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

[bold grove]

damn they got shit on

[viscid zephyr]

ho chi minh just wanted it more

he was fighting invaders of his country for literally 60+ years straight

[bold grove]

no wonder they named a city after him

[viscid zephyr]

from his 20s until he was elderly

[bold grove]

dang... what a life

[viscid zephyr]

he was an admirer of george washington because they were both liberators of their people from colonial oppression and he had a portrait of washington in his office

while we were actively at war with him

[bold grove]

yeah.. but at that point the united states government had already strayed so far away from the ideas of george washington

GW would be rolling in his grave

[viscid zephyr]

ho chi minh's military career was objectively far more impressive than washington's though

[radiant grove]

lounge-2

anyways what did i miss

[bold grove]

the entire histry of the vietnam war

[radiant grove]

damn

[bold grove]

i mean washingtojn was just an average british military general

ho chi is a legend

[viscid zephyr]

we literally lived long enough to see ourselves become what we hated

[bold grove]

so fucking unfortunate

if anythingwe did some things that were much wrose than the british

like we intered random japanese people here

for what reason

[radiant grove]

"please no political parties"

[bold grove]

yeah what a joke

[radiant grove]

we yall just ignored him so hard

[viscid zephyr]

we also rounded up all the inuits in alaska and put them in concentration camps deeper in the usa as well so that imperial japan couldnt kidnap them and turn them into spies

ppl forget that one

[bold grove]

i remember reading about that yeah

[radiant grove]

im not responsible im not a citizen in usa

[viscid zephyr]

tbf imperial japan did actually kidnap a few alaskan inuits

they invaded alaska

[bold grove]

they did

alot of fucked up shit lmao

[radiant grove]

also japanese camps

i think that existed

[viscid zephyr]

only time mainland usa has been invaded since war of 1812

[prisma magnet]

Oh god HOW did this chat devolve from osdev to Japan invading Alaska

[bold grove]

the japanese committed their own holocaust

[viscid zephyr]

who cares

[bold grove]

based autistic people

[radiant grove]

i know

[bold grove]

yeah fr

[viscid zephyr]

OMG DISCUSSION IS ORGANIC I JUST REALIZED HTIS FOR THE FIRST TIME CUZ IM A BABY

[radiant grove]

im talking AMERICAN japanese camps

[bold grove]

yeah fr

no we must talk about page tables

[radiant grove]

uhh

[viscid zephyr]

also most discord moderators

[prisma magnet]

No we must talk about the population of North American River Otter and how it relates to MOSFETs

[radiant grove]

moving my eax to her cr0 till she paging init ... ok

[viscid zephyr]

nothing pisses me off more than discord mods seeing a discussion organically evolve into something off topic in an on topic channel, and nobody is trying to use the on topic channel for anything else, and they go "erm. please move to general"

thats just power tripping imo

serves zero purpose

just destroys an ongoing conversation (people never actually move, it always ends there)

for no gain

[radiant grove]

fr

[viscid zephyr]

other than ego stroking

[bold grove]

yeah fr

my thread is a safe space for any bullshit banter

[radiant grove]

x86 register tier list when

[bold grove]

im the moderator here fr

[radiant grove]

mine too

[viscid zephyr]

im glad the mods here dont usually do that but it has happened a few times

[radiant grove]

my thread always turns into osdev-misc-2

[bold grove]

yeah

[thorn dust]

it's pure poison

i really can't stand it

[radiant grove]

deadass

[bold grove]

its mainly people who arent mods doing it too

becuase 5they think they will get brownie points

but it is annoying when youre on topic and someone starts talking about something off topic over the original conversation

but if there is an actual conversation

[bold grove]

who cares

[viscid zephyr]

yes this is the problem

[thorn dust]

whenever i've seen a conversation redirected never once have i seen it survive the redirection

[bold grove]

exactlyyy

conversation is natural

[thorn dust]

some servers have this stupid bot that posts a link to the "correct channel" but it still doesn't do anytihing

[radiant grove]

true

even if its an on topic convo

[bold grove]

yo dont tell people irl who are talking to go to another room bexcause its not the "conversation topic X room"

[viscid zephyr]

or if someone decides to bring up something on topic while a conversation is ongoing that had become off topic, the former should of course take priority

but if neither of these are happening then i see no reason for moderators to become involved

[radiant grove]

release deadly neurotoxin otherwise

[small hound]

what colorscheme do you use in your terminal

i use campbell (from windows)

[bold grove]

campbell windiows

[prisma magnet]

This is why Discord needs to implement a feature to view the population of parrots

[small hound]

copycat

[viscid zephyr]

theyre called moderators bc theyre meant to moderate discussion. for some reason some of them decide to become terminators who terminate discussion

[bold grove]

isnt it the default

[bold grove]

yeah. the mods here are mostly good tho

some outliers but

not many issues here

plus the actual chat is moderated well

any obscene stuff/ slurs

gets taken care of quickly

i only talk in this server becuase so many otyhers have annoying ass mods

[prisma magnet]

ToaruOS has affected brie cheese in a lot of ways

[radiant grove]

i remember seeing a thread where some guy said the n word and then right after a mod is like "ok thats it the guy's getting banned"

he was never seen since

[thorn dust]

it's really such a dispiriting experience to be having a conversation and this happens

sometimes i think it means you rattled the moderators' sensibilities by having a different position to them, and they redirect out of spite

[viscid zephyr]

it can also be laziness

they get annoyed conversation is happening at all while theyre busy and cant pay attention

i think its pretty thoughtless and a sign of a terrible person to have petty power though

its incompetence plain and simple

dont tell THEM that though!

