EvalynOS

[viral bison]

their id is always 0

[hazy saddle]

yeah ima fix that

[viral bison]

anyway, if you are using single core, i have added a function to get the current thread in those patches

if you wanna apply them

[hazy saddle]

btw @viral bison my int 0x69 syscalls are my non POSIX ones

that im gonna axe eventualy

[viral bison]

k

[hazy saddle]

yoooo

[viral bison]

gg

[hazy saddle]

Now I gotta port a libc

[viral bison]

mlibc

[hazy saddle]

It looks scary

[viral bison]

it's not

[hazy saddle]

So how do I do it

[viral bison]

there is a guide

#curated-resources

[hazy saddle]

Before I push I’m gonna make a Mandelbrot program and a generic k thread for making user programs and a function

And then we can have some fun

My kernel is free from the curses

I meaaannnn

I kind of load an elf

[viral bison]

lol

[hazy saddle]

I did read that mlibc will mainline ports for other kernels

I doubt mine is going in there tho

[viral bison]

idk, maybe

[hazy saddle]

they still synced for some reason @viral bison

[viral bison]

lol idk

[hazy saddle]

and and

the first one waits for the ither to catch up

[viral bison]

that's strange

[hazy saddle]

its as if some varribles are being shared??

but they properly move places

[viral bison]

did you apply my patches?

[hazy saddle]

Let me try those

apply to a fresh tree right

[viral bison]

yes

[hazy saddle]

@viral bison your patchset does the same

[viral bison]

hmm

[hazy saddle]

leme give you my patch set with the debuggable bad apple

[viral bison]

oh i think i know

the app is resetting the PIT cycles

[hazy saddle]

OHHH RIGHT

[viral bison]

you should not do that

[hazy saddle]

[viral bison]

btw, keep editing my patches

[hazy saddle]

[viral bison]

those have some more fixes

[hazy saddle]

ooo like what else?

[viral bison]

the vmm share pages fix

and a ctx switch one

tho i forgot to change this back to this

and remove the *--stack for the pagemap pasisng under it

[hazy saddle]

IT WORKS

[viral bison]

perfect

[hazy saddle]

redoing the bad apple for int match and not needing to touch the PIT

[viral bison]

you could still get the PIT ticks

[hazy saddle]

Oh I meant as in reprogramming it

[viral bison]

and convert the base freq to the amount of ticks needed

[hazy saddle]

It can still read it

I wana say thank you so much for the help @viral bison

[viral bison]

np

[hazy saddle]

@viral bison I have a really stupid idea but I wana add a kernel level compositor

[viral bison]

huuh

don't do that XD

[hazy saddle]

Not a full WM though 😭

But like being able to slice off bits of the frame buffer

And manage virtual ones

[viral bison]

make a graphic thread

[hazy saddle]

I can do that in userspace tho

And il setup my shared memory system

@viral bison wana know a fun fact about the bad apple demo I have

[viral bison]

?

[hazy saddle]

I made it

Before I had a VMM

Or a PMM

[viral bison]

oooof

[hazy saddle]

And before any userspace

And when I was making my own bootloader

FUCK THAT SHIT BTW

ALL HAIL LIMINE

[viral bison]

XD

i might make my own BL too later on

when i have my own graphics lib

so i can make it very good looking

[hazy saddle]

Oh btw wana know somthing really hacky I did

I was trying to get somthing to run on an embedded Linux device

But it only runs things as root

And my app says no no no

[viral bison]

XD

[hazy saddle]

https://tenor.com/view/scripulous-fingore-mario-new-super-mario-bros-gif-27542964

Soooo

I made a shim .so file you preload

And it lies to it and it will tell it it’s UID 1000

And GID 1000

And that lets it run

XD

[viral bison]

XD

[hazy saddle]

That’s the other thing I need to put in my public domain repo

Because fuck this shit it’s ass

I don’t want credit

I don’t need it

Just take it and go

Oh I need to get the insane rant I made when I gave up that project

I made a timeline

From my shit memory but yeah a

@viral bison

If you open a project and seee this

What are you doing

[viral bison]

wtf is that

[hazy saddle]

Me going insane

[viral bison]

yeah, i can see XD

[hazy saddle]

I spent 3 sleepless days working on that project

[viral bison]

maybe you should sleep more often

[hazy saddle]

I should

It’s 5am rn tho

[viral bison]

i'm 6h ahead if you, lol

[hazy saddle]

This is also where I get some of my best work done

A lot of my kernel was made at 2am with the power of two entire two liters of Dr Pepper in my veins

This is how I code

[viral bison]

i'm immune to caffeine

so that doesn't work for me

or well, caffeine makes me even more sleepier

[hazy saddle]

But yeah I try and get shit done

And that’s why my commits are somtimes a little large

I do tons of manual fuzzing on my code

[viral bison]

i make small commits, i have the fear of my SSD killing itself again

[hazy saddle]

This is a real commit I did btw

[viral bison]

WHAT THE FUCK

280k LINES IN 47 FILES???

[hazy saddle]

I code a lot

(This is when I added bad apple and refactored a shit ton)

(When I did this commit I also did a limine conversion)

(And an entire project refactor from a single file)

(It was massive)

[viral bison]

oof

[hazy saddle]

Wait shit @viral bison

I don’t have a USB to validate this on real hardware

Mine dispersed

[viral bison]

lol

i have one

where is the iso?

[hazy saddle]

Let me merge my bad apple into your patch set and then I will generate a new patch

[hazy saddle]

My OS dosnt use ISOs

You take a USB and make a fat32 gpt partition

And mark it as esp and boot

And put the proper files on it plainly

And it boots

[viral bison]

k

[hazy saddle]

It’s honestly pretty neat?

