = 0; i < pages; i++) {
        uintptr_t src_vaddr = src_base + i * PAGE_SIZE;
        uintptr_t dst_vaddr = dst_base + i * PAGE_SIZE;

        uint32_t src_pd_idx = _pd_index(src_vaddr);
        uint32_t src_pt_idx = _pt_index(src_vaddr);

        if (!RESOLVE_RECURSIVE_PD[src_pd_idx].present) continue;

        pte_t* src_pt = RESOLVE_RECURSIVE_PT(src_pd_idx);
        if (!src_pt[src_pt_idx].present) continue;

        uintptr_t src_paddr = src_pt[src_pt_idx].frame_addr << PAGE_SIZE_SHIFT;

        page_attrs_t attrs = {
            .present = src_pt[src_pt_idx].present,
            .rw      = src_pt[src_pt_idx].rw,
            .user    = src_pt[src_pt_idx].user
        };

        void* dst_frame = pmm_alloc_page();
        if (!dst_frame) return -1;

        flop_memcpy(dst_frame, (void*)src_paddr, PAGE_SIZE);
        attrs.frame_addr = ((uintptr_t)dst_frame) >> PAGE_SIZE_SHIFT;

        if (map_page(dst_vaddr, (uintptr_t)dst_frame, attrs) < 0) {
            // Rollback previous mappings
            for (size_t j = 0; j < i; j++) {
                uintptr_t undo_vaddr = dst_base + j * PAGE_SIZE;
                unmap_page(undo_vaddr);
            }
            pmm_free_page(dst_frame);
            return -1;
        }
    }

    return 0;
}

cancer fr

[wild glen]

ok im done with the homewo- WHAT THE FUCK

cancer indeed

[hushed briar]

just look at any vmm.c in any os

its just fucked

i added a bunch of protections features too

my shit is 1k lines of code for no reason

[wild glen]

ok now im scared of adding vmm

[hushed briar]

nah

you only need 2 functions

[wild glen]

oh ok

[hushed briar]

for a "vmm"

[wild glen]

also what do i do with the pmm once i have virtual memory

it shouldnt just vaporize

[hushed briar]

wdym

no

the vmm is just an abstraction of physical memory

[wild glen]

true

[hushed briar]

so you need a pmm

[wild glen]

i still allocate physical memory

[hushed briar]

to give you pages

yes

[wild glen]

then make it pretty to make it virtual

[hushed briar]

i guess you can say it like that yes

[wild glen]

idk

[hushed briar]

you just need to be able to map pages to a virtual address

which isnt crazy

[wild glen]

not sure but iirc paging or whatever allows to like block off access to the kernel in user space

[wild glen]

fair

[hushed briar]

yes

isolation

[wild glen]

from yesterday i have this absolute fart

[hushed briar]

beatiful

[wild glen]

doesnt map actual memory addresses actually

[hushed briar]

valid? free? 🗣️

[wild glen]

im gonna add a parameter in the struct for that

[wild glen]

valid? free? owner size start

mem location

first is base16 second is base10

too much effort for a debug table that will get sent to the shadow realm shortly after using it

i need to write a spreadsheet editor for nullium after i get to userspace and shit

size is in pages btw

[hushed briar]

I think you’re overthinking alot of this

Wanna look at my impl

[wild glen]

i mean sure ig

the table is helpful though

its all debugging and stuff

[hushed briar]

No this is good

https://github.com/amar454/floppaos/blob/main/mem/pmm.c

This file is a banger

[wild glen]

truly

i made an allocator that cuts exactly how much you ask instead of just the middle

[hushed briar]

Oh

I see

That isn’t a buddy then tho

[wild glen]

:0

[hushed briar]

Because buddy blocks are supposed to be powers of 2

But

It’s fine lol

No one cares

[fluid moon]

i did

[wild glen]

then what did i make lmao

[hushed briar]

A memory allocator

Which works

[wild glen]

i made my own algorithm i should trademark it

[hushed briar]