[thorn dust]

unfortunately it's widespread

[bold grove]

or you might get banned

[viscid zephyr]

theyll incompetently inflame the issue until they have an excuse to incompetently ban you

[thorn dust]

we all know exactly which server we have in mind when we talk about this kind of moderation, and exactly the conversations we've had that were terminated this way

[bold grove]

i was on this embedded systems server and i got banned because i suggested that the i386 isnt the best embedded processor

[prisma magnet]

Fufu and egusi has affected Bosnian OS Development Time rates by a LOT!

[bold grove]

💀

[prisma magnet]

PMAO WHAT??

[viscid zephyr]

there are a few mods here who rarely appear in chat and whenever they do its to kill a conversation by doing this

i wont name names

but theres a few mods here i literally only ever see when they do that

its infrequent but it occurs

[bold grove]

yeah we know who too

[radiant grove]

what

[prisma magnet]

let's not name them

but we know who

[radiant grove]

i dont and now i am curious

or i forgot idk

[prisma magnet]

Emacs is really nice for developing cat food devices

[thorn dust]

are you turkish, albanian, chechen, texan or what?

[bold grove]

guys im gonna stream on twitch soon

if you wanna watch

[radiant grove]

stream what and where

me interested

whats the channel name

[bold grove]

im gonna stream osu on my channel amarat

https://www.twitch.tv/amarat

[radiant grove]

oh aight

i found a mod for ultrakill that allows to change style rank names

[bold grove]

whats ultrakill

[radiant grove]

video game

idk

shooter game

shoot people

they die

doom eternal on steroids

and also lots of blood because lore (and gameplay) reasons

[small hound]

stream osdev fr

[radiant grove]

truly

[bold grove]

i was thinkking about doing that

but i realized it would be alot of me reading

and sitting

and banging my head on my desk

OSDev ™

[bold grove]

would you watch an osdev sstream

[small hound]

there are plenty that i already watch

so peak

[bold grove]

really?

who

[small hound]

https://www.youtube.com/renerebe

[radiant grove]

idk

[small hound]

Watch AI engineer use 100 CPU threads & 240GB RAM to finish porting AMD ROCm AI to RISCV-64!

[bold grove]

imo this is a cool use of ai

[radiant grove]

it might be boring as shit but idk

[radiant grove]

:0

[bold grove]

i remember binder made a functioning C compiler using LLM's

@hollow igloo

[radiant grove]

vibe coders i just dont understand

writing something cool with ai as a shits and giggles thing is really cool

but making it a career is just kinda weird

[bold grove]

making it a career is crazy

yall ever just have intrusive thoughts of deleting your whole repo, banishing it from your brain, and starting over?

[radiant grove]

i do

[bold grove]

smp is really making me consider it

because i wrote alot of my os without smp in mind

and now its just a process of migrating everything

[radiant grove]

i honestly just sometimes want to be reborn as a big ceo so i can have free slave workers and not pay taxes

[bold grove]

dont we all

sometimes i wanna be a hunter gatherer in the steppes of mongolia and throat sing for my wife

[small hound]

subset of C

[bold grove]

still cool though

they have lots of potential

[radiant grove]

llm's are confusing

its hard to make them understand you

like a person might immediatelly understand your request but an ai is big stupid so it wont

[bold grove]

they dont really speak human language

they speak what they think to be human language

i mean

theres a phenonmena where

if you make two ai's talk to each other

they will evntually start talking in a labguage only they can understand

and it sounds like gibberish to us measly human mortals

which is really fucking scary

[radiant grove]

gibberlink

i swear that video was a shitpost

there is no way that wasnt programmed intentionally

[bold grove]

no ive literally recreated it

[small hound]

what

[small hound]

it's intentional

it's a half satire thing

this is fully intentional

and by design

[bold grove]

yeah its not as intense as the meme makes it seem but its still crazy

they end up just speaking weirdly

[small hound]

https://en.wikipedia.org/wiki/Gibberlink

[small hound]

proof?

https://github.com/PennyroyalTea/gibberlink powered by typescript

[radiant grove]

guessed it

[small hound]

this would be ridiculously infeasible for LLMs to just go and do on their own

[bold grove]

Nah it was a drunk experiment me and my girl did and it was really off

[small hound]

what 💔

[bold grove]

Like we made our chat gpts talk to each other and the language kept getting harder to understand

And eventually it stopped

[small hound]

huh

me when I turn the chatgpt temperature slider to the maximum

[radiant grove]

it probably actually relies on a human to be understandable

[hollow igloo]

just barely... and I could have done it by hand many times faster.. but IMO it's important to understand the tools and how to use them

[hollow igloo]

youtube.com/watch?v=jM8dCGIm6yc

[bold grove]

This song makes me feel like taking over half the world

[bold grove]

ah yes

my favorite source file

cpu/cpushit.asm

[bold grove]

I feel like you’d enjoy this

[radiant grove]

enjoy what?

[radiant grove]

it makes the cpu shit

[bold grove]

cpushit.asm

[radiant grove]

truly

this is a cpushit of all time

what does this even do

floating point stuff?

whats xchg

[bold grove]

This part of it yes

[bold grove]

assembly version of atomic_exchange

[radiant grove]

oh

[bold grove]

It’s smp stuff

[radiant grove]

mk

[bold grove]

You know what atomics are?

