OBOS (not vibecoded) | Operating System Development | Page 30

thick jolt Dec 24, 2024, 9:16 PM

#

probs ur other codebase relys on that or smthin

#

mmm

#

i could give waitable objects a new field

#

called interrupted

#

and check if that was enabled

#

but again idk that may be stupid too

#

where is the waitable_header structure defined

#

i dont have clangd working

#

and i dont want to bear -- make the project

#

or whatever build system u use

#

hold on i have a search feature

#

found it

#

omar do u think adding a new field called interrupted is a good idea

#

i want to check this againist you

#

whenever u come back let me know

#

im not gonna touch anything until i know what i am thinking of is a good idea or no, i really do NOT wanna nyauxify obos

#

code is very well written btw

#

ill give you this

flint idol Dec 24, 2024, 9:21 PM

#

lgtm

#

that's what I was thinking too

#

but I didn't wanna tell you the solution

weak kestrel Dec 24, 2024, 9:22 PM

#

@thick jolt did you get muted

thick jolt Dec 24, 2024, 9:22 PM

#

yes

#

😭

weak kestrel Dec 24, 2024, 9:22 PM

#

L

flint idol Dec 24, 2024, 9:22 PM

#

a note: not all waitable objects have use_signaled=true

#

which causes the signaled field to be ignored

#

(I really need to document my code)

thick jolt Dec 24, 2024, 9:23 PM

#

flint idol but I didn't wanna tell you the solution

yea

#

im adding it

#

?

#

oh right

#

okay but its still complaining

#

even after i changed it to 0

#

weird?

flint idol Dec 24, 2024, 9:24 PM

#

it's one bit

#

so : 1

white mulch Dec 24, 2024, 9:24 PM

#

why would you have a field with 0 bit width

thick jolt Dec 24, 2024, 9:24 PM

#

OHHHH

white mulch Dec 24, 2024, 9:24 PM

#

you can't store anything in 0 bits

thick jolt Dec 24, 2024, 9:24 PM

#

sorry i did not understand what that meant

#

i dont do those in C

weak kestrel Dec 24, 2024, 9:24 PM

#

white mulch you can't store anything in 0 bits

you can store nothing in 0 bits

flint idol Dec 24, 2024, 9:24 PM

#

thick jolt i dont do those in C

for good reason

white mulch Dec 24, 2024, 9:24 PM

#

weak kestrel you can store nothing in 0 bits

well yeah but that's not very useful lol

weak kestrel Dec 24, 2024, 9:24 PM

#

:p

thick jolt Dec 24, 2024, 9:24 PM

#

flint idol for good reason

why so

flint idol Dec 24, 2024, 9:25 PM

#

they're just bad

thick jolt Dec 24, 2024, 9:25 PM

#

yea probs

#

anyways

#

now that i added this field

#

in abort_wait

flint idol Dec 24, 2024, 9:25 PM

#

for stuff like fixed-bit structures

thick jolt Dec 24, 2024, 9:25 PM

#

what i could do is just make interruptable true or whatever

#

signal the object

#

and then we can check

flint idol Dec 24, 2024, 9:27 PM

#

thick jolt what i could do is just make interruptable true or whatever

you're gonna abstract this behind a separate function right?

#

and not make the user set that field then call signal threads

thick jolt Dec 24, 2024, 9:27 PM

#

flint idol you're gonna abstract this behind a separate function right?

abort_wait will do this

flint idol Dec 24, 2024, 9:27 PM

#

ah okay

thick jolt Dec 24, 2024, 9:27 PM

#

whats the naming scheme for your functions

flint idol Dec 24, 2024, 9:27 PM

#

don't forget to name it in obos' naming convention

thick jolt Dec 24, 2024, 9:27 PM

#

like CoreS

#

and stuff

flint idol Dec 24, 2024, 9:28 PM

#

#1141057599584878645 message

#

here

thick jolt Dec 24, 2024, 9:28 PM

#

hold on brb

#

i need to tell my parents my voice is back

flint idol Dec 24, 2024, 9:28 PM

#

you could that as CoreH_AbortWaitingThreads

thick jolt Dec 24, 2024, 9:31 PM

#

flint idol you could that as `CoreH_AbortWaitingThreads`

return type of void correct

flint idol Dec 24, 2024, 9:31 PM

#

obos_status

thick jolt Dec 24, 2024, 9:31 PM

#

okay

flint idol Dec 24, 2024, 9:31 PM

#

implement in wait.c pls

thick jolt Dec 24, 2024, 9:31 PM

#

okay

#

flint idol Dec 24, 2024, 9:35 PM

#

wrong

#

set interrupted to true

#

then just return CoreH_SignalWaitingThreads(obj, true, false);

#

otherwise you're repeating code

thick jolt Dec 24, 2024, 9:36 PM

#

how is this

flint idol Dec 24, 2024, 9:36 PM

#

look again

thick jolt Dec 24, 2024, 9:37 PM

#

whoops sorry

flint idol Dec 24, 2024, 9:37 PM

#

np

thick jolt Dec 24, 2024, 9:37 PM

#

huh

#

OHHHHH

#

WAIT

#

#

hold on

#

okay now i have the function @flint idol

flint idol Dec 24, 2024, 9:42 PM

#

wooo

#

now add the proper support for it in the waiting functions

thick jolt Dec 24, 2024, 9:43 PM

#

in CoreH_SignalWaitingThreads

flint idol Dec 24, 2024, 9:43 PM

#

no

#

the waitonobject function

#

I can do wait on objects since that is more complicated

thick jolt Dec 24, 2024, 9:43 PM

#

hold on

lean glen Dec 24, 2024, 9:44 PM

#

just do

#

if(interrupted)

#

return ABORTED

thick jolt Dec 24, 2024, 9:44 PM

#

i added this

#

now if it isnt

flint idol Dec 24, 2024, 9:44 PM

#

wrong

thick jolt Dec 24, 2024, 9:44 PM

#

what do i do here

flint idol Dec 24, 2024, 9:44 PM

#

just do

#

if (obj->interrupted) return OBOS_STATUS_ABORTED after the block function is called

thick jolt Dec 24, 2024, 9:45 PM

#

done

#

will i make a pr?

flint idol Dec 24, 2024, 9:46 PM

#

now test it

#

of course

#

make a mutex in one thread

thick jolt Dec 24, 2024, 9:46 PM

#

how do i test it 😭

flint idol Dec 24, 2024, 9:46 PM

#

on another

thick jolt Dec 24, 2024, 9:46 PM

#

where

flint idol Dec 24, 2024, 9:46 PM

#

flint idol on another

thread try to acquire it

#

basically:

mutex mut;
// thread one
acquire(mut)
interrupt_wait(WAITABLE_OBJECT(mut));
// thread two
// calls into wait on object
assert (acquire(mut) != ABORTED);```

thick jolt Dec 24, 2024, 9:48 PM

#

in which function

flint idol Dec 24, 2024, 9:48 PM

#

nowhere

#

you need to test it

thick jolt Dec 24, 2024, 9:48 PM

#

kentry?

flint idol Dec 24, 2024, 9:48 PM

#

sure

thick jolt Dec 24, 2024, 9:48 PM

#

okay

flint idol Dec 24, 2024, 9:48 PM

#

arch/x86_64/entry.c

#

go to the bottom of the file

#

and add it somewhere

thick jolt Dec 24, 2024, 9:52 PM

#

do i need to create threads or

flint idol Dec 24, 2024, 9:52 PM

#

that is in the completely wrong place

#

and not only that

thick jolt Dec 24, 2024, 9:52 PM

#

oh

flint idol Dec 24, 2024, 9:52 PM

#

that was pseudocode

#

the thing I sent

thick jolt Dec 24, 2024, 9:52 PM

#

sorry

flint idol Dec 24, 2024, 9:52 PM

#

it's np

thick jolt Dec 24, 2024, 9:53 PM

#

flint idol that is in the completely wrong place

which place is correct

flint idol Dec 24, 2024, 9:53 PM

#

bottom of the file

thick jolt Dec 24, 2024, 9:53 PM

#

oaky

flint idol Dec 24, 2024, 9:54 PM

#

wait if you're gonna make a pull request

#

how don't you have a fork

#

afaik that's the only way to pr if you don't have write access

#

to a repo

thick jolt Dec 24, 2024, 9:54 PM

#

fuck

#

but wait

flint idol Dec 24, 2024, 9:54 PM

#

just fork it

thick jolt Dec 24, 2024, 9:54 PM

#

flint idol bottom of the file

like bottom bottom of the file

flint idol Dec 24, 2024, 9:54 PM

#

then you can git remote set-url origin link/to/fork

flint idol Dec 24, 2024, 9:54 PM

#

thick jolt like bottom bottom of the file

yes

#

but above the inline asm

#

around line 860

lean glen Dec 24, 2024, 9:55 PM

#

860 lines

#

in the entry file

#

wtf

flint idol Dec 24, 2024, 9:55 PM

#

no, 950

flint idol Dec 24, 2024, 9:55 PM

#

lean glen in the entry file

it's really unorganized

thick jolt Dec 24, 2024, 9:56 PM

#

here???