[viral bison]

i prefer an iso

so i can use it with my ventoy usb

tho i still have like 4 or 5 more usbs

[hazy saddle]

i dunno how to package it right ngl

i hope my changes diddnt fuck up the timing too bad on the bad apple

[viral bison]

which files do i have to copy?

these?

@hazy saddle

ok, i booted in

i'm in the terminal

[hazy saddle]

Yup

[hazy saddle]

Now let’s test it

Can bad apple properly run twice

Or 3 times

[hazy saddle]

Let’s see how far it can go before exploding

[viral bison]

time to test

no sound

[hazy saddle]

It works through PC speaker

Your laptop may not have one?

[viral bison]

i have integrated speakers

btw, the 2 bad apples are overwriting themselves

lol

[hazy saddle]

Huh

They should not be

[hazy saddle]

WHT

all i did was exit qemu to discord

[viral bison]

they are not overwriting in qemu

but it is failing a bit after

while on the laptop it's still going

[hazy saddle]

in my qemu it keeps going

[viral bison]

didn't it fail there?

[hazy saddle]

thast was with user input

i get this happening if i run credits

[viral bison]

oh wait

[hazy saddle]

and then spam enter

[viral bison]

my laptop is using the one with my patches only

[hazy saddle]

your patches only should overlap

my patches make it not overlap

[viral bison]

i only ran make all

and copied the elf in the main dir

not the bin dir

it pagefaulted on rodata

wtf

[hazy saddle]

https://tenor.com/view/rick-morty-rick-and-morty-20minutes-adventure-quick-adventure-gif-22364490

[viral bison]

XD

[hazy saddle]

So what do we gotta do

[viral bison]

idk

[hazy saddle]

FUCKKKKK

[viral bison]

they are still overwriting on my laptop, wtf

[hazy saddle]

Even with my patches?

[viral bison]

ye

[hazy saddle]

Check the code did it apply right

[viral bison]

yes, if i run it via qemu it works

[hazy saddle]

WTF

[viral bison]

oh wait what

[hazy saddle]

[hazy saddle]

Huh

[viral bison]

it gets a thread id -1

or wait no

thread id 0 is the shell

[hazy saddle]

Yes

@viral bison

I am clueless on the issue

[viral bison]

idk

[hazy saddle]

Does your patch set still change the TSS?

Rather the RSP0

https://tenor.com/view/there's-nothing-we-can-do-gif-14390726165144583214

Welp @viral bison back in stasis the project goes?

Or maybe @devout path can save us

[neon drum]

Have you fixed the BAD APPLE fault?

[hazy saddle]

It can run twice

But it has some random issues

[neon drum]

Twice 😭

[hazy saddle]

This one

[hazy saddle]

Well any number of times*

But it randomly crashes from certain inputs

[hazy saddle]

And this crash

And there are issues between real hardware and VM

[neon drum]

Hmm

Aight bye

My brain is fucked

[hazy saddle]

This is the current patch set rn

[hazy saddle]

Damn

Will you be back later perhaps

[neon drum]

Idk

[hazy saddle]

Mood

[viral bison]

Yes

[hazy saddle]

What happens if you DONT do that

As a wild idea

[rancid viper]

what is the status anyway

[viral bison]

We fixed the original bug

Now there is another

[hazy saddle]

But we got new ones

[viral bison]

It tries to execute code from ro data

[hazy saddle]

99 little bugs in the code, 99 little bugs. Take one down, patch it around 117 little bugs in the code.

[rancid viper]

well if you want to update a branch I could take a look

[hazy saddle]

Grab the main branch and apply these patches

This is how I do development as I don’t like “bad” code on the repo

[rancid viper]

bugfix is bad?

[hazy saddle]

Well there are other bugs

[rancid viper]

sure but these are there currently too, wouldnt fixing one be better than none

[hazy saddle]

Do you have limine installed

[rancid viper]

either way its ur repo

[rancid viper]

it will never be in that path

ill fix it somehow

[hazy saddle]

It is on my system

[rancid viper]

I use nix

[hazy saddle]

I should edit that script to support more than one distro

[rancid viper]

you could pull the binaries from github

[hazy saddle]

Even for EDK2 OVMF?

[rancid viper]

or codeberg now I suppose

[rancid viper]

no..

not ur job to provide ovmf imo

[hazy saddle]

But it might not always been in the past specified by my script

[rancid viper]

where do u even use the binary

I cant find it

found it

[viral bison]

it works if you run make all

at least, it got all the files it needed

and worked

[rancid viper]

you probably have limine installed

[viral bison]

oh, maybe that

[hazy saddle]

Limine and OVMF are used by QEMU so it it loads like a UEFI system?

The path limine is put in gets loaded automatically by OVMF

[rancid viper]

no I mean

I fixed it

for me

[hazy saddle]

Huh neat

[rancid viper]

so how do I reproduce this issue

[viral bison]

run badapple 2 times

and let them play

[rancid viper]

I ran it three times

they have now played for a couple minutes

[viral bison]

show

[rancid viper]

?

[viral bison]

wtf

which patches are you using?

[rancid viper]

the ones evalyn told me to use

[viral bison]

patches2.patch?

[rancid viper]

yes

[viral bison]

strange, it's going fine for me too rn

nope

it nuked itself

the framebuffer had enough

[rancid viper]

works fine for me

[rancid viper]

are u guys running under kvm

[viral bison]

this is after a few mins

[viral bison]

ye

if you do make run it starts under kvm

[rancid viper]

well

idk I cant reproduce the bug

[viral bison]

it seems to fail randomly every time

it kept going for a bit

again

[rancid viper]

well

[viral bison]

and then the framebuffer got nuked

and now i got a GPF

wth

on the pit interrupt

[hazy saddle]

Type on the keyboard too

That’s also part of it

[viral bison]

i didn't have to do it

[hazy saddle]

Run the CREDITS command and spam enter also caused it for me

[viral bison]

why that attribute?