[radiant grove]

unstoppable operation iirc

[small hound]

someone forgot the lock prefix 🤨

wait you dont need it nvm

i forgor

[bold grove]

_______________atomic_set_and_test

[prisma magnet]

lost me at leave

[small hound]

If a memory operand is referenced, the processor’s locking protocol is automatically implemented for the duration of the exchange operation, regardless of the presence or absence of the LOCK prefix or of the value of the IOPL.

💥

[bold grove]

This is just a file full of cpu shit

That’s why I called it that

Basically

Has everything

[prisma magnet]

reminds me of when i named a variable

wait can i say this

probably yes or no

"howcanibehomophobicmybitchisgay'

[bold grove]

LMAO

THATS HILARIOUS

and no, you won’t get banned for that lol

If anything that ISNT homophobic

[prisma magnet]

suure

i wrote that variable in my terrible lexer

[bold grove]

In my new vmm I have a variable called “page_to_shit_out”

[radiant grove]

at least capitalize it to be inclusive

[prisma magnet]

in that same lexer i also have vars named fuck and useless

[radiant grove]

all my debug variables are named france

[prisma magnet]

?? why 😭

[radiant grove]

oui

[prisma magnet]

page_to_shit_out

why just why

i also wrote this masterpiece in my acpi thing

[prisma magnet]

also why 8 bytes of 0

[radiant grove]

oh so uh it was offset to the left for some reason

limine couldnt scan the framebuffer tag (i use mb2)

so i just added a few blank items to shift it back

[prisma magnet]

and why are you doing this in asm

[radiant grove]

multiboot2

why would i use limine's boot protocol

[prisma magnet]

limine with mb2

legal

but actually no youre under arrest

[radiant grove]

WRONG

im not arrested because i said so

[unborn latch]

do you align 8 before each tag

[unborn latch]

@flint cobalt flex your variable names

this is your time

https://tenor.com/view/skeleton-gif-26826812

[radiant grove]

tried it

didnt work

[flint cobalt]

no

[radiant grove]

if it works it works

if anyone ever is gonna change it im bringing the wrath of zeus onto your house

[unborn latch]

https://tenor.com/view/caught-an-exception-error-coding-gif-5259200162938284431

[bold grove]

void fuckshit();

[unborn latch]

int sex

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

[bold grove]

This is how I pregamed floppa os months ago:

Me in green

[prisma magnet]

Everyone looks mildly concerned

Or high

Or mental

Or all of the above

[bold grove]

Yes we were all high

[prisma magnet]

All of the above

Got it

[bold grove]

Yeah

[prisma magnet]

Do not.

Ok

Do not.

[small hound]

🤑

[radiant grove]

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

[bold grove]

Yo do you have the speech bubble meme png

[small hound]

nop

[bold grove]

Do you have the bubble image

[radiant grove]

linus

[bold grove]

Okay guys

I’m gonna actually do the big ahh commit

Tn

[bold grove]

@small hound I’m doing tlb shootdown

[small hound]

💥

[bold grove]

So basically the concept is to invalidate the page mapping for the current cpu then send an ipi to tell all the other CPUs to invalidate it too?

Am I correct or

[small hound]

that's the naive approach yes, it can get you a simple, slow, but functional TLB shootdown

[bold grove]

Is it good enough

Like

Is that standard

[small hound]

it can be good enough if you have 4 cores but as soon as you start scaling beyond that things get hairy

[bold grove]

I just have to register an interrupt handler for the ipi right?

[small hound]

it's not standard no, real kernels use much more advanced things

[bold grove]

I figured

But in terms of functionality

Is my concept decent

[small hound]

indeed

[small hound]

it will be "simple, functional, and slow", so if that's fine with you then yeah it would be decent

you can always make it more complete later on

[bold grove]

Well I’m not going for complexity or anything like that

@small hound do you register your ipi handler before or after initializing smp

[small hound]

I do it when I IDT init... ok

you should already have some dynamic ISR registration method

[bold grove]

Okay so ap’s don’t need to be started before that, got it

I do

[small hound]

👍

[small hound]

are you doing per core IDTs?

[bold grove]

I haven’t gotten that far but that’s the plan

[small hound]

sounds unnecessary

linux doesn't do that

[bold grove]

Okay bet

[small hound]

just one global IDT

[bold grove]

I won’t do that then

[small hound]

the main thing is you get very little benefit out of that

just added complexity

[bold grove]

idt_set_entry()™

[bold grove]

Yeah I figured

[small hound]

or isr_register for something more arch agnostic lol

[bold grove]

Yeah but my shit is ia32 or leave the room

[small hound]

💥

[bold grove]

Jk jk

Not a bad idea

[bold grove]

I have 8, is that fine?

[small hound]

I guess you'll find out after you implement and test it, it varies from kernel to kernel

[bold grove]

Yeah true

I’m also gonna do proper mutexes after this

Is making a purely cooperative scheduler insane

[small hound]

not unheard of

[bold grove]

I kinda want it to be yield() only lmao

Fuck ticks

[small hound]

I mean we obviously have stuff like language runtimes that do that but in kernels it's not unheard of

[small hound]

how come

[bold grove]

Mmmm idk because funny

[small hound]

whimsical even

[bold grove]

I just saw some bizarre kernel that had one and it looked interesting

Forgot the name but

It was from like 2007

And it supported xorg

Which is insane

[small hound]

macos?

https://apple.stackexchange.com/questions/435943/did-the-macintosh-os-in-1997-really-not-have-multitasking

it was cooperative only

[bold grove]

no but that is a notable example of one

[small hound]