So who cares

Idk exactly what that would be called

Someone with more knowledge than me could probably tell you

[fluid moon]

next is an actual VMM that divides memory per region

[wild glen]

i mean its like hard disk partitions

[hushed briar]

Day #294839272837 of people not knowing what a vmm is

[fluid moon]

that’s the only place I know where I should put a virtual address region map

[hushed briar]

That’s just a free list allocator

[wild glen]

damnit

[hushed briar]

Which is cool

[wild glen]

not damnit

[hushed briar]

Nothing wrong with those

Shit, Linux used one for years

[wild glen]

tried to write a buddy allocator but i got a freelist ;-;

im this bad at understanding things

[fluid moon]

educate me please

[hushed briar]

It’s an abstraction of physical memory

[wild glen]

someone teach me to read outside of school man this is absolutely braindead

lmfao

anyways

[hushed briar]

Meaning that it presents memory to applications and the kernel in a way to where each process has its own “address space”

[fluid moon]

which is where I should divide virtual address spaces, no?

[fluid moon]

exactly

[hushed briar]

What do you mean divide virtual address spaces

Dividing what

[wild glen]

(#osdev-misc-2?) /j

[fluid moon]

mapping more than dividing

[hushed briar]

Well then sure

[hushed briar]

Mb mb

[wild glen]

nah its fine

yall can talk

[hushed briar]

My thread has a lot of that happening sometimes

[wild glen]

fair

[hushed briar]

It’s mainly just me being schizophrenic tho

[wild glen]

shit happens 😭

[fluid moon]

cool

terry

[wild glen]

elephant simulator

https://tenor.com/view/templeos-terry-davis-gif-19184988

[wild glen]

for easier access this is reversed

it grows from the top

[hushed briar]

Good

[wild glen]

indeed?

[hushed briar]

Well

Now the big question

Does it work?

[wild glen]

uh

i think

lemme check

if i can alloc then i can go make a free-er

[hushed briar]

Lol

[wild glen]

The

id say it works

would these interfere though

actually nah ion think they would

tell people "size means end address"

[wild glen]

FREE

FREE MEMORY FOR FREE

time to merge it

[hushed briar]

Pmm lore

[wild glen]

ibm ultrabay specifications when

[wild glen]

it can merge now

before:

after:

time to test it

add a few memory segments then do stuff to them

so the merger works fine

except it doesnt merge all the free segments

ж

[hushed briar]

Why is size the same

[wild glen]

are you sure

look closer

im done with page allocation

yay

do i do paging next or do i add a malloc()

[hushed briar]

Wait nvm

[hushed briar]

Paging

You need that anyway to make a proper malloc()

[wild glen]

oh

fair

lemme just do page allocation for kernel space

[hushed briar]

Yeah that’s fine

Although

A good kernel heap allocator is really nice

Because imagine you need a 30 byte allocation

You would do a page alloc

And waste

4066 bytes

With only 3 being used

Internal fragmentation

[wild glen]

mk

not sure if i add one but thanks for the info

[hushed briar]

Well

If you wanna waste all your memory sure

No actual sane kernel does that tho

[wild glen]

rip

[wild glen]

do what

not do the heap allocator or do it

[hushed briar]

Not do it

[wild glen]

i think i can do a heap actually

theres a point in it

other than you telling that all the cool people are using it

github push

added physical memory allocator

[hushed briar]

No it’s just logically better to not waste a page size worth of memory every time you need an allocation

Like

Imagine how many small allocations there is for thread structs

Imagine each of them taking a whole ass page

[wild glen]

im gonna make system processes use cooperative multitasking

this only includes the file explorer (im gonna call it the browser for shits and giggles) and system menu and error messages really

nothing else

[lethal summit]

no preemption?

crazy

[wild glen]

thats for literally everything else

[lethal summit]

huh

[wild glen]

absolutely everything except the fucking file explorer is gonna be preemtive

[lethal summit]

what

[vocal root]

did classic macOS/aux even use pre-emptive multitasking

[gloomy musk]

I don't think so?