[rancid viper]

uh

dont do that

[viral bison]

something is nuking the stack

and i think the pit has something to do with it

[hazy saddle]

[hazy saddle]

Dosnt unused just tell it that I don’t use it

And it should still deal with the stack right

[viral bison]

i meant the interrupt one

[hazy saddle]

Maybe it’s about time I do interupts in assembly isn’t it

[viral bison]

yes

[hazy saddle]

Good thing I have a good reference I can steal from since I’m working on a wiki with some friends as a side project and it has some about that

[hazy saddle]

What’s the generated code from clang here

I wana see what it’s doing to the stack

I’m in bed right now crying eating whip cream btw

So might take a bit to get up

[viral bison]

go to sleep

it's better for now

[hazy saddle]

Whipped cream lmao

[hazy saddle]

Yup I should lmao

I’m work on this in the afternoon when I wakeup

And then hopefully we should be okay

@viral bison alright so plan
When I wake up I’m going to get assembly based interupts working properly and then if this fixes the bug things should be able to be properly cleaned up and pushed

I’m gonna order some USBs

Wait second wind is hitting

[viral bison]

k

[hazy saddle]

i got an extra 30 minutes of energy

[viral bison]

lol

[hazy saddle]

@viral bison would this be correct?

extern dispatch_interupt

%macro ISR 1
global do_isr%1
do_isr%1:
    cli
    push 0
    push %1
    jmp dispatch_interupt_asm
%endmacro

%macro ISR_ERR 1
global do_isr%1
do_isr%1:
    cli
    push %1
    jmp dispatch_interupt_asm
%endmacro

dispatch_interupt_asm:
    push rax
    push rbx
    push rcx
    push rdx
    push rsi
    push rdi
    push rbp
    push r8
    push r9
    push r10
    push r11
    push r12
    push r13
    push r14
    push r15

    mov rdi, rsp
    call dispatch_interupt

    pop r15
    pop r14
    pop r13
    pop r12
    pop r11
    pop r10
    pop r9
    pop r8
    pop rbp
    pop rdi
    pop rsi
    pop rdx
    pop rcx
    pop rbx
    pop rax

    add rsp, 16
    iretq

; Exceptions
ISR_ERR 0x08     ; DF
ISR_ERR 0x0D     ; GPF
ISR_ERR 0x0E     ; PF

; Hardware
ISR     0x20     ; PIT
ISR     0x21     ; PS/2
ISR     0x24     ; Serial #1 & #3

; Spurious Interupts
ISR     0x27     ; Spurious PIC1
ISR     0x2F     ; Spurious PIC2

; System Call
ISR 0x69         ; Non POSIX System Calls

[viral bison]

it's incomplete

[hazy saddle]

what am i missing?

[viral bison]

basically all of the interrupts

[hazy saddle]

i am aware

im matching pairty with my C code

WHICH

[viral bison]

yeah well

[hazy saddle]

is fine because

[viral bison]

that won't work

[hazy saddle]

\

it should be fine right?

[viral bison]

pagefault won't trigger

[hazy saddle]

wym

[viral bison]

bc it's an exception

you can't change its interrupt

[hazy saddle]

i know?

[viral bison]

then why you set it to 0x69

[hazy saddle]

thats not 0x69??

ISR_ERR 0x0E ; PF

thats page faults

[viral bison]

oh wait

i did not see it

[hazy saddle]

[viral bison]

btw, syscall is not called like that in x64

[hazy saddle]

ehhh whom cares

[viral bison]

interrupts were the very old method

now there is the syscall instruction

[hazy saddle]

ik its the old method but i dunno how to make that work right

[viral bison]

you need to set some MSRs

to use the syscall instruction

[hazy saddle]

i know but the docs were confusing for them

[viral bison]

we talked about them a while ago in #osdev-misc-0

bc i did not uderstand those docs too

[hazy saddle]

discord isnt wanting to bring up the message right 😭

[viral bison]

search for my messages

[hazy saddle]

leme finish this up first tho

oh also bonus part of this @viral bison is interupt driven panics will give me wayyy more accurate reg dumps

[viral bison]

ye

[hazy saddle]

gonna have to rework a good bit of shi rn but its gonna be good

[hazy saddle]

wow and my adhd meds were taken on time

i think i got this

il keep this one in C rn because its a taken function and should be OKAY @viral bison

[viral bison]

i think asm is better

tho why the leave instruction?

[hazy saddle]

il convert later™

[hazy saddle]

idfk it works

this shit is so unstable i dont wana touch it with a 10ft pole

[viral bison]

XD

this is my syscall handler

[hazy saddle]

that looks super close to my ISR hander code

[viral bison]

bc it basically is

[hazy saddle]

yeah lmfao

[viral bison]

i wanna get all the regs

[hazy saddle]

[viral bison]

so i can support syscalls from multiple oses

[hazy saddle]

ooo smart

my friend had the reverse idea

if an app attempted a sys call from another os

it would know the process was compimised

so it would lock it down and lie to the process to make it suffer

[viral bison]

XD

[hazy saddle]

eg malloc would return invalid kernel pointers or other shit

[viral bison]

my future plan is to support windows app as good as possible

[hazy saddle]

ala reactOS?