I don't really think you gain anything from cooperative only scheduling, besides you can always make standard scheduling that is preemptive and cooperative and then just turn off ticks with some syscall or whatever

so you can just build a standard scheduler and then have a way to switch to cooperative only mode or something

not super hard to do

I mean it is quite literally just a

void lapic_timer_disable() {
    uint32_t lvt = lapic_read(LAPIC_REG_LVT_TIMER);
    lvt |= LAPIC_LVT_MASK;
    lapic_write(LAPIC_REG_LVT_TIMER, lvt);
}```

with some fancy stuff and other wrappers around it

[bold grove]

So like for video games and stuff I would probably want to do that

To get the highest frame rate

[small hound]

or you can make realtime threads that disable the scheduler tick :)

this way when they finish running (block or the thread terminates) the scheduler tick comes back to run the timesharing threads

and then it goes away again when the realtime thread runs once again

[unborn latch]

what do you mean

like you want the bubble

speechmeme.com

[bold grove]

Thank you

[bold grove]

POV floppaos

[torpid viper]

cpushit.asm

[torpid viper]

whyyyyy?????

[torpid viper]

done plus d'info

[bold grove]

are you france .. i mean french

[torpid viper]

yes

[bold grove]

Ahhhh

I’m Bosnia… Bosnian

[bold grove]

Which region

I went to île de France

And champagne

[torpid viper]

"ile de france"

[bold grove]

So Paris

[torpid viper]

yeah really close to paris

[torpid viper]

what contient is this on ?

[radiant grove]

europe

[bold grove]

8 CPUs seems to work fine

I’ll post some tests later

[paper cloak]

Imagine parler français

[paper cloak]

Scalability issues will rise when you get hundreds of tlb shootdowns

[bold grove]

Yeah I figured but I’ll only support 8

[small hound]

#define MAX_CPUS 8

[paper cloak]

I said hundreds of tlb shootdowns

Not hundreds of cores

[bold grove]

Ohhh

[paper cloak]

Mine is 1024 I think

[small hound]

lame

[paper cloak]

Nvm it's 4096

[bold grove]

How do you do a more efficient shootdown

[small hound]

there we go

[small hound]

queues of shootdown pages

[bold grove]

Ahhhh

For each core?

[paper cloak]

rcu is the limitation here

[bold grove]

I’m assuming

[small hound]

yea and also don't send the page to everyone if you can guarantee some don't have it in their tlb

like if you have a process and it only ever has run on core 0 there is no need to do any shootdowns at all for its pages

[paper cloak]

Ideally you want to reduce tlb shootdowns

As they are inherently very expensive

[bold grove]

Yeah invlpg is a slow instruction

[small hound]

and so is the IPI

[bold grove]

I’m pretty sure it’s probably the slowest one period

[small hound]

false

wbinvd

[bold grove]

At least that’s the worst one I know

Jesus

[small hound]

you havent seen many x86 instructions i guess

there's also the more interesting ones like rdseed

[bold grove]

No im more of a C guy lol

[small hound]

what does that have to do with not knowing cpu instructions tho

💥

[paper cloak]

mov cr3

[bold grove]

I don’t know every mnemonic and what it does is what I’m saying

[small hound]

i would say wbinvd and rdseed are the more famous ones tho

understandable

[paper cloak]

is rdseed that slow?

[hollow igloo]

1200 cycles ish

with 64bit 2500

[paper cloak]

also it depends on how you count instruction speed

Like wbinvd has very slow side effects

But is the execution of it actually slow?

[hollow igloo]

movnti sucks

cpuid is like 100

fbstp 288

xsave 339

xsavec 623

so xsavec is the slowest direct instruction on zen4

no... I fucked up.. .thats is the number of microops

[small hound]

invlpg seems to take ~200ish clock cycles

idk I'm relying on kvm with -cpu host, might be very wrong

[hollow igloo]

bunch of moves which are around 1000

[bold grove]

Big things are coming my friends. @crimson plover knows

[rigid root]

U cant leave us hanging

[bold grove]

I’m making a new version called CaracalOS that is long mode that uses @crimson plover ‘s bootloader :)

[rigid root]

Thats cool

[bold grove]

Yes and since my shit was 32 bit x86 a lot of the code can be copied from FloppaOS

So I guess this will be a descendant of FloppaOS just in long mode

For example my pmm can literally stay the exact same

Same with my synchronization primitives, etc

[rigid root]

Makes sense

[bold grove]

I’m in Hawaii tho so the progress won’t be as quick as it has been lol

I’m getting sushi in Hawaii rn yall wanna see?

[crimson plover]

for sure

[bold grove]

I’m a big back so I forgor

But my dad took a nice photo of me

Also

I have a sneak preview (cool amarat)

Amarat with glasses

Faceass

[rigid root]

damn leave some for me

[radiant grove]

:0

[bold grove]

I ate all that spicy tuna already 😭😭

[rigid root]

[bold grove]

I got kava… it’s like a traditional Hawaiian/pacific island drink that gets you buzzed

[small hound]

https://github.com/karlstav/cava lore

[bold grove]

Dude this shit got me buzzed as hell

[small hound]

you are drunked

[bold grove]

Mom takes nice photos

I’ll post some progress tonight

Non alcoholic cocktail btw

[blissful grove]

hell yeah, enjoy hawaii 🍸

[bold grove]

hey gang

i got the bootloader working

tomrrow will be source tree structuring!

good night chat

[radiant grove]

https://tenor.com/view/im-a-big-fan-lewi-lewis-fan-spinning-nikocado-avacado-gif-21664606