[wild glen]

it didnt lololol

[lethal summit]

nope

[vocal root]

so if this is a/ux style co-operative multitasking is fine 😭

[wild glen]

true

a/ux was a unix though

[lethal summit]

oh right I forgot about that part of the project

ok then lgtm Have Fun

[wild glen]

i just realized this shit will not be a unix 😭

maybe i should just make it a classic mac os-like os

im just gonna keep the classic mac os part of it

theres still gonna be a text ode

mode

and bash

and shit

[gloomy musk]

UNIX is overrated

[wild glen]

indeed

Nullium - a classic Mac OS like OS (some unix elements present)

more accurate

its gonna have preemptive multitasking

but the file explorer is just built different

[wild glen]

how do you load paging

or like a page directory

how do you even use this shit afterwards

im confused

the article on the wiki (osdev.org) uses gas while i use nasm

[lethal summit]

just convert the syntax

[wild glen]

wtf is 8(%esp) supposed to mean in nasm

i can understand everything else

[lethal summit]

mov eax, [esp + 8]

[wild glen]

oh

gnu is evil

full stop

thx a lot

it triple faults

https://tenor.com/view/tense-table-smash-mad-gif-13656077

[gloomy musk]

[wild glen]

tmrw im gonna fix my shitty paging

apple's liquid glass looks like total farts imo so im gonna make a cooler one

like excuse me what the fuck are these icons

[wild glen]

should the menu panel be at the top of the screen

note that even if it was at the bottom it functions nothing like the windows one

fuck it im moving it up to the top

[wild glen]

paging shits itself

as usual

i hate it

i hate it and i need it

[gloomy musk]

😭

[wild glen]

i dont want my fucking kernel to be rewritten by a program that wont be exactly cash money

i really just want this specifically

i dont care for anything else i want isolation from the kernel

[gloomy musk]

Segmentation

[hushed briar]

Don’t even

Stop right there

💀

[gloomy musk]

🤣

ye I'm obviously joking, pls don't use segmentation

[wild glen]

i forgot what that

[gloomy musk]

good, keep it that way

[wild glen]

https://tenor.com/view/gutterman-ultrakill-me-when-me-when-the-time-for-bed-gif-9428411444002163535

[hushed briar]

The memory management strategy before paging

Even tho paging existed for 30 years before intel started supporting it

[wild glen]

oh

[hushed briar]

We took a big ass step backwards

[wild glen]

intel chose to be smartasses?

ok yeah that

[hushed briar]

Yeah

We had paging computers in the 60’s

And Intel didn’t support it until the late 80’s

[wild glen]

hardware or software

[hushed briar]

Uh

[wild glen]

386 moment

[hushed briar]

That’s what I’m targeting

Although paging was still optional

You could still do segmentation

For example Microsoft DOS used segmentation

[wild glen]

dos is like the first x86 operating system it just uses whatever came out first

fair

[hushed briar]

I wouldn’t even call Dos an os in the traditional sense lol

[wild glen]

wdym

[hushed briar]

I mean I guess it’s an os but it’s pretty useless compared to most operating systems

[wild glen]

its still an os

operating systems are bloated to hell these days

even linux

[hushed briar]

There’s no multitasking, no memory protection of any kind, only ring 0, lack of hardware abstraction

[hushed briar]

Linux is notoriously bloated lol

[wild glen]

ok true lmao

[hushed briar]

I main the core kernel code is good for sure

But drivers/

Is just fucked

Also just the naming on the drivers too

[wild glen]

that folder is ultimate abomination of intel, nvidia, ati, and 5000 shitty programmers

[hushed briar]

Wtf is half of this shit

[wild glen]

half is an understatement

this is like reading chinese

[gloomy musk]

what the heck is allat

[wild glen]

nullium goal 2: dont make the driver folder look like the one in linux

[hushed briar]

Mine is somewhat sane

[wild glen]

(the first one is to make it look good compared to mac os 26 (just call it 16 god fucking damn it))

[wild glen]

sane

[gloomy musk]

okie, see, that makes sense

[hushed briar]