[viral bison]

similar, yeah

[hazy saddle]

you gonna have an entire UI framwork in the kernel

[viral bison]

i wanna try to make my os run apps from windows, macos, linux

and my own os's apps

[hazy saddle]

what about my os

once i get better systemcalls

[viral bison]

can be added

[hazy saddle]

[viral bison]

i think i'll use modules for that

[hazy saddle]

would be neat if you could have a hint system call maybe?

so if apps are aware of it they can tell you what layer to use

and i could compile my crt0 to emit that

[viral bison]

well, my os will try to look at the header to know which os the app was made for

[hazy saddle]

ahhh

but i have no headers

/s

[viral bison]

you'll need to make elfs

(they are not that complicated to load)

[hazy saddle]

[clang] (-Wincompatible-pointer-types) Incompatible pointer types passing 'struct interrupt_frame *' to parameter of type 'struct interrupt_frame *' huh

[viral bison]

wtf

[hazy saddle]

src/kernel/interupts/gp_fault.h:4:33: warning: ‘struct interrupt_frame’ declared inside parameter list will not be visible outside of this definition or declaration
    4 | extern void gp_fault_isr(struct interrupt_frame* frame);
      |                                 ^~~~~~~~~~~~~~~
src/kernel/interupts/interupts.c: In function ‘dispatch_interupt’:
src/kernel/interupts/interupts.c:10:26: error: passing argument 1 of ‘gp_fault_isr’ from incompatible pointer type [-Wincompatible-pointer-types]
   10 |             gp_fault_isr(frame);
      |                          ^~~~~
      |                          |
      |                          struct interrupt_frame *
src/kernel/interupts/gp_fault.h:4:50: note: expected ‘struct interrupt_frame *’ but argument is of type ‘struct interrupt_frame *’
    4 | extern void gp_fault_isr(struct interrupt_frame* frame);
      |                          ~~~~~~~~~~~~~~~~~~~~~~~~^~~~~

 ✘ evalyn@EvalynPC  ~/Documents/Programming/C/evalynOS   master ±  grep -Rnw . -e "struct interrupt_frame"
./src/kernel/interupts/gp_fault.h:4:extern void gp_fault_isr(struct interrupt_frame* frame);
./src/kernel/interupts/page_fault.h:5:extern void page_fault_isr(struct interrupt_frame*);
./src/kernel/interupts/interupts.c:4:void dispatch_interupt (struct interrupt_frame *frame) {
./src/kernel/interupts/double_fault.c:4:void double_fault_isr(struct interrupt_frame* frame) {
./src/kernel/interupts/interupts.h:30:struct interrupt_frame {
./src/kernel/interupts/interupts.h:37:extern void dispatch_interupt (struct interrupt_frame *frame);
./src/kernel/interupts/generic.c:4:void generic_isr(struct interrupt_frame* frame) {
./src/kernel/interupts/page_fault.c:4:void page_fault_isr(struct interrupt_frame* frame) {
./src/kernel/interupts/generic.h:3:extern void generic_isr(struct interrupt_frame* frame);
./src/kernel/interupts/double_fault.h:3:extern void double_fault_isr(struct interrupt_frame* frame);
./src/kernel/interupts/gp_fault.c:4:void gp_fault_isr(struct interrupt_frame* frame) {
./src/kernel/utils/panic.c:139:void panic(char* message, struct interrupt_frame* frame) {
./src/kernel/utils/panic.h:8:void panic(char* message, struct interrupt_frame* frame);

[viral bison]

maybe a missing include?

[hazy saddle]

no?

WTF

[severe marlin]

Just gonna step in bc I noticed, but gp_fault.h also needs to include whatever header the struct is defined in

[hazy saddle]

wait why isnt it

huh

[severe marlin]

That's where the first warning is coming from at least

[hazy saddle]

ty

how did that end up like that

[hazy saddle]

i do wonder why x86 dosnt have an instruction to push all GPRs to the stack?

its CISC right

[viral bison]

idk

[hazy saddle]

i ❤️ switch/case statments

[viral bison]

oooooof

[hazy saddle]

its fine lmao

[viral bison]

this is an easier way

[hazy saddle]

btw is my kernel code organized well

i dont have arch spesifc ISR folders because they are all x86 spesific

[viral bison]

it should be good

[hazy saddle]

well shit

This is with assembly based interupts btw

[viral bison]

well, time to find the issue

[hazy saddle]

Want the patch file?

[viral bison]

not rn

trying to finish implementing syscalls

and running the first userspace app

[hazy saddle]

ooo nice

oooooo

Hey atleast working on this helps with your stuff too

[viral bison]

XD

[hazy saddle]

@viral bison what should i throw breakpoints on in GDB btw

ima try and debug

[viral bison]

idk really

try to breakpoint the badapple code

near to where it errors

[hazy saddle]

im gonna breakpoint the exception handlers and then i can see the stack trace maybe

it crashes if i type too much on the keyboard

[viral bison]

don't you already print it?

[viral bison]

it crashes by also doing nothing

just play bad apple 1 time

[hazy saddle]

its some kind of interupt

[viral bison]

and after a bit the framebuffer is nuked

[hazy saddle]

and my stacktrace failed to generate

#0  panic (message=<optimized out>, frame=<optimized out>) at src/kernel/utils/panic.c:409
#1  0xffffffff8000c0df in dispatch_interupt_asm ()
#2  0x0000000000000008 in ?? ()
#3  0x0000000000000206 in ?? ()
#4  0xffff810000030950 in ?? ()
#5  0x0000000000000000 in ?? ()

[viral bison]

yeah well, if the stack gets nuked

[hazy saddle]

yeah entire thing gets nuked

bump alloc pool is in here somwhere

prolly junk tho

@viral bison ERROR : 00000000 00000000 00000000 00001011
this aparently means its a segment related error coming from an external thing from the IDT at index 0b1

for the GPF error

[viral bison]

oh

[hazy saddle]

let me try and repro the PF one

[devout path]

wat hi

[hazy saddle]

hello

we got more than one user app running

but other bugs

[devout path]

wat help do u guys need

[hazy saddle]

debugging

[devout path]

lemme back tread

[devout path]

owh

[hazy saddle]

ooo fun we got one of these

the panic panic

[viral bison]

yeah i got a few of those

[hazy saddle]

serial manages to keep the logs properly tho

[devout path]

dis main problem?