[bold grove]

kiss me

[radiant grove]

questionable statement at best

[bold grove]

Buzzkill /j

[radiant grove]

https://tenor.com/view/social-credit-credit-social-уважение-плюс-уважение-gif-1626328442317885176

[bold grove]

https://tenor.com/view/winnie-the-pooh-hungry-dinner-gif-14590837

[bold grove]

Xi jinping on his way to eat you after making that joke

[radiant grove]

evil

[bold grove]

Hi guys

Gonna work on kernel

[radiant grove]

kermbrl

[bold grove]

I got sushi again

And I remembered to take a photo

[radiant grove]

nice

[bold grove]

FloppaOS thread became food picture thread real

[radiant grove]

FloppaBlog

[bold grove]

FloppaOSIsFarting

So I’ve gotten some stuff done

With the new fart

[radiant grove]

what stuff

[radiant grove]

The NewFart Microkernel Technology is a revolutionary new system, allowing deep integration of Apple Intelligence in its' core, now powering over 6 million MacBooks worldwide

[bold grove]

Well I’ve got long mode, got the new bootloader working, ported my pmm and started on my new vmm

[bold grove]

Powered by FloppaOS

Imagine the day where every computer runs FloppaOS

I think they would all blow up

[radiant grove]

im addicted to microkernels now they're somewhat peak

[bold grove]

Lol we all had a phase of that

I was convinced that I should change mine to a microkernel

[radiant grove]

be me
works at nuclear reactor
horse installs REACTOS?!?!?!?!?!?!?!?!?!?!!?
WHAT THE FUCK
reactor explodes

[radiant grove]

m

i kinda already changed it

[bold grove]

Do you have userspace

[radiant grove]

idk i might just have a shit ton of hard disk and storage drivers to boot the everything

[radiant grove]

no

[bold grove]

Okay then your kernel isn’t a microkernel, by default it’s monolithic

[radiant grove]

ok fair

[bold grove]

You cannot have a microkernel without running drivers in userspace

At that point it’s a buzzword

[radiant grove]

oh ok

my plan is to run drivers in userspace

[bold grove]

Maybe you should read up again lol

[bold grove]

Yeah nothing wrong with that

[radiant grove]

no no i know what a microkernel is

minor brainfart there

[bold grove]

It’s okay lol

[radiant grove]

im planning it to be a microkernel like run 99% of the drivers in userspace

hybrid ke-

[bold grove]

I forgot what initrd meant for like 5 minutes earlier because I didn’t have my morning coffee lol

[bold grove]

Don’t you fucking dare in my thread

Keep that cancer away from me

[bold grove]

I feel like 99% is all of them unless you literally have 100 drivers and only run one outside of userspace

[radiant grove]

i meant General Assembly resolution 39/46 of the Geneva Convention
-# (Convention against Torture and Other Cruel, Inhuman or Degrading Treatment or Punishment)

oops

[bold grove]

Hahahahahaha

Balkan Slavs don’t follow the Geneva conventions

More like Geneva suggestion

[radiant grove]

uhhhhhhhh

maybe idk

[bold grove]

I don’t even think most Linux kernel images have 100 drivers lol

But there’s definitely more than 100 in the whole kernel

[radiant grove]

i just realized i dont think i can write a microkernel that actually works on things

nvm

because i need like a shit load of hard drive access points

to get like 99% of the drivers

a microkernel runs all its' drivers in userspace but idk if i should run the fucking ps/2 driver in userspace or some other thing

ykw im gonna move it to user space when i get there

Theas Fucking Keybordbs

[bold grove]

Bro

I have so much to tell yall

[bold grove]

Who’s kernel is this 💀💀

[radiant grove]

mine

Joe

[bold grove]

I found this funny code comment in my os

// TODO pt2: fix clunky bits of the allocator such as alignment, guarding, and setting the heap vmm region 
// so uhhh basically everything but it works so I don't care for now. 

[radiant grove]

all comments on my pmm are crash outs

[bold grove]

I love yall just saying

Thank you for caring about my dumb os

Floppa fart

Good night

Okay maybe not good night

Maybe a little kernel dev first

I decided on a name

FloppaOSx64

[crimson plover]

drop the x

[torpid viper]

yes

[bold grove]

Bet

[radiant grove]

drop the floppa

OS64

[bold grove]

Just drop everything

Name it “ “

[unborn latch]

no

Name it ““

[bold grove]

Good point

[bold grove]

I sent this to @crimson plover

But imma explain my concept for the vfs

I’m gonna represent physical memory regions in the vfs
I also will represent virtual memory regions in the vfs
I'm really aiming to make an abstraction of the hardware in the vfs
I think it'll make my life a lot easier
Like each slab will be represented with a vts object
I'm gonna try to make a new standard
Like not Unix at all
Butttt | plan to make a posix compatibility layer
I think I'll make something similar to wsll... where Linux syscalls are represented using the equivalent windows syscall
I kinda like NT in some ways, especially its VFS and I like the concept of the registry I am gonna do something similar But it won't be nt-like at all
Just some inspiration
I'm thinking about calling it the Keywall
"Keys" represent registries
And are hierarchal
I have this written down somewhere
Keys will also be... you guessed it... represented in the VFS

There’s my vfs yap :)

[paper cloak]

WSL is not a translation layer

It's a whole-ass virtualized Linux

WSL1 was a translation layer, which was cooler from a technical point of view but it sucked

Also NT doesn't have a VFS

[bold grove]

Well not in the way Unix does

But the registry, security control, task scheduler, all that shit is part of the vfs

[bold grove]

I meant to say wsl1

Which converted Linux syscalls to windows syscalls

[paper cloak]

No what

These are all their own subsystems lol

Pretty much

The scheduler/dispatcher is in the kernel (Ke)

[bold grove]

Yeah but the registry is inherently a virtual file system no?