flint idol Dec 24, 2024, 9:56 PM

#

flint idol but above the inline asm

.

thick jolt Dec 24, 2024, 9:56 PM

#

okay

flint idol Dec 24, 2024, 9:56 PM

#

the inline asm is just a test program

#

for syscalls

thick jolt Dec 24, 2024, 9:56 PM

#

okay

#

now

flint idol Dec 24, 2024, 9:56 PM

#

also you need to make it multithreaded

#

it's the only way

#

to test it

thick jolt Dec 24, 2024, 9:57 PM

#

okay

flint idol Dec 24, 2024, 9:57 PM

#

but unfortunately, my threading api is bumcheeks

thick jolt Dec 24, 2024, 9:57 PM

#

nooo

flint idol Dec 24, 2024, 9:59 PM

#

thick jolt <:nooo:750466870993223750>

// not psuedocode
thread* thr = CoreH_ThreadAllocate(nullptr);
thread_ctx thr_ctx = {};
void* stack =  Mm_VirtualMemoryAlloc(&Mm_KernelCtx, nullptr, 0x10000, 0, VMA_FLAGS_KERNEL_STACK, nullptr, nullptr);
CoreS_SetupThreadContext(&thr_ctx, (uintptr_t)thread_entry_here, (uintptr_t)&any_user_data_here, true, stack, 0x10000);
CoreH_ThreadInitialize(thr, THREAD_PRIORITY_NORMAL, Core_DefaultThreadAffinity, &thr_ctx);
CoreH_ThreadReady(thr);

#

take this snippet

thick jolt Dec 24, 2024, 10:00 PM

#

okay now that ive created

#

a thread

#

i will create another as well?

flint idol Dec 24, 2024, 10:01 PM

#

nah you only need one

thick jolt Dec 24, 2024, 10:01 PM

#

okay

lean glen Dec 24, 2024, 10:01 PM

#

why CoreH

#

what is CoreS

#

what

#

wa

flint idol Dec 24, 2024, 10:01 PM

#

as you already have another thread: the kernel main thread

flint idol Dec 24, 2024, 10:01 PM

#

lean glen why CoreH

Core+Helper

flint idol Dec 24, 2024, 10:01 PM

#

lean glen what is CoreS

Core+arch Specific

thick jolt Dec 24, 2024, 10:01 PM

#

whats the api for creating a mutex

lean glen Dec 24, 2024, 10:01 PM

#

https://tenor.com/view/spongebob-eye-bleach-gif-11574607

Tenor

flint idol Dec 24, 2024, 10:02 PM

#

thick jolt whats the api for creating a mutex

mutex mut = MUTEX_INITIALIZE();```

thick jolt Dec 24, 2024, 10:02 PM

#

okay

flint idol Dec 24, 2024, 10:02 PM

#

acquire:
Core_MutexAcquire(&mut);

#

release:
Core_MutexRelease(&mut);

thick jolt Dec 24, 2024, 10:02 PM

#

okay

flint idol Dec 24, 2024, 10:02 PM

#

and to convert a mutex to a waitable_header*:

#

WAITABLE_HEADER(mut);

thick jolt Dec 24, 2024, 10:03 PM

#

flint idol basically: ```c mutex mut; // thread one acquire(mut) interrupt_wait(WAITABLE_OB...

okay and then call waitforobjects or smthin?

flint idol Dec 24, 2024, 10:03 PM

#

that's implicitly called when you acquire a mutex

#

and it's already acquired

thick jolt Dec 24, 2024, 10:03 PM

#

okay

#

so now i need to call my abort_wait

#

yes?

flint idol Dec 24, 2024, 10:04 PM

#

yes

#

after you acquire the mutex in thread one

#

which is the kmain thread (where the entry code executes)

#

then you also must acquire the mutex in thread two

thick jolt Dec 24, 2024, 10:04 PM

#

but u said i only needed one thread

flint idol Dec 24, 2024, 10:04 PM

#

flint idol after you acquire the mutex in thread one

maybe it's better to acquire the mutex in thread two first, and abort the wait there

flint idol Dec 24, 2024, 10:04 PM

#

thick jolt but u said i only needed one thread

I only said you needed to create one thread

thick jolt Dec 24, 2024, 10:05 PM

#

ohhh

#

cause this kmain is already a thread

flint idol Dec 24, 2024, 10:05 PM

#

ye

thick jolt Dec 24, 2024, 10:05 PM

#

okay brb i might be gone for an hour or 2

flint idol Dec 24, 2024, 10:05 PM

#

sure

thick jolt Dec 24, 2024, 10:15 PM

#

flint idol release: `Core_MutexRelease(&mut);`

how do i get the kentrys mutex

flint idol Dec 24, 2024, 10:15 PM

#

through the user data

flint idol Dec 24, 2024, 10:16 PM

#

flint idol ```c // not psuedocode thread* thr = CoreH_ThreadAllocate(nullptr); thread_ctx t...

you see that one parameter here

thick jolt Dec 24, 2024, 10:16 PM

#

flint idol you see that one parameter here

yes

#

i dont have auto complete working so

#

whats the field im looking for?

flint idol Dec 24, 2024, 10:17 PM

#

&mut

#

change it to that

thick jolt Dec 24, 2024, 10:17 PM

#

okay

flint idol Dec 24, 2024, 10:17 PM

#

no

thick jolt Dec 24, 2024, 10:17 PM

#

oh

flint idol Dec 24, 2024, 10:18 PM

#

pass &mut there

vale nymph Dec 24, 2024, 10:18 PM

#

wtf rust

flint idol Dec 24, 2024, 10:18 PM

#

no

#

mut->mutex

thick jolt Dec 24, 2024, 10:18 PM

#

so its a function i can call?????

#

oh no

flint idol Dec 24, 2024, 10:18 PM

#

huh???????

thick jolt Dec 24, 2024, 10:18 PM

#

nvm

thick jolt Dec 24, 2024, 10:20 PM

#

flint idol pass &mut there

i dont understand im sorry

#

😭

flint idol Dec 24, 2024, 10:20 PM

#

know c?

thick jolt Dec 24, 2024, 10:20 PM

#

yes

flint idol Dec 24, 2024, 10:20 PM

#

you know how YOU'RE CALLING A FUNCTION

#

oops caps lock

thick jolt Dec 24, 2024, 10:20 PM

#

what is the function called

flint idol Dec 24, 2024, 10:20 PM

#

&any_user_data_here is a placeholder

thick jolt Dec 24, 2024, 10:20 PM

#

oh

#

so change that with my mutex?

#

yes or no?

flint idol Dec 24, 2024, 10:21 PM

#

yesa

thick jolt Dec 24, 2024, 10:21 PM

#

okay

flint idol Dec 24, 2024, 10:21 PM

#

&mut

thick jolt Dec 24, 2024, 10:22 PM

#

flint idol Dec 24, 2024, 10:22 PM

#

yup

#

actually

#

acquire the mutex after having started the thread

thick jolt Dec 24, 2024, 10:22 PM

#

okT

#

okay

flint idol Dec 24, 2024, 10:23 PM

#

and in both threads after having acquired the mutex, add ur interrupt wait function

#

and one of those waits should fail

thick jolt Dec 24, 2024, 10:23 PM

#

like this

flint idol Dec 24, 2024, 10:23 PM

#

with aborted

#

if you did all correctly

#

you see where I add thread_entry_here

#

that's also a placeholder

thick jolt Dec 24, 2024, 10:23 PM

#

yea

#

so the entry will be what?

#

exactly

#

js to make sure

flint idol Dec 24, 2024, 10:24 PM

#

the entry point of your other thread

#

the one u want to start

thick jolt Dec 24, 2024, 10:25 PM

#

the one im running in right now?

flint idol Dec 24, 2024, 10:25 PM

#

nay

#

read again

thick jolt Dec 24, 2024, 10:25 PM

#

i havent created 2 threads hiold on

thick jolt Dec 24, 2024, 10:26 PM

#

flint idol the entry point of your other thread

ohhh

#

nvm i dont understand

#

im sorry.

#

wait

#

do i put test_program into it

#

probs no

#

sorry

flint idol Dec 24, 2024, 10:28 PM

#

you are making another thread, yes

#

for hte other part of the test

#

?