[hazy saddle]

well kinda

[viral bison]

no

the stack is getting nuked

[hazy saddle]

that dosnt happen on qemu but it will on rea; hwardware right

[devout path]

oke ima take my time and read

[hazy saddle]

but this is most important

[devout path]

wat

[viral bison]

ye

[hazy saddle]

your also gonna want the most recent patch file?

is this a fucking reserved bit being set FLAGS : 00000000 00000001 00000000 00000010

[devout path]

can ya'll summarize the problem for me huhu

im like

brain dedge rn

[hazy saddle]

when running userspace apps

the kernel can shit its pants

and either PF or GPF

[devout path]

what the sigma?

randomly?

[hazy saddle]

randomly

[devout path]

HUH

[hazy saddle]

yes

[devout path]

what's the stack trace

[hazy saddle]

fucked

[devout path]

random every time?

[hazy saddle]

yup

[viral bison]

not really random

i kept getting the same pagefault

but it was at a sub 0x1000 addr

[hazy saddle]

but it can also somtimes GPF

[devout path]

okay that's at least great

[viral bison]

ye

that too

it can be GPF or pagefault

[devout path]

wat patch are ya'll runnin on

[viral bison]

patches2.patch

[devout path]

the github? then patches2.patch?

[hazy saddle]

i have a patches3.patch i need to generate

[hazy saddle]

yes

do git diffs have untracked files?

[viral bison]

no

[devout path]

no

[hazy saddle]

how do i make it do that then

[viral bison]

add the files

[devout path]

can ya'll send a

nvm

[hazy saddle]

this is getting a lil big

[devout path]

heh-

thanks

wat

[hazy saddle]

hm?

[devout path]

do i really need bootx64.efi on my /usr/share

when i did make run it said that

oh wait

im actually kinda dumb rn

[devout path]

;-; my ovmf name is different form here

i've been manually changing some stuff XD

wat

i can't seem to make this pf

i'll be back in like 30 mins to see what's happenin here

@hazy saddle

this must have to do with how the pages work ngl

[hazy saddle]

hm?

[devout path]

i'll add more debug to ur pf if that's oke

[hazy saddle]

yeah yeah ofc

i have a spot for that actualy

[devout path]

im jus wildly guessin here but it must've been somethin that had to do with the corelation of the page entry table and the userspace

[hazy saddle]

part of the stack trace code has checks for the kind of fault it is

[devout path]

yep

[hazy saddle]

you can reorder it a bit and use that

hey @devout path atleast my kernel is less buggy than minecraft bedrock

[devout path]

TRUE

XDDDDDDDDDDDD

[hazy saddle]

Honestly me you and diamond could probably remake bedrock in a few weeks and have something better

Honestly that could be a fun group project after our kernels are in a decent shape

[devout path]

to be honest, we could LMAO

[hazy saddle]

Make a from scratch vulkan game engine?

Mine is in 2d and I need to learn a lot more vulkan to fix the thing

[devout path]

YEAH

that would be awesome ngl

[hazy saddle]

I did a c++ project with hot reloadable modules even

Could use some of that tech with it

It is LGPLv3 but it’s my own code so I can do whatever I want with it

[devout path]

daangg

[hazy saddle]

So when developing stuff could be modular and they can be reloaded like parts of the event loop

There is only one bug that I found with it too

If you delete a module

And then reload it gets unlosded

But if you put it back

And then reload to attempt to get it back

It will crash

But that’s not HORRIBLE

And if the system dosnt need to be fully dynamic it will reduce issues

Or cross threads

Because I also had a thread pool

[devout path]

XDDD

[hazy saddle]

But yeah what seems to be the happs here

[devout path]

i wrote debug and my guess WASN'T CORRECT LMAO

hmmmmmmmmmmmmmmmmmm

[hazy saddle]

https://tenor.com/view/there's-nothing-we-can-do-gif-14390726165144583214

[devout path]

i thought it would be that the pages are being assigned as user space

[viral bison]

yep, lol

[devout path]

then i realized even after changing the vmm.c a bit

cuz i thought

[viral bison]

vulkan is pain tho

[devout path]

I THOUGHT it was faulty

[devout path]

my bad gng

[hazy saddle]

Yeah but it would be good for preformance for a Minecraft like game

[hazy saddle]

Nah nah I’m joking

[devout path]

compare it to j a v a :)

XDD

[hazy saddle]

Because the pipeline can be tailored for block shor

Could even have threads for generating optimized geometry too?

So that’s offloaded

[devout path]

apparently..

have anyone tried

patch3 in real hardware?

[hazy saddle]

My USB stick is gone

Missing

[devout path]

wanna know what's funny?

it's working perfectly

LMAO

[viral bison]

i tried patch2

[hazy saddle]

HUH

[devout path]

HAHAHAHAHAHAH

PATCH 2 SEEMS TO HAVE SOMETHING

[hazy saddle]

QEMU or KVM bug?

[devout path]

SOMETHING THAT I DON'T KNOW LMAO

[hazy saddle]

let me try TCG?

[devout path]

TRY PATCH 3

i did every thing i had to do for analysis

and im 80% sure it's all fine

i thought it was the vmm.c being a lil bitch

i skimmed it and foudn some questionable stuff

but when i put it in debug it had same result

[hazy saddle]

it crashes even faster on TCG

[devout path]

even if it was coded like that

HUH

IT'S CRASHING FOR U?

[devout path]

dude intel is built diff i guess LMAO

[hazy saddle]

in QEMU and TCG or KVM

[devout path]

im runnin on KVM rn

IM SO CONFUSED XD

wait..

i have a different bootx64 efi

nah most likely not because of that no?\

what in the actual fuck is this LMAO

[hazy saddle]

here is mine

[viral bison]

i can't build it

[hazy saddle]

it should be limines?