At least that’s what I see it is

It’s hierarchical

Or I guess a form of a vfs

[paper cloak]

No?

Where are the files

[bold grove]

Each registry entry is a vfs object no? Idk maybe I throught of it wrong

[paper cloak]

It's just a hierarchical structure with named objects, like a VFS is

[paper cloak]

NT does not have a VFS

[bold grove]

I see. I thought that was just kinda referred to as a vfs

[viscid zephyr]

it sorta does

[bold grove]

In a context like this

[viscid zephyr]

not in the unixlike way though

[paper cloak]

He wasn't referring to the object namespace

[torpid viper]

it does

[paper cloak]

the object namespace anyway is not a very elegant design

[paper cloak]

the object namespace has similar functionality to a VFS but it is pretty far from a traditional unix-like VFS

[torpid viper]

i'm not talking about this

[paper cloak]

The thing where you have to re-parse the entire path when you enter a different filesystem (or something like that, I haven't read about it in a while)

[torpid viper]

the thing that allow to mount stuff as a:/ b:/ and other is a vfs

not a very complex one

but still it's a vfs

• the vfs handle some special devices

[paper cloak]

Isn't A:/ just translated to like \Device\Disk whatever

[torpid viper]

like COM0 and stuff

[paper cloak]

In the object namespace

[torpid viper]

then in that case it's a vfs

[paper cloak]

No, how?

[paper cloak]

This

Anyway, there is better stuff to steal from NT imo

[bold grove]

I like push locks a lot

Astral has one

[paper cloak]

turnstiles tho

[viscid zephyr]

TURNSTILE GANG DESTROYS PUSHLOCK GANG

[paper cloak]

Also astral's is broken afaik?

[viscid zephyr]

he fixed it

[paper cloak]

Ah great

[bold grove]

No it works I tried it earlier

But yeah it was borked for a bit

[bold grove]

Truth nuke

Probably easier to implement anyways

[paper cloak]

A nice thing to steal from NT is generic handles/capabilities instead of just FDs

Also the layering is pretty nice imo

And irql

(those are not all unique NT concepts)

[bold grove]

NT does them well tho yeah

[viscid zephyr]

pushlocks are easier to implement

initially

when you add priority inheritance is when the picture changes

then pushlocks are harder

and turnstiles are easier

[paper cloak]

No this is very easy

Just implement autoboost

[viscid zephyr]

autoboost is hard

[paper cloak]

Skill issue

literally the easiest thing ever

[viscid zephyr]

fym skill issue none of us even know how it works other than the broad strokes

[paper cloak]

You didn't try hard enough

Will if you ever work at Microsoft will you come back to this server with autoboost wisdom

[fervent flare]

It's essentially priority inheritance.

[north hull]

this is common with turnstiles

[north hull]

with autoboost there is a very unclear moment in the case when all autoboost entries exhausted

theres no longer a limit of 6 entries but apparently there's still some limit since EtwTraceAutoBoostEntryExhaustion exists

[bold grove]

All I know about those is that they try to recover from a deadlock using priority inversion

[north hull]

wdym

[small hound]

this not how deadlock recovery works

linux workqueues have this funny deadlock recovery with rescuer threads

Regular work processing on a pool may block trying to create a new worker which uses GFP_KERNEL allocation which has slight chance of developing into deadlock if some works currently on the same queue need to be processed to satisfy the GFP_KERNEL allocation. This is the problem rescuer solves.

When such condition is possible, the pool summons rescuers of all workqueues which have works queued on the pool and let them process those works so that forward progress can be guaranteed.

This should happen rarely.

🤪 pretty silly

[bold grove]

Idk it’s just something I read from a windows article about the topic

[paper cloak]

Deadlocks are another set of problems

They avoid priority inversion

[bold grove]

I see

[small hound]

maybe u misheard something or something misspoke

💥

deadlock recovery is definitely an interesting thing tho and there are places where they can be recovered from (like the linux workqueue GFP_KERNEL allocation deadlock)

[bold grove]

Man honestly I might switch back to vga text mode and I don’t care what anyone says about it

[paper cloak]

A functioning system shouldn't have deadlocks at all

[small hound]

linux isn't functional I guess

[paper cloak]

There's probably a cleaner way to do it

But it's slower or something so they just hack around it

[small hound]

I guess they could have the gfp_kernel allocation spawn a separate thread/wake a separate thread on that path

[viscid zephyr]

Locking related deadlocks yeah but that's not quite what this is

[hollow igloo]

where did you find a deadlock?

[radiant grove]

he lost losing to flanterm :'(

[small hound]

https://docs.kernel.org/core-api/workqueue.html

Regular work processing on a pool may block trying to create a new worker which uses GFP_KERNEL allocation which has slight chance of developing into deadlock if some works currently on the same queue need to be processed to satisfy the GFP_KERNEL allocation. This is the problem rescuer solves.

When such condition is possible, the pool summons rescuers of all workqueues which have works queued on the pool and let them process those works so that forward progress can be guaranteed.