I feel like these folder names should be EASYto understand

[wild glen]

mine looks like this

[hushed briar]

Yeah that looks fine

I prefer folders for drivers

[wild glen]

some drivers are in the arch folder

these are somewhat universal

as in they work both for x86_64 and i386

[hushed briar]

Mmmm

[wild glen]

allegedly

(outb always existed)

[hushed briar]

You drivers shouldn’t be arch specific period

[gloomy musk]

I have yet to get to the point of having any drivers lol

[wild glen]

is gdt a driver

its really just boot init stuff

[hushed briar]

No it’s an array of integers

[wild glen]

congratulations

i did everything right

in the arch folder is just paging (ptsd), gdt, idt, system entrance etc

[hushed briar]

That’s what should be there yeah

Drivers aren’t arch specific

[wild glen]

i guess

idk if theres an outb instruction in like powerpc

[hushed briar]

There’s an equivalent

Write a byte to a port

That is universally available

[wild glen]

thats what outb is

so the evil ps2-x86 will work on literally anything if outb gets ported to [insert whatever]?

awesome

[hushed briar]

Something like that

Yes

[wild glen]

cool

making an x86 thing currently though

the most no shit thing i said

[hushed briar]

X86 is definitely a mess tho

I feel like any big hardware/software will be messy

Like the Intel SDM is extremely cumbersome

To read

And a lot of the mnemonics don’t even make sense

[lethal summit]

false

ARM has no in/outb

the IRQ abstraction here is also x86 specific

[hushed briar]

Not an exact equivalent

But you need some way to write to data ports lmao

[lethal summit]

which means it wont work at all

those ports are MMIO on arm

[hushed briar]

Oh yeah mmio I forgot arm does that

[lethal summit]

you'd want ps2_read/write functions that compile to in/out on x86 and whatever else on ARM

so this code is correct for being PS2-x86.c

[hushed briar]

Yeah true but it wouldn’t work for arm

[lethal summit]

you wouldnt entirely rewrite it for Arm, you would just introduce arch-agnostic abstractions like a ps2_read/write

[hushed briar]

So it would be a mmio write and read rather than outb and stuff

Makes sense

[lethal summit]

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

[wild glen]

IT TRIPLE FAULTS AGAIN

IM GONNA FUCKING KILL SOMOENE

is there any fucking sensible instruction on paging anywhere on the fucking internet

https://tenor.com/view/tense-table-smash-mad-gif-13656077

[hushed briar]

Not really

Show me your code

Maybe I can help

Nah but fr I can try to help

[wild glen]

a sec

unsigned long *page_directory = (unsigned long*) 0x9C000;
unsigned long *page_table = (unsigned long*) 0x9D000; 

extern unsigned long read_cr0();
extern unsigned long read_cr3();

extern void write_cr0(unsigned long);
extern void write_cr3(unsigned long);

void init_Paging(){
    unsigned long address=0;
    unsigned int i;

    // only first 4mb of memory
    for(i=0; i<1024; i++)
    {
    page_table[i] = address | 3;
    address = address + 4096;
    };

    page_directory[0] = (unsigned long)page_table; 
    page_directory[0] = page_directory[0] | 3;

    for(i=1; i<1024; i++)
    {
    page_directory[i] = 0 | 2; // attribute set to: supervisor level, read/write, not present(010 in binary)
    };

    write_cr3((unsigned long)page_directory); 
    write_cr0(read_cr0() | 0x80000000); 
}

followed the thing on the bona fide osdev site

this is total farts

global read_cr0
read_cr0:
    mov eax, cr0
    retn

global write_cr0
write_cr0:
    push ebp
    mov ebp, esp
    mov eax, [ebp+8]
    mov cr0,  eax
    pop ebp
    retn

global read_cr3
read_cr3:
    mov eax, cr3
    retn

global write_cr3
write_cr3:
    push ebp
    mov ebp, esp
    mov eax, [ebp+8]
    mov cr3, eax
    pop ebp
    retn

assembly

your implementation is somehow way simpler than whatever the fuck everyone else is doing

@hushed briar