[devout path]

i downloaded recent limine's binaries

[hazy saddle]

hows your assembler

[devout path]

cuz i don't hav eit in usr/share

[devout path]

wat huh

[devout path]

can ya'll try this simple fix

[hazy saddle]

im on latest

[devout path]

this is recent patch btw @viral bison

[viral bison]

i applied patches3

[devout path]

wat da

patches 3 worked for me

[viral bison]

do i have to apply patches.patch too?

[devout path]

no i don't think so ;-;

[hazy saddle]

apply it to fresh

[devout path]

i only applied it to a new instance

[viral bison]

i stashed the old ones

[devout path]

in vmm.c

[viral bison]

so it was clean

[devout path]

change | to &

[devout path]

now it did a double PANIC HUH

ok im gonna stop guessing and actually check this out

@hazy saddle

#include <stdint.h>

#include <interupts/interupts.h>
#include <utils/panic.h>

#define PF_PRESENT (1 << 0)     // Page is not present
#define PF_WRITE (1 << 1)       // Fault was a write
#define PF_USER (1 << 2)        // Fault from user mode
#define PF_RESERVED (1 << 3)    // Reserved bit was set
#define PF_FETCH (1 << 4)       // Instruction fetch fault

__attribute__((interrupt))
void page_fault_isr(struct interrupt_frame* frame, uint64_t error) {
    // my head hurts
    uint64_t faulting_addr;
    asm volatile("mov %%cr2, %0" : "=r"(faulting_addr));
 
     int present = (error & PF_PRESENT) ? 1 : 0;
    int write = (error & PF_WRITE) ? 1 : 0;
    int user = (error & PF_USER) ? 1 : 0;
    int reserved = (error & PF_RESERVED) ? 1 : 0;
    int fetch = (error & PF_FETCH) ? 1 : 0;
    volatile uint64_t pf_addr = faulting_addr;
    volatile int pf_flags = error;
    volatile int pf_present = present;
    volatile int pf_write = write;
    volatile int pf_user = user;
    volatile int pf_reserved = reserved;
    volatile int pf_fetch = fetch;
    
    (void)pf_addr;
    (void)pf_flags;
    (void)pf_present;
    (void)pf_write;
    (void)pf_user;
    (void)pf_reserved;
    (void)pf_fetch;
    
    panic("Page Fault", 0x0E, frame, error, PANIC_FLAGS_FRAME | PANIC_FLAGS_ERROR | PANIC_FLAGS_VECTOR);
}

can u add this to ur page_fault.c

just check if i messed something up LMAO

[hazy saddle]

void page_fault_isr(struct interrupt_frame* frame, uint64_t error) {
huh

it dosnt take in an error anymore

and it shouldnt be attribute((interrupt))??

did you not get the assembly intterupts code?

[devout path]

wat

huh

[hazy saddle]

did your patch apply right?

[devout path]

#include <stdint.h>

#include <interupts/interupts.h>
#include <utils/panic.h>

attribute((interrupt))
void page_fault_isr(struct interrupt_frame* frame, uint64_t error) {
panic("Page Fault", 0x0E, frame, error, PANIC_FLAGS_FRAME | PANIC_FLAGS_ERROR | PANIC_FLAGS_VECTOR);
}

wasn't this the code earlier

[hazy saddle]

that code got goned reduced to atoms

[devout path]

oops i think it didn't ;-;

o h

w e l p back to le drawing board

[hazy saddle]

it should be like this

#include <interupts/interupts.h>
#include <utils/panic.h>

void page_fault_isr(struct interrupt_frame* frame) {
    panic("Page Fault", frame);
}

[devout path]

heh nope ;-;

wait gimme a second

[hazy saddle]

i got a whole

extern dispatch_interupt

%macro ISR 1
global isr%1
isr%1:
    cli
    push 0
    push %1
    jmp dispatch_interupt_asm
%endmacro

%macro ISR_ERR 1
global isr%1
isr%1:
    cli
    push %1
    jmp dispatch_interupt_asm
%endmacro

dispatch_interupt_asm:
    push rax
    push rbx
    push rcx
    push rdx
    push rsi
    push rdi
    push rbp
    push r8
    push r9
    push r10
    push r11
    push r12
    push r13
    push r14
    push r15

    mov rdi, rsp
    call dispatch_interupt

    pop r15
    pop r14
    pop r13
    pop r12
    pop r11
    pop r10
    pop r9
    pop r8
    pop rbp
    pop rdi
    pop rsi
    pop rdx
    pop rcx
    pop rbx
    pop rax

    add rsp, 16
    iretq

; Exceptions
ISR_ERR 0x08           ; DF
ISR_ERR 0x0D           ; GPF
ISR_ERR 0x0E           ; PF

; Hardware
ISR     0x20           ; PIT
ISR     0x21           ; PS/2
ISR     0x24           ; Serial #1 & #3

; Spurious Interupts
ISR     0x27           ; Spurious PIC1
ISR     0x2F           ; Spurious PIC2

; System Call
ISR     0x69           ; Non POSIX System Calls

; Others
ISR     0x16           ; Reserved (Used as placeholder for generic)

[devout path]

lemme rm -rf real quick

oopsie

anyways lemme make sure my patch actually applied first

[hazy saddle]

[devout path]

XD

oops

heh...............................................................

welp

back to the drawing board

[devout path]

i tried everythin

i'll get back on this tmr

i need sleep xd it's like almost 12

[hazy saddle]

damn

[devout path]

@hazy saddle

@hazy saddle

@hazy saddle

heh i ain't sleeping apparently

[hazy saddle]

oh shit what

[devout path]

YEP SAME REACTION

I ADDED SOME DEBUG XD

i ran that then it did that

yk what's funny i added an infinite loop but uuh it just went to that

[hazy saddle]

how tf did it fail to load the bin file

[devout path]

0 damn clue

honestly i have no clue

oh wait

i think

nvm

nahi have no clue honestly

[hazy saddle]

can you change the retuen values of the VFS layers?

see which one fails

eg

make these -2

[devout path]

gime a sec

[hazy saddle]

@devout path is bad apple running atleast once if at all?