This should happen rarely.

it's recoverable

[hollow igloo]

yea... thats not a deadlock

really starvation

[small hound]

hmmm ok I guess they just described it differently

[hollow igloo]

picture this...

you have tasks A, B, C, and D

A needs to complete before B, C, and D

they get inserted into the queue as
Head -> B, C, D, A
If you have 3 workers then they will pickup B, C, and D
and it will never progress

might not be the exact issue here, but sounds similar

[bold grove]

Drinking Hawaiian monster flavor on my last day in Hawaii :)

[bold grove]

No issues over here in Hawaii 😎😎😎

[rigid root]

bro flexing

[violet anchor]

does it taste of Hawaii

[hollow igloo]

you keep keep beaches and sand... I like mountains and trees

[violet anchor]

Hawaii has mountains that spray molten rock. the best kind of mountains. tends to set fire to trees though

[bold grove]

Don’t worry, I live in the Pacific Northwest, that’s all we’ve got here :)

[bold grove]

Sure did

[bold grove]

It’s my last day here I had to a little

[bold grove]

man i fucked up paging somewhere

[radiant grove]

courier ptsd

[bold grove]

What do you think about the font I made

[radiant grove]

1.) YOOOO nice font
2.) Times New Roman

[bold grove]

I was kinda going for that

[radiant grove]

floppa os if it was medieval:

[bold grove]

Work has continued

[torpid viper]

when GPU

[bold grove]

gpu init - not ok

[foggy cedar]

GPU: Segmentation fault

[bold grove]

Okay I’m making the operating system less shit

So I decided to keep on with this kernel for now

Just because I want it to be good first

And it was not good before

[bold grove]

my logs are fucking atrocious

@hexed notch 💀

I have the worst meme os

[radiant grove]

nah they're fine enough

look at mine

[hexed notch]

I mean, they don't look too bad lol

[bold grove]

Am I tripping or is that just a purple screen

[radiant grove]

yes thats just a purple screen

peak logs fr

i dont need these filthy logs anyways

how's progress doing tho

[bold grove]

Yeah no they’re pretty fucking braindead

[bold grove]

https://tenor.com/view/взгляд-2000-ярдов-война-war-soldier-gif-3632617944134077161

[radiant grove]

it's two thousand yard stare nevermind

[bold grove]

Nah it’s just shit

I completely removed my vmm basically

[radiant grove]

installing limine broke my windows xp install

big pain

[bold grove]

Also I removed my vmm because it was shit

[hexed notch]

fair enough 🤣

[radiant grove]

rip vmm

[bold grove]

Yeah it was horrendous

[radiant grove]

56342nd time's the charm tho

[bold grove]

I’ve made a tlb shootdown

@small hound

[bold grove]

Okay so we have a vmm again

I’ll push it sometime

I was thinking about doing modules

But eh I don’t need them

Gonna work on scheduler

And all that good stuff

I’m doing tlb shootdown by sending an ipi to each cpu

Eventually I’ll make this lockless

By having per cpu allocators

Or something idk

[bold grove]

okay so i worked on it

heres the start

no multicore yet though bc of invlpg

mapping works

My vmm region just has a pointer to the pd and the next region

Dude I’m fucking cooking

Okay so now I have to

Make this multicore

When I get back to my pc

[radiant grove]

damn

[bold grove]

I’ve made functions to copy and nuke page maps

Which was the weirdest part

[radiant grove]

void vmm_fuckshit();

[radiant grove]

how so

[bold grove]

Copying a pagemap requires you to copy all the mappings

Which can be kinda involved

[radiant grove]

oh

right

[bold grove]

Nuking a pagemap is fine

You just need to make sure you don’t leak anything

[bold grove]

void* return_vmm_fart(vmm_region * fuck);

[bold grove]

Aren’t you making a vmm

[radiant grove]

well i have a vmm but it cant clone pagemaps

[bold grove]

How you planning on forking then

[radiant grove]

add a clone feature

[bold grove]

Yoink mine

Actually don’t

Please don’t

I still treat all my pds and pts as 32 bit ints

Don’t even use a pde struct

[radiant grove]

uhhhhh

same :(

[bold grove]

I mean that’s what they are

[radiant grove]

wait pds and pts as in pagetables?

[bold grove]

But it’s not as clean

[radiant grove]

and pagedirs?

[bold grove]

Yeah

[radiant grove]

damn

ok same my code is ass

[bold grove]

I mean that’s what they are

[radiant grove]

pointers are uint32_t's

but like

fancy

[radiant grove]

and also my code works probably way differently

like beyond different

[bold grove]

No. Just as a rule of thumb don’t yoink my code

It’s shite

Well

I want a decently featured vmm

I’ve got

Mapping and un mapping (

Regions

Cloning and destroying them

[radiant grove]

nice

[paper cloak]

CoW?

[bold grove]

Yeah

[paper cloak]

how do you do it

refcount?

[bold grove]

Wait no I don’t do copy on write

I just clone the page tables

[paper cloak]

boring

[bold grove]

Is it much more involved to do copy on write

[paper cloak]

depends

your copy routine will be much simpler

but your page fault handler will do more things

[bold grove]

Yeah mine currently just says “fuck you” and halts

[bold grove]

Yeah my copy routine is actually really complicated

[gentle flicker]

I do, and it sucks, I just clone the regions whenever a page fault is handled and whenever it's requested it is freed independently from the parent address space :)

[paper cloak]

wha

you should use a per-page refcount

and you know that if you have a write fault with ref > 1 and the paged is marked CoW then you need to make a new mapping and copy the data there

[gentle flicker]

yep, I probably worded it wrong

once a region is cloned and flagged as CoW and a write page fault occurs, the physical address is retrieved, a new page is allocated and the data is copied into it then a new mapping that points to the newly allocated page overwrites the old one

but I don't keep any refcount, as I said it sucks, it's held together with duct tape and not optimal at all

[bold grove]

Okay mapping

And shit

Is done kinda

Now I’m working on tlb shootdown

[bold grove]

okay i pushed stuff

not tlb yet

heres pagemap cloning

heres my (simple) vmm region data structure

its interesting that after the rewrite my vmm init became much simpler

@torpid viper vmm rewrite!