#

MAKE THAT THE NETRY POINT

#

oops caps

#

lock

thick jolt Dec 24, 2024, 10:28 PM

#

flint idol lock

mkae the entry point the lock???

flint idol Dec 24, 2024, 10:28 PM

#

wut

#

wut

#

wut

#

wut

#

wut

quaint rivetBOT Dec 24, 2024, 10:28 PM

#

Anti-spam Message

Possible spam detected for user: oberrow. Please contact a moderator to be unmuted.

thick jolt Dec 24, 2024, 10:28 PM

#

SORRY

#

@real pecan

#

@white mulch

barren merlin Dec 24, 2024, 10:29 PM

#

.!unmute @flint idol

quaint rivetBOT Dec 24, 2024, 10:29 PM

#

@flint idol is not muted.

barren merlin Dec 24, 2024, 10:29 PM

#

next time ban

thick jolt Dec 24, 2024, 10:30 PM

#

wait really????

white mulch Dec 24, 2024, 10:30 PM

#

oberrow or nyaux not get timed out every day challenge meme

thick jolt Dec 24, 2024, 10:30 PM

#

barren merlin next time ban

your serious?

flint idol Dec 24, 2024, 10:30 PM

#

he's talking to me

thick jolt Dec 24, 2024, 10:30 PM

#

yes

#

but your serious??

barren merlin Dec 24, 2024, 10:30 PM

#

no

#

I'm not

thick jolt Dec 24, 2024, 10:30 PM

#

oh thank god

barren merlin Dec 24, 2024, 10:30 PM

#

I'm not evil

flint idol Dec 24, 2024, 10:31 PM

#

can someone explain to him

#

how thread entry points work

torpid wigeon Dec 24, 2024, 10:31 PM

#

who

thick jolt Dec 24, 2024, 10:31 PM

#

i know how they work

#

i just dont understand what entry point you want me to give it

flint idol Dec 24, 2024, 10:32 PM

#

the one u make

#

for the test

thick jolt Dec 24, 2024, 10:32 PM

#

okay so ill create a new function in the file

#

okay

#

okay i made one

flint idol Dec 24, 2024, 10:34 PM

#

make it static

#

and name it whatever

#

no need to follow the naming convention

thick jolt Dec 24, 2024, 10:34 PM

#

void static?

flint idol Dec 24, 2024, 10:34 PM

#

also give it mutex* mut as a parameter

torpid wigeon Dec 24, 2024, 10:34 PM

#

why is the void upper case

thick jolt Dec 24, 2024, 10:34 PM

#

okay

#

my bad

flint idol Dec 24, 2024, 10:34 PM

#

thick jolt void static?

static void

thick jolt Dec 24, 2024, 10:34 PM

#

thanks

torpid wigeon Dec 24, 2024, 10:35 PM

#

40 stars on obos

flint idol Dec 24, 2024, 10:35 PM

#

flint idol also give it `mutex* mut` as a parameter

because of ur userdata this works

flint idol Dec 24, 2024, 10:35 PM

#

torpid wigeon 40 stars on obos

indeed

thick jolt Dec 24, 2024, 10:35 PM

#

there

flint idol Dec 24, 2024, 10:36 PM

#

why are you doing &mut->mutex

#

just &mut

#

it's that simple

thick jolt Dec 24, 2024, 10:36 PM

#

sorry

flint idol Dec 24, 2024, 10:36 PM

#

np

thick jolt Dec 24, 2024, 10:36 PM

#

its &mut

flint idol Dec 24, 2024, 10:36 PM

#

indeed

#

after acquire don't forget to call your interrupt wait function

thick jolt Dec 24, 2024, 10:37 PM

#

my abort_wait

flint idol Dec 24, 2024, 10:37 PM

#

yes

thick jolt Dec 24, 2024, 10:37 PM

#

nono

#

okay

#

after mutex acquire?

flint idol Dec 24, 2024, 10:37 PM

#

yes

thick jolt Dec 24, 2024, 10:37 PM

#

okay

flint idol Dec 24, 2024, 10:37 PM

#

also do the same thing in the other thread you start

#

acquire+abort_wait

#

then Core_ExitCurrentThread

#

also you need to make sure the wait gets aborted properly

#

by checking the status of acquire

thick jolt Dec 24, 2024, 10:38 PM

#

okay

#

do i remove mutex acquire probs not right?

flint idol Dec 24, 2024, 10:38 PM

#

if (obos_is_error(status)) printf("fail");

flint idol Dec 24, 2024, 10:38 PM

#

thick jolt do i remove mutex acquire probs not right?

nop

thick jolt Dec 24, 2024, 10:38 PM

#

kk

flint idol Dec 24, 2024, 10:39 PM

#

btw

#

in mutex acquire you will need to make some changes

#

when it calls Core_WaitOnObject it never checks the status

thick jolt Dec 24, 2024, 10:39 PM

#

okay

flint idol Dec 24, 2024, 10:39 PM

#

thick jolt okay

acquire it then abort

thick jolt Dec 24, 2024, 10:40 PM

#

okay

#

leave the one in this thread right?

flint idol Dec 24, 2024, 10:40 PM

#

the idea here is to create two threads racing for the mutex, and the one that doesn't get it first will simply hang acquiring it

thick jolt Dec 24, 2024, 10:40 PM

#

okay

flint idol Dec 24, 2024, 10:40 PM

#

thick jolt leave the one in this thread right?

yes

thick jolt Dec 24, 2024, 10:40 PM

#

yea i understand

flint idol Dec 24, 2024, 10:40 PM

#

flint idol the idea here is to create two threads racing for the mutex, and the one that do...

and then the one that does get the mutex aborts the other mutex waiting

thick jolt Dec 24, 2024, 10:41 PM

#

okay

flint idol Dec 24, 2024, 10:41 PM

#

not quite there yet

#

obos_status status = Core_MutexAcquire(mut);

#

and also: WAITABLE_OBJECT(*mut)

thick jolt Dec 24, 2024, 10:41 PM

#

okay done

flint idol Dec 24, 2024, 10:42 PM

#

so it should look like:

obos_status status = Core_MutexAcquire(mut);
CoreH_AbortWaitingThreads(WAITABLE_OBJECT(*mut));```

thick jolt Dec 24, 2024, 10:42 PM

#

gotcha

#

thats all done

flint idol Dec 24, 2024, 10:42 PM

#

and the same thing on the first thread, but you don't dereference any pointers

thick jolt Dec 24, 2024, 10:42 PM

#

okay

flint idol Dec 24, 2024, 10:42 PM

#

to mut

#

then you compile obos

#

and in theory, status in one of the threads will be aborted

#

which you can printf

thick jolt Dec 24, 2024, 10:43 PM

#

flint idol Dec 24, 2024, 10:43 PM

#

printf("%d", status);

#

don't forget

thick jolt Dec 24, 2024, 10:43 PM

#

in arch_kernelmainbootstrap

flint idol Dec 24, 2024, 10:43 PM

#

yup

thick jolt Dec 24, 2024, 10:44 PM

#

flint idol Dec 24, 2024, 10:44 PM

#

now you will most likely run into a small bug with mutexes

#

because acquire doesn't forward the status of WaitOnObject

#

anyway gtg eat

thick jolt Dec 24, 2024, 10:44 PM

#

so lets make someone changes to that

flint idol Dec 24, 2024, 10:44 PM

#

flint idol because acquire doesn't forward the status of WaitOnObject

you should be able to figure this out easily

flint idol Dec 24, 2024, 10:45 PM

#

flint idol because acquire doesn't forward the status of WaitOnObject

if (status_of_wait_object != success) return status_of_wait_object; // then do the other shit

#

Gl and godspeed

lean glen Dec 24, 2024, 10:45 PM

#

thick jolt leave the one in this thread right?

Why is everything green

thick jolt Dec 24, 2024, 10:45 PM

#

idk

flint idol Dec 24, 2024, 10:45 PM

#

Oh wait I don't gtg eat

#

Yet

#

Brb though

thick jolt Dec 24, 2024, 10:47 PM

#

okay

flint idol Dec 24, 2024, 10:59 PM

#

Bqck

#

*back

thick jolt Dec 24, 2024, 11:03 PM

#

making changes to acquire

#

is this correct @flint idol

#

wait

#

no nvm

#

i have no idea

#

@flint idol

#

OOHOHOHOH WAIT

#

okay

#

@flint idol

#

i did this

#

is this correct yes or no

#

in acquiring

#

i think hes gone

#

hello?

#

also having these build errors

#

@flint idol ??????

flint idol Dec 24, 2024, 11:35 PM

#

thick jolt <@841762704766140486>

Yes

#

Huh weird

flint idol Dec 24, 2024, 11:36 PM

#

thick jolt also having these build errors

Gcc version?

thick jolt Dec 24, 2024, 11:36 PM

#

13.3.0

flint idol Dec 24, 2024, 11:37 PM

#

Hmm

#

Try clang

#

Reconfigure with -DOBOS_USE_CLANG=ON

#

You need to delete the cmake cache

#

Tho

thick jolt Dec 24, 2024, 11:37 PM

#

how

#

i dont use cmake

flint idol Dec 24, 2024, 11:38 PM

#

rm CMakeCache.txt

thick jolt Dec 24, 2024, 11:38 PM

#

flint idol Dec 24, 2024, 11:39 PM

#

rm build/CMakeCache.txt

thick jolt Dec 24, 2024, 11:39 PM

#

doesnt exist

#

found it

flint idol Dec 24, 2024, 11:39 PM

#

Can you ls build

thick jolt Dec 24, 2024, 11:39 PM

#

x86_64

flint idol Dec 24, 2024, 11:39 PM

#

Oh ok

flint idol Dec 24, 2024, 11:39 PM

#

thick jolt found it

Now delete it

thick jolt Dec 24, 2024, 11:39 PM

#

flint idol Now delete it

i did

flint idol Dec 24, 2024, 11:40 PM

#

flint idol Reconfigure with -DOBOS_USE_CLANG=ON

Do this

#

And obos will be built with clang

thick jolt Dec 24, 2024, 11:40 PM

#

okay

#

oh

#

ill change the names

flint idol Dec 24, 2024, 11:41 PM

#

#include <locks/wait.h> at the top of the file maybe?