[devout path]

n o p e

[hazy saddle]

It works for me???

[devout path]

huh

wdym

like it runs?

it returns fs_read is the problem with me

[hazy saddle]

Bad Apple runs

Is the initramfs building right?

Ok ok

There is an issue with my builds

If the bad apple compile fails

It fails silently

Cd into the folder and run the script manually

[devout path]

i mean it does run but it pf's

[devout path]

hmm me?

[hazy saddle]

Yes you

If the fs call is failing

Check the tar file?

[devout path]

oke wait

o h

@hazy saddle

y e a

@hazy saddle bad apple is indeed missing

[hazy saddle]

Try running the script manually

For the bad apple build

[devout path]

wat's the script

is it compile user apps

[hazy saddle]

The one under src/user/badapple

You need to cd there too

[devout path]

i kinda get it now

i don't have clang

ah yes i completely forgot

im not on my fedora

im on ubuntu

my bad my bad

thanks for pointin that out

[hazy saddle]

Isn’t the main kernel also compiled with clang

[devout path]

w a i t.

yeah i just realized

huh

[hazy saddle]

Huh

[devout path]

now this is even more confusing LMAO

cuz i don't have clang on the os that i am using rn

philippine servers are absolute garbage

oh damn i built with gcc

fuckin

holy

BALLS

[hazy saddle]

It’s the power of hopes and dreams

[devout path]

LMAO

I BUILT WITH GCC

[hazy saddle]

GCC works but aparently it’s shit now lmao

So I’m using clang

[devout path]

okay it's running

[devout path]

and it takes 3gb to make a cross compiler :)

fuck that ngl

[hazy saddle]

What the fuck is a cross compiler!!!!!!!

https://tenor.com/view/sfm-soldier-tf2-meme-american-gif-24728385

I don’t use one

I probably should

But ehhhh

[devout path]

XD

[hazy saddle]

How does cross compiler work on clang

[devout path]

oh idk i think clang already has one

u don't have to build it like how u did with gcc

im not entirely sure i always used gcc XD

[rancid viper]

clang is a cross-compiler

by nature

it uses llvm as its backend so

[devout path]

there u go thank u so much

[rancid viper]

to be clear it wont have a target for your kernel by default obviously

so u still have to do that if thats what is being discussed

[devout path]

jimmir@jimmir-workstation:~/debugevalynOS$ addr2line -e ./kernel.elf 0xffffffff8000fa9f
??:?
jimmir@jimmir-workstation:~/debugevalynOS$

[rancid viper]

a compiler able to target a different platform than the host platform

[devout path]

this ain't useful at all lmao

[hazy saddle]

I was making somthing called a joke

[devout path]

XD

LMAOOOOOOOOO

[hazy saddle]

XD

[rancid viper]

did not make that very clear

[hazy saddle]

Does that need to be complied from scratch or is it a conf file thing

[rancid viper]

needs to be compiled

[obtuse loom]

depends on which compiler you wanna use

gcc needs to be rebuilt

[hazy saddle]

Is there a good way of distributing that for people? Or do I make people spend like an hour comping clang

[obtuse loom]

clang can often do it with a few command line flags

what do you want to target

[rancid viper]

their own kernel

[obtuse loom]

oh okay then you can just use clang

no rebuild needed

[hazy saddle]

Yeah and I already use clang with a ton of command line arguments so it should already be fine

[obtuse loom]

It takes way more than an hour nowadays lmao

[obtuse loom]

okay cool

as long as the flags are right then yes

[hazy saddle]

I used to use native GCC with the same arguments and it worked

[obtuse loom]

lucky 😂😂

[devout path]

w a t

[hazy saddle]

The flags were like

[devout path]

heh

[hazy saddle]

https://tenor.com/view/random-bullshit-go-moon-knight-random-random-bullshit-gif-25725102

Everything

Every flag

Oh you should see my first build script I had for this kernel before it got serious

https://github.com/EvalynGoemer/evalynOS/blob/00421909b5ef629ed39a9108a0ca2b017fc4dfba/make.sh

[rancid viper]

oh so targeting freestanding

[hazy saddle]

Actualy @devout path take a look at that and the bad apple build script

I wonder if you can see how similar they are

[devout path]

what did u do before u even messed this up

[devout path]

hmm?

[hazy saddle]

Well before only one bad apple could run at once

Now there are two

Is it even running properly for you now?

Compiling etc

[devout path]

yes i can run badapple

[hazy saddle]

Run it again

And then run credits a bunch of times until it decides to crash

[devout path]

yep i got a gpf

that's why i did addr2line earlier

then it returned a whopping ?:??

very insightful

[hazy saddle]

This shits fucked

[devout path]

truly

im skimming files rn

what the fuck

[obtuse loom]

this looks fine besides not disabling xmm register usage

[devout path]

the moment i go out of qemu it pf's

yeah

[hazy saddle]

I actually did use FPU registers in some of my kernel code at one point

[obtuse loom]

nevermind then ☠️

[hazy saddle]

Like on purpose

[devout path]

heh

[hazy saddle]

I said I am going to use the floating point registers

And then I used them

And it worked

[devout path]

actually i do nothing then it pfs

[hazy saddle]

Quite queer

[devout path]

AYO

UWOAH

IT'S BEEN 5 MINS

NO CRASH

@hazy saddle

[hazy saddle]

It’s random innit?