[bold grove]

@radiant grove heres how i did the copy pagemap

[bold grove]

yeah my copy routine rn is way too complicated lolll

yayyyyy

now we do nothing

[bold grove]

im actually satisfied rn

for the most part

with how things are going

[torpid viper]

cool !

do you have CoW ?

[bold grove]

no

i just copy mappings

[torpid viper]

still cool

[bold grove]

heres how i copy mappings @torpid viper

[torpid viper]

good

i think a little bit of arch abstraction would have been better

[bold grove]

Yeah I’ve been thinking about that

I don’t have any hardware abstraction layer

It’s just ia32

[torpid viper]

you should try to do it early

trust me this will avoid a lot of shit

[bold grove]

what if i want only one arch

or is it good to still have a hal

regardless

[crimson plover]

nah

if u are just tryina finish ur x86 kernel dont do the hal

just finish it and do a hal for the x86_64 one

[bold grove]

okay thank you

[bold grove]

I added aslr

And this thing called randframe

Which randomizes all of the physical frames within a virtual region

Kinda just extra security

Idk I thought the idea was cool

[gentle flicker]

heap allocated region descriptors?

[gentle flicker]

really cool

[radiant grove]

cool

[radiant grove]

flbopa os

[bold grove]

Is that bad

[gentle flicker]

nah, I’m just used to make the vmm before the heap allocator

[bold grove]

I added more stuff

Into the vmm

More physical page randomization within the vmm

And cloning

With randomization

[bold grove]

Fair

I made my vmm before the heap back in the day

I just rewrote my vmm

And my heap doesn’t rely on my vmm

[bold grove]

Im gonna work on expanding my region descriptors

[bold grove]

Gn chat

[bold grove]

Yeah region descriptors are getting better

[bold grove]

Hey gang

No work today on the os most likely

I’ve been doing a lot of vmm stuff

Kinda a bit burnt out for now

We’re gonna work on smp tomorrow though

And actually integrating it

First I’ll need proper trampoline code which

I think I’ve almost figured it out

[bold grove]

i kinda made my mapping functions better

[gentle flicker]

your codebase honestly looks pretty organized

I could never

[bold grove]

Really….?

[gentle flicker]

yeah, you should look at mine, it's a mess of namespaces and naming styles

[fallow orbit]

do you plan to port the kernel?

[bold grove]

Port the kernel?

[fallow orbit]

oops. To other architectures

[bold grove]

Yeah I will

At some point

[bold grove]

I’m probably just gonna stick to single core

For a while because

I need to relearn scheduling concepts

Mine just doesn’t work how it should

[bold grove]

Working on scheduler

[torpid viper]

good

[bold grove]

I’m making a round robin with preemption

[bold grove]

Okay I got it working

I’m gonna make a video game now

Just to prove the concept

[bold grove]

Well I’m shit at that

Still working on it

[radiant grove]

oh no

floppogames

[bold grove]

I’m gonna make a word processor too

[radiant grove]

flord

[torpid viper]

amazing name

[radiant grove]

[bold grove]

Ok gang

Scheduler has some issues lol

I’m doing preemption with the PIT

Kinda running into an issue where certain threads get starved

Man fucking hell

I have so many ideas for the scheduler but I am not at my computer or at home 😔

[crimson plover]

this is why we need to design the fucken pcb

lol

unfortunately this weekend im not at home

next one I might be but then im going to paris

so basically fuckall time this month

[bold grove]

I’m so down

I wish I was home 😔

[crimson plover]

yessir

timezones are a bitch

[paper cloak]

What pcb

[bold grove]

Me and wux (and a few friends) are designing something cool

I don’t wanna give it away without @crimson plover permission

[radiant grove]

3 days later im going to see "FLOPPABOT DESTROYS THE MOON WITH THE SHEER UNRIVALED POWER OF THE SUN" in the news

[bold grove]

In userspace that is

[crimson plover]

cyberdeck type of thing

we need to make the main board the SOC is going to slot into

[bold grove]

Okay new scheduler is up

So

It’s able to handle multiple cores but for now it’s just preemptive with the PIT timer

[bold grove]

So we have priority handling

And also I have a way of pushing starved threads to the front of the queue when they wanna run

[bold grove]

So basically each thread gets a starting quantum of 10 ticks then if it gets starved it gets a gold quantum of 20 ticks

[bold grove]

I came up with this shit in a dream

I think it’s valid

Threads aren’t getting starved anymore

[paper cloak]

Then other threads get starved?

[viscid zephyr]

I still am

I'm dieting

Thanks for being insensitive

[bold grove]

No, it updates constantly

The gold quantum gets set back to the base after the starved thread runs

[paper cloak]

Idk why you need this

If they all run at the same timeslice then there is no starvation

[bold grove]

No basically I up the priority of low priority threads whenever it starts wanting to run. The new priority is called the effective priority

[viscid zephyr]

There can be if he has priorities

[paper cloak]

yes that is true

I didn't know he had priorities tho

[paper cloak]

What does wanting to run mean

[bold grove]

I added those