#

idk

#

I do that tho

#

ur toolchain is cursed

thick jolt Dec 24, 2024, 11:42 PM

#

now its just this

flint idol Dec 24, 2024, 11:42 PM

#

you define this in wait.h, right?

thick jolt Dec 24, 2024, 11:42 PM

#

i didnt put it in wait.h

flint idol Dec 24, 2024, 11:42 PM

#

well no wonder then

thick jolt Dec 24, 2024, 11:42 PM

#

🤦‍♂️

#

flint idol Dec 24, 2024, 11:43 PM

#

bruh

#

execute_process(COMMAND x86_64-elf-gcc -print-file-name=no-red-zone/libgcc.a OUTPUT_VARIABLE LIBGCC)
if (LIBGCC STREQUAL "no-red-zone/libgcc.a")
    # Use normal libgcc
    execute_process(COMMAND x86_64-elf-gcc -print-file-name=libgcc.a OUTPUT_VARIABLE LIBGCC)
endif()```

#

this is how I get it..

#

can you run x86_64-elf-gcc -print-file-name=no-red-zone/libgcc.a

#

and show me the output

thick jolt Dec 24, 2024, 11:44 PM

#

flint idol Dec 24, 2024, 11:45 PM

#

hmm

#

can you go to line 52 of src/build/x86_64/toolchain.cmake and add message("hai")

#

and reconfigure cmake

#

after deleting cache

real pecan Dec 24, 2024, 11:46 PM

#

thick jolt oh

WTH, obos Thor reference

thick jolt Dec 24, 2024, 11:46 PM

#

flint idol can you go to line 52 of src/build/x86_64/toolchain.cmake and add `message("hai"...

okay

thick jolt Dec 24, 2024, 11:47 PM

#

flint idol can you go to line 52 of src/build/x86_64/toolchain.cmake and add `message("hai"...

???

flint idol Dec 24, 2024, 11:47 PM

#

I gave you an exact path

#

you didn't use it

#

relative to the repo's root

thick jolt Dec 24, 2024, 11:48 PM

#

oh

#

#

@flint idol

thick jolt Dec 25, 2024, 12:01 AM

#

real pecan WTH, obos Thor reference

real

#

trl

real pecan Dec 25, 2024, 12:01 AM

#

Thats managarms kernel btw

thick jolt Dec 25, 2024, 12:02 AM

#

ik

real pecan Dec 25, 2024, 12:05 AM

#

This is still funny

flint idol Dec 25, 2024, 12:11 AM

#

@real pecan obos curse

flint idol Dec 25, 2024, 12:12 AM

#

flint idol and reconfigure cmake

Do this @thick jolt

thick jolt Dec 25, 2024, 12:12 AM

#

flint idol can you go to line 52 of src/build/x86_64/toolchain.cmake and add `message("hai"...

id did that

#

there was no hi

#

flint idol Dec 25, 2024, 12:12 AM

#

Did you reconfigure cmake and delete the cache

thick jolt Dec 25, 2024, 12:13 AM

#

yes

flint idol Dec 25, 2024, 12:13 AM

#

Rather, delete the cache then reconfigure the cache

real pecan Dec 25, 2024, 12:13 AM

#

Lol

real pecan Dec 25, 2024, 12:13 AM

#

flint idol <@353593948523986945> obos curse

The beginning

thick jolt Dec 25, 2024, 12:13 AM

#

flint idol Rather, delete the cache then reconfigure the cache

i did

flint idol Dec 25, 2024, 12:14 AM

#

thick jolt i did

The message should be printed at configure time

thick jolt Dec 25, 2024, 12:14 AM

#

flint idol The message should be printed at configure time

OHHH

flint idol Dec 25, 2024, 12:14 AM

#

If it isn't then I have an idea of what the bug is

#

If it is then I also have an idea of what the bug is

thick jolt Dec 25, 2024, 12:15 AM

#

hold on

#

im like playing cs 2 with streaks and laura, after this match ill do so

#

its 2-2

flint idol Dec 25, 2024, 12:15 AM

#

Bruh ok

thick jolt Dec 25, 2024, 12:15 AM

#

wdym bruh ok 😭

real pecan Dec 25, 2024, 12:16 AM

#

What's your premier rating

thick jolt Dec 25, 2024, 12:16 AM

#

zero

#

havent played cs

#

before

real pecan Dec 25, 2024, 12:17 AM

#

Ohh

flint idol Dec 25, 2024, 12:17 AM

#

thick jolt wdym bruh ok 😭

Btw did you give up on nyaux or smth

#

Why'd you wanna contrib to obos

thick jolt Dec 25, 2024, 12:17 AM

#

flint idol Btw did you give up on nyaux or smth

I DIDNT GIVE UP ON NYAUX

#

😭

real pecan Dec 25, 2024, 12:17 AM

#

Nyaux obos merge when

thick jolt Dec 25, 2024, 12:17 AM

#

I JUST FELT NICE

#

😭

flint idol Dec 25, 2024, 12:17 AM

#

Lol was joking

thick jolt Dec 25, 2024, 12:18 AM

#

:sib

#

😭 😭

torpid wigeon Dec 25, 2024, 12:18 AM

#

thick jolt I DIDNT GIVE UP ON NYAUX

news to me

thick jolt Dec 25, 2024, 12:18 AM

#

lmao

torpid wigeon Dec 25, 2024, 12:18 AM

#

what happened to rust

thick jolt Dec 25, 2024, 12:19 AM

#

bro

#

i write nyaux in c now

#

im on my 3rd rewrite

torpid wigeon Dec 25, 2024, 12:19 AM

#

L

thick jolt Dec 25, 2024, 12:19 AM

#

to which im not giving up

real pecan Dec 25, 2024, 12:19 AM

#

W

thick jolt Dec 25, 2024, 12:19 AM

#

WDYM L

#

😭

#

C IS A W

flint idol Dec 25, 2024, 12:19 AM

#

torpid wigeon L

May

#

*nay

#

W

torpid wigeon Dec 25, 2024, 12:20 AM

#

nyaux you need rust handholding

weary hound Dec 25, 2024, 12:20 AM

#

thick jolt C IS A W

add a ++ and then i might agree

flint idol Dec 25, 2024, 12:20 AM

#

Is the right answer

torpid wigeon Dec 25, 2024, 12:20 AM

#

weary hound add a ++ and then i might agree

🗣️ GET OUT

short mortar Dec 25, 2024, 12:20 AM

#

C is painful

flint idol Dec 25, 2024, 12:20 AM

#

Skill issue

torpid wigeon Dec 25, 2024, 12:20 AM

#

https://www.youtube.com/watch?v=8WVBT7FrgJI

YouTube

Ttv_thereaper22

Get out sound effect!!

▶ Play video

thick jolt Dec 25, 2024, 12:20 AM

#

skill issue

torpid wigeon Dec 25, 2024, 12:20 AM

#

yea exactly what you're having

thick jolt Dec 25, 2024, 12:20 AM

#

torpid wigeon nyaux you need rust handholding

https://www.youtube.com/watch?v=8WVBT7FrgJI

YouTube

Ttv_thereaper22

Get out sound effect!!

▶ Play video

flint idol Dec 25, 2024, 12:20 AM

#

https://tenor.com/view/savages-pocahontas-gif-5458002

Tenor

torpid wigeon Dec 25, 2024, 12:21 AM

#

rust gave you some sort of harness troll2

thick jolt Dec 25, 2024, 12:21 AM

#

with rust i had more UB

#

then anything

short mortar Dec 25, 2024, 12:21 AM

#

I'm thinking about writing stuff in Ada for pmOS

flint idol Dec 25, 2024, 12:21 AM

#

torpid wigeon rust gave you some sort of harness <:troll2:898264292124655617>

Trust me it was the opposite

flint idol Dec 25, 2024, 12:21 AM

#

short mortar I'm thinking about writing stuff in Ada for pmOS

-🦭

short mortar Dec 25, 2024, 12:21 AM

#

short mortar I'm thinking about writing stuff in Ada for pmOS

Best language

flint idol Dec 25, 2024, 12:21 AM

#

Ada is a W

short mortar Dec 25, 2024, 12:22 AM

#

I know SQL, it should be similar

flint idol Dec 25, 2024, 12:22 AM

#

I wouldn't say that if ironclad weren't written in it

torpid wigeon Dec 25, 2024, 12:22 AM

#

i'd rather use V

real pecan Dec 25, 2024, 12:24 AM

#

Didnt go well for mint

short mortar Dec 25, 2024, 12:25 AM

#

short mortar I'm thinking about writing stuff in Ada for pmOS

I also had Go runtime almost working...

real pecan Dec 25, 2024, 12:25 AM

#

JVM when

torpid wigeon Dec 25, 2024, 12:25 AM

#

wasm

short mortar Dec 25, 2024, 12:26 AM

#

After I get a disk driver

flint idol Dec 25, 2024, 12:26 AM

#

When uJVM

real pecan Dec 25, 2024, 12:26 AM

#

Lmao

short mortar Dec 25, 2024, 12:26 AM

#

And a shell

short mortar Dec 25, 2024, 12:26 AM

#

flint idol When uJVM

uUSB first

real pecan Dec 25, 2024, 12:26 AM

#

Pmos running Minecraft playing on a server running on managarm

torpid wigeon Dec 25, 2024, 12:27 AM

#

uAPI headers need to be finished first

#

stare

real pecan Dec 25, 2024, 12:27 AM

#

trl

short mortar Dec 25, 2024, 12:27 AM

#

real pecan Pmos running Minecraft playing on a server running on managarm

Running in QEMU on Linux

white mulch Dec 25, 2024, 12:27 AM

#

jvm has some weird instructions that idk how to implement

#

like invokedynamic

flint idol Dec 25, 2024, 12:27 AM

#

Just write the drivers for obos then everyone needs to write shims for the obos driver interface

torpid wigeon Dec 25, 2024, 12:28 AM

#

@flint idol finish pci headers for infy

flint idol Dec 25, 2024, 12:28 AM

#

torpid wigeon <@841762704766140486> finish pci headers for infy

cp obos/src/oboskrnl/driver_interface/pci.h uAPI/include

#

Done

torpid wigeon Dec 25, 2024, 12:28 AM

#

fake

white mulch Dec 25, 2024, 12:30 AM

#

minecraft on obos when

#

or first more like openjdk on obos when

real pecan Dec 25, 2024, 12:30 AM

#

real pecan Lol

Oberrow back then: scared to touch acpi
Oberrow now: first hobby os with S3 suspend

thick jolt Dec 25, 2024, 12:40 AM

#

im back

thick jolt Dec 25, 2024, 12:41 AM

#

flint idol can you go to line 52 of src/build/x86_64/toolchain.cmake and add `message("hai"...

#

@flint idol

flint idol Dec 25, 2024, 12:54 AM

#

Bruh

flint idol Dec 25, 2024, 12:54 AM

#

real pecan Oberrow back then: scared to touch acpi Oberrow now: first hobby os with S3 susp...

Real

thick jolt Dec 25, 2024, 12:55 AM

#

i did it exactly

flint idol Dec 25, 2024, 12:55 AM

#

I will be with you in a bit

thick jolt Dec 25, 2024, 12:55 AM

#

kk

flint idol Dec 25, 2024, 1:05 AM

#

thick jolt

back

#

line 50 of the same file, insert this:
message(${LIGCC})

#

then delete cache and reconfigure

thick jolt Dec 25, 2024, 1:07 AM

#

flint idol line 50 of the same file, insert this: `message(${LIGCC})`

oaky

flint idol Dec 25, 2024, 1:08 AM

#

thick jolt oaky

btw, if 9+47 is airplane, and your boat has a flat tire as you fly through the desert, what should you have in your back pocket?

thick jolt Dec 25, 2024, 1:09 AM

#

thick jolt Dec 25, 2024, 1:09 AM

#

flint idol btw, if 9+47 is airplane, and your boat has a flat tire as you fly through the d...

😭

#

what

flint idol Dec 25, 2024, 1:09 AM

#

thick jolt

wth

thick jolt Dec 25, 2024, 1:10 AM

#

flint idol wth

i have no fucking idea

flint idol Dec 25, 2024, 1:10 AM

#

well I know the workaround

#

replace that if line with:

if (LIBGCC STREQUAL "no-red-zone/libgcc.a" OR LIBGCC STREQUAL "")

#

line 51

#

@thick jolt

thick jolt Dec 25, 2024, 1:12 AM

#

okay

#

okay

#

flint idol Dec 25, 2024, 1:14 AM

#

thick jolt

clear cache and reconfigure

#

and remove the message line at line 50

thick jolt Dec 25, 2024, 1:14 AM

#

flint idol clear cache and reconfigure

i did

#

ok

#

i did hto

flint idol Dec 25, 2024, 1:14 AM

#

good

#

if that doesn't work, try replacing the if at line 50 with:

if ("${LIBGCC}" STREQUAL "no-red-zone/libgcc.a" OR "${LIBGCC}" STREQUAL "")

#

@thick jolt

thick jolt Dec 25, 2024, 1:16 AM

#

oaky

flint idol Dec 25, 2024, 1:17 AM

#

while this does lead me to raise the question of why your gcc doesn't have the no-red-zone libgcc, we can ignore that for now

thick jolt Dec 25, 2024, 1:19 AM

#

flint idol if that doesn't work, try replacing the if at line 50 with: ```cmake if ("${LIBG...

still

#

can u just compile my changes @flint idol and test them please

flint idol Dec 25, 2024, 1:21 AM

#

send patch

#

git format-patch

#

don't forget to make a commit

thick jolt Dec 25, 2024, 1:21 AM

#

flint idol `git format-patch`

okay

flint idol Dec 25, 2024, 1:21 AM

#

then you can pop that in here and I will apply it and test it for you

thick jolt Dec 25, 2024, 1:23 AM

#

i ran git format-patch

#

where are the patches

flint idol Dec 25, 2024, 1:23 AM

#

in the repo root

thick jolt Dec 25, 2024, 1:23 AM

#

i see nothing

#

only ur patches

flint idol Dec 25, 2024, 1:23 AM

#

did you make a commit

#

?

thick jolt Dec 25, 2024, 1:23 AM

#

yes

#

i can make a fork

flint idol Dec 25, 2024, 1:24 AM

#

can you ls in the repo root

flint idol Dec 25, 2024, 1:24 AM

#

thick jolt i can make a fork

don't do that yet

thick jolt Dec 25, 2024, 1:24 AM

#

flint idol Dec 25, 2024, 1:24 AM

#

git log

#

make sure

#

you committed

thick jolt Dec 25, 2024, 1:24 AM

#

#

yep is my commit

flint idol Dec 25, 2024, 1:25 AM

#

git format-patch HEAD

thick jolt Dec 25, 2024, 1:25 AM

#

still nothing

flint idol Dec 25, 2024, 1:25 AM

#

git format-patch HEAD~1

thick jolt Dec 25, 2024, 1:26 AM

#

okay i have the patches

flint idol Dec 25, 2024, 1:26 AM

#

send

thick jolt Dec 25, 2024, 1:26 AM

#

📎 0001-yep.patch

#

apply them and try

flint idol Dec 25, 2024, 1:28 AM

#

compiling

thick jolt Dec 25, 2024, 1:30 AM

#

kk

flint idol Dec 25, 2024, 1:33 AM

#

good news mr. nyaux man

#

after fixing a bug in my code

#

your stuff worked

#

partially

thick jolt Dec 25, 2024, 1:34 AM

#

YESSS

#

wdym partially

#

whats wrong

flint idol Dec 25, 2024, 1:35 AM

#

you forgot to add an exit thread statement

#

to the test thread

thick jolt Dec 25, 2024, 1:36 AM

#

flint idol you forgot to add an exit thread statement

oh lol

#

well u can fix that

flint idol Dec 25, 2024, 1:36 AM

#

yeah

thick jolt Dec 25, 2024, 1:36 AM

#

and then

#

u can commit this

#

!!

#

make sure to give me credit

#

😎

flint idol Dec 25, 2024, 1:37 AM

#

ofc

#

flint idol Dec 25, 2024, 1:39 AM

#

thick jolt u can commit this

I will amend your commit for it to be your github account

thick jolt Dec 25, 2024, 1:39 AM

#

flint idol I will amend your commit for it to be your github account

okayokay

#

😎

flint idol Dec 25, 2024, 1:39 AM

#

then commit my code making the fixes

#

to the test

#

although there is a slight problem, sometimes the status for both threads is SUCCESS

thick jolt Dec 25, 2024, 1:40 AM

#

flint idol although there is a slight problem, sometimes the status for both threads is SUC...

oh no

#

nooo

flint idol Dec 25, 2024, 1:40 AM

#

I will debug that

thick jolt Dec 25, 2024, 1:40 AM

#

okay

flint idol Dec 25, 2024, 1:41 AM

#

bruh

#

    if (obj->interrupted) 
        return OBOS_STATUS_ABORTED;
    CoreH_ThreadBlock(curr, true);```

#

you're supposed to do that after the block

flint idol Dec 25, 2024, 1:42 AM

#

flint idol to the test

I think I'll actually remove the test actually

#

I don't usually push code with a test in it

#

and I will be making a proper commit message

thick jolt Dec 25, 2024, 1:48 AM

#

flint idol you're supposed to do that *after* the block

my bad

#

lol

#

😭

thick jolt Dec 25, 2024, 1:48 AM

#

flint idol and I will be making a proper commit message

oki

#

let me know when u commit

flint idol Dec 25, 2024, 1:48 AM

#

thick jolt let me know when u commit

https://github.com/OBOS-dev/obos/commit/99388ca792f60b1bf570d4b1e51a2841aacd2e76

GitHub

Implement interruptable waits. · OBOS-dev/obos@99388ca

Core: Add CoreH_AbortWaitingThreads.

thick jolt Dec 25, 2024, 1:49 AM

#

yesss

#

😎

flint idol Dec 25, 2024, 1:50 AM

#

will take a bit for you to be on the contributors list

#

because bah github

flint idol Dec 25, 2024, 2:04 AM

#

thick jolt yesss

if you want to do anything else

#

I have some other TODOs for you

thick jolt Dec 25, 2024, 2:04 AM

#

flint idol if you want to do anything else

i probs wont contribute much as i have my own kernel to work

#

on

#

but

flint idol Dec 25, 2024, 2:04 AM

#

not pressuring you

thick jolt Dec 25, 2024, 2:04 AM

#

ofc now

#

not*

#

but whenever im free

flint idol Dec 25, 2024, 2:05 AM

#

just if you're ever bored

thick jolt Dec 25, 2024, 2:05 AM

#

yea

#

ill contribute

#

add me to the contributes list lol

flint idol Dec 25, 2024, 2:05 AM

#

github is stoobid

#

so it might take a bit

thick jolt Dec 25, 2024, 2:05 AM

#

frfr

#

all good

flint idol Dec 25, 2024, 2:05 AM

#

and if it's ultra stupid I'll add you to the credits

thick jolt Dec 25, 2024, 2:13 AM

#

flint idol and if it's ultra stupid I'll add you to the credits

okay

flint idol Dec 25, 2024, 2:14 AM

#

// NOTTODO: Test.```
lol

#

in arch/x86_64/ssignal.c

flint idol Dec 25, 2024, 2:18 AM

#

thick jolt okay

if you're ever bored again and feel like contributing to obos, here are some current TODOs that I have for you:

fix the m68k port
fix the gdb stub
rewrite the disk swap interface
make the driver loader mostly platform-independant

thick jolt Dec 25, 2024, 2:20 AM

#

flint idol if you're ever bored again and feel like contributing to obos, here are some cur...

okay thanks

frank flare Dec 25, 2024, 7:48 AM

#

flint idol in arch/x86_64/ssignal.c

if my code works first-try that means it's actually not working

flint idol Dec 25, 2024, 12:44 PM

#

#

after I finish this small thing in obos

#

I will be doing vfs syscalls

flint idol Dec 25, 2024, 1:07 PM

#

I am currently fixing a bug related to standby/dirty lists and shared pages.
Previously, it would blindly assume that it could put any page on standby/dirty, which included shared pages.
This behaviour is incorrect, as if a page is shared, it's possible, and very likely, that the other shared page(s), are still using it while it's supposed to be treated as standby memory, or memory that will become standby (i.e., dirty pages).

#

The solution was to add another reference counter to struct page (physical page ofc)

#

// A reference count of virtual pages that have this page paged in.
// Must always be <= refcount.
_Atomic(size_t) pagedCount;```

#

this vmm work inspires me to unfuck my pagecache

#

it's not buggy, just shitty

flint idol Dec 25, 2024, 2:31 PM

#

this branch is getting out of hand

white mulch Dec 25, 2024, 2:31 PM

#

maybe you should merge it lol

flint idol Dec 25, 2024, 2:31 PM

#

I want to get my syscalls finished first

#

and fix any regressions in the m68k port

#

in a separate discord server, I predicted that I would have ~70 syscalls by the end of this

#

and I can never forget TTYs

#

which I need to implement

#

basically everything here

#

plus this vmm work

flint idol Dec 25, 2024, 4:21 PM

#

after getting myself to stop procrastinating

#

I am now going to finish the new api to the vmm

#

streaming obos in voice one

#

for those who care for some reason

torpid wigeon Dec 25, 2024, 4:47 PM

#

those who care 💀

flint idol Dec 25, 2024, 4:48 PM

#

this shit is so boring

thick jolt Dec 25, 2024, 4:53 PM

#

cant believe my code for obos worked first try ish

flint idol Dec 25, 2024, 4:53 PM

#

rea;

#

real

thick jolt Dec 25, 2024, 4:54 PM

#

i knew nothing about thread blocking btw

#

crazy

flint idol Dec 25, 2024, 4:54 PM

#

mhm

#

crazy

thick jolt Dec 25, 2024, 4:54 PM

#

flint idol streaming obos in voice one

still streaming?

flint idol Dec 25, 2024, 4:54 PM

#

I can in a bit

#

actually call me

thick jolt Dec 25, 2024, 4:55 PM

#

okay

astral mason Dec 25, 2024, 5:56 PM

#

huh?

thick jolt Dec 25, 2024, 5:56 PM

#

astral mason huh?

WHERES UR MANGA PLUS OS

astral mason Dec 25, 2024, 5:59 PM

#

thick jolt WHERES UR MANGA PLUS OS

Manga I haven't drawn any panels for yet lol, still working on the script. OS, er, I forgot about XD, I think I was trying to figure out how to read a Ext2 file path

#

Lemme find my OS' thread

#

OK Idk how to link to a thread but it's called AquaOS

flint idol Dec 25, 2024, 6:04 PM

#

https://discord.com/channels/440442961147199490/1192919089635668068

astral mason Dec 25, 2024, 6:04 PM

#

thank you

flint idol Dec 25, 2024, 9:33 PM

#

I should have Mm_MapViewOfUserMemory implemented

#

I will not be testing this.

flint idol Dec 26, 2024, 12:03 AM

#

uh oh

#

obos in debug mode broke

#

it's clang doing dumb shit with kasan

#

#

it looks like it's poisoning the stack

#

#

https://cdn.discordapp.com/emojis/816178972672131083.gif?size=64

#

if (OBOS_USE_CLANG)
    add_compile_options($<$<COMPILE_LANGUAGE:C,CXX>:-Wno-error=unknown-pragmas>
                        $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-unknown-pragmas>
                        $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-error=unknown-attributes>
                        $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-error=c23-extensions>)
    if (OBOS_ENABLE_KASAN)
        set (OBOS_ENABLE_KASAN OFF)
        message (WARNING "OBOS does not support Clang's ASAN implementation. If you really care about this, make a PR, as I don't.")
    endif()
endif()```

#

fixed it

white mulch Dec 26, 2024, 12:16 AM

#

have you tried setting the asan call threshold to 0?

flint idol Dec 26, 2024, 12:16 AM

#

I have no idea what that means

white mulch Dec 26, 2024, 12:17 AM

#

-mllvm -asan-instrumentation-with-call-threshold=0 should get rid of clang making inline shadow map checks if that's what its doing

flint idol Dec 26, 2024, 12:17 AM

#

this is only to the compiler right?

white mulch Dec 26, 2024, 12:17 AM

#

yes

flint idol Dec 26, 2024, 12:17 AM

#

k

#

I'll try

#

still crashes

#

resorting to plan B

flint idol Dec 26, 2024, 12:18 AM

#

flint idol ```c if (OBOS_USE_CLANG) add_compile_options($<$<COMPILE_LANGUAGE:C,CXX>:-Wn...

aka this

#

I love this stack trace

#

poor man's stack trace

#

#define OffsetOfReg(reg) ((uintptr_t)&(((ioapic_registers*)nullptr)->reg))

could clang be treating this as UB

#

and cause UBSan to yell at me?

#

I'm just gonna spam OBOS_NO_UBSAN until it boots

#

actually I'll just properly fix the bugs

weary hound Dec 26, 2024, 12:43 AM

#

flint idol ```c #define OffsetOfReg(reg) ((uintptr_t)&(((ioapic_registers*)nullptr)->reg)) ...

local man reinvents offsetof

flint idol Dec 26, 2024, 12:43 AM

#

why the hell did my 2nd monitor just die on me nooo

thick jolt Dec 26, 2024, 12:43 AM

#

flint idol I love this stack trace

can u just bt

real pecan Dec 26, 2024, 12:44 AM

#

flint idol ```c #define OffsetOfReg(reg) ((uintptr_t)&(((ioapic_registers*)nullptr)->reg)) ...

Just use offsetof??

thick jolt Dec 26, 2024, 12:44 AM

#

for the love of allah

real pecan Dec 26, 2024, 12:44 AM

#

It has a built-in even

flint idol Dec 26, 2024, 12:44 AM

#

real pecan Just use offsetof??

I did

flint idol Dec 26, 2024, 12:44 AM

#

thick jolt can u just bt

t'was broken

thick jolt Dec 26, 2024, 12:44 AM

#

flint idol t'was broken

so ur stack trace is broken

#

trl

#

corrupt stack

#

/j /j

flint idol Dec 26, 2024, 12:44 AM

#

flint idol why the hell did my 2nd monitor just die on me <:nooo:750466870993223750>

it's back

flint idol Dec 26, 2024, 12:45 AM

#

thick jolt corrupt stack

it was actually a stack overflow

#

trl

thick jolt Dec 26, 2024, 12:45 AM

#

😭

flint idol Dec 26, 2024, 12:45 AM

#

so yeah it turns out indexing an array at index -1 is bad

#

(whoops)

real pecan Dec 26, 2024, 12:45 AM

#

Works in python

flint idol Dec 26, 2024, 12:46 AM

#

and in C

#

but UBSan doesn't like it

#

and also it happens to be a bug

#

it boots now

#

https://github.com/OBOS-dev/obos/pull/27
Decided to finally make a draft PR for the userspace-work branch

GitHub

Userspace work by oberrow · Pull Request #27 · OBOS-dev/obos

Countless changes to the kernel, adding userspace, rewriting the VMM, PCI interface, power management stuff, and many other miscellaneous changes.

#

since it's getting pretty close to getting merged

#

time to continue vfs syscalls

#

I hope to get these done within a couple hours

#

should be simple enough, right?

real pecan Dec 26, 2024, 1:07 AM

#

flint idol since it's getting pretty close to getting merged

LETSFUCKINGGOOOOOO

flint idol Dec 26, 2024, 1:08 AM

#

We should be in bash by the end of the month

oberrow, august 2024

#

maybe I'm mistaken, but I think we're not in bash

real pecan Dec 26, 2024, 1:08 AM

#

Well u did a lot of other things

flint idol Dec 26, 2024, 1:08 AM

#

ye

#

I got "a bit sidetracked"

#

but it was for the better

real pecan Dec 26, 2024, 1:10 AM

#

Ye

flint idol Dec 26, 2024, 1:22 AM

#

is it bad that read is able to read even if the target buffer is mapped as RO KEKW

#

void* kbuf = Mm_MapViewOfUserMemory(CoreS_GetCPULocalPtr()->currentContext, (void*)buf, nullptr, nBytes, 0, &status);
if (obos_is_error(status))
    return status;

#

Mm_MapViewOfUserMemory ignores any protection flags the user set

#

so when the kernel gets it's view of user memory, it can do whatever the hell it wants

#

of course the change in prot flags isn't reflected in user-mode

flint idol Dec 26, 2024, 1:25 AM

#

flint idol is it bad that `read` is able to read even if the target buffer is mapped as RO ...

like you could do something like Vfs_FdRead(file, "heheheh I am evil", 5, NULL);

#

and the first five bytes of the string literal would be over written with whatever was in the file

#

KEKW

#

maybe I should fix that

#

I'll add a parameter called "respectUserProtection"

#

which does the following:

when checking if the user pages exist, if anyone of the user pages has different protection flags than the ones the kernel desires in it's mapping, it return OBOS_STATUS_PAGE_FAULT

#

    if (respectUserProtection)
    {
        if (kprot & OBOS_PROTECTION_EXECUTABLE)
        {
            OBOS_Error("Kernel is doing shady things, refusing to map user pages as executable inside the kernel address space. If all is in your favour, this is a bug, otherwise it's malware.");
            return false;
        }
        // If the kernel wants an RW mapping, but the user range is RO, then fail.
        if (!!(kprot & OBOS_PROTECTION_READ_ONLY) < rng->prot.ro)
            return false;

    }```

#

it's fair to refuse mapping kernel pages as executable if they're being shared from user mode

#

for safety

lean glen Dec 26, 2024, 1:37 AM

#

why are you sharing kernel pages

flint idol Dec 26, 2024, 1:37 AM

#

sorry, it's the other way around

#

user pages are being shared to kernel mode

lean glen Dec 26, 2024, 1:38 AM

#

why

flint idol Dec 26, 2024, 1:38 AM

#

kpti

torpid wigeon Dec 26, 2024, 1:39 AM

#

https://tenor.com/view/orange-cat-staring-orange-cat-staring-cat-gif-13724146065807985297

Tenor

flint idol Dec 26, 2024, 1:39 AM

#

otherwise I'd need to do very large amounts of memcpy

#

and allocation

#

for reading files

#

and writing files

#

so I thought to myself, why don't I just have a function to map user pages into the kernel context temporarily, so I can just pass the pages mapped in kernel mode to the vfs' read/write functions

#

so that's what I did

lean glen Dec 26, 2024, 1:43 AM

#

well then you control the protection

#

idk why you add a check

#

you need to read into something, just set the write bit

#

you need to write, only set the read bit

flint idol Dec 26, 2024, 1:44 AM

#

which is exactly what I did

#

what I added was a way to make sure the user can't do something like this:
Vfs_FdRead(file, "heheheh I am evil", 5, NULL);
and it returning success and doing the read into the RO string literal

#

with that, I should have my vfs syscalls implemented

#

implementing syscalls is really just a bunch of code copy+pasting when they just call into kernel functions with the exact same API

flint idol Dec 26, 2024, 1:56 AM

#

flint idol

I promised a redesign of my I/O interface

#

but I think it's better to do that after merging this branch

#

#1141057599584878645 message
Next thing on the list is partition probing syscalls

#

void OBOS_PartProbeAllDrives(bool check_checksum);
obos_status OBOS_PartProbeDrive(struct dirent* ent, bool check_checksum);```
that will correspond almost directly to a call to either of these functions

#

in fact, the first function can be directly made a syscall

#

without glue code

#

ok finished those syscalls

thick jolt Dec 26, 2024, 2:01 AM

#

bros doing syscalls

#

while im suffering in the nyaux thread

#

nooo

flint idol Dec 26, 2024, 2:02 AM

#

meme

thick jolt Dec 26, 2024, 2:02 AM

#

😡

flint idol Dec 26, 2024, 2:02 AM

#

driver interface syscalls now

#

driver_id *Drv_LoadDriver(const void* file, size_t szFile, obos_status* status);
obos_status Drv_StartDriver(driver_id* driver, thread** mainThread);
obos_status Drv_UnloadDriver(driver_id* driver);```

#

obos_status Drv_PnpLoadDriversAt(dirent* directory, bool wait);```

#

these are the functions I will expose

#

and I will add another syscall:

handle Sys_FindDriverByName(const char* name /* assumed to be 64-bytes at max */);```

#

also:

// Returns the next driver in the list
// The handle 'curr' is not automatically closed.
handle Sys_EnumerateLoadedDrivers(handle curr);

#

as well as:

obos_status Sys_QueryDriverName(handle drv, const char* namebuf, size_t sznamebuf /* need not be over 64 */);```

short mortar Dec 26, 2024, 2:20 AM

#

flint idol ```c void* kbuf = Mm_MapViewOfUserMemory(CoreS_GetCPULocalPtr()->currentContext,...

I though you were supposed to just let the kernel page fault on itself and handle that there

flint idol Dec 26, 2024, 2:21 AM

#

well two things:

I can't do that, I do KPTI (user pages and kernel pages are separated through completely different CR3s, as opposed to just a bit in the PTE)
even if I could do that, I wouldn't, as I want to be able to give the user a status telling it it page faulted, which could prove difficult if I just let it page fault

short mortar Dec 26, 2024, 2:22 AM

#

You add a shim function?

flint idol Dec 26, 2024, 2:22 AM

#

wdym?

short mortar Dec 26, 2024, 2:23 AM

#

Like save the registers, make pf handler switch to some function which returns error

#

I also kinda have that problem, but my issues are because of nonpreemptive kernel

flint idol Dec 26, 2024, 2:23 AM

#

short mortar Like save the registers, make pf handler switch to some function which returns e...

short mortar Dec 26, 2024, 2:24 AM

#

Otherwise mapping and unmapping have a huge performance hit

flint idol Dec 26, 2024, 2:24 AM

#

I'd rather map than allocate then do a slow ass memcpy_k_to_usr

#

at least for large sized IO

short mortar Dec 26, 2024, 2:24 AM

#

Also kpti assumes userspace pages are still mapped

flint idol Dec 26, 2024, 2:25 AM

#

does it?

short mortar Dec 26, 2024, 2:25 AM

#

(at least that's how I've been understanding it)

flint idol Dec 26, 2024, 2:25 AM

#

that's a misunderstanding

#

you only have the bare minimum of the kernel mapped in a user context

short mortar Dec 26, 2024, 2:26 AM

#

Yes

flint idol Dec 26, 2024, 2:26 AM

#

i.e., only enough to make a switch to the actual kernel context+ring3

short mortar Dec 26, 2024, 2:26 AM

#

And then another mapping with full user context + full kernel

flint idol Dec 26, 2024, 2:26 AM

#

well I haven't been doing that (oops)

short mortar Dec 26, 2024, 2:27 AM

#

#

And then you can copy efficiently

#

I'm not sure, but I think switching page tables is still a barrier against side channels (?)

#

(I'm not an expert and don't have a lot of interest in security vulnerabilities, besides the basics)

flint idol Dec 26, 2024, 2:28 AM

#

what I do is have it switch to the kernel's CR3 in the IRQ handler when coming from ring3

#

and then do all user memory access through a special function

#

memcpy_k_to_usr+memcpy_usr_to_k

short mortar Dec 26, 2024, 2:29 AM

#

Yeah, that's reasonable

#

But you have a flag which protects you from reading user memory from ring 0

flint idol Dec 26, 2024, 2:30 AM

#

ye, SMAP

#

and SMEP

#

which I enable if they exist

short mortar Dec 26, 2024, 2:30 AM

#

So in your functions, you can just disable that temporarely and read user memory as needed (?)

flint idol Dec 26, 2024, 2:30 AM

#

nay

#

they have a cursed implementation

short mortar Dec 26, 2024, 2:30 AM

#

Without inducing costs of page faults

flint idol Dec 26, 2024, 2:31 AM

#

    for (size_t i = 0; i < count; )
    {
        MmS_QueryPageInfo(ctx->pt, (uintptr_t)usr_src + i, nullptr, &usrphys);
        if (!usrphys)
        {
            // Mm_VirtualMemoryUnlock(ctx, futex, sizeof(*futex));
            return OBOS_STATUS_PAGE_FAULT;
        }
        size_t currCount = bytes_left > OBOS_PAGE_SIZE ? OBOS_PAGE_SIZE-((usrphys+usroffset)%OBOS_PAGE_SIZE) : (size_t)bytes_left;
        memcpy((void*)((uintptr_t)k_dest + i), MmS_MapVirtFromPhys(usrphys+usroffset), currCount);
        usroffset = 0;
        i += currCount;
        bytes_left -= currCount;
    }```

#

it does this

#

for usr_to_k

#

and vice versa for k_to_usr

short mortar Dec 26, 2024, 2:31 AM

#

That's what I do as well

#

kinda

flint idol Dec 26, 2024, 2:32 AM

#

TODO: Could it be possible to include these functions in the user CR3, then drop to the user CR3 to do the k_to_usr memcpy or usr_to_k memcpy?

short mortar Dec 26, 2024, 2:34 AM

#

How many cycles do pagefaults take on modern CPUs?

flint idol Dec 26, 2024, 2:34 AM

#

#x86

short mortar Dec 26, 2024, 2:34 AM

#

For small stuff, you can have some sort of scratch area

#

Which is mapped to both

#

So you first copy there from kernel to userspace-facing mapping

#

And then to userspace userspace

flint idol Dec 26, 2024, 2:35 AM

#

hmm

flint idol Dec 26, 2024, 3:10 AM

#

ok so driver interface syscalls are done

#

next up: TTYs-

#

oh

#

nooo

flint idol Dec 26, 2024, 3:13 AM

#

flint idol I got "a bit sidetracked"

it's only been two hours 😭

#

TTYs are tomorrow nooo

#

https://cdn.discordapp.com/emojis/816178972672131083.gif?size=64

#

then I can merge the branch

#

theoretically

#

wait

#

I'VE BEEN SAVED

#

signal-related syscalls

#

void OBOSS_SigReturn(interrupt_frame* frame);
obos_status OBOS_Kill(struct thread* as, struct thread* thr, int sigval);
obos_status OBOS_SigAction(int signum, const sigaction* act, sigaction* oldact);
obos_status OBOS_SigSuspend(sigset_t mask);
obos_status OBOS_SigPending(sigset_t* mask);
enum {
    SIG_BLOCK,
    SIG_SETMASK,
    SIG_UNBLOCK,
};
obos_status OBOS_SigProcMask(int how, const sigset_t* mask, sigset_t* oldset);
obos_status OBOS_SigAltStack(const uintptr_t* sp, uintptr_t* oldsp);```

#

those are the signal-related APIs I will expose

#

to user mode

#

Current roadmap:

~~Signal syscalls~~
~~Directory syscalls~~
exec syscall
TTYs
merge userspace-work
OBOS x86_64 gcc target
mlibc
bash

lean glen Dec 26, 2024, 3:31 AM

#

then you're gonna realize everything's broken because you didnt test anything

flint idol Dec 26, 2024, 3:31 AM

#

indeed

lean glen Dec 26, 2024, 3:31 AM

#

cryingmeme

flint idol Dec 26, 2024, 3:31 AM

#

that was the plan

flint idol Dec 26, 2024, 1:02 PM

#

Signal syscalls

#

ok

#

uhh

#

should be relatively easy

flint idol Dec 26, 2024, 2:10 PM

#

done

#

theoretically since these are just practically memcpys then just calling into the appropriate signal function

#

they should work

#

https://cdn.discordapp.com/emojis/816178972672131083.gif?size=64

#

now dirent syscalls would be simple, if I could find the part of the POSIX spec for them

#

there it is

thick jolt Dec 26, 2024, 2:33 PM

#

u will hit bash soon

flint idol Dec 26, 2024, 2:33 PM

#

Yes

flint idol Dec 26, 2024, 2:33 PM

#

flint idol Current roadmap: - ~~Signal syscalls~~ - ~~Directory syscalls~~ - exec syscall -...

roadmap here

thick jolt Dec 26, 2024, 2:33 PM

#

where as nyaux will be left in the dust struggling to fix bug after bug

flint idol Dec 26, 2024, 2:33 PM

#

hopefully by the new year

thick jolt Dec 26, 2024, 2:33 PM

#

nooo

short mortar Dec 26, 2024, 2:33 PM

#

thick jolt u will hit bash soon

Nyaux bash twice before pmOS

thick jolt Dec 26, 2024, 2:33 PM

#

short mortar Nyaux bash twice before pmOS

nahhh wont happen

#

ill probs hit another bug

#

after this one

#

then after that one

short mortar Dec 26, 2024, 2:33 PM

#

I need to implement working set

thick jolt Dec 26, 2024, 2:34 PM

#

yea we def gonna encounter a LOT of bugs

short mortar Dec 26, 2024, 2:34 PM

#

Then AHCI driver

flint idol Dec 26, 2024, 2:34 PM

#

thick jolt then after that one

-# you start working on obos trl

thick jolt Dec 26, 2024, 2:34 PM

#

flint idol -# you start working on obos <:trl:829316641191886888>

no meme

short mortar Dec 26, 2024, 2:34 PM

#

Before that I'll probably get caught in writing some other random drivers

#

Then decide to "improve" VMM again

thick jolt Dec 26, 2024, 2:34 PM

#

anyways no more distractions back to debugging nyaux

flint idol Dec 26, 2024, 2:35 PM

#

say, what if I didn't want to implement alphasort

#

posix mandates it

#

but idk if my kernel needs it as a syscall

#

or if mlibc does it for me

short mortar Dec 26, 2024, 2:37 PM

#

This should probably be done in libc

flint idol Dec 26, 2024, 2:38 PM

#

hmm

#

mlibc doesn't

#

even have sysdeps for any of those dirent functions

#

only sys_read_entries

devout niche Dec 26, 2024, 2:43 PM

#

flint idol even have sysdeps for any of those dirent functions

which ones are you thinking of?

#

the read entries one does all that mlibc needs the kernel (or posix server) to do

flint idol Dec 26, 2024, 2:48 PM

#

devout niche which ones are you thinking of?

these

#

stuff like opendir, readdir, closedir, etc.

devout niche Dec 26, 2024, 2:49 PM

#

flint idol stuff like opendir, readdir, closedir, etc.

not to worry, mlibc will do all those

flint idol Dec 26, 2024, 2:49 PM

#

well what is sys_read_entries supposed to do?

#

I know it reads the dirents

#

but what exactly is the format of the struct it's reading into

devout niche Dec 26, 2024, 2:50 PM

#

struct dirent

flint idol Dec 26, 2024, 2:50 PM

#

hold on, I think I found it:

typedef struct dirent {
    uint32_t inode; // Inode number
    uint32_t type;
    char name[NAME_MAX]; // Filename
} dirent_t;```

#

so this

devout niche Dec 26, 2024, 2:51 PM

#

i think mlibc has its own definition and expects in that form

flint idol Dec 26, 2024, 2:51 PM

#

ah ok

#

thanks

#

I'm sure it'd be perfectly fine if I always set inode to zero

#

right

#

what could possibly go wrong

#

I'm going to need to make stat syscalls

#

why did no one tell me I could something like man stat.3type

#

to find the struct stat format

#

int sys_read_entries(int handle, void *buffer, size_t max_size,
        size_t *bytes_read);```
is buffer allowed to be NULL here?

#

and if it is, do I just set *bytes_read = however many bytes it would read

real pecan Dec 26, 2024, 3:14 PM

#

see which stubs call into this and read the posix spec

flint idol Dec 26, 2024, 3:14 PM

#

real pecan see which stubs call into this and read the posix spec

it's not posix, iirc

lean glen Dec 26, 2024, 3:16 PM

#

it is

#

no

#

nvm

#

getdents

#

it's the equivalent of getdents

#

which is used by readdir

#

so read readdir

flint idol Dec 26, 2024, 3:17 PM

#

k

#

thanks

#OBOS (not vibecoded)