Try smashing the keyboard a bit

[devout path]

heh.......................................................................................................................................

it is not crashing

i think i found it

[obtuse loom]

needs more in-kernel floating point

[hazy saddle]

What about my set of 60 randomized kernel panic messages

[hazy saddle]

Ooo

[obtuse loom]

is it true random? 👀

[devout path]

idk if ur going to judge my fix but uuh

it works LMAO

badapple sounds like deathcore to me now

dang

[hazy saddle]

Nah it’s just based of the TSC at the time of panic and some random algorithm

[hazy saddle]

What’s the fix?

[devout path]

well

i made the context switch save evry registr

and then

i made it so that

the initial thread matches to the first switch and returns to actual code instead of whatever the fuck mlao

so yeah it's been 15 mins

no crash XD

[hazy saddle]

@viral bison is the one that made the thing not save every register

[viral bison]

you shouldn't do that tho

that's strange

[devout path]

i m e a n : )

[hazy saddle]

What does this mean

[devout path]

it w o r k s : )

well...

take a look for urself

[hazy saddle]

But isn’t that how preemption is done?

Don’t you have to save all registers

[devout path]

global thread_switch
global switch_to_user

section .text

thread_switch:
    pushfq
    push rax
    push rbx
    push rcx
    push rdx
    push rsi
    push rdi
    push rbp
    push r8
    push r9
    push r10
    push r11
    push r12
    push r13
    push r14
    push r15

    mov [rdi], rsp
    mov rsp, rsi

    pop r15
    pop r14
    pop r13
    pop r12
    pop r11
    pop r10
    pop r9
    pop r8
    pop rbp
    pop rdi
    pop rsi
    pop rdx
    pop rcx
    pop rbx
    pop rax
    popfq
    ret

USER_STACK_TOP equ 0x0000000080000000

switch_to_user:
    mov ax, 0x30 | 3
    mov ds, ax
    mov es, ax
    mov fs, ax
    mov gs, ax

    push 0x30 | 3
    mov rax, USER_STACK_TOP
    push rax
    push 0x200
    push 0x28 | 3
    push 0x4000
    iretq

[viral bison]

i don't save all the regs and it works

[devout path]

#include "stddef.h"
#include <stdint.h>
#include <stdlib.h>
#include <string.h>

#include <scheduler/scheduler.h>
#include <scheduler/switch.h>
#include <filesystem/filesystem.h>
#include <memory/vmm.h>
#include <memory/pmm.h>
#include <drivers/x86_64/ports.h>
#include <drivers/x86_64/gdt.h>
#include <utils/panic.h>

uint64_t STACK_SIZE = 65536;

void task_quit() {
    while (1) {

    }
}

struct thread_node* threads = NULL;
int next_thread_id = 0;

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

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

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

    new_thread->is_user_task = 0;
    new_thread->threadId = next_thread_id;
    next_thread_id++;

    if (pagemap != NULL) {
        new_thread->pagemap = pagemap;
    } else {
        new_thread->pagemap = kernel_pagemap;
    }

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

    if (entry_point == NULL) {
        entry_point = task_quit;
    }
    *--stack = (uint64_t)task_quit;

    // heh mlao
    *--stack = (uint64_t)entry_point; 
    *--stack = 0x202;                

    *--stack = 0; // rax
    *--stack = 0; // rbx
    *--stack = 0; // rcx
    *--stack = 0; // rdx
    *--stack = 0; // rsi
    *--stack = 0; // rdi
    *--stack = 0; // rbp
    *--stack = 0; // r8
    *--stack = 0; // r9
    *--stack = 0; // r10
    *--stack = 0; // r11
    *--stack = 0; // r12
    *--stack = 0; // r13
    *--stack = 0; // r14
    *--stack = 0; // r15

don't judge

[devout path]

: )

[viral bison]

why task quit too

[devout path]

i'll leave u guys to figure that out

im brain fried

it's working ; )

[devout path]

    new_thread->rsp = (uint64_t)stack;

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

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

    shouldSchedule = 1;
}

void schedule() {
    outb(0x20, 0x20);

    asm volatile("cli");

    struct thread *previous_thread = threads->thread;
    threads = threads->next_thread;
    struct thread *current_thread = threads->thread;

    vmm_switch_to(current_thread->pagemap);

    tss.rsp0 = (uint64_t)current_thread->stack_top;

    thread_switch(&previous_thread->rsp, current_thread->rsp);
}

struct thread *get_current_thread() {
    return threads->thread;
}

this the extention to the scheduler.c

im actually brain cooked rn

im gonna look at a window and refresh

brb

[hazy saddle]

what happens if a task is using r9. it context switches. and another clobbers it

[devout path]

im back

[viral bison]

the problem is that the user registers are saved on the stack

[devout path]

so the code earlier only set a partial register set and had a stack layout that d i d n t m at ch ;-;

[hazy saddle]

[viral bison]

so changing there should not have any effect

[devout path]

heh : )

[viral bison]

for some reason

[devout path]

my brain fried for dis : )

im gonna look at window more

[hazy saddle]

what is 202 here *--stack = 0x202;

[devout path]

basically this bug now fixed pls do try

[viral bison]

flags

[hazy saddle]

but what flags

[viral bison]

rflags

for the kernel

[hazy saddle]

but it changed

@devout path

[devout path]

0x202 is the initial rflags correct

HUH

WHY IS IT WORKING FOR ME

[viral bison]

it prob wasn't the regs

[devout path]

did u remove the

[viral bison]

as i said

[devout path]

remove the thing

the

no

it is the regs

u have to remove one thing

the

sti

fuck

what was it again

bro i didn't take notes

wait

[hazy saddle]

wait wairt

[hazy saddle]

i got it

[devout path]

idk just compare

[viral bison]

but the user registers are not touched in the ctx switch

[devout path]

oke nice

[devout path]

jus compare it here

i mean

[hazy saddle]

i had a fuckup

[devout path]

been running for a while now so

do i give patch file?

[hazy saddle]

i got it i think

[devout path]

oh

[hazy saddle]

its not crashing

[devout path]

heh